article
stringlengths 0
745k
| abstract
stringlengths 0
92.7k
|
|---|---|
near - ir spectroscopy has been playing a unique role in our understanding of the agn phenomenon for several reasons .
first , it includes a wealth of emission lines and stellar absorption features not observed in the optical , which are useful for studying the physical properties of the emission gas and to determine the age and metallicity of the nuclear stellar population as , for instance , the strong co bandheads in the h and k bands ( e.g. , schreiber 1998 ; origlia , moorwood , & oliva 1993 ) .
second , extinction by dust is attenuated by a factor of ten relative to that of the optical , allowing to probe depths unreachable at shorter wavelengths .
third , it is a transition region , where the continuum from the central source no longer dominates while the thermal continuum produced by dust and stars becomes important . in the last years
, the interest for observation in this interval has increased , and thanks to the availability of cross - dispersed ( xd ) spectrographs , it is possible to study the whole 0.8 - 2.4@xmath0 m region in a single observation , avoiding the aperture and seeing effects that usually affects @xmath3 spectroscopy done in long - slit single band observations . with the above in mind , here we present the first spatially resolved xd spectroscopy covering the interval 0.8 - 2.4@xmath0 m made for ngc1068 .
this object is one of the nearest and probably the most intensely studied seyfert 2 galaxy .
observations in all wavelength bands from radio to hard x - rays have formed a uniquely detailed picture of this source .
ngc 1068 has played a unique role in the development of unified scenarios for seyfert 1 and seyfert 2 galaxies ( antonucci & miller 1985 ) , in the study of molecular gas in the nuclear region of seyferts ( e.g. , myers & scoville 1987 ; tacconi et al .
1994 ) , and in elucidating the importance of star formation activity coexistent with the active galactic nucleus ( agn ) , on both larger ( e.g. , telesco & decher 1988 ) and smaller ( macchetto et al . 1994 ; thatte et al . 1997 ) scales .
ngc 1068 also hosts a prominent narrow - line region ( nlr ) that is approximately co - spatial with a linear radio source with two lobes ( wilson & ulvestad 1983 ) .
the narrow emission line region has been extensively characterized from subarcsecond clouds probed by the hubble space telescope ( hst ) ( evans et al .
1991 ; macchetto et al .
1994 ) to the ionization cone and extended emission - line region ( pogge 1988 ; unger et al .
1992 ) extending to radii of at least 30@xmath4 ( 1@xmath4=72 pc at the distance of 14.4 mpc ; tully 1988 ) . from approximately 2 `` southwest to the nucleus to 4 '' northeast , emission lines exhibit multiple components ( cecil et al .
1990 ; crenshaw & kraemer 2000a ) .
broad lines were found to be approximately 2500 - 4000 km s@xmath1 wide , which may be linked to those that are found in polarized light and believed to be reflected light from the inner broad line region ( blr ) .
narrow lines consist of a pair of red and blue components .
optical studies of [ oii ] and [ nii ] line profiles suggest that the separation of these two components varies across the conical nlr . between @xmath52``.5 and 4''.5 northeast from the nucleus ,
uv line emission is redshifted relative to the systemic value , a pattern that is interpreted as the expansion of the plasma in the radio lobe ( axon et al .
1998 ) .
the mechanisms powering the gas of the nlr , i.e. photoionization from the nucleus or shocks produced by jets , have long been under debate .
dopita & sutherland ( 1996 ) and bicknell et al . (
1998 ) proposed that the emission in the nlr may be entirely caused by shocks .
velocity splitting over 1000 km@xmath1 , reported by axon et al .
( 1998 ) , in the vicinity of some of the bright emission - line knots provides evidence that fast shocks exist in the nlr of ngc 1068 .
however , more recent hst data ( crenshaw & kraemer 2000b ; cecil et al . 2002 ; mazzalay et al . 2009 ) found that the emission line ratios are consistent with photoionization instead of shock heating mechanism . on the other hand ,
models accounting for both photoionization from the central radiation and shocks were required to explain both the continuum and emission - lines observed in the spectra of active galaxies ( e.g. , contini , rodrguez - ardila & viegas 2003 , rodrguez - ardila , contini & viegas 2005 ) .
the scenario in ngc 1068 is thus clearly complex , and observational constraints from all wavelengths need to be put together if we want to understand the processes taking place in this galaxy .
spatially resolved nlr spectroscopy in the nir can contribute to the solution of the ngc 1068 puzzle . here , we present the first observations on this source covering simultaneously the 0.8@xmath6 m region at moderate spectral and spatial resolution .
our data are thus able to map a wavelength interval rich in emission lines with a large range of ionization and absorption features from the stellar population not only from the nucleus but also from the circumnuclear region , poorly studied in the literature . in 2 a description of the observations
the main characteristics of the nuclear and extended spectra are presented in 3 and 4 , respectively .
photoionization model predictions for the most intense lines are discussed in 5 , whereas the calculations of the central mass based on stellar features appear in 6 .
finally , our concluding remarks are presented in 7 .
the spectra were obtained at the nasa 3 m infrared telescope facility ( irtf ) in october 30 , 2007 .
the spex spectrograph ( rayner et al . , 2003 ) was used in the short cross - dispersed mode ( sxd , 0.8 - 2.4 @xmath0 m ) .
the detector employed consisted of a 1024x1024 aladdin 3 insb array with a spatial scale of 0.15``/pixel . a 0.8''x 15 `` slit oriented in the north - south direction was used , providing a spectral resolution of 360 km / s .
this value was determined both from the arc lamp and the sky line spectra and was found to be constant with wavelength along the observed spectra . during the observations the seeing was @xmath51.3 '' .
observations were done nodding in an object - sky - sky - object pattern with integration time of 180 s per frame and total on - source integration time of 12 minutes .
after the galaxy , the a0v star hd18571 was observed as telluric standard and to flux calibrate the object .
the spectral reduction , extraction and wavelength calibration procedures were performed using spextool , the in - house software developed and provided by the spex team for the irtf community ( cushing , vacca & rayner , 2004 ) .
the s / n of the spectra varies from @xmath543 at the nuclear aperture to @xmath513 at the most external apertures .
intermediate regions have s / n of @xmath520 .
figure 1 shows the position of the slit superimposed in the galaxy contours obtained from galliano et al .
( 2003 ) . from our data
, we found that the light peak of the nuclear aperture in the k - band coincides in position with that determined by galliano et al .
( 2003 ) for the same band .
the gray contours show the 6 cm emission ( gallimore et al . 1996 ) and the red dotted contours show the 20 @xmath7 image ( alloin et al .
2000 ) . for ngc 1068
, 17 extractions were made along the spatial direction : one , of radius 0.8 , centered at the peak of the k - band distribution , and eight more , of 0.2 in radius , at each side of it .
in addition , in order to study in detail the variations of line profiles and continuum emission with distance , the nuclear aperture were subdivided into seven smaller apertures ( called nuc@xmath8a to nuc@xmath8 g in figure 1 ) of 0.2 of radius each .
we are aware that these smaller apertures can not be used for flux estimates because they are oversampled .
the 1-d spectra were then corrected for telluric absorption and flux calibrated using xtellcor ( vacca , cushing & rayner , 2003 ) , another in - house software developed by the irtf team .
finally , the different orders of the galaxy spectrum were merged to form a single 1-d frame .
it was later corrected for redshift , determined from the average z measured from the most conspicuous lines .
no galactic extinction correction was applied because it is negligible at nir ( a@xmath9=0.112 ) .
image ( alloin et al
. 2000),width=332 ] .
below we describe in detail the main features , both in emission and absorption , detected in the nuclear region of ngc 1068 , which correspond to the spectrum covering [email protected] , centered on the maximum of the galaxy profile .
the resulting nuclear spectra in the zj , h and k ranges are plotted in figures 2 , 3 and 4 , respectively , where the wavelengths have been translated to the observer s rest frame .
an inspection to figures 2 , 3 and 4 allows us to state that the most conspicuous features in the nuclear spectrum are [ siii]@[email protected],0.953 @xmath7 , [email protected] @xmath7 , pa@xmath12 , [ feii]@xmath111.643 @xmath7 , pa@xmath13 , and the coronal lines [ six]@xmath111.430 @xmath7 , and [ sivi]@xmath111.962 @xmath7 .
pa@xmath14 and br@xmath14 are also present , but very weak .
pa@xmath14 is in the wing of the very strong he i line .
pa@xmath13 could not be measured because it partially falls in the gap between orders .
emission from h@xmath15 molecule is clearly detected , whose most prominent lines are the transitions h@xmath15 1 - 0 ( s2 ) and h@xmath15 1 - 0 ( s1 ) , although only h@xmath15 1 - 0 ( s1 ) could be measured .
1 - 0 ( s3 ) is not clear in the nuclear spectra because it is too close to the [ sivi ] line , which is very broad .
the flux of h@xmath15 can be used to derive the mass of hot molecular gas , as shown in section 4.3 .
several other lines of lower intensity are detected in the zj spectra , namely [ cai]@xmath110.985 @xmath7 , [ sviii]@xmath110.991 @xmath7 , [email protected] @xmath7 , [ sii]@xmath111.032 @xmath7 ( a blend of 4 [ sii ] transitions ) , [ pii]@xmath111.188 @xmath7 , [ feii]@xmath111.257 @xmath7 and the coronal line [ six]@xmath111.252 @xmath7 . wherever possible , emission line fluxes for all features were measured assuming that the line profiles can be represented by a single or a sum of gaussian profiles .
the continuum underneath each line was fit by a low - order polynomium , usually a straight line . the smart routine ( higdon et al .
2004 ) , a software developed by the infrared spectrograph ( irs ) instrument team , originally designed to perform real - time processing and analysis of irs data , was used for this purpose .
the results are shown in table 1 .
errors on the fluxes were obtained from the values given by smart , which is calculated based on the residual of the fit and the flux calibration error , which is about 10 % for the infrared .
the detection of [ pii ] with considerable strength is particularly interesting . as explained by mazzalay & rodrguez - ardila ( 2007 ) , in the sun
this element is a factor @xmath51000 times less abundant than carbon . in the nuclear spectrum of ngc 1068
we find a ratio of [ pii]/[ci ] equal to 0.8 .
if the p / c abundance is about solar , the [ pii ] line should not be present , unless carbon , like iron , might be locked in grains .
this indicates that phosphorus may be overabundant by a factor of 10 - 20 compared to its solar abundance , relative to hydrogen .
the observed line widths are between 800 - 1000 km s@xmath1 .
this is in agreement with what is found in the optical region ( pelat & alloin 1980 ) .
although ngc 1068 has been considered a prototypical seyfert 2 galaxy , its forbidden lines are much wider than those of typical seyfert 2s . moreover , some emission - line profiles show multiple components ( pelat & alloin 1980 , alloin et al .
1983 , meaburn & pedlar 1986 ) also seen in the nir as we show in section 4.1 .
@|l|ccc & @xmath11 & flux & ew + line & ( @xmath7 ) & ( 10@xmath16 ergs @xmath17 s@xmath1 ) & ( @xmath7 ) + & 0.907 & 7.50 @xmath18 0.77 & 0.0085 @xmath18 0.0008 + @xmath19s iii@xmath20 $ ] & 0.953 & 19.80 @xmath18 2.03 & 0.0242 @xmath18 0.0026 + @xmath19c i@xmath20 $ ] & 0.985 & 0.91 @xmath18 0.11 & 0.0024 @xmath18 0.0003 + @xmath19s viii@xmath20 $ ] & 0.991 & 0.91 @xmath18 0.11 & 0.0015 @xmath18 0.0002 + he ii & 1.012 & 1.61 @xmath18 0.17 & 0.0023 @xmath18 0.0006 + @xmath19s ii@xmath20 $ ] & 1.032 & 2.81 @xmath18 0.36 & 0.0035 @xmath18 0.0008 + he i & 1.083 & 17.00 @xmath18 1.76 & 0.0327 @xmath18 0.0071 + pa@xmath14 & 1.093 & 0.97 @xmath18 0.12 & 0.0026 @xmath18 0.0004 + @xmath19pii@xmath20 $ ] & 1.188 & 0.78 @xmath18 0.14 & 0.0010 @xmath18 0.0002 + @xmath19s ix@xmath20 $ ] & 1.252 & 0.88 @xmath18 0.12 & 0.0021 @xmath18 0.0003 + @xmath19fe ii@xmath20 $ ] & 1.257 & 0.91 @xmath18 0.13 & 0.0023 @xmath18 0.0012 + pa@xmath12 & 1.282 & 2.24 @xmath18 0.29 & 0.0033 @xmath18 0.0007 + @xmath19si x@xmath20 $ ] & 1.430 & 1.59 @xmath18 0.18 & 0.0021 @xmath18 0.0007 + @xmath19fe
ii@xmath20 $ ] & 1.643 & 0.72 @xmath18 0.09 & 0.0008 @xmath18 0.0002 + h@xmath15 1 - 0 s(3 ) & 1.956 & 0.54 @xmath18 0.07 & 0.0003 @xmath18 0.0001 + @xmath19si vi@xmath20 $ ] & 1.962 & 3.39 @xmath18 0.37 & 0.0030 @xmath18 0.0009 + h@xmath15 1 - 0 s(1 ) & 2.121 & 0.35 @xmath18 0.05 & 0.0002 @xmath18 0.0001 + br@xmath14 & 2.165 & 0.42 @xmath18 0.06 & 0.0002 @xmath18 0.0001 + @xmath19ca viii@xmath20 $ ] & 2.322 & 1.01 @xmath18 0.23 & 0.0005 @xmath18 0.0002 + several absorption features are clearly seen in the h- and k - band nuclear spectrum ( figures 3 and 4 ) .
in particular , we observe the following ones : mgi 1.50 @xmath7 , co 1.58 @xmath7 , sii 1.59 @xmath7 , co 1.62 @xmath7 , mgi 1.71 @xmath7 , co 2.29 @xmath7 , co 2.35 @xmath7 and co 2.38 @xmath7 .
these features are identified after their main contributor , although , depending on the dominant stellar type , other features may become dominant .
the co 1.62 @xmath7 can be used to obtain a rough estimate of the contribution of the stellar population flux to the h band .
this is done by measuring the depth of the @xmath21co(6 - 3 ) overtone bandhead at 1.62 @xmath7 .
we found that the observed depth is 12@xmath22 @xmath18 1@xmath22 of the continuum .
this can be compared to the @xmath5 20@xmath22 that is typically expected for a population of gkm supergiants that dominates the h band light for a stellar population older than 10@xmath23 yr ( schinnerer et al .
it can thus be roughly estimated that about 60@xmath22 @xmath18 5@xmath22 of the total flux in the h band continuum is due to gmk supergiants .
following schinnerer et al .
( 1998 ) , although these stars dominate the stellar emission in the h band , about one - third of the total stellar flux comes from stars of other stellar classes .
therefore , 90@xmath22 @xmath18 8@xmath22 of the nuclear h band continuum could be of stellar origin .
this is in reasonable agreement with the 74@xmath22 @xmath18 8@xmath22 determined by cid fernandes et al .
( 2001 ) for this object , based on optical spectroscopy .
this means that the h - band continuum is essentially dominated by the stellar population , with only @xmath5 10 - 20 @xmath22 due to the agn .
this calculations show how the nir spectroscopy is a potential tool to unveil hidden starburst components , either in highly obscured objects or in those in which the optical continuum emission is dominated by the agn component ( as in seyfert 1 galaxies ) .
one way of estimating the age of the stellar population in the nir is using the co index defined by ivanov et al .
( 2000 ) , for the absorption feature starting at 2.29 @xmath7 .
this index is supposed to be insensitive to extinction and to possible uncertainties in the continuum shape of the infrared spectra and it is defined as @xmath24 where @xmath25 and @xmath26 are the average fluxes within a bandwidth of 0.01 @xmath7 , centered on the blue continuum and on the bandhead respectively .
for the nuclear aperture we obtained an index of -0.01 @xmath18 0.05 .
this value is extremely low for a seyfert 2 galaxy , according to the sample of ivanov et al .
notice that in a seyfert galaxy nucleus , the late - type stellar features are substantially diluted by non - stellar nuclear emission .
ngc 1068 is probably an extreme case , since its nuclear spectrum shows a very steep increase in the k band continuum flux with wavelength ( figure 4 ) , much steeper than what is expected for a non - thermal power - law source .
the k band excess of emission is probably due to thermal radiation from hot dust ( thatte et al . 1997 ) . in order to minimize the influence of the shape of the continuum
, we normalized the spectrum in this region using a linear fit in the region between 2.20 and 2.26 @xmath7 , where there are no apparent features . after dividing the spectrum by this fit
the co index is recalculated .
the new value is 0.01 @xmath18 0.05 .
thus , the low value is not caused by the shape of the spectrum and suggests that there is no evidence for a strong starburst , in agreement with the results of ivanov et al .
( 2000 ) for seyfert galaxies , but contrary to what was found in the h band .
nevertheless , it is really hard to estimate the true dilution of the bandhead by the continuum . the best way would be to do a full stellar population fitting , including a non - stellar component .
this is out of the scope of this paper , but is presented in detail in a follow - up paper ( martins et al .
in addition to the nuclear spectra discussed in the preceding section , the spatial information contained in the slit allowed us to extract the spectra corresponding to the narrow line region and to the extended line region of ngc 1068 to the limits of @xmath5350 pc north and south of the nucleus .
gas kinematics as well as extinction and the molecular and [ feii ] lines will be studied in more detail in the following sections .
the spectra for all integrated apertures are presented in figures 5 - 7 , where the wavelengths are in the observer s rest frame .
it is easy to see that the most conspicuous lines show a double peaked profile .
one component is in the rest frame of the galaxy ( primary component , hereafter ) while the other ( second component , hereafter ) is either blueshifted towards the north of the nucleus or redshifted to the south .
the relative intensity of the peaks also varies with distance , as shown in figure 8 for [ siii]@xmath110.953 @xmath7 , the strongest line in our spectra .
it can be seen that [ siii ] is detected at distances of up to @xmath5350 pc from the nucleus . as explained in sec . 2 , for the purpose of comparing the line profiles we divided the nuclear spectra ( 1.6 aperture ) into seven smaller apertures ( see fig . 1 ) .
the aperture named `` nuc@xmath27 '' , centered on the peak of the continuum light distribution , is extracted over the central 0.4 `` window and the remaining six with apertures 0.2 '' wide .
although the integrated flux of these smaller apertures are oversampled because the extraction window is smaller than the seeing value , this approach allows us to study line profiles , which clearly show changes along the spatial direction at scales down to 20 pc .
a comparison of the line profiles in velocity space observed in the most conspicuous emission lines is presented in figure 9 . in this figure
, the continuum underneath each line was adjusted using a linear interpolation and then subtracted .
it can be seen that the coronal line [ sviii ] shows a profile very similar to that of [ siii ] , and second components can be clearly seen .
the molecular h@xmath15 lines also show scarce evidence of a second component , not so clear since the line is weak .
they are also narrower than the emission features of other species outside the nucleus , which indicates that it might be formed in a different region .
indeed , nir observations has shown that the h@xmath15 emission is distributed mainly along a direction perpendicular to the ionizing cone ( galliano & alloin 2002 ) .
the [ feii ] lines seem to follow the h@xmath15 lines more closely .
[ nirzoom ] we did not identify a broad component with fwhm @xmath5 2000 km s@xmath1 in any line , although they have been reported on fuse observations ( zheng et al .
2008 ) . that may be because fuse slit collects emission from a block of regions spanning several arcseconds perpendicular to the conical axis and the total line emission from these regions may be blended into one broad component .
since we are observing much smaller regions with each aperture , we minimize this effect .
not all emission lines have double peaked profiles .
this is the case of high ionization lines such as [ sviii ] and [ six ] as well as molecular lines of h@xmath15 .
moreover , low ionization lines like [ cai ] or [ pii ] also seem to be devoided of a double - peak structure but these are faint lines
. the lack of splitting may be either by intrinsic reasons or due to low s / n . in figure 10 ,
the relative flux of the most conspicuous emission lines are shown as a function of the distance to the center .
the fluxes were normalized relative to the nucleus to allow the comparison between the different lines .
negative distances represent the northern direction and positive distances the southern one . for double peaked lines
, the upper panel shows the distribution of the primary component while the bottom panel that of the secondary component .
it can be seen that the second component of [ siii ] 0.953@xmath0 m and hei 1.083@xmath0 m are observed up to 200 pc from the nucleus .
indeed , it seems that there is a maximum of the second components between 100 - 150 pc from the center . as for the primary components , evidence of emission up to 300 pc from the center
is observed although it becomes too much faint to be measured at 3@xmath28 level .
note that the [ feii ] emission is the most extended of all lines analyzed . the flux gas distribution shown in figure 10 and mapped by the first time in the literature with data in the interval 0.8@xmath29
m confirms previous reports in the optical and mir in the sense that the northern zone has stronger emission than the southern one ( mazzalay et al .
2009 ; geballe et al .
this can be attributed to the strong extinction towards the observer . in this scenario , the ionization cone
is partially hidden from our view because it crosses the host galaxy plane .
note also that high ionization lines like [ six ] 1.430@xmath0 m and [ sivi ] 1.954@xmath0 m are detected only up to @xmath5200 pc from the nucleus .
this is a clear indication that the gas is more ionized near the nucleus .
the observed emission features of the extended region were also measured with the software smart .
the results are shown in tables 2 and 3 .
errors were estimated the same way as in table 1 .
@|l||c cccccccc + & & + line & @xmath11 ( @xmath7 ) & ap01 & ap02 & ap03 & ap04 & ap05 & ap06 & ap07 & ap08 + @xmath19s iii@xmath20 $ ] & 0.907 & 5.81 @xmath18 1.10 & 1.63 @xmath18 0.27 & 1.28 @xmath18 0.25 & 1.31 @xmath18 0.22 & 1.04 @xmath18 0.19 & 0.45 @xmath18 0.08 & 0.61 @xmath18 0.06 & 0.23 @xmath18 0.03 + @xmath19s iii@xmath20 $ ] & 0.953 & 14.21 @xmath18 1.80 & 6.66 @xmath18 0.85 & 3.95 @xmath18 0.71 & 3.93 @xmath18 0.70 & 1.84 @xmath18 0.33 & 0.96 @xmath18 0.14 & 1.90 @xmath18 0.24 & 1.22 @xmath18 0.15 + @xmath19c i@xmath20 $ ] & 0.985 & 1.60 @xmath18 0.26 & 0.48 @xmath18 0.09 & 0.45 @xmath18 0.11 & 0.64 @xmath18 0.13 & - & - & - & - + @xmath19s viii@xmath20 $ ] & 0.991 & 0.21 @xmath18 0.07 & - & - & - & - & - & - & - + he ii & 1.012 & 1.12 @xmath18 0.37 & 1.00 @xmath18 0.15 & 0.63 @xmath18 0.16 & 0.61 @xmath18 0.12 & 0.34 @xmath18 0.10 & - & 0.11 @xmath18 0.02 & 0.22 @xmath18 0.03 + @xmath19s ii@xmath20 $ ] & 1.032 & 3.50 @xmath18 0.77 & 1.28 @xmath18 0.28 & 0.81 @xmath18 0.16 & - & - & - & 0.19 @xmath18 0.03 & - + he i & 1.083 & 11.48 @xmath18 1.56 & 3.64 @xmath18 0.48 & 2.36 @xmath18 0.32 & 2.38 @xmath18 0.43 & 1.43 @xmath18 0.34 & 0.77 @xmath18 0.11 & 0.84 @xmath18 0.12 & 0.58 @xmath18 0.08 + pa@xmath14 & 1.093 & 1.13 @xmath18 0.19 & 0.74 @xmath18 1.45 & 0.58 @xmath18 0.13 & 0.62 @xmath18 0.14 & - & 0.07 @xmath18 0.01 & 0.08 @xmath18 0.01 & - + @xmath19pii@xmath20 $ ] & 1.188 & 1.10 @xmath18 0.19 & 0.82 @xmath18 0.12 & 0.64 @xmath18 0.10 & 1.01 @xmath18 0.13 & - & - & - & - + @xmath19s ix@xmath20 $ ] & 1.252 & 0.39 @xmath18 0.08 & 0.11 @xmath18 0.02 & 0.72 @xmath18 0.10 & 0.22 @xmath18 0.09 & - & - & - & - + @xmath19fe ii@xmath20 $ ] & 1.257 & 1.39 @xmath18 0.28 & 1.06 @xmath18 0.15 & - & 0.37 @xmath18 0.08 & 0.28 @xmath18 0.07 & 0.22 @xmath18 0.05 & 0.36 @xmath18 0.05 & - + pa@xmath12 & 1.282 & 2.22 @xmath18 0.41 & 1.04 @xmath18 0.14 & 1.06 @xmath18 0.15 & 0.65 @xmath18 0.12 & 0.39 @xmath18 0.09 & 0.13 @xmath18 0.03 & 0.12 @xmath18 0.02 & - + @xmath19si x@xmath20 $ ] & 1.430 & 1.43 @xmath18 0.27 & 1.25 @xmath18 1.05 & - & - & - & - & - & - + @xmath19fe ii@xmath20 $ ] & 1.643 & 0.76 @xmath18 0.59 & 1.27 @xmath18 0.17 & 0.57 @xmath18 0.07 & 0.58 @xmath18 0.08 & 0.33 @xmath18 0.06 & 0.15 @xmath18 0.03 & - & - + @xmath19si vi@xmath20 $ ] & 1.962 & 3.28 @xmath18 0.41 & 1.06 @xmath18 0.17 & 0.17 @xmath18 0.03 & 0.16 @xmath18 0.03 & - & - & 0.23 @xmath18 0.04 & 0.12 @xmath18 0.01 + h@xmath15 1 - 0 s(1 ) & 2.121 & 0.65 @xmath18 0.13 & 0.32 @xmath18 0.04 & 0.24 @xmath18 0.03 & 0.25 @xmath18 0.03 & 0.14 @xmath18 0.02 & 0.10 @xmath18 0.02 & 0.12 @xmath18 0.02 & 0.10 @xmath18 0.01 + br@xmath14 & 2.165 & - & - & - & - & - & & - & - + @xmath19ca viii@xmath20 $ ] & 2.322 & 0.21 @xmath18 0.03 & - & - & - & - & & - & - + + & & + line & @xmath11 ( @xmath7 ) & ap01 & ap02 & ap03 & ap04 & ap05 & ap06 & ap07 & ap08 + @xmath19s iii@xmath20 $ ] & 0.907 & 2.13 @xmath18 0.44 & 2.77 @xmath18 0.52 & 1.25
@xmath18 0.13 & 1.24 @xmath18 0.24 & - & - & - & - + @xmath19s iii@xmath20 $ ] & 0.953 & 4.16 @xmath18 0.81 & 4.81 @xmath18 0.87 & 2.87 @xmath18 0.31 & 2.87 @xmath18 0.54 & 0.76 @xmath18 0.16 & 0.65 @xmath18 0.14 & - & - + he ii & 1.012 & 0.17 @xmath18 0.31 & - & - & - & - & - & - & - + he i & 1.083 & 4.09 @xmath18 0.79 & 4.46 @xmath18 0.82 & 2.20 @xmath18 0.56 & 2.18 @xmath18 0.43 & 0.69 @xmath18 0.14 & 0.29 @xmath18 0.07 & 0.25 @xmath18 0.07 & - + pa@xmath12 & 1.282 & 0.20 @xmath18 0.04 & 0.29 @xmath18 0.06 & 0.18 @xmath18 0.09 & 0.76 @xmath18 0.16 & - & - & - & - + @xmath19si vi@xmath20 $ ] & 1.962 & - & 0.76 @xmath18 0.14 & 0.35 @xmath18 0.11 & 0.39 @xmath18 0.09 & - & - & - & - + @|l||c cccccccc + & & + line & @xmath11(@xmath7 ) & ap09 & ap10 & ap11 & ap12 & ap13 & ap14 & ap15 & ap16 + @xmath19s iii@xmath20 $ ] & 0.907 & 17.96 @xmath18 2.06 & 11.73 @xmath18 1.18 & 6.50 @xmath18 0.96 & 2.53 @xmath18 0.33 & 1.74 @xmath18 0.24 & 1.43 @xmath18 0.19 & 1.32 @xmath18 0.18 & 0.56 @xmath18 0.09 + @xmath19s iii@xmath20 $ ] & 0.953 & 46.14 @xmath18 6.96 & 30.03 @xmath18 3.01 & 16.33 @xmath18 2.10 & 7.46 @xmath18 0.99 & 4.83 @xmath18 0.62 & 2.94 @xmath18 0.39 & 3.25 @xmath18 0.45 & 1.24 @xmath18 0.15 + @xmath19c i@xmath20 $ ] & 0.985 & 2.38 @xmath18 0.42 & 1.85 @xmath18 0.27 & 1.39 @xmath18 0.20 & 0.88 @xmath18 0.13 & 0.53 @xmath18 0.08 & 0.55 @xmath18 0.07 & - & - + @xmath19s viii@xmath20 $ ] & 0.991 & 1.86 @xmath18 0.30 & 1.40 @xmath18 0.21 & 0.67 @xmath18 0.10 & 0.21 @xmath18 0.04 & - & - & - & - +
he ii & 1.012 & 4.25 @xmath18 0.70 & 2.19 @xmath18 0.30 & 1.09 @xmath18 0.16 & 0.68 @xmath18 0.10 & 0.50 @xmath18 0.08 & 0.57 @xmath18 0.08 & - & - + @xmath19s ii@xmath20 $ ] & 1.032 & 6.36 @xmath18 1.05 & 4.15 @xmath18 0.57 & 2.30 @xmath18 0.30 & 0.90 @xmath18 0.12 & 0.69 @xmath18 0.09 & 0.30 @xmath18 0.05 & - & - + he i & 1.083 & 29.19 @xmath18 4.34 & 17.39 @xmath18 1.75 & 9.87 @xmath18 1.33 & 4.75 @xmath18 0.65 & 2.31 @xmath18 0.30 & 2.19 @xmath18 0.30 & 1.98 @xmath18 0.22 & 0.74 @xmath18 0.13 + pa@xmath14 & 1.093 & 2.46 @xmath18 0.31 & 1.45 @xmath18 0.20 & 1.10 @xmath18 0.19 & 0.72 @xmath18 0.13 & 0.30 @xmath18 0.05 & 0.38 @xmath18 0.05 & - & - + @xmath19pii@xmath20 $ ] & 1.188 & 2.01 @xmath18 0.22 & 1.46 @xmath18 0.19 & 1.01 @xmath18 0.13 & 1.01 @xmath18 0.13 & 0.39 @xmath18 0.06 & - & - & - + @xmath19s ix@xmath20 $ ] & 1.252 & 2.31 @xmath18 0.28 & 1.68 @xmath18 0.23 & 0.81 @xmath18 0.12 & 0.99 @xmath18 0.16 & 0.36 @xmath18 0.06 & 0.22 @xmath18 0.04 & 0.15 @xmath18 0.03 & - + @xmath19fe
ii@xmath20 $ ] & 1.257 & 2.40 @xmath18 0.29 & 2.06 @xmath18 0.28 & 1.95 @xmath18 0.27 & 1.14 @xmath18 0.19 & 1.15 @xmath18 0.17 & 0.77 @xmath18 0.10 & 0.51 @xmath18 0.07 & - + pa@xmath12 & 1.282 & 4.81 @xmath18 0.70 & 2.44 @xmath18 0.32 & 1.02 @xmath18 0.42 & 0.85 @xmath18 0.12 & 0.52 @xmath18 0.08 & 0.27 @xmath18 0.04 & 0.24 @xmath18 0.04 & - + @xmath19si x@xmath20 $ ] & 1.430 & 2.41 @xmath18 0.30 & 1.06 @xmath18 0.15 & 0.76 @xmath18 0.11 & 0.45 @xmath18 0.09 & - & - & - & - + @xmath19fe ii@xmath20 $ ] & 1.643 & 1.91 @xmath18 0.26 & 1.71 @xmath18 0.25 & 1.80 @xmath18 0.21 & 1.43 @xmath18 0.16 & 1.00 @xmath18 0.12 & 0.79 @xmath18 0.09 & 0.78 @xmath18 0.10 & - + @xmath19si vi@xmath20 $ ] & 1.962 & 5.51 @xmath18 0.69 & 2.42 @xmath18 0.36 & 0.86 @xmath18 0.12 & 0.43 @xmath18 0.10 & 0.13 @xmath18 0.03 & - & - & - + h@xmath15 1 - 0 s(1 ) & 2.121 & 0.76 @xmath18 0.10 & 0.53 @xmath18 0.06 & 0.32 @xmath18 0.04 & 0.21 @xmath18 0.04 & 0.13 @xmath18 0.02 & 0.08 @xmath18 0.01 & 0.15 @xmath18 0.03 & - + br@xmath14 & 2.165 & 0.99 @xmath18 0.12 & 0.58 @xmath18 0.09 & - & 0.16 @xmath18 0.04 & - & 0.12 @xmath18 0.01 & - & - + @xmath19ca viii@xmath20 $ ] & 2.322 & 1.14 @xmath18 0.17 & 0.39 @xmath18 0.09 & - & - & - & - & - & - + + & & + line & @xmath11 ( @xmath7 ) & ap09 & ap10 & ap11 & ap12 & ap13 & ap14 & ap15 & ap16 + @xmath19s iii@xmath20 $ ] & 0.907 & 2.49 @xmath18 0.33 & 1.66 @xmath18 0.22 & 1.20 @xmath18 0.16 & 2.29 @xmath18 0.33 & 1.18 @xmath18 0.18 & 0.55 @xmath18 0.08 & 0.43 @xmath18 0.07 & - + @xmath19s iii@xmath20 $ ] & 0.953 & 5.01 @xmath18 0.70 & 3.90 @xmath18 0.52 & 3.07 @xmath18 0.45 & 3.33 @xmath18 0.44 & 1.89 @xmath18 0.28 & 1.20 @xmath18 0.18 & 0.88 @xmath18 0.13 & - + @xmath19s viii@xmath20 $ ] & 0.991 & 0.47 @xmath18 0.07 & - & - & - & - & - & - & - + he ii & 1.012 & - & - & - & - & - & - & - & - + he i & 1.083 & 4.23 @xmath18 0.58 & 3.64 @xmath18 0.49 & 1.98 @xmath18 0.27 & 1.99 @xmath18 0.29 & 2.07 @xmath18 0.23 & 0.58 @xmath18 0.09 & - & - + pa@xmath12 & 1.282 & 0.58 @xmath18 0.09 & 0.53 @xmath18 0.08 & 0.62 @xmath18 0.09 & - & 0.19 @xmath18 0.31 & 0.15 @xmath18 0.03 & - & - + @xmath19si x@xmath20 $ ] & 1.430 & 0.34 @xmath18 0.05 & 0.25 @xmath18 0.04 & - & - & - & - & - & - + @xmath19fe ii@xmath20 $ ] & 1.643 & 0.47 @xmath18 0.06 & 0.46 @xmath18 0.07 & - & - & - & - & - & - + @xmath19si vi@xmath20 $ ] & 1.962 & 2.69 @xmath18 0.35 & - & - & - & - & - & - & - + the extinction towards the nucleus and circumnuclear region was determined from the comparison of the predicted and observed emission line ratios assuming the standard extinction law of cardelli et al .
( 1989 ) for r@xmath30 = 3.1 . from our observed lines ,
two line ratios are extinction indicators : pa@xmath12/pa@xmath14 and [ feii ] 1.257@xmath0m/1.643@xmath0 m .
we added both components of pa@xmath12 for this purpose .
the reason for this is that pa@xmath14 has no second components , probably because it is a weak line and it was not possible to distinguish them .
when using only the primary component of pa@xmath12 , many ratios give unphysical results .
pa@xmath13 could not be used in this calculation because it was severely affected by telluric transmission .
moreover , br@xmath14 was only detected in the central aperture . for hydrogen we used the intrinsic ratios for the case b given by osterbrock ( 1989 ) .
the pair of [ feii ] lines used to estimate the extinction share the same upper level , meaning that the intensity ratio depends only on the energy differences between the lines and their einstein - a coefficients , making this line ratio a useful probe of reddening .
the intrinsic value of the [ feii ] 1.257@xmath0m/1.643@xmath0 m ratio was taken from bautista & pradhan ( 1998 ) , calculated from nussbaumer & storey ( 1988 ) transition probabilities .
table 4 shows the values of extinction found for each aperture .
for the errors in this table we only considered the errors in the line measurements and not in the flux calibration .
@ccccc line & pa@xmath12/pa@xmath14 & e(b - v ) & [ feii]1.257/1.643 & e(b - v ) + ap07 & 1.49 @xmath18 0.31 & @xmath31 0.01 & - & - + ap06 & 1.84 @xmath18 0.48 & 0.13 @xmath18 0.25 & 1.54 @xmath18 0.49 & @xmath31 0.92 + ap05 & - & - & 0.85 @xmath18 0.22 & 1.65 @xmath18 0.26 + ap04 & 2.27 @xmath18 0.55 & 1.05 @xmath18 0.26 & 0.64 @xmath18 0.17 & 2.65 @xmath18 0.26 + ap03 & 1.31 @xmath18 0.55 & 0.82 @xmath18 0.56 & 1.26 @xmath18 0.16 & 0.28 @xmath18 0.13 + ap02 & 1.83 @xmath18 0.41 & 0.09 @xmath18 0.22 & 1.12 @xmath18 0.31 & 0.69 @xmath18 0.28 + ap01 & 1.98 @xmath18 0.51 & 0.44 @xmath18 0.17 & 1.83 @xmath18 0.51 & @xmath31 0.10 + nuc & 2.31 @xmath18 0.24 & 1.13 @xmath18 0.10 & 1.26 @xmath18 0.15 & 0.26 @xmath18 0.12 + ap09 & 1.96 @xmath18 0.32 & 0.39 @xmath18 0.12 & 1.26 @xmath18 0.16 & 0.28 @xmath18 0.13 + ap10 & 2.05 @xmath18 0.34 & 0.60 @xmath18 0.13 & 1.20 @xmath18 0.23 & 0.44 @xmath18 0.20 + ap11 & 1.50 @xmath18 0.38 & @xmath31 0.03 & 1.09 @xmath18 0.10 & 0.80 @xmath18 0.10 + ap12 & 1.17 @xmath18 0.35 & - & 0.80 @xmath18 0.11 & 1.90 @xmath18 0.14 + ap13 & 2.33 @xmath18 0.49 & 1.18 @xmath18 0.18 & 1.15 @xmath18 0.16 & 0.60 @xmath18 0.14 + ap14 & 1.11 @xmath18 0.55 & @xmath31 0.05 & 0.98 @xmath18 0.26 & 1.17 @xmath18 0.71 + ap15 & - & - & 0.65 @xmath18 0.12 & 2.60 @xmath18 0.19 + the results point to extinction variations up to a factor of 4 within the nlr .
however , the e(b - v ) derived from the hydrogen ratios suffer from two major problems : the pa@xmath12 line may be contaminated with the [ feii ] 1.278 @xmath7 , leading to an overestimation of the reddening determined from pa@xmath12/pa@xmath14 .
also , the pa@xmath14 line is in the wing of the much stronger hei 1.083 @xmath7 line .
therefore , problems with the deblending of these two lines could lead to larger uncertainties in this calculation .
note that the uncertainties quoted are determined only from the s / n around the region containing the lines of interest and they do not take into account potential deblending issues .
differences between the reddening obtained from the [ feii ] lines and hydrogen lines are expected , since these lines are probably not emitted from the same region .
keep in mind that the uncertainties in the e(b - v ) from the [ feii ] lines can also be large since these lines are intrinsically weak .
the analysis of kraemer et al .
( 1998 ) indicates that there may be some dust mixed with gas in various amounts in this galaxy .
veilleux et al .
( 1997 ) obtained a value of e(b - v ) = 0.34 using the balmer decrement .
our average result for the nlr from both determinations is above this value .
this might be an indication that the nir might come from dustier , deeper regions of the nlr .
the extinction obtained from the two line ratios for the various apertures are shown in figure 11 .
the nuclear region seems to be more dusty than the adjacent regions but secondary peaks of extinction are also observed , with maximums around 150 - 200 pc from each side of the nucleus .
these positions are also close to the region where the second component of the emission lines has a maximum , and may be associated to the presence of dense clouds mixed with dust in this region .
emission from h@xmath15 molecules are observed in most apertures of ngc 1068 , and has been the subject of extensive analysis by different authors .
reunanen et al .
( 2003 ) suggest a star formation origin for these lines as a result of uv heating from ob stars . whether the heating mechanism is thermal or
fluorescence is not clear because neither the 2 - 1 s(1 ) 2.247@xmath0 m nor the 1 - 0 s(0 ) 2.223@xmath0 m emission were detected in the nucleus .
they derived an upper limit of 5300 k for the vibrational temperature in the inner few hundred of parsecs of this object .
we used their method to estimate the mass of the hot molecular gas present in each aperture . assuming t = 2000k , a transition probability a@xmath32 = 3.47 x 10@xmath33 s@xmath1 ( turner et al .
1977 ) and the population fraction in the @xmath34 = 1 , j=3 level f@xmath35 = 0.0122 ( scoville et al . 1982 ) , the h@xmath15 mass was obtained from @xmath36 where d is the distance in mpc , i@xmath37 is the observed flux of h@xmath15 @xmath112.121 @xmath7 in erg @xmath17 s@xmath1 and a@xmath38 is the 2.2-@xmath7 extinction calculated through the e(b - v ) from table 4 .
the mass m@xmath39 is given in m@xmath40 .
cccc south & m@xmath39 [ m@xmath40 ] & north & m@xmath39 [ m@xmath40 ] + ap08 & 11.8 & ap16 &
+ ap07 & 13.5 & ap15 & 49.8 + ap06 & 16.8 & ap14 & 15.1 + ap05 & 30.9 & ap13 & 189.5 + ap04 & 82.0 & ap12 & 65.7 + ap03 & 31.4 & ap11 & 52.0 + ap02 & 48.7 & ap10 & 72.6 + ap01 & 78.1 & ap09 & 98.2 + nuc & 449.8 & & + the results for each aperture are given in table 5 .
we have to point here that it is now well known that the distribution of h@xmath15 ( perpendicular to the nlr ) is so that we miss a large fraction of it in our slit ( sanchez et al .
2009 , galliano et al .
these values are in general agreement with the values obtained by reunanen et al .
( 2003 ) for seyfert 1 and 2 galaxies , although the nuclear value we obtained is much smaller than what they obtained for ngc 1068 .
several reasons may be invoked to explain the difference .
the aperture they use to do the analysis is much larger than ours , encompassing all our apertures .
as can be seen in table 4 , the extinction varies and the higher values are not in the center .
reunanen et al .
probably got an average of these values .
in addition , they obtained a@xmath38 from the continuum extinction , assuming that the difference for the line extinction would not be large .
this may be true for many seyfert galaxies , but the continuum of the nuclear region of ngc 1068 seems to be extremely dusty , as can be seen in figure 7 , by the inclination of the spectrum . using the extinction we derived from the emission lines ( section 4.2 ) we obtained a much smaller value for the hot molecular gas mass .
galliano et al . ( 2003 ) analyze the possibility of explaining their observations of the mir emission with diffuse dust in the ionization cone .
this model is discarded because the diffuse dust would shield the central engine as well as the blr from any direction , requiring an ionizing mechanism other than photoionization to explain the nlr emission . on the other hand
, they favor a model of thick dust clouds in the cone , with a small covering factor , heated by the central source radiation , without discarding additional heating mechanisms by shocks generated by the interaction of radio jets and gas clouds . in order to explain the observed molecular emission perpendicular to the cone axis , galliano et al.(2003 ) built a numerical model combining a molecular disc and an axisymmetrical , no coplanar , x - ray absorber ( see their figure 9 ) , reached by x - rays from the central source .
this two component model reproduces quite well the main features observed in the h2 and co emission maps .
more recently , the excitation of near infrared h@xmath15 and fe ii lines have been discussed by rodrguez - ardila et al .
( 2004 ) based on observations covering the inner 300 pc of most seyfert galaxies and minimizing contamination by the host galaxy .
they showed that the diagnostic diagram h@[email protected]/br@xmath14 versus [ feii]@xmath111.257 /pa@xmath12 proposed by larkin et al .
( 1998 ) is an efficient tool to separate objects by the level of nuclear activity , i.e. , agn , starburst or liner . by including more objects , rodrguez - ardila , riffel & pastoriza ( 2005 ) further test the efficiency of the diagram . for ngc 1068
, we obtained 0.7 @xmath31 h@[email protected]/br@xmath14 @xmath31 1.3 and 0.6 @xmath42 [ feii]@xmath111.257 /pa@xmath12 @xmath42 2.8 .
our results are shown in figure 12 ( filled circles ) together with those observed by rodrguez - ardila 2004 and 2005 ( crosses ) ; the dashed lines show the limits suggested by these authors .
the observational data for ngc 1068 corresponding to the nucleus , as well as to different locations along the slit , are in the agn region of the diagram .
this reinforces the results found previously , i.e. , that this diagram is a helpful tool for classifying objects according to their degree of excitation , and shows that line ratio intensities from slits at different distances from the center can be also used .
in addition , the [ feii]@xmath111.257/[pii]@xmath111.188 line ratio can be used as a tracer of the ionization mechanism of the emitting gas and depends only on the fe / p abundance . following oliva et al .
( 2001 ) , in photoionized gas , like hii regions and planetary nebulae , this line ratio is less than 2 , while in shock dominated gas its value is higher than 20 . the high value would be due to the evaporation of the iron - based grains in the shock front , increasing the fe abundance relative to p in the low - ionization region .
our data show 0.70 @xmath42 [ fe ii]/[p ii ] @xmath42 3.0 .
such low values might indicate that , either like we mention in section 3.1 , the high value of [ pii ] observed here might be due to an overabundance of phosphorus , or most of the fe is locked in grains and the emitting gas is photoionized .
2.121/br@xmath14 versus [ fe ii ] 1.257 /pa@xmath12 .
filled circles represent our data for ngc 1068 .
crosses to data obtained by rodrguez - ardila et al .
( 2004 , 2005).,width=328 ]
in the last few years , photoionization models have been largely used to analyze the physical conditions of the nlr of ngc 1068 . some of these studies tried to reproduce the observed emission - line intensities in a wide wavelength range ( alexander et al .
2000 , kraemer & crenshaw 2000a , kraemer & crenshaw 2000b , spinoglio et al . 2005 ) , while others were based on selected lines ( groves et al .
2004 , groves et al . 2006 , martins , viegas & gruenwald 2003 , rodrguez - ardila et al .
recently , osaki ( 2009 ) focus his analysis on the kinematic and excitation structure of the nlr of ngc 1068 .
although the basic photoionization codes used in the literature are very similar , the choice of the input parameters by different authors may change - in particular , the choice of the filling factor . since one - dimensional photoionization codes are unable to properly deal with an inhomogeneous gas distribution , the optical depth at each cloud shell is multiplied by a chosen filling factor ( see , for instance , osterbrock 1989 ) .
the net effect is to increase the size of the low ionized zone as well as to decrease the average temperature of the cloud , which could mislead the conclusions . following martins et al .
( 2003 ) , in the present paper we adopt a multicloud model which should provide a better simulation of the inhomogeneous nlr of ngc 1068 . another important issue concerning photoionization models is the use of the so - called ionization parameter u to characterize the intensity of the ionizing radiation continuum . in fact , u is the ratio of the density of the ionizing photons reaching the cloud to the gas density .
it is a convenient free parameter when using diagnostic diagrams to analyze emission - line ratios coming from different galaxies .
however , in photoionization models for a given nlr where the distance d from the ionizing radiation source as well as the source luminosity are known , u is not a free parameter anymore but it is constrained by the source luminosity , the distance from the ionizing radiation source and the gas density ( ferguson et al . 1997 ) . as shown in tables 1 , 2 and 3 ,
the main emission line intensities were obtained as a function of the distance to the center .
these measured line intensities result from the contribution of all the clouds in the line of sight covered by each aperture .
it is not in the scope of this paper to provide a specific model for the observed data from each aperture .
our objective here is to give a scenario to try to explain the observations in general .
thus , in the following , we search for photoionization models which can reproduce the observed range of intensities .
first attempts to model the emission - line region with a homogeneous gas photoionized by the nuclear radiation failed to reproduce the observed nir line intensities .
as shown below , at least two density regions are necessary to explain the observed data . in the following ,
we present a two - cloud model for the nir emission - lines using the photoionization code aangaba ( gruenwald & viegas 1992 ) . the model must reproduce the observed coronal lines [ six]@xmath111.252 , [ sviii]@xmath110.991 , [ six]@xmath111.430 , and [ s vi]@xmath111.962 , as well as the [ siii]@xmath430.907 , 0.953 , and [ sii]@xmath431.032 emission lines ( tables 1 , 2 and 3 ) .
the sulfur line intensity ratios are not directly dependent on the adopted chemical abundance .
thus , they are first used to limit the other input parameters , i. e. , u ( ionization parameter ) , n ( density ) and d ( distance from the ionizing source ) , given the ionizing radiation luminosity for ngc 1068 , l@xmath44 .
the model results are then compared to other line ratios , including those for [ fexi]@xmath110.7889 , [ fex]@xmath110.6374 , [ fevii]@xmath110.6087 ( rodrguez - ardila et al .
we assume a solar chemical abundance ( grevesse & anders 1989 ) . for ngc 1068
, the ionizing radiation luminosity is about 10@xmath45 erg / s assuming an uv power - law index equal to -1.7 and an x - ray index of -0.5 ( pier et al . 1994 , if the reflection factor is 0.01 ) . in order to fit the high - ionization lines ,
a large value of u is usually required . however , due to the relationship between u , n , r and l@xmath44 , a high u requires a low n , and a low distance from the ionizing source . in the models presented by the different authors referred above , the density value varies from 10@xmath46 @xmath47 ( groves et al .
2006 ) to 10@xmath48 @xmath47 ( ozaki 2009 ) , although some of them use filling factors as low as 6.5@xmath4910@xmath50 . from the observed coronal emission line intensities we note that for gas clouds extending to 200 pc , values of u @xmath51 0 are required to fit high - ionization lines , and could be reached only for densities lower than 10@xmath46 @xmath47 ( rodrguez - ardila et al .
2006 , figure 12 ) . after running a series of models for a homogeneous low density gas around the ionizing radiation source
, we selected n = 1 @xmath47 and u = 10 to better fitting the coronal lines ; such a model is called model a. the line intensities relative to h@xmath12 obtained for model a , for an optically thick cloud ( @xmath52 ) , are given on the second column of table 6 .
however , such a model also produces a large semi - ionized region , resulting in a [ siii]/[s ii ] line ratio too high .
thus the final model for the low density gas should be matter bound ( see below ) . while the coronal line ratios are more sensitive to the u parameter , the [ siii]/[sii ] line intensity ratio is more dependent on the gas density .
the observed values indicate that densities 10@xmath48 @xmath47 or higher are required to reproduced the observed line ratio .
notice that high density clouds ( 10@xmath53 @xmath47 ) are also favored by ozaki ( 2009 ) when modeling the structure of the nlr of ngc 1068 using the optical emission - lines [ fevii ] , [ oiii ] , and [ oi ] , relative to h@xmath54 .
thus , we assume that high density clouds ( 10@xmath48 @xmath47 ) are radiation bound , located farther from the central source and being shielded from the uv , but not from the x - ray , by the low density gas producing the coronal lines .
the results for a model using an ionizing spectrum with e @xmath55 200 ev are shown in table 6 ( model b ) .
in addition to [ siii ] and [ sii ] lines , this high density gas produces also the [ sivi ] emission line , but not the other coronal lines as expected .
the high density clouds contribute mainly to the low ionization emission lines . in order to obtain the fraction f of the contribution of the high density clouds relative to the coronal gas
, we use the [ sviii]/[siii ] observed line ratios ( table 1 , 2 and 3 ) . since [ siii ] comes both from models a and b , such line ratio depends on the optical depth at the lyman limit of the low density region .
as noticed above , the [ sivi ] line could partially come from the high density clouds .
however their contribution is much smaller than that from the low density region , so that the [ six]/[siiv ] ratio is practically independent of f. as said above , the low density cloud must be matter bound for not emitting too much [ sii ] or [ siii ] .
the best results are obtained with @xmath56 for model a between 1 and 5 . in table 6 , the line intensity ratios for @xmath56 = 5 and 1
are listed ( third and fourth columns , respectively ) .
for @xmath56 of model a equal 1 and 5 , the corresponding fractions f are equal to 0.078 and 0.065 , respectively , leading to a luminosity ratio of h@xmath54 between the two components equal to 10.2 and 38.7 .
the range of the observed infrared line ratios is shown in table 7 , compared to the results of the two possible two - cloud models , i. e. , the combination of model b with model a ( for @xmath56 = 1 and 5 ) .
notice that detailed models for each aperture may require a range of densities .
lcccc & mod a & mod a & mod a & mod b + & @xmath52 & @xmath56=5 & @xmath56=1 & @xmath57 + @xmath19s ii@xmath20/h\beta $ ] & 3.5e-3 & 6.2e-5 & 9.2e-6 & 7.6e+0 + @xmath19s iii@xmath20/h\beta$ ] & 9.1e-1 & 4.2e-1 & 9.1e-2 & 7.1e+1 + @xmath19s viii@xmath20/h\beta$ ] & 3.2e-2 & 4.5e-2 & 1.4e-1 & 0.0e+0 + @xmath19s ix@xmath20/h\beta$ ] & 7.9e-2 & 1.1e-1 & 3.5e-1 & 0.0e+0 + @xmath19si vi@xmath20/h\beta$ ] & 6.6e-2 & 9.1e-2 & 2.5e-1 & 4.2e-5 + @xmath19si x@xmath20/h\beta $ ] & 2.1e-2 & 2.9e-2 & 9.0e-2 & 0.0e+0 + @xmath19fe vii@xmath20/h\beta$ ] & 1.1e-1 & 1.5e-1 & 4.5e-1 & 0.0e+0 + @xmath19fe x@xmath20/h\beta $ ] & 4.9e-2 & 6.8e-2 & 2.2e-1 & 0.0e+0 + @xmath19fe xi@xmath20/h\beta $ ] & 2.2e-2 & 3.1e-2 & 9.7e-2 & 0.0e+0 + lccc & observation & multi - cloud model & + & & @xmath56=1 & @xmath56=5 + @xmath19s iii@xmath20$]/s@xmath19 ii@xmath20 $ ] & + 0.8 to + 1.2 & + 1.0 & + 1.1 + @xmath19s viii@xmath20$]/@xmath19s iii@xmath20 $ ] & -2.0 to -1.5 & -1.7 & -1.7 + @xmath19s ix@xmath20$]/@xmath19s viii@xmath20 $ ] & + 0.0 to + 0.7 & + 0.4 & + 0.4 + @xmath19si x@xmath20$]/@xmath19si vi@xmath20 $ ] & -0.4 to + 0.0 & -0.4 & -0.5 + @xmath19fe x@xmath20$]/@xmath19fe vii@xmath20 $ ] & -0.5 to -0 4 & -0.3 & -0.3 + @xmath19fe xi@xmath20$]/@xmath19fe x@xmath20 $ ] & -0.9 to -0.2 & -0.4 & -0.4 + all line ratios are in the range of the observational data ( tables 1 , 2 and 3 ) , including the coronal iron lines from rodrguez - ardila et al .
( 2006 ) . in that paper , only models with density @xmath55 10@xmath58 @xmath47 were used , and pure photoionization models could not reproduce the line ratios . here
we relax this lower limit , and most of the ratios could be reproduced .
this , however , does not discard the possibility of having other ionization mechanisms playing a role . to put a stronger constrain on this issue
it would be necessary to fit a large amount of emission - lines in a broader range of wavelengths , which is out of the scope of this paper .
we have used the stellar absorption features to estimate the mass of the central region inside the extraction apertures .
the velocity dispersion was estimated using the same velocity distribution and methodology proposed by van der marel & franx ( 1993 ) .
the @xmath59 and @xmath60 coefficients , which measure the distortions from a gaussian distribution , are very uncertain and therefore we restrict our discussion to the velocity dispersions . the results were obtained using as template a mixture of standard stellar spectra with the same resolution as our nuclear spectra of ngc1068 and high s / n .
the spectral region was the same used by tamura et al .
( 1991 ) for this same galaxy , around 2.26 @xmath7 , where there are strong molecular absorptions due
to co. this region seems adequate since the interstellar absorption is small and there are no emission lines .
errors in the velocity dispersions were obtained using the s / n of the spectra .
an example of the fit is shown in figure 13 .
the results are shown in figure 14 and we can observe that the velocity dispersion rises strongly in the central region , probably due to the agn contribution . moreover in this very central region the light contribution of the agn and the hot dust contaminates the stellar continuum and makes it difficult to measure the velocity dispersion . away from the center
the nuclear contamination is much lower in our extracted spectra and we observe that the velocity dispersion remains quite large for a typical spiral galaxy . nevertheless moving outwards to 100 pc from the nucleus the velocity dispersion is quite close to the values already reported in the literature . using these determinations and the virial theorem we give a very crude estimate showing how the central mass depends on the adopted nuclear distance .
our approach consists in using a simple model that can be resolved analytically and compared to the data .
basically the model consists of an stellar core with uniform density , @xmath61 , extending to a maximum radial distance @xmath62 .
the total mass of this stellar core is @xmath63 and in the central region we add a punctual mass , m@xmath64 , that simulates the presence of the agn and its accretion disk .
the total mass profile can be readily estimated as @xmath65 where @xmath66 represents the radial distance .
assuming that the velocity distribution of the stellar component is random and isotropic we derive that @xmath67 where @xmath68(r ) is the velocity dispersion and @xmath14 = m@xmath64/m@xmath69 is a measure of the relative importance of the central mass compared with the core stellar distribution .
however we only have access to the average kinetic energy of the stars projected in the line of sight of the observer . weighting this component by its luminosity contribution , or by mass for a constant mass - luminosity ratio
, we have @xmath70 and assuming that the density of the stellar core is uniform , we finally obtain our estimation for the observed line of sight velocity dispersion as @xmath71\end{aligned}\ ] ] for large values of @xmath14 the central mass dominates and we simply have a keplerian profile which is shown as the blue continuous curve in figure 14 .
we can see that the observed profile is completely inconsistent using the usual bh mass estimation ( @xmath72 ) found in the literature ( lodato & bertin 2002 ) .
to explain the high velocity dispersion of 255 km / s in the central region we need to include an stellar core of @xmath73 extending to approximately 200 pc from the nucleus shown as the dotted line . therefore a value @xmath74 would be a more appropriated contribution of the central mass
we carried out long - slit spectroscopy in the near - ir of ngc 1068 and presented here for the first time the whole range from 0.9 to 2.4@xmath7 , extending 15 " in the nuclear region .
the nuclear emission shows many important emission lines , from low ionization to coronal lines , molecular lines and absorption features .
we show that many emission lines are composed by two narrow components , one in the galaxy rest frame , and the other blueshifted in the south and redshifted in the north apertures .
the relative intensity of the peaks also varies and they seem to be correlated with the jet of the galaxy .
no broad component ( @xmath5 2000 ) was identified in any line profile .
not all emission lines have double peaked profiles and some lines are present only near the center , particularly the high ionization lines .
this is a clear indication that the emitting gas is more ionized in the nuclear region .
we estimated the extinction based on paschen hydrogen and [ feii ] emission lines .
e(b - v ) is highest in the nucleus then in the adjacent regions , but there are peaks around 150 - 200 pc from each side of the nucleus .
this is also close to the region where the second components of the emission lines have a maximum , and may be associated with dense clouds around these regions .
emission is very extended , but it is difficult to identify the ionization process because only the 1 - 0(s1 ) line could be measured .
simple photoionization models were used to try to explain the emission lines , and a first approximation shows that the line ratios can be explained with pure photoionization , as long as low density clouds are present .
a more detailed analysis however is necessary , to rule out other ionization mechanisms .
we also used the stellar signatures to determine the mass of the central region , obtaining a stellar core of 1.5@xmath4910@xmath75m@xmath40 .
this research has been partially supported by the brazilian agency fapesp ( 2007/04316 - 1 ) .
ara acknowledges partial support of cnpq through grant 308877/2009 - 8 .
the authors would like to thank dr .
sueli viegas for the helpful insights and comments on the paper .
we also thank the anonymous referee for valuable comments .
the irs was a collaborative venture between cornell university and ball aerospace corporation funded by nasa through the jet propulsion laboratory and ames research center .
alexander , t. , lutz , d. , sturm , e. , genzel , r. , sternverg , a. , & netzer , h. 2000 , apj , 536 , 710 alloin , d. , pelat , d. , boksenberg , a. , & sargent , w.l.w .
1983 , apj , 275 , 493 alloin , d. , pantin , e. , lagage , p.o . , & granato , g.l .
2000 , a&a , 363 , 926 antonucci , r.r.j , & miller , j.s .
1985 , 297 , 621 axon , d.j . ,
marconi , a. , capetti , a. , macchetto , f.d . ,
schreier , e. , & robinson , a. 1998 , apj , 496 , l75 bautista , m.a .
& pradhan , a.k .
1998 , apj , 492 , 650 bicknell , g. , dopita , m.a . ,
tsvetanov , a.i . , & sutherland , r.s .
1998 , apj , 495 , 680 cardelli , j.a . ,
clayton , g.c . , mathis , j.s .
1989 , apj , 345 , 245 cecil , g. , bland , j. , & tully , r.b .
1990 , apj , 355 , 70 cecil , g. , dopita , m.a . , groves , b. , wilson , a.s .
, ferruit , p. , pcontal , e. , & binnet , l. 2002 , apj , 568 , 627 cid fernandes , r. , heckman , t. , schmitt , h. , gonzlez delgado , r. , & storchi - bergmann , t. 2001 , apj , 558 , 81 crenshaw , d.m .
& kraemer , s.b .
2000a , apj , 532 , 247 crenshaw , d.m .
& kraemer s.b . 2000b , apj , 532 , l10 cushing , m.c . ,
vacca , w.d . , & rayner , j.t .
2004 , pasp , 116 , 362 dopita , m.a . & sutherland , r.s .
1996 , apjs , 102 , 161 evans , i.n . ,
ford , h.c . ,
kinney , a.l .
, et al . 1991 , apj , 369 , l27 ferguson , j.w . ,
korista , k. t. , & ferland , g. j. 1997 , apjs , 110 , 287 galliano , e. , alloin , d. , granato , g.l . ,
& villar - martn , m. 2003 , a&a , 412 , 615 gallimore , j.f . ,
baum , s.a . ,
odea , c.p . , & pedlar , a. 1996 , apj , 458 , 136 grevesse , n. , & anders , e. 1989 , in aip conf .
183 , cosmic abundances of matter ( new york : aip ) , 1 groves , b.a . ,
cecil , g. , ferruit , p. , & dopita , m.a .
2004 , apj , 611 , 786 groves , b. , dopita , m. , & sutherland , r. 2006 , a&a , 458 , 405 gruenwald , r. , & viegas , s. m. 1992 , apjs , 78 , 153 higdon , s.j.u . , devost , d. , higdon , j.l . , et al .
2004 , pasp , 116 , 975 ivanov , v.d . ,
rieke , g.h . ,
groppi , c.e . , alonso - herrero , a. , rieke , m.j . ,
& engelbrach , c.w .
2000 , apj , 545 , 190 kraemer , s. b. , & crenshaw , d. m. 2000a , apj , 532 , 256 kraemer , s. b. , & crenshaw , d. m. 2000b , apj , 544 , 763 kraemer , s.b . ,
ruiz , j.r .
, & crenshaw , d.m .
1998 , apj , 508 , 232 larkin , j. e. , armus , l. , knop , r. a. , soifer , b. t. , & matthews , k. 1998 , apjs , 114 , 59 lodato & bertin 2002 macchetto , f. , capetti , a. , sparks , w.b . ,
axon , d.j . ,
boksenberg , a. 1994 , apj , 435 , 15 martins , l.p . ,
viegas , s.m . & gruenwald , r. 2003 martins , l.p .
, riffel , r. , ardila , a. , gruenwald , r. , de souza , r. 2010 , submitted to mnras mazzalay , x. & rodrguez - ardila , a. 2007 , a&a , 463 , 445 meaburn , j. & pedlar , a. 1986 , a&a , 159 , 336 myers , s.t . , & scoville , n.z .
1987 , apj , 312 , 39 nussbaumer , h. & storey , p.j .
1988 , a&a , 193 , 327 oliva et al .
2001 , a&a , 369 , l5 origlia , l. , moorwood , a.f.m . , oliva , e. 1993 , a&a , 280 , 5360 osaki , s. 2009 , pasj , 61 , 259 osterbrock , d.e .
1989 , astrophysics of gaseous nebulae and active galactic nuclei ( mill valley : university science books ) pelat , d. & alloin , d. 1980 , a&a , 81 , 172 pier , e.a .
, antonucci , r. , hurt , t. , kriss , g. , & krolik , j. 1994 , apj , 428 , 124 pogge , r.w .
1988 , apj , 328 , 519 rayner , j.t .
, toomey , d.w . ,
onaka , p.m. , denault , a.j . ,
stahlberger , w.e . ,
vacca , w.d . ,
cushing , m.c .
, & wang , s. 2003 , pasp , 155 , 362 reunanen , j. , kotilainen , j.k . , & prieto , m. a. 2003 , mnras , 343 , 192 rodrguez - ardila , a. , pastoriza , m.g . ,
viegas , s. , sigut , t.a.a . , & pradhan , a.k .
2004 , a&a , 425 , 457 rodrguez - ardila , a. , riffel , r. , & pastoriza , m. g. 2005 , mnras , 364 , 1041 rodrguez - ardila , a. , prieto , m. a. , viegas , s. , & gruenwald , r. 2006 , apj , 653 , 1098 schinnerer , e. , eckart , a. , & tacconi , l.j .
1998 , apj , 500 , 147 schreiber , a.w .
scoville , n.z . , hall , d.n.b . ,
keinmann , s.g . , & ridgway , s.t .
1982 , apj , 253 , 136 spinoglio , l. , malkan , m. a. , smith , h. a. , gonzalez - alfonso , e. , & fischer , j. 2005 , apj , 623 , 123 tacconi , l.j . ,
genzel , r. , blietz , m. , cameron , m. , harris , a.i . , madden , s. 1994 , apj , 426 , 77 tamura , m. , kleinmann , s.g . ,
scoville , n.z . , &
joyce , r.r .
1991 , apj , 371 , 131 telesco , c.m . , & decher , r. 1988 , apj , 334 , 573 thatte , n. , quirrenbach , r. , maiolino , r. , & tecza , m. 1997 , apj , 490 , 238 tully 1988 turner , j. , kirby - docken , k. , & dalgarno , a. 1977 , apjs , 35 , 281 unger , s.w . ,
lewis , j.r . ,
pedlar , a. , axon , d.j .
1992 , mnras , 258 , 371 vacca , w.d . ,
cushing , m.c . & rayner , j.t .
2003 , pasp , 115 , 389 van der marel , r.p . & franx , m. 1993 , apj , 407 , 525 veilleux , s. , goodrich , r.w . , & hill , g.j .
1997 , apj , 477 , 631 wilson , a.s . &
ulvestad , j.s .
1983 , apj , 275 , 8 zheng , w. , wang , j - x . ,
kriss , g.a . ,
sahnow , d. , allen , m. , dopita , m. , tsvetanov , z. , bicknell , g. 2008 , apj , 686 , 881 | we report the first simultaneous zjhk spectroscopy on the archetypical seyfert 2 galaxy ngc 1068 covering the wavelength region 0.9 to 2.4@xmath0 m .
the slit , aligned in the ns direction and centred in the optical nucleus , maps a region 300 pc in radius at sub - arcsec resolution , with a spectral resolving power of 360 km s@xmath1 .
this configuration allow us to study the physical properties of the nuclear gas including that of the north side of the ionization cone , map the strong excess of continuum emission in the k - band and attributed to dust and study the variations , both in flux and profile , in the emission lines .
our results show that ( 1 ) mid- to low - ionization emission lines are splitted into two components , whose relative strengths vary with the position along the slit and seem to be correlated with the jet . ( 2 ) the coronal lines are single - peaked and are detected only in the central few hundred of parsecs from the nucleus .
( 3 ) the absorption lines indicate the presence of intermediate age stellar population , which might be a signicant contributor to the continuum in the nir spectra . ( 4 ) through some simple photoionization models we find photoionization as the main mechanism powering the emitting gas . ( 5 ) calculations using stellar features point to a mass concentration inside the 100 - 200 pc of about 10@xmath2 .
[ firstpage ] galaxies : active - galaxies : individual ( ngc 1068 ) - galaxies : seyfert - infrared : galaxies |
the parrondo effect refers to a reversal in direction of some system parameter when two similar dynamics are combined .
it was first described by j. m. r. parrondo in 1996 in the context of games of chance : he showed that there exist two losing games that can be combined to win .
the games were originally intended as a pedagogical model of the flashing brownian ratchet .
early work focussed on capital - dependent ( harmer and abbott @xcite ) and history - dependent ( parrondo , harmer , and abbott @xcite ) games for a single player .
multi - player games were introduced by toral @xcite , including games with spatial dependence and games with redistribution of wealth .
toral s @xcite parrondo games with one - dimensional spatial dependence rely on an integer @xmath4 and three probability parameters , @xmath5 , @xmath6 , and @xmath7 .
there are @xmath0 players labeled from 1 to @xmath0 and arranged in a circle in clockwise order . at each turn
, one player is chosen at random to play . in game @xmath8
, he tosses a @xmath9-coin if @xmath10 of his two nearest neighbors are winners ( @xmath11 ) . a player s status as winner or loser depends on the result of his most recent game .
the player wins one unit with heads and loses one unit with tails .
the game can be initialized arbitrarily .
game @xmath12 is the special case of game @xmath8 in which @xmath13 . for @xmath14 , game @xmath15 ,
often denoted by @xmath16 , is a random mixture of games @xmath12 and @xmath8 .
( at each turn a coin with @xmath17 is tossed , and game @xmath12 is played if heads appears , game @xmath8 if tails . ) we let @xmath18 , @xmath19 , and @xmath20 be the equilibrium mean profits per turn ( to the ensemble of @xmath0 players ) in games @xmath12 , @xmath8 , and @xmath15 . of course , @xmath21 .
we say that the _ parrondo effect _ occurs if @xmath22 and @xmath23 ( two fair or losing games combine to win ) and the _ anti - parrondo effect _ occurs if @xmath24 and @xmath25 ( two fair or winning games combine to lose ) .
mihailovi and rajkovi @xcite modeled the three games by introducing markov chains in a state space with @xmath2 states .
they were able to study the games analytically for @xmath26 .
ethier and lee @xcite used a state space reduction method that allowed them to study the games analytically for @xmath27 , an improvement that led to the conjecture that the mean profits @xmath19 and @xmath20 converge as @xmath28 .
( this was subsequently proved in @xcite under certain conditions . )
xie et al .
@xcite introduced a more spatially dependent version of game @xmath12 , which we will refer to as game @xmath29 .
this game amounts to a loss of one unit by a randomly chosen player together with a win of one unit by a randomly chosen nearest neighbor of that player .
( this differs from the description in @xcite but it is probabilistically equivalent . )
li et al .
@xcite used a state space reduction method to study games @xmath29 , @xmath8 , and @xmath30 analytically . however , their method results in markov chains that are not directly related to the model of xie et al .
let us explain what we mean by state space reduction , and what is required for it to be successful . in general
, consider an equivalence relation @xmath31 on a finite set @xmath32 . by definition ,
@xmath31 is _ reflexive _ ( @xmath33 ) , _ symmetric _ ( @xmath34 implies @xmath35 ) , and _ transitive _ ( @xmath34 and @xmath36 imply @xmath37 ) .
it is well known that an equivalence relation partitions the set @xmath32 into _ equivalence classes_. the set of all equivalence classes , called the _ quotient set _ , will be denoted by @xmath38 .
let us write @xmath39:=\{y\in e : y\sim x\}$ ] for the equivalence class containing @xmath40 .
then @xmath41:x\in e\}$ ] . now suppose @xmath42 is a ( time - homogeneous ) markov chain in @xmath32 with transition matrix @xmath43 .
in particular , @xmath44 for all @xmath45 and @xmath46 . under what conditions on @xmath43 is @xmath47,[x_1],[x_2],\ldots$ ]
a markov chain in the `` reduced '' state space @xmath38 ?
a sufficient condition , apparently due to kemeny and snell @xcite , is that @xmath43 be _ lumpable _ with respect to @xmath31 . by definition
, this means that , for all @xmath48 , @xmath49}p(x , y')=\sum_{y'\in[y]}p(x',y').\ ] ] moreover , if holds , then the markov chain @xmath47,[x_1],[x_2],\ldots$ ] in @xmath38 has transition matrix @xmath50 given by @xmath51,[y]):=\sum_{y'\in[y]}p(x , y').\ ] ] notice that ensures that is well defined . for parrondo games with one - dimensional spatial dependence , the state space , assuming @xmath4 players ,
is @xmath52 which has @xmath2 states .
a state @xmath53 describes the status of each of the @xmath0 players , 0 for losers and 1 for winners .
we can also think of @xmath32 as the set of @xmath0-bit binary representations of the integers @xmath54 , thereby giving a natural ordering to the vectors in @xmath32 .
ethier and lee @xcite used the following equivalence relation on @xmath32 : @xmath55 if and only if @xmath56 for a permutation @xmath57 of @xmath58 belonging to the dihedral group @xmath59 of order @xmath60 generated by the rotations and reflections of the players . they verified the lumpability condition , with the result that the size of the state space was reduced by a factor of nearly @xmath60 for large @xmath0 .
it should be noted that a sufficient condition for the lumpability condition in this setting is @xmath61 or for all @xmath57 in a subset of @xmath59 that generates @xmath59 .
li et al .
@xcite reduced the state space much further by effectively using the following equivalence relation on @xmath32 : @xmath55 if and only if @xmath62 for some permutation @xmath57 of @xmath58 .
equivalently , @xmath55 if and only if @xmath63 and @xmath64 have the same number of 1s
. however , here the lumpability condition fails .
( this is intuitively clear because the equivalence relation does not respect the spatial structure of the players . )
thus , the markov chains in @xmath38 constructed in @xcite , which are based on @xmath65,[y]):=\frac{1}{|[x]|}\sum_{x'\in[x]}\bigg[\sum_{y'\in[y]}p(x',y')\bigg]\ ] ] rather than , are not directly related to the model of xie et al . @xcite . to put it another way ,
there is no theoretical basis for such markov chains and the results derived from them .
our aim in this paper is not simply to point out a weakness in earlier work , but to provide accurate results as well for the model of toral @xcite as modified by xie et al .
@xcite .
actually , as originally formulated , the model had four probability parameters @xmath5 , @xmath6 , @xmath7 , and @xmath66 corresponding to the various configurations of the two nearest neighbors , namely @xmath67 , @xmath68 , @xmath69 , and @xmath70 . however , we assume throughout that @xmath71 ( and we relabel the remaining parameters @xmath72 as @xmath73 ) . there are three reasons for doing this .
first , the parrondo region is a subset of the parameter space , and a three - dimensional region is easier to visualize than a four - dimensional one .
second , the equivalence relation of ethier and lee @xcite described above requires this assumption ( otherwise the dihedral group of order @xmath60 would have to be replaced by a cyclic group of order @xmath0 and the state space reduction would be less effective ) .
third , all previous computational works on parrondo games with one - dimensional spatial dependence @xcite have made this assumption . the smallest number of players for which there is a distinction between the equivalence relation of @xcite and that of @xcite is @xmath74 , so we consider that case first in section [ n=4 ] .
we treat the general case by the method of @xcite in section [ el_method ] and by the method of @xcite in section [ li_method ] . in section [ parrondo ] we use computer graphics to sketch the parrondo and anti - parrondo regions for @xmath1 as well as their approximations based on the methods of li et al .
finally , section [ conclusions ] summarizes the main conclusions .
to make this less abstract , let us consider separately the case of four players , that is , @xmath74 .
we begin with game @xmath8 .
the markov chain in @xmath32 for game @xmath8 depends on three parameters , @xmath75 $ ] , so that the parameter space is the unit cube @xmath76 ^ 3 $ ] .
there are 16 states ( namely , the four - bit binary representations of the integers 015 ) and the transition matrix has the form @xmath77 where the diagonal entries are chosen to make the row sums equal to 1 : @xmath78 and @xmath79 for @xmath11 .
this is consistent with eq .
( 12 ) of xie et al .
@xcite . for the equivalence relation of @xcite mentioned above ,
there are six equivalence classes , namely @xmath80 , @xmath81 , @xmath82 , @xmath83 , @xmath84 , and @xmath85
. the lumpability condition is easily verified .
for example , denoting the states by their decimal representations ( 015 ) , the equivalence classes are @xmath86 , @xmath87 , @xmath88 , @xmath89 , @xmath90 , and @xmath91 , and the probabilities of a transition from a state in the equivalence class @xmath88 to the equivalence class itself are @xmath92 which are equal .
moreover , the reduced transition matrix ( with rows and columns labeled by the equivalence classes and ordered as just indicated ) is @xmath93 both matrices , @xmath94 and @xmath95 , can be simplified ( using @xmath96 for @xmath11 ) , but there is a good reason to leave them in this form for now , as we will see . for the equivalence relation of @xcite mentioned above , there are five equivalence classes , namely @xmath80 , @xmath81 , @xmath97 , @xmath84 , and @xmath85
. the lumpability condition fails .
for example , denoting the states by their decimal representations ( 015 ) , the probabilities of a transition from a state in the equivalence class @xmath98 to the equivalence class itself are @xmath99 which are not , in general , equal .
the idea of li et al .
@xcite is to use the average of these probabilities , namely @xmath100 , in @xmath95 ( cf . ):
@xmath101 which is consistent with eq .
( 3 ) of li et al . @xcite .
next we need some notation .
we denote by @xmath102 the matrix @xmath94 with @xmath103 replaced by @xmath104 for @xmath11 .
similarly , we denote by @xmath105 the matrix @xmath95 with @xmath103 replaced by @xmath104 for @xmath11 .
let @xmath106 be the unique stationary distribution of @xmath94 , and let @xmath107 be the unique stationary distribution of @xmath95 .
( we assume that they exist , which rules out certain boundary cases such as @xmath108 , @xmath6 arbitrary , and @xmath109 . ) then , as shown in @xcite using the lumpability property ( see @xcite for more detail ) , the mean profit per turn in game @xmath8 is given by @xmath110 where @xmath111 is the column vector of 1s of the appropriate dimension .
this equation shows that , as far as @xmath19 is concerned , nothing is lost in reducing the state space from @xmath32 to @xmath38 , provided the lumpability condition is satisfied ( as in the case of ) . however , when this condition is not satisfied ( as in the case of ) , the relationship between @xmath94 and @xmath95 is less clear , and in particular the second equation in fails .
straightforward computations yield , for the equivalence relation of @xcite , @xmath112\nonumber\\ & & \;/[3 + 6 p_0 - 2 p_2 - 3 p_0 ^ 2 - 2 p_0 p_2 - p_2 ^ 2 + 12
p_0 ^ 2 p_2 - 4 p_0 p_2 ^ 2 - 8 p_0 ^ 2 p_2 ^ 2\nonumber\\ & & \quad { } - 4 ( 1-p_0 + p_2 + 2 p_0 ^ 2 + 2 p_0 p_2 ) ( 1-p_2 ) p_1\nonumber\\ & & \quad { } + 2 ( 1 + 4 p_0 - p_0 ^ 2 - 2 p_0 p_2 - p_2 ^ 2 ) p_1 ^ 2].\end{aligned}\ ] ] on the other hand , if we use the equivalence relation of @xcite , we can compute the right - hand side of , but it yields only an approximation of @xmath19 , which we denote by @xmath113 : @xmath114\nonumber\\ & & \;/[9 + 24 p_0 - 12 p_2 + 9 p_0 ^ 2 - 32 p_0 p_2 + 3 p_2 ^ 2 - 8 p_0 ^ 2 p_2 + 8 p_0 p_2 ^ 2\nonumber\\ & & \quad{}- 2 ( 6 - p_0 - 7 p_2 - p_0 ^ 2 + p_2 ^ 2 )
p_1 + 4 ( 1 + p_0 - p_2 ) p_1 ^ 2].\end{aligned}\ ] ] the fair surface @xmath115 based on and its approximation @xmath116 based on are graphed in figure [ elvli , b4,c4](a ) .
the difference is significant .
+ -1 cm ( a)2.1in(b ) we turn next to game @xmath29 .
again there are 16 states ( namely , the 4-bit binary representations of the integers 015 ) and the transition matrix has the form @xmath117 the corresponding matrix in eq .
( 3 ) of xie et al .
@xcite has errors at six of the 256 entries .
labeling rows and columns by 015 , the errors occur at entries @xmath118 , @xmath119 , @xmath120 , @xmath121 , @xmath122 , and @xmath123 . using the equivalence relation of @xcite with six equivalence classes , the lumpability condition holds and the reduced transition matrix ( with rows and columns labeled by the equivalence classes and ordered as previously indicated ) is @xmath124 using the equivalence relation of @xcite with five equivalence classes , the lumpability condition fails , but using yields the transition matrix @xmath125 as in eq .
( 1 ) of li et al .
@xcite .
the transition matrices corresponding to game @xmath30 are @xmath126 let @xmath127 be the unique stationary distribution of @xmath128 , and let @xmath129 be the unique stationary distribution of @xmath130 .
since every play of game @xmath29 results in a profit of 0 to the ensemble of four players , @xmath131 and @xmath132 and hence @xmath133 this allows us to evaluate @xmath134 let us assume that @xmath135 .
using the equivalence relation of @xcite , we find from that @xmath136\nonumber\\ & & \ ; { } /[2 ( 315 + 175 p_0 - 125 p_2 - 22 p_0 ^ 2 - 10 p_0 p_2 - 12 p_2 ^ 2 + 48 p_0 ^ 2 p_2 - 32 p_0 p_2 ^ 2 \nonumber\\ & & \quad { } - 16 p_0 ^ 2 p_2 ^ 2 ) - 4 ( 25 - 2 p_0 + 8 p_2 + 17 p_0 ^ 2 + 12 p_0 p_2 - 13 p_2 ^ 2 - 8 p_0 ^ 2 p_2 \nonumber\\ & & \quad { } - 8 p_0 p_2 ^ 2 ) p_1 + 8 ( 14 + 7 p_0 - 3 p_2 - p_0 ^ 2 - 2 p_0 p_2 - p_2 ^ 2 ) p_1 ^ 2],\end{aligned}\ ] ] whereas , using the equivalence relation of @xcite , the right - hand side of yields the approximation @xmath137\nonumber\\ & & \;{}/[2 ( 168 + 137 p_0 - 107 p_2 + 19 p_0 ^ 2 - 80 p_0 p_2 + 13 p_2 ^ 2 - 8 p_0 ^ 2 p_2 + 8 p_0 p_2 ^ 2)\nonumber\\ & & \quad { } - 4 ( 15 - 6 p_0 - 12 p_2 - p_0 ^ 2 + p_2 ^ 2 ) p_1 + 8 ( 3 + p_0 - p_2 ) p_1 ^ 2].\end{aligned}\ ] ] the fair surface @xmath138 based on and its approximation @xmath139 based on are graphed in figure [ elvli , b4,c4](b ) above . the difference is again significant .
we can use the formulas found above to sketch the parrondo and anti - parrondo regions for games @xmath29 , @xmath8 , and @xmath140 .
this is done in figure [ region_n=4 ] .
we notice a curious feature of the regions that did not appear for games @xmath12 , @xmath8 , and @xmath15 .
when projected onto the @xmath141 plane , there is a small apparently circular region in which the parrondo and anti - parrondo regions on opposite sides on the line @xmath142 are inverted .
however , this region is neither circular nor elliptical , but rather is determined by a polynomial of degree 8 in @xmath5 and @xmath7 .
the markov chain formalized by mihailovi and rajkovi @xcite keeps track of the status ( loser or winner , 0 or 1 ) of each of the @xmath4 players of game @xmath8 , which was described in section [ intro ] .
its state space is the product space @xmath143 with @xmath2 states .
let @xmath144 , or , in other words , let @xmath145 be the number of winners among the two nearest neighbors of player @xmath146 when the state is @xmath63 .
of course @xmath147 and @xmath148 because of the circular arrangement of players . also , let @xmath149 be the element of @xmath32 equal to @xmath63 except at the @xmath146th component .
for example , @xmath150 .
the one - step transition matrix @xmath94 for this markov chain depends not only on @xmath0 but on three parameters , @xmath75 $ ] .
it has the form @xmath151 and @xmath152 where @xmath79 for @xmath11 and empty sums are 0 .
the markov chain is irreducible and aperiodic if @xmath153 for @xmath11 . under slightly weaker assumptions ( see @xcite ) , the markov chain is ergodic , which suffices .
for example , if @xmath154 and @xmath153 for @xmath155 , or if @xmath153 for @xmath156 and @xmath157 , then ergodicity holds . the markov chain of xie et al .
@xcite keeps track of the status ( loser or winner , 0 or 1 ) of each of the @xmath4 players of game @xmath29 .
recall that a player is chosen at random , and he then loses one unit while a randomly chosen nearest neighbor of that player wins one unit .
its state space is again @xmath32 . for @xmath158 ,
let @xmath159 be the element of @xmath32 whose @xmath160th component is equal to @xmath161 if @xmath162 , 0 if @xmath163 , and 1 if @xmath164 ; and let @xmath165 be the element of @xmath32 whose @xmath160th component is equal to @xmath161 if @xmath166 , 0 if @xmath163 , and 1 if @xmath167 .
of course @xmath147 and @xmath148 .
for example , @xmath168 .
starting from state @xmath53 , player @xmath146 is chosen with probability @xmath169 , and a transition to state @xmath159 or state @xmath165 occurs , each with probability 1/2 .
the one - step transition matrix @xmath170 for this markov chain in @xmath32 therefore has the form @xmath171,\ ] ] where @xmath172 is the kronecker delta ( @xmath173 if @xmath174 ; @xmath175 otherwise ) .
the markov chain is not irreducible because states @xmath176 and @xmath111 can not be reached , but it does have a unique stationary distribution because it is irreducible when restricted to @xmath177 . using the equivalence relation of ethier and lee @xcite , we can verify the lumpability condition for the transition matrix @xmath94 by checking that @xmath178 where the third equality uses @xmath179 we can verify that holds for @xmath180 and for @xmath181 , which suffices because these two permutations generate @xmath59 . for @xmath170 , we can verify the lumpability condition by observing that , if @xmath180 , then @xmath182\\ & = & ( 2n)^{-1}\sum_{i=1}^n[\delta((\bm x_\sigma)^{\sigma^{-1}(i),-1},\bm y_\sigma)+\delta((\bm x_\sigma)^{\sigma^{-1}(i),1},\bm y_\sigma)]\\ & = & ( 2n)^{-1}\sum_{i=1}^n[\delta((\bm x^{i,-1})_\sigma,\bm y_\sigma)+\delta((\bm x^{i,1})_\sigma,\bm y_\sigma)]\\ & = & ( 2n)^{-1}\sum_{i=1}^n[\delta(\bm x^{i,-1},\bm y)+\delta(\bm x^{i,1},\bm y)]\\ & = & p_{a'}(\bm x,\bm y)\end{aligned}\ ] ] since @xmath183 . if @xmath181 , then the same sequence of identities holds , the only distinction being that @xmath184 .
since the lumpability condition is satisfied for @xmath170 , it is also satisfied for @xmath128 .
a fairly explicit formula for @xmath95 is given in @xcite .
first , define the function @xmath185 by @xmath186):=x_1+x_2+\cdots+x_n$ ] ; it counts the number of 1s in each element of an equivalence class .
then @xmath187,[\bm y])=\begin{cases}n^{-1}\big(\sum_{i : x_i=0}q_{m_i(\bm x)}+\sum_{i : x_i=1}p_{m_i(\bm x)}\big)&\text{if $ [
\bm y]=[\bm x]$}\\ n^{-1}\sum_{i : x_i=1,\bm x^i\sim\bm y}q_{m_i(\bm x)}&\text{if $ s([\bm y])=s([\bm x])-1$}\\
n^{-1}\sum_{i : x_i=0,\bm x^i\sim\bm y}p_{m_i(\bm x)}&\text{if $ s([\bm y])=s([\bm x])+1$}\\ 0&\text{otherwise}\end{cases}\ ] ] for all @xmath188,[\bm y]\in \bar{e}$ ] .
this generalizes .
a formula for @xmath189 is @xmath190,[\bm y])=\begin{cases}(2n)^{-1}\sum_{i=1}^n\big(1_{[\bm y]}(\bm x^{i,-1})+1_{[\bm y]}(\bm x^{i,1})\big)&\text{if $ |s([\bm y])-s([\bm x])|\le1$}\\ 0&\text{otherwise}\end{cases}\ ] ] for all @xmath188,[\bm y]\in \bar{e}$ ] .
this generalizes . as in section [ n=4 ] , let @xmath106 be the unique stationary distribution of @xmath94 , and let @xmath107 be the unique stationary distribution of @xmath95 .
we denote by @xmath102 the matrix @xmath94 with @xmath103 replaced by @xmath104 for @xmath11 .
similarly , we denote by @xmath105 the matrix @xmath95 with @xmath103 replaced by @xmath104 for @xmath11 .
then the mean profit per turn in game @xmath8 is given by @xmath191 the transition matrices corresponding to game @xmath30 are @xmath126 let @xmath127 be the unique stationary distribution of @xmath128 , and let @xmath129 be the unique stationary distribution of @xmath130 .
since every play of game @xmath29 results in a profit of 0 to the ensemble of @xmath0 players , @xmath131 and @xmath132 and hence @xmath133 then the mean profit per turn in game @xmath140 is given by @xmath192 there is a technical issue that must be addressed to fully justify our results .
the strong law of large numbers of ethier and lee @xcite does not apply directly because the payoffs are not completely specified by the one - step transitions of the markov chain .
( for example , a transition from @xmath63 to @xmath63 could correspond to a payoff of @xmath193 or @xmath194 in game @xmath8 ; @xmath193 , @xmath195 , or @xmath194 in game @xmath140 . )
we addressed this issue for game @xmath8 in @xcite by augmenting the state space . a similar approach , but with a different augmentation , works for game @xmath140 .
we let @xmath196 .
the state is @xmath197 if @xmath63 describes the status of each player , @xmath146 is the label of the next player to play , and @xmath198 is the next game to be played ( @xmath199 for game @xmath12 and @xmath200 for game @xmath8 ) .
the new one - step transition matrix @xmath201 has the form @xmath202/2,\end{aligned}\ ] ] where @xmath203 and @xmath204 . for each transition
there is an associated payoff , specifically @xmath205 and @xmath206 .
this allows us to show that @xmath207 using the slln and @xmath208 , from which follows .
assuming the equivalence relation in which two states in @xmath32 are equivalent if they have the same number of 1s , li et al .
@xcite found some rather elegant formulas for the matrices @xmath189 and @xmath95 ( defined using ) . both are tridiagonal matrices with rows and columns indexed by @xmath209 .
specifically , @xmath210 @xmath211 where @xmath212 , and @xmath213 and @xmath214 . this generalizes .
moreover , @xmath215 @xmath216 where @xmath217 and @xmath79 for @xmath11 ; finally , @xmath218 and @xmath219 and @xmath220 .
( defining @xmath221 directly instead of as a complementary probability as in simplifies the definition of @xmath105 .
) this generalizes . since li et al .
@xcite did not include proofs , we sketch a proof of : consider a uniformly distributed sequence of 0s and 1s of length @xmath0 .
then @xmath222\\ & & \quad{}+(1/2)\e[\text{proportion of positions with a 0 followed by a 0}\mid i\text { 1s}]\\ & = & ( 2n)^{-1}\e[\text{number of positions with a 0 preceded by a 0}\mid i\text { 1s}]\\ & & \quad{}+(2n)^{-1}\e[\text{number of positions with a 0 followed by a 0}\mid i\text { 1s}]\\ & = & \p(\text{0s in positions $ n-1 $ and $ n$ } \mid i\text { 1s})\\ & = & \frac{\binom{n-2}{i}}{\binom{n}{i}},\end{aligned}\ ] ] where the third equality uses an exchangeability argument .
the other formulas are proved analogously .
these equations determine @xmath223 .
as before , we denote by @xmath105 the matrix @xmath95 with @xmath103 replaced by @xmath104 for @xmath11 . since every play of game @xmath29 results in a profit of 0 to the ensemble of @xmath0 players , @xmath132 and hence @xmath224 let @xmath107 be the unique stationary distribution of @xmath95 on @xmath38 , and let @xmath129 be the unique stationary distribution of @xmath130 on @xmath38 .
this allows us to evaluate @xmath225 these are the approximations of li et al .
@xcite to the exact @xmath19 and @xmath226 . as in section [ el_method ] ,
the strong law of large numbers in @xcite does not apply directly , so the markov chains in @xmath227 described above must be augmented .
we explain the procedure for game @xmath8 ; the procedure for game @xmath140 is similar .
we let @xmath228 .
the state is @xmath229 if @xmath146 is the number of 1s and @xmath230 is the profit obtained in the transition to state @xmath146 . in ,
notice that @xmath221 is the sum of two fractions , which we denote by @xmath231 and @xmath232 , respectively .
the new one - step transition matrix @xmath233 has the form @xmath234 this allows us to show that @xmath235 using the slln and @xmath236 , from which the first equation in follows .
we can now compare the parrondo regions for games @xmath29 , @xmath8 , and @xmath140 of xie et al .
@xcite with those for games @xmath12 , @xmath8 , and @xmath15 of toral @xcite .
we can also compare the parrondo regions for games @xmath29 , @xmath8 , and @xmath140 computed exactly with those computed using the approximation of li et al .
let us first compare them for certain parameter values .
table [ toral ] assumes that @xmath237 , as did toral @xcite .
we can compute mean profits for @xmath238 . notice that the parrondo effect appears for games @xmath12 , @xmath8 , and @xmath15 , as well as for games @xmath29 , @xmath8 , and @xmath140 , when @xmath239 and @xmath240 .
however , the approximate formulas give a very different conclusion : there is no effect for @xmath241 , an anti - parrondo effect for @xmath242 , and no effect for @xmath243 .
tables [ vector2 ] and [ vector3 ] treat the two other cases studied in @xcite . in both cases
the parrondo effect appears for games @xmath12 , @xmath8 , and @xmath15 as well as for games @xmath29 , @xmath8 , and @xmath140 , at least for @xmath244 , whereas the approximate formulas yield misleading results .
# = # cccccc @xmath0 & @xmath19 & @xmath113 & @xmath20 & @xmath226 & @xmath245 + @@@3 & @xmath246@@ & @xmath246@@ & @xmath247@ & @xmath248@ & @xmath248@ + @@@4 & # 0.0799608@@ & @xmath249@@ & # 0.0171357@ & # 0.0156538@ & @xmath250@ + @@@5 & @xmath251@ & @xmath252@@ & # 0.00405176 & # 0.00565126 & @xmath253@ + @@@6 & @xmath254@@ & @xmath255@@ & # 0.00463310 & # 0.01343312 & @xmath256@ + @@@7 & # 0.00350598@ & @xmath257@ & # 0.00482261 & # 0.00680337 & @xmath258@ + @@@8 & # 0.000698188 & @xmath259@ & # 0.00479021 & # 0.00678290 & @xmath260@ + @@@9 & @xmath261@ & @xmath262@ & # 0.00479036 & # 0.00678314 & @xmath263@ + @@10 & @xmath264 & @xmath265@ & # 0.00479099 & # 0.00678338 & @xmath266 + @@11 & @xmath267 & @xmath268@ & # 0.00479089 & # 0.00678336 & @xmath269 + @@12 & @xmath270 & @xmath271@ & # 0.00479089 & # 0.00678336 & @xmath272 + @@13 & @xmath273 & @xmath274@ & # 0.00479089 & # 0.00678336 & @xmath275 + @@14 & @xmath276 & @xmath277 & # 0.00479089 & # 0.00678336 & @xmath278 + @@15 & @xmath279 & @xmath280 & # 0.00479089 & # 0.00678336 & @xmath281 + @@16 & @xmath282 & # 0.000341368 & # 0.00479089 & # 0.00678336 & @xmath283 + @@17 & @xmath284 & # 0.000760068 & # 0.00479089 & # 0.00678336 & @xmath285 + @@18 & @xmath286 & # 0.00113478@ & # 0.00479089 & # 0.00678336 & @xmath287 + @@19 & @xmath288 & # 0.00147194@ & # 0.00479089 & # 0.00678336 & @xmath289 + @@20 & & # 0.00177683@ & & & @xmath290 + @100 & & # 0.00652292@ & & & # 0.00329682 + @500 & & # 0.00748377@ & & & # 0.00430074 + 2500 & & # 0.00767594@ & & & # 0.00449892 + @=@ # = # cccccc @xmath0 & @xmath19 & @xmath113 & @xmath20 & @xmath226 & @xmath245 + @@@3 & # 0.0710383@ & # 0.0710383 & # 0.0297791@ & # 0.0525560@@ & # 0.0525560 + @@@4 & @xmath291@ & # 0.0485411 & # 0.00241457 & # 0.000952648 & # 0.0363651 + @@@5 & # 0.00257895 & # 0.0398300 & # 0.00818232 & # 0.00765099@ & # 0.0295117 + @@@6 & @xmath292@ & # 0.0349801 & # 0.00721881 & # 0.0136825@@ & # 0.0256872 + @@@7 & @xmath293 & # 0.0318731 & # 0.00736816 & # 0.00691714@ & # 0.0232447 + @@@8 & @xmath294 & # 0.0297097 & # 0.00734464 & # 0.00691038@ & # 0.0215492 + @@@9 & @xmath295 & # 0.0281159 & # 0.00734835 & # 0.00691100@ & # 0.0203035 + @@10 & @xmath296 & # 0.0268928 & # 0.00734776 & # 0.00691094@ & # 0.0193494 + @@11 & @xmath297 & # 0.0259243 & # 0.00734786 & # 0.00691095@ & # 0.0185954 + @@12 & @xmath298 & # 0.0251385 & # 0.00734784 & # 0.00691095@ & # 0.0179843 + @@13 & @xmath299 & # 0.0244881 & # 0.00734784 & # 0.00691095@ & # 0.0174792 + @@14 & @xmath300 & # 0.0239408 & # 0.00734784 & # 0.00691095@ & # 0.0170546 + @@15 & @xmath301 & # 0.0234740 & # 0.00734784 & # 0.00691095@ & # 0.0166927 + @@16 & @xmath302 & # 0.0230711 & # 0.00734784 & # 0.00691095@ & # 0.0163806 + @@17 & @xmath303 & # 0.0227198 & # 0.00734784 & # 0.00691095@ & # 0.0161086 + @@18 & @xmath303 & # 0.0224108 & # 0.00734784 & # 0.00691095@ & # 0.0158696 + @@19 & @xmath303 & # 0.0221369 & # 0.00734784 & # 0.00691095@ & # 0.0156577
+ @@20 & & # 0.0218925 & & & # 0.0154688 + @100 & & # 0.0183996 & & & # 0.0127773 + @500 & & # 0.0177477 & & & # 0.0122767 + 2500 & & # 0.0176191 & & & # 0.0121780 + @=@ # = # cccccc @xmath0 & @xmath19 & @xmath113 & @xmath20 & @xmath226 & @xmath245 + @@@3 & @xmath304@@ & @xmath304@@@ & @xmath305 & # 0.0250737 & # 0.0250737 + @@@4 & @xmath306@ & @xmath307@@@ & # 0.0108365@ & # 0.0175362 & # 0.0217807 + @@@5 & @xmath308@ & @xmath309@@@ & # 0.0141217@ & # 0.0169208 & # 0.0202632 + @@@6 & @xmath310@ & @xmath311@@ & # 0.0147166@ & # 0.0336654 & # 0.0193758 + @@@7 & @xmath312 & @xmath313@@ & # 0.0148223@ & # 0.0168224 & # 0.0187918 + @@@8 & @xmath314 & @xmath315@@ & # 0.0148408@ & # 0.0168213 & # 0.0183779 + @@@9 & @xmath316 & @xmath317@@ & # 0.0148441@ & # 0.0168212 & # 0.0180692 + @@10 & @xmath318 & @xmath319@@ & # 0.0148446@ & # 0.0168212 & # 0.0178301 + @@11 & @xmath320 & @xmath321@@ & # 0.0148447@ & # 0.0168211 & # 0.0176394 + @@12 & @xmath322 & @xmath323@ & # 0.0148447@ & # 0.0168211 & # 0.0174837 + @@13 & @xmath324 & @xmath325 & # 0.0148448@ & # 0.0168211 & # 0.0173543 + @@14 & @xmath326 & # 0.00451782@ & # 0.0148448@ & # 0.0168211 & # 0.0172449 + @@15 & @xmath327 & # 0.00908041@ & # 0.0148448@ & # 0.0168211 & # 0.0171513 + @@16 & @xmath328 & # 0.0130734@@ & # 0.0148448@ & # 0.0168211 & # 0.0170703 + @@17 & @xmath329 & # 0.0165933@@ & # 0.0148448@ & # 0.0168211 & # 0.0169994 + @@18 & @xmath330 & # 0.0197165@@ & # 0.0148448@ & # 0.0168211 & # 0.0169370 + @@19 & @xmath331 & # 0.0225043@@ & # 0.0148448@ & # 0.0168211 & # 0.0168815 + @@20 & & # 0.0250062@@ & & & # 0.0168319 + @100 & & # 0.0610154@@ & & & # 0.0161143 + @500 & & # 0.0674780@@ & & & # 0.0159784 + 2500 & & # 0.0687331@@ & & & # 0.0159515 + another distinction is the rate at which the means
@xmath19 , @xmath20 , and @xmath226 converge .
notice that @xmath20 and @xmath226 converge very rapidly , having stabilized to six significant digits by @xmath332 , 12 , or 13 in each case .
@xmath19 converges a little more slowly .
by @xmath333 , it has stabilized to three significant digits in the first case , six in the second case , and four in the third case . as a consequence ,
it is unnecessary to extend these calculations to larger @xmath0 .
if it were possible , the results would be virtually identical .
contrast that with the situation for the approximate means @xmath113 and @xmath245 of li et al . @xcite .
as one can see from tables [ toral][vector3 ] , the convergence is _ much _ slower .
next , we use computer graphics to sketch , for games @xmath29 , @xmath8 , and @xmath140 , the parrondo and anti - parrondo regions when @xmath1 . see figure [ region_el ] .
we can compare these figures with those of @xcite , for games @xmath12 , @xmath8 , and @xmath15 , when @xmath334 .
( the reason we can go further in the present case than we did in @xcite is that we found a new method for sketching graphs without the need for explicit formulas . )
the figures for games @xmath29 , @xmath8 , and @xmath140 are distinctively different from those for games @xmath12 , @xmath8 , and @xmath15 . in both cases ,
the general shape of the parrondo and anti - parrondo regions does not change much , once @xmath335 . in figure [ region_li ] , we sketch , for games @xmath29 , @xmath8 , and @xmath140 , the approximate parrondo and anti - parrondo regions when @xmath336 and @xmath337 , based on the approximation of li et al . @xcite . we see clearly that the approximation is poor .
+ @xmath338 @xmath74 @xmath339 + + @xmath340 @xmath341 + + @xmath342 @xmath343 + @xmath338 @xmath74 @xmath339 + + @xmath340 @xmath341 @xmath342 + + @xmath343 @xmath344 @xmath337 +
* parrondo games with @xmath4 players and one - dimensional spatial dependence were introduced by toral @xcite and studied analytically by mihailovi and rajkovi @xcite for @xmath26 and by ethier and lee @xcite for @xmath27 .
the additional seven cases in the latter work led to the conjecture that the mean profits converge as @xmath28 , and this was subsequently proved in @xcite under certain conditions .
the reason that larger @xmath0 could be treated was not because of faster computers but because of a state space reduction method that requires for its justification that a lumpability condition be satisfied by the markov chains describing the parrondo games . *
xie et al .
@xcite modified toral s model by changing the nonspatial game @xmath12 to one with spatial dependence .
li et al .
@xcite studied these parrondo games using a state space reduction method that reduces the size of the state space from @xmath2 states to @xmath3 states .
however , the markov chains describing these modified parrondo games fail to satisfy the lumpability condition , with the result that the reduced markov chains are inconsistent with the model of xie et al .
* toral s model as modified by xie et al . @xcite is studied in this paper using the methods introduced in @xcite .
the lumpability condition is satisfied and so the reduced markov chains are consistent with the model of xie at al .
this allows us to compare our exact results for this model with the approximate ones of li et al .
we find that their approximation is poor and that their results are misleading .
this can be seen from tables [ toral][vector3 ] or by comparing figures [ region_el ] and [ region_li ] . the same is true for large @xmath0 as well , as tables [ toral][vector3 ] suggest .
* li et al .
@xcite justified their approximation on the grounds that exact computations are impossible for large @xmath0 .
while that may be true , we would argue that computations for large @xmath0 are unnecessary .
the parrondo region stabilizes rather quickly as @xmath28 .
by @xmath333 , mean profits for the parrondo games under consideration have stabilized to three or more significant digits . although we can not extrapolate with absolute mathematical certainty , it seems safe to do so based on the computations that have been done so far .
see tables [ toral][vector3 ] , for example . | we study toral s parrondo games with @xmath0 players and one - dimensional spatial dependence as modified by xie et al .
specifically , we use computer graphics to sketch the parrondo and anti - parrondo regions for @xmath1 .
our work was motivated by a recent paper of li et al . , who applied a state space reduction method to this model , reducing the number of states from @xmath2 to @xmath3 .
we show that their reduced markov chains are inconsistent with the model of xie et al . _
key words and phrases _ : markov chain , equivalence class , lumpability , dihedral group , parrondo s paradox , cooperative parrondo games . |
quasi - periodic variations were observed in the hard x - ray ( hxr ) emission of many flares (; see also the review of and references therein ) . in our previous papers ( ( paper i ) , @xcite ( paper ii ) ) we investigated quasi - periodic oscillations in flares with periods @xmath2 s , but in paper i we have found three flares with periods @xmath3 s. they turned out to be large arcade flares .
investigation of the quasi - periodic oscillations in such large flares is very important , since their large sizes allow us to investigate the structure of the oscillation volume more comprehensively .
unfortunately , appropriate observations of the x - ray oscillations in such large flares are very rare . in paper
iii @xcite we investigated such a large arcade flare of 2 november 1991 .
its hxr light curve is shown in figure [ fig1]a .
we have found there a direct observational evidence that the strong hxr pulse at 16:34 - 16:35 ut is the result of dense plasma coming from the chromospheric evaporation and flowing into the acceleration volume located within the loop - top cusp structure .
in paper iii we have also found that a. the precipitation of accelerated electrons from the cusp structure is strongly asymmetric , _
i.e. _ there is strong difference in the precipitation into the northern and southern arms ( `` legs '' ) of the flaring loop ; b. there are significant changes of the precipitation with time ; c. the properties of precipitation depend on the energy of accelerated electrons .
it has been shown that these complicated properties of the precipitation can be easily explained in terms of our model of oscillating magnetic traps ( see paper iii ) . in the present paper
we investigate a large arcade flare of 2 march 1993 .
its hxr light curve is shown in figure [ fig1]b .
we see close similarity of the light curves in figures [ fig1]a and [ fig1]b which indicates that there is close similarity in the impulsive - phase development of these two flares .
section [ sec2 ] contains the analysis of observations , section [ sec3 ] presents the discussion of the decay phase of the flare , and summary of the paper is given in section [ sec4 ] .
in the present paper we investigate a large arcade flare which occurred at the eastern limb on 2 march 1993 .
it was a long duration event ( lde ) of goes class m5.1 ( see figure [ fig2 ] ) .
the hxr light curves , recorded by the _
yohkoh hard x - ray telescope _
( hxt ; ) and the _ compton gamma ray observatory burst and transient source experiment _ ( cgro / batse ; ) are shown in figures [ fig3 ] and [ fig4]a , respectively .
the nominal energy range of the batse observations is @xmath4 kev , but in paper iii we have found that actual batse energy range was @xmath5 kev .
the hxrs began to rise at 21:01 ut ( beginning of the impulsive phase ) .
unfortunately , there were no _ yohkoh _ soft - x - ray ( sxr ) observations for the flare impulsive phase .
some sxr images were available only for the late phase of the flare decay ( see section [ sec23 ] ) .
we have investigated the quasi - periodicity of the hxr pulses using our standard method described in papers ii and iii .
we have calculated the normalized time series : @xmath6 where @xmath7 is the measured hxr flux and @xmath8 is the running average of @xmath7 . the red curve in figure [ fig4]a shows @xmath8 calculated with averaging time @xmath9s .
the normalized time series , @xmath10 , is shown in figure [ fig4]b .
next we have measured time intervals , @xmath11 , between successive hxr peaks and calculated the period , @xmath12 , and its standard ( r.m.s . ) deviation , @xmath13 .
our criterion of a quasi - periodicity is @xmath14 .
the values of @xmath15 and @xmath13 are given in table [ tab1 ] .
[ see equation ( 1)].,scaledwidth=100.0% ] cccc & + & + + @xmath15 [ s ] & amp @xmath16 & @xmath15 [ s ] & amp @xmath16 + & & & + 94@xmath178 & 0.49 & 134 & 0.56 + @xmath15 is the mean time - interval between successive hxr peaks , and + amp @xmath16 is the mean value of full amplitude of function @xmath10 + measured for individual hxr peaks .
figures [ fig3 ] and [ fig4 ] , together with table [ tab1 ] , show clear quasi - periodicity of the hxr pulses during the impulsive phase rise ( 21:01 - 21:12 ut ) . table [ tab1 ] also shows that the period @xmath15 is longer near the impulsive phase maximum than during the impulsive phase rise .
we explain this effect as being due to quick increase in the density inside the acceleration volume , which causes significant decrease in the alfv@xmath18n speed , @xmath19 ( see section [ sec3 ] ) .
the light curves in figures [ fig3 ] and [ fig4]a can be divided into two components : the hxr pulses and a `` quasi - smooth '' component ( emission buried beneath the pulses in the figures ) .
the full amplitude , amp @xmath16 , of function @xmath10 [ equation ( 1 ) ] for a hxr pulse is a measure of the ratio of the pulse intensity to the quasi - smooth component ; the mean values of amp @xmath16 are given in table 1 .
our interpretation of the quasi - smooth component has been given in paper ii , where we explained it as a result of superposition of the emission generated by many magnetic traps whose oscillations are shifted in phase .
the sensitivity of the _ yohkoh _ hxt was moderate and the counting rates were low in this flare ( see figure [ fig3 ] ) ; therefore it was necessary to apply rather long integration time in the reconstruction of hxr images .
figure [ fig5 ] shows the 23 - 33 kev image recorded during the maximum of hxr emission .
we see the high flaring loop above the eastern solar limb ( its altitude was about 45 mm ) .
there was a triangular cusp structure , bpc , at the top and strong footpoint sources , f1 and f2 , on the solar disc .
the position of the footpoints indicates that the plane of the flaring loop was tilted from the plane of the image , _
i.e. _ that there is a significant geometrical foreshortening in the north - south dimension of the loop .
we have seen in papers i and iii that the sources b and c were located at the places where the cusp structure was connected with the arcade channel . .
bpc is a triangular cusp structure at the top of flaring loop . f1 and f2 are the flare footpoints on the solar disc.,scaledwidth=80.0% ] figure [ fig6 ] shows a sequence of 23 - 33 kev images .
this long sequence allowed us to investigate asymmetry in electron distribution and its changes in the precipitation in some detail .
between 21:06 and 21:08 ut ( figures [ fig6]a - c ) there was a clear coupling between the top source p and the footpoint f1 . during this time interval
a gradual increase in the intensity of the p source was seen , which is certainly the result of the increase in density due to the chromospheric evaporation from the footpoint f1 .
about 21:10 ut ( figure [ fig6]d ) had begun strong precipitation toward the footpoint f2 .
this caused chromospheric evaporation , inflow of dense plasma into the acceleration volume , increase in the number of accelerated electrons and generation of intense hxr pulses .
this moment of time ( @xmath2021:10 ut ) is analogous to 16:31 ut in the flare of 2 november 1991 ( see paper iii ) when the footpoint source f had appeared ( see figure 6 in that paper ) .
figure [ fig7 ] shows a sequence of the _ yohkoh _ hxt 14 - 23 kev images . in the images the loop - top sources p and b
dominate . in some images ( figures [ fig7]d , [ fig7]f , and [ fig7]g ) these sources are connected ( not resolved ) .
the footpoint sources are conspicuous only in the first three images ( figures [ fig7]o , [ fig7]a , and [ fig7]b ) and during the hxr maximum ( figure [ fig7]e ) .
after 21:17 ut the loop - top source p dominated ( figures [ fig7]i , [ fig7]j , and [ fig7]k ) .
figure [ fig7 ] ( continued ) figures [ fig6 ] and [ fig7 ] show comparison of significant differences in the precipitation of @xmath2015 kev and @xmath2025 kev electrons . according to our model ( see papers i - iii )
the cusp structure was filled with many oscillating magnetic traps . during compression of a trap
its trap ratio , @xmath21 ( @xmath22 is the magnetic field strength at the ends of the trap , @xmath23 is the strength at the middle of the trap ) , decreased and reached its minimum value , @xmath24 , during the maximum of compression .
then the electrons inside the trap reached their maximum energy and the efficiency of their precipitation reached its maximum ( see papers i - iii ) .
the maximum of compression is different in different traps , _ i.e. _ their @xmath24 values are different .
the electrons inside the traps undergoing weaker compression ( higher @xmath24 values ) achieved lower energies ( @xmath2015 kev ) and the efficiency of their precipitation was low ( see figure [ fig7 ] ) .
the electrons within the traps undergoing strong compression ( low @xmath24 values ) achieved higher energies ( @xmath2025 kev ) and they efficiently precipitated toward the footpoints ( figure [ fig6 ] ) .
this means that the ensemble of the traps which was responsible for generation of most of the @xmath2015 kev electrons , was different from the ensemble providing the @xmath2025 kev electrons .
figure [ fig7 ] also shows that most of the @xmath2015 kev electrons emitted their energy and were thermalized within the bpc cusp structure .
the strong source b in these images indicates that the traps which dominated in the generation of @xmath2015 kev electrons had good connection with the arcade channel at b. characteristic features of the hxr impulsive phase are the asymmetry in the precipitation of accelerated electrons from the cusp structure toward the footpoints and changes in the asymmetry with time and with the energy of the electrons ( see papers i - iii ) .
these features are clearly seen in figure [ fig6 ] of the present paper . during the hxr pulses 21:06 - 21:10 ut the precipitation toward the footpoint f1 dominated ( figures [ fig6]a - c ) .
about 21:10 ut ( figure [ fig6]d ) we see the development of the footpoint f2 and about 21:12 ut ( figure [ fig6]e ) the footpoints f1 and f2 had similar intensities . during the hxr maximum ( 21:13 ut , figure [ fig6]f ) the precipitation toward the footpoint f1 again dominated , but during the second strong hxr pulse ( 21:15 ut , figure [ fig6]g ) the precipitation toward the two footpoints was again of similar magnitude . in later images
( after 21:16 ut , figure [ fig6]h ) , the precipitation toward the footpoint f2 dominated .
these complicated variations in the asymmetry of precipitation of accelerated electrons can be adequately explained in terms of our model of oscillating magnetic traps .
a. the asymmetry in the precipitation for an hxr pulse may arise if the axis of symmetry of the magnetic traps ( approximately , it is the line joining source p with the middle of the segment bc ) is not perpendicular to bc .
a small deviation of the axis of symmetry from being perpendicular to bc will introduce small difference in the maximum compression ( @xmath25 ) at the opposite ends of the traps .
these small differences in the @xmath25 values lead to significant differences in the precipitation of electrons , since the efficiency of precipitation steeply depends on @xmath25 ( see section 3 in paper iii ) .
hence , large asymmetry in precipitation may result from moderate deviation of the axis of symmetry of the magnetic traps from being perpendicular to the line bc .
an observation that footpoint intensities f1@xmath26f2 means that @xmath27 in the cusp , f1@xmath20f2 means @xmath28 , and f1@xmath29f2 means @xmath30 .
b. changes in the asymmetry of precipitation ( f1/f2 ratio ) mean that the direction of the axis of symmetry of the traps changes in time . again , small changes in the direction can cause large changes in the asymmetry of precipitation , since small changes in @xmath25 lead to large changes in precipitation .
x - ray images investigated in papers i and iii have shown two important features of the large arcade flares : a. the triangular ( `` cusp '' ) bpc structure was magnetically connected with the arcade channel and therefore the accelerated electrons were able to penetrate into the channel and heated it .
b. the triangular cusp covered only a part of the length of arcade channel , _
i.e. _ the channel was significantly longer than the extension of the cusp measured along the channel ( see an example in figure [ fig8 ] ) .
this indicates that the energy which penetrated into the arcade channel was efficiently transferred along the channel by thermal conduction .
unfortunately , the arcade channel could not be seen during the impulsive phase of the flare of 2 march 1993 , since no sxr images are available for this phase .
there were no sxr observations for the impulsive phase of the investigated flare of 2 march 1993 .
some sxr images , recorded with the thin al.1 filter , were available only for the late decay phase of the flare ( figure [ fig9 ] ) .
we see a long arcade of sxr loops at the altitude of about 95 mm .
the response function of the _ yohkoh _ sxt observations with the al.1 filter weakly depends on temperature in the wide range of temperatures @xmath31 mk @xcite .
therefore all plasma having @xmath32 mk efficiently contributed to the recorded emission .
the distribution of the intensity in the sxr images displays the distribution of the sxr emitting plasma .
figure [ fig10 ] shows the comparison of a sxr arcade image with an impulsive - phase hxr image .
the hxr image was recorded at 21:08 ut , when the arcade channel was at the level of b and c sources ( see figure [ fig5 ] ) .
figure [ fig10 ] shows that the bpc cusp structure covered only a small part of the arcade length ( compare this figure with figure [ fig8 ] ) . in figure [ fig11 ]
we show the time variation of the temperature , @xmath33 , and emission measure , em , derived from the standard goes observations ( figure [ fig2 ] ) . during the quick increase of temperature ( 21:00 - 21:20 ut ; impulsive phase ) the cusp structure and
the arcade channel had been filled with the plasma coming from the chromospheric evaporation .
between 21:20 and 22:00 ut the temperature decreased , but the emission measure continued to increase .
this means that the energy release decreased , but it was sufficiently high to maintain the chromospheric evaporation .
the decay phase of the investigated flare started at about 22:00 ut .
in steady - state loops the following condition is fulfilled : @xmath34 ( @xmath35 is the heating rate per unit volume , @xmath36 is the radiative loss per unit volume , the integrals are over the whole volume of a loop ) , _ i.e. _ the total heating is balanced by the total radiative loss from the loop .
the `` scaling law '' of is an analytical expression of the energy balance [ equation ( 2 ) ] .
it can be written in the following form ( see ): @xmath37 where @xmath38 is the mean electron number density in @xmath39 , @xmath33 is the temperature at the top of the loop in k , and @xmath40 is the semilength of the loop in cm .
it is very important that coronal loops have an efficient mechanism which allows them to satisfy the energy balance condition [ equation ( 2 ) ] , _ i.e. _ to achieve a steady state .
here we briefly describe this mechanism . if @xmath41 then the flux of energy which reaches the footpoints by thermal conduction is large and it generates chromospheric evaporation .
this increases density in the loop and increases @xmath42 which allows the loop to achieve the energy balance [ equation ( 2 ) ] . if , on the other hand , @xmath43 the flux of energy which reaches the footpoints is low , since most of the energy is emitted above the footpoints .
therefore the pressure at the footpoints is too low to balance the weight of plasma contained in the loop and some amount of plasma precipitates to the chromosphere ( this is seen in numerical simulations described in ) .
the density in the loop and @xmath42 decrease and this allows the loop to achieve the energy balance [ equation ( 2 ) ] .
this mechanism of self - regulation of coronal loops is quick in comparison with slow evolution of a flare during the slow decay phase .
for example , numerical simulations have shown that a loop of semilength @xmath44 mm fits a change in heating rate in the time of only 5 min @xcite . in previous papers we investigated flare evolution in @xmath45 vs. @xmath46 or @xmath45 vs. @xmath47 diagnostic diagrams ( see and references therein ) .
we have found that during the slow decay phase flares evolve along the line of steady - state loops , _
i.e. _ the line described by equation ( 3 ) with @xmath48 .
this indicates that during this evolution flares are close to the steady state [ equation ( 2 ) ] with slowly decreasing heating , _
i.e. _ decreasing values of the integrals in equation ( 2 ) ( quasi - steady - state , or qss , evolution of the loops ) .
these results have been supported by numerical simulations of loops with slowly decreasing heating @xcite .
these results can be summarized as follows : a. during the slow decay phase ( after 22:30 ut in figure [ fig11 ] ) the loops seen in figure [ fig9 ] slowly evolved along a sequence of steady states ( qss evolution ) .
b. the loops were continuously heated to support this qss evolution . how this continuous heating of the loops seen in figure [ fig9 ] was performed ?
our proposed explanation is the following : a. magnetic reconnection occurred at the tops of the loops and between the loops and arcade channel .
this provided continuous heating of the loops .
b. new magnetic loops which were generated by the reconnection comprised only a small fraction of the volume of loops seen in figure [ fig9 ] and they quickly achieved the energy balance [ equation ( 2 ) ] due to the self - regulation described above . therefore the reconnecting loops did not disturb much the slow qss evolution of the loops .
this interpretation is supported by the close correlation between the temperature and emission measure during the slow decay phase ( @xmath49 in agreement with equation ( 2 ) and numerical simulations described in ) .
the two main processes of thermal energy loss from a flare kernel are thermal conduction and radiative losses .
the rates of these losses , calculated per unit volume , are ( see ): @xmath50}\quad\ ] ] and @xmath51}\quad\ ] ] where @xmath52 is the radius of the x - ray kernel , @xmath40 is the length of the flaring loop `` legs '' and @xmath53 is the radiative loss function . when the observed temperature changes are slow , _
i.e. _ the values of @xmath54 are small , the loss of energy from the kernel is compensated for by the heating @xmath35 : @xmath55 we have applied equations ( 3)-(6 ) to the beginning of the decay phase ( 22:30 ut in table [ tab2 ] ) and also to the temperature maximum ( 21:18 ut ) , since the change in the heating was slow also then ( small @xmath54 means small @xmath56 ) . in figure [ fig5 ]
we have measured the radius of the cusp bpc structure , @xmath57 mm , and the length of the loop `` legs '' , @xmath58 mm .
using equations ( 3)-(6 ) we have calculated the values of physical parameters which are given in table [ tab2 ] .
.physical parameters for the time interval 21:18 - 22:30 ut [ cols="^,^,^,^,^,^ " , ] @xmath59 this value has been obtained by scaling @xmath60 @xmath39 + ( lower row ) according to the change in @xmath61 , since the scaling + law [ equation ( 3 ) ] is not valid before the decay phase .
table [ tab2 ] shows that the energy release within the cusp ( magnetic reconnection at the edges of the cusp , acceleration of the electrons and their thermalization ) steeply decreased after the temperature maximum .
figure [ fig11]a shows that after 22:00 ut the decrease in the energy release was much slower which allowed the flare to develop the long qss decay phase .
we also see in table [ tab2 ] that @xmath62 , _ i.e. _ the conductive loss of energy was larger than the radiative loss ( see also ) .
hence , we can use the following simple approximation in further estimates : @xmath63 .
this gives @xmath64 .
the small value of the power index in the last formula implies that significant changes in @xmath35 induce only weak changes in @xmath33 ( see table [ tab2 ] ) .
figure [ fig9 ] indicates that the energy release during the decay phase occurred in the arcade of sxr loops .
it is most probable that the weak energy release which maintained the temperature at @xmath65 mk was mostly due to reconnection between the arcade loops and the arcade channel .
the bright arcade loops in figure [ fig9 ] are the places where the reconnection was most efficient .
we have applied the above equations ( 3)-(6 ) to one of the arcade loops .
we have measured the radius in the southern loop - top kernel , @xmath66 mm , and the length of the `` leg '' connecting the kernel with the loop footpoint , @xmath67 mm .
we have obtained @xmath68 erg@xmath39s@xmath69 and @xmath70 @xmath39 .
hence , to maintain the long - duration decay phase , weak energy release in the arcade loops was sufficient .
important information is contained in the time variation of the emission measure , em@xmath1 ( figure [ fig11]b ) .
this figure shows that during the impulsive phase a large amount of hot plasma had been accumulated within the flaring system ( the flaring loop and arcade channel ) , and during the decay phase this amount of plasma slowly and smoothly decreased ( the small peak in em@xmath1 about 00:25 ut is of instrumental origin ; the peaks after 03:00 ut are due to other flares ) .
the relationship between the physical parameters during the decay phase was the following : @xmath71 where @xmath72 ( here changes in the loop semilength , @xmath40 , and changes in the volume , @xmath73 , of sxr emitting plasma have been assumed to be of minor importance . )
combination of the relationships [ equation ( 8) ] gives : @xmath74^{1.14 } \quad\mbox{or}\quad e_{\rm h}(t ) \sim [ { \rm em}(t)]^{0.88}.\ ] ] this simple relationship stresses the fact that the slow decrease in em during the decay phase is due to slow decrease in the heating @xmath35 .
in other words , the observed slow and smooth decrease of the emission measure , em@xmath1 , indicates that the thermal energy release , @xmath75 , decreased slowly and smoothly during the long decay phase .
it is most probable that this weak energy release was mostly due to the reconnection between the magnetic arcade loops and the arcade channel .
in paper iii we have found clear observational evidence that the strong hxr pulses at the hxr maximum ( see figure [ fig1]a ) were the result of inflow of dense plasma ( coming from the chromospheric evaporation ) into the acceleration volume inside the cusp structure . for the investigated flare of 2 march 1993 we were not able to monitor the inflow of plasma into the cusp , since we had no sxr imaging observations for the flare impulsive phase .
we can , however , obtain simple estimates for the increase in density inside the cusp using only the hxr light curves . in table
[ tab1 ] we see that the time interval , @xmath11 , between the pulses increased during the hxr maximum ( @xmath76 , where @xmath77 is the period before 21:10 ut , and @xmath78 is the period during the hxr maximum ) .
we assume that this increase in the period is the result of increase in density and we consider two extreme cases : a. we assume that @xmath79 inside the traps , where @xmath80 is the magnetic field strength and @xmath81 is the pressure .
then the magnetic field does not change significantly during the increase in pressure in the traps and we have : @xmath82 or @xmath83 where @xmath19 is the alfv@xmath18n speed .
b. we assume that @xmath84 .
then the increase in pressure causes broadening of the traps and the magnetic field in the traps decreases . for this estimate we assumed that @xmath85 where @xmath86 is the magnetic field strength before the hxr maximum and @xmath87 is the field strength during the maximum . then @xmath88 combination of equations ( 12 ) and ( 13 ) gives : @xmath89 and therefore @xmath90 _ i.e. _ @xmath91 on the other hand
, we can estimate the ratio of the densities directly from the ratio , @xmath92 , of the hxr fluxes ( @xmath93 is the flux between 21:05 - 21:08 ut and @xmath94 is the flux at the hxr maximum ) . for the 23 - 33 kev flux the ratio was @xmath95
( see figure [ fig3 ] ) . during these time intervals
the emission came mainly from the flare footpoints ( see figure [ fig6 ] ) , hence it is proportional to the number of precipitating electrons per second which , in turn , is proportional to the electron number density , @xmath38 , within the cusp .
hence , @xmath96 the last estimate is independent of the previous two . therefore , we consider these estimates to be a confirmation that the density has quickly increased by a factor between 1.4 and 2.0 ( the mean value is [email protected] ) and the longer period , @xmath11 , at the hxr maximum is the result of this increase in density .
the sequence of hxr images for the impulsive phase of 2 march 1993 flare has allowed us to investigate asymmetry in the precipitation of accelerated electrons from the acceleration volume within the cusp toward the flare footpoints . according to our model of acceleration of the electrons in oscillating magnetic traps , the precipitation was most efficient during the maximum compression of the traps ( see papers i and iii ) .
the asymmetry of precipitation shows that the maximum of compression was different at the opposite ends of the traps , _ i.e. _ the axis of symmetry of the traps was slightly inclined toward one of the ends of the traps ( see paper iii ) .
the changes in the asymmetry from one hxr pulse to another indicate that the inclination of the axis of symmetry of the traps changed .
sxr images , which were available for the late decay phase , show a long arcade of sxr loops ( figure [ fig9 ] ) . in figure
[ fig10 ] a contrast between the `` slim '' shape of the flaring loop and the long arcade is seen .
this contrast is enhanced by the fact that there were no sxr images for the flare impulsive phase and therefore the arcade channel was not seen during this phase . in other flares
such sxr images show a connection between the cusp and arcade channel and the heating of the channel by the cusp ( see figure [ fig8 ] ) .
similar cases like in figure [ fig8 ] will be shown in our next paper .
important information about the evolution of a flare during slow decay phase is contained in the time variation of the temperature , @xmath0 , and emission measure , em@xmath1 .
this information is the following : a. weak heating occurs during the slow decay phase and it slowly decreases ; b. the decrease in the heating determines slow and smooth decrease in em ; c. the coupling between the heating and the amount of the hot plasma makes the flare evolve along a sequence of quasi - steady states during the slow decay phase ( qss evolution ) .
this paper is dedicated to professor antoni opolski on the occasion of his 100th birthday .
professor opolski was the main person at our astronomical institute for many years and he efficiently stimulated the development of solar research at the institute .
we are very thankful to him for his well - wishing support
. the _ yohkoh _
satellite is a project of the institute of space and astronautical science of japan .
compton gamma ray observatory _ is a project of nasa .
the authors are very thankful to the anonymous referee for her / his important remarks which helped to improve this paper .
we acknowledge financial support from the polish national science centre grant 2011/03/b / st9/00104 . | a large arcade flare of 2 march 1993 has been investigated using x - ray observations recorded by the _ yohkoh _ and goes satellites and the _ compton gamma ray observatory_. we analyzed quasi - periodicity of the hard - x - ray ( hxr ) pulses in the flare impulsive phase and found close similarity between the quasi - periodic sequence of the pulses with that observed in another large arcade flare of 2 november 1991 .
this similarity helped to explain the strong hxr pulses which were recorded at the end of the impulsive phase , as due to an inflow of dense plasma ( coming from the chromospheric evaporation ) into the acceleration volume inside the cusp . in hxr images
a high flaring loop was seen with a triangular cusp structure at the top , where the electrons were efficiently accelerated . the sequence of hxr images allowed us to investigate complicated changes in the precipitation of the accelerated electrons toward the flare footpoints .
we have shown that all these impulsive - phase observations can be easily explained in terms of the model of electron acceleration in oscillating magnetic traps located within the cusp structure . some soft - x - ray ( sxr )
images were available for the late decay phase .
they show a long arcade of sxr loops .
important information about the evolution of the flare during the slow decay phase is contained in the time variation of the temperature , @xmath0 , and emission measure , em@xmath1 .
this information is the following : i ) weak heating occurs during the slow decay phase and it slowly decreases ; ii ) the decrease in the heating determines slow and smooth decrease in em ; iii ) the coupling between the heating and the amount of the hot plasma makes the flare evolve along a sequence of quasi - steady states during the slow decay phase ( qss evolution ) . |
hexagonal boron nitride ( @xmath0-bn ) has the same honeycomb lattice as graphite with two atoms per unit cell and similar lattice parameters . due to this similarity ,
boron nitride materials have attracted a growing interest in line with the development of low - dimensional carbon - related materials . similarly to carbon , bn materials can be synthesized either as nanotubes ( one - dimensional ( 1d ) form ) @xcite or as monolayers and/or multilayers ( two - dimensional ( 2d ) form).@xcite in the following we focus on this latter form .
2d layers of carbon , namely graphene sheets , display extraordinary electronic properties which open unanticipated routes for a new generation of electronic devices .
however , the electron mobility of supported graphene typically falls short of that of suspended graphene , due to detrimental effects of substrate disorder and adsorbents .
@xcite facing this problem , @xmath0-bn layers are of particular interest as support or capping layers of graphene .
they combine several properties : they are insulating ( @xmath0-bn is a large gap semiconductor due to the polar bn bond ) , @xcite they display an especially compatible layered @xmath2 structure with that of graphene , they have a low concentration of charges impurities and they can be very flat due to an easy cleavage . owing to these properties ,
graphene transferred on bn layers displays an electron mobility at room temperature of , which is the highest reported value for a supported graphene @xcite and very close to that of suspended graphene . @xcite beyond the high mobility of graphene supported on bn , their excellent lattice matching is promising for the realization of heterostructures of these materials for vertical transport stacking , in which graphene layers act as tunable metallic electrodes for the bn quasi - ideal tunnel barrier .
@xcite these promising perspectives have been demonstrated by pioneering experiments done using sheets mechanically exfoliated from both graphite and @xmath0-bn single crystals . in the future ,
@xmath0-bn and graphene based devices and heterostructures would most probably use chemical vapor deposited ( cvd ) polycrystalline films and sheets .
their performances would only be achieved via an accurate control of the defects in both graphene and bn layers and of the layers engineering .
while the electronic properties of graphene have been well described theoretically and investigated experimentally , this is not the case of bn layers and even of @xmath0-bn .
this is due to both the scarcity of high quality materials and to the nature of their electronic properties dictated by the large gap .
it is thus a basic issue to understand the spectroscopic properties of atomically thin @xmath0-bn layers and their intrinsic defects , which is the focus of this paper .
in contrast to graphene , usual spectroscopic characterization techniques such as raman are not easy to manipulate or they provide poor information when used for @xmath0-bn .
absorption and luminescence spectroscopies have been shown to be the most direct approach to investigate the electronic properties of bn materials , due to their large gap . to this aim , dedicated cathodolumnescence and photoluminescence experiments
have been recently developed and applied to bn powders and single crystals .
@xcite both theoretical calculations @xcite and the most recent excitation photoluminescence experiments on single crystals @xcite converge to establish the band gap of @xmath0-bn near .
furthermore , it is now commonly accepted that @xmath0-bn optical properties are dominated by huge excitonic effects . the near - band - edge luminescence spectrum is composed of two series of lines . referring to measurements done on single crystals in ref .
[ ] , they are defined as the @xmath3 and @xmath1 series .
the four higher energy lines , labeled @xmath4 to @xmath5 , located between 5.7 and , are attributed to the excitons , whereas the lower energy ones , labeled @xmath6 to @xmath7 , between 5.4 and , are assigned to excitons trapped to structural defects . @xcite the excitons in @xmath0-bn are more of frenkel - type than of wannier - type ( as in others usual semiconductors , such as aln with a gap ) . _
ab initio _ calculations indeed predict that the spatial extension of the exciton wavefunction is of the order of one @xmath0-bn atomic layer.@xcite moreover the experimental stokes shift of observed for the @xmath5-line suggests its self - trapping , @xcite consistent with the very localized view of the frenkel exciton .
to complete this view , the effect of a reduction in the @xmath0-bn thickness down to the atomic level has to be analyzed .
up to now , only scarce studies deal with the optical properties of nanometer - thick bn layers . an optical absorption edge between 5.6 and at room temperature is reported , @xcite _ i.e. _ in the same range than in bulk @xmath0-bn .
only two studies report near - band edge recombination luminescence , with no correlation to the bn layer thickness under investigation .
@xcite in this paper we present the first study of the luminescence properties of single bn nanosheets , with well - known thickness , by combining atomic force microscopy ( afm ) and cathodoluminescence ( cl ) measurements .
bn nanosheets were prepared by mechanical exfoliation of small @xmath0-bn crystallites of a polycrystalline powder .
this material offers the advantage to give access at the same time to the intrinsic optical response of the crystallite as well as to the effect of grain boundaries and the crystallite thickness on this response .
an advanced characterization of the starting bulk material is first presented and its near - band - edge recombinations observed by cl are discussed with respect to those of the single crystal .
then the luminescence of the exfoliated bn sheets is presented and discussed as a function of their thickness .
the bulk material exfoliated in this paper is the high purity trs bn@xmath8 st - gobain commercial power puhp1108 , used in cosmetic applications .
the @xmath0-bn crystallites are already shaped as flakes with large ( 00.1 ) surfaces of typically diameter , which is particularly convenient for the exfoliation process .
their thickness is about .
one of these crystallites is shown in the scanning electron microscope ( sem ) image of .
this powder was synthesized at high temperature from boric acid and a nitrogen source .
our measurements on this powder are compared with the ones obtained from a single crystal of the best available quality , synthesized by a high - pressure high - temperature ( hpht ) crystal process @xcite and provided by taniguchi _
_ this reference sample is hereafter referred to as the hpht sample .
exfoliation of few - layer @xmath0-bn was carried out by mechanical peeling following the same method used for graphene .
@xcite the powder is applied to an adhesive tape , whose repeated folding and peeling apart separates the layers .
these thin layers are then transferred on a si wafer covered with of , which is the optimal thickness for imaging bn flakes with maximum contrast .
@xcite prior to the @xmath0-bn deposition , cr / au localization marks were formed on the wafer by means of uv - lithography using an az5214e photo - resist and joule evaporation .
@xcite the marks facilitate the localization of the flakes for the different measurements to be achieved on a given flake . still prior to transferring the layers ,
the wafer is chemically cleaned with acetone and isopropanol , followed by several minutes of exposure to a strong plasma ( , @xmath9 ) .
this last step eliminates most contaminants but also renders the surface hydrophilic,@xcite which has the drawback of facilitating the inclusion of a thin water layer between the flakes and the substrate.@xcite as we shall see later , this complicates the layer thickness measurement done by atomic force microscopy ( afm ) .
the surface morphology of the @xmath0-bn exfoliated layers was investigated by atomic force microscopy with a dimension 3100 scanning probe microscope ( brukers ) operating in tapping mode with commercial mpp-11100 probes .
the cathodoluminescence of the @xmath0-bn samples was analyzed at low temperature in the ( i ) spectroscopic , ( ii ) imaging and ( iii ) spectral mapping modes , using an optical system ( horiba jobin yvon sa ) installed on a high resolution jeol7001f field - emission scanning electron microscope .
the samples are mounted on a gatan cryostat sem stage and cooled down to @xmath10 with a continuous flow of liquid helium .
they are excited by electrons accelerated at with a beam current as low as .
the cl emission is collected by a parabolic mirror and focused with mirror optics on the entrance slit of a -focal length monochromator .
the all - mirror optics combined with a suitable choice of uv detectors and gratings ensures a high spectral sensitivity down to .
a silicon charge - coupled - display ( ccd ) camera is used to record spectra in mode ( i ) and as well for the spectral mapping mode ( iii ) . in mode ( iii ) , also referred as the hyperspectral imaging acquisition mode , the focused electron beam is scanned step by step with the _ hjy cl link _ drive unit ( co - developed and tested at gemac ) and synchronized with the ccd camera to record one spectrum for each point . the spectrometer is also equipped with a uv photomultiplier on the lateral side exit for fast monochromatic cl imaging ( _ i.e. _ image of the luminescence at a given wavelength ) in mode ( ii ) . for all spectra reported in this paper , it was checked that the linewidths are not limited by the spectral resolution of the apparatus ( in the best conditions ) . in the case of @xmath0-bn
exfoliated layers , cl was only performed in mode ( i ) and using a fast _ e_-beam scanning of a 1.5@xmath11 area on the sample instead of using a fixed focused beam , in order to minimize the irradiation dose and the _ e_-beam induced modifications .
the @xmath0-bn crystallites are often made of a few single crystals separated by grain boundaries , as for another powder studied in a previous work.@xcite owing to the electron diffraction patterns recorded by transmission electron microscopy ( tem ) on the specimen presented in , each grain could be precisely orientated along the ( 00.1 ) zone axis ( not shown here , for experimental details , see ref .
the tem analysis shows that the two main grains of the crystallite in are slightly tilted one with respect to the other .
they are separated by the grain boundary labeled # 1 .
-bn materials in the near - band edge region : at the top is a reference spectrum of a hpht high quality crystal , @xcite compared to # 1 , registered in the grain boundary area delimited by the red rectangle ( averaged over 15 spectra ) and to # 2 on the main grain of the crystallite in the area delimited by the green square ( averaged over 240 spectra ) , as labeled in .
the specimen temperature is about . ]
cl spectra recorded on different areas ( labelled # 1 and # 2 ) of the crystallite in are shown in .
they are compared to the one recorded on the hpht reference sample and hereafter referred to as the reference spectrum .
it is worth mentioning that this reference sample precisely displays the same exciton recombination energies than the ones observed in photoluminescence experiments done on a single crystal of the same origin as reported in ref .
spectrum # 1 in is recorded at the grain boundary of the @xmath0-bn crystallite .
it presents the same features as the reference sample , with clear @xmath3and @xmath1 series .
the @xmath1 lines are slightly shifted toward lower energies by about when compared to the reference spectrum .
since the s lines are not shifted , such a decrease of the emission energy is attributed to a different binding energy of the exciton to the grain boundary , rather than to a change of bandgap energy .
moreover the @xmath3 and @xmath1 series appear broader than in the reference sample , possibly due to an inhomogeneous distribution of residual strain in the sample .
this interpretation is supported by the slight fluctuations of the recombination energies of a few mev , observed from point to point in a cl mapping ( not shown here ) .
moreover we remark that the relative intensities of @xmath3 and @xmath1 series are very different when compared to the reference spectrum .
spectrum # 1 is dominated by the @xmath1 series .
it is known that this series is linked to the presence of structural defects such as grain boundaries , dislocations , or stacking faults .
@xcite to confirm this interpretation , monochromatic images have been registered , at energy centered on the @xmath5 line ( at , ) and on the d4 series ( at , ) .
both @xmath1 and @xmath3 emission series are slightly enhanced ( by a factor of two in average ) at the crystallite edges .
this enhancement confirms previous observations on single @xmath0-bn crystals by kubota _
_ , @xcite and is consistent with their interpretation based on a more efficient light scattering at the grain edges . more interestingly , a clear one - to - one correlation is observed between the @xmath1 series emission and the grain boundary locations ( ) .
to better quantify this view and to dispose of a comparison criterion , we define the @xmath12 ratio as the ratio of amplitudes between the @xmath7 and the @xmath5 peaks ( at and , resp . 227 and ) . it can be viewed as an indicator of the defect density , as introduced by watanabe _
_ @xcite and in the same way than in conventional semiconductors for the impurities content .
@xcite owing to cl spectral mapping facilities , the @xmath12 ratio is evaluated in each point of the studied @xmath0-bn crystallite of , and shown in .
it exhibits a variation of two orders of magnitude ( 0.2 20 ) .
this reveals the extremely high impact of structural defects in bulk @xmath0-bn .
it is also remarkable that in the central part of the upper grain , the @xmath12 ratio is four times lower than in our reference hpht sample ( @xmath13 ) , suggesting a better structural quality .
as the @xmath1 series almost vanishes , new intrinsic features are revealed , which were hindered by the @xmath1 series in previous studies .
indeed in the cl spectrum # 2 of recorded from the central part of the grain , three new lines are detected at 5.619 , 5.457 and .
they are shifted by @xmath14 ( with @xmath15 ) times ( ) from the @xmath16 band , a period close to the ( ) optical phonon mode at the center of the brillouin zone .
@xcite they are also characterized by a monotonous decay in intensity upon increasing @xmath14 .
we thus attribute these new lines to three phonon replica of the dominant @xmath3 exciton recombinations .
we notice that the energy of the second phonon replica almost coincides with the @xmath7 line . as a consequence ,
the @xmath12 ratio as defined here , has an intrinsic lower value , which corresponds to the ratio between the @xmath17 replica and the fundamental @xmath3 line .
the observation of phonon replica attests to the existence of a strong electron - phonon coupling occurring in this material , as evidenced previously on a deep impurity band .
@xcite turning now to the sheets exfoliated from the bulk material studied in section iiia , several sheets have been studied from which we show three representative samples in .
the surface of each @xmath0-bn flake was scanned by afm in the tapping mode in order to measure its thickness , consecutively to the cl studies .
the sheet surface is not completely flat , with some prominent dots being visible .
most probably they are adhesive tape glue residues from the exfoliation process .
some small 10- size holes are also observable .
it is not clear whether they were already in the bulk material or whether they are due to electron exposure as reported in the literature .
indeed when etching under an electron beam in a transmission electron microscope , even at a low acceleration voltage , they are often observed .
@xcite aside from this , the flakes appear to be well spread on top of the substrate . [
cols="<,^,^,^",options="header " , ] particular attention has been paid to the thickness determination because of the water adsorbed at the surface and trapped under the bn flakes .
this problem is well known in the case of graphene on @xcite and already mentioned for @xmath0-bn on .
@xcite due to the presence of the water layer , the height profiles recorded by afm on steps at the flake edges overestimate the @xmath0-bn thickness .
the special geometry of sample a , which displays a folded part on itself , gives a key to solving this problem .
the step height of the folded part provides a thickness measurement independent of the amount of water trapped between @xmath0-bn and , and this measurement has been used as the reference for the thickness measurement as follows .
the @xmath0-bn flake thickness is found to be equal to a value of averaged over five measurements of the step height along the edge ( see ) .
the standard deviation is found to be below , the interplanar distance along the @xmath18 axis in @xmath0-bn , indicating that the thickness of flake a is homogeneous . knowing the @xmath0-bn interplanar distance and assuming an integer value of the layer number
, we consider that the exfoliated sample a is 6 atomic layers thick .
then the water layer thickness is deduced by difference with the step edges and is found equal to @xmath19 over seven measurements .
the thickness of samples b ( ) and c ( ) is measured assuming a homogenous water layer thickness over the wafer .
the results are summarized in .
one can notice that the areas scanned during the cl acquisition appear as dark rectangles in afm images , which indicates a lower height .
this could be interpreted by an electron beam induced etching . as we can also observe a lower height on the , it is more probably a reduction of the volume of the amorphous silica under irradiation . @xcite
the cl spectra of the three exfoliated @xmath0-bn layers presented above are shown in .
first , we observe that the cl intensity decreases when decreasing the @xmath0-bn layer thickness . since a nanometer - thick @xmath0-bn layer is transparent to the electron beam , the electron - hole generation in such a thin layer is directly proportional to its thickness . in spite of this effect ,
exfoliated layers show a significant luminescence in comparison to their nanometric thickness , which is a result in itself . indeed in semiconductor nanostructures , the surface to volume ratio , being extremely large
, often results in non radiative surface recombinations , responsible for the quenching of luminescence .
@xcite our results indicate that such a surface effect is weak in @xmath0-bn , probably due to the @xmath2 character of the bonding combined with the strong localization of the exciton at the atomic layer scale .
this is an advantage of this material compared to other semiconductors , for which passivating the surface is necessary for enhancing the luminescence .
@xcite the cl spectra of exfoliated layers are dominated by the @xmath7 and @xmath20 lines related to the @xmath1 series .
weaker and unresolved @xmath21 and @xmath22 lines correspond to the @xmath3 series , clearly attributed by comparison with the reference hpht spectrum . looking in more detail at the excitonic recombination emission ,
we observe a shift from for the reference spectrum to 5.479 , 5.477 and for exfoliated samples c ( ) , b ( ) and a ( ) respectively .
for the 8 and -thick samples , the blueshift is observed for the defect - related @xmath1 lines but also for the @xmath3 lines , indicating a modification in the intrinsic exciton emission energy ( change in bandgap and/or in exciton binding energy ) . for the thinnest sample ,
the intrinsic @xmath3 line emission is not observed possibly because of a too low signal - to - noise ratio .
-bn layer a ( ) , b ( ) and c ( ) compared to the normalized reference spectrum ( from the hpht crystal ) .
the zero of the cl spectra was shifted for clarity . ] to interpret our results , it has to be emphasized first that a calculation including excitonic effects still needs to be performed to describe the influence of the number of atomic layers on @xmath0-bn optical properties .
however , inside bulk @xmath0-bn crystals , the strong localization of the exciton wavefunction around a single atomic basal plane has been evidenced by arnaud
_ et al . _
@xcite it indicates that the regime of quantum confinement is still far to be reached with bn sheets in the range of 6- , and that exciton recombination properties are probably close to the ones of bulk @xmath0-bn . it is also interesting to consider the theoretical work of wirtz _ et al .
_ @xcite it deals with the optical response of a single sheet of @xmath0-bn as a function of the intersheet distance . when the interdistance between @xmath0-bn planes becomes large , the situation tends to uncoupled bn sheets of atomic thickness .
the calculation results as follows : the exciton binding energy increases when the intersheet distance increases , but simultaneously the quasiparticle band gap increases , so that the net result is a weak blueshift , about , in the error bar of such calculations .
similar effects have been predicted in the case of bn nanotubes when their diameter decreases , @xcite and more recently in nanoribbons .
@xcite these arguments indicate that the change of the exciton recombination energy in monoatomic bn sheets should be weak compared to the bulk and should not exceed a few tenths of ev anyway .
our experimental results show a slight increase ( @xmath23 ) in the @xmath1-series exciton recombination energy from the bulk to the thinnest sample ( ) , consistent with these theoretical predictions .
further calculations as well as experiments for less than six layers would be of great interest in the future to see if larger effects are observed . when comparing the exfoliated materials with the bulk from which they were exfoliated , it is remarkable that excitonic recombinations of exfoliated layers are systematically dominated by the @xmath1 series . in the bulk source materials ,
various crystallites have been investigated .
we found that the @xmath12 ratio varies from one crystallite to another but stays between 0.2 and 0.5 .
by contrast , the @xmath12 ratio is found to be equal to about 4 in the 8 and sheets ( samples b and c ) and to about 10 in the sheet ( sample a ) , that is hundred times larger than in the bulk crystallites .
these results are summarized in .
as discussed previously , this increase of the @xmath12 ratio indicates a dramatic increase in the density of defects .
they are probably induced by the exfoliation process , even if we can not exclude layer deformation induced by the water layer freezing arising during cl experiments at , or etching under the electron beam .
these results raise the question of the suitability of the mechanical cleavage exfoliation process .
@xmath0-bn has indeed been proven to be extremely fragile and subject , for instance , to easy stacking fault generation.@xcite the exfoliation process may be a violent mechanical experiment for such a material and can be obviously incriminated as the source of new defects responsible for the @xmath12 increase .
the nature of the substrate is also questionable .
the roughness and surface charges of the used substrate can affect the @xmath0-bn band structure and degrade some properties as for graphene .
@xcite ratio in the three exfoliated layers , compared with the bulk material , consisting of the crystallites from which the layers were exfoliated . ]
in conclusion , we have studied high quality @xmath0-bn crystallites by cathodoluminescence and compared them with exfoliated sheets in the near - band edge region around . first owing to the high quality of the crystallites
, we could clearly discriminate the luminescence of high structural quality areas ( dominated by the @xmath3 series ) and that of defective zones ( dominated by the @xmath1 series ) . from this , for evaluating the structural defect density in @xmath0-bn samples , we define a simple quantitative parameter as the ratio of the intensity of the @xmath1 series over that of the @xmath3 series .
furthermore the observation of high quality areas has revealed new intrinsic features , attributed to phonon replica of the excitonic recombinations .
second , investigation of the emission of exfoliated sheets has shown that the reduction in thickness induces a slight increase of the exciton emission energy .
these observations are consistent with theoretical results obtained on single layers and on bn nanotubes .
this explanation still deserves to be confirmed as one can not exclude for the moment surface and strain effects .
the dominant defect - related emission of exfoliated samples also questions the exfoliation process and the role played by the substrate .
this indicates that special care should be taken in the sample preparation .
presently , our measurements evidence that in @xmath0-bn , exfoliation can degrade some of the properties and should be used as a guideline to further optimize this process
. further experiments on free - standing sheets are thus in progress , with the advantage to make possible correlations between the emission and structural features owing to tem observations .
more generally this work reports luminescence spectra of a semiconductor only a few atomic layers thick , which is not common and can be of great interest for future optical applications .
t. taniguchi and k. watanabe from nims , japan , are warmly acknowledged for providing one of their hpht crystals .
the authors would like to thank s. pouget from cea grenoble for x - ray experiments , c.vilar from gemac for her technical help on the cathodoluminescence - sem set - up , and h. mariette from the npsc for fruitful discussions .
the research leading to these results has received funding from the european union seventh framework programme under grant agreement n^o^604391 graphene flagship .
it has been also supported by grant from the mission interdisciplinaire of the cnrs ( challenge `` graphene '' of the program g3n ) and by the federative research program `` graphene '' of onera .
thanks cnano rhne - alpes and ile - de - france for financial support .
59ifxundefined [ 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 * * , ( ) link:\doibase 10.1021/nl1023707
[ * * , ( ) ] link:\doibase 10.1021/nl1022139 [ * * , ( ) ] link:\doibase 10.1021/nl4012529 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1021/nl2005115 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1002/pssr.201105190 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1063/1.3599708 [ * * , ( ) ] link:\doibase 10.1021/nl200758b [ * * , ( ) ] link:\doibase 10.1063/1.3665405 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.101.096802 [ * * , ( ) ] http://dx.doi.org/10.1038/nnano.2008.199 [ * * , ( ) ] link:\doibase 10.1126/science.1218461 [ * * , ( ) ] link:\doibase 10.1021/nl1039499 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1021/nl203249a [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1021/nn301940k [ * * , ( ) ] link:\doibase 10.1039/c2jm15109j [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1021/nl301061b [ * * , ( ) ] link:\doibase 10.1063/1.4731203 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1002/smll.201001628 [ * * , ( ) ] _ _ , @noop ph.d .
thesis , ( ) link:\doibase 10.1063/1.3611394 [ * * , ( ) ] link:\doibase 10.1021/nl070613a [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1103/physrevb.83.073201 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1103/physrevlett.109.205502 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physrevlett.108.075501 [ * * , ( ) ] link:\doibase 10.1103/physrevb.85.045440 [ * * , ( ) ] link:\doibase 10.1063/1.363379 [ * * , ( ) ] link:\doibase 10.1063/1.3519980 [ * * , ( ) ] link:\doibase 10.1557/jmr.2011.211 [ * * , ( ) ] link:\doibase 10.1021/jp9530562 [ * * , ( ) ] link:\doibase 10.1021/ja970754 m [ * * , ( ) ] link:\doibase 10.1021/nl302490y [ * * , ( ) ] link:\doibase 10.1063/1.2364885 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1063/1.3625922 [ * * , ( ) ] | hexagonal boron nitride ( @xmath0-bn ) and graphite are structurally similar but with very different properties .
their combination in graphene - based devices is now of intense research focus , and it becomes particularly important to evaluate the role played by crystalline defects on their properties . in this paper , the cathodoluminescence ( cl ) properties of hexagonal boron nitride crystallites are reported and compared to those of nanosheets mechanically exfoliated from them .
first the link between the presence of structural defects and the recombination intensity of trapped excitons , the so - called @xmath1 series , is confirmed .
low defective @xmath0-bn regions are further evidenced by cl spectral mapping ( hyperspectral imaging ) , allowing us to observe new features in the near - band - edge region , tentatively attributed to phonon replica of exciton recombinations .
second the @xmath0-bn thickness was reduced down to six atomic layers , using mechanical exfoliation , as evidenced by atomic force microscopy . even at these low thicknesses ,
the luminescence remains intense and exciton recombination energies are not strongly modified with respect to the bulk , as expected from theoretical calculations indicating extremely compact excitons in @xmath0-bn . |
since the detection of the first extrasolar giant planet ( egp ) , 51 peg b @xcite , considerable effort has gone into both observing and modeling properties of the atmospheres of the close - in `` hot jupiter '' planets .
spitzer space telescope _ now provides us a unique probe into the mid - infrared emission from these exotic atmospheres . in this letter
we examine mid infrared spectra of transiting planets hd 209458b and hd 189733b , which were recently observed with _ spitzer _ s infrared spectrograph ( irs ) by @xcite and @xcite , respectively .
the spectra were obtained from @xmath27.5 to 14@xmath3 m around the time of secondary eclipse .
the most prominent conclusion of these works is that strong absorption arising from h@xmath4o , expected short - ward of 10@xmath3 m , was not seen for either planet .
possible explanations put forward for the relatively featureless spectra included a lack of h@xmath4o in these atmospheres , masking by a continuum of opaque clouds , or an isothermal temperature structure .
recently @xcite published an 8@xmath3 m light curve for hd 189733b that includes an observation of the secondary eclipse .
these observations used _ spitzer _
s infrared array camera ( irac ) .
we suggest that this data point is inconsistent with the irs spectrum , that water vapor is present and absorbs from 6.5 - 10@xmath3 m , and that the published , low s / n spectrum of hd 189733b may not reflect the true planet spectrum .
we have computed model atmospheres and spectra of hd 189733b and hd 209458b in an effort to shed additional light on these exciting observations .
we will also briefly discuss l- and t - type brown dwarfs , whose atmospheres span the same temperature range as hot jupiters .
we suggest that the generally very good fits of our models to these high signal - to - noise irac and irs datasets supports the veracity of the underlying physics and chemistry of our planet models .
two teams employed the irs instrument in `` short - low '' configuration to obtain low resolution spectra of parent stars hd 209458 and hd 189733 around the time of planetary secondary eclipse .
each group compared the spectrum of the system when the planet s day side was still visible ( before and after the secondary eclipse ) to the stellar spectrum when the planet was behind the star .
the observed enhancement in flux seen when the planet was visible was attributed to the planet , thus providing a measurement of the planet - to - star flux ratio .
these exciting and difficult observations are at the leading edge of spectral characterization of exoplanets .
@xcite observed two secondary eclipse events of planet hd 189733b and from these observations published a planet - to - star flux ratio from 7.57 to 14.71@xmath3 m .
they found a flux ratio spectrum that is essentially flat across these wavelengths with a mean eclipse depth of @xmath5 .
the subset of their data with the highest signal - to - noise is shown in _
@xcite observed two secondary eclipses of planet hd 209458b and published a flux ratio spectrum from 7.5 to 13.2@xmath3 m .
the derived average ratio is shown in _
c_. these authors also find a spectrum that could be characterized as mostly flat .
however , they also detect increases in the flux ratio from 9.3 - 10@xmath3 m , as well as a narrow emission feature at 7.7@xmath3 m , which they attribute to emission from a stratosphere .
in addition to spectra , we plot _ spitzer _ secondary eclipse photometric data . these points
include the irs blue peak - up at 16@xmath3 m @xcite and the irac 8@xmath3 m point @xcite for hd 189733b , and the multiband imaging photometer for spitzer ( mips ) point at 24@xmath3 m for hd 209458b @xcite .
we note that in _ a _ the irac 8@xmath3 m point clearly plots below the irs data , which we ll discuss in detail in 2.2 . in order to further the study of these observations and their constraints on the character of each planet s atmosphere
we have computed model atmospheres and generated low resolution spectra .
we employ a 1d model atmosphere code that has been used for a variety of planetary and substellar objects .
recently it has been used for brown dwarfs @xcite and egps @xcite .
it explicitly includes both incident radiation from the parent star ( if present ) and thermal radiation from the planet s atmosphere .
the radiative transfer solving scheme was developed by @xcite and was first applied to irradiated giant planet atmospheres by @xcite .
we use the elemental abundance data of @xcite and compute chemical equilibrium compositions following lodders & fegley ( 2002 , 2006 ) .
our opacity database is described in @xcite .
the stellar models are from @xcite and the system parameters are from @xcite for hd 189733b and @xcite for hd 209458b .
the planet models presented here are cloud - free .
we plot model planet - to - star flux ratios in . for hd 189733b
we show two one - dimensional models .
the solid red curve is for a day - side model that assumes all incident flux is efficiently redistributed over the entire planet , meaning that the incident flux is multiplied by a factor of 1/4 @xcite .
following conventions adopted in solar system atmosphere modeling ( e.g. * ? ? ?
* ) , here and in our previous papers @xcite the incident flux is decreased by incorporating a multiplicative factor @xmath6 of 1/2 due to the day - night average , and another factor of 1/2 arises from the mean 60@xmath7 stellar zenith angle ( @xmath3 , the cosine of this angle , is 0.5 ) that is correct for global average insolation conditions @xcite . with this prescription more stellar flux
is absorbed at low pressures , leading to a warmer upper atmosphere and shallower temperature gradient , compared to a model where incident flux is directly multiplied by 1/4 with @xmath3=1.0 .
the greater absorption of incident flux at lower pressure is due to the twice longer path length for incident photons to reach a given pressure . in reality , these differences in 1d models could be swamped by a more complex day - side temperature structure for these planets . here , for models with a warmer day side ,
the incident flux is multiplied by a factor @xmath6 between 1/2 and 1 , while @xmath3 remains 0.5 .
the dash - dot curve in _
c _ is for a warmer model , where @xmath8 , meaning that only 20% of the absorbed incident flux is assumed to be advected to the night side .
both models predict an absorption feature due to h@xmath4o short - ward of 10@xmath3 m .
_ b _ shows absorption cross sections for h@xmath4o , co , and ch@xmath9 . at these temperatures ( @xmath21200 k ) , co is strongly favored at the expense of ch@xmath9 .
the dashed model in _ a _ is a good ( @xmath21@xmath10 ) fit to the band - averaged 8@xmath3 m @xcite and 16@xmath3 m @xcite secondary eclipse observations .
we note that although @xcite claim their flat ratio spectrum is consistent with a blackbody , _ a _ shows that it is actually bluer than a blackbody .
_ c _ shows the mean @xcite data set . in solid red
we plot the @xmath11 1d model spectrum from @xcite .
note that the predicted `` downturn '' in the ratio spectrum short - ward of 10@xmath3 m is actually quite small in this 1d model , less than that predicted by other modelers ( e.g. , * ? ? ?
* ; * ? ? ?
* ) , for reasons that are not yet clear , but could be related to the angle of incident radiation . given the s / n of this dataset and the relatively small predicted depth of the model h@xmath4o absorption feature , it is not clear that absorption due to h@xmath4o can be dismissed .
dynamical processes can act to produce isothermal profiles in the upper atmosphere which would also suppress the depths of absorption features . in solid blue
we plot the spectrum of the day side of the @xcite 3d dynamical model of hd 209458b , as computed in @xcite .
the isothermal atmosphere is due to a super - rotating west - to - east equatorial jet ; as cool night side gas is blown around the planet s limb this gas is more readily warmed at lower pressures where the radiative time constant is shorter @xcite , leading to day side _ p - t _ profiles that are more isothermal than one would predict for a radiative equilibrium model @xcite .
neither the 1d nor 3d model has the increase in flux ratio seen from @xmath12@xmath3 m , as neither model has the strong temperature inversion that would be necessary to create emission features . in _
b _ we plot the mie absorption efficiency of 1@xmath3 m silicate particles of forsterite ( mg@xmath4sio@xmath9 ) and enstatite ( mgsio@xmath13 ) .
larger particles have a grayer spectrum and do not exhibit the canonical silicate feature @xcite .
the location of this possible emission feature in the hd 209458b spectrum is suggestive , but it will take higher signal - to - noise observations to determine if silicates have indeed been detected .
hd 209458b is a better candidate than hd 189733b for having silicate clouds in its visible atmosphere , due to the higher stellar insolation and hotter atmosphere . as was noted above and shown in _
a _ , the irac observation at 8@xmath3 m by @xcite indicates a downturn that is in fact consistent with h@xmath4o vapor absorption .
we now take a closer look in . our model ( red )
is a good fit to the data , while in blue we show a @xcite model , as taken from @xcite , that appears to also be a good fit to the photometry , albeit with a much deeper water feature . binning the 31 irs data points from 7.57 to 9.39@xmath3 m ( which overlaps the irac band ) into a single point
, we find an eclipse depth of @xmath14 , @xmath25.3@xmath10 removed from the knutson et al . irac point .
if variation in planetary or stellar flux were to account for this difference ( e.g. * ? ? ?
* ; * ? ? ?
* ) , the variability would have to be @xmath225% near 8@xmath3 m .
if there is no variability and both the irs spectrum and irac data point are accurate as published , we can examine how small the flux ratio must be short - ward of the bluest irs point at 7.57@xmath3 m .
we have created synthetic data with the same spectral resolution of the irs data set . in triangles
we show a constant flux ratio of 0.00156 between 6.3@xmath3 m ( the blue end of the irac filter ) and 7.57@xmath3 m . in boxes we show a synthetic ratio that decreases linearly from the 0.00430 binned value .
the ratio must fall to _ zero _ by 6.79@xmath3 m to give a band - average eclipse depth that agrees precisely with the irac data point . in _ b _ we showed that the only potential absorbers predicted from equilibrium chemistry from 6 to 10@xmath3 m are h@xmath4o and ch@xmath9 ,
neither of which possess a sharp increase in opacity short - ward of 7.7@xmath3 m .
with the assumption that the irac secondary eclipse depth is accurate , and given the severe absorption feature that would be needed just short of 7.57@xmath3 m to fit this point given the irs spectrum , we conclude that the published irs spectrum short - ward of 10@xmath3 m does not reflect the true spectrum at these wavelengths , where the irac data point suggests absorption by water vapor .
one possibility toward reconciling the data sets would be to shift the entire irs spectrum downward .
indeed , for hd 209458b , @xcite were unable to accurately determine the absolute depth of the secondary eclipse with the irs , and therefore tied their spectrum to an unpublished irac 8@xmath3 m observation of the secondary eclipse . given this issue , the absolute depths of the @xcite data may also require a shift .
however , the downward shift that would be required ( @xmath225% ) would lead to much poorer agreement between this data set and the @xcite point at 16@xmath3 m .
in addition , although both teams corrected for slit losses , this was only done in a gray manner .
since the point spread function is larger at longer wavelengths , it is possible that slit losses would be more pronounced at the red end of the spectrum .
greater flux at longer wavelengths , together with a downward scaling , may be able to bring the observed spectrum for hd 189733b into agreement with the photometric points .
future work on all of these data sets may be beneficial , as , for example , the relative photometry for these measurements is more than an order of magnitude more precise than irac s initial specifications @xcite .
the _ p - t _ conditions encountered by the atmospheres of hot jupiters is coincident with that found in the atmospheres of l- and t - type brown dwarfs . although the actual _ p - t _ profiles of irradiated egps are of course different , the underlying atmospheric physics and chemistry are similar and indeed the actual computational models are in some cases virtually identical , such that a comparison of brown dwarf models to data can illuminate theoretical areas that are well understood .
comparisons of our group s models to near- and mid - infrared spectra of l- and t - dwarfs , including superb observations with the irs , have generally found very good to excellent matches with data @xcite .
specifically our `` cloudy '' l - dwarf and `` clear '' t dwarf models ( * ? ? ?
* m. s. marley et al . in prep . )
fit the depths of the water bands lying between y , j , h , and k bands as well as the @xmath15 water band @xcite expected in the irs spectra of hot jupiters .
there are , however , departures from the data in regions where molecular opacities are poorly known as well as some systematic spectral shape differences in k and n bands . since the marley et al .
models have been demonstrated to fit the near and mid infrared data of the cloud - free t dwarfs we gain confidence in our radiative transfer @xcite , opacity database @xcite , and chemistry calculations @xcite .
this experience gives us confidence these areas are becoming well understood .
of course , since isolated brown dwarfs lack incident flux that drive atmospheric dynamics and photochemistry , these additional areas may remain challenging topics of hot jupiter characterization for some time .
it is in deviations from 1d or simple day / night models that these aspects will become apparent .
the common conclusion of the @xcite and @xcite secondary eclipse observations was the apparent lack of absorption due to water vapor .
however , for hd 189733b we have shown that photometry indicates a planet - to - star flux ratio that is not as featureless as indicated by the published irs data .
if variability in planetary flux is to make up for this discrepancy , it must reach @xmath225% .
the apparent downturn in the flux ratio from 16 to 8@xmath3 m implied by photometry is consisent with a model that predicts absorption due to water shortward of 10@xmath3 m .
as was shown in , there is considerable diversity in feature depths among models , perhaps due to differences in chemistry , opacities , or treatment of incoming stellar flux . for hd 189733b , here
we have shown that our model is a good match to 8 and 16@xmath3m - band photometry .
in addition , both our 1d and 3d models fit the hd 209458b observations reasonably well . our conclusion that water vapor absorption is present in these atmospheres
is supported by the recent analysis of the @xcite transit data set by @xcite , who concluded that water absorption in present in the atmosphere of hd 209458b in a band at @xmath16 nm .
of course the characterization of the atmospheres of these planets is just beginning .
both planets have been observed in all 4 irac bands , but this data has not yet been published .
additional irs spectroscopic observations for hd 209458b ( deming and collaborators ) and hd 189733b ( grillmair and collaborators ) will be obtained this year .
importantly , these observations will go down to 5.5@xmath3 m , fully overlapping the irac 8@xmath3 m band , which will allow for more accurate normalization of the obtained spectra .
+ + we thank j. richardson , m. cushing , d. deming , c. grillmair , d. charbonneau , and the referee for useful discussions .
j. j. f. acknowledges the support of a spitzer fellowship from nasa and m. s. m. from the nasa origins and planetary atmospheres programs . | published mid infrared spectra of transiting planets hd 209458b and hd 189733b , obtained during secondary eclipse by the infrared spectrograph ( irs ) aboard the _
spitzer space telescope _ , are predominantly featureless . in particular these flux ratio spectra do not exhibit an expected feature arising from water vapor absorption short - ward of @xmath0 . here
we suggest that , in the absence of flux variability , the spectral data for hd 189733b are inconsistent with @xmath1m - photometry obtained with _
spitzer _
s infrared array camera ( irac ) , perhaps an indication of problems with the challenging reduction of the irs spectra .
the irac point , along with previously published secondary eclipse photometry for hd 189733b , are in good agreement with a one - dimensional model of hd 189733b that clearly shows absorption due to water vapor in the emergent spectrum .
we are not able to draw firm conclusions regarding the irs data for hd 209458b , but spectra predicted by 1d and 3d atmosphere models fit the data adequately , without adjustment of the water abundance or reliance on cloud opacity .
we argue that the generally good agreement between model spectra and irs spectra of brown dwarfs with atmospheric temperatures similar to these highly irradiated planets lends confidence in the modeling procedure . |
gravitational interaction between galaxies is now considered to be one of the primary mechanisms of galaxy evolution .
major mergers , composed of two or more galaxies with comparable masses , are widely believed to provide a way to form elliptical and spheroidal galaxies @xcite .
furthermore , galaxy interaction is known to increase star formation activities as seen in the increasing fraction of tidally distorted morphologies in ultraluminous infrared galaxies ( ulirgs : l@xmath4 l@xmath5 , sfr @xmath6 m@xmath5 yr@xmath7 ; * ? ? ?
in addition , the number of merging galaxies appears to increase at high redshifts ( e.g. * ? ? ? * ) .
the obvious first step in characterizing this process and the response of the cold gas and its relation to merger induced starburst activity is to observe nearby merging galaxies .
the antennae galaxies ( ngc 4038/9 ) is the nearest colliding galaxy pair observable from the northern hemisphere located at a distance of 22 mpc @xcite .
numerical simulations of the antennae ( e.g. * ? ? ?
* ) suggested that the antennae are at the intermediate stage of the collision .
@xcite suggest that the two spiral galaxies passed the initial pericenter 210 myr ago , and the final coalescence will take place in about 100 myr .
the antennae have been observed at all available wavelengths from radio to x - ray .
@xcite observed the antennae with the wide field planetary camera 2 on the _ hubble space telescope _
( hst ) and identified thousands of stellar clusters which have formed during the merging process .
they found that most of the star formation occurs in the form of massive compact star clusters , which are formed from supergiant molecular clouds .
previous observations have found that most of the active star forming regions are concentrated in the region where the two galaxies overlap .
herschel_-pacs maps at 70 , 100 and 160 @xmath8 revealed that both nuclei are brighter than the hii regions in the arms , but the brightest emission in three bands comes from the overlap region @xcite . the mid - infrared ( mir ) and far - infrared ( fir ) emission traces buried star formation , which is obscured by dust and therefore is invisible at optical wavelengths .
they estimated that the integral sfr is @xmath9 22 m@xmath5 yr@xmath7 and the sfr in the southernmost knot of the overlap region with an area of 4.68 kpc@xmath10 is @xmath9 3 m@xmath5 yr@xmath7 .
they also found that several emission knots in the overlap region are in different evolutionary stages .
low angular resolution observations have revealed strong [ cii ] emission , which arises mainly from photo dissociation regions created by far - ultraviolet photons from hot young stars , near the overlap region @xcite .
in addition , the southernmost knot of the overlap region is currently the most active area with a very young stellar complex .
@xcite observed 4 and 6 cm radio continuum emission using the vla and estimated that the overall star formation rate seems to be a factor of @xmath9 20 higher than normal galaxies .
they found that the compact radio source with the strongest thermal radio emission requires @xmath9 5000 o5 stars to account for the free - free emission .
they estimate a supernova rate of @xmath9 0.2 yr@xmath7 if the sources with steep radio spectra are dominated by non - thermal emission produced in supernova remnants . the distribution and the characteristics of molecular gas have been investigated in detail using mm / submm telescopes around the world . @xcite obtained the co ( 10 ) map using the nrao 12 m single - dish telescope and found that the distribution of the co ( 10 ) emission is similar to those of the mir and fir emission at a kiloparsec scale .
@xcite identified five supergiant molecular complexes ( sgmcs ) with masses of ( 36 ) @xmath2 10@xmath11 m@xmath5 in the overlap region using the @xmath1co ( 10 ) map obtained at the owens valley radio observatory ( ovro ) . @xcite
have obtained single - dish maps at the nobeyama 45 m telescope in the co ( 10 ) line and using the james clark maxwel telescope ( jcmt ) in the co ( 21 ) and co ( 32 ) lines .
each nucleus contains a molecular mass of @xmath12 m@xmath5 and the overlap region accounts for approximately 42@xmath13 of total @xmath1co ( 10 ) flux of the inner disk of the antennae .
they found that the co integrated intensity ratios are high in the overlap region .
this is the first case where high co ( 32)/co ( 10 ) ratios are found outside a galactic nucleus .
furthermore , @xcite obtained co ( 10 ) and co ( 21 ) maps using the iram 30 m millimeter radio telescope and co ( 32 ) map using the heinrich hertz telescope .
the total molecular gas mass of the system ( @xmath9 10@xmath14 m@xmath5 ) is about twice the total gas mass of the milky way .
however the molecular gas mass within the 1 kpc nucleus of ngc 4038 and ngc 4039 exceeds that of the central 1 kpc of the milky way by a factor of almost 100 , revealing high gas concentration into the two nucleus .
@xcite also derived the line ratios ranging between 0.50 and 0.66 , which are lower than the ratios estimated by @xcite .
the purpose of our work is to investigate the physical properties and the kinematics of the co ( 32 ) emitting molecular complexes in the antennae using @xmath15 ( 1@xmath16 corresponds to about 107 pc ) resolution co ( 32 ) observations .
the co ( 10 ) rotational transition has a critical density of @xmath17 @xmath18 and an upper energy level @xmath19 5 k , whereas the co ( 32 ) rotational transition has @xmath20 @xmath18 and @xmath19 33 k. thus the co ( 32 ) emission is a better tracer of denser and warmer molecular gas than co ( 10 ) emission .
this paper is organized as follows .
we describe our observations in 2 and our results in 3 . in 4 , we provide a discussion of our results .
we present the properties and the distribution of identified molecular complexes ( 4.1 ) , the co ( 32)/(10 ) brightness temperature ratio ( 4.2 ) , the kinematics in ngc 4039 ( 4.3 ) , and a possible molecular gas bubble in ngc 4038 ( 4.4 ) .
we summarize and conclude this paper in 5 .
we conducted co ( 32 ) observations of main bodies of the ngc 4038/9 with the submillimeter array ( sma ; * ? ? ?
* ) on march 2005 and march 2008 in the compact configuration and december 2008 in the extended configuration .
the data were obtained using a five - pointing mosaic where the phase centers were chosen to cover the main co ( 32 ) emission peaks seen in the jcmt map of @xcite .
the coordinates of the phase centers are shown in table [ tb : fov ] and the fields of view are shown in figure [ fig : im ] ( right ) . the total on source
observing time was approximately 10 hours .
the sis receivers were tuned to the frequency of the co ( 32 ) line ( @xmath21=345.796 ghz ) redshifted to the systemic velocity of 1634 km / s .
the primary beam of the array at this frequency is @xmath22 .
the correlator had a 2 ghz total bandwidth with a 0.8 mhz ( 0.7 km s@xmath7 ) frequency resolution .
the quasar 3c 279 was observed for phase and amplitude calibration .
absolute flux calibration was performed using uranus .
the uncertainty of flux calibration is 20 @xmath13 .
data inspection and calibration was carried out using the mir package written in idl and imaging was done using the miriad package .
the synthesized beam size is @xmath23 , which corresponds to 150 pc @xmath2 120 pc with a position angle of 29.4@xmath24 .
natural weighting of the visibilities is used to maximize the sensitivity , and the achieved rms noise level in 10 km s@xmath7 channel maps is 38 mjy beam@xmath7 . the continuum emission is not detected with an upper limit of 17.7 mjy ( 3 @xmath25 ) .
the co ( 32 ) integrated intensity map overlaid on optical / mir images is shown in figure [ fig : im ] .
in addition , we compare the co ( 32 ) map with multiwavelength images from radio and to x - ray ( see appendix a ) . in general , the locations of the co ( 32 ) emitting molecular complexes are consistent with the positions of the sgmcs detected by @xcite in co ( 10 ) , but the higher angular resolution achieved in the co ( 32 ) map shows details that were unseen in the co ( 10 ) image .
both galaxies show central co ( 32 ) concentrations near the nuclei and have one gas arm roughly distributed along the dust lanes toward southwest seen in the hst 435 nm image ( figure [ fig : im ] ( left ) ) . the strongest co ( 32 ) emission is associated with the nucleus of ngc 4038 with a peak flux density ( 10 km s@xmath7 velocity resolution ) of 2.0 @xmath26 0.4 jy beam@xmath7 .
the co ( 32 ) emission near the nucleus of ngc 4039 is composed of at least two distinct large scale components separated by 560 pc .
the overlap region consists of numerous emission peaks , with complex structure seen in larger ( @xmath9 1 kpc ) molecular complexes .
spatial correlation with optical emission is poor in most of the molecular complexes in the overlap region , mostly because these regions are strongly affected by dust extinction .
however , the general association with the _ spitzer _ 8 @xmath27 m emission ( figure [ fig : im ] ( right ) ) , which traces the dusty regions associated with co ( 32 ) emitting molecular clouds , is much better than the optical association .
the total integrated co ( 32 ) flux of main body of the antennae is 4.4 @xmath2 10@xmath28 jy km s@xmath7 .
the two nuclei of ngc 4038 and ngc 4039 contain 1.4 @xmath2 10@xmath28 and 9.4 @xmath2 10@xmath10 jy km s@xmath7 , respectively .
the total flux in the overlap region is 2.1 @xmath2 10@xmath28 jy km s@xmath7 , which is approximately half of the total integrated flux over the entire galaxy .
the percentage of missing flux derived by comparing the sma and jcmt flux @xcite are 55@xmath13 in ngc 4038 , 35@xmath13 in ngc 4039 and 67@xmath13 in the overlap region .
the low missing flux in ngc 4039 suggests the compact nature of the co ( 32 ) emitting clouds , while the relatively large missing flux suggests the presence of a significant extended ( @xmath29 3 kpc ) component in the overlap region .
the central @xmath30 pc in ngc 4038 is unresolved by the @xmath31 beam , but the strong signal allows us to use the extended configuration data alone for a higher angular resolution image ( figure [ fig : extended ] ) .
the synthesized beam size is @xmath32 by adopting natural weighting of the visibilities .
most of the diffuse and extended emission below 15@xmath25 in the @xmath31 resolution map made using all data ( figure [ fig : im ] ) is resolved out in the high resolution map .
the deconvolved size of the main component is 230 pc ( @xmath33 ) @xmath2 150 pc ( @xmath34 ) , and the fwhm velocity width is 110 km s@xmath7 .
the brightness temperature @xmath35 ( the peak intensity ) scales with the beam filling factor @xmath36 and the molecular gas temperature @xmath37 . from the peak intensity of 13 k and assuming that the typical molecular gas temperature @xmath37 of the co ( 32 ) emission is @xmath38 30 k @xcite , the beam filling factor is @xmath38 0.4 .
thus the distribution of the co ( 32 ) emission is likely more compact than the region observed with the @xmath39 pc beam .
the velocity fields of ngc 4038 , ngc 4039 , and the overlap region are shown in figure [ fig : mom1 ] ( a ) , ( b ) and ( c ) , respectively . at this spatial resolution ,
the velocity at the k@xmath40-band nucleus of ngc 4038 @xcite is @xmath9 80 km s@xmath7 lower than the surrounding medium , which may suggest a rotation or gas streaming into the nuclear region along the arm . in contrast , the co ( 32 ) emitting gas near the nucleus of ngc 4039 shows a large velocity gradient ( 310 km s@xmath7
kpc@xmath7 ) from the north - west to south - east , with the velocity ranging from 1560 km s@xmath7 to 1640 km s@xmath7 .
furthermore , the gas component distributed along the optical dust lane of ngc 4039 also shows a steep velocity gradient of 170 km s@xmath7 kpc@xmath7 .
the implications of these kinematic features will be discussed in 4.3 . in the overlap region ,
the velocity of the gas ranges from v@xmath41 = 1300 to 1650 km s@xmath7 .
the large scale kinematics in the overlap region show a north - south velocity gradient , which is consistent with the velocity field seen in the lower resolution jcmt map @xcite .
the new sma map reveals that each molecular component shows complex kinematics , with velocity gradients in excess of 100 km s@xmath7 kpc@xmath7 in each molecular complex , some as large as 300 km s@xmath7 kpc@xmath7 .
the channel maps of the high resolution cube of ngc 4038 with 10 km / s velocity resolution are shown in figure [ fig : cmap_ex ] .
the rms noise level ( 1@xmath25 ) is 70 mjy beam@xmath7 . at this angular and velocity resolution ,
the morphology of the emission feature is highly complex around the nucleus , with the emission distributed in a semicircle around the known nonthermal radio source dominated by synchrotron radiation from supernova remnants @xcite .
the radio source does not have detectable co ( 32 ) associated at an rms of 70 mjy beam@xmath7 which corresponds to a molecular gas mass limit of 9 @xmath2 10@xmath42 m@xmath5 .
further discussion on the possible origin of this co ( 32 ) arc is provided in 4.4 .
we identified molecular complexes by applying the automatic clump identification algorithm clumpfind @xcite to the co ( 32 ) data .
one of programs in clumpfind , clfind , works on 3d ( ra , dec , velocity ) data cubes , searches for local peaks of emission , and follows them down to lower intensity levels .
then it decomposes the data cube into clumps in which the emission is concentrated . while the 2@xmath25 threshold is the recommended value for identifying robust molecular clumps @xcite
, our experiments find that the total number of identified molecular complexes varies by 15@xmath13 for a 10@xmath13 deviation from the 2@xmath25 threshold .
this is likely because the program falsely detects ( @xmath43 ) sidelobes of the real emission .
we find that the number of identified clumps only changes by a few percent for a threshold of @xmath44 .
thus we chose 2.6@xmath25 for robustly detecting molecular clumps with high significance in the following analysis .
a total of 57 molecular complexes are identified , 33 ( @xmath45 ) of which are located in the overlap region .
the properties of the molecular complexes such as position , systemic velocity ( @xmath46 ) , radiusv(@xmath47 ) , and velocity dispersion ( @xmath48 ) were measured using the program clstats @xcite and are summarized in table [ tb : clump ] ( appendix b ) .
we estimate the error in the position , radius , and velocity to be @xmath49 , 44 pc , and 5 km s@xmath7 , respectively .
the radii of the molecular complexes ranges between 85 and 348 pc , whereas the velocity dispersion ranges between 6 and 36 km s@xmath7 .
the molecular gas mass is derived by ; @xmath50 where @xmath51 is the co luminosity - to - h@xmath52 mass conversion factor , @xmath53 is the velocity integrated flux , @xmath54 is the observing frequency , and @xmath55 is the luminosity distance @xcite .
we used @xmath51 = 4.8 m@xmath5/k km s@xmath7 pc@xmath10 @xcite scaled by the average co ( 32)/(10 ) line intensity ratio for each region ( see 4.2 ) .
the molecular masses are estimated to be 10@xmath56
10@xmath11 m@xmath5 .
in addition , the virial mass of the molecular complexes are estimated from @xmath57 , where @xmath47 and @xmath58 is the radius and the velocity dispersion of the molecular complex , and @xmath59 is the gravitational constant .
the molecular complex is gravitationally bound when their @xmath60 is smaller than unity .
we find that 84 @xmath13 of the molecular complexes are gravitationally bound , including the most massive molecular complex located north of the ngc 4038 nucleus .
we note that the velocity dispersions used here are likely overestimated by a factor of 2 3 ( 4.1.3 ) , and hence the derived virial masses are also significantly overestimated .
this suggests that almost all of the molecular complexes observed at this spatial resolution have @xmath61 1 and that they are gravitationally bound .
the distribution of the radii of the molecular complexes in each region are shown in figure [ fig : r_hist ] .
the kolmogorov - smirnov test gave p - value = 0.002 ( table [ tb : p - value ] ) for the comparison of the distribution between ngc 4038 and ngc 4039 , suggesting that the two populations are significantly different .
there are molecular complexes with large radii ( @xmath62 180 pc ) in ngc 4038 , but not in ngc 4039 .
the median radii in ngc 4038 and ngc 4039 are 210 and 120 pc , respectively , and thus the molecular complexes near the nucleus of ngc 4038 are on average 1.8 @xmath26 0.7 times larger than those in ngc 4039 . the size distribution in the overlap region range between 90 and 350 pc .
the average size of the molecular complexes in the southern part of the overlap region ( declination @xmath63 -18@xmath64 53@xmath65 00@xmath66 ) is 1.3 times larger than the average size of those in the overlap region . in addition , the p - value = 0.975 and 0.017 ( table [ tb : p - value ] ) for the comparison of the distribution between the southern part of the overlap region and ngc 4038 , and ngc 4039 , respectively .
this suggests that the radii between ngc 4038 and the southern part of the overlap region are similar . @xcite
estimated the sfrs of seven individual emission knots from their _ herschel _ maps .
knot k1 , which corresponds to the southern part of the overlap region , has the highest sfr ( 2.78 m@xmath5 yr@xmath7 ) and knot nn , which corresponds to the central region of ngc 4038 , has the second highest sfr ( 1.17 m@xmath67yr@xmath7 ) .
therefore the larger molecular complexes , which are possibly formed by clustering around local peaks of the gravitational potential wells , may be used as the signposts of current dusty star forming regions .
the velocity dispersions of the molecular complexes in each region are shown in figure [ fig : sigma_hist ] .
we conducted a kolmogorov - smirnov test on the distribution of the velocity dispersions , and obtained p - value = 0.664 0.973 ( table [ tb : p - value ] ) , suggesting that the velocity dispersions among the three different regions are similar .
in ngc 4039 , molecular complexes with larger velocity dispersions ( 25 35 km s@xmath7 ) are located within 500 pc from the nucleus and those with a smaller velocity dispersion ( 10 25 km s@xmath7 ) are located in the gas arm . comparing the molecular complexes in the antennae with clumps in m64 which is a molecule - rich spiral galaxy located at 4.1 mpc @xcite and the galaxy @xcite , the molecular complexes with the same velocity dispersion have an order of magnitude larger molecular mass in the antennae . on the gmc scales ,
it is known that the radius is proportional to the velocity dispersion @xcite .
@xcite find a correlation between the radii and velocity dispersions of gmcs in m 33 which is a normal spiral galaxy located at 850 kpc , and the correlation appears to hold even in the outer disk @xcite .
in other instances , @xcite showed that the gmcs in the large magellanic cloud ( lmc ) follow a similar correlation , but it is offset from that found for the gmcs in the inner galaxy @xcite . in order to investigate if such correlation exists in the antennae at gma scales , we plot the relation between the radii and the velocity dispersions of the identified molecular complexes ( figure [ fig : r - sigma ] ) .
the correlation coefficients derived from the least squares fit to the different regions are 0.68 in ngc 4038 , 0.25 in ngc 4039 , and 0.01 in the overlap region , suggesting the absence of a strong correlation in each region of the antennae . at these scales ,
the velocity dispersion likely reflects the relative bulk motion of a collection of co ( 32 ) emitting clumps , and not the intrinsic velocity dispersions pertaining to the individual gmcs .
hence the absence of a strong correlation at size scales of gmas is somewhat expected , especially in the dynamically disturbed parts of the galaxy such as the overlap region .
we investigate further the relation between radius and velocity dispersion of molecular complexes in ngc 4038 because the correlation coefficient in ngc 4038 is the highest among three regions . by using clumpfind on the high resolution data
, we identified 11 molecular complexes that are mostly located 100 pc north of ngc 4038 .
their radii range between 40 and 80 pc and are about the same size as the gmcs in m 33 .
the relatively tight correlation between radius and velocity dispersion is seen in the gmcs detected in a quiescent spiral galaxy m 33 @xcite ( figure [ fig : r - sigma ] ) .
however , the velocity dispersions of the molecular complexes in ngc 4038 are 2 3 times larger than the gmcs in m 33 and distributed between 10 and 30 km s@xmath7 .
this is further evidence suggesting that the velocity dispersion does not reflect the intrinsic velocity dispersion but the relative bulk motion of co ( 32 ) emitting clumps . finally , we note that the lack of significant difference in the velocity dispersion of molecular complexes between ngc 4038 and ngc 4039 may be a result of limited sensitivity and angular resolution .
the clump identification routine is unable to separate the apparently smooth and continuous distribution of molecular gas into smaller clumps . the green circles in figure [ fig : cluster ]
show the positions of the identified molecular complexes and the crosses represent the positions of 50 most luminous ( at v - band ) and/or the 50 most nir - bright clusters whose ages are older than 1 myr @xcite .
overall , only 17 out of 57 molecular complexes ( 30 @xmath13 ) host one or more of these star clusters with a median age of 4 myr estimated from the ages of the star clusters @xcite , not considering projection effects and extinction . in a related study
, @xcite identified star clusters using the vlt k@xmath40-band and found that about 70 @xmath13 of the star clusters with masses @xmath68 are younger than 10 myr and most of the young clusters with the age @xmath63 6 myr are located in the overlap region .
although @xcite suggest that the co ( 10 ) obtained at a spatial resolution of @xmath69 in the overlap region shows spatial correlation with the location of the star clusters , our co ( 32 ) emission map with a spatial resolution of @xmath31 shows a clear offset from the star clusters , assuming that the star clusters identified using the v - band image overlap with those identified using the k@xmath40-band image .
the offset is evident in the southern part of the overlap region , where the co ( 32 ) emission is strong and the co line ratio is relatively high .
hence , the absence of a significant spatial correlation between the locations of the molecular complexes and star clusters may suggest that the molecular gas traced in co ( 32 ) emission corresponds to the regions of future star formation , or regions where star formation occurred in the past @xmath9 1 myr because no star cluster which is younger than 1 myr correspond to the molecular complexes .
this timescale is similar to the typical timescale for disassociation from the molecular cloud which is under 3 myr @xcite .
we investigated the integrated brightness temperature co ( 32)/(10 ) ratios using the sma co ( 32 ) and ovro co
( 10 ) data @xcite . while we assume the same beam filling factor for both transitions , it is possible that the true distribution of the co ( 32 ) emitting cloud is more clumpy and concentrated than the clouds emitting co ( 10 ) emission .
therefore , the peak line ratios we estimate here are lower limits .
we clipped the visibilities so that both data have the same shortest uv range ( minimum uv distance = 10.48 k@xmath70 = 9.13 m ) and then convolved the co ( 32 ) data to the angular resolution ( [email protected] @xmath2 [email protected] ) of the co ( 10 ) image before estimating the line ratios .
finally , the line ratios are clipped at the 3@xmath25 level .
the resultant ratio map is shown in figure [ fig : ratio ] .
we separate the ratio map into five regions ( ngc 4038 , ngc 4039 , complex 1 , complex 2 , and complex 3 ) and estimate the mean ratio for each region .
the mean integrated intensity ratio ranges from 0.3 to 0.6 ( table [ tb : ratio ] ) . here
the error bars are derived from the uncertainties in the flux calibration .
on average , the co line in the antennae is not thermalized up to @xmath72 = 32 .
they are slightly less than the average ratios estimated by @xcite using single dish telescopes , which ranged between 0.67 and 0.91 .
in contrast , the line ratios are close to unity in the southern part of complex 3 and the two nuclei .
@xcite derived 0.93 for the central regions of 12 dusty u / lirgs in the local universe , which is also in agreement with the starburst galaxies arp220 and m82 where the derived ratios are @xmath73 ( e.g. * ? ? ?
thus , the average excitation conditions analyzed at @xmath9 1 kpc scale in the antennae are close to the average properties of nearby galaxies ( e.g. 0.44 is the average ratio among 7 nearby galaxies ; * ? ? ?
* ) , except for two localized regions where the excitation conditions are similar to the central regions of starburst galaxies .
finally , we mention two issues that likely affect the analysis and interpretation of the line ratios .
one issue is the inapplicability of the rayleigh- jeans approximation , especially for the higher frequency co ( 32 ) line . in such case
the brightness temperature ratio can be less than unity ( 0.8 0.9 for a temperature of 40 k ; see @xcite ) , assuming a thermalized and optically thick , single component cloud .
the other possibility arises from the filling factor difference between the co ( 32 ) and co ( 10 ) lines .
such a multi - component cloud may be more realistic than adopting a single component cloud , and higher angular resolution alma / aca observations of both lines should provide us with the exact distribution of each line tracers in the future .
the mean @xmath1co ( 32)/(10 ) ratio of ngc 4038 is 0.6 @xmath26 0.2 , which is the highest among the averages of the five complexes .
the spatial distribution of the ratio presented in figure [ fig : ratio ] shows that the ratio is 0.75 in the center , and this is two times higher than the ratios in the outer region .
the regions with particularly high line ratios correspond well with the peaks of the mir and nir emission , qualitatively suggesting the correlation between high line ratio and embedded star formation .
this is consistent with observations in nearby galaxies where the ratio appears to increase from the outer disk to the central regions ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
a similar trend is also seen in the milky way galaxy , where the ratio increases from @xmath74 in the inner disk to @xmath75 towards the galactic center where the star formation is more active @xcite .
furthermore , signatures of more recent star formation activity in ngc 4038 are seen in optical and x - ray observations ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , suggesting the coexistence of a variety of star clusters formed at various stages of the tidal activity .
the line ratio near the nuclear region of ngc 4039 is high ( @xmath76 ) , despite a significantly lower mean ratio ( [email protected] ) .
the star formation activity in the central region of ngc 4039 is only modest ( sfr = 0.50 m@xmath77 yr@xmath7 ; * ? ? ?
* ) , which is only half of the value derived in ngc 4038 .
the absence of the br@xmath78 line in the stellar continuum is an indication of the dominance of old giants and red supergiants , further suggesting inactivity @xcite .
thus it is not likely that the high line ratio in the nuclear region is produced by current star formation .
the co line could be collisionally heated and excited up to high energy levels if ngc 4039 harbors an agn @xcite .
a bright diffuse x - ray emission region is associated with the k@xmath40-band nucleus within 220 pc , and also with the region with high line ratio .
the radio continuum has a steep spectrum , which suggests non - thermal emission arising from supernova remnants @xcite , or x - ray binaries @xcite .
furthermore , @xcite found that the observed infrared continuum is consistent with a pure starburst model and concluded that the spectrum of the nucleus is produced by emission from dust that is heated by star formation alone .
thus , these multi - wavelength properties of ngc 4039 provide evidence against the presence of an agn .
another possible cause for the high line ratio is gas collision arising from violent gas kinematics . as shown in 3 , the distribution of the co ( 32 ) gas in ngc 4039
is composed of two main components ; the nuclear concentration and the single - arm gas that appears to stream along the optical dust lane .
although the spatial resolution of the ratio image is limited to 500 pc , the region where the two components connect is consistent with the position that displays the highest ratio .
the properties of the molecular complexes located at this region are markedly different from the average properties of ngc 4039 , where the radii of the two molecular complexes are larger .
in addition , the velocity dispersion of the molecular complexes in this regions are @xmath79 larger than the average of ngc 4039 .
furthermore , the recession velocities ( see table [ tb : clump ] ) of these molecular complexes are consistent with the velocity of the clouds with the highest ratio .
this observational evidence suggests a scenario where the molecular gas flows into the central disk ( see 4.4 ) , collides , and results in high excitation .
future high resolution observations of shock tracers is necessary to confirm this scenario .
the mean @xmath1co ( 32)/(10 ) ratio is 0.4@xmath26 0.1 in complex 1 , 0.3@xmath26 0.1 in complex 2 , and 0.5@xmath26 0.1 in complex 3 .
although the average ratios are all consistent within the errors , gradients in the line ratio are seen in each complex ( figure [ fig : ratio ] ) , with a particularly steep gradient in complex 3 .
@xcite derived the fuv
24 @xmath27 m broad - band spectral energy distribution ( sed ) and found that the uv / optical emission gradually become redder and weaker from complex 1 ( which corresponds to their region 6 ) to complex 3 ( region 4 ) .
this suggests that the ratio of young star population ( @xmath63 10 myr ) to intermediate / old star population ( @xmath62 10 myr ) increases from complex 1 to 3 . the continuum emission at mir wavelengths in complex 1 ( sgmc 1 in @xcite ) and complex 2 ( sgmc 2 ) are flat , but complex 3 ( sgmc 35 ) contains strong and rising continuum emission @xcite .
the mid - infrared continuum emission in complex 3 can be produced by massive o stars .
similarly , strong fir continuum emission seen in complex 3 ( k1 ) has the highest sfr and sfe and maximum dust temperature @xcite .
this observational evidence suggests that complex 3 is likely to be a very young complex and star forming histories are different from complex 1 and 2 .
this is consistent with the higher co line ratio in complex 3 .
the overlap region of the antennae was observed with a single pointing in the @xmath1co ( 21 ) line at the sma . from our analysis of the archival data , the distribution of the co ( 32 ) and co ( 21 ) maps
are similar overall , but subtle differences are seen particularly in complex 3 .
we found that the co ( 21 ) peak in complex 3 is located 100 pc north of the co ( 32 ) peak , and the co ( 10 ) peak is further displaced toward the north by 200 pc as shown in figure [ fig : complex3 ] .
@xcite detected an h@xmath52o maser within 100 pc of the co ( 32 ) peak , and estimated from the vla 3.6 cm continuum that the ionized gas is equivalent to 2000 5000 o@xmath80 stars .
furthermore , the most massive cluster is embedded 350 pc southeast of complex 3 @xcite .
these are observational indicators that suggest the earliest stages of active star formation in the southern part of complex 3 are entangled with more mature clusters that were formed 1 myr ago .
the high line ratio coincides with this region , suggesting that heating of the interstellar medium due to uv emission from newly born stars is responsible for collisionally exciting the co gas to the @xmath72=32 level .
the spatial distribution of the co ( 32 ) emission in ngc 4039 ( figure [ fig : pv_n9 ] ( left ) ) clearly shows two distinct molecular gas components that are separated by a projected distance of 560 pc .
one component ( component a ; see figure [ fig : pv_n9](left ) ) is apparently distributed around the nucleus and the other component ( component b ; see [ fig : pv_n9](left ) ) is distributed along the dust lane seen in the b - band image .
the masses of components a and b are 5.2 @xmath2 10@xmath11 and 4.5 @xmath2 10@xmath11 m@xmath5 , respectively .
the pv diagram of component a is shown in figure [ fig : pv_n9 ] ( right ) .
the systemic recession velocity of ngc 4039 determined from our data is 1610 km s@xmath7 .
the mean gas velocity of component a is blue- and redshifted by -50 and + 30 km s@xmath7 with respect to the k@xmath40-band nucleus to the west and east side , respectively . then assuming that component a is in the plane of the galaxy , this suggests that component a rotates around the galactic center of ngc 4039 .
we compare the dynamical mass and molecular mass of component a and find that the disk is dynamically stable , assuming that component a is a disk with a rotation velocity of 40 km s@xmath7 and the disk inclination of @xmath81 = 60@xmath24 @xcite .
@xcite found a high occurrence of rotating 100 pc scale nuclear disks or rings from their molecular gas observations of ulirgs .
while the kinematic signature of a disk in ngc 4039 is less obvious than the ulirgs in @xcite mainly because of a lower s / n ratio , presence of such nuclear disks may be a common phenomenon in the nuclear regions of colliding systems . a clear characteristic seen in ngc 4039 is that this component appears to rotate in the opposite direction from the galactic rotation defined by the direction of the spiral arms seen in optical images , which is defined from the direction of the trailing arm .
@xmath82-body / smoothed particle hydrodynamic ( sph ) simulation performed by @xcite predicts that a non - axisymetric potential can enhance the stellar torques exerted on the gas , causing gas inflows into the central regions in a merging process .
these inflows may be kinematically distinct , as shown in the simulation studies by @xcite .
component b shows the predicted kinematic signature of gas streaming along the arm ( @xmath9 1 kpc ) , as shown in figure [ fig : mom1 ] .
in addition , a high line ratio is seen at the location where components a and b connect and the velocity dispersion of the molecular complexes are @xmath79 larger than the average of ngc 4039 as shown in 4.2.1 . if component b is indeed related to gas inflow , it is then possible that high gas excitation is caused by collisions between the inflowing gas component and the counter - rotating disk .
the high resolution image ( @xmath83 ; figure [ fig : extended ] ) allows us a detailed investigation of the molecular gas surrounding the nuclear region of ngc 4038 .
the molecular and virial masses of the central component are 3.2 @xmath2 10@xmath11 m@xmath5 and 2.0 @xmath2 10@xmath11 m@xmath5 , respectively , suggesting that this component is gravitationally bound .
furthermore , by using clumpfind on the high resolution data , we identified 11 molecular complexes that are mostly located 100 pc north of the galaxy center .
the virial ratios for all of the clumps are less than unity , suggesting gravitationally bound clouds . a nonthermal radio source with an energy equivalent to 115 times cassiopeia a is located 110 pc north of the k@xmath40-band peak ( square sign in figure [ fig : extended ] ) ( neff @xmath84 ulvestad 2000 ) .
the distribution of co ( 32 ) appears to form an arc centered around this radio source , with an apparent absence of significant co ( 32 ) emission at the radio source .
similar molecular bubbles have been found in m 82 ( e.g. * ? ? ?
* ) and ngc 253 @xcite .
they attributed it to supernova blasts that swept away the molecular gas from the surrounding medium .
following @xcite , we estimated the kinetic energy needed to create the bubble using , @xmath85 , where @xmath86 ( @xmath87 is the hydrogen number density of the prebubble medium , and @xmath88 is the the atomic mass of hydrogen ) is the initial mass density of the interstellar matter , @xmath89 is the size of a bubble , and @xmath58 is the expansion velocity of molecular gas .
we derived @xmath89 from the distance between the nonthermal radio source and the co ( 32 ) peak in figure [ fig : extended ] , which is 110 pc , and used the offset from systemic velocity at the co ( 32 ) peak as @xmath58 of 30 km s@xmath7 .
the estimated energy is @xmath90 j , adopting @xmath91 @xmath92 , which is a typical density of interstellar matter in the galaxy and nearby galaxies .
this energy corresponds to the kinetic energy released by 10 100 supernova explosions and also to the released energy estimated from the radio continuum .
thus a high frequency of recent supernova explosions is a possible cause of the apparent absence in the co ( 32 ) distribution .
we present interferometric @xmath1co ( @xmath72 = 32 ) observations toward main bodies of the antennae galaxies ( ngc 4038/9 ) obtained using the sma .
the image shows gas concentration at both nuclei and the overlap region , and the emission is roughly distributed along the dust lanes seen in the b - band image .
the integrated intensity over main body of the antennae is 4.4 @xmath2 10@xmath28 jy km s@xmath7 and the overlap region contains almost half of the emission .
we detect using the clumpfind algorithm 57 molecular complexes , 33 of which located in the overlap region . adopting a galactic co - to - h@xmath52 conversion factor and the estimated co ( 32)/(10 ) line intensity ratios ,
the molecular complexes appear to have masses of 10@xmath56
10@xmath11 m@xmath5 .
while we find that 58 @xmath13 ( 33/57 ) of the spatially resolved gmas are located in the overlap region , only @xmath93 are spatially coincident with the star clusters detected in the optical and/or nir images , suggesting that the bulk of the co ( 32 ) emission traces the regions with very recent or near future star formation activity .
we find that the gmas in the antennae do not follow the well known radius - velocity dispersion relation obeyed by gmcs in milky way and other nearby galaxies .
the molecular complexes with large radii are distributed in ngc 4038 and the southern part of the overlap region , where the star formation rates are high .
therefore , the larger molecular complexes , which are possibly formed by clustering around local peaks of the gravitational potential wells , may be used as the signposts of current dusty star forming regions . in ngc 4039
, the molecular complexes with the larger velocity dispersion ( 2535 km s@xmath7 ) are located within 500 pc of the nucleus and those with the smaller velocity dispersion ( 1025 km s@xmath7 ) are located in the gas arm .
we estimated the co ( 32)/(10 ) brightness temperature ratio using the sma co ( 32 ) and ovro co
( 10 ) data , assuming a same beam filling factor for both transitions .
the mean ratios range between 0.3 and 0.6 , which suggests that , on average , the co line in the antennae is not thermalized up to @xmath72 = 32 .
however the line ratios in two nuclei and the southern of the overlap region are close to unity in some localized regions .
the high ratio in the nucleus of ngc 4038 and the overlap region is likely caused by star formation because the line ratio becomes high for clouds that are close to active star forming regions .
in contrast , the nucleus of ngc 4039 has a relatively inactive star forming region .
one possibility for causing the high line ratio in the nucleus of ngc 4039 is a presence of a hidden agn .
the other is gas collision between the inflowing gas and the counter - rotating central disk .
the atacama large millimeter / submillimeter array ( alma ) will begin scientific observations in the second half of 2011 .
the antennae is one of the best targets observed using the alma because it passes near zenith at the alma site and the beam pattern will be a nearly circular shape . using the superb sensitivity and angular resolutions offered by alma , we can observe other molecular lines including high-@xmath72 co in the antennae and study the properties of molecular gas quantitatively .
we thank the anonymous referee for useful comments .
we also thank sachiko onodera for providing the data of gmcs in m 33 and giving helpful comments , which improved our discussion .
fun barnes , j. e. , & hernquist , l. 1996 , , 471 , 115 bigiel , f. , bolatto , a. d. , leroy , a. k. , blitz , l. , walter , f. , rosolowsky , e. w. , lopez , l. a. , & plambeck , r. l. 2010 , , 725 , 1159 boone , f. , et al .
2011 , , 525 , a18 brandl , b. r. , et al .
2009 , , 699 , 1982 brogan , c. , johnson , k. , & darling , j. 2010 , , 716 , l51 bundy , k. , fukugita , m. , ellis , r. s. , targett , t. a. , belli , s. , & kodama , t. 2009 , , 697 , 1369 downes , d. , & solomon , p. m. 1998 , , 507 , 615 fabbiano , g. , zezas , a. , king , a. r. , ponman , t. j. , rots , a. , & schweizer , f. 2003 , , 584 , l5 fazio , g. g. , et al .
2004 , , 154 , 10 fukui , y. , et al .
2008 , , 178 , 56 gao , y. , lo , k. y. , lee , s .- w . , & lee , t .- h .
2001 , , 548 , 172 gilbert , a. m. , et al .
2000 , , 533 , l57 gil de paz , a. , et al .
2007 , , 173 , 185 greve , t. r. , papadopoulos , p. p. , gao , y. , & radford , s. j. e. 2009 , , 692 , 1432 harris , a. i. , baker , a. j. , zonak , s. g. , sharon , c. e. , genzel , r. , rauch , k. , watts , g. , & creager , r. 2010 , , 723 , 1139 ho , p.
t. p. , moran , j. m. , & lo , k. y. 2004 , , 616 , l1 iono , d. , yun , m. s. , & mihos , j. c. 2004 , , 616 , 199 iono , d. , et al .
2009 , , 695 , 1537 israel , f. p. 2009
, , 506 , 689 klaas , u. , nielbock , m. , haas , m. , krause , o. , & schreiber , j. 2010 , , 518 , l44 koester , b. , stoerzer , h. , stutzki , j. , & sternberg , a. 1994 , , 284 , 545 krips , m. , et al .
2011 , , 736 , 37 lada , c. j. , & lada , e. a. 2003 , , 41 , 57 larson , r. b. 1981 , , 194 , 809 mao , r .- q . , schulz , a. , henkel , c. , mauersberger , r. , muders , d. , & dinh - v - trung 2010 , , 724 , 1336 matsushita , s. , et al .
2004 , , 616 , l55 mengel , s. , lehnert , m. d. , thatte , n. , & genzel , r. 2002 , , 383 , 137 mengel , s. , lehnert , m. d. , thatte , n. , & genzel , r. 2005 , , 443 , 41 mihos , j. c. , bothun , g. d. , & richstone , d. o. 1993 , , 418 , 82 neff , s. g. , & ulvestad , j. s. 2000 , , 120 , 670 nikola , t. , genzel , r. , herrmann , f. , madden , s. c. , poglitsch , a. , geis , n. , townes , c. h. , & stacey , g. j. 1998 , , 504 , 749 oka , t. , nagai , m. , kamegai , k. , tanaka , k. , & kuboi , n. 2007 , , 59 , 15 onodera , s. 2009 , ph.d .
thesis , the university of tokyo rieke , g. h. , et al .
2004 , , 154 , 25 rosolowsky , e. , & blitz , l. 2005 , , 623 , 826 rosolowsky , e. , engargiola , g. , plambeck , r. , & blitz , l. 2003 , , 599 , 258 sakamoto , k. , et al .
2006 , , 636 , 685 sanders , d. b. , & mirabel , i. f. 1996 , , 34 , 749 schulz , a. , henkel , c. , muders , d. , mao , r. q. , rllig , m. , & mauersberger , r. 2007 , , 466 , 467 schweizer , f. , et al .
2008 , , 136 , 1482 solomon , p. m. , & barrett , j. w. 1991 , dynamics of galaxies and their molecular cloud distributions , 146 , 235 solomon , p. m. , rivolo , a. r. , barrett , j. , & yahil , a. 1987 , , 319 , 730 solomon , p. m. , & vanden bout , p. a. 2005 , , 43 , 677 toomre , a. 1977 , in the evolution of galaxies and stellar populations , ed .
tinsley & r.b .
larson ( new haven : yale univ . ) , 401 toomre , a. , & toomre , j. 1972 , , 178 , 623 van der tak , f. f. s. , black , j. h. , schier , f. l. , jansen , d. j. , & van dishoeck , e. f. 2007 , , 468 , 627 weedman , d. w. , et al .
2005 , , 633 , 706 whitmore , b. c. , et al .
2010 , , 140 , 75 whitmore , b. c. , zhang , q. , leitherer , c. , fall , s. m. , schweizer , f. , & miller , b. w. 1999 , , 118 , 1551 williams , j. p. , de geus , e. j. , & blitz , l. 1994 , , 428 , 693 wilson , c. d. , scoville , n. , madden , s. c. , & charmandaris , v. 2000 , , 542 , 120 wilson , c. d. , et al .
2009 , , 693 , 1736 zezas , a. , fabbiano , g. , baldi , a. , schweizer , f. , king , a. r. , ponman , t. j. , & rots , a. h. 2006 , , 166 , 211 zezas , a. , fabbiano , g. , rots , a. h. , & murray , s. s. 2002 , , 577 , 710 zezas , a. , fabbiano , g. , rots , a. h. , & murray , s. s. 2002 , , 142 , 239 zhang , h .- x . , gao , y. , & kong , x. 2010 , , 401 , 1839 zhu , m. , seaquist , e. r. , & kuno , n. 2003 , , 588 , 243 ) contour map overlaid on the hst 435 nm image ( left ) and the spitzer 8 @xmath27 m image ( right ) .
the contour levels are 6 jy km s@xmath7@xmath2 2 , 3 , 5 , 7 , 9 , 11 , 13 , 15 , 20 , 25 .
the galaxy and region names are labeled in the left figure .
the circles in the right figure show the primary beam centered on each pointing .
the ellipse in the lower right corner of both figures shows the synthesized beam size ( @xmath94 ) .
these integrated intensity maps are clipped at 2@xmath25 level using the aips task , momnt .
, width=396 ] ) contour map overlaid on the hst 435 nm image ( left ) and the spitzer 8 @xmath27 m image ( right ) .
the contour levels are 6 jy km s@xmath7@xmath2 2 , 3 , 5 , 7 , 9 , 11 , 13 , 15 , 20 , 25 .
the galaxy and region names are labeled in the left figure .
the circles in the right figure show the primary beam centered on each pointing .
the ellipse in the lower right corner of both figures shows the synthesized beam size ( @xmath94 ) .
these integrated intensity maps are clipped at 2@xmath25 level using the aips task , momnt .
, width=366 ] @xmath2 5 , 7 , 9 , 11 .
the small ellipse in the lower right corner indicates the higher angular resolution ( @xmath95 ) .
the grey contour map is the co ( 32 ) map made with all the data .
the contour levels are 6 jy km s@xmath7 @xmath2 5 , 7 , 9 , 11 , 13 , 15 , 20 , 25 .
the large ellipse in the lower right corner indicates the angular resolution ( @xmath96 ) .
the cross sign shows the galactic center defined by the vlt k@xmath40-band image and the square sign shows the nonthermal radio source @xcite . ,
width=377 ] -band image .
, width=302 ] -band image .
, width=302 ] -band image . , width=321 ] = 70 mjy beam@xmath7 ) @xmath2 3 , 4 , 5 , 6 , 7 , 8 , 9 , 10 .
the cross sign shows the galactic center and the box sign shows the radio source . , width=453 ] m contour map overlaid on the hst h@xmath97 image .
the contour levels are 40 , 50 , 60 , 70 , 80 , and 90 @xmath13 of the peak intensity .
the upper , middle , and lower circles show the co ( 10 ) , the co ( 21 ) , and the co ( 32 ) peaks , respectively .
the upper cross shows the location of the h@xmath52o maser observed by @xcite and the lower cross shows the location of the most massive cluster identified by @xcite . , width=302 ] -band peak of ngc 4039 shown the green circle in the left figure .
the contour levels are 6 jy km a@xmath7 @xmath2 2 , 3 , 5 , 7 , 9 . , width=302 ] -band peak of ngc 4039 shown the green circle in the left figure .
the contour levels are 6 jy km a@xmath7 @xmath2 2 , 3 , 5 , 7 , 9 .
, width=302 ] [ htbp ] .the phase centers [ cols="^,^,^",options="header " , ] addtoresettablesection addtoresetfiguresection
we compare the co ( 32 ) map with multiwavelength images from radio and to x - ray .
figure [ fig : stamp ] shows the co ( 32 ) contour maps of five regions ( ngc 4038 , ngc 4039 , complex 1 , complex 2 , and complex 3 ) overlaid on 14 different images .
the co ( 10 ) map was taken with the ovro interferometer by @xcite .
the strongest emission in all regions is located in ngc 4038 and the emission in complex 2 is strongest in the overlap region . @xcite
conducted fir observations at 70 , 100 and 160 @xmath27 m using the _ herschel_-pacs .
the brightest emission in three bands comes from complex 3 .
complex 1 and complex 2 together become the brightest area in the 160 @xmath27 m map .
the 24 @xmath27 m and four near - infrared ( nir ) at 3.6 , 4.5 , 5.8 , and 8.0 @xmath27 m images were obtained with the mips @xcite and the irac @xcite on board the _ spitzer space telescope _ , respectively .
we use the basic calibrated data obtained using the leopard software .
the strongest emission in all bands , especially 24 @xmath27 m , is located in complex 3 . the h@xmath97 image ( f656n ) , and the b - band image ( f435w )
have been observed using the acs on board the hst @xcite .
we use the archival data obtained from the mast ( multimission archive at stsci ) .
there is no optically bright emission in complex 2 and complex 3 .
both the far - ultraviolet ( fuv ; @xmath9 1516 @xmath98 ) and the near - ultraviolet ( nuv ; @xmath9 2267 @xmath98 ) images were taken with the _ galaxy evolution explorer _ ( galex ) ultraviolet atlas of nearby galaxies distributed by @xcite .
the strong uv emission is detected west of the peak in complex 3 , where the bright optical , nir and mir emission is also observed .
we obtained the archival x - ray image which @xcite observed with _
the bright x - ray emission is associated with the nucleus of ngc 4039 .
cccccccc [ tb : clump ] + + i d & ra ( j2000 ) & dec ( j2000 ) & @xmath46 & @xmath47 & @xmath48 & @xmath99 & @xmath100 + & [ h m s ] & [ @xmath64 @xmath101 @xmath16 ] & [ km s@xmath7 ] & [ pc ] & [ km s@xmath7 ] & [ @xmath104 m@xmath77 ] & [ @xmath104 m@xmath77 ] + + i d & ra ( j2000 ) & dec ( j2000 ) & @xmath46 & @xmath47 & @xmath48 & @xmath99 & @xmath100 + & [ h m s ] & [ @xmath64 @xmath101 @xmath16 ] & [ km s@xmath7 ] & [ pc ] & [ km s@xmath7 ] & [ @xmath104 m@xmath77 ] & [ @xmath104 m@xmath77 ] + 1 & 12 01 55.36 & -18 52 49 & 1414 & 130 & 21 & 16 & 2.6 + 2 & 12 01 55.53 & -18 52 47 & 1514 & 242 & 22 & 30 & 5.3 + 3 & 12 01 55.37 & -18 52 48 & 1374 & 157 & 22 & 15 & 3.4 + 4 & 12 01 55.42 & -18 52 49 & 1504 & 126 & 16 & 13 & 1.5 + 5 & 12 01 55.53 & -18 52 45 & 1364 & 162 & 14 & 8.2 & 1.4 + 6 & 12 01 55.50 & -18 52 45 & 1394 & 146 & 35 & 8.2 & 8.1 + 7 & 12 01 55.30 & -18 52 51 & 1464 & 180 & 25 & 4.9 & 5.1 + 8 & 12 01 55.38 & -18 52 48 & 1464 & 130 & 27 & 4.4 & 4.4 + 9 & 12 01 55.48 & -18 52 52 & 1474 & 110 & 16 & 1.5 & 1.4 + 10 & 12 01 55.38 & -18 52 51 & 1364 & 151 & 18 & 2.4 & 2.3 + 11 & 12 01 54.96 & -18 52 53 & 1464 & 181 & 21 & 30 & 3.5 + 12 & 12 01 54.74 & -18 52 52 & 1444 & 205 & 24 & 24 & 5.4 + 13 & 12 01 54.78 & -18 52 55 & 1604 & 185 & 29 & 12 & 7.0 + 14 & 12 01 54.78 & -18 52 53 & 1524 & 116 & 28 & 5.0 & 4.0 + 15 & 12 01 54.77 & -18 52 53 & 1544 & 190 & 34 & 6.6 & 9.8 + 16 & 12 01 54.77 & -18 52 57 & 1644 & 92 & 9 & 2.6 & 0.34 + 17 & 12 01 55.01 & -18 52 53 & 1564 & 139 & 33 & 3.6 & 7.0 + 18 & 12 01 55.12 & -18 52 54 & 1504 & 153 & 22 & 7.7 & 3.2 + 19 & 12 01 55.00 & -18 53 06 & 1494 & 239 & 31 & 30 & 10 + 20 & 12 01 54.92 & -18 53 03 & 1554 & 208 & 28 & 15 & 7.2 + 21 & 12 01 55.09 & -18 53 02 & 1514 & 129 & 12 & 4.0 & 0.87 + 22 & 12 01 54.68 & -18 53 03 & 1454 & 263 & 17 & 6.3 & 3.4 + 23 & 12 01 54.89 & -18 53 00 & 1584 & 143 & 11 & 3.5 & 0.81 + 24 & 12 01 54.92 &
-18 53 00 & 1514 & 137 & 20 & 5.1 & 2.5 + 25 & 12 01 55.02 & -18 53 01 & 1504 & 105 & 22 & 4.4 & 2.3 + 26 & 12 01 54.96 & -18 53 08 & 1614 & 243 & 13 & 6.0 & 1.9 + 27 & 12 01 54.96 & -18 53 08 & 1584 & 203 & 36 & 4.9 & 12 + 28 & 12 01 54.66 & -18 53 05 & 1474 & 226 & 31 & 7.3 & 9.9 + 29 & 12 01 55.18 & -18 52 40 & 1574 & 348 & 15 & 19 & 3.4 + 30 & 12 01 54.66 & -18 52 45 & 1584 & 251 & 21 & 7.8 & 4.9 + 31 & 12 01 54.44 & -18 52 43 & 1614 & 147 & 9 & 1.3 & 0.57 + 32 & 12 01 54.26 & -18 53 08 & 1414 & 132 & 16 & 4.6 & 1.5 + 33 & 12 01 54.33 & -18 53 08 & 1454 & 203 & 15 & 7.1 & 2.1 + 34 & 12 01 53.16 & -18 52 02 & 1624 & 256 & 29 & 61 & 10 + 35 & 12 01 53.22 & -18 52 04 & 1584 & 199 & 24 & 34 & 5.0 + 36 & 12 01 53.17 & -18 52 06 & 1624 & 273 & 26 & 31 & 8.5 + 37 & 12 01 53.49 & -18 52 06 & 1634 & 218 & 25 & 4.6 & 6.0 + 38 & 12 01 52.61 & -18 51 59 & 1714 & 93 & 10 & 1.6 & 0.42 + 39 & 12 01 53.26 & -18 52 04 & 1674 & 225 & 22 & 7.6 & 4.8 + 40 & 12 01 52.97 & -18 51 60 & 1634 & 206 & 27 & 4.2 & 6.8 + 41 & 12 01 53.18 & -18 52 06 & 1684 & 217 & 12 & 3.0 & 1.4 + 42 & 12 01 52.85 & -18 52 06 & 1634 & 180 & 13 & 3.1 & 1.5 + 43 & 12 01 52.76 & -18 52 07 & 1634 & 89 & 6 & 1.3 & 0.13 + 44 & 12 01 53.65 & -18 53 09 & 1604 & 190 & 31 & 21 & 8.1 + 45 & 12 01 53.32 & -18 53 09 & 1604 & 107 & 31 & 12 & 4.6 + 46 & 12 01 53.33 & -18 53 13 & 1624 & 122 & 18 & 5.7 & 1.8 + 47 & 12 01 53.65 & -18 53 11 & 1684 & 130 & 28 & 7.4 & 4.5 + 48 & 12 01 53.21 & -18 53 11 & 1644 & 120 & 28 & 8.1 & 4.4 + 49 & 12 01 53.45 & -18 53 09 & 1584 & 178 & 33 & 10 & 8.6 + 50 & 12 01 53.16 & -18 53 12 & 1704 & 85 & 16 & 2.6 & 1.0 + 51 & 12 01 53.17 & -18 53 12 & 1684 & 113 & 24 & 5.2 & 2.8 + 52 & 12 01 53.25 & -18 53 18 & 1724 & 142 & 20 & 2.2 & 2.6 + 53 & 12 01 53.36 & -18 53 17 & 1704 & 120 & 10 & 2.8 & 0.55 + 54 & 12 01 53.81 & -18 53 13 & 1644 & 156 & 19 & 3.3 & 2.6 + 55 & 12 01 53.34 & -18 53 11 & 1604 & 119 & 17 & 3.1 & 1.5 + 56 & 12 01 53.06 & -18 53 15 & 1704 & 111 & 12 & 1.9 & 0.77 + 57 & 12 01 53.28 & -18 53 16 & 1684 & 115 & 18 & 2.1 & 1.6 + | we present a @xmath0 ( 100 pc ) resolution @xmath1co ( 32 ) map of the nearby intermediate stage interacting galaxy pair ngc 4038/9 ( the antennae galaxies ) obtained with the submillimeter array .
we find that half the co ( 32 ) emission originates in the overlap region where most of the tidally induced star formation had been previously found in shorter wavelength images , with the rest being centered on each of the nuclei .
the gross distribution is consistent with lower resolution single dish images , but we show for the first time the detailed distribution of the warm and dense molecular gas across this galaxy pair at resolutions comparable to the size of a typical giant molecular complex . while we find that 58% ( 33/57 ) of the spatially resolved giant molecular associations ( gmas ; a few @xmath2 100 pc ) are located in the overlap region , only @xmath3 spatially coincides with the optically detected star clusters , suggesting that the bulk of the co ( 32 ) emission traces the regions with very recent or near future star formation activity .
the spatial distribution of the co ( 32)/co ( 10 ) integrated brightness temperature ratios mainly range between 0.3 and 0.8 , which suggests that on average the co ( 32 ) line in the antennae is not completely thermalized and similar to the average values of nearby spirals . a higher ratio is seen in both nuclei and the southern complexes in the overlap region .
higher radiation field associated with intense star formation can account for the nucleus of ngc 4038 and the overlap region , but the nuclear region of ngc 4039 show relatively little star formation or agn activities and can not be easily explained .
we show kinematical evidence that the high line ratio in ngc 4039 is possibly caused by gas inflow into the counter - rotating central disk . |
the asteroseismology of red giant stars has become a highlight of the corot and _ kepler _ space missions ( for general reviews see e.g. * ? ? ?
* ; * ? ? ?
one feature that has made these stars interesting , is the presence of non - radial oscillation modes that reveal properties of the stellar cores .
the non - radial modes have a mixed nature , behaving like acoustic ( or p ) modes in the envelope with pressure acting as the restoring force , and as gravity ( or g ) modes in the core region with buoyancy being the restoring force ( e.g. * ? ? ?
the p- and g - mode cavities are separated by an evanescent zone , which the waves can tunnel through from either side .
the exact p- and g-`mixture ' , or flavor , of a mixed mode depends on its frequency and spherical degree , @xmath0
. modes with frequencies close to the acoustic resonant frequencies of the stellar envelope tend to be more p - mode like , while those far from the acoustic resonances are much more g - mode like .
the latter therefore probe deeper into the stellar interior compared to the former .
how much the flavor changes from mode to mode across the acoustic resonances depends on the overall coupling between the envelope and the core .
the overall aspects of mode mixing in red giants arising from this coupling is well understood theoretically @xcite .
observationally , the dipole modes ( @xmath1 ) have turned out to be particularly useful probes of the core because of their stronger coupling between core and envelope .
the characterization of dipole mixed modes @xcite led to the discovery that red giant branch stars can be clearly distinguished from red clump stars @xcite , and to the detection of radial differential rotation @xcite .
modes of higher spherical degree are also mixed , but the weaker coupling makes it difficult to observe the modes with strong g - mode flavor .
the quadrupole modes ( @xmath2 ) we observe , and to a larger degree the octupole modes ( @xmath3 ) , are therefore on average more acoustic compared to the dipole modes , and hence less sensitive to the stellar core . one particular observational result about the dipole modes posed an intriguing puzzle .
the ensemble study of a few hundred _ kepler _ red giants by @xcite showed that a few dozen stars
about 20% of their sample had significantly lower power in the dipole modes than ` normal ' stars .
however , no significant suppression of higher degree modes was reported , leading to the conclusion only dipole modes were affected .
recent theoretical work has proposed that the mechanism responsible for the mode suppression results in almost total trapping of the mode energy that leaks into the g - mode cavity @xcite .
they put forward magnetic fields in the core region of the stars as the most plausible candidate for the suppression mechanism .
this interpretation was further supported by the observation that mode suppression only occurs in stars above 1.1m@xmath4 , with an increasing fraction up to 50 - 60% for slightly more massive stars , all of which hosted convective cores during their main sequence phase ; strongly pointing to a convective core dynamo as the source of the mode suppressing magnetic field @xcite .
both @xcite and @xcite focused their analysis on dipole modes .
however , the theory by @xcite does allow one to predict the magnitude of the suppression for higher degree modes .
agreement with observations of these modes would provide important support for the proposed mechanism . in this paper , we use 3.5 years of _ kepler _ data of over 3,600 carefully selected red giant branch stars to investigate the mode suppression in the non - radial modes of degree @xmath1 , 2 , and 3 , and compare theoretical predictions with our observed mode visibilities .
we use the modules for experiments in stellar evolution ( mesa , release # 7456 , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) to compute stellar evolution models of low - mass stars from the zero age main sequence to the tip of the red giant branch .
non - rotating models have been computed using an initial metallicity of @xmath5 with a mixture taken from @xcite and adopting the opal opacity tables @xcite .
we calculate convective regions using the mixing - length theory with @xmath6 .
the boundaries of convective regions are determined according to the schwarzschild criterion .
we calculate the expected visibilities for dipole , quadrupole , and octupole modes in models with masses @xmath7m@xmath4 , @xmath8m@xmath4 , @xmath9m@xmath4 , @xmath10m@xmath4 , and @xmath11m@xmath4 following @xcite .
according to the theory , the ratio of suppressed mode power to normal mode power is @xmath12^{-1 } \ , , \ ] ] where @xmath13 is the large frequency separation , @xmath14 is the radial mode lifetime measurable from the observed frequency power spectrum ( e.g. * ? ? ? * ) ) , and @xmath15 is the wave transmission coefficient through the evanescent zone .
@xmath15 is calculated via @xmath16 \ , .\ ] ] here , @xmath17 and @xmath18 are the lower and upper boundaries of the evanescent zone , @xmath19 is the lamb frequency squared , @xmath20 is the buoyancy frequency , @xmath21 is the angular wave frequency , and @xmath22 is the sound speed .
we calculate @xmath13 and the frequency of maximum power , @xmath23 , using the scaling relations of @xcite and @xcite with the solar references values , @xmath13@xmath24@xmath25hz and @xmath23@xmath26@xmath25hz .
our initial selection of stars was based on the analysis by @xcite of 13,412 red giants from which we adopt their measurements of @xmath13 and @xmath23 , as well as their estimates of stellar mass based on the combination of those seismic observables with photometrically - derived effective temperatures using scaling relations ( see * ? ? ?
* for details ) .
in order to select only red giant branch stars from this sample we follow the approach by @xcite .
we selected all 3,993 stars with @xmath23 @xmath27@xmath25hz and @xmath28m@xmath4 , which based on stellar models is expected to select only red giant branch stars ( e.g. * ? ? ? * ) .
however , this selection do not take measurement uncertainties in @xmath23 and @xmath29 into account , possibly introducing some helium - core burning stars into our sample towards the lower end of the @xmath23 bracket .
we therefore performed a few additional steps to further reduce the possible ` contamination ' by helium - burning stars in our sample .
first , we derived the offset from zero of the harmonic series of radial modes , known as @xmath30 in the asymptotic relation , @xmath31 , where @xmath32 is the mode frequency and @xmath33 is the radial order .
this offset is known to be different for helium - burning stars @xcite relative to red giant branch stars of similar density , and hence has diagnostic power to distinguish the two types of stars . to measure @xmath30 we calculated frequency power spectra for each star using long - cadence ( @xmath34minutes ) _ kepler _ data obtained between 2009 may 2 and 2012 oct 3 , corresponding to observing quarters 014 , or a total of about 54,000 data points per star .
we used the method by @xcite to fit and remove the background noise profile of the spectra , and then selected the central @xmath35@xmath13spectral range around @xmath23 , which we folded with an interval of @xmath13 .
the folded spectrum was smoothed by a gaussian function with a full - width - half - maximum of 0.1@xmath13 , and correlated with a model of the spectrum ( figure [ foldedspectrum ] ) .
-wide region around @xmath23 of a normal star ( kic2141255 ) .
regions dominated by modes of degrees , @xmath36 , 1 , 2 , and 3 are indicated .
the red curve shows the model shifted to the position resulting in the largest correlation with the data .
( b ) correlation versus shift between model and data .
( c ) and ( d ) the same as ( a ) and ( b ) , but for a star with suppressed non - radial modes ( kic4348666 ) .
[ foldedspectrum],title="fig:",width=321 ] -wide region around @xmath23 of a normal star ( kic2141255 ) .
regions dominated by modes of degrees , @xmath36 , 1 , 2 , and 3 are indicated .
the red curve shows the model shifted to the position resulting in the largest correlation with the data .
( b ) correlation versus shift between model and data .
( c ) and ( d ) the same as ( a ) and ( b ) , but for a star with suppressed non - radial modes ( kic4348666 ) .
[ foldedspectrum],title="fig:",width=321 ] each spherical degree in the model was described by a lorentzian profile with relative heights of 1.0 , 0.5 , and 0.8 for the @xmath36 , @xmath1 , and @xmath2 modes , and widths of 5% , 10% , and 5% of @xmath13 , respectively .
the lorentzian profiles were centered relative to each other with the one representing quadrupole modes being 0.12@xmath13 to the left of the radial mode profile , and the one representing the dipole modes being 0.52@xmath13 to the right of the radial mode following the results by @xcite ( see figure [ foldedspectrum ] ) .
the shift between data and model providing the largest correlation was adopted as @xmath37 . only the 3,721 stars with @xmath30 within @xmath38 of the @xmath13-@xmath30 relation by @xcite ( @xmath39 ) were kept in our sample ( see figure [ epsilon ] ) . versus @xmath30 of the 3,993 red giants from the @xcite sample with @xmath23 @xmath27@xmath25hz and @xmath40m@xmath4 .
the solid red curve shows the relation by @xcite , and the dotted red curves show the relation shifted by @xmath41 .
[ epsilon],width=332 ] second , we cross matched our remaining sample with the stars identified as burning helium by @xcite and @xcite based on their dipole - mode period spacings .
we found 86 helium - burning stars this way , which we removed .
we note that all but eight of these helium - burning contaminants had @xmath23@xmath42@xmath25hz ( and all @xmath43@xmath25hz ) .
although our remaining sample could potentially still include some helium - burning stars if they were not in the @xcite and @xcite samples , the above cross match indicates such contaminants would most likely have @xmath23 @xmath44@xmath25hz . as a final check , we visually inspected the power spectra of the remaining sample ( 3,635 stars ) , which led to the removal of 24 stars that appeared to be helium - core burning or had spectra of such poor quality ( low signal - to - noise or bad window function ) that the stellar classification was ambiguous .
to measure the mode visibilities , we first needed to identify the regions in the power spectra dominated by the different modes . here , we used the @xmath30 values found in the previous step ( sect .
[ sampleselection ] ) to locate the four central radial modes closest to @xmath23 , and followed the approach by @xcite for masking the parts of the spectra dominated by @xmath36 , @xmath1 , @xmath2 , and @xmath3 modes .
the regions we chose were @xmath45 for @xmath36 , @xmath46 for @xmath1 , @xmath47 for @xmath2 , and @xmath48 for @xmath3 . in this way ,
the entire 4@xmath13-wide central part of the spectrum was divided up into distinct segments associated with either @xmath36 , @xmath1 , @xmath2 , or @xmath3 modes . in figure
[ examplespectra ] we show a couple of representative spectra with the results of the masking , identifying the different mode degrees .
see also figure 1 in @xcite who used the exact same scheme . , which partly merges the power from individual mixed modes into one broad peak around each acoustic resonance .
the colored regions indicate the integrated power associated with each mode of degree @xmath36 , 1 , 2 , and 3 .
the horizontal dashed curves shows the estimated noise profile ( a ) star with normal mode visibility , @xmath49 , of the non - radial modes .
( b ) star with low mode visibilities in both dipole and quadrupole modes , but normal octupole mode visibility .
[ examplespectra],width=321 ] finally , the mode visibilities , @xmath49 , of the non - radial modes were derived as the ratio of the total mode power of the segments associated with a given degree relative to the total radial mode power . for this
we used the background corrected spectra described in sect .
[ sampleselection ] .
we note that the uncertainty in the background estimation introduces measurement scatter on top of the intrinsic spread in the visibilities , potentially pushing intrinsically low values of @xmath49 below zero .
rejecting those stars or artificially pushing them up to @xmath50 would be statistically incorrect and bias our measurements of the average @xmath49 .
in figure [ vis_vs_numax ] we show our measurements of the dipole mode visibility as function of @xmath23 .
we denote the stars below the dotted line as the dipole - suppressed sample , as opposed to the ` normal ' stars .
as shown in @xcite and @xcite , the predicted visibilities of the suppressed dipole modes ( solid lines ) match the observations remarkably well , and with normal stars on average being less massive . here
, we show that the prediction for the dipole - mode suppression is insensitive to stellar mass for the typical mass range of the _ kepler _ red giants ( the solid black curves fall almost on top of each other ) , which is also what we observe in the data . the slight dependence on mode lifetime , @xmath14 ,
is illustrated by comparing the black curves ( all @xmath51days ) with the gray curves based on a @xmath10m@xmath4 track just above ( @xmath52days ) and below ( @xmath53days ) the black curves .
we find no variation in the average visibility ( @xmath54 ) and its scatter for the normal stars as a function on @xmath23 . ) for 3,611 red giants below the red giant branch luminosity bump .
stars evolve from right to left , and their approximate @xmath55 values are shown on the top axis .
the rms scatter in @xmath56 is 0.26 ( or 20% ) for the normal stars .
the color of each symbol indicates the stellar mass according to the scale on the right - hand side .
the solid black lines are the predicted visibilities from 1.1m@xmath4 , 1.3m@xmath4 , 1.5m@xmath4 , 1.7m@xmath4 , and 1.9m@xmath4 models adopting an average mode lifetime of 20 days in agreement with @xcite ( sect , [ theory ] ) .
the solid gray curves are the predicted visibilities from 1.7m@xmath4 models for mode lifetimes of 10 days ( highest @xmath56 ) and 40 days ( lowest @xmath56 ) .
the dotted fiducial line , adopted from @xcite , separates dipole - suppressed and normal stars . the observed dipole visibilities are identical to those by @xcite .
[ vis_vs_numax],width=321 ] in an attempt to determine if the dipole - suppressed sample is somehow distinct other than by their mass @xcite , we do not find them to be statistically different in terms of their distance and position on sky . if we focus on the mass range @xmath57/m@xmath4 @xmath58 , where we find the highest number and fraction ( 5060% ) of dipole - suppressed stars @xcite , the dipole - suppressed sample is on average slightly more metal poor ( by @xmath59 dex ) than the normal stars , and slightly hotter ( by @xmath60k ) ( based on sdss - apogee dr12 , * ? ? ?
the quadrupole and octupole visibilities are shown in figures [ vis2_vs_numax ] ( a ) and ( b ) , respectively . here
, we only show the dipole - suppressed sample as filled symbols , to make it easier to distinguish them from the normal population shown as empty symbols . while less dramatic than the dipole modes , the quadrupole modes show significant reduction in mode power for the dipole - suppressed sample .
the least evolved stars ( @xmath61@xmath23/@xmath25hz @xmath62 ) show on average @xmath63 , compared to @xmath64 for all the normal stars , corresponding to 49% reduction in power .
although there is overall good agreement between the predictions and the observations , the trend of the visibility as a function of @xmath23 is predicted to be steeper than observed if we , as done here , assume a fixed mode lifetime along the red giant branch ( figure [ vis2_vs_numax](a ) ) .
we used a lifetime of @xmath65days , which is representative for the average value found by @xcite .
this assumption could , at least partly , explain the discrepancy , because the observed mode lifetimes are most likely increasing as red giants evolve towards lower @xmath23 and lower @xmath66 .
the variation in mode lifetime was found to be roughly 50% by ( * ? ? ?
* their figure 7 ) across a sample spanning a @xmath23 range of @xmath67@xmath25hz ( their table a.2 ) .
our sample span a much larger @xmath23range , and for reference we therefore also show the @xmath10m@xmath4 tracks in gray with factors of two difference in mode lifetime ; 10 and 40 days , respectively . based on this , we conclude that an increasing mode lifetime along the red giant branch could possibly explain the apparent discrepancy of the predicted versus observed trend seen in figure [ vis2_vs_numax](a ) .
a mode lifetime of @xmath68days at @xmath23 @xmath69@xmath25hz and a lifetime of @xmath70days at @xmath23 @xmath71@xmath25hz would provide a better match to the data .
however , we note that the predicted visibilities of the suppressed sample depends somewhat on the overshoot beyond classical schwarzschild implemented in the stellar models ( cantiello et al .
2016 , in prep . ) , so we defer a more thorough analysis to future work , but point out here the possibility of constraining either the mode lifetime or the overshoot given independent measurements of the other . .
the notation follows that of figure [ vis_vs_numax ] except only the dipole - suppressed stars ( those below the dotted line in figure [ vis_vs_numax ] ) are shown by filled symbols . the observed rms scatter in @xmath72 is 0.14 ( or 20% ) for the normal stars .
( b ) the same as panel ( a ) but for octupole modes . here , the observed @xmath73 scatter is 0.05 ( 84% ) .
[ vis2_vs_numax],title="fig:",width=321 ] .
the notation follows that of figure [ vis_vs_numax ] except only the dipole - suppressed stars ( those below the dotted line in figure [ vis_vs_numax ] ) are shown by filled symbols .
the observed rms scatter in @xmath72 is 0.14 ( or 20% ) for the normal stars .
( b ) the same as panel ( a ) but for octupole modes . here , the observed @xmath73 scatter is 0.05 ( 84% ) . [ vis2_vs_numax],title="fig:",width=321 ] turning to the octupole modes ( figure [ vis2_vs_numax](b ) ) , we see no significant effect of the suppression , as anticipated by theory ( solid black curves ) because they are almost purely acoustic envelope modes .
we find the average visibilities to be @xmath74 for the normal stars , and
@xmath75 for the dipole - suppressed sample , which are the same within @xmath76 . in figure [ vis2_vs_numax](b )
we do not show the tracks of different mode lifetimes because they are essentially indistinguishable from the black curves .
both the evolution of the predicted mode visibility as a function of @xmath23 ( for fixed spherical degree ) and its dependence on mode degree ( for fixed @xmath23 ) are essentially determined by the difference in coupling between the core and envelope as the star evolves ( or as seen by modes of different degree ) .
the key aspect of this predicted behavior is that there is perfect trapping of the mode energy leaking into the core by the suppression mechanism .
hence , given the agreement reported here between theory and observations for both @xmath2 and @xmath3 modes , in addition to what is seen for dipole modes , supports this notion that the mode suppression mechanism indeed traps almost , if not , all the energy that leaks into the core region .
the theoretical predictions presented in figures [ vis_vs_numax ] and [ vis2_vs_numax ] are derived relative to non - suppressed stars , and are hence scaled to the observed average visibilities of the normal stars .
we note that the absolute scale of the observed visibilities reported here are not directly comparable with expectations for the normal stars @xcite because we simply integrated the power over fixed regions of the spectrum in terms of @xmath13 . due to the presence of mixed modes , which strictly speaking
exist across the entire spectrum , small amounts of power ( from gravity - dominated modes far from the acoustic resonant frequencies ) from one degree of modes will be present in the regions that we associate with modes of another degree .
direct comparisons would therefore need to be based on simultaneous fitting of model profiles to all the modes in the spectrum , which is currently not practically possible for large numbers of stars @xcite .
similarly , any differences between the average visibilities reported here and those by @xcite , are most likely due to slight different choices for the mode integration regions in the two studies .
we have measured oscillation mode power in over 3,600 red giant stars , and compared our results to theoretical predictions based on the `` magnetic greenhouse '' mechanism proposed by @xcite for dipole mode suppression .
our results from modes of spherical degrees @xmath77 , provide strong support for one of the main assertions of the theory behind the mode suppression all mode energy leaking into the g - mode cavity of a star is efficiently trapped or dissipated . specifically , we confirm the up to almost 100% suppression for dipole modes previously shown and measure for the first time up to 49% suppression on average of the quadrupole modes and no suppression of the octupole modes .
we find it likely that a variation in the mode lifetime along the red giant branch could explain the small difference in slope between our prediction and the observations in the @xmath72 versus @xmath23diagram .
however , the variation required needs to be confirmed by measurements of the mode lifetime of radial modes for a robust sample of stars with @xmath23 around 220@xmath25hz and 70@xmath25hz .
further tests of the magnetic greenhouse effect could be implemented with determinations of the lifetimes of the suppressed dipole modes , or the measurement of suppressed dipole mode visibilities in helium - burning clump stars ( cantiello et al .
2016 in prep . )
we acknowledge the entire _ kepler _ team , whose efforts made these results possible .
d.s . is the recipient of an australian research council future fellowship ( project number ft140100147 ) .
j.f . acknowledges support from nsf under grant no .
ast-1205732 and through a lee dubridge fellowship at caltech . r.a.g .
acknowledge the support of the european community s seventh framework programme ( fp7/2007 - 2013 ) under grant agreement no . 269194 ( irses / ask ) , the anr- idee ( n anr-12-bs05 - 0008 ) , and from the cnes .
d.h . acknowledges support by the australian research council s discovery projects funding scheme ( project number de140101364 ) and support by the national aeronautics and space administration under grant nnx14ab92 g issued through the kepler participating scientist program .
this project was supported by nasa under tcan grant number nnx14ab53 g , and the nsf under grants phy 11 - 25915 and ast 11 - 09174 .
funding for the stellar astrophysics centre is provided by the danish national research foundation ( grant agreement no . : dnrf106 ) .
the research is supported by the asterisk project ( asteroseismic investigations with song and kepler ) funded by the european research council ( grant agreement no . : 267864 ) . , m. , grevesse , n. , & sauval , a. j. 2005 , in astronomical society of the pacific conference series , vol .
336 , cosmic abundances as records of stellar evolution and nucleosynthesis , ed .
t. g. barnes , iii & f. n. bash , 25 , p. g. , bedding , t. r. , mosser , b. , stello , d. , garcia , r. a. , kallinger , t. , hekker , s. , elsworth , y. , frandsen , s. , carrier , f. , de ridder , j. , aerts , c. , white , t. r. , huber , d. , dupret , m. , montalbn , j. , miglio , a. , noels , a. , chaplin , w. j. , kjeldsen , h. , christensen - dalsgaard , j. , gilliland , r. l. , brown , t. m. , kawaler , s. d. , mathur , s. , & jenkins , j. m. 2011 , science , 332 , 205 , p. g. , montalban , j. , kallinger , t. , de ridder , j. , aerts , c. , garca , r. a. , hekker , s. , dupret , m .- a . , mosser , b. , eggenberger , p. , stello , d. , elsworth , y. , frandsen , s. , carrier , f. , hillen , m. , gruberbauer , m. , christensen - dalsgaard , j. , miglio , a. , valentini , m. , bedding , t. r. , kjeldsen , h. , girouard , f. r. , hall , j. r. , & ibrahim , k. a. 2012 , , 481 , 55 , t. r. , mosser , b. , huber , d. , montalbn , j. , beck , p. , christensen - dalsgaard , j. , elsworth , y. p. , garca , r. a. , miglio , a. , stello , d. , white , t. r. , de ridder , j. , hekker , s. , aerts , c. , barban , c. , belkacem , k. , broomhall , a. , brown , t. m. , buzasi , d. l. , carrier , f. , chaplin , w. j. , di mauro , m. p. , dupret , m. , frandsen , s. , gilliland , r. l. , goupil , m. , jenkins , j. m. , kallinger , t. , kawaler , s. , kjeldsen , h. , mathur , s. , noels , a. , aguirre , v. s. , & ventura , p. 2011
, , 471 , 608 , e. , stello , d. , huber , d. , bedding , t. r. , bonanno , a. , brogaard , k. , kallinger , t. , benomar , o. , white , t. r. , mosser , b. , basu , s. , chaplin , w. j. , christensen - dalsgaard , j. , elsworth , y. p. , garca , r. a. , hekker , s. , kjeldsen , h. , mathur , s. , meibom , s. , hall , j. r. , ibrahim , k. a. , & klaus , t. c. 2012 , , 757 , 190 , d. , bedding , t. r. , stello , d. , mosser , b. , mathur , s. , kallinger , t. , hekker , s. , elsworth , y. p. , buzasi , d. l. , de ridder , j. , gilliland , r. l. , kjeldsen , h. , chaplin , w. j. , garca , r. a. , hale , s. j. , preston , h. l. , white , t. r. , borucki , w. j. , christensen - dalsgaard , j. , clarke , b. d. , jenkins , j. m. , & koch , d. 2010 , , 723 , 1607 , t. , hekker , s. , mosser , b. , de ridder , j. , bedding , t. r. , elsworth , y. p. , gruberbauer , m. , guenther , d. b. , stello , d. , basu , s. , garca , r. a. , chaplin , w. j. , mullally , f. , still , m. , & thompson , s. e. 2012 , , 541 , a51 , b. , benomar , o. , belkacem , k. , goupil , m. j. , lagarde , n. , michel , e. , lebreton , y. , stello , d. , vrard , m. , barban , c. , bedding , t. r. , deheuvels , s. , chaplin , w. j. , de ridder , j. , elsworth , y. , montalban , j. , noels , a. , ouazzani , r. m. , samadi , r. , white , t. r. , & kjeldsen , h. 2014 , , 572 , l5 , b. , elsworth , y. , hekker , s. , huber , d. , kallinger , t. , mathur , s. , belkacem , k. , goupil , m. j. , samadi , r. , barban , c. , bedding , t. r. , chaplin , w. j. , garca , r. a. , stello , d. , de ridder , j. , middour , c. k. , morris , r. l. , & quintana , e. v. 2012 , , 537 , a30 | the asteroseismology of red giant stars has continued to yield surprises since the onset of high - precision photometry from space - based observations .
an exciting new theoretical result shows that the previously observed suppression of dipole oscillation modes in red giants can be used to detect strong magnetic fields in the stellar cores .
a fundamental facet of the theory is that nearly all the mode energy leaking into the core is trapped by the magnetic greenhouse effect .
this results in clear predictions for how the mode visibility changes as a star evolves up the red giant branch , and how that depends on stellar mass , spherical degree , and mode lifetime . here
, we investigate the validity of these predictions with a focus on the visibility of different spherical degrees .
we find that mode suppression weakens for higher degree modes with an average reduction in the quadrupole mode visibility of up to 49% for the least evolved stars in our sample , and no detectable suppression of octupole modes , in agreement with the theoretical predictions .
we furthermore find evidence for the influence of increasing mode lifetimes on the measured visibilities along the red giant branch , in agreement with previous independent observations .
these results support the theory that strong internal magnetic fields are responsible for the observed suppression of non - radial modes in red giants .
we also find preliminary evidence that stars with suppressed dipole modes on average have slightly lower metallicity than normal stars .
stars : fundamental parameters stars : oscillations stars : interiors stars : magnetic field |
the series of states of accretion disks called the radiatively inefficient accretion flows ( riaf ) forms an optically - thin , under - luminous ( usually radiating at a small fraction of the eddigton luminosity of each central object ) branch in the accretion - rate vs. surface - density diagram .
another separate branch exists in a more optically - thick ( i.e. , large surface - density ) domain and continues from the standard - disk series to the slim - disk series , via a thermally unstable part , as the accretion rate increases ( e.g. , @xcite ) . specifically for the riaf theories , a more detailed description can be found , e.g. , in @xcite .
the main efforts to take the effects of ordered magnetic fields into account in the accretion disk theories may be divided into two classes . in one class ,
the presence in the disk of only a toroidal field with even polarity ( i.e. , the polarity is the same on both upper and lower sides of the equatorial plane ) is taken seriously .
the resulting magnetic pressure is added to gas pressure to support the disk against the vertical component of gravity .
further , if the @xmath0-prescription ( @xcite ) with respect to the total pressure is adopted for a viscosity , an additional viscous extraction of angular momentum passing through the disk plane becomes possible .
for example , the modifications of the standard - disk theory ( e.g. , @xcite ) and of riafs ( e.g. , @xcite ) have been discussed , respectively , in relation to some controversial spectral features seen in cataclysmic variables and to the state transitions seen in galactic black - hole x - ray binaries . in the other class , on the contrary
, the presence of both poloidal and toroidal components of an ordered field are taken seriously . since the toroidal component is considered to appear as a result of dragging of the vertical field lines by the rotational motion of the disk , its polarity reverses on both sides of the equatorial plane ( i.e. , an odd polarity ) .
thus , the toroidal component develops mainly outside the disk and vertically compresses the disk against gas pressure .
moreover , such a configuration makes it possible to vertically extract the angular momentum by the maxwell stress .
this point is essential in relation to the production of astrophysical jets ( e.g. , @xcite ; @xcite ; @xcite ) often observed to emanate from the vicinity of the disk inner edge . in most of the analytic models that is addressed to the formation of jets ,
however , the magnetic field is not treated self - consistently with the fluid disk .
self - consistent inclusion of an ordered magnetic field into riaf states has been performed in a series of works by the present author ( for a review , see @xcite ; hereafter referred to as paper i ) . in this model ,
a twisted magnetic field works to extract angular momentum from the disk plasma , and the resistive dissipation converts the available gravitational energy into heat .
this makes a good contrast with the usual riaf models , in which only turbulent magnetic fields are included , and the fluid viscosity plays an essential role in converting energy and extracting angular momentum .
therefore , we call the former the resistive - riaf model , distinguished from the latter , the viscous - riaf model
. it should be mentioned also that there is another series of studies in which the presence of an ordered magnetic field is treated self - consistently ( for a review , see @xcite ) . although its relation to riafs is not so clear , ferreira and his coworkers discuss an inner region of the accretion disk threaded by a global magnetic field .
their main interest is in the local ( i.e. , at a given radius ) mechanisms to launch magnetohydrodynamic ( mhd ) jets , and the details of vertical transport of energy and angular momentum are investigated . on the other hand ,
the present concern of the resistive - raif model is to show how the energy can be supplied to the jet launching site from wide aria of an accretion disk .
this paper is a direct follow - up of paper i that has been devoted to discuss the appearance of the poynting flux near the inner edge of a resistive - riaf , which may lead to the jet launching .
however , the discussion was based on the inward extrapolation of an outer asymptotic solution whose accuracy is not necessarily guaranteed in the inner region .
moreover , the outer solution has been derived by assuming a specific condition , which we call hereafter the extended riaf condition ( equation [ 9 ] in paper i or [ [ eqn : exriaf ] ] below ) .
this condition may seem rather arbitrary or artificial .
therefore , we give it up in the present paper . instead ,
according to the spirit of this condition , we first obtain several asymptotic solutions in the outer region of an accretion disk , which are equivalent to each other within the accuracy to the first order in the smallness parameter @xmath1 ( the definition will be given in the next section ) . under the above situation , the criterion to sift a specific solution from others would be the wideness of its applicability range .
thus , we are naturally led to examine the behavior of these outer solutions in the opposite limit of small radius , and find that only one case among them becomes exact also in this limit .
namely , the selected one becomes accurate not only in the limit of large radius but also in that of small radius .
therefore , it may be called a global solution , although it is still an approximate one at middle radii .
this finding is indeed a great improvement since we can discuss global operation of such accretion flows based on this much secure ground than before .
another advantage of this improved solution is that the expressions for all relevant physical quantities are written analytically in closed forms .
the organization of this paper is as follows . in section 2 , the variable - separated version of the governing equations
are summarized , extracted from paper i for convenience . as plausible examples of asymptotic solutions at large radii ,
four possibilities are derived in section 3 without employing the extended riaf condition . by examining the behavior of these outer asymptotic solutions in the opposite limit of small radius , we find in section 4 that there is one and only one case in which the same expressions become asymptotically exact also in this limit .
full expressions for the relevant quantities within this global solution are derived also in this section . using these expressions , we calculate and discuss in section 5 the local energy budgets of a few types . as summarized in the final section
, the obtained results clearly show the appearance of preferable circumstances for jet launching .
as for notation , we completely follow paper i , and hence adopt spherical polar coordinates , ( @xmath2 , @xmath3 , @xmath4 ) . the normalized versions of the radius and co - latitude are , @xmath5 , and @xmath6 , respectively . here , @xmath7 denotes the inner - edge radius of the accretion disk .
all physical quantities have been expressed in the variable - separated forms , equations ( 16)-(30 ) in paper i , should read @xmath8 . ] within the geometrically - thin disk approximation where the half opening - angle of the accretion disk is very small , i.e. , @xmath9 .
the fundamental equations for the radial - part functions in quasi - stationary problems ( for the definition , see paper i ) are summarized below .
* leading equations * * magnetic flux conservation : @xmath10 * ampre s law : @xmath11 @xmath12 @xmath13 * ohm s law : @xmath14 @xmath15 @xmath16 where @xmath17 @xmath18 * mass continuity : @xmath19 * equation of motion : @xmath20 \tilde{v}_r^2 = \tilde{v}_{\varphi}^2 - v_{\rm k}^2 , \label{eqn : eomr } \end{aligned}\ ] ] @xmath21 @xmath22 where @xmath23 * equation of state : @xmath24 * subsidiary equations * * faraday s law : @xmath25 @xmath26 @xmath27 * charge density : @xmath28 one of the characteristic aspects of our treatment is that ohm s law is used directly , without substituted into faraday s law . the poloidal and toroidal magnetic - reynolds numbers , equations ( [ eqn : rep ] ) and ( [ eqn : ret ] ) , have been introduced in rewriting ohm s law . the final expression for @xmath29
can be derived by using equation ( 73 ) of paper i. the expression ( [ eqn : eomr ] ) for the radial component of equation of motion ( eom ) follows from equation ( 93 ) of paper i , but the second term on the right - hand side ( rhs ) of the latter has been shifted to the left - hand side ( lhs ) . the derivation may become evident , when one is referred to the identity @xmath30 in equation ( [ eqn : eomr ] ) , @xmath31 denotes the kepler velocity @xmath32 with @xmath33 and @xmath34 being the gravitational constant and the mass of a central object , respectively .
the expression ( [ eqn : eomph ] ) for the azimuthal component of eom has been quoted from equation ( 102 ) of paper i. we often refer to the expansion parameter @xmath35 that becomes very small at large radii , where @xmath36 and @xmath37 are the radial and azimuthal components of the velocity , respectively .
the suffix vpf means that the quantity is evaluated by the lowest - order solution , the vanishing poynting - flux ( vpf ) solution ( see paper i for details ) .
the original riaf condition , @xmath38 has been assumed in the asymptotic region at large radii .
this condition means that the ratio of the pressure gradient force to the gravity should be a constant in this region , which guarantees a characteristic nature of optically thin riafs , i.e. , a virial - like temperature .
combined with the equation ( 68 ) of paper i , which is the lowest - order version of the @xmath2-component of eom in the power series of @xmath1 , we have derived a lowest - order solution ( @xcite ) in the asymptotic region . since the poynting flux vanishes identically in this solution ( i.e. , the vpf solution ) , it can not explain the jet launching which is commonly expected for the accretion disks of the resistive - riaf type . this is because the electrodynamic launching surely requires the supply of jet - driving power through the poynting flux . in order to overcome this difficulty
, we tried in paper i to improve the accuracy of the solution by taking the first - order corrections in @xmath1 into account . in this connection , we have replaced the original riaf condition by the extended riaf condition , @xmath39 here , @xmath40 is a constant which controls the radial profiles of relevant physical quantities in the vpf solution , and in this sense plays a similar role to the polytropic index that replaces the energy transfer equation ( see paper i ) .
in contrast to the original riaf condition , the extended one does not have firm grounds to stand on , except that it becomes identical with the original one in the vpf limit .
therefore , we can not reject the criticism that it is only a makeshift policy , and if possible , such an obscure postulate should be avoided in deriving higher - order solutions .
when the above postulate has been removed , the only remaining requirement is that any improved solution should coincide with the vpf solution in the limit of vanishing corrections . under such circumstances , the new policy we adopt here is to portion out @xmath41 and @xmath42 from equation ( [ eqn : eomph ] ) so as to reproduce a term that is proportional to @xmath43 or @xmath44 , as the leading contribution from the partial - pressure gradient force appearing in the identity ( [ eqn : id_1 ] ) .
this requirement is equivalent to the condition ( [ eqn : orriaf ] ) or ( [ eqn : exriaf ] ) , respectively , as far as the leading - order terms are concerned .
even if we follow this policy , the solution is not determined uniquely , and anyway the process of finding solutions becomes necessarily a kind of trial and error . in the following subsections ,
we show four examples of successful trials .
they are all different from the outer asymptotic solution obtained in paper i , which has been derived under the restriction of extended riaf condition .
although only one of them leads to a final global solution , we dare to mention all of the new results .
we believe that such a description will be helpful for the reader to become familiar with subtle insight into the strategies in finding solutions and to experience how the type of an accretion flow ( i.e. , sub- or trans - critical infall ) is actually determined .
if that had been omitted , the description of this paper would become very abrupt and less understandable . in specifying the radial profiles of the velocity and magnetic fields ,
it is natural to first follow the results obtained in paper i. the velocity components have been written as @xmath45 @xmath46 where @xmath47 is a positive constant whose value is specified in the course of discussion , and @xmath48 the profile of the @xmath2-component for the magnetic field has been fixed as @xmath49 at the disk outer edge @xmath50 , the magnitude of this component becomes comparable to the externally imposed uniform field , @xmath51 , as guaranteed by the relation @xmath52 .
the above specifications result in the derivatives @xmath53 and further from equations ( [ eqn : fcont ] ) , ( [ eqn : rep ] ) and ( [ eqn : eomph ] ) , respectively , @xmath54 @xmath55 and @xmath56 it is evident that the above set of selections for @xmath37 and @xmath57 is effective in keeping the expression ( [ eqn : ret_1 ] ) rather simple .
* case 1*. when we portion out @xmath42 and @xmath41 from equation ( [ eqn : ret_1 ] ) as @xmath58 @xmath59 we obtain the relation @xmath60 reproducing one of our aimed form . since the above determined @xmath42 is a constant smaller than unity for any value of @xmath40 in the allowed range , @xmath61 ( see @xcite ) , the accretion flow in this case is said to be a ` sub - critical ' ( i.e. , @xmath62 ) .
actually in our treatment , however , no such criticality exists at @xmath63 . ] infall .
although @xmath64 as confirmed from @xmath65 , the solution obtained in case 1 does not belong to a vpf solution since @xmath66 ( i.e. , @xmath67 ) .
the form of @xmath68 follows from the definition of @xmath41 : @xmath69 which yields @xmath70 at large radii , the first term on the rhs of equation ( [ eqn : lnbp_1 ] ) is the quantity of @xmath71 while the second and third terms are of @xmath72 .
they are referred to , respectively , as @xmath73 and @xmath74 .
then , shifting only the leading - order term resulting from ( [ eqn : lnbp_1 ] ) to the rhs in the @xmath2-component of eom ( [ eqn : eomr ] ) , we have @xmath75 on the other hand , the lhs of eom ( [ eqn : eomr ] ) becomes @xmath76\ \tilde{v}_r^2 \nonumber \\ & \simeq & -\left\{\frac{2n+9}{2(2n+1 ) } - 2a\right\ } \tilde{v}_r^2 .
% \label{eqn : lhs_1}\end{aligned}\ ] ] in the final line of the this equation , only the lowest - order terms in @xmath77 have been kept in the curly brackets in order to match the form of the rhs .
finally , equating both sides , we can fix the value of @xmath47 as @xmath78 * case 2*. the other option for portioning out of @xmath42 and @xmath41 is @xmath79 @xmath80 in which we have @xmath68 of the form @xmath81 and the relations @xmath82 @xmath83 differently from case1 , the accretion flow in this case is a ` trans - critical ' ( see the footnote 2 ) infall because it starts at a subcritical velocity ( i.e. , @xmath84 ) at the disk outer edge @xmath50 ( @xmath85 ) and reaches a super - critical value ( i.e. , @xmath86 ) at small radii ( @xmath87 ) .
after shifting only the term including @xmath73 to the rhs of equation ( [ eqn : eomr ] ) , we have @xmath88 the remaining lhs becomes @xmath89 according to the same procedure as in case 1 . equating both sides of this equation , we obtain @xmath90 which requires @xmath91 since @xmath47 should be positive definite . in this subsection , we discuss fairly different types of expressions for the radial profiles of the velocity and magnetic fields .
the new guideline in specifying them is to pay special attention to the identity @xmath92 where @xmath93 these are monotonic functions as shown in figure 1 , and their behaviour at large radii is @xmath94 within the accuracy to the first order in @xmath77 .
first , the radial profile of @xmath57 is specified as @xmath95 it is easy to confirm that equations ( [ eqn : br_2 ] ) and ( [ eqn : br_1 ] ) are equivalent within the accuracy to the first order in @xmath77
. then , we have @xmath96 from the relation @xmath97 in order to keep equation ( [ eqn : eomph ] ) simple , we are naturally led to select @xmath98 which gives @xmath99 for the radial velocity component , it turns out after some trials and errors that the specification @xmath100 and hence @xmath101 is very interesting to examine .
then , the velocity ratio remains the same as in subsection 3.2 , and we find very simple expressions for the magnetic reynolds numbers : @xmath102 similarly to the discussion in the previous subsection , there are again two possibilities in portioning out @xmath42 and @xmath41 from the latter of equation ( [ eqn : ret_2 ] ) .
* case 3*. the first possibility is the specifications , @xmath58 @xmath103 which describe a subcritical accretion flow analogously to case 1 . in this case
, we have the results @xmath104 and @xmath105 @xmath106 in the last equation derived above , the two terms on the rhs of the logarithmic derivative of @xmath68 are both of order unity at large radii , since @xmath107 as @xmath108 . therefore , we shift the whole term containing @xmath109 on the lhs of eom ( [ eqn : eomr ] ) to the rhs : @xmath110 \\ & & \qquad \simeq -\frac{6a}{2n+1}\ \tilde{v}_r^2 , \nonumber % \label{eqn : rhs_3}\end{aligned}\ ] ] where the last expression is the limiting form at large radii calculated with the aid of equation ( [ eqn : lgfs ] ) .
note that all the leading - order terms in the square brackets cancel out in this limit and we have only a first order term ( @xmath111 ) .
the terms remaining on the left is @xmath112\ \tilde{v}_r^2 \nonumber\\ & = & -\frac{1}{2(2n+1)}\left\{4n+10-(2n+1)s\right\}\ \tilde{v}_r^2 \\ & \simeq & -\frac{2n+9}{2(2n+1)}\ \tilde{v}_r^2 , \nonumber % \label{eqn : lhs_3}\end{aligned}\ ] ] where the last expression is also the approximate form at large radii .
therefore , this equation holds asymptotically at large radii , as far as @xmath113 * case 4*. the last option in our consideration is the profiles @xmath114 @xmath115 which describe a trans - critical accretion flow analogously to case 2 . in this case
, we have the results @xmath116 and @xmath117 @xmath118 as in case 3 , both terms resulting from the above logarithmic derivative of @xmath68 are of order unity , and hence the term containing this factor in the @xmath2-component of eom should be shifted altogether to the right .
then , the rhs becomes @xmath119 where the last expression is the limiting form at large radii .
again , note that the leading - order terms in the curly brackets have been completely cancelled out in this limit . on the other hand ,
the lhs becomes @xmath120\ \tilde{v}_r^2 } % \nonumber \\ & = & -\frac{1}{6}\ v_{\rm k ,
in}^2\xi^{-2}\ \left[\ 2(2n+1 ) \right .
\nonumber \\ & & \qquad\qquad\qquad \left .
+ \left\{8-(2n+1)s\right\}\xi^{-1}f\right ] \label{eqn : lhs_4 } \\ & \simeq & -\frac{2n+9}{2(2n+1)}\ \tilde{v}_r^2 .
\nonumber \end{aligned}\ ] ] therefore , the equation holds in the outer asymptotic region , when @xmath121
in the previous section , we have obtained four different sets of asymptotic solutions at large radii .
although these sets have different expressions for any one of the relevant physical quantities ( and their components ) , they are equivalent within the accuracy to the first order in @xmath1 ( @xmath122 ) . as far as we remain only in this outer asymptotic region
, we can not therefore judge which type of the accretion flows ( e.g. , the sub - critical or trans - critical type ) , is more likely to fit for the reality .
for the resolution of this problem , considerations from a global point of view are needed .
thus , we are led to examine the behavior of the above sets in the opposite limit of small radius . fortunately , as shown below , there is one and only one case ( case 4 ) in which the same set also serves as the asymptotic solution at small radii .
this means that this set can be regarded as a global solution , though the accuracy may be somewhat poor in the middle region . in order to discuss the small radius limit ,
we note here that @xmath123 in the first three cases discussed in the pervious section , the asymptotic behavior is different on both sides of the @xmath2-component of eom .
indeed we obtain , lhs@xmath124 and rhs@xmath125 in case 1 ; lhs@xmath124 and rhs@xmath125 in case 2 ; lhs@xmath124 and rhs@xmath111 in case 3 . however , in case 4 , equations ( [ eqn : lhs_4 ] ) and ( [ eqn : rhs_4 ] ) yield , respectively , @xmath126 @xmath127 then equating both sides , we obtain @xmath128 which specifies the value of @xmath47 in the asymptotic solution at small radii . if the values of @xmath47 in the two asymptotic regions at large and small radii ( i.e. , equations [ [ eqn : ala ] ] and [ [ eqn : asm ] ] ) coincide , the asymptotic solutions match smoothly and become a global solution .
this actually happens when @xmath129 the presence of a select value of @xmath40 may suggest that a preferable thermodynamic circumstance is required for realization of the state described by our global solution .
the global solution indicates that the infall has a trans - critical nature as seen from equation ( [ eqn : d_4 ] ) .
however , this should not be interpreted as a restriction on the radial profiles of infall velocity , since it is fixed always by equation ( [ eqn : vr_2 ] ) for both cases 3 and 4 .
rather , it should be interpreted as a restriction on the profile of the characteristic velocity @xmath130 , and hence , on those of @xmath57 and @xmath131 .
other physical quantities ( than @xmath36 , @xmath37 , @xmath57 , @xmath132 , and @xmath68 ) are derived straightforwardly as follows .
it should be noted that all quantities are written in closed forms , i.e. , not in the forms of truncated power series in @xmath77 .
we obtain from equation ( [ eqn : defd ] ) @xmath133 and further substituting it in equation ( [ eqn : mcont ] ) , @xmath134 equation ( [ eqn : vthf ] ) indicates that the generation of a wind from an accretion disk is directly controlled by the parameter @xmath40 ( differently from the result in paper i ) .
the direction of the wind is upward ( i.e. , vertically outgoing ) when @xmath135 , and downward ( i.e. , vertically ingoing ) when @xmath136 .
the above selected value , @xmath137 , means the presence of a medium - strength upward wind .
it can be seen in figure 1 that , as far as the profiles of the velocity components are concerned , they are essentially the same as in paper i. ( 85 mm , 95mm)figure1.eps the pressure and temperature are calculated respectively from equations ( [ eqn : eomth ] ) and ( [ eqn : eos ] ) as @xmath138 @xmath139 the deviation of temperature from the virial form is expected only in a middle region and remains to be rather small .
the results for every component of the current density and the electric field follow from ampre s law and ohm s law : @xmath140 @xmath141 @xmath142 @xmath143 @xmath144 @xmath145 similarly to the case of @xmath146 , the coefficient of @xmath147 is proportional to @xmath40 .
it seems rather natural from the viewpoint of current closure that the value of @xmath40 determined in equation ( [ eqn : na ] ) is no - zero .
finally , we cite the component expressions for the poynting flux , including their @xmath148-dependences since they have been dropped in paper i by accident .
@xmath149 @xmath150 @xmath151 ( 85 mm , 95mm)figure2.eps as it turns out from figure 2 , @xmath152 and @xmath153 are negative everywhere , but @xmath154 changes sign from slightly positive ( outgoing ) to negative ( incoming ) as the radius decreases across @xmath155 . in order to see this fact more clearly
, the formula in the curly brackets in equation ( [ eqn : pfth ] ) is shown as @xmath156 in the figure .
taking the @xmath148-dependece of @xmath157 also into account , we can say that the poynting flux in the poloidal plane flows from the wide outer region ( @xmath158 ) into the narrow inner region ( @xmath159 ) , almost along the surface of an accretion disk .
it should be noted that @xmath154 here has the opposite sign to that obtained in paper i , in the inner region where @xmath154 becomes very large .
this means that the extrapolation of an outer solution that is not the global solution can actually lead to an erroneous result .
before proceeding to energy budgets , it is convenient to introduce the mass accretion rate through a vertical cross - section of arbitrary radius , @xmath160 where the approximation @xmath161 has been used , extrapolating the functional form near the equator even to large @xmath148 regions ( i.e. , neglecting the @xmath8 term ) .
this is because the @xmath148-dependences are reliable only near the disk midplane owing to the adopted method of approximation ( see paper i ) .
since we need a co - latitudinal integration in the above calculation , this may cause some worry about the accuracy of the result .
however , the most important thing is a finiteness of the integral and its precise value does not matter on the essence of the following discussions . reflecting the presence of the vertical flows ,
@xmath162 varies with @xmath163 like @xmath164 .
it is a constant only when @xmath165 , i.e. , there is no wind . the accretion rate at the outer edge of an accretion disk
is given explicitly as @xmath166 because , at this radius , @xmath167 can be approximated by its asymptotic form , @xmath168 .
first substituting @xmath169 , @xmath170 and @xmath171 in the above definition , and then solving for @xmath172 , we obtain @xmath173 this equation will be used to eliminate @xmath172 in the following subsections .
the heating rate due to the joule dissipation of the electric current per unit volume is calculated as @xmath174 in obtaining the approximate expression in the first line of the above equations , we have neglected @xmath175 since @xmath176 , @xmath177 , and also neglected a term containing @xmath8 dependence . in obtaining the expression on the third line , @xmath178 and @xmath172 have been eliminated with the aids of equation ( 73 ) in paper i and equation ( [ eqn : b0 ] ) above , respectively . on the other hand ,
the advection cooling rate per unit volume is written as @xmath179 where @xmath180 is the enthalpy for an ideal gas . similarly to the case of @xmath181 , the approximate expression on the first line of the above equations is obtained by neglecting the @xmath8 terms . however , the heating and cooling terms given above does not balance generally .
this is because i ) there may exist the exchange of thermal energy between neighboring fluid elements through conduction and/or convection ( these are expressed as non - adiabaticities ) , even if the radiation loss may be negligibly small , or ii ) the balance can not be achieved in principle , implying the break - down of the model .
the amount of this discrepancy is @xmath182 when it is expressed in terms of an additional heating ( or a cooling if it is negative ) .
there are two distinct terms in the above result ; the term which contains @xmath40 and which does not .
the former dominates over the latter in the outer asymptotic region ( i.e. , in the vpf limit ) . in this asymptotic region , this term has been interpreted as due to the non - adiabaticities ( @xcite ) .
our global solution indicates that both terms enhance in a narrow region at around @xmath183 .
even in such inner regions , the former drives an upward wind ( i.e. , @xmath135 ) when there is additive heating ( i.e. , @xmath91 ) , and vice versa , as seen in equation ( [ eqn : vthf ] ) . on the other hand
, the latter term always contributes to a cooling .
although it is not so certain , this fact may suggests that the latter corresponds to a radiative cooling that may enhance at such an inner region .
as shown in figure 3 , the advection cooling becomes largely exceeds the joule heating where @xmath184 .
the cause of this enhancement is in the monotonic increase of @xmath36 and @xmath146 toward the center . in principle
, it is impossible to balance the advective cooling even by supplying heat through conduction or convection , as far as the cooling there exceeds the peak value of the joule heating . in this sense
, the present model does not seem to hold in the region where @xmath185 .
thus , we are consistently led to the conclusion that there is an inner edge of an accretion disk at @xmath186 .
the energy equation for the electromagnetic field is known as the poynting theorem : @xmath187 where @xmath188 ( see paper i ) is the electromagnetic energy density , and @xmath189 is the poynting flux .
the two terms on the right - most side of equation ( [ eqn : pyt ] ) represent losses of the electromagnetic energy through the joule dissipation and through work done by the magnetic force on the fluid , respectively .
when we concentrate our attention mainly to the region near the disk midplane , we can neglect @xmath8 terms as before .
the contribution from @xmath190 vanishes in this process , and the poynting equation finally reduces to the form @xmath191 as the above definition of @xmath192 includes a minus sign , it means the work done by fluid on the electromagnetic force ( hence , on the field ) . the explicit expression of @xmath193 can be reached either by calculating the right - hand side of the expression on the middle line or by calculating its left - hand side directly according to its definition , confirming that the poynting equation actually holds in our case . since @xmath194 at any finite radius , the fluid motion is doing work on the electromagnetic field everywhere in the disk . as seen from figure 2 , the vertical poynting flux ,
@xmath154 , is positive in the region @xmath158 while negative in @xmath159 .
this implies that @xmath195 where @xmath158 and @xmath196 where @xmath159 .
namely , in the wide outer region ( @xmath158 ) the fluid in the disk is doing work that exceeds the local dissipation @xmath197 and the excess is gathered into the small inner region ( @xmath159 ) through the poynting flux ( see also the discussion at the end of the previous section ) .
thus , in the inner region , energy input through the poynting flux can largely exceeds local supply through the work @xmath198 , and both are thermalized as the joule dissipation ( i.e. , @xmath199 ) .
in other words , the outer main disk is driving the innermost region electrodynamically , suggesting that , if a more careful treatment of the vertical structure and flow in this region is introduced in the model , the launching of an mhd jet could be obtained . within the present status of our accretion disk model ,
however , this ability is wasted only on the joule dissipation there . in a sense ,
the new global picture of our model is reasonable because it closes within itself . on the other hand , the previous picture stated in paper i is less persuasive , because it does not close unless assuming a presence of some external component ( e.g. , an mhd jet ) that is driven by the large positive ( i.e. , outgoing ) poynting flux mainly emanating from the innermost disk .
( 85 mm , 95mm)figure3.eps the binding energy per unit volume of a fluid element is described by the bernoulli sum ( , ) , @xmath200 where the approximations of geometrically - thin disk and that of respecting midplane have been used again in obtaining the expression on the second line . calculating this quantity in terms of our global solution
, we obtain @xmath201 the asymptotic values for large and small @xmath163 are @xmath202 the upper line of the above equation reproduces the vpf result ( @xcite ) .
as seen in figure 3 , the global solution predicts that @xmath203 remains positive everywhere ( cf .
, ; however see also , @xcite ; @xcite ; @xcite ; @xcite ; @xcite ) and becomes divergently large in the limit of @xmath204 , even when @xmath165 .
this fact suggests that the accretion flows described by our global solution should be ejected , at least in its fraction , before it reaches the center .
summarizing the discussions in the previous section , we have reached the conclusions that i ) the accretion state characterized under the name of resistive - riaf does not seem to extend into the region @xmath184 , ii ) the main disk ( @xmath158 ) is driving the innermost region ( @xmath159 ) electrodynamically , and iii ) infalling matter always stay unbound and can not reach the gravitational center , as a whole . judging from these evidences ,
the most probable scenario is an ejection of the infalling matter , at least in its fraction , at around @xmath186 which may be regarded as the inner edge of an accretion disk .
thus , it has been shown clearly that one of the most preferable circumstances necessary for the mhd jet launching is actually prepared within our resistive - riaf model .
abramowicz , m. , lasota , j .-
, & igumenshchev , i.v . , 2000 , , 314 , 775 beckert , t. , 2000 , apj , 539 , 223 begelman , m. c. , & pringle , j. e. , 2007 , , 375 , 1070 blandford , r. d. , & begelman , m. c. , 1999 , , 303 , l1 blandford , r. d. , & payne , d. g. , 1982 , , 199 , 883 bondi , h. , 1952 , mnras , 112 , 195 ferreira , j. , 2008 , new astron .
, 52 , 42 kaburaki , o. , 2000 , apj , 531 , 210 kaburaki , o. , 2001 , apj , 563 , 505 kaburaki , o. , 2012 , pasj , 64 , 39 , paper i kato , s. , fukue , j. , & mineshige , s. , 2008 , black - hole accretion disks : towards a new paradigm ( kyoto university press , kyoto ) lubow , s.h . , papaloizou , j. c. b. & pringle , j. e. 1994 , , 268 , 1010 nakamura , k. e. , 1998 , , 50 , l11 narayan , r. & mcclintock , j. e. 2008 , new astron .
51 , 733 narayan , r. & yi , i. 1994 , apj , 428 , l13 narayan , r. & yi , i. 1995 , apj , 444 , 231 oda , h. , machida , m. , nakamura , k. e. , matsumoto , r. & narayan , r. 2012 , , 64 , 15 parker , e. n. , 1960 , apj , 132 , 175 pudritz , r. e. & norman , c. a. 1983 , apj , 274 , 677 turolla , r. & dullemond , c. p. 2000 , , 531 , l49 | in our recent paper , we have obtained a model solution to the problem of radiation - inefficient accretion flows ( riafs ) in a global magnetic field ( so called , resistive riaf model ) , which is asymptotically exact in outer regions of such flows forming accretion disks . when extrapolated inwardly , the model predicts a local enhancement of the vertical poynting flux within a small radius that may be regarded as the disk inner - edge .
this fact has been interpreted as the origin of power source for the astrophysical jets observationally well - known to be ejected from this type of accretion disks .
since the accuracy of the solution may become rather poor in such inner regions , however , the ground of this assertion may not seem to be so firm . in the present paper
, we develop a sophisticated discussion for the appearance of jet - driving circumstances , based on a much more firm ground by deriving a global solution in the same situation .
although the new solution still has an approximate nature , it becomes exact in the limits not only of large radius but also of small radius .
the analytic results clarify that the electrodynamic power is gathered by the poynting flux , from outer main - disk region to feed the innermost part of an accretion disk .
the injected power largely exceeds the local supply of work by the fluid motion . |
the multiple descriptions ( md ) problem was proposed in the late seventies and has been studied extensively since , yielding a series of advances , ranging from the derivation of asymptotic bounds @xcite to practical approaches for multiple descriptions quantizer design @xcite .
it was originally viewed as a method to cope with channel failures , where multiple source descriptions are generated and sent over different paths .
the encoder generates @xmath2 descriptions for transmission over @xmath2 available channels .
it is assumed that the decoder receives a subset of the descriptions perfectly and the remaining are lost , as shown in fig .
[ fig : basic_md ] .
the objective of the md problem is to design the encoders ( for each description ) and decoders ( for each possible received subset of the descriptions ) , with respect to an overall rate - distortion ( rd ) trade - off .
the subtlety of the problem is due to the balance between the full reconstruction quality versus quality of individual descriptions ; or noise free quality versus the amount of redundancy across descriptions needed to achieve resilience to descriptions loss .
channel multiple descriptions setup : each description is received error free or is completely lost at the decoder[fig : basic_md ] ] one of the first achievable regions for the 2-channel md problem was derived by el - gamal and cover ( ec ) in 1982 @xcite .
it follows from ozarow s results in @xcite that the ec region is complete for the 2-channel quadratic gaussian md problem , i.e. , when the source is gaussian and the distortion measure is mean squared error ( mse ) .
it was further shown by ahlswede in @xcite that the ec region is complete for a cross - section of the 2-description md setup , called the ` no - excess rate regime ' , a scenario wherein the central decoder receives information at the minimum sum rate .
this led to the popular belief that the ec achievable region is complete for the general 2-channel md problem .
however zhang and berger ( zb ) in @xcite proved a then surprising result , that the ec scheme is strictly sub - optimal in general .
in particular , they showed that for a binary source under hamming distortion , sending a common codeword in both descriptions can achieve points that are strictly outside the ec region .
while introducing a common codeword implies explicit redundancy among the two descriptions , this codeword assists in better coordination between the descriptions leading to a strictly larger rd region .
several researchers have since focused on extending ec and zb to the @xmath0channel md problem @xcite .
an achievable scheme , due to venkataramani , kramer and goyal ( vgk ) @xcite , directly builds on ec and zb , and introduces a combinatorial number of refinement codebooks , one for each subset of the descriptions .
motivated by zb , a _
common codeword is also shared by all the descriptions , which assists in better coordination of the messages , improving the rd trade - off . for the @xmath0channel quadratic gaussian problem ,
it was shown by wang and viswanath in @xcite that a special case of the vkg coding scheme , where no common codeword is sent , achieves the minimum sum rate when only the individual and the central distortion constraints are imposed .
in particular , they showed that a ` correlated quantization ' based encoding scheme , which is an extension of the ozarow s encoding mechanism to the @xmath2-descriptions problem , achieves the minimum sum rate for the cross - section involving constraints only on the individual and central distortions .
it was also shown recently by chen in @xcite that , in fact , this approach leads to the complete region for this particular cross - section .
pradhan , puri and ramachandran ( ppr ) considered a practically interesting cross - section of the general @xmath0channel md problem in @xcite called the ` symmetric md problem ' wherein it is assumed that the rates of all the descriptions are equal and the distortion is a function only of the ` number ' of descriptions received rather than which particular subset is received .
they proposed a new coding scheme leveraging principles from distributed source coding @xcite , and particularly slepian and wolf s random binning techniques , and showed that , for this symmetric cross - section , the proposed encoding scheme improves upon the vkg region .
tian and chen derived a new coding scheme in @xcite for the symmetric md problem which further extends the ppr region .
it was also shown in @xcite that this region is very close to complete for the symmetric quadratic gaussian md problem .
recently , wang and viswanath @xcite derived a coding scheme based on the vkg and the ppr encoding principles and showed its sum - rate optimality for certain cross - sections of the quadratic gaussian problem wherein only 2 layers of distortions are imposed . moreover , in @xcite , song et.al . showed that , for the quadratic gaussian setting , when the minimum sum rate is attained subject to two levels of distortion constraints ( with the second level imposed on the complete set of descriptions ) , the ppr scheme also leads to the minimum achievable distortion at the intermediate levels .
in this paper we present a new encoding scheme involving `` combinatorial message sharing '' ( cms ) , where a unique common codeword is sent in ( shared by ) each subset of the descriptions , thereby introducing a combinatorial number of _ shared codebooks _
, along with the refinement codebooks of @xcite .
the common codewords enable better coordination between descriptions , providing an improved overall rd region .
we derive an achievable region for cms and show that it subsumes vkg for general sources and distortion measures .
moreover , we show that cms achieves a strictly larger region than vkg for all @xmath3 , if there exists a 2-description subset for which zb achieves points strictly outside the ec region .
in particular , cms achieves strict improvement for a binary source under hamming distortion .
we then show a surprising result : cms strictly outperforms vkg for a general class of sources and distortion measures , which includes several scenarios in which , for every 2-description subset , the zb and the ec regions coincide .
in particular , we show that for a gaussian source under mse , cms achieves points strictly outside the vkg region .
this result is in striking contract to the corresponding 2-descriptions setting .
optimality of ec for the 2-descriptions setup has led to a natural conjecture that common codewords do not play a necessary role in quadratic gaussian md coding , and all the achievable regions characterized so far neglect the common layer ( see , e.g. , @xcite ) . in this paper , we show that the common codewords in cms play a critical role in achieving the complete rd region for several asymmetric cross - sections of the general @xmath0channel quadratic gaussian md problem .
we note that the principles underlying cms have been shown in precursor work to be useful in related applicational contexts of routing for networks with correlated sources @xcite and data storage for selective retrieval @xcite .
we further note that a preliminary version of the results in this paper appeared in @xcite , @xcite and @xcite .
we also note that , in this paper , we focus on generalizing the vkg coding scheme using a combinatorial number of shared messages for the md problem with general sources and distortion measures .
however , the cms principle can be extended to incorporate random binning based encoding techniques , to utilize the underlying symmetry in the problem setup .
preliminary results in this direction have recently appeared in @xcite and form part of our current research focus .
the cms scheme introduces a structured mechanism for generating random codebooks that lead to controlled redundancy across transmitted messages .
the cms scheme is only one of the several possible generalizations of the zb scheme to @xmath2-descriptions .
the fundamental idea of introducing structured codes and controlled redundancy for @xmath0channel multiple descriptions problem is not new and has been explored in several publications in the past , including @xcite and some of these schemes have been shown to outperform vkg for certain sources and distortion measures .
however , to the best of our knowledge , for a gaussian source , under mse , cms is the only generalization of the vkg scheme that achieves a strictly larger rate - distortion region .
there are other encoding schemes that can achieve points strictly outside the vkg region , such as the ppr scheme , but are not strict generalizations vkg in the sense that their achievable region has not been shown to always subsume the entire vkg rate - distortion region .
hence there are fundamentally new insights to be gained by studying cms .
the rest of the paper is organized as follows . in section
[ sec : prior_results ] , we formally state the @xmath0channel md setup and briefly describe the approaches and regions of ec @xcite , zb @xcite and vkg @xcite . to keep the notation simple , we first describe in section [ sub:3-descriptions - scenario ] , the cms scheme for the 3 descriptions scenario and extend it to the general case in section [ sub : l - channel - case ] .
we then prove in section [ sec : strict - improvement1 ] that cms achieves strict improvement over vkg whenever there exists a 2-description subset for which zb achieves points outside the ec region . in section [ sec : strict - improvement2 ] we prove that cms outperforms vkg for a fairly general class of source - distortion measure pairs , including a gaussian source under mse . finally in section [ sec : gaussian - mse - setting ] , we derive new results for the @xmath0channel quadratic gaussian md setup and show that cms achieves the complete rd region for several asymmetric cross - sections of the general problem .
we use uppercase letters to denote random variables ( e.g. , @xmath4 ) and lowercase letters to denote their realizations ( e.g. , @xmath5 ) .
we use script letters to denote sets , alphabets and subscript indices ( e.g. , @xmath6 ) .
a sequence of @xmath7 independent and identically distributed ( iid ) random variables is denoted by @xmath8 and its realization by @xmath9 .
@xmath10 denotes the set of all subsets ( power set ) of any set @xmath11 and @xmath12 denotes the set cardinality .
note that @xmath13 .
@xmath14 denotes the set complement ( the universal set will be explicitly specified when not obvious ) . for two sets @xmath15 and @xmath16
, we denote the set difference by @xmath17 .
all constant random variables , that take a single deterministic value , are denoted by @xmath18 . throughout the paper , random variables , distortions and rates
are indexed using sets , e.g. , @xmath19 , where @xmath20 takes values such as @xmath21 , @xmath22 , @xmath23 , @xmath24 etc .
these subscript indices are always subsets of the set @xmath25 , where @xmath2 denotes the number of descriptions , and satisfy all the standard properties of sets . for ease of notation
, we often drop the braces and the comma whenever it is obvious .
for example , @xmath26 , @xmath27 , @xmath28 and @xmath29 are abbreviated as @xmath30 and @xmath31 , respectively .
next , let @xmath11 be a set of subscript indices .
then , we use the shorthand @xmath32 to denote the set of variables @xmath33 ( e.g. , if @xmath34 , then @xmath35 .
note the difference between @xmath36 and @xmath37 .
@xmath36 is a set of variables , whereas @xmath37 is a single variable .
we use the notation in @xcite to denote all standard information theoretic quantities . with a slight abuse of notation
, we use @xmath38 to denote the entropy of a discrete random variable or the differential entropy of a continuous random variable .
we first give a formal definition of the @xmath0channel md problem .
a source produces an iid sequence @xmath39 , taking values over a finite alphabet @xmath40 .
we denote @xmath41 .
there are @xmath2 encoding functions , @xmath42 , which map @xmath8 to the descriptions @xmath43 , where @xmath44 takes values in the set @xmath45 .
the rate of description @xmath46 is defined as : @xmath47 description @xmath46 is sent over channel @xmath46 and is either received at the decoder error - free or is completely lost .
there are @xmath48 decoding functions for each possible received combination of the descriptions @xmath49 , @xmath50 , where @xmath51 takes values in a finite set @xmath52 , and @xmath53 denotes the null set .
the distortion at the decoder when a subset @xmath20 of the descriptions is received is : @xmath54\label{eq : dist_main_constraint}\ ] ] where @xmath55 , is a well - defined bounded single letter distortion measure .
we say that the rd tuple @xmath56 is achievable if there exist @xmath2 encoding functions with rates @xmath57 and @xmath48 decoding functions yielding distortions @xmath58 . the closure of the set of all achievable rd tuples is defined as the ` _
_ @xmath2-channel multiple descriptions rd region _ _ ' dimensions .
] , denoted hereafter by @xmath59 .
] the first achievable region for the 2-channel md problem was given by el - gamal and cover in 1982 @xcite .
their region is denoted here by @xmath60 .
it is given by the convex closure over all tuples @xmath61 for which there exist auxiliary random variables @xmath62 , @xmath63 and @xmath64 ( defined over arbitrary finite alphabets @xmath65 , @xmath66 and @xmath67 , respectively ) , jointly distributed with @xmath4 and functions @xmath68 such that , @xmath69,\,\,\mathcal{k}\subseteq\{1,2\},\mathcal{k}\neq\emptyset\label{eq : egc}\end{aligned}\ ] ] note that the original ec description assumes @xmath70 to be identity functions ( i.e. , @xmath71 ) . however , the two characterizations are equivalent and we use the one described above as it is easier to relate to the @xmath2-channel achievable regions .
we also note that the function @xmath72 can , in general , be made to depend on @xmath73 and @xmath64 .
however , the overall rd region remains unchanged . to see this ,
let us say , for some joint distribution , @xmath72 depends on @xmath73 and @xmath64 .
we can always construct a new joint distribution with the random variable @xmath64 set equal to @xmath74 .
this new joint distribution satisfies all the rate and distortion conditions in ( [ eq : egc ] ) with the new @xmath72 being an identity function .
the order of codebook generation is shown in fig .
[ fig : egc_zb](a ) .
as stated in the introduction , the ec region can be shown to be complete for the 2-descriptions quadratic gaussian md problem @xcite and for the ` no excess rate ' regime for general sources and distortion measures @xcite .
the achievability of the above region is described as follows .
@xmath75 codewords of @xmath62 and @xmath76 codewords of @xmath63 are generated , each of length @xmath7 , using the marginal distributions of @xmath62 and @xmath63 , respectively .
for each pair @xmath77 of codewords , a codeword for @xmath64 is generated according to the conditional density @xmath78 . given a sequence @xmath8
, the encoder looks for a triplet of codewords that are jointly typical with the observed sequence .
information at rate @xmath79 is sent in description 1 and at rate @xmath80 in description 2 .
on receiving either of the the two descriptions , the decoder estimates the source based on the corresponding codeword of @xmath62 or @xmath63 . however , if it receives both the descriptions , it estimates the source based on @xmath64 .
el - gamal and cover showed that , for joint typicality encoding , the probability of error asymptotically approaches zero if the rates satisfy conditions in ( [ eq : egc ] ) .
as the distortion measures are bounded , joint typicality assures the distortion constraints to be satisfied .
given the optimality of ec for the no - excess rate scenario and for the quadratic gaussian setup , it was naturally conjectured that it is optimal in general .
however , zhang and berger @xcite proposed a new coding scheme and an associated rate - distortion region for the 2-channel md problem and showed that it strictly subsumes @xmath60 .
in particular , they showed that for a binary symmetric source under hamming distortion , points outside @xmath60 can be achieved by sending a common codeword in both the descriptions .
this extended region is denoted here by @xmath81 .
they showed that , any tuple is achievable for which there exist auxiliary random variables will become evident later ] @xmath82 , jointly distributed with @xmath4 , for which there exist functions @xmath83 such that , @xmath84,\,\,\mathcal{k}\subseteq\{1,2\},\mathcal{k}\neq\emptyset\label{eq : zb}\end{aligned}\ ] ] the zb region is given by the closure of all such rate - distortion tuples . again note that the zb region is described differently in @xcite , and does not include @xmath64 in its characterization .
however , it was recently shown in @xcite that the above characterization is equivalent to the original one in @xcite . further note
that the above region is convex and does not require ` convexification ' as is the case for @xmath60 .
the codebooks are generated as follows : first , @xmath85 codewords of @xmath86 , each of length @xmath7 , are generated according to the marginal density of @xmath86 .
conditioned on each codeword of @xmath86 , @xmath87 codewords of @xmath62 and @xmath88 codewords of @xmath63 are generated according to the conditional densities @xmath89 and @xmath90 , respectively .
for each codeword tuple @xmath91 , a codeword for @xmath64 is generated according to the conditional density @xmath92 .
similar to ec , descriptions 1 and 2 carry information about codewords of @xmath62 and @xmath63 , respectively .
however , along with these ` _ _ private _ _ ' messages , both the descriptions carry a common component , which is the codeword of @xmath86 .
although this common codeword introduces explicit redundancy , it helps to co - ordinate the two messages well and therefore provides better overall efficiency . the encoding structure of the random variables
is shown in fig .
[ fig : egc_zb](b ) .
note the functional difference of the two random variables @xmath64 and @xmath86 .
the codeword corresponding to @xmath86 is sent in _ both _ the descriptions , whereas the information corresponding to the codeword of @xmath64 is ( implicitly ) _ split _ between the two descriptions .
we therefore call @xmath86 the _ common random variable _ and @xmath64 as the ` _ _ refinement random variable _ _ ' . random variables @xmath62 and @xmath63 form the so called ` _ _ base layer random variables _ _ ' . ) indicates ` common random variable ' .
red ( @xmath93 ) indicates ` base layer random variables ' and blue ( @xmath94 @xmath95 ) indicates ` refinement random variables ' .
[ fig : vkg-1 ] ] an extension of @xmath60 and @xmath81 to the @xmath0channel setup was proposed by venkataramani et al . in @xcite .
the resulting region is denoted here by @xmath96 , and described as follows .
let @xmath97 be any set of @xmath98 random variables distributed jointly with @xmath4 .
then , an rd tuple is said to be achievable if there exist functions @xmath99 such that : @xmath100\label{eq : vkg}\end{aligned}\ ] ] @xmath101 .
the closure of the achievable tuples over all such @xmath98 random variables is @xmath96 . here
, we only present an overview of the encoding scheme .
the order of codebook generation of the auxiliary random variables is shown in fig .
[ fig : vkg-1 ] .
first , @xmath102 codewords of @xmath103 are generated using the marginal distribution of @xmath103 .
conditioned on each codeword of @xmath103 , @xmath104 codewords of @xmath105 are generated according to their respective conditional densities .
next , for each @xmath106 , a single codeword is generated for @xmath107 conditioned on @xmath108 @xmath109 .
note that to generate the codebook for @xmath37 , we first need the codebooks for all @xmath110 and @xmath103 . on observing a typical sequence @xmath8 , the encoder tries to find a jointly typical codeword tuple one from each codebook .
codeword index of @xmath105 ( at rate @xmath111 ) is sent in description @xmath46 . along with the ` _ _ private _ _ ' message , each description also carries a ` _ _ shared message _ _ ' at rate @xmath112 , which is the codeword index of @xmath103 . hence , the rate for each description is @xmath113 .
vkg showed that , to ensure finding a set of jointly typical codewords with the observed sequence , the rates must satisfy ( [ eq : vkg_rate ] ) .
it then follows from standard arguments ( see , e.g. , @xcite ) that , if the random variables also satisfy ( [ eq : vkg ] ) , then the distortion constraints are met .
note that @xmath103 is the _ only _ shared random variable . @xmath114
form the base layer random variables and all @xmath115 form the refinement layers .
observe that the codebook generation follows the order : shared layer @xmath116 base layer @xmath116 refinement layer .
also , observe that the random variable @xmath103 plays the role of @xmath86 in the zb scheme ) can be set to be a deterministic function of all the remaining random variables without affecting the overall achievable region ( for the 2-channel md case , this corresponds to setting @xmath64 to be a function of @xmath117 and @xmath63 ) .
a similar result can be proven even for the proposed coding scheme .
however , we continue to use this last layer in our theorems to avoid additional notation . ] .
in this section , we describe the proposed encoding scheme and derive the new achievable region . to simplify notation and understanding
, we first describe the scheme for the 3-descriptions case and offer intuitive arguments to show the achievability of the new region .
we then extend the arguments to the @xmath0channel case and provide formal proofs as part of theorem [ thm : main ] . ]
the encoding order for the 3-descriptions vkg scheme is shown in fig . [ fig:3_des](a ) .
recall that the common codeword helps in coordinating the 3 descriptions .
the vkg encoding scheme employs _
one _ common codeword ( @xmath103 ) that is sent in all the @xmath2 descriptions .
however , when dealing with @xmath3 descriptions , restricting to a single shared message could be suboptimal .
the cms scheme therefore allows for ` combinatorial message sharing ' , i.e. , a common codeword is sent in each ( non - empty ) subset of the descriptions .
the shared random variables are denoted by ` @xmath118 ' .
the base and the refinement layer random variables are denoted by ` @xmath119 ' .
for the 3 descriptions scenario , we have 11 auxiliary random variables which include 3 additional variables over the vkg scheme .
these variables are denoted by @xmath120 and @xmath121 .
the codeword corresponding to @xmath122 is sent in _ all _ the descriptions @xmath123 .
for example , the codeword of @xmath86 is sent in both the descriptions @xmath124 and @xmath125 .
this introduces a new layer in the encoding structure as shown in fig .
[ fig:3_des ] ( b ) .
this extra encoding layer provides an additional degree of freedom in controlling the redundancy across the messages , and hence leads to an improved rate - distortion region .
the codebook generation is done as follows .
first , the codebook for @xmath126 is generated containing @xmath127 independently generated codewords , each generated according to the marginal @xmath128 . then codebooks for @xmath120 and @xmath121 (
each containing @xmath85 , @xmath129 and @xmath130 codewords , respectively ) are independently generated conditioned on each codeword of @xmath126 , according to the respective conditional densities .
next , the base layer codebooks for @xmath105 , @xmath131 ( each containing @xmath104 codewords ) are generated conditioned on the codewords of all @xmath122 such that @xmath123 .
for example , the codebooks for @xmath62 are generated conditioned on the codewords of @xmath120 and @xmath126 .
note that each codebook of @xmath62 contains @xmath87 codewords and there are such @xmath132 codebooks .
the refinement layer codewords are generated similar to the vkg scheme .
however , the codewords for @xmath37 are now generated conditioned not only on the codewords of @xmath133 and @xmath103 , but also on the codewords of all @xmath134 such that @xmath135 .
for example , a single codeword for @xmath64 is generated conditioned on each codeword tuple of @xmath136 .
note that , overall , there are @xmath137 codewords of @xmath64 that are generated . similarly codewords for @xmath138 , @xmath139 and @xmath31 are generated conditioned on codewords of @xmath140 , @xmath141 and @xmath142 , respectively .
the encoder , on observing @xmath8 , tries to find a codeword from the codebook of @xmath126 that is jointly typical with @xmath8 . using typicality arguments , it is possible to show that the probability of not finding such a codeword approaches zero if @xmath143 .
let us denote the selected codeword of @xmath126 by @xmath144 .
the encoder next looks at the codebooks of @xmath86 , @xmath145 and @xmath121 , which were generated conditioned on @xmath144 , to find a triplet of codewords which are jointly typical with @xmath146 . it can be shown using arguments similar to @xcite ( formal proof is given in theorem 1 ) , that the probability of not finding such a triplet approaches zero if the following conditions are satisfied @xmath147 : @xmath148 let us denote these codewords by @xmath149 , @xmath150 and @xmath151 .
the encoders next step is to find an index tuple @xmath152 such that @xmath153 @xmath154 is jointly typical with @xmath155 . here , @xmath156 denotes the @xmath157th codeword from the codebook of @xmath62 and @xmath158 denotes the codeword of @xmath64 generated conditioned on @xmath159 .
similar notation is followed for @xmath160 , etc . again using arguments similar to @xcite
, we can show that the probability of not finding such an index tuple approaches zero if @xmath161 : @xmath162 where @xmath163 .
we denote the codewords corresponding to this index tuple by @xmath164 .
note that , in the above illustration , we assumed that the encoder finds jointly typical codewords in a sequential manner , i.e. , it first finds a codeword of @xmath126 , then finds jointly typical codewords from the codebooks of @xmath165 and so on .
this was done only for the ease of understanding . in theorem
[ thm : main ] , we derive the conditions on rates for the encoder to find typical sequences from all the codebooks jointly ( at once ) .
the conditions on the rates for joint encoding is generally weaker ( the region is larger ) than that for sequential encoding .
the encoder sends the index of the base layer codewords in the corresponding descriptions .
it also sends the index of codewords corresponding to @xmath122 in _ all _ the descriptions @xmath123 .
for example , in description 1 , the encoder sends the indices corresponding to @xmath166 and @xmath167 .
similarly , descriptions 2 and 3 carry indices corresponding to @xmath168 and @xmath169 , respectively .
therefore , rate for description @xmath46 is : @xmath170 conditions on @xmath171 can be obtained by substituting bounds from ( [ eq:3d_1 ] ) and ( [ eq:3d_2 ] ) .
the decoder , on receiving a subset of descriptions , estimates @xmath8 based on the refinement layer codeword corresponding to the received subset .
for example , if the decoder receives the descriptions 1 and 2 , it estimates @xmath8 as @xmath172 for some function @xmath72 .
it follows from standard arguments @xcite that , if there exist functions @xmath99 satisfying the following constraints , then distortion vector @xmath173 is achievable . @xmath174\label{eq:3d_4}\ ] ] an achievable rd region for the 3-descriptions md setup is obtained by taking the closure of the achievable tuples over all such 11 auxiliary random variables , jointly distributed with @xmath4 . )
indicates ` shared random variable ' .
red ( @xmath93 ) indicates ` base layer random variables ' and white ( @xmath175@xmath95 ) indicates ` refinement random variables ' .
[ fig : l_channel_cms ] ] the fundamental idea here is same as the 3-descriptions scenario , albeit involving more complex notation .
the codebook generation is done in an order as shown in fig .
[ fig : l_channel_cms ] .
first , the codebook for @xmath103 is generated .
then , the codebooks for @xmath122 , @xmath176 are generated in the order @xmath177 .
this is followed by the generation of the base layer codebooks , i.e , @xmath37 , @xmath178 .
then , the refinement layer codebooks corresponding to @xmath37 , @xmath176 are generated in the order @xmath179 .
each codebook is generated conditioned on a subset of the previously generated codewords .
the specifics of codebook generation will be described as part of the proof of theorem [ thm : main ] . before stating the theorem , we define the following subsets of @xmath180 : @xmath181 let @xmath182 be any non - empty subset of @xmath183 with @xmath184 . we define the following subsets of @xmath185 and @xmath186 : @xmath187
we also define : @xmath188 note that @xmath189 .
next we consider subsets of @xmath190 and define some important notation .
let @xmath191 be any subset of @xmath190 .
we denote by @xmath192_{1}$ ] the set of all elements of @xmath191 of cardinality
@xmath124 , i.e. , : @xmath193_{1}=\{\mathcal{k}:\mathcal{k}\in q,\,|\mathcal{k}|=1\}\ ] ] finally , we say that @xmath194 if it satisfies the following property .
let @xmath191 be a subset of @xmath195 .
if , for every set @xmath20 that belongs to @xmath191 , all the sets @xmath196 also belong to @xmath191 , then we say that @xmath194 , i.e. , @xmath194
if @xmath197 : @xmath198 let @xmath199 be any set of @xmath200 random variables jointly distributed with @xmath4 . let @xmath201 such that @xmath194 .
we define : @xmath202_{1}}h\left(v_{\mathcal{k}}|\{v\}_{\mathcal{i}_{|\mathcal{k}|+}(\mathcal{k})}\right)\nonumber \\ & & + \sum_{\mathcal{k}\in2^{[\mathcal{q}]_{1}}-\emptyset}h\left(u_{\mathcal{k}}|\{v\}_{\mathcal{i}_{1+}(\mathcal{k})},\{u\}_{2^{\mathcal{k}}-\emptyset-\mathcal{k}}\right)\nonumber \\ & & -h\left(\{v\}_{\mathcal{q}-[\mathcal{q}]_{1}},\{u\}_{2^{[\mathcal{q}]_{1}}-\emptyset}|x\right)\label{eq : alpha_defn_1 - 1}\end{aligned}\ ] ] we follow the convention @xmath203 .
next we state the rate - distortion region achievable by the cms scheme for the @xmath0descriptions framework .
[ thm : main]let @xmath199 be any set of @xmath200 random variables jointly distributed with @xmath4 , where @xmath37 and @xmath122 take values in some finite alphabets @xmath204 and @xmath205 , respectively @xmath206 .
let @xmath207 be the set of all subsets of @xmath190 satisfying ( [ eq : cond_q_main ] ) and let @xmath208 and @xmath209 be @xmath48 auxiliary rates satisfying : then , the rd region for the @xmath0channel md problem contains the rates and distortions for which there exist functions @xmath99 , such that , @xmath210\label{eq : dist_condition_thm}\end{aligned}\ ] ] the closure of the achievable tuples over all such @xmath200 random variables is denoted by @xmath211 .
@xmath211 can be extended to continuous random variables and well - defined distortion measures using techniques similar to @xcite .
we omit the details here and assume that the above region continues to hold even for well behaved continuous random variables ( for example , a gaussian source under mse ) .
@xmath211 is convex , as a time sharing random variable can be embedded in @xmath103 .
_ codebook generation _ : suppose we are given @xmath212 and @xmath99 satisfying ( [ eq : dist_condition_thm ] ) .
the codebook generation begins with @xmath103 .
we independently generate @xmath102 codewords of @xmath103 , denoted by @xmath213 , according to the density @xmath214 . for each codeword @xmath215
, we independently generate @xmath216 codewords of @xmath122 @xmath217 , according to @xmath218 .
we denote these codewords by @xmath219 @xmath220 .
this procedure for generating the codebooks of the shared random variables continues .
@xmath216 codewords of @xmath122 are independently generated for each codeword tuple of @xmath221 according to @xmath222 .
these codewords are denoted by @xmath223 .
note that to generate the codebooks for @xmath122 @xmath224 , we need the codebooks of @xmath225 . the codebook generation follows the order indicated in fig .
[ fig : l_channel_cms ] .
once all the codebooks of shared random variables are generated , the codebooks for the base layer random variables are generated .
for each codeword tuple of @xmath226 , @xmath104 codewords of @xmath105 are generated independently according to @xmath227 and are denoted by @xmath228 @xmath229 .
then the codebooks for the refinement layers are formed by assigning a single codeword @xmath230 to each @xmath231 and @xmath232 .
these codewords are generated according to @xmath233 .
the encoder , on observing a typical sequence @xmath8 , attempts to find a set of codewords , one for each variable , such that they are all jointly typical .
if the encoder succeeds in finding such a set , from the typical average lemma @xcite , it follows that the average distortions are smaller than @xmath234 . however , if the encoder fails to find such a set , the average distortions are upper bounded by @xmath235 ( as the distortion measures are assumed to be bounded ) .
hence , if the probability of finding a set of jointly typical codewords approaches 1 , the distortion conditions ( [ eq : dist_condition_thm ] ) are satisfied .
we show in appendix a that if the rates @xmath236 and @xmath111 satisfy conditions ( [ eq : aux_rate_cond_thm ] ) , this probability approaches 1 .
next , recall that the encoder sends the codewords of @xmath122 ( at rate @xmath236 ) in all the descriptions @xmath123
. it also sends the codewords of @xmath105 ( at rate @xmath111 ) in description @xmath46 .
therefore the rate of description @xmath46 is given by ( [ eq : rate_condition_thm ] ) , proving the theorem .
we note that the characterization of the achievable rd region in theorem [ thm : main ] involves auxiliary rates @xmath111 and @xmath236 .
this makes it hard to prove general converse results .
however deriving an explicit characterization only in terms of @xmath171 is more involved and deriving such bounds will be considered as part of future work .
in this section we show that the cms scheme achieves points strictly outside @xmath96 whenever there exists a 2-description subset for which the zb region achieves points outside the ec region . as a result of this theorem
, it follows that for a binary symmetric source under hamming distortion measure , cms strictly outperforms vkg .
[ thm : zb_imp](i ) the rate - distortion region achievable by the cms scheme is always at least as large as the vkg achievable region , i.e. : @xmath237 ( ii ) @xmath238 , the cms scheme achieves points that are not achievable by vkg for any source and distortion measures for which there exists a 2-description subset such that the zhang - berger scheme achieves points outside the el - gamal - cover region .
i.e. : @xmath239 for some 2-description subset . specifically , for a binary symmetric source under hamming distortion , cms achieves a strictly larger rate - distortion region compared to vkg @xmath238 .
in section [ sec : strict - improvement2 ] we will show a more surprising result that cms achieve points strictly outside @xmath96 even in scenarios where @xmath240 for every 2-description subset .
moreover , we will show that cms outperforms vkg for a gaussian source under mse , a setting for which it is well known that the ec and zb regions coincide .
part ( i ) of the theorem is a straightforward corollary of theorem [ thm : main ] .
it follows directly by setting @xmath241 , @xmath206 such that @xmath242 in @xmath211 ( where @xmath18 denotes a constant ) . substituting in ( [ eq : rate_condition_thm ] ) , and setting the auxiliary rate @xmath243 , we get the same conditions as ( [ eq : vkg ] ) . part ( ii ) of the theorem , in fact , follows directly from zhang and berger s result . to see this , without loss of generality ,
let descriptions @xmath23 be the 2-description subset for which zb achieves points outside ec region .
let us consider a cross - section of the @xmath0channel md problem where we only consider @xmath79 , @xmath80 , @xmath244 , @xmath245 and @xmath246 and set the remaining rates to @xmath247 while ignoring all the remaining distortion constraints .
clearly , the cms scheme leads to the zb region as the common layer codeword @xmath86 is sent as part of both descriptions 1 and 2 .
however , observe that , vkg leads to ec as @xmath103 must be set to constant to ensure @xmath248 .
although this example is sufficient to prove part ( ii ) of the theorem , it is arguable that this particular cross - section of the @xmath0channel md setup is degenerate and hence this proof has little value in practice .
we therefore provide a more general proof where all the descriptions carry non - trivial information and all rates @xmath249 are greater than @xmath247 . we prove ( ii ) for @xmath250 . note that once we prove that the cms scheme achieves a strictly larger region for some @xmath251 , then it must be true for all @xmath252 .
consider a 3-descriptions md problem where descriptions @xmath23 form the 2-description subset for which zb achieves points outside ec region .
we denote the vkg and cms achievable regions by @xmath253 and @xmath254 , respectively .
we now consider a particular cross - section of these regions where we only constrain @xmath255 and @xmath256 .
we remove the constraints on all other distortions , i.e. , we allow @xmath257 and @xmath258 to be @xmath259
. equivalently , we can think of a 3 descriptions md problem with a particular channel failure pattern , wherein only one of the following sets of descriptions can reach the decoder reliably : @xmath260 as shown in fig .
[ fig:3_des-1 ] .
we denote the set of all achievable points for this setup using vkg and cms by @xmath261 and @xmath262 , respectively .
( a ) equivalent model for the cross - section considered to prove that the cms scheme achieves strict improvement over the vkg scheme for a binary symmetric source under hamming distortion measure . ] observe that with respect to the first 2 descriptions , we have a simple 2-descriptions problem , while descriptions @xmath263 represent a successive refinement setup . we know from the results of zhang and berger that a common codeword sent among the first two descriptions leads to a strictly better rd trade - off for these two descriptions .
however , the vkg coding scheme can not use this common codeword unless it is also sent as part of the third description , which is redundant for the successive refinement framework involving descriptions @xmath263 , since description @xmath264 is never received without description @xmath124 . on the other hand , cms allows for a unique common codeword to be sent within every subset of the transmitted descriptions .
specifically , it can send a common codeword which is only shared by the first two descriptions , achieving the zb advantage without adding redundancy to description @xmath264 .
this argument is in fact sufficient to prove the claim .
we rewrite it formally below .
consider a point @xmath265 in the zb region that is not in the ec region .
let the rate of the common codeword that is sent as part of both the descriptions at @xmath266 be denoted by @xmath267 .
denote by @xmath268 the following cross - section of @xmath261 : @xmath269 and for the corresponding cross - section of @xmath262 , the infimum @xmath270 rate by @xmath271 .
as @xmath266 is achievable by zb and not by ec , the constraint @xmath272 forces a common codeword to be sent in descriptions 1 and 2 . to use this common codeword
, vkg requires it to be sent as part of description 3 as well leading to decoder @xmath263 receiving the same codeword twice .
however , cms sends this common codeword only as part of descriptions 1 and 2 leading to a smaller rate for the third description .
hence it follows that @xmath273 is smaller than @xmath274 by at least the rate of this common codeword which is equal to @xmath267 .
it can be verified from @xcite that for a binary symmetric source under hamming distortion measure : @xmath275 where @xmath276 denotes the binary entropy function .
hence , we have shown that the cms region is strictly larger than vkg region whenever there exists a 2-descriptions subset for which zb outperforms ec . moreover ,
the gap between cms and vkg regions is at least as large as the gap between the ec and the zb regions for the corresponding 2-descriptions subset .
it is clear from theorem [ thm : zb_imp ] that for any given distribution for @xmath4 and distortion measures , if there exists a 2-description subset such that @xmath277 , then cms strictly outperforms vkg . in this section we show a more surprising result that the common layer codewords in cms play a critical role in achieving a strictly larger region for a fairly general ( and larger ) class of source distributions and distortion measures .
what makes it particularly interesting is the fact that under mse , a gaussian source ( @xmath278 ) belongs to this class , i.e. , we will show that cms achieves strictly larger rd region compared to vkg for the @xmath0channel quadratic gaussian md problem @xmath279 .
this result is in striking contrast to the corresponding results for the 2-descriptions setting .
it follows from ozarow s results in @xcite that for a gaussian source under mse , the ec coding scheme achieves the complete rd region , i.e. , sending a common codeword among the two descriptions does not provide any improvement in the rd region ( @xmath280 ) .
this result has lead to a natural belief that common codewords do not play a role in the @xmath0channel quadratic gaussian md problem and all the explicit characterizations for the achievable regions in the past make no use of common random variables ( see for example , @xcite ) .
surprisingly , in the @xmath0descriptions framework , the common codewords in cms can be used constructively for better coordination between the descriptions leading to achievable points outside the vkg region .
we begin by defining a set of random variables , denoted by @xmath281 , that plays a critical role in the subsequent theorems .
let us consider a two descriptions md setup .
for any given distortion measures , @xmath282 , we say that @xmath283 , if there exists an operating point @xmath61 that belongs to @xmath81 , but _ can not _ be achieved by an ` independent quantization ' mechanism using the zb coding scheme .
a formal definition of @xmath281 is given below .
* [ definition : defn_zzb]@xmath281 * : let @xmath284 .
let us denote @xmath285 to be the rate - distortion region achievable by the zb coding scheme when the closure in ( [ eq : zb ] ) if defined only over joint distributions for the auxiliary random variables satisfying the following conditions refers to ` independent quantization ' ] : @xmath286 & \leq & d_{\mathcal{k}},\,\,\mathcal{k}\in\{1,2,12\}\nonumber \\ i(u_{12};x|u_{1},u_{2},v_{12 } ) & < & \epsilon\label{eq : zzb-1}\end{aligned}\ ] ] we say that @xmath283 , if : @xmath287 i.e. , if @xmath288 is strictly subsumed in @xmath81 .
note that if @xmath289 , then clearly @xmath283 , i.e. , if there is a strict suboptimality in the zb region when the closure in ( [ eq : zb ] ) is defined only over joint distributions for the auxiliary random variables satisfying : @xmath290 & \leq & d_{\mathcal{k}},\,\,\mathcal{k}\in\{1,2,12\}\nonumber \\ i(u_{12};x|u_{1},u_{2},v_{12 } ) & = & 0\label{eq : zzb-1 - 1}\end{aligned}\ ] ] then @xmath283 .
observe that the last constraint @xmath291 can be replaced with @xmath292 , without any loss of optimality .
this follows from the fact that for any given joint distribution @xmath293 that satisfies all the conditions in ( [ eq : zzb-1 - 1 ] ) , it is possible to construct a conditional distribution as follows : @xmath294\\ 0 & \mbox{otherwise } \end{cases } & \end{aligned}\ ] ] the joint distribution @xmath295 achieves the smallest distortion for @xmath246 .
therefore , it follows that the constraint @xmath291 can be replaced with @xmath292 , without any loss of optimality .
this simplified definition of @xmath281 will be used in the proof of theorem [ thm : general_cms-1 ] .
we will show in theorem [ thm : general_cms-2 ] that @xmath296 , @xmath297 .
we will later provide more intuition on the underlying relation between @xmath298 and the reason for strict improvement of cms over vkg .
hereafter , we will often drop the parenthesis and abbreviate @xmath298 by @xmath281 , whenever the distortion measures are obvious . ] before stating the result we describe the particular cross - section of the rd region that we will use to prove strict improvement in theorem [ thm : general_cms-2 ] .
consider a 3-descriptions md setup for a source @xmath4 wherein we impose constraints only on distortions @xmath299 and set the limit on the rest of the distortions , @xmath300 to @xmath259 .
this cross - section is schematically shown in fig .
[ fig:3_des_new-1 ] . to illustrate the gains underlying cms ,
let us further restrict to the setting where we impose @xmath301 and @xmath302 .
we also assume that the distortion measures @xmath303 and @xmath304 are same as the distortion measures @xmath305 and @xmath306 , respectively .
the points in this cross - section , achievable by vkg and cms , are denoted by @xmath307 and @xmath308 , respectively .
we note that the symmetric setting is considered _ only _ for ease of understanding the proof .
the arguments can be easily extended to the asymmetric framework .
this particular symmetric cross - section of the 3-descriptions md problem is equivalent to the corresponding 2-descriptions problem , in the sense that , one could use any coding scheme to generate bit - streams for descriptions 1 and 2 , respectively .
description 3 would then carry a replica ( exact copy ) of the bits sent in description 1 . due to the underlying symmetry in the problem setup , the distortion constraints at all the decoders are satisfied .
hence an achievable region based on the zb coding scheme can be derived as follows .
let @xmath309 be any random variables jointly distributed with @xmath4 and taking values over arbitrary finite alphabets .
then the following rd - region is achievable for which there exist functions @xmath310 such that @xmath311 , @xmath301 , @xmath302 and : @xmath312,\,\,\mathcal{k}\subseteq\{1,2\},\mathcal{k}\neq\emptyset\nonumber\label{eq : rd(x)-1}\end{aligned}\ ] ] the closure of achievable rd - tuples over all random variables @xmath309 is denoted by @xmath313 . in the following theorem
, we will show that @xmath314 .
we also show that the vkg coding scheme _ can not _ achieve the above rd region , i.e. , @xmath315 , if @xmath283 . throughout proof of theorem [ thm : general_cms-2 ] , we use the notations @xmath316 , @xmath317 and @xmath318 to denote the rate - region cross - sections of @xmath307 , @xmath308 and @xmath313 at distortions @xmath319 , respectively .
we note that in theorem [ thm : general_cms-2 ] , we focus only on the 3-descriptions setting . however , the results can be easily extended to the general @xmath0descriptions scenario .
also note that @xmath308 could be strictly larger than @xmath313 , in general .
[ thm : general_cms-2](i ) for the setup shown in fig .
[ fig:3_des_new-1 ] the cms scheme achieves @xmath313 , i.e. , @xmath314 .
\(ii ) if @xmath283,then there exists points in @xmath313 that * can not * be achieved by the vkg encoding scheme , i.e. , , it directly follows from ( i ) and ( ii ) that @xmath297 for the @xmath0channel md problem @xmath279 , if @xmath320 . _
_ _ intuition _
: we first provide an intuitive argument to justify the claim . due to the underlying symmetry in the setup
the cms scheme introduces common layer random variables @xmath321 and @xmath322 .
it then sends the codeword of @xmath145 is both descriptions 1 and 3 ( i.e. , @xmath323 ) .
hence it is sufficient for the encoder to generate enough codewords of @xmath324 ( conditioned on @xmath126 ) to maintain joint typicality with the codewords of @xmath145 .
however , the vkg scheme is forced to set the common layer random variable @xmath145 to a constant .
thus , in this case , the encoder needs to generate enough number of codewords of @xmath63 ( and @xmath64 ) so as to maintain joint typicality individually with the codewords of @xmath62 and @xmath325 , which are now generated independently conditioned on @xmath126 .
it is possible to show that this entails some excess rate on @xmath63 , unless @xmath326 are pairwise independent conditioned on @xmath126 .
however , if @xmath320 , then enforcing independence of @xmath62 and @xmath63 conditioned on @xmath126 leads to a strictly smaller rate - distortion region .
therefore , @xmath327 , the vkg scheme leads to a strictly smaller rate - distortion region compared to the cms scheme .
part ( i ) of the theorem is straightforward to prove .
we set , @xmath328 and the rest of the random variables in theorem [ thm : main ] to constants .
it is easy to verify that the rate - distortion region given in ( [ eq : rd(x)-1 ] ) is achievable by the cms scheme using the above joint distribution for the auxiliary random variables and by choosing the auxiliary rate @xmath329 .
next , let @xmath283 with respect to the given distortion measures , i.e. , there exists a strict suboptimality in the zb region when the closure of rates is defined only over joint densities for the auxiliary random variables satisfying ( [ eq : zzb-1 ] ) . for ease of understanding the proof
, we make a simplifying assumption that there exists at least one ` corner point ' in the zb region that is not achievable using joint densities satisfying ( [ eq : zzb-1 ] ) .
specifically , let the distortions be @xmath330 , respectively , at the three decoders .
we assume that : @xmath331 i.e. , if @xmath332 , then the infimum over all @xmath80 required to achieve @xmath330 , using joint densities satisfying ( [ eq : zzb-1 ] ) , is strictly larger than the infimum over all @xmath80 achievable using zb .
however , it is important to note that the proof can be easily extended to the general case of strict suboptimality occurring at any intermediate point . we will briefly discuss the extension towards the end of the proof . towards proving ( ii )
, we consider one particular boundary point of ( [ eq : rd(x)-1 ] ) and show that this can not be achieved by the vkg encoding scheme .
let @xmath333 and @xmath334 and @xmath246 be fixed .
define the following quantity : @xmath335 the corresponding quantity defined for the cms region is given by : @xmath336 we will show that : @xmath337 similarly , we use the notation @xmath338 to denote the same quantity defined over @xmath318 , i.e. , @xmath339 from cms region characterization in theorem [ thm : main ] , @xmath340 is given by the solution to the following optimization problem : @xmath341 where the infimum is over all joint distributions @xmath342 , and conditional distributions @xmath343 for which there exists a function @xmath344 such that : @xmath345 & \leq & d_{1}\label{eq : cms_13_constraints}\end{aligned}\ ] ] i.e. , @xmath346 achieves a close - to - optimal reconstruction of @xmath4 at @xmath244 and @xmath347 is any distribution for which there exists functions @xmath348 and @xmath72 satisfying the distortion constraints for @xmath245 and @xmath246 , respectively .
we next specialize and restate @xmath307 for the considered cross - section .
let @xmath349 be any random variables jointly distributed with @xmath4 taking values on arbitrary alphabets .
then , the following rate - distortion tuples are achievable by the vkg scheme for which there exist functions @xmath350 , such that : @xmath351 @xmath352 the closure of the above rate - distortion region over all joint distributions leads to @xmath353 . observe that there are no distortion constraints imposed on @xmath256 and @xmath258 .
this allows us to simplify the region further , without any loss of optimality .
first , the random variables @xmath138 and @xmath31 can be set to constants .
this is because they do not appear in any of the distortion constraints and setting them to constants leads to a larger rate - region for the given distribution over all other random variables .
this step simplifies the constraints on @xmath354 and @xmath355 to : @xmath356 as a next step of simplification , we restrict the closure of the region to be taken only over joint densities , @xmath357 , that satisfy the following constraint : @xmath358 we note that this restriction does not lead to any loss in @xmath307 for this particular cross - section .
this is because , for any given joint distribution @xmath359 , we can construct another joint distribution that satisfies ( [ eq : u_1223_mark ] ) and leads to a larger rate - distortion region . to see this , consider the joint distribution constructed as : @xmath360 which satisfies ( [ eq : u_1223_mark ] ) .
observe that this joint distribution satisfies the same distortion constraints as @xmath361 .
moreover , it leads to the same rate constraints as @xmath361 , except for the constraint on @xmath355 .
however , the constraint it imposes on @xmath355 is weaker than that imposed by @xmath361 and hence the rate - distortion region is larger than that achievable by @xmath361 .
therefore , it is sufficient to consider only those joint distributions that satisfy ( [ eq : u_1223_mark ] ) for @xmath307 .
we next impose the constraints @xmath362 and @xmath363 in ( [ eq : simplification_1 ] ) .
this enforces the conditional density @xmath364 to satisfy the following constraints : @xmath365 & \leq & d_{1}\nonumber \\
e\left[d_{1}(x,\psi_{3}(v_{123},u_{3}))\right ] & \leq & d_{1}\nonumber \\
i(u_{1};u_{3}|x , v_{123 } ) & < & \epsilon\label{eq : vkg_13_constraints}\end{aligned}\ ] ] where the last condition is required to satisfy the constraint on @xmath354 in ( [ eq : simplification_1 ] ) . therefore , using ( [ eq : simplification_1 ] ) and ( [ eq : u_1223_mark ] ) , @xmath366 can be written as the solution to the following optimization problem : @xmath367 where the infimum is over all conditional densities @xmath357 satisfying ( [ eq : vkg_13_constraints ] ) for which there exist functions @xmath368 satisfying the distortion constraints in ( [ eq : rd_vkg_dist ] ) . summarizing what we have so far : @xmath369 where the infimum for @xmath366 is subject to ( [ eq : vkg_13_constraints ] ) and the infimum for @xmath340 is subject to ( [ eq : cms_13_constraints ] ) .
we first note that the optimization objective for @xmath366 is always greater than or equal to the optimization objective for @xmath340 , for a given joint distribution over the random variables .
this follows from the following standard information theoretic inequalities : @xmath370 moreover , the constraints for @xmath340 are a subset of the constraints for @xmath366 . therefore , for @xmath371 to be equal to @xmath372 , there should exist a small enough @xmath333 for every @xmath373 , such that : @xmath374 we will next show that this can not happen if @xmath283 .
i.e. , we will show that @xmath375 is lower bounded by a value that is strictly greater than @xmath247 .
we prove this claim by contradiction .
we assume that it is possible to find a small enough @xmath333 for every @xmath373 such that ( [ eq : eps_delta_constraint ] ) holds and then arrive at a contradiction .
let @xmath376 .
let @xmath357 be any conditional distribution that satisfies the constraints in ( [ eq : vkg_13_constraints ] ) and satisfies : @xmath377 where the superscript @xmath378 has been added to emphasize that the mutual information is with respect to the distribution @xmath378 .
the distribution @xmath378 achieves the optimization objective in ( [ eq : rvkg_cms_1 ] ) , within @xmath379 of @xmath366 . for ( [ eq : eps_delta_constraint ] ) to hold
, the following condition must be satisfied by all conditional distributions @xmath380 that satisfy the constraints in ( [ eq : cms_13_constraints ] ) : @xmath381 we will next show that this leads to a contradiction if @xmath283 . towards proving that , we set @xmath382 .
observe that @xmath378 is a valid candidate for @xmath383 as the constraints in ( [ eq : cms_13_constraints ] ) are a subset of ( [ eq : vkg_13_constraints ] ) . with @xmath384 , for ( [ eq : more_eps_delta ] ) to hold , we need : @xmath385 first , observe that the constraint @xmath386 implies that @xmath387 . however , as @xmath378 satisfies ( [ eq : vkg_13_constraints ] ) , we have @xmath388 . on substituting ,
we get : @xmath389 i.e. , @xmath390 which implies that : @xmath391 hence , from ( [ eq : contra_cond ] ) and ( [ eq : contra_cond2 - 1 - 1 ] ) , for ( [ eq : eps_delta_constraint ] ) to hold , we need the following constraints to be satisfied for the joint distribution @xmath378 that achieves close to optimality in ( [ eq : rvkg_cms_1 ] ) : @xmath392 & \leq & d_{\mathcal{k}},\,\,\mathcal{k}\in\{2,3,23\}\end{aligned}\ ] ] under the limits of @xmath393 , @xmath394 and @xmath395 , if @xmath283 , the above constraints imply that there is a strict sub - optimality in the zb region . from our assumption in ( [ eq : corner_point_assumption ] ) , it immediately follows that : @xmath396 where @xmath338 is defined in ( [ eq : thm3_2 - 1 - 1 ] ) .
however , from ( i ) of the theorem , it follows that : @xmath397 this leads to a contradiction as it implies that ( [ eq : eps_delta_constraint ] ) is not satisfied .
it therefore follows that if @xmath283 , cms achieves a strictly larger rd region compared to vkg , proving the theorem .
we note that if the strict sub - optimality in ( [ eq : corner_point_assumption ] ) exists at any other boundary point of the zb region , the above proof holds by changing the definitions of @xmath366 , @xmath340 and @xmath338 accordingly .
we only consider this particular corner point in this proof for ease of understanding .
* discussion : * a direct consequence of the above theorem is that , if @xmath320 , then the common layer codewords of cms are needed to achieve strict improvement in the region , i.e. , if @xmath320 , @xmath398 , where @xmath399 denotes the vkg region when the common layer random variable ( denoted by @xmath103 ) is set to a constant . to a constant in vkg is equivalent to setting all the common layer random variables to constants in cms .
] in fact , it is possible to show that , whenever @xmath400 , @xmath401 , where @xmath402 is similarly defined as the set of all random variables @xmath4 for which there exists an operating point , with respect to the given distortion measures , that _ can not _ be achieved by an ` independent quantization ' mechanism using the ec coding scheme , i.e. , if there exists an operating point in the ec region that _ can not _ be achieved by the ec coding scheme using a joint density for the auxiliary random variables satisfying : @xmath403 & \leq & d_{\mathcal{k}}\,\,\,\mathcal{k}\in\{1,2,12\}\nonumber \\
u_{12 } & = & f(u_{1},u_{2})\label{eq : zec}\end{aligned}\ ] ] where @xmath404 is any deterministic function would be in lines of the definition of @xmath281 in definition [ definition : defn_zzb ] .
we state the simpler version here for brevity . ] .
note that the set @xmath281 is a subset of @xmath402 .
it is possible to construct distributions and distortion measures such that @xmath281 is a strict subset of @xmath402 , but as the construction does not provide any useful insights into the md problem , we choose to omit the details . also observe that if @xmath405 , the concatenation of two independent optimal quantizers is optimal in achieving a joint reconstruction . while this condition could be satisfied for specific values of @xmath334 and @xmath246 , it is seldom achieved _ for all _ values of @xmath330 .
though such sources are of some theoretical interest , the multiple descriptions encoding for such sources is degenerate . hence with some trivial exceptions , it can be asserted that the common layer codewords in cms can be used to achieve a strictly larger region ( compared to not using any common codewords ) for all sources and distortion measures , @xmath279 .
we note in passing that , for the two descriptions setting , zhang and berger studied an important cross - section of the problem in @xcite called the `` no excess marginal rate '' setting , where @xmath332 and @xmath406 .
they derived upper and lower bounds on the achievable @xmath246 and showed that the gap is negligible for a binary source under hamming distortion measure .
the constraints @xmath407 imposed in the proof of theorem [ thm : general_cms-2 ] resemble the constraints imposed in @xcite and hence the results in @xcite may seem relevant to the setting considered in this paper . however , it is important to note that in the proof of theorem [ thm : general_cms-2 ] , @xmath80 can be greater than @xmath408 . as the interaction is only between descriptions @xmath23 and @xmath24 ,
the problem considered here is not directly related to the ` no excess marginal rate ' case and requires new tools to prove the results . at a first glance ,
it is tempting to conclude from ozarow s results in @xcite that under mse a gaussian random variable belongs to both @xmath402 and @xmath281 .
however , the formal proof is non - trivial
. we will formally show in the next section that this is indeed the case and an independent quantization scheme ( with or without a common codeword ) leads to strict suboptimality for the 2-descriptions quadratic gaussian md problem .
in this section we present a series of new results for the @xmath0descriptions quadratic gaussian md problem .
we will first show that cms achieves a strictly larger rd region , by proving that under mse , a gaussian source belongs to @xmath281 .
we then use similar encoding principles to derive the complete rate region in several asymmetric distortion regimes . throughout this section
, we will assume that @xmath278 and the distortion at all the decoders is the squared error , i.e. , @xmath409 . before stating the results formally , we review ozarow s result for the 2-descriptions md setting .
ozarow showed that the complete region for the 2-descriptions gaussian md problem can be achieved using a ` correlated quantization ' scheme which imposes the following joint distribution for @xmath410 in the ec scheme : @xmath411 @xmath412 , where @xmath413 and @xmath414 are zero mean gaussian random variables independent of @xmath4 with covariance matrix @xmath415 , and the functions @xmath416 are given by the respective mse optimal estimators , e.g. , @xmath417 $ ] .
the covariance matrix @xmath415 is set to satisfy all the distortion constraints .
specifically , the optimal @xmath415 is given by : @xmath418\label{eq : ozarow_k-1}\ ] ] where @xmath419 and the optimal @xmath420 , denoted by @xmath421 , is given by ( see @xcite ) : @xmath422\nonumber \\ d_{12}^{max } & = & d_{1}d_{2}/(d_{1}+d_{2}-d_{1}d_{2})\nonumber \\ \pi & = & ( 1-d_{1})(1-d_{2})\label{eq : other_defn-1}\end{aligned}\ ] ] we denote the complete gaussian - mse @xmath2-descriptions region by @xmath423 .
the characterization of @xmath424 is given in @xcite ( see also @xcite ) and we omit restating it explicitly here for brevity . equipped with these results , we next show that cms achieves points outside the vkg region , by proving that a gaussian source under mse belongs to @xmath281 .
[ thm : general_cms-1](i ) cms achieves the * complete * rd region for the symmetric 3-descriptions quadratic gaussian setup shown in fig . [ fig:3_des_new-1 ] .
@xmath425 where @xmath426 is defined by : @xmath427 where @xmath421 is defined in ( [ eq : other_defn-1 ] ) .
\(ii ) the vkg encoding scheme can not achieve all the points in the region , i.e. , it follows from ( i ) and ( ii ) that @xmath297 for the @xmath0channel quadratic gaussian md problem @xmath238 .
to prove achievability using cms , we set , @xmath428\label{eq : cor1_1}\end{aligned}\ ] ] where @xmath413 and @xmath414 are zero mean gaussian random variables independent of @xmath4 with covariance matrix @xmath415 and the functions @xmath416 are given by the respective mse optimal estimators , e.g. , @xmath417 $ ] .
we set all the remaining auxiliary random variables to constants .
it follows directly that the rate - distortion region given in ( [ eq : symm_rd ] ) is achievable by the cms scheme . following the footsteps of ozarow in @xcite , it is straightforward to show that the above region is also complete for the symmetric setup considered .
note that ozarow s results suggest that if @xmath429 , then an ` independent quantization ' scheme does not achieve the smallest @xmath246 .
it might be tempting to conclude that ( ii ) follows directly from this observation and theorem [ thm : general_cms-2 ] .
however , a closer inspection reveals that the above argument holds only after we prove the optimality of gaussian codebooks under ` independent quantization ' mechanism .
we relegate the proof of ( ii ) to appendix b as the underlying principles are quite orthogonal to the rest of the paper .
note that , as @xmath430 , a gaussian source under mse belongs to @xmath402 .
hence , the ` correlated quantization ' scheme ( an extreme special case of vkg ) which has been proven to be complete for several cross - sections of the @xmath0descriptions quadratic gaussian md problem @xcite , is strictly suboptimal in general .
in this section we show that cms achieves the complete rd region for several cross - sections of the general quadratic gaussian @xmath0channel md problem .
we again begin with the 3-descriptions case and then extend the results to the @xmath2 channel framework . recall the setup shown in fig .
[ fig:3_des-1 ] , i.e , a cross - section of the general 3-descriptions rate - distortion region wherein we impose constraints only on distortions @xmath299 and set the rest of the distortions , @xmath300 to @xmath124 . here
we consider the general asymmetric case , i.e. @xmath431 and @xmath432 and show that the cms scheme achieves the complete rate region in several distortion regimes . in the following theorem , without loss of generality we assume that @xmath433 .
if @xmath434 , then the theorem holds by interchanging ` @xmath124 ' and ` @xmath264 ' everywhere .
let @xmath246 be any distortion such that @xmath435 .
we define @xmath436 as : @xmath437 where @xmath438 @xmath439 and @xmath440 where @xmath421 is defined in ( [ eq : other_defn-1 ] ) . in the following theorem , we will show that cms achieves the complete rate - region if @xmath441 .
[ thm : sum_rate_gauss]for the setup shown in fig .
[ fig:3_des-1 ] , let @xmath433 .
then , \(i ) cms achieves the complete rate - region if : @xmath442 where @xmath443 is defined in ( [ eq : defn_d23 ] ) .
the rate region is given by : @xmath444 where @xmath445 is defined in ( [ eq : delta_defn ] ) .
\(ii ) moreover , cms achieves the minimum sum - rate if one of the following hold : \(a ) for a fixed @xmath246 , @xmath446 \(b ) for a fixed @xmath447 , @xmath448 we note that the above rate region _ can not _ be achieved by vkg .
we omit the details of the proof here as it can be proved in same lines as the proof of theorem [ thm : general_cms-1 ] .
an achievable rate - distortion region can be derived for general distortions using the encoding principles we derive as part of this proof .
however , it is hard to prove outer bounds if the conditions in ( [ eq : sum_rate_thm1 ] ) are not satisfied and hence we omit stating the results explicitly here .
both cms and vkg achieve the complete rate region when @xmath449 and @xmath450 , where @xmath451 and @xmath452 are defined in ( [ eq : other_defn-1 ] ) .
it can be easily verified that in this case an independent quantization scheme is optimal and the complete achievable rate - region is given by @xmath453 .
it follows from the above theorem that cms achieves the minimum sum - rate whenever @xmath302 for any @xmath454 .
we begin with the proof of ( i ) .
the proof of ( ii ) then follows almost directly from ( i ) .
first we show the converse , which is quite obvious .
conditions on @xmath455 follow from the converse to the source coding theorem .
conditions on @xmath456 and @xmath457 follow from ozarow s results in @xcite , to achieve @xmath330 using descriptions @xmath23 and to achieve @xmath458 using descriptions @xmath24 at the respective decoders .
we next prove that cms achieves the rate region in ( [ eq : sum_rate_thm2 ] ) if ( [ eq : sum_rate_thm1 ] ) holds .
we first give an intuitive argument to explain the encoding scheme .
description 3 carries an rd - optimal quantized version of @xmath4 ( which achieves distortion @xmath459 ) .
description 1 carries all the bits embedded in description 3 along with ` refinement bits ' which assist in achieving distortion @xmath433 .
this entails no loss in optimality as a gaussian source is successively refinable under mse @xcite .
description 2 then carries a quantized version of the source which is correlated with the information in descriptions 1 and 3 .
we will show that if @xmath460 , then the correlations can be set such that description 2 is optimal with respect to both descriptions 1 and 3 .
formally , to achieve the rate region in ( [ eq : sum_rate_thm2 ] ) , we set the auxiliary random variables in the cms coding scheme as follows : @xmath461 and the functions @xmath462 as the respective mse optimal estimators , where @xmath463 are zero mean gaussian random variables independent of @xmath4 with a covariance matrix : @xmath464\label{eq : sum_rate_pf_4}\ ] ] where @xmath465 , @xmath466 , @xmath467 .
the correlation coefficient @xmath420 is set to achieve distortion @xmath246 , i.e. @xmath468 defined in ( [ eq : other_defn-1 ] ) .
let us denote by @xmath469 .
observe that the encoding for descriptions 2 and 3 resembles ozarow s correlated quantization scheme with @xmath470 and @xmath471 .
let us denote the correlation coefficient between @xmath414 and @xmath472 be @xmath473 .
we have the following equation relating @xmath420 and @xmath473 ( which is equivalent to ( [ eq : defn_rho ] ) ) : @xmath474 note that the above relation is derived using the independence of @xmath475 and @xmath476 , which follows from our choice of @xmath477 .
hence the minimum distortion @xmath447 achievable using the above choice for the joint density of the auxiliary random variables is given by : @xmath478 we next derive the rates required by this choice of @xmath477 .
application of theorem [ thm : main ] using the above joint density leads to the following achievable rate region for any given distortions @xmath479 : @xmath480 @xmath481 @xmath482 where @xmath483 follows from the markov chain @xmath484 , @xmath485 from the independence of @xmath4 and @xmath486 and @xmath487 from the independence of @xmath475 and @xmath476 . at a first glance , it might be tempting to conclude that the region for the tuple @xmath488 in ( [ eq : cms_gauss_ach_region ] ) is equivalent to the region given by ( [ eq : sum_rate_thm2 ] ) .
this is not the case in general as the equations in ( [ eq : cms_gauss_ach_region ] ) have an implicit constraint on the auxiliary rates @xmath489 .
however , we will show that if @xmath490 , then the two regions are indeed equivalent .
we denote the rate region given in ( [ eq : sum_rate_thm2 ] ) by @xmath491 and the region in ( [ eq : cms_gauss_ach_region ] ) by @xmath492 . clearly , @xmath493 , as any @xmath488 that satisfies ( [ eq : cms_gauss_ach_region ] ) also satisfies ( [ eq : sum_rate_thm2 ] ) .
we need to show that @xmath494 . towards proving this claim , note that both @xmath491 and @xmath492 are convex regions bounded by hyper - planes .
hence , it is sufficient for us to show that all the corner points of @xmath491 lie in @xmath492 .
clearly , @xmath491 has 6 corner points denoted by @xmath495 @xmath496 defined as : @xmath497 to prove @xmath494 , we need to prove that every corner point @xmath498 is achieved by some non - negative @xmath499 such that @xmath500 .
we set @xmath501 and @xmath502 and show that we can always find @xmath503 satisfying ( [ eq : cms_gauss_ach_region ] ) such that @xmath504 .
let us first consider the points @xmath505 given by : @xmath506 this can be achieved by using the following auxiliary rates , @xmath507 , @xmath508 and @xmath509 it is easy to verify that @xmath503 if @xmath490 .
hence @xmath510 .
let us next consider the points @xmath511 given by : @xmath512 again it is straightforward to show that @xmath513 belongs to @xmath492 .
finally , we consider the remaining two points @xmath514 and @xmath515 .
@xmath514 is given by : @xmath516 where @xmath517 .
consider the following auxiliary rates : @xmath508 , @xmath507 and @xmath518 .
clearly the first three constraints in ( [ eq : cms_gauss_ach_region ] ) are satisfied by these auxiliary rates .
the following inequalities prove that the last two constraints are also satisfied by these rates and hence @xmath519 .
@xmath520 @xmath521 next consider @xmath515 : @xmath522 using same arguments as before , it can be shown that @xmath523 by using the following auxiliary rates : @xmath508 , @xmath507 and @xmath524 .
therefore , it follows that @xmath525 and hence cms achieves the complete rate region , proving ( i ) .
we next prove ( ii)(a ) .
it follows from ( i ) that the following rate point is achievable @xmath526 : @xmath527 also observe that @xmath526 , @xmath528 and hence a lower bound to the sum rate is @xmath529 .
therefore the above point achieves the minimum sum rate @xmath526 .
the proof of ( ii)(b ) follows similarly by noting that if @xmath448 , the minimum sum rate is given by @xmath530 which is achieved by the point : @xmath531 this proves the theorem .
it is interesting to observe that the optimal encoding scheme introduces common codewords ( creates an interaction ) between descriptions 1 and 3 , even though these two descriptions are never received simultaneously at the decoder .
while common codewords typically imply redundancy in the system , in this case , introducing them allows for better co - ordination between the descriptions leading to a smaller rate for the common branch .
we will use similar principles in the following section for the @xmath0descriptions framework and show that the cms scheme achieves the complete rd region for several distortion regimes . and @xmath532 .
the blue points correspond to the region of distortions wherein the cms scheme achieves the complete rate - region and the green points represent the region where the cms scheme achieves the minimum sum rate [ fig : example :- this - figure ] . ]
we consider an asymmetric setting where @xmath533 and @xmath532 .
[ fig : example :- this - figure ] shows the regime of distortions where cms achieves the complete rate - region and minimum sum rate .
the blue region corresponds to the set of distortion pairs @xmath534 wherein the cms rate - region is complete .
the green region denotes the minimum sum rate points .
it is clearly evident from the figure that cms achieves the minimum sum rate for a fairly large regime of distortions .
we next extend the coding scheme described in the previous section to the @xmath0descriptions framework . observe that , in all the setups considered so far , we have used only a single common layer random variable .
we will see , in the proof of theorem [ thm : boundary_points_l ] , that multiple common layer random variables are necessary to achieve the complete rate region of several cross - sections of the rate - distortion region .
we first describe the particular cross - section we are interested in and then state the result as part of theorem [ thm : boundary_points_l ] .
we impose only the following distortion constraints : @xmath535 and limit the remaining distortions to 1 .
this essentially corresponds to an @xmath0descriptions framework wherein each of the descriptions are either received individually or the first description is received along with another description from the set @xmath536 .
observe that the 3-descriptions framework considered in the previous section is a special case of this @xmath0channel setup . in the following theorem
, we assume without loss of generality that @xmath537 .
the results follow accordingly for any other permutation of the ordering .
[ thm : boundary_points_l]consider a cross - section of the @xmath0descriptions quadratic gaussian rate - distortion region , wherein we only impose distortions @xmath535 . without loss of generality , assume that @xmath537 holds . then the cms scheme achieves the following complete rate - region : @xmath538 if the distortions @xmath539 satisfy the following conditions : @xmath540 where @xmath541 , @xmath542 and @xmath543 are given by : @xmath544 it is possible to show that the cms scheme achieves the minimum sum rate in several distortion regimes similar to theorem [ thm : sum_rate_gauss ]
. however , we omit explicitly stating the distortion regimes here as they can be derived directly from the above theorem . following the arguments as in the proof of theorem [ thm : sum_rate_gauss ]
, the converse for the rate region in ( [ eq : sum_rate_l_thm1 ] ) follows directly using ozarow s converse results in @xcite . the encoding scheme which achieves this sum rate resembles that in theorem [ thm : sum_rate_gauss ] .
first , a quantized version of the source , optimized to achieve @xmath545 , is sent in description @xmath2 .
this information is also sent as part of all descriptions @xmath546 .
then , refinement information is sent as part of description @xmath547 which helps in achieving a lower distortion @xmath548 .
this first layer of refinement information is also sent as part of all descriptions @xmath549 .
such an encoding mechanism is repeated successively .
the refinement information generated at layer @xmath46 is sent in description @xmath46 , as well as all descriptions @xmath550 .
note that successive refinability of a gaussian source under mse @xcite ensures that there is no rate loss in encoding the source using such a mechanism .
finally , the first description carries enough information to achieve @xmath244 and @xmath551 , when the respective descriptions are received at the decoder .
specifically , we choose the following joint density for the auxiliary random variables : @xmath552 @xmath553 we set all the other auxiliary random variables to constants and set the functions @xmath462 to be the respective optimal mse estimators where @xmath554 are zero mean jointly gaussian random variables with the covariance matrix given by : @xmath555\ ] ] where @xmath465 , @xmath466 and @xmath556 , @xmath557 .
@xmath420 is equal to @xmath421 ( @xmath421 is defined in eq . ( [ eq : other_defn-1 ] ) ) .
this induces a correlation of @xmath558 ( defined in ( [ eq : rho_l_dist ] ) ) between @xmath62 and @xmath559 .
observe that for any fixed @xmath543 , any distortion tuple satisfying the following conditions is achievable : @xmath560 next , we have from theorem [ thm : main ] that the following rates are achievable : @xmath561 : @xmath562 following similar arguments as in theorem [ thm : sum_rate_gauss ] , it follows directly that the above rate - region is equivalent to ( [ eq : sum_rate_l_thm1 ] ) , completing the proof .
a novel encoding scheme for the general @xmath0channel multiple descriptions problem was proposed which results in a new achievable rate - distortion region that subsumes the achievable region due to venkataramani , kramer and goyal .
the proposed encoding scheme adds controlled redundancy by including a common codeword in every subset of the descriptions .
the common codewords assist in better coordination between the descriptions leading to a strictly larger region for a fairly general class of sources and distortion measures . in particular
, we showed that the proposed scheme achieves a strictly larger rate - distortion region for a gaussian source under mse and for a binary symmetric source under hamming distortion for all @xmath1 .
we further showed that cms achieves the complete rate - distortion region for several asymmetric cross - sections of the @xmath0channel quadratic gaussian md problem .
the possible impact of the new ideas presented here , on practical multiple descriptions encoder design and on certain special cases of the general setting , such as the symmetric multiple descriptions problem , will be studied as part of the future work .
10 r. venkataramani , g. kramer , and v. goyal .
multiple descriptions coding with many channels . , 49(9):21062114 , sep 2003 .
a. gamal and t.m .
achievable rates for multiple descriptions . , it-28:851857 , nov 1982 .
z. zhang and t. berger .
new results in binary multiple descriptions . , it-33:502521 , jul 1987 .
r. ahlswede .
the rate - distortion region for multiple descriptions without excess rate .
, 31(6):721 726 , nov 1985 .
l. ozarow . on a source - coding problem with two channels and three receivers .
, 59(10):19091921 , dec 1980 .
j. wang , j. chen , l. zhao , p. cuff , and h. permuter . on the role of the refinement layer in multiple description coding and scalable coding .
, 57(3):14431456 , mar 2011 .
pradhan , r. puri , and k. ramchandran .
n - channel symmetric multiple descriptions - part i : ( n , k ) source - channel erasure codes . ,
50(1):47 61 , jan . 2004 .
r. puri , s.s .
pradhan , and k. ramchandran .
n - channel symmetric multiple descriptions - part ii : an achievable rate - distortion region . , 51(4):1377 1392 , apr 2004 .
h. wang and p. viswanath .
vector gaussian multiple descriptions with individual and central receivers .
, 53(6):2133 2153 , jun 2007 .
j. chen .
rate region of gaussian multiple description coding with individual and central distortion constraints .
, 55(9):3991 4005 , sep 2009 .
design of multiple description scalar quantizers .
, 39(3):821 834 , may 1993 .
p. koulgi , s.l .
regunathan , and k. rose .
multiple description quantization by deterministic annealing .
, 49(8):2067 2075 , aug 2003 .
h. wang and p. viswanath .
vector gaussian multiple description with two levels of receivers . , 55(1):401 410 , jan 2009 .
s. mohajer , c. tian , and s.n .
asymmetric multilevel diversity coding and asymmetric gaussian multiple descriptions .
, 56(9):4367 4387 , sep 2010 .
d. slepian and j. k. wolf .
noiseless coding of correlated information sources . , 19:471480 , jul 1973 .
a. d. wyner and j. ziv . the rate - distortion function for source coding with side information at the decoder . , 22:110 ,
jan 1976 .
c. tian and j. chen .
new coding schemes for the symmetric k -description problem . , 56(10):53445365 , oct 2010 .
c. tian , s. mohajer , and s.n .
approximating the gaussian multiple description rate region under symmetric distortion constraints .
, 55(8):3869 3891 , aug 2009 . j. chen l. song , s. shuo .
a lower bound on the sum rate of multiple description coding with symmetric distortion constraints . , 60:75477567 , dec 2014 .
k. viswanatha , e. akyol , and k. rose . on optimum communication cost for joint compression and dispersive information . in _ proceedings of ieee information theory workshop ( itw ) _ , pages 15 , sep 2010 .
j. nayak , s. ramaswamy , and k. rose .
correlated source coding for fusion storage and selective retrieval . in _ proceedings of ieee international symposium on information theory _ , pages 9296 , sep 2005 .
k. viswanatha , e. akyol , and k. rose .
combinatorial message sharing for a refined multiple descriptions achievable region . in _ proceedings of ieee international symposium on information theory proceedings ( isit ) _ , pages 1312 1316 , aug 2011 .
k. viswanatha , e. akyol , and k. rose .
a strictly improved achievable region for multiple descriptions using combinatorial message sharing . in
_ proceedings of ieee information theory workshop ( itw ) _ , pages 100 104 , oct 2011 .
e. akyol , k. viswanatha , and k. rose .
combinatorial message sharing and random binning for multiple description coding . in _ proceedings of ieee international symposium on information theory proceedings ( isit ) _ , pages 1371 1375 , jul 2012 .
e. akyol , k. viswanatha , and k. rose . on random binning versus conditional codebook methods in multiple descriptions coding . in _ proc .
ieee information theory workshop ( itw ) _ , sep 2012 .
e. akyol , k. viswanatha , and k. rose . on the role of common codewords in quadratic gaussian multiple descriptions coding in _ proceedings of ieee international symposium on information theory proceedings ( isit ) _
, pages 1367 1371 , jul 2013 .
f. shirani and s.s .
an achievable rate - distortion region for the multiple descriptions problem . in _
ieee international symposium on information theory ( isit ) _
, pages 576580 , june 2014 .
t.m . cover and j.a .
wiley - interscience , 1991 .
a. gamal and y kim . .
cambridge university press , 2011 .
the rate - distortion function for source coding with side information at the decoder ii : general sources .
, 38:6080 , jul 1978 .
z. zhang and t. berger .
multiple description source coding with no excess marginal rate .
, 41(2):349357 , mar 1995 .
r. zamir .
gaussian codes and shannon bounds for multiple descriptions .
, 45(7):2629 2636 , nov 1999 .
equitz and t.m .
successive refinement of information .
, 37(2):269 275 , mar 1991 .
han and k. kobayashi .
a unified achievable rate region for a general class of multiterminal source coding systems . , it-26:277288 , may 1980 .
j. thomas .
feedback can at most double gaussian multiple access channel capacity ( corresp . ) .
, 33(5):711 716 , sep 1987 .
we follow the notation and the notion of strong typicality defined in @xcite . we refer to @xcite ( section 3 ) for formal definitions and basic lemmas associated with typicality . _
_ let @xmath563 denote the event of an encoding error .
we have : @xmath564 from standard typicality arguments @xcite , we have @xmath565 as @xmath566 .
hence , it is sufficient to find conditions on the rates to bound @xmath567 . towards finding conditions on the rate to bound @xmath567 ,
we define event @xmath568 , for any index tuple @xmath569 , as , @xmath570 @xmath571 , where for any set @xmath194 ( defined in ( [ eq : cond_q_main ] ) ) , @xmath572 denotes the codeword tuple @xmath573 and @xmath574 denotes the tuple @xmath575_{1},\mathcal{k}\neq\emptyset\}$ ] .
define random variables @xmath576 we have @xmath577 where @xmath578 . from chebyshev s inequality , it follows that : @xmath579\leq\frac{4\mbox{var}(\psi)}{\left(e(\psi)\right)^{2}}\label{eq : p_3 - 1}\ ] ] from standard typicality arguments , we can bound @xmath580 as follows : @xmath581 where for any set @xmath191 satisfying ( [ eq : cond_q_main ] ) , we define : @xmath202_{1}}h\left(v_{\mathcal{k}}|\{v\}_{\mathcal{i}_{|\mathcal{k}|+}(\mathcal{k})}\right)\nonumber \\ & & + \sum_{\mathcal{k}\in2^{[\mathcal{q}]_{1}}-\emptyset}h\left(u_{\mathcal{k}}|\{v\}_{\mathcal{i}_{1+}(\mathcal{k})},\{u\}_{2^{\mathcal{k}}-\emptyset-\mathcal{k}}\right)\nonumber \\ & & -h\left(\{v\}_{\mathcal{q}-[\mathcal{q}]_{1}},\{u\}_{2^{[\mathcal{q}]_{1}}-\emptyset}|x\right)\label{eq : alpha_defn_1}\end{aligned}\ ] ] we follow the convention , @xmath203 . next consider @xmath582 where , @xmath583\nonumber \\
= \sum_{\{j\}_{\mathcal{j}(\mathcal{l})}}\sum_{\{k\}_{\mathcal{j}(\mathcal{l})}}p\left[\mathcal{a}(\{j\}_{\mathcal{j}(\mathcal{l})})\cap\mathcal{a}(\{k\}_{\mathcal{j}(\mathcal{l})})\right]\label{eq : e_xi_s}\end{aligned}\ ] ] the probability in ( [ eq : e_xi_s ] ) depends on whether @xmath584 and @xmath585 are equal for a subset of indices .
let @xmath571 such that @xmath586 .
observe that , due to the hierarchical structure in the conditional codebook generation mechanism , for @xmath587 to hold , @xmath191 must satisfy ( [ eq : cond_q_main ] ) , i.e. , it must be such that , @xmath588 it follows from codebook generation that given the codeword tuple @xmath589 , tuples @xmath590 and @xmath591 are independent and identically distributed .
hence we can rewrite the probability in ( [ eq : e_xi_s ] ) as : @xmath592\nonumber \\ & & = \left(p\left[\mathcal{a}(\{j\}_{\mathcal{j}(\mathcal{l})-\mathcal{q}})\biggl|\mathcal{a}(\{j\}_{\mathcal{q}}\right]\right)^{2}\nonumber \\ & & \times p\left[\mathcal{a}(\{j\}_{\mathcal{q}})\right]\label{eq : upper_bound_p_var}\\ & & = \left(p\left[\mathcal{a}(\{j\}_{\mathcal{j}(\mathcal{l})})\biggl|\mathcal{a}(\{j\}_{\mathcal{q}}\right]\right)^{2}\nonumber \\ & & \times p\left[\mathcal{a}(\{j\}_{\mathcal{q}})\right]\\ & & = \frac{\left(p\left[\mathcal{a}(\{j\}_{\mathcal{j}(\mathcal{l})})\right]\right)^{2}}{p\left[\mathcal{a}(\{j\}_{\mathcal{q}})\right]}\end{aligned}\ ] ] note that if @xmath593 ( i.e. , @xmath584 are @xmath585 are not equal for any subset of the indices ) , then @xmath594=\left(p\left[\mathcal{a}(\{j\}_{\mathcal{j}(\mathcal{l})})\right]\right)^{2}$ ] .
hence , we bound @xmath595 by ( see @xcite for a similar argument ) : @xmath596\nonumber \\ & & \times\left(p\left[\mathcal{a}(\{j\}_{\mathcal{j}(\mathcal{l})})\biggl|\mathcal{a}(\{j\}_{\mathcal{q}})\right]\right)^{2}\biggl\}\nonumber \\ & = & \sum\biggl\ { n(\mathcal{q})\frac{\left(p\left[\mathcal{a}(\{j\}_{\mathcal{j}(\mathcal{l})})\right]\right)^{2}}{p\left[\mathcal{a}(\{j\}_{\mathcal{q}})\right]}\biggl\}\end{aligned}\ ] ] where the summation is over all @xmath201 such that ( [ eq : cond_q_main ] ) holds ( i.e. , over all @xmath194 ) and @xmath568 denotes the event that @xmath597 .
@xmath598 denotes the total number of ways of choosing @xmath584 and @xmath585 such that they overlap in the subset @xmath191 .
an upper bound on @xmath598 is given in eq .
( [ eq : upper_bound_ratee_var ] ) . also observe that the term corresponding to @xmath593 gets canceled with the ` @xmath599 ' term in @xmath595 leaving a summation over all non - empty @xmath194 .
@xmath600_{1}}r_{\mathcal{k}}^{''}+\sum_{l\in[\mathcal{q}]_{1}}r_{l}^{'}\right)}\prod_{\mathcal{k}\in\mathcal{j}(\mathcal{l})-\mathcal{q}-[\mathcal{j}(\mathcal{l})-\mathcal{q}]_{1}}2^{nr_{i,\mathcal{k}}^{'}}(2^{nr_{i,\mathcal{k}}^{'}}-1)\prod_{l\in[\mathcal{j}(\mathcal{l})-\mathcal{q}]_{1}}2^{nr_{i , l}^{'}}(2^{nr_{i , l}^{'}}-1)\nonumber \\ & \leq & 2^{n\{\sum_{\mathcal{k}\in\mathcal{q}-[\mathcal{q}]_{1}}r_{i,\mathcal{k}}^{''}+\sum_{l\in[\mathcal{q}]_{1}}r_{i , l}^{'}+2\sum_{\mathcal{k}\in\mathcal{j}(\mathcal{l})-\mathcal{q}-[\mathcal{j}(\mathcal{l})-\mathcal{q}]_{1}}r_{i,\mathcal{k}}^{''}+2\sum_{l\in[\mathcal{j}(\mathcal{l})-\mathcal{q}]_{1}}r_{i , l}^{'}\}}\nonumber \\ & = & 2^{n\{2\sum_{\mathcal{k}\in i_{1+}}r_{i,\mathcal{k}}^{''}+2\sum_{l\in\mathcal{l}}r_{i , l}^{'}-\sum_{\mathcal{k}\in\mathcal{q}-[\mathcal{q}]_{1}}r_{i,\mathcal{k}}^{''}-\sum_{l\in[\mathcal{q}]_{1}}r_{i , l}^{'}\}}\label{eq : upper_bound_ratee_var}\end{aligned}\ ] ] on substituting the upper bound for @xmath598 , we have : @xmath601_{1}}r_{\mathcal{k}}^{''}-\sum_{l\in[\mathcal{q}]_{1}}r_{l}^{'}\right)+3n\epsilon}\biggl\}\label{eq : var_xi}\end{aligned}\ ] ] where the summation is over all @xmath602 such that ( [ eq : cond_q_main ] ) holds , i.e. , over all @xmath194 . inserting ( [ eq : var_xi ] ) and
( [ eq : e_xi ] ) in ( [ eq : p_3 - 1 ] ) , we get , @xmath603_{1}}r_{\mathcal{k}}^{''}-\sum_{l\in[\mathcal{q}]_{1}}r_{l}^{'}\right)+5n\epsilon}\label{eq : p_4}\ ] ] where the summation is over all @xmath602 such that ( [ eq : cond_q_main ] ) holds , i.e. , over all @xmath194 .
@xmath604 can be made arbitrarily small if @xmath605_{1}}r_{\mathcal{k}}^{''}+\sum_{l\in[\mathcal{q}]_{1}}r_{l}^{'}>\alpha(\mathcal{q})\,\,\forall\mathcal{q}\in\mathcal{q}^{*}$ ] .
in this appendix , we will show that a gaussian random variable , under mse distortion belongs to both @xmath402 and @xmath281 and hence cms achieves points strictly outside the vkg region . throughout this section ,
we use the notation @xmath606 , to denote the set of all jointly gaussian distributions .
first , let us recall the definitions of @xmath281 and @xmath402 .
@xmath281 is defined in definition [ definition : defn_zzb ] as the set of all random variables @xmath4 for which there exists a strict sub - optimality in the zb region ( with respect to the given distortion measures ) when the closure of the rates in ( [ eq : zb ] ) is defined only over joint distributions that satisfy the following conditions : @xmath290 & \leq & d_{\mathcal{k}},\,\,\mathcal{k}\in\{1,2,12\}\nonumber \\
u_{12 } & = & f(u_{1},u_{2},v_{12})\label{eq : zzb-1 - 1 - 1}\end{aligned}\ ] ] where @xmath404 is some deterministic function .
note that the statement in definition [ definition : defn_zzb ] is more general than the above description . however , for a gaussian source , under mse , it is possible to show strict sub - optimality in zb region even for @xmath607 , which leads to the above constraints for the joint distributions .
similarly , @xmath4 is said to belong to @xmath402 if there exists a strict sub - optimality in the ec region when the closure of the rates in ( [ eq : egc ] ) is defined only over joint distributions that satisfy the following conditions : @xmath608 & \leq & d_{\mathcal{k}},\,\,\mathcal{k}\in\{1,2,12\}\nonumber \\
u_{12 } & = & f(u_{1},u_{2})\label{eq : zzb-1 - 1 - 1 - 1}\end{aligned}\ ] ] where @xmath404 is some deterministic function .
we next show that a gaussian source under mse belongs to both @xmath402 and @xmath281 .
we first prove the result for @xmath402 and then extend similar arguments for @xmath281 .
\1 ) * proof for @xmath402 * : we first give an intuitive argument to justify the claim and then follow it up with the formal proof .
it follows from ozarow s results ( see also @xcite ) that there exists a regime of distortions @xmath330 for which the following rate region is achievable ( and complete ) : @xmath609 i.e. , there is no excess rate incurred due to encoding the source using two descriptions .
equivalently , there exists a regime of distortions for which the excess sum rate term in the ec region ( i.e. , @xmath610 ) must be set to zero to achieve the complete rate - region .
we will show that , if we restrict the optimization to conditionally independent joint densities , then it is impossible to simultaneously satisfy all the distortions and achieve @xmath611 .
we next provide a formal proof .
our objective is to show that for the 2-descriptions quadratic gaussian md problem , there is a strict sub - optimality if we perform the ec encoding scheme and restrict the optimization to joint densities satisfying the following conditions : for some functions @xmath613,@xmath614 and @xmath404 .
we denote by @xmath615 , the closure of all the achievable rd tuples using the ec encoding scheme , over all joint densities satisfying ( [ eq : ec_cms ] ) .
we need to show that @xmath615 is strictly smaller than @xmath424 .
we consider one point in @xmath424 and show that it is not contained in @xmath615 .
let the distortions be such that @xmath616 holds .
this regime of distortions has been termed as the ` high distortion regime ' in the literature and the complete rate region under these constraints is given by ( [ eq : high_dist_region ] ) ( see for example @xcite ) .
let us consider the following rate point : @xmath617 .
we will show that this point in not contained in @xmath615 when @xmath616 .
we first rewrite the ec region for any joint density over the random variables @xmath618 and functions @xmath613 and @xmath614 : @xmath619,\mathcal{k}\in2^{\{1,2\}}-\emptyset\nonumber\label{eq : egc-1}\end{aligned}\ ] ] observe that , to achieve @xmath620 , we have to impose the joint density for @xmath621 and the function @xmath622 to be rd optimal at @xmath244 .
this requires @xmath621 to be jointly gaussian and @xmath622 to be the optimal estimator of @xmath4 given @xmath62 .
note that , if @xmath623 is not gaussian ( with @xmath622 being the mmse estimator ) , it is possible to construct a jointly gaussian distribution which achieves the same distortion with smaller rate .
for example , one could construct a distribution @xmath624 , with variance of @xmath625 equal to the variance of @xmath626 . as a gaussian distribution achieves maximum entropy among all distributions with the same variance , @xmath627 , leading to a smaller rate for @xmath79 .
therefore , we can assume that @xmath628 .
we denote this joint density by @xmath629 .
it follows that the infimum @xmath80 achievable using an independent quantization scheme when @xmath630 is given by : @xmath631 where the infimum is over all joint densities @xmath632 and functions @xmath633 and @xmath404 satisfying ( [ eq : ec_cms ] ) , where @xmath628 is rd optimal at @xmath244 .
we will next show that @xmath634 . where the infimum is over all jointly gaussian densities @xmath638 satisfying the following conditions : @xmath639 & \leq & d_{2}\nonumber \\
e\left[(x-\tilde{u}_{12})^{2}\right ] & \leq & d_{12}\label{eq : ec_cms-1}\end{aligned}\ ] ] consider any joint density @xmath641 and functions @xmath613,@xmath642@xmath404 , satisfying ( [ eq : ec_cms ] ) , where @xmath621 are distributed according to a jointly gaussian density which is rd optimal at @xmath244 .
we now construct a joint density @xmath643 that achieves a smaller value for the quantity in ( [ eq : r_ec_iq ] ) and is in @xmath606 .
consider @xmath644 such that @xmath645 , i.e. , the random variables @xmath646 have the same covariance matrix as @xmath647 , given @xmath621 .
clearly , all the conditions in ( [ eq : ec_cms-1 ] ) are satisfied by the joint density @xmath648 . we need to show that the quantity in ( [ eq : r_ec_iq ] ) is smaller for the joint density @xmath648 compared to @xmath378 .
recall that , for a fixed covariance matrix , a gaussian distribution over the relevant random variables maximizes the conditional entropy @xcite .
hence , we have : @xmath649 leading to @xmath650 . equipped with this result , we continue with our proof to show that @xmath651 .
consider the following series of inequalities : @xmath652 where the infimum is over all @xmath653 satisfying ( [ eq : ec_cms-1 ] ) .
note that ( a ) follows from the fact that @xmath654\leq d_{12}$ ] ( and therefore @xmath655 ) .
( b ) follows from the fact that @xmath656 are jointly gaussian satisfying the markov condition @xmath657 , where @xmath658 and @xmath659 .
therefore , @xmath660 , proving that there is a strict sub - optimality if we restrict the ec coding scheme to an independent quantization mechanism for a gaussian random variable under mse .
hence a gaussian random variable , under mse , belongs to @xmath402 .
\2 ) * proof for @xmath281 * : our next objective is to prove that a gaussian random variable under mse also belongs to @xmath281 .
equivalently , we need to show that there is a strict sub - optimality in the zb region when we restrict the optimization to joint densities satisfying the following constraints : the proof follows in very similar lines to that for @xmath402 , the main difference being that all the quantities are now defined conditioned on every value of @xmath86 . recall that the zb region achievable using any joint density @xmath662
is given by : @xmath663,\,\,\mathcal{k}\subseteq\{1,2\},\mathcal{k}\neq\emptyset\nonumber\end{aligned}\ ] ] before we continue with the proof , we take a small digression and prove certain properties of the zb region .
we summarize these properties in the following lemma .
[ lemma : end_lemma]the closure of the rd - tuples in @xmath81 can be restricted to joint distributions that satisfy the following markov chain conditions , without any loss in the achievable region : @xmath664 let @xmath662 be any joint density , which may or may not satisfy the above properties , and let @xmath613 and @xmath614 be functions that satisfy the respective distortion constraints .
we will construct another joint density @xmath665 that satisfies ( [ eq : zb_end_markov ] ) and achieves the same rd region .
towards constructing such a joint density , we set : @xmath666 first , observe that the above choice of the joint density satisfies ( [ eq : zb_end_markov ] ) .
also observe that , @xmath667 , @xmath668 , @xmath669 and @xmath670 .
hence , the rate region achievable by @xmath361 is the same as that of @xmath671 .
moreover , all the distortion constraints can be satisfied using the same functions @xmath613 and @xmath614 .
hence it follows that the entire zb region can be achieved by considering a closure only over random variable that satisfy ( [ eq : zb_end_markov ] ) .
equipped with these results , we continue our proof .
let us again consider the corner point @xmath617 for some @xmath330 satisfying @xmath616 , and show that it is not achievable by zb when we restrict the joint densities to satisfy ( [ eq : zb_cms_1 ] ) .
note that , as a result of lemma ( [ lemma : end_lemma ] ) and corollary ( [ cor : last ] ) , it is sufficient to consider joint densities that also satisfy ( [ eq : zb_end_markov ] ) .
observe that , as @xmath673 , @xmath620 can be achieved only by joint densities that satisfy @xmath674 .
hence , to prove that a gaussian random variable under mse belongs to @xmath281 , it is sufficient to show that @xmath620 is not achievable when we restrict the optimization to joint densities satisfying ( [ eq : zb_cms_1 ] ) , ( [ eq : zb_end_markov ] ) and @xmath674 .
d_{\mathcal{k } } & \geq & e\left[(x-\psi_{\mathcal{k}}(u_{\mathcal{k}}))^{2}\right],\,\,\mathcal{k}\subset\{1,2\},\mathcal{k}\neq\emptyset\nonumber \\ & = & e\left[e\left[(x-\psi_{\mathcal{k}}(u_{\mathcal{k}}))^{2}\bigl|v_{12}\right]\right]\label{eq : zb_v12_split}\end{aligned}\ ] ] we will next show that the optimization can be further restricted to joint densities @xmath678 such that @xmath679 given @xmath680 , @xmath681 .
first , as @xmath4 is independent of @xmath86 , @xmath682 is gaussian @xmath681 .
recall that @xmath620 is obtained by first minimizing @xmath79 , followed by minimizing @xmath80 given @xmath79 subject to all the distortion constraints . from ( [ eq : zb_v12_split ] ) , we have : @xmath683 where the infimum is over all joint densities @xmath684 satisfying the distortion constraint @xmath244 . towards proving that @xmath628 given @xmath680 , @xmath681 , let @xmath684 be any joint density over the given alphabets and let function @xmath344 achieve the distortion constraint at @xmath244 .
consider the following joint density @xmath685 , such that @xmath686 are jointly gaussian given @xmath680 and @xmath687 , @xmath681 . observe that this joint density satisfies the distortion constraint for @xmath244 .
moreover , it achieves a smaller rate , as a gaussian density over the relevant random variables maximizes the conditional entropy for a fixed covariance matrix , i.e. , @xmath688 , @xmath681 .
therefore , to achieve minimum @xmath689 , we can consider only densities wherein @xmath686 are jointly gaussian given @xmath86 .
we next focus on minimizing @xmath80 given @xmath630 . from ( [ eq : zb_v12_split ] ) , we have : @xmath690-r_{1 } \bigl\ } \label{eq : last_eq}\end{aligned}\ ] ] where the infimum is over all joint densities @xmath691 satisfying ( [ eq : zb_cms_1 ] ) , ( [ eq : zb_end_markov ] ) and @xmath674 and where @xmath686 are jointly gaussian given @xmath680 , @xmath681 ( required to minimize @xmath79 ) .
it is easy to show using similar arguments that , to achieve the infimum in ( [ eq : last_eq ] ) , it is sufficient to consider joint densities where @xmath679 given @xmath680 and @xmath692 , @xmath681 . to summarize what we have
so far , we have shown that to achieve minimum @xmath79 followed by minimum @xmath80 in @xmath672 , it is sufficient to consider only joint densities @xmath662 that satisfy the following properties : 1 .
@xmath693 , @xmath292 and @xmath662 satisfies all the distortion constraints 2 .
@xmath674 3 .
@xmath618 are jointly gaussian given @xmath680 @xmath681 4 .
the markov chain conditions in ( [ eq : zb_end_markov ] ) are satisfied 5 .
@xmath62 achieves rd - optimality at @xmath244 we will show that the point @xmath620 can not be achieved by any joint density that satisfies the above properties . observe that , given @xmath680 , @xmath694 are jointly gaussian random variables satisfying the markov condition @xmath695 .
this leads to two possibilities for the joint distribution @xmath662 : \i ) @xmath696 such that @xmath697 and @xmath698 : as @xmath694 are jointly gaussian , this leads to the conclusion that @xmath699 and hence @xmath700 .
this clearly implies that there is excess rate on @xmath80 and hence point @xmath620 is not achievable .
\ii ) @xmath681 , either @xmath701 or @xmath702 .
this implies that the alphabet space of @xmath86 can be divided into two disjoint subsets such that , when @xmath86 takes values in the first set , @xmath4 and @xmath63 are independent , and when @xmath86 takes values in the second set @xmath4 and @xmath62 are independent .
let the two sets be denoted by @xmath703 and @xmath704 , respectively .
we have : @xmath705 however , recall that the joint distribution can be restricted to satisfy , @xmath674 , @xmath706 and @xmath707 .
hence , we have : @xmath708 @xmath709 where ( a ) and ( b ) follow from ( [ eq : end3 ] ) .
this implies that @xmath621 and @xmath710 must be pair - wise independent and hence the distortion constraints on @xmath244 and @xmath245 can not be satisfied .
kumar b. viswanatha(s08 ) received his phd degree in 2013 in the electrical and computer engineering department from university of california at santa barbara ( ucsb ) , usa .
he is currently working for qualcomm research center in san diego . prior to joining qualcomm , he was an intern associate in the equity volatility desk at goldman sachs co. , new york , usa .
his research interests include multi - user information theory , wireless communications , joint compression and routing for networks and distributed compression for large scale sensor networks .
emrah akyol(s03 , m12 ) received the ph.d .
degree in 2011 in electrical and computer engineering from the university of california at santa barbara . from 2006 to 2007 , he held positions at hewlett - packard laboratories and ntt docomo laboratories , both in palo alto , where he worked on topics in video compression . currently , dr . akyol is a postdoctoral researcher in the department of electrical and computer engineering , university of illinois urbana - champaign . prior to that , he was a post doctoral researcher at electrical engineering department at university of southern california and at university of california - santa barbara .
his current research is on the interplay of networked information theory , communications and control .
he then joined the department of electrical and computer engineering , university of california at santa barbara , where he is currently a professor .
his main research activities are in the areas of information theory and signal processing , and include rate - distortion theory , source and source - channel coding , audio and video coding and networking , pattern recognition , and non - convex optimization .
he is interested in the relations between information theory , estimation theory , and statistical physics , and their potential impact on fundamental and practical problems in diverse disciplines . | this paper presents a new achievable rate - distortion region for the general l channel multiple descriptions problem .
a well known general region for this problem is due to venkataramani , kramer and goyal ( vkg ) @xcite .
their encoding scheme is an extension of the el - gamal - cover ( ec ) and zhang - berger ( zb ) coding schemes to the l channel case and includes a combinatorial number of refinement codebooks , one for each subset of the descriptions . as in zb
, the scheme also allows for a single common codeword to be shared by all descriptions .
this paper proposes a novel encoding technique involving ` combinatorial message sharing ' ( cms ) , where every subset of the descriptions may share a distinct common message .
this introduces a combinatorial number of shared codebooks along with the refinement codebooks of @xcite .
these shared codebooks provide a more flexible framework to trade - off redundancy across the messages for resilience to descriptions loss .
we derive an achievable rate - distortion region for the proposed technique , and show that it subsumes the vkg region for general sources and distortion measures .
we further show that cms provides a strict improvement of the achievable region for any source and distortion measures for which some 2-description subset is such that zb achieves points outside the ec region .
we then show a more surprising result : cms outperforms vkg for a general class of sources and distortion measures , including scenarios where the zb and ec regions coincide for all 2-description subsets .
in particular , we show that cms strictly improves on vkg , for the @xmath0channel quadratic gaussian md problem , for all @xmath1 , despite the fact that the ec region is complete for the corresponding 2-descriptions problem .
consequently , the ` correlated quantization ' scheme ( an extreme special case of vkg ) , that has been proven to be optimal for several cross - sections of the @xmath0channel quadratic gaussian md problem , is strictly suboptimal in general . using the encoding principles derived ,
we show that the cms scheme achieves the complete rate - distortion region for several asymmetric cross - sections of the @xmath0channel quadratic gaussian md problem .
multiple descriptions coding , source coding , rate - distortion theory , combinatorial message sharing |
matter - wave bright solitons stable solitary waves whose density is greater than the background one have recently been realized in a quasi one - dimensional ( 1d ) bose - einstein condensate ( bec ) @xcite . in 1d
, the quantum kinetic pressure counterbalances an attractive interatomic interaction , allowing a stable bright soliton to be formed .
however , in 2d or higher dimensions , bright solitons are always unstable against collapse or expansion as long as the attractive interaction is constant in time @xcite .
it has been demonstrated by abdullaev _
et al . _
@xcite and by us @xcite that by rapid oscillations of interaction between attractive and repulsive through , e.g. , the feshbach resonance @xcite , a bec can be stabilized in a quasi 2d space without a radial trapping potential .
the mechanism of the stabilization is similar to that of an inverted pendulum , in which a bob is stabilized above the vertically oscillating pivot @xcite .
however , the stabilization of a bec in 3d free space has been elusive due to complex dynamical instabilities inherent in nonlinear nonequilibrium systems . in the present paper , we show that a droplet @xcite of a bec can be stabilized in 3d free space by using oscillating interactions . in order to simulate experimental situations at finite temperature
, we take into account the effect of dissipation caused by the interaction between a bec and a thermal cloud .
this can be incorporated in mean field theory by a non - unitary term added to the gross - pitaevskii ( gp ) equation .
this model has successfully been used to describe damping of collective modes @xcite and vortex nucleation @xcite .
we find that the dissipation plays a crucial role in the stabilization of a bec droplet by suppressing the dynamical instabilities , and show that the droplet can indeed be stabilized for a realistic amount of dissipation .
this paper is organized as follows .
section [ s : gp ] introduces the gp equation with oscillating interactions , and shows that the dependence of the dynamics on the oscillation frequency can be eliminated by the scaling property when a trapping potential vanishes .
section [ s : num ] presents the main results of this paper , in which a bec droplet is shown to be stabilized in 3d free space by use of oscillating interactions .
the stability diagrams with respect to the interaction and dissipative parameters are also obtained .
section [ s : var ] studies the variational analysis , and sec .
[ s : conc ] concludes this paper .
we consider the gp equation in the presence of dissipation given by @xcite @xmath0 where @xmath1 is a phenomenological dissipation constant which is to be determined experimentally , and the trapping potential @xmath2 and the s - wave scattering length @xmath3 are controlled to vary in time .
the last term on the right - hand side of eq .
( [ gp ] ) guarantees the normalization @xmath4 with @xmath5 being the number of atoms , where the chemical potential is given by @xmath6 for simplicity , we assume an isotropic trapping potential @xmath7 where @xmath8 is unity at @xmath9 , then decreases towards zero , and vanishes at some time .
the s - wave scattering length is made to oscillate as @xmath10 where @xmath11 ramps up from 0 to 1 .
the ramp functions @xmath8 and @xmath11 are introduced to avoid initial nonadiabatic disturbances that cause dynamic instabilities .
these functions must be chosen so that neither collapse nor expansion occurs in the course of the ramping .
normalizing the length , time , energy , and wave function by @xmath12 , @xmath13 , @xmath14 , and @xmath15 , respectively , we obtain the normalized form of the gp equation : @xmath16 where @xmath17 , and the wave function is normalized as @xmath18 .
it should be noted that once the trapping potential is switched off [ @xmath19 , eq . ( [ gpn ] ) no longer depends on @xmath20 explicitly .
this implies that the @xmath20 dependence can be eliminated by appropriate scaling of the time , length , and interaction .
the dynamics at driving frequency @xmath21 can be reduced to those at @xmath20 by rescaling @xmath22 , @xmath23 , and @xmath24 .
for example , if large @xmath20 is unavailable experimentally , we have only to increase the strength of interaction and enlarge the size of the bec droplet .
we numerically solve the gp equation ( [ gpn ] ) using the crank - nicholson scheme @xcite , where the following simple forms of the ramp functions are used : @xmath25 we use a quadratic function for @xmath11 because it makes initial nonadiabatic disturbances smaller than linear ramps .
figure [ f : stable ] shows the time evolution of the peak density @xmath26 and the monopole moment @xmath27 for @xmath28 and @xmath29 .
( left axis ) and monopole moment @xmath30 ( right axis ) for the oscillating interaction @xmath31 with @xmath32 .
the initial state is the noninteracting ground state in a trapping potential @xmath33 .
then the interaction is gradually turned on , and the trapping potential is simultaneously turned off as shown in the inset .
( b ) a magnification of ( a ) .
the inset shows the density profile @xmath34 from @xmath35 to @xmath36 . the dotted line is a sine function for comparison of the phase .
, width=317 ] the dissipation constant @xmath1 is taken to be 0.03 @xcite , and the ramp parameters are @xmath37 and @xmath38 .
the initial state is chosen to be the noninteracting ground state for a trapping potential with @xmath39 .
the density and width oscillate at the driving frequency .
[ the black bands in fig .
[ f : stable ] ( a ) represent rapid oscillations of @xmath40 and @xmath41 which are beyond the resolution limit . ]
we note that the amplitudes and mean values of both @xmath42 and @xmath41 converge to constant values , indicating that a bec droplet is dynamically stabilized .
the initial disturbances last for a relatively long time shown as slow oscillations or ripples of @xmath42 and @xmath41 in fig .
[ f : stable ] ( a ) , which gradually decay due to dissipation . from the time evolution of the density profile @xmath34 [ inset in fig .
[ f : stable ] ( b ) ] , we find that the droplet exhibits breathing - mode oscillations with a large amplitude .
this indicates that large oscillations of mass flow persist even after stabilization .
we find from fig .
[ f : stable ] ( b ) that the phase of the breathing - mode oscillations of the droplet lags slightly behind that of the oscillating interaction shown as the dotted curve .
this retardation is considered to be due to dissipation because there is no phase difference in the dissipation - free 2d case @xcite , where @xmath41 and @xmath42 are in - phase and out - of - phase with the oscillating interaction .
we gradually changed @xmath43 and @xmath44 from those of the stable droplet state in fig .
[ f : stable ] , and found distinct types of instabilities as shown in fig .
[ f : unstable ] .
( left axis ) and monopole moment @xmath30 ( right axis ) .
the initial state is prepared as in fig .
[ f : stable ] ( a ) , and @xmath45 are linearly changed from @xmath46 to @xmath47 in ( a ) , @xmath48 in ( b ) , and @xmath49 in ( c ) during @xmath50 .
the insets in ( b ) and ( c ) are magnifications of @xmath51 in the dashed squares and the density profiles @xmath34 at the times indicated by a and b. , width=317 ] when @xmath52 is decreased , a droplet expands because of a decrease in an attractive interaction as shown in fig .
[ f : unstable ] ( a ) .
the expansion occurs also by an increase in @xmath44 , since the effective repulsive interaction due to the oscillation is proportional to @xmath53 [ see eq .
( [ eom2 ] ) ] .
when @xmath52 is increased and @xmath44 is decreased , the system becomes dynamically unstable against the slow oscillation and eventually expands away as shown in fig .
[ f : unstable ] ( b ) .
the expansion occurs , despite the fact that an increase in @xmath52 enhances the attraction and a decrease in @xmath44 suppresses the repulsion .
the slow oscillation is seen in the profile of the maxima of @xmath41 ( the ripple in the upper edge ) , while the minima is almost constant .
figure [ f : unstable ] ( c ) shows an instability that arises for large @xmath52 and @xmath44 .
since the modulation at half frequency grows , this instability is considered to arise from parametric resonance .
the density profiles in the inset show that higher radial modes are excited , which leads to destabilization of the droplet .
we prepare a stable droplet in the same manner as in fig .
[ f : stable ] , and change the parameters to various values to obtain the stability diagram shown in fig .
[ f : contour ] . ,
and the interaction is linearly changed during @xmath50 as in fig .
[ f : unstable ] .
( a ) the black region corresponds to the one in which a droplet expands during the change of the interaction , and the white region corresponds to the one in which a droplet survives until at least @xmath54 .
the behaviors in the regions `` instability i , ii , and iii '' correspond to those in fig .
[ f : unstable ] ( a ) , ( b ) , and ( c ) , respectively .
( b ) the parameter @xmath1 is also linearly changed as well as the interaction during @xmath50 from @xmath55 to @xmath56 .
, width=317 ] the parameters @xmath43 and @xmath44 are changed linearly during @xmath57 . we define the lifetime of the droplet as the duration between @xmath58 and the time at which the droplet begins to expand indefinitely . the regions referred to as `` instability i , ii , and iii '' in fig .
[ f : contour ] correspond to the instabilities demonstrated in fig .
[ f : unstable ] ( a ) , ( b ) , and ( c ) .
we find from fig .
[ f : contour ] ( a ) that the stable region is surrounded by the three distinct types of instabilities .
the boundary between the `` stable '' and `` instability i '' regions are sharp while the `` instability ii '' gradually sets in as @xmath52 increases and @xmath44 decreases .
the collapse occurs for large @xmath52 and small @xmath44 , but the corresponding parameter regime lies outside fig .
[ f : contour ] . in fig
[ f : contour ] ( b ) , @xmath1 is linearly decreased from 0.03 to 0.01 as @xmath52 and @xmath44 are ( also linearly ) changed .
the instability regions enlarge as compared with those of fig [ f : contour ] ( a ) , and the stability region disappears . from this result , we conclude that the stability region exists only for @xmath59 . however , we can not exclude the possibility that other stable states exist , since we do not investigate the entire functional space . in numerical calculations of the time evolution , we must pay special attention to the boundary effect .
since the trapping potential is absent , the atoms that escape from the droplet spread out , reflect at the boundary , and return to the droplet region , producing spurious boundary effects .
for example , when the dirichlet boundary condition is imposed at @xmath60 , the lifetime of the droplet ( whose definition is the same as in fig .
[ f : contour ] ) for @xmath28 and @xmath61 becomes @xmath62 , while the correct value is @xmath63 .
therefore , the spatial cutoff in the numerical analysis must be much larger than the size of the droplet .
the numerical simulations that we have presented so far have been carried out under the assumption of the spherical symmetry of the system . however , this geometry can not generate multipole dynamical instabilities which destroy a bec droplet . in order to check the absence of such multipole instabilities , we have performed full 3d calculations by discretizing the space of @xmath64 size in our dimensionless unit into a @xmath65 mesh . a limited spatial size due to
our computational power causes spurious boundary effects and a rough mesh produces numerical errors . nevertheless , as shown in fig .
[ f:3d ] , snapshots of time evolution of a bec droplet appear isotropic and do not show any multipole instability . in units of @xmath66 .
, width=317 ] we have thus confirmed that a bec droplet can be stabilized without multipole instabilities .
the gaussian variational wave function well describes the dynamical stabilization of becs in 2d free space @xcite . we examine a variational method in 3d using the gaussian trial function as @xcite @xmath67 where @xmath68 is the variational parameter that characterizes the size of the condensate . substituting eq .
( [ gauss ] ) into the action @xmath69 we obtain the equation of motion for @xmath70 as @xmath71 where @xmath72 .
we separate @xmath70 into the slowly varying part @xmath73 and the rapidly oscillating part @xmath74 as @xmath75 . according to ref .
@xcite , an effective potential for @xmath73 is given by @xmath76 , where @xmath77 is the amplitude of the oscillating `` force '' for @xmath73 and @xmath20 is its frequency . from eq .
( [ eom ] ) , @xmath77 corresponds to @xmath78 .
since we are using a system of units in which @xmath79 , the effective potential becomes @xmath80 , and then the equation of motion for @xmath73 reads @xmath81 this equation agrees with that in ref .
@xcite when @xmath20 is much larger than the characteristic frequencies of the system .
thus , the oscillating interaction blocks the collapse by the effective potential proportional to @xmath82 .
the effective potential @xmath83 has a local minimum for appropriate values of @xmath84 and @xmath85 .
for instance , when @xmath86 and @xmath87 , @xmath83 has a local minimum at @xmath88 .
in fact , by numerically solving the equation of motion ( [ eom ] ) for @xmath70 , we find that a stable solution exists
. however , fig .
[ f : contour ] ( b ) implies that the system is unstable against the dynamical instability for @xmath89 .
this indicates that the simple gaussian function ( [ gauss ] ) is insufficient to describe the instability of the system .
figure [ f : diff ] illustrates the difference between the density profiles of the stable droplet ( @xmath90 in fig .
[ f : stable ] ) and the gaussian functions fitted to them .
( solid curves ) of the stable droplet for @xmath28 and @xmath29 ( @xmath91 in fig .
[ f : stable ] ) when the peak density becomes ( a ) maximal and ( b ) minimal .
the dashed curves are the gaussian functions fitted to the density profiles .
, width=317 ] we adopt the least squares fitting of @xmath92 to @xmath93 with fitting parameter @xmath94 , which is an appropriate fitting in 3d .
figure [ f : diff ] shows that the numerically exact wave functions significantly deviate from the gaussian functions even in the stable droplet .
we note that the gaussian function in fig [ f : diff ] ( a ) is almost the same as that in fig [ f : diff ] ( b ) .
this indicates that the outer region , which has large weight in the fitting due to the factor @xmath95 , is almost stationary , and the mass flow occurs mainly around the center of the droplet .
thus , more appropriate functions are needed to correctly describe the phenomena .
we have studied a bose - einstein condensate with an oscillating interaction with dissipation , and shown that a bec droplet is stabilized in 3d free space with realistic dissipation , @xmath55 .
the frequency of the oscillating interaction must be much faster than the characteristic frequency of the system , but it can be a moderate frequency according to the scaling property discussed in sec . [
for instance , the situation in fig .
[ f : stable ] can be realized in a condensate of @xmath96 @xmath97 atoms by the s - wave scattering length @xmath98 [ nm ] with @xmath99 hz , where a trapping potential with frequency 3.3 hz is removed in 25 ms , and the final size of the droplet becomes a few micrometers .
such an oscillation of interaction can be easily realized in experiments using the feshbach resonance @xcite . under gravity
, a condensate falls after the trapping potential is turned off . in the above example , the gravitational sag in the initial trapping potential is @xmath100 cm , and the condensate falls about 1 mm until the trapping potential vanishes .
the effect of gravity can be canceled out using the magnetic levitation @xcite , which enables us to observe the longtime behavior of the droplet .
a `` gaseous bec droplet '' found in this paper is different from the usual condensate in that it coheres by itself without the help of the trapping potential .
this self - trapped matter wave might exhibit interesting dynamics , such as in collective mode , collapsing dynamics , and vortex nucleation , which will be discussed elsewhere
. _ note added _ while preparing the present paper , a similar paper @xcite appeared which also reports stabilization of a bec in 3d free space by an oscillating interaction .
the strength of interaction studied in ref .
@xcite is much larger than that of ours . in our parameter regime
, a droplet can be stabilized adiabatically by simple ramp functions with realistic dissipation . on the other hand , in ref @xcite a droplet
is stabilized without dissipation by following a complicated ramp scheme to reach a final state .
this work was supported by the special coordination funds for promoting science and technology , a 21st century coe program at tokyo tech `` nanometer - scale quantum physics '' , and a grant - in - aid for scientific research ( grant no .
15340129 ) from the ministry of education , science , sports , and culture of japan .
an illuminating account of this stabilizing mechanism is found in : r. p. feynman , r. b. leighton , and m. l. sands , _ the feynman lectures on physics _ , ( addison - wesley , massachusetts , 1964 ) , vol .
ii , chap . 29 . | we show that a droplet of a bose - einstein condensate can be dynamically stabilized in free space by rapid oscillations of interatomic interactions between attractive and repulsive through , e.g. , the feshbach resonance .
energy dissipation , which is present in realistic situations , is found to play a crucial role to suppress dynamical instabilities inherent in nonlinear nonequilibrium systems . |
a spinor bose - einstein condensate is a system that possesses both superfluid and magnetic order @xcite .
experiments are able to explore transitions between phases of different magnetic order by controlling the quadratic zeeman energy ( @xmath0 ) that shifts the magnetic sublevels @xcite .
such a system presents a rich playground for studies of non - equilibrium phenomena , including defect formation related to the rate that the phase transition is crossed @xcite , through to the late - time coarsening dynamics describing how the small domains produced by the quench anneal towards the equilibrium state @xcite .
studies have largely focused on the spin properties of the system , leaving the superfluid dynamics unexplored .
we note that in binary condensates , which can be regarded as a pseudo spin-@xmath1 system , there has been work on both spin coarsening @xcite and aspects of the superfluid behavior ( e.g. analyzed via kinetic energy spectra @xcite ) .
the development of various techniques for measuring spin @xcite and superfluid correlations ( e.g. see @xcite ) demonstrate that it is possible to study the evolution of both types of order in spinor experiments .
the relationship between spin and superfluid order has been considered in the equilibrium properties of spinor condensates .
for example , in ref . @xcite it was found that for a large spin - dependent interaction , ferromagnetism emerges at a higher temperature than condensation . here
we study the superfluid and spin ordering dynamics of a ferromagnetic spin-1 condensate quenched into a easy - axis ferromagnetic phase , where the spin order preferentially aligns ( or anti - aligns ) along the quantization axis defining the quadratic zeeman energy shift .
initially small domains form , but grow as time passes . at sufficiently late times
this coarsening dynamics can enter a universal scaling regime @xcite : the correlation function of the order parameter collapses to a universal ( time - independent ) form when space is scaled by a characteristic length @xmath2 , which yields the dynamic critical exponent @xmath3 . here
we find that the spin and superfluid order exhibit universal scaling , but do not develop identically .
the spin order spreads more rapidly across the system than the superfluid order , with different dynamic critical exponents for each type of ordering . in this paper
we consider two types of quenches . the first ( standard ) quench ( e.g. see @xcite ) is to start with an equilibrium non - magnetic ( polar ) condensate at large @xmath0 and then to suddenly quench to a negative @xmath0 value where the easy - axis magnetic phase is the ground state .
we also propose a second type of quench that utilizes the atomic physics toolbox of coherent manipulations : this differs from the first quench in that a @xmath4 spin rotation is applied to the initial state immediately prior to @xmath0 being quenched .
both quenches produce the same late time coarsening behavior , but we show that the second quench has less heating and the early - time dynamics ( where local order develops ) concludes more rapidly , suggesting this may be more suitable for experiments where long times are difficult to access .
we finally consider the areas of the individual spin domains that form after the quench and how these evolve in time .
our results show that once the domains develop in the system their areas vary over many orders of magnitude . at sufficiently long times
the domain area distribution decays as a power law @xmath5 for sufficiently large @xmath6 , where @xmath6 is the domain area .
this can be related to ideas of percolation theory @xcite , as has also been pointed in in recent work on the immiscibility phase transition in a binary condensate @xcite .
the outline of the paper is as follows . in sec .
[ secformalism ] we introduce the dynamical formalism used to simulate the quench dynamics .
we introduce the two different quench types considered and discuss the role and nature of the unstable excitations that drive the initial phase transition dynamics .
our main results are presented in sec .
[ secresults ] .
first we consider the early - time dynamics to emphasize the differences in how local spin order emerges for the two quench types .
we then turn to considering the spin and superfluid order , as described by the relevant two - point correlation functions .
we demonstrate that in the late time dynamics these correlation functions exhibit dynamical scaling , albeit with different growth laws .
then we compute the spin domain areas , extract the domain size distribution and investigate the scaling of this distribution as a function of domain size and time .
finally we conclude in sec .
[ secconclude ] .
the system we consider is a homogeneous quasi - two - dimensional ( quasi-2d ) spin-1 condensate described by the hamiltonian @xcite@xmath7\!.\end{aligned}\ ] ] here @xmath8 is a three component spinor describing the condensate amplitude in the three spin hyperfine sublevels ( @xmath9 ) and @xmath0 is the quadratic zeeman shift arising from the presence of an external field along @xmath3 .
the interactions are described by a density dependent term @xmath10 and a spin - density dependent term @xmath11 , where @xmath12 and @xmath13 are density - dependent and spin - dependent coupling constants , @xmath14 is the total number density , and @xmath15 is the spin density , with @xmath16 being the spin-1 matrices .
the dynamics of the system is described by the spin-1 gross - pitaevskii equation ( gpe ) @xmath17 for the system to be mechanically stable we require @xmath18 , and we additionally restrict our attention here to the case of ferromagnetic interactions , i.e. @xmath19 , as realized in @xmath20rb condensates @xcite . in spinor condensate experiments
the quasi-2d regime has been realized by using a trapping potential with tight confinement in one direction ( e.g. see @xcite ) .
our interest is in homogeneous systems where the phase transition dynamics are simpler , noting that recent experiments have realized flat - bottomed optical traps for this purpose @xcite ( also see @xcite ) .
magnetic phase diagram for a ferromagnetic spin-1 condensate with zero @xmath3-magnetization .
the spheres show the direction of magnetization in the three states .
( a ) for @xmath21 the magnetization lies along the @xmath22 axis and the state is termed _ easy - axis_. ( b ) for @xmath23 the magnetization lies in the transverse ( @xmath24-@xmath25 ) plane and the state is termed _ easy - plane_. quenches to this phase are not analyzed here but can be found in refs .
( c ) for @xmath26 the @xmath27 levels are unoccupied and the system is unmagnetized .
this state is termed _
polar_. the directed line on the phase diagram indicates the quench we consider in this paper , where the zeeman energy is suddenly reduced from an initial value @xmath28 where the ground state is polar , to a value @xmath21 where the system favours easy - axis spin ordering.,scaledwidth=50.0% ] the ground state of eq .
( [ spinh ] ) depends on the value of @xmath0 ( e.g. see @xcite ) relative to the characteristic spin energy @xmath29 , where @xmath30 is the ( uniform ) condensate density . a schematic phase diagram and representation of the ground states important to this paper
are shown in fig .
[ phasediag ] .
there are two ferromagnetic phases ( a ) and ( b ) , which differ in their symmetries , and a non - magnetized polar phase ( c ) .
of the two ferromagnetic phases we only consider the easy - axis phase here as this phase admits well - defined spin domains .
we briefly review the relevant order parameters for our simulations , which we express in the form @xmath31 where @xmath32 is a global ( superfluid ) phase , and @xmath33 is a normalized spinor ( i.e. @xmath34 ) . for high values of the quadratic zeeman energy ( @xmath26 ) the the ground state is the non - magnetized polar phase , with normalized spinor @xmath35 for which @xmath36 .
this state , or a spin rotated form of this state , is the initial condition for our quench .
our interest is how the system then reorders when @xmath0 is quenched to a negative value , where the ground state is an easy - axis ferromagnetic phase with normalized spinor @xmath37 corresponding to magnetization of + 1 or -1 along @xmath3 , respectively .
these states break the @xmath38 ( @xmath3-reflection ) symmetry of the hamiltonian .
we now introduce the two types of quench we use to transition the system from the polar to the easy - axis phase . + * quench 1 ( q1 ) * : in this quench the quadratic zeeman energy is simply set to a negative value at @xmath39 .
here the polar state is dynamically unstable and decays , initially by developing transverse magnetization . in this case
an energy per particle of @xmath40 is liberated ( i.e. the excess energy that the polar state has over the easy - axis ferromagnetic state for @xmath21 ) , and available to heat the system . + * quench 2 ( q2 ) * :
in this quench a spin rotation of @xmath41 , is performed producing a so - called anti - ferromagnetic state : @xmath42 after the spin rotation has been preformed the quadratic zeeman energy is immediately set to a negative value .
we note that this rotation ( using rf pulses ) and quench to negative @xmath0 in a spin-1 experiment has been reported in ref .
@xcite . for this initial condition the energy per particle available to heat
the system is smaller than for q1 and is independent of @xmath0 ( for @xmath21 ) , i.e. @xmath43 ( color online ) bogoliubov spectra for the post - quench initial states emphasizing the dynamically unstable modes .
spectra of the @xmath44 initial state produced by the q1 quench at ( a ) @xmath45 and ( b ) @xmath46 .
spectra of the @xmath47 initial state produced by the q2 quench at ( a ) @xmath45 and ( b ) @xmath46 .
the lines indicate the real parts of the energies , while the dashed lines indicate the imaginary part .
the phonon ( @xmath48 ) and magnon ( @xmath49 ) excitation branches are labelled in the plots , noting that the magnon branches are degenerate in ( a ) and ( b ) .
these results are for the case @xmath50 .
, scaledwidth=50.0% ] insight into the initial post - quench dynamics is also provided through understanding the bogoliubov quasi - particle excitations of the initial state . for a spin-1 system
there are three excitation branches , which can be typically classified as a phonon branch and two magnon ( or spin - wave ) branches . in discussing these modes
we adopt the naming conventions used in @xcite and refer to the phonon branch as @xmath51 and the magnon branches as @xmath52 .
the instabilities of the initial state following the quench are revealed by the imaginary parts in the quasi - particle energies @xmath53 .
as our system is mechanically stable ( and @xmath54 ) the phonon branch always remains stable , and exhibits a rapidly rising linear spectrum @xmath55 , where @xmath56 is the speed of sound . for the q1 quench the initial non - equilibrium state ( p - phase ) at @xmath21
has a bogoliubov spectrum with two unstable magnon branches . for our case of zero magnetization along @xmath3 these magnon branches and a condensate with zero @xmath3-magnetization .
] are degenerate with the dispersion relation @xmath57 where @xmath58 . for @xmath21
these magnon branches have dynamically unstable modes ( i.e. the quasi - particle energies @xmath59 are imaginary ) for the range of @xmath60 values @xmath61 these magnon modes have amplitude in the @xmath62 sub - levels and lead to the decay of the polar phase by the development of transverse magnetization . as the value of @xmath0 changes the character of the unstable modes remains qualitatively the same [ see figs .
[ figspectrum](a ) and ( b ) ] , with the quantitative change that the unstable modes shift to larger @xmath60 as @xmath0 becomes more negative . for the q2 quench the initial state ( af - phase )
has two magnon branches of qualitatively different behavior @xmath63 the @xmath64 branch consists of modes that have amplitude in the @xmath62 sublevels giving rise to longitudinal magnetic fluctuations , while the @xmath65 branch modes have amplitude in the @xmath66 sublevel giving rise to transverse fluctuations .
the unstable modes for the @xmath64 branch are independent of @xmath0 and occur for the @xmath60 range @xmath67 e.g. , see figs .
[ figspectrum](c ) and ( d ) .
the @xmath65 branch has a finite region of unstable modes @xmath68 for @xmath69 ( noting that the lower bound is 0 for @xmath21 ) . at @xmath70
[ see figs .
[ figspectrum](d ) ] , this branch becomes stable , and the condensate dynamics is driven entirely by the unstable modes in the @xmath64 branch .
we simulate the dynamics using the spin-1 gpe ( [ spingpes ] ) with noise added to the initial state to seed the growth of symmetry breaking domains . for the initial state of the q1 quench we take @xmath71 where @xmath72 is a uniform ( zero - momentum ) polar condensate and
@xmath73 is a small noise field used to seed the dynamical instabilities that occur after the quench .
the precise form of noise used is based on the truncated wigner formalism @xcite and is described in ref .
@xcite . for the q2 quench the spin rotation @xmath74
is applied to state ( [ eqpic ] ) to yield the initial condition .
we use a condensate areal density of @xmath75 , where @xmath76 is the spin healing length .
we evolve the spin-1 gpe for the spinor field @xmath77 on a 2d square grid with dimensions @xmath78 covered by an @xmath79 grid of equally spaced points . here
most of our results are calculated on a grid of linear size @xmath80 with @xmath81 points .
we evolve the spin-1 gpe using a fourth - order symplectic method that uses fast fourier transforms to evaluate the kinetic energy operators with spectral accuracy @xcite .
the quadratic zeeman energy is set at the final quench value @xmath82 for the duration of the simulation dynamics , so that the quench is effectively instantaneous at @xmath39 from the initial condition for the case of the q1 and q2 quench , respectively . in order for the post - quench dynamics to become universal
it is necessary to simulate the system over many spin times @xmath83 .
indeed , here we evolve out to times of up to @xmath84 , and on such long time - scales the use of a highly accurate symplectic algorithm ensures we obtain accurate solutions that conserve energy and magnetization .
all the results presented here are for @xmath85 .
this choice is made to ensure our interactions are in a similar regime to the binary condensate immiscibility simulations of ref .
@xcite , noting that the mapping of binary condensate parameters onto effective density and spin dependent interactions is discussed in ref .
we have also performed simulations for @xmath86 to verify that our results are qualitatively unchanged with a smaller spin - dependent interaction .
growth of local magnetization following the q1 and q2 quench to ( a ) @xmath45 and ( b ) @xmath46 . in all cases
the ( local ) magnetization is calculated as a spatial average over the system at each time , i.e. @xmath87 for @xmath88 ( plain lines ) and @xmath89 ( lines with symbols ) .
other parameters as in fig .
[ figspectrum ] . , scaledwidth=50.0% ] for both the q1 and q2 quenches the initial state is unmagnetized , but the unstable modes lead to the spin density growing exponentially .
the initial post - quench dynamics depends upon the kind and depth ( i.e. @xmath0 ) of the quench .
we take the _ early - time _ regime to be the initial time period over which the longitudinal magnetization locally develops . for our parameters ( see fig . [ figstdynamics ] ) this regime extends up to @xmath90 . for times after this the longitudinal magnetization is the dominant component of the magnetization , and the dynamics can be then be analysed in terms of spin domain coarsening , which is the focus of the next subsection .
we provide examples of the early - time dynamics of the local magnetization in fig .
[ figstdynamics ] for the q1 and q2 quenches to two different final values of @xmath0 .
we generally observe that for the q2 quench the longitudinal magnetization develops more rapidly and approaches a larger value at late times , compared to the q1 case .
this second observation can be understood as arising from the additional heating in the q1 quench ( as discussed in sec .
[ secphasesquenches ] ) which tends to reduce the ground state order .
notably , for even deeper quenches ( @xmath91 ) than presented here we find that local longitudinal magnetization obtained after q1 quench decreases as the final value of @xmath0 becomes more negative , while the q2 quench is insensitive to the value of @xmath0 . in fig .
[ figdomains ] we compare the magnetized domains formed near the end of the early - time dynamics .
these results emphasize that the q2 quench produces larger and less noisy domains following the quench .
longitudinal spin density at a time of @xmath92 after the ( a ) q1 and ( b ) q2 quench to @xmath45 .
the white line in each image indicates the spin correlation length @xmath93 defined in the text , with a value of ( a ) @xmath94 and ( b ) @xmath95.the spin density is only shown over a subregion of the simulation , with other parameters as in fig .
[ figspectrum ] .
, scaledwidth=50.0% ] we can obtain qualitative insight into other aspects of the dynamics in fig . [ figstdynamics ] through considering the unstable excitations of the initial states , discussed in sec
. [ secunstablemodes ] . for the q1 quench to @xmath45 [ fig .
[ figstdynamics](a ) ] and @xmath96 [ fig .
[ figstdynamics](b ) ] the dynamics proceeds in a qualitatively similar manner .
the two degenerate magnon branches [ eq .
( [ eqmagnonp ] ) ] have unstable sections , and the growth of these excitations causes the transfer of atoms into the @xmath62 sublevels ( i.e. spin - mixing ) , leading to the exponential growth of transverse magnetization .
after this process saturates ( also see @xcite ) the longitudinal magnetization then develops as the transverse magnetization decays . for the q2 quench the dynamics depends more strongly on the final @xmath0 value . for @xmath69 [ e.g. fig .
[ figstdynamics](a ) ] both the magnon branches [ eqs . ( [ eqmagnonaf1 ] ) and ( [ eqmagnonaf2 ] ) ]
are unstable . as discussed in sec .
[ secmagnonsq2 ] , these magnons have transverse and longitudinal magnetic character and the simulation results show that both components of magnetization initially grow in a similar manner until saturation . for a deeper quench to @xmath91 , only the longitudinal magnon branch [ eq .
( [ eqmagnonaf1 ] ) ] is unstable , and we see [ fig .
[ figstdynamics](b ) ] that the longitudinal magnetization develops with negligible growth in transverse magnetization .
this case has similarities to the immiscibility transition in a binary condensate @xcite , in that the initial state occupies the @xmath62 sublevels and the dynamics causes the condensate density in these two sublevels spatially seperate .
in contrast for shallow quenches ( @xmath69 ) the unstable magnon causing transverse magnetization allows spin mixing to transfer some of the @xmath62 atoms into the @xmath66 sublevel . in the early - time dynamics local longitudinal magnetization
is established by small domains forming of microscopic ( @xmath97 ) dimensions ( cf .
[ figdomains ] ) .
these domains then grow by coarsening dynamics .
the average spatial distribution of domains in the system is captured by computing the relevant two - point correlation function .
here we consider the relevant spin correlation functions @xmath98 where @xmath99 is the transverse magnetization density and the average is taken at a time @xmath100 after the quench .
the functions @xmath101 and @xmath102 characterize longitudinal and transverse magnetization , respectively . for our case of quenches to the easy - axis phase @xmath101
is the ( spin ) _ order parameter
_ correlation function of interest for coarsening ( e.g. see @xcite ) .
we also consider @xmath102 to demonstrate the contrasting behavior for transverse magnetisation which does not order . here
we also address the superfluid order as a second type of order that develops in the system .
this order is characterized by the global phase coherence of the atomic fields , and for the easy - axis phase this is captured by the correlation function @xmath103 in expressions ( [ gz])-([g0 ] ) we have utilized translational invariance to spatially average the correlation function .
additionally , we use spatial isotropy of the correlation functions to perform an angular average over all points at a distance @xmath104 and further improve statistical sampling by averaging over 32 simulation trajectories conducted with different initial noise .
correlation functions for q1 quench to @xmath45 .
the longitudinal spin correlation function ( a ) at various times and ( b ) scaled by the length scale @xmath105 to reveal correlation function collapse .
the superfluid correlation function ( c ) at various times and ( d ) scaled by @xmath106 .
( e ) the evolution of the length scales .
results with symbols out to @xmath107 are for simulations with @xmath108 and @xmath81 points , whiles the dashed lines out to @xmath84 are from simulations with @xmath109 with @xmath81 points . , scaledwidth=50.0% ]
correlation functions for q2 quench to @xmath45 .
the longitudinal spin correlation function ( a ) at various times and ( b ) scaled by the length scale @xmath105 to reveal correlation function collapse . the superfluid correlation function ( c ) at various times and ( d ) scaled by @xmath106 .
( e ) the evolution of the length scales .
results with symbols out to @xmath107 are for simulations with @xmath108 and @xmath81 points , whiles the dashed lines out to @xmath84 are from simulations with @xmath109 with @xmath81 points . , scaledwidth=50.0% ] the temporal evolution of the @xmath101 and @xmath110 correlation functions is shown in figs . [ figcorrq1 ] and [ figcorrq2 ] for both types of quench .
the length scale over which the correlation function decays can be taken to define a characteristic domain size .
for @xmath101 we take this length scale @xmath105 to be the first zero crossing of @xmath111 .
the @xmath110 and @xmath102 ( not shown ) correlation functions do not usually have a zero crossing so we take @xmath106 to be the position where the correlation peak in @xmath110 decays to 0.25 of its maximal value ( occurring at @xmath112 ) , and using the same procedure we define the length scale @xmath113 for @xmath102 . as time progresses the length scales @xmath93 and @xmath114
are seen to grow as the longitudinal spin order and superfluid order extends over large regions [ figs .
[ figcorrq1](e ) and [ figcorrq2](e ) ] .
in contrast the length scale @xmath113 , which is not associated with the ground state order of the system , does not grow and remains comparable to the microscopic length scale @xmath115 .
we investigate whether the system exhibits dynamic scale invariance , i.e. whether correlations of the order parameters at late times collapse onto a single universal curve @xmath116 when lengths are scaled by a characteristic length scale @xmath117 @xmath118 using the length scales @xmath105 and @xmath106 we demonstrate correlation function collapse for sufficiently late times in both the spin order [ see figs . [ figcorrq1](b),[figcorrq2](b ) ] and the superfluid order [ see figs . [ figcorrq1](d ) , [ figcorrq2](d ) ] . for both quenches at late times
( @xmath119 ) @xmath93 is seen to grow as @xmath120 as found previously @xcite .
the superfluid order ( length scale @xmath114 ) grows in a similar manner to @xmath93 for intermediate times , but eventually ( @xmath121 ) grows more slowly .
our best fit to the exponent gives @xmath122 .
this result is slower than the @xmath123 scaling expected from dissipative xy - model dynamics @xcite , where the logarithmic correction to the @xmath124 exponent originates from the presence of free vortices . for the case of a quasi-2d scalar condensate damle _
et al . _
@xcite found @xmath125 , albeit for a higher temperature regime ( close to the critical temperature for vortex unbinding ) and using small simulation grids .
recent results of karl _ et al . _
@xcite reveal even slower dynamics ( @xmath126 ) for particular initial condition of a quasi-2d condensate .
we have performed simulations for deeper quenches to @xmath46 and for the case of a larger interaction ratio @xmath127 to verify that the spin and superfluid ordering occurs with the same exponents .
we noted in sec .
[ secearlytime ] that local order developed faster in the early - time dynamics for the q2 quench relative to q1 . comparing figs .
[ figcorrq1](e ) and [ figcorrq2](e ) we also observe that for times @xmath128 the values of @xmath93 and @xmath114 for the q2 quench are appreciably larger that those of the q1 quench at the same time , and grow at a more steady rate [ the @xmath93 correlation lengths at @xmath129 are also shown in fig .
[ figdomains ] ] .
the @xmath110 correlation function at @xmath130 in fig . [ figcorrq2](c ) shows that the system has long - range superfluid order at early times ( i.e. @xmath110 approaches a constant non - zero value for large @xmath104 ) .
this order eventually decays ( @xmath121 ) and the @xmath110 behavior is similar to that observed in the q1 quench .
the long - range order is established by the initial spin rotation used in q2 quench , which rotates the coherent polar state into @xmath131 sublevels . in the postquench dynamics ,
this coherence is rapidly destroyed by the immiscibility dynamics leading to spin domain formation . in figs . [
figcorrq1](e ) and [ figcorrq2](e ) we also show results for a calculation with a grid of @xmath81 and @xmath109 ( i.e. with a grid point spacing about 60% larger than the other results ) .
the greater spatial range in this case allows us to simulate longer ( we need to ensure that @xmath132 to avoid finite size effects ) .
the results from these calculations are shown as dashed lines and are seen to lie on top of the main results for characteristic length scales associated with the order parameter ( i.e. @xmath93 and @xmath114 ) , but differ appreciably from the main results for the behavior of @xmath113 .
this nicely demonstrates the universality of the order parameter coarsening dynamics , while emphasizing that the transverse magnetization is a disordered degree of freedom that is sensitive to the microscopic details ( in this case it is largely determined by the thermalization of the @xmath66 spin waves , see @xcite ) .
the binary character of the easy - axis spin order allows clear identification of spin domains ( e.g. see fig .
[ figdomains ] ) . in this subsection
we consider the properties of the domains produced as characterized by the domain area distribution @xmath133 where @xmath134 is the average number of domains of area @xmath6 to @xmath135 per unit system area at time @xmath100 . in simulations we define positive domains as connected spatial regions where the @xmath22 spin density is positive to be part of a positive domain . ] , and all locations satisfying this are set to a value of + 1 , and all other locations are set to a value of zero . on this _ binary image _ we apply the hoshen - kopelman algorithm @xcite to label the clusters of contiguous + 1 cells .
similarly , a binary image can be constructed for the negative domains ( with @xmath136 ) . analyzing the combined set of positive and negative magnetized domains at time
@xmath100 we compute their areas , and bin this to determine @xmath133 . to improve statistics we use 32 trajectories to average for @xmath133 .
domain area distributions for the ( a)-(b ) q1 and ( c)-(d ) q2 quenches .
the domains area distribution is shown at early - times as a function of ( a ) , ( c ) area and of ( b ) , ( d ) scaled - area @xmath137 , where @xmath93 is the longitudinal spin correlation length evaluated as discussed in sec .
[ seccoarsening ] .
the insets in ( b ) and ( d ) use the same axis range as the main plots , and indicate results at later times for comparison . , scaledwidth=50.0% ] our results for @xmath133 at various times are shown in fig .
[ figdomainareas ] .
while the late time behavior of the domain distributions is reasonably similar [ see insets to figs .
[ figdomainareas](b ) and ( d ) ] , the early - time behavior is strikingly different .
notably , even for times as short as @xmath138 in the q2 case , @xmath133 is seen to extend out to areas of order the system size [ i.e. @xmath139 , see fig .
[ figdomainareas](c ) ] , which indicates that the system can have a percolating domain that connects across the system .
in contrast , such percolating domains for the q1 quench only develop on long time scales [ i.e. @xmath140 in fig . [ figdomainareas](a ) ] .
this difference likely originates from the nature of the @xmath141 magnon branch for the q2 quench which has unstable wavelengths extending from @xmath97 up to the system size [ e.g. see fig .
[ figspectrum](c ) ] . in comparison
the unstable modes for the q1 quench are restricted to a finite wavelength range around @xmath142 [ e.g. see fig . [
figspectrum](a ) ] . we can apply dynamic scaling to the domain size distribution . since the @xmath101 correlation function grows as @xmath105 , that the domain areas should grow as @xmath143 , and if dynamic scaling holds we would expect @xmath144 to be a universal time - independent distribution .
we perform this scaling in figs .
[ figdomainareas](b ) and ( d ) . at late - times
( insets to those figures ) we see reasonable collapse of the area distribution , however at earlier - times ( @xmath145 ) only the q2 quench exhibits scaling collapse over a wide range of areas .
we also indicate on the scaled results the @xmath146 power law , which provides a reasonable fit to the distribution for sufficiently large domains . such power law behavior in domain size ( or cluster size " ) distribution
is predicted in percolation theory near the percolation transition with @xmath147 in 2d .
normally in percolation theory the occupation probability ( @xmath148 ) is varied to pass through the transition .
for the spin-1 system this is controlled by the total ( conserved ) @xmath3-magnetization , and for our results ( where the total @xmath3-magnetization is zero ) the probability of a location being in a positive ( or negative ) domain is @xmath149 , which is approximately the percolation threshold condition in 2d . in order to explore the percolation transition it would be necessary to vary the total @xmath3-magnetization , which would in turn vary the relative portion of two domain types .
such a study has been performed using simulations of a segregating binary condensate , where the effective magnetization is controlled by varying the relative portion of the two atomic species @xcite .
this study presented strong evidence for a percolation threshold at @xmath150 .
we have considered the spin and superfluid ordering of a spin-1 condensate . we have found that in the coarsening dynamics these two types of order exhibit dynamic scaling , but develop with different dynamic critical exponents .
we find that ( and as previously observed in refs .
@xcite ) the spin - order develops with a growth law of @xmath151 , consistent with a binary fluid in the inertial hydrodynamic regime @xcite .
our new results here concern the superfluid ordering , which we find to grow much more slowly as @xmath152 . indeed , our simulations out to @xmath84 reveal that @xmath114 is almost an order of magnitude smaller than @xmath93 [ see figs . [ figcorrq1](e ) and [ figcorrq2](e ) ] .
future work will involve developing an understanding for the slower superfluid growth rate , and whether aspects of the spin domains ( e.g. shedding of various types of vortices at the spin domain walls , or other types of topological defects ) play a role .
another direction for investigation is to consider our observed dynamics in the context of pre - thermalization and turbulence theory , which have proven fruitful directions of research for the immiscibility transition in a binary condensate @xcite ( also see @xcite ) .
in addition to the standard q1 quench , in which the quadratic zeeman energy is suddenly changed , we propose and simulate a second type of quench ( q2 ) in which a spin rotation is applied prior to the zeeman energy being changed . in the late - time dynamics both quenches
exhibit the same ordering dynamics , demonstrating the universal nature of the coarsening dynamics
. however , the q2 quench has favourable properties for exploring domain formation : ( i ) the domains form more rapidly and with less heating ; ( ii ) the domains percolate at early times .
these properties may make the q2 quench better suited to experiments ( e.g. , with @xmath20rb condensates ) , where the small spin - dependent interaction makes it challenging to explore long - time post - quench dynamics .
for example , some of the short - time predictions we have made for domain formation and the behavior of the spin domain area distribution should be feasible to study in current experiments .
the authors acknowledge useful discussions and assistance from h. takeuchi , l. williamson , and l. symes .
pbb acknowledges the contribution of nz escience infrastructure ( nesi ) high - performance computing facilities , and support from the marsden fund of the royal society of new zealand .
50ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ] + 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty \doibase http://dx.doi.org/10.1016/j.physrep.2012.07.005 [ * * , ( ) ] link:\doibase 10.1103/revmodphys.85.1191 [ * * , ( ) ] http://dx.doi.org/10.1038/nature05094 [ * * , ( ) ] link:\doibase 10.1103/physreva.79.043631 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.107.195306 [ * * , ( ) ] link:\doibase 10.1103/physreva.84.063625 [ * * , ( ) ] link:\doibase 10.1103/physreva.89.033631 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.98.160404 [ * * , ( ) ] link:\doibase 10.1103/physreva.76.043613 [ * * , ( ) ] link:\doibase 10.1103/physreva.75.013621 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.99.120407 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.99.130402 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.110.165301 [ * * , ( ) ] link:\doibase 10.1103/physreva.90.033604 [ * * , ( ) ] link:\doibase 10.1103/physrevb.76.104519 [ * * , ( ) ] link:\doibase 10.1103/physreva.81.053612 [ * * , ( ) ] link:\doibase 10.1103/physreva.88.013630 [ * * , ( ) ] link:\doibase 10.1103/physreva.91.053609 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.116.025301 [ * * , ( ) ] link:\doibase 10.1103/physreva.94.023608 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.113.095702 [ * * , ( ) ] link:\doibase 10.1103/physreva.88.063615 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.95.050401 [ * * , ( ) ] http://dx.doi.org/10.1038/35003132 [ * * , ( ) ] link:\doibase 10.1126/science.1138807 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.102.170401 [ * * , ( ) ] link:\doibase 10.1126/science.1258676 [ * * , ( ) ] http://dx.doi.org/10.1038/ncomms7162 [ * * ( ) ] link:\doibase 10.1103/physreva.81.053617 [ * * , ( ) ] link:\doibase 10.1080/00018739400101505 [ * * , ( ) ] @noop _ _ ( , ) link:\doibase 10.1103/physreva.92.043608 [ * * , ( ) ] link:\doibase 10.1007/s10909 - 016 - 1543 - 7 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.81.742 [ * * , ( ) ] link:\doibase 10.1143/jpsj.67.1822 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.92.140403 [ * * , ( ) ] http://stacks.iop.org/0295-5075/36/i=1/a=007 [ * * , ( ) ] link:\doibase 10.1103/physreva.84.023606 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.115.015301 [ * * , ( ) ] link:\doibase 10.1103/physreva.90.053616 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physreve.93.053309 [ * * , ( ) ] link:\doibase 10.1103/physreva.58.4836 [ * * , ( ) ] link:\doibase 10.1103/physreve.47.1525 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.84.1503 [ * * , ( ) ] link:\doibase 10.1103/physreva.54.5037 [ * * , ( ) ] @noop ( ) , link:\doibase 10.1103/physrevb.14.3438 [ * * , ( ) ] link:\doibase 10.1103/physreva.31.1103 [ * * , ( ) ] link:\doibase 10.1103/physreva.93.033620 [ * * , ( ) ] | in this paper we study the coarsening dynamics of a spinor condensate quenched into an easy - axis ferromagnetic phase by a sudden change in the quadratic zeeman energy .
we show that applying a spin rotation prior to changing the zeeman energy accelerates the development of local order and reduces heating .
we examine the longitudinal spin ordering and the superfluid ordering of the system and show that the respective order parameter correlation functions exhibit dynamic scaling in the late time dynamics .
our results also demonstrate that these two types of order grow at different rates , i.e. with different dynamic critical exponents .
the spin domain area distribution is calculated and is shown to have power law scaling behavior expected from percolation theory . |
for tricritical phenomena , the highest dimension in which non - classical behaviour can be observed is @xmath0 @xcite .
consequently , 2d tricritical phenomena has been the subject of a large number of previous investigations , employing a wide variety of techniques , including series expansions @xcite , mean - field theory @xcite , renormalisation group ( rg ) @xcite , transfer matrix @xcite , monte carlo simulations @xcite and monte carlo rg methods @xcite . to date
, however , this interest has focused almost exclusively on lattice - based spin models such as the blume - capel model or the spin-@xmath1 next - nearest - neighbour ising model . in this paper
, we report the first detailed simulation study of 2d tricritical behaviour in an off - lattice spin fluid model . the model we consider is a simplified representation for a liquid of two - state molecules , and has been the subject of a number of previous studies in both its classical and quantum regimes @xcite . in the present work , however , we shall consider only the classical limit , for which the configurational energy is given by : @xmath2 with @xmath3 and where @xmath4 is chosen to be a hard disk potential with diameter @xmath5 .
the distance - dependent spin coupling parameter @xmath6 is assigned a square well form : @xmath7 the phase diagram of this model is expected to correspond to the situation depicted schematically in figure [ fig : pdschem ] . for high temperatures , there exists a line of ising critical points ( the so - called ` critical line ' ) separating a ferromagnetic fluid phase from a paramagnetic fluid phase .
the particle density varies continuously across this line .
as one follows the critical line to lower temperatures , however , the size of the particle density fluctuations grows progressively . precisely at the tricritical point ,
the fluctuations in both the particle density and magnetisation are simultaneously divergent .
lowering the temperature still further results in a phase separation between a low density paramagnetic gas and a high density ferromagnetic liquid .
for subtricritical temperatures , the phase transition between these two phases is first order .
owing to the interplay between the density and magnetisation fluctuations , the tricritical properties of the spin fluid system are expected to differ qualitatively from those on the critical line .
general universality arguments @xcite predict that for a given spatial dimensionality , fluids with short - ranged interactions should exhibit the same tricritical properties as lattice - based spin systems .
however , since fluids possess a continuous translational symmetry that lattice models do not , this proposal needs be checked .
additionally , experience with ` ordinary ' ( ising ) critical behaviour in simple fluids such as the lennard - jones fluid @xcite , shows that the reduced symmetry of fluids can profoundly influence certain non - universal aspects of the critical properties .
principal among these , are the directions of the relevant scaling fields associated with the fixed point , and the distribution functions of observables such as the particle density and energy .
it is thus of interest to assess the extent of these ` field - mixing ' effects in the tricritical fluid and to compare it with the situation at the liquid - vapour critical point of simple fluids .
an accurate determination of the universal forms of the tricritical scaling operator distribution is also of considerable value .
such distributions are _ unique _ to a universality class and hence knowledge of their forms would be of considerable practical utility to future simulation studies of 2d tricriticality , serving as they do to simplify the computational task of locating the tricritical parameters .
moreover , as we shall see , the forms of the scaling operator distribution functions can impart important physical insight into the nature of the tricritical fluctuations .
our paper is broadly organised as follows . in section [ sec : meth ] we describe the finite - size scaling methods and other computational techniques employed in the study .
we then proceed in section [ sec : res ] to detail the application of these techniques to monte carlo simulations of both the 2d spin fluid model described above , and the 2d blume - capel .
the simulations yield accurate estimates of the location of the tricritical point for both models , as well as the universal forms of the tricritical scaling operator distributions and the directions of the relevant scaling fields . in the case of the spin fluid model , the estimates for the tricritical point parameters
are compared with the results of a mean field calculation .
finally section [ sec : concs ] details our conclusions .
the techniques we employ in this work have been previously developed in the context of simulation studies of ising critical phenomena in a variety of fluid models , including a decorated lattice gas model @xcite , a lattice model for polymer mixtures @xcite , and both the two and three - dimensional lennard - jones fluids @xcite . in common with the strategy pursued in these previous works
, we have chosen to work within the grand canonical ensemble , use of which affords effective treatment of the particle density fluctuations which are a central feature of fluid critical behaviour . let us assume our system to be contained in a volume @xmath8 , with @xmath0 in the simulations to be described later .
the grand partition function is given by @xmath9 } \label{eq : pf}\ ] ] where @xmath10 is the particle number , @xmath11 is the inverse temperature , @xmath12 is the chemical potential and @xmath13 is the uniform applied magnetic field .
the observables of chief concern to the present study are the ( reduced ) particle density @xmath14 the configurational energy density ( which we express in units of @xmath15 ) @xmath16 and the magnetisation .
@xmath17 the coarse - grained behaviour of the system in the vicinity of the tricritical point is controlled by three relevant scaling fields @xcite which we denote @xmath18 and @xmath19 . in general these scaling fields
are each expected to comprise linear combinations of the three thermodynamic fields @xmath20 and @xmath13 @xcite . for the spin fluid model considered in this work ,
however , the configurational energy is invariant with respect to sign reversal of the spin degrees of freedom .
this special symmetry implies that the tricritical point lies in the symmetry plane @xmath21 , and that the scaling field @xmath19 coincides with the magnetic field @xmath13 , being orthogonal to the @xmath22 plane containing the other two scaling fields , @xmath23 and @xmath24 .
thus we can write @xmath25 where the subscript @xmath26 signifies tricritical values and the parameters @xmath27 and @xmath28 are system - specific ` field mixing ' parameters that control the directions of the scaling fields in the @xmath12@xmath29 plane .
the scaling fields @xmath23 and @xmath24 are depicted schematically in figure [ fig : pdschem](b ) .
one sees that @xmath23 is tangent to the coexistence curve at the tricritical point @xcite , so that the field mixing parameter @xmath28 may be identified simply as the limiting tricritical gradient of the coexistence curve .
the scaling field @xmath24 , on the other hand , is permitted to take a general direction in the @xmath12@xmath29 plane which does not necessarily have to coincide with any special direction of the phase diagram @xcite .
conjugate to each of the scaling fields are scaling operators , defined by the requirements @xmath30 from which it follows ( utilising equations [ eq : pf ] and [ eqn : scafldsa][eqn : scafldsc ] ) , that @xmath31 \\ { \cal e } & = & \frac{1}{1-sr}[u - r\rho ] \end{aligned}\ ] ] motivated by our experience with ordinary critical phenomena in simple fluids
@xcite we make the following finite - size scaling _
@xcite for the limiting ( large l ) near - tricritical distribution of @xmath32 @xmath33 where @xmath34 is a universal scaling function , the @xmath35 are non - universal metric factors and the @xmath36 are the standard tricritical eigenvalue exponents @xcite . precisely at the tricritical point ,
the tricritical scaling fields vanish identically and the last three arguments of equation [ eq : ansatz ] can be simply dropped , yielding @xmath37 where @xmath38 is a universal and scale invariant function characterising the tricritical fixed point . in what follows we
shall explicitly test the proposed universality of equation [ eq : ptri ] for the case of the spin fluid model , by obtaining the form of @xmath38 and comparing it with that for the tricritical 2d blume - capel model .
the monte - carlo simulations of the spin fluid model were performed using a metropolis algorithm within the grand canonical ensemble .
particle insertions and deletions were carried out using the prescription of adams @xcite . when attempting particle insertions the spin of the candidate particle was randomly assigned the value @xmath39 or @xmath40 with equal probability .
spin flip attempts were performed at the same frequency as insertion / deletion attempts . in order to facilitate efficient computation of interparticle interactions , the periodic simulation space of volume
@xmath41 was partitioned into @xmath42 cubic cells each of side @xmath43 , corresponding to the interaction range of the interparticle potential ( cf .
equation [ eq : pot ] ) .
this strategy ensures that interactions emanating from particles in a given cell extend at most to particles in the @xmath44 neighbouring cells .
we chose to study three system sizes corresponding to @xmath45 and @xmath46 , containing , at coexistence , average particle numbers of @xmath47 , @xmath48 and @xmath49 respectively .
for the @xmath45 and @xmath46 system sizes , equilibration periods of @xmath50 monte carlo transfer attempts _ per cell _
( mcs ) were utilised , while for the @xmath51 and @xmath52 system sizes up to @xmath53 mcs were employed .
sampling frequencies ranged from @xmath54 mcs for the @xmath55 system to @xmath56 mcs for the @xmath52 system .
production runs amounted to @xmath57 mcs for the @xmath55 and up to @xmath58 mcs for the @xmath52 system size . at coexistence
the average acceptance rate for particle transfers was approximately @xmath59 , while for spin flip attempts the acceptance rate was approximately @xmath60 . during the production runs ,
the joint probability distribution @xmath32 was obtained in the form of a histogram . in order to increase computational efficiency ,
the histogram extrapolation technique @xcite was employed .
use of this technique permits histograms obtained at one set of model parameters to be reweighted to yield estimates appropriate to another set of model parameters .
the method is particularly effective close to a critical point where , owing to the large fluctuations , a single simulation permits extrapolation over the entire critical region . as an aid to locating the tricritical point ,
the cumulant intersection method was employed @xcite .
the fourth order cumulant ratio @xmath61 is a quantity that characterises the form of a distribution @xcite .
it is defined in terms of the fourth and second moments of a given distribution @xmath62 the tricritical scale invariance of the distributions @xmath63 and @xmath64 , ( as expressed by equation [ eq : ptri ] ) , implies that at the tricritical point ( and modulo corrections to scaling ) , the cumulant values for all system sizes should be equal .
the tricritical parameters can thus be found by measuring @xmath61 for a number of temperatures and system sizes along the first order line , according to the prescription given below . precisely at the tricritical temperature
the curves of @xmath61 corresponding to the various system sizes are expected to intersect one another at a single common point . in seeking to confirm the proposed universality linking the tricritical point of the 2d spin fluid model to those of 2d lattice models , it is first necessary to determine the tricritical operator distribution functions for a simple lattice model . to this end
we have also performed a simulation study of the blume - capel model on a periodic square lattice , the hamiltonian of which is given by @xmath65 with @xmath66 . here @xmath13 is a uniform magnetic field and @xmath67 is the so - called ` crystal field ' . as with the 2d spin fluid model , the symmetry of the configurational energy under spin sign reversal implies that the tricritical point lies in the symmetry plane @xmath21 .
previous mcrg @xcite investigations place the tricritical point of the 2d blume - capel model at @xmath68 , while a more recent transfer matrix study @xcite gives @xmath69 . using these estimates as an initial guide
, we performed extensive monte carlo simulations of the model using a vectorised metropolis algorithm on a cray - ymp .
four system sizes of linear extent @xmath70 were studied and following equilibration , runs ranging from @xmath71 monte carlo sweeps ( mcs ) for the l=12 system , to @xmath72 mcs for the l=40 system were performed .
the quantities measured in the course of these runs were @xmath73 here we note that @xmath74 plays the same rle as the density in the spin fluid model , being discontinuous across the first order line but continuous on the critical line .
this fact is most clearly evident in the lattice gas representation of the blume - capel model , where the crystal field @xmath67 appears as a chemical potential @xcite . during the simulations ,
the joint distribution @xmath75 was collected in the form of a histogram . to determine the locus of the first order line ( and
hence locate the tricritical point in which it formally terminates ) , the coexistence symmetry criterion for the operator distribution @xmath76 was utilised .
this criterion is the analogue for asymmetric first order transitions , of the order - parameter distribution symmetry condition applicable to symmetric first order transitions such as that of the subcritical simple ising model @xcite . for a given temperature @xmath29 , the first order transition point
can thus be located by tuning the crystal field @xmath67 and the value of the field mixing parameter @xmath27 , within the histogram reweighting scheme , until the operator distribution @xmath76 is symmetric in @xmath77 .
the first order line and its finite - size analytical extension @xcite was determined in this way for temperatures in the range @xmath78@xmath79 , and for each of the @xmath80 system sizes .
the corresponding values of the cumulant ratio @xmath81 _ along _ this coexistence line are presented in figure [ fig : bc_ul ] as a function of the temperature . to within numerical uncertainties
the cumulant values for each system size intersect at a common temperature , which we estimate as @xmath82 .
the corresponding estimate for the tricritical field is @xmath83 .
clearly these values are in excellent agreement with the aforementioned estimates of previous studies . in the following subsection
we compare the measured forms of the tricritical operator distributions @xmath84 and @xmath64 of the 2d blume - capel , with those of the 2d spin fluid model .
the procedure for locating the tricritical point of the 2d spin fluid model followed the same pattern as that for the 2d blume - capel model described above , except that in the present case no prior estimates for the tricritical point were available .
it was thus necessary to search for the approximate location of the tricritical point by performing a number of short runs in which a temperature was chosen and the chemical potential tuned .
observations of the behaviour of the density from these short runs suggested that the tricritical point lay close to the parameters @xmath85 .
having obtained an approximate estimate of the location of the tricritical point , long runs were carried out for each of the three system sizes @xmath86 . as with the blume - capel model , the coexistence symmetry condition
was then applied to the operator distribution @xmath76 in conjunction with the histogram reweighting scheme , in order to determine the first order line and its analytic extension in the @xmath12@xmath29 plane .
the locus of this line is shown in figure [ fig : mu_t ] , while the measured values of @xmath87 along this coexistence line are shown in figure [ fig : fl_ul ] for the three system sizes @xmath86 .
clearly the curves of figure [ fig : fl_ul ] have a single well - defined intersection point , from which we estimate the tricritical temperature as being @xmath88 .
the associated estimate for the chemical potential is @xmath89 .
typical near - tricritical configurations for the @xmath52 system are shown in figure in figure [ fig : collapse ] we present the forms of the operator distributions @xmath90 , @xmath76 and @xmath64 corresponding to the designated values of the tricritical parameters .
the value of the field mixing parameter @xmath28 implicit in the definition of @xmath91 , was assigned the value @xmath92 , as obtained from the measured gradient of the phase boundary in the @xmath12@xmath29 plane at the tricritical point .
the value of the field mixing parameter @xmath27 was assigned , as previously described , so that @xmath76 satisfied the symmetry condition .
however , the resulting estimates of @xmath27 were found to exhibit a systematic finite - size dependence .
this effect has also been previously noted ( albeit with much reduced magnitude ) in a recent study of critical phenomena in the lennard - jones fluid , and is traceable to the finite - size dependence of the average critical energy @xcite .
for the three system sizes , @xmath86 , we found @xmath93 respectively .
interestingly , these values are at least an order of magnitude smaller than those measured at the critical point of the 2d and 3d lennard - jones fluid , a finding which we discuss further in section [ sec : concs ] .
this smallness implies that the scaling field @xmath24 almost coincides with the @xmath12 axis of the phase diagram .
also included in figure [ fig : collapse ] are the measured tricritical operator distributions for the 2d blume - capel model . in accordance with convention
, all the operator distributions have been scaled to unit norm and variance .
clearly in each instance and for each system size , the operator distributions collapse extremely well onto one another as well as onto those of the tricritical blume - capel model .
the measured scaling operator distributions also serve to furnish estimates of the eigenvalue exponents @xmath94 and @xmath95 characterising the three relevant scaling fields .
these exponents are accessible via the respective finite - size scaling behaviour of tricritical distributions of @xmath90 , @xmath76 and @xmath64 .
specifically , consideration of the scaling form [ eq : ptri ] shows that the typical size of the fluctuations in a given operator @xmath96 vary with system size like @xmath97 .
comparison of the standard deviation of a given operator distributions as a function of system size thus affords estimates of the appropriate exponents . from the measured variance of the spin fluid operator distributions , we find @xmath98 . as far as @xmath99 and @xmath100
are concerned , these estimates are in excellent agreement with exact conjectures @xmath101 @xcite .
the situation regarding the eigenvalue exponent @xmath95 , on the other hand , is not so satisfactory , there being a sizable discrepancy with the exact value of @xmath102 .
this discrepancy stems , we believe , from two sources .
firstly , since the operator distribution @xmath64 is highly sensitive with respect to the designation of the value of the field mixing parameter @xmath28 implicit in the definition of @xmath91 , small uncertainties in the estimate of @xmath28 can lead to significant larger errors in the measured variance of @xmath64 .
a similar effect was also previously observed at the liquid - vapour critical point of the lennard - jones fluid @xcite .
secondly , and as we discuss in section [ sec : concs ] , the near gaussian character of @xmath64 signifies the absence of strong fluctuations in the @xmath91 , in which case it is questionable whether a finite - size scaling can be reliable applied to @xmath64 at all . in view of this problem
we have adopted a rather different approach for measuring @xmath95 based on the scaling properties of @xmath81 , a quantity which does show strong fluctuations and which is also insensitive to the designation of the field mixing parameter @xmath28 .
it can be shown @xcite , that the maximum slope of the cumulant ratio @xmath103 near @xmath104 varies with system size like @xmath105 . using the histogram extrapolation technique ,
we have obtained the temperature dependence of this slope for the spin fluid model .
the results yield the estimate @xmath106 , which agrees to within error with the exact conjecture .
turning now to the observables @xmath107 and @xmath108 , we plot in figure [ fig : flbare ] the measured distributions of these quantities at the assigned tricritical parameters .
here we note that as is the case with ordinary critical phenomena in the lennard - jones fluid @xcite , the energy distribution @xmath109 differs qualitatively in form from the operator distribution @xmath64 .
this finding , the origin of which is explained in detail in reference @xcite , reflects the coupling of the tricritical energy fluctuations to the density , the latter of which are stronger and thus dominate for large @xmath110 .
the influence of this coupling is also discernible as a small asymmetry in the tricritical density distribution .
for the average tricritical density we find @xmath111 , while for the average tricritical energy density we find @xmath112 .
the average magnetisation is of course strictly zero on symmetry grounds .
finally in this subsection , table [ tab : uls ] summarises the measured values of the fourth order cumulant ratios @xmath113 for each of the three scaling operator distributions at the assigned values of the tricritical parameters . in this section
we describe a simple mean - field calculation of the tricritical parameters of the 2d spin fluid .
let @xmath114 denote the pressure in the system for a given chemical potential @xmath12 and volume @xmath115 in the thermodynamic limit , where the equilibrium density @xmath116 is given by the ratio of the average number of particles @xmath117 and the volume @xmath115 , @xmath118
. then @xmath119 \label{mf1}\ ] ] where @xmath120 is the free energy per volume , @xmath121 \;\;. \label{mf2}\ ] ] in the mean field approximation we assume an interaction between a spin @xmath122 and an effective field @xmath123 , where @xmath124 is the effective coordination number and the @xmath10-particle hamiltonian is written as @xmath125 s_i \label{mf2a}\ ] ] and the mean field free energy @xmath126 is @xmath127 \label{mf2b}\ ] ] @xmath128 \label{mf3}\ ] ] @xmath129 is the free energy of a classical system with hamiltonian @xmath130 and the second term on the right hand side of eq .
( [ mf3 ] ) reaches its minimum at @xmath131 \label{mf4}\ ] ] since the coordination number @xmath124 in the fluid is not fixed we approximate the effective field on one particle by @xmath132 where the fluid correlation function @xmath133 is taken from the percus yevick solution for hard discs , which can be found numerically .
two phase coexistence between a gas phase at density @xmath134 and a liquid phase phase at density @xmath135 is given by equal pressure in the two phases @xmath136 this condition determines the mean field chemical potential @xmath137 for phase coexistence via eq .
( [ mf1 ] ) . implementing this criterion numerically
, we find @xmath138 @xmath139 .
clearly this estimate for @xmath140 seriously overestimates the measured tricritical temperature ( @xmath141 ) showing that mean field calculations of this type can not be relied upon to provide accurate tricritical data , at least for 2d systems .
in summary we have demonstrated that the tricritical ordering operator distributions of the 2d spin fluid can be mapped into excellent correspondence with those of the 2d blume - capel model .
the existence of such a mapping represents perhaps the most stringent test of universality .
there can thus be little doubt that despite their very different microscopic character , the two systems do indeed share a common fixed point . with regard to the scaling operator distribution themselves
, we note that the form of @xmath142 is ( to within the precision of our measurements ) essentially gaussian , as evidenced by the very small value of the cumulant ratio @xmath143 .
this guassian behaviour implies that the tricritical fluctuations in @xmath91 are extremely weak and is thus a consequence of the central limit theorem .
this weakness is further manifest in the very small measured values of the field mixing parameter @xmath27 , as well as in the near absence of asymmetry in the tricritical density distributions , a situation which contrast markedly with that of the 2d and 3d lennard - jones fluids , where much stronger field mixing effects are observed in the density distribution @xcite . finally , the tricritical form of @xmath144 merits special comment .
we note that the _ three - peaked _ form of this distribution differs radically from the universal magnetisation distribution of the critical ising model , which is strongly _
double_-peaked in two dimensions @xcite .
the existence of a three peaked structure for tricritical phenomena reflects the additional coupling that arises between the magnetisation and the density fluctuations .
specifically , the central peak corresponds to fluctuation to small density , which are concomitant with an overall reduction in the magnitude of the magnetisation ( cf .
figure [ fig : config ] ) . were one , however , to depart from the tricritical point along the critical line , these density fluctuations would gradually die out and a crossover to a magnetisation distribution having the double - peaked ising form would be expected . in future work
we intend to investigate the nature of this crossover in detail . a generous allocation of cray - ymp computer time at the rhrk , universitt kaiserslautern and hlrz , jlich is gratefully acknowledged .
the authors have benefitted from helpful discussions with k. binder , b. dnweg , k.k .
mon and m. mller .
nbw acknowledges the financial support of the commission of the european community ( erb chrx ct-930 351 ) .
pn thanks the dfg for financial support ( heisenberg foundation ) see e.g. r.m .
stratt , j. chem . phys . * 80 * , 5764 ( 1988 ) ; d. marx , p. nielaba and k. binder , phys . rev .
* b47 * , 7788 ( 1993 ) ; p.
de smedt , p. nielaba , j.l .
lebowitz , j. talbot and l. dooms , phys .
* a38 * 1381 ( 1988 ) and references therein . it should be noted that for large values of the argument @xmath145 a crossover occurs to standard ising critical behaviour on the critical line .
such crossover phenomena can also be explicitly incorporated within the finite - size scaling framework , see eg . k. binder and h .-
deutsch , europhys .
* 18 * , 667 ( 1992 ) . | monte carlo simulations are used to investigate the tricritical point properties of a 2d spin fluid .
measurements of the scaling operator distributions are employed in conjunction with a finite - size scaling analysis to locate the tricritical point and determine the directions of the relevant scaling fields and their associated tricritical exponents .
the scaling operator distributions and exponents are shown to match quantitatively those of the 2d blume - capel model , confirming that both models belong to the same universality class .
mean - field calculations of the tricritical point properties are also compared with the simulation measurements .
epsf pacs numbers 64.60fr , 64.70.fx , 05.70.jk |
a total coloring of a graph @xmath2 is an assignment of colors to the vertices and edges of @xmath2 such that every pair of adjacent / incident elements receive different colors .
a @xmath3-total coloring of a graph @xmath2 is a total coloring of @xmath2 from a set of @xmath3 colors .
the minimum positive integer @xmath3 for which @xmath2 has a @xmath3-total coloring , denoted by @xmath4 , is called the total chromatic number of @xmath2 .
it is easy to see that @xmath5 for any graph @xmath2 by looking at the color of a vertex with maximum degree and its incident edges .
the next step is to look for a brooks - typed or vizing - typed upper bound on the total chromatic number in terms of maximum degree .
it turns out that the total coloring version of maximum degree upper bound is a difficult problem and has eluded mathematicians for nearly 50 years .
the most well - known speculation is the total coloring conjecture , independently raised by behzad @xcite and vizing @xcite , which asserts that every graph of maximum degree @xmath6 admits a @xmath7-total coloring .
this conjecture remains open , however , many beautiful results concerning it have been obtained ( cf .
in particular , the total chromatic number of all outerplanar graphs has been determined completely by zhang et al .
@xcite and that of all series - parallel graphs @xcite has been determined completely by wu and hu @xcite .
a graph is pseudo - outerplanar if each of its blocks has an embedding in the plane so that the vertices lie on a fixed circle and the edges lie inside the disk of this circle with each of them crossing at most one another .
for example , @xmath8 and @xmath9 are both pseudo - outerplanar graphs .
this notion was introduced by zhang , liu and wu in @xcite , where the edge - decomposition of pseudo - outerplanar graphs into forests with a specified property was studied . in this paper , we prove that the total chromatic number of every pseudo - outerplanar graph with maximum degree @xmath0 is exactly @xmath1 and thus the total coloring conjecture holds for all pseudo - outerplanar graphs .
to begin with , let us review an useful structural property of pseudo - outerplanar graphs which was proved in @xcite .
we shall prove the theorem by induction on @xmath31 .
so we assume that @xmath2 is 2-connected and thus @xmath32 .
set @xmath33 . in the next ,
we complete the proof by verifying the following four claims ; they imply a contradiction to lemma [ struc ] .
suppose , to the contrary , that @xmath36 .
consider the graph @xmath37 . by induction
, @xmath38 has an @xmath39-total coloring @xmath40 .
denote the other neighbor of @xmath41 by @xmath42 .
now erase the color on @xmath41 and color the edge @xmath11 with @xmath43 .
this is possible since @xmath44 .
denote the extended coloring at this stage still by @xmath40 .
then at last we color @xmath41 with @xmath45 , which is also possible since @xmath46 .
otherwise , @xmath48 admits an @xmath39-total coloring by induction and every edge of the 4-cycle has at least two available colors since it is incident with at most @xmath49 colored elements .
this implies that one can extend the coloring of @xmath38 to the four edges @xmath50 and @xmath51 since every 4-cycle is 2-edge - choosable . at last ,
the two vertices @xmath41 and @xmath52 can be easily colored since they are both of degree two .
suppose , to the contrary , that @xmath2 contains such a 4-cycle .
we consider the graph @xmath37 that has an @xmath39-total coloring @xmath40 by induction .
one can find that the only obstacle of extending @xmath40 to @xmath11 is the case when @xmath54 and @xmath55 . without loss of generality ,
let @xmath56 and @xmath57 . if @xmath58 , then we recolor @xmath41 by 4 ( note that @xmath59 ) and then color @xmath11 by 5 .
so we assume @xmath60 .
similarly we can prove @xmath61 .
therefore , we can recolor @xmath52 by 1 and then color @xmath11 by 6 , a contradiction .
suppose , to the contrary , that @xmath2 contains such a 7-path .
we consider the graph @xmath62 , which admits an @xmath39-total coloring @xmath40 by induction .
denote by @xmath63 the set of available colors to properly color an edge @xmath64 .
it is easy to see that @xmath65 and @xmath66 , moreover , @xmath67 and @xmath68 .
if @xmath69 , without loss of generality assume that @xmath70 and @xmath71 , then color @xmath72 with 1 and @xmath73 with @xmath47 .
if @xmath74 , then color @xmath75 with @xmath76 and @xmath77 with 1 .
if @xmath78 , then color @xmath77 with @xmath79 and @xmath75 with @xmath80 .
in each case we have @xmath81 .
thus we can color @xmath51 with @xmath82 and @xmath83 with @xmath84 such that @xmath85 . at this stage ,
the three vertices @xmath86 and @xmath42 can be easily colored since they are all of degree two .
so we assume that @xmath87 .
now we firstly color @xmath72 and @xmath73 by 1 .
if @xmath78 , then color @xmath77 with @xmath88 and @xmath75 with @xmath80 .
the current extended coloring satisfies that @xmath81 .
therefore , we can color the remaining elements similarly as before .
so we assume that @xmath74 .
this implies that @xmath89 and thus we can exchange the colors on @xmath72 and @xmath90 . by doing
so we obtain a new coloring satisfying @xmath91 and therefore we can extend this partial coloring to @xmath2 by a same argument as above .
the graph ( a ) in figure [ fig ] is a pseudo - outerplanar graph , since it has a pseudo - outerplanar drawing as ( b ) .
one can easy to check that it is a graph with maximum degree 3 and total chromatic number 5 .
thus the upper bound for @xmath6 in corollary [ main ] , although probably not the best possible , can not be less than 4 . | a graph is pseudo - outerplanar if each of its blocks has an embedding in the plane so that the vertices lie on a fixed circle and the edges lie inside the disk of this circle with each of them crossing at most one another . in this paper , the total coloring conjecture
is completely confirmed for pseudo - outerplanar graphs . in particular , it is proved that the total chromatic number of every pseudo - outerplanar graph with maximum degree @xmath0 is @xmath1 .
+ keywords : pseudo - outerplanar graph , total coloring , maximum degree . |
the bloch frequency ( @xmath0 ) of a biased sl is given by @xmath1 where @xmath2 is the applied electric field and @xmath3 is the superlattice period .
superlattice transport at low temperatures is characterised by two regimes . at low electric fields where @xmath4 , where @xmath5 is the sl miniband width , miniband transport through extended sl states dominates . in the simple esaki - tsu miniband transport
model@xcite electron drift velocity ( @xmath6 ) is described by a scattered bloch oscillator in 1-dimension , @xmath7 , where electron mobility @xmath8 , @xmath9 and @xmath10 is the scattering time .
the main mechanisms which contribute to @xmath10 are phonon scattering , impact ionisation and interface roughness scattering . at high electric fields however @xmath11 causing the miniband to split into localised wannier - stark - ladder ( wsl ) states@xcite and
consequently miniband transport is no longer the dominant process . despite the critical role of energy relaxation processes in superlattices the magnetophonon effect@xcite has only received a small amount of attention .
it has been extensively studied in bulk@xcite and 2-dimensional@xcite semiconductor systems , where magnetophonon resonances ( mpr ) are observed as magnetoresistance oscillations caused by resonant scattering of electrons by optical phonons .
the dominant electron - phonon coupling in all iii - v systems is with the lo phonon due to the large electric polarisation associated with these modes .
conservation of crystal momentum limits the lo phonon scattering to phonons near the brillouin zone center such that the lo phonon energy is essentially mono - energetic .
the effects of mpr on the resistivity depend strongly on the relative configuration of the applied electric and magnetic fields .
in the transverse case , ( @xmath12 ) , resonant momentum relaxation causes magnetoresistance maxima to be observed at precisely the mpr resonance condition @xmath13 where @xmath14 is the cyclotron frequency , @xmath15 is the lo phonon frequency and @xmath16 is an integer . at b - fields which satisfy
the mpr condition strong inelastic scattering occurs between zero momentum states separated in landau index by @xmath16 . in the longitudinal configuration ( @xmath17 ) considered in this report
the mpr is typically more complicated as direct lo phonon emission and absorption processes between zero momentum states do not relax the electron momentum in the electric - field direction . in bulk materials resonances
occur due to an interplay of different indirect scattering processes@xcite .
extensive experimental observations of longitudinal mpr ( lmpr ) in bulk iii - v systems@xcite reveal resistivity minima which are displaced to @xmath18-fields somewhat below condition [ eqn : mpr ] .
calculations for superlattices have suggested both that resistivity minima@xcite should occur at fields slightly below the resonance condition and that maxima could occur@xcite at the resonance condition .
when @xmath5 is significantly below the lo phonon energy polyanovskii@xcite has also suggested that transitions from the top to the bottom of the miniband will generate new superlattice magnetophonon resonances ( slmpr ) at the condition @xmath19 where resonant lo phonon scattering between areas of high density of states ( dos ) at the top and bottom of the superlattice miniband enhances the current .
in contrast to the predictions of theory all experimental observations of mpr in superlattices@xcite have assigned mpr features as resistance maxima , predominantly in samples with low @xmath5 and using the gaas / alas system .
a significant enhancement of the resonant peak intensity , compared with bulk material , has been reported and attributed to the effect of the superlattice band structure .
_ et al.__@xcite also observed that low @xmath5 samples exhibited plateaus in the oscillatory part of the magnetoresistance trace which it was suggested may provide evidence for polyanovskii s predictions .
detailed studies by gassot _
_ et al.__@xcite also assigned lmpr peaks as maxima in resistance but reported significant deviations from the predicted mpr conditions .
the analysis of the gaas / alas system is further complicated by the presence of resonances attributed to both the gaas and alas lo phonons , which are approximately 25% different in energy .
this paper reports studies of the magnetophonon effect in the type - ii inas / gasb system which has a low carrier effective mass ( m@xmath20m@xmath21 ) allowing the study of transport at cyclotron energies considerably above the lo phonon energy ( @xmath22mev ) .
the analysis of the resonance is also simplified because the lo phonon energies of the inas and gasb layers are almost exactly equal .
tunnelling between adjacent superlattice layers is dominated by interband coupling and is strongly * k * dependant .
the interband coupling to the barrier valence band states is strongly reduced for higher landau index ( @xmath23 ) llmb s resulting in narrower miniband widths@xcite ( @xmath24 ) and suppressing domain formation . throughout this report
we will refer to all experimentally observed resonances as lmpr features so as to distinguish them from the mpr condition .
this is important as the longitudinal configuration typically produces resonances slightly shifted from the mpr condition .
p * simulation results for sample 4561 ( _ a _ ) landau miniband fan calculated from average of spin up and down levels .
the dashed line marks the lo phonon energy and fundamental mpr condition .
( _ b _ ) miniband dispersion for the 1@xmath25 and 2@xmath26 sl minibands at @xmath27 t .
( _ c _ ) plot of the magnetic field dependence of the first four landau minibands.[fig:4561ll ] ] experiments were performed on 100 period undoped inas / gasb superlattices grown by movpe .
samples were grown on _
n_-type inas substrates ( carrier density @xmath28 @xmath29 estimated from shubnikov - de haas measurements ) with @xmath30 inas buffer and cap layers . the ratio of inas / gasb ( @xmath31 ) is estimated from growth rates measured using an _ in - situ _ surface photo - absorption ( spa ) technique@xcite .
samples were also characterized using x - ray diffraction ( xrd ) allowing calculation of the superlattice period ( @xmath32 ) . for vertical transport measurements 150@xmath33
m mesas were defined using standard lithographic and wet etching techniques .
ohmic contacts were made to sample and substrate by evaporating 5 nm of chromium and 150 nm of gold .
previous measurements@xcite on the same samples have identified that transport is best described by a number of active periods which is less than the nominal 100 periods grown .
reduction in the number of periods is caused by the formation of inas inversion layers at the superlattice - substrate and superlattice - cap interfaces , figure [ fig : samples - potential ] ( _ b _ ) .
low temperature measurements on these samples identify 2-dimensional electron gases with carrier densities @xmath34 @xmath35 , from shubnikov - de haas like oscillations , which is consistent with the formation of this structure .
electron wavefunctions in the miniband couple with the inversion layer states causing a widening of the superlattice miniband in some of the first and last periods of the structure .
the broadened miniband states are highly conducting and allow electrons to be easily injected into the miniband .
the broadened miniband also exhibits negligible voltage drop when the structure is biased resulting in the reduction of the number of active periods observed(@xmath36 ) .
all measurements were performed in dc mode with two contacts as displayed schematically in figure [ fig : samples - potential](_a _ ) .
@xmath37 measurements in magnetic fields up to 19.5 t were performed using a superconducting magnet system and a _ keithley _ 236 smu .
higher magnetic fields were obtained using non - destructive magnets in the clarendon laboratory kurti pulsed magnetic field facility .
the magnetic field pulse was recorded using a pick - up coil situated 2 mm below the sample .
induced voltages during the magnetic field pulse are removed by averaging two pulses of opposite polarity .
pulsed field measurements were performed with _ gage compuscope _ 5mhz transient recorder cards .
.sample characteristics.[tab : growth ] [ cols="^,^,^,^,^",options="header " , ] sample characteristics were simulated with * k@xmath38p * theory@xcite solved using the envelope function approximation@xcite in momentum space@xcite .
sample characteristics are summarised in table [ tab : growth ] .
fourier transform magnetoabsorption spectroscopy was used to observe the superlattice energy gap which allows refined estimation of the ratio @xmath31 and shows that miniband width estimates in table [ tab : growth ] are correct within to @xmath39% . *
k@xmath38p * simulation of the superlattice landau miniband fan are later used to predict the position of mpr features .
an example of a sl landau miniband fan is presented in figure [ fig:4561ll ] for sample 4561 .
all samples display a large separation between the first and second conduction minibands ( @xmath40mev for 4561 ) such that for the bias values used in this study only conduction through the fundamental miniband need be considered , figure [ fig:4561ll ] ( _ b _ ) .
simulation results also clearly display a strong suppression of miniband width with increasing magnetic field which is stronger for higher index landau levels due to the higher energy of these states , figure [ fig:4561ll ] ( _ c _ ) .
characteristic of sample 4577 and 4562 at @xmath41k and 3756 at @xmath42k .
[ fig : i - vesaki - tsu ] ] @xmath43 curves for three samples are displayed in figure [ fig : i - vesaki - tsu ] and are characteristic of the response of all superlattices studied in this report .
we observe conventional miniband transport characteristics with a region of ohmic miniband transport at low bias followed by a miniband transport peak identified as a shoulder in the @xmath43 trace ( for example at @xmath44mv for sample 4577 and @xmath45mv for sample 3756 ) . for sample 4577
we observe a second shoulder at @xmath46mv which is attributed to electron - phonon resonances in the stark hopping regime@xcite as have been discussed in a previous publication on hot - electron mpr observed in the wsl regime for the same sample set@xcite .
curves for sample 4577 at @xmath41k for a range of bias .
( _ b _ ) plot of @xmath47 for sample 4577 with @xmath41k and a range of sample bias .
vertical dashed lines mark the observed lmpr and slmpr features .
[ fig:4577-rb ] ] @xmath37 curves for sample 4577 ( @xmath48mev ) are displayed in figure [ fig:4577-rb ] ( _ a _ ) for bias in the range @xmath49mv and @xmath41k .
@xmath37 traces are analysed by plotting @xmath47 to remove the monotonically increasing background resistance and identify weak resonances , figure [ fig:4577-rb ] ( _ b _ ) . in plots of @xmath47 maxima
correspond to resistivity minima in the equivalent @xmath37 trace . in this paper
we will use the simplification that the additional resistivity maxima and minima caused by the resonant scattering processes refer to minima and maxima in @xmath47 respectively , a simplification in discussion often used when discussing mpr features@xcite .
plots of @xmath47 display strong maxima features at @xmath50 t and @xmath51 t .
estimates of the @xmath52 mpr condition for sample 4577 using @xmath53 calculations predict a resonance at @xmath54 t .
the appearance of two resonant peaks close to the conventional magnetophonon condition predicted by @xmath53 calculations suggests that the resonant behavior may be significantly more complex than has previously been assumed .
the main resistivity maximum lies close to the conventional mpr condition , equation [ eqn : mpr ] , whilst the resistivity minimum which might be attributed to lmpr is at @xmath55 t shifted to @xmath56 below the conventional mpr condition .
the second minimum at @xmath57 t is a possible example of the new slmpr resonance given by equation [ eqn : polyanovskii ] .
this assignment is supported by the fact that both lmpr and slmpr features are significantly suppressed with increasing bias beyond @xmath58200 mv when the superlattice moves into the wsl regime , and so all further measurements were restricted to the miniband transport range of bias .
the feature at @xmath59 t is likely to be the @xmath60 slmpr resonance .
assignment of the 6.3 t feature to @xmath60 lmpr is however inconclusive as a residual shubnikov de - haas oscillation peak for the bulk substrate occurs at approximately @xmath61 t and complicates analysis in this low b - field region . for a range of superlattice samples.,scaledwidth=45.0% ] in order to test this assignment results are examined for samples with a range of miniband widths as shown in figure [ fig : allplots ] .
the strong features identified for analysis are selected for repeatability between successive measurements and in comparison with data taken for steady fields up to @xmath62 t , as shown for sample 4577 in figure [ fig:4577-rb ] .
resistivity minima ( @xmath47 maxima ) which are candidates for lmpr and slmpr features are indicated by arrows . as the miniband width increases the separation of the slmpr feature from the conventional mpr resonance increases progressively .
all features show a characteristic mpr temperature dependence with maximum amplitude in the range @xmath63k .
p * calculations alongside experimentally observed resonances .
the x - axis indicates the sl miniband width at @xmath27 t calculated with *
k@xmath38 p*. note that the * k@xmath38p * calculated resonance positions take account of the reduction in sl miniband width with increasing magnetic field which becomes particularly significant for the three highest miniband width samples .
as there is no predicted lineshape for the resonances the errors are estimated from 40% full width half maximum ( fwhm ) of the features .
shaded region indicates the parameter space experimentally probed.,scaledwidth=45.0% ]
the magnetic field positions of the observed lmpr and slmpr minima and maxima features are compared with the values predicted by @xmath53 calculations and equations [ eqn : mpr ] and [ eqn : polyanovskii ] in figure [ fig : slmpr ] .
the @xmath53 predicted magnetic field value for the mpr condition falls with increasing @xmath5 due to the decrease in the superlattice band gap , which causes the effective mass to decrease .
we observe that resistivity maxima ( @xmath47 minima ) occur at fields slightly above the conventional mpr resonance position .
the experimentally observed resistivity minima labelled as @xmath52 lmpr are typically seen when @xmath64 in good agreement with bulk lmpr resonances@xcite . by contrast
the slmpr features move up in field due to the increased contribution from @xmath5 in equation [ eqn : polyanovskii ] and show excellent agreement with the predicted slmpr positions when account is taken of the magnetic field dependence of the miniband width .
as the miniband width increases we expect a continuous transition to conventional bulk magnetophonon behavior once @xmath65 since the superlattices may be considered to act as a bulk 3-dimensional system if all thermal and cyclotron excitations are much smaller than the superlattice miniband width within this regime electrons are unable to probe the upper portion of the superlattice miniband and carriers only experience a dispersion similar to the parabolic bulk case .
it would therefore be expected that such samples display no slmpr features .
@xmath53 calculation results for samples 3756 and 4520 indicate that if observed the slmpr would be located at @xmath67 t and @xmath68 t respectively . in experiment ( figure [ fig : allplots ] ) both exhibit resistance minima features at around @xmath69 t which correspond well with the predicted @xmath70 mpr resonances which is also particularly strong in the longitudinal configuration for bulk materials@xcite suggesting that the transition to bulk behavior has occurred for these structures .
this will be discussed further later in this report .
the physical origins of the lmpr and slmpr features in low miniband width samples ( @xmath71 ) can be understood by considering the lo phonon absorption / emission processes which are allowed for different @xmath14 values , as shown in figure [ fig : slmprv2 ] .
inter - miniband lo phonon scattering between the @xmath72 and @xmath73 llmb s is only permitted in the range @xmath74 due to the forbidden energy gaps in the system .
note also that the miniband width of any given llmb has a significant magnetic field dependence ( @xmath24 ) further complicating this schematic picture .
transitions for which final or initial states are at @xmath75 and/or the mini - brillouin zone boundary @xmath76 have high scattering rates due to the large density of states at these @xmath77 .
we can consider the scattering of electrons in the positive @xmath77 region of the @xmath72 llmb dispersion which contribute to the transport current .
it is clear that lo phonon absorption transitions will result in final states at @xmath78 and @xmath79 such that the average velocity of a final state is zero .
the average of lo phonon scattering events therefore relaxes the electron momentum and the resistance will be proportional to scattering rate . the scattering rate will increase approaching the mpr condition as lo phonon scattering is allowed over a greater proportion of the dispersion . at the mpr condition the lo phonon scattering rate is at its greatest . if @xmath80 scattering will be allowed anywhere within the superlattice dispersion . in reality
@xmath81 for the inas / gasb system restricting scattering at the mini - brillouin zone boundary . above the mpr condition the scattering rate
is reduced with increasing b - field as lo phonon transitions from the center of the superlattice dispersion are increasingly forbidden . at the slmpr condition ( @xmath82 ) only scattering from @xmath76 in the @xmath72 llmb to @xmath75 in the @xmath73 llmb
is allowed .
the relaxation therefore links two high density of states regions of the dispersion relation resulting once again in a high scattering rate . for superlattice miniband transport , however , the current flow is dependent on more than simply the immediate scattering rate .
the contribution to current flow also depends on the subsequent ballistic motion in k - space under electric field acceleration , since the velocity is strongly dependent upon where the carrier is in the dispersion relation .
these factors must be considered to fully understand the form of mpr features .
we consider the relative effects of acceleration of carriers in initial and final scattered states to determine the effect of the scattering upon transport current .
acceleration in an electric field following lo phonon scattering has the effect of decreasing drift velocity below the mpr condition since the carrier velocity is increasing with @xmath83 in the lower landau level and decreasing with @xmath83 in the upper level ( figure [ fig : slmprv2](b ) ) .
conversely the drift velocity increases above the mpr condition . at the mpr condition
final and initial states are comparable such that subsequent acceleration does not alter the total transport current . scattering at the slmpr condition has the most dramatic and significant consequence that it prevents the occurrence of bragg scattering and the subsequent suppression of transport current due to cycling of the carriers through the negative drift velocity section of the dispersion relation . above @xmath84 the coupling of the @xmath72 and @xmath73 llmbs through lo phonon scattering is removed and intra - llmb scattering dominates .
this reverses carrier momentum and quenches transport@xcite resulting in a significant increase in resistance following the slmpr feature .
the characteristics of the lmpr and slmpr features are therefore determined by interplay of the scattering rate and the subsequent ballistic motion of the transport electrons .
this simple analysis indicates that lmpr is likely to appear as a resistance maximum at the mpr condition due to the high scattering rate and slmpr as a resistance minimum followed by a large increase in resistance . in order to understand the different contributions to the magnetoresistance , modelling of the experimental results
was therefore performed using semiclassical monte carlo simulations of the miniband transport , assuming that the effects of wannier - stark localisation are not significant .
simulations follow the method outlined by henriques _ _
et al.__@xcite .
limited inelastic acoustic phonon scattering in an energy window of 1mev ( selected to be less than @xmath5 ) was introduced to ensure that the conductivity at high @xmath18-fields is small but non - zero .
it is found that changing this window significantly alters the background magnetoresistance but that the lmpr and slmpr features remain .
simulation results are displayed for superlattice band structures approximating to that of samples 4561 , 4579 and 3756 in figure [ fig : andre2 ] .
the landau miniband energy widths have been approximated with the function @xmath85 where @xmath86 and @xmath87 are parameters obtained from fits to the results of @xmath53 simulations .
both samples are simulated with @xmath88k and electric field @xmath89 such that results are close to the ohmic miniband transport regime but show some ballistic behavior .
the simulation results for low miniband widths ( figure [ fig : andre2 ] ( _ a _ ) and ( _ b _ ) ) display two clear features : firstly , there is a pronounced minimum in the drift velocity ( corresponding to a resistivity maximum ) at the conventional mpr condition and secondly there are peaks in drift velocity at the slmpr conditions .
these results show excellent agreement with the schematic picture previously discussed .
the simulations identify that the magnetotransport is dependent on both the scattering rate and the positions on the dispersion curves of the initial and final states .
resistance maxima at the mpr are caused by the increased scattering rate .
the resistance minima at the slmpr are formed from the relative effects of electric field acceleration in the initial and final scattering states .
similar simulation results are observed for other samples which show that the position of the slmpr feature is well described by equation [ eqn : polyanovskii ] , shifting progressively away from the lmpr feature with increasing miniband width .
the simulation results change significantly when the miniband width becomes large ( @xmath65 ) as observed for sample 3756 , figure [ fig : andre2 ] ( _ c _ ) .
we observe that resistivity minima now occur at the @xmath52 and @xmath70 mpr conditions .
this is in good agreement with the bulk approximation in which the lmpr is observed as a resistivity minimum slightly below the mpr condition@xcite . for sample 3756 the @xmath90 mpr feature
is particularly enhanced by proximity to the slmpr condition .
the slmpr feature is however suppressed above the @xmath90 mpr condition due to efficient intra - miniband lo phonon emissions .
our overall conclusion therefore is that the modelling provides strong support for the attribution of the high field resistivity minimum to the predicted slmpr feature and that this is strongly supported by the clear dependence of the slmpr position on miniband width .
by contrast , the attribution of a maximum or minimum in resistance to the precise mpr condition is not found to hold experimentally .
the stronger feature is found to be a maximum at fields a few percent above the mpr condition .
a resistivity minimum is found in all samples slightly below the mpr condition , in an analogous behavior to that seen in bulk material .
the non - observation of the slmpr feature in the earlier gaalas based work is probably due to the complications caused by the presence of the two phonon modes and the relatively small range of miniband widths studied .
in summary we have demonstrated a new form of magnetophonon resonance which occurs due to the scattering of carriers from the top of the ground state miniband to the bottom of the next llmb . this new resonance known as superlattice mpr ( slmpr )
was first predicted by polyanovskii in 1983 .
the resonance condition corresponds to taking carriers from the top of the miniband dispersion where they will have a negative differential velocity and moving them to the bottom of the band where they can be recycled through the positive conduction regime .
parts of this work has been supported by euromagnet under the eu contract rii3-ct-2004 - 506239 of the 6th framework structuring the european research area , research infrastructures action and by the brazilian agency cnpq under contract 308116/2004 - 6 . | we report an experimental study of miniband magnetoconduction in semiconducting inas / gasb superlattices . for samples with miniband widths below the longitudinal optical phonon energy
we identify a new superlattice magnetophonon resonance ( slmpr ) caused by resonant scattering of electrons across the mini - brillouin zone .
this new resonant feature arises directly from the superlattice dispersion and total magnetic quantisation ( energetic decoupling ) of the superlattice landau level minibands .
semiconductor superlattices ( sl s ) comprise alternating layers of two or more semiconductor materials , leading to the formation of continuous energy bands in the growth direction called minibands .
the reduced brillouin zone and energy band width of the sl allows measurements that probe parameter spaces which are inaccessible in bulk semiconductors .
total quantisation of the superlattice energy scheme can be achieved by the application of a large magnetic field which suppresses inter - landau level miniband ( llmb ) scattering and allows the realisation of a quasi 1-dimensional or quantum box sl ( qbsl ) regime .
this has led to strong interest in the sl magnetoresistance and transport characteristics @xcite .
the sl miniband structure can be engineered such that in the qbsl regime optical phonon scattering is limited@xcite by using narrow minigap and miniband widths , leaving only weak acoustic - phonon processes to dissipate the electron energy . in this report
we investigate magnetotransport properties of inas / gasb superlattices in the miniband transport regime . in a previous publication@xcite we investigated hot - electron magnetophonon resonance caused by
the lo phonon mediated hopping between landau wannier - stark states at low temperatures . in this report
we study longitudinal magnetophonon resonances caused by the resonant emission / absorption of longitudinal optical ( lo ) phonons in the miniband transport regime for a range of sl structures at high tempertures . through a systematic study using different miniband widths
we identify a new form of magnetophonon resonance which provides evidence for the energetic de - coupling of sl landau level minibands leading to suppression of optical phonon scattering . |
over the past few years various authors have considered the possibility of large extra dimensions as a solution to the discrepancy between the @xmath1 tev scale of standard model physics and the apparent mass scale associated with gravitational interactions @xmath2gev@xmath3 . in these models ,
the standard model particles are confined to a 3 + 1 dimensional brane whilst gravity is free to inhabit all 3 + 1+n dimensions .
the expression for the observed value of g in four dimensions is given by @xmath4 @xmath5 and so it is possible to obtain arbitrarily small values of the higher dimensional fundamental scale @xmath6 provided the radii @xmath7 of the higher dimensions are large enough .
factorisable higher dimensions ( for example higher dimensions compactified on @xmath8 or @xmath9 ) have associated with them an infinite tower of massive kk states which can be produced during high energy scattering of standard model particles on the brane , a phenomenon which may be investigated at the lhc @xmath10 .
the possibility of this kk mode generation has already led to strong constraints on the size and number of higher dimensions from astro - particle physics .
as one of the mechanisms expected to be responsible for the generation of the kk modes is gravi - bremsstrahlung from nucleon - nucleon scattering , the core of a supernova would seem to be a good place to look for the effect .
however , the observed neutrino pulse from supernova 1987a successfully accounted for most of the energy of the collapse .
it was therefore possible to place constraints on the energy lost via this mechanism to kk modes and consequently the radius and number of the extra dimensions @xmath11 .
one would expect the same mechanism to apply to nucleon - nucleon scattering in the early universe ] .
the kk modes corresponding to phenomenologically interesting values of the size of the extra dimensions have lifetimes larger than the age of the universe @xmath12 and dissipate as matter , i.e. @xmath13 .
this extra injection of matter leads to an earlier matter - radiation equality which in turn leads to a more rapid decline in the cmb temperature .
the increased cooling rate means that by the time the cmb has cooled to 2.73k the universe is still be much too young to hold the objects we observe in ours .
it is these constraints that are investigated in this paper .
in @xmath14 the angle averaged emmissivity of the plasma into kk modes @xmath15 from a non - degenerate gas of nucleons is shown to be given by the expression @xmath16 where t is temperature , @xmath17 is the isospin averaged nucleon mass @xmath18 mev , @xmath19 is given in the appendix , and the radius of the large extra dimensions r is given by rearranging equation ( @xmath20 ) @xmath21 in ( @xmath22 ) the nucleon - nucleon cross section @xmath23 is evaluated at the peak of the centre of mass energy distribution @xmath24 where @xmath25 so the cross section is evaluated at @xmath26 .
the total s - wave nucleon - nucleon scattering cross section is given by @xmath27 @xmath28 where @xmath29 is the scattering length and @xmath30 is the effective range .
we can obtain approximate values for these parameters by using those experimentally observed for the neutron - neutron @xmath31 channel , @xmath32 and @xmath33 @xcite .
now we can express the cross section at the peak of the energy distribution in mev@xmath34 @xmath35^{2}}\ ] ] where t is measured in mev . at this stage
it is worth noting that although the cross section decreases with temperature by a factor of about 2 between 1 mev and 200 mev we will still expect the emissivity to be very temperature dependant because of equation ( [ emiss ] ) .
in order to ensure that we are correct in using the normal 4 dimensional frw equations for our cosmology , we need to make sure that the density of matter in the bulk remains lower than the fundamental planck density @xmath36 by the factor @xmath37 @xmath38 . in this paper where we start with no kk modes , the total density in the bulk
is given by the density lost from the brane divided by the volume of large extra dimensions .
assuming a toroidal compactification manifold these constraints reduce to @xmath39 this inequality is easily satisfied throughout our calculations .
secondly , the rate at which radiation energy density on the brane will evaporate into kk modes in the bulk was obtained in @xmath4 by dimensional analysis .
one can then obtain a temperature at which the cooling of radiation is dominated by this evaporation into the bulk as opposed to the normal cooling due to cosmological expansion .
this temperature , @xmath40 , serves as a maximum temperature up to which the universe can be considered normal. comparing cooling rates : @xmath41 we are only considering events which occur below 170 mev and the case with one large extra dimension has been ruled out by gravitational experiments for all but very high values of @xmath6 .
consequently , only the case with 2 large extra dimensions has a value of @xmath40 below 170 mev and then only for values of @xmath6 below approximately 8 tev which have been ruled out by the supernova work @xmath11 .
having checked the above two validity arguments we can continue with the analysis .
the energy conservation equation for the kk modes in an expanding universe is @xmath42 to convert this into a differential equation with respect to temperature , we use the standard expression for the radiation dominated epoch , @xmath43 where @xmath44 is the total number of relativistic degrees of freedom in the plasma @xmath45
. these degrees of freedom are given in the appendix .
the radiation energy density is written @xmath46 then we can write @xmath47 the temperature range over which we shall consider the production of kk modes will be from the end of the qcd phase transition ( 170 mev ) downwards . the temperature sensitivity of the kk mode generation rate means the vast majority of any mode production occuring in this range will take place at temperatures just below @xmath48 , or whatever upper temperature we run the equations from .
because the detailed calculations of @xmath14 have been made for the nucleon - nucleon system , we should only use them once the qcd phase transition is over .
that is not to say that there will be no kk mode production by gravi - bremsstrahlung before the phase transition in the quark - gluon plasma for a first approach to this calculation ] , merely that that has not been explicitly calculated at the time of publication and the cosmology of a universe with large extra dimensions becomes increasingly uncertain as one pushes back towards higher energies .
the constraints obtained in this paper will assume that there are no kk modes excited at @xmath48 , only those generated below this temperature will be considered .
inclusion of any kk modes generated above @xmath48 would only strengthen the constraints obtained below .
the qcd phase transition is well below the nucleon mass threshold so the neutrons and protons are non - relativistic and the distribution boltzmann . @xmath49 here g is the number of degrees of freedom , 4 for a nucleon , and @xmath17 is the mass of the nucleon .
any chemical potential corresponding to baryon number conservation will already be small and will decrease with the expansion so will not be included in these calculations . finally there is a multiplicity factor @xmath50 which takes into account the fact that both n - n and n - p scattering will give rise to kk modes . bringing these results together
, the rate of energy density generated in the form of kk modes with respect to temperature @xmath51 will be given by @xmath52 the ratio of time vs. the epoch of radiation - matter equality @xmath53 is @xmath45 @xmath54 and the temperature of radiation @xmath55 is inversely proportional to the scale factor @xmath29 . we know that the temperature of the cosmic microwave radiation today is equal to @xmath56 mev . if we also assume that the minimum age of the universe is @xmath57 as determined by the mean observed age of globular clusters @xmath58 then by using equation ( @xmath59 ) we can obtain a maximum temperature for @xmath60 , the temperature at radiation - matter equality .
@xmath61 this is the limit we have to ensure is not broken by an early @xmath53 due to overproduction of kk modes .
limits on the minimum value of @xmath6 obtained for n=2 , 3 and 4 for values of @xmath48=100 , 135 and 170 mev are shown in table @xmath62 .
these values of @xmath48 are quite low and therefore conservative as a higher @xmath48 rapidly increases the number of modes produced .
an example of a situation previously permitted by sn1987a calculations was @xmath63 tev .
as can be seen in figure [ fig2 ] this is no longer permitted even if one assumes that @xmath48 is as low as 100 mev as it leads to an unacceptably high @xmath60 .
if one allows the kk mode production to begin at 170 mev then @xmath6 must be greater than @xmath64 tev for n=2 .
this is no longer a solution of the hierarchy problem . even if one were to claim that the calculations in this paper were only valid for temperatures far below the qcd phase transition there would still be many other mechanisms by which kk modes could be produced via the bremsstrahlung or annihilation processes of other particles @xmath65 .
it is possible that these constraints could be avoided by some new cosmological evolution which occurs above 1 mev to repect nucleosynthesis but at a temperature below about 100 mev to avoid overproduction of kk modes . until that is achieved
gravi - bremsstrahlung processes in the context of large extra dimensions appear to be at odds with cosmology .
99 arkani - hamed , n . , dimopoulos , s. and dvali , g. , phys.rev . *
d59 * , 086004 , ( 1999 ) ahern , s.c . ,
norbury , j.w . and
poyser , j.w .
, phys.rev . * d62*,116001 , ( 2000 ) cullen , s. and perelstein , m. , phys.rev.lett.*83 * , 268 , ( 1999 ) , barger , v. , han , t. , kao , c. and zhang , r.j .
, phys.lett.*b461 * , 34 , ( 1999 ) hanhart , c. , phillips , d.r .
, reddy.s . and savage , m.j
. , nucl - th/0007016 hall , l.j . and smith , d. , phys.rev .
* d60 * , 085008 , ( 1999 ) han , t. , lykken , j.d . and
zhang , r - j . , phys.rev . *
d59 * , 105006 , ( 1999 ) wilson , r. , the nucleon - nucleon interaction , john wiley , ( 1963 ) de teramond , g.f . and gabioud , b. , phys.rev .
* c36 * , 691 , ( 1987 ) kaloper , n. and linde , a. , phys.rev . * d59 * , 101303 , ( 1999 ) , lyth , d.h . ,
hep - ph/0003120 kolb , e.w . and turner , m.s . , the early universe , addison wesley , ( 1994 ) arndt , d. , nucl - th/0102040 krauss , l.m . , phys.rept . * 333 * , 33 - 45 , ( 2000 ) peacock , j.a . , cosmological physics , cambridge university press , ( 1999 ) | we calculate the production of massive kaluza - klein(kk ) modes via nucleon - nucleon gravi - bremsstrahlung in the early universe .
overproduction of these states would result in early matter domination and therefore a low age for the universe so it is possible to place constraints on the number and size of large extra dimensions .
the constraints are stronger than those from sn1987a - for 2 large extra dimensions and @xmath0170 mev , we show the fundamental scale must be larger than 1,000 tev . |
ngc 4636 is one of the nearest and , at @xmath22@xmath3 , one of the most x - ray luminous `` normal '' ellipticals .
ngc 4636 lies in the outskirts of the virgo cluster , 10 or 2.6 mpc on the sky to the south of m87 , for a distance to ngc 4636 of 15 mpc ( tonry et al .
2001 ) . as found for most luminous , slowly rotating ellipticals , the optical surface brightness of ngc 4636 flattens in the inner regions .
the central region , as seen in short exposures and with hst , has low eccentricity and is classed as an e0 .
however , as noted by sandage ( 1961 ) , at low surface brightness the galaxy is flattened ( e4 ) .
ngc 4636 has ionized gas in its core , but is unusual in that the gas velocities are uncorrelated with and significantly larger than those of the stars ( caon , macchetto & pastoriza , 2000 ) . a weak , extended radio source ( @xmath4 )
is observed at the galaxy center ( birkinshaw & davies 1985 , stanger & warwick 1986 ) , while loewenstein et al .
( 2001 ) place an upper limit of @xmath5 on nuclear x - ray emission .
einstein x - ray images first showed that , like other luminous elliptical galaxies , ngc 4636 was surrounded by an extensive hot gas corona ( forman , jones , & tucker 1985 ) .
while einstein observations allowed only an emission - weighted temperature to be determined , with rosat and asca , a modest increase in the gas temperature with radius was found , abundances in the halo were measured and a very extended x - ray component was detected ( awaki et al .
1994 , trinchieri et al .
1994 , matsushita et al .
1997 , finoguenov & jones 2000 , buote 2000 ) . from an einstein hri observation ,
stanger & warwick ( 1986 ) found an asymmetric gas distribution that they suggested could be the result of erratic large scale gas flows .
the chandra observatory allows us to study the structure of the x - ray halo around ngc 4636 with a limiting resolution of @xmath650 pc . in this paper
we present an analysis of acis observations that reveal unusual structures in the x - ray halo and describe a possible mechanism for producing these features .
ngc 4636 was observed with acis - s for 53 ksec on 26 - 27 january 2000 ( obsid 323 ) and with acis - i for 11 ksec on 4 - 5 december , 1999 ( obsid 324 ) .
we filtered the observations by selecting asca grades 0 , 2 , 3 , 4 and 6 and eliminating intervals of high background ( `` flares '' ) as well as bright columns or pixels due to instrumental effects or cosmic ray afterglows .
the remaining `` good '' times are 41286 seconds in the acis - s observation and 5989 seconds in the acis - i observation .
1 shows the region of the acis - s3 ccd centered on ngc4636 in the energy range from 0.5 to 2.0 kev .
this image shows a bright central region surrounded by armlike structures . at a distance of 15
mpc , these structures are observed to extend @xmath68 kpc from the galaxy center . while the features toward the northeast and southwest are the brightest and appear symmetric around the galaxy center , the acis - s image shows a third , fainter arm northwest of the galaxy , that shares symmetries with the two brighter x - ray arms .
in particular the sharp edges along the bright southwest and faint northwest arms define a parabola that also traces the southeast edge of the bright galaxy core .
part of this parabola is mirrored in the bright ne arm .
in addition to these features , fig . 1 shows structure in the eastern part of the halo .
in particular east of the galaxy center is a low surface brightness region , with 1 elongated regions of enhanced emission marking its northeast and southeast boundaries . on smaller scales , as fig .
2 shows , the x - ray emission in the galaxy core is elliptical ( position angle @xmath6320 ) and has a 10 long by 1 wide `` gap '' running north - south , where the surface brightness is about half that in the adjacent regions . to better define the halo structures , we generated a radial profile of the x - ray emission excluding the brighter regions noted above , as well as point sources , and used this profile to produce a smooth , radially symmetric , nonparametric two dimensional model for the surface brightness . fig .
1 ( right ) shows the result of subtracting this model from the image on the left and smoothing the remaining emission .
most prominent in this figure are the enhanced regions of emission and the cavities to the east and west of the nucleus . fig .
1 also shows extended x - ray emission that is brighter to the west of the galaxy .
this is likely the very extended emission found from rosat and asca observations ( trinchieri et al .
1994 and matsushita et al .
although this extended emission was suggested to be symmetric around ngc 4636 , the extended emission in the chandra acis - s image ( and in the acis - i image ) is brighter west of the galaxy . to study the bright armlike structures , we generated surface brightness distributions across the ne and sw arms and measured the gas temperature in several regions along and across these arms .
as an example , fig . 3 ( left ) shows the x - ray surface brightness distribution projected along a rectangle ( @xmath7 ) centered 65from the nucleus and aligned parallel to the `` leading '' edge of the sw arm . as suggested in fig . 1 ,
the projections show that the `` leading '' edges are remarkably sharp .
both arms show changes in brightness by about a factor of two on scales of a few arcseconds .
the surface brightness falls more gradually toward the trailing edge . to measure the gas temperature in these features , and throughout ngc 4636 , we first identified and excluded regions around 127 point sources in the acis - s3 ccd
. to identify sources we generated a 0.5 to 2.0 kev image without filtering for background flares .
chandra s excellent angular resolution results in small `` cells '' for point sources , particularly near the telescope aim point .
thus for point source detection , the benefit of the longer exposure time gained by including times of high background outweighs the modest increase in the local background .
we omitted point sources with at least nine counts .
we also examined a `` harder '' 2.0 - 5.0 kev image to insure that no highly absorbed sources were missed .
we extracted spectra for a variety of regions from the `` cleaned image '' and fit these using xspec over the energy range from 0.5 to 3 kev .
although the focus of this paper is on the features in the hot halo , we first comment on the overall temperature structure . for several regions within the central bright core , we measured gas temperatures of 0.5 kev ( with 90% uncertainties less than 0.03 kev ) .
outside the central region , we observe a gradual rise in temperature from 0.5 to 0.7 kev within a radius of three arcminutes .
this is consistent with the modest rise in temperature with radius reported from rosat .
for the ne and sw arms , we measured the gas temperature both along and across these features . along the arms
, we found the same increase in gas temperature from 0.5 to 0.7 kev observed elsewhere in the halo . to reduce any biases that this radial temperature change could produce in looking for tempera- ture changes across the arms , we extracted the spectra across each arm in two regions , centered @xmath8 and @xmath9 from the galaxy center . for each region , we extracted spectra in a series of rectangles ( @xmath10 by @xmath11 ) whose long sides were parallel to the sharp edge of each arm .
for the arm regions closer to the galaxy center , where the halo emission is bright , we found no significant changes in the gas temperature across the arms .
in particular , in the inner spectral regions of the ne arm , the best fit temperatures range from 0.57 to 0.65 kev ( with 90% uncertainties of 0.05 kev ) , while in the sw inner arm , the best fit temperatures range from 0.60 to 0.68 kev .
however , in the outer region of the sw arm , the spectra show significant changes in the gas temperature across the arm ( see fig . 3 right ) . for this region of the sw arm ,
if we take the temperature measurement south of the edge as the ambient temperature and fit the temperature in the arm with two components , one fixed at the temperature and normalization of the ambient region , the temperature of the second component is @xmath12 kev . in the ne arm , the temperature measured in the rectangle along the leading edge of the arm had the highest temperature ( @xmath13 kev ) , but this was not significantly higher than in other regions .
the smaller and less significant temperature changes in the ne arm may be due to contamination from emission from the other structures in the eastern part of the halo .
while the x - ray features in ngc 4636 are so far unique , they share properties with structures seen in the hot gas in other galaxies and in clusters .
the symmetric , parabolic regions of brighter emission resemble the x - ray bright filaments observed around the radio lobes in the elliptical galaxy m84 ( finoguenov and jones 2000 ) and in the hydra a and perseus clusters ( mcnamara et al .
2000 , fabian et al .
however in ngc 4636 , no large radio lobes are observed .
in addition while the sharpness of the edges of the ne and sw arms appears similar to the sharp edges found along `` fronts '' in clusters ( markevitch et al .
2000 , vikhlinin et al . 2001 ) , the cluster `` fronts '' are cold , while those in ngc 4636 are hot . while the presence of sharp fronts suggests the possibility of an ongoing merger , the east - west symmetry of the halo structures , the similarity of this structure to that seen around radio lobes , as well as the lack of a disturbed morphology in the stellar core or in the stellar velocities ( caon et al .
2000 ) led us to consider a nuclear outburst as the underlying cause .
in particular we interpret the bright sw arm , the fainter nw arm and the bright ne arm as the projected edges of paraboloidal shock fronts expanding about an east west axis through the nucleus .
a shock model is also consistent with the evacuated cavities to the east and west of the central region ( fig 1 right ) . with this interpretation
, we used the size as well as the surface brightness and temperature measurements of the sw arm to constrain a simple shock model . from our spectral measurements of the sw arm ,
the gas outside the arm has an ambient temperature @xmath14 kev . taking the 1 kev temperature of the second spectral component in the sw arm as the postshock temperature
, we can estimate the strength of the shock .
such a temperature jump by a factor of 1.75 arises from a shock having a density jump of a factor of 2 .
the corresponding pressure jump is a factor of 3.5 and the mach number is 1.73 , giving a shock velocity of 725 km s@xmath15 .
we computed the evolution of a shock front produced by a point explosion of @xmath16 ergs in a uniform density gas .
the state of this model is completely determined by the strength of the shock .
we chose a time when the density jump is a factor of two for comparison with the observations . in our model , the surface brightness profile is determined by embedding the shocked region in a uniform cube of unshocked gas .
the size of this cube relative to the size of the shocked region determines the amount of emission from unshocked gas , and hence the size of the jump in surface brightness across the shock front .
the length of the cube , @xmath17 , is determined as the depth of gas , with the density of 0.012 @xmath18 measured at @xmath19 from the nucleus , that gives the observed surface brightness at this radius .
using beta model parameters for the undisturbed gas measured from the x - ray surface brightness profile ( @xmath20 , core radius @xmath21 ) gives @xmath22 ( 10.4 kpc ) . at a radius of 65 , the diameter of the shocked region in ngc 4636 is @xmath23 , so the ratio of @xmath17 to the shock radius is @xmath24 . with this value
, we obtain the surface brightness distribution across the shock shown in fig .
we note that the shape of the shock front is not sensitive to the shock strength .
the agreement between the surface brightness profiles derived from the observations and from the model supports our interpretation of these features as shock fronts .
closer to the galaxy , where the preshock pressure is greater , the shock would be weaker with a smaller temperature jump .
this may account for the lack of an observed temperature jump in the inner arm regions .
our analysis assumes that the symmetry axis of the shocks lies in the plane of the sky .
however , as shown in fig 1 and 2 , close to the center of the galaxy , the bright rim of the western parabola can be traced to the southeast of the nucleus , while the edge of the eastern parabola can be traced to the northwest of the nucleus .
this overlap of the bases of the parabolas ( the shock fronts ) implies that the symmetry axis can not be exactly in the plane of the sky . to determine the axis orientation
, we would need the intrinsic shape of the shocked regions .
this is constrained by the agreement between the observed and model shock profiles , and by the expectation that the shocked regions are not highly elongated .
both suggest that the symmetry axis is not far from the plane of the sky , but detailed modelling is needed to constrain this further .
both the size and symmetry of the apparent shocks point immediately to the nucleus as their energy source . scaling from our simple model ,
the age of the shocks is @xmath25 years and the total energy driving them is @xmath26 ergs .
this implies a modest mean nuclear output of @xmath27 .
however the absence of either strong x - ray or radio emission suggests that the agn output is not constant over time .
if the power of the outburst were near the eddington limit for a @xmath28 black hole ( magorrian et al .
1998 ) , then the required energy could be released in only @xmath29 years .
this raises the possibility of very brief periods of quasar rebirth .
we suggest that the other emission features in ngc 4636 s eastern halo originated from another nuclear outburst . the east - west asymmetry may result from projection effects or because the shocks have broken through to the east , but so far have been confined by the denser medium west of the galaxy .
such outbursts could cause the large chaotic velocities found in the emission line gas .
the narrow waist in the shocked region close to the nucleus suggests , in the simplest interpretation , that the energy driving the shocks was injected off - center , possibly via jets . although the shape of the shocked regions favors off - center injection , the presence of only a weak , small - scale radio source in ngc 4636 argues against jets .
alternatively , since the speed of a strong shock is @xmath30 , if the unshocked gas is sufficiently dense close to the plane of the nucleus ( perpendicular to the symmetry axis ) , then energy injected close to the nucleus can be channeled to produce shocks of this general shape .
in fact , there is no real distinction between these ; it is simply a matter of how far from the nucleus the energy is injected .
the hot gas in the centers of elliptical galaxies has a high density and thus a short cooling time .
when the gas cools appreciably it forms a dense disk due to rotation ( brighenti and mathews 2000 ) .
if this cooled gas fuels a nuclear outburst that deposits energy close to the nucleus , the presence of the dense disk would channel the energy and produce shocks like those observed in ngc 4636 .
thus ngc 4636 appears to demonstrate the energy feedback process invoked to prevent the deposition of large quantities of cooled gas in the centers of galaxies and clusters ( tabor & binney 1993 , churazov et al .
2001 , david et al .
2001 ) . in a galaxy with the x - ray luminosity of ngc 4636
, outbursts would need to occur every @xmath31 y to prevent the accumulation of a significant amount of cooled gas . if outbursts causing shocks occur every @xmath32
@xmath33 y , then ngc 4636 and other ellipticals could be in this shock phase 1 10% of the time .
if the outburst is fueled by cooled gas , then the repetition time must equal the central cooling time of the hot gas following an outburst
. a typical elliptical galaxy , with a central cooling time of a @xmath34 years , would be about half way through this cycle .
we recall that ngc 4636 s x - ray luminosity is unusually high for its absolute magnitude compared to other early type galaxies , ( forman , jones & tucker 1985 ) .
this high x - ray luminosity may be due to the advanced state of cooling that brought on the outburst , and may be further enhanced by the shocks .
when the gas halo returns to hydrostatic equilibrium , its x - ray luminosity could decline by an order of magnitude or more , in which case the estimated outburst intervals would be much less frequent .
such excursions in x - ray luminosity would help explain the broad @xmath35
@xmath36 relation for elliptical galaxies .
interactions between radio sources and the hot interstellar and intergalactic medium in elliptical galaxies and clusters have been observed in a number of systems , e.g. cen a ( kraft et al . 1999 ) , m84 ( finoguenov & jones 2000 ) , perseus ( bohringer et al .
1993 ) , hydra a ( mcnamara et al .
2000 ) , and a2052 ( blanton et al . 2001 ) .
however , in these examples , the main observational characteristic is displacement of hot gas by the radio emitting lobes , while in approximate pressure equilibrium with the gas . in ngc 4636
we see evidence for a nuclear outburst , but no large radio lobes and only a relatively weak ( 10@xmath37 ) central radio source .
we suggest that the more common x - ray and radio lobe structures ( e.g. m84 ) represent a later stage of the process seen in ngc 4636 .
this raises the issue of whether the nuclear outbursts that provide significant heating are primarily radio outbursts , or whether radio outbursts occur in a phase following such an outburst .
the latter scenario is consistent with the modest ratio of jet mechanical power to radio power found in hydra a ( mcnamara et al . 2000 ) . in summary
, the chandra image shows new features in the x - ray halo of ngc 4636 that can be explained as the results of shocks produced by nuclear outbursts .
we suggest ngc 4636 is an example of the feedback process that prevents the deposition of significant amounts of cold gas by cooling flows .
the lack of extended radio jets or lobes suggests that the energy from the agn may be released by some more direct process , rather than via a radio outburst .
we acknowledge stimulating discussions with e. churazov , r. mushotzky , and h. donnelly .
this work was supported by nasa contracts nas8 - 38248 , nas8 - 39073 , the chandra science center , and the smithsonian institution .
pejn gratefully acknowledges the hospitality of the center for astrophysics . | chandra images show symmetric , 8 kpc long , arm - like features in the x - ray halo surrounding ngc 4636 .
the leading edges of these features are sharp and are accompanied by temperature increases of @xmath0% .
these properties , along with their scale and symmetry , suggest that the arm - like structures are produced by nuclear outburst driven shocks .
we interpret these observations as part of a cycle in which cooling gas originally fueled a nuclear outburst about @xmath1 years ago leading to shocks that reheat the cooling gas , thus preventing the accumulation of significant amounts of cooled gas in the galaxy center and temporarily starving the central agn . |
a decade of experimental signals@xcite for qcd hybrids ( in particular with @xmath0 ) has culminated in the claimed observation of three such states at brookhaven@xcite in the last year .
the question of the nature of qcd hybrids has thus become topical .
furthermore , lattice gauge calculations are now at the point of accurately determining light hybrid masses . in view of these developments
, it is of interest to compare models of strong ( low energy ) qcd with lattice data to determine their viability and to explicate and guide current experimental efforts .
it is often stated that a hybrid is a hadron consisting of valence quarks and glue .
however , one must specify what is meant by the notion of valence glue for this statement to be useful .
there are two broad ideas in this regard : it is some sort of string or flux tube@xcite or it is an effective constituent confined by a bag@xcite or potential@xcite . as an example of the importance of choosing correct degrees of freedom , we mention the simple problem of determining the number of components of a constituent gluon .
it has been suggested that a massive constituent gluon should be transverse so as to maintain consistency with yang s theorem@xcite .
however it was noted that this is inconsistent with the requirements of lorentz invariance .
thus , for example , @xmath1 glueballs are expected to exist and lattice calculations indicate that they are quite heavy ( roughly 3 gev)@xcite .
such a state may not be constructed from two transverse constituent gluons ( yang s theorem ) and therefore may be expected to have a mass of roughly @xmath2 gev .
however massive vector gluons have no such constraint and one therefore expects them to have a mass of approximately @xmath3 gev .
the nature of the appropriate effective degrees of freedom for glue can only be determined by a long process of calculation and comparison with experimental and lattice data .
there are , however , a few indications that low energy glue is string - like .
perhaps the most compelling of these are lattice calculations of energy , action , or field densities between static color sources which are reminiscent of flux tubes@xcite .
an intriguing clue is also provided by the spin splittings of heavy quarkonia . it is known that an effective interaction free of long range exchange spin - orbit coupling is needed to reproduce the mass splitting of the @xmath4 heavy quarkonium multiplets
. however , an analysis of qcd in the heavy quark limit convincingly demonstrates that obtaining such an effective potential requires that low energy glue must be string - like@xcite .
alternatively , pointlike models of low energy glue have a long history , originating with mit bag model calculations of barnes@xcite and others@xcite .
horn and mandula@xcite were the first to consider a potential constituent glue model of hybrids .
their hybrids consisted of constituent quarks and pointlike , massless , spinless , and colorless glue in a nonrelativistic potential model .
the confining potential was taken to be linear with a string tension given by the ratio of color casimir operators : @xmath5 .
the authors noted that the two body @xmath6 potential is anti - confining in the color octet channel and has a repulsive coulomb spike at short distances .
they argued that this is unphysical and hence choose to neglect this term in the interaction .
it is clear that a great many simplifying assumptions have gone into the construction of this model .
it is our purpose to compare a more sophisticated version of the model to lattice data to learn something about these assumptions . in the following
we employ a model field theoretic hamiltonian of qcd .
the model incorporates linear confinement at low energy and evolves into perturbative qcd at high energy .
a nontrivial vacuum is used to generate constituent quark and gluon masses .
the eigenvalue equation is derived for a @xmath7 system where the quarks are static .
the resulting adiabatic potential surfaces are then compared to recent lattice results .
we conclude that the simple picture of glue as a pointlike constituent particle reproduces the general behaviour of the lattice results but fails to yield the correct level orderings .
furthermore , other models which regard the gluonic degrees of freedom as pointlike ( eg . , @xcite , @xcite ) do not contain sufficient degrees of freedom to generate all of the adiabatic surfaces .
thus constituent glue models appear to fail to describe hybrids .
this stands in contrast to string or bag - like models which , although disagreeing on details , capture the rough structure of the lattice data .
the starting point for our description of hybrids is the following model hamiltonian : @xmath8 \psi({\bf x } ) + { 1\over 2 } \int d{\bf x } \left [ |{\bf e}^a({\bf x})|^2 + |{\bf b}^a(x ) |^2 \right ] \nonumber \\ & + & { 1\over 2}\int d{\bf x } d{\bf y } \rho^a({\bf x})v(|{\bf x } - { \bf y}| ) \rho^a ( { \bf y})\end{aligned}\ ] ] where the color charge density is @xmath9 and the potential is given by @xmath10 and @xmath11 .
the quark mass appearing in this hamiltonian is the current mass . to be phenomenologically successful
, constituent quark masses must be generated in some way .
this may be achieved by employing a bcs vacuum ansatz ; the gap equation which follows from minimizing the vacuum energy density , @xmath12 , ( where @xmath13 represents the bcs trial vacuum ) gives rise to a low energy constituent quark mass of roughly 200 mev@xcite .
a similar calculation in the glue sector yields a gluon dispersion relation which is well - approximated by @xmath14 with @xmath15 mev and @xmath16 gev .
one sees that a constituent gluon mass of approximately 800 mev has been generated@xcite .
hadrons are then constructed on top of the bcs vacuum @xmath13 by employing a basis truncation ( typically tamm dancoff or random phase ) and solving the resulting bethe - salpeter equation .
this approach has been used to derive the low lying spectrum of glueballs and agrees remarkably well with lattice data@xcite .
it should be noted that the dynamical gluons are transverse so that yang s theorem holds and the difficulty mentioned in the introduction does not arise . in the following ,
the parameters of the model are fixed to the @xmath17 potential derived in the lattice calculation of juge , kuti , and morningstar@xcite ( see fig . 3 below ) .
the fit yields @xmath18 and @xmath19 gev@xmath20 .
the final parameter , @xmath21 serves as an ultraviolet cutoff on the linear confinement potential .
its value was determined in ref .
@xcite by fitting the gluon condensate , and will be set at 4 gev in the following .
we shall consider the static quark limit in the remainder of this work so that quark masses ( and the quark sector of the bcs vacuum ) become irrelevant .
the model presented here may be considered as a simplified version of the coulomb gauge qcd hamiltonian where the effects of nonabelian gauge couplings have been modeled by the linear confinement term .
furthermore , second order transverse gluon exchange is suppressed by the heavy quark masses . in principle , this approach allows the elimination of the ultraviolet scale @xmath21 via renormalization .
a method for achieving this which is appropriate for nonperturbative hamiltonian - based calculations is described in ref .
@xcite . in the tamm dancoff approximation ,
hybrids are constructed as @xmath22 excitations of the bcs vacuum . for the heavy hybrids considered here , the ( anti)quarks serve as static color sources ( sinks ) and the gluons are constituent particles as described above .
we choose to work in the diatomic molecule " basis because this facilitates comparison with the lattice results of ref .
thus the hybrid state may be written as @xmath23 wigner rotation functions are written as @xmath24 in this expression , @xmath25 is the distance between the @xmath17 pair , @xmath26 is the gluon helicity , and @xmath27 is the radial hybrid wavefunction in momentum space .
the gluon polarization wavefunction , @xmath28 is given by @xmath29 and @xmath30 .
the two cases , @xmath311 , -1 , represent transverse magnetic and transverse electric hybrids respectively . finally , @xmath32 is the projection of the gluonic angular momentum onto the @xmath17 axis , @xmath33 is the total angular momentum of the gluon , and @xmath34 labels the radial basis state .
we note that employing helicity basis gluon creation operators makes this expression significantly more compact than the canonical basis .
the salpeter equation which follows from this ansatz and the hamiltonian in eqn .
( 1 ) may be obtained from the following matrix element : @xmath35 \nonumber \\ & + & { 3 \over 8 } \int \int d \vec q d \vec k\ , \varphi_{n_g ' j_g'}^*(k ) \varphi_{n_g j_g}(k)\ , v(k - q ) \left [ { \omega(k)^2 + \omega(q)^2 \over \omega(k ) \omega(q ) } ( 1 + ( \hat q \cdot \hat k)^2 ) \right ] \nonumber \\ & - & { 3 \over 4 } \int\int d \vec q d \vec k\ , \varphi_{n_g ' j_g'}^*(q)\varphi_{n_g j_g}(k ) \ , v(k - q ) \left({\rm e}^{i { r\over 2 } \cdot ( k - q ) } + { \rm e}^{-i { r\over 2}\cdot ( k - q ) } \right ) { \omega(k ) + \omega(q)\over \sqrt{\omega(k ) \omega(q)}}\times \nonumber \\ & & \times { \cal d}^{j_g*}_{\lambda \lambda'}(\hat k ) { \cal d}^{j_g'}_{\lambda'\sigma'}(\hat q ) \sqrt{2 j_g + 1 \over 4 \pi } \sqrt{2 j_g'+1\over 4 \pi } \chi_{\lambda \lambda'}^\xi \chi_{\sigma \sigma'}^{\xi ' } { \cal d}^1_{\mu \lambda}(\hat k ) { \cal d}^{1*}_{\mu\sigma}(\hat q ) \nonumber \\ & + & \left({1\over 6 } v(r ) + { 4\over 3}v(0)\right ) { \cal n}_{n_g ' j_g ' n_g j_g } \delta_{j_g ' j_g } \delta_{\lambda ' \lambda } \delta_{\xi ' \xi } \end{aligned}\ ] ] an overall @xmath36 is understood and @xmath37 is the wavefunction normalization factor .
the two extra wigner rotation matrices arise from converting the cartesian basis implicit in eqn .
( 1 ) and in the expression for the color current to the helicity basis , @xmath38 , where @xmath39 and @xmath40 are canonical and helicity polarization vectors respectively .
the first term in this expression is the gluon kinetic energy , the second is the gluon self energy , the third is the gluon potential , and the fourth is the @xmath6 potential and self energies for static quarks in a color octet .
the presence of the gluon and quark self energies assures the infrared finiteness of the salpeter equation .
this appears to be a general feature of color singlet states in our approach@xcite .
the last task is to identify the diatomic quantum numbers used to label the states . in the jacob - wick convention the action of parity on glue
is given by @xmath41 where @xmath42 is the azimuthal angle of @xmath43 and @xmath44 is the intrinsic gluon parity .
thus the hybrid states given in eqn .
( 4 ) are eigenstates of gluonic parity with @xmath45 a reflection of the glue degrees of freedom through a plane containing the @xmath17 axis leaves the hamiltonian invariant and when acting on the states it takes @xmath46 . for @xmath47 one
thus has doubly degenerate states .
we call this operation @xmath48-parity and perform it by setting @xmath49 and taking @xmath50 which may be achieved by a parity operation followed by a rotation through @xmath51 about the y - axis . the action of @xmath48 on a single transverse gluon state , @xmath52 is therefore given by , @xmath53 where @xmath54 .
for the states defined in eq .
( [ basis ] ) one has @xmath55 where the relations @xmath56 and @xmath57 were used . for @xmath58 the y - diagonal states may thus be written as @xmath59 where @xmath60 and , @xmath61 for @xmath62 we simply have , @xmath63 to allow for easy comparison with conventions used elsewhere , we summarize the transformation properties of @xmath64 states under total parity , @xmath65 , and charge conjugation @xmath66 , including the quark degrees of freedom . @xmath67 @xmath68 where @xmath69 = @xmath70 = @xmath71 = -1 .
the states introduced in eq .
( [ ybasis ] ) are therefore also eigenstates of combined @xmath72 , with @xmath73 after dividing out the quark portion of @xmath72 we are left with @xmath74 the salpeter equation is solved by expanding the radial wavefunction in a complete basis and by diagonalizing the resulting hamiltonian matrix .
the evaluation of the matrix elements is greatly facilitated by performing the angular integrals analytically .
furthermore , we found it expedient to do the remaining numerical integrals with the potential in position space since the integrand is less oscillatory for large argument in this case . the plane waves and potential were expanded in a double series of wigner functions yielding a total of ten wigner functions .
the angular integral then evaluates to a sum over a product of six clebsch - gordon coefficients .
as discussed above eigenstates of the hamiltonian may be labelled with the projection of the angular momentum onto the @xmath17 axis , the product of gluonic parity with charge conjugation , @xmath75 and the y - parity eigenvalue .
for @xmath58 the two y - parity eigenstates are degenerate .
states are therefore denoted by @xmath76 where @xmath77 are denoted by @xmath78 , @xmath79 , @xmath80 ; @xmath81 = @xmath82 or @xmath83 for even or odd parity respectively ; and @xmath84 = @xmath85 . the total gluonic angular momentum
, @xmath33 is not a good quantum number .
we have found however , that the sum over radial basis states and gluon angular momentum saturates quickly .
for example , fig 1 . shows the ( approximately ) exact wavefunction for @xmath86 with @xmath87 and a variational single gaussian orbital . the next contribution to the @xmath86 eigenstate
has @xmath88 , this is shown as the dotted line in the figure . evidently the eigenstate is dominated by the lowest gluonic angular momentum component and the single gaussian approximation is quite accurate .
this remains true for all @xmath25 studied here , with the @xmath88 component rising to 13% of the wavefunction for @xmath89 gev@xmath90 .
these observations support traditional approximations made for hybrids : the use of simple gaussian wavefunctions @xcite and the truncation of the adiabatic schrdinger equation to lowest @xmath33@xcite . to fig . 1 .
@xmath86 wavefunctions .
exact result ( dashed line ) , single gaussian approximation ( solid line ) , @xmath91 component ( dotted line ) .
the results for the gluon spectrum are presented as a function of the static quark separation in figs . 2 and 3 .
they are plotted in terms of the lattice scale , @xmath92 = 2.32 gev@xmath90@xcite and the potentials have all been normalized by subtracting an overall constant given by @xmath93 . in figs .
2a - c we compare the recent lattice results from ref .
@xcite with the predictions of the flux tube model @xcite .
the flux tube model was motivated by the strong coupling limit of the qcd lattice hamiltonian .
it is based upon a nonrelativistic , small oscillation approximation to motion of the colored links in a topological sector where there are no overlapping ( color representations of higher dimension ) or disconnected links . under parity ( charge )
conjugation the spatial ( color ) orientation of lattice links is reversed , thus the nonrelativistic `` beads '' of the flux tube model are assumed to flip orientation with respect to the position of the quarks under parity and to have positive intrinsic charge conjugation parity . the lowest solid line in fig .
2a is a fit to the @xmath17 potential given by @xmath94 which corresponds to the ground state lattice @xmath95 potential .
the other solid lines in fig .
2 show the flux tube potential as given by @xmath96 with @xmath97 and @xmath98 .
the latter represents the total number of right - handed ( @xmath99 ) and left - handed ( @xmath100 ) transverse phonon modes weighted by the phonon momentum ( @xmath101 ) .
we note that the authors of ref .
@xcite included a coulomb term in this expression , which is incompatible with the lattice results .
the flux tube model predicts the first excited @xmath102 to be split by @xmath103 from the ground states , and two degenerate @xmath104 and @xmath105 potentials at @xmath106 . in the flux tube model the lowest @xmath79 state
is predicted to be the @xmath107 .
it is split from the @xmath6 ground state by @xmath108 and is followed by the @xmath103 , @xmath109 potential .
the two lowest @xmath80 states , @xmath110 and @xmath111 correspond to @xmath103 and @xmath106 respectively .
the flux tube model fits the first excited state , @xmath107 quite well over a wide range of the quark separation .
it is , however , the only surface to do so at small distance .
furthermore , this may be a fluke due to the particular choice of the short distance cutoff of the @xmath112 term employed in ref .
alternatively , at large distances , the system is expected to act like some sort of string with an excitation energy given by @xmath112 .
the splittings do indeed appear reasonable for all the surfaces considered .
it is , however , disconcerting that the @xmath79 surfaces diverge from the flux tube model predictions for @xmath113 .
this must be taken as an indication that the simple flux tube model considerations fail for more complex gluonic configurations . in fig.3a - c
we plot the results of our calculations for @xmath78 , @xmath79 and the @xmath80 potentials with the flux tube results ( solid lines ) for comparison .
it is apparent that the @xmath109 surface lies below the @xmath107 surface while the @xmath111 lies below the @xmath110 .
this is opposite to the lattice results , indicating that the model fails to reproduce the expected level orderings .
the @xmath78 levels are also permuted with respect to the lattice calculations at small @xmath25 .
we note , however , that the correct level orderings may be reinstated by simply flipping the intrinsic parity of the gluon ( set @xmath114 in the expressions above ) .
the resulting surfaces agree reasonably well with the lattice at small distances .
for example , the @xmath107 and @xmath109 potentials are roughly 1 and 2 at the origin in both calculations .
we obtain values of approximately 3 and 3.5 for the @xmath111 and @xmath110 surfaces at the origin , similar to the lattice values of 2 and 3 respectively .
the agreement persists to intermediate @xmath115 , after which it is apparent that the model results approach the linear regime much too quickly with respect to the lattice .
this is a strong indication that more degrees of freedom become active as the quark separation increases beyond @xmath116 fm .
this is certainly sensible from the flux tube model point of view , where a large number of degrees of freedom are necessary to construct string phonons .
our results include the contribution due to direct interaction between the static quarks ( the last term of eqn .
this term was set to zero by horn and mandula@xcite because it anti - confines and therefore may be expected to not produce a flux tube between the quarks .
however , we note that it is responsible for producing a hybrid potential slope equal to that of the @xmath17 ground state potential , in keeping ( roughly ) with the lattice data .
thus we have chosen to retain this term .
note however , that this implies that there is a short distance repulsive coulomb spike which should appear at very short quark separation .
the appearance of this spike is , however , unphysical because the hybrid may emit a gluon and convert into a @xmath17 color singlet and a low lying glueball and this is energetically favorable for small @xmath115 .
equivalently , @xmath117 for @xmath118 fm .
thus the coulomb spike mentioned above is irrelevant and one should see , instead , a coulomb core for small @xmath115 .
this effect may be easily incorporated into the model presented here by allowing for coupling to the @xmath119 channel .
notice that no core is visible in the lattice data ( especially in @xmath107 , which is measured down to @xmath120 fm ) .
the authors of ref .
@xcite are currently examining this issue@xcite . to to to
the model presented here agrees moderately well with lattice calculations of adiabatic potential surfaces .
however , it disagrees in detail . in particular
, it is necessary to ignore the intrinsic parity of the gluon to obtain the expected level orderings .
although one may argue that this is in keeping with the expectations for the parity of a lattice link operator ( such as is employed in the development of the flux tube model ) , the lack of consistency is disconcerting .
furthermore , the model predicts surfaces which become linear too quickly with respect to the lattice .
we expect this flaw to persist in all constituent glue potential models ( such as @xcite and @xcite ) .
this is a strong indication that more degrees of freedom are required to describe soft glue at large interquark separation .
it is important to note that models which employ single spinless gluons do not contain sufficient degrees of freedom to reproduce the potential spectrum . in particular , single bead " flux tube models ( @xcite ) can not make @xmath109 or @xmath121 states while three dimensional bead models ( @xcite ) can not make @xmath122 states . thus including gluon spin is a minimal necessity in this class of models ( although the level ordering problem must be overcome as discussed above ) .
we also note that it is possible for spherical bag models to reproduce the lattice calculations at small quark separation , but that they fail at large @xmath25 .
furthermore , flux tube model or nambu - goto string models@xcite reproduce the lattice reasonably well for intermediate to large quark separations , but do not perform well at small @xmath25 or in detail at large @xmath25 . in particular
the potential separation does not appear to be the expected @xmath112 and the slopes do not appear to agree .
finally , the bag model of ref .
@xcite works reasonably well over all quark separations , although problems remain to be resolved in the @xmath78 states and for large @xmath25 . in summary
, it appears that some sort of flux tube is required to explain the lattice adiabatic hybrid potential surfaces at large quark separation .
this is in keeping with the conclusions of ref .
@xcite , where flux tubes were required to explain the spin splittings in heavy quarkonia . however , any flux tube model must attempt to incorporate the small @xmath25 behaviour of the potential surfaces ( and the seemingly anomalous behaviour at large @xmath25 ) .
this appears to be an indication that different degrees of freedom are required at small distances ( as is incorporated in , for example , the bag model of ref .
finally , the model presented here ( with the gluon parity reversal ) works moderately well in describing the potential surfaces and could provide a useful starting point to simple models of gluonic hadron properties ( note that the early transition to a linear potential seen in figs .
3 should not be relevant since the hadron wavefunction is exponentially suppressed in this region ) .
thus we expect that the success of the previous glueball spectrum calculation@xcite in this model ( where the gluon parity was irrelevant ) was not a fluke .
several benefits of the model are particularly relevant to hybrids ; light quarks may be easily incorporated , the effects of coupled channels may be examined , and the effects of light quark coupling to virtual transverse or coulomb gluons may be included .
the latter effect is of interest because it is excluded in quenched lattice calculations and may have a substantial effect on the hidden flavor hybrid spectrum .
the authors are grateful to c. morningstar for discussions and for providing us the preliminary lattice data of ref .
this work was supported by the doe under grants de - fg02 - 96er40944 ( es ) and de - fg02 - 87er40365 ( as ) .
p. hasenfratz , r.r .
horgan , j. kuti , and j.m .
richard , phys .
95b * , 299 ( 1981 ) ; t. barnes , f.e .
close , and f. de viron , nucl
. phys . * b224 * , 241 ( 1983 ) ; m. chanowitz and s. sharpe , nucl . phys . *
b222 * , 211 ( 1983 ) . | hybrid meson energies are calculated in the static quark limit with the dynamical quark model ( dqm ) . in the dqm , transverse gluons are represented as effective constituents with a dynamically generated mass .
hybrid masses are determined within the tamm - dancoff approximation for the resulting relativistic salpeter equation .
although the general features of the adiabatic potential surfaces correspond with lattice data , the results disagree on level orderings .
similar problems appear to exist in all constituent glue models of hybrids .
we conclude that constituent gluons do not accurately represent soft gluonic degrees of freedom .
the steps necessary to correct this deficiency are discussed . |
geometries in general relativity can possess various pathological features , and it is of interest how string theory , as a quantum mechanical theory of gravity , might or might not remove these pathologies .
curvature singularities are the most well - known problematic features , but another interesting class is closed timelike curves ( ctcs ) , which apparently enable time travel and render the usual notions of causal time evolution untenable . hawking conjectured that no geometry initially free of closed timelike curves would develop them , the so - called chronology protection conjecture @xcite ; but whether a quantum theory of gravity might remove closed timelike curves that are already classically present is another question . in a separate development ,
there has been considerable interest in using the ads / cft correspondence to build models of condensed matter physics ( for reviews , see @xcite ) .
the schrdinger spacetime , formulated in @xcite and embedded in string theory in @xcite , is of interest due to its isometry group matching the non - relativistic conformal symmetry @xcite of certain ultracold gases at unitarity , although a number of issues with the application of the geometry remain ( for a recent discussion of some aspects , see for example @xcite ) .
the schrdinger spacetime may be generated from the poincar patch of anti - de sitter space ( ads ) through the so - called
null melvin twist " @xcite , generalized to a wide class of cases in @xcite ( see also @xcite ) .
nonzero temperature is implemented through acting the twist on a geometry containing a black hole .
the appearance of rotation in such spacetimes is interesting and potentially relevant towards a description of rotating ultracold gases ; for a relativistic example , see @xcite .
it is also interesting from the point of view of gravity to understand what results from acting the null melvin twist on a broader class of spacetime metrics .
a natural step is then to apply the null melvin map to a rotating black hole .
these black holes , however , live in global anti - de sitter space , and the only spacelike isometries available to perform the melvin map are rotational isometries , distinct from the translational symmetries used to generate the original schrdinger spacetimes and its descendants . to understand the kind of spacetime that results , we study here the simplest case containing the essential features : the geometry resulting from acting the null melvin twist on global anti - de sitter space using a rotational isometry . since global ads
holographically realizes the same field theory as the poincar patch of ads , but living on a sphere instead of flat space , it is natural to hypothesize that the geometry that results in this case will realize the same non - relativistic field theory as the original schrdinger spacetime , but living on a sphere .
we find the resulting geometry is a homogeneous spacetime with a large isometry group , very similar to the symmetry group for the schrdinger spacetime , with the differences in the algebraic structure suitable for the space to indeed be dual to the non - relativistic system living on a sphere .
moreover , the original schrdinger spacetime can be recovered through a coordinate limit of blowing up the size of the sphere on the novel geometry .
however , there is a surprising feature of this new geometry : closed timelike curves exist .
these ctcs pass through every point in the spacetime , but have a minimal radius .
this aspect is strongly reminiscent of gdel spacetimes , both the original geometry @xcite and its supersymmetric descendants ( see for example @xcite ) .
consequently the spacetime we generate shares properties of both schrdinger and gdel systems , and we refer to it as the gdel - schrdinger spacetime .
one is thus led to ask : is this gdel - schrdinger spacetime pathological due to the closed timelike curves , or does string theory mitigate them in some way ?
we consider two distinct but complementary arguments suggesting that string theory does indeed avoid the difficulties of the closed timelike curves .
first , one may consider a holographic screen in the spirit of the covariant holographic conjecture @xcite .
for ordinary global ads space , this screen lives at the boundary , and hence it is natural to identify the boundary with the location " of the dual field theory degrees of freedom ; the entire spacetime is then holographically described by these modes . in the gdel - schrdinger case , however , the screen sits at a finite radius , interior to the smallest closed timelike curves ; a similar situation holds for the original gdel geometry and some of its supersymmetric descendants @xcite .
thus the degrees of freedom can be taken to live at the screen , and only the interior ( ctc - free ) part of the geometry is associated to the holographic dual description .
moreover , it is possible to place a brane probe inside the chronologically challenged region ; since a brane is a nonlocal object , it can sample an entire closed timelike curve and hence should be aware of the pathology .
it was described by dyson @xcite how for certain black hole spacetimes with closed timelike curves , probe branes may develop a repulson - type instability @xcite which one would expect to be resolved by the branes condensation , analogous to the enhanon mechanism @xcite .
a similar test was performed in the case of a supersymmetric gdel spacetime @xcite , where it was found that a bps supertube probe develops a negative kinetic term in the ctc region , implying an analogous instability . in close analogy
, we find that a d3-brane dual giant graviton " probe develops negative kinetic terms precisely in the acausal region , and thus would be expected to condense .
we thus hypothesize that while the geometry we obtain is a solution to type iib supergravity , it is not a true solution to type iib string theory . incorporating the condensed brane backreaction ,
a new solution sourced by the condensed branes should exist at large radius , while the small - radius geometry would persist ; according to the holographic screen argument , only this region would be necessary to describe the dynamics experienced by a given observer .
( for further work on chronology protection in string theory , see @xcite-@xcite . ) the null melvin map acting on rotating black holes would then produce holographic duals of rotating non - relativistic conformal field theories , whose analogous ctcs are presumably removed by a similar mechanism . in section [ melvinsec ] , we review the null melvin twist and the original schrdinger spacetime . in section [ globalsec ]
we apply the null melvin twist to global anti - de sitter space to generate the gdel - schrdinger spacetime , calculate the isometries and their algebra , and obtain ordinary schrdinger as a limit .
we demonstrate the existence of closed timelike curves larger than a certain critical radius in section [ screensec ] , compare the geometry to the gdel and supersymmetric gdel spacetimes , and demonstrate that like those geometries , an observer at the origin of gdel - schrdinger is associated to a holographic screen living inside the critical radius for ctcs .
finally , in section [ branesec ] we identify d3-brane giant graviton solutions in the gdel - schrdinger geometry and show that they develop negative kinetic terms precisely at the critical radius , suggesting an instability towards condensation of branes . in section [ conclusionsec ] we conclude .
the null melvin twist is an operation involving compositions of coordinate shifts and two t - dualities , which maps solutions to solutions in type iib supergravity @xcite .
it was used by @xcite to generate the schrdinger spacetime from ordinary anti - de sitter space in the poincar patch , with an added black hole leading to the finite - temperature generalization .
the melvin map was generalized to include its action on charged black holes and other , broader classes of spacetimes in @xcite ; our discussion follows @xcite .
( for further developments generalizing schrdinger spacetimes , see for example @xcite . )
the null melvin twist requires as input a ten - dimensional metric with timelike isometry @xmath3 and two spacelike isometries @xmath4 and @xmath5 .
the solutions we are interested in have a product form with a more general compact sasaki - einstein space , but since our focus will be on phenomena in @xmath6 , we will not do so . ]
@xmath7 , where @xmath8 and @xmath9 are timelike and spacelike coordinates on @xmath6 , respectively , and @xmath10 is the hopf coordinate on @xmath11 : @xmath12 where the 1-form @xmath13 defines the fibration of @xmath11 over @xmath14 .
the initial solution also has nonzero @xmath15 , @xmath16 with other rr fields and the nsns potential @xmath17 vanishing , and the dilaton constant @xmath18 .
one may always write the five - dimensional metric as @xmath19 with @xmath20 and @xmath21 running over the remaining three coordinates , and @xmath22 ; note the cross terms between the @xmath8 or @xmath9 and the other three coordinates are absorbed into @xmath23 , while a cross term between @xmath8 and @xmath9 is written as an explicit @xmath24 component .
the null melvin twist then produces a new string frame type iib supergravity solution , characterized by a parameter @xmath25 , given by the nsns sector @xmath26 with the rr sector unchanged , where @xmath27 is a scalar function and @xmath28 is a 1-form on @xmath6 : @xmath29 and the new five - dimensional metric is @xmath30 with @xmath31 . in the limit @xmath32 , the original spacetime is recovered .
one may also write the difference between the original metric and the new metric in terms of the one - form @xmath28 : @xmath33 the effective five - dimensional theory from dimensionally reducing on the ( squashed ) @xmath11 is a theory of a metric , a massive gauge field , and since @xmath34 , a single scalar : @xmath35 this action is a truncation of a three - scalar lagrangian described in @xcite ; we have set the ads scale to unity . when @xmath36 , this is just 5d einstein gravity with a cosmological constant .
the mode @xmath37 has no kinetic term in this string frame , but regains one in einstein frame , with @xmath38 : @xmath39 we next review the five - dimensional schrdinger solution and its symmetry group .
five - dimensional anti - de sitter space in the poincar patch may be written as ( again , the ads scale is unity ) : @xmath40 with @xmath41 .
defining @xmath42 we may apply the null melvin twist , resulting in the expressions @xmath43 and the massive vector @xmath44 since @xmath45 , the dilaton remains constant at @xmath46 , so we need not distinguish string and einstein frame metrics .
it is standard to reexpress this solution in terms of light cone coordinates , @xmath47 in terms of which we obtain the metric @xmath48 with @xmath49 , and the 1-form @xmath50 this is the schrdinger solution , so called because its isometries coincide with the non - relativistic schrdinger conformal symmetry group with dynamical exponent @xmath51 , in @xmath52 spatial dimensions .
the utility of non - relativistic conformal field theories ( nrcfts ) to ultracold gases at unitarity was described in @xcite , and the dual geometry ( [ pschro ] ) was first described in @xcite , and lifted to a solution of type iib string theory in @xcite .
we will denote the schrdinger solution as @xmath53 .
the schrdinger spacetime @xmath53 possesses nine isometries , @xmath54 corresponding in the dual field theory to the hamiltonian @xmath55 , particle number @xmath56 , momentum generators @xmath57 , @xmath58 rotations @xmath59 , galilean boosts @xmath60 , dilatations @xmath61 and special conformal transformations @xmath62 .
several subgroups of the algebra are noteworthy .
the commutators of @xmath55 , @xmath61 and @xmath62 are , @xmath63 = 2 i h \ , , \quad \quad [ c , h ] = i d \ , , \quad \quad [ c , d ] = 2 i c \,,\end{aligned}\ ] ] which upon defining @xmath64 form the generators @xmath65 of an @xmath66 subalgebra @xmath67 = i ( \eta_{ac } j_{bd } - \eta_{bc } j_{ad } + \eta_{bd } j_{ac } - \eta_{ad } j_{bc } ) \,,\end{aligned}\ ] ] with @xmath68 and @xmath69 .
furthermore , in @xmath70 spatial dimensions the @xmath71 constitute the generators of the standard @xmath70-dimensional rotational algebra @xmath72 , @xmath73 = i ( \delta_{ik } m_{jl } - \delta_{jk } m_{il } + \delta_{jl } m_{ik } - \delta_{il } m_{jk } ) \,,\end{aligned}\ ] ] with @xmath52 in our case , and the @xmath74 and @xmath72 subalgebras commute , @xmath75 = 0 \,.\end{aligned}\ ] ] these factors are linked through the momentum @xmath57 and boosts @xmath60 , which are manifestly each vectors under @xmath72 and may be combined into an object @xmath76 which is also a doublet under @xmath74 : @xmath77 = -{i \sigma^1\over 2 } { \cal p}_i \ , , \quad \quad [ j_{01 } , { \cal p}_i ] = - { \sigma^2\over 2 } { \cal p}_i \ , , \quad \quad [ j_{02 } , { \cal p}_i ] = { i \sigma^3\over 2 } { \cal p}_i \,.\end{aligned}\ ] ] finally the @xmath57 and @xmath60 each commute with themselves , but the mutual commutator is centrally extended by @xmath56 : @xmath78 = 0 \ , , \quad \quad [ k_i , k_j ] = 0 \ , , \quad \quad [ k_i , p_j ] & = & i \delta_{ij } n \,,\end{aligned}\ ] ] where @xmath56 commutes with everything .
thus the poincar algebra may be thought of as @xmath79 ( the @xmath80 factor being @xmath56 ) augmented by the spinor - vector @xmath81 which commutes onto the @xmath80 .
we now turn to the focus of our study , the geometry generated by applying the null melvin twist to a rotational isometry in global anti - de sitter space , which for reasons that will become apparent we will call the five - dimensional gdel - schrdinger geometry , or @xmath82 . before describing this spacetime , we will review global ads , its symmetries and holographic interpretation .
anti - de sitter space can be presented in a number of coordinate systems , some of which span the whole manifold and some of which do not .
the poincar patch metric ( [ pads ] ) , while very useful for ads / cft applications , does not cover the entire manifold as geodesics exit it in finite proper time .
global anti - de sitter space , in contrast , is locally the same geometry , but presented in a coordinate system that covers the entire , geodesically complete spacetime . in @xmath83 dimensions , the global ads metric
may be written @xmath84 where @xmath85 is the metric on the round @xmath86-sphere .
anti - de sitter space in @xmath83 dimensions possesses a @xmath87 isometry group , corresponding to the ( relativistic ) conformal group of the dual @xmath70-dimensional field theory .
different subgroups of @xmath87 are more apparent in different coordinate presentations .
the poincar patch metric ( [ pads ] ) makes manifest the foliation of the space with respect to @xmath88 spanned by @xmath8 and the @xmath89 , and the corresponding @xmath90 lorentz subgroup and the @xmath70 translations are clearly visible .
the boundary of the poincar patch is a copy of @xmath88 , and so the dual field theory lives on this space .
the remaining generators of @xmath87 correspond to the dilatation and special conformal transformations in the dual conformal field theory .
the schrdinger isometries of @xmath53 constitute a nine - dimensional subgroup of @xmath91 preserving ( among others ) the isometries used in the melvin map . in global ads , on the other hand , the @xmath72 subgroup of @xmath87 acts in the usual way on the @xmath86-sphere , while @xmath8-translations form the @xmath58 subgroup .- dimensional space , @xmath8 emerges as a periodic coordinate , and one typically passes to the universal cover to generate an infinite temporal extent
. we will not be careful distinguishing the @xmath58 symmetry of compact time from the @xmath92 of the universal cover . ]
the space is a foliation by @xmath93 ( the universal cover of @xmath94 ) , on which the boundary theory lives ; thus @xmath72 and @xmath3 are the geometrical symmetries of the dual field theory , and the remaining generators of @xmath87 are the conformal enhancement . the ads / cft interpretation is that the poincar and global ads cases are dual to the same cft , only living on different spaces , @xmath88 vs. @xmath93 .
note that it is rather nontrivial that the geometric symmetries of the two cases ( which are quite distinct for a non - conformal theory ) are enhanced to the identical symmetry group @xmath87 by the additional conformal generators .
the natural question then arises : can one find something similar for the schrdinger case , a kind of global schrdinger " spacetime , which describes the same dual field theory but living on a sphere instead of a plane ?
spacetime . ] as we shall see , the geometry resulting from applying the null melvin twist to a rotational generator in global ads seems an appropriate candidate for such a spacetime , though the full group is not exactly the same as in the @xmath53 schrdinger case . for global @xmath95
, we may take the coordinates to be @xmath96 the coordinates for the @xmath97 emphasize the hopf fibration of @xmath98 , parameterized by @xmath99 , over the @xmath100 parameterized by @xmath101 and @xmath102 . to apply the melvin map
, we must pick a pair of isometries .
global ads has all the same isometries as the poincar patch , so in principle one could perform the same melvin twist that led to @xmath53 ( [ pschro ] ) by using a translation " generator .
we are interested in the result if a rotation generator on the @xmath97 is used instead .
there are several motivations for exploring this possibility .
the first is finding a gravitational dual for the nrcft on a sphere , as described above .
another is understanding the implementation of rotation in the background of the nrcft , for which we want to explore the action of the melvin map on rotating black holes . the most general rotating black hole in 5d ads space preserves only three isometries , @xmath3 , @xmath103 and @xmath104 .
we may study the null melvin twist applied to a rotational isometry of ads as the simplest case of the class of spaces that include the melvinization of the rotating black holes . even choosing to pick a rotational direction
, there is a certain amount of freedom still in defining the isometry .
our philosophy will be to pick the isometry so as to make the resulting geometry resemble the schrdinger solution as much as possible .
we will define the rotational isometry , the @xmath9 direction , as proportional to the hopf fiber direction @xmath99 : @xmath105 where @xmath106 is a constant , so that @xmath107 then we have @xmath108 and thus @xmath109\ , , \\
we see that to avoid @xmath27 going negative ( and the string coupling becoming imaginary ) , we need @xmath110 . to further constrain this parameter , we note that the schrdinger solution @xmath53 has a constant dilaton . this may be achieved in the present case by demanding @xmath111 giving @xmath112 to avoid the unphysical result of vanishing or negative @xmath27 . as with @xmath53 ,
the resulting constant dilaton means we need not distinguish the einstein and string frame metrics . note that a more general choice of rotational isometry ( [ choosepsi ] ) involving both @xmath99 and @xmath102 also results in a varying dilaton .
then the melvinized metric is @xmath113 which we can also write as @xmath114 along with the massive vector field and dilaton , @xmath115 we may also present the metric more compactly in terms of the massive vector field as in ( [ metricam ] ) , @xmath116 with @xmath117 the global ads metric given in ( [ globalads ] ) , and where we have defined @xmath118 we note that in this presentation the melvin parameter @xmath25 appears exclusively in the scale @xmath119 ; the limit @xmath32 returning to global ads corresponds to @xmath120 .
unlike the schrdinger case , where @xmath25 can be absorbed into a redefinition of the light - cone coordinates ( [ lightcone ] ) , here @xmath25 defines a radial scale which has physical meaning relative to the unit ads radius .
it is natural to expect that this ratio will be a nontrivial parameter in a spacetime dual to a theory living on a sphere , which possesses an extra length scale corresponding to the sphere s radius . we will call this spacetime the five - dimensional gdel - schrdinger geometry , or @xmath82 , though so far we have not justified the connection either to the schrdinger solution @xmath53 nor to gdel spacetimes . in the remaining parts of this section
we will do the former , while the latter will follow in the next section .
as already described , five - dimensional ads space has an @xmath91 symmetry group , where in the global coordinate system the natural subgroups @xmath121 act in the usual way on the @xmath97 and the global time coordinate @xmath8 .
the remaining 8 generators span a @xmath122 of @xmath121 ; viewing @xmath123 as @xmath124 , they appear as a @xmath125 of @xmath126 .
the null melvin map breaks a number of these symmetries ; however , nine of the fifteen isometries of global ads in fact survive , the same number that is present for the schrdinger space @xmath53 .
@xmath127 is broken , but since the new angular term in ( [ mgads ] ) is proportional to the hopf fiber @xmath128 in the fibration of @xmath97 over @xmath100 , the @xmath129 acting on the @xmath100 is preserved , as is the @xmath58 of the hopf coordinate @xmath99 . thus @xmath130 ;
the @xmath58 of global time translations is preserved as well .
this @xmath131 is generated by @xmath132 : @xmath133 where we have defined the light cone coordinates @xmath134 with @xmath135 .
the remaining four isometries are a subset of the eight off - diagonal " generators in global ads , forming a @xmath136 of @xmath137 , while being invariant under @xmath138 ; as with @xmath53 , @xmath56 commutes with all generators .
they are most compactly expressed in the boyer - lindquist coordinates @xmath139 as @xmath140 where @xmath141 we note that all 9 generators are independent of @xmath25 , and so go over smoothly to ads generators in the @xmath32 limit .
the commutators not determined by the symmetry assignments given above are the mutual commutators of the @xmath27s , given by @xmath142 = -4 l_3 \ , , \quad & \quad [ k_1 , k_3 ] = - 6 h + n - 4 l_2 \ , , \quad&\quad [ k_1 , k_4 ] = 4 l_1 \ , , \\ { } [ k_2 , k_3 ] = 4 l_1 \ , , \quad & \quad [ k_2 , k_4 ] = - 6 h + n + 4 l_2 \ , , \quad&\quad [ k_3 , k_4 ] = - 4 l_3 \ , .
\nonumber\end{aligned}\ ] ] we see that @xmath56 appears as a central generator , as in the schrdinger case , but @xmath55 and the @xmath143 appear on the right - hand - sides as well . in schrdinger holography , the space on which the dual field theory lives is the space obtained keeping both the radial coordinate and the coordinate associated to @xmath56 fixed .
taking @xmath144 constant , @xmath145 constant we have a copy of @xmath146 spanned by @xmath101 , @xmath102 and @xmath147 .
meanwhile , the @xmath82 symmetry algebra based on @xmath131 is quite similar to the schrdinger algebra based on @xmath148 , but with differences appropriate to the dual theory living on the new space .
the schrdinger @xmath149 ( generated by @xmath55 , @xmath62 and @xmath61 ) has been replaced by the spatial @xmath129 of the @xmath100 ; in the schrdinger case the hamiltonian is part of @xmath149 while @xmath150 forms an @xmath58 , while in the gdel - schrdinger case , it is the hamiltonian generator that forms the @xmath58 , describing the time direction of the dual .
analogously to the case comparing poincar and global ads , the generators that are spatial " relative to the boundary geometry are reshuffled amongst those corresponding to conformal enhancement .
it is thus natural to interpret the boundary cft as living on @xmath146 on which @xmath137 naturally acts .
the other difference between the two symmetry algebras is the commutators of other four generators . in both cases
these four form a @xmath136 of @xmath151 or @xmath152 ; however in the @xmath53 case , their mutual commutators include only the central element @xmath56 , while in the @xmath82 case , the other generators are present on the right - hand side as well .
the large isometry group is transitive , guaranteeing that the spacetime is homogeneous ; any point can be reached from any other by the application of symmetry transformations , implying that all points are equivalent .
one can see this by noting the @xmath27 generators may be used to translate in @xmath153 , then at fixed @xmath153 one may move in @xmath101 using @xmath154 and @xmath155 , before reaching any desired @xmath8 , @xmath99 , @xmath102 using @xmath55 , @xmath56 and @xmath156 .
it is possible to take a coordinate limit of the gdel - schrdinger metric to obtain the original schrdinger solution ; the same sort of limit works in obtaining the poincar patch representation of ads from the global solution .
one imagines taking a point at large @xmath153 so the three - spheres are large compared to the ads radius ; recall we have set this radius to unity .
one must then look at small angle fluctuations , so the complete three - sphere is no longer visible .
geodesics fall out of the poincar patch in finite proper time , so we consider small fluctuations of @xmath8 as well . finally , to preserve the schrdinger character and not revert to ordinary anti - de sitter space , we take the melvin length scale @xmath119 to be large such that @xmath157 remains finite , implying @xmath119 is much larger than the ads scale , which also characterizes the radius of the sphere on which the dual nrcft lives .
this can be implemented with the coordinate redefinition @xmath158 and light - cone coordinates in schrdinger may be assembled in the usual way , @xmath159 substituting into the gdel - schrdinger metric ( [ mgads ] ) and taking the limit @xmath160 , and finally dropping all primes in the end , returns precisely the @xmath53 schrdinger metric ( [ pschro ] ) . since the gdel - schrdinger spacetime is homogeneous , there is nothing special about the point we consider , and any point reduces locally to the spacetime @xmath53 when @xmath161 .
one naturally expects that the nine isometries of gdel - schrdinger go over into the original schrdinger isometries , and indeed this is the case .
looking only at leading behavior in the small @xmath162 limit , only the isometries of mass dimension 1 appear at @xmath163 : @xmath164 with the @xmath165 s also giving various combinations of @xmath166 , @xmath167 , @xmath168 and @xmath169 .
the remaining generators are present , but at higher order ; those of mass dimension 0 appear at @xmath170 , and the one of mass dimension @xmath171 at @xmath172 .
we can extract them all by taking appropriate linear combinations : @xmath173 dropping the primes , these are precisely the generators ( [ schroisoms ] ) of schrdinger spacetime from section [ melvinsec ] .
note that the algebra is modified by the limit , since the @xmath53 and @xmath82 symmetry groups are not the same ; for example , the three - sphere rotations @xmath155 and @xmath156 do not commute in @xmath82 , but in the limit taken , they become commuting translations @xmath57 in @xmath53 . thus we have justified that the gdel - schrdinger spacetime has the appropriate symmetry algebra and limit to holographically encode the same nrcft as @xmath53 , but on the space @xmath146 .
we note that the schrdinger geometry @xmath53 is known to be nonsupersymmetric @xcite .
given the limit relating them just described , this suggests that the gdel - schrdinger spacetime may lack supersymmetry as well ; however , this has not been explicitly verified .
there are also known to be generalizations of the schrdinger geometry with unbroken supersymmetry @xcite .
given the similarity between @xmath82 and several supersymmetric geometries described in the following sections , and the local stability associated to supersymmetric solutions , it would be useful to perform this analysis more explicitly .
so far we have argued for the similarities between this gdel - schrdinger geometry and the schrdinger space @xmath53 ; we shall now justify the
gdel " part of the name by demonstrating the existence of closed timelike curves in close analogy to known gdel - type spacetimes , and investigate the holographic screens of the geometry .
we wish to demonstrate that the gdel - schrdinger geometry possesses closed timelike curves ( ctcs ) .
this is easy to do .
the @xmath102-direction is naturally periodic with periodicity @xmath174 .
moving in this direction only on the @xmath82 geometry , we have @xmath175 this is manifestly spacelike at small @xmath153 , and spacelike everywhere for the global ads limit of @xmath120
. however , in the gdel - schrdinger spacetime the direction becomes timelike at @xmath176 the angles giving the smallest value of the right - hand - side are @xmath177 .
( note that the @xmath178 direction is not degenerate within the three - sphere at these points ; instead one of the boyer - lindquist coordinates ( [ blcoords ] ) @xmath179 or @xmath180 becomes degenerate , with the other linear combination giving an @xmath98 , which is the closed timelike curve we discover . )
thus the transition radius to a zone with ctcs is @xmath181 the radial scale parameter we defined previously ( [ definerctc ] ) , justifying the subscript .
this radius obviously moves off to infinity as we take @xmath32 .
of course , the spacetime is homogeneous , so there are closed timelike curves going through every point .
what the argument above demonstrates is that these ctcs have a certain minimum size in the radial coordinate ; speaking loosely , only a region of space of characteristic size @xmath119 or larger will be able to sample a complete closed timelike curve .
the fact that the spacetime is homogeneous but possesses closed timelike curves is highly reminiscent of the original gdel spacetime , as well as its higher - dimensional descendants .
all these geometries are naively pathological due to the closed timelike curves .
however , a number of ideas for rescuing or reinterpreting the various gdel solutions have been put forward , and we can apply these to our geometry as well . in the remainder of this section , we will consider the calculation of a holographic preferred screen for an observer at the origin in gdel - schrdinger spacetime , and show that it occurs in the region free from closed timelike curves , suggesting that the holographic description of the observer s dynamics is independent of the ctcs . in the next section ,
we consider how a condensation of brane giant gravitons may modify the acausal region of the solution to remove the ctcs altogether .
the original gdel solution @xcite is a four - dimensional geometry solving einstein s equations in the presence of pressureless dust and a negative cosmological constant .
the metric may be written @xmath182 with periodic coordinate @xmath102 @xcite .
the geometry is homogeneous , and orbits of the @xmath102 coordinate become closed timelike curves for @xmath183 .
although the gdel spacetime has a global time coordinate @xmath147 , it has no global time function , as the constant-@xmath147 surfaces are not everywhere spacelike .
certain properties of the gdel universe , notably the homogeneity and closed timelike curves , are also present in the spacetime we are considering .
an even closer cousin is the five - dimensional supersymmetric generalization of the gdel spacetime , discovered by gauntlett , gutkowski , hull , pakis and reall @xcite .
this is a solution to five - dimensional minimal supergravity with metric @xmath184 where @xmath185 is the rotation parameter , along with a nontrivial massless gauge field .
following @xcite , we shall refer to this spacetime as @xmath186 .
it may be lifted to a solution of eleven - dimensional supergravity , and is t - dual to a type iib pp - wave spacetime .
this geometry shares a number of features with our gdel - schrdinger spacetime , the most immediately obvious being the cross - term between the time coordinate @xmath8 and the hopf fiber of the @xmath97 .
they are both homogeneous , and also share the existence of closed timelike curves , which are present for @xmath186 starting at a radius @xmath187 furthermore , the symmetry groups are virtually identical .
both possess a nine - dimensional group of isometries consisting of @xmath131 along with four more generators transforming in the @xmath188 .
the only difference is that for @xmath186 these four generators commute except for closing onto the central extension , as was the case for @xmath53 , in contrast with the @xmath189 case , where the @xmath190 generators appear on the right - hand sides as well , as shown in equation ( [ kcomm ] ) .
other differences are that @xmath186 is supported by a massless gauge field and reverts to minkowski space as the rotation parameter is set to zero , while @xmath189 features a massive gauge field and a cosmological constant , and becomes global ads in the non - rotating limit .
thus in a sense , our geometry is an anti - de sitter version of the gdel spacetime .
there is also the question of whether @xmath82 is indeed nonsupersymmetric , and if so , whether there exist other solutions with properties analogous to @xmath82 ( in particular , closed timelike curves and asymptotic ads space ) that possess supersymmetry as well .
the covariant holographic principle of bousso @xcite states that the size of a codimension - two area @xmath191 bounds the entropy ( or degrees of freedom ) passing through a light sheet " generated by the contracting null rays orthogonal to @xmath191 .
in general two of the four possible null congruences orthogonal to @xmath191 ( two future - going , two past - going ) will be shrinking ; in weak gravity this is the
inward "- moving congruence in each the past and the future , but this can change in the presence of strong gravity .
a light - sheet terminates when the expansion ceases being negative ; when the null - energy condition is satisfied , this can only happen at a caustic where the curves pass through each other . for a given observer , one may take the families of null rays emanating from any point on the observer s worldline and follow them until they cease expansion ; the area @xmath191 that is the locus of zero expansion is called a preferred screen " , with the family of null rays forming a light sheet for the screen .
the union of all the observer s preferred screens forms a hypersurface called the preferred screen - hypersurface ; generally this hypersurface is simply referred to as a preferred screen itself , and we will follow this convention . the preferred screen holographically encodes a region consisting of all the light sheets associated to the null congruences , including the observer s worldline ; the region so encoded is called the holographic domain of the screen . for anti - de sitter space and similar spacetimes , the preferred screen for all observers is the boundary , and the holographic domain of the boundary is all of spacetime ; thus there is a universal , observer - independent holographic boundary theory describing the entire geometry .
this is not necessarily the case for a general spacetime .
notably , for de sitter space , each individual observer has a distinct preferred screen coinciding with that observer s cosmological horizon , whose holographic domain does not span the entire space but instead is of finite size , suggesting each observer only has access to a finite number of degrees of freedom .
another example in an ads - type space is the karch - randall geometry of an ads@xmath192 brane inside ads@xmath193 @xcite , where part of the space is in the holographic domain of the ads@xmath192 brane , and the rest is in the domain of the remainder of the ads@xmath193 boundary @xcite .
one expects that for a general geometry , holography will be observer - dependent in some way . as discussed in @xcite , holography in gdel - type universes
is an example of such observer - dependence . for the supersymmetric gdel universe
, congruences of null rays expanding from the observer at @xmath194 do not expand to infinity but instead reach precisely the radius @xmath195 where closed timelike curves appear , called there the velocity - of - light surface , before contracting again .
the surface of zero expansion , that is the preferred screen for the @xmath194 observer , was found to be at @xmath196 by homogeneity , this analysis holds for any point in the space .
thus no single observer s preferred screen contains a complete closed timelike curve within its holographic domain , suggesting that the dynamics experienced by this observer may be consistently described by a theory living on the screen that is ignorant of the existence of the ctcs ; for further discussion , see @xcite .
we will now demonstrate that a completely analogous story obtains for the gdel - schrdinger spacetime .
consider an observer at @xmath194 ; again , homogeneity will imply the same obtains for any other point .
the simplest way to find the screen is to note that given the large symmetry group of the @xmath82 spacetime , in particular the preserved @xmath190 acting on the squashed three - sphere , we may infer that the preferred screen will be an @xmath144 constant hypersurface .
( we will demonstrate a more rigorous method momentarily . ) taking the metric on a slice @xmath197 constant , @xmath198 constant , @xmath199 which is just the metric of the squashed three - sphere , we may calculate the area as a function of @xmath153 , which is @xmath200 which as @xmath120 is just the volume of the ordinary three - sphere .
starting from the observer at @xmath194 , the area increases until one reaches the maximal area , which is the location of the screen : @xmath201 for larger @xmath153 the area decreases until it goes to zero at @xmath119 , heralding the appearance of closed timelike curves as @xmath8 fails to be a global time function ; see figure [ areafig ] . , constant-@xmath153 slice of the gdel - schrdinger spacetime .
it peaks at the screen location @xmath202 before going to zero at @xmath203 . ]
we note that the relationship between the screen location and the ctc radius is numerically identical to that for the susy gdel spacetime ( [ gscreen ] ) , further demonstrating the similarity between these two geometries .
one expects that a similar argument would work for more general spaces with closed timelike curves , and sufficient symmetry . whenever the area of the would - be screen contains the direction that develops the closed timelike curve
, the onset of ctcs will send the area to zero .
given the area is also zero at the origin of the radial coordinate and growing as one moves away , there must be a maximal area somewhere before the ctc radius , which is the location of the screen . to check that this is the correct result for the location of the preferred screen , and to better understand the geometry
, we may explicitly find the null congruences heading away from the @xmath194 observer and calculate where their expansion vanishes .
consider null geodesics passing through the origin ; let a given geodesic be parameterized by @xmath204 with @xmath205 the affine parameter .
three constants of the motion are immediately obtainable from the fact that the coordinates @xmath8 , @xmath99 and @xmath102 are absent from the metric : @xmath206 in terms of which we can solve for @xmath207 , @xmath208 and @xmath209 : @xmath210 the norm of the tangent vector to the geodesic is then @xmath211 this must vanish for a null geodesic ; given the first two terms , this is only possible at @xmath194 if @xmath212 .
thus as one would expect for a geodesic passing through the origin , both angular momenta are zero . for a null geodesic , we may also rescale the affine parameter to impose @xmath213 .
since the total @xmath129 angular momentum is zero , the rotational symmetry generators @xmath154 and @xmath155 ( [ symgens ] ) must also have zero inner product with the geodesic tangent vector . requiring this for arbitrary @xmath153 and @xmath101 forces @xmath214 .
the only remaining component is then @xmath215 , and solving for it gives us @xmath216 it is already apparent a solution will only exist for @xmath217 .
one can solve the resulting equations to find the solution for null geodesics passing through the origin , @xmath218 with @xmath101 and @xmath102 constant , where @xmath219 and @xmath220 are arbitrary initial values and we took @xmath221 at @xmath222 .
since the spacetime is homogeneous , geodesics passing through any point behave the same way .
thus as expected for a gdel universe , null geodesics spiral outward , advancing in both @xmath153 and @xmath99 as time progresses , while staying fixed on the two - sphere parameterized by @xmath101 and @xmath102 , until they reach a maximum radius of @xmath223 , after which they begin contracting again .
they contract in fashion symmetric to the expansion until they reach @xmath194 , where the process repeats .
the radial oscillation as a function of the rotation in @xmath99 ( or equally well the affine parameter ) is displayed in figure [ geodesicsfig ] .
direction , which is directly proportional to the affine parameter .
the radius of the preferred screen is the red horizontal line . ]
we may finally calculate the expansion @xmath101 of the null geodesic congruence , whose vanishing defines the preferred screen : @xmath224 which goes to zero at @xmath225 precisely the location of maximal area ; this indeed agrees with our previous result for the location of the preferred screen .
it is interesting to note how gdel - schrdinger inherits properties both of anti - de sitter space as well as the gdel - type geometries . in ads , massive geodesics through the origin
move out to an energy - dependent maximum radius before re - collapsing , while the lightlike geodesics make it all the way to the boundary ; in a sense ads s primary characteristic is that it traps massive particles but allows massless ones to reach infinity in finite time . in gdel spacetimes , in contrast , the null geodesics are trapped as well , and re - collapse to the origin in finite time , while rotating .
this is another way of describing the observation that anti - de sitter space is holographically represented by a screen at the boundary , while gdel - type universes have a holographic screen ( for the observer at the origin ) at a finite radius . in sum
, we have demonstrated the close kinship between the gdel - schrdinger spacetime and the supersymmetric gdel universe @xmath186 ; the original gdel universe has a similar set of preferred screens @xcite , as will other , similar universes .
the existence of the finite - sized preferred screen implies that any individual observer s dynamics may be calculable without having to include the ctcs , which are too large to fit inside any single holographic domain .
this offers a kind of chronology protection .
it may not be fully satisfactory , however , since it still leaves us with a global geometry in which the ctcs are present .
one may wonder whether string theory directly addresses their existence , or only skirts it in the holographic description . in the next section
, we will argue that a condensation of branes should change the geometry in the region with closed timelike curves , presumably removing them , without affecting the holographic domain of the preferred screen .
hawking @xcite proposed the existence of a chronology protection conjecture , that closed timelike curves are prevented from developing in geometries that begin without them ; see also for example @xcite .
it is another question , however , to ask whether string theory or any other theory of quantum gravity might resolve " closed timelike curves in spacetimes that classically possess them at all times .
it was suggested by dyson @xcite that in certain cases where the classical geometry has ctcs , the dynamics of branes may generate an enhanon - type mechanism that changes the geometry to remove the acausal behavior ; in her example , the closed timelike curves lived within a compact region , with chronologically safe geometry at infinity .
a case closer to ours , where the ctcs appear outside a certain radius , was provided by drukker , fiol and simn ( dfs ) @xcite , involving a condensation of supertubes in a four - dimensional gdel - type geometry .
one is naturally led to ask whether dyson s enhanon could operate in the gdel - schrdinger system as well , and as we shall show , the answer is that it does . beyond the original supersymmetric gdel universe @xmath186 ,
a number of other gdel - type geometries that preserve some supersymmetry and appear as solutions to supergravity equations of motion have been found .
, see @xcite . ] a geometry solving the equations of type iia supergravity was described by harmark and takayanagi @xcite with the form @xmath226 , where the metric and @xmath227-form fields are trivial on the @xmath228 , and @xmath229 has metric @xmath230 and @xmath227-form field strengths @xmath231 again we have a gdel - type spacetime , with closed timelike curves manifesting themselves for @xmath232 .
an analysis of the area of constant-@xmath147 , constant @xmath153-slices implies a preferred screen for the observer at the origin at @xmath233 again less than the critical radius for the appearance of ctcs . as pointed out in @xcite , a natural string theory probe of a geometry with closed timelike curves
is a brane , since the brane s finite extent is capable of wrapping the entire acausal curve .
dfs found that a suitable brane probe of the @xmath229 geometry is a supertube @xcite .
a supertube is an object with d0-brane and fundamental string charge , as well as d2-brane dipole moment ; it may be realized as a d2-brane with both electric @xmath234 and magnetic fields @xmath235 on its world volume , which act as sources for the f - string charge @xmath236 and d0-brane charge @xmath237 respectively . for a tube filling @xmath147 , @xmath9 and @xmath102
we have the worldvolume field strength @xmath238 and the associated charges , @xmath239 with @xmath240 the momentum conjugate to @xmath234 . for given conserved charges , the crossed worldvolume fields generate an angular momentum , @xmath241 which stabilizes the tubular brane against collapse at a finite @xmath242 , @xmath243 the solution to the equations of motion has @xmath244 for any @xmath245 .
thus one can have a supertube at any radius by tuning the conserved charges appropriately .
drukker , fiol and simn considered a supertube embedded in the gdel universe @xmath229 and found that the above analysis , originally performed in flat space , obtains equally well with the @xmath246 term added to the magnetic field .
the supertube is a source for the same fields that are turned on in the background , and preserves some supersymmetry , making it a natural probe to consider .
however , considering radial fluctuations of the supertube around its minimum radius in the @xmath229 geometry , they found kinetic terms @xmath247 with coefficient @xmath248 which is positive - definite for @xmath249 , but passes through zero precisely at the ctc radius and becomes negative at larger @xmath153 .
dfs argued that such a negative kinetic term is a signal that the moduli space metric no longer contains the proper degrees of freedom , analogous to the repulson singularity @xcite . continuing the analogy ,
the repulson should be removed by the condensation of branes with negative kinetic terms , in this case the supertubes , and replaced by a new enhanon " geometry @xcite .
dfs suggested that the spacetime sourced by a collection of supertubes was the proper new geometry outside the ctc radius , and thus that after the condensation of branes the final geometry would be @xmath229 for @xmath249 , and the supertube - backreacted geometry for @xmath250 .
the resulting spacetime is free of closed timelike curves .
it is then natural to ask whether an analogous repulson " instability exists in the gdel - schrdinger spacetime .
we will show that this is indeed the case .
the brane playing the role of the supertube will be a cousin , the dual giant graviton ; appropriately enough for the rotating " gdel geometries , this probe also possesses angular momentum . before studying the giant graviton in @xmath82 , we recall its description in global ads .
the original giant graviton solution in @xmath251 consists of a d3-brane propagating in time and wrapped on an @xmath252 while moving at constant speed on an equator of @xmath11 transverse to the @xmath97 @xcite . shortly thereafter ,
the dual " giant graviton solution was formulated , which wraps the @xmath97 in global @xmath95 instead , while propagating in global time and moving on an equator of @xmath11 @xcite .
this solution may be seen as follows .
the action for a d3-brane , including only terms relevant in this paper , consists of the born - infeld and chern - simons terms , @xmath253 ) } + \mu_3 \int p[c_4 ] \,,\end{aligned}\ ] ] where @xmath254 is the brane charge and tension , and the symbol @xmath255 reminds us to pull back the spacetime fields to the brane worldvolume .
for the global ads solution , we have the metric @xmath256 with @xmath117 given in ( [ globalads ] ) and @xmath257 , @xmath258 , a constant dilaton and @xmath259 where the ellipsis indicates parts polarized entirely on the @xmath11 which are not relevant for us .
we consider a brane moving along the @xmath11 hopf fiber with constant velocity , @xmath260 defining the lagrangian @xmath261 by @xmath262 and using that the pullback of the metric includes the term @xmath263_{\tau\tau } = g_{\tau \tau } + \omega^2 g_{\chi\chi}$ ] , we find @xmath264 note that for the poincar case , @xmath265 is replaced by @xmath266 and the potential vanishes at @xmath267 , leading to a moduli space for any @xmath153 . in the global case , one can find the conserved angular momentum , @xmath268 and the hamiltonian is @xmath269 this energy functional is minimized for given @xmath270 at either @xmath221 ( the trivial solution ) or @xmath271 by choosing the appropriate angular momentum , a giant graviton solution can be set up at any @xmath153 .
we note the close analogy to the supertube , where an extended brane is also stabilized at finite radius by angular momentum ; the relationship @xmath272 is identical in both cases .
moreover , for a giant graviton solution one has @xmath273 , analogous to @xmath244 for the supertube .
one can imagine taking a global ads spacetime with a giant graviton in it , and performing the null melvin twist .
the twist consists of coordinate redefinitions and two t - dualities .
since the melvin map acting on the spacetime leaves @xmath274 unchanged and does not add any other rr fields , one might speculate that the d3-brane giant graviton ( which is a source for @xmath274 ) might exist essentially unchanged in the gdel - schrdinger spacetime .
we will find that this is true with the important caveat that giant gravitons localized in the region with ctcs have negative kinetic terms , and hence manifest a repulson - type sickness . rather than attempting to act the null melvin twist on the d3-brane
, we will instead consider directly d3-brane solutions in the gdel - schrdinger spacetime .
the ten - dimensional metric is now @xmath275 with @xmath276 the 5d gdel - schrdinger metric ( [ mgads ] ) .
the dilaton is @xmath277 and there is now a @xmath235-field @xmath278 we again assume a d3-brane wrapped on the ( now squashed ) @xmath97 , propagating in @xmath8 and moving on @xmath10 with constant velocity @xmath279 .
the pullback of the metric again includes a @xmath280 term , @xmath281_{\tau\tau } = g_{\tau \tau } + \omega^2 g_{\chi\chi } = g_{\tau\tau } + { \omega^2 \over 1 - \beta^2 } \,.\end{aligned}\ ] ] the @xmath235-field has one index polarized in the d3-brane directions and one along the @xmath10 direction , resulting in a pullback that is nontrivial , @xmath282 = { \beta \omega \over 1- \beta^2 } { r^2 \over 2 } d\tau \wedge ( d \psi + \cos \theta d \phi ) \,.\end{aligned}\ ] ] putting this together , we find the simple result , @xmath283 ) } = { 1\over \sqrt{1 - \beta^2 } } \left(r^3 \sqrt{1 + r^2 - \omega^2 } \right ) \,,\end{aligned}\ ] ] which when combined with the dilaton ( [ gsdilaton ] ) leads to the conclusion that the lagrangian for the giant graviton in the gdel - schrdinger geometry is identical to the ads case , @xmath284 thus we can again have a solution at any @xmath285 by choosing a suitable angular momentum .
one can trace this initially surprising simplification to the relationship ( [ metricam ] ) between @xmath286 and @xmath287 .
let us now turn to the radial fluctuations around such a giant graviton solution .
adding a velocity @xmath288 , we now obtain @xmath281_{\tau\tau } = g_{\tau \tau } + \dot{r}^2 g_{rr } + \omega^2 g_{\chi\chi } \,,\end{aligned}\ ] ] we find the lagrangian @xmath289 which expanding in powers of @xmath290 becomes @xmath247 where we have dropped terms higher order in @xmath291 , and @xmath292 this is positive - definite for @xmath293 ( the global ads case ) , and for nonzero @xmath25 it remains positive for small @xmath153 . however , the kinetic term coefficient becomes zero and then negative at @xmath294 , or @xmath295 thus precisely in the region where closed timelike curves are present , the branes develop a repulson - type instability , suggesting the solution given is not the true vacuum in that region .
thus we expect the dual giant gravitons to condense and generate a new spacetime , presumably free of the chronological issues related to the presence of closed timelike curves .
we have used the null melvin twist on a rotational isometry of anti - de sitter space to generate a geometry with properties of both the gdel and schrdinger spacetimes .
this gdel - schrdinger spacetime is homogeneous with a large isometry group , appropriate to describe a non - relativistic conformal field theory on @xmath146 and reducing to the schrdinger geometry in a suitable limit .
this geometry possesses closed timelike curves , but we have argued that it may be useful nonetheless .
the preferred screen associated to an observer at the origin encloses a holographic domain small enough that no complete closed timelike curve is contained within it , suggesting the observer - dependent holographic dual description of this observer s experience is free of acausal pathology .
meanwhile , a repulson - type instability in the dynamics of probe dual giant gravitons manifests itself precisely in the region with closed timelike curves , implying that the true geometry will be modified by brane condensation in this region , while leaving the holographic domain of the observer s screen intact .
this is further evidence that string theory engages in a kind of chronology protection , as described by dyson @xcite , removing closed timelike curves by means of an enhanon - type mechanism .
a number of questions naturally present themselves .
the most immediate is , what is the spacetime after giant graviton brane condensation , the gdel - schrdinger version of the enhanon ? for the case of @xmath229 ,
evidence was given that the condensation of supertubes results in a domain wall solution , with @xmath229 inside the wall , and the geometry sourced by the supertube backreaction outside @xcite .
it is possible that a similar result occurs here , although we note a distinction between the cases : the supertube is a source for all the relevant spacetime @xmath227-form gauge fields due to the electric and magnetic fields turned on , while our giant graviton has no worldvolume fields , and thus does not seem to be a source for a @xmath17 field as found in the @xmath82 background .
a related question involves the action of the null melvin twist , which is made out of coordinate shifts and t - dualities .
one would expect both of these operations to map proper solutions to proper solutions , and indeed , as a supergravity solution the gdel - schrdinger spacetime is consistent .
it contains no strong curvatures or singularities and may exist at arbitrarily weak coupling , and thus does not appear to immediately suggest string theory corrections need to be added . indeed ,
the instability is nonlocal , involving the complete closed timelike curves and the extended branes , and thus the corrections go beyond string perturbation theory to the nonperturbative arena .
understanding better how the pathological properties of the gdel - schrdinger solution arise from the action of t - dualities on anti - de sitter space would shed light on the nature of how string theory relates to and resolves non - local pathologies , and to the nonperturbative nature of string theory in general .
it is also interesting to ask how the brane condensation relates to the homogeneity of the gdel - schrdinger spacetime .
since every point is equivalent , observers through every point perceive their own preferred screen beyond which closed timelike curves manifest .
the condensation of branes leading to a new enhanon geometry appears to break this homogeneity .
does the brane condensation indicate an actual spontaneous symmetry breaking , with a single observer becoming preferred , or does each observer continue to perceive their own slice of @xmath82 , each of which being in some sense gauge equivalent to the others , in a manner analogous to horizon complementarity ?
more understanding of these questions would be very useful in understanding how string theory implements holography and removes pathological features ; for related thoughts , see @xcite .
the question of the relationship between supersymmetry and the gdel - schrdinger space - time is also of interest .
the schrdinger geometry @xmath53 is nonsupersymmetric , but several of the gdel geometries we have described possess supersymmetry , as do generalizations of the schrdinger spacetime @xcite .
the ability to reach @xmath53 as a coordinate limit of gdel - schrdinger suggests the latter is nonsupersymmetric , but this should be verified explicitly . if it is indeed nonsupersymmetric , one is naturally led to wonder whether the gdel - schrdinger geometry , like the original gdel spacetime , admits generalizations ( possessing closed timelike curves and anti - de sitter asymptotics , for example ) that are supersymmetric .
certain gdel universes are known to be t - dual to pp - wave geometries @xcite .
a natural additional question is then whether the gdel - schrdinger geometry has a dual description as a pp - wave or any other spacetime .
a class of pp - wave spacetimes can be obtained as limits of anti - de sitter space and its cousins ; since the null melvin twist generates gdel - schrdinger from ads using coordinate limits and t - dualities , such a relationship would not be surprising . finally , it is interesting to speculate on the application of the closed timelike curves to schrdinger holography .
the null melvin twist acting on rotating black holes in ads will generate gdel - like spacetimes dual to spinning nrcfts that possess ctcs , and string theory presumably resolves them in a fashion analogous to that described here .
one is led to wonder what , if any , meaning the closed timelike curves and their nonperturbative removal have in the non - gravitational dual .
these issues are of considerable interest for further exploration .
we would like to thank a. adams , s. de alwis , s. kachru , i. klebanov , and c. rosen for helpful discussions .
the work of c.m.b . and o.d .
was supported by the department of energy under grant no .
de - fg02 - 91-er-40672 .
j. mcgreevy , `` holographic duality with a view toward many - body physics , '' adv .
high energy phys . * 2010 * , 723105 ( 2010 ) [ arxiv:0909.0518 [ hep - th ] ] .
m. ammon , `` gauge / gravity duality applied to condensed matter systems , '' fortsch .
phys . * 58 * , 1123 - 1250 ( 2010 ) .
j. p. gauntlett , j. b. gutowski , c. m. hull , s. pakis , h. s. reall , `` all supersymmetric solutions of minimal supergravity in five- dimensions , '' class .
quant . grav .
* 20 * , 4587 - 4634 ( 2003 ) .
[ hep - th/0209114 ] .
n. drukker , b. fiol and j. simon , `` godel s universe in a supertube shroud , '' phys .
* 91 * , 231601 ( 2003 ) [ arxiv : hep - th/0306057 ] . c. a. r. herdeiro , `` special properties of five - dimensional bps rotating black holes , '' nucl .
phys . * b582 * , 363 - 392 ( 2000 ) .
[ hep - th/0003063 ] . c. v. johnson and h. g. svendsen , `` an exact string theory model of closed time - like curves and cosmological singularities , '' phys .
d * 70 * , 126011 ( 2004 ) [ arxiv : hep - th/0405141 ] .
w. h. huang , `` instability of tachyon supertube in type iia godel spacetime , '' phys .
b * 615 * , 266 ( 2005 ) [ arxiv : hep - th/0412096 ] .
m. s. costa , c. a. r. herdeiro , j. penedones and n. sousa , `` hagedorn transition and chronology protection in string theory , '' nucl .
b * 728 * , 148 ( 2005 ) [ arxiv : hep - th/0504102 ] .
l. dyson , `` studies of the over - rotating bmpv solution , '' jhep * 0701 * , 008 ( 2007 ) .
[ hep - th/0608137 ] .
j. raeymaekers , d. van den bleeken and b. vercnocke , `` relating chronology protection and unitarity through holography , '' jhep * 1004 * , 021 ( 2010 ) [ arxiv:0911.3893 [ hep - th ] ] . j. raeymaekers , d. van den bleeken , b. vercnocke , `` chronology protection and the stringy exclusion principle , '' jhep * 1104 * , 037 ( 2011 ) .
[ arxiv:1011.5693 [ hep - th ] ] .
u. theis , `` free lunch from t - duality , '' arxiv:1002.4758 [ hep - th ] . b. s. kim and d. yamada , `` properties of schroedinger black holes from ads space , '' jhep * 1107 * , 120 ( 2011 ) [ arxiv:1008.3286 [ hep - th ] ] .
n. banerjee , s. dutta , d. p. jatkar , `` geometry and phase structure of non - relativistic branes , '' class .
* 28 * , 165002 ( 2011 ) .
[ arxiv:1102.0298 [ hep - th ] ] .
p. kraus and e. perlmutter , `` universality and exactness of schrodinger geometries in string and m - theory , '' jhep * 1105 * , 045 ( 2011 ) [ arxiv:1102.1727 [ hep - th ] ] .
m. blau , j. hartong , b. rollier , `` geometry of schrodinger space - times , global coordinates , and harmonic trapping , '' jhep * 0907 * , 027 ( 2009 ) .
[ arxiv:0904.3304 [ hep - th ] ] .
s. a. hartnoll and k. yoshida , `` families of iib duals for nonrelativistic cfts , '' jhep * 0812 * , 071 ( 2008 ) [ arxiv:0810.0298 [ hep - th ] ] . n. bobev , a. kundu and k. pilch , `` supersymmetric iib solutions with schrodinger symmetry , '' jhep * 0907 * , 107 ( 2009 ) [ arxiv:0905.0673 [ hep - th ] ] . a. donos and j. p. gauntlett , `` solutions of type iib and d=11 supergravity with schrodinger(z ) symmetry , '' jhep * 0907 * , 042 ( 2009 ) [ arxiv:0905.1098 [ hep - th ] ] . h. ooguri and c. -s .
park , `` supersymmetric non - relativistic geometries in m - theory , '' nucl .
b * 824 * , 136 ( 2010 ) [ arxiv:0905.1954 [ hep - th ] ] .
s. w. hawking , g. f. r. ellis , `` the large scale structure of space - time , '' cambridge university press , cambridge , 1973 . | we show that the null melvin map applied to a rotational isometry of global anti - de sitter space produces a spacetime with properties analogous to both the gdel and schrdinger geometries .
the isometry group of this gdel - schrdinger spacetime is appropriate to provide a holographic dual for the same non - relativistic conformal theory as the ordinary schrdinger geometry , but defined on a sphere .
this spacetime also possesses closed timelike curves outside a certain critical radius .
we show that a holographic preferred screen for an observer at the origin sits at an interior radius , suggesting that a holographic dual description would only require the chronologically consistent region . additionally , giant graviton probe branes experience repulson - type instabilities in the acausal region , suggesting a condensation of branes will modify the pathological part of the geometry to remove the closed timelike curves .
colo - hep-566 0.5 in charles max brown@xmath0 and oliver dewolfe@xmath1 0.4 in _
@xmath2 department of mathematics and physics + kentucky state university + carver hall 131 , 400 e. main st .
+ frankfort , ky 40601 , usa _ 0.2 in _
@xmath1 department of physics + university of colorado + 390 ucb + boulder , co 80309 , usa _ 0.5 in |
during the past decades several studies have revealed a connection between the mass of black holes hosted in galactic nuclei and the properties of their host galaxies ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
moreover , the redshift evolution of the cosmic star formation rate ( sfr ) and smbh accretion rate density are very similar @xcite . this co - evolution " has led various authors to investigate a possible connection between the presence of active galactic nuclei ( agn , which traces the smbh growth ) and the star formation properties of galaxies .
a correlation was indeed observed between star formation and nuclear activity in numerous works at high agn luminosities ( e.g * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , while at low agn luminosities this link is more debated , with @xcite reporting no significant difference in the sfr between active and inactive galaxies , while @xcite found a slight enhancement for agn - hosting galaxies .
they also reported that the enhancement of star formation activity in agn with respect to the bulk of inactive galaxies disappeared if quiescent galaxies were discarded , i.e. agn are more likely hosted in star forming galaxies .
the gas content is often regarded as a more fundamental property of galaxies , with respect to the sfr .
the sfr is tightly related to the gas content through the schmidt - kennicutt relation ( * ? ? ?
* ; * ? ? ?
* sk relation hereafter ) .
currently , one of the most favored scenarios is that the cosmic evolution of the star formation rate in galaxies is mostly a consequence , through the sk relation , of the more fundamental evolution of their ( molecular ) gas content ( e.g. * ? ? ?
* ; * ? ? ?
* ) . within the context of agns , gas is the fundamental ingredient both for nuclear activity and star formation .
possible differences in terms of star formation properties between agns hosts and inactive galaxies could be due to more fundamental differences in terms of gas content , as recently argued by @xcite and @xcite .
therefore , it is most important to obtain information on the gas content of agn host galaxies , possibly as a function of galaxy properties ( e.g. stellar mass ) and redshift .
the molecular gas content can be inferred from the luminosity of the co millimeter transitions , by assuming a proper co - to - h@xmath2 conversion factor . however , co observations are very time consuming , and surveys of large samples are extremely difficult and time demanding .
alternatively , the total ( molecular and atomic ) gas mass can be derived from the dust content , inferred from the fir - submm sed , by assuming a dust - to - gas ratio ( dgr ; e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the uncertainties on the dust - to - gas ratio and its dependence on metallicity are similar to those affecting the co - to - h@xmath2 conversion factor , making the two methods comparable in terms of accuracy , at least at metallicities 12+log(o / h)@xmath38.0 @xcite . in this work
we exploit the dust method for measuring the gas masses in agn host galaxies . in particular
, gas masses are obtained from the dust mass derived from the fir sed of several hundred agn host galaxies at z@xmath01 , along with a control sample of normal galaxies selected in the same stellar mass and redshift ranges .
the aim of this work is to investigate differences in terms of gas content between agn hosts and the bulk of the galaxy population ( i.e. star forming and quiescent galaxies ) , in bins of stellar mass and redshift , to avoid potential biases caused by the dependency of the gas content on these two quantities .
we make use of a stacking procedure to increase the luminosity completeness of the studied samples ( [ stack ] ) .
we selected a sample of agn at @xmath4 ( as well as normal galaxies for the control sample ) in the cosmos , goods - s and goods - n fields .
the choice of these fields was driven by the wide multiwavelength coverage provided by a number of surveys , which is crucial to derive reliable properties ( redshift , stellar mass , star - formation rate and dust mass ) .
we used the far - infrared ( fir ) data from the pacs evolutionary probe ( pep , @xcite ) and the herschel multi - tiered extra - galactic survey ( hermes , @xcite ) programs , which cover the three fields used in this work .
the former was performed with the pacs camera ( 70 , 100 and @xmath5 ; @xcite ) while the latter with the spire camera ( 250 , 350 and @xmath6 ; @xcite ) , both on board of the herschel space observatory @xcite .
herschel catalogues are based on prior information on mips @xmath7 positions and fluxes .
the pep catalogue was described by @xcite and @xcite , while the hermes catalogue was presented by @xcite . since only the goods - s field
was observed at 70 @xmath8 , following ( * ? ? ?
* s13 hereafter ) , who reported that the exclusion of that band does not significantly affect the results , we will not use the 70 @xmath8 data , to use a consistent procedure among all the fields . for the goods - s field we used the optical / near - ir photometric data from the goods - music catalogue @xcite and the x - ray counterpart information from the 4 ms _ chandra _ deep field south ( cdf - s ) main catalogue @xcite .
in the goods - n field we collected the multiwavelength data from the pep team catalogue @xcite and the x - ray data from the 2 ms _ chandra _ deep field north catalogue @xcite .
finally , in the cosmos field we used the @xcite and @xcite multiwavelength catalogues and the _ chandra _ @xcite and xmm @xcite cosmos optical identification catalogues , complemented by the photometric redshifts presented by @xcite .
all the catalogues are supplemented with spectroscopic or photometric redshifts .
photometric redshift for inactive galaxies in cosmos and goods - n lacking redshift information in the above - mentioned catalogues were computed by using the eazy code @xcite .
we applied two selection criteria to the parent sample in the three fields : 1 ) signal - to - noise ratio ( snr ) @xmath9 in the k - band ; 2 ) @xmath10 .
the k - band selection ensures that we can derive reliable estimates for the stellar mass ( following @xcite ) , while the redshift cut ensures that the herschel bands are not probing rest - frame wavelengths that may be significantly affected by agn heating and reprocessing ( which may affect our estimation of the dust masses ) .
indeed , @xcite have shown that out to z@xmath111 the pacs colors are consistent with those typical of star forming galaxies , while some agn contamination to the 100@xmath12 m may occur at higher redshift .
once these selections are applied , we divided the resulting objects into agn and a galaxy samples . in cosmos , we considered an x - ray detected object to be an agn following its optical classification @xcite or best - fitting sed template @xcite or if its absorption - corrected luminosity in the rest - frame @xmath13 band , where available , is @xmath14 .
this threshold is a compromise between the sample size and the contamination by purely powerful star - forming galaxies @xcite .
the luminosities were collected from @xcite and @xcite , who performed a spectral analysis on a sample of bright sources in the _
chandra_-cosmos @xcite and xmm - cosmos @xcite catalogues , respectively . intrinsic luminosities of all the xmm - cosmos selected sources were also derived by the xmm - cosmos team using the observed fluxes and hardness ratio hr ( see * ? ? ?
in goods - s we assumed the @xcite classification and luminosity for the x - ray sources . in goods -
n we used the x - ray detected agn sample of @xcite .
the luminosities of the x - ray selected agn for which this information could be retrieved are in the range @xmath15 , and only @xmath16 per cent of them have @xmath17 . given the limiting fluxes of the x - ray samples , the most obscured agn can still remain undetected . in order to at least partially recover them , we applied the irac power - law selection by @xcite .
the resulting agn sample consists in 801 objects , of which 631 x - ray selected ( in any x - ray band considered in the above - mentioned catalogues ; 486 in cosmos and 145 in goods - s+n ) and 255 irac selected ( 241 in cosmos and 14 in goods - s+n ) .
eighty - five agn ( 78 in cosmos and 7 in goods - s+n ) are in common between the two selections .
all the sources which do not satisfy any of the agn selection criterion in each field are automatically classified as inactive galaxies ( @xmath18 objects ) .
in this section we describe how the physical parameters for our sample of agn and galaxies were derived . in order to avoid selection effects due to redshift , stellar mass and luminosity affecting the dust mass ( and hence the gas mass ) measurement and distribution
, we adopted a stacking technique to derive the average fir fluxes in bins of stellar mass and redshift .
this method ensures a high level of completeness , since only a small fraction of sources are individually detected by herschel , especially in the longest wavelength bands .
stellar masses for agn - hosting galaxies in the sample were derived using the same method as in @xcite , where several consistency tests were also performed .
the observed optical photometry was fitted with a combined library of stellar synthetic templates @xcite and pure agn emission templates affected by levels of absorption @xcite associated with different column densities @xmath19 . in order to break the degeneracies between the different fit parameters ,
the agn were classified into type i ( unabsorbed ) and type ii ( absorbed ) sources on the basis of optical spectroscopic classification ( where available ) , best - fitting spectral energy distribution ( sed ) template from @xcite and intrinsic absorption derived from x - ray data analysis @xcite .
sources for which all these priors are lacking , as well as x - ray undetected agn ( i.e. spitzer selected agn ) were considered as compton - thin type ii agn . moreover , by averaging the two templates of moderate absorption , the four @xcite templates were reduced to three : unabsorbed ( @xmath20 ) , compton - thin absorbed ( @xmath21 ) and compton - thick absorbed ( @xmath22 ) . only the most suitable template , based on the above classification ,
was fitted to each agn . in particular , the unabsorbed , absorbed compton thin and absorbed compton thick templates were used for 100 , 694 and 7 sources , respectively .
the fit was performed through a @xmath23 minimization up to @xmath24 rest - frame , since the @xcite templates do not include emission from dust reprocessing , assuming an exponentially declining star - formation history and a salpeter imf for the stellar component .
each flux was weighted by the inverse of the photometric error and the redshift of each object was fixed during the fit .
the stellar mass was derived by the best - fitting stellar component alone .
agn and galaxies are divided into 6 stellar mass bins ( from @xmath25 to 12 ) and 4 redshift bins .
the number of agn and galaxies in each bin is reported in table [ tab1 ] .
hereafter , the stellar mass corresponding to each bin is assumed to be the mean stellar mass of the objects in the bin , with uncertainty equal to one standard deviation ( typically @xmath26 dex for both agn and galaxies ) .
we adopt the stacking procedure implemented by @xcite , where details on the stacking procedure are given ( see also @xcite , @xcite and @xcite ) . however , in contrast with santini et al . , we do not bin in star formation rate ( sfr ) , since our goal is to investigate if differences in terms of gas content between agns and normal galaxies may actually be at the origin of their claimed sfr differences , hence we do not want to include any _ a priori _ selection or binning on the sfr .
details on the stacking procedure are given in @xcite .
here we only briefly summarize the basic steps .
for each herschel band we excluded areas on the map where the integration time is lower than half the maximum , to avoid regions of high noise level .
then , for each @xmath27 bin containing at least 10 objects , we stack on the residual image ( in which all the 3@xmath1 detected objects were subtracted ) of each herschel band at the position of the undetected sources in the bin , weighting with the inverse of the square of the error map .
fluxes on the stacked images were measured through a psf fitting ( for the pacs bands ) or from the value of the central pixel ( on the spire images ) .
errors on the stacked fluxes were computed through a bootstrap procedure .
finally , the average flux s in each herschel band is computed as : @xmath28 where @xmath29 is the stacked flux of the @xmath30 undetected objects in the bin and @xmath31 is the flux of each of the @xmath32 detected objects at the @xmath33 confidence level .
[ cols= " > , > , > , > , > , > , > " , ] since gas accretion onto super massive black holes ( smbh ) is the process at the origin of nuclear activity , whatever the mechanisms that drives the gas into the central regions are , it is not surprising that agns are preferentially hosted by gas rich galaxies ( see also * ? ? ? * ) .
indeed , the conditions for gas accretion are statistically easier to be fulfilled in presence of a larger gas content . beyond these simple statistical arguments
, models have been proposed that ascribe agn secular fuelling to disk instabilities @xcite , which are stronger in gas rich disks .
our results support this scenario .
as discussed above , the gas content is also the fundamental ingredient driving star formation in galaxies , through the sk relation .
several works have found a relation between strong nuclear activity and enhanced sfr with respect to inactive galaxies ( e.g * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
@xcite and @xcite found that the herschel detection fraction for agn is higher than for galaxies and concluded that agn are more likely to be hosted by star forming galaxies . given the dependency of the sfr on the gas content and the result obtained by us , the enhanced star formation in agn galaxies appears to be primarily the result of a larger ( on average ) gas content , with respect to the bulk of the galaxy population ( star forming and quiescent ) at similar stellar masses , as already suggested by @xcite and @xcite . however , the differences in terms of sfr are less clear with respect to the differences found by us in terms of gas content , probably because of the additional spread introduced by the sk relation , and the contribution of triggering mechanisms ( e.g. galaxy interactions ) , which may affect the star formation efficiency per unit gas mass .
it is beyond the scope of this paper to further discuss the level and nature of enhanced sfr in agn hosts .
the main important result of our work is that agn host galaxies are characterized by much larger amounts of gas , strongly suggesting that generally agn activity in galaxies is simply fostered by a larger content of gas , without invoking specific triggering mechanisms .
bins.,width=302 ] bins.,width=302 ]
making use of the wide multiwavelength data available in the cosmos , goods - s and goods - n fields , we selected a sample of agn and galaxies at @xmath34 .
a stacking procedure on the herschel maps was implemented to derive the average herschel fluxes in bins of stellar mass and redshift .
the stacked fir fluxes were then used to derive the average dust mass and , under reasonable assumptions , the average gas mass of agn hosts and normal galaxies .
finally we compared the average gas mass of agn hosts and inactive galaxies in the same @xmath35 bins .
we find that , at a given stellar mass and redshift , agns are hosted in galaxies much more gas - rich than inactive ones .
the difference is strongest at low stellar masses , @xmath36 , where the gas mass in agn hosts is on average ten times higher than in normal galaxies ( a result significant at @xmath37 ) .
significantly higher gas masses , relative to the normal galaxy population , are however also observed in agn hosts with stellar masses higher than @xmath38 .
taken altogether , in nearly all stellar mass and redshift bins agn host galaxies have higher gas content than normal galaxies ; in almost half of the sample the gas fraction of agn host galaxies is more than three times higher than in normal galaxies .
our result strongly suggests that the likelihood of having an agn in a galaxy is primarily given by the amount of gas in the host galaxy , while dynamical triggering processes ( bars , galaxy mergers and interactions ) likely play a secondary role , at least in the luminosity range probed by us .
this result can be interpreted in simple statistics terms that it is more likely that a gas cloud falls into the potential of the supermassive black hole if there are overall more gas clouds in the host galaxy .
more elaborated models , in which secular fuelling of agns is caused by disk instabilities , which are stronger in more gas rich disks , are also supported by our results .
we acknowledge support from the italian space agency under the asi - inaf contract i/009/10/0 and from inaf under the con- tract prin - inaf-2012 .
fv thanks b. luo for kindly providing chandra i d s of music counterparts in goods - s and i. delvecchio for useful discussion on sed fitting procedure .
this paper uses data from _ herschel _ s photometers pacs and spire .
pacs has been developed by a consortium of institutes led by mpe ( germany ) and including uvie ( austria ) ; ku leuven , csl , imec ( belgium ) ; cea , lam ( france ) ; mpia ( germany ) ; inaf - ifsi / oaa / oap / oat , lens , sissa ( italy ) ; iac ( spain ) .
this development has been supported by the funding agencies bmvit ( austria ) , esa - prodex ( belgium ) , cea / cnes ( france ) , dlr ( germany ) , asi / inaf ( italy ) , and cicyt / mcyt ( spain ) .
spire has been developed by a consortium of institutes led by cardiff university ( uk ) and including univ .
lethbridge ( canada ) ; naoc ( china ) ; cea , lam ( france ) ; ifsi , univ .
padua ( italy ) ; iac ( spain ) ; stockholm observatory ( sweden ) ; imperial college london , ral , ucl - mssl , ukatc , univ .
sussex ( uk ) ; and caltech , jpl , nhsc , univ .
colorado ( usa ) .
this development has been supported by national funding agencies : csa ( canada ) ; naoc ( china ) ; cea , cnes , cnrs ( france ) ; asi ( italy ) ; mcinn ( spain ) ; snsb ( sweden ) ; stfc , uksa ( uk ) ; and nasa ( usa ) . | we have investigated the gas content of a sample of several hundred agn host galaxies at z@xmath01 and compared it with a sample of inactive galaxies , matched in bins of stellar mass and redshift .
gas masses have been inferred from the dust masses , obtained by stacked herschel far - ir and sub - mm data in the goods and cosmos fields , under reasonable assumptions and metallicity scaling relations for the dust - to - gas ratio .
we find that agns are on average hosted in galaxies much more gas rich than inactive galaxies . in the vast majority of stellar mass bins , the average gas content of agn hosts is higher than in inactive galaxies .
the difference is up to a factor of ten higher in low stellar mass galaxies , with a significance of 6.5@xmath1 . in almost half of the agn sample
the gas content is three times higher than in the control sample of inactive galaxies .
our result strongly suggests that the probability of having an agn activated is simply driven by the amount of gas in the host galaxy ; this can be explained in simple terms of statistical probability of having a gas cloud falling into the gravitational potential of the black hole .
the increased probability of an agn being hosted by a star - forming galaxy , identified by previous works , may be a consequence of the relationship between gas content and agn activity , found in this paper , combined with the schmidt - kennicutt law for star formation .
[ firstpage ] methods : data analysis galaxies : active galaxies : ism |
the database was meticulously constructed by visually inspecting images of purported pne ( sourced from major catalogues , simbad and other literature sources ) in the supercosmos halpha survey ( shs , parker et al .
2005 ) , the int photometric halpha survey ( iphas , drew et al .
2005 ) and the supercosmos sky surveys ( sss , hambly et al .
2001 ) . during this process
both pne and obvious non - pne ( e.g. galaxies , photographic flaws , hii regions , duplicates , etc . ) were added to the database .
the non - pne were flagged as different object types to keep them separate and to keep track of objects already considered for inclusion in the database .
this laborious process allowed us to make the most accurate estimate to date of the total number of galactic pne , @xmath1 , instead of the usual approach of adding total numbers found in new catalogues .
we find @xmath2 , though this upper limit may be reduced as more non - pne are identified after we further check and refine our database before its first release .
our value is based on currently published catalogues and will also be revised upwards once future pn discoveries are published .
figure [ fig : fig1 ] depicts the current galactic distribution of pne .
the database will be accessible through a web - based interface that allows users to search for pne by object name ( iau pn g and common name ) and j2000 coordinates ( single or list format ) .
static releases will also be made intermittently to vizier for virtual observatory compatibility .
results from the web - interface may be browsed in a gallery format with thumbnail images of each pn , as well as a variety of other formats .
groups of pne may be selected based on their coordinates , object size , name , catalogue and so on .
links to online resources are also made available including e.g. the eso archive , vizier and staralt .
the first release of the database is planned for early 2012 and is intended to produce a working database with the cleanest and largest set of entries for published pne .
this will allow for queries to be made , selecting large samples of pne to be studied in a unified fashion with accurate coordinates and pn g designations .
the initial dataset will include at least image cut - outs from the shs and sss ( in jpeg and fits format ) , secgpn spectra in fits format ( acker et al .
1992 ) , radial velocities ( e.g. durand et al . 1998 ) and central star magnitudes ( e.g. tylenda et al .
additions to the database will mostly include large observational datasets in refereed publications that are of broad interest to the whole sample .
acker , a. , marcout , j. , ochsenbein , f. , stenholm , b. , & tylenda , r. 1992 , garching : european southern observatory , 1992 drew , j. e. , greimel , r. , irwin , m. j. , et al .
2005 , mnras , 362 , 753 durand , s. , acker , a. , & zijlstra , a. 1998 , a&as , 132 , 13 hambly , n. c. , macgillivray , h. t. , read , m. a. , et al .
2001 , mnras , 326 , 1279 jacoby , g. h. , kronberger , m. , patchick , d. , et al .
2010 , pasa , 27 , 156 miszalski , b. , et al .
2008 , mnras , 384 , 525 parker , q. a. , phillipps , s. , pierce , m. j. , et al .
2005 , mnras , 362 , 689 parker , q. a. , et al .
2006 , mnras , 373 , 79 tylenda , r. , acker , a. , raytchev , b. , stenholm , b. , & gleizes , f. 1991 , a&as , 89 , 77 viironen , k. , et al .
2009 , a&a , 504 , 291 | since the unifying strasbourg - eso catalogue of galactic planetary nebulae + ( secgpn ) a large number of new discoveries have been made thanks to improved surveys and discovery techniques .
the increasingly heterogeneous published population of galactic pne , that we have determined totals @xmath0 2850 pne , is becoming more difficult to study on the whole without a centralised repository .
we introduce a consolidated and interactive online database with object classifications that reflect the latest multi - wavelength data and the most recent results .
the extensible database , hosted by the centre de donnees astronomique de strasbourg ( cds ) , will contain a wealth of observed data for large , well - defined samples of pne including coordinates , multi - wavelength images , spectroscopy , line intensities , radial velocities and central star information .
it is anticipated that the database will be publicly released early 2012 . |
the achievable rate for a time - selective fading channel depends on what channel state information ( csi ) is available at the receiver and transmitter .
namely , csi at the receiver can increase the rate by allowing coherent detection , and csi at the transmitter allows adaptive rate and power control ( e.g. , see ( * ? ? ?
6 ) ) . obtaining csi at the receiver and/or transmitter requires overhead in the form of a pilot signal and feedback .
we consider a correlated time - selective flat rayleigh fading channel , which is unknown at both the receiver and transmitter .
the transmitter divides its power between a pilot , used to estimate the channel at the receiver , and the data .
given an average transmitted power constraint , our problem is to optimize the instantaneous pilot and data powers as functions of the time - varying channel realization .
our performance objective is a lower bound on the achievable rate , which accounts for the channel estimation error . power control with channel state feedback , assuming the channel is perfectly known at the receiver , has been considered in @xcite .
there the focus is on optimizing the input distribution for different channel models using criteria such as rate maximization and outage minimization .
optimal power allocation in the presence of channel estimation error has been considered in @xcite .
the problem of optimal pilot design for a variety of fading channel models has been considered in @xcite . there
the pilot power and placement , once optimized , is fixed and is not adapted with the channel conditions . a key difference here
is that the transmitter uses the csi to adapt _ jointly _ the instantaneous data and pilot powers .
because the channel is correlated in time , adapting the pilot power with the estimated channel state can increase the achievable rate .
we also remark that although we analyze a single narrowband fading channel , our results apply to a set of parallel fading gaussian channels , where the average power is split over all channels .
we start with a correlated block fading model in which the sequence of channel gains is gauss - markov with known statistics at the receiver . the channel estimate is updated at the beginning of each block using a kalman filter , and determines the power for the data , and the power for the pilot symbols in the succeeding coherence block .
optimal power control policies are specified implicitly through a bellman equation @xcite .
other dynamic programming formulations of power control problems have been presented in @xcite , although in that work the channel is either known perfectly ( perhaps with a delay ) , or is unknown and not estimated . because an analytical solution to the bellman equation appears to be difficult to obtain , we study a diffusion limit in which the correlation between successive coherence blocks tends to one and the average power goes to zero .
( this corresponds to a wideband channel model in which the available power is divided uniformly over a large number of parallel flat rayleigh fading sub - channels . ) in this limit , the gauss - markov channel becomes a continuous - time ornstein - uhlenbeck process @xcite , and the bellman equation becomes a partial differential equation ( pde ) .
a diffusion equation is also derived , which describes the evolution of the state ( channel estimate and the associated error variance ) , given a power allocation policy . in this limit
, we show that given a peak power constraint for the pilot power , the optimal pilot power control policy is a switching policy ( `` bang - bang '' control ) : the pilot power is either the maximum allowable or zero , depending upon the current state .
hence the optimal pilot power control policy requires at most one feedback bit per coherence block .
also , the optimal data power control policy is found to be a variation of waterfilling @xcite
. other work in which the wireless channel is modeled as a diffusion process is presented in @xcite .
the switching points for the optimal policy form a contour in the state space , which is referred to as the _ free boundary _ for the corresponding pde . solving this pde then falls in the class of _ free boundary problems _ @xcite .
we show that in the diffusion limit the system state becomes confined to a narrow region along the boundary .
furthermore , the associated probability distribution over the boundary is exponential .
that enables a numerical characterization of the boundary shape .
our results show that the average pilot power should decrease as the channel becomes more severely faded .
we observe that the optimal switching policy is equivalent to adapting the pilot symbol insertion rate with _ fixed _
pilot symbol energy .
the optimal pilot insertion rate as a function of the channel estimate is then determined by the shape of the free boundary .
we show that the boundary shape essentially shifts pilot power from more probable ( faded ) states to less probable ( good ) states .
furthermore , the boundary shape guarantees that the channel estimate is sufficiently accurate to guide the power adaptation .
numerical results show that pilot power adaptation can provide substantial gains in achievable rates ( up to a factor of two ) .
the gains are more pronounced at low snrs and with fast fading channels .
although these results are derived in the limit of large bandwidth ( low snr ) , monte carlo simulations show that they provide an accurate estimate of the performance when the bandwidth is large but finite ( a few hundred coherence bands ) .
moreover , the optimal switching policy in the diffusion limit accurately approximates the optimal pilot power control policy for the discrete - time model , and provides essentially the same performance gains relative to constant pilot power . to limit the overall feedback rate
, we also consider combining the adaptive pilot power with `` on - off '' data power control , which also switches between a fixed positive value and zero .
( hence that also requires at most one bit feedback per coherence block . )
the corresponding optimal free boundaries are computed , and results show that this scheme gives negligible loss in the achievable rate .
the next section presents the system model and section [ discrete ] formulates the pilot optimization problem as a dynamic program .
section [ cont ] presents the associated diffusion limit and the corresponding bellman equation .
the optimal policy is then characterized in sections [ solanal]-[water ] with optimal data power control , and in section [ on - off ] with optimal on - off data power control .
numerical results showing free boundaries and the corresponding performance are also presented in sections [ water ] and [ on - off ] .
training overhead is discussed in section [ tl ] , and conclusions and remaining issues are discussed in section [ conc ] .
we start with a block fading channel model in which each coherence block contains @xmath0 symbols , consisting of @xmath1 pilot symbols and @xmath2 data symbols .
the vector of channel outputs for coherence block @xmath3 is given by @xmath4 where @xmath5 and @xmath6 are , respectively , vectors containing the pilot and data symbols , each with unit variance , and @xmath7 and @xmath8 are the associated pilot and data powers .
the noise @xmath9 contains circularly symmetric complex gaussian ( cscg ) random variables , and is white with covariance @xmath10 .
the channel gain @xmath11 is also cscg , is constant within the block , and evolves from block to block according to a gauss - markov process , i.e. , @xmath12 where @xmath13 is an independent cscg random variable with mean zero and variance @xmath14 , and @xmath15 $ ] determines the correlation between successive blocks .
we will assume that @xmath16 and @xmath17 are known at the receiver .
the training energy per symbol in block @xmath3 is defined as @xmath18 , where @xmath19 . in what follows
, it will be convenient to write @xmath7 as @xmath20 .
the receiver updates the channel estimate during each coherence block with a kalman filter @xcite , given the model ( [ eq : gm ] ) and the pilot symbols , and relays the estimate back to the transmitter .
the feedback occurs between the pilot and data symbols , and is assumed to occupy an insignificant fraction of the coherence time .
we re - write the noise vector @xmath9 as @xmath21^\dag$ ] where @xmath22 and @xmath23 are , respectively , @xmath24 and @xmath25 vectors and @xmath26^\dag$ ] denotes hermitian transpose .
the channel estimate @xmath27 and estimation error @xmath28 evolve according to the following kalman filter updates : @xmath29 where @xmath30 it is straightforward to show that the channel estimate @xmath31 in ( [ eq : kalest ] ) does not depend on @xmath1 .
( what is important is the total pilot energy per coherence block . )
hence the data rate is maximized by taking @xmath32 with fixed @xmath33 ( i.e. , the training power @xmath34 ) .
training therefore requires an overhead of @xmath35 fraction of the channel uses .
we ignore this overhead for the time being and focus on optimizing the training power @xmath33 .
this issue is revisited in sec .
[ tl ] .
we wish to determine @xmath8 and @xmath33 , which maximize the achievable rate .
specifically , the channel estimate @xmath31 and variance @xmath36 determine the data power in the _ current _ coherence block , @xmath8 , and the pilot power in the _ next _ coherence block , @xmath37 .
we assume that the transmitter codes over many coherence blocks , and use the following lower bound on ergodic capacity , which accounts for channel estimation error , as the performance objective @xcite , @xmath38 where @xmath39 . in the next section , we formulate the joint pilot and data power optimization problem , and subsequently characterize the optimal power control policy implicitly as the solution to a discrete - time bellman equation .
the pilot and data power control problem can be stated as @xmath40 \\
\text{subject to : } & \,\,\label{eq : pconst } \limsup_{n \rightarrow \infty}e\left [ \frac{1}{n } \sum_{i=0}^{n-1 } \,{\epsilon}_i + \frac{1}{n}\sum_{i=0}^{n-1 } \,p_i \right ] \le \,p_{av},\\ \text{and } & \qquad\epsilon_i \le { \epsilon_{max } } , \end{aligned}\end{aligned}\ ] ] where the expectation is over the sequence of channel gains
. we have imposed an additional peak power constraint on the training power .
this is a discrete - time markov control problem , so that the solution can be formulated as an infinite - horizon dynamic program with an average value objective .
the system state at time ( block ) @xmath3 is @xmath41 , and the action maps the state to the power pair @xmath42 . to see that @xmath43 is the system state , note that @xmath44 in and @xmath31 are independent random variables , hence it follows from and that the probability distribution of @xmath45 is determined only by @xmath43 and the action @xmath46 .
the process @xmath47 is therefore a markov chain driven by the control @xmath48 .
the average power constraint in can be included in the objective through a lagrange multiplier giving the relaxed problem @xmath49\right]\ ] ] where @xmath50 is chosen to enforce the constraint .
if there exists a bounded function @xmath51 and a constant @xmath52 , which satify the bellman equation @xmath53 \right ] , \ ] ] then an optimal policy maximizes the right - hand side @xcite .
the function @xmath54 is called an `` auxiliary value function '' , and @xmath52 is the maximum value of the objective in .
the expectation @xmath55 $ ] is over the conditional probability of @xmath45 given @xmath56 and action @xmath57 . using the channel state evolution equations derived in section [ model ]
, we have @xmath58 = \int_0^{\infty}\ , v(u , \theta_{i+1})\ , f_{{\hat{\mu}}_{i+1 } | s_i } ( u ) du\ ] ] where @xmath59 is the conditional density of @xmath60 given @xmath61 , and @xmath62 is given by . from it
follows that @xmath59 is ricean with noncentrality parameter @xmath63 and variance @xmath64 \theta_{i+1 } + r^2 \hat{\mu}$ ] , where @xmath62 and @xmath65 are given by and , respectively .
the bellman equation is an integral fixed point equation , and appears to be difficult to solve analytically . to gain insight into properties of optimal policies , we consider the following scaling , corresponding to a low snr regime : 1 .
time is scaled by the factor @xmath66 , where @xmath67 is large , so that each coherence block of @xmath0 symbols corresponds to @xmath68 time units .
therefore one time unit in the scaled system contains @xmath69 coherence blocks , or equivalently @xmath70 channels uses .
the correlation between adjacent coherence blocks is @xmath71 , where @xmath72 is a constant .
hence this correlation goes to one as @xmath73 ( equivalently , the channel coherence time goes to zero ) , but with fixed correlation between blocks separated by @xmath70 channel uses .
3 . to maintain constant energy over @xmath70 channel uses , the average power @xmath74 , data power @xmath8 , training power @xmath33 and the maximum training power @xmath75 are each scaled by @xmath66 . in the limit as @xmath73
, it can be shown that the discrete - time , complex , gauss - markov process @xmath76 , given by , converges weakly to a continuous - time ornstein - uhlenbeck diffusion process @xmath77
( e.g. , see ( * ? ? ?
furthermore , the limiting channel process satisfies the stochastic differential equation ( sde ) @xmath78 where @xmath79 is complex brownian motion , and we assume that the initial state @xmath80 is a cscg random variable with zero mean and variance @xmath17 .
this is a gauss - markov process , which is continuous in probability , and has autocorrelation function @xmath81 where @xmath82 is the lag between the time samples of the channel .
hence @xmath72 determines how fast the channel varies relative to the symbol rate .
for example , the end - to - end normalized correlation across an interval of @xmath83 ( or , equivalently @xmath70 channel uses ) is @xmath84 .
the diffusion limit considered can be interpreted as _ zooming out _ on the channel and associated data transmission .
segments of the discrete channel process then become `` compressed '' in time , but with increasing correlation between successive coherence blocks so that the channel autocorrelation remains fixed .
prior work , which advocates the use of diffusion models for wireless channels is presented in @xcite . in this limit
the kalman filter continuously estimates the channel process , and the pilot and data powers are continuously updated based on continuous feedback .
( we will see that to achieve optimal performance the feedback need not be continuous . ) the optimal power control policy in the diffusion limit can then be interpreted as an approximation for the optimal discrete - time policy .
( this will be illustrated numerically . )
the power scaling by @xmath66 can be interpreted as introducing @xmath70 parallel , independent and statistically identical _ sub - channels _ over which the power is equally split .
hence the low snr regime corresponds to a _ wideband _ channel . in the diffusion
limit the channel estimate and estimation error updates given by and , respectively , become the dynamic equations @xmath85 where @xmath86 is a complex brownian motion independent of @xmath79 , and @xmath87 is the pilot power at time @xmath88 . a heuristic derivation of , and from the discrete - time equations , and is given in appendix [ ap : lmt ] . note
that both @xmath89 and @xmath90 are complex .
the following sde defining the evolution of the channel estimate @xmath91 can be obtained from and a straightforward application of ito s lemma ( * ? ? ?
4 ) : @xmath92\,dt + 2 \left[\theta ( t ) \sqrt{\frac{\epsilon ( t)}{{\sigma_z}^2 } } \hat{h}_r(t)\right ] d\overline{b}_{r}(t ) + 2 \left[\theta ( t ) \sqrt{\frac{\epsilon ( t)}{{\sigma_z}^2 } } \hat{h}_j(t)\right ] d\overline{b}_{j}(t),\ ] ] where @xmath93 and @xmath94 are , respectively , the real and imaginary parts of the complex brownian motion @xmath86 . we can re - write as @xmath95\,dt + \theta ( t ) \sqrt{\frac{\epsilon ( t ) { \hat{\mu}}(t)}{{\sigma_z}^2 } } \,\ , d\underline{b}(t),\ ] ] where @xmath96 is a _ real_-valued standard brownian motion .
[ lm : cont ] given @xmath97 $ ] , the state process @xmath98 , which is the solution to the stochastic differential equations and , has continuous sample paths . the proof is given in appendix [ ap : cont ] .
note that this lemma does not require the control input @xmath99 to be continuous in time .
this observation will be useful in the subsequent discussion .
we now consider the continuous - time limit of the optimization problem .
if the data power for the @xmath100 discrete coherence block is @xmath101 , then for large ( but finite ) @xmath70 , the objective @xmath102 becomes close to the continuous objective @xmath103 $ ] , where @xmath104 is given by and the index @xmath3 corresponds to time @xmath88 .
our problem is then to choose @xmath105 and @xmath106 , as a function of the state @xmath107 , to maximize the accumulated rate function ( over time ) , averaged over the channel process @xmath77 .
equivalently , we can maximize the scaled objective @xmath108 $ ] ( corresponding to the sum rate over @xmath70 parallel sub - channels ) .
a difficulty is that this objective is unbounded as @xmath109 . to simplify the analysis , we first take @xmath110 , which lower bounds the objective for all @xmath111 ( and corresponds to scaling up the power in the diffusion limit ) . after characterizing the optimal policy
we then replace the objective with the preceding scaled objective with fixed @xmath70 to generate numerical results .
we discuss the rate of growth of the scaled rate objective as @xmath109 .
] we therefore rewrite the discrete - time optimization as the continous - time control problem @xmath112 \\ \text{subject to : } & \,\ , \label{eq : pconst1 } \limsup_{t \rightarrow \infty}e\left [ \frac{1}{t } \int_{0}^{t } \,{\epsilon}(t)\,dt + \frac{1}{t}\int_{0}^{t } \,p(t)\,dt \right ] \le \,p_{av},\\ \text{and } & \qquad { \epsilon}(t ) \le { \epsilon_{max}}. \end{aligned}\end{aligned}\ ] ] analogous to , the bellman equation can be written as ( see @xcite ) @xmath113 \right\}\ ] ] where @xmath114 is the generator of the state process @xmath115 with pilot power @xmath105 ( * ? ? ?
7 ) , and is given by @xmath116 = \frac{e[dv]}{dt } = a + \epsilon b\ ] ] where @xmath117\end{aligned}\ ] ] and the dependence on @xmath88 is omitted for notational convenience . here
we ignore existence issues , and simply assume that there exists a bounded , continuous , and twice differentiable function @xmath54 satisfying .
note that @xmath54 is unique only up to a constant ( * ? ? ?
* ch . 4),@xcite .
[ thm ] given the pilot power constraint @xmath118 $ ] , the optimal pilot power control policy is given by @xmath119 in words , optimal pilot power control is achieved by a switching ( bang - bang ) policy .
this follows immediately from substituting the generator @xmath120 , given by - , into , i.e. , @xmath121\ ] ] where @xmath122 $ ] .
substituting into gives the final version of the bellman equation @xmath123 where @xmath124 .
an alternative way to arrive at and is to take the diffusion limit of the discrete - time bellman equation .
this alternative derivation is given in appendix [ ap : alt ] .
it is easily shown that the optimal data power allocation is @xmath125 \\ & = & \biggr ( \frac{- \lambda { \sigma_z}^2 ( 2 \theta + \hat{\mu } ) + \sqrt{\delta } } { 2 \lambda \theta ( \hat{\mu } + \theta ) } \biggr ) ^ { + } \label{eq : palloc}\end{aligned}\ ] ] where @xmath126 and @xmath50 determines @xmath74 in .
note that @xmath127 for @xmath128 .
this power allocation is the same as that obtained in @xcite , which considers a fading channel with constant estimation error , as opposed to the time - varying estimation error in our model .
from theorem [ thm ] the optimal pilot power control policy is determined by the switching boundary in the state space @xmath129 , which is defined by the condition @xmath130 .
this is referred to as a `` free boundary '' condition for the bellman pde @xcite .
the dynamical behavior of the optimal pilot power control policy is illustrated in fig [ fig : sysdyn1 ] .
the vertical and horizontal axes correspond to the state variables @xmath131 and @xmath132 , respectively .
the shaded region , @xmath133 , is the region of the state space in which @xmath134 , and @xmath135 in the complementary region @xmath136 .
these two regions are separated by the free boundary , @xmath137 .
the penalty factor @xmath50 determines the position of this boundary , and the associated value of @xmath74 .
the vertical line @xmath138 in the figure corresponds to the estimation error variance @xmath139 , which results from taking @xmath140 for all @xmath88 .
clearly , in steady state the estimation error variance can not be lower than this value , hence the steady - state probability density function ( pdf ) of the state @xmath141 is zero for @xmath142 . substituting @xmath140 in and setting @xmath143 gives @xmath144 suppose that the initial state is in @xmath145 .
with @xmath146 the state evolution equations and become @xmath147 and @xmath148 .
this implies that the state trajectory is a straight line towards the point @xmath149 until it hits the free boundary , as illustrated in fig .
[ fig : sysdyn1 ] .
therefore , for @xmath150 , @xmath50 must be selected so that the point @xmath149 lies in @xmath133 .
otherwise , the state trajectory eventually drifts to @xmath149 and stays there , corresponding to @xmath135 for all @xmath88 , @xmath151 ( because @xmath152 ) , and @xmath153 .
if the trajectory hits the free boundary below the point @xmath154 , then it is pushed back into @xmath145 .
this is because at the boundary @xmath155 and for @xmath156 , the drift term in is negative , namely , @xmath157 .
otherwise , if the trajectory hits the boundary above point @xmath154 , it continues into @xmath133 and settles along the line @xmath138 where the drift @xmath158 . for the discrete - time model with small , but positive @xmath159 , the state trajectory zig - zags around the boundary , as shown in fig .
[ fig : sysdyn1 ] . hence if the free boundary @xmath137 intersects @xmath138 at point @xmath154 , then in steady state the probability mass must be concentrated along the curve @xmath160 .
this is verified through monte carlo simulations and illustrated in fig .
[ fig : trace ] .
points in the state space are shown corresponding to a realization generated from and with @xmath161 ( so that @xmath162 ) .
the preceding discussion suggests that the steady - state probability associated with states not on the curve defined by the free boundary and @xmath163 tends to zero in the continuous - time limit .
this is stated formally in the next section .
we also remark that in region @xmath145 the pde is a `` transport equation '' @xcite , which has an analytical solution containing an arbitrary function of a single variable .
determining this function and the constant @xmath52 appears to be difficult , so that we will take an alternative ( more direct ) approach to determining the free boundary .
in this section we characterize the steady - state behavior of the state trajectory with the optimal switching ( bang - bang ) training policy , and compare with some simpler policies . in particular , we give the first - order pdf over the free boundary , which we subsequently use to compute the optimized boundary explicitly .
we will denote the free boundary as @xmath164 for @xmath165 . to simplify the analysis we make the following assumptions : * ( p1 ) the free boundary @xmath166 is a continuously differentiable curve such that @xmath167 * ( p2 )
the function @xmath168 is one - to - one , i.e. , for any @xmath169 such that @xmath170 , @xmath171 .
note that ( p1 ) requires @xmath172 to be large enough so that the entire free boundary ( @xmath137 in fig .
[ fig : sysdyn1 ] ) lies to the right of @xmath173 .
( that is , they do not intersect . )
the condition on the derivative of the free boundary curve is mild .
geometrically , it implies that the region enclosed by the free boundary @xmath137 and point @xmath174 is convex .
this condition is indeed satisfied by the optimized free boundaries computed in later sections .
[ lm : strip ] let the pilot power as a function of the state @xmath175 be given by @xmath176 then for any @xmath177 , the solution to and satisfies @xmath178 the proof is given in appendix [ ap : strip ] .
the theorem implies that for large @xmath88 the state @xmath175 moves along the free boundary @xmath179 .
hence for the discrete - time system with large @xmath70 , the state is typically confined to a narrow strip around the free boundary .
[ ssp ] given the pilot power control , the steady - state probability of training conditioned on the channel estimate @xmath180 is @xmath181}{[\theta_{\epsilon}(u)]^2{\epsilon_{max}}},\ ] ] and the steady - state pdf of the channel estimate @xmath182 is @xmath183 the proof is given in appendix [ ap : ssp ] . from the average training power for the pilot power control scheme can be computed as @xmath184 ^ 2 } f_{\hat{\mu}}(u ) \ , du.\ ] ] therefore if @xmath75 is large enough so that @xmath185 for all @xmath186 , then neither the pdf @xmath187 nor the average training power @xmath188 depends on @xmath75 .
this is because as @xmath172 increases , the probability of training , given by , decreases so that the average training power given @xmath182 , namely @xmath189 , remains unchanged , this overhead is reduced by increasing @xmath172 .
however , for the diffusion approximation to be accurate , @xmath190 must be small , hence @xmath172 can not be too large . ] .
in addition , we observe that @xmath191 is independent of the correlation parameter @xmath72 .
now consider the case in which the free boundary is constrained to be vertical , that is , @xmath192 .
this still corresponds to a switching policy , but where the variance of the channel estimation error is constrained to be a constant , independent of the channel estimate . from the steady - state pdf of @xmath182
is exponential , i.e. , @xmath193 , and the average training power is @xmath194 . now consider the _ constant _ pilot power control policy , where @xmath195 ( constant ) for all @xmath88 .
substituting @xmath196 into and setting @xmath143 implies that the steady - state estimation error variance @xmath197 for all @xmath88 .
in addition , the steady - state pdf of @xmath182 is exponential with mean @xmath198 .
hence for a given average training power , constant pilot power can give _
exactly _ the same estimation error and steady - state pdf as the switching policy with a vertical boundary .
both schemes therefore achieve the same rate with a total power constraint ( ignoring overhead due to pilot insertion ) .
we will see that the optimized boundary is _ not _ vertical , which implies that adaptive pilot power control can perform better than constant pilot power .
we also observe that the same performance as the optimal switching policy can be achieved by _ continuously _ varying the training power as a function of @xmath182 .
namely , taking @xmath199 gives the same steady - state pdf and training power as in and , respectively .
however , this scheme corresponds to feeding back the pilot power as a sequence of real numbers , which in principle require infinite precision . in contrast
, the switching policy can be implemented by fixing the training power and varying the rate at which pilot symbols are inserted .
the transmitter therefore does not need to know the exact value of the channel estimate .
more specifically , the optimal switching policy inserts pilots of power @xmath75 with probability @xmath200 ( or equivalently , once every @xmath201 coherence blocks ) when the channel estimate @xmath180 .
this requires at most one bit per coherence block to inform the transmitter whether or not to train in the next block .
( of course , the feedback can be substantially reduced by exploiting channel correlations . )
the switching policy therefore requires fewer training symbols than continuous pilot power control , which requires a pilot symbol every coherence block .
from the preceding discussion the optimal pilot policy is determined by the free boundary . here
we compute the free boundary by observing that this boundary must maximize the rate objective , assuming a switching policy for the pilot power .
a difficulty is that the rate objective of interest is @xmath202 , whereas the steady - state probabilities in theorem [ ssp ] were derived in the limit as @xmath73 . in what follows we use the asymptotic probabilities in theorem [ ssp ] to approximate the steady - state probabilities corresponding to large but finite @xmath70 .
simulation results have shown that the resulting free boundary is insensitive to the choice of @xmath70 in the objective .
also , subsequent simulation results in section [ wf : num ] show that the analytical performance results accurately predict the performance of the corresponding discrete - time model with the optimal switching policy when @xmath70 is a few hundred . with the preceding approximation for large but finite @xmath70
the optimal free boundary can be computed as the solution to the following functional optimization problem , @xmath203 f_{{\hat{\mu}}}(u ) du \\
\text{subject to : } & \label{eq : pconst_ss } \int_{0}^{\infty } p(u ) f_{{\hat{\mu}}}(u)du + \int_{0}^{\infty } \epsilon(u ) \,f_{{\hat{\mu}}}(u)\,du \le \,p_{av},\\ \text{and } & \qquad \qquad \theta_{\epsilon}(u ) \ge
\theta^\star \quad \text{for } \,\ , u \ge 0 , \end{aligned}\end{aligned}\ ] ] where @xmath204}{[\theta_{\epsilon}(u)]^2}$ ] can be interpreted as the average training power when the estimate @xmath180 .
the objective is the achievable rate averaged over the free boundary since for large @xmath70 , the entire probability mass becomes concentrated _ on _ the boundary . to proceed , define the lagrangian function as @xmath205 = n r(p / n , u , \theta ) - \lambda \left(p + \frac{2 \rho \sigma_z^2 ( { \sigma_h^2}- \theta)}{\theta^2 } \right)\ ] ] analogous to , the optimization problem can be re - stated as @xmath206 f_{{\hat{\mu}}}(u)du\ ] ] such that @xmath207 and @xmath50 is chosen to satisfy the power constraint with equality .
it is shown in appendix [ prop - wf - fbd ] that the solution is given by @xmath208 where @xmath139 is defined in , and @xmath209 satisfies @xmath210 + l_\lambda[np_d(u , \theta_f(u ) , \lambda ) , u , \theta_f(u ) ] = \\ \int_{u}^\infty l_\lambda[np_d(v , \theta_f(v ) , \lambda ) , v , \theta_f(v ) ]
\frac{\exp \left(- \int_{u}^v\frac{1}{{\sigma_h^2}- \theta_f(s)}ds\right)}{{\sigma_h^2}- \theta_f(v)}dv\end{gathered}\ ] ] for @xmath211 , where @xmath212 = -\frac{n\,p^2 u}{(p\theta + p u + n\sigma_z^2)(p\theta+ n\sigma_z^2 ) } + \frac{2 \lambda \rho \sigma_z^2(2\sigma_h^2 - \theta)}{\theta^3}\ ] ] and @xmath213 is given by .
the optimal data power allocation is given by @xmath214 .
the condition gives the value @xmath209 as a functional of the free boundary @xmath215 for @xmath216 .
hence we can compute the boundary numerically via a backward recursion provided that @xmath209 is known for large values of @xmath217 .
moreover , it can be shown that @xmath209 is a decreasing function of @xmath217 and as @xmath218 , it converges to a constant value , that is , @xmath219 .
taking the limit @xmath218 on both sides of , this value can be shown to satisfy @xmath220 hence as long as @xmath75 is large enough so that @xmath221 , the free boundary is given by @xmath209 , independent of @xmath75 .
substituting @xmath222 into , and using the fact that @xmath223 for all @xmath186 ( since @xmath221 ) , and @xmath224 , we obtain @xmath225 ^ 2}{[\theta_{\epsilon}^\star(0)]^3 } = \frac{1}{4 \rho \sigma_z^2 } \left[\frac{\bar{r}}{\lambda } - p_{av}\right].\ ] ] where @xmath226 is the optimized objective in .
note that this relation depends on @xmath70 only through the water - filling level @xmath50 .
clearly , the left - hand side of is positive , which implies that we should have @xmath227 . also , as @xmath228 ( low snrs ) , @xmath229 and @xmath50 increases .
similarily , the right - hand side decreases to zero as @xmath230 ( fast fading ) , so that in both cases @xmath231 .
the preceding approach to computing the optimal free boundary relies on the asymptotic pdf of the state in theorem [ ssp ] .
alternatively , it is possible to solve the continuous - time bellman equation directly .
this is potentially useful for other scenarios in which the steady - state distribution is more difficult to obtain . a challenge , however , is that the optimized free boundary is unknown _ a priori _ ,
i.e. , it is obtained as part of the solution .
hence none of the standard numerical methods for solving pdes , which rely on specified boundary conditions , can be directly applied .
it is shown in appendix [ varinq ] that a numerical solution to the free boundary bellman equation can be obtained by re - formulating the problem as a quadratic program .
that method can be used to obtain a solution to a general class of free boundary problems and does not require knowledge of the steady - state statistics .
however , such a numerical computation has high complexity , and can be sensitive to parameter variations . in particular , the free boundary obtained from that method is often irregular ( not smooth ) due to discretization and finite - precision effects . here
we present some numerical examples of free boundaries obtained by solving the optimization problem along with performance results .
the analytical ( diffusion ) results are also compared with results from a monte carlo simulation of the discrete - time system .
to solve we discretize the @xmath182 axis and also truncate it at a value @xmath232 .
for all of the results in this section @xmath233 , and @xmath234 ( 11.76 db ) .
the snr is then @xmath235 .
[ [ free - boundary - examples ] ] free boundary examples + + + + + + + + + + + + + + + + + + + + + + fig . [
fig : fbdn1000 ] shows free boundaries corresponding to @xmath236 and @xmath237 for different snrs .
the channel correlation with lag 1000 is therefore @xmath238 , corresponding to relatively fast fading .
( we abuse notation in this section by referring to @xmath182 as a particular realization of the channel estimate . )
also shown are the optimized vertical boundaries with a switching policy ( i.e. , @xmath239 with optimized @xmath240 ) .
recall that the performance with the vertical boundary is the same as training with a constant fraction of power , which results in the estimation error @xmath240 . for each boundary
the data power allocation is given by the optimal water - filling power allocation in .
the free boundaries are shaped so that the estimation error is larger for small values of @xmath182 , and smaller for larger values of @xmath182 .
the reason for this is that the pdf of @xmath182 is larger when @xmath182 is close to zero where the instantaneous rate @xmath104 is small .
( in fact , @xmath153 for @xmath241 . )
allowing larger estimation errors for small @xmath182 ( relative to the vertical boundary ) therefore does not significantly reduce the overall ergodic rate , whereas it saves a significant amount of training power .
furthermore , shifting the savings in training power to larger values of @xmath182 reduces the estimation error for those values , thereby increasing the rate ( since the rate increases with @xmath182 ) .
it will be shown in sec [ onoff : analytic ] that with an optimized on - off power allocation , for large enough @xmath70 the achievable rate for the free boundary control depends on the shape of the free boundary only through the harmonic mean of the function @xmath242 .
this can also be used to show that the boundary has the general shape shown in fig .
[ fig : fbdn1000 ] .
of course , the estimation error can not be too large for small @xmath182 , since otherwise @xmath182 may decreases to zero .
( note from that as @xmath243 , the density @xmath244 becomes concentrated around @xmath245 . ) also , fig .
[ fig : fbdn1000 ] shows that at smaller snrs @xmath246 is closer to @xmath247 , so that the free boundary is more skewed .
the curvature of the boundary near @xmath248 is a numerical artifact due to truncation of the boundary at @xmath249 .
namely , this curvature disappears as @xmath250 increases , since as discussed in the last section , @xmath251 is a decreasing function of @xmath182 and approaches the value @xmath252 as @xmath182 becomes large .
[ [ gain - in - achievable - rate ] ] gain in achievable rate + + + + + + + + + + + + + + + + + + + + + + + fig .
[ fig : n1000_cap ] compares the rate objective with the optimized free boundary to that achieved with the optimized vertical boundary at different snrs . with @xmath253 ( fast fading )
these results show that optimized pilot power control gives substantial gains at low snrs ( e.g. , a factor of two at an snr of @xmath254 ) .
the percentage gain diminishes with the snr . with @xmath255 ,
corresponding to a correlation of 0.6 with lag 1000 , the gain in achievable rate is relatively small .
the optimized free boundary gives the most gain at low snrs and fast fading , since in that region the training power constitutes a larger percentage of the total power budget .
[ [ comparison - with - monte - carlo - simulations ] ] comparison with monte carlo simulations + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + fig . [
fig : n1000_cap ] also compares the analytical results from the diffusion model with the performance of the original discrete - time model obtained from monte carlo simulations . specifically , the discrete - time system - was simulated with the kalman filter estimator - and the switching policy for pilot power defined by the optimized free and vertical boundaries .
the comparison in fig .
[ fig : n1000_cap ] shows that the analytical results for the optimized free boundary underestimates the achievable rate by about 15 - 20% .
the simulation and numerical optimization give nearly the same values with the optimized vertical boundary .
in addition to the parameters @xmath70 and @xmath72 , which are the same as for the analytical results , for the discrete - time model another parameter is @xmath0 , the number of samples per coherence block . recall that the diffusion model is obtained in the limit as @xmath256 , hence for fixed @xmath70 the analytical results should be more accurate for smaller values of @xmath0 ( corresponding to higher correlations between successive channel gains ) . however , smaller values of @xmath0 incur more overhead , since the training and channel state feedback occur each coherence block .
( we discuss this further in sec .
[ tl ] . ) the results in fig .
[ fig : n1000_cap ] correspond to @xmath257 . to see the effect of varying @xmath0 with fixed @xmath70 , fig .
[ fig : n1000_eavg_diffm ] compares the optimized average training power obtained via analysis and monte carlo simulations for different values of @xmath0 with an snr of 7 db .
as expected , the two curves grow apart as @xmath0 increases , but are reasonably close for @xmath258 .
fig [ fig : n1000_sspdf ] compares the simulated steady - state pdf of the channel estimate @xmath182 with @xmath259 with the asymptotic pdf .
the two curves nearly overlap .
further results show that achievable rates computed from the analytical model nearly match the simulated rates with the vertical boundary over a wide range of @xmath0 , whereas for the optimized boundary the difference remains similar to that shown in fig .
[ fig : n1000_cap ] . fig
[ fig : cap_diffn ] compares analytical and simulated results as a function of @xmath70 with @xmath260 . for a fixed @xmath74 and @xmath72 ,
as @xmath70 decreases , the snr per sub - channel increases and the channel varies at a faster rate since the correlation across @xmath70 channel uses is fixed at @xmath84 .
the achievable rate increases with @xmath70 due to the increase in training and channel state feedback .
[ on - off ] we show that the rate increases as @xmath261 . ) again the analytical results closely match the simulated results with the optimized vertical boundary , and underestimate the achievable rate by @xmath262 with the optimized free boundary . , the scaled rate objective increases without bound .
this occurs even though from theorem [ ssp ] the steady - state distribution of the system state converges to the exponential distribution . ]
the plots for average training power become close for @xmath263 .
an important consequence of the optimal switching policy for pilot power control is that it requires no more than one bit feedback per coherence block
. however , optimal _ data _ power control still requires infinite - precision feedback .
therefore to reduce the overall feedback rate , we now consider on - off data power allocation , which also requires at most one bit feedback per coherence block .
( the feedback could be reduced further by exploiting the time correlation of the channel . ) for the optimization problem we therefore set @xmath264 where the threshold @xmath265 will be optimized , and restate as @xmath266 where @xmath267 is the steady - state pdf of the channel estimate given by , @xmath268 is the average training power and @xmath269 . given a free boundary @xmath270 for @xmath186 , where @xmath271 , the objective in can be re - written as @xmath272 where we have used the fact that the total power constraint is satisfied with equality .
using , the probability of data transmission is @xmath273 furthermore , the rate in can be bounded as @xmath274 where the upper bound follows by replacing @xmath164 by @xmath275 , using jensen s inequality @xcite , and the fact that @xmath276 . observing
that @xmath277 is a decreasing function of @xmath265 , it can be shown that the upper bound is maximized with threshold @xmath278 such that @xmath279 where @xmath280 is an increasing function of @xmath70 .
( an exact description of @xmath281 is unecessary for the following analysis . )
we can re - write as @xmath282 where @xmath283 is the _ harmonic mean _ of @xmath284 for @xmath285 $ ] , that is , along the free boundary truncated at the threshold value @xmath278 . since @xmath286
, implies that @xmath278 grows as @xmath261 .
substituting @xmath278 into , we observe that the upper and lower bounds have the same asymptotic growth rate , so that the rate also has this growth rate , given by implies that @xmath287 . ]
@xmath288 since @xmath278 maximizes the upper bound in , this is the growth rate of the achievable rate .
we observe that this @xmath261 growth in achievable rate is the same as the growth in achievable rate for parallel rayleigh fading channels ( in frequency or time ) with a sum power constraint and perfect channel knowledge at the transmitter ( e.g. , see @xcite ) .
this is because the coherence blocks correspond to separate degrees of freedom ( i.e. , the transmitter can choose whether or not to transmit over each block ) , and the number of coherence blocks increases linearly with @xmath70 . for our model
the associated constant is @xmath289 , which accounts for channel estimation error , and depends on the channel correlation @xmath72 .
this product therefore determines the shape of the free boundary .
( note also that @xmath278 depends on the free boundary . )
namely , choosing boundary points closer to @xmath247 reduces @xmath188 , but also reduces the harmonic mean @xmath290 , and vice versa .
the optimal boundary balances @xmath188 and @xmath290 by shifting training power from small values of @xmath182 to larger values , as discussed previously in sec .
[ wf : num ] .
[ fig : fbdonoff ] shows free boundaries at different snrs obtained by solving the optimization problem numerically for @xmath291 and @xmath292 .
also shown are the optimized vertical boundaries with on - off data power control .
as with water - filling , the free boundary is shaped to save training power when @xmath182 is small ( high probability region ) and re - distribute it to the instances when @xmath182 is large ( low probability region ) .
the boundaries shown here are more irregular , due to the discontinuous data power allocation .
the shape of the boundary for @xmath293 is a straight line , but does not affect the objective since the rate depends on the harmonic mean for @xmath294 .
[ fig : n200_cap ] shows plots of achievable rates versus snr with the optimized free and vertical boundaries and on - off data power control .
plots corresponding to the optimal waterfilling data power allocation are also shown for comparison .
these results show that the performance with the optimized on - off power allocation are nearly the same as with water - filling . also shown
are the rates obtained via monte carlo simulations of the discrete - time system with the optimized boundary .
those are again higher than the rates calculated from the diffusion model , whereas the simulated rates with the vertical boundary closely match the analytical results .
so far we have ignored the time overhead due to the channel uses that are occupied by the training symbols . here
we restate the pilot power control problem taking this overhead into account . a switching policy for the pilot power
requires that one of the @xmath0 channel uses in a coherence block is a training symbol whenever the transmitter is directed to train . if the channel estimate for the coherence block is @xmath182 , then the probability of training ( as discussed in sec .
[ steadystate ] ) is given by @xmath295 , where @xmath296 .
therefore the original optimization problem can be reformulated , taking the training overhead in to account , by replacing the rate objective with @xmath297 f_{{\hat{\mu}}}(u)du .
\label{eq : thpt}\ ] ] of course , if either @xmath75 or @xmath0 is large , then the training symbol overhead is negligible and the problem reduces to .
otherwise , the overhead term will influence the free boundary and ergodic rate .
specifically , it will reduce the optimal training power @xmath298 ( so that the boundary shifts towards @xmath299 ) , since the overhead penalty is proportional to the training power .
[ fig : n200_cap_tl ] shows plots of the rate objective in versus snr with optimized free and vertical boundaries . for this figure @xmath291 and @xmath300 , corresponding to a worst - case loss in throughput due to training overhead . also , @xmath292 and @xmath234 ( 11.76 db ) .
the data power control is assumed to be on - off and only the analytical results ( obtained by maximizing ) are shown .
( note that the channel state pdf is still given by theorem [ thm ] . ) at low snrs the average training power and associated overhead are small , so that taking the overhead into account does not significantly affect the rate . at high snrs ( around 10 db )
the training overhead reduces the achievable rate by about @xmath301 with both the free and vertical boundaries .
the percentage improvement provided by the free boundary relative to the vertical boundary remains approximately the same .
a final remark is that when symbol overhead is taken into account , the throughput associated with the vertical switching policy is no longer the same as that associated with constant power control .
that is because the switching policy requires on average @xmath302 channel uses for training every coherence block , whereas constant power control requires one channel use for training every coherence block .
of course , this savings in overhead for the switching policy comes at the cost of feedback .
we have studied achievable rates for a correlated rayleigh fading channel , where both the data and pilot power are adapted based on estimated channel gain . in low snr and fast fading scenarios the pilot power constitutes a substantial fraction of the total power budget , so that pilot power adaptation can provide a substantial gain in achievable rates . by taking a diffusion limit ,
corresponding to low snrs ( or wideband channel ) and high correlation between consecutive channel realizations , several insights were obtained about the optimal pilot power control policy .
namely , it was shown that a policy that switches between zero and peak training power is optimal , and that the training power should be reduced when the channel is bad and increased when the channel becomes good .
the optimal policy in the diffusion limit was also explicitly characterized , and shown to provide a significant increase in achievable rate for low snrs and fast fading . for the discrete - time system of interest
the switching policy is equivalent to maintaining constant pilot symbol power , but inserting pilot symbols less frequently when the channel estimate is weak ( and vice versa ) .
when combined with on - off data power control , this requires finite feedback , and achieves essentially the same performance with the optimal ( water - filling ) data power control .
of course , the csi feedback required for optimal data and pilot power control can be substantially reduced by exploiting the correlation between successive coherence blocks .
several modeling assumptions have been made , which could be relaxed in future work .
for example , we have assumed that the receiver knows the statistical model of the channel . in practice ,
the receiver may assume ( or estimate ) a model , such as , which is mismatched to the actual channel statistics .
an issue then is how sensitive this overall performance is to this mismatch .
also , the first - order rayleigh fading model might be replaced with other fading models ( e.g. , ricean , nakagami , and higher - order autoregressive models ) .
additional issues may arise when considering other channel models .
for example , here we have imposed a power constraint , which is averaged over many coherence blocks .
the results can therefore be directly applied to parallel fading channels where the total power constraint is split among the channels . however , for a frequency - selective channel the total power summed over parallel channels might instead be constrained per coherence block .
other extensions and applications of diffusion models to multi - input multi - output ( mimo ) and multiuser channels remain to be explored .
substituting @xmath303 into and ignoring terms with higher power of @xmath159 , we obtain @xmath304 in the diffusion limit the noise @xmath305 can be modeled as @xmath306 , where @xmath79 is a standard complex brownian motion .
hence as @xmath307 , the preceding equation becomes .
substituting @xmath308 in gives @xmath309 replacing @xmath33 by @xmath310 in and substituting @xmath159 for @xmath311 gives @xmath312 combining and , ignoring the @xmath313 term , gives @xmath314 which becomes as @xmath307 .
substituting for @xmath16 and replacing @xmath33 by @xmath310 , can be re - written as , @xmath315\ ] ] where , @xmath316 is a zero mean unit variance cscg random variable independent of @xmath13 .
the term @xmath317 is a cscg random variable with mean zero and variance @xmath318 ( \delta t)^2 $ ] and hence can be ignored .
modeling @xmath319 as @xmath320 , where @xmath86 is a standard complex brownian motion independent of @xmath79 gives as @xmath307 .
defining the @xmath321 state vector as @xmath322^\dag$ ] , and can be re - written as , @xmath323\ , dt + \mathbf{v}[\hat{h}_r(t ) , \hat{h}_j(t ) , \theta(t ) ] \mathbf{\bar{b}}(t)\ ] ] where the drift and variance are given by @xmath324^\dag \\
\mathbf{v}(\hat{h}_r , \hat{h}_j , \theta ) & = & \text{diag } \left[\theta \sqrt{\frac{{\epsilon}}{2{\sigma_z^2 } } } , \,\ , \theta \sqrt{\frac{{\epsilon}}{2{\sigma_z^2 } } } , \,\ , 0\right]\end{aligned}\ ] ] respectively , the dependence on time is dropped for notational convenience , and the three entries of the vector @xmath325 are independent , real - valued , standard brownian motions . from ( * ? ?
* theorem 5.2.1 ) , given and , the solution to exists and is continuous in @xmath88 provided that the following two conditions are satisfied : @xmath326 @xmath327 where for any matrix @xmath328 with @xmath329 entry @xmath330 , @xmath331 , and @xmath332 are constants .
condition is called the linear dominance property and is called the lipschitz property . given @xmath333 , we have @xmath334 + 4\rho^2 \sigma_h^4 } + \sqrt{\frac{{\epsilon_{max}}\theta^2}{{\sigma_z^2}}}\\ & \le & \left(2\rho+\frac{{\epsilon_{max}}{\sigma_h^2}}{{\sigma_z^2}}\right ) \sqrt{\hat{h}_r^2 + \hat{h}_j^2+\theta^2 } + \sqrt{4\rho^2 \sigma_h^4}+ \sqrt{\frac{{\epsilon_{max}}\theta^2}{{\sigma_z^2}}}.\end{aligned}\ ] ] so that is satisfied . similarly for @xmath335
we have @xmath336 ^ 2(\theta_1-\theta_2)^2 } + \sqrt{\frac{{\epsilon_{max}}}{{\sigma_z^2}}(\theta_1-\theta_2)^2}.\end{gathered}\ ] ] so that is satisfied . since the solution to and , @xmath98 , is a continuous function of @xmath337 , it must also be continuous in @xmath88 .
we first rewrite the discrete - time bellman equation as @xmath338 - v(\hat{\mu } , \theta ) \right\},\ ] ] where @xmath339 - v(\hat{\mu } , \theta ) = \int_0^{\infty}\ , \left[v(u , \theta_{i+1 } ) - v(\hat{\mu } , \theta ) \right]\ , f_{\hat{\mu}_{i+1 } | s_i}(u ) du .\ ] ] assuming that @xmath51 is a continuous and smooth function , we can expand @xmath340 around @xmath341 via the taylor series @xmath342\\ + \textrm{higher - order terms}\end{gathered}\ ] ] where all the derivatives are computed at @xmath341 . as stated in sec .
[ discrete ] , @xmath343 conditioned on @xmath43 is ricean , so that @xmath344 where @xmath345 is the zeroth - order modified bessel function of the first kind and @xmath346 where @xmath62 and @xmath65 are given by and , respectively , with @xmath36 replaced by @xmath132 .
the first two moments are ( * ? ? ?
2 ) , @xmath347 & = & r^2\,\hat{\mu } + { \sigma_o}^2\\ e[\hat{\mu}_{i+1}^2 | ( \hat{\mu},\theta ) ] & = & 2 { \sigma_o}^4\,\left[1 + 2 \bigg ( \frac{r^2 \hat{\mu}}{{\sigma_o}^2 } \bigg ) + \frac{1}{2 } \bigg ( \frac{r^2 \hat{\mu}}{{\sigma_o}^2 } \bigg)^2 \right ] \label{eq : mom2 } \ ] ] next we take the diffusion limit . substituting @xmath348 in and replacing @xmath349 by @xmath350 gives @xmath351 making these substitutions in - gives @xmath352 & = & \left [ - 2 \rho \hat{\mu } + { \theta}^2 \frac{\epsilon}{{\sigma_z}^2}\right ] \delta t + o(\delta t^2)\\ e[({\hat{\mu}}_{i+1 } - \hat{\mu})^2 | ( \hat{\mu},\theta ) ] & = & \left [ 2 \hat{\mu } { \theta}^2 \frac{\epsilon}{{\sigma_z}^2}\right ] \delta t + o(\delta
t^2 ) \ ] ] it is easily shown that the higher - order moments @xmath353
\leq o(\delta t^2)$ ] for @xmath354 , hence we can ignore the higher - order terms in .
substituting @xmath355 and taking the diffusion limit , can be re - written as @xmath356 \delta t\ ] ] substituting into and combining with - gives @xmath357 - v(\hat{\mu } , \theta ) = \frac{\partial v}{\partial \hat{\mu } } \left [ - 2 \rho \hat{\mu } + { \theta}^2 \frac{\epsilon}{{\sigma_z}^2 } \right ] \delta t \\
+ \frac{\partial v}{\partial \theta } \left [ - 2 \rho \theta + 2 \rho { \sigma_h}^2 - { \theta}^2 \frac{\epsilon}{{\sigma_z}^2 } \right ] \delta t + \frac{\partial^2 v}{\partial \hat{\mu}^2 } \left [ { \theta}^2 \frac{\epsilon}{{\sigma_z}^2}\,\hat{\mu } \right ] \delta t\end{gathered}\ ] ]
lastly , @xmath52 , @xmath358 , @xmath359 and @xmath360 can be multiplied by @xmath159 without changing the original optimization problem . applying this scaling in , substituting into , and multiplying the entire equation by @xmath361 and letting @xmath307 gives .
define the _ distance _ from the free boundary at time @xmath88 as @xmath362 .
irrespective of the initial state , due to the drift term in and time continuity of the state process ( lemma [ lm : cont ] ) , with probability one there exists a finite time instant such that the state lies on the free boundary . without loss of generality , rename that instance as @xmath363 so that @xmath364 .
for any @xmath177 let @xmath365 .
by continuity of the state process , @xmath366 and there exists a @xmath367 . if @xmath368 , bang - bang control implies that @xmath369 and the dynamical equations and simplify to @xmath370 and @xmath371 .
then @xmath372dt,\ ] ] and since @xmath368 implies @xmath373 , we have @xmath374dt$ ] . thus given condition , we have @xmath375 whenever @xmath376 . however , this contradicts the fact that @xmath377 .
therefore we can not have a @xmath378 , which implies that @xmath379 for any @xmath177 .
next we show that @xmath380 . for any continuous and twice differentiable function
@xmath381 we must have ( * ? ? ?
7 ) @xmath382 \right\ } = 0,\ ] ] where the expectation is over the steady - state distribution of the state @xmath383 .
the generator @xmath384 $ ] is defined as in , except that the function @xmath385 is replaced by @xmath386 .
we choose @xmath386 to be a function of @xmath132 only , i.e. , @xmath387 , so that @xmath388 = a_{\epsilon}[w(\theta ) ] = w ' ( \theta ) \left[2 \rho ( { \sigma_h^2}-\theta ) - \frac{\epsilon \theta^2}{{\sigma_z^2}}\right]\ ] ] where @xmath389 denotes the derivative of @xmath390 with respect to @xmath132 .
let @xmath391 denote the steady - state probability of training given that the channel estimate is @xmath217 , as in .
rewriting as @xmath392 \right ] = 0 , \ ] ] and evaluating the inner conditional expectation using the preceding result that @xmath393 for any @xmath394 gives @xmath395 \right\ } \\
+ e_{{\hat{\mu } } } \left\ { p({\hat{\mu } } ) e_{\theta|{\hat{\mu}};\theta\ge\theta_{\epsilon}({\hat{\mu } } ) } \left[w'(\theta)\left(2 \rho ( { \sigma_h^2}-\theta ) - \frac{\epsilon_{max } \theta^2}{{\sigma_z^2}}\right)\right]\right\ } = 0\end{gathered}\ ] ] where @xmath396 $ ] denotes the expectation over @xmath132 given that the estimate is @xmath182 and @xmath397 .
choosing @xmath390 such that @xmath398 $ ] and substituting in gives @xmath399 + e_{{\hat{\mu } } } [ p({\hat{\mu } } ) ] = 0 , \ ] ] which implies @xmath400 = e_{{\hat{\mu } } } \left[p({\hat{\mu}})\left[\frac{\epsilon_{max } \theta_{\epsilon}({\hat{\mu}})^2}{{\sigma_z^2 } } - 2 \rho ( { \sigma_h^2}-\theta_{\epsilon}({\hat{\mu } } ) ) \right]\right ] .\ ] ] next choose @xmath401 so that @xmath402 .
for this choice gives @xmath403 \right ] = e_{{\hat{\mu } } } \left[p({\hat{\mu } } ) e_{\theta|{\hat{\mu}};\theta \ge \theta_{\epsilon}({\hat{\mu } } ) } \left(\frac{\epsilon_{max } \theta^2}{{\sigma_z^2 } } -2 \rho ( { \sigma_h^2}-\theta)\right)\right].\ ] ] we now argue by contradiction that for any @xmath394 , @xmath404 almost everywhere ( a.e . ) in the set @xmath405 .
if this were not the case , then we must have @xmath406 over a subset in @xmath407 with positive measure . since @xmath408 is a positive increasing function of @xmath132 , implies @xmath409 > e_{{\hat{\mu } } } \left [ p({\hat{\mu}})\left(\frac{\epsilon_{max } \theta_{\epsilon}({\hat{\mu}})^2}{{\sigma_z^2 } } - 2 \rho ( { \sigma_h^2}-\theta_{\epsilon}({\hat{\mu } } ) ) \right)\right]\ ] ] with strict inequality , which contradicts .
hence this establishes the proposition for any @xmath394 .
as in appendix [ ap : strip ] , we use the fact that for any continuous and twice differentiable function @xmath381 , we have ( * ? ? ?
7 ) @xmath410 \right\ } = 0,\ ] ] where the expectation is over the steady - state distribution of @xmath383 and the generator @xmath384 $ ] is given by with @xmath385 replaced by @xmath386 .
choosing @xmath411 in to be a function of @xmath132 only and applying proposition [ lm : strip ] gives @xmath412 = 0\ ] ] where @xmath413.\ ] ] next we observe that @xmath414 a.e .
in the set @xmath415 . if this were not the case , then since @xmath416 is a one - to - one function , we could choose @xmath417 such that @xmath418 , which would make the left - hand side of strictly positive .
therefore setting @xmath414 gives the steady - state probability of training given @xmath182 shown in .
we now solve for the steady - state pdf @xmath267 .
choosing @xmath419 , a continuous and twice differentiable function of @xmath182 only , and applying the generator gives @xmath420 = - 2\rho { \hat{\mu}}w_2'({\hat{\mu } } ) + \left[w_2'({\hat{\mu } } ) + { \hat{\mu}}w_2''({\hat{\mu}})\right ] \,\frac{\epsilon \,\theta^2}{{\sigma_z^2}}.\ ] ] the necessary condition can now be written as @xmath421f_{{\hat{\mu}}}(u ) du = 0,\ ] ] where @xmath422 ^ 2{\epsilon_{max}}}{{\sigma_z^2 } } p(u ) \quad \textrm{and } \quad d(u ) = u \frac{[\theta_{\epsilon}(u)]^2{\epsilon_{max}}}{{\sigma_z^2 } } p(u).\ ] ] we can further choose @xmath423 to satisfy the following properties : @xmath424}{du } w_2(u ) = 0\end{aligned}\ ] ] and using integration by parts we can re - write as @xmath425 - \frac{d}{du}[c(u)f_{{\hat{\mu}}}(u)]\right ) du = 0.\ ] ] since this condition must be satisfied for any such @xmath426 , we have @xmath427 - \frac{d}{du}[c(u)f_{{\hat{\mu}}}(u ) ] = 0 , \qquad \textrm{a.e . }
\,\,\,u \ge 0.\ ] ] substituting into gives the differential equation @xmath428f_{{\hat{\mu}}}(u)\bigg\ } = 0 , \quad \textrm{a.e . }
\,\,\,u \ge 0\ ] ] which can be further simplified as @xmath429u \frac{df_{{\hat{\mu}}}(u)}{du } + 2\rho u \left(1 - \frac{d \theta_{\epsilon}(u)}{du}\right)f_{{\hat{\mu}}}(u ) + k = 0,\ ] ] where @xmath430 is a constant .
this is a first - order ordinary differential equation with solution @xmath431 \int_{0}^{u}\frac{\exp[i(t)]}{2\rho\left[{\sigma_h^2}-\theta_{\epsilon}(t)\right ] t } dt + k_1 \exp[-i(u)],\ ] ] where @xmath432 - \log[{\sigma_h^2}-\theta_{\epsilon}(0)]\end{aligned}\ ] ] and @xmath433 is another constant , which needs to be determined .
since @xmath267 is a pdf , we must have @xmath434 , which implies @xmath435 .
this is because the first integral in is unbounded , that is , @xmath436}{2\rho({\sigma_h^2}-\theta_{\epsilon})t } dt \ge \int_0 ^ 1\frac{1}{2\rho{\sigma_h^2}t } dt = \infty.\ ] ] in addition we must have @xmath437 , which implies @xmath438 . substituting these values into gives .
first we fix the free boundary @xmath164 and optimize the data power allocation . for any @xmath165 setting the derivative of the objective function with respect to @xmath439 to zero
gives the optimal power allocation @xmath440 . substituting this @xmath441 into and taking the derivative with respect to @xmath164 gives the optimality condition @xmath442 \cdot \frac{\partial p^\star({\hat{\mu}})}{\partial \theta_{\epsilon}({\hat{\mu } } ) } + \frac{\partial l_{\lambda}}{\partial \theta } [ p^\star({\hat{\mu } } ) , { \hat{\mu } } , \theta_{\epsilon}({\hat{\mu } } ) ] \right]f_{{\hat{\mu}}}({\hat{\mu } } )
= \\ - l_{\lambda}[p^\star({\hat{\mu } } ) , { \hat{\mu } } , \theta_{\epsilon}({\hat{\mu } } ) ] \frac{f_{{\hat{\mu}}}({\hat{\mu}})}{{\sigma_h^2}- \theta_{\epsilon}({\hat{\mu } } ) } + \frac{1}{[{\sigma_h^2}-\theta_{\epsilon}({\hat{\mu}})]^2}\int_{{\hat{\mu}}}^{\infty } l_{\lambda}[p^\star(v ) , v , \theta_{\epsilon}(v ) ] f_{{\hat{\mu}}}(v ) dv.\end{gathered}\ ] ] note that @xmath443 for @xmath241 so that @xmath444 .
for @xmath445 we have @xmath446 = 0 $ ] . therefore reduces to with @xmath447 replaced by @xmath448 .
the additional constraint @xmath449 implies .
we first observe that can be written as the _ variational inequality _
@xcite @xmath450 a solution to is a solution to and vice versa .
now consider the following optimization problem , @xmath451 d\theta du \nonumber\\\ ] ] @xmath452 where @xmath453 for @xmath454 and @xmath455 . if @xmath456 , @xmath457 , and @xmath458 , then the solution to is a solution to . also , a solution to with zero objective value is a solution to .
the second term in the objective function is included to regularize the numerical solution .
the effect of this term can be controlled by changing the weights @xmath459 and @xmath460 .
these weights affect both the accuracy of the results and also the rate at which the non - linear optimization algorithm converges .
the training region is where @xmath461 .
therefore the free boundary can be obtained by solving numerically given values for @xmath50 and @xmath72 . | we consider data transmission through a time - selective , correlated ( first - order markov ) rayleigh fading channel subject to an average power constraint .
the channel is estimated at the receiver with a pilot signal , and the estimate is fed back to the transmitter .
the estimate is used for coherent demodulation , and to adapt the data and pilot powers .
we explicitly determine the optimal pilot and data power control policies in a continuous - time limit where the channel state evolves as an ornstein - uhlenbeck diffusion process , and is estimated by a kalman filter at the receiver .
the optimal pilot policy switches between zero and the maximum ( peak - constrained ) value ( `` bang - bang '' control ) , and approximates the optimal discrete - time policy at low signal - to - noise ratios ( equivalently , large bandwidths ) .
the switching boundary is defined in terms of the system state ( estimated channel mean and associated error variance ) , and can be explicitly computed . under the optimal policy , the transmitter conserves power by decreasing the training power when the channel is faded , thereby increasing the data rate .
numerical results show a significant increase in achievable rate due to the adaptive training scheme with feedback , relative to constant ( non - adaptive ) training , which does not require feedback .
the gain is more pronounced at relatively low snrs and with fast fading .
results are further verified through monte carlo simulations .
limited - rate feedback , gauss - markov channel , channel estimation , adaptive training , wideband channel , diffusion approximation , free boundary problems , bang - bang control , variational inequalities . |
in recent years a number of authors have investigated the microlensing of extended stellar sources .
( hereafter snw95 ) have shown that the light curves of extended sources can exhibit a significant chromatic dependence , essentially because limb darkening renders the effective radius of the star a function of wavelength .
thus , in addition to improving constraints on the lens parameters , modelling the microlensing of extended sources provides a powerful tool for gravitational imaging stellar surfaces . in this contribution
we describe how microlensing light curves of extended sources may be used as a test of stellar atmosphere models .
we generate artificial light curves , assuming a particular model atmosphere , and use the backus - gilbert numerical inversion method to estimate the radial stellar intensity profile from the observed light curves .
the ( time dependent ) integrated flux , @xmath0 , from an extended stellar source of radius , @xmath1 , lensed by a point lens is given by ( c.f .
snw95 ) @xmath2 where @xmath3 is the projected distance from the lens to the element of the stellar surface and the amplification function , @xmath4 , takes its usual analytic form for a point source .
( note that @xmath3 is a function of @xmath5 , @xmath6 and @xmath7 ) .
if we assume that the projected stellar surface displays circular symmetry
i.e. @xmath8 we may write the above equation in the form @xmath9 where @xmath10 and the kernel function , @xmath11 , is obtained by integrating over @xmath6 .
a solution to this integral equation for @xmath12 can be obtained by applying the backus - gilbert inversion procedure .
this method takes account of the smoothing effect of the kernel function and reconstructs a regularised estimator of @xmath12 which optimises the trade - off between the bias and variance of the estimator . for details of the method
in order to test the feasibility of reconstructing stellar surface profiles , we assumed a simple , linear , limb darkening law with coefficient , @xmath13 .
we generated microlensed light curves of typically 100 data points , with an even sampling rate , and gaussian noise added to the photometry at a level of typically 2% of the baseline flux .
in most cases we took the impact parameter and the einstein radius equal to the stellar radius .
we considered two model atmospheres : @xmath14 ; @xmath15 ; @xmath16 , @xmath17 @xmath18 ; @xmath19 ; @xmath20 , @xmath21 where the johnson @xmath22 and @xmath23 band linear limb darkening coefficients are from .
we carried out inversions of the @xmath22 and @xmath23 band intensity profiles for these models , and investigated the effect of changing the impact parameter , stellar radius , light curve sampling rate and photometric accuracy .
figure [ fig : fig1 ] illustrates the reconstructed @xmath23 band profile for model ( 2 ) , for the case of the impact parameter and stellar radius equal to the einstein radius of the lens .
the errors on the recovered solution are determined from the covariance matrix of the backus - gilbert estimator .
we can see that the reconstructed profile is significantly biased for @xmath24 , due to the smoothing effect of the kernel function , but the true profile is well recovered over the interval @xmath25 .
we now briefly summarise the results of varying the stellar and lens parameters .
* the inversions generally recover the true profiles well over the interval @xmath25 , for a wide range of @xmath26 , @xmath27 and @xmath13 . *
increasing the einstein radius improves significantly the inversions for @xmath28 , but no improvement is seen for @xmath24 . * reducing the impact parameter ( i.e. a transit event ) significantly improves the accuracy of the reconstruction for @xmath28 and reduces the bias for @xmath24 . for impact parameters greater than the stellar radius , however , the reconstruction deteriorates rapidly for all @xmath29 . *
even with photometric errors of 10% a reasonable recovery of @xmath12 is still obtained over the interval @xmath25 ; on the other hand , reducing the errors to only 0.2% does _ not _ improve the recovery for @xmath24 , however .
this is because the bias is primarily due to the ill - posedness of the kernel over this range , and not due to the photometric errors . *
the reconstructions become unacceptably noisy when the number of light curve data points is reduced to @xmath30 , but there is little further improvement in accuracy above @xmath31 .
our results indicate that with realistic light curve sampling and photometric errors one can accurately reconstruct , at least in part , the multicolour radial intensity profiles of extended stellar sources from their microlensed light curves , provided that the impact parameter of the lens is comparable to the stellar radius .
the smoothing properties of the kernel function result in a biased solution for @xmath24 , unless the lensing event is a transit with small impact parameter .
nevertheless , the accurate recovery over the interval @xmath25 is a robust result over a wide range of stellar temperatures and limb darkening coefficients . despite the narrow width of this ` good fit ' annulus , it is still adequate to usefully discriminate between different model atmospheres
e.g. two models with the same temperature but with @xmath15 and @xmath32 .
thus , we conclude that broad band microlensed photometric light curves are a powerful tool for investigating extended stellar sources and testing model stellar atmospheres , and form a useful adjoint to spectroscopic and polarimetric microlensing signatures .
we are currently investigating the application of inversion techniques to more realistic model atmospheres and stellar intensity profiles . | we investigate the feasibility of reconstructing the radial intensity profile of extended stellar sources by inverting their microlensed light curves . using a simple , linear , limb darkening law as an illustration
, we show that the intensity profile can be accurately determined , at least over the outer part of the stellar disc , with realistic light curve sampling and photometric errors .
the principal requirement is that the impact parameter of the lens be less than or equal to the stellar radius .
thus , the analysis of microlensing events provides a powerful method for testing stellar atmosphere models .
gravitational lensing , stars : atmospheres |
the grassmann manifolds ( sometimes , referred to as simply the grassmannian ) which parametrize vector subspaces of fixed dimensions of a given vector space are fundamental objects , and appear in many areas of mathematics .
their study dates back to plcker who , in the 19th century , considered the case of vector subspaces of dimension @xmath3 of a space of dimension @xmath4 . in this note
we study the integral cohomology of _ real _ grassmann manifolds .
the cohomology of _ complex _ grassmann manifolds is well - known but the real case poses additional challenges .
we give an explicit description of the betti numbers and the incidence graph which leads to the integral cohomology .
there is a well - known decomposition of a real grassmann manifold into _
schubert cells_. the schubert cells can be parametrized with young diagrams and these young diagrams will then also appear in a co - chain complex to calculate the cohomology . in this note
the young diagrams are filled in with a letter @xmath1 in a checkered pattern and this gives rise to a power of @xmath1 for each young diagram and therefore to a polynomial by considering certain _ alternating _ sum .
this polynomial is computed explicitly and , it is shown to contain the information of all the betti numbers of the real grassman manifolds .
that is , it gives the poincar polynomial of the manifold .
there is also an alternative construction of these polynomials in terms of varieties over a finite field @xmath5 where this time @xmath1 is a power of a prime number .
one can consider varieties over @xmath5 that naturally correspond to the _ real _ graassmann manifolds and then count the number of @xmath5 points .
this again gives rise to a polynomial in @xmath1 which agrees with the polynomial that was obtained directly from the young diagrams .
the results in this paper were initially motivated by considering real grassmann manifolds in terms of the kp hierarchy @xcite .
the real grassmannian @xmath0 is defined by @xmath6 let @xmath7 be a basis of @xmath8 , and @xmath9 a basis of @xmath10-dimensional subspace @xmath11 .
then there exists a full rank @xmath12 matrix @xmath13 such that we have @xmath14 where @xmath15 is the transpose of the matrix @xmath13 .
since the left action of @xmath16 on @xmath13 does not change the subspace @xmath17 , @xmath0 can be expressed as @xmath18 where @xmath19 is the set of all @xmath12 matrices of rank @xmath10 .
the @xmath16-action puts @xmath13 in the canonical form called the reduced row echelon form ( rref ) . for example
, the rref for the generic matrix has the form , @xmath20 the 1 s in the matrix are called the pivots of the rref . in general , each @xmath10-element subset @xmath21 of @xmath22=\{1,2,\ldots , n\}$ ] is called the _
schubert symbol _ and represents the index set of the pivots for the matrix @xmath13 .
then the set of all @xmath10-element subsets @xmath23}{k}$ ] gives a parametrization for the schubert decomposition of @xmath0 , i.e. @xmath24}{k}}\omega^\circ_{i}\ ] ] where @xmath25 is the schubert cell , and @xmath26 means that the rref of @xmath13 has the pivot index @xmath27 . let @xmath28 denote the set of minimal coset representatives defined by ( see @xcite ) @xmath29 where @xmath30 is the length of @xmath31 , and @xmath32 is the maximal parabolic subgroup generated by @xmath33 since there is a bijection between @xmath23}{k}$ ] and @xmath28 , the schubert decomposition can be also expressed in terms of @xmath28 , i.e. @xmath34 where we identify @xmath35 .
more precisely , we have @xmath36 , i.e. @xmath37 .
note that the dimension of @xmath38 is given by dim@xmath39 .
we label the schubert cell @xmath38 for @xmath40 by the young diagram @xmath41 . given a schubert symbol @xmath42 , first define @xmath43 with @xmath44 ( note @xmath45 ) and @xmath46 .
this @xmath47 gives a partition of number @xmath48 .
then the young digram with shape @xmath47 is a collection of @xmath48 top - left - justified boxes with @xmath49 boxes in the @xmath50-th row from the bottom as shown below : ( 130,65 ) ( 5,55)(1,0)145 ( 5,45)(1,0)112 ( 5,35)(1,0)91 ( 5,25)(1,0)91 ( 5,15)(1,0)71 ( 5,5)(1,0)71 ( 5,5)(0,1)50 ( 20,5)(0,1)50 ( 35,5)(0,1)50 ( 76,5)(0,1)10 ( 60,5)(0,1)10 ( 96,25)(0,1)10 ( 80,25)(0,1)10 ( 117,45)(0,1)10 ( 100,45)(0,1)10 ( 50,45)(0,1)10 ( 10,29)@xmath51 ( 22,29)@xmath52 ( -5,29)@xmath50 ( 98,29)@xmath53 ( 81,29)@xmath54 ( 10,49)@xmath55 ( -5,49)@xmath10 ( 22,49)@xmath56 ( -5,38)@xmath57 ( -5,18)@xmath57 ( 65,38)@xmath57 ( 101,49)@xmath58 ( 67,49)@xmath59 ( 45,29)@xmath59 ( 20,20 ) ( 47,18)@xmath57 ( 10,9)@xmath60 ( -5,9)@xmath61 ( 45,9)@xmath59 ( 78,9)@xmath62 ( 25,9)@xmath63 ( 61,9)@xmath64 ( 119,48)@xmath65 ( 5,59)@xmath66 ( 21,59)@xmath67 ( 100,59)@xmath68 ( 67,59)@xmath59 ( 11,-2)@xmath61 ( 26,-2)@xmath69 ( 45,-2)@xmath59 ( 61,-2)@xmath70 in the diagram , one can consider a lattice path starting from the bottom left corner and ending to the top right conner with the label @xmath71 in the counterclockwise direction , then the indices in the schubert symbol @xmath72 appear at the vertical paths . in the case of @xmath73 ,
i.e. the top cell , corresponding to the longest element @xmath74 , we write @xmath75 representing the @xmath76 rectangular diagram . here a reduced expression of @xmath77 can be written by @xmath78 where each @xmath79 is the product of @xmath80 in the @xmath50-th row from the bottom .
note that each @xmath41 is a sub - diagram of the @xmath81 of the top cell , and we have the inclusion relation , @xmath82 ( notice the direction of the bruhat order in this paper . )
we then define : based on a weak bruhat order , we define the _ weak bruhat graph _
, @xmath83 that is , the vertices are gives by the young diagrams @xmath41 , and the edges are defined by @xmath84 for some @xmath85 .
this graph is also referred to as _ young s lattice _ of @xmath0 .
below we define additional graphs @xmath86 , @xmath87 whose vertices have parameters involving the young diagrams @xmath41 .
the blanket notational convention is that if there is an edge in any of these graphs involving a pair @xmath88 then there is a simple reflection @xmath89 such that @xmath90 with @xmath91 .
these graphs can therefore be considered as subgraphs of the weak bruhat graph even if this is not explicitly restated .
[ gr25 ] the graph @xmath92 is given by @xmath93[8][$s_2 $ ] & \longrightarrow & \young[2][8][$s_2$,$s_3 $ ] & \longrightarrow & \young[3][8][$s_2$,$s_3$,$s_4 $ ] \\[-2.5ex]\\ & & \downarrow & & \downarrow & & \downarrow \\[-2ex]\\ & & \young[1,1][8][$s_2$,$s_1 $ ] & \longrightarrow & \young[2,1][8][$s_2$,$s_3$,$s_1 $ ] & \longrightarrow & \young[3,1][8][$s_2$,$s_3$,$s_4$,$s_1 $ ] \\[-2ex ] \\ & & & & \downarrow & & \downarrow \\[-2ex]\\ & & & & \young[2,2][8][$s_2$,$s_3$,$s_1$,$s_2 $ ] & \longrightarrow & \young[3,2][8][$s_2$,$s_3$,$s_4$,$s_1$,$s_2 $ ] \\[-2ex]\\ & & & & & & \downarrow \\[-2ex]\\ & & & & & & \young[3,3][8][$s_2$,$s_3$,$s_4$,$s_1$,$s_2$,$s_3 $ ] \end{matrix}\ ] ] each arrow @xmath94 represents the action @xmath89 to the pivot set , and new box added by this action is labeled by @xmath89 . for example , @xmath95 in the top left one .
notice that a reduced expression of @xmath40 can be obtained by following the paths @xmath94 starting from @xmath96 to the diagram @xmath41 .
for example , the diagram at the right side in the second row can be written by @xmath97 depending on the three different paths .
let us first denote the schubert variety marked by @xmath41 as @xmath98 let @xmath99 denote the schubert class associated to @xmath100 , i.e. @xmath101 $ ] .
then the co - chain complex of @xmath0 is given in the form , @xmath102 the co - boundary operator , @xmath103 , gives the form , @xmath104}~\sigma_{w'},\ ] ] where the coefficient @xmath105 $ ] is called the _ incidence _ number and it takes either @xmath106 or @xmath107 ( see e.g. @xcite ) .
we denote the incidence graph of @xmath0 by @xmath108=\pm 2\right\}.\ ] ] then identifying @xmath99 with the young diagram @xmath41 in the bruhat graph @xmath109 , the _ arrow _ @xmath94 in @xmath109 can be considered as the action of the co - boundary operator @xmath110 .
the incidence graph @xmath111 is then obtained by giving specific information on the incidence numbers for those arrows .
the main result of this paper is to provide a `` _ simple _ '' representation of the incidence graph in terms of the bruhat graph with those specific incidence numbers : for this purpose , let us first define : [ weighted ] we define a @xmath112-_weighted _ schubert variety @xmath113 where @xmath114 is defined as follows : * insert the letter @xmath112 and @xmath115 alternatively in the boxes of the young diagram @xmath41 , such that @xmath112 locates at the northwest conner box of the diagram .
that is , the first box with @xmath55 has @xmath1 , then the boxes with @xmath116 have @xmath1 , and the boxes with @xmath117 have 1 .
* then @xmath114 for each diagram @xmath41 filled with @xmath61 and @xmath1 in the previous step is defined by @xmath118 we will also need to refer to the @xmath1 weighted schubert variety @xmath119 defined in exactly the same way but now the boxes with @xmath116 have @xmath61 , and the boxes with @xmath117 have @xmath1 . to make a distinction , the first arrangement of @xmath1 s in the young diagram
will be referred to as _ standard _ and the second as @xmath1-_shifted_. we will show in subsection [ proof ] below that the monomial @xmath120 can be expressed a in terms of the hecke algebra action , a @xmath1-deformation of the weyl group @xmath121 . for each @xmath40 , one can easily find the formula of @xmath114 .
[ etaformula ] the function @xmath114 is given by @xmath122 where @xmath123 if @xmath124even , and @xmath125 if @xmath124odd .
[ graphg ] we define a graph @xmath86 for the @xmath1-weighted schubert varieties by modifying the bruhat graph @xmath109 , @xmath126 that is , identifying each young diagram with @xmath1 s as the @xmath1-weighted schubert variety , and replacing the edge @xmath94 between @xmath127 and @xmath128 in the bruhat graph @xmath109 with @xmath129 if @xmath130 .
there is an alternative graph @xmath87 in which an edge @xmath131 corresponds instead to @xmath132 , i.e. @xmath133 the graph @xmath134 is given by @xmath135[8][$q$ ] & { \longrightarrow } & \young[2][8][$q$,$1 $ ] & \longrightarrow & \young[3][8][$q$,$ 1$,$q$ ] \\[-2.5ex]\\ & & { \downarrow } & & { \downarrow } & & { \downarrow } \\[-2ex]\\ & & \young[1,1][8][$q$,$1 $ ] & { \longrightarrow } & \young[2,1][8][$q$,$1$,$1 $ ] & \longrightarrow & \young[3,1][8][$q$,$1$,$q$,$1 $ ] \\[-2ex ] \\ & & & & \downarrow & & \downarrow \\[-2ex]\\ & & & & \young[2,2][8][$q$,$1$,$1$,$q$ ] & \longrightarrow & \young[3,2][8][$q$,$1$,$q$,$1$,$q$ ] \\[-2ex]\\ & & & & & & { \downarrow } \\[-2ex]\\ & & & & & & \young[3,3][8][$q$,$1$,$q$,$1$,$q$,$1 $ ] \end{matrix}\ ] ] then we have the main theorem of this paper : [ main ] the graph @xmath86 with the edges @xmath131 having the incidence number @xmath136 ( others are zero ) gives the _ incidence _
graph @xmath111 of the real grassmann manifold @xmath0 . based on this theorem
, we can compute the @xmath137-cohomology of @xmath0 .
for example , from the above example , one can easily compute the @xmath138-cohomology of @xmath139 , and we have @xmath140 note @xmath139 is not orientable .
this is always true if @xmath141 is odd , that is , we have : @xmath0 is orientable _ iff _
@xmath142 is even .
since the box at the southeast conner of the top young diagram , @xmath143 , has the letter @xmath112 iff @xmath142 is _
even_. this implies the incidence number @xmath144=0 $ ] for the longest element @xmath145 .
the proof of theorem [ main ] relies on the description given in @xcite on the incidence graph for real flag manifolds .
in @xcite , each local system on the real flag manifold is parametrized in terms of a vector of signs @xmath146 , which encodes the structure of the local system ( see definition 4.3 of @xcite ) .
the sign @xmath147 indicates that the local system is _
constant_along @xmath148 ( restricted to a fiber which corresponds to the flag manifold of an @xmath149 ) .
similarly , the sign @xmath150 corresponds to the local system which is _ twisted _ along @xmath148 . in the case of cohomology with constant coefficients ,
we define an action of the weyl group on the signs starting with @xmath151 representing a _ constant _ local system : [ wactions ] ( * ? ? ?
* definition 2.10 ) each simple reflection @xmath152 acts on the sign @xmath153 in the vector @xmath154 as @xmath155 where @xmath156 is the cartan matrix of @xmath157 .
that is , @xmath158 and @xmath159 for @xmath160 . with the @xmath121-action on @xmath161
, we also define a number @xmath162 with @xmath163 by @xmath164 1+\hat\eta(w,\epsilon ) \qquad&\text{if}\quad s_i(w(\epsilon))\ne w ( \epsilon ) .
\end{array}\right.\ ] ] the @xmath165 is then simply the number of times that the simple reflections in a fixed reduced expression of @xmath77 have acted non - trivially ( changed a sign ) starting with @xmath166 ( see definition 4.6 and proposition 4.2 of @xcite ) .
the numbers @xmath167 have the algebraic description in terms of the hecke algebra @xmath168\otimes_{\mathbb{z}}\mathbb{z}[w]$ ] ( see section 4 of @xcite for the details ) .
we now recall the connection with the incidence graph @xmath111 .
there is an edge @xmath131 between two vertices , @xmath169 related by a simple reflection @xmath148 exactly when the @xmath148-action does not change the signs , i.e. @xmath170 starting with @xmath171 .
then defining a weighted schubert class by @xmath172 , we have the following theorem which is proven in @xcite , and also in theorem 4.9 in @xcite : with @xmath171 , the incidence graph @xmath111 is expressed by @xmath173 that is , one can replace @xmath105\ne 0 $ ] with the equation @xmath174 . for @xmath0
there are exactly two initial signs @xmath175 corresponding to local systems on @xmath176 .
let @xmath177 denote those initial signs which are defined by @xmath178 as explained above , @xmath179 represents a constant local system , while @xmath180 represents a non - constant local system .
then we have : [ lema2 ] for any @xmath40 , one can show that 1 .
@xmath181 if and only if @xmath182 , and 2 .
@xmath183 if and only if @xmath184 .
we give an outline of the proof by induction on @xmath142 using the weak bruhat graph @xmath109 ( see example [ gr25 ] ) .
we first consider the partial graph consisting of the young diagrams @xmath41 with one row , i.e. @xmath185 , or in terms of the schubert symbol @xmath186 which is @xmath187 in terms of the signs representing local systems , we compute @xmath188 with @xmath189 .
it is easy to check with a direct calculation that lemma [ lema2 ] holds for this partial graph .
one recovers the incidence graph of @xmath190 in this first step .
this graph has an alternating pattern in @xmath131 .
this edges @xmath131 can be simultaneously described in terms of agreement of @xmath114 and in terms of the @xmath121-action on signs .
lemma [ lema2 ] is then true for the edges in this partial graph .
if one starts with @xmath191 then the graph starts with @xmath131 and this corresponds to a @xmath1-shifted arrangement which is the the incidence graph of @xmath192 with local coefficients .
we can now decompose the graph @xmath109 into a disjoint union of sub - graphs .
each sub - graph consists of the young diagrams with the same first row having @xmath193 boxes for @xmath194 .
this sub - graph can be identified with @xmath195 after removing the first common top row ( in the example [ gr25 ] , those subgraphs correspond to the columns of @xmath92 ) .
we note , however , that if start with a young diagram with a _ standard _
@xmath1 arrangement , the young diagrams in the decomposition , excluding that the common top row , correspond to @xmath196 but are @xmath1-_shifted_. this is a trivial observation related to the fact that @xmath1 s are placed in an alternating pattern and we are starting from the second row which now has @xmath1 and not @xmath61 at the beginning .
we thus have a decomposition into a disjoint union of subgraphs associated to @xmath197 .
then by induction , lemma [ lema2 ] holds for the subgraphs and this allows us to complete the proof .
this lemma with the definition of @xmath167 then leads to the following proposition : [ lema1 ] we have @xmath198 and @xmath199 .
this proposition gives the proof of theorem [ main ] .
that is , we have @xmath200 by identifying @xmath201 with @xmath202 .
also , for @xmath87 , we have the incidence graph @xmath203 corresponding to a non - constant local system , i.e. @xmath204 as the simplest but important example ( used in the proof as the first step ) , we give the incidence graph of @xmath205 whose the longest element is @xmath206 .
the graph @xmath207 is given by @xmath208[8][$q$]~\overset{s_2}{\longrightarrow}~\young[2][8][$q$,$1$]~\overset{s_3}{\longrightarrow}~\young[3][8][$q$,$1$,$q$]\quad{\longrightarrow}\quad\cdots\ ] ] recall that the boxes marked by @xmath209 with odd @xmath193 has @xmath1 and others have @xmath61 .
then we recover the well - known results of the integral cohomology of @xmath210 , @xmath211 also the graph @xmath212 is given by @xmath213[8][$1$]~\overset{s_2}{\longrightarrow}~\young[2][8][$1$,$q$]~\overset{s_3}{\longrightarrow}~\young[3][8][$1$,$q$,$1$]\quad{\longrightarrow}\quad\cdots\ ] ] then we have the cohomology with twisted coefficient @xmath214 , @xmath215 as we will show in section [ poincare ] , this cohomology is related to the homology of the non - orientable case @xmath216 with @xmath217odd via the poincar - verdier duality , i.e. @xmath218
let @xmath2 be a finite field with @xmath1 elements . here
@xmath1 is a power of a prime .
then based on the @xmath1-weighted schubert variety defined in the previous section , one can find the number of @xmath2 points on @xmath0 .
we first recall that in the complex case , the @xmath2-points of @xmath219 is given by @xmath220_q!:=[k]_q[k-1]_q\cdots[1]_q$ ] and @xmath221_q$ ] is the @xmath1-analog of @xmath10 , @xmath222_q:=\frac{1-q^k}{1-q}=1+q+\cdots+q^{k-1},\ ] ] then the @xmath223-binomial coefficient @xmath224_{t^2}$ ] gives the poincar polynomial @xmath225 of @xmath226 .
the real grassmanian @xmath0 can then be described as @xmath227 , where @xmath228 and the levi factor @xmath229 of a maximal parabolic subgroup @xmath230 containing the borel subgroup @xmath231 of upper triangular matrices in @xmath232 .
we can consider @xmath227 over different fields such as the algebraically closed fields fields such as @xmath233 or the finite field @xmath234 .
this allows us to consider @xmath235 and the number of points @xmath236 .
we refer to as the @xmath2-points on @xmath0 . here
we give an explicit formula of the @xmath2-points in terms of @xmath114 describing the number of @xmath1 s in the young diagram @xmath41 .
let us first give : ( * ? ? ?
* definition 3.1)[polyq ] we define a polynomial @xmath237 by @xmath238 then from tyne incidence graph @xmath86 , one can find the explicit formula of the polynomial @xmath237 . for @xmath0
, the polynomial @xmath237 takes the following form , * if @xmath239 equals @xmath240 or @xmath241 , then @xmath242_{q^2},\ ] ] * if @xmath243 , then @xmath244_{q^2}.\ ] ] .
let @xmath41 be a young diagram satisfying the following condition : * if @xmath239 equals @xmath240 or @xmath241 , then @xmath41 consists of only the sub - diagram given by @xmath245[8][$q$,$1$,$1$,$q$].\ ] ] * if @xmath243 , @xmath41 consists of the above sub - diagram _ and _ the hook diagram , @xmath246 with @xmath247 , e.g. for @xmath248 , @xmath249 and the young diagram is given by @xmath250 , @xmath251[8][$q$,$1$,$q$,$1$,$q$,$1$,$q$].\ ] ] the number of @xmath1 s in this hook diagram is @xmath193 , i.e. it has the weight @xmath252 , and the length ( total number of boxes ) of the diagram is @xmath253 . then for those young diagrams in the graph @xmath86 , the arrows coming in or going out are just @xmath254 , not @xmath131 .
that is , @xmath114 in the @xmath1-weighted schubert variety @xmath255 changes under those nontrivial @xmath121-actions , and those @xmath77 contribute the sum .
one can also show that those are only ones which contribute the sum .
then for example , in the case @xmath256 in the case ( i ) , one can see that the polynomial @xmath237 agrees with @xmath257 in , which is the formula in ( i ) .
other cases in ( i ) can be obtained in the similar manner . for the case ( ii ) , i.e. @xmath243 , there are two types of young diagrams which contribute the sum @xmath237 .
the young diagrams having just type ( i ) give the second term @xmath258_{q^2}$ ] in @xmath237 ( @xmath259 sign is due to @xmath260 ) , and those with the hook diagram give @xmath261_{q^2}$ ] , the first term in @xmath237 .
one can also define the polynomial @xmath262 using @xmath263 by @xmath264 this polynomial will be important for the non - orientable case of @xmath0 with @xmath217odd , and we have : [ q - poly ] for @xmath265 , the polynomial @xmath262 takes the form , @xmath266_{q^2}=q^s\,p(q),\ ] ] where @xmath267 if @xmath268 , and @xmath269 if @xmath270 .
we then have the following theorem showing that @xmath237 is related to the @xmath2 points on @xmath0 : [ points ] we assume @xmath271
. then we have @xmath272 in the following subsection , we give an outline of the proof based on the action of the hecke algebra and the frobenius eigenvalues in etale cohomology ( the weil tale cohomology theory ) .
let us some explicit examples : * @xmath273 * @xmath274 * @xmath275 we recall that the number of points @xmath276 is given by a lefshetz fixed - point theorem applied to the frobenius map , @xmath277 on @xmath278 with the algebraically closed field @xmath279 .
that is , we have the formula relating @xmath276 to the frobenius eigenvalues , @xmath280 where @xmath281 is the tale cohomology with compact support of @xmath282 with the values in the @xmath283-adic number field @xmath284 .
we also have the following proposition which corresponds to proposition 6.1 in @xcite : [ powersq ] the induced action of the frobenius map on @xmath285 with the @xmath283-adic number field @xmath286 has the eigenvalues of the form @xmath287 where @xmath288 is an integer . by definition [ polyq ] of the polynomial @xmath237 , proposition [ powersq ] then implies @xmath289 then to compute @xmath236
, we simply have to make a replacement in the polynomial @xmath237 corresponding to the _ dual _ of @xmath290 .
duality has the effect of replacing @xmath291 instead of @xmath1 to take into account the action of frobenius on the dual space .
then there is the additional standard shift by @xmath292 because of the frobenius eigenvalues on stalks of the coefficient @xmath293 , the dual of a constant sheaf in the poincar duality formula .
thus we have @xmath294 .
these observations leads to the proof of theorem [ points ] : we need to consider the cases with @xmath217even ( orientable case ) and @xmath217odd ( non - orientable case ) : 1 .
assume that @xmath142 is even .
by the poincar duality , @xmath295 is the dual of @xmath296 .
as has been already discussed above , this simply corresponds to considering the polynomial @xmath294 .
+ we now use standard properties of the @xmath1 deformation of binomial coefficients .
we note that if @xmath297 then @xmath298 .
we obtain @xmath299 2 .
assume that @xmath142 is odd , say @xmath300 .
there are two cases , i.e. @xmath268 and @xmath270 . by the poincar - verdier duality ( see ix.4 and vi.3 in @xcite )
, @xmath301 is the dual of @xmath302 where @xmath303 is a twisted local system .
we then have @xmath304 with @xmath305 . here
the @xmath1-shifted polynomial @xmath262 are given in proposition [ q - poly ] , i.e. @xmath306 with @xmath267 if @xmath268 , and @xmath269 if @xmath270 .
then we have @xmath307 , and obtain @xmath308 .
for @xmath0 , the poincar polynomial , denoted by @xmath309 , is given by @xmath310 here we show that @xmath311 can be obtained from the polynomial @xmath237 defined in the previous section .
[ b ] the poincar polynomial for @xmath0 is given as follows : * if @xmath239 equals @xmath312 or @xmath313 , we have @xmath314_{t^4 } .\ ] ] * if @xmath243 , we have @xmath315_{t^4}.\ ] ] where @xmath316_{q^2}$ ] , that is , replace @xmath317 by @xmath318 and set @xmath319 .
note in particular that for the case ( i ) , we have @xmath320 these polynomials then lead to the well - known formulas of the euler characteristic for @xmath0 , that is , we have : the euler characteristic @xmath321 has the following form .
* if @xmath239 equals @xmath240 or @xmath241 , then @xmath322 * if @xmath243 , then @xmath323 when the grassmannian @xmath0 is orientable , @xmath142 is even , we have the _ poincar duality _ @xmath324 in the non - orientable case , i.e. @xmath142 is odd , we define the incidence graph with twisted coefficients @xmath214 , denoted by @xmath325 , that is , this graph is obtained by the young diagrams with @xmath1-shifted arrangement .
the graph then gives the cohomology of twisted coefficient , @xmath326 which then gives the homology of the non - orientable grassmannian @xmath0 , via the _ poincar - verdier duality _
( see e.g. section ix.4 and vi.3 in @xcite ) , @xmath327 the graph @xmath328 is given by @xmath329[8][$1 $ ] & { \longrightarrow } & \young[2][8][$1$,$q$ ] & { \longrightarrow } & \young[3][8][$1$,$ q$,$1 $ ] \\[-2.5ex]\\ & & { \downarrow } & & { \downarrow } & & { \downarrow } \\[-2ex]\\ & & \young[1,1][8][$1$,$q$ ] & { \longrightarrow } & \young[2,1][8][$1$,$q$,$q$ ] & { \longrightarrow } & \young[3,1][8][$1$,$q$,$1$,$q$ ] \\[-2ex ] \\ & & & & { \downarrow } & & { \downarrow } \\[-2ex]\\ & & & & \young[2,2][8][$1$,$q$,$q$,$1 $ ] & { \longrightarrow } & \young[3,2][8][$1$,$q$,$1$,$q$,$1 $ ] \\[-2ex]\\ & & & & & & { \downarrow } \\[-2ex]\\ & & & & & & \young[3,3][8][$1$,$q$,$1$,$q$,$1$,$q$ ] \end{matrix}\ ] ] which is obtained by exchanging @xmath330 in the @xmath1-weighted schubert varieties in the incidence graph @xmath86 . the cohomology obtained from this graph is @xmath331 h^1(\gr(2,5),\mathcal{l})=\mathbb{z}_2 , & h^5(\gr(2,5),\mathcal{l})=\mathbb{z}_2 , \\[1.0ex ] h^2(\gr(2,5),\mathcal{l})=\mathbb{z } , & h^6(\gr(2,5),\mathcal{l})=\mathbb{z}. \\[1.0ex ] h^3(\gr(2,5),\mathcal{l})=\mathbb{z}_2 , & { } \end{array}\ ] ] then the homology group @xmath332 via the poincar - verdier duality gives @xmath333 notice that the homology generators are the pontryagin classes in @xmath334 and @xmath335 .
recall ( see e.g. @xcite ) that the cohomology ring of the _ complex _ grassmannian @xmath219 is given by @xmath336}{\{c\cdot\bar{c}=1\}},\ ] ] where @xmath337 and @xmath338 with the chern classes @xmath339 . * if @xmath239 equals @xmath240 or @xmath241 , then @xmath345}{\{p\cdot \bar{p}=1\}},\ ] ] where @xmath346 and @xmath347 with the pontryagin classes @xmath348 .
* if @xmath349 , then @xmath350}{\{p\cdot \bar{p}=1 , r^2\}},\ ] ] where @xmath348 and the element @xmath351 corresponds to the schubert class @xmath99 with the hook diagram @xmath352 . | we give an explicit and simple construction of the incidence graph for the integral cohomology of _ real _ grassmann manifold @xmath0 in terms of the young diagrams filled with the letter @xmath1 in checkered pattern .
it turns out that there are two types of graphs , one for the trivial coefficients and other for the twisted coefficients , and they compute the homology groups of the orientable and non - orientable cases of @xmath0 via the poincar - verdier duality .
we also give an explicit formula of the poincar polynomial for @xmath0 and show that the poincar polynomial is also related to the number of points on @xmath0 over a finite field @xmath2 with @xmath1 being a power of prime which is also used in the young diagrams .
partially supported by nsf grant dms-1108813 . ] |
experimental information on nuclei far from stability constitutes a challenging test for the applicability of nuclear models and serves as a guidance for improvement of the models .
radioactive ion beam facility plays a vital role in gathering such information . in a recent experiment at the oak ridge hribf , padilla - rodal _ et al . _
measured @xmath6 in @xmath7ge @xcite , which provides a complete view of @xmath8 data extended to the neutron - rich region , reaching the major shell closure at @xmath9 .
it was also demonstrated @xcite that the observed @xmath5 values in @xmath7ge can be reproduced by using a shell model with the model space @xmath10 for both protons and neutrons .
hence this work showed that the structure change in the exotic mass region could be microscopically studied by means of large - scale shell model calculation , which has become available only recently .
large - scale shell model calculations have been applied to the study of @xmath0 magicity in refs .
@xcite .
the work of padilla - rodal _ et al .
_ has stimulated our interest in the long - standing problem of structure change along the ge isotopic chain , particularly in the vicinity of @xmath0 .
the experimental @xmath8 data shown in fig . 4 of ref .
@xcite suggest that @xmath11ni can be regarded as a quasi - doubly - closed - shell nucleus , but the nature of the @xmath0 subshell closure is gradually destroyed as the proton number outside the @xmath12 orbit increases .
the variation of @xmath8 along the @xmath13 axis reminds us of the old problem , much discussed in the 1970 s and 1980 s , of the structure change from the @xmath14 to @xmath15 ge isotope .
references were listed , for instance , in ref .
@xcite published in 1987 .
a notable irregularity at @xmath0 was observed in the @xmath16 and @xmath17 reaction cross sections @xcite , which is characterized by a drastic drop of the second @xmath2 excitation energy in @xmath18ge .
this is an unusually low excited @xmath2 state , showing a similar character as that of typical doubly - closed - shell nuclei . at @xmath0 ,
the drop of @xmath19 appears in coincidence with a remarkable enhancement of @xmath8 .
these experimental data suggest that the structure of @xmath18ge with @xmath0 is very different from that of @xmath11ni studied in refs .
@xcite .
the structure changes in ge isotopes were discussed in terms of pairing correlation and shape change ( or shape coexistence ) by using various collective models @xcite .
the interacting boson model @xcite reproduced the observed energy levels and @xmath5 values for @xmath20ge .
however , a satisfactory microscopic explanation has not been seen .
most of the early studies in the literature limited their discussion on the configuration change within the @xmath21 shell .
contributions from the @xmath1 orbit were very little mentioned .
these early studies are insufficient according to the new knowledge on the @xmath0 magicity obtained in refs .
@xcite , which has emphasized that nucleons in the @xmath1 orbit play a crucial role in determining the structure of the @xmath0 isotones .
thus , as a necessary requirement , a microscopic study for the structure in ge isotopes should include the @xmath1 orbit .
in the previous publications @xcite , we have demonstrated the feasibility of a shell model description for @xmath22ge ( including odd - mass isotopes ) , and for some other neighboring nuclei of the mass region . in these studies , large - scale shell model calculations ( up to dimension @xmath23 ) were performed in the model space @xmath10 , the same model space as that of ref .
the aim of the present article is twofold : to fill the gap in the ge isotopic chain that the previous shell model calculations have neglected , and to look for possible structure reasons that may explain the unsolved puzzle in the middle of the chain around @xmath0 .
the extended pairing plus quadrupole ( @xmath24 ) hamiltonian @xcite is employed in the present work .
this hamiltonian has recently been proposed and tested through several shell model applications .
it should be stressed that this isospin - invariant hamiltonian includes strong proton - neutron interactions in both @xmath25 and @xmath26 channel .
the single - particle energies in our model are @xmath27=0.00 , @xmath28=0.77 , @xmath29=1.11 , and @xmath30=2.50 ( in mev ) . in the present work that deals with a long isotopic chain , we use the @xmath31-dependent interaction strengths : @xmath32 mev for the monopole pairing force , @xmath33 mev for the quadrupole - quadrupole ( @xmath34 ) force , and @xmath35 mev for the octupole - octupole ( @xmath36 ) force .
these are the same parameters as those in ref .
the hamiltonian contains also five monopole correction terms ( see ref .
the effective charges are @xmath37 and @xmath38 . ,
@xmath39 , and @xmath19 for the ge isotopes from @xmath40 to 50 . calculated @xmath8 with two sets of effective charge are shown for comparison.,width=264,height=264 ] our results are presented in fig .
[ fig1 ] , where calculated @xmath8 , @xmath39 , and @xmath19 for the ge isotopes from @xmath41ge ( @xmath40 ) to @xmath42ge ( @xmath9 ) are compared with data .
the calculation reasonably reproduces @xmath8 and @xmath19 at both ends of the isotopic chain with @xmath13 = 34 , 36 and @xmath13 = 48 , 50 , but fails to describe those in between .
especially , the calculation could not reproduce the increasing trend of @xmath8 over @xmath0 and the drop of @xmath19 around this neutron number .
we can get an improved agreement with the @xmath8 data for @xmath43ge and @xmath44ge ( see fig .
[ fig1 ] ) if we enlarge the effective charges to @xmath45 and @xmath46 ( @xmath46 is taken from ref .
@xcite and @xmath45 is fixed so as to reproduce @xmath8 in @xmath42ge ) .
however , the use of larger effective charges destroys the good agreement for @xmath41ge and @xmath11ge already obtained in ref .
moreover , the rise in effective charges does not seem to have any positive effect on enhancing @xmath8 at @xmath0 .
it is thus clear that the problem must stem from deeper structure reasons , which can not be resolved simply by a global fit through effective charges .
the above results indicate an inadequacy either in the interaction employed in the calculation , or in the model space for @xmath47ge in question . at present
, there is no available effective interaction better than the extended @xmath24 interaction for ge isotopes in the model space @xmath10 .
we thus proceed our study with the extended @xmath24 interaction and look for any possible reasons for this discrepancy .
the enhancement of @xmath8 in the @xmath0 isotones from @xmath11ni to @xmath44sr has been qualitatively explained by langanke _
_ @xcite using the shell model monte carlo ( smmc ) approach , where a @xmath24 hamiltonian ( with no monopole terms ) is used but their model space is larger , which includes also the @xmath4 shell .
it must be noted that the smmc calculation gives only total @xmath5 strength to excited @xmath48 states and does not give the energy of the @xmath49 state .
the smmc calculation predicted a large neutron occupation number in the @xmath1 orbit ( @xmath50 ) , in contrast to our small value ( @xmath51 ) obtained for @xmath18ge .
this gives us a hint that inclusion of the @xmath52 orbits may significantly enhance the role of the @xmath1 orbit through the pairing and quadrupole correlations when the fermi level moves toward the @xmath4 shell .
the deficiency in our calculation may be due to the insufficient model space for the nuclei where neutrons start to occupy the @xmath53 and @xmath1 orbits . the @xmath1 occupancy of neutrons is investigated in @xmath11ni with the ordinary shell model @xcite , in which the @xmath1 orbit is excluded from the model space for protons .
the value @xmath54 obtained in ref .
@xcite is considerably different from the smmc result @xmath55 @xcite .
in another recent work @xcite , we have investigated @xmath11ni in the neutron model space @xmath10 assuming a @xmath56ni core .
the shell model for neutrons also gives the value @xmath54 , similar to that in ref .
these shell model calculations show that the ground state of this nucleus remains a transitional character to a superfluid phase .
it is thus interesting to investigate whether the delicate structure changes from the @xmath14 to @xmath15 ge isotopes are caused by strong pairing correlations .
we continue our study within the @xmath57 space as this model space is at present the largest possible for a full shell model diagonalization . within this truncated space
, contributions from the @xmath58 orbits must be expressed in terms of effective interaction .
we have made testing calculations with different monopole corrections , which effectively change the neutron @xmath1 occupation . however , moderate modifications for the monopole corrections can not produce a significant structure change near @xmath59 . on the other hand
, it is expected that the effects of the @xmath58 orbits tend to strengthen the pairing and quadrupole correlations .
we therefore consider to enlarge the quadrupole matrix element @xmath60 to mimic any effect from the missing @xmath58 orbits .
note that the quadrupole matrix elements between the @xmath61 and @xmath62 orbits are nonzero and those between the @xmath21 and @xmath63 shells are zero ( the @xmath64 orbit nearest to @xmath61 has the largest quadrupole matrix element with @xmath61 ) .
based on the above considerations , we write the effective quadrupole matrix element for the @xmath1 orbit by multiplying a factor @xmath65 @xmath66 physically , an @xmath67 factor strengthens the @xmath34 force with respect to the @xmath1 orbit , which is consistent with the expectation that the ge isotopes become deformed at the beginning of the @xmath1 orbit and that a strong @xmath34 force contributes to lowering of the @xmath1 orbit .
a stronger @xmath60 increases also the effective charges selectively for the @xmath1 nucleons . in our shell model treatment , we transform the @xmath34 force into multipole - pairing - type interactions in the particle - particle channel . thus effectively , a stronger @xmath34 force with respect to the @xmath1 orbit enhances also the pairing correlations , which allows more nucleons to jump to the @xmath1 orbit . , @xmath39 , and @xmath19 in @xmath3ge as @xmath30 approaches @xmath29.,title="fig:",width=264,height=241 ] , @xmath39 , and @xmath19 in @xmath3ge as @xmath30 approaches @xmath29.,title="fig:",width=264,height=241 ] , @xmath39 , and @xmath19 in @xmath3ge as @xmath30 approaches @xmath29.,title="fig:",width=264,height=241 ] the effects from the @xmath58 orbits are considered to be isotope dependent and should become larger as the fermi level moves up .
thus in our trial calculations , we assume an empirical expression for @xmath65 , @xmath68 . with these considerations ,
the calculations indeed achieve a great improvement in excitation energies @xmath39 and @xmath19 , especially for @xmath69ge .
the calculated @xmath8 values remain good at the beginning of the chain ( @xmath13 = 34 and 36 , because the configurations are almost exhausted by the @xmath21 shell ) , and are enhanced at the end with @xmath13 = 46 @xmath70 50 .
these improvements support the use of effective quadrupole matrix element @xmath71 .
however , @xmath8 for @xmath72ge are again not enhanced .
this indicates that the microscopic content in the wave - functions has not been improved . as a matter of fact ,
the neutron and proton occupation numbers @xmath73 and @xmath74 obtained for @xmath18ge ( @xmath0 ) by using @xmath71 can not bear a comparison with the large values @xmath75 and @xmath76 reported in ref .
the large @xmath8 value in @xmath18ge corresponds to a large occupation of the @xmath1 orbit , which must be due to the pairing correlations in addition to the @xmath34 force .
in fact , the unusually low energy of the second @xmath2 state in @xmath18ge implies a delicate balance between the pairing correlations and the energy gap @xmath77 .
it has been pointed out by schiffer _
_ @xcite and utsuno _ et al . _
@xcite that the effective energy " of the intruder states changes rapidly around a closed subshell in neutron - rich nuclei , and these changes were discussed in terms of the spin - orbit interaction or the spin - isospin dependence of nucleon - nucleon interaction . in our case , the change of the effective energy gap @xmath77 could be attributed to the pairing correlations in the @xmath4 shell .
a good effective hamiltonian in a full shell model should be able to describe the balance microscopically . in the present work with the truncated model space
, we try to understand the problem by simulating the underlying physics .
namely , we adjust the energy gap @xmath78 and study any resultant change in @xmath8 and @xmath19 .
we study variations of @xmath8 and @xmath19 in @xmath69ge as a function of single - particle separation energy between the @xmath1 and @xmath53 orbits . in the calculation
, we use the enhancement factor @xmath79 for @xmath80ge , and a slightly larger one @xmath81 for @xmath82ge .
interesting results are obtained as shown in fig .
the circled areas represent the best choices in parameter that can describe the observed @xmath8 , and at the same time , give correct sequences of @xmath83 and @xmath49 in @xmath84ge .
figure [ fig2 ] indicates clearly the necessity of a rapid lowering of the single - particle energy @xmath30 from the original value 2.5 mev in @xmath85ge , to @xmath86 2.0 mev in @xmath43ge , to @xmath86 1.75 mev in @xmath18ge , and to @xmath86 1.5 mev in @xmath82ge .
thus the calculation demonstrates that the observed enhancement in @xmath8 and variation in @xmath19 correlate closely with the single - particle energy separation .
data can not be understood unless the effective energy gap @xmath78 near @xmath0 is set to be small . on the other hand , the calculated @xmath8 in @xmath18ge and @xmath82ge
can not reach the observed large values unless the enlarged matrix element @xmath71 is used .
in particular , the correct sequence of the @xmath49 and @xmath83 states in @xmath82ge can not be obtained without narrowing the effective gap @xmath77 to a sensitive region in the parameter space , which implies an extremely delicate balance between the pairing correlations and the energy separation @xmath77 . , @xmath39 , and @xmath19 in @xmath87ge as @xmath30 approaches @xmath29.,width=264,height=241 ] although a meaningful parameter choice can be found for the isotopes up to @xmath82ge , it is difficult to go beyond them and find good parameters that can simultaneously reproduce @xmath8 and @xmath19 for @xmath87ge , as illustrated in fig .
this indicates limitations in our renormalization approach within the truncated model space missing @xmath58 orbits above @xmath1 . as we shall see in sect .
iv , a small energy separation @xmath88 causes nucleons to jump to the @xmath1 orbit easily , and with more nucleons in the @xmath1 orbit the use of @xmath71 enhances @xmath5 values in @xmath84ge
. however , the situation seems to be different between @xmath84ge and @xmath87ge . in @xmath84ge ,
the @xmath1 orbit is occupied by at most four neutrons . in @xmath87ge , however , about half of the neutron @xmath1 orbit is occupied . with an additional neutron pair in the @xmath1 orbit
, @xmath71 enhances @xmath5 exaggeratedly .
in such a case , our renormalization approach fails to work properly .
it would be necessary for the description of the @xmath89 isotopes to explicitly include upper orbits such as @xmath90 , as suggested in ref .
below , we continue our discussion but focus the attention on the structure change in @xmath43ge , @xmath18ge , and @xmath82ge only . , @xmath39 , and @xmath19 with experimental data.,width=264,height=264 ] with the above adjustment in @xmath77 and the use of effective quadrupole matrix element @xmath71 , we are able to reproduce correctly all the observed variations in @xmath8 , @xmath19 , and @xmath39 for @xmath69ge , as summarized in fig .
[ fig4 ] . here
, the energy separation @xmath78 is not required to change for the @xmath41ge and @xmath11ge calculations if @xmath71 with @xmath91 is used .
figure [ fig4 ] shows that the variation trend of @xmath92 energy is also correctly reproduced .
of course , our simulated calculation does not provide a final answer to the problem , but is quite suggestive .
it should be stressed that the set of circles in fig .
[ fig2 ] is the sole selection in parameter that can describe the experimental data .
the important information extracted from the above discussions is that in order to describe all the observed quantities in ge isotopes , it is necessary to have enough @xmath1 contributions in the wave - functions . next , let us study how the @xmath1 proton and neutron occupations correlate with the observations and how they vary in @xmath3ge .
we calculate occupation numbers @xmath93 for protons and neutrons as functions of the energy separation @xmath78 , and focus our discussion on the results in the circled area that have reproduced data .
figure [ fig5 ] shows that in @xmath43ge , a considerable number of neutrons occupy the @xmath1 orbit ( @xmath94 ) while protons are scarcely found in this orbit . for the ground state of @xmath18ge ,
the occupation numbers become @xmath95 and @xmath96 .
these occupation numbers are consistent with those of ref .
@xcite , though our value @xmath97 is larger .
the result indicates a superfluid state for neutrons in @xmath18ge , in contrast to the narrow subshell closure in @xmath11ni @xcite .
it is notable that in the circled area for @xmath18ge , the coupling between the @xmath98 and @xmath49 states becomes significant and a considerable number of protons occupy the @xmath1 orbit , which shows a qualitative difference from the structure in @xmath43ge .
the structure in @xmath82ge again differs qualitatively from @xmath18ge .
as @xmath30 approaches @xmath29 , the proton occupation numbers @xmath99 for @xmath98 and @xmath49 are reversed . for @xmath100 mev
with which the calculated @xmath8 are comparable with data , more than four neutrons occupy the @xmath1 orbit and protons occupy this orbit also considerably , in sharp contrast to the situation in @xmath43ge . and @xmath101 in @xmath3ge as @xmath30 approaches @xmath29.,title="fig:",width=160,height=188 ] and @xmath101 in @xmath3ge as @xmath30 approaches @xmath29.,title="fig:",width=160,height=188 ] and @xmath101 in @xmath3ge as @xmath30 approaches @xmath29.,title="fig:",width=160,height=188 ] thus the structure differences appearing in our calculation for the two lowest @xmath2 states , @xmath98 and @xmath49 , may offer a new explanation for the old problem observed in @xmath16 and @xmath17 reactions .
our simulated calculation predicts that the proton structure in the @xmath1 orbit is very different among the states in @xmath3ge .
the proton @xmath1 occupation changes from a negligible amount in @xmath43ge to a considerable amount in @xmath102ge , and in addition , with a different distribution in the @xmath98 and @xmath49 states .
obviously , these results can not be obtained without the presence of strong proton - neutron correlations in the model , which are important for nuclei where protons and neutrons occupy the same shell .
the energy difference @xmath103 mev used in our calculation , which is adopted so as to well describe @xmath85ge @xcite and also consistent with the value used in ref .
@xcite , is rather small .
still , for the isotopes up to @xmath43ge , the proton occupation number @xmath99 remains small , in contrast to the strong configuration mixing within the @xmath21 shell including the @xmath53 orbit . in this sense , for protons when @xmath104 , there exists a gap between the @xmath105 shell and @xmath1 .
however , as clearly seen from fig .
[ fig5 ] , the proton occupation number @xmath99 , together with the neutron occupation number @xmath101 , increases suddenly in @xmath18ge and @xmath82ge .
resonance " effect of @xmath99 and @xmath106 can be best attributed to strong proton - neutron correlations .
similar effects are obtained in the smmc calculations for the @xmath0 isotones @xcite , where a larger energy difference @xmath107 ( or 1.97 ) mev is used but upper @xmath58 orbits above @xmath1 are included in the model space .
a considerably large number of @xmath99 starts from @xmath18ge ( where @xmath108 is only equal to 32 and @xmath13 takes the magic number 40 ) .
we may thus conclude that the proton @xmath1 occupancy must be caused by the proton - neutron correlations in the @xmath63 shell . our treatment with small @xmath78 and large @xmath60
is considered to mimic the effects , since the mechanism is not properly included in the model .
it should be noted in this regard that our isospin - invariant hamiltonian treats dynamically the strong proton - neutron interactions .
in conclusion , inspired by the recent experimental advances in @xmath5 measurement for neutron - rich ge isotopes , we have carried out a systematical shell model calculation for @xmath109ge in the model space @xmath10 .
the calculations have shown that the strong enhancement of @xmath8 and the unusually low excitation of the second @xmath2 state near @xmath0 can be explained only with sufficient occupation of protons and neutrons in the @xmath1 orbit .
the simulated calculations that mimic such effects have suggested a possible understanding of the structure change from @xmath43ge to @xmath82ge in terms of rapid increase in the number of @xmath1 protons and neutrons .
this could be the source for the long - standing problem that has not yet been understood .
the isotopic dependence of the @xmath1 proton occupation must have an origin of strong proton - neutron interactions . in principle , it is a very difficult question whether all these effects can be expressed in an effective interaction .
the calculations in ref .
@xcite pointed out that the @xmath90 orbit above the @xmath1 makes a considerable contribution when @xmath110 in cr and fe isotopes apart from the proton magic number @xmath111 .
the smmc calculations in the @xmath112 space @xcite also suggested an important role of the @xmath90 orbit in the enhancement of @xmath1 occupancies in those nuclei that have sufficient number of protons in the upper @xmath113 shell . according to the extended @xmath24 model
, we can suggest that the pairing and quadrupole - quadrupole correlations in the major shell @xmath4 make it easier for protons and neutrons to be excited from the @xmath113 shell to the @xmath1 orbit . the present investigation with enlarged @xmath60 and decreased energy gap @xmath78 is along the line of this thought .
the problem may be thoroughly explained if shell model calculations are performed in a two - major - shell - space including both @xmath113 and @xmath4 . | the long - standing problem of dramatic structure change near @xmath0 in ge isotopes is investigated by means of large - scale shell model calculation .
the analysis of simulated calculations suggests a possible understanding of the problem in terms of rapid increase in the @xmath1 proton and neutron occupation .
the observed variation in excitation of the second @xmath2 state in @xmath3ge appears to correlate closely with the @xmath1 occupations induced by strong proton - neutron interactions .
the enhancement of the @xmath1 occupancies is probably due to correlations in the @xmath4 shell .
, , , @xmath3ge , structure change , @xmath5 , second @xmath2 energy , shell model 27.50.+e , 23.20.lv , 21.10.re , 21.60.cs |
the discovery of very high redshift and primeval galaxies has been a ` holy - grail ' for cosmologists for many years @xcite . amongst numerous approaches to finding forming galaxies and understanding their subsequent evolution , faint imaging surveys , for both number counts @xcite and to find individual interesting objects selected by various colour criteria ( e.g. multicolour searches for high redshift quasars )
are well established techniques @xcite . among the more interesting results to emerge from these studies
has been the discovery , at blue wavelengths , of many more galaxies than standard ` no evolution ' models predict the so called ` faint blue excess ' @xcite . however , various spectroscopic ( glazebrook et al . , lilly et al . ) and hst - based morphological studies ( driver et al . , glazebrook et al . ) have shown this excess population to reside at only intermediate redshifts ( @xmath5 ) , and to consist of dwarf late - type spiral / irregular galaxies which undergo substantial luminosity and colour evolution between @xmath6 and the present time .
on the other hand , these same studies have shown the redder early - type ( e / s0abc ) population to have undergone little if any evolution in this period .
the formation epoch of these galaxies , therefore , must lie at considerably greater redshift , even if the nature of the faint blue excess has now been determined .
in this context , the director of the hubble space telescope ( hst ) assigned a substantial allocation of discretionary time ( @xmath7 150 orbits ) to a project aimed at obtaining very deep wide - field & planetary camera 2 ( wfpc-2 ) images of a single ` random ' field ( position 12h 36 m
49.4s + 62d 12 58 , j2000 @xcite ; hereafter referred to as the ` hubble deep field ' ( hdf ) ) .
four filters were chosen for these observations , f300w , f450w , f606w and f814w .
the data , which were taken over the 13-day period 1995 december 1830 , were made public immediately @xcite and we present here the results of a programme to search for high redshift counterparts of normal galaxies using the presence of the lyman break at 912 in the f300w filter as an indicator of redshift . in the next section
we describe our selection technique . in section 3
we present our results , while in section 4 we discuss the properties of the objects we find .
we draw our conclusions in section 5 .
the hdf extends to magnitude limits ( on the stmag system - 21.10 where @xmath8 is the flux in ergscm@xmath2s@xmath2@xmath2 received in the bandpass ] ) of 26.7 , 28.8 , 30.1 , and 30.3 in the f300w , f450w , f606w , and f814w filters , respectively , for a 5 @xmath10 detection in 16 connected 0.04 pixels @xcite . for the present study
we wish to detect and then examine the morphological properties of high redshift galaxies .
we identify high redshift candidates by looking for a substantial magnitude difference between the f300w and f450w filters . for galaxies with 2.6 @[email protected] , the f300w passband will correspond to rest - frame wavelengths shorter than the lyman cutoff at 912 and the f450w filter to longer wavelengths .
these high redshift galaxies will thus be undetectable at f300w or will have anomalously red f300w - f450w colours .
objects at still higher redshift will additionally have f450w suppressed .
a similar method has been used successfully by steidel et al to look for companions to high redshift quasars and damped absorption - line systems , and by guhathakurta et al to set a limit on the redshift of the faint - blue - galaxy population .
additionally , the spectral energy distribution of objects at high redshift is likely to be fairly flat redward of the lyman cutoff @xcite , an assumption that is borne out by recent keck spectroscopy of z@xmath7 3 galaxies @xcite .
our selection technique is thus to look for all those objects with f300w - f450w @xmath11 2 , and remaining colours ( f450w - f606w and f606w - f814w ) between -0.6 and 0.6 .
consideration of the colour - colour diagram in @xcite and the detailed photometric modeling of fukugita et al demonstrate that it is very unlikely for lower redshift galaxies of any type to have such colours .
objects not detected in the f300w filter are treated as if they have a magnitude of 27.7 , corresponding to the 2@xmath10 upper limit , for the purposes of this selection .
we are thus constrained to objects with an f450w magnitude of 25.7 or brighter for our candidates , which usefully guarantees that they will be bright enough for us to study their morphology .
examining the flux zero points for these filters shows that for an f300w - f450w colour @xmath11 2 , the flux ratio between the f450w and f300w bands is @xmath11 6 .
this is higher than could be caused by the 4000 break or by absorption in the ly@xmath12 forest where the continuum suppression factor at these redshifts is only about 50% @xcite .
this selection was based on object catalogs derived from the combined ` drizzled ' hdf images @xcite using the automated photometry program sextractor written by e. bertin .
a detailed description of the techniques this program employs in the detection and photometry of objects is given by @xcite ; its specific application to hst wfpc-2 images is discussed by @xcite . in using it here , sextractor was first run on a combined f606w+f814w image to provide a deep ` master ' list of objects .
it was then run on each of the individual f300w , f450w , f606w , and f814w images , the resulting catalogs being matched and merged with the master version . in all cases , a detection threshold of 1.3@xmath10 ( where @xmath10 is the standard deviation of the background noise distribution ) and a minimum area after convolution with a 0.3arcsec top
hat filter of 0.05arcsec@xmath13 ( 30 connected pixels ) were adopted .
this resulted in an average of 200 , 420 , 510 , and 420 objects being detected on each wfc chip in f300w , f450w , f606w , and f814w , respectively .
sextractor computes several different types of magnitude for each detected object but we worked solely with the corrected isophotal ( mag_isocor ) values as they appeared to be the most stable at the faintest limits .
this particular magnitude provides an estimate of the total flux from an object by taking the light measured within the threshold isophote and then , by assuming a gaussian profile , making a small ( @xmath14% ) correction for the light lost outside this isophote . as a check ,
these magnitudes were compared with the ` total ' magnitudes available in catalogs generated at the space telescope science institute using the focas package @xcite . with the exception of the f300w band ( see below ) , excellent agreement was found between the two , both in zero point and scale .
the scatter observed between the two sets of magnitudes was also consistent with the random photometric errors expected from photon statistics .
it was noticed that sextractor and focas both had difficulties in reliably detecting and measuring objects on the f300w images .
a visual inspection revealed that many of the detected objects were spurious and yet a considerable number of real objects were missed .
this is undoubtedly due to the problems experienced in flat - fielding the f300w data and removing scattered light . a conspicuous checker - board pattern peculiar to the drizzling scheme was also present around the edges of the frames .
these problems were dealt with by using the iraf apphot routine to measure aperture magnitudes at each of the positions in our master catalogue , setting the aperture diameter equal to the major axis length computed for each image on the combined f606w+f814w frame .
it was these f300w aperture magnitudes that were used in the selection process .
using the above colour and magnitude criteria , we searched our sextractor catalogue for candidate high redshift objects .
we found 9 candidates at [email protected] which satisfied our colour criteria .
visual inspection of all these candidates led us to reject one of these because it lay near the edge of one of the ccd chips and was severely contaminated by spurious flux at the edge of the dithering pattern .
basic data on the objects is presented in table 1 .
images of the objects are shown in figure 1a , using the combined f606w+f814w frame .
corresponding f300w images are shown in figure 1b , demonstrating the lack of any detections , and thus the presumed presence of the ly break in this band .
we also used a similar search technique to find candidate objects at still higher redshift by looking for a flat f606w -f814w colour combined with f450w - f606w @xmath15 and little or no flux in the f300w band .
this would select for objects with 3.9@xmath0z@xmath05 , though with additional complications from the ly@xmath12 forest absorption which becomes stronger .
no such objects were found .
[ images ] [ mags ]
we found 8 objects on the 3 wfc chips likely to be in the range 2.6@[email protected] .
we now calculate the number - density , absolute magnitude and physical size of these objects ; in doing so , we assume them all to be at the mid - point of this redshift range ( ie .
@xmath16 ) and adopt an h@xmath1=100 , q@xmath1=0.5 ( i.e. einstein - de - sitter ) cosmology except where stated otherwise . a total of 8 objects within the 4.7 sq .
arcmin area of the hdf gives a surface density for these objects of 1.7 arcmin@xmath17 . at matching magnitudes we find a similar surface density of 0.4 arcmin@xmath17 to that of the r@xmath7 25 3@xmath0z@xmath18 galaxies discussed by steidel et al @xcite .
for the assumed redshift and cosmology , the number density of high redshift galaxies in the hdf corresponds to a comoving number density of 0.005 h@xmath19mpc@xmath20 ( where h@xmath21 h@xmath1/100 ) ; in comparison , nearby bright ( @xmath22 ) galaxies have a number density of 0.015 mpc@xmath20 @xcite .
the similarity of these two numbers suggests that we may finally be seeing the progenitors of nearby bright galaxies .
the corresponding number for a q@xmath1=0.05 universe is 0.0009 h@xmath19mpc@xmath20 .
absolute ` blue ' magnitudes were determined from the apparent f450w magnitudes using the k - correction calculated by cowie et al . for a star - forming ( ` spiral ' )
galaxy at this redshift .
this is based on local galaxy seds from coleman , wu & weedman ( 1980 ) in the optical , and mobasher , ellis & sharples in the infrared .
we also apply the + 0.4 magnitude correction between the f450w and standard b band appropriate for a flat spectrum object using the information provided by holtzman et al .
these absolute magnitudes must be treated with great caution as neither the applicability of this particular k - correction to our sample of high - redshift candidates nor the required knowledge of the far - uv spectral energy distribution of different present - day types are well determined .
indeed the k correction is similarly large and uncertain for all the filters discussed here since the objects lie at such a high redshift .
this notwithstanding , it is of much interest that the values found are at the upper end of the luminosity function of local galaxies with our values being about 1.4 magnitudes brighter if a value of q@xmath1=0.05 is adopted . ]
( see figure 2 ) . in figure 3
we show the sizes of the major axis of the objects measured at an isophote 3 magnitudes fainter than the peak surface brightness .
in contrast to the luminosities , these appear to be significantly lower ( median diameter @xmath74kpc ) than those of their present - day counterparts .
a similarly small - sized population of sources was noted by dressler et al . in a galaxy cluster found serendipitously on a wfpc-2 image and thought to be associated with a qso at z=2.2 .
[ tab1 ] [ cols="^,^,^,^,^,^,^,^,^",options="header " , ] all magnitudes are on the st magnitude system .
no galaxies were detected in the f300w filter , giving a 2 @xmath10 magnitude limit @xmath2327.7 .
names follow the pattern chip no .
no . so that the first object detected on chip 2 is 2p1 .
absolute magnitudes are calculated assuming a k correction appropriate for spiral galaxies and the appropriate magnitude shift from f450w to b given in @xcite .
values are given for h@xmath1 = 100 km@xmath2mpc@xmath24 and q@xmath1=0.5 .
for q@xmath1 = 0.05 , absolute magnitudes get brighter by 1.4 mags , and sizes increase by a factor of 2 , while for h@xmath1=50 , corresponding figures are 1.5 mags brighter and 2 times bigger .
the object selection is sufficiently bright in the longer wavelength filters that the images are deep enough to allow examination of the morphologies of the selected galaxies .
inspection of the images in figure 1a immediately demonstrates that all of these objects are disturbed systems , with multiple nuclei , jets , plumes , and asymmetric flux distributions .
indeed , the structures observed are similar to interacting / merging galaxies observed locally e.g. clements et al . .
it has been known for some time that interactions and mergers between galaxies can trigger bursts of star formation @xcite .
these morphological signatures are also common at intermediate redshifts where hst and other high - resolution observations have revealed objects of this type amongst the blue ` butcher - oemler ' population in clusters ( couch et al . , dressler et al . ) , the faint blue field population @xcite , and mjy radio sources @xcite .
theoretical models of galaxy formation and evolution ( e.g. kauffman ) currently favour the idea that the galaxies we see today were formed as a result of the merging / accretion of a number of subunits .
if this is the case then each such accretion event is likely to be accompanied by the vigorous star formation we see in local mergers .
the disturbed , multiple - component nature of the hdf high redshift galaxies presented here suggests that this is a wide - spread occurrence at these very early epochs .
the close match between the number densities of these objects and local galaxies might indicate that we are at last seeing the epoch where most of the galaxies we see today were being assembled from such subunits .
alternatively , if there has been considerable number - density evolution in galaxies at lower redshifts ( 0.5 1 ) , as might be indicated by the faint - blue - galaxy population , then we might just be seeing occasional starbursts in a population of more numerous small galaxies .
the effect of viewing these objects in their restframe uv ( @xmath7 1670 ) is also another factor that must be considered .
we might just be seeing the hottest , brightest regions of active star formation in a larger , lower surface brightness structure that we can not see . observations of the hdf at ir wavelengths , particularly those conducted with the new nicmos camera soon to be installed on hst , will be invaluable in addressing the problem of the observed restframe wavelength . for example , the centre of the @xmath25 window corresponds , at these redshifts , to @xmath7 5100 , conveniently in the restframe optical emission .
observations at these wavelengths from the ground will also be useful in better constraining the spectral nature of these objects ; we shall be conducting these in the forthcoming months .
we have presented results on a sample of objects very likely to lie at 2.6@[email protected] .
we find that these objects have similar co - moving number density to nearby bright galaxies .
however they are significantly more luminous and smaller than nearby galaxies .
almost all of our high redshift objects show signs of a disturbed and multi component morphology , which we interpret as being due to merging processes taking place during the assembly of a galaxy .
+ + * acknowledgments * + + we congratulate the team that devised and conducted the hdf observations on a job well done , and are especially grateful to the director of the hst , bob williams , for organizing such an excellent programme .
we would also like to thank the eso / ecf hdf group for many useful discussions , and especially bob fosbury , richard hook , rene mendez and hans - martin adorf for their help .
we also thank the anonymous referee for many helpful comments .
dlc is funded by an eso fellowship and wjc by an eso visitors position .
williams , r.e . , adorf , h .-
, blacker , b. , dickinson , m. , dixon , w.v . ,
ferguson , h.c .
, fruchter , a.s . , giavalisco , m. , gilliland , r. , heyer , i. , hook , r.n . ,
lucas , r.a .
, mcelroy , d.b . , petro , l. , postman , m. : 1996 , astron . j. , ( in preparation ) . | we present the results of colour - selection of candidate high redshift ( 2.6 @xmath0 z @xmath0 3.9 ) galaxies within the hubble deep field based on the ly - break at 912 .
we find 8 such objects in the region , giving a comoving number density comparable to that of nearby bright galaxies ( for a flat q@xmath1=0.5 , h@xmath1=100 kms@xmath2mpc@xmath2 universe ) .
we provide basic data on the properties of these objects , and show that despite their absolute magnitude being significantly brighter than l@xmath3 ( typically m@xmath4 = -22 ) , they are generally smaller than nearby galaxies . furthermore , visual inspection of their images shows that they are all highly disturbed systems , with multiple nuclei , tails and plumes , suggesting that they are undergoing merging processes similar to most nearby starburst galaxies .
theoretical models suggest that galaxies form by accumulation of numerous subcomponents , and we suggest that we are seeing this process underway in these objects .
it is thus possible that the epoch of galaxy formation might have been discovered .
galaxies : formation - galaxies : primeval - galaxies : interactions - galaxies : starburst |
effects of physical dimensionality on crystallization are among the most important problems in condensed matter physics . as was established by landau and pierls @xcite , the long - range crystalline order in 2d is universally destroyed by thermal fluctuations .
the problem has been revisited in 1970s by kosterlitz and thouless @xcite-@xcite , who have shown that crystals do exist in 2d , within a new topological definition .
nevertheless , their melting temperature is believed to be always lower than in 3d systems ( assuming the same interparticle potential ) .
one of the manifestations of this effect is the phenomenon of surface melting : a microscopic liquid layer normally appear at the interface of a crystalline solid well below its bulk melting temperature . in this paper , we describe a remarkable example in which the thermal fluctuation _ stabilize _ a 2d crystalline solid , embedded in 3d physical space , with respect to transition to an alternative 3d structure .
the model system discussed below has been introduced in the recent work by one of us @xcite , in order to describe dna assisted self
assembly of colloids .
its essential ingredients are cohesive interparticle interactions and medium range soft core repulsion . the binary system of same size spheres ( a and b ) discussed in ref .
@xcite , combines the repulsive potential @xmath0 acting between same type particles , with a
b attraction .
as was shown at that work , both interactions may be induced by properly designed dna .
it was found that this colloid
dna mixture may exhibit an unusually diverse phase diagram as a function of two control parameter : the relative strength of attraction and repulsion , and aspect ratio @xmath1 ( @xmath2 is the range of repulsive potential @xmath3 , and @xmath4 is the particle diameter ) . among various self
assembled phases expected for that system , it was especially striking to find quasi-2d membrane with the in plane square order ( sq ) .
in other words , according to our calculations , this 3d system may prefer to self
assemble into a lower dimensional structure .
we will refer to this phenomenon as spontaneous self compactification .
of course , there are other known examples of self compactified structures in condensed matter , such as lipid membranes @xcite .
however , our case is quite unique because it is based on _
isotropic pair potentials _ ( in contrast to anisotropic interactions between lipids , or covalent bonding of carbon atoms in graphite ) .
note also that at the found sq phase is not a stacking of weakly coupled layers ( lamella like ) , but rather an isolated membrane like structure . in our early calculations , we have only accounted for the interplay of repulsive and attractive energies , while the thermal fluctuations were totally ignored .
even though this approximation is applicable when the characteristic energies considerably exceed @xmath5 , the entropic effects are expected to be significant in any realistic case . given that fluctuations are known to strongly affect 2d crystals
, one might wonder whether the phenomenon of self compactification will still be present if the fluctuations are introduced .
below we present the detailed study of this question .
before going into a specific example , we describe our model and its generic features .
we start with particles packed into ideal crystalline lattices , whose fluctuation free energies are to be compared .
the interparticle potentials will be replaced with linear springs , whose spring constants correspond to the second derivatives of the corresponding potentials .
the first derivatives of the potentials will give rise to a pre
existing stresses in those springs .
we consider only very short range interactions , so that any connections beyond second nearest neighbors will be neglected .
let @xmath6 and @xmath7 be spring constants for the first and second nearest neighbor bonds , respectively .
repulsion between the second nearest neighbors induces tension @xmath8 in @xmath7springs , which should be balanced by an appropriate compressive force @xmath9 @xmath10 in @xmath6springs .
if the interaction range @xmath2 is much shorter than nearest neighbors distance @xmath4 , one expects @xmath11 , and @xmath12 .
this gives rise to a hierarchy of elastic constants in this system : @xmath13 @xmath14 .
below , we will use harmonic analysis to diagonalize the phonon hamiltonian and find the fluctational contributions to free energies of the competing phases . as an example
, we compare 2d square lattice ( sq ) ( embedded in 3d physical space ) to an alternative three dimensional phase with a very similar local structure .
figure [ sq ] shows this 3d counterpart of sq ( referred below as `` dual phase '' ) , which also has four nearest neighbors lying in one plane around each particle .
the difference from sq is that now there are eight , rather than four second nearest neighbors for each site . because of this difference in the number of repulsive bonds
, sq phase is generally preferred energetically over its 3d dual , at the zero temperature limit .
below , we will see whether the free energy balance between the two phases may be reversed by the thermal fluctuations .
before we present the exact results for sq , and its 3d - dual phase , we discuss certain general features of this problem . given the hierarchy of the spring constants , one can distinguish between several kinds of phonon modes .
namely , there are _ hard _ modes , which involve deformations of strong @xmath15springs , and _ soft _ modes which only depend on spring constant @xmath7 .
there are @xmath16 degrees of freedom per particle , and @xmath17 of them correspond to hard modes , both for 2d lattice and its 3d dual phase . in 3d structure ,
the third mode is the soft one ( @xmath7-mode ) .
however , if there were no stresses @xmath18 and @xmath8 , within harmonic approximation there would be no restoring force for the out - of - plane fluctuations in the 2d structure .
this means that in 2d lattice the effective spring constant for the transverse modes should be of the order of @xmath19 ( since @xmath20 , we may justifiably call these modes of 2d lattice @xmath21 ) .
because of the replacement of the soft modes with the supersoft ones , we do expect the entropy of the quasi2d structure to be higher , which will further shift the free energy balance to its favor .
in addition , in small@xmath22 limit , the out - of - plane modes in 2d structure will universally have dispersion @xmath23 ( associated with bending rigidity ) .
this will give another negative correction to free energy compared to dispersion @xmath24 of regular acoustic phonons .
note that these arguments are quite universal , and can be applied beyond the particular case of sq lattice .
we now proceed with the discussion of the specific example , 2d sq vs. its 3d - dual phase .
the latter 3d structure may be obtained from a simple cubic lattice , by removing the particles occupying sites . @xmath25 and @xmath26 . in this geometry
, the equilibrium condition requires @xmath27 , compared to @xmath28 for sq .
note that it is tension @xmath8 , not @xmath18 which is expected to be nearly identical for the two competing structures , since it is given by derivative of the potential , @xmath29 ( taken at the distance of the second nearest neighbor ) .
in contrast , @xmath18 can be considerably varied by relatively small deformation of the strong @xmath6spring .
an obvious choice of unit cell for 2d lattice is a square containing one atom . for the 3d structure ,
consider a cube containing eight smaller cubes , as shown on figure [ sq](b ) .
the particles are located at @xmath30 , where @xmath31 is @xmath32 , @xmath32 , @xmath33 , @xmath34 , @xmath35 , @xmath36 .
the translational symmetry in this structure is generated by vectors @xmath37 , @xmath38 and @xmath39 , i. e. the cube on the picture corresponds to two unit cells .
given the translational symmetry , the particle at @xmath39 is equivalent to @xmath40 ( we label them as type a particles ) , @xmath41 to @xmath32 ( type b ) and @xmath34 to @xmath42 ( type c ) .
we can now obtain hamiltonian @xmath43 and calculate the phonon contribution to free energy .
the number of particles per a unit cell is , @xmath44 for sq and @xmath45 for its 3d - dual . for the 2d phase , displacement of a particle at (
m , n ) is @xmath46 . for the 3d phase , the displacements there are three families of modes , corresponding to the three types of non - equivalent particles , a , b , and c : @xmath47 the phonon free energy has general form @xmath48 \exp \left ( -\frac{h}{k_{b}t}\right ) \label{free - en}\ ] ] here index @xmath49 parameterizes all the modes for a given wave vector @xmath50 , and hamiltonian @xmath43
is given by@xmath51by performing the gaussian integration over @xmath52 , one can transform eq .
( [ free - en ] ) into : @xmath53 \right ) \label{dets}\ ] ] the integration here is performed over a single brilluen zone , i.e. each component of the wave vector runs from @xmath54 to @xmath55
. note that in our regime ( @xmath56 and @xmath57 ) , one can neglect the coupling between hard and soft modes , which allows one represent the determinants entering eq .
( [ dets ] ) in factorized form : @xmath58 \simeq \det \left [ \hat{\gamma}^{\left ( hard\right ) } \left ( \mathbf{q}% \right ) \right ] \det \left [ \hat{\gamma}^{\left ( soft\right ) } \left ( \mathbf{% q}\right ) \right ] $ ] .
furthermore , the hard modes have essentially identical spectra in the both structures ( corresponding to one - dimensional chain of @xmath15-springs ) .
therefore , the difference of the fluctuational free energies is mainly due to the soft mode contributions . in the case of sq phase
, there is only one soft mode : @xmath59 = \hat{\gamma}_{zz}\left ( \mathbf{q}\right ) = \frac{4\sqrt{2}\tilde{% \tau}}{a}\sin ^{2}\left ( \frac{q_{1}}{2}\right ) \sin ^{2}\left ( \frac{q_{2}}{% 2}\right)\]]for the 3d dual phase , 3 out of 9 modes are soft , namely @xmath60 : @xmath61\]]here @xmath62,@xmath63 $ ] , and @xmath64 .
thus , @xmath65 \simeq 8\tilde{\kappa}^{3}\left ( 1 + 3\alpha \right ) ^{3}\left [ 1-\varpi_{1}\left ( \frac{1+\alpha } { 8\left ( 1 + 3\alpha \right ) } \right ) ^{2}+\varpi_{2}\left ( \frac{1+\alpha } { 8\left ( 1 + 3\alpha \right ) } \right ) ^{3}\right]\ ] ] here @xmath66 , and @xmath67 .
now we can calculate the fluctuational contribution into the free energy difference between the two structures : @xmath68 \label{result}\ ] ] consistent with the above general arguments , we obtain @xmath69 , i.e. the entropic effects enhance the stability of the self compactified phase . one can identify the physical origin of each of the three contributions at the above result .
the first logarithmic term is due to an entropic gain of a single fluctuation particle with all of its neighbors fixed .
one can see that these fluctuations are enhanced in 2d case ( for an arbitrary value of parameter @xmath70 ) , mainly due to the smaller number of second nearest neighbors there .
the second term , @xmath71 represent an additional gain due to @xmath72 dispersion of the soft acoustic phonon in sq phase .
in contrast , the soft modes in 3d phase correspond to optical phonons , whose dispersion gives rise to the negligible third term in eq .
( [ result ] ) .
since the effects responsible for the dominant contributions to our result , eq .
( [ result ] ) , are not specific for the above example , we expect our conclusions to be rather universal , and applicable to other self compactified phases .
in fact , the very same binary system introduced in ref .
@xcite , provides other examples of such phases , e. g. _ honeycomb _ lattice .
this structure has several 3d counterparts @xcite , some of which are shown on fig.([honey ] ) .
we expect the thermal fluctuations to enhance the stability of honeycomb with respect to transition to alternative 3d structures , as in the case of sq .
another aspect of our observation is that the region of the phase diagram corresponding to quasi-2d structures ( like sq or honeycomb ) , should expand when thermal fluctuations are introduced .
however , in this study we have not discuss the transition between solid and liquid phases .
in fact , the strong fluctuations in 2d structures indicate that they should melt earlier than 3d solid phases , which is consistent with the classical picture .
in other words , the thermal fluctuations enhance the stability of quasi-2d latices with respect to transition to 3d crystalline structures , but not to the disordered liquid phase .
note also that the self - compactified structure need not to be an ideal crystal , even topologically .
introduction of defects such as disclination or dislocation , normally associated with 2d melting , is not expected affect our arguments , as long as the systems remains effectively two - dimensional .
that is because the dominant contribution to the free energy is associated with length scales of the order of the interparticle distance .
hence , the macroscopic properties and conformation of these lattices are not relevant for the discussed phenomenon . as long at the large - scale behavior
is concern , one might expect their properties to be similar to those of tethered ( solid - like ) membranes @xcite .
it should be noted that the non - zero bending rigidity is known to stabilize such a membrane with respect to crumpling transition .
one may give a simple qualitative interpretation to the obtained results .
the self compactification in zero - fluctuation limit corresponds to the regime of the interparticle repulsion which is too weak to destabilize the 2d structure , yet strong enough to cause an overall repulsion between two such layers .
introduction of thermal fluctuations results in an additional effective repulsion between these 2d layers , an effect very similar to helfrich interaction between conventional membranes @xcite , helfrich . | we discuss the phenomenon of spontaneous self compactification in a model colloidal system , proposed in a recent work on dna mediated self assembly .
we focus on the effect of thermal fluctuations on the stability of membrane - like self assembled phase with in - plane square order .
surprisingly , the fluctuations are shown to enhance the stability of this quasi2d phase with respect to transition to alternative 3d structures . *
pacs numbers : 64.70.kb , 64.70.nd * |
in this paper , we report an investigation of the non - equilibrium relaxation kinetics in the vortex glass phase of layered disordered type - ii superconductors . since struik s original investigations,@xcite many glassy systems have been found to exhibit physical aging phenomena , which have attracted considerable interest during the past decades.@xcite recently , it has been realized that glass - like relaxation and aging can in fact be found in many other systems.@xcite glassy materials feature extremely long relaxation times which facilitates the investigation of aging phenomena in real as well as in numerical experiments .
our definition of physical aging here entails two fundamental properties : first , we require relaxation towards equilibrium to be very slow , typically characterized by a power law decay , observable in a large accessible time window @xmath0 ; here @xmath1 denotes an appropriate short microscopic time scale , whereas @xmath2 is the much larger equilibration time for the macroscopic system under consideration .
second , a non - equilibrium initial state is prepared such that the kinetics is rendered non - stationary ; thus , time - translation invariance is broken , and two - time response and correlation functions depend on both times @xmath3 and @xmath4 independently , not just on the elapsed time difference @xmath5 . in this context , @xmath3 is often referred to as waiting time , and @xmath6 as observation time .
in addition , in the limit @xmath7 many aging systems are characterized by the emergence of dynamical scaling behavior.@xcite the physics of interacting vortex lines in disordered type - ii superconductors is remarkably complex and has been a major research focus in condensed matter physics in the past two decades .
it has been established that the temperature vs. magnetic - field phase diagram displays a variety of distinct phases.@xcite a thorough understanding of the equilibrium and transport properties of vortex matter is clearly required to render these materials amenable to optimization with respect to dissipative losses , especially in ( desirable ) high - field applications .
investigations of vortex phases and dynamics have in turn enriched condensed matter theory , specifically the mathematical modeling and description of quantum fluids , glassy states , topological defects , continuous phase transitions , and dynamic critical phenomena .
an appealing feature of disordered magnetic flux line systems is their straightforward experimental realization which allows direct comparison of theoretical predictions with actual measurements .
the existence of glassy phases in vortex matter is well - established theoretically and experimentally.@xcite the low - temperature abrikosov lattice in pure flux line systems is already destroyed by weak point - like disorder ( such as oxygen vacancies in the cuprates ) .
the first - order vortex lattice melting transition of the pure system @xcite is then replaced by a continuous transition into a disorder - dominated vortex glass phase.@xcite here , the vortices are collectively pinned , displaying neither translational nor orientational long - range order.@xcite in addition , there is now mounting evidence for a topologically ordered dislocation - free bragg glass phase at low magnetic fields or for weak disorder;@xcite and an intriguing intermediate multidomain glass state has been proposed .
@xcite unambiguous signatures of aging in disordered vortex matter have also been identified experimentally : for example , du _ et al .
_ recently demonstrated that the voltage response of a 2h - nbse@xmath8 sample to a current pulse depended on the pulse duration @xcite ( see also ref .
out - of - equilibrium features of vortex glass systems relaxing towards their equilibrium state were studied some time ago by nicodemi and jensen through monte carlo simulations of a two - dimensional coarse - grained model system;@xcite however , this model applies to very thin films only since it naturally disregards the prominent three - dimensional flux line fluctuations .
more recently , three - dimensional langevin dynamics simulations of vortex matter were employed by olson _
_ @xcite and by bustingorry , cugliandolo , and domnguez @xcite ( see also ref . [ ] ) in order to investigate non - equilibrium relaxation kinetics , with quite intriguing results and indications of aging behavior in quantities such as the two - time density - density correlation function , the linear susceptibility , and the mean - square displacement .
rom and domnguez extended these studies to monte carlo simulations of the three - dimensional gauge glass model at the critical temperature.@xcite we remark that it is generally crucial for the analysis of out - of - equilibrium systems to carefully investigate alternative microscopic realizations of their dynamics in order to probe their actual physical properties rather than artifacts inherent in any mathematical modeling .
indeed , different mathematical and numerical representations of non - equilibrium systems rely on various underlying _ a priori _ assumptions that can only be validated _ a posteriori_. it is therefore imperative to test a variety of different numerical methods and compare the ensuing results in order to identify those properties that are generic to the physical system under investigation . in this paper
we employ metropolis monte carlo simulations for a three - dimensional interacting elastic line model to investigate the relaxation behavior in the physical aging regime for systems with uncorrelated attractive point defects .
we strive to employ parameter values that describe high-@xmath9 superconducting materials such as ybco , and limit our investigations to low magnetic fields and temperatures ( typically @xmath10 k ) in order for our disordered elastic line model to adequately represent a type - ii superconductor with realistic material characteristics .
thus we address a parameter and time regime wherein the slow dynamics is dominated by the gradual build - up of correlations induced by an intricate interplay of repulsive vortex interactions and attractive point pinning sites .
our work differs in crucial aspects from other recent studies [ ] . as in refs .
[ ] , we consider the in type - ii superconductors physically relevant situation where all defects serve as genuinely attractive and localized pinning sites for vortices , in the sense that they locally reduce the chemical potential ( or equivalently , suppress the superconducting transition temperature ) .
our pinning potential landscape is therefore characterized by large flat regions in space , where the vortices feel no pinning force , interspersed with small attractive potential minima of extension @xmath11 , much smaller than the london penetration depth @xmath12 that sets the vortex - vortex interaction range .
this is to be contrasted with the model used in refs .
[ ] , which is rather motivated by studies of interfaces in random environments that are described by gaussian distributions for the disorder strength.@xcite consequently these models inevitably incorporate both attractive and repulsive disorder , which can be viewed as mimicking a sample with a very high density of point defects .
alternatively , such a coarse - grained representation of pinning centers forming a continuous disorder landscape certainly becomes appropriate at elevated temperatures near @xmath9 , since then the pinning range is set by the coherence length @xmath13 , which diverges as the critical point is approached .
thus , a random medium description is best - suited for investigations of critical phenomena , and also more easily amenable to field - theory representations . at low temperatures , however , where @xmath14 , our modeling of the localized pinning centers appears more realistic , and we furthermore remark that in this scenario repulsive defects would introduce different physics in the non - equilibrium relaxation and aging kinetics of vortices in superconductors , such as flux bunching in regions devoid of such disorder .
we therefore carefully exclude any repulsive pinning sites .
in addition , the temperatures used in refs .
[ ] appear to be considerably higher than those studied in our present work .
further differences can be found in the length of the vortex lines ( in refs .
[ ] rather short lines were considered ) , in the boundary conditions , as well as in the initial preparation of the system .
we characterize the aging properties of the interacting and pinned flux lines through several different two - time quantities , namely the vortex density - density autocorrelation function , the flux line height - height autocorrelation function , and the transverse vortex mean - square displacement .
investigating the influence of weak point defects , we find that the non - equilibrium relaxation properties of magnetic flux lines in disordered type - ii superconductors are governed by various crossover effects that reflect the competition between pinning and repulsive interactions . in the long - time limit and for not too large pinning strengths , the dynamics is manifestly similar to that observed in structural glasses .
the structure of this paper is as follows : in section [ sec2 ] we describe our model and the monte carlo simulation algorithm , and define the quantities of interest for our study .
our data and principal results are presented in section [ sec3 ] . in order to disentangle the different contributions to the non - equilibrium relaxation dynamics of our system
, we first separately elucidate the effects of attractive pinning centers and of long - range vortex - vortex interactions , before we endeavour to analyze and understand their intriguing interplay as reflected in the vortex system s relaxation kinetics .
finally , we discuss our findings in section [ sec4 ] , and compare them with other studies .
we consider a three - dimensional vortex system in the london limit , where the london penetration depth is much larger than the coherence length .
we model the vortex motion by means of an elastic flux line free energy described in ref . [ ] , see also , e.g. , refs .
[ ] . the system is composed of @xmath15 flux lines in a sample of thickness @xmath16 .
the effective model hamiltonian @xmath17 is defined in terms of the flux line trajectories @xmath18 , with @xmath19 , and consists of three components , namely the elastic line tension , the repulsive vortex - vortex interaction , and the disorder - induced pinning potential : @xmath20 here , the elastic line stiffness is @xmath21 , with @xmath12 and @xmath13 denoting the london penetration depth and coherence length in the crystallographic ab plane ( we assume the magnetic field along the c axis ) , and the anisotropy parameter ( effective mass ratio ) @xmath22 .
the energy scale is set by @xmath23 , where @xmath24 is the magnetic flux quantum .
the expression for the elastic line energy in eq .
( [ hamilt ] ) is valid in the limit @xmath25 .
the purely in - plane repulsive interaction potential ( consistent with the extreme london limit ) between flux line elements is given by the modified bessel function of zeroth order , @xmath26 , which diverges logarithmically as @xmath27 , and decreases exponentially for @xmath28 .
these vortex interactions are truncated at half the system size , which is in turn chosen sufficiently big such that numerical artifacts due to this cut - off length are minimized .
we model point pinning centers through square potential wells with radius @xmath11 and strength @xmath29 at @xmath30 defect positions .
( for additional details , see refs . [ ] . )
our simulation parameter values @xcite were chosen corresponding to typical material parameters for ybco as listed in appendix d of ref .
[ ] . in the following , lengths and energies are reported relative to the effective defect radius @xmath11 and interaction energy scale @xmath31 ( using cgs units ) , and time in monte carlo steps ( mcs ) , where one mcs correspond to @xmath32 proposed updates of the flux line elements , with @xmath15 the number of flux lines and @xmath16 the number of layers .
we set the pinning center radius @xmath33 , anisotropy parameter @xmath34 , and , as is appropriate at low temperatures , @xmath35 , and @xmath36 .
then @xmath37 ( in cgs units of energy / length ) , and the energy scale in the line tension term becomes @xmath38 .
we systematically vary the pinning strength @xmath29 between @xmath39 and @xmath40 .
usually , our simulations are performed at temperature @xmath41 k , which corresponds to @xmath42 .
thermal excitation energies are thus small compared to the elastic and pinning energies , and at equilibrium we therefore expect the system to be deep in the glassy regime .
we do not allow for flux line cutting and reconnection processes in our simulations of a low - temperature and dilute vortex system .
we apply the standard metropolis monte carlo simulation algorithm in three dimensions with a discretized version of the above effective hamiltonian ( [ hamilt]).@xcite the system contains @xmath43 flux lines in @xmath16 layers , with a distance @xmath11 between consecutive layers ; and an equal number @xmath44 of point pinning centers that are however randomly distributed within the layer , with mean separation @xmath45 ; in comparison , the triangular vortex lattice spacing would be @xmath46 in our dilute system .
we apply periodic boundary conditions in all three space directions , as we are mainly interested in bulk properties .
this is to be contrasted with refs .
[ ] , where free boundary conditions were used along the c axis . we have systematically changed @xmath16 between @xmath10 and @xmath47 in order to carefully monitor finite - size effects .
the in - plane system size is @xmath48 = [ \frac{2}{\sqrt{3 } } \times 8 \lambda_{\rm ab } , 8 \lambda_{\rm ab}]$ ] ; the dimensions of the @xmath49 plane were chosen such that in the absence of disorder the system accommodates a regular triangular flux lattice . in the absence of defects
, we have tested that initially randomly placed vortices properly equilibrate to form a triangular abrikosov flux lattice .
we have also checked that there are no appreciable effects due to the sharp cut - off of the vortex interactions at @xmath50 . in order to investigate aging phenomena in the system with uncorrelated point disorder ,
the vortices are prepared in an out - of - equilibrium state : straight flux lines are initially ( at @xmath51 ) placed at random locations in the system .
the vortex lines are subsequently allowed to relax at the temperature @xmath52 k for a duration @xmath3 , the ` waiting ' time , before we start measuring two - time quantities for @xmath4 , see fig . [
algorithm ] .
( this is again different to refs .
[ ] , where the vortex lines were equilibrated at high temperatures inside the vortex liquid phase before the subsequent quench to lower temperatures . ) our waiting times extend up to @xmath53 mcs , whereas the total length of a simulation run is typically @xmath54 .
, straight vortex lines are initialized far out of equilibrium by placing them randomly in the system at @xmath41 k in the presence of weak point defects .
the vortex lines are then allowed to relax for different waiting times @xmath3 , before various two - time quantities , such as the height autocorrelation function @xmath55 , are measured . ]
aging phenomena can generally be adequately characterized through the study of two - time quantities . in our work
we put special emphasis on a range of observables that allow us to rather comprehensively monitor the distinct relaxation processes in vortex matter that originate from pinning to attractive point defects and repulsive interaction forces , respectively , and their intricate competitive interplay . the height - height autocorrelation function and
mean - square displacement represent two quantities that are routinely studied in the context of interface fluctuations and non - equilibrium growth processes .
@xcite separating the time - dependent position of the flux line @xmath56 in the @xmath57th layer into its @xmath58 and @xmath59 components , @xmath60 , the two - time height - height autocorrelation function can be written as @xmath61 where @xmath62 , and similarly for @xmath63 .
the brackets @xmath64 here denote both an average over the noise history , i.e. , over the sequential realizations of random number sequences , as well as a configurational average over defect distributions and initial positions of the straight vortex lines at the outset of the simulation runs . the two - time mean - square displacement in the @xmath49 planes , transverse to the external magnetic field , can similarly be cast in the form @xmath65 \ .\end{aligned}\ ] ] we remark that other related quantities that essentially contain the same information , are the two - time roughness function or the two - time structure factor.@xcite unfortunately , both the height autocorrelation and the mean - square displacement are probably not easily accessible in experiments on type - ii superconductors , except perhaps through low - angle neutron scattering .
much better suited for an experimental study is likely the ( connected ) two - time vortex density - density autocorrelation function that can formally be written as @xmath66 where @xmath67 represents the local flux density per unit area at position @xmath68 , with constant uniform average @xmath69 . following an initially random placement , the repulsive vortex interactions cause positional rearrangements , such that one would expect a temporal decay of the density autocorrelation function . in our simulations , we realize the vortex density autocorrelation function in the following way:@xcite as before we start with randomly placed straight vortex lines at @xmath70 and let the system subsequently relax up to waiting time @xmath3 . a density count for the vortex line elements
is then generated by setting a circular area , with a radius equal to @xmath71 at the location of each vortex line element @xmath72 at @xmath73 .
typical values for @xmath74 range from @xmath75 to @xmath76 . as time
@xmath6 elapses , we count the number of vortex line elements still in their circles , generating a time sequence of occupation numbers @xmath77 , with @xmath78 or @xmath79 , and @xmath80 by construction . due to the repulsive vortex interactions ,
flux line elements tend to move away from their initial positions , whence @xmath81 can be @xmath39 at a later time if the vortex line element leaves the prescribed circle . in the presence of pinning centers
, vortex line elements will become trapped inside the defects over a long time , causing @xmath81 to preferentially remain @xmath79 .
this quantity is then averaged over the @xmath82 different vortex segments and many distinct defect distributions and initial configurations , yielding @xmath83 fig .
[ correlation ] illustrates the algorithm for calculating the density autocorrelation function.@xcite we have checked the results for different values of @xmath74 and found that within a reasonable range the precise choice of @xmath74 does not affect the results in the long - time aging regime where @xmath7 .
( large open circles ) at different observation times.@xcite ( a ) at @xmath73 , each vortex line element by construction resides inside its own circle and gives a count @xmath84 ( @xmath85 here ) .
( b ) at later times @xmath4 , the repulsive vortex interactions cause the flux line elements to move away from their initial positions .
this results in a smaller occupation number @xmath86 ( @xmath87 in this example ) .
( c ) at long times @xmath7 , it is possible that all vortex line elements have left the circle , @xmath88 , which results in a complete decorrelation .
for instance , the quantity @xmath89 in the pictures at these three times is evaluated to be @xmath10 at @xmath73 , @xmath90 at @xmath4 , and @xmath39 for @xmath7 . ]
in order to fully understand the non - equilibrium relaxation processes and aging phenomena in disordered type - ii superconductors at low temperatures , we found it imperative to carefully disentagle the dynamical contributions originating from the repulsive interactions between the vortex lines and from their pinning to attractive point defects .
we start our discussion with free flux lines , mainly in order to validate our code by comparing our data with the theoretically expected behavior and earlier work .
we will then separately consider the effects of attractive point pins and of the long - range repulsive vortex interactions , before we at last venture to study the interplay of these two competing mechanisms to induce or relax correlations in the system .
the relaxation kinetics of a single free elastic vortex line constitutes a valuable benchmark to check our program as this case can be easily understood by recalling that in the presence of thermal noise a fluctuating interface that tries to minimize its line tension should be described in the continuum limit by the linear edwards
wilkinson equation.@xcite as the fluctuations in the transverse @xmath58 and @xmath59 directions are independent random variables for our free line , we expect the results for the free vortex to be described by the one - dimensional version of that well - known stochastic equation ( @xmath91 below stands for either @xmath58 or @xmath59 ) : @xmath92 where @xmath93 represents a gaussian white noise with zero mean and covariance @xmath94 , @xmath95 is the line stiffness ( equal to @xmath96 here ) , and @xmath97 the temperature of the heat bath .
the edwards wilkinson equation , as well as a range of microscopic models belonging to the same dynamic universality class , has been studied extensively .
starting from a straight line , one first observes a short - time regime with uncorrelated fluctuations , which is rapidly replaced by a correlated intermediate - time interval characterized by a non - trivial power law increase of the line roughness . after a crossover time that algebraically
depends on the system size , this correlated regime finally reaches the steady - state or saturation regime .
two - time quantities have also been studied in the context of the edwards
wilkinson equation,@xcite , facilitated by the fact that a full analytical analysis is possible for the linear stochastic equation ( [ eq : ew ] ) .
for example , in the correlated regime of the one - dimensional edwards wilkinson equation the following exact expression for the height - height autocorrelation function has been derived:@xcite @xmath98 \ , \label{eq : corr_ew}\ ] ] where @xmath99 is a known constant .
the detailed crossover properties of two - time quantities in the region between the correlated and saturated regimes have been carefully investigated in ref .
and ( d - f ) @xmath100 planes , and averaged over typically 200 independent monte carlo simulation runs ; ( a ) and ( d ) : mean - square displacement , ( b ) and ( e ) : height - height autocorrelation function , ( c ) and ( f ) : vortex density - density autocorrelation function ( measured with @xmath101 ) .
data obtained for different waiting times are shown . for @xmath102 layers the system rapidly reaches the steady state , and time - translation invariance
is recovered , i.e. , the two - time quantities only depend on the time difference @xmath103 . for @xmath100 aging and dynamic scaling prevail throughout the simulation time window .
the full line in ( e ) indicates the exact expression ( [ eq : corr_ew ] ) derived from the edwards wilkinson equation .
the dashed lines in ( d ) and ( f ) indicate the predicted asymptotic power laws with exponents @xmath104 and @xmath105 , respectively . here and in the following figures error bars are much smaller than the symbol sizes . ] in fig .
[ fig3 ] , we display our monte carlo simulation results for our elastic vortex line model when both the vortex interaction and the defect pinning are switched off , i.e. , only the first contribution in ( [ hamilt ] ) is retained .
one immediately notices a striking difference between the behavior of a `` thin film '' composed of only a few layers ( such as @xmath102 , see figs .
[ fig3]a[fig3]c ) and `` bulk '' systems consisting of many layers ( @xmath100 , see figs . [ fig3]d[fig3]f ) . in the former case
, the system rapidly evolves into the steady - state regime , yielding two - time quantities that only depend on the elapsed time difference @xmath103 . as a result ,
the transverse displacements in the @xmath58 and @xmath59 directions perform simple random walks , as is revealed by the linear increase of the mean - square displacement with time , see fig .
[ fig3]a . for the larger bulk system ,
the correlated regime persists throughout the duration of our simulations , and both waiting and observation times reside well within that extended intermediate regime .
this gives rise to aging and dynamical scaling : time - translation invariance is broken , and all the two - time quantities display full - aging scaling.@xcite for each two - time observable we find the following scaling behavior ( here given for the height autocorrelation function @xmath55 ) : @xmath106 where @xmath107 represents an aging scaling exponent and @xmath108 denotes an associated scaling function that follows a power law decay for large arguments .
for the height - height autocorrelation we have @xmath109.@xcite in fig .
[ fig3]e we explicitly compare our numerically determined scaling function with the expression ( [ eq : corr_ew ] ) resulting from the direct solution of the edwards wilkinson equation ( full line ) , and obtain perfect agreement . summarizing , we see that the free vortex line fluctuations are indeed aptly described by the one - dimensional edwards wilkinson equation ( [ eq : ew ] ) .
we also observe a strong dependence on the system s extension @xmath16 in the magnetic field direction , i.e. , the vortex length : on the time scale of our simulations , the stationary regime is almost immediately reached when the system consists of only a few layers ; in contrast , for larger bulk systems aging and dynamical scaling are observed easily .
this points to quite distinct relaxation behavior in thin superconducting films and thicker bulk samples .
we decided to avoid the additional complications stemming from the crossover between the correlated and steady - state regimes in our present study , and to rather focus on system sizes sufficiently large that no finite - size effects ( no crossover to the steady state ) are observed on the accessed time scales .
properties of smaller systems and possible experimental consequences for thin superconducting films will be discussed in a separate publication .
intuitively , one anticipates pinning centers to strongly influence the thermal fluctuations of our elastic flux lines .
indeed , attractive forces emanating from the pinning centers will tend to localize vortex segments , and thus ultimately suppress thermal fluctuations . depending on the pinning strength
, flux line elements will end up spending an appreciable amount of time close to a pinning center .
therefore , compared to freely fluctuating lines , a marked _ increase _ of correlations as function of time must be expected .
before we proceed to analyze the influence of pinning centers in more detail , we need to stress that we exclusively consider attractive point defects , in accordance with the physics of disordered type - ii superconductors in the low - temperature regime . a recent study [ ] addressed the relaxation and aging properties of elastic lines subjected to a random potential , corresponding to both attractive and repulsive pinning centers .
whereas a gaussian disorder strength distribution certainly is a good model for disordered ferromagnets , its relevance for relaxation processes in disordered type - ii superconductors at low temperatures is less obvious . vs. time @xmath6 ( in mcs ) for different values of the pinning strength @xmath110 , for systems of size @xmath111 .
in the initial time regime , the pinning centers attract the vortex segments as revealed by an increase of the slope of @xmath112 as compared to a pure system . at later times the localization of the flux line elements induced by the pinning yields a strong decrease of this slope .
the data shown result from averaging over typically 100 independent runs .
the dashed lines indicate the times @xmath113 , @xmath114 , @xmath115 , and @xmath116 mcs , see fig .
[ fig5 ] below . ]
let us start by looking at the mean - square displacement @xmath112 , with @xmath117 , which gives a measure of the ( squared ) distance traveled by the flux line elements since the initial preparation of the system . in fig .
[ fig4 ] we compare the behavior of a free elastic line with that of flux lines subject to pinning centers of various strengths @xmath118 .
the presence of attractive pins clearly gives rise to different regimes .
the flux lines are rapidly attracted by the point defects , which yields an increase of the slope in the log - log plots of @xmath112 vs. time @xmath6 .
this continues until some pinning strength - dependent crossover time at which the slope decreases even below the value of the free line , signifying the confinement of localized vortex segments to the vicinity of the pins .
as one would expect , this crossover time decreases for increasing pinning strengths .
for @xmath119 , @xmath112 remains essentially unchanged , which indicates that for non - interacting lines there exists a critical pinning strength above which thermal fluctuations are no longer sufficiently strong to allow the vortex line elements to escape from the defects .
these different regimes also manifest themselves when two - time quantities are considered , as seen in fig .
[ fig5 ] , where we have plotted the data according to the free - line scaling behavior .
of course , it is not to be expected that these scaling laws remain valid when attractive defects are added to the system , but this representation of our data facilitates the following discussion .
we first remark , see figs .
[ fig5]a and [ fig5]d , that the change of the slope of @xmath112 translates into deviations of the mean - square displacement @xmath120 from the free - line scaling that can be readily understood .
for example , for @xmath121 the time intervals @xmath122 $ ] , @xmath123 $ ] , and @xmath124 $ ] , used to compute @xmath120 for the waiting times @xmath125 , @xmath126 , and @xmath114 , respectively , correspond to the time regime with increasing local slopes of @xmath112 , compare fig .
this yields a shift of @xmath127 to higher values .
as the crossover time of @xmath112 lies inside the interval @xmath128 $ ] , the converse behavior is observed for @xmath129 and even larger waiting times , with a shift of @xmath127 to lower values .
this effect is more pronounced for larger pinning strengths , since then the crossover of @xmath112 takes place earlier . and ( d - f ) @xmath130 : ( a ) and ( d ) mean - square displacement ; ( b ) and ( e ) height - height autocorrelation function ; ( c ) and ( f ) vortex density - density autocorrelation ( with @xmath101 ) .
the shown data result from averaging over typically 100 independent runs in systems of size @xmath131 . for direct comparison with fig .
[ fig3 ] , the data are plotted according to the scaling properties of free elastic lines . whereas an approximate scaling prevails for small waiting times ( especially for the mean - square displacement and the vortex density - density autocorrelation ) , strong deviations emerge for larger waiting times . in ( e ) the apparent collapse of the height autocorrelation function for the largest waiting times is merely caused by the scale used in that figure . ]
the strongest influence of point defects and largest deviations from free elastic lines are observed in the height autocorrelation function , figs .
[ fig5]b and [ fig5]e . as the flux line elements are trapped by the pinning centers , their transverse in - plane displacements
become diminished , which leads to an increase of the correlations as a function of waiting time .
in addition , the decay of @xmath55 as a function of @xmath6 is much slower for larger values of @xmath3 . for larger pinning strengths and long waiting times
we even observe non - monotonic temporal evolution , as the trapped flux lines experience an increase of the height correlations .
finally , the vortex density - density autocorrelation turns out to be the least sensitive among our two - time observables to the presence of pinning centers , at least for comparatively small waiting times , see figs .
[ fig5]c and [ fig5]f .
indeed , for moderate values of @xmath3 one still observes the free - line scaling behavior ; this is of course a consequence of our prescription for the computation of this correlation function , namely setting a circular area with fixed radius around every vortex line element at time @xmath3 : as long as only a few line elements are captured by a point defect , the scaling of @xmath132 remains approximately unchanged . only when the majority of the vortex segments become trapped , does this localization induce a strong enhancement of the vortex density - density correlations .
the shown data result from averaging over typically 100 independent runs in systems of size @xmath133 .
the data display simple aging scaling , with different exponents for the two quantities . ] to conclude this section , we note that the non - equilibrium relaxation physics is drastically different when both attractive and repulsive pinning centers are implemented .
as studied in ref .
[ ] ( see also refs . [ ] ) an elastic line in a random potential is characterized by a time - dependent correlation length that crosses over from an early time power law growth to an asymptotic logarithmic growth .
consequently , two - time quantities display an apparent simple aging scaling with effective exponents that depend on temperature and on the randomness .
we have verified that we obtain similar results as in ref .
[ ] when using both attractive and repulsive defects in our model and monte carlo algorithm . indeed , as shown in fig .
[ fig6 ] , the time - dependent correlation length gives rise to simple aging scaling of our two - time quantities , with effective exponents that display a dependence on temperature and on the distribution of the pinning strengths .
these crossover features also capture most of the relevant properties of disordered ferromagnets undergoing phase ordering.@xcite in refs .
[ ] , disordered type - ii superconductors in the low - temperature phase were modeled by a corresponding model with random pins that are either attractive or repulsive .
however , the physical realization relevant to materials is that of purely attractive pins , similar to those studied in our present work . yet since the properties of elastic lines strongly depend on the nature of the pinning centers , any conclusions regarding the non - equilibrium relaxation properties of disordered type - ii superconductors at low temperatures that are inferred from models with both attractive and repulsive defects should be viewed with some scepticism . vs. time @xmath6 ( in mcs ) for interacting vortex lines with different values of the pinning strength @xmath110 ( system size @xmath134 ; the data result from averaging over typically 100 independent runs ) .
due to the long - range interactions , the flux lines aim to maximize their separations , thus yielding values of @xmath112 that are two orders of magnitude larger than in the absence of repulsive forces , compare fig .
these displacements are impeded and vortex motion eventually stopped by the caging constraints of neighboring lines and pinning to attractive point defects .
the ( red ) dashed line displays our data for a pure system , in the absence of pinning centers .
the vertical dashed lines indicate the times @xmath113 , @xmath114 , @xmath115 , and @xmath116 mcs . ] in the absence of disorder our system composed of interacting flux lines evolves toward a regular triangular abrikosov lattice . as we start our simulations by deposing initially straight lines at random positions , large displacements of the flux line elements are expected , as the system tries to minimize the long - range in - plane repulsive vortex interaction energy .
the ensuing dynamic regimes are again nicely captured by the mean - square displacement @xmath112 which takes on values that are two orders of magnitude larger than in the absence of interactions , see the ( red ) dashed line in fig .
while the flux line segments experience these large displacements , @xmath112 displays an approximate power law increase with time , with an effective exponent of @xmath135 . once the majority of vortices have reached the vicinity of their final equilibrium positions , the slope of @xmath112 starts to gradually decrease . ) .
data obtained for different waiting times @xmath3 are shown , typically obtained from averaging over 500 independent simulation runs ; the system size is @xmath131 . ( a ) , ( c ) , and ( e ) display the unscaled data in a log - log plot , whereas ( b ) , ( d ) , and ( f ) show the approximate scaling observed for not too large waiting times .
this scaling regime corresponds to the intermediate time window where @xmath112 exhibits a power law increase as a function of time , with an ( effective ) exponent @xmath135 , see fig .
[ fig7 ] . ]
the two - time quantities reveal both the initial - time regime as well as the crossover at later times , see fig .
the mean - square displacement @xmath120 yields a reasonably good data collapse for the smaller waiting times with the scaling exponent @xmath136 that follows from the slope of @xmath112 in that regime .
when the observation time @xmath6 exceeds the crossover time , this scaling breaks down .
instead the growth rate of @xmath120 decreases strongly with increasing @xmath3 , even resulting in a crossing of the curves for different waiting times .
the behavior of @xmath120 is mirrored by that of the vortex density autocorrelation function : for small waiting times @xmath3 , scaling is achieved with exponent @xmath137 , whereas for larger waiting times the decay of the correlation slows down as @xmath3 increases . from these results
we infer that the vortex density - density autocorrelation contains essentially the same physical information as the mean - square displacement .
interestingly , the height - height autocorrelation function displays a different scaling for smaller waiting times , given by the scaling exponent @xmath138 of the free line , see figs .
[ fig8]c and [ fig8]d .
this means that during the initial rearrangement of the vortex lines the height fluctuations are essentially the same as for the free line .
only when the vortices come close to their equilibrium positions does the character of the correlations change , reflecting the presence of long - range repulsive forces .
we are now ready to study the combined effects of repulsive flux line interactions and point defect pinning during the non - equilibrium relaxation of vortex matter in disordered type - ii superconductors .
we again begin by first considering the mean - square displacement @xmath112 , see fig .
adding very weak attractive defects , e.g. , with @xmath121 , has only a very minor effect on the time evolution of @xmath112 .
strengthening the point pins leads to a smaller rate of increase for the mean - square displacement , see the curve for @xmath139 . at early times
the flux lines are still displaced from their initial positions , as the vortex interactions try to establish an abrikosov lattice . however , at intermediate times these displacements are impeded by the defects that noticeably slow down vortex motion . as a result
the system tends to a new ( quasi-)equilibrium state that balances these two competing mechanisms . for even stronger
pinning , the moving flux lines become rapidly trapped by the disorder and the system gradually freezes into a blocked configuration . in fig .
[ fig7 ] this is clearly the case for both @xmath140 and @xmath141 . ) .
the pinning strength here is @xmath130 , the system size @xmath131 , and the data result from averaging over typically 800 independent simulation runs .
the waiting times @xmath3 for the vortex density autocorrelation are the same as those for the mean - square displacement . for the height autocorrelation function ,
the waiting times range from @xmath142 to @xmath53 , the values of @xmath3 being doubled between consecutive curves .
the ( normalized ) height - height autocorrelation function , after a crossover , displays the same qualitative behavior as that encountered in structural glasses . for large values of @xmath5 ,
the height autocorrelation assumes the stretched exponential form ( [ eq : str_exp ] ) with @xmath143 , as shown by the dashed line overlying the @xmath53 data . ]
the most interesting scenario naturally emerges for intermediate defect strengths .
indeed , when the pins are very weak , the rearrangement of the flux lines is barely affected , and all studied two - time quantities quantitatively display the same behavior as in a pure system . on the other hand , when the defects are too strong , the flux line elements remain firmly attached to the pinning centers , and a frozen configuration ensues .
the non - trivial behavior at intermediate pinning strengths is studied in more detail in fig .
[ fig9 ] for the case @xmath130 . the data for @xmath120 and @xmath132 shown in figs .
[ fig9]a and [ fig9]c are readily understood by comparing them to the corresponding results for the pure case , see figs .
[ fig8]a and [ fig8]e . the difference between these data sets is the absence of the early - time regime where @xmath112 has an approximately constant slope , see fig .
consequently the data with @xmath130 do not allow any data collapse , not even for the smallest waiting times considered .
however , this is the only noticeable difference , and the behavior for larger waiting times is qualitatively the same as for @xmath144 , except that the decrease in slope of @xmath120 for larger values of @xmath3 is stronger when @xmath145 .
however , a completely different picture emerges for the evolution of the normalized height - height autocorrelation function . as shown in fig .
[ fig9]b , for waiting times @xmath3 larger than a certain crossover value , @xmath55 exhibits the typical two - step relaxation of a structural glass : an initial time - translation invariant regime , which corresponds to the so - called @xmath146 relaxation in glasses and only depends on the elapsed time difference @xmath103 , is followed by a slow decay that is usually referred to as @xmath74 relaxation in the glass literature.@xcite in the long - time limit we can fit this slow decay to a stretched exponential @xmath147 \
, \label{eq : str_exp}\ ] ] with @xmath148 , and a waiting - time dependent decorrelation time @xmath149 . for our different waiting times
we obtain a consistent value @xmath150 for the stretching exponent in eq .
( [ eq : str_exp ] ) .
this emergence of a characteristic two - step glass - like relaxation is very intriguing .
obviously , the flux lines do not settle into a stable microstate even after their lateral displacements have become strongly reduced owing to the capture by the attractive pinning centers and the caging due to their repelling neighboring vortices . instead , as a consequence of the two competing relaxation mechanisms , collective dynamics and slow decorrelation sets in that yields the typical two - step relaxation dynamics of a glass .
we also note the intriguing shape of the normalized height autocorrelation function in the crossover regime .
indeed , at intermediate waiting times @xmath55 displays a strongly non - monotonic behavior , with a maximal value that even _ exceeds _ the value @xmath151 at @xmath73 .
this remarkable feature points to a fundamental change in the nature of the emerging correlations , which is due to the trapping of vortex segments in the vicinity of the defects and the subsequent balancing of the competition between the attractive pinning and the repulsive interactions .
our three - dimensional monte carlo investigation of relaxation processes in disordered type - ii superconductors has allowed us to gain a thorough understanding of the non - equilibrium properties of these technologically important materials .
we find the relaxation processes to be dominated by the interplay of two competing interactions , namely the pinning of the flux line elements to attractive point defects and the long - range mutual repulsion of the vortices .
this competition generates various crossover scenarios that we have discussed systematically .
the most interesting regime emerges for pinning centers of intermediate strength , for which we observe a distinguished two - step relaxation and a final slow , stretched - exponential decay of the height - height autocorrelation function .
this behavior is reminiscent of that encountered in structural glasses , clearly demonstrating that disordered type - ii superconductors subject to point defects indeed display pronounced glassy behavior at low temperatures , again justifying the term `` vortex glass '' for this frustrated pinned low - temperature phase .
we remark that our results are at variance with recent investigations based on three - dimensional london langevin dynamics simulations , where standard aging and dynamical scaling behavior of two - time quantities was observed,@xcite akin to the relaxation features of elastic lines in a random medium.@xcite however , these studies , in addition to difference in sample preparation , system size , and boundary conditions in the @xmath57-direction , employed a coarse - grained continuous random medium model of disordered type - ii superconductors that becomes adequate near the normal- to superconducting transition , but does not realistically capture superconducting materials at low temperatures for which isolated defects such as oxygen vacancies always induce a local suppression of the transition temperature and therefore constitute attractive localized pinning centers for vortices .
as our study shows , dynamical scaling no longer prevails for purely attractive point pinning centers , but instead much richer glassy relaxation dynamics sets in . as already mentioned in the introduction
, it is essential for investigations of non - equilibrium systems to study different dynamics and their algorithmic implementations in order to ensure that any ensuing results indeed describe actual physical properties of the system rather than numerical artifacts .
we have therefore recently begun to implement corresponding london
langevin dynamics simulations for our elastic line model ( [ hamilt ] ) with exclusively attractive pinning centers.@xcite our first tentative findings are in complete agreement with the monte carlo simulation results reported in this paper : they too show the emergence of glass - like behavior , with a slow , stretched - exponential decay at long times .
an in - depth analysis of this dynamics is currently in progress ; this comparative study also aims at matching the different microscopic time scales implicit in monte carlo and langevin dynamical simulations .
our current study can readily be expanded in various directions .
our results are valid in the regime where all time - dependent length scales remain small compared to the size of the system .
however , many transport and relaxation experiments are carried out on thin superconducting films rather than bulk samples . in our model
, a finite ( small ) number of layers introduces a dominant new length scale that substantially changes the relaxation processes , leading to additional crossover features .
we also note that other types of defects can be experimentally realized , ranging from parallel and splayed columnar pins to planar defects , and combinations thereof with point disorder .
it is an open and intriguing problem to understand how these different defect configurations influence the out - of - equilibrium relaxation processes in type - ii superconductors .
a detailed understanding of the relaxation phenomena in superconducting materials may facilitate characterization and optimization of samples with respect to pinning and flux transport .
finally , in all transport applications the flux lines are driven across the samples by external currents , which at long times yields a non - equilibrium steady state replacing the thermal equilibrium state that emerges without drive . following similar lines as in the present study
, one should be able to also analyze the relaxation properties of driven disordered type - ii superconductors in a comprehensive manner .
we plan to address these and related problems in the future .
we are indebted to thananart klongcheongsan for his original contributions during his ph.d .
dissertation work , and for preparing figure [ correlation ] .
we thank sebastian bustingorry for useful correspondence , and ulrich dobramysl for interesting and helpful discussions .
this work was supported by the u.s .
department of energy , office of basic energy sciences ( doe bes ) under grant no .
de - fg02 - 09er46613 .
m. henkel and m. pleimling , in _ rugged free energy landscapes : common computational approaches in spin glasses , structural glasses and biological macromolecules _ , ed .
w. janke , lecture notes in physics * 736 * , 107 ( springer , 2008 ) .
t. giamarchi and p. le doussal , phys .
* 72 * , 1530 ( 1994 ) ; phys .
b * 52 * , 1242 ( 1995 ) ; phys . rev
. lett . * 76 * , 3408 ( 1996 ) ; phys .
b * 55 * , 6577 ( 1997 ) ; t. klein , i. joumard , s. blanchard , j. marcus , r. cubitt , t. giamarchi , and p. le doussal , nature * 413 * , 404 ( 2001 ) .
m. nicodemi and h.j .
jensen , phys .
lett . * 86 * , 4378 ( 2001 ) ; j. phys . a : math
* 34 * , l11 ( 2001 ) ; europhys . lett . * 54 * , 566 ( 2001 ) ; j. phys .
a : math . gen . * 34 * , 8425 ( 2001 ) ; phys . rev .
b * 65 * , 144517 ( 2002 ) . | we study the non - equilibrium relaxation kinetics of interacting magnetic flux lines in disordered type - ii superconductors at low temperatures and low magnetic fields by means of a three - dimensional elastic line model and monte carlo simulations . investigating the vortex density and height autocorrelation functions as well as the flux line mean - square displacement , we observe the emergence of glassy dynamics , caused by the competing effects of vortex pinning due to point defects and long - range repulsive interactions between the flux lines .
our systematic numerical study allows us to carefully disentangle the associated different relaxation mechanisms , and to assess their relative impact on the kinetics of dilute vortex matter at low temperatures . |
scanning probe microscopy ( spm ) is nowadays an essential technique to measure local electronic properties of mesoscopic structures .
scanning tunnel microscopy ( stm ) , consisting in probing the sample surface with a metallic tip , is the most popular among all spm techniques .
stm experiments have first enabled the detection of standing wave pattern in confined surface electron systems such as quantum corrals.@xcite the physical interpretation of such images is immediate since they derive from direct tunneling between the surface electrons and the tip .
hence , stm images represent the density of states of the system at a given position of the scanning probe.@xcite later on , another spm technique , named scanning gate microscopy ( sgm ) , has been introduced in order to obtain similar information for structures buried under an insulating layer .
this technique consists in scanning a biased tip over the sample surface .
the gate acts as a local electrostatic ( repulsive or attractive ) potential on the electronic system and allows to obtain two - dimensional ( 2d ) conductance ( or resistance ) images of the scanned area as a function of the tip position . at the present time ,
sgm or an alternative technique called scanning capacitance microscopy ( scm ) have been adopted to investigate the physics of quantum points contacts,@xcite quantum dots,@xcite carbon nanotubes,@xcite open billiards@xcite and edge states in the integer quantum hall regime.@xcite sgm on inas nanowires has evidenced the presence of multiple quantum dots inside the structure corresponding to circular coulomb blockade peaks in the conductance plots.@xcite from a theoretical point of view , numerical simulations based on the keldysh - green s functions formalism have been exploited to demonstrate wavefunction imaging in open quantum dots@xcite and to associate conductance images to electron flow in quantum point contacts.@xcite recently , isophase lines for electrons in an electrostatic aharonov - bohm ( ab ) experiment@xcite and local - density - of - states ( ldos ) mapping inside a coherent ab ring@xcite have been reported . in the latter work ,
both experimental curves and numerical simulations have found a linear dependence of the conductance variation on the tip voltage .
interestingly , clear wavefunction images were obtained only in this case,@xcite suggesting to limit the imaging technique to the linear - response regime .
it was also noticed that the simulated conductance maps were not significantly influenced by the particular shape adopted to mimic the effective perturbing potential due to the scanning gate .
since sgm is essentially an indirect measurement of the 2d electron gas ( 2deg ) properties , a physical interpretation of conductance images is not immediate .
therefore , in this paper we try to clarify the physical meaning of sgm experiments and isolate the experimental conditions under which the linear response - regime holds .
in section [ s2 ] we present an analytical model which provides a simple physical interpretation of the sgm conductance images in the single - channel transmission regime .
we find that the conductance correction due to the scanning tip is related to the hilbert transform of the local density of states ( ldos ) of the system .
moreover , we analyze the spatial and energetic conditions on the perturbing potential under which this direct relation is maintained . in section [ s3 ]
we present 2d numerical simulations of a realistic quantum ring characterized by the presence of randomly localized impurities .
our calculations are based on a recursive green s functions method and illustrate the correspondence between ldos and conductance images of the system in such a complex case .
in particular , we address the interesting case for which the wavefunction of the system is dominated by localized states due to charged impurities or by recursive semi - classical orbits .
section [ s4 ] summarizes the relevant results .
in this section , we evaluate the effect of a local perturbation on the total transmission of a non - interacting system connected to single - channel leads .
we adopt the landauer - bttiker transport theory assuming the zero
temperature limit.@xcite we consider a multi - states system connected to one - dimensional ( 1d ) leads and neglect the effect of inelastic scattering and electron - electron interactions .
these assumptions are valid in the case of low - temperature experiments on mesoscopic samples working in the phase coherent transport regime.@xcite we model our system as a chain of @xmath1 sites with the @xmath2-th on - site potential @xmath3 coupled to two semi - infinite reservoirs with fixed chemical potentials . the coupling strength with the left and the right reservoir is denoted by @xmath4 , where @xmath5 is the coupling with the leads and @xmath6 is the density of states of the reservoirs , respectively .
moreover , each site is coupled only to its nearest neighbor with coupling constant @xmath7 , which for simplicity we assume not to depend on the position . since the linear conductance of a system with fermi energy @xmath8
is determined by the propagator from site 1 to site @xmath1 , we simply need to evaluate the element @xmath9 of the retarded green s function of the system @xmath10 . by adopting the common assumption of a large band - width coupling with the reservoirs
, the conductance @xmath11 reads @xmath12 where @xmath13 is the quantum of conductance.@xcite further , we will use the well known fact that the ldos at @xmath14 is proportional to the imaginary part of the diagonal elements of the retarded green s function @xmath15 as @xmath16 , \label{ldos}\ ] ] which derives from the pole structure of the green s function.@xcite in order to relate the conductance in eq .
( [ trans ] ) and the ldos in eq .
( [ ldos ] ) we exploit the dyson equation @xmath17 where @xmath18 is the unperturbed green s function of the isolated sites and @xmath19 corresponds to the hopping connection between sites . from eq .
( [ dyson ] ) one can obtain the relation between the propagator @xmath20 and @xmath21 or @xmath22 as@xcite @xmath23 hence , by recursively applying eq .
( [ dyson2 ] ) to express @xmath24 in terms of @xmath21 , we can rewrite eq .
( [ trans ] ) as @xmath25 where the index @xmath2 can have any value between 1 and @xmath1 .
further , since we are interested in studying the stationary regime of transport , it is convenient to write the green s functions in the energy space representation by performing a fourier transform .
we have @xmath26^{-1}$ ] and @xmath27^{-1}=[(g^{\rm r}_{i , i})^{-1}-\lambda_i+i\gamma_i]^{-1}$ ] , where @xmath28 is a positive infinitesimal number , @xmath29 is the self - energy including the interactions with the system on the local @xmath30th green s function.@xcite we now consider the effect on the transmission of a local perturbing potential which can be physically associated with the action of a charged tip scanning the system . we mimic the effect of the charged tip at position @xmath14 as a renormalization of the on - site energy @xmath31 , where @xmath32 is the perturbing potential . assuming that the potential strength is sufficiently `` weak '' , the conductance correction is linear in @xmath32 and reads @xmath33 .
\label{corr}\ ] ] by comparing eq .
( [ corr ] ) with eq .
( [ ldos ] ) we find that the conductance variation due to the action of a local perturbing potential and the ldos are related in the same way as the real and the imaginary part of the green s function @xmath34 , namely by a kramers - kronig relation .
mathematically , this is expressed in terms of the hilbert transform @xmath35= { \cal h } \ { { \rm i m } \left [ g_{i , i}^{\rm r } ( \omega ) \right ] \ } = \frac{1}{\pi } { \cal p } \int d\omega ' \frac{1}{\omega-\omega'}{\rm i m } \left [ g_{i , i}^{\rm r}(\omega')\right]$ ] , where @xmath36 stands for the cauchy principal part .
therefore , in the linear - response regime , the conductance ( transmission ) correction is proportional to the hilbert transform of the local density of states .
notice that the sign of the conductance correction is determined by both the perturbing potential @xmath32 and by the relative position of the fermi energy with respect to @xmath37 .
the physical origin of such a relationship can be understood by noticing that the main effect of the local perturbing potential @xmath32 is to renormalize the real part of the self - energy . by analyzing eq .
( [ corr ] ) we can deduce the conditions of validity of the ldos-@xmath0 correspondence . in order to deal with a `` weak '' perturbation giving rise to a linear correction
the condition @xmath38 has to hold . a condition on the extension of the perturbing potential
is given by the spatial periodicity of the wavefunction and limits the spatial range of the effective potential below the half - fermi wavelength @xmath39 .
nevertheless , notice that in the multichannel transmission case that we discuss in section [ s3 ] , the spatial periodicity of the ldos can be larger than @xmath39 .
interestingly , in the single - channel transmission regime the ldos is usually a sinusoidal function of the spatial coordinate , whereas the hilbert transform of such a function alters the original function by only a @xmath40 phase shift .
more precisely , we have that the hilbert transform adds a @xmath41 phase for right - going states and @xmath42 for left - going states .
this justifies the use of a scanning local perturbation , like that obtained with sgm experiments , as an efficient tool to image the ldos of the system . for a double - barrier system .
the distance between the two barriers of height 0.1 ev is @xmath43 nm , the effective mass is @xmath44 with @xmath45 the free electron mass and the fermi energy is @xmath46 ev .
the perturbing potential used to obtain ( b ) is a square repulsive ( red line ) or attractive ( blue line ) barrier with height @xmath47 and width @xmath48 nm . dashed vertical lines indicate the maximum position of ldos ( black line ) and @xmath0 ( red or blue line).,width=307 ] in order to give a practical application of the model we study the simple case of a symmetric double barrier . in fig .
[ figure1 ] we show the ldos of the unperturbed system and the conductance correction due to a scanning potential acting as a local scatterer .
the conductance is computed by means of eq .
( [ trans ] ) after numerical calculation of the total green s function .
the local perturbing potential is assumed to be a rectangular - shaped barrier with height @xmath49 ( or @xmath50 ) and width @xmath48 nm smaller than @xmath51 nm .
let us discuss the @xmath0-ldos correspondence separately for the two regions inside and outside the two barriers .
we model the two barriers by using the same scattering matrix @xmath52 which can be parametrized as @xmath53 where @xmath54 and @xmath55 are real phases ( @xmath56 if time - reversal symmetry holds ) and @xmath57 , @xmath58 , with @xmath59 transmission probability . in order to compute the ldos inside the two barriers we evaluate the wavefunction for left- and right - injected states by imposing the proper boundary conditions .
we obtain @xmath60 $ ] and @xmath61 $ ] , with @xmath62 the fermi wavevector and @xmath63 the distance between the two barriers . by estimating the ldos of the system as the independent sum of left - injected and right - injected states we obtain @xmath64 $ ] .
the hilbert transform of this function adds a @xmath41 phase shift to the partial ldos corresponding to the left - injected states and a @xmath42 phase shift to the partial ldos corresponding to the right - injected states .
it reads @xmath65 $ ] and it has the same spatial dependence of @xmath66 . an identical result is numerically recovered in fig . [ figure1 ] by computing the conductance variation due to a square perturbing potential .
notice that a further sign on the conductance variation can be determined by the choice of an attractive or repulsive potential .
the effect of the local scatterer is different in the spatial region outside the two barriers . on the left ( right ) of the barriers
the ldos shape is determined only by left ( right)-injected states which gives @xmath67 .
in fact on the left ( right ) of the barrier the right ( left)-injected states contribute as a constant to the ldos .
the hilbert transform of such a function gives the same expression with a @xmath68 phase shift for right- and -left - injected states .
again , this is the result found numerically in fig .
[ figure1 ] .
finally , we address the case the fermi energy equals @xmath37 it turns out that @xmath69=0 $ ] and the system is `` on resonance '' .
this condition gives a maximum of the transmission probability .
in such a case eq .
( [ corr ] ) vanishes and the correction to the transmission is given by the second order in @xmath32 .
it reads @xmath70 therefore , in the resonant case the correction to the transmission is a monotonic function of the local density of states @xmath71 and has a fixed sign .
we focus now on the simulation of an ab ring since such a structure is particularly suited to study phase coherence effects in mesoscopic physics.@xcite further , we consider a ring obtained from an ingaas / inalas heterostructure@xcite with @xmath72 nm for the width openings and @xmath73 nm and @xmath74 nm for the inner and outer radius , respectively .
the effective mass of carriers is assumed to be @xmath44 with @xmath45 the free electron mass .
for the present case we analyze an ideal ring without the presence of any defect .
we adopt a tight - binding model to describe the hamiltonian of the system.@xcite the numerical technique is based on the calculation of the retarded green s function of the system by using a recursive technique.@xcite the method is quite general and allows to treat the possible influence of spin - orbit coupling on the system.@xcite the presence of an external magnetic field can be modeled by adding peierl s phases to the hopping elements of the hamiltonian.@xcite nm for the width openings and @xmath73 nm and @xmath74 nm for the inner and the outer radius , respectively .
we have adopted a fermi energy @xmath75 mev in ( a ) , ( b ) , ( c ) , and @xmath76 mev in ( d ) , ( e ) , ( f ) , whereas a maximum for the transmission energy stays at @xmath77 mev .
therefore , figures on the left stay below and figures on the right stay above a resonance . for the first - line couple ( a ) , ( d ) the perturbing potential mimicking the effect of the charged tip
is modeled as @xmath78 ^ 2\}$ ] with @xmath79 and @xmath80 nm . for the couple ( b ) , ( e ) and the couple ( c ) , ( f ) we have adopted the same asymmetric perturbing potential @xmath81 ^ 2+[2(x - y)/\sigma]^2\}$ ] with @xmath79 , but with different spatial extension determined by @xmath80 nm and @xmath82 nm , respectively . ]
we study the conductance variation of such an ab ring under the effect of a superimposed perturbing potential @xmath83 . in figs .
[ figure2](a ) , [ figure2](d ) we have first assumed a long - range lorentzian function @xmath78 ^ 2\}$ ] , where @xmath84 , the coordinates @xmath85 and @xmath86 have the tip position as origin , @xmath87 is the maximum potential perturbation and @xmath88 determines the decay rate of the potential . notice that the spatial extension of such a potential is larger than @xmath89 . however , since recent experimental results on open quantum dots have pointed out that , in some cases , a significant spatial asymmetry characterizes the tip - induced potential,@xcite we have also adopted an asymmetric shape for @xmath83 . in figs .
[ figure2](b ) , [ figure2](e ) and figs .
[ figure2](c ) , [ figure2](f ) we have assumed the potential @xmath90 ^ 2+[2(x - y)/\sigma]^2\}$ ] with @xmath80 nm and @xmath82 nm , respectively .
conductance maps on the left side of fig .
[ figure2 ] have a fermi energy slightly smaller than a resonance point for which the ring has maximal transmission @xmath91 , whereas on the right side they have a fermi energy slightly larger than the resonance
. since both energies are close each other , the mapped conductances are very similar and differ only for having opposite signs in complete agreement with eq .
( [ corr ] ) .
here , we want to emphasize that , even if the correct potential shape of the perturbing potential should be obtained by computing the electrostatics of the total system , still the conductance variation does not depend on the particular shape of the potential in the linear - response regime under study.@xcite concerning the effect of an asymmetric tip - induced potential we observe that the small asymmetry in figs .
[ figure2](b ) , [ figure2](e ) can be considered negligible with respect to figs .
[ figure2](a ) , [ figure2](b ) , obtained with a symmetric potential . on the contrary
, it is relevant in figs .
[ figure2](c ) , [ figure2](f ) where the spatial extension of the perturbing potential become larger than @xmath92 nm giving rise to a non - linear behavior of the conductance .
we conclude that the effect of an asymmetric potential can be minimized by providing a tip - induced perturbation with limited spatial extension .
in this section we extend the results obtained for the single - channel transmission to the case the conductance is the sum of several independent channels as expressed by the landauer - bttiker formula @xmath93 , where @xmath94 is the transmission probability of the @xmath95-th conducting transverse mode .
this is the most common experimental situation .
@xcite in particular , we want to verify whenever the ldos-@xmath0 correspondence depicted in section [ s2 ] is valid for realistic 2d structures .
a first complication arises from the spatial resolution of the scanning tip . indeed ,
in the multi - channel transmission case the ldos structure is usually very complex and presents a spatial periodicity related to half the fermi wavelength , whereas in experiments the extension of the perturbing potential is typically larger than @xmath39 .
thus , one can expect that the spatial resolution of the tip - induced potential is smaller than the spatial extension of conducting modes related to lowest energies , but larger than those related to highest energies .
this would lead to properly take into account only conducting modes corresponding to the lowest energies and lose a satisfactory correspondence with the ldos .
another problem concerns the interpretation of conductance maps in terms of hilbert transform of the ldos .
generally , in the multi - channel transmission case it is not possible to obtain a direct correspondence like in eq .
( [ corr ] ) or ( [ res ] ) since the effect of the local perturbing potential due to the scanning tip is different for each conducting channels contributing to the conductance .
however , it often occurs that the ldos and hence the conductance are dominated by one or few close transverse modes . in these specific cases the same ldos-@xmath0 correspondence for the one - channel case
is recovered provided that the conditions found in section [ s2 ] are valid .
this means that the spatial extension of the tip - induced potential has to be smaller than the spatial periodicity of the dominant channel _ and _ the potential height @xmath87 has to be smaller than the corresponding energy level .
2.5 v and ( b ) averaged conductance variation @xmath96 as a function of the tip voltage @xmath97 measured in different device regions indicated in ( a ) . a clear linear dependence is present only in the region i inside the ring .
the 2deg distance of the tip is @xmath98= 50 nm , @xmath99 k and a magnetic field of @xmath100 t is applied perpendicularly to the surface.,width=307 ] from an experimental point of view we demonstrate the feasibility of the linear - response regime in fig . [
figure3](b ) , where an averaged conductance variation as a function of the tip voltage is shown for a coherent quantum ring .
this is fabricated from a ingaas / inalas heterostructure using electron beam lithography and wet etching .
the 2deg is located 25 nm below the surface , and its low temperature ( @xmath59 ) electron density and mobility are @xmath101 m@xmath102 and 10 m@xmath103/vs , respectively .
the ring inner and outer ( lithographic ) radii are 120 nm and 290 nm ( this sample is the same as sample r1 in ref . ) .
notice that a clear linear dependence on the tip voltage is experimentally found only if the tip scans the region i inside the ring , where the former conditions are expected to hold . in summary ,
specific situations exist such that the total wavefunction of the system adopts a simple and regular shape and the conditions for a satisfactory ldos-@xmath0 correspondence are recovered . in the following of this section we give a few relevant examples and illustrate them with experimental data and numerical simulations .
we first focus on a realistic ab ring with the same sizes as the one studied in section [ s2 ] , but with a random distribution of symmetry - breaking defects .
these defects are probably responsible for the asymmetric conductance images in ref . , where the wavefunction inside the ring is strongly influenced by localized states . in fig .
[ figure4 ] we display a typical conductance map for the same ring of fig . [ figure3 ] .
we notice that the conductance map inside the ring is characterized by radial asymmetric fringes in both fig .
[ figure3](a ) and fig .
[ figure4 ] . .
the dashed line indicates a depletion region at the border of the ring . @xmath97=0.5
v , @xmath98= 50 nm , @xmath104 k and @xmath105 t. , width=307 ] here , we model the defects by superimposing a randomly distributed analytical potential @xmath106 to the structure profile .
this takes into account the electrostatic effects of defects with charge @xmath107 at distance @xmath108 from the 2deg and reads @xmath109 } \ , , \
] ] with @xmath110 and @xmath111.@xcite in our simulations we have adopted a random distribution with a 90 @xmath112m@xmath102 numerical density at a fixed distance of 20 nm from the 2deg . in fig . [ figure5 ]
we show the impurity - induced potential profile used in our calculations .
nm , @xmath73 nm and @xmath74 nm ) .
these impurities are mainly localized around the border of the ring.,width=307 ] further , we have chosen a fermi energy range for which more than ten conducting channels are present in the openings and more than one hundred inside the ring .
the first favorable situation occurs when the ldos is determined by the position of these defects . by computing the ldos of the ab ring with the addition of randomly distributed impurities we have found several images where the squared wavefunction of the system is dominated by such localized states .
nm , @xmath73 nm and @xmath74 nm ) with randomly distributed impurities .
the presence of localized states is clearly visible in both of them .
the perturbing potential mimicking the effect of the charged tip is modeled as the long range potential @xmath78 ^ 2\}$ ] with @xmath113 and @xmath114 nm .
other parameters are @xmath115 ev and @xmath116=0.,width=307 ] as an example , in fig .
[ figure6 ] we show the ldos of the ab ring with the presence of randomly distributed impurities for one of such specific energies , as well as the corresponding conductance variation due to a lorentzian perturbing potential . in this case
the ldos and the conductance variation are characterized by radial asymmetric fringes similar to those of fig .
[ figure4 ] .
even if the effective spatial extension of the tip - induced potential is larger than @xmath117 nm the conductance map accurately reproduces the ldos shape due to the large spatial extension of the localized states .
another favorable situation occurs in the presence of semi - classical orbits inside the mesoscopic structure.@xcite they correspond to quasi - classical states that scar the total wavefunction of the system and recur periodically as the fermi energy or an external magnetic field is varied.@xcite nm,@xmath73 nm and @xmath74 nm ) with randomly distributed impurities . in the inset the fourier transform is plotted.,width=307 ] in fig .
[ figure7 ] we plot the conductance of the ab ring with impurities as a function of the fermi energy as well as its fourier transform ( inset ) . in the latter plot it is possible to identify a sharp maximum at a frequency of 400 1/ev corresponding to the occurrence of a dominant pattern in the ldos .
such a quasi - classical orbit is plotted in fig .
[ figure8](a ) and was found to recur with a frequency of 400 1/ev . also in this case the large and uniform spatial extension of the wavefunction permits to recover a clear correspondence with the conductance map shown in fig . [ figure8](b ) .
nm , @xmath73 nm and @xmath74 nm ) with the presence of randomly distributed impurities .
semi - classical orbits are well visible in both .
the perturbing potential mimicking the effect of the charged tip is modeled as a the long range potential @xmath78 ^ 2\}$ ] with @xmath118 ev , @xmath49 and @xmath80 nm.,width=307 ]
we have presented a theoretical analysis of scanning gate microscopy as an experimental tool to image the ldos of mesoscopic samples fabricated from buried 2d electron systems like aharonov - bohm rings .
we have shown that in the single - channel transmission case a generalized kramers - kronig relation holds between the ldos of the system and the conductance variation due to the scanning tip .
such relationship is valid in the linear - response regime where the tip induced potential is so weak that the energy levels of the system are unchanged . in this case
we have shown a striking correspondence between ldos and @xmath0 images for a symmetric double - barrier system and an ab ring . in the multi - channel transmission case
an exact ldos-@xmath0 correspondence is less straightforward due to the deficient spatial resolution of the tip - induced potential even if the perturbing potential is smaller than the energy levels of the system .
however , in some special cases for which the total wavefunction presents a large and uniform spatial extension , such correspondence can still be recovered .
in particular , we have shown a nice agreement between ldos and conductance images when the ldos is dominated by localized states or by semi - classical periodic orbits responsible for wavefunction scarring effects .
we would like to thank m. governale for useful discussions and suggestions .
this work is supported by the _ pnano 2007 _ program of the _ agence nationale de la recherche _ ( `` micatec '' project ) .
b. h. acknowledges funding by the fnrs ( belgium ) . | we study the relationship between the local density of states ( ldos ) and the conductance variation @xmath0 in scanning - gate - microscopy experiments on mesoscopic structures as a charged tip scans above the sample surface .
we present an analytical model showing that in the linear - response regime the conductance shift @xmath0 is proportional to the hilbert transform of the ldos and hence a generalized kramers - kronig relation holds between ldos and @xmath0 .
we analyze the physical conditions for the validity of this relationship both for one - dimensional and two - dimensional systems when several channels contribute to the transport .
we focus on realistic aharonov - bohm rings including a random distribution of impurities and analyze the ldos-@xmath0 correspondence by means of exact numerical simulations , when localized states or semi - classical orbits characterize the wavefunction of the system . |
kepler-22b is a recently validated extrasolar planet detected by the _ kepler mission _ via the transit technique ( @xcite , b12 hereafter ) . orbiting a star with 75% of solar luminosity once every 290days
, kepler-22b receives an insolation just 10 - 15% greater than that received by the earth and thus was claimed to be the first transiting planet discovered to orbit within the so - called habitable - zone of its parent sun ( b12 ) .
although habitable - zone exoplanets have been detected by radial velocity surveys @xcite , kepler-22b is a fascinating object thanks to the plethora of follow - up opportunities afforded by transiting planets @xcite .
this excitement is evident in the preceding astronomical literature with speculation about the planet s habitability @xcite , a presently undetected exomoon s habitability @xcite and considerable efforts to redefine the so - called habitable - zone ( e.g. @xcite ) . despite the excitement generated by the pioneering discovery of kepler-22b by b12
, the planet can not be considered `` earth - like '' given its radius of 2.4@xmath0 .
observationally , the composition of the planet is essentially unconstrained since the upper limit on the planetary mass is quite broad at @xmath7@xmath2 ( 2@xmath8 ) ( b12 ) .
the most well - characterized planet with a radius similar to kepler-22b would seem to be gj1214b @xcite with a radius of 2.7@xmath0 , which has a low bulk density indicating that it can not have a rock - dominated composition @xcite .
although kepler-22b may not be a rocky habitable earth - like world , there is a distinct possibility for the system to yet maintain such a world if kepler-22bpossesses a large moon .
recently , there has been considerable speculation about the potential habitability of such a moon @xcite since it is thought that exomoons should be detectable around transiting planets using kepler @xcite . despite this speculation ,
we note that , to our knowledge , no - one has ever conducted a search for an exomoon in this system .
indeed , we note that there has never even been a search for an exomoon around a habitable - zone _ candidate _
planet , let alone a validated one , in the present literature .
the `` hunt for exomoons with kepler '' ( hek ) project @xcite is presently the only systematic program we are aware of attempting to observationally identify extrasolar satellites . searching for such signals is computationally challenging due to the highly multimodal ,
complex and high - dimensional parameter space one is faced with @xcite , the need for calculating the bayesian evidence integral @xcite and the photodynamic forward modeling required to simulate exomoon signals @xcite . to provide some context
, the light curve fits presented in this work required 49.7years of cpu time . to date , eight kepler objects of interest ( kois ) have been analyzed for evidence of extrasolar moons each with null detections leading to upper limits on the satellite - to - planet mass ratio of @xmath9-@xmath10% @xcite . in this work
, we present an analysis of the first habitable - zone planet by hek .
this target was identified as a target - of - opportunity by the project because it i ) is in the habitable - zone ii ) is a validated planet iii ) has radial velocity measurements iv ) has a quiet , bright ( @xmath11 ) host star v ) has a host star with asteroseismology constraints ( b12 ) . in this work ,
we include several new modes to thoroughly explore the exomoon parameter space .
these include retrograde ( [ sub : retro ] ) and eccentric moon solutions ( [ sub : eccentricity ] ) , informative and uninformative limb darkening priors ( [ sub : ldpriors ] ) , applying bayesian model averaging ( [ sub : bma ] ) , high - resolution fitting ( [ sub : hires ] ) , leveraging the radial velocities ( [ sub : rvs ] ) and asteroseismology constraints ( [ sub : astero ] ) and injecting and retrieving synthetic moon signals ( [ sec : injected ] ) .
we also take this opportunity to provide updated constraints on the mass , radius , composition , habitability and orbit of kepler-22b using new _ kepler _ data ( [ sec : planetonly ] ) .
in the discovery paper of kepler-22b ( b12 ) , three transits were detected by the _ kepler mission _ occurring in quarters 1 , 4 and 7 ( q1 , q4 and q7 ) . since this time
, an additional three transits should have been observed by _ kepler _ : a @xmath12 transit in q11 , a @xmath13 transit in q14 , and a @xmath14 transit in q17 .
the @xmath12 transit was successfully recorded but unfortunately the @xmath13 was not due to a data gap in the q14 time series .
the @xmath14 transit is unlikely to have been recorded because it occurred during a safe mode event from may @xmath15 to may @xmath14 2013 .
further , soon after this _ kepler _ lost functionality of a second reaction wheel on may @xmath16 , putting future science observations in doubt .
therefore , it is quite possible that the four transits of kepler-22b analyzed in this paper will be the only transits ever observed by _
kepler_. the first transit was observed in long - cadence ( lc ) mode only , whilst the latter three have short - cadence ( sc ) data .
we downloaded the reduced data from the mikulski archive for space telescopes ( mast ) . in this work , we always make use of the `` raw '' ( labelled as `` sap_flux '' in the header ) data processed by the data analysis working group ( dawg ) pipeline ( see accompanying data release notes for details ) .
the `` raw '' data has been processed using pa ( photometric analysis ) , which includes cleaning of cosmic ray hits , argabrightenings , removal of background flux , aperture photometry and computation of centroid positions . for the sake of brevity
, we do not reproduce the details of the pa pipeline here , but direct those interested to @xcite and the data release handbooks .
the _ kepler _ photometry contains several systematic effects which require detrending before a precise transit light curve analysis can be conducted .
these effects can be instrumental ( such as focus drift , pointing tweaks , safe modes , etc ) or astrophysical ( such as flaring , rotational modulations , etc ) . in this work
, we utilize the cofiam ( cosine filtering with autocorrelation minimization ) algorithm described in @xcite for detrending these various effects .
cofiam can be thought of as a high - pass , low - cut periodic filter optimized to undisturb all periodicities at or below the so - called `` protected timescale '' , @xmath17 .
the algorithm builds upon initial applications of cosine filtering by @xcite for corot photometry and @xcite for _ kepler _ photometry .
cofiam regresses the following sum of harmonic functions to the time series : @xmath18 , \label{eqn : linearcosinefilter}\end{aligned}\ ] ] where @xmath19 is the total baseline of the data under analysis , @xmath20 are the time stamps of the data , @xmath21 & @xmath22 are model variables and @xmath23 is the highest harmonic order . for
any given @xmath17 , one may define the maximum number of distinct harmonic cosines functions ( @xmath24 ) to regress to the data without disturbing the timescale @xmath17 as : @xmath25 however , choosing any @xmath26 will also protect the timescale @xmath17 . exploiting this fact , cofiam explores every possible variation ( in total there are @xmath27 variations ) . the detrended light curve after each cofiam regression
is trimmed to within twice the `` window '' timescale ( @xmath28 ) of the predicted transit time and then a final linear slope is fitted through the data ( excluding the transit itself ) to serve as a final normalization .
after every detrending , we compute the autocorrelation on a 30minute timescale using the durbin - watson metric : @xmath29 where @xmath30 are the residuals and @xmath31 is the number of data points .
the value of @xmath32 always lies between 0 and 4 , with 2 representing an absence of autocorrelation , @xmath33 implying otherwise .
cofiam therefore selects the value of @xmath23 which minimizes @xmath34 ( hence `` autocorrelation minimization '' ) .
cofiam is applied to each transit epoch individually and the inputted data is pre - screened for any sharp , discontinuous offsets ( due to say a pointing tweak ) as well as a general outlier rejection phase using median filtering .
we never attempt to stitch two quarters together since there is no reason to expect the detrending function to be smooth between rolls of the spacecraft ( which occur every quarter ) . in this work , we chose @xmath35 where @xmath36 is the first - to - fourth contact transit duration .
@xcite showed that @xmath36 is the lowest periodicity in the fourier transform of a transit and so we multiply this by three to provide a small cushion .
the `` window '' timescale is defined to be @xmath37 $ ] ( the 1.2 factor is again a cushioning factor ) where @xmath38 is the hill sphere timescale . for this
we define @xmath39^{1/3 } [ p_{b*}/(2\pi)]$ ] where @xmath40 is the planet s orbital period , @xmath41 is the planet s mass and @xmath42 is the stellar mass . for all of these quantities we simply used the best - fit values quoted in b12 ( for @xmath41 we used the 3@xmath8 upper mass limit ) .
we stress that the detrending function is regressed independently of the later transit fits .
this is done because the form of the detrending function is varied and we typically try between @xmath43-@xmath44 different models per transit . to perform the detrending in conjunction with the transits
would thus require around @xmath45 unique models i.e. @xmath46 $ ] for kepler-22b .
given the very costly compuational demands of even a single photodynamical fit , this would be unrealistic with current computational capabilities .
after detrending with cofiam , we find @xmath47 , @xmath48 , @xmath49 and @xmath50 and the detrending functions are plotted over the pa data in figure [ fig : qaraw ] . + +
we first fit and model the light curve as being due to a single planet transiting the host star , to serve as a benchmark for the subsequent planet - with - moon fits .
the planet - only light curve is modeled as was done in @xcite , using the popular @xcite algorithm .
we assume a quadratic limb darkening law for the host star , where the specific intensity is described by @xmath51 where @xmath52 , @xmath53 is the normalized radial coordinate on the stellar disk and @xmath54 & @xmath55 are the limb darkening coefficients .
planet - with - moon models are generated using the photodynamic luna algorithm @xcite , which inherently accounts for the various timing effects on the planet induced by a moon @xcite .
we direct the reader to @xcite for details on how the photodynamical light curves are computed .
model light curves are fitted to the data using the multimodal nested sampling algorithm multinest @xcite . by using this code ,
we not only obtain parameter posteriors but also the bayesian evidence of each model attempted , thus allowing for later bayesian model selection and model averaged posteriors .
bayesian model selection is a crucial tool in detecting exomoons due to the roughly doubling of free parameters combined with the complex parameter space @xcite . in general
, the methods described above are the same as those implemented in our previous hek surveys ( e.g. see @xcite ) .
notably , the choice of priors follows that of @xcite and are shown in table [ tab : priors ] .
however , in studying kepler-22b , we introduced several new changes to our methodology to improve both the accuracy of our derived results and the speed of the computations .
these modifications are discussed in what follows . in this work ,
we use two different priors for the limb darkening coefficients : i ) informative priors ii ) uninformative priors .
the informative priors are computed using a monte carlo forward modeling of stellar atmosphere models .
first , we draw a random normal variate from the effective temperature ( @xmath56 ) and stellar surface gravity ( @xmath57 ) and then compute the associated quadratic limb darkening coefficients using the @xcite stellar atmosphere model database convolved with the _ kepler _ bandpass ( see @xcite for more details ) .
we repeat this process until @xmath58 fair realizations of @xmath54 and @xmath55 have been computed . for @xmath56 and @xmath57
, we used the quoted values from b12 of @xmath59k and @xmath60 but doubled the uncertainties for both . the @xmath58 realizations of @xmath54-@xmath55 form a joint prior probability for the limb darkening coefficients .
for computational expedience , it is desirable to characterize this joint prior probability with a simple analytic form , such as a bivariate gaussian , rather than calling a stellar atmosphere model at every realization in the light curve fits .
however , a bivariate gaussian is not ideal since our @xmath54-@xmath55 joint prior probability displays a strong covariance ( correlation coefficient of -0.97 ) . despite this , a bivariate gaussian can be applied by re - parameterizing the limb darkening coefficients . to this end , we perform a principal components analysis ( pca ) on the joint prior probability of @xmath54-@xmath55 and re - parameterize the terms into the orthogonal components @xmath61 and @xmath62 following the suggestion of @xcite : @xmath63 where we derived @xmath64 from the pca . using this transformation , we computed a joint prior probability for @xmath61-@xmath62 with negligible covariance . the lack of covariance , combined with the fact the resulting distributions are unimodal and approximately symmetric means that a bivariate gaussian is now a reasonable analytic approximation . regressing a gaussian to each cumulative density function ( cdf )
, we derive normal priors of @xmath65 and @xmath66 . in figure
[ fig : u1u2prior ] , we compare the pdf in @xmath54 and @xmath55 predicted by these priors relative to the actual prior probability initially generated , where one can see the good agreement . the second type of prior we try is an uninformative one .
@xcite recently showed that uniform priors in @xmath54-@xmath55 can be drawn , without ever sampling unphysical limb darkening coefficients , by applying the `` triangular sampling '' technique .
this is achieved by re - parameterizing the coefficients to @xmath67-@xmath68 and sampling uniformly between zero and unity in each , where @xmath69 employing triangular sampling is approximately twice as efficient than the uninformative priors used previously in @xcite .
this is because previously we used uniform priors in @xmath70 and @xmath71 , which spends exactly half of its time sampling from unphysical ( and thus rejected ) parameter ranges @xcite . in principle
, cofiam should provide a light curve with an out - of - transit baseline flux equal to exactly unity . in practice ,
even with cofiam s final stage linear normalization , it is prudent to regress a baseline normalization factor for each transit epoch simultaneous to the transit parameters .
one major benefit of doing so is that the uncertainty of the baseline normalization is propagated into the uncertainty of the other transit parameters . in the past
, we have treated the normalization factor as a free parameter in multinest like any other term . for planets with many transits
epochs , this is problematic since the regression now includes a very large number of free parameters leading to laborious cpu times .
the fact that a relatively minor nuisance parameter causes this dramatic slow - down makes the situation somewhat impractical . in this work , we introduce a new refinement to the hek fitting strategy which simply treats the baseline parameters ( @xmath72 ) as a set of nuisance parameters . rather than marginalizing over these terms , every realization of the transit model simply optimizes the baseline parameters to the data .
this is easily achieved using a weighted linear minimization which is both extremely quick and guaranteed to always find the global minimum . in principle , one should wish to marginalize over nuisance parameters rather than simply maximize the likelihood , but in practice we find this procedure has little effect on the other transit parameters and leads to a significant speed - up .
unlike previous surveys , we decided to investigate the potential for eccentric orbits for both the planet and the moon due to the high - priority nature of kepler-22b .
eccentric fits always require dramatically more time to explore due to i ) the introduction of two new free parameters ii ) the requirement to solve kepler s transcendental equation @xmath73trillions of times during the course of a single fit .
eccentric orbits can also produce unphysical scenarios , unless limits are enforced on the allowed range in @xmath74 ( eccentricity of the planet - satellite barycentre around the host star ) and @xmath75 ( eccentricity of the satellite around the planet - satellite barycentre ) .
first of all , parameter realizations reproducing collisions should be avoided .
we therefore enforce that @xmath76 , i.e. the periastron separation between the planet and the satellite exceeds two planetary radii ( we use two planetary radii since the satellite s radius is @xmath77 ) .
we use two planetary radii because the satellite s radius can be up to one planetary radius in our fits .
similarly , for the planetary orbits , we enforce @xmath78 .
additionally , @xcite showed that the maximum stable planet - satellite separation is attenuated for eccentric orbits .
@xcite showed that all satellites must conform to : @xmath79 where @xmath80 is equal to the semi - major axis of the satellite relative to the planet in units of the hill radius . under the assumption that @xmath81 , @xcite showed that @xmath82 , where @xmath83 is the period of the satellite around the planet - satellite barycentre .
this allows us to enforce the @xcite condition as : @xmath84 these constraints are optimistic in the sense that the orbits are assumed to be coplanar and the moon is simply a test particle .
more realistic solutions can be found by exploring the three - body solution with non - zero masses and three - dimensional geometry ( e.g. see @xcite ) , but we reserve these more detailed dynamical investigations for the final posterior analyses .
we also enforce the same limits on the planet and satellite densities used in @xcite .
these various limits are all imposed by simply rejecting any trials reproducing such behavior . due to the computationally much higher cost of exploring eccentric solutions , we only explored eccentric - planet or eccentric - moon solutions but not both simultaneously , as such fits were found to be prohibitively cpu expensive .
since the hill stability of a putative moon decreases strongly with increasing eccentricity for both terms , we do not envisage it likely that this significantly hinders our detection ability .
in @xcite we did not explore retrograde orbits for the exomoon , which luna defines as being when @xmath85 , following the definition used for a planet transiting a star . ] . in this work ,
we present a method to explore such orbits whilst still maintaing an isotropic prior in @xmath86 .
one difficultly in doing this is that an isotropic prior is uniform in @xmath87 but this trigonometric function is not uniquely defined over the interval @xmath88 .
for example , @xmath89 may refer to @xmath90 or @xmath91 .
our resolution to this is to introduce an auxiliary cosine term which we dub @xmath92 and is defined over the interval @xmath93 .
the inverse of this function is given by : @xmath94 & \text{if } -1 < \cos'i_{sb } \leq + 1 , \\
\cos^{-1}[\cos'i_{sb}-2]+\pi & \text{if } + 1 < \cos'i_{sb } \leq + 3 .
\end{cases } \label{eqn : inclination}\ ] ] by enforcing a uniform prior in @xmath93 and inverting the above , we reproduce an isotropic prior in inclination and still explore the full range of inclinations ( i.e. both prograde and retrograde orbits ) .
unlike the previous kois analyzed in survey i @xcite , kepler-22 has asteroseismology constraints available .
consequently , there exists a strong empirical prior on @xmath95 , the mean stellar density of @xmath96g@xmath97 ( b12 ) . in all fits ,
we treat @xmath95 as a normal prior following this measurement .
usually , the light curve fitted eccentricity and stellar density are highly degenerate but the presence of a strong prior on the latter breaks this degeneracy .
this trick was first noted in @xcite and was dubbed `` single - body asterodensity profiling '' by the authors , or sap .
we are therefore able to provide sap - derived eccentricity constraints in the light curve fits provided in this work .
we also note that b12 are able to combine their precise @xmath95 measurement with stellar evolution isochrone models to provide precise estimates of the stellar radius and mass ( @xmath98 and @xmath42 ) .
although these are not model parameters in any our fits , we can use them later to derive physical parameters . as part of the campaign to validate the planetary nature of kepler-22b
, b12 obtained sixteen high resolution spectra of the target star between 17 august 2010 and 25 august 2011 using the hires spectrometer on the keck i 10 m telescope , which we utilize in this work .
the spectra yield @xmath99ms@xmath100 precision radial velocities and show no clear detection of kepler-22b .
the null - detection is consistent with a low - mass planet , as expected for the radius of kepler-22b . by leveraging these radial velocities
, we can constrain various parts of the parameter space explored by multinest , such as excluding massive planets or moderate - mass planets on highly eccentric orbits . including
the rvs requires only two extra free parameters to our model , @xmath101 and @xmath102 ( the radial velocity jitter ) , since all of the other orbital parameters are included in our fits anyway .
the jitter term behaves like an additional error which is simply added in quadrature to the reported rv uncertainties . unlike b12
, we do not choose a fixed @xmath102 value of @xmath103ms@xmath100 but rather we fit the parameter using a modified jeffrey s prior , as advocated by @xcite .
following @xcite , we choose the inflection point of the modified jeffrey s prior to be equal to the median rv uncertainty and set the maximum to be equal to twice the range of the reported rvs .
the same prior is also used for @xmath101 .
the radial velocities can be used to derive the planetary mass , @xmath41 , given a known stellar mass , @xmath42 . to determine @xmath41 , we solve the cubic equation of the well - known mass - function using equation 3.18 of @xcite .
whilst some authors , such as b12 , have advocated exploring negative @xmath101 solutions to avoid positive - biases in fitting joint rv + transit data , this is actually unnecessary for multinest as the nested sampling technique does not suffer boundary condition biases , unlike markov chains .
a new feature we implement in this work for the hek project is the exploration of negative @xmath104 solutions i.e. negative radii moons .
negative radii moons are of course not physically plausible and our implementation simply involves flipping the transit signal component caused by the moon in such cases .
we exploit this trick as a vetting test such that solutions favoring a negative radius moon can be easily dismissed .
this replaces the previous test of @xcite of fitting for zero - radius moons and comparing the bayesian evidence .
we find that negative - radius moon exploration is more efficient computationally since we do not require another additional moon fit purely for vetting purposes .
we select a uniform prior on the radius ratio term of @xmath105 .
._planet - moon parameters used in light curve fits and their associated priors .
@xmath106 is a uniform prior between @xmath107 and @xmath108 .
@xmath109 is a gaussian prior with a mean of @xmath107 and standard deviation @xmath108 .
@xmath110 is a modified jeffrey s prior with a maximum at @xmath108 and an inflection point at @xmath107 .
@xmath111 represents uninformative limb darkening coefficient priors ; informative priors are discussed in
[ sub : ldpriors ] . _ [ cols="^,^ " , ] [ tab : fakeevidence ] as was done previously for kepler-22b itself , we should now investigate whether the posterior distribution allows us to infer the habitability of our injected moon .
given the complex parameter degeneracies and low - to - moderate signal - to - noise of the injected moon signal , it is not immediately clear that we are able to definitively determine the moon s habitability , given the broad and multimodal nature of the moon parameter posteriors .
this is further complicated by the fact that moons possess several extra sources and sinks of thermal energy compared to their host planets .
if their orbit around the planet is elliptical , they can experience tidal heating , which can significantly affect surface temperatures even for small eccentricities @xcite .
they also suffer frequent stellar eclipses by their host planet , which can act as an effective energy sink @xcite .
both of these phenomena are extremely sensitive to the dynamical properties of the planet - moon pair .
as recently demonstrated by @xcite using latitudinal equilibrium balance models ( lebms ) , even the orbital direction of the moon can have consequences for a telluric exomoon s climate .
the habitable zone of an exomoon is therefore a non - trivial multi - parameter manifold , and it is not generally the case that if the host planet is partially or completely uninhabitable , the moon must be also .
we therefore can not rely on the habitable zone calculations of @xcite ( as was done for kepler-22b ) to determine the habitability of the exomoon .
however , this does not preclude the possibility of a robust statistical analysis .
we employ the same lebms as described in @xcite to investigate how well the posteriors allow us to recover the moon s habitability .
the lebm allows us to evolve the moon s surface temperature as a function of latitude assuming it is earth - like in composition . in the simulation ,
the moon is subject to insolation from the host star , tidal heating from the host planet , eclipses by the planet , and infrared atmospheric cooling .
the circulation of heat in the atmosphere is modeled by a diffusion equation ( see also @xcite and @xcite ) . by using each realization of the posteriors as a set of inputs for a lebm simulation
, we can construct a distribution of exomoon climates , which can then be compared to the climate derived from the `` true '' injected parameters .
each lebm simulation is classified in the following fashion @xcite : 1 . _ * habitable moons * _ - these moons possess a habitable surface that covers at least 10% of the total area .
this figure is time - averaged , and the standard deviation in habitable surface over this time is less than 10% of the mean .
* hot moons * _ - these moons display a surface habitability fraction of less than 10% , and typically possess surface temperatures above 373k across all seasons , and are therefore conventionally uninhabitable .
3 . _ * snowball moons * _ - these moons also display surface habitability of less than 10% , and have undergone a snowball transition to a state where the entire moon is frozen , and are therefore conventionally uninhabitable .
* variable moons * _ - these moons are similar to the habitable moons class above , but the standard deviation in habitable surface over time is high , greater than 10% of the mean . the simulation run using the injected signal yields a stable , habitable climate with a mean temperature of approximately 320k .
the moon is on a close to circular orbit , and hence tidal heating does not play a significant role in the resulting climate .
how does this compare to the data retrieved from the posteriors ?
figure [ fig : habhist_lebm ] shows the resulting classification of all posteriors .
around 75% of all simulations run using the posterior data results in a habitable moon .
the next most populous classification is the cold moon case - this is somewhat unsurprising , as the so - called `` snowball '' transition ( generated by a rapid increase in albedo as surface water freezes ) acts as a positive feedback mechanism , making cool climates colder , typically with no restorative warming mechanism to combat it .
there is no analogous positive feedback at the other temperature extreme - indeed , hot climates possess a strong negative feedback mechanism through infrared cooling , and as a result the hot classification is less populated . as a further consequence of these differing feedback mechanisms
, there is a small collection of moons which are classified as variable , the majority of which tend towards being too hot rather than too cold .
we can see this in more detail in figure [ fig : tallhist_lebm ] , which shows the minimum , maximum and mean temperatures exhibited by each simulation .
the majority of simulations possess minima and maxima which fit inside the 100k temperature range where water is expected to be liquid on a telluric planet s surface .
there is a small tail at @xmath112 k , but this is not as populous as the `` snowball '' population with mean temperatures around 200k .
the rapid albedo transition at 273k ensures that few moons can maintain steady surface temperatures in this regime , giving rise to the gap centered on @xmath113k .
the mode of the mean temperature distribution is close to the `` true '' mean temperature of 320k , with a long tail out to around 450k .
k and @xmath114k respectively.__,width=8 ] in short , we can say with moderate confidence that we can reliably identify whether we expect a detected moon to be habitable or otherwise , given the posterior distribution , i.e. modeling the surface temperature using the posterior data gives similar results to those produced by modeling the injected signal .
this being said , we should be cognizant of the long tail of the resulting temperature distributions produced , and we should also note that one in six posterior runs identify the moon as uninhabitable .
we first begin by discussing the fits attempted under the assumption that a putative exomoon follows a circular orbit around the planet kepler-22b ( although the planet itself is permitted to have an eccentric orbit ) . in total , there are four fits we attempted which fall into this category : a circular orbit planet with and without informative limb darkening priors ( models @xmath115 and @xmath116 respectively ) and an eccentric orbit planet with and without informative limb darkening priors ( models @xmath117 and @xmath118 respectively ) .
none of these four fits show particularly significant model preferences , being in the range 1 - 2@xmath8 favorable over a simple planet - only fit ( see table [ tab : evidence ] ) . as with @xcite
, we consider only models above @xmath10@xmath8 to be considered as candidate exomoons .
therefore , we find no compelling evidence for a circular orbit exomoon around kepler-22b . out of the four models , @xmath118 shows the highest significance at 2.3@xmath8 and we show the corresponding maximum a - posteriori light curve fit from this model in figure [ fig : bestfits_ep ] , which has a @xmath119 better fit for @xmath120 data points .
although the signal is not significant , we note that the solution corresponds to a moderately eccentric planet of @xmath121 and mass @xmath122@xmath2 ( derived purely from moon dynamical model , not from rvs ) , which is consistent with that derived from the rvs .
the corresponding satellite would have a radius of @xmath123@xmath0 and mass of @xmath124@xmath2 with a broad density posterior spanning @xmath125g@xmath97 .
we estimate that only 64.7% of trials yield a physically plausible density by using the internal structure models of @xcite . as with our previous survey @xcite ,
despite finding a non - detection we derive upper limits on the mass - ratio for a putative exomoon .
applying bayesian model averaging over the four planet - only models and the four circular - orbit moon models ( i.e. eight models in total ) allows us to compute a model - averaged posterior for the mass ratio of @xmath126 .
the 95% upper quantile of this posterior gives @xmath127 i.e. 6.2% .
for the radius - ratio posterior , @xmath104 , we note that 97.8% of trials were positive ( recall we explored negative radius solutions ) , corresponding to 2.0@xmath8 significance using the odds - ratio test given by equation 4 in @xcite .
unlike previous planets studied , we have a reliable mass constraint for the host planet from the radial velocities meaning we can translate this mass ratio into a physical mass upper limit . further more , kepler-22b is a validated small - radius planet and so must conform with physically plausible mass - radius constraints .
we therefore generate a posterior for @xmath128 in absolute units and then eliminate any trials where either the joint posterior of @xmath41-@xmath129 or @xmath128-@xmath129 corresponds to a density exceeding the maximum mass stripping limit of an iron - rich planet derived by @xcite .
we find that this eliminates 5% of the posterior trials . from this
, we estimate that @xmath130@xmath2 to 95% confidence .
we therefore conclude that kepler-22b is highly unlikely to host an earth - like moon on a circular orbit .
both of the eccentric moon models , @xmath131 and @xmath132 , are strongly favored over the planet - only models at @xmath133@xmath8 and thus are significant enough to be further considered as a candidate , with the latter model being slightly preferred . in a @xmath134 sense ,
the maximum a - posteriori realization from model @xmath132 is superior to planet - only fit @xmath135 at @xmath136 and the circular moon fit @xmath116 at @xmath137 .
the light curve fits appear to be heavily influenced by a transit - like feature occurring prior to the transit of kepler-22b in the long - cadence data of q1 ( see fig [ fig : bestfits_es ] ) .
unlike the circular orbit fits , the eccentric fits have the flexibility to both explain this feature and yet remain comptable with the rest of the time series .
although our bayesian model selection approach penalizes models for using more free parameters , such as these eccentric fits , our likelihood function assumes pure gaussian noise and so we must remain cautious in interpreting what could simply be a time - correlated noise feature of unknown origin .
this is particularly salient in light of the fact cofiam identified q1 has having the greatest degree of autocorrelation out of the four transits used in this work ( see [ sub : cofiam ] ) .
one of the key tests described in @xcite for vetting such systems is to inspect the derived posteriors and see if they are physically plausible or not .
the bayesian model averaged posteriors of all ten models attempted are dominated by the eccentric moon fits due to the large odds ratios and so we inspect these posteriors to vet these solutions .
the dynamically derived planet density appears rather high at @xmath138g@xmath97 yielding @xmath139@xmath2 .
this may be compared with the rv derived @xmath41 from the same posteriors of @xmath140@xmath2 and @xmath141 to 95% confidence , from which we conclude the result is slightly incompatible . before commenting on the putative exomoon s composition , we note that the orbit appears excited with @xmath142 and inclined @xmath143 ( highly multimodal posterior ) from the planet s orbital plane . for the composition , we find an unusually low density of @xmath144g@xmath97 , which favors a bulk density below that of water / ice .
using the minimum atmospheric height method of @xcite and using the derived moon radius of @xmath145@xmath0 , we find that 93.3% of trials yield a radius exceeding a pure water / ice composition moon .
the only way to explain this situation is an extended atmosphere but given that @xmath146@xmath2 , the moon would rapidly lose an atmosphere , given the equilibrium temperature of @xmath147k .
given that kepler-22 is not a young star ( b12 ) and so presumably the exomoon has had plenty of time for an atmosphere to have escaped already , we must invoke a continuously replenishing atmosphere to explain these values .
the orbit of the inclined , eccentric moon can also be shown to be unstable .
the hill stability of inclined low - mass binaries in the three body problem was explored extensively in @xcite .
@xcite presented expressions for the maximum stable eccentricity of an inclined binary in the case where the combined planet - moon mass is much less than that of the host star , as is applicable for the kepler-22b system .
using equation 25 of this work , which has a complex dependency with @xmath128 , @xmath41 , @xmath42 , @xmath86 , @xmath148 and @xmath149 , we estimate that 99.97% of the joint posterior realizations exhibit hill instability ( see figure [ fig : eshisto ] ) .
driving this determination is the wide - separation of the putative moon s orbit of @xmath150@xmath129 combined with the inclined , eccentric nature of the orbit .
we therefore consider that the solution is a false - positive , likely induced by residual autocorrelation in the long - cadence data of q1 .
this false transit signal can only be fitted by driving to high eccentricities .
one effect of high exomoon eccentricity is to increase the r.m.s .
ttv and tdv amplitudes induced by the moon on the planet @xcite . since no ttvs or tdvs occur in the kepler-22b data ( figure [ fig : ttvs ] ) , the algorithm is forced to set the moon to be a very low - mass object in order to explain the full data set .
it has been shown that the @xmath133@xmath8 eccentric exomoon signal is a false positive i.e. spurious detection .
spurious detections can not be used to derive upper limits for reasons discussed in @xcite and so we do not use these results to derive our excluded moon mass limits presented here .
this is equivalent to assigning a negligible prior model probability to models @xmath132 and @xmath131 .
performing bayesian model averaging over the remaining eight models provides reliable upper limits on the presence of an exomoon .
we find that @xmath151% to 95% confidence which can be converted into a physical mass estimate by leveraging the @xmath152 constraint from the radial velocities and @xmath42 from asteroseismology to give @xmath153@xmath2 to 95% confidence .
due to the large number of results in this paper , we summarize the most important findings below : * we have conducted the first search for an exomoon around a habitable - zone exoplanet and find no compelling evidence for a companion to kepler-22b .
furthermore , we constrain @xmath153@xmath2 to 95% confidence . *
we have demonstrated that an earth - like moon would be detectable around kepler-22b with the current data to 8.3@xmath8 by signal injection and recovery . *
we have further shown that such an injected moon would very likely have conditions suitable for liquid water on the surface through latitudinal energy balance modelling ( lebm ) of the posteriors samplings .
* we have introduced several new improvements to the hek methodology , including exploring negative - radius , retrograde and eccentric moon solutions , free limb darkening sampled from a dirichlet prior and bayesian model averaging .
* for the first time , the radial velocities of kepler-22b have been fitted with a model accounting for free eccentricity and free stellar jitter . combined with an updated ephemeris from a new transit in q11 , we derive a tighter mass constraint on kepler-22b of @xmath154@xmath2 to 95% confidence . * utilizing single - body asterodensity profiling ( sap ) and the radial velocities , we provide the first constraints on the orbital eccentricity of kepler-22b with @xmath155 and @xmath156 to 95% confidence . * from our refined transit model , we estimate that kepler-22b has a @xmath5% probability of lying within the empirical habitable - zone but a @xmath6% probability of lying within the conservative habitable - zone ( as defined by @xcite ) .
we derive an insolation of @xmath157@xmath158 .
the above results were derived through photodynamical modeling and multimodal nested sampling regression requiring 49.7years of cpu time@xmath159 . compared to other systems analyzed during the hek project @xcite , this load is higher than average due to the high - resolution fitting mode employed here ( see [ sub : hires ] ) .
nevertheless , this highlights the unique computational challenges of seeking exomoons .
we find no evidence for an earth - like exomoon around kepler-22b and yet have shown that the present data can easily detect such an object via signal injection .
current observations therefore dictate that kepler-22b does not possess an earth - like habitable moon .
our results then , combined with the very robust measurement of the planet s radius , mean that kepler-22 does not possess an earth analog .
this does not mean that the system possesses no options for an inhabited world , with notable possiblities being a smaller , presently undetectable moon ( e.g. @xmath160@xmath2 ) or a possible ocean on kepler-22b .
however , it is now clear that this is not the location to find a second earth . to date , nine systems have been surveyed for exomoons by the hek project @xcite and no detections have been made with most cases yielding detection sensitivities of @xmath161 .
we caution that the number of systems analyzed remains too small to draw any meaningful conclusions about the occurrence of large moons , @xmath162 , but this is the ultimate goal of our project . in coming work
, two future surveys will focus on i ) m - dwarf host star planetary candidates and ii ) planets exhibiting repeated visual anomalies .
slowly then , the landscape of the frequency of exomoons will be revealed .
this work made use of the michael dodds computing facility .
we thank the anonymous reviewer for their thoughtful comments which improved the quality of our manuscript .
dmk is funded by the nasa carl sagan fellowships .
df gratefully acknowledges support from stfc grant st / j001422/1 .
jh and gb acknowledge partial support from nsf grant ast-1108686 and nasa grant nnx12ah91h .
dn acknowledges support from nsf ast-1008890 .
99 anglada - escud , g. et al . , 2012 , apj , 751 , l16 anglada - escud , g. , rojas - ayala , b. , boss , a. p. , weinberger , a. j. & lloyd , j. p. , 2013 ,
a&a , 551 , 48 balan , s. t. & lahav , o. , 2009 , mnras , 394 , 1936 borucki , w. j. et al . , 2012 , apj , 745 , 120 ( b12 ) charbonneau , d. et al . , 2009 ,
nature , 462 , 891 domingos , r. c. , winter , o. c. & yokoyama , t. , 2006 , mnras , 373 , 1227 donnison , j. r. , 2010 , p&ss , 58 , 1169 feroz , f. , hobson , m. p. & bridges , m. , 2009a , mnras , 398 , 1601 feroz , f. , gair , j. r. , hobson , m. p. & porter , e. k. , 2009b , cqg , 26 , 215003 forgan , d. & kipping , d. m. , 2013 , mnras , accepted ( astro - ph:1304.4377 ) gilliland . , r. et al .
, 2010 , apj , 713 , 160 vladilo , g. , murante , g. , silva , l. , provenzale , a. , ferri , g. & ragazzini , g. , 2013 , apj , 767 , 65 heller , r. , 2012 , a&a , 545 , l8 heller , r. & barnes , r. , 2012 , arxiv:1210.5172 heller , r. & barnes , r. , 2013 , asbio , 13 , 18 kipping , d. m. , 2009 , mnras , 392 , 181 kipping , d. m. , 2009 , mnras , 396 , 1797 kipping , d. m. , fossey , s. j. & campanella , g. , 2009 , mnras , 400 , 398 kipping , d. m. , 2010 , mnras , 408 , 1758 kipping , d. m. et al . , 2010 , apj , 725 , 2017 kipping , d. m. & bakos , g. a. , 2011 , apj , 730 , 50 kipping , d. m. & bakos , g. a. , 2011 , apj , 733 , 36 kipping , d. m. , 2011 , mnras , 416 , 689 kipping , d. m. , 2011 , phd thesis , university college london ( astro - ph:1105.3189 ) kipping , d. m. , dunn , w. , jasinksi , j. m. & manthri , v. p. , 2012
, mnras , 421 , 1166 kipping , d. m. , bakos , g. . ,
buchhave , l. , nesvorn , d. & schmitt , a. 2012 , apj , 750 , 115 kipping , d. m. , hartman , j. , buchhave , l. , schmitt , a. , bakos , g. . &
nesvorn , d. , 2013 , apj , 770 , 101 kipping , d. m. , spiegel , d. s. , sasselov , d. , 2013 , mnras , 434 , 1883 kipping , d. m. , 2013 , mnras , accepted ( astro - ph:1308.0009 ) kopparapu , r. k. et al . , 2013 , apj , 765 , 131 kurucz r. , 2006 , stellar model and associated spectra ( http://kurucz.harvard.edu/grids.html ) mandel , k. & agol , e. , 2002 , apj , 580 , 171 marcus , r. a. , sasselov , d. , hernquist , l. & stewart , s. t. 2010 , apj , 712 , 73 mazeh , t. & faigler , s. , 2010 , a&a , 521 , 59 neubauer , d. , vrtala , a. , leitner , j. j. , gertrude f. m. , hitzenberger , r. , 2012 , p&ss , 73 , 397 nesvorn , d. kipping , d. m. , buchhave , l. a. , bakos , g. a. , hartman , j. & schmitt , a. , 2012 , science , 336 , 1133 parkinson , d. & liddle , a. 2013 , sam , 6 , 3 pal , a. , 2008 , mnras , 390 , 281 reynolds , r. t. , mckay , c. p. & kasting , j. f. , 1987 , adspr , 7 , 125 scharf , c. a. , 2006 , apj , 648 , 1196 spiegel , d. s. , menou , k. & scharf , c. a. , 2008 , apj , 681 , 1609 southworth , j. , wheatley , p. j. & sams , g. , 2007 , mnras , 379 , 11 vogt , s. s. , butler , r. p. , rivera , e. j. , haghighipour , n. , henry , g. w. , williamson , m. h. 2010 , apj , 723 , 954 waldmann , i. p. , tinetti , g. , drossart , p. , swain , m. r. , deroo , p. & griffith , c. a. , 2012 , apj , 744 , 35 williams , d. m. & kasting , j. f. , 1997 , icarus , 129 , 254 winn , j. n. 2010 , arxiv:1001.2010 zeng , l. & sasselov , d. , 2013 , pasp , submitted ( astro - ph:1301.0818 ) zsom , a. , seager , s. & de wit , j. , 2013 , apj , submitted ( astro - ph:1304.3714 ) | kepler-22b is the first transiting planet to have been detected in the habitable - zone of its host star . at 2.4@xmath0 ,
kepler-22b is too large to be considered an earth - analog , but should the planet host a moon large enough to maintain an atmosphere , then the kepler-22 system may yet possess a telluric world . aside from being within the habitable - zone , the target is attractive due to the availability of previously measured precise radial velocities and low intrinsic photometric noise , which has also enabled asteroseismology studies of the star . for these reasons , kepler-22b was selected as a target - of - opportunity by the `` hunt for exomoons with kepler '' ( hek ) project . in this work , we conduct a photodynamical search for an exomoon around kepler-22b leveraging the transits , radial velocities and asteroseismology plus several new tools developed by the hek project to improve exomoon searches .
we find no evidence for an exomoon around the planet and exclude moons of mass @xmath1@xmath2 to 95% confidence . by signal injection and blind retrieval
, we demonstrate that an earth - like moon is easily detected for this planet even when the time - correlated noise of the data set is taken into account .
we provide updated parameters for the planet kepler-22b including a revised mass of @xmath3@xmath2 to 95% confidence and an eccentricity of @xmath4 by exploiting single - body asterodensity profiling ( sap ) .
finally , we show that kepler-22b has a @xmath5% probability of being within the empirical habitable - zone but a @xmath6% probability of being within the conservative habitable - zone . |
among the extremely remarkable properties of the fractional quantum hall ( fqh ) effect @xcite a major role is played by the emergence of anyonic excitations carrying fractional charge and statistics @xcite . in particular , quasiparticle ( qp ) excitations for states belonging to the laughlin @xcite and jain sequence @xcite are predicted to have abelian exchange statistics .
more intriguingly , some of the proposed models for the filling factor @xmath0 @xcite predict the emergence of excitations with charge @xmath1 , and multiples , with possible non - abelian properties @xcite .
these predictions paved the way to possible applications of non - abelian anyons in fault - tolerant topological quantum computation ( see @xcite and references therein ) .
unfortunately , as far as we know , a direct confirmation of fractional statistics is still lacking even if different proposal are reported in the literature @xcite .
so far , evidence of the fractional statistics is indirect , essentially based on the evidence of the existence of fractional charges @xcite .
a great experimental effort has been devoted in the last years to access fractional charges through shot noise measurements in quantum point contact ( qpc ) geometry starting with the seminal works of refs .
@xcite . in this direction ,
composite states ( for example @xmath2 ) showed a quite universal phenomenology leading to a crossover between two different value of the effective charge as a function of temperature or bias @xcite .
this behaviour has been explained in terms of competition of two charge carriers : the agglomerates and the single - qp @xcite .
similar arguments hold as well for the state at @xmath0 @xcite .
other interpretations based on edge state reconstruction @xcite , local filling factor effects @xcite or tunnelling amplitude non - linearities @xcite have also been proposed .
unfortunately , in the discussed measurements , the contributions associated to the various excitations are typically mixed because these studies are conducted at very low frequencies ( almost dc ) .
therefore , it is useful to find alternative methods to address the excitations separately . a possible way is to consider the noise at finite frequency ( f.f . )
indeed , this quantity presents resonant singular behaviour ( such as peaks or dips ) in correspondence of the josephson frequency @xmath3 associated to each charge carrier @xmath4 with @xmath5 the applied bias .
this is an independent method to measure the charge of the fractional excitations in the system that has not yet be experimentally explored for fqh states so far .
indeed , for sufficiently low temperatures @xmath6 , it allows to separate the different charge contributions realising a sort of qp spectroscopy .
intriguingly enough , f.f .
noise could efficiently address also other properties , like the scaling dimensions associated to each qp excitation using only a bias scan at fixed frequency as we will show .
this combination of information has the potential to give further constraints on the edge state model @xcite , and finally to address the topological order of the bulk ground state @xcite .
first theoretical steps in this direction were done on symmetrised noise @xcite , typical quantity considered at low frequencies . in such case the expected features associated to the state at @xmath0 @xcite as well as the possibility to access the contribution associated to the different qps predicted by theoretical models @xcite have been investigated . however , at high frequencies @xmath7 , quantum effects become relevant and the symmetrised f.f .
noise is only one possible choice among different experimental quantities addressed by different protocols . indeed , in such regime , one has to identify the relevant quantity measured with the specific setup under investigation @xcite .
hereafter we get inspiration by the proposal of lesovik and loosen @xcite where a model based on a resonant lc circuit is discussed in order to extract non - symmetrised current - current correlations .
recent experiments carried out for a two dimensional electron gas qpc in absence of magnetic field fully agree with theoretical predictions @xcite . since the resonant circuit coupled to a qpc is the prototypical measurement scheme in fqh experiments @xcite it appears quite obvious to explore the same physics in this contest .
we have recently investigated the same setup for an abelian fqh states @xcite , and here we consider it to the case @xmath0 analysing the signatures of the different non - abelian phases ( pfaffian and anti - pfaffian ) . the goal of the present paper is to analyse in details the expected f.f . measured noise for a realistic situation .
the effects associated to the temperature of the fqh qpc system and the lc detector on the visibility are also carefully considered .
the paper is divided as follows : in section 2 we discuss the non - abelian models for edge states pointing out similarities and differences in term of the most dominant fractionally charged excitations . in section 3
we discuss the definition of the noise properties for a qpc in the weak - backscattering regime and the definition of the noise power measured in the proposed setup . in section 4
we discuss the result for the measured noise power obtained for two considered non - abelian models and we also compare them with the well know symmetrised noise . finally we inspect the effect of changing the detector temperature @xmath8 .
we start recalling the two more accredited models of composite edge states at filling factor @xmath0 @xcite : the pfaffian ( p ) @xcite and the disorder dominated anti - pfaffian ( ap ) @xcite .
the associated lagrangian densities are given by the sum of charged ( @xmath9 ) and neutral contributions(@xmath10 ) , namely @xmath11 , with ( @xmath12 ) _
c=- _ x _ c(_t_c+v_c _ x _ c ) [ lagrangian_charged ] and _
n =- i(_t + v_n _ x)- _ x _ n ( _ t _
n + v_n _ x _ ) .
[ lagrangian_neutral ] they both describe an hall fluid at filling factor @xmath13 with two additional filled landau levels , playing the role of the _ vacuum _ of the theory , according to the conventional decomposition @xmath14 .
the charged bosonic field @xmath15 is related to the electron number density through @xmath16 , while @xmath17 is a bosonic neutral field and @xmath18 represents a neutral majorana fermion in the ising sector @xcite .
the parameter @xmath19 denotes the direction of propagation of neutral modes with respect to the charged ones .
in particular , for @xmath20 the modes are co - propagating , while for @xmath21 they are counter - propagating .
the two models differ in the neutral sector @xmath10
@xcite , with @xmath22 and @xmath23 for p , @xmath24 and @xmath21 for ap .
the propagation velocities of the charged and neutral modes are indicated with @xmath25 and @xmath26 respectively . due to the hidden symmetry of the neutral sector of ap model @xcite in the disorder dominated phase , one has the same velocity @xmath26 both for the bosonic @xmath27 and the fermionic neutral modes @xmath28 .
moreover one may reasonably assume a larger charge velocity @xmath29 @xcite .
the quantization of the above bosonic fields is given by the commutation relation = i_c / n sgn(x - y ) , with @xmath30 and @xmath31 , while the majorana fermion commutes with both .
operators destroying an excitation along the edge can be written as @xcite @xmath32}\ , , \label{psi}\end{aligned}\ ] ] with @xmath33 integer numbers and where the operator in the ising sector @xmath34 can be the identity operator @xmath35 , the majorana fermion @xmath18 or the spin operator @xmath36 .
they are associated to the fact that the excitation charge is even or odd multiple of the fundamental charge of the model @xmath37 .
indeed all the excitations described by previous operators have charge @xmath38 and we call them @xmath39-agglomerates @xcite . the single - valuedness properties of the phase acquired by an @xmath39-agglomerate with respect to the operation of encircling an electron in the bulk , force @xmath39 and @xmath40 to be : even integers for @xmath41 or @xmath28 , and odd integers for @xmath42 @xcite . notice that the presence of @xmath43 in the operator leads to non - abelian statistical properties important for fault - tolerant quantum computation as determined by the fusion rules @xcite . the zero temperature time dependent green s functions associated to the operators of the ising sector and the charged and neutral bosonic fields are @xcite ( 0,t ) ( 0,0 ) & = ( 1+i _ t)^-_,&[gfising ] + _ s(0,t ) _
s(0,0 ) & = -|_s| & s= , [ gfbosonic ] with @xmath44 , @xmath45 and @xmath46 the conformal weights of the field in the ising sector @xcite and we introduced the energy bandwidths @xmath47 , with @xmath48 a finite length cut - off . in the following
we will assume @xmath49 as the largest energy scale of the model .
from the long - time behaviour of the imaginary time two - point green s function and in the scaling @xmath50 . from now on indeed we will consider for any @xmath39-agglomerate only the qp operators with the minimal scaling dimension compatible with the single - valuedness requirement . ]
t_^(m)_l ( ) _ l^(m)^(0)||^-2^(m)_ll= we can extract the scaling dimensions @xcite of the @xmath39-agglomerates _
p^(m)=_+ m^2;_ap^(m)=_+ m^2+n^2 [ delta ] which depends on the model considered . therefore
, for the single - qp with minimal charge @xmath1 ( @xmath51 , @xmath42 and only for ap @xmath52 ) one has respectively _
p^(1)=;_ap^(1)= , [ 1_8 ] while the @xmath53-agglomerate excitation with charge @xmath54 ( @xmath55 , @xmath56 , @xmath41 ) , with a scaling dimension driven by the charged mode contribution only with _ p^(2)=_ap^(2)=. [ deltaagg ] these values indicate the single - qp as the most dominant excitation at low energy in the p case , while in the ap case single - qp and 2-agglomerate have equal relevance with the same scaling dimensions @xcite .
the latter situation is quite general and valid for all anti - read - rezayi states @xcite .
all other excitations , with higher charges , have higher scaling dimensions and can be safely neglected in what follows .
it is worth to mention that interactions with the external environment can lead to renormalizations of the scaling parameters with remarkable consequences on the transport properties ( see ref .
@xcite for a better discussion ) . in the following , for sake of simplicity
, we will focus on the unrenormalised case only despite the method may be generalised to the renormalised case .
once characterised the excitations of the considered models for @xmath0 , we can investigate the associated f.f .
backscattering noise in the qpc geometry shown in fig .
a similar measurement scheme was proposed for the first time by lesovik and loosen in ref .
@xcite . here
, the qpc is subjected to a bias voltage v and coupled to a resonant lc circuit , playing the role of the detector ( with measurement frequency @xmath57 ) , via an impedance matching circuit ( see dashed box in fig .
[ fig1 ] ) . with strong magnetic field
the impedance matching in the system is a challenging technological problem @xcite therefore it is advantageous to suppose to work at fixed resonant frequency @xmath58 assuming a very high quality factor of the detector .
is applied to the qpc and @xmath39-agglomerate excitations can tunnel between the edges .
the two circuits are matched in impedance via a coupling circuit ( inside dashed line ) .
here we assume they are kept at two different temperatures @xmath59 and @xmath8 as indicated in the picture . in the figure
we explicitly show also with grey boxes a low - pass ( lp ) and high - pass ( hp ) filters which make possible to directly couple the dc bias to qpc and , at the same time , to deviate the high - frequency components toward the lc circuit . ]
we focus on the simple two terminal geometry in the weak backscattering limit , where @xmath39-agglomerate tunnelling processes can be treated separately . in real systems a four terminal version of this setup is required , however this does nt change the main result obtained with this simplified version .
the point - like tunnelling of a generic @xmath39-agglomerate between the right - propagating ( @xmath60 ) and the left - propagating ( @xmath61 ) edge can be described through the tunnelling hamiltonian ( @xmath62 ) [ tun ] h^(m)_t , l=
t_m _ l , + ^(m)(0)^(m)_l , -^(0)+h.c .
, where @xmath63 is the @xmath39-agglomerate tunnelling amplitude ( assumed energy independent ) .
the finite bias @xmath5 between the two edges can be included in our formalism through the gauge transformation @xmath64 , where , @xmath65 is the josephson frequency associated to the fundamental charge @xmath66 @xcite . from the tunnelling hamiltonian in ( [ tun ] )
one can easily derive the backscattering current operator associated to the @xmath39-agglomerate @xcite i^(m)_b , l(t)=i me^ * ( t_m e^im_0t_l,+^(m)(0,t)^(m)_l,-^(0,t)-h.c . ) .
the contribution to the averaged backscattering current of the @xmath39-agglomerate at lowest order in the tunnelling can be easily written in terms of the tunnelling rates @xmath67 @xcite i^(m)_b , l = m e^*(1- e^-m_0/t)^(m)_l ( _ 0 ) , [ current_bal ] with the average @xmath68 taken over the quantum statistical ensemble .
the rates can be also evaluated analytically at low temperatures @xmath69 @xcite ^(m)_l ( _ 0 ) & = & t^_m+_l , -1e^ + & & b(_m+_l , -i ; _
m+_l , + i ) [ eq : gamma ] where @xmath70 is the euler beta function . here ,
@xmath71 and @xmath72 depend on the variables @xmath73 and @xmath40 which characterise the tunnelling excitation for the specific p or ap model considered ( see discussion around ( [ 1_8 ] ) and ( [ deltaagg ] ) ) . note that for @xmath74 asymptotic expansion shows that @xmath75 , with the expected power law dependences of the rates from the bias energy @xmath76 and the scaling dimension @xmath50 as usually happen in the luttinger liquid theory . for higher bias value @xmath77 the power - law does not depend anymore on the neutral components and
one finds @xmath78 where the power - law scaling is determined only by the charge of the @xmath39-agglomerates .
the low - energy analytical result presented corresponds to the standard golden rule rate for the tunnelling processes and one may eventually calculate it also with numerical methods following the prescription of ref .
@xcite .
the proper quantity to consider in order to investigate the current fluctuations of the qpc coupled to the resonant circuit is the non - symmetrised noise @xcite s^(m)_+()=^+_- dt e^i t i^(m)_b , l(0 ) i^(m)_b , l(t ) , where we have introduced the back - scattering current fluctuation @xmath79 . on the noise power since its definition is exactly the same for the two models . ]
this quantity represents , for @xmath80 , the noise power emitted by the system into the detector .
the corresponding absorptive part is given by s^(m)_-()=^+_- dt e^i t i^(m)_b , l(t ) i^(m)_b , l(0 ) = s_+^(m)(- ) [ s_minus ] .
with these quantities it is easy to calculate the f.f .
symmetrised noise @xcite usually considered in literature [ eq : ssym ] s^(m)_sym()&= & ^+_- d t e^i t \ { i^(m)_b , l(t ) , i^(m)_b , l(0)}= s^(m)_+()+s^(m)_- ( ) [ eq : s_sym ] having indicated with @xmath81 the anticommutator . at lowest order in the tunnelling amplitudes and using standard keldysh formalism also
the non - symmetrised noise can be expressed in terms of the qpc tunnelling rates @xcite s^(m)_+(,_0)= , [ s_plus_rate ] with a peculiar combination of the frequency @xmath58 and the bias voltage @xmath76 in the arguments of the golden rule rates . the detector of fig .
[ fig1 ] represents a concrete measurement scheme to access current fluctuations at high frequencies . in the following
we will focus on the regime where the qpc temperature @xmath59 is lower than the frequency ( quantum limit ) and the bias ( shot noise limit ) , i.e. @xmath82 .
this allow to investigate the fractional qp contributions via a sort of spectroscopy .
the measurable quantity is the spectral power measured in the amplifier chain ( grey area in fig .
[ fig1 ] ) , which is proportional to the variation of the energy stored in the lc before and after the switching on of the lc - qpc coupling . from now on we will indicate it as measured noise @xmath83 where with @xmath58 we indicate the frequency of the lc circuit and with @xmath76 the qpc bias . at lowest order in the coupling @xmath84
it can be expressed as @xcite s^(m)_meas(,_0)= k \{s^(m)_+(,_0)+n_b ( ) } , [ s_meas ] where the non - symmetrised noise qpc spectra for the @xmath39-agglomerate are @xmath85 of ( [ s_plus_rate ] ) . here ,
@xmath86 $ ] the bose distribution describing the equilibrium state of the lc detector and @xmath87 the detector inverse temperature . in general @xmath8
can be different from the system temperature @xmath59 since system and detector are _ weakly _ coupled .
we wish finally recall that this quantity can be also investigated using a strategy similar to the definition of the excess noise which further simplify the impedance matching problem at the level of the lc - qpc coupling ( see ref .
@xcite for details ) . to consider all contributions due to tunnelling of different @xmath39-agglomerate , in weak tunnelling
, one can directly sum them ( @xmath88 ) s_i(_0)= _ m s^(m)_i(_0 ) , [ noise_sum ] where from now on we suppress the explicit dependence on the lc frequency @xmath58 since in the following discussion is always kept fixed .
we conclude this part noting that these results suggest that f.f .
noise is a spectroscopy tool for different tunnelling charges .
( in units of @xmath89 ) for the p ( left ) and the ap ( right ) model at @xmath0 as a function of the bias @xmath90 and at fixed frequency @xmath58 .
the qpc temperature associated to each curve is indicated in the legend .
( bottom panels ) corresponding derivatives @xmath91 .
the curve at the lowest temperature ( @xmath92 mk ) has been omitted for better visibility .
other parameters expressed in temperature scale are : @xmath93 mk , @xmath94 mk , @xmath95 mk , @xmath96 k and @xmath97.,title="fig : " ] ( in units of @xmath89 ) for the p ( left ) and the ap ( right ) model at @xmath0 as a function of the bias @xmath90 and at fixed frequency @xmath58 . the qpc temperature associated to each curve
is indicated in the legend .
( bottom panels ) corresponding derivatives @xmath91 .
the curve at the lowest temperature ( @xmath92 mk ) has been omitted for better visibility .
other parameters expressed in temperature scale are : @xmath93 mk , @xmath94 mk , @xmath95 mk , @xmath96 k and @xmath97.,title="fig : " ] + ( in units of @xmath89 ) for the p ( left ) and the ap ( right ) model at @xmath0 as a function of the bias @xmath90 and at fixed frequency @xmath58 . the qpc temperature associated to each curve is indicated in the legend .
( bottom panels ) corresponding derivatives @xmath91 .
the curve at the lowest temperature ( @xmath92 mk ) has been omitted for better visibility .
other parameters expressed in temperature scale are : @xmath93 mk , @xmath94 mk , @xmath95 mk , @xmath96 k and @xmath97.,title="fig : " ] ( in units of @xmath89 ) for the p ( left ) and the ap ( right ) model at @xmath0 as a function of the bias @xmath90 and at fixed frequency @xmath58 .
the qpc temperature associated to each curve is indicated in the legend .
( bottom panels ) corresponding derivatives @xmath91 .
the curve at the lowest temperature ( @xmath92 mk ) has been omitted for better visibility .
other parameters expressed in temperature scale are : @xmath93 mk , @xmath94 mk , @xmath95 mk , @xmath96 k and @xmath97.,title="fig : " ] the results concerning the measured f.f .
noise @xmath98 in ( [ s_meas ] ) and ( [ noise_sum ] ) at @xmath0 are shown in fig .
[ fig2 ] ( upper panels ) as a function of the qpc bias @xmath76 for different qpc temperatures @xmath59 , keeping fixed the resonant circuit temperature @xmath8 .
we discuss only the behaviour at positive bias @xmath99 since the noise is a symmetric function of the qpc bias @xmath76 .
analogies and differences between the p ( upper left panel ) and ap ( upper right panel ) models become evident from a direct comparison . starting from the lowest temperature case ( @xmath92 mk , black curves ) we observe that both models show a flat behaviour at @xmath100 , which is a clear signature of the lack of contribution of ground states fluctuations in the considered measurement scheme @xcite .
the little deviation from zero are associated to the mismatch between the systems and detector temperature which can be always cancelled when @xmath101 @xcite .
steep jumps associated to the @xmath53-agglomerate contribution appear at @xmath102 showing an identical profile in both models , which reflects the same scaling dimension of the two model for that excitation ( see ( [ deltaagg ] ) ) .
different is the spike associated with the single - qp occuring at @xmath103 .
indeed , they are much more high and sharp in the p case with respect to the ap reflecting a lower scaling dimension of qp excitation for the p model .
this feature could quite clearly distinguish between the two models .
however the temperature should be kept quite low , since increasing it the differences are progressively less marked .
eventually some signature survive only by considering the bias derivative of this quantity ( see bottom panels of fig . [ fig2 ] ) .
the previous behaviours can be explained in a simple way in the quantum limit ( @xmath104 ) for the detector and the shot - noise limit ( @xmath105 ) for the system . in this case one
has the contributions of single and double excitations for the two models ( @xmath62 ) @xcite [ shift ] s_meas(_0)_1 * * _ l^(1)(_0-)+_2 * * _ l^(2)(2_0- ) with @xmath106 and @xmath107 constant prefactors and the explicit expression of the rates are reported in ( [ eq : gamma ] ) for @xmath108 .
this result confirms the same scaling for the @xmath53-agglomerate in the two models , but different behaviours for the single - qp contributions ( see ( [ 1_8 ] ) and ( [ deltaagg ] ) ) .
as we observed after ( [ eq : gamma ] ) for high biases @xmath109 the scaling dimensions are determined only by the charged part that is the same for the two models .
this is another reason why conventional scaling analysis , which is typically done in asymptotic regime , would fails in detecting the differences between the two models especially when the neutral mode bandwidth @xmath110 is quite small . ] . by increasing the system temperature @xmath59 , keeping fixed the one of the detector ( @xmath93 mk ) ,
the peaked structure become progressively smoothened due to a rounding of the singularities . in this regime , the differences in the measured f.f .
noise become clear only by looking the derivative with respect to the qpc bias @xmath91 as shown in the bottom panels of fig .
[ fig2 ] . by focusing on the blue curves , and keeping in mind the different scale in the ordinates between the two panels
, one can observe again the similarity of peaks associated to the @xmath53-agglomerate ( @xmath102 ) .
however concerning the single - qp ( @xmath103 ) , the difference in the scaling leads to a pronounced peak followed by a stronger dip in the p case with respect to the ap case .
( dashed lines ) and the symmetrised noise and @xmath111 ( solid lines ) for different values of the ratio @xmath112 ( indicated in the legend ) .
( left panel ) the f.f .
noises @xmath113 with @xmath114 in units of @xmath89 .
note that @xmath115 for the symmetrised noise .
( right panel ) the bias derivatives @xmath116 in units @xmath117 .
other parameters are : @xmath118 mk , @xmath119 mk , @xmath95 mk and @xmath96 k. , title="fig : " ] ( dashed lines ) and the symmetrised noise and @xmath111 ( solid lines ) for different values of the ratio @xmath112 ( indicated in the legend ) .
( left panel ) the f.f .
noises @xmath113 with @xmath114 in units of @xmath89 .
note that @xmath115 for the symmetrised noise .
( right panel ) the bias derivatives @xmath116 in units @xmath117 .
other parameters are : @xmath118 mk , @xmath119 mk , @xmath95 mk and @xmath96 k. , title="fig : " ] until now all the plots were done for fixed ratio @xmath120 .
however this parameter is unknown and may change for any specific experimental realization .
for this reason we present in fig .
[ fig3 ] with dashed lines @xmath98 ( right panel ) and its bias derivative @xmath121 ( left panel ) as a function of bias with changing @xmath122 values .
we concentrate mainly on the ap model but similar considerations can be repeated in the p case . as expected , increasing this parameter progressively enhances the @xmath53-agglomerate contribution with respect to the single - qp but still leave both visible at different bias values .
this is particularly true looking the bias derivative .
this result shows the convenience of proposed setup in order to address the presence of the two different charged excitations also when eventually one of the contribution is deeply suppressed in comparison of the other due to non - universal effects .
( in units of @xmath89 ) for the p ( left ) and the ap ( right ) model for the hall state at @xmath0 as a function of the voltage and at fixed frequency ( @xmath123 ) varying the detector temperature ( see legend ) .
( bottom panels ) corresponding derivatives @xmath91 .
other parameters are ( in temperature units where necessary ) : @xmath118 mk , @xmath94 mk , @xmath95 mk , @xmath96 k and @xmath97.,title="fig : " ] ( in units of @xmath89 ) for the p ( left ) and the ap ( right ) model for the hall state at @xmath0 as a function of the voltage and at fixed frequency ( @xmath123 ) varying the detector temperature ( see legend ) .
( bottom panels ) corresponding derivatives @xmath91 .
other parameters are ( in temperature units where necessary ) : @xmath118 mk , @xmath94 mk , @xmath95 mk , @xmath96 k and @xmath97.,title="fig : " ] + ( in units of @xmath89 ) for the p ( left ) and the ap ( right ) model for the hall state at @xmath0 as a function of the voltage and at fixed frequency ( @xmath123 ) varying the detector temperature ( see legend ) .
( bottom panels ) corresponding derivatives @xmath91 .
other parameters are ( in temperature units where necessary ) : @xmath118 mk , @xmath94 mk , @xmath95 mk , @xmath96 k and @xmath97.,title="fig : " ] ( in units of @xmath89 ) for the p ( left ) and the ap ( right ) model for the hall state at @xmath0 as a function of the voltage and at fixed frequency ( @xmath123 ) varying the detector temperature ( see legend ) .
( bottom panels ) corresponding derivatives @xmath91 .
other parameters are ( in temperature units where necessary ) : @xmath118 mk , @xmath94 mk , @xmath95 mk , @xmath96 k and @xmath97.,title="fig : " ] in order to make this statement more quantitative , in fig .
[ fig3 ] , we compare @xmath98 with the f.f . symmetrised noise @xmath124 ( solid lines ) in ( [ eq : ssym ] ) and their bias derivatives ( in right panel ) @xcite . here , again , we keep fixed the frequency @xmath58 , changing the bias @xmath76 that is , by far , the most convenient protocol at the investigated ghz range .
we see that the @xmath124 , as a function of the bias , is unable to detect the two singularities associated to the two fractional charges even in the bias derivatives
. in particular , changing the tunnelling amplitude ratio @xmath122 , the quantity seems only affected with a common multiplicative factor demonstrating that the signature of the two excitations is mainly mixed in that quantity .
this supports the idea that in this bias dependent protocol @xmath124 is not useful especially in comparison to @xmath98 .
this statement can be easily verified by looking at both the left and right panel of fig .
[ fig3 ] .
finally , in order to find a signature of the charged excitations without identifying their scaling dimension , we could vary the lc detector temperature @xmath8 in order to increase the sensibility for charge detection @xcite . in fig .
[ fig4 ] we show that this approach works for both the two non - abelian edge models .
increasing @xmath8 increases the height of the jump in @xmath98 ( top panels ) which correspond also to an increase of the height of the peak in the derivatives ( bottom ) .
note that since the coupling with the detector is assumed weak ( no poisoning from the detector ) the width of the peaks is only slightly influenced by the detector temperature , preserving the resolving power of the discussed bias spectroscopy .
then the crucial limiting factor to the bias spectroscopy is the qpc temperature @xmath59 . as shown in fig .
[ fig4 ] the power to distinguish between the ap and p model is not essentially compromised by the detector temperature since all the feature characterising the model in terms of @xmath98 seems mutually amplified .
increasing @xmath8 is an interesting resource in order to increase detection efficiency , in this perspective .
we have investigated the behaviour of f.f . emitted power @xmath98 of a resonant lc circuit weakly coupled to a qpc built in a quantum hall bar at filling factor @xmath0 .
we showed that the emitted power is represented in terms of the non - symmetrised noise components of the quantum hall qpc weighted by the bosonic distribution of the resonant lc circuit .
we have inspected the different predictions of the pfaffian and anti - pfaffian non - abelian edge states models for this quantity .
we showed that this setup can detect and discriminate between the dominant and sub - dominant fractionally charged excitations looking at the bias dependence at fixed ghz frequencies .
we also discussed the advantage to use this measurement protocol in comparison to the f.f .
symmetrised noise .
finally we demonstrated how the sensibility of the proposed setup can be increased varying the lc detector temperature @xmath8 .
we thank c. glattli , w. belzig and p. solinas for useful discussions .
we acknowledge the support of the miur - firb2012 - project hybridnanodev ( grant no.rbfr1236vv ) , eu fp7/2007 - 2013 under rea grant agreement no 630925 coheat , miur - firb2013 project coca ( grant no . rbfr1379ux ) , and the cost action mp1209 .
a.b acknowledges support from stm 2015 from cnr and victoria university of wellington where partially work was done .
10 tsui d c 1999 _ rev .
phys . _ * 71 * 891 stern a 2008 _ ann .
phys . _ * 323 * 204 laughlin r b 1983 _ phys .
lett . _ * 50 * 1395 jain j k 1989 _ phys .
lett . _ * 63 * 199 willett r , eisenstein j p , stormer h l , tsui d c , gossard a c and english j h 1987 _ phys .
lett . _ * 59 * 1776 moore g and read n 1991 _ nucl .
b _ * 360 * 362 nayak c , simon s h , stern a , freedman m and das sarma s 2008 _ rev .
phys . _ * 80 * 1083 bishara w , bonderson p , nayak c , shtengel k and slingerland j k 2009 _ phys .
b _ * 80 * 155303 stern a , rosenow b , ilan r and halperin b i 2010 _ phys .
b _ * 82 * 085321 willett r l , pfeiffer l n and west k w 2010 _ proc
sci . _ * 106 * 8853 rosenow b , simon s h 2012 _ phys .
b _ * 85 * 201302 de picciotto r , reznikov m , heiblum m , umansky v , bunin g and mahalu d 1997 _ nature _ * 389 * 162 saminadayar l , glattli d c , jin y and etienne b 1997 _ phys .
lett . _ * 79 * 2526 chung y c , heiblum m and umansky v 2003 _ phys .
lett . _ * 91 * 216804 bid a , ofek n , heiblum m , umansky v and mahalu d 2009 _ phys .
lett . _ * 103 * 236802 dolev m , gross y , chung y c , heiblum m , umansky v , and mahalu d 2010 _ phys
b _ * 81 * 161303 ferraro d , braggio a , merlo m , magnoli n and sassetti m 2008 _ phys
* 101 * 166805 ferraro d , braggio a , magnoli n and sassetti m 2010 _ phys . rev . b _ * 82 * 085323 ferraro d , braggio a , magnoli n and sassetti m 2010 _ new j. phys . _ * 12 * 013012 carrega m , ferraro d , braggio a , magnoli n and sassetti m 2011 _ phys . rev . lett . _ * 107 * 146404 wang j , meir y and gefen y 2013 _ phys .
lett . _ * 111 * 246803 roddaro s , pellegrini v , beltram f , biasiol g , and sorba l 2004 _ phys .
lett . _ * 93 * , 046801 roddaro s , pellegrini v , beltram f , pfeiffer l n , and west k w 2005 _ phys .
_ * 95 * 156804 ( 2005 ) hashisaka m , ota t , muraki k , and fujisawa t 2015 _ phys .
lett . _ * 114 * , 056802 shtanko o , snizhko k and cheianov v 2014 _ phys .
b _ * 89 * 125104 smits o , slingerland j k and simon s h 2014 _ phys . rev .
b _ * 89 * 045308 rogovin d and scalapino d j 1974 _ ann . phys . _
* 86 * 1 radu i p , miller j b , marcus c m , kastner m a and pfeiffer l n , west k w 2008 _ science _ * 320 * 899 yang g and feldman d e 2014 _ phys , rev .
b _ * 90 * 161303 susskind l 1995 _ j. math .
phys . _ * 36 * 6377 chamon c , freed d e and wen x g 1995 _ phys .
_ b * 51 * 2363 chamon c , freed d e and wen x g 1996 _ phys . rev .
_ b * 53 * 4033 bena c and nayak c 2006 _ phys . rev .
b _ * 73 * 155335 carrega m , ferraro d , braggio a , magnoli n and sassetti m 2012 _ new j. phys . _ * 14 * 023017 bednorz a , bruder c , reulet b , belzig w 2013 _ phys . rev .
* 110 * 250404 lesovik g b and loosen r 1997 _ jetp lett . _ * 65 * 295 zakka - bajjani e , segala j , portier f , glattli d c , cavanna a and jin j 2007 _ phys .
_ * 99 * 236803 ferraro d , carrega m , braggio a and sassetti m 2014 _ new j. phys . _
* 043018 fendley p , fisher m p a and nayak c 2007 _ phys . rev . b _ * 75 * 045317 lee s -s , ryu s , nayak c and fisher m p a 2007 _ phys . rev . lett . _ * 99 * 236807 levin m , halperin b i and rosenow b 2007 _ phys .
lett . _ * 99 * 236806 boyarsky a , cheianov v and froehlich j 2009 _ phys . rev .
b _ * 80 * 233302 hu z -x , rezayi e h , wan x and yang k 2009 _ phys . rev .
b _ * 80 * 235330 bishara w , fiete g a and nayak c 2008 _ phys . rev .
b _ * 77 * 241306(r ) ginsparg p 1989 _ applied conformal field theory _ , les houches lectures , ed .
brezin e and zinn - justin j ed .
( amsterdam : north holland ) weiss u 1999 _ quantum dissipative system _ ( singapore : world scientific ) cuniberti g , sassetti m and kramer b 1996 _ j. phys . : condens
* 8 * l21 braggio a , grifoni m , sassetti m and napoli f 2000 _ europhys . lett . _ * 50 * 236 braggio a , sassetti m , and kramer b 2001 _ phys .
lett . _ * 87 * 146802 kane c l and fisher m p a 1992 _ phys .
lett . _ * 68 * 1220 braggio a , ferraro d , magnoli n 2012 _ phys .
_ * t151 * 014052 braggio a , ferraro d , carrega m , magnoli n and sassetti m 2012 _ new j. phys . _ * 14 * 093032 altimiras c , parlavecchio o , joyez p , vion d , roche p , esteve d , portier f 2013 _ appl
lett . _ * 103 * 212601 altimiras c , parlavecchio o , joyez p , vion d , roche p , esteve d , portier f 2014 _ phys . rev . lett . _
* 112 * 236803 martin t 2005 _ course 5 noise in mesoscopic physics _ les houches , france gavish u , levinson y , and imry y _ phys .
b _ * 67 * 10637 r aguado r and l. p.
kouwenhoven l p 2000 _ phys . rev .
lett . _ * 84 * , 1986 creux m , crpieux a and martin t 2006 _ phys .
b _ * 74 * 115323 zazunov a , creux m , paladino e , crpieux a and martin t 2007 _ phys .
lett . _ * 99 * 066601 chevallier d , jonckheere t , paladino e , falci g and martin t 2010 _ phys
b _ * 81 * 205411 blanter ya
m and bttiker m 2000 _ phys .
rep . _ * 336 * 1 gavish u , yurke b , and imry y 2004 _ phys .
lett . _ * 93 * 250601 gavish u , levinson y and imry y 2001 _ phys
lett . _ * 87 * 216807 | we investigate the finite frequency noise of a quantum point contact at filling factor @xmath0 using a weakly coupled resonant lc circuit as a detector .
we show how one could spectroscopically address the fractional charged excitations inspecting separately their charge and scaling dimensions .
we thus compare the behaviour of the pfaffian and the anti - pfaffian non - abelian edge states models in order to give possible experimental signatures to identify the appropriate model for this fractional quantum hall states .
finally we investigate how the temperature of the lc resonant circuit can be used in order to enhance the sensibility of the measurement scheme . |
ultra - cold atom gases provide a unique opportunity to study basic models of strongly interacting many - body systems in a highly controlled experimental framework @xcite . after the pioneering realization of the superfluid - to - mott insulator transition of ultra - cold bosonic atoms @xcite ,
recently different groups reported the realization of a mott - insulating state also for ultra - cold @xmath2k fermionic atoms on three - dimensional simple - cubic lattices @xcite .
a next major step would be the realization and identification of a low - temperature antiferromagnetically ordered nel state in such systems .
in fact , various recent studies explored the prospects of realizing the antiferromagnetic nel state in an ultra - cold gas of fermionic atoms confined to an optical lattice @xcite .
a quantity that is important in assessing the relevant temperature scale is the upper bound on the entropy allowed in order to transfer the atoms into the antiferromagnetically ordered state . for a three - dimensional fermionic hubbard model on a simple - cubic lattice at half - filling , estimates of the entropy @xmath3 as a function of temperature , as well as the ratio @xmath4 between the onsite repulsion @xmath5 and the hopping amplitude @xmath6 , were obtained within a single - site dynamical mean - field theory ( dmft ) approach @xcite . as reported in ref .
, dmft however overestimates the nel temperature by about @xmath7 in the intermediate coupling regime , as compared to direct quantum monte carlo simulations based on systems with up to @xmath8 lattice sites @xcite .
obtaining in a similar way the entropy from direct quantum monte carlo simulations is challenging , and typically involves integration of the specific heat over an extended temperature range .
however , in the limit of large @xmath4 , the spin physics of the hubbard model is well known to be described by a nearest - neighbor spin-1/2 quantum heisenberg model , with an antiferromagnetic exchange coupling @xmath9 , obtained within perturbation theory around the strong coupling limit @xmath10 .
this model is accessible to large scale quantum monte carlo simulations and moreover as shown below
it is possible to provide precise values of the entropy in this large-@xmath5 limit .
in particular , one can obtain the value of the critical entropy density ( i.e. the entropy per lattice site ) @xmath11 , below which antiferromagnetic order emerges .
nevertheless , thus far no systematic quantum monte carlo study concerning the entropy has been reported . on the other hand , different estimates of @xmath11 for the heisenberg antiferromagnet
have been put forward recently . in ref . , a schwinger boson approach @xcite was employed , leading to the estimate that @xmath11 is about @xmath7 of the mean - field value @xmath12 .
a more recent study @xcite reported a value of @xmath11 obtained from a fluctuation - corrected mean - field approach , leading to a reduction in @xmath11 of only @xmath13 from the mean - field value @xmath14 .
this rather large discrepancy among current estimates of @xmath11 calls for a clarification based on numerically exact simulations of the heisenberg model .
( dashed lines ) and @xmath15 ( solid lines ) .
error bars for the shown data are below the line thickness .
vertical lines show the transition temperatures @xmath16 and @xmath17 for the antiferromagnet and the ferromagnet , respectively .
green dashed - dotted curves denote fits to the algebraic low-@xmath18 scaling @xmath19 ( @xmath20 ) for the antiferromagnet ( ferromagnet ) .
, width=321 ] here , we present an unbiased estimate of the temperature dependence of the entropy and in particular the critical entropy density by combining extended ensemble quantum monte carlo simulations with a finite - size scaling analysis based on the known critical properties of the quantum heisenberg model . for comparison , we consider in the following both the ferromagnetic and the antiferromagnetic spin-1/2 heisenberg model on the three - dimensional simple - cubic lattice .
these models are described by a hamiltonian @xmath21 where @xmath22 denotes a spin-1/2 degree of freedom on lattice site @xmath23 , and @xmath24 is nearest - neighbor exchange interaction , with @xmath25 ( @xmath26 ) , for the antiferromagnetic ( ferromagnetic ) case . from previous studies , the locations of the transition temperatures to the ordered phases
have been determined as @xmath27 @xcite ( nel temperature ) for the antiferromagnet , and @xmath28 @xcite ( curie temperature ) for the ferromagnet , based on quantum monte carlo simulations .
in order to extract the temperature dependence of the entropy , we use an optimized extended ensemble approach @xcite , that is based on a generalization of the wang - landau @xcite algorithm to the case of quantum monte carlo simulations @xcite , performed within the stochastic series expansion representation @xcite using multi - cluster deterministic loop updates @xcite . within this approach ,
one obtains monte carlo estimates of the expansion coefficients @xmath29 of the high - temperature series expansion of the partition function @xmath30 in the inverse temperature @xmath31 , @xmath32 for a given system of @xmath33 lattice sites . here , @xmath34 denotes the linear size of the simulated finite cube , and we employ periodic boundary conditions in all directions . from the expansion coefficients @xmath29 , the free energy @xmath35 the internal energy @xmath36 and the entropy @xmath37
are obtained as continuous functions of the temperature @xmath18 . in practice , as discussed in ref .
, we obtain the expansion coefficients up to a upper cutoff @xmath38 , that is chosen sufficiently large in order to reliably calculate the thermodynamic properties of the finite system based on the coefficients @xmath29 , @xmath39 down to a given temperature scale @xmath40 . the required @xmath38 scales linear in both @xmath40 and the system size @xmath41 . in our simulations , we consider finite systems up to @xmath42 , for which @xmath43 is required in order to obtain thermodynamic properties down to a temperature of about @xmath44 . repeated simulations with independent random streams results in statistical error bars on the thermodynamic quantities . with the inverse linear system size @xmath45 for the spin-1/2 quantum heisenberg antiferromagnet and ferromagnet on the simple - cubic lattice .
the quantum monte carlo data are shown by symbols , and the dashed lines represent fits of the finite - size data for @xmath46 to the finite - size scaling form in eq .
the results for @xmath11 in the thermodynamic limit are indicated by symbols as well .
, width=321 ]
in fig . 1 , we present the results for the entropy from our simulations for both the antiferromagnetic and the ferromagnetic heisenberg model . for both cases ,
we show data for systems with @xmath47 and @xmath15 . outside a narrow range around the transition temperature ,
the entropy curves obtained for the two different system sizes compare well , and thus essentially represent the behavior of the entropy density in the thermodynamic limit . also shown in fig
. 1 are fits of the numerical data to the low-@xmath18 scaling @xmath19 and @xmath20 for the antiferromagnet and ferromagnet , respectively .
this difference in the low-@xmath18 scaling behavior relates to the linear ( quadratic ) dispersion of the magnetic spin - wave excitations in the antiferromagnetic ( ferromagnetic ) case @xcite .
however , in order to extract the critical entropy @xmath11 in the thermodynamic limit , we need to perform a finite - size scaling analysis of the finite - size data @xmath48 taken at the transition temperature . for this purpose
, we use a finite - size scaling ansatz @xmath49 where @xmath50 is the critical exponent of the correlation length , and @xmath51 the dimensionality of the system . starting from @xmath52
, this relation follows from the scaling behavior of the free energy @xmath53 with the correlation length @xmath54 , that scales as @xmath55 when the reduced temperature @xmath56 , by the usual finite - size scaling ansatz , that at criticality @xmath54 relates to the linear system size @xmath34 . for the current case of a three - dimensional system ( @xmath57 ) with @xmath58 symmetry , @xmath59 @xcite .
figure 2 shows that the finite - size data indeed fits well to the above scaling form in the range of system sizes @xmath46 . from the extrapolated values , we obtain @xmath60 for the antiferromagntic model , whereas for the ferromagnet , a larger value of @xmath61 results .
we also performed an algebraic fit without biasing the exponent , which returned identical values of @xmath11 within the statistical error bars . while the ferromagnet has a critical temperature @xmath17 that is about @xmath62 _ lower _ than @xmath16 , it orders already below an entropy that is about @xmath63 _ higher _ than the critical entropy of the antiferromagnet .
the higher critical entropy directly exhibits a reduced efficiency of thermal disorder in destroying long - range magnetic order .
the lower transition temperature of the ferromagnetic is due to a significantly enhanced production of entropy upon heating up the system , which follows from both fig . 1 , as well as from the mentioned difference in the low-@xmath18 behavior of the entropy .
based on numerically exact quantum monte carlo simulations , we obtained the temperature dependence of the entropy for both the antiferromagnetic and the ferromagnetic spin-1/2 heisenberg model on the simple - cubic lattice .
focusing on the region close to the magnetic ordering transition , we employed a finite - size scaling analysis to extract the critical entropy density below which long - ranged magnetic order sets in . in case of the antiferromagnetic model , our result for the critical entropy is consistent with a @xmath7 reduction from the mean - field value , that was estimated recently from a schwinger boson calculation@xcite .
this clarifies among the different values of the critical entropy reported previously@xcite . in comparison
, we found the ferromagnet to be more robust to thermal disorder , which at criticality is about @xmath63 larger for the ferromagnet than for the antiferromagnet in terms of the critical entropy .
we consider the current results for the temperature dependence of the magnetic entropy of value also for future calculations of the entropy for the full fermionic hubbard model beyond the strongly interacting regime .
we thank n. blmer , a. muramatsu , and l. pollet for discussions , and acknowledge the allocation of cpu time on the hlrs stuttgart and nic jlich supercomputers .
support was also provided through the dfg within sfb / trr 21 . | we analyze the temperature dependence of the entropy of the spin-1/2 heisenberg model on the three - dimensional simple - cubic lattice , for both the case of antiferromagnetic and ferromagnetic nearest neighbor exchange interactions . using optimized extended ensemble quantum monte carlo simulations
, we extract the entropy at the critical temperature for magnetic order from a finite - size scaling analysis . for the antiferromagnetic case , the critical entropy density equals @xmath0 , whereas for the ferromagnet
, a larger value of @xmath1 is obtained .
we compare our simulation results to estimates put forward recently in studies assessing means of realizing the antiferromagnetic nel state in ultra - cold fermion gases in optical lattices . |
peeling is a kind of fracture that has been studied experimentally in the context of adhesion and is a technologically important subject .
experimental studies on peeling of an adhesive tape mounted on a cylindrical roll are usually in constant pull speed condition @xcite .
more recently , constant load experiments have also been reported @xcite .
early studies by bikermann @xcite , kaeble @xcite have attempted to explain the results by considering the system as a fully elastic object .
this is clearly inadequate as it ignores the viscoelastic nature of the glue at the contact surface and therefore can not capture many important features of the dynamics .
the first detailed experimental study of maugis and barquins @xcite show stick - slip oscillations within a window of pull velocity with decreasing amplitude of the pull force as a function of the pull velocity .
further , these authors report that the pull force shows sinusoidal , sawtooth and highly irregular ( chaotic as these authors refer to ) wave patterns with increasing velocities . more recently ,
gandur _ et al .
_ have carried out a dynamical time series analysis of the force waveforms , as well as those of acoustic emission signals and report chaotic force waveforms at the upper end of the pull velocities @xcite .
one characteristic feature of the peeling process is that the experimental strain energy release rate shows two stable branches separated by an unstable branch .
stick - slip behavior is commonly observed in a number of systems such as jerky flow or the portevin - le chatelier ( plc ) effect @xcite , frictional sliding @xcite , and even earthquake dynamics is thought to result from stick - slip of tectonic plates @xcite .
stick - slip is characterized by the system spending most part of the time in the stuck state and a short time in the slip state , and is usually seen in systems subjected to a constant response where the force developed in the system is measured by dynamically coupling the system to a measuring device .
one common feature of such systems is that the force exhibits negative flow rate characteristic " ( nfrc ) .
models which attempt to explain the dynamics of such systems use the macroscopic phenomenological nfrc feature as an input , although the unstable region is not accessible .
this is true for models dealing with the dynamics of the adhesive tape as well .
to the best of our knowledge , there is no microscopic theory which predicts the origin of the nfrc macroscopic law except in the case of the plc effect @xcite ( see below ) . as there is a considerable similarity between the peeling of an adhesive tape and the plc effect , it is useful to consider the similarities in some detail .
the plc effect refers to a type of plastic instability observed when samples of dilute alloys are deformed under constant cross head speeds @xcite .
the effect manifests itself in the form of a series of serrations in a range of applied strain rates and temperatures .
this feature is much like the peeling of an adhesive tape .
other features common to these two situations are : abrupt onset of the large amplitude oscillations at low applied velocities with a gradually decreasing trend and nfrc , which in the plc effect refers to the existence of negative strain rate sensitivity of the flow stress . in the case of the plc effect , the physical origin of the negative strain rate sensitivity is attributed to the ageing of dislocations and their tearing away from the cloud of solute atoms .
recently , the origin of the negative srs has been explicitly demonstrated as arising from competing time scales of pinning and unpinning in the ananthakrishna s model @xcite . in the case of adhesive tape
, the origin of nfrc can be attributed to the viscoelastic behavior of the fluid .
( constant load and constant load rate experiments are possible in the plc also . )
while simple phenomenological models based on nfrc explain the generic features of the plc effect @xcite , there appears to be some doubts if the equations of motion conventionally used in the present case of peeling are adequate to describe the velocity jumps @xcite .
indeed , these equations of motion are singular and pose problems in the numerical solutions . apart from detailed experimental investigation of the peeling process , maugis and barquins @xcite , have also contributed substantially to the understanding of the dynamics of the peeling process . however , the first dynamical analysis is due to hong and yue @xcite who use an n " shaped function to mimic the dependence of the peel force on the rupture speed .
they showed that the system of equations exhibits periodic and chaotic stick - slip oscillations .
however , the jumps in the rupture speed are introduced _ externally _ once the rupture velocity exceeds the limit of stability @xcite .
thus , the stick - slip oscillations are _ not _ obtained as a natural consequence of the equations of motion .
therefore , in our opinion the results presented in ref .
@xcite are the artifacts of the numerical procedure followed .
et al . _
@xcite interpret the stick - slip jumps as catastrophes .
again , the belief that the jumps in the rupture velocity can not be obtained from the equations of motion appears to be the motivation for introducing the action of discrete operators on the state of the system to interpret the stick - slip jumps @xcite , though they do not demonstrate the correctness of such a framework for the set of equations .
lastly , there are no reports that explain the decrease in the amplitude of the peel force with increasing pull speed as observed in experiments .
_ as there is a general consensus that these equations of motion correctly describe the experimental system , a proper resolution of this question ( on the absence of dynamical jumps in these equations ) assumes importance_. the purpose of this paper is to show that the dynamics of stick - slip during peeling can be explained using a differential - algebraic scheme meant for such singular situations @xcite and demonstrate the rich dynamics inherent to these equations . in what follows
we first derive the equations of motion ( used earlier @xcite ) by introducing an appropriate lagrangian for the system .
then , we use an algorithm meant to solve differential - algebraic equations @xcite and present the results of our simulations for various parameter values .
one of our major findings is that inertia has a strong influence on the dynamics .
in addition , following the dynamization scheme similar to the one used in the context of the plc effect @xcite , we suggest that the peel force depends on the applied velocity . using this form of peel force leads to the decreasing nature of the magnitude of the pull force as a function of applied velocity . for certain values of the inertia
, we find canard type solutions .
these numerical results are captured to a reasonable accuracy using a set of approximations valid in different regimes of the parameter space .
even though , our emphasis is on demonstrating the correctness of these equations of motion and richness of the inherent dynamics that capture the qualitative features of the peeling process , we also attempt to make a comparison of the experimental results mentioned above to the extent possible .
for the sake of completeness , we start by considering the geometry of the experimental setup shown schematically in fig .
[ fig1 ] . an adhesive roll of radius @xmath0 is mounted on an axis passing through @xmath1 normal to the paper and is pulled at a constant velocity @xmath2 by a motor positioned at @xmath3 with a force @xmath4 acting along @xmath5 .
let the distance between @xmath1 and @xmath3 be @xmath6 , and that between the contact point @xmath7 to @xmath3 be @xmath8 .
the point @xmath7 moves with a local velocity @xmath9 which can undergo rapid bursts in the velocity during rupture .
the force required to peel the tape is usually called the force of adhesion denoted by @xmath10 .
the two measured branches referred to earlier , are those of the function @xmath10 in a steady state situation of constant pulling velocity ( i.e. , there are no accelerations ) .
the line @xmath8 makes an angle @xmath11 with the tangent at the contact point @xmath7 .
the point @xmath7 subtends an angle @xmath12 at @xmath1 , with the horizontal line @xmath13 .
we denote the elastic constant of the adhesive tape by @xmath14 , the elastic displacement of the tape by @xmath15 , the angular velocity by @xmath16 and the moment of inertia of the roll by @xmath17 .
the angular velocity itself is identified by @xmath18 .
the geometry of the setup gives @xmath19 and @xmath20 which further gives , @xmath21 .
the total velocity @xmath2 at @xmath22 is then made up of three contributions @xcite , given by @xmath23 which gives @xmath24 following standard methods in mechanics , it is straightforward to derive the equations of motion for @xmath12 and @xmath16 by considering @xmath25 as the generalized co - ordinates .
the corresponding lagrangian of the system can be written as @xmath26 ^ 2 - { k\over2}u^2 .
\label{lagran}\end{aligned}\ ] ] we write the dissipation function as @xmath27 where @xmath28 physically represents the peel force which we assume is dependent on rupture speed as well as the pull speed assumed to be derivable from a potential function @xmath29 .
the physical origin of this is due to the competition between the internal relaxation time scale of the viscoelastic fluid and the time scale determined by the applied velocity @xcite .
when the applied velocity is low , there is sufficient time for the viscoelastic fluid to relax .
as we increase the applied velocity , the relaxation of the fluid gets increasingly difficult and thus behaves much like an elastic substance .
the effect of competing time scales is well represented by deborah number @xcite which is the ratio of time scale for structural relaxation to the characteristic time scale for deformation . indeed , in the studies on hele - shaw cell with mud as the viscous fluid
, one observes a transition from viscous fingering to viscoelastic fracturing @xcite with increasing rate of invasion of the displacing fluid . as stated in the introduction
, the existing models do not explain the decreasing amplitude of pull force .
similar feature observed in the plc serrations has been modeled using a scheme referred to as dynamization of the negative strain rate sensitivity ( srs ) of the flow stress @xmath30 @xcite , where @xmath31 is the plastic strain rate .
based on arguments similar to the preceding paragraph , they modify this function to depend on the applied strain rate , @xmath32 , i.e. , the negative srs of the flow stress is taken to be @xmath33 such that the gap between the maximum and the minimum of the function @xmath34 decreases with increasing @xmath32 . following this , we consider @xmath10 to depend on @xmath2 also , in a way that the gap in @xmath10 decreases as a function of the pull speed @xmath2 ( fig .
[ fig2 ] ) . using the lagrange equations of motion , @xmath35 we obtain the same set of ordinary differential equations as in ref .
@xcite given by @xmath36 \label{flow3},\end{aligned}\ ] ] with an algebraic constraint @xmath37 ( the last equation results from the elimination of two second order equations for @xmath12 . ) in eqs .
( [ flow2 ] ) , ( [ flow3 ] ) , and ( [ constr ] ) we have used @xmath38 . while eqs .
( [ flow1])-([flow3 ] ) are differential equations , eq .
( [ constr ] ) is an algebraic constraint necessitating the use of differential - algebraic scheme to obtain the numerical solution @xcite . the fixed point of eqs .
( [ flow1 ] ) , ( [ flow2 ] ) , ( [ flow3 ] ) , and ( [ constr ] ) is given by @xmath39 .
( for numerical solution , in the above equations we have actually used @xmath40 in place of @xmath41 . )
this point is stable for @xmath42 and unstable for @xmath43 . as @xmath2 is varied such that the sign of @xmath44 changes from negative to positive value , the system undergoes a hopf bifurcation and a limit cycle appears .
the limit cycles reflect the abrupt jumps between the two positive slope branches of the function @xmath28 .
the singular nature of these equations becomes clear if one were to consider the differential form of eq .
( [ constr ] ) given by @xmath45 , \label{vdot } \\ & \simeq & [ \dot f ( 1 + \alpha ) + f \dot \alpha]/f^{\prime } , \label{vdotapprox}\end{aligned}\ ] ] where the prime denotes the derivative with respect to @xmath9 .
equations ( [ vdot ] ) with eq . ( [ flow1 ] ) , ( [ flow2 ] ) , and ( [ flow3 ] ) [ or ( [ flow3 ] ) ] constitute the full set of evolution equations for the vector @xmath46 . however , it is clearly singular at points of extremum of @xmath28 , requiring an appropriate numerical algorithm .
we note that eqs.([flow1 ] ) , ( [ flow2 ] ) , ( [ flow3 ] ) , and ( [ constr ] ) can be written as @xmath47 where @xmath48 , @xmath49 is a vector function that governs the evolution of @xmath50 and @xmath51 is a singular `` _ mass matrix _ ''
@xcite given by , @xmath52 equation ( [ xdae ] ) is a differential - algebraic equation ( dae ) and can be solved using the so called singular perturbation technique @xcite in which the singular matrix @xmath51 is perturbed by adding a small constant @xmath53 such that the singularity is removed .
the resulting equations can then be solved numerically and the limit solution obtained as @xmath54 .
we have checked the numerical solutions for @xmath53 values ranging from @xmath55 to @xmath56 in some cases and the results do not depend on the value of @xmath53 used as long as it is small .
the results presented below , however , are for @xmath57 .
we have solved eq .
( [ xdae ] ) using a standard variable - order solver , matlab ode15s program .
we have parametrized the form of @xmath28 as @xmath58 to give values of the extremum of the peel velocity that mimic the general form of the experimental curves @xcite .
the measured strain energy release rate @xmath59 from stationary state measurements is shown in fig .
the decreasing nature of the gap between the maximum and minimum of @xmath28 for increasing @xmath2 is clear from fig .
[ the values of @xmath28 could not be correctly determined as @xmath59 is in j / m@xmath60 requiring more details . however , the value of @xmath61 is closer to ref . @xcite and the jumps in @xmath9 are similar to those in experiments . ]
the reason for using the form given by eq .
( [ fvv ] ) is that the effects of dynamization are easily included through its dependence on the pulling velocity while more complicated terms are required to mimic completely the experimental curve ( particularly the flat portion ) .
however , we stress that the trend of the results remains unaffected when the actual experimental curve is used except for the magnitude of velocity jumps and the force values .
plane for @xmath62 .
the corresponding @xmath28 is shown by a solid curve .
( b ) a phase space trajectory in the @xmath63 plane for @xmath64 and @xmath65 . ( c ) a plot of @xmath66 for @xmath67 and @xmath65 .
( d ) a plot of @xmath68 ( period 4 ) for @xmath69 and @xmath65 .
( units of @xmath9 , @xmath2 are in m / s , @xmath4 in @xmath70 , @xmath17 in kg m@xmath60 and @xmath71 in s. ) , title="fig:",width=302,height=151 ] plane for @xmath62 .
the corresponding @xmath28 is shown by a solid curve .
( b ) a phase space trajectory in the @xmath63 plane for @xmath64 and @xmath65 . ( c )
a plot of @xmath66 for @xmath67 and @xmath65 .
( d ) a plot of @xmath68 ( period 4 ) for @xmath69 and @xmath65 .
( units of @xmath9 , @xmath2 are in m / s , @xmath4 in @xmath70 , @xmath17 in kg m@xmath60 and @xmath71 in s. ) , title="fig:",width=302,height=151 ] plane for @xmath62 .
the corresponding @xmath28 is shown by a solid curve .
( b ) a phase space trajectory in the @xmath63 plane for @xmath64 and @xmath65 . (
c ) a plot of @xmath66 for @xmath67 and @xmath65 .
( d ) a plot of @xmath68 ( period 4 ) for @xmath69 and @xmath65 .
( units of @xmath9 , @xmath2 are in m / s , @xmath4 in @xmath70 , @xmath17 in kg m@xmath60 and @xmath71 in s. ) , title="fig:",width=302,height=151 ] plane for @xmath62 .
the corresponding @xmath28 is shown by a solid curve .
( b ) a phase space trajectory in the @xmath63 plane for @xmath64 and @xmath65 . ( c ) a plot of @xmath66 for @xmath67 and @xmath65 .
( d ) a plot of @xmath68 ( period 4 ) for @xmath69 and @xmath65 .
( units of @xmath9 , @xmath2 are in m / s , @xmath4 in @xmath70 , @xmath17 in kg m@xmath60 and @xmath71 in s. ) , title="fig:",width=302,height=151 ] we have studied the dynamics of the system of equations for a wide range of values of the parameters .
we have found that transients for some regions of parameters space take considerable time to die out .
the results reported here are obtained after these long transients are omitted .
these equations exhibit rich dynamics , some even unanticipated . here
we report typical results for two important parameters , namely the pull velocity @xmath2 ( m / s ) and the inertia @xmath17 ( kg m@xmath60 ) , keeping the elastic constant of the tape @xmath72 n / m , @xmath73 m and @xmath74 m @xcite .
the influence of @xmath14 will also be mentioned briefly .
( henceforth , we drop the units for the sake of brevity . )
we find that the observed jumps of the orbit in the @xmath9-@xmath4 plane occur in a fully dynamical way .
more importantly , we find all the three possibilities namely , the orbit can jump when it approaches the limit of stability , before or beyond that permitted by @xmath28
. the dynamics can be broadly classified into low , intermediate and high regimes of inertia .
+ ( i ) low inertia . here also , there are three regimes : low , intermediate , and high pull velocity .
+ ( a ) consider keeping inertia @xmath17 at a low value ( say @xmath75 ) and @xmath2 also at a low value ( say , near the top , say @xmath76 ) .
here we observe regular saw tooth form for the pull force @xmath4 .
the phase plot in the @xmath4-@xmath9 plane is as shown in fig .
the corresponding function @xmath28 is also shown by the continuous curve .
we see that the trajectory jumps almost instantaneously from @xmath77 to @xmath78 on reaching the maximum of @xmath28 ( or from @xmath79 to @xmath80 when it reaches the minimum ) .
the system spends considerably more time on @xmath81 compared to that on @xmath82 .
however , this feature of jumping of the trajectory at the limit of stability is only true for small values of @xmath17 and when @xmath2 is near the limit of stability . at slightly higher pull velocity ,
say @xmath83 , even for small @xmath17 , say @xmath84 , the jumps occur even before reaching the top or bottom ( the points @xmath77 and @xmath79 ) as can be seen from fig .
[ fig3](b ) for @xmath67 . the small amplitude high frequency oscillations seen in the phase plots [ fig .
[ fig3](a ) , and [ fig3](b ) ] on the branch @xmath81 are due to the inertial effect , i.e. , finite value of @xmath17 .
these oscillations are better seen on the @xmath66 plot shown in fig .
for these values of parameters , the system is aperiodic .
+ b ) as we increase @xmath2 , even as the saw tooth form of @xmath4 is retained , various types of periodic orbits [ period 4 shown in fig . [ fig3](d ) for @xmath85 as well as irregular orbits are seen . in both cases ( periodic as well as chaotic ) the trajectory jumps from high velocity branch ( @xmath82 ) to the low velocity branch before traversing the entire branch or sometimes going beyond the values permitted by @xmath10 .
the value of @xmath4 at which the orbit jumps is different for different cycles . for @xmath86 ,
at high velocity , say @xmath87 , the phase plot is periodic .
+ plane for a single cycle for @xmath88 and @xmath67 .
the corresponding @xmath28 is shown by a thick solid curve .
( b ) corresponding plots of @xmath66 , ( c ) the pull force @xmath68 ( period 8) and ( d ) the peel velocity @xmath89 .
( units of @xmath9 , @xmath2 are in m / s , @xmath4 in n , @xmath17 in kg m@xmath60 and @xmath71 in s. ) , title="fig:",width=302,height=188 ] plane for a single cycle for @xmath88 and @xmath67 .
the corresponding @xmath28 is shown by a thick solid curve .
( b ) corresponding plots of @xmath66 , ( c ) the pull force @xmath68 ( period 8) and ( d ) the peel velocity @xmath89 .
( units of @xmath9 , @xmath2 are in m / s , @xmath4 in n , @xmath17 in kg m@xmath60 and @xmath71 in s. ) , title="fig:",width=302,height=151 ] plane for a single cycle for @xmath88 and @xmath67 .
the corresponding @xmath28 is shown by a thick solid curve .
( b ) corresponding plots of @xmath66 , ( c ) the pull force @xmath68 ( period 8) and ( d ) the peel velocity @xmath89 .
( units of @xmath9 , @xmath2 are in m / s , @xmath4 in n , @xmath17 in kg m@xmath60 and @xmath71 in s. ) , title="fig:",width=302,height=188 ] plane for a single cycle for @xmath88 and @xmath67 .
the corresponding @xmath28 is shown by a thick solid curve .
( b ) corresponding plots of @xmath66 , ( c ) the pull force @xmath68 ( period 8) and ( d ) the peel velocity @xmath89 .
( units of @xmath9 , @xmath2 are in m / s , @xmath4 in n , @xmath17 in kg m@xmath60 and @xmath71 in s. ) , title="fig:",width=302,height=151 ] \(ii ) intermediate and high inertia .
+ ( a ) as the results of small @xmath2 for intermediate and high inertia are similar , we illustrate the results for @xmath88 and @xmath90 .
the @xmath63 phase plot , @xmath91 and @xmath9 are shown in fig .
[ fig4](a)-[fig4](d ) .
consider , fig .
[ fig4](a ) showing a typical phase space trajectory for a single cycle .
the corresponding function @xmath28 is also shown by the thick continuous curve .
we see that the maximum ( and minimum ) value of @xmath4 is larger ( or smaller ) than that allowed by @xmath28 .
[ this feature holds when the inertia is in the intermediate regime also , though the values of maxima ( minima ) of @xmath4 are not significantly larger ( less ) than @xmath92 ( @xmath93 ) .
] when the trajectory jumps from @xmath81 to @xmath82 at the highest value of @xmath4 for the cycle , the trajectory stays on @xmath82 for a significantly shorter time compared to the small inertia case ( @xmath65 ) and jumps back to @xmath81 well before @xmath4 has reached the minimum of @xmath28 , i.e. , @xmath94 is much smaller than @xmath95 .
the pull force @xmath4 cascades down through a series of back and forth jumps between the two branches till the lowest value of @xmath4 for the cycle is reached .
note that @xmath4 at the point @xmath96 is less than @xmath93 .
for the sake of clarity , two different portions of the trajectory are marked @xmath97 and @xmath98 corresponding to the top and bottom regions of the plot .
the corresponding points are also identified on the @xmath68 plot . after reaching @xmath96
, the orbit jumps to @xmath99 on @xmath81 , the trajectory decides to move up all the way till @xmath4 reaches a maximum value ( larger than @xmath92 , the point @xmath100 ) without jumping to the @xmath82 branch .
this part of @xmath4 as a function of time , which is nearly linear on @xmath81 , ( i.e. , the segment @xmath101 ) displays a noticeable sinusoidal modulation .
the sinusoidal form is better seen in @xmath12 [ fig . [ fig4](b ) ] .
note that the successive drops in @xmath4 are of increasing magnitude .
the jumps between the two branches in the @xmath63 plane are seen as bursts of @xmath9 [ fig .
[ fig4](d ) ] .
for these values of parameters , the system is periodic .
+ ( b ) as we increase @xmath2 , the sinusoidal nature of @xmath4 and @xmath12 becomes more clear with its range becoming larger reaching a nearly sinusoidal at @xmath87 for large @xmath17 .
[ the range @xmath101 in fig . [ fig4](c ) expands .
compare fig .
[ fig5](a ) . ]
the magnitude of @xmath94 on the @xmath82 branch for small @xmath2 and moderately or large @xmath17 , gradually decreases with increasing @xmath2 .
the magnitude of @xmath94 itself decreases as @xmath17 is increased . in the limit of large @xmath2 and @xmath17 ,
the drops in @xmath102 and @xmath12 become quite small which are now located near the maxima and minima of these curves .
this is shown in figs .
[ fig5](a ) and [ fig5](b ) .
the sinusoidal nature is now obvious even in @xmath68 unlike for smaller @xmath2 and @xmath17 where it is clear only in @xmath66 for the low @xmath9 branch .
note that for @xmath103 , the nature of @xmath104 is nearly flat .
this induces certain changes in the @xmath63 phase plot that are not apparent in @xmath4 and @xmath12 .
the jumps between the two branches are now concentrated in a dense band at low and high values of @xmath4 . in this case , the maximum ( minimum ) value of @xmath4 is significantly larger ( less ) than @xmath92 ( @xmath93 ) .
these rapid jumps between the branches manifest as jitter at the top and bottom of @xmath4 and @xmath12 . for @xmath105 and @xmath106 .
( b ) corresponding plot of @xmath66 .
the inset shows an expanded plot of decreasing trend of @xmath66 .
( c ) corresponding plots of phase space trajectory that reflects the chaotic nature and ( d ) the peel velocity @xmath89 .
( @xmath9 , @xmath2 are in m / s , @xmath4 in n , @xmath17 in kg m@xmath60 and @xmath71 in s. ) , title="fig:",width=340,height=188 ] for @xmath105 and @xmath106 .
( b ) corresponding plot of @xmath66 .
the inset shows an expanded plot of decreasing trend of @xmath66 .
( c ) corresponding plots of phase space trajectory that reflects the chaotic nature and ( d ) the peel velocity @xmath89 .
( @xmath9 , @xmath2 are in m / s , @xmath4 in n , @xmath17 in kg m@xmath60 and @xmath71 in s. ) , title="fig:",width=340,height=207 ] for @xmath105 and @xmath106 .
( b ) corresponding plot of @xmath66 .
the inset shows an expanded plot of decreasing trend of @xmath66 .
( c ) corresponding plots of phase space trajectory that reflects the chaotic nature and ( d ) the peel velocity @xmath89 .
( @xmath9 , @xmath2 are in m / s , @xmath4 in n , @xmath17 in kg m@xmath60 and @xmath71 in s. ) , title="fig:",width=340,height=207 ] for @xmath105 and @xmath106 .
( b ) corresponding plot of @xmath66 .
the inset shows an expanded plot of decreasing trend of @xmath66 .
( c ) corresponding plots of phase space trajectory that reflects the chaotic nature and ( d ) the peel velocity @xmath89 .
( @xmath9 , @xmath2 are in m / s , @xmath4 in n , @xmath17 in kg m@xmath60 and @xmath71 in s. ) , title="fig:",width=340,height=188 ] unlike for small @xmath2 [ fig . [ fig4](a ) ] , the nature of the trajectory in fig .
[ fig5](c ) is different . after reaching a critical value of @xmath4 near the maximum value of @xmath4 ( the point @xmath100 )
, the orbit spirals upwards and then descends down till another critical value of @xmath4 ( the point @xmath107 ) is reached .
having reached @xmath107 , the orbit monotonically comes down till @xmath108 where it jumps to the ab branch . beyond this point
, it again spirals upwards till the point @xmath99 is reached .
thereafter , @xmath4 monotonically increases till @xmath100 is reached .
the regions @xmath101 and @xmath109 are the regions where @xmath4 shows a near sinusoidal form .
the regions @xmath110 and @xmath111 are the regions where the orbit jumps between the branches rapidly . these manifest themselves as bursts of @xmath9 which tend to bunch together almost into a band .
[ compare fig .
[ fig4](d ) with fig .
[ fig5](d ) .
] it is interesting to note that the jumps between the two branches occur exactly at points where @xmath112 , even when the maximum ( minimum ) of @xmath4 are higher ( lower ) than that allowed by the stationary curve @xmath28 .
the variables are aperiodic for the set of parameters .
the phase plots appear to be generated by an effective @xmath28 that is being cycled .
[ this visual feeling is mainly due to the fact that jumps between the branches still occur at the maximum and minimum of @xmath28 . ]
the influence of @xmath14 is generally to increase the range of the pull force @xmath4 as can be easily anticipated and to decrease the associated time scale
. it may be desirable to comment on the similarity of the nature of the force waveforms displayed by the model equations with those seen in experiments .
as mentioned in the introduction , apart from qualitative statements on the waveforms in ref .
@xcite ( such as periodic , sawtooth etc .
, which are seen in the model as well ) , it should be stressed that there is a paucity of quantitative characterization of the waveforms . in this respect , the study by gandur _ et al . _
@xcite fills the gap to some extent .
these authors have carried out a dynamical analysis of the time series for various values of the pull velocities ( for a fixed value of the inertia corresponding to their experimental roller tape geometry ) . in order to compare this result ,
we have calculated the largest lyapunov exponent for a range of values of @xmath17 and @xmath2 .
the region of chaos is in the domain of small pull velocities @xmath2 when @xmath17 is small .
the maximum lyapunov exponent turns out to be rather high , typically around 7.5 bits / s in contrast to the small values reported in ref .
the large magnitude of the positive exponent in our case can be traced to the large changes in the jacobian , as @xmath113 varies over several order of magnitude(@xmath114 ) as a function of the peeling velocity and hence as a function of time .
in contrast , hong _ et al . _ use an @xmath70 shaped curve where @xmath113 is constant ( and small ) on both low and high @xmath2 branches .
however , these large values of lyapunov exponents are consistent with rather high values obtained by gandur _
@xcite from time series analysis of the pull force .
we also find chaos for intermediate and high inertia in the region of high velocities where the value of the lyapunov exponent is small , typically 0.5 .
the small value here again can be traced to the small changes in @xmath113 at high velocities .
it must be mentioned that comparison with experiments is further complicated due to the presence of a two parameter family of solutions strongly dependent on both @xmath17 and @xmath2 .
thus , the phase diagram is complicated , i.e. , the sequence of solutions encountered in the @xmath17-@xmath2 plane as we change @xmath2 or @xmath17 or both does not in general display any specific ordering of periodic and chaotic trajectories ( see fig . 1 of ref .
@xcite ) usually found in the well known routes to chaos .
( for instance @xmath115 periods should be observed before the odd periods @xcite . ) indeed , in our model , we find the odd periods 3,5,7 etc , on increasing @xmath2 ( or @xmath17 ) , without seeing all the @xmath115 periods .
( these odd periods also imply chaos at parameter values prior to that corresponding to these periods . ) in view of this , a correct comparison with experiments requires an appropriate cut in the @xmath116 plane consistent with the experimental values of @xmath17 and @xmath2 even where they are given .
however , as the values of @xmath17 are not provided , full mapping of chaotic solutions is not possible .
( we also note that gandur _
@xcite use a different tape from that used in ref .
@xcite , as is clear from the instability range , leading additional difficulties in comparison . )
one quantitative result that can be compared with experiment is the decreasing trend of the force drop magnitude .
we have calculated the magnitude of the force drops during stick - slip phase as a function the pull velocity @xmath2 for both low ( @xmath86 ) and high ( @xmath88 ) inertia cases .
figure [ fig6 ] shows the monotonically decreasing trend of average @xmath117 as @xmath2 is increased , for both small and large @xmath17 , a feature observed in experiments @xcite .
these two distinct behaviors are a result of the dynamization of @xmath28 as in eq .
( [ fvv ] ) . as a function of the pull speed @xmath2 , for two distinct values of @xmath17 .
the dashed line corresponds to @xmath118 while the dotted line corresponds to @xmath65 .
( @xmath9 , @xmath2 are in m / s , @xmath4 in n , @xmath17 in kg m@xmath60 and @xmath71 in s. ) , width=302,height=188 ] finally , as another illustration of the richness of the dynamics seen in our numerical simulations , we show in fig .
[ fig7 ] , a plot of an orbit that sticks to unstable part of the manifold before jumping back to the @xmath81 branch .
such solutions are known as canards @xcite .
though canard type of solutions are rare , we have observed them for high values of @xmath17 and low values of @xmath2 . in our case ,
such solutions are due to the competition of time scale due to inertia and that due to @xmath9 .
this again illustrates the influence of inertia of the roll on the dynamics of peeling .
it is clear that these equations exhibit rich and complex dynamics .
a few of these features are easily understandable , but others are not .
for instance , the saw - tooth form of @xmath4 for low inertia and low pull velocity can be explained as resulting from the trajectory sticking to stable part of @xmath28 and jumping only when it reaches the limit of stability . for these parameter values , as the time spent by the system is negligible during the jumps between the branches @xmath81 and @xmath82 ( and vice versa ) , the system spends most of the time on the branch @xmath81 and much less on @xmath82 due to its steep nature .
then , from eq .
( [ flow3 ] ) , it is clear that we should find a sawtooth form whenever the peel velocity @xmath9 jumps across the branch to a value of @xmath9 larger than the pull velocity @xmath2 .
however , several features exhibited by these system of equations are much too complicated to understand .
we first list the issues that need to be explained .
+ ( i ) small @xmath17 .
+ ( a ) we find high frequency tiny oscillations superposed on the linearly increasing @xmath4 [ on the @xmath81 branch or better seen in the @xmath12 plot fig . [ fig3](c ) ] .
this needs to be understood .
+ ( b ) the numerical solutions show that the influence of inertia can be important _ even for small _ @xmath17 and small @xmath2 .
for instance , the jumps between @xmath81 and @xmath82 branches occur even before @xmath4 reaches the extremum values of @xmath10 .
\(ii ) for intermediate and high values of inertia , for low @xmath2 case .
+ ( a ) we observe several relatively small amplitude saw tooth form of @xmath4 on the descending part of the pull force @xmath4 .
these appear as a sequence of jumps between the two branches in the @xmath63 plane which we shall refer to as the jumping mode " .
a proper estimate of the magnitude of @xmath94 is desirable .
\(b ) in addition , there appears to be a critical value of @xmath4 for a given cycle below which the return jumps from ab to cd stop and one observes a monotonically increasing trend in @xmath4 [ @xmath101 in fig .
[ fig4](c ) ] .
+ ( iii ) high i and high @xmath2 .
\(a ) the jumps between the branches occur at a very high frequency [ fig .
[ fig5](c ) ] and now are located near the extremum values of @xmath4 and @xmath12 .
but these regions are separated by a stretch where the orbit monotonically increases on the @xmath81 branch and monotonically decreases on the @xmath82 branch .
we need to elucidate the underlying causes leading to the switching between the jumping mode and monotonically increasing or decreasing mode .
\(b ) for large @xmath2 , say @xmath87 and large @xmath17 ( fig . [ fig5 ] ) , the extent of values of @xmath68 range between @xmath119 and @xmath120 much beyond @xmath104 whose range is around 300 .
this feature is less dominant for small @xmath17 and small @xmath2 case .
as the dynamics is described by a coupled set of differential equations with an algebraic constraint , the results are not transparent .
we first attempt to get insight into the complex dynamics through some simple approximations valid in each of the regimes of the parameters .
solution of these approximate equations will require appropriate initial values for the relevant variables which will be provided from the exact numerical solutions . due to the nature of approximations ,
the results are expected to capture only the trend and order of magnitudes of the effects that are being calculated .
but as we will show , even the numbers obtained match quite closely with the exact numerical results .
our idea is to capture the dynamics through a single equation ( as far as possible or at most two as in the high @xmath17 and @xmath2 case ) by including all the relevant time scales and solve the relevant equation _ on each branch .
_ for this we note that the equation for @xmath12 and @xmath9 play a crucial role as the inertial contribution appears only through eqs .
( [ flow1 ] ) and ( [ flow2 ] ) and the time spent by the system is controlled by the equation for @xmath9 , eq . ( [ vdotapprox ] ) .
using eqs .
( [ flow1 ] ) and ( [ flow2 ] ) , we get @xmath121 the general equation for @xmath12 can be written down by using eq .
( [ vdotapprox ] ) , in eq .
( [ genalpha ] ) , we get @xmath122 } { rf^{\prime } } , \label{genalpha1}\\ & \simeq & -\frac{fr\alpha } { i } -\frac { [ \dot f + f \dot\alpha]}{rf^{\prime}}. \label{airy1 } % & \simeq & -\frac{f\alpha } { i } -\frac{f \dot\alpha}{rf^{\prime}}.\end{aligned}\ ] ] in obtaining eq .
( [ airy1 ] ) , we have used @xmath123 which is valid except for high @xmath17 and high @xmath2 .
further , in most cases , we can drop @xmath124 as the magnitude of this term is small and use @xmath125 . to be consistent we use @xmath126 .
we note however that even for high @xmath17 and high @xmath2 where @xmath12 is not small , dropping @xmath127 and @xmath128 causes only 10% error .
+ * case i , small @xmath17 * on the low velocity branch @xmath81 , as @xmath129 is small in eq .
( [ flow1 ] ) , we can drop @xmath130 term in eq .
[ genalpha ] .
thus , @xmath131 note that for the low inertia case , @xmath132 approximation is clearly justified [ see eq .
( [ flow2 ] ) ] . using this equation
, we first get an idea of the relevant time scales as @xmath17 is increased . + *
case a * consider the low velocity branch @xmath81 where the small amplitude high frequency oscillations are seen on the nearly linearly increasing part of @xmath4 [ given by @xmath133 , see for instance fig .
[ fig3](b ) ] .
a rough estimate of this time spent on this branch is obtained by @xmath134 . using @xmath135 and @xmath136 , [ from fig . [ fig3](b ) ] , we get @xmath137 ( compared to the correct value of @xmath138 which we shall obtain soon ) which is much larger than the period of the high frequency oscillation
. thus , we could take the local value @xmath4 for the purpose of calculating the period of the high frequency oscillation .
consider the orbit at the lowest value of @xmath4 for which we can use @xmath139 .
then using eq .
( [ approxalpha ] ) , the frequency @xmath140 for @xmath84 which gives the period of oscillation @xmath141 .
this agrees very well with the exact numerical value @xmath142 .
this frequency decreases when the force reaches the maximum value @xmath143 to @xmath144 giving @xmath145 which is again surprisingly close to the numerical value @xmath146 . in the numerical solutions
, we find that the period gradually decreases [ see fig .
[ fig3](c ) ] .
this feature is also easily recovered by using @xmath147 .
this leads to an additional term in the equation of motion for @xmath12 in eq .
( [ approxalpha ] ) , @xmath148 r\alpha / i , \label{airy}\ ] ] where @xmath71 is the time required for @xmath4 to reach @xmath61 starting from @xmath149 . here
again the @xmath9 term can be dropped .
if @xmath149 was absent , the equation has the airy s form .
( note that for this case also we could assume @xmath150 and @xmath151 . )
though this equation does not have an exact solution , we note that we could take @xmath12 to have a sinusoidal form with @xmath152 where @xmath4 is treated as a slowly increasing parameter .
( this assumption works quite well . )
the above equation captures the essential features of the numerical solution .
the numerical solution of eq .
( [ airy ] ) ( as also this representation ) gives the decreasing trend of the small amplitude high frequency oscillations .
( note that the airy equation itself gives a decreasing amplitude @xcite . )
we note that eq .
( [ approxalpha ] ) is valid on the @xmath81 branch where @xmath9 is small even for high inertia and small @xmath2 case .
thus , we may be able to recover the gross time scales using this equation .
our numerical results show that as we increase the inertia , @xmath12 exhibits a sinusoidal form on the @xmath81 branch [ see fig . [ fig4](b ) ] , although one full cycle is not seen .
we note that though the value of @xmath12 is much larger than that for small @xmath17 , we can still use the above equations [ eq .
( [ approxalpha ] ) and ( [ airy ] ) ] . on this branch
@xmath4 increases from a value @xmath153 to a maximum @xmath151 . for large @xmath118
( and @xmath67 ) , we get a rough estimate of the period by using the mean value of @xmath154 in eq .
[ approxalpha ] .
this gives a period @xmath155 which already agrees satisfactorily with the numerically exact value @xmath156 = 0.11 considering the approximation used ( i.e. , using the mean @xmath4 ) .
a better estimate can be obtained by using eq .
( [ airy ] ) .
for the high @xmath17 and @xmath2 case , fig .
5 for @xmath87 shows that the wave forms are nearly sinusoidal except for a jitter at the top and bottom . for this case , @xmath104 is nearly flat over the entire range of values of @xmath9 , with a value @xmath157 300 . here , even on the ab branch , we can not ignore the @xmath130 term in eq .
( [ genalpha ] ) .
however , one sees that as @xmath158 and @xmath159 which suggest that to the leading order , we could ignore the @xmath130 term .
this gives the period @xmath160 . from fig .
[ fig5](b ) , considering only the monotonically decreasing part ( @xmath101 ) , the value of @xmath161 read off from the figure compares reasonably well with this value . for the @xmath82 branch ,
as @xmath9 is not small , the @xmath162 term appears to be important in eq .
( [ genalpha ] ) .
some idea of when this term is important can be had by looking at the time scales arising from inertia , namely , @xmath163 and the coefficient of the damping term , @xmath164 in eq .
( [ airy1 ] ) .
consider @xmath67 for @xmath165 and @xmath166 .
the period obtained by assuming the mean value of @xmath167 in @xmath163 gives @xmath168 for @xmath86 compared to @xmath169 for @xmath106 .
these numbers can be compared with the time scale @xmath170 which is 0.01 ( where we have used @xmath171 from numerical simulations for @xmath83 ) .
this shows that for high inertia the damping coefficient @xmath164 in eq .
( [ airy1 ] ) is important .
we will discuss this issue in more detail later .
+ * case b * now we focus on the origin of jumps between the branches .
we note that the jumps from @xmath82 to @xmath81 ( or vice versa ) occur only when the peel velocity @xmath9 reaches a value where @xmath172 .
this also means that the time scale on each branch , whether it spends only a short time or not , is controlled by the equation for @xmath9 .
however , clearly the influence of inertia needs to be included . here
we present an approximate equation for @xmath9 which is valid in the various limits of the parameters : @xmath173/f^{\prime } , \label{fullvdot}\\ & \simeq & \frac { [ k ( v - v ) + ( f_{in } + k ( v - v)t ) \dot \alpha)]}{f^{\prime } } , \label{approxvdot}\end{aligned}\ ] ] where the time @xmath71 is time spent on the branch considered ( low or high @xmath2 ) . in eq .
( [ approxvdot ] ) , we have again used @xmath174 and @xmath175 with the same approximation used in eq .
( [ airy1 ] ) .
we now attempt to obtain correct estimates of the time spent by the orbit on each branch starting with the least complicated situation of the low inertia and small @xmath2 . for this case , on the low velocity branch , one can use the sinusoidal solution for @xmath12 , namely @xmath176 , where @xmath49 is a phase factor which also includes the contribution arising from the jump as well and @xmath177 with @xmath178 .
both @xmath179 and @xmath49 needs to be supplied .
alternately , one can use eq .
( [ approxalpha ] ) with eq .
( [ approxvdot ] ) for which we provide @xmath179 and @xmath180 at the point from the exact numerical solutions .
we stress that this procedure is _ not _ equivalent to solving all the equations , as the only equation we use is eq .
( [ approxvdot ] ) with the form of @xmath12 already determined from the equation for @xmath12 .
[ we note here that though we have used the sinusoidal form of @xmath12 along with the initial conditions on @xmath181 , it is simpler to supply the initial conditions @xmath182 and use eq .
( [ approxalpha ] ) . ]
we note here that @xmath183 is a crucial factor that determines the time at which the orbit jumps from one branch to the other .
equation ( [ approxvdot ] ) needs to be integrated from @xmath184 to @xmath185 that are determined by the pulling velocity @xmath2 , i.e. , the form of @xmath28 . for the low @xmath9 branch @xmath183 term makes a significant contribution for the time spent by the trajectory on @xmath81 .
indeed , one can obtain the order of magnitude of the time spent by the orbit on ab by using a crude approximation for @xmath186 .
this can be easily integrated from @xmath187 , to @xmath188 which already gives @xmath189 .
this number is comparable to the numerically exact value @xmath138 .
a correct estimate can be obtained by using @xmath183 from eq .
( [ fvv ] ) with the sinusoidal form @xmath12 or eq .
( [ approxalpha ] ) .
( we have used @xmath190 from the numerical simulations for @xmath67 and @xmath191}$ ] with @xmath192 . )
this gives nearly the exact numerical value of @xmath193 .
in fact , this solution also captures the oscillatory growth nature of @xmath9 quite accurately .
the approximate form of @xmath89 ( continuous line ) along with the numerically exact solution ( dotted line ) are shown in fig .
[ fig8 ] . using @xmath194 in @xmath195 gives @xmath196 and @xmath197 which is in good agreement with the exact numerical value of @xmath198 .
it is interesting to note that this value is much less than @xmath199 [ see fig . [ fig3](b ) ] or equivalently @xmath94 is less than @xmath200 , what is also observed in our exact numerical simulation .
the underlying mechanism of jumping of the orbit before @xmath4 reaches @xmath92 also becomes clear from the analysis ( fig .
[ fig8 ] ) .
we note that the magnitude of the oscillatory component in @xmath9 grows till it reaches @xmath201 permitted by @xmath202 .
then , the orbit has to jump to @xmath82 .
thus , the approximate solution gives an insight into the cause of the orbit jumping even before @xmath4 reaches @xmath92 ( for small i ) . for the @xmath82 branch also , the dominant term is @xmath183 .
indeed , any reasonable function which has the same geometrical form of @xmath10 shown in fig . [ fig2 ] will give good results for @xmath203 .
using the correct form of @xmath183 , we get @xmath204 which is close to the exact result .
this again gives correct magnitude of @xmath205 .
in addition the nature of the @xmath89 obtained by this approximation is close to the exact numerical solution shown in the inset of fig .
8 . for @xmath65 and @xmath67 for the ab branch .
the inset shows a similar comparison of @xmath89 on the cd branch .
( @xmath9 , @xmath2 are in m / s , @xmath17 in kg m@xmath60 and @xmath71 in s.),width=302,height=226 ] * case ii , intermediate and high @xmath17 and low @xmath2 * the most difficult feature of our numerical solutions to understand is the dynamical mechanism leading to a series of drops in the pull force seen on the descending branch of @xmath68 for intermediate and high values of inertia and for a range of @xmath2 values .
consider the high inertia and low @xmath2 case ( say @xmath206 and @xmath67 ) shown in fig .
as stated earlier , there are two different issues that need to be understood here .
first , the series of small force drops @xmath94 and second the monotonic increasing nature of @xmath4 on the @xmath81 branch . in this case , as already discussed , the coefficient of @xmath207 , namely , @xmath164 term in eq .
( [ airy1 ] ) determines the time scale on @xmath82 , while on @xmath81 , the term @xmath208 dominates .
thus , the general equation valid for this case is @xmath209 where we use @xmath210 .
[ note that we have dropped @xmath211 term from eq .
( [ airy1 ] ) as this term does not have any dependence on @xmath12 or @xmath212 . ]
we start with the cascading effect .
consider the orbit when it is at the highest value of @xmath213 on the @xmath82 branch for which we can drop @xmath208 term . as @xmath183 is a function of @xmath9 , and @xmath4
also depends on time , it appears that we need to use coupled equations @xmath214 with eq .
( [ approxvdot ] ) .
however , the numerical solution of these equations show that one can make further approximation by taking @xmath183 to be constant taken at @xmath215 and @xmath216 , as the time spent on this branch is very small .
the error in using this approximation is within @xmath217 . indeed , using @xmath218 and numerically integrating eq .
( [ approxvdot ] ) , along with eq .
( [ airy2 ] ) from @xmath219 to @xmath220 gives @xmath221 .
this compares reasonably with the numerical value of @xmath222 . using this
we get @xmath223 which compares well with the numerically exact value 19.8 . at this point
the orbit jumps to the low velocity @xmath81 branch ( to the point @xmath224 ) .
thus , as @xmath203 is small , for all practical purposes , we can ignore the dependence of @xmath10 on @xmath9 and @xmath4 on @xmath71 and use @xmath12 to be an exponentially decreasing function for analytical estimates .
these analytical estimates already give reasonably accurate numbers . on the @xmath81 branch ,
the dominant time scale is determined by @xmath208 , and we can use the approximate sinusoidal form in eq .
( [ approxvdot ] ) , or eq . ( [ approxalpha ] ) along with eq .
( [ approxvdot ] ) for the time evolution from the point @xmath224 . integrating from @xmath225 to @xmath226 with the appropriate initial values @xmath227 ( or @xmath228 ) and @xmath229
, gives @xmath230 which again compares very well with exact numerical value @xmath231 .
this gives @xmath232 .
the procedure for calculating the time spent by the orbit on @xmath82 and @xmath81 is the same and we find that successive values of @xmath94 increases which is again consistent with what is seen in figs .
[ fig4](a ) and [ fig4](c ) .
continuing this procedure , we find that a minimum value of @xmath233 for the cycle is reached .
now consider the time evolution of @xmath4 on @xmath81 that should lead to a monotonically increasing nature as seen in the numerically exact solution .
as this point corresponds to the point at which the dynamics switches from the jumping mode to the monotonically increasing nature of @xmath4 ( i.e. , the stretch @xmath101 ) , we discuss this in some detail .
for the point @xmath99 , we have used the initial condition @xmath234 and integrating eq .
( [ approxvdot ] ) and eq .
( [ approxalpha ] ) ( or the sinusoidal form of @xmath12 ) from @xmath235 to @xmath236 gives @xmath237 .
this is nearly the value 0.114 obtained from the exact numerical integration .
this gives @xmath238 and @xmath239 which compares very well with the exact numerical value .
in addition , the growth form of @xmath9 obtained from this approximation ( continuous line ) agrees very well with that of the exact numerical solution ( dotted line ) as shown in fig .
the discrepancy seen in the figure can be reduced for instance if we include the terms neglected in eq .
( [ genalpha1 ] ) such as @xmath240 and using @xmath241 in eq .
( [ fullvdot ] ) .
now we come to the crucial question
. how does the system know that it has to go from @xmath99 to @xmath100 , while just during the previous visit to the point @xmath14 on @xmath81 branch lead only to a small increase in @xmath94 [ fig . [
fig4](a ) and [ fig4](c ) ] before jumping to @xmath82 ? to understand this , we recall that on @xmath81 , a sinusoidal solution is allowed .
first , one can notice a few differences in the initial conditions between the point @xmath99 and @xmath14 .
for the point @xmath14 , the initial conditions taken from the exact numerical solution are @xmath242 and @xmath243 .
( @xmath244 ) , while for the point @xmath99 , @xmath245 .
however , for @xmath12 to begin a sinusoidal form , the initial value of @xmath246 is much higher than the natural slope .
the local slope for any sinusoidal form is maximum when the variable is close to zero . in fig .
[ fig4](b ) , the sinusoidal form starts when @xmath12 is close to zero ( @xmath247 at @xmath99 ) where the local slope should be close to the maximum value . near @xmath248 , the local slope is the product of the maximum amplitude of @xmath12 , say , @xmath249 ( in the sinusoidal stretch @xmath101 ) and @xmath250 .
( we have assumed @xmath251 by dropping the phase factor . )
thus , one should have @xmath252 when @xmath248 . using the value @xmath253 from exact numerical solution and @xmath254 at @xmath255
, we find that @xmath256 near @xmath248 . indeed
, this is satisfied only at @xmath99 where @xmath257 .
[ note that @xmath12 is not symmetric around zero due to the presence of @xmath9 in eq .
( [ flow1 ] ) which has been ignored for the purpose of present discussion .
] however , @xmath258 at @xmath14 is significantly higher than the slope permitted for @xmath12 to start a sinusoidal sojourn .
this forces the orbit to make one more small loop ( @xmath81 to @xmath82 and back ) so that the initial value of @xmath180 is commensurate for @xmath12 to start a sinusoidal form .
indeed , the initial values of @xmath212 at all the earlier visits to @xmath81 branch keep decreasing until it reaches a value that is consistent to begin the sinusoidal growth . once this is satisfied ,
the monotonic increasing behavior from @xmath99 to @xmath100 is seen . as we will show this is the mechanism operating for high @xmath17 and @xmath2 case .
+ * case iii , high @xmath17 and @xmath2 * for this case , even on the @xmath81 branch , @xmath162 can not be ignored in eq .
( [ genalpha ] ) and thus one needs to use coupled eqs .
( [ approxvdot ] ) and ( [ genalpha ] ) .
calculations follow much the same lines and give correct values for @xmath194 and @xmath94 on both the branches during the rapid jumps . again
, we need to answer when exactly does the system know to switch from a rapid jumping mode to monotonically increasing on @xmath81 or decreasing mode on @xmath82 ? consider the last of the rapid jumps from @xmath82 to @xmath81 ( just prior to the point @xmath99 ) in fig . [
fig5](c ) .
the corresponding point in the @xmath12 plot [ fig .
[ fig5](b ) ] is shown on an expanded scale in the inset . from this figure , it is clear that @xmath207 has a positive slope at @xmath14 , though of small magnitude while at @xmath99 , it has a value -9.7 .
the latter is close to the natural ( negative ) slope of @xmath12 when it begins the descending branch of the sinusoidal form . on the other hand ,
the slope of @xmath12 is positive at @xmath14 and hence will not allow the growth to change over from a jumping mode to the sinusoidal growth form for @xmath12 .
one can note that the slopes at points of all the earlier visits to ab [ see fig .
[ fig5](b ) inset ] keep decreasing till the slope becomes negative required for the monotonically decreasing trend of @xmath12 .
this is exactly the same mechanism for @xmath106 and @xmath67 also , for the low @xmath9 branch , except that in this case , even the sign of the slope is incompatible for all the points prior to @xmath99 in fig .
[ fig5 ] c. the mechanism operating on @xmath82 ( i.e. , at the switching from jumping mode to monotonically decreasing nature of @xmath4 ) is essentially the same but arguments are a little more involved and hence they are not presented . with numerically exact solution ( dotted line ) for @xmath259 .
( @xmath9 , @xmath2 are in m / s , @xmath17 in kg m@xmath60 and @xmath71 in s. ) , width=302,height=188 ] now , we consider the causes leading to the maximum and minimum values taken by @xmath4 being much more than permitted by @xmath28 . as
this is dominant for @xmath87 , we illustrate this using fig . [ fig5](a ) and [ fig5](c ) .
we first note that eq .
( [ constr ] ) constrains the dynamically changing values of @xmath68 and @xmath66 to the stationary values of @xmath28 .
clearly , this implies that @xmath260 . a rough estimate of @xmath61 can be obtained by @xmath261 with @xmath149 determined by @xmath262 .
this relation can be easily verified by using the numerical values of @xmath12 .
for instance , for @xmath87 and @xmath106 , @xmath263 and @xmath264 .
this gives @xmath265 while the numerical value from the phase plot for this case gives @xmath266 which is very close .
similarly , using @xmath267 and @xmath268 , we get @xmath269 which compares well with the numerical value of 180 .
we have verified this relation is respected for various values of @xmath2 and @xmath17 . for small @xmath17 , @xmath12 is small , we should not find much difference between @xmath61 ( @xmath149 ) and that of @xmath10 .
we first summarize the results before making some relevant remarks .
we have carried out a study of the dynamics of an adhesive roller tape using a differential - algebraic scheme used for singular set of differential equations .
the algorithm produces stick - slip jumps across the two dissipative branches as a consequence of the inherent dynamics .
our extensive simulations show that the dynamics is much richer than anticipated earlier .
in particular the influence of inertia is shown to be dramatic .
for instance , even at low inertia , for small values of @xmath2 , the influence of inertia manifests with jumps of the orbit occurring even before @xmath4 reaches @xmath92 ( or @xmath93 ) which is quite unexpected .
more dominant is its influence for high @xmath17 both for low @xmath2 and high @xmath2 , though it is striking for the latter case . following the reasoning used in the plc effect
, we introduce a dynamized curve @xmath28 as resulting from competing time scales of internal relaxation and imposed pull speed .
the modified peel force function leads to the decreasing trend in the magnitude of @xmath270 with increasing pull velocity , a feature observed in experiments .
we have also recaptured the essential features of the dynamics by a set of approximations valid in different regimes of the parameter space .
these approximate solutions illustrate the influence of various time scales such as that due to inertia , the elasticity of the tape and that determined by the stationary peel force @xmath28 .
we also find the unusual canard type of solutions . here , it is worthwhile to comment on the dynamical features of the model .
the numerical results themselves are too complex to understand .
a striking example of this is the series of force drops seen on the descending branch of the pull force [ fig .
[ fig4](c ) ] .
this result is hard to understand as it would amount to a partial relaxation of the pull force .
however , a partial relaxation is only possible in the presence of another competing time scale ( other than the imposed time scale ) .
another example is the jumping of the orbit for low @xmath17 case , from @xmath81 to @xmath271 and vice versa even before the pull force reaches the extremum values of @xmath28 .
for this reason , we have undertaken to make this complex dynamics transparent using a set of approximations .
the basic idea here is to solve a single equation ( or at most two equations as in the high @xmath17 and @xmath2 case ) which incorporates all the relevant time scales .
this method not only captures all the results to within 10% error but it also clearly brings out the regimes of parameter space where these time scales become important .
this analysis also shows that the time scale due to inertia of the roller tape shows up even for low @xmath17 which comes as a surprise as one expects that for low inertia , the orbit should stick to the stationary peel function .
( recall that for low inertia , equations have been approximated by lienard type of equations by maugis and barquins @xcite . )
our approximate equations demonstrate that a crucial role in inducing the jumps _ even at low inertia _ is played by the high frequency oscillations resulting from the inertia of the roller tape . as for high inertia ( both for low and high pull velocities ) , the time scale due to inertia is responsible for the partial relaxation of @xmath4 as shown
. a few comments may be in order on the bursting type of oscillations in the peel velocity .
bursting type of oscillatory behavior are commonly seen in neuro - biological systems @xcite .
conventionally , bursting type oscillations arise in the presence of homoclinic orbit @xcite .
such bursting type of oscillations have also been modeled using one dimensional map @xcite .
however , it is clear that the mechanism for bursting type of oscillations in our case is different . in our case
, this arises due to the fact that the orbit is forced to jump between the stable manifolds as a result of competing time scale of inertia and the time scale for the evolution of @xmath9 .
( we note that the latter itself includes more than one time scale [ see eq . ( [ approxvdot ] ) ] , namely the contribution from the slopes of the stable parts of the stationary curve @xmath28 and that due to elasticity of the tape . )
the bunching of the spikes in @xmath9 is the result of @xmath28 becoming flat for large @xmath2 and @xmath17 .
one other comment relates to canard type solutions .
figure [ fig7 ] shows one such solution .
as mentioned , these type solutions arise from sticking to the unstable manifold .
in fact , a similar type of solution is seen in fig . [
fig5](c ) .
as noted earlier , all the jumps from @xmath82 to @xmath81 or vice versa always occur when the peel velocity reaches the limiting value where @xmath272 .
however , it can be seen from this figure , the orbit starting from @xmath107 monotonically decreases well into the unstable part of @xmath10 .
thus , this solution also has the features of canards .
it must be stated that our approximate solutions can not capture the behavior of canards . finally , the results presented in this paper are on the nature of dynamics of the model equations which so far had defied solution . however , comparison with experiments has been minimal largely due to the paucity of quantitative experimental findings as stated earlier .
our analysis shows that the model predicts periodic , saw - tooth @xcite , as well as chaotic solutions as reported in @xcite .
the high magnitude of the lyapunov exponents for the chaotic solutions in the low pull velocities is consistent with that reported earlier @xcite .
we note that the other quantitative experimental feature reported by refs.@xcite is the decreasing trend of the average force drop magnitudes as a function of the pull velocity is also captured by our model ( fig .
[ fig6 ] ) , a result that holds for both low and high inertia .
this result is a direct consequence of the dynamization of the peel force function , i.e. , dependence of the peel force on the pull velocity .
we note here that the complex dynamics at high velocities ( see fig .
[ fig5 ] ) is a direct result of the unstable part of dynamized curve , @xmath28 , shrinking to zero . to the best of our knowledge ,
this is first time the result in fig .
[ fig6 ] has been explained .
as the hypothesis of dynamization captures the decreasing trend of the force drops , it also suggests that the underlying mechanism of competing time scales responsible for the peel force depending on the pull velocity is likely to be correct as in the plc effect . clearly , a rigorous derivation of the peel force function from microscopic considerations that includes the effect of the viscoelastic glue at the contact point is needed to understand the dynamics appropriately .
the authors wish to thank a. s. vasudeva murthy of tifr , bangalore for useful discussions on dae algorithm .
rd and am wish to thank m. bekele of addis ababa university , ethiopia and m. s. bharathi of brown univ . ,
usa for stimulating and friendly discussions .
this work is financially supported by the department of science and technology , new delhi , india under the grant sp / s2k-26/98 . | we investigate the dynamics of peeling of an adhesive tape subjected to a constant pull speed .
we derive the equations of motion for the angular speed of the roller tape , the peel angle and the pull force used in earlier investigations using a lagrangian . due to the constraint between the pull force , peel angle and the peel force
, it falls into the category of differential - algebraic equations requiring an appropriate algorithm for its numerical solution .
using such a scheme , we show that stick - slip jumps emerge in a purely dynamical manner .
our detailed numerical study shows that these set of equations exhibit rich dynamics hitherto not reported .
in particular , our analysis shows that inertia has considerable influence on the nature of the dynamics .
following studies in the portevin - le chatelier effect , we suggest a phenomenological peel force function which includes the influence of the pull speed . this reproduces the decreasing nature of the rupture force with the pull speed observed in experiments .
this rich dynamics is made transparent by using a set of approximations valid in different regimes of the parameter space .
the approximate solutions capture major features of the exact numerical solutions and also produce reasonably accurate values for the various quantities of interest . |
physical layer security has been a very active area of research in information theory .
see @xcite and @xcite for overviews of recent progress in this field .
a basic model of physical layer security is a wiretap / broadcast channel @xcite with two receivers , a legitimate receiver and an eavesdropper .
both the legitimate receiver and the eavesdropper channels are assumed to be _ known _ at the transmitter . by exploring the ( statistical ) difference between the legitimate receiver channel and the eavesdropper channel , one may design coding schemes that can deliver a message reliably to the legitimate receiver while keeping it asymptotically perfectly secret from the eavesdropper . while assuming the transmitter s knowledge of the legitimate receiver channel might be reasonable ( particularly when a feedback link is available ) , assuming that the transmitter knows the eavesdropper channel is _ unrealistic _ in most scenarios .
this is mainly because the eavesdropper is an _
adversary _ , who usually has no incentive to help the transmitter to acquire its channel state information .
hence , it is critical that physical layer security techniques are designed to withstand the _ uncertainty _ of the eavesdropper channel . in this paper
, we consider a communication scenario where there are _ multiple _ possible realizations for the eavesdropper channel . which realization will actually occur is _ unknown _ to the transmitter .
our goal is to design coding schemes such that the number of _ secure _ bits delivered to the legitimate receiver depends on the _ actual _ realization of the eavesdropper channel .
more specifically , when the eavesdropper channel realization is weak , _ all _ bits delivered to the legitimate receiver need to be secure .
in addition , when the eavesdropper channel realization is strong , a prescribed _ part _ of the bits needs to _ remain _ secure .
we call such codes _ security embedding codes _ , referring to the fact that high - security bits are now embedded into the low - security ones .
we envision that such codes are naturally useful for the secrecy communication scenarios where the information bits are _ not _ created equal : some of them have more security priorities than the others and hence require stronger security protections during communication .
for example , in real wireless communication systems , control plane signals have higher secrecy requirement than data plane transmissions , and signals that carry users identities and cryptographic keys require stronger security protections than the other signals .
a key question that we consider is at what expense one may allow part of the bits to enjoy stronger security protections .
note that a naive " security embedding scheme is to design two separate secrecy codes to provide two different levels of security protections , and apply them to two separate parts of the information bits . in this scheme
, the high - security bits are protected using a stronger secrecy code and hence are communicated at a lower rate .
the overall communication rate is a _ convex _ combination of the low - security bit rate and the high - security bit rate and hence is lower than the low - security bit rate .
moreover , this rate loss becomes larger as the portion of the high - security bits becomes larger and the additional security requirement ( for the high - security bits ) becomes higher .
the main result of this paper is to show that it is possible to have a significant portion of the information bits enjoying additional security protections _ without _ sacrificing the overall communication rate .
this further justifies the name security embedding , " as having part of the information bits enjoying additional security protections is now only an added bonus . more specifically ,
in this paper , we call a secrecy communication scenario _ embeddable _ if a _
fraction of the information bits can enjoy additional security protections without sacrificing the overall communication rate , and we call it
_ perfectly embeddable _ if the high - security bits can be communicated at _ full _ rate ( as if the low - security bits do not exist ) without sacrificing the overall communication rate .
key to achieve optimal security embedding is to _ jointly _ encode the low - security and high - security bits ( as opposed to separate encoding as in the naive scheme ) .
in particular , the low - security bits can be used as ( part of ) the _ transmitter randomness _ to protect the high - security bits ( when the eavesdropper channel realization is strong ) ; this is a key feature of our proposed security embedding codes .
the rest of the paper is organized as follows . in sec .
[ sec : wtc ] , we briefly review some basic results on the secrecy capacity and optimal encoding scheme for several classical wiretap channel settings .
these results provide performance and structural benchmarks for the proposed security embedding codes . in sec .
[ sec : mswtc ] , an information - theoretic formulation of the security embedding problem is presented , which we term as _ two - level security wiretap channel_. a coding scheme that combines rate splitting , superposition coding , nested binning and channel prefixing is proposed and is shown to achieve the secrecy capacity region of the channel in several scenarios .
based on the results of sec .
[ sec : mswtc ] , in sec .
[ sec : gmswtc ] we study the engineering communication models with real channel input and additive white gaussian noise , and show that both scalar and independent parallel gaussian ( under an individual per - subchannel average power constraint ) two - level security wiretap channels are _ perfectly embeddable_. in sec .
[ sec : mswtc2 ] , we extend the results of sec . [
sec : mswtc ] to the _ wiretap channel ii _ setting of ozarow and wyner @xcite , and show that two - level security wiretap channels ii are also _ pefectly embeddable_. finally , in sec .
[ sec : con ] , we conclude the paper with some remarks .
consider a discrete memoryless wiretap channel with transition probability @xmath0 , where @xmath1 is the channel input , and @xmath2 and @xmath3 are the channel outputs at the legitimate receiver and the eavesdropper , respectively ( see fig .
[ fig : wtc ] ) .
the transmitter has a message @xmath4 , uniformly drawn from @xmath5 where @xmath6 is the block length and @xmath7 is the rate of communication .
the message @xmath4 is intended for the legitimate receiver , but needs to be kept asymptotically perfectly secret from the eavesdropper .
mathematically , this secrecy constraint can be written as @xmath8 in the limit as @xmath9 , where @xmath10,\ldots , z[n])$ ] is the collection of the channel outputs at the eavesdropper during communication .
a communication rate @xmath7 is said to be _ achievable _ if there exists a sequence of codes of rate @xmath7 such that the message @xmath4 can be reliably delivered to the legitimate receiver while satisfying the asymptotic perfect secrecy constraint .
the largest achievable rate is termed as the _ secrecy capacity _ of the channel .
a discrete memoryless wiretap channel @xmath0 is said to be _ degraded _ if @xmath11 forms a markov chain in that order .
the secrecy capacity @xmath12 of a degraded wiretap channel was characterized by wyner @xcite and can be written as @xmath13 \label{eq : cs - wyn}\ ] ] where the maximization is over all possible input distributions @xmath14 .
the scheme proposed in @xcite to achieve the secrecy capacity is _ random binning _ , which can be described as follows .
consider a codebook of @xmath15 codewords , each of length @xmath6 .
the codewords are partitioned into @xmath16 bins , each containing @xmath17 codewords . given a message @xmath18 ( which is uniformly drawn from @xmath5 ) , the encoder _
randomly _ and uniformly chooses a codeword @xmath19 in the @xmath18th bin and sends it through the channel .
the legitimate receiver needs to decode the entire codebook ( and hence recover the transmitted message @xmath18 ) , so the overall rate @xmath20 can not be too high . on the other hand , the rate @xmath21 of the sub - codebooks in each bin represents the amount of external randomness injected by the transmitter ( transmitter randomness ) into the channel and hence needs to be sufficiently large to confuse the eavesdropper . with an appropriate choice of the codebooks and the partitions of bins ,
it was shown in @xcite that any communication rate @xmath7 less than the secrecy capacity is achievable by the aforementioned random binning scheme .
for a _ general _ discrete memoryless wiretap channel @xmath0 where the channel outputs @xmath2 and @xmath3 are _ not _ necessarily ordered , the random binning scheme of @xcite is _ not _ necessarily optimal . in this case , the secrecy capacity @xmath12 of the channel was characterized by csiszr and krner @xcite and can be written as @xmath22 \label{eq : cs - ck}\ ] ] where @xmath23 is an auxiliary random variable satisfying the markov chain @xmath24 .
the scheme proposed in @xcite is to first _ prefix _
the channel input @xmath1 by @xmath23 and view @xmath23 as the input of the _ induced _ wiretap channel @xmath25 . applying the random binning scheme of @xcite to
the induced wiretap channel @xmath26 proves the achievability of rate @xmath27 for any given joint auxiliary - input distribution @xmath28 . in communication engineering , communication channels
are usually modeled as discrete - time channels with real input and additive white gaussian noise .
consider a ( scalar ) gaussian wiretap channel where the channel outputs at the legitimate receiver and the eavesdropper are given by @xmath29 here , @xmath1 is the channel input which is subject to the average power constraint @xmath30)^2 \leq p \label{eq : apc}\ ] ] @xmath31 and @xmath32 are the channel gains for the legitimate receiver and the eavesdropper channel respectively , and @xmath33 and @xmath34 are additive white gaussian noise with zero means and _ unit _ variances .
the secrecy capacity of the channel was characterized in @xcite and can be written as @xmath35^+ \label{eq : cs - lh}\ ] ] where @xmath36^+:=\max(0,x)$ ] .
note from that @xmath37 if and only if @xmath38 .
that is , for the gaussian wiretap channel , asymptotic perfect secrecy communication is possible if and only if the legitimate receiver has a larger channel gain than the eavesdropper . in this case , we can equivalently write the channel output @xmath3 at the eavesdropper as a degraded version of the channel output @xmath2 at the legitimate receiver , and the random binning scheme of @xcite with _ gaussian _ codebooks and _ full _ transmit power achieves the secrecy capacity of the channel .
a closely related engineering scenario consists of a bank of @xmath39 independent parallel scalar gaussian wiretap channels @xcite . in this scenario ,
the channel outputs at the legitimate receiver and the eavesdropper are given by @xmath40 and @xmath41 where @xmath42 here , @xmath43 is the channel input for the @xmath44th subchannel , @xmath45 and @xmath46 are the channel gains for the legitimate receiver and the eavesdropper channel respectively in the @xmath44th subchannel , and @xmath47 and @xmath48 are additive white gaussian noise with zero means and _ unit _ variances .
furthermore , @xmath49 are independent for @xmath50 so all @xmath39 subchannels are independent of each other .
two different types of power constraints have been considered : the average individual per - subchannel power constraint @xmath51)^2 \leq p_l , \quad l=1,\ldots , l \label{eq : pcons - s}\ ] ] and the average total power constraint @xmath52)^2\right ] \leq p. \label{eq : pcons - t}\ ] ] under the average individual per - subchannel power constraint , the secrecy capacity of the independent parallel gaussian wiretap channel is given by @xcite @xmath53 where @xmath54 is defined as in .
clearly , any communication rate less than the secrecy capacity can be achieved by using @xmath39 separate scalar gaussian wiretap codes , each for one of the @xmath39 subchannels .
the secrecy capacity , @xmath55 , under the average total power constraint is given by @xmath56 where the maximization is over all possible power allocations @xmath57 such that @xmath58 .
a waterfilling - like solution for the optimal power allocation was derived in ( * ? ? ?
* th . 1 )
, which provides an efficient way to numerically calculate the secrecy capacity @xmath55 .
consider a discrete memoryless broadcast channel with three receivers and transition probability @xmath59 .
the receiver that receives the channel output @xmath2 is a legitimate receiver .
the receivers that receive the channel outputs @xmath60 and @xmath61 are two possible realizations of an eavesdropper .
assume that the channel output @xmath61 is _ degraded _ with respect to the channel output @xmath60 , i.e. , @xmath62 forms a markov chain in that order .
therefore , the receiver that receives the channel output @xmath60 represents a stronger realization of the eavesdropper channel than the receiver that receives the channel output @xmath61 .
the transmitter has two independent messages : a high - security message @xmath63 uniformly drawn from @xmath64 and a low - security message @xmath65 uniformly drawn from @xmath66 . here
, @xmath6 is the block length , and @xmath67 and @xmath68 are the corresponding rates of communication . both messages @xmath63 and @xmath65 are intended for the legitimate receiver , and need to be kept asymptotically perfectly secure when the eavesdropper realization is weak , i.e. , @xmath69 in the limit as @xmath9 . in addition , when the eavesdropper realization is strong , the high - security message @xmath63 needs to remain asymptotically perfectly secure , i.e. , @xmath70 in the limit as @xmath9 .
a rate pair @xmath71 is said to be _ achievable _ if there is a sequence of codes of rate pair @xmath71 such that both messages @xmath63 and @xmath65 can be reliably delivered to the legitimate receiver while satisfying the asymptotic perfect secrecy constraints and .
the collection of all possible achievable rate pairs is termed as the _ secrecy capacity region _ of the channel .
[ fig : mswtc ] illustrates this communication scenario , which we term as _ two - level security wiretap channel_. the above setting of two - level security wiretap channel is closely related to the traditional wiretap channel setting of @xcite .
more specifically , without the additional secrecy constraint on the high - security message @xmath63 , we can simply view the messages @xmath63 and @xmath65 as a single ( low - security ) message @xmath4 with rate @xmath72 . and
the problem reduces to communicating the message @xmath4 over the traditional wiretap channel with transition probability @xmath73 . by the secrecy capacity expression ,
the maximum achievable @xmath72 is given by @xmath74 \label{eq : cs_ck2}\ ] ] where @xmath23 is an auxiliary random variable satisfying the markov chain @xmath75 .
similarly , without needing to communicate the low - security message @xmath65
( i.e. , @xmath76 ) , the secrecy constraint reduces to @xmath77 which is implied by the secrecy constraint since @xmath78 due to the markov chain . in this case
, the problem reduces to communicating the high - security message @xmath63 over the traditional wiretap channel with transition probability @xmath79 . again , by the secrecy capacity expression , the maximum achievable @xmath67 is given by @xmath80 \label{eq : cs_ck3}\ ] ] where @xmath81 is an auxiliary random variable satisfying the markov chain @xmath82 .
based on the above connections , we may conclude that a two - level security wiretap channel @xmath59 is _ embeddable _ if there exists a sequence of coding schemes with a rate pair @xmath71 such that @xmath72 is equal to and @xmath83 , and it is _ perfectly embeddable _ if there exists a sequence of coding schemes with a rate pair @xmath71 such that @xmath72 is equal to and @xmath67 is equal to . an important special case of the two - level security wiretap channel problem considered here is when the channel output @xmath61 is a constant signal . in this case , the secrecy constraint becomes _ obsolete _ , and the low - security message @xmath65 becomes a _
regular _ message without any secrecy constraint .
the problem of simultaneously communicating a regular message and a confidential message over a discrete memoryless wiretap channel was first considered in @xcite , where a single - letter characterization of the capacity region was established .
for the _ general _ two - level security wiretap channel problem that we consider here , both high - security message @xmath63 and low - security message @xmath65 are subject to asymptotic perfect secrecy constraints , which makes the problem potentially much more involved .
the following theorem provides two _ sufficient _ conditions for establishing the achievability of a rate pair for a given discrete memoryless two - level security wiretap channel .
[ thm : dm1 ] consider a discrete memoryless two - level security wiretap channel with transition probability @xmath59 that satisfies the markov chain .
a nonnegative pair @xmath71 is an achievable rate pair of the channel if it satisfies @xmath84 for some input distribution @xmath14 .
more generally , a nonnegative pair @xmath71 is an achievable rate pair of the channel if it satisfies @xmath85 for some joint distribution @xmath86 , where @xmath87 and @xmath23 are auxiliary random variables satisfying the markov chain @xmath88 and such that @xmath89 .
clearly , the sufficient condition can be obtained from by choosing @xmath90 and @xmath87 to be a constant .
hence , is a more general sufficient condition than .
the sufficient condition can be proved by considering a _ nested _ binning scheme that uses the low - security message @xmath65 as part of the transmitter randomness to protect the high - security message @xmath63 ( when the eavesdropper channel realization is strong ) .
the more general sufficient condition can be proved by considering a more complex coding scheme that combines rate splitting , superposition coding , nested binning and channel prefixing . a detailed proof of the theorem is provided in sec .
[ pf : thm1 ] .
the following corollary provides sufficient conditions for establishing that a two - level security wiretap channel is ( perfectly ) embeddable .
the conditions are given in terms of the existence of a joint auxiliary - input random triple and are immediate consequences of theorem [ thm : dm1 ] .
a two - level security wiretap channel @xmath59 is _ embeddable _ if there exists a pair of auxiliary random variables @xmath87 and @xmath23 satisfying the markov chain @xmath88 and such that @xmath89 , @xmath28 is an _
optimal _ solution to the maximization program , and @xmath91 , and it is _ perfectly embeddable _ if there exists a pair of auxiliary random variables @xmath87 and @xmath23 satisfying the markov chain @xmath88 and such that @xmath89 , @xmath28 is an _
optimal _ solution to the maximization program , and @xmath92 is equal to .
if , in addition to the markov chain , we also have the markov chain @xmath93 in that order , the sufficient condition is also necessary , leading to a precise characterization of the secrecy capacity region .
the results are summarized in the following theorem ; a proof of the theorem can be found in appendix [ app:1 ] .
[ thm : dm2 ] consider a discrete memoryless two - level security wiretap channel with transition probability @xmath59 that satisfies the markov chains and .
the secrecy capacity region of the channel is given by the set of all nonnegative pairs that satisfy for some joint distribution @xmath86 , where @xmath87 and @xmath23 are auxiliary random variables satisfying the markov chain @xmath88 .
if , in addition to the markov chains and , we also have the markov chain @xmath94 in that order , the ( weaker ) sufficient condition also becomes necessary , leading to a simpler characterization of the secrecy capacity region ( which does _ not _ involve any auxiliary random variables ) .
the results are summarized in the following theorem ; a proof of the theorem can be found in appendix [ app:2 ] .
[ thm : dm3 ] consider a discrete memoryless two - level security wiretap channel with transition probability @xmath59 that satisfies the markov chains , and .
the secrecy capacity region of the channel is given by the set of all nonnegative pairs that satisfy for some input distribution @xmath14 .
we first prove the weaker sufficient condition by considering a nested binning scheme that uses the low - security message @xmath65 as part of the transmitter randomness to protect the high - security message @xmath63 ( when the eavesdropper channel realization is strong ) .
we shall consider a random - coding argument , which can be described as follows .
fix an input distribution @xmath14 .
_ codebook generation .
_ randomly and independently generate @xmath95 codewords of length @xmath6 according to an @xmath6-product of @xmath14 .
randomly partition the codewords into @xmath96 bins so each bin contains @xmath97 codewords .
further partition each bin into @xmath98 subbins so each subbin contains @xmath17 codewords . label the codewords as @xmath99 where @xmath100 denotes the bin number , @xmath101 denotes the subbin number within each bin , and @xmath44 denotes the codeword number within each subbin . see fig . [
fig : nb ] for an illustration of the codebook structure . _ encoding .
_ to send a message pair @xmath102 , the transmitter _ randomly _
( according to a uniform distribution ) chooses a codeword @xmath103 from the subbin identified by @xmath102 and sends it through the channel . _
decoding at the legitimate receiver .
_ given the channel outputs @xmath104 , the legitimate receiver looks into the codebook @xmath105 and searches for a codeword that is jointly typical @xcite with @xmath104 . in the case
when @xmath106 with high probability the transmitted codeword @xmath103 is the _ only _ one that is jointly typical with @xmath104 ( and hence can be correctly decoded ) . _ security at the eavesdropper .
_ note that each bin corresponds to a message @xmath107 and contains @xmath97 codewords , each randomly and independently generated according to an @xmath6-product of @xmath14 .
for a given message @xmath107 , the transmitted codeword is randomly and uniformly chosen from the corresponding bin ( where the randomness is from both the low - security message @xmath65 and the transmitter s choice of @xmath108 ) . following @xcite , in the case when @xmath109 we have @xmath110 tends to zero in the limit as @xmath111 .
furthermore , each subbin corresponds to a message pair @xmath102 and contains @xmath17 codewords , each randomly and independently generated according to an @xmath6-product of @xmath14 . for a given message pair @xmath102 ,
the transmitted codeword is randomly and uniformly chosen from the corresponding subbin ( where the randomness is from the transmitter s choice of @xmath108 ) . again , following @xcite , in the case when @xmath112 we have @xmath113 tends to zero in the limit as @xmath111 .
eliminating @xmath21 from using fourier - motzkin elimination , we can conclude that any rate pair @xmath71 that satisfies is achievable .
next we prove the more general sufficient condition by considering a coding scheme that combines rate splitting , superposition coding , nested binning and channel prefixing .
we shall once again resort to a random - coding argument , which can be described as follows .
fix a joint auxiliary - input distribution @xmath114 with @xmath89 and @xmath115 .
split the low - security message @xmath65 into two independent submessages @xmath116 and @xmath117 with rates @xmath118 and @xmath119 , respectively . _ codebook generation . _
randomly and independently generate @xmath120 codewords of length @xmath6 according to an @xmath6-product of @xmath121 .
randomly partition the codewords into @xmath122 bins so each bin contains @xmath123 codewords .
label the codewords as @xmath124 where @xmath100 denotes the bin number , and @xmath101 denotes the codeword number within each bin .
we shall refer to the codeword collection @xmath125 as the @xmath87-codebook . for each codeword @xmath124 in the @xmath87-codebook , randomly and independently generate @xmath126 codewords of length @xmath6 according to an @xmath6-product of @xmath127 .
randomly partition the codewords into @xmath96 bins so each bin contains @xmath128 codewords .
further partition each bin into @xmath129 subbins so each subbin contains @xmath17 codewords . label the codewords as @xmath130 where @xmath131 indicates the base codeword @xmath132 from which @xmath130 was generated , @xmath44 denotes the bin number , @xmath133 denotes the subbin number within each bin , and @xmath134 denotes the codeword number within each subbin
. we shall refer to the codeword collection @xmath135 as the @xmath23-subcodebook corresponding to base codeword @xmath124 .
see fig .
[ fig : nb2 ] for an illustration of the codebook structure . _ encoding .
_ to send a message triple @xmath136 , the transmitter _ randomly _
( according a uniform distribution ) chooses a codeword @xmath137 from the @xmath138th bin in the @xmath87-codebook .
once a @xmath137 is chosen , the transmitter looks into the @xmath23-subcodebook corresponding to @xmath137 and _ randomly _ chooses a codeword @xmath139 from the subbin identified by @xmath140 .
once a @xmath139 is chosen , an input sequence @xmath19 is generated according to an @xmath6-product of @xmath141 and is then sent through the channel . _
decoding at the legitimate receiver . _
given the channel outputs @xmath104 , the legitimate receiver looks into the @xmath87-codebook and its @xmath23-codebooks and searches for a pair of codewords @xmath142 that are jointly typical @xcite with @xmath104 .
in the case when @xmath143 with high probability the codeword pair selection @xmath144 is the only one that is jointly typical @xcite with @xmath104 . _
security at the eavesdropper .
_ to analyze the security of the high - security message @xmath63 and the submessage @xmath117 at the eavesdropper , we shall assume ( for now ) that both the submessage @xmath138 and the codeword selection @xmath145 are known at the eavesdropper .
note that such an assumption can only _ strengthen _ our security analysis . given the base codeword @xmath145 , the encoding of @xmath107 and @xmath146 using
the corresponding @xmath23-subcodebook is identical to the nested binning scheme considered previously ( with additional channel prefixing ) .
thus in the case when @xmath147 we have @xmath148 in the limit as @xmath9 .
the equalities in and are due to the fact that @xmath149 and @xmath116 are independent . from we
may conclude that @xmath150 in the limit as @xmath9 . to analyze the security of the submessage @xmath116 ,
note that each bin in the @xmath87-codebook corresponds to a message @xmath138 and contains @xmath120 codewords , each randomly and independently generated according to an @xmath6-product of @xmath121 . for a given submessage @xmath138 ,
the codeword @xmath145 is randomly and uniformly chosen from the corresponding bin ( where the randomness is from the transmitter s choice of @xmath151 ) .
note from that the rate of each @xmath23-subcodebook is greater than @xmath152 . following ( * ? ? ?
* lemma 1 ) , we have @xmath153 in the limit as @xmath9 .
putting together and , we have @xmath154 which tends to zero in the limit as @xmath9 . finally , note that the overall communicate rate @xmath68 of the low - security message @xmath65
is given by @xmath155 eliminating @xmath21 , @xmath118 and @xmath119 from
, , and @xmath156 using fourier - motzkin elimination , simplifying the results using the facts that 1 ) @xmath89 , 2 ) @xmath157 which is due to the markov chain , and 3 ) @xmath158 and @xmath159 which are due to the markov chain @xmath88 , and letting @xmath160 , we may conclude that any rate pair @xmath71 satisfying is achievable .
this completes the proof of theorem [ thm : dm1 ] .
consider a discrete - time two - level security wiretap channel with real input @xmath1 and outputs @xmath2 , @xmath60 and @xmath61 given by @xmath161 where @xmath31 , @xmath162 and @xmath163 are the corresponding channel gains , and @xmath33 , @xmath34 and @xmath164 are additive white gaussian noise with zero means and unit variances .
assume that @xmath165 so the receiver that receives the channel output @xmath60 represents a stronger realization of the eavesdropper channel than the receiver that receives the channel output @xmath61 .
the channel input @xmath1 is subject to the average power constraint .
we term the above communication scenario as _
( scalar ) gaussian two - level security wiretap channel_. the following theorem provides an explicit characterization of the secrecy capacity region .
[ thm : mgwtc ] consider the ( scalar ) gaussian two - level security wiretap channel .
the secrecy capacity region of the channel is given by the collection of all nonnegative pairs @xmath71 that satisfy @xmath166 where @xmath54 is defined as in .
_ proof : _ we first prove the converse part of the theorem . recall from sec .
[ sec : mswtc - mod ] that without transmitting the low - security message @xmath65 ( which can only increase the achievable rate @xmath67 ) , the problem reduces to communicating the high - security message @xmath63 over the traditional wiretap channel @xmath167 . for the gaussian two - level security wiretap channel
, the problem reduces to communicating the high - security message @xmath63 over the gaussian wiretap channel with channel outputs @xmath2 and @xmath60 given by @xmath168 we thus conclude that @xmath169 for any achievable rate @xmath67 .
similarly , ignoring the additional secrecy constraint for the high - security message @xmath63 ( which can only enlarge the achievable rate region @xmath170 ) , we can simply view the messages @xmath63 and @xmath65 as a single message @xmath4 with rate @xmath72 . in this case
, the problem reduces to communicating the message @xmath4 over the traditional wiretap channel @xmath171 . for the gaussian two - level security wiretap channel
, the problem reduces to communicating the message @xmath4 over the gaussian wiretap channel with channel outputs @xmath2 and @xmath61 given by @xmath172 we thus conclude that @xmath173 for any achievable rate pair @xmath71 . to show that any nonnegative pair @xmath71 that satisfies is achievable , let us first consider two simple cases .
first , when @xmath174 , both @xmath175 and @xmath176 are equal to zero ( c.f . definition ) . so does not include any positive rate pairs and hence there is nothing to prove .
next , when @xmath177 , @xmath178 and reduces to @xmath179 since the high - security message @xmath63 does not need to be transmitted , any rate pair in this region can be achieved by using a scalar gaussian wiretap code to encode the low - security message @xmath65 .
this has left us with the only case with @xmath180 .
for the case where @xmath180 , the achievability of any rate pair in follows from that of by choosing @xmath1 to be gaussian with zero mean and variance @xmath181 .
this completes the proof of the theorem . @xmath182
the following corollary follows directly from the achievability of the corner point @xmath183 of .
scalar gaussian two - level security wiretap channels under an average power constraint are perfectly embeddable .
[ fig : cssg ] illustrates the secrecy capacity region for the case where @xmath184 .
also plotted in the figure is the rate region that can be achieved by the naive scheme that uses two gaussian wiretap codes to encode the messages @xmath63 and @xmath65 separately .
note that the corner point is strictly outside the naive " rate region , which illustrates the superiority of nested binning over the separate coding scheme .
consider a discrete - time two - level security wiretap channel which consists of a bank of @xmath39 independent parallel scalar gaussian two - level security wiretap channels . in this model , the channel outputs are given by @xmath40 , @xmath185 and @xmath186 where @xmath187 here
, @xmath43 is the channel input for the @xmath44th subchannel , @xmath45 , @xmath188 and @xmath189 are the corresponding channel gains in the @xmath44th subchannel , and @xmath47 , @xmath48 and @xmath190 are additive white gaussian noise with zero means and unit variances .
we assume that @xmath191 for all @xmath50 , so the receiver that receives the channel output @xmath60 represents a stronger realization of the eavesdropper channel in _ each _ of the @xmath39 subchannels than the receiver that receives the channel output @xmath61 . furthermore , @xmath192 , @xmath50 , are independent so all @xmath39 subchannels are independent of each other .
we term the above communication scenario as _ independent parallel gaussian two - level security wiretap channel_. the following theorem provides an explicit characterization of the secrecy capacity region under an average individual per - subchannel power constraint . [ thm : mpgwtc ] consider the independent parallel gaussian two - level security wiretap channel where the channel input @xmath1 is subject to the average individual per - subchannel power constraint .
the secrecy capacity region of the channel is given by the collection of all nonnegative pairs @xmath71 that satisfy @xmath193 where @xmath54 is defined as in .
_ proof : _ we first prove the converse part of the theorem . following the same argument as that for theorem [ thm : mgwtc ] , we can show that @xmath194 for any achievable secrecy rate pair @xmath71 . by the secrecy capacity expression for the independent parallel gaussian wiretap channel under an average individual per - subchannel power constraint
, we have @xmath195 substituting into proves the converse part of the theorem . to show that any nonnegative pair @xmath71 that satisfies is achievable , let us consider _ independent _ coding over each of the @xmath39 subchannels .
note that each subchannel is a scalar gaussian two - level security wiretap channel with average power constraint @xmath196 and channel gains @xmath197 .
thus , by theorem [ thm : mgwtc ] , any nonnegative pair @xmath198 that satisfies @xmath199 is achievable for the @xmath44th subchannel .
the overall communication rates are given by @xmath200 substituting into proves that any nonnegative pair @xmath71 that satisfies is achievable .
this completes the proof of the theorem .
similar to the scalar case , the following corollary is an immediate consequence of theorem [ thm : mpgwtc ] .
independent parallel gaussian two - level security wiretap channels under an average individual per - subchannel power constraint are perfectly embeddable .
the secrecy capacity region of the channel under an average total power constraint is summarized in the following corollary .
the results follow from the well - known fact that an average total power constraint can be written as the _ union _ of average individual per - subchannel power constraints , where the union is over all possible power allocations among the subchannels .
consider the independent parallel gaussian two - level security wiretap channel where the channel input @xmath1 is subject to the average total power constraint .
the secrecy capacity region of the channel is given by the collection of all nonnegative pair @xmath71 that satisfies @xmath201 for some power allocation @xmath57 such that @xmath58 .
[ fig : cspg ] illustrates the secrecy capacity with @xmath202 subchannels where @xmath203 as we can see , under the average total power constraint , the independent parallel gaussian two - level security wiretap channel is embeddable but _ not _ perfectly embeddable .
the reason is that the optimal power allocation @xmath204 that maximizes @xmath205 is _ suboptimal _ in maximizing @xmath206 . by comparison , under the average individual per - subchannel power constraint , the power allocated to each of the subchannels
is fixed so the channel is always perfectly embeddable .
in sec . [ sec : wtc ] we briefly summarized the known results on a classical secrecy communication setting known as wiretap channel . a closely related classical secrecy communication scenario is _ wiretap channel ii _ , which was first studied by ozarow and wyner @xcite . in the wiretap channel
ii setting , the transmitter sends a binary sequence @xmath207 of length @xmath6 _ noiselessly _ to an legitimate receiver .
the signal @xmath208 received at the eavesdropper is given by @xmath209 where @xmath210 represents an erasure output , and @xmath211 is a subset of @xmath212 of size @xmath213 representing the locations of the transmitted bits that can be accessed by the eavesdropper .
if the subset @xmath211 is _ known _ at the transmitter , a message @xmath4 of @xmath214 bits can be noiselessly communicated to the legitimate receiver through @xmath215 .
since the eavesdropper has no information regarding to @xmath216 , _ perfectly _ secure communication is achieved _ without _ any coding .
it is easy to see that in this scenario , @xmath214 is also the _ maximum _ number of bits that can be reliably and perfectly securely communicated through @xmath6 transmitted bits . an interesting result of @xcite is that for any @xmath115 , a total of @xmath217 bits can be reliably and _ asymptotically perfectly _ securely communicated to the legitimate receiver even when the subset @xmath211 is _ unknown _ ( but with a fixed size @xmath213 ) a priori at the transmitter . here , by asymptotically perfectly securely " we mean @xmath218 in the limit as @xmath9 . unlike the case where the subset @xmath211 is known ,
coding is _ necessary _ when @xmath211 is unknown a priori at the transmitter . in particular
, @xcite considered a random binning scheme that partitions the collection of all length-@xmath6 binary sequences into an appropriately chosen _ group code _ and its cosets .
for the wiretap channel setting , as shown in sec .
[ sec : mswtc ] , a random binning scheme can be easily modified into a nested binning scheme to efficiently embed high - security bits into low - security ones . the main goal of this section
is to extend this result from the classical setting of wiretap channel to wiretap channel ii .
more specifically , assume that a realization of the subset @xmath211 has two possible sizes , @xmath219 and @xmath220 , where @xmath221 .
the transmitter has two independent messages , the high - security message @xmath63 and the low - security message @xmath65 , uniformly drawn from @xmath64 and @xmath66 respectively .
when the size of the realization @xmath211 is @xmath220 , both messages @xmath63 and @xmath65 need to be secure , i.e. , @xmath222 in the limit as @xmath9 .
in addition , when the size of the realization of @xmath211 is @xmath219 , the high - security message @xmath63 needs to remain secure , i.e. , @xmath223 in the limit as @xmath9 .
we term this communication scenario as _ two - level security wiretap channel ii _
, in line with our previous terminology in sec .
[ sec : mswtc ] . by the results of @xcite , without needing to communicate the low - security message @xmath65 , the maximum achievable @xmath67 is @xmath224 . without the additional secrecy constraint @xmath223 on the high - security message @xmath63 ,
the messages @xmath225 can be viewed as a single message @xmath4 with rate @xmath72 , and the maximum achievable @xmath72 is @xmath226 .
the main result of this section is to show that the rate pair @xmath227 is indeed achievable , from which we may conclude that two - level security wiretap channels ii are _ perfectly _ embeddable .
moreover , perfect embedding can be achieved by a nested binning scheme that uses a _ two - level _ coset code .
the results are summarized in the following theorem .
two - level security wiretap channels ii are perfectly embeddable .
moreover , perfect embedding can be achieved by a nested binning scheme that uses a two - level coset code .
_ proof : _ fix @xmath115 . consider a binary parity - check matrix @xmath228\ ] ] where the size of @xmath229 is @xmath230 and the size of @xmath231 is @xmath232 .
let @xmath233 be a one - on - one mapping between @xmath234 and the binary vectors of length @xmath235 , and let @xmath236 be a one - on - one mapping between @xmath237 and the binary vectors of length @xmath238
. for a given message pair @xmath102 , the transmitter randomly ( according to a uniform distribution ) chooses a solution @xmath19 to the linear equations @xmath239= \left [ \begin{array}{c } s_1(m_1 ) \\
s_2(m_2 ) \end{array } \right ] \label{eq : gc}\ ] ] and sends it to the legitimate receiver . when the parity - check matrix @xmath240 has _ full _
( row ) rank , the above encoding procedure is equivalent of a nested binning scheme that partitions the collection of all length-@xmath6 binary sequences into bins and subbins using a two - level coset code with parity - check matrices @xmath241 .
moreover , let @xmath242 be the columns of @xmath240 and let @xmath243 .
define @xmath244 as the dimension of the subspace spanned by @xmath245 and @xmath246 when the size of the realization of @xmath211 is @xmath220 , by ( * ? ? ?
* lemma 4 ) we have @xmath247 note that the low - security message @xmath65 is uniformly drawn from @xmath248 .
so by , for a given high - security message @xmath107 , the transmitted sequence @xmath19 is randomly chosen ( according to a uniform distribution ) as a solution to the linear equations @xmath249 .
if we let @xmath250 be the columns of @xmath229 and define @xmath251 where @xmath252 is the dimension of the subspace spanned by @xmath253 , we have again from ( * ? ? ?
* lemma 4 ) @xmath254 when the size of the realization of @xmath211 is @xmath219 .
let @xmath255 when we have either @xmath240 does _ not _ have full rank , or @xmath256 , or @xmath257 , and let @xmath258 otherwise . by using a randomized argument that generates the entries of @xmath240 independently according to a uniform distribution in @xmath259
, we can show that there exists an @xmath240 with @xmath258 for sufficiently large @xmath6 ( see appendix [ app:3 ] for details ) .
for such an @xmath240 , we have from and that @xmath260 when the size of the realization of @xmath211 is @xmath220 , and @xmath261 when the size of the realization of @xmath211 is @xmath219 . letting @xmath9 and @xmath160 ( in that order )
proves the achievability of the rate pair @xmath227 and hence completes the proof of the theorem .
in this paper we considered the problem of simultaneously communicating two messages , a high - security message and a low - security message , to a legitimate receiver , referred to as the security embedding problem .
an information - theoretic formulation of the problem was presented . with appropriate coding architectures
, it was shown that a significant portion of the information bits can receive additional security protections without sacrificing the overall rate of communication .
key to achieve efficient embedding was to use the low - security message as part of the transmitter randomness to protect the high - security message when the eavesdropper channel realization is strong .
for the engineering communication scenarios with real channel input and additive white gaussian noise , it was shown that the high - security message can be embedded into the low - security message at full rate without incurring any loss on the overall rate of communication for both scalar and independent parallel gaussian channels ( under an average individual per - subchannel power constraint ) .
the scenarios with multiple transmit and receive antennas are considerably more complex and hence require further investigations . finally , note that even though in this paper we have only considered providing two levels of security protections to the information bits , most of the results extend to multiple - level security in the most straightforward fashion . in the limit when the security levels change continuously , the number of secure bits delivered to the legitimate receiver would depend on the realization of the eavesdropper channel even though such realizations are unknown a priori at the transmitter .
first note that when @xmath262 forms a markov chain in that order , we have @xmath89 for any jointly distributed @xmath263 that satisfies the markov chain @xmath88 . to show that the sufficient condition is also necessary , let @xmath71 be an achievable rate pair .
following fano s inequality @xcite and the asymptotic perfect secrecy constraints and , there exists a sequence of codes ( indexed by the block length @xmath6 ) of rate pair @xmath71 such that @xmath264 where @xmath265 in the limit as @xmath9 .
following and , we have @xmath266\\ & = & h(m_1|z_1^n)-h(m_1,m_2|y^n ) \\ & \leq & h(m_1,m_2|z_1^n)-h(m_1,m_2|y^n ) \\ & = & i(m_1,m_2;y^n)-i(m_1,m_2;z_2^n ) . \end{aligned}\ ] ] let @xmath267 , @xmath268,\ldots , y[i-1])$ ] , @xmath269, ... ,z_1[n])$ ] and @xmath270 : = ( y^{i-1},z_{1,i+1}^n)$ ] .
we further have @xmath271|y^{i-1 } ) - i(m;z_1[i]|z_{1,i+1}^n)\right]\\ & \stackrel{(a)}= & \sum_{i=1}^n \left[i(m;y[i]|y^{i-1},z_{1,i+1}^n ) - i(m;z_1[i]|y^{i-1},z_{1,i+1}^n)\right]\\ & = & \sum_{i=1}^n \left[i(m;y[i]|u[i ] ) - i(m;z_1[i]|u[i])\right]\\ & = & n\left[i(m;y[q]|u[q],q ) - i(m;z_{1}[q]|u[q],q)\right]\\ & = & n\left[i(m , u[q],q;y[q]|u[q],q ) - i(m , u[q],q;z_{1}[q]|u[q],q)\right]\\ & = & n\left[i(v[q];y[q]|u[q],q ) - i(v[q];z_{1}[q]|u[q],q)\right]\end{aligned}\ ] ] where ( a ) is due to the csiszr - krner sum equality ( * ? ? ?
* lemma 7 ) , @xmath272 is a standard time - sharing variable @xcite , and @xmath273:=(m , u[q],q)$ ] . following and , we have @xmath274\\ & = & i(m;y^n)-i(m;z_2^n)\\ & = & \sum_{i=1}^n \left[i(m;y[i]|y^{i-1 } ) - i(m;z_2[i]|z_{2,i+1}^n)\right]\\ & \stackrel{(b)}= & \sum_{i=1}^n \left[i(m;y[i]|y^{i-1},z_{2,i+1}^n ) - i(m;z_2[i]|y^{i-1},z_{2,i+1}^n)\right]\\ & = & \sum_{i=1}^n \left[i(m , y^{i-1},z_{1,i+1}^n , z_{2,i+1}^n;y[i ] ) - i(m , y^{i-1},z_{1,i+1}^n , z_{2,i+1}^n;z_2[i])\right]\\ & & \quad\quad -\sum_{i=1}^n \left[i(y^{i-1},z_{2,i+1}^n;y[i ] ) - i(y^{i-1},z_{2,i+1}^n;z_2[i ] ) \right]\\ & & \quad\quad -\sum_{i=1}^n \left[i(z_{1,i+1}^n;y[i]|m , y^{i-1},z_{2,i+1}^n ) - i(z_{1,i+1}^n;z_2[i]|m , y^{i-1},z_{2,i+1}^n)\right]\\ & \stackrel{(c)}\leq & \sum_{i=1}^n \left[i(m , y^{i-1},z_{1,i+1}^n , z_{2,i+1}^n;y[i ] ) - i(m , y^{i-1},z_{1,i+1}^n , z_{2,i+1}^n;z_2[i])\right]\\ & \stackrel{(d)}= & \sum_{i=1}^n \left[i(m , y^{i-1},z_{1,i+1}^n;y[i ] ) - i(m , y^{i-1},z_{1,i+1}^n;z_2[i])\right]\\ & = & \sum_{i=1}^n \left[i(m , u[i];y[i ] ) - i(m , u[i];z_2[i])\right]\\ & = & n\left[i(m , u[q];y[q]|q ) -i(m , u[q];z_{2}[q]|q)\right]\\ & = & n\left[i(m , u[q],q;y[q ] ) -i(m , u[q],q;z_{2}[q])-\left(i(y[q];q)-i(z_2[q];q)\right)\right]\\ & = & n\left[i(v[q];y[q ] ) -i(v[q];z_{2}[q])-\left(i(y[q];q)-i(z_2[q];q)\right)\right]\\ & \stackrel{(e)}\leq & n\left[i(v[q];y[q ] ) -i(v[q];z_{2}[q])\right]\end{aligned}\ ] ] where ( b ) follows from the csiszr - krner sum equality ( * ? ? ?
* lemma 7 ) , ( c ) is due to the markov chain , ( d ) is due to the markov chain , and ( e ) follows again from the markov chain and the fact that the channel is memoryless . finally , we complete the proof of the theorem by letting @xmath275,q)$ ] , @xmath276 $ ] , @xmath277 $ ] , @xmath278 $ ] , @xmath279 $ ] , @xmath280 $ ] and @xmath9 .
as shown in theorem [ thm : dm2 ] , when we have the markov chains and , there exists a random triple @xmath263 satisfying the markov chain @xmath88 and such that @xmath281 and @xmath282 .
in fact , the sum rate @xmath72 can be further bounded from above as @xmath283\\ & \stackrel{(a)}= & i(x;y)-i(x;z_2)-\left[i(x;y|v)-i(x;z_2|v)\right]\\ & \stackrel{(b ) } \leq & i(x;y)-i(x;z_2)\end{aligned}\ ] ] where ( a ) follows from the markov chain @xmath284 , and ( b ) follows from the markov chain so @xmath285 .
when we further have the markov chain , @xmath67 can be further bounded from above as @xmath286\\ & \stackrel{(c)}= & i(v;y)-i(v;z_1 ) - \left[i(u;y)-i(u;z_1)\right]\\ & \stackrel{(d)}\leq & i(v;y)-i(v;z_1)\\ & = & i(v , x;y)-i(v , x;z_1)-\left[i(x;y|v)-i(x;z_1|v)\right]\\ & \stackrel{(e)}\leq & i(v , x;y)-i(v , x;z_1)\\ & \stackrel{(f)}= & i(x;y)-i(x;z_1)\end{aligned}\ ] ] where ( c ) follows from the markov chain @xmath287 , ( d ) and ( e ) follow from the markov chain so @xmath288 and @xmath289 , and ( f ) follows from the markov chain @xmath290 .
this completes the proof of the theorem .
to show that there exists a parity - check matrix @xmath240 such that @xmath258 , it is sufficient to show that @xmath291 where @xmath292 denotes the expectation of a random variable @xmath1 .
let @xmath293 and @xmath294 for @xmath295 . by the union bound
, we have @xmath296 by ( * ? ? ?
* lemma 6 ) , @xmath297 for sufficiently large @xmath6 . by (
* lemma 5 ) , for any @xmath243 such that @xmath298 @xmath299 since the total number of different subsets of @xmath212 is @xmath300 , we have @xmath301 for @xmath302 . substituting and into
proves that @xmath291 for sufficiently large @xmath6 and hence completes the proof .
r. liu , t. liu , h. v. poor , and s. shamai ( shitz ) , new results on multiple - input multiple - output gaussian broadcast channels with confidential messages,"_ieee trans .
inf . theory _ , submitted for publication .
available online at http://arxiv.org/abs/1101.2007 y. k. chia and a. el gamal , 3-receiver broadcast channels with common and confidential messages , " _ ieee trans .
inf . theory _ , submitted for publication .
available online at http://arxiv.org/abs/0910.1407 | this paper considers the problem of simultaneously communicating two messages , a high - security message and a low - security message , to a legitimate receiver , referred to as the security embedding problem .
an information - theoretic formulation of the problem is presented . a coding scheme that combines rate splitting , superposition coding , nested binning and channel prefixing
is considered and is shown to achieve the secrecy capacity region of the channel in several scenarios .
specifying these results to both scalar and independent parallel gaussian channels ( under an average individual per - subchannel power constraint ) , it is shown that the high - security message can be embedded into the low - security message at full rate ( as if the low - security message does not exist ) without incurring any loss on the overall rate of communication ( as if both messages are low - security messages ) .
extensions to the wiretap channel ii setting of ozarow and wyner are also considered , where it is shown that perfect " security embedding can be achieved by an encoder that uses a two - level coset code . |
the unified model for seyfert galaxies proposes that all types of seyfert galaxy are fundamentally the same , however , the presence of an optically thick structure obscures the broad line region ( blr ) in many systems . in this paper
it is assumed that , in the majority of seyfert 2s , this structure is a dusty molecular torus " although other galactic structures ( e.g. dust lanes / starbursts , see malkan , gorjian and tam , 1998 ) can perform the same role . in this scenario
the classification of a seyfert 1 or seyfert 2 galaxy ( seyfert 1broad permitted lines , seyfert 2narrow permitted lines ) depends on the inclination of the torus to the line of sight ( antonucci , 1993 ) .
probably the most convincing evidence for this model comes from optical spectropolarimetry . using this technique , the scattered emission from the blr of many seyfert 2s
is revealed in the form of broad lines in the polarised flux ( e.g. antonucci and miller , 1985 , young et al , 1996 , heisler , lumsden and bailey , 1997 ) . in this unified picture
the high energy central source emission ( optical to x - ray continuum ) is absorbed by the dust within the torus which re - emits this energy at infrared ( ir ) wavelengths .
independent strong support has been given by hard x - ray ( hx , 2 to 10 kev ) , near - ir and mid - ir observations ( e.g. turner et al , 1997 , risaliti , maiolino and salvati , 1999 , alonso - herrero , ward , kotilainen , 1997 and clavel et al , 2000 ) showing that seyfert 2s are generally characterised by strong absorption whilst seyfert 1 galaxies are relatively unabsorbed .
heisler , lumsden and bailey ( 1997 , hereafter hlb ) performed an optical spectropolarimetric study of a well defined and statistically complete iras 60@xmath4 m selected seyfert 2 sample to determine the statistical detectability of polarised broad lines .
the objects were selected at 60@xmath4 m to reduce the possibility of biasing due to torus inclination / extinction effects and all objects were observed to the same signal to noise to ensure similar detection thresholds . in this study a striking relationship between the detectability of polarised broad h@xmath0 and the iras @xmath1 flux ratio was found where only those galaxies with warm iras colours ( @xmath54.0 ) showed a hidden broad line region ( hblr ) .
both seyfert 2 galaxy types were found to be well matched in terms of redshift , overall polarisation and detection rate of compact nuclear radio emission .
therefore , without any apparent contradictory evidence , hlb suggested that the iras @xmath1 ratio provides a measure of the inclination of the torus to the line of sight : in a cool seyfert 2 the torus is so highly inclined that even the broad line scattering screen is obscured .
i present here hx evidence that suggests this picture is incorrect and provide a new view that is consistent with other observations .
the picture presented by hlb appears reasonable . assuming that the seyfert torus is optically thick at mid - ir wavelengths ( e.g. pier and krolik , 1993 , granato and danese , 1994 , efstathiou and rowan - robinson , 1995 ) the mid - ir to far - ir flux ratio should vary depending upon the inclination of the torus to the line of sight .
a simple prediction of this picture is that seyfert 1 galaxies should show warmer colours than seyfert 2 galaxies .
the mean iras @xmath1 flux ratios of seyfert galaxies from the iras 60@xmath4 m selected bright galaxy sample ( bgs , soifer et al , 1989 ) , as classified by kim et al ( 1995 ) and using data from the literature ( see section 3 ) are [email protected] and [email protected] for seyfert 1s and 2s respectively .
the seyfert 1s do not statistically show warmer colours in this sample .
however , it could be argued that these ratios are biased by differences in the star formation between seyfert 1s and 2s or optical depth effects in the seyfert 2 nuclei , biasing the ratio towards warm objects . in any case
the most direct test of the inclination picture is made with hx observations .
one of the key supports of the unified model comes from hx observations where the nuclear extinction is directly determined from the observed spectral slope .
seyfert 1 galaxies are characterised by little or no absorption 20@xmath7log(@xmath8)@xmath721 @xmath9 whilst seyfert 2s have significant , sometimes extreme , absorption 22@xmath7log(@xmath8)@xmath725 @xmath9 ( e.g. turner et al , 1997 and risaliti , maiolino and salvati , 1999 ) .
although the hx properties of seyferts are too poorly known to allow a detailed measure of the torus inclination , as the cool seyfert 2s are more highly inclined than the warm seyfert 2s in the hlb interpretation , they should statistically show higher column densities . to date 13 of the galaxies in the hlb sample
have been observed with either bepposax or asca .
the other 3 objects have been observed by einstien or in the heao1/a survey . in the case of the heao1/a objects
only upper limits could be placed .
for these two galaxies ( ngc34 and ngc1143 ) i have used the upper limits and unextincted [ oiii]@xmath105007 emission line fluxes to predict their nuclear extinction using the diagnostic diagram of bassani et al ( 1999 ) .
the distribution of hx derived column densities are shown in figure 1 and presented in table 1 .
lclcccr + galaxy & z & @xmath11 & @xmath12/@xmath13 & @xmath14/@xmath15 & hblr ?
& @xmath8 + + ngc0034 & 0.01978 & -14.9@xmath16 & 7.01 & 25.0 & no & @xmath1723.0@xmath18 + ngc1068 & 0.00379 & -12.7 & 2.07 & 6.2 & yes & @xmath1924.0@xmath20 + ngc1143 & 0.02822 & -15.0@xmath16 & 8.37 & 11.0 & no & @xmath1722.0@xmath18 + i05189 - 2524 & 0.04256 & -14.5 & 3.97 & 5.8 & yes & 22.7@xmath20 + ngc4388 & 0.00842 & -14.8 & 2.96 & 5.5 & yes & 23.6@xmath20 + ic3639 & 0.01096 & -14.1@xmath16 & 3.32 & 4.6 & yes & @xmath1924.0@xmath21 + i13197 - 1627 & 0.01718 & -14.0 & 2.06 & 4.8 & yes & 23.9@xmath20 + ngc5135 & 0.01372 & -14.4@xmath16 & 7.04 & 7.1 & no & @xmath1924.0@xmath20 + i19254 - 7245 & 0.06171 & -15.1@xmath16 & 4.42 & 10.0 & no & 23.3@xmath22 + ic5063 & 0.01135 & -13.3 & 1.36 & 5.9 & yes & 23.4@xmath20 + ngc7130 & 0.01615 & -15.4@xmath16 & 7.78 & 8.7 & no & @xmath1924.0@xmath21 + ngc7172 & 0.00859 & -16.4@xmath16 & 7.50 & 6.8 & no & 22.9@xmath20 + ngc7496 & 0.00550 & -15.4@xmath16 & 5.28 & 5.6 & no & 22.7@xmath23 + ngc7582 & 0.00525 & -14.1@xmath16 & 7.63 & 8.9 & no & 23.1@xmath20 + ngc7590 & 0.00532 & -16.5@xmath16 & 8.32 & 17.8 & no & @xmath720.9@xmath20 + ngc7674 & 0.02901 & -14.9 & 2.95 & 4.4 & yes & @xmath1924.0@xmath20 + + _ notes .
the h@xmath0 emission line flux is in log(ergs@xmath24@xmath9 ) and refers to the observed polarised broad h@xmath0 flux ( p = predicted flux ) ; the hblr refers to whether polarised broad h@xmath0 emission has been detected ; @xmath8 in log(@xmath25
. references .
( a ) bassani et al , 1999 , ( b ) pappa , georgantopoulos and stewart , 1999 , ( c ) risalti , maiolino and salvati , 1999 , ( d ) polletta et al , 1996 , ( e ) kruper , urry and canziares , 1990 _ the derived column densities show that an optically thick structure exists in both the warm and cool seyfert 2 galaxy types , with the notable exception of ngc7590 which may be a seyfert 1 with galactic dust obscuring the blr , however , signficantly there is very little difference in the distribution of column densities .
the overall distribution is similar to that found for the [ oiii]@xmath105007 selected seyfert sample of risaliti , maiolino and salvati ( 1999 ) suggesting that the far - ir selects seyferts in a reasonably unbiased manner : @xmath2635% of the objects are compton thick ( i.e. log(@xmath27 @xmath9 ) , the mean log(@xmath8 ) for the whole sample is [email protected] @xmath9 and the mean for the warm and cool seyfert 2s are [email protected] @xmath9 and [email protected] @xmath9 ( [email protected] @xmath9 if ngc7590 is excluded ) respectively .
the hypothesis of hlb could still be retained if the cool seyfert 2s are compton thick and the determined column densities refer to the extinction suffered by the scattered emission , however , the mean log([oiii]/hx ) of [email protected] and [email protected] for the warm and cool seyfert 2s respectively suggest that this is not the case ( see bassani et al , 1999 ) .
therefore the most direct conclusion to make from this hx analysis is that the iras @xmath1 flux ratio does not indicate the inclination angle of the torus in seyfert 2s .
if the iras @xmath1 colour ratio is not an indicator of the inclination of the dusty torus , what does this colour ratio imply ?
a natural starting point is to compare the hlb seyfert properties to those of non - seyfert galaxies .
a good comparison is the bgs sample which is selected at the same wavelength as the hlb sample ( 60@xmath4 m ) and has a very similar flux limit . the bgs sample is partially classified by kim et al ( 1995 ) using the optical emission line ratio technique ( e.g. baldwin , phillips and terlevich , 1981 ) . to increase the number of classified objects i have taken these observations and other optical spectroscopic observations from the literature , classifying 77% of the bgs sample , see table 2 .
these galaxies have been classified using the emission line diagnostics of veilleux and osterbrock ( 1987 ) and the mode classification for each galaxy is adopted . for brevity only the [ nii]@xmath106583/h@xmath0 vs [
oiii]@xmath105007/h@xmath28 diagram is shown here , see figure 2 .
cccc + hii & liners & agn & no emission lines + + 62% & 25% & 12% & 1% + + a logical first step is to compare the iras colours , see figure 3 .
the ir warm region is clearly dominated by seyfert 2s although the cool region also contains hii and liner galaxies with a wide range of iras colours .
the cool seyfert 2s can not be distinguished from the hii and liner galaxies in terms of their iras emission ( note there are no seyfert 2s with log(@xmath290.9 due to the hlb selection criteria ) .
the distribution of hx derived nuclear column densities suggest an optically thick structure ( i.e. the torus ) exists in both seyfert 2 galaxy types which , according to the unified model , should emit thermally at ir wavelengths . therefore , although the hx emission shows that a seyfert nucleus is present in the cool seyfert 2s , the ir emission from the torus must be dominated by galactic emission in the large iras apertures ( as previously suggested by alexander et al , 1999 ) .
additional evidence for this picture is found in the distribution of optical emission line ratios where the cool seyfert 2s have , on average , weaker [ oiii]/h@xmath28 emission , see figure 2 . assuming that both the warm and cool seyfert 2s have the same basic seyfert nucleus and galactic emission , the lower mean emission line ratio in the cool seyfert 2s implies a larger ratio of galactic to seyfert activity .
indeed in one galaxy ( ngc7496 ) the observed emission line ratio is consistent with that of an hii galaxy even though it clearly has hx emission and therefore a seyfert nucleus .
hii galaxies produce mid - ir pah emission ( see puget and leger , 1989 ) whilst pure seyfert nuclei do not . therefore in the above scenario the pah equivalent width should vary based on the iras @xmath1 colour ratio .
indeed lutz et al ( 1998 ) found that this occurs in the extremely luminous object class of ulirgs although this test has not been performed for seyfert 2 galaxies .
unfortunately suitable mid - ir observations are not available for the complete hlb sample although seyfert 2 observations from clavel et al ( 2000 ) and hii galaxy observations from rigopoulou et al ( 1999 ) provide the general picture , see figure 4 .
the distribution of pah emission is reasonably well segregated with the cool seyfert 2s occupying a similar region to the hii galaxies although there is not a tight correlation .
this may be because the galactic emission can be a combination of both galactic cirrus emission , which show strong pah features , and more intense star formation emission , which show weak pah features ( e.g. laurent et al , 1999 ) .
the cool seyfert 2s have iras colours , optical emission line ratios and pah emission suggestive of a weak seyfert nucleus and a dominant galactic component . can the spectropolarimetric results be explained in the same way ? as the [ oiii]@xmath105007 emission line is a good indicator of the strength of the seyfert nucleus ( e.g. alonoso - herrero et al , 1997 , bassani et al , 1999 ) it should be possible to estimate the strength of polarised broad h@xmath0 in the cool seyfert 2s . indeed the [ oiii]@xmath105007 emission line flux , corrected for galactic extinction and then re - extincted to the wavelength of h@xmath0 to simulate the probable extinction suffered by the h@xmath0 emission , shows a good correlation with polarised broad h@xmath0 flux , see figure 5 and table 1 . in general the predicted fluxes of the cool seyfert 2s are a factor @xmath263 lower than the warm seyfert 2s and in many cases fall below that easily obtainable on a 4 metre telescope . in two cases
the predicted fluxes are similar to those found for warm seyfert 2s and in one of these objects ( ngc7582 ) broad h@xmath0 has been directly observed ( aretxaga et al , 1999 ) .
however , in determining the detectability of polarised broad h@xmath0 there is another factor to consider as the galaxies were observed to the same signal to noise and therefore the detectability of a hblr also depends on the optical brightness of the galaxy . the mean optical ( v band ) magnitudes are [email protected] and [email protected] for the warm and cool seyfert 2s respectively , showing that there is no statistical difference in their overall optical brightnesses .
therefore as the polarised broad h@xmath0 fluxes are lower in the cool seyfert 2s , whilst the overall fluxes are not , the sensitivity to polarised broad h@xmath0 should be worse by a factor of @xmath263 in the cool seyfert 2s .
these two factors can account for the lack of detected hblrs in the cool seyfert 2s and agree with the overall hypothesis that the differences between the seyfert 2 galaxy types are that the cool seyfert 2s have a weak seyfert nucleus and a dominant galactic component .
the weaker seyfert emission in the cool seyfert 2s should lead to a lower observed polarisation , however , hlb claim no statistical difference between the seyfert 2 galaxy types .
this apparent inconsistency can be reconciled as there is another component to the observed polarisation : galactic polarisation .
galactic polarisation is proportional to the measured galactic extinction ( serkowski , mathewson and ford , 1975 and scarrott , 1996 ) and as the cool seyfert 2s have a higher galactic extinction ( i.e. as measured by the h@xmath0/h@xmath28 emission line ratio ( ward et al , 1987 ) , see hlb and table 1 ) they should also have a larger galactic polarisation component . an obvious question to raise
is why do the cool seyfert 2s have a higher galactic extinction ?
a natural explanation for this would be that the galactic disc is more highly inclined in the cool objects , however the mean inclinations ( measured from dss images ) of 43@xmath630 and 42@xmath625 degrees for the warm and cool seyfert 2 galaxies respectively show that this is not the case .
therefore there must be another component contributing to the higher extinction in the cool seyfert 2s .
analysis of dss images of the hlb sample suggest that cool seyfert 2s are more likely to have nuclear dust lanes .
dust lanes are known to be very efficient at causing galactic polarisation ( see centaurus a , hough et al , 1987 ) and would also account for the higher galactic extinction in the cool seyfert 2s .
high resolution images of the hlb sample are required to test this scenario .
hlb showed a dichotomy in the detectability of polarised broad h@xmath0 in seyfert 2 galaxies where only those objects with warm iras colours showed polarised broad line emission .
they suggested that those seyfert 2s with cool iras colours had highly inclined tori which obscured the broad line scattering screen .
i have presented evidence against this picture and suggested that the ir dichotomy is due to the relative strength of galactic to seyfert emission .
the main results of this study are : \(i ) the hard x - ray derived column densities of the warm and cool seyfert 2s are statistically the same and therefore the iras @xmath1 flux ratio can not imply the inclination of the torus in the standard unified model .
\(ii ) the cool seyfert 2s have iras colours and pah emission similar to hii galaxies and optical nuclear emission line ratios consistent with a mixture of seyfert and galactic activity . as the hard x - ray observations suggest
that both seyfert 2 galaxy types have similar nuclear properties , the differences in the iras @xmath1 flux ratios are most probably due to a weaker seyfert component and a more dominant galactic component in the cool seyfert 2s .
\(iii ) the difference in the detectability of polarised broad h@xmath0 is most probably due to a larger component of optical starlight in the cool seyfert 2s .
the reason for the lack of difference in observed polarisation between both seyfert 2 types is probably due to a larger galactic polarisation component in the cool galaxies caused by their larger galactic extinction .
i thank the tmr network ( fmrx - ct96 - 0068 ) for support through a postdoctoral grant . this research has made use of the nasa / ipac extragalactic database ( ned ) , which is operated by the jet propulsion laboratory , california instutite of technolody , under contract with nasa .
i am extremely grateful to the many people who have commented on earlier drafts of this paper particularly prof .
matt malkan , the referee , and dr .
stuart young .
i am in debt to the late dr .
charlene heisler for interesting and enthusiastic discussions whilst working on this topic . | an optical spectropolarimetric study has shown that the detectability of polarised broad h@xmath0 in seyfert 2 galaxies is correlated to the iras @xmath1 flux ratio where only those seyfert 2s with warm " iras colours show polarised broad line emission .
it was suggested that those seyfert 2s with cool " iras colours have highly inclined tori which obscure the broad line scattering screen .
i present here hard x - ray observations inconsistent with this picture showing that the derived column densities of warm and cool seyfert 2 galaxies are statistically the same .
i classify the bright galaxy sample to produce a non - seyfert comparison .
the analysis of the properties of these galaxies with the seyfert 2s suggest that the iras @xmath1 flux ratio is consistent with implying the relative strength of galactic to seyfert emission .
i show that this new picture can account for the absence of polarised broad h@xmath0 in the cool seyfert 2s .
# 1 2_2h@xmath2 2_h_2@xmath3 # 1.#2 # 1#2 # 1.#2 # 1#2 # 1.#2 # 1#2 # 1h#2m#3s # 1^h#2^m#3^s # 1d#2m#3s # 1#2#3 # 1h#2m#3.#4s ( ra ) # 1^h#2^m#3#4 # 1d#2m#3.#4s ( dec ) # 1#2#3#4 # 1.#2 # 1^m -0.55em.0.22em#2 = = = = = = = = # 1 # 1 # 1 # 1 @mathgroup@group @mathgroup@normal@groupeurmn @mathgroup@bold@groupeurbn @mathgroup@group @mathgroup@normal@groupmsamn @mathgroup@bold@groupmsamn = `` 019 = ' ' 016 = `` 040 = ' ' 336 = " 33e = = = = = = = = # 1 # 1 # 1 # 1 = = = = = = = = [ firstpage ] polarization - galaxies : active - infrared : galaxies - galaxies : seyfert |
plasma inhomogeneities across the magnetic field in the presence of finite - size charged grains causes a wide class of instabilities of an inhomogeneous dusty plasma called gradient instabilities .
such instabilities can be studied in the approximation on magnetic field where we have parallel straight field lines in order to simplify our treatment .
we look for instabilities in the very low frequency regime where a new spectrum instabilities and waves appear , induced by the dust collective dynamics : dust - acoustic - waves ( daws ) , dust - ion - acoustic - waves ( diaws ) , etc .
the frequency of daws are around 10 hz as determined in the laboratory and lower in astrophysical plasmas [ 1,2 ] . in the case that grains are in the micron range we expect a non - ideal behavior due to the fact that the particulate are highly charged and intermolecular forces could play certainly an important role . in order to discuss this problem
we compare the ideal properties with the simple hard - core model and in a next work we will use a better model by means of of the square - well model and the pad rational approximant to the equation of state [ 3 ] for hard - sphere gas , that in our knowledge is more realistic as the simple application of the van der waals equation of state [ 4 ] . in this paper
we show an analysis of the electrostatic waves and instabilities growth rates in a weakly non - ideal magnetized dusty plasma with density and temperature gradients , ignoring charge fluctuation .
as introduced before , the non - ideal behavior is characterized by the hardcore model defined by @xmath0 or in similar manner by the square - well model given by the ree and hoover expression [ 5 ] .
in this paper we introduce a new numerical treatment in combination with a more realistic formulation of the equation of state to simulate weak non ideal effects in order to analyze inhomogeneous vlasov - dusty plasma systems where a linearized dispersion relation is obtained . due to the lower frequency range ( @xmath1 )
, enough energy can be transferred from the particle to the wave and instabilities can be generated . in order to get an adequate linear dispersion relation with a magnetic field given by @xmath2 for maxwellian multi - species plasmas ( electron , ion and dust )
, we introduce our well known and very accurate multipolar approximation [ 6 ] for the @xmath3 dispersion function . in the presence of a magnetic field
we have the distribution function of the species @xmath4 , solution for the kinetic equation @xmath5 in the time dependent following form[7,8 ] @xmath6 { \bf \nabla } \phi ` ( r(t^{\prime } ) ) \cdot \frac{\partial f_{o\alpha } } { \partial { \bf v(}t^{\prime } { \bf ) } } dt^{\prime } \ ] ] where @xmath7 now , the dispersion relation in terms of the dielectric susceptibilities , in the low frequency approximation ( @xmath1 ) is @xmath8 where , @xmath9\ ] ] with : @xmath10 further , in order to simplify our expressions , we use : @xmath11 now , using the following identity for the dispersion function @xmath3 @xmath12,}$\nonumber}\ ] ] we obtain after several cumbersome algebraic manipulations the dielectric susceptibility in the form @xmath13 \right ) \right\ } \right ] \,\ ] ] in order to put our dispersion relation in a dimensionless form , we introduce following suitable definitions : @xmath14 now , using those results and assuming that @xmath15 we can write down eq.(3 ) as @xmath16 in the non ideal case ( dust ) we introduce a relation that in principle express the non ideal behavior of the system in terms of the pressure in the form @xmath17 given by the hard - core model . this model is taken for simplicity . a better model , as mentioned before , will be introduced in a future work .
now , following definitions are also useful @xmath18 those relations are very convenient by writing the full dispersion relation[4 ] .
in fact we have @xmath19 for the non - ideal case . for the ideal one ,
we use the well known relation @xmath20 , and in a similar way we get @xmath21 where @xmath22 .
two special cases can be worked out : + a ) density gradient equal to zero @xmath23 , that means , @xmath24 + +
b ) temperature gradient equal to zero @xmath25 , that means , @xmath26 + further we can introduce following relations in order to express dielectric susceptibilities in a suitable forms @xmath27 @xmath28 using those relations we arrive to the dispersion relation for the case b where we get : @xmath29\ ] ] @xmath30\ ] ] @xmath31\ ] ] where @xmath32\lambda_{p } $ ] and @xmath33 . in a similar way , it is possible to include the terms for case a , where we shall have @xmath34 introducing now the multipolar approximation to @xmath35 we can get a polynomial expression in the well known form[9 ] @xmath36 where coefficients @xmath37 and @xmath38 are functions of the system parameters .
such an expression is easy to solve and with high accuracy to find roots of the numerator .
an analysis of these solutions spectra permit us to give the imaginary parts @xmath39 in function of @xmath40 , which represent the growth rate instabilities .
the quasi - neutrality equation for dusty plasmas can be approached by a simplified one due to the high state of charge of the dust grains @xmath41 and the electron susceptibility can be neglected in the dispersion relation .
the range of the main parameters in the study of the low frequency oscillation of dust grains is established by the approximations that conduced to the simplified dispersion relation @xmath42 unstable dust oscillations ( @xmath43 ) are found for @xmath44 , @xmath45 . at the present time , we only give the results for the density gradient case ( _ i.e. _ @xmath46 ) . for slightly inhomogeneous plasmas with normalized density gradient length @xmath47 ,
the shape of the dust instability ( @xmath48 ) curve as function of the perpendicular to magnetic field wavelength ( @xmath49 ) is similar to that for ions , previously studied [ 8 ] . ) and for a relatively inhomogeneous one ( @xmath50).,width=359,height=359 ] the maximum value of the instability increases and narrows with the state of charge of the dust @xmath51 but decreases and get wider with the mass . for typical laboratory light dusty plasmas (
@xmath52 , @xmath53 ) the instability of dust acoustic or electrostatic waves is narrower and smaller than that for ions in figure 1 the peak of the left corresponds to the typical shape of instability of slightly inhomogeneous plasmas , while the right region of instability appears for density gradient lengths of the order of a hundred of debye lengths ( @xmath54 ) . for higher density gradients ( @xmath55 ) , this new instability region is wider and so high as the typical one . for even higher density gradients ( @xmath56 ) ,
figure 2 shows that the new right region gives a higher instability .
this figure also shows the effect of the non ideality of the plasma .
necessary condition for the exhibition of dust acoustic waves . for typical laboratory dust radius the @xmath57 parameter of the hard core potential equation of state , is of the order of @xmath58 . and for typical values of ion density of @xmath59 ( and corresponding @xmath60 , by quasi - neutrality relation ) , it appears a new intermediate instability region which can reach a maximum for denser plasmas ( @xmath61 ) or larger dust particles ( @xmath62 )
. this maximum is limited for the relation for dust collective behavior @xmath63
1 . j. h. chen , j. b. du , and lin i. , _
j. phys_. d : appl .
phys . * 27 * , 296(1994 ) 2 .
a. barkan , r. l. merlino , and n. dangelo , _ phys .
plasmas _ , * 2 * , 3563(1995 ) 3 .
reichl l.e . , _ a modern course in statistical physics
_ , edward arnold , 1991 . 4
. n. n. rao , `` _ frontier in dusty plasmas _ '' , y. nakamura , t. yakota , and p.k .
shukla , eds . , elsevier science b.v,(2000 ) 5 .
f. h. ree and w. g. hoover , _ j. chem .
phys _ , * 40 * , 939(1964 ) 6 .
p. martn et al . , j. math
* 21 * , 280 ( 1980 ) 7 .
mikhailovsky a. b. , _ handbook of plasma physics _ , rosenbluth and sagdeev eds . , north holland , amsterdam , 1983 . 8 .
a. galeev et al .
soviet physics .
jetp * 17 * , 615 ( 1963 ) 9 .
j. puerta and c. cereceda , proc .
icppp * 1 * , 94 - 97 ( 1996 ) | in this paper we introduce an algebraic form of the dispersion relation for a non ideal inhomogeneous dusty plasma in order to improve drastically the calculation of the drift instability growth rate .
this method makes use of the multipole approximation of the z dispersion function , previously published , and valid for the entire range . a careful analysis of the solutions spectra of this kind of polynomial equation permits us to calculate easily the growth rate of the drift instability for the ion - dust and dust acoustic mode .
the value of the parallel to magnetic field wavelength for which the instability reaches the maximal value is carefully localized and discussed .
the unstable dust - ion and dust acoustic mode are discriminated and analyzed in function of the density gradient , te / ti - ratio , and dust grain radius . |
while the average treatment effect can be easily estimated without bias in randomized experiments , treatment effect heterogeneity plays an essential role in evaluating the efficacy of social programs and medical treatments .
we define treatment effect heterogeneity as the degree to which different treatments have differential causal effects on each unit . for example , ascertaining subpopulations for which a treatment is most beneficial ( or harmful ) is an important goal of many clinical trials .
however , the most commonly used method , subgroup analysis , is often inappropriate and remains one of the most debated practices in the medical research community [ e.g. , @xcite ] .
estimation of treatment effect heterogeneity is also important when ( 1 ) selecting the most effective treatment among a large number of available treatments , ( 2 ) designing optimal treatment regimes for each individual or a group of individuals [ e.g. , @xcite ] , ( 3 ) testing the existence or lack of heterogeneous treatment effects [ e.g. , @xcite ] , and ( 4 ) generalizing causal effect estimates obtained from an experimental sample to a target population [ e.g. , @xcite ] . in all of these cases , the researchers must infer how treatment effects vary across individual units and/or how causal effects differ across various treatments .
two well - known randomized evaluation studies in the social sciences serve as the motivating applications of this paper .
earlier analyses of these data sets focused upon the estimation of the overall average treatment effects and did not systematically explore treatment effect heterogeneity .
first , we analyze the get - out - the - vote ( gotv ) field experiment where many different mobilization techniques were randomly administered to registered new haven voters in the 1998 election [ @xcite ] .
the original experiment used an incomplete , unbalanced factorial design , with the following four factors : a personal visit , 7 possible phone messages , 0 to 3 mailings , and one of three appeals applied to visit and mailings ( civic duty , neighborhood solidarity , or a close election ) .
the voters in the control group did not receive any of these gotv messages .
additional information on each voter includes age , residence ward , whether registered for a majority party , and whether the voter abstained or did not vote in the 1996 election . here
, our goal is to identify a set of gotv mobilization strategies that can best increase turnout . given the design , there exist 193 unique treatment combinations , and the number of observations assigned to each treatment combination ranges dramatically , from the minimum of 4 observations ( visited in person , neighbor / civic - neighbor phone appeal , two mailings , with a civic appeal ) to the maximum of @xmath0 ( being visited in person , with any appeal ) .
the methodological challenge is to extract useful information from such sparse data .
the second application is the evaluation of the national supported work ( nsw ) program , which was conducted from 1975 to 1978 over 15 sites in the united states .
disadvantaged workers who qualified for this job training program consisted of welfare recipients , ex - addicts , young school dropouts , and ex - offenders .
we consider the binary outcome indicating whether the earnings increased after the job training program ( measured in 1978 ) compared to the earnings before the program ( measured in 1975 ) .
the pre - treatment covariates include the 1975 earnings , age , years of education , race , marriage status , whether a worker has a college degree , and whether the worker was unemployed before the program ( measured in 1975 ) .
our analysis considers two aspects of treatment effect heterogeneity .
first , we seek to identify the groups of workers for whom the training program is beneficial .
the program was administered to the heterogeneous group of workers and , hence , it is of interest to investigate whether the treatment effect varies as a function of individual characteristics .
second , we show how to generalize the results based on this experiment to a target population . such an analysis is important for policy makers who wish to use experimental results to decide whether and how to implement this program in a target population . to address these methodological challenges ,
we formulate the estimation of heterogeneous treatment effects as a variable selection problem [ see also @xcite ] .
we propose the squared loss support vector machine ( l2-svm ) with separate lasso constraints over the pre - treatment and causal heterogeneity parameters ( section [ secmodel ] ) .
the use of two separate constraints ensures that variable selection is performed separately for variables representing alternative treatments ( in the case of the gotv experiment ) and/or treatment - covariate interactions ( in the case of the job training experiment ) . not only
do these variables differ qualitatively from others , they often have relatively weak predictive power .
the proposed model avoids the ad - hoc variable selection of existing procedures by achieving optimal classification and variable selection in a single step [ e.g. , @xcite ] .
the model also directly incorporates sampling weights into the estimation procedure , which are useful when generalizing the causal effects estimates obtained from an experimental sample to a target population . to fit the proposed model with multiple regularization constraints , we develop an estimation algorithm based on a generalized cross - validation ( gcv ) statistic .
when the derivation of an optimal treatment regime rather than the description of treatment effect heterogeneity is of interest , we can replace the gcv statistic with the average effect size of the optimal treatment rule [ @xcite ] . the proposed methodology with the gcv statistic does not require cross - validation and hence is more computationally efficient than the commonly used methods for estimation of treatment effect heterogeneity such as boosting [ @xcite ] , bayesian additive regression trees ( bart ) [ @xcite ] , and other tree - based approaches [ e.g. , @xcite , @xcite ] . while most similar to a bayesian logistic regression with noninformative prior [ @xcite ] , the proposed method uses lasso constraints to produce a parsimonious model .
to evaluate the empirical performance of the proposed method , we analyze the aforementioned two randomized evaluation studies ( section [ secapplications ] ) .
we find that personal visits are uniformly more effective than any other treatment method , while sending three mailings with a civic duty message is the most effective treatment without a visit . in addition , every mobilization strategy with a phone call , but no personal visit , is estimated to have either a negative or negligible positive effect . for the job training study , we find that the program is most effective for low - education , high income non - hispanics , unemployed blacks with some college , and unemployed hispanics with some high school .
in contrast , the program would be least effective when administered to old , unemployed recipients , unmarried whites with a high school degree but no college , and high earning hispanics with no college . finally , we conduct simulation studies to compare the performance of the proposed methodology with that of various alternative methods ( section [ secsimulations ] ) .
the proposed method admits the possibility of no treatment effect and yields a low false discovery rate , when compared to the nonsparse alternative methods that _ always _ estimate some effects . despite reductions in false discovery ,
the method remains statistically powerful .
we find that the proposed method has a comparable discovery rate and competitive predictive properties to these commonly used alternatives .
in this section we describe the proposed methodology by presenting the model and developing a computationally efficient estimation algorithm to fit the model . we describe our method within the potential outcomes framework of causal inference .
consider a simple random sample of @xmath1 units from population @xmath2 , with a possibly different target population of inference @xmath3 .
for example , the researchers and policy makers may wish to apply the gotv mobilization strategies and the job training program to a population , of which the study sample is not representative .
we consider a multi - valued treatment variable @xmath4 , which takes one of @xmath5 values from @xmath6 where @xmath7 means that unit @xmath8 is assigned to the control condition . in the gotv study
, we have a total of 193 treatment combinations ( @xmath9 ) , whereas the job training program corresponds to a binary treatment variable ( @xmath10 ) .
the potential outcome under treatment @xmath11 is denoted by @xmath12 , which has support @xmath13 .
thus , the observed outcome is given by @xmath14 and we define the causal effect of treatment @xmath15 for unit @xmath8 as @xmath16 . throughout
, we assume that there is no interference among units , there is a unique version of each treatment , each unit has nonzero probability of assignment to each treatment level , and the treatment level is independent of the potential outcomes , possibly conditional on observed covariates [ @xcite ] .
such assumptions are met in randomized experiments , which are the focus of this paper . under these assumptions
, we can identify the average treatment effect ( ate ) for each treatment @xmath15 , @xmath17 .
in observational studies , additional difficulty arises due to the possible existence of unmeasured confounders .
one commonly encountered problem related to treatment effect heterogeneity requires selecting the most effective treatment from a large number of alternatives using the causal effect estimates from a finite sample .
that is , we wish to identify the treatment condition @xmath15 such that @xmath18 is the largest , that is , @xmath19
. we may also be interested in identifying a subset of the treatments whose ates are positive .
when the number of treatments @xmath20 is large as in the gotv study , a simple strategy of subsetting the data and conducting a separate analysis for each treatment suffers from the lack of power and multiple testing problems .
another common challenge addressed in this paper is identifying groups of units for which a treatment is most beneficial ( or most harmful ) , as in the job training program study .
often , the number of available pre - treatment covariates , @xmath21 , is large , but the heterogeneous treatment effects can be characterized parsimoniously using a subset of these covariates , @xmath22 . this problem can be understood as identifying a sparse representation of the conditional average treatment effect ( cate ) , using only a subset of the covariates .
we denote the cate for a unit with covariate profile @xmath23 as @xmath24 , which can be estimated as the difference in predicted values under @xmath11 and @xmath7 with @xmath25 .
the sparsity in covariates greatly eases interpretation of this model .
we next turn to the description of the proposed model that combines optimal classification and variable selection to estimate treatment effect heterogeneity . for the remainder of the paper
, we focus on the case of binary outcomes , that is , @xmath26 .
however , the proposed model and algorithm can be extended easily to nonbinary outcomes by modifying the loss function .
we choose to model binary outcomes with the l2-svm to illustrate our proposed methodology because it presents one of the most difficult cases for implementing two separate lasso constraints .
as we discuss below , our method can be simplified when the outcome is nonbinary ( e.g. , continuous , counts , multinomial , hazard ) or the causal estimand of interest is characterized on a log - odds scale ( with a logistic loss ) . in particular , readily available software can be adapted to handle these cases [ @xcite ] . in modeling treatment effect heterogeneity
, we transform the observed binary outcome to @xmath27 .
we then relate the estimated outcome @xmath28 and the estimated latent variable @xmath29 , as @xmath30 @xmath31 is an @xmath32 dimensional vector of treatment effect heterogeneity variables , and @xmath33 is an @xmath34 dimensional vector containing the remaining covariates .
for example , when identifying the most efficacious treatment condition among many alternative treatments , @xmath31 would consist of @xmath20 indicator variables ( e.g. , different combinations of mobilization strategies ) , each of which is representing a different treatment condition .
in contrast , @xmath33 would include pre - treatment variables to be adjusted ( e.g. , age , party registration , turnout history ) .
similarly , when identifying groups of units most helped ( or harmed ) by a treatment , @xmath31 would include variables representing interactions between the treatment variable ( e.g. , the job training program ) and the pre - treatment covariates of interest ( e.g. , age , education , race , prior employment status and earnings ) . in this case
, @xmath33 would include all the main effects of the pre - treatment covariates .
thus , we separate the causal heterogeneity variables of interest from the rest of the variables .
we do not impose any restriction between main and interaction effects because some covariates may not predict the baseline outcome but do predict treatment effect heterogeneity .
finally , we choose the linear model because it allows for easy interpretation of interaction terms .
however , the researchers may also use the logistic or other link function within our framework . in estimating @xmath35 , we adapt the support vector machine ( svm ) classifier and place separate lasso constraints over each set of coefficients [ @xcite ] .
our model differs from the standard model by allowing @xmath36 and @xmath37 to have separate lasso constraints .
the model is motivated by the qualitative difference between the two parameters , and also by the fact that often causal heterogeneity variables have weaker predictive power than other variables .
specifically , we formulate the svm as a penalized squared hinge - loss objective function ( hereafter l2-svm ) where the hinge - loss is defined as @xmath38 [ @xcite ] .
we focus on the l2-svm , rather than the l1-svm , because it returns the standard difference - in - means estimate for the treatment effect in the absence of pre - treatment covariates . with two separate @xmath39 constraints to generate sparsity in the covariates , our estimates
are given by @xmath40 are pre - determined separate lasso penalty parameters for @xmath36 and @xmath37 , respectively , and @xmath41 is an optional sampling weight , which may be used when generalizing the results obtained from one sample to a target population .
our objective function is similar to several existing lasso variants but there exist important differences .
for example , the elastic net introduced by @xcite places the same set of covariates under both a lasso and ridge constraint to help reduce mis - selections among correlated covariates .
in addition , the group lasso introduced by @xcite groups different levels of the same factor together so that all levels of a factor are selected without sacrificing rotational invariance . in contrast , the proposed method places separate lasso constraints over the qualitatively distinct groups of variables .
the l2-svm offers two different means to estimate heterogeneous treatment effects .
first , we can predict the potential outcomes @xmath12 directly from the fitted model and estimate the conditional treatment effect ( cte ) as the difference between the predicted outcome under the treatment status @xmath15 and that under the control condition , that is , @xmath42 .
this quantity utilizes the fact that the l2-svm is an optimal classifier [ @xcite ] .
second , we can also estimate the cate . to do this
, we interpret the l2-svm as a truncated linear probability model over a subinterval of @xmath43 $ ] .
while it is known that the svm does not return explicit probability estimates [ @xcite ] , we follow work that transforms the values @xmath44 to approximate the underlying probability [ @xcite ] .
specifically , let @xmath45 denote the predicted value @xmath44 truncated at positive and negative one .
we estimate the cate as the difference in truncated values of the predicted outcome variables , that is , @xmath46 . while this cate estimate is not precisely a difference in probabilities , the method provides a useful approximation and returns sensible results that comport with probabilistic estimates of the cate . with an estimated cate for each covariate profile
, the cate for any covariate profile can be estimated by simply aggregating these estimates among corresponding observations .
our algorithm proceeds in three steps : the data are rescaled , the model is fitted for a given value of @xmath47 , and each fit is evaluated using a generalized cross - validation statistic .
_ rescaling the covariates_. lasso regularization requires rescaling covariates [ @xcite ] . following standard practice ,
we standardize all pre - treatment main effects by centering them around the mean and dividing them by standard deviation .
higher - order terms are recomputed using these standardized variables . for causal heterogeneity variables ,
we do not standardize them when they are indicator variables representing different treatments .
when they represent the interactions between a treatment indicator variable and pre - treatment covariates , we interact the ( unstandardized ) treatment indicator variable with the standardized pre - treatment variables .
_ fitting the model_. the l2-svm is fitted through a series of iterated lasso fits , based on the following two observations .
first , we note that for a given outcome @xmath48 , @xmath49 .
thus , the svm is a least squares problem on a subset of the data .
second , for a given value of @xmath50 , rescaling @xmath51 and @xmath52 allows the objective function to be written as a lasso problem , with a tuning parameter of 1 , as @xmath53 where @xmath54 , @xmath55 , @xmath56 , and @xmath57 .
this allows a fitting strategy via the efficient lasso algorithm of @xcite .
specifically , each iteration of our algorithm consists of fitting a model on the set of `` active '' observations @xmath58 and updating this set .
we now describe the proposed algorithm in greater detail .
first , the values of tuning parameters are selected , @xmath59 .
we then select the initial values of the coefficients and fitted values as follows , @xmath60 and @xmath61 for all @xmath8 .
this places all observations in the initial active set , so that @xmath62 .
next , for each iteration @xmath63 let @xmath64 denote the set of active observations , and @xmath65 represent the number of observations in this set . define @xmath66 and @xmath67 as the centered versions of @xmath68 and @xmath69 ( around their respective mean ) , respectively , using only the observations in the current active set @xmath70 .
similarly , we use @xmath71 to denote the centered value of @xmath72 using only the observations in @xmath70 .
then , at each iteration @xmath73 , the algorithm progresses in three steps .
we update the lasso coefficients as @xmath74 the fitted value is updated as @xmath75 where the intercept is @xmath76 .
these steps are repeated until @xmath35 converges .
work concurrent to ours has developed an algorithm for the regularization path of the l2-svm [ @xcite ] .
future work can combine our work and that of @xcite in order to estimate the whole `` regularization surface '' implied by a model with two constraints . _ selecting the optimal values of the tuning parameters_. we choose the optimal values of the tuning parameters , @xmath77 , based on a generalized cross - validation ( gcv ) statistic [ @xcite ] , so that the model fit is balanced against model dimensionality .
the number of nonzero elements of @xmath78 provides an unbiased degree - of - freedom estimate for the lasso [ @xcite ] .
our gcv statistic is defined over the observation in the active set @xmath79 , as follows : @xmath80 where the second equality follows from the fact that the observations outside of @xmath79 does not affect the model fit , that is , @xmath81 for @xmath82 . given this gcv statistic , we use an alternating line search to find the optimal values of the tuning parameters .
first , we fix @xmath83 at a large value , for example , @xmath84 , effectively setting all causal heterogeneity parameters to zero [ @xcite ] .
next , @xmath85 is evaluated along the set of widely spaced grids , for example , @xmath86 , with the value producing the smallest gcv statistic selected .
given the current estimate of @xmath85 , @xmath83 is evaluated along the set of widely spaced grids , for example , @xmath86 , and the @xmath85 that produces the smallest gcv statistic is selected .
we alternate in a line search between the two parameters to convergence .
after convergence , the radius is decreased based on these converged parameter values , and the precision is increased to the desired level , for example , @xmath87 .
the final estimates of coefficients are estimated given the converged value of @xmath47 .
the use of this gcv statistic is reasonable when exploring the degree to which the treatment effects are heterogeneous .
however , if the goal is to derive the optimal treatment rule , the researchers may wish to directly target a particular measure of the performance of the learned policy .
for example , following @xcite and @xcite , we could use the largest average treatment effect as the statistic ( known as `` value statistic '' ) for cross - validation .
in addition to its computational burden , one practical difficulty of cross - validation based on the value statistic is that when the total number of causal heterogeneity variables is large and the sample size is relatively small as in our applications , we may not have many observations in the test sample that actually received the same treatment as the one prescribed by the optimal treatment rule .
in addition , the training sample may have empty cells so that they do not predict treatment effects for some individuals in the test set .
this makes it difficult to apply this procedure in some situations .
the use of the gcv statistic is also computationally efficient as it avoids cross - validation .
comparing computation times across competitors is difficult , as they can vary dramatically depending on the number of cross - validating folds ( boosting , lasso ) , iterations ( boosting ) , the number of mcmc draws and number of trees ( bart ) , and desired precision in estimating the tuning parameters ( lasso , svm ) .
the general pattern from our simulations is that the computational time for the proposed method is significantly greater than the bayesian glm , tree , and the cross - validated logistic lasso with a single constraint .
it is comparable to bart and significantly less than cross - validated boosting .
finally , in a recent paper , @xcite propose a method that is related to ours .
there are several important differences between the two methods .
first , we are primarily interested in feature selection using lasso penalties , whereas @xcite focus on prediction using an @xmath88 penalty . while we use a simple parametric model with a large number of features , @xcite place their method within a nonparametric , reproducing kernel framework . in kernelizing the covariates
, they achieve better prediction but at the cost of difficulty in interpreting precisely which features are driving the treatment rule .
second , we use two separate lasso penalties for causal heterogeneity variables and pre - treatment covariates , whereas @xcite do not make this distinction .
third , our tuning parameter is a gcv statistic , which eliminates the computational burden of cross - validation as used by @xcite
in this section we apply the proposed method to two well - known field experiments in the social sciences .
first , we analyze the get - out - the - vote ( gotv ) field experiment where 69 mobilization techniques were randomly administered to registered new haven voters in the 1998 election [ @xcite ] .
it is known in the gotv literature that there is substantial interference among voters in the same household [ @xcite ] .
thus , to avoid the problem of possible interference between voters , we focus on @xmath89 voters in single voter households where @xmath90 voters of them belong to the control group and hence did not receive any of these gotv messages . addressing this interference issue fully requires an alternative experimental design where the treatment conditions correspond to different number of voters within the same household who receive the gotv message ( our method is still applicable to the data from this experimental design because it can handle multi - valued treatments ) .
for the purpose of illustration , we also ignore the implementation problems documented in @xcite and analyze the most recent data set . in our specification , the causal heterogeneity variables @xmath31 include the binary indicator variables of 192 treatment combinations , that is , @xmath91 .
we include a set of noncausal variables @xmath33 , which consist of the main effect terms of four pre - treatment covariates ( age , member of a majority party , voted in 1996 , abstained in 1996 ) , their two - way interaction terms , and the square of the age variable , that is , @xmath92 .
@lcccd2.2@ & + & + * visit * & * phone * & * mailings * & * appeal type * & + yes & no & 0 & any & 3.06 + yes & no & 3 & civic & 2.64 + yes & no & 3 & close & 2.31 + yes & civic , close & 0 & close , neighbor & 2.04 + yes & no & 12 & close & 1.60 + yes & no & 3 & civic & 1.50 + yes & civic , close & 13 & civic , close & 1.46 + yes & none , civic / neighbor , neighbor & 12 & neighbor & 1.46 + yes & none , civic & @xmath93 & civic , neighbor & 1.46 + no & no & 3 & civic & 1.17 + no & no & 2 & civic & 1.14 + no & close & @xmath94 & close & 0.81 + yes & civic / blood , neighbor , neighbor / civic & 3 & civic , neighbor & 0.80 + no & close & 2 & close & 0.74 + no & no & 3 & close & 0.70 + yes & civic / blood , civic & 13 & civic & 0.53 + no & no & 3&neighbor & 0.04 + yes & civic / blood & 13 & civic & -0.64 + no & neighbor / civic & 3 & neighbor&-0.65 + no & civic / blood & @xmath95 & civic & -0.91 + no & civic / blood & 2 & civic & -0.99 + no & civic / blood , civic & @xmath95 & civic & -2.07 + no & no & 2 & close&-2.08 + no & civic / blood & 2 & civic & -2.14 + no & civic , civic / blood , neighbor & 1 & civic , neighbor & -2.60 + no & neighbor & 2 & neighbor & -2.67 + no & neighbor & 3 & neighbor & -3.24 + no & civic & 01 & neighbor & -3.49 + no & civic & 2 & neighbor & -3.56 + no&civic & 3 & neighbor & -4.12 + of the 192 possible treatment effect combinations , 15 effects are estimated as nonzero ( see table [ ggcoefs ] in ) .
as these coefficients range from main effects to four - way interactions , they are difficult to interpret . instead , we present the estimated treatment effect for every treatment combination in table [ ggeffects ] . some of our results are consistent with the prior analysis .
first , canvassing in person is the most effective gotv technique .
this result can be obtained even from a simple use of the difference - in - means estimator ( estimated 2.69 percentage point with @xmath15-statistic of 2.68 ) .
second , every mobilization strategy that consists of a phone call and no personal visit is estimated with a nonpositive sign , suggesting that the marginal effect of a phone call is either zero or slightly negative .
most prominently , phone messages with a neighborhood appeal or civic appeal decrease turnout .
the proposed method also yields finer findings than the existing analyses .
for example , since personal canvassing is expensive , campaigns may be interested in the most effective treatment that does not include canvassing .
we find that three mailings with a civic responsibility message and no phone calls or personal visits increase turnout marginally by 1.17 percentage point .
this result is similar to the one independently obtained in another study [ @xcite ] .
three mailings with other appeals produce smaller effects ( 1.17 and 0.04 percentage point increase , resp . ) .
finally , the proposed method , upon considering all possible treatments , produces clear prescriptions .
first , in the presence of canvassing , any additional treatment ( phone call , mailing ) will lessen the canvassing s effectiveness .
if voters are canvassed , they should not receive additional treatments .
second , if voters are not canvassed , they should be targeted with three mailings with a civic duty appeal .
any other treatment combination will be less cost - effective , and may even suppress turnout .
next , we apply the proposed methodology to the national supported work ( nsw ) program .
our analysis focuses upon the subset of these individuals previously used by other researchers [ @xcite ] where the ( randomly selected ) treatment and control groups consist of 297 and 425 such workers , respectively .
we consider two aspects of treatment effect heterogeneity .
first , we seek to identify the groups of workers for whom the training program is beneficial .
the program was administered to a heterogeneous group of workers and , hence , it is of interest to investigate whether the treatment effect varies as a function of individual characteristics .
second , we show how to generalize the results based on this experiment to a target population . such an analysis is important for policy makers who wish to use experimental results to decide whether and how to implement this program in a target population . for illustration , we generalize the experimental results to the 1978 panel study of income dynamics ( psid ) , which oversamples low - income individuals . within this psid sample
, we focus on 253 workers who had been unemployed at some point in the previous year to avoid severe extrapolation .
this subsample is labeled psid-2 in @xcite .
the differences across the two samples are substantial .
the psid respondents are on average older ( @xmath96 vs. @xmath97 years old ) and more likely to be married ( @xmath98 vs. @xmath99 ) and have a college degree ( @xmath100 vs. @xmath101 ) than nsw participants .
the proportion of blacks in the psid sample ( @xmath102 ) is much less than in the nsw sample ( @xmath103 ) .
in addition , on average , psid respondents earned more income ( $ 7600 ) than nsw participants ( $ 3000 ) .
all differences , except for proportion hispanic , are statistically significant at the 5% level . in our model ,
the matrix of noncausal variables , @xmath52 , consists of 45 pre - treatment covariates .
these include the main effects of age , years of education , and the log of one plus 1975 earnings , as well as binary indicators for race , marriage status , college degree , and whether the individual was unemployed in 1975 .
we also use square terms for age and years of education , and every possible two - way interactions among the pre - treatment covariates are included .
the matrix of causal heterogeneity variables @xmath51 includes the binary treatment and interactions between this treatment variable and each of the 39 pre - treatment covariates .
this yields @xmath104 and @xmath105 .
using this specification , we first fit the model to the nsw sample to identify the subpopulations of workers for whom the job training program is beneficial .
second , we generalize these results to the psid sample and estimate the ate and cate for these low - income workers .
unfortunately , sampling weights are not available in the original data and , hence , for the purpose of illustration , we construct them by fitting a bart model , using @xmath52 as predictors [ @xcite ] .
we then take the inverse estimated probability of being in the nsw sample as the weights for the proposed method [ @xcite ] . to facilitate comparison between the unweighted and weighted models , we standardize the weights to have a mean equal to one . a weight greater than one signifies an observation that is weighted more highly in the psid model than in the nsw model .
this allows us to assess the extent to which differences in identified heterogeneous effects reflect underlying differences in the covariate distributions between the nsw and psid samples . after fitting the model to the unweighted and weighted nsw samples
, the cate is estimated using the covariate value of each observation , that is , @xmath106 .
the sample average of these cates yields an ate estimate of @xmath107 and @xmath108 percentage points for the nsw and psid samples , respectively .
nonzero coefficients from the fitted models are shown in table [ lalondecoefs ] of the .
as with the previous example , interpreting high order interactions is difficult . thus , we present the groups of workers who are predicted to experience the ten highest and lowest treatment effects of the job training program in the nsw ( table [ nswhighlow ] ) and psid sample ( table [ psidhighlow ] ) .
the groups most helped , and hurt , by the treatment were identified by matching the observations in these tables to the nonzero coefficients . across both tables , unemployed hispanics and
highly educated , low - earning non - hispanics are predicted to benefit from the program .
similarly , workers who were older and employed and whites with a high school degree are identified as those who are negatively affected by the program .
weights marked with asterisks in each table indicate heterogeneous effects that are not identified in the other table .
for example , unemployed blacks with some college are identified as beneficiaries only in table [ nswhighlow ] , while married whites with no high school degree only appear in table [ psidhighlow ] .
this difference is explained by the fact that unemployed blacks with some college make up 2.7% of the nsw sample but only 0.4% of the psid sample .
similarly , married whites with no high school degree make up 15.8% of the psid sample and are identified in table [ psidhighlow ] , but only make up 0.1% of the nsw sample and are not identified in table [ nswhighlow ] .
indeed , when generalizing the results to a different population , large groups in that population are more likely to be selected for heterogeneous treatment effects .
weighting allows us to efficiently estimate heterogeneous treatment effects in a target population .
in this section we conduct two simulation studies to evaluate the performance of the proposed method relative to the commonly used methods : bart ( r package ` bayestree ` ) , bayesian logistic regression with a noninformative prior ( r package ` arm ` ) , conditional inference trees [ @xcite ; r package ` party ` ] , boosting with the number of iterations selected by cross - validation ( as implemented in r package ` ada ` ) , and logistic regression with a single lasso constraint and cross - validation on the `` value '' statistic [ @xcite ; r package ` glmnet ` ] .
the first set of simulations corresponds to the situation where the goal is to select a set of the most effective treatments among many alternatives .
the second set considers the case where we wish to identify a subpopulation of units for which a treatment is most effective . in both cases ,
we assume that the treatment @xmath4 is independent of the observed pre - treatment covariates @xmath109 .
the logistic lasso method is only applied to the second set of simulations for the reason mentioned in section [ subsecalgorithm ] . finally ,
for each scenario , we examine 4 different sample sizes between @xmath110 and @xmath111 and run @xmath112 simulations .
we conduct simulations for selecting a set of the best treatments among a large number of available treatments .
we use two settings , one with the correct model specification and the other with misspecified models , where un - modeled nonlinear terms are added to the data generating process . in the simulations with correct model specification , we have one control condition , 49 distinct treatment conditions , and 3 pre - treatment covariates .
that is , @xmath31 consists of 49 treatment indicator variables and @xmath33 is a vector of 3 pre - treatment covariates plus an intercept , that is , @xmath113 and @xmath114 . among 49 treatments , 3 of them
have substantive effects ; the ate is approximately equal to @xmath115 , @xmath116 , and @xmath117 percentage points , respectively .
the remaining 46 treatment indicator variables have nonzero but negligible effects , with the average effect sizes ranging within @xmath118 percentage point .
in contrast , all pre - treatment covariates are assumed to have substantial predictive power .
we independently sample the pre - treatment covariates from a multivariate normal distribution with mean zero and a randomly generated covariance matrix .
specifically , an @xmath119 matrix , @xmath120 $ ] , was generated with @xmath121 and the covariance matrix is given by @xmath122 .
the design matrix for the 49 treatment variables is orthogonal and balanced .
the true values of the coefficients are set as @xmath123 and @xmath124 , where @xmath125 denotes 47 remaining coefficients drawn from a uniform distribution on @xmath126 $ ] .
finally , the outcome variable @xmath127 is sampled according to the following model ; @xmath128 with @xmath129 selected such that the magnitude of the ates roughly equals the values specified above .
for the simulations with an incorrectly specified model , we include unmodeled nonlinear terms based on the pre - treatment covariates in the data generating process .
specifically , @xmath33 now includes the interaction term between the first and second pre - treatment covariates and the square term of the third pre - treatment covariate as well as the main effect term for each of the three covariates .
these higher - order terms are used to generate the data , but not included as covariates in fitting any model . as before , the outcome variable is generated after an affine transformation in order to keep the size of the ates approximately equal to the pre - specified levels given above .
figure [ figfirstsims ] summarizes the results in terms of false discovery rate ( fdr ) and discovery rate ( dr ) separately for the largest and substantive effects .
we define discovery as estimating the largest effect ( three largest effects ) as the largest effect ( nonzero effects ) with the correct sign .
similarly , false discovery occurs when the largest effect is not correctly discovered _ and _ at least one coefficient is estimated to be nonzero .
fdr may not equal one minus dr because the former is based only on the simulations where at least one coefficient is estimated to be nonzero .
the first row presents fdr for the largest effect whereas the second row presents its dr .
similarly , the third row plots the fdr for the three largest effects while the fourth row presents their dr .
note that fewer than three nonzero effects may be estimated.=-1 the results show that across simulations the proposed method ( ` svm ` ; solid lines ) has a smaller fdr while its dr is competitive with other methods .
the comparison with bart reveals a key feature of our method .
the proposed method dominates bart in fdr regardless of model specification .
the largest estimated effect from bart identifies the largest effect slightly more frequently , but at the cost of a higher fdr . despite its low fdr ,
our method maintains a competitive dr . for many methods ,
model misspecification increases fdr and reduces dr .
for this reason , we recommend erring in favor of including too many rather than too few pre - treatment covariates in the model . unlike three of its competitors , boosting , conditional inference trees , and bayesian glm
, the performance of the proposed method improves as the sample size increases . boosting and trees
both attain an fdr and dr of zero as the sample size grows .
the tree focuses in on the largest effects , not identifying any small effects as the sample size grows .
this may be due to the fact that trees do not converge asymptotically to the true conditional mean function unless the underlying function is piecewise constant ( though they converge to the minimal risk ) [ see , e.g. , @xcite ] .
the boosting algorithm uses trees as base learners , which may be leading to the deteriorating performance in identifying small effects .
the performance of bayesian glm also declines with increasing sample size , because we are not considering uncertainty in the posterior mean estimates . to address this issue
, one must use some @xmath130-value based regularization , such as using a @xmath130-value threshold of @xmath131 ( see figure [ figusvsbayes ] in the next set of simulations for illustration ) . in the second set of simulations , we consider the problem of identifying groups of units for which a treatment is beneficial ( or harmful ) . here
, we are interested in identifying interactions between a treatment and observed pre - treatment covariates .
the key difference between this simulation and the previous one is that in the current setup causal heterogeneity variables ( treatment - covariate interactions ) may be correlated with each other as well as other noncausal variables .
the previous simulation setting assumes that causal heterogeneity variables ( treatment indicators ) are independent of each other and other variables . in this simulation
, we also include a comparison with the logistic regression with a single lasso constraint and the maximal ten - fold cross - validation on the `` value '' statistic [ @xcite ] .
this statistic is the expected benefit from a particular treatment rule [ see also @xcite ] . note that in the previous simulation , due to a large number of treatments , cross - validation on this statistic is not feasible .
-value less than @xmath132 ) . here
, discovery is defined as estimating the largest effect ( four largest effects ) as the largest effect ( nonzero effects ) with the correct sign .
the top and bottom plots in the first ( second ) column present the fdr and dr for the largest effect ( three largest effects ) , respectively . ] in the current simulation , we have a single treatment condition , that is , @xmath133 , and 20 pre - treatment covariates @xmath109 .
the pre - treatment covariates are all based on the multivariate normal distribution with mean zero and a random variance - covariance matrix as in the previous simulation study , with five covariates then discretized using 0.5 as a threshold .
causal heterogeneity variables @xmath31 consist of 20 treatment - covariate interactions plus the main effect for the treatment indicator ( @xmath134 ) , while @xmath33 is composed of the main effects for the pre - treatment covariates ( @xmath135 ) . given this setup ,
we generate the outcome variable @xmath136 in the same way as in section [ subsecbest ] according to the linear probability model .
there are 4 pre - treatment covariates that interact with the treatment in a systematic manner .
as before , we apply an affine transformation so that an observation whose values for these two covariates are one standard deviation above the mean has the cate of roughly @xmath137 and @xmath138 percentage points .
that is , we set @xmath139 and @xmath140 where the @xmath125 denotes uniform draws from @xmath126 $ ] .
figure [ figusvsbayes ] compares the fdr and dr for our proposed method ( ` svm ` ; solid lines ) with those for the logistic lasso ( ` lasso ` ) and bayesian logistic regression ( ` glm ` ; dotted and dashed lines ) .
for the bayesian glm , we consider two rules : one based on posterior means of coefficients ( dashed lines ) and the other selecting coefficients with @xmath130-values below @xmath132 ( dotted lines ) . unlike the simulations given in section [ subsecbest ] , neither bart , boosting , nor conditional inference trees provide a simple rule for variable selection in this setting and hence no results are reported .
the interpretation of these plots is identical to that of the plots in figure [ figfirstsims ] . in the left column , the top ( bottom ) plot presents fdr ( dr ) for the largest effect , whereas that of the right column presents fdr ( dr ) for the four largest effects .
when compared with the bayesian glm , the proposed method has a lower fdr for both largest and four largest estimated effects .
the @xmath130-value thresholding improves the bayesian glm , and yet the proposed method maintains a much lower fdr and comparable dr .
relative to the lasso , the proposed method is not as effective in considering the largest estimated effect except that it has a lower fdr when the sample size is small .
however , when considering the four largest estimated effects , the proposed method maintains a lower fdr than the lasso , and a comparable dr .
this result is consistent with the fact that the value statistic targets the largest treatment effect while the gcv statistic corresponds to the overall fit.=-1 to further evaluate our method , we consider a situation where each method is applied to a sample and then used to generate a treatment rule for each individual in another sample . for each method ,
a payoff , characterized by the net number of people in the new sample who are assigned to treatment and are in fact helped by the treatment , is calculated .
to represent a budget constraint faced by most researchers , we specify the total number of individuals who can receive the treatment and vary this number within the simulation study.=-1 specifically , after fitting each model to an initial sample , we draw another simple random sample of 2000 observations from the same data generating process . using the result from each method , we calculate the predicted cate for each observation of the new sample , @xmath141 , and give the treatment to those with highest predicted cates until the number of treated observations reaches the pre - specified limit . finally , a payoff of the form @xmath142 is calculated for all treated observations of the new sample where @xmath143 is the true cate .
this produces a payoff of @xmath144 if a treated observation is actually helped by the treatment , @xmath145 if the observation is harmed , and @xmath146 for untreated observations . as a baseline , we compare each method to the `` oracle '' treatment rule , @xmath147 , which administers the treatment only when helpful . we have also considered an alternative payoff of the form @xmath148 , representing how much ( rather than whether ) the treatment helps or harms .
the results were qualitatively similar to those presented here.=-1 @ld4.0d3.0d3.0d3.0@ & + & + & & & & + & -2 & 11 & 22 & 42 + & -19 & -4 & 8 & 21 + & -18 & 2 & 15 & 28 + & -20 & -7 & 7 & 34 + & -1 & 10 & 18 & 40 + & 2 & 2 & 2 & 5 + & -123 & -121 & -121 & -116 +
the results from the simulation are presented in table [ tabpayoff ] .
the table presents a comparison of payoffs , by method , as a percentage of the optimal oracle rule , which is considered as @xmath149 .
the bottom row presents the outcome if every observation were treated , indicating that in this simulation the average treatment effect is negative but there exists a subgroup for which treatment is beneficial . the proposed method ( ` svm ` )
narrowly dominates boosting ( ` boost ` ) , and both the proposed method and boosting noticeably outperform all other competitors , except conditional inference trees ( ` tree ` ) at sample size 250 . at larger sample sizes ,
however , the tree severely underfits .
while the proposed method and boosting perform similarly by a predictive criterion , boosting does not return an interpretable model .
we also find that ` svm ` outperforms ` lasso ` , which is consistent with the fact that the gcv statistic targets the overall performance while the value statistic focuses on the largest treatment effect . if administering the treatment is costless , the proposed method generates the most beneficial treatment rule among its competitors .
figure [ figposneg ] presents the results across methods and sample sizes in the presence of a budget constraint .
the left column shows the proportion of treated units that actually benefit from the treatment for each observation considered for the treatment in the order of predicted cate ( the horizontal axis ) .
the oracle identifies those who certainly benefit from the treatment and treats them first .
the middle column shows the proportion of treated units that are hurt by the treatment . here , the oracle never hurts observations and hence is represented by the horizontal line at zero .
the right column presents the net benefit by treatment rule , which can be calculated as the difference between the positive ( left column ) and negative ( middle column ) effects .
each row presents a different sample size to which each method is applied .
the figure shows that when the sample size is small , the proposed method assigns fewer observations a harmful treatment , relative to its competitors . for moderate and large sample sizes
, the proposed method dominates its competitors in both identifying a group that would benefit from the treatment and avoiding treating those who would be hurt .
this can be seen from the plots in the middle column where the result based on the proposed method ( ` svm ` ; solid thick lines ) stays close to the horizontal zero line when compared to other methods .
similarly , in the right column , the results based on the proposed method stay above other methods .
when these lines go below zero , it implies that a majority of treated observations would be harmed by the treatment .
the disadvantage of the proposed method is its conservativeness .
this can be seen in the left column where at the beginning of the percentile the solid thick line is below its competitors for small sample sizes .
this difference vanishes as the sample size increases , with the proposed method outperforming its competitors . in sum , when used to predict a treatment rule for out - of - sample observations , the proposed method makes fewer harmful prescriptions and often yields a larger net benefit than its competitors .
estimation of heterogeneous treatment effects plays an essential role in scientific research and policy making . in particular , researchers often wish to select the most efficacious treatments from a large number of possible treatments and to identify individuals who benefit most ( or are harmed ) by treatments .
estimation of treatment effect heterogeneity is also important when generalizing experimental results to a target population of interest .
the key insight of this paper is to formulate the identification of heterogeneous treatment effects as a variable selection problem . within this framework ,
we develop a support vector machine with two separate sparsity constraints , one for a set of treatment effect heterogeneity parameters of interest and the other for observed pre - treatment effect parameters .
this setup addresses the fact that in many applications , pre - treatment covariates are much more powerful predictors than treatment variables of interest or their interactions with covariates . in addition , unlike the existing techniques such as boosting and bart , the proposed method yields a parsimonious model that is easy to interpret .
our simulation studies show that the proposed method has low false discovery rates while maintaining competitive discovery rates .
the simulation study also shows that the use of our gcv statistic is appropriate when exploring the treatment effect heterogeneity rather than identifying the single optimal treatment rule .
@ld3.2d3.2@ & & + treatment intercept & 6.92 & 6.67 + _ main effects _ + age & 0.00 & -0.83 + married & 1.32 & 3.39 + white & 0.00 & 0.10 + _ squared terms _
+ age@xmath150 & -0.03 & -0.09 + education@xmath150 & 0.89 & 0.86 + _ interaction terms _ + no hs degree , unemployed in 1975 & -1.06 & 0.00 + white , married & 0.00 & 26.16 + white , no hs degree & 25.35 & 30.65 + hispanic , logged 1975 earnings & -49.36 & -62.15 + black , logged 1975 earnings & 8.29 & 0.00 + white , education & 0.00 & -1.41 + married , education & 4.90 & 12.11 + married , logged 1975 earnings & 0.00 & 5.72 + education , unemployed in 1975 & 7.52 & 9.59 + age , education & 0.00 & -0.47 + age , black & -0.56 & 0.00 + age , hispanic & 0.00 & 0.34 + age , unemployed in 1975 & 3.30 & 4.79 + a number of extensions of the method developed in this paper are possible .
for example , we can accommodate other types of outcome variables by considering different loss functions . instead of the gcv statistic we use ,
alternative criteria such as aic or bic statistics as well as more targeted quantities such as the average treatment effect for the target population can be employed .
while we use lasso constraints , researchers may prefer alternative penalty functions such as the scad or adaptive lasso penalty .
furthermore , although not directly examined in this paper , the proposed method can be extended to the situation where the goal is to choose the best treatment for each individual from multiple alternative treatments . finally , it is of interest to consider how the proposed method can be applied to observational data [ e.g. , see @xcite who develop a doubly robust estimator for optimal treatment regimes ] and longitudinal data settings where the derivation of optimal dynamic treatment regimes is a frequent goal [ e.g. , @xcite ] .
the development of such methods helps applied researchers avoid the use of ad hoc subgroup analysis and identify treatment effect heterogeneity in a statistically principled manner .
an earlier version of this paper was circulated under the title of `` identifying treatment effect heterogeneity through optimal classification and variable selection '' and received the tom ten have memorial award at the 2011 atlantic causal inference conference .
we thank charles elkan , jake bowers , kentaro fukumoto , holger kern , michael rosenbaum , and sherry zaks for useful comments . the editor , associate editor , and two anonymous reviewers provided useful advice . | when evaluating the efficacy of social programs and medical treatments using randomized experiments , the estimated overall average causal effect alone is often of limited value and the researchers must investigate when the treatments do and do not work .
indeed , the estimation of treatment effect heterogeneity plays an essential role in ( 1 ) selecting the most effective treatment from a large number of available treatments , ( 2 ) ascertaining subpopulations for which a treatment is effective or harmful , ( 3 ) designing individualized optimal treatment regimes , ( 4 ) testing for the existence or lack of heterogeneous treatment effects , and ( 5 ) generalizing causal effect estimates obtained from an experimental sample to a target population . in this paper , we formulate the estimation of heterogeneous treatment effects as a variable selection problem .
we propose a method that adapts the support vector machine classifier by placing separate sparsity constraints over the pre - treatment parameters and causal heterogeneity parameters of interest .
the proposed method is motivated by and applied to two well - known randomized evaluation studies in the social sciences .
our method selects the most effective voter mobilization strategies from a large number of alternative strategies , and it also identifies the characteristics of workers who greatly benefit from ( or are negatively affected by ) a job training program . in our simulation studies ,
we find that the proposed method often outperforms some commonly used alternatives . |
debris disks older than @xmath010 myr containing dust at temperatures @xmath1100 k are extremely rare @xcite .
when warm dust does appear , it is likely to be from a stochastic event , perhaps akin to our own solar system s `` late heavy bombardment , '' at about 600 myr after formation .
late episodes of dust production may signal the presence of a planetary system undergoing architectural reconfiguration @xcite .
bd+20 307 is one of the few examples of a non - young star with hot debris @xcite .
it has a ring of dust at @xmath00.5 au @xcite . in order to understand the implications of the large amount of close - in dust ,
it would help greatly to know the age of the star .
@xcite used the li 6707equivalent width and chromospheric activity to suggest an age of @xmath0300 myr .
new observations reported here show that bd + 20 307 is actually a spectroscopic binary .
i reexamine the evidence for the age of the star .
the magellan inamori kyocera echelle ( mike ) spectrograph on the clay ( magellan ii ) telescope was used to observe bd + 20 307 on three consecutive nights 2007 october 24 - 26 ( ut ) .
the 0.35 wide @xmath2 5 long slit provided a resolution of about 55,000 at wavelengths 3400 7250 .
seeing was @xmath00.5 on the first two nights and 0.81.2 on the third .
on all three nights , data were obtained with an iodine cell in place to facilitate looking for planets around the star , and on the first night an observation without the iodine cell was also obtained .
i do not use the iodine lines for the radial velocity analyses that follow .
an observing log is given in table [ tab_observlog ] .
llc 2004 aug 24 & 08:28:20 & 700 + 2007 oct 24 & 04:35:20 & 600 + 2007 oct 25 & 04:45:43 & 600 + 2007 oct 26 & 04:12:33 & 600 + the spectra were flattened , extracted and wavelength calibrated using the mike pipeline written by d. kelson with methods described in @xcite .
the two observations from the first night were averaged . the signal - to - noise ratio ( s / n ) per pixel was @xmath1100 for wavelengths @xmath14000 on the first two nights , except in the region of maximal iodine absorption around 5000 .
the s / n was about 50% worse on the third night due to the worse seeing .
bd + 20 307 was also observed on 2004 august 24 with the echelle spectrograph on the 2.5 m du pont telescope at las campanas observatory .
these data cover wavelengths @xmath040009000 and have a resolution of about 25,000 and s / n of 30100 .
the data were extracted and calibrated using standard iraf tasks .
heliocentric and barycentric velocities were calculated with the rvsao package in iraf .
two sets of lines are clearly visible in all three nights of mike data . to obtain the velocities of the double - lined spectroscopic binary , cross - correlations with a synthetic spectrum with effective temperature 6000 k and
log(g)=5.0 were performed .
this spectrum was generated using r. o. gray s spectrum code and line listand a castelli - kurucz model atmosphere with solar metalicity .
the xcsao package in iraf was used to compute the cross - correlations , and the two peaks were fit with parabolas in idl .
the uncertainty in the velocities was computed as the standard deviation of the velocities in the 40 different orders used .
results are reported in table [ tab_binaryvel ] and shown in figure [ fig_binaryvel ] .
on all three nights , the primary star produced a higher cross correlation peak .
llll primary & -1.63 @xmath3 0.46 & -51.34 @xmath3 0.62 & 4.92 @xmath3 0.79 + secondary & -17.49 @xmath3 0.85 & 38.03 @xmath3 0.64 & -21.24 @xmath3 0.67 + the same cross - correlation was done for the lower resolution du pont spectrum .
a double - peaked cross - correlation appears only for the lowest two orders ( 40004150 ) .
i do not consider this detection of the binary reliable , and do not include these rv data in the analyses which follow . on all nights a li
6707 line was detected from the primary star ( figure [ fig_lithium ] ) .
the equivalent widths were computed using direct integration over the lines relative to the combined continua from the two stars .
uncertainties from the pipeline reduction were used to give the statistical uncertainty .
an additional systematic uncertainty was estimated by choosing different methods of finding the continuum and recomputing the equivalent widths . a 3@xmath4 upper limit on the secondary s li line
was placed using the data from 2007 october 25 , when the two stars were separated by 89 km @xmath5 ( 2 ) .
these equivalent widths are given in the first two columns of table [ tab_li_eqw ] .
lllll 2004 aug 24 & 41 @xmath3 3 & & 70 @xmath3 5 & + 2007 oct 24 & 35 @xmath3 2 & & 60 @xmath3 4 & + 2007 oct 25 & 34 @xmath3 2 & @xmath66 ( 3@xmath4 ) & 58 @xmath3 4 & @xmath614 ( 3@xmath4 ) + 2007 oct 26 & 33 @xmath3 2 & & 56 @xmath3 4 & + the continuum normalized spectrum on 2007 october 25 , which has the maximum separation of the two stars , was fit with a combination of synthetic spectra calculated with spectrum from castelli - kurucz model atmospheres with solar metalicity .
free parameters were two effective temperatures , a single @xmath7 , a single log(g ) , and two normalizations .
one blue and one red region of the spectrum were fit
40994360 and 6282 - 6549 .
the best fit in both cases had t@xmath8=6500 k and 6250 k for the primary and secondary stars , respectively and log(g)=5.0 .
contours of chi - square indicate the uncertainty is within 250 k ( the gridding of the models ) in t@xmath9 .
the lines are measurably broader than the thar calibration lamp lines at the same wavelengths .
the best fit models had @xmath10 km s@xmath11 . to compute the stars li equivalent widths relative to their own stellar continua
, the flux ratio of the two stars must be obtained at 6707 .
synthetic spectra were fit as above to the region at 6645 6835 .
the best fit flux ratio was 1.397 @xmath3 0.007 .
however , this statistical uncertainty probably underestimates the systematics from how the models are calculated . applying this flux ratio to the measured equivalent widths makes the primary and secondary equivalent widths increase by factors of 1.709 and 2.387 , respectively .
the computed equivalent widths are given in the last two columns of table [ tab_li_eqw ] .
the spectra were examined for evidence of chromospheric activity .
both stars show weak central reversals on their ca h and k lines ( figure [ fig_calcium ] ) that change in velocity along with the stars .
the balmer lines show no central reversals . .
the velocities were fit for the stellar mass ratio ( secondary / primary ) yielding 0.91 @xmath3 0.02 and the line of sight radial velocity yielding @xmath12 .
although the small number of observations prohibit the calculation of the full binary orbit , i tested whether a circular orbit could fit the velocities .
the best fit produced reasonable residuals ( @xmath61 km @xmath13 ) for an orbit with a period of 3.448 days , @xmath14=-8.4 km @xmath13 , consistent with the fit above , and mass ratio 0.93 , again consistent with the fit above .
this orbit , shown in figure [ fig_binaryvel ] yields @xmath15 for each star 0.13 and 0.12 for the primary and secondary stars respectively .
i estimate the true masses of the stars as 1.3 and 1.2 m@xmath16 based on the effective temperatures ( 6500 and 6250 k ) obtained in the spectral fit described in 3 and the mk spectral - type calibration given in @xcite ; these masses are uncertain by @xmath00.1 m@xmath16 . calculating the inclination from these masses
yield 27.8 @xmath3 0.8@xmath17 .
the fit , assuming the inclination of 28@xmath17 , gives semi - major axes for the binary orbits of 4.4 and 4.7 @xmath2 10@xmath18 km .
the stars should have radii of 1.31.6 r@xmath16 , so their separation is about 9 stellar radii .
of primary interest for understanding the dusty debris disk around bd+20 307 is the age of the system .
although the orbit is not well measured , a reasonable fit is obtained with e=0 .
binaries of any age with periods less than @xmath08 days are found in circular orbits @xcite .
so , e=0 is consistent with the 3.4 day period and is no indication of the age of the system .
the formation of close binaries may depend on the presence of another star in the system ; @xcite find that 80% of binaries with periods @xmath67 days are actually in higher order multiple systems .
it is unknown whether bd+20 307 has another , wider , component , but this should be investigated .
taking @xmath19 from the spectral modeling , the rotational velocity of the stars is 21 km s@xmath11 . a standard radius for a t@xmath20=6250 k star would be 1.3 r@xmath16 , but solar - like stars with active chromospheres have somewhat larger radii than their non - active counterparts @xcite .
stellar radii of 1.3 1.6 r@xmath16 correspond to rotational periods of 3 3.7 d. the close agreement between the rotation period and the orbital period suggests that the stars are tidally synchronized .
the timescale for synchronization for stars of mass similar to the sun is given by @xcite , and for equal mass components is @xmath21 where p is the orbital period .
given a period of 3.5 d , t@xmath22 is 100 myr .
this provides a lower limit to the age of the system
. tidal coupling in a close binary can give rise to enhanced magnetic activity @xcite .
so , the central reversals observed in both stars ca h and k lines need not arise from youth but could be from their interaction . for stars with deep convective zones
, there is a tight correlation between lithium abundance with age and spectral type , but for stars with t@xmath20 @xmath0 6500 k , there is little li depletion by the age of the hyades @xcite .
a typical li equivalent width for a star with b - v of 0.5 in the hyades is 90 m , 1.5 times that for bd+20 307 .
furthermore , tidally locked systems in the hyades show inhibited li depletion @xcite .
the situation for older stars is more complex .
stars even hotter than the sun apparently do deplete li after reaching the main sequence .
@xcite summarize the li abundances in open clusters of various ages , including the wide spread in abundances among members of old ( @xmath11 gyr ) clusters .
hot stars can show substantial depletions .
for example in the 1 gyr old cluster ngc 3960 , stars with @xmath23 ( t@xmath24 6250 k ) show a range ew(li ) of @xmath610 to 80 m @xcite .
initial metalicity can have a large effect on main sequence li abundance , but given their tight orbit , the two stars in bd+20 307 presumably formed from the same molecular cloud material at the same metalicity .
@xcite found that @xmath020% of wide field binaries showed disparate li .
stars may also be polluted by li and other elements during planetary evolution ( e.g. * ? ? ?
* ) , which could account for differences in li abundance between wide binaries and within clusters that should have the same initial metalicity .
especially for stars with small convective zones , like the components of bd+20 307 , material deposited should last for the entire main sequence lifetime .
the tightness of the bd+20 307 orbit means that any planets must be circumbinary .
if disparate pollution were to occur , it might be expected to favor the more massive star .
it is indeed the primary star that shows the greater li line strength . taken together , the modest li line strength of the primary and very low li line strength for the secondary argue for an old age of the system , certainly greater than that of the hyades and probably @xmath1 1 gyr . in this case , the dust generating event happened quite late . in the solar system ,
the last era of major planetesimal removal happened at an age 600 - 800 myr @xcite .
it is further tempting to speculate on what happens to the dust ring over time .
poynting - robertson and solar wind drag should cause the grains at 0.55 au to fall onto the central stars on timescales @xmath61000 yr @xcite .
meteorites in the solar system contain li , no matter where they formed in the original proto - planetary disk @xcite .
thus the opportunity seems to exist for substantial stellar pollution with li by the dust around bd+20 307 .
perhaps this dust is being swept up by a body orbiting between the ring and the binary and thus can not pollute the surface of the stars .
a detailed abundance study of the stars could reveal the extent of any pollution .
bd+20 307 is a spectroscopic binary composed of two similar stars of late f spectral - type in a short , rotationally synchronized orbit of @xmath03.4 days .
standard chromospheric - age calibrations for single solar - type stars are therefore not applicable to this system .
the binary is likely quite old , @xmath1 1 gyr , in order to allow the two stars to develop quite different lithium equivalent widths .
the collision that produced the circumbinary dust ring must have happened recently , given the short timescale for removal of material at 0.5 au , and could be expected to pollute the envelopes of the stars .
that they have such low lithium could be evidence that the dust does not actually reach them .
mathieu , r. d. , latham , d. w. , mazeh , t. , duquennoy , a. , mayor , m. , & mermilliod , j .- c .
1992 , in binaries as tracers of stellar formation .
editors , a. duquennoy & m. mayor ( cambridge : cambridge university press ) , 278 i acknowledge paul butler and mercedes lopez - morales for assistance in setting up mike , john debes for performing the observations , johanna teske for running the data through the pipeline , and george preston for obtaining the du pont spectra .
thanks also to lopez - morales and evgenya shkolnik for helpful discussions and ben zuckerman for comments on the manuscript .
this work was supported in part by a spitzer space telescope grant through jpl and the carnegie node of the nasa astrobiology institute . | three epochs of high resolution spectra of the star show that it is a short period ( @xmath03.4 day ) spectroscopic binary of two nearly identical stars . surprisingly , the two stars , though differing in effective temperature by only @xmath0250 k and having a mass ratio of 0.91 , show very different li line equivalent widths . a li 6707 line
is only detected from the primary star , and it is weak .
this star is therefore likely to be older than 1 gyr .
if so , the large amount of hot circumbinary dust must be from a very large and recent , but very late evolutionarily , collision of planetesimals . |
according to the large extra - dimension scenario , there exists a chance for higher dimensional black holes to be created at the lhc @xcite .
hence , the higher dimensional black holes have been intensively investigated .
it should be noted that the einstein theory of gravity is not the most general one in higher dimensions . in four dimensions ,
the einstein gravity can be deduced by assuming the general coordinate covariance and the absence of higher derivative terms larger than the second order in the lagrangian . in higher dimensions , however , the same assumptions lead to lovelock theory of gravity @xcite .
in addition to this mathematical motivation , we have a physical motivation to consider lovelock gravity .
in fact , at the energy scale of black hole production , the einstein gravity is not reliable any more .
it is widely believed that string theory is valid at the fundamental scale .
remarkably , string theory can be consistently formulated only in ten dimensions .
as is well known , string theory reduces to einstein gravity in the low energy limit . in string theory , however , there are higher curvature corrections in addition to the einstein - hilbert term .
thus , it is natural to extend einstein gravity into those with higher power of curvature in higher dimensions .
it is lovelock gravity that belongs to such class of theories . in lovelock gravity
, it is known that there exist static spherically symmetric black holes @xcite .
hence , it is reasonable to suppose black holes produced at the lhc are of this type . in five or six dimensions
, the lovelock gravity reduces to the so - called gauss - bonnet gravity which has static and spherically symmetric black hole solutions @xcite .
the stability analysis of gauss - bonnet black holes under scalar , vector , and tensor perturbations has been performed @xcite .
it is shown that there exists the scalar mode instability in five dimensions , the tensor mode instability in six dimensions , and no instability in other dimensions @xcite . in more than six dimensions , however , we need to consider more general lovelock gravity . for example , when we consider ten dimensional black holes , we need to incorporate the third and fourth order lovelock terms into the action .
indeed , when we consider black holes at the lhc , it is important to consider these higher order lovelock terms .
the stability of black holes in any order lovelock gravity has been studied @xcite .
it turned out that small black holes are unstable in any dimensions . in order to understand properties of black holes in lovelock gravity , it is important to study qnms of black holes .
the qnfs of gauss - bonnet gravity have already been investigated @xcite .
thus , the aim of this paper is to calculate qnfs of the stable large black holes in lovelock gravity using the wkb method @xcite .
the qnfs of lovelock black holes have been calculated for the special background @xcite .
since the metric is analytically known for such cases , there is no difficulty in using the wkb - method for obtaining qnfs . in general , however , a problem arises because the metric of the black hole is only implicitly given by an algebraic equation .
hence , the primary aim of this paper is to give a general formalism to calculate qnfs of black holes in lovelock gravity .
the other purpose of this paper is to calculate qnfs of lovelock black holes in seven and eight dimensions and reveal effects of higher order lovelock terms on qnfs .
the organization of the paper is as follows . in sec .
ii , we introduce lovelock gravity and lovelock black holes . in sec .
iii , we review the perturbation theory of lovelock black holes . in sec .
iv , we explain the wkb method for obtaining qnfs .
there , we propose a novel method adapted to lovelock gravity for calculating qnfs .
then , we present numerical results and extract many interesting features from the results .
the final section is devoted to the conclusion .
in this section , we briefly review lovelock gravity in higher dimensions @xcite and derive static black hole solutions determined by an algebraic equation .
lovelock gravity is characterized by the general coordinate covariance and the absence of terms with higher than the second order derivatives in the equations of motion .
the lagrangian satisfying these properties can be constructed from @xmath0th lovelock terms @xmath1 defined by the product of riemann tensors _ ( m ) ^_1 _ 1 _ m _ m__1 _ 1 _ m _ m r^_1 _ 1 _ _ 1 _ 1 r^_m _ m _ _ m _ m , where we used the totally antisymmetric kronecker delta ^_1 _ p__1 _ p | cccc ^_1__1 & ^_2__1 & & ^_p__1 + ^_1__2 & ^_2__2 & & ^_p__2 + & & & + ^_1__p & ^_2__p & & ^_p__p @xmath2 is written by _ d
-2 + ^k_m=1 _ ( m ) , where @xmath3 is a cosmological constant , @xmath4 represents the coupling constants of lovelock gravity and @xmath5 is a parameter defined by k .
this lagrangian is called the lovelock lagrangian .
we can set @xmath6 without losing generality .
the action obviously has the general coordinate invariance .
it is also straightforward to see the absence of higher derivative terms larger than the second order derivatives . performing the variation with respect to the metric
, we obtain the lovelock tensor @xmath7 defined as ^d _
= ^ _ - ^k_m=1 ^_1 _ 1 _ m _ m__1 _ 1 _ m _ m r^_1 _ 1 _ _ 1 _ 1 r^_m _ m _ _ m _ m , where we used the bianchi indentity to eliminate the terms with derivative of the riemann tensor .
this tensor is equal to the einstein tensor in @xmath8 and the einstein - gauss - bonnet tensor in @xmath9 .
thus , lovelock theory can be regarded as a natural generalization of einstein theory .
it is known that there exist static black hole solutions in lovelock gravity @xcite . to obtain the solutions
, we put an ansatz ds^2 = g _
x^ x^ = - f(r ) t^2 + r^2 + r^2 _ ij x^i x^j , where @xmath10 is the metric of the sphere in @xmath11 dimensions . using this ansatz
, we can calculate the lovelock tensor which must be zero in vacuum .
it is convenient to define a new variable @xmath12 as ( r ) = .
then , we can integrate equations of motion and obtain ( ) = , [ algebraic_eq ] where the polynomial @xmath13 is defined as ( ) = \ { c - ( k=1 ) , + + ^k_q=2 + - ( k 2 ) . .
[ kettei ] here , @xmath14 is related to adm mass of black hole as follows : m_adm . in this paper , we will focus on the asymptotically flat spacetime . to realize the asymptotically flat spacetime , we set @xmath15 and @xmath16 for all @xmath0 .
thus , @xmath13 becomes a monotonically increasing function of @xmath17 with zero at @xmath18 .
since @xmath18 is achieved at @xmath19 , we can realize asymptotically flat spacetime with this setup .
it should be stressed that eq.([algebraic_eq ] ) implicitly define @xmath20 , namely , the metric function @xmath21 . in other words , we can not obtain the metric in an analytic form in general .
in the previous section , we have presented black hole solutions in lovelock gravity .
now , we consider a scalar field on this spacetime and linear metric perturbations around this background spacetime . we show that there exists a master equation in any case .
the master equation is an equation for the master variable by which all other components of the metric can be derived . in the case of the scalar field ,
the master equation is nothing but the klein - gordon equation . in the case of metric perturbations
, one can also obtain the master equations .
the explicit relation between the master variable and the other metric components can be found in @xcite .
once we obtain the master equations which are separable , we can analyze the stability of black holes and calculate qnfs of black holes .
since we are considering static background , the time - dependence of the fields can be separated by the factor e^-i t . moreover , due to the spherical symmetry , angular dependence of the master variable can be separated by using harmonics on the sphere .
thus , the master equation can be reduced to the schr@xmath22dinger type equation for the radial direction as + ( x ) = 0 , where we have defined the tortoise coordinate @xmath23 as follows : = , for quasinormal mode analysis , the shape of the effective potential @xmath24 is our main objective . in the rest of this section
, we write down the various effective potentials we will use in this paper .
as we mentioned , the master equation of a scalar field is the klein - gordon equation .
the klein - gordon equation with the mass @xmath25 in curved spacetime reads ( t , r , ) = 0 , where @xmath26 represents the coordinates in a @xmath27-dimensional sphere .
the spectrum of spherical harmonics @xmath28 on a @xmath27-dimensional sphere is given by y_l ( ) = - l(l+n-1 ) y_l ( ) , with an integer @xmath29 .
after separating variables , we obtain the following equation + ( x ) = 0 . here , a prime represents a differentiation with respect to the radial coordinate @xmath30 .
therefore , in this case we get the effective potential as follows : v(r ) = f(r ) ( ^2 + f(r ) + f(r ) + ) .
the metric perturbations in lovelock gravity was fully analyzed in @xcite .
there , the metric perturbations are classified into tensor , vector , and scalar types according to the symmetry on the sphere
. then , the master equations for metric perturbations of tensor- , vector- , and scalar - type have been derived . in this article
, we simply use the effective potentials obtained in @xcite . in the following ,
we use the functions @xmath31 and @xmath32 defined by t(r ) & & r^n-1 _ ( ) , + x(r ) & & . for the absence of the ghost ,
we need to assume @xmath33 for @xmath34 .
we call it the @xmath35-condition .
the master equation for tensor type perturbations can be written as v_t(r ) = f(r ) .
it is known that there exists the instability of small black holes under tensor perturbations in even dimensions .
the master equation for vector perturbations reads v_v(r ) = f(r ) .
there exists no instability in this sector .
finally , the master equation for scalar perturbations takes the following form v_s(r ) & = & f(r ) . in this case
, it is also known that there exists the instability of small black holes under scalar perturbations in odd dimensions .
the lovelock black holes are unstable in various ways .
the instability of black holes caused by the ghost is characterized by the violation of the @xmath35-condition .
as we mentioned , there are other types of instability under tensor and scalar type perturbations .
the instability comes from the negative region of the effective potential . for this class of black holes
, we can not use the wkb method because this method can be applicable to the positive definite effective potential with single peak . in this paper
, we calculate the qnfs of stable black holes by excluding the cases with the negative regions or with the multipeaks in the potential .
in this section , we present a novel method for the calculation of quasinormal modes in lovelock gravity .
quasinormal modes are fundamental vibration modes around a black hole . these modes are obtained by solving the master equation under the appropriate boundary condition .
the general formalism of calculating qnfs by using wkb - approximation has been proposed by schutz and will @xcite and subsequently developed by many people @xcite . here , we summarize the main points of the wkb - method for qnfs . in general , we can divide the region into two regions . the region @xmath36 ( @xmath37 ) is the one ranging from top of the effective potential @xmath38 to the horizon of the black hole .
the region @xmath39
( @xmath40 ) is the one ranging from top of the potential @xmath38 to the far outside of the black hole , i.e. , infinity .
the wave traveling to the potential is called an ingoing wave and the wave traveling from the potential is called outgoing wave . in each region , the solutions of the master equation can be expressed as \ {
c _ i = z^in_i^in_i + z^out_i ^out_i , + + _ ii = z^in_ii^in_ii + z^out_ii ^out_ii , . where @xmath41 and @xmath42 represent ingoing and outgoing wave , respectively .
the boundary condition for obtaining quasinormal modes is that there are no ingoing waves : z^in_i = z^in_ii= 0 .
since there are two conditions , only discrete complex eigenvalues are allowed . in the @xmath43th - order wkb - method , we approximate the function @xmath44 defined by q(x ) ^2 - v(x ) . in terms of @xmath45th - order taylor expansion around the maximum of the potential as follows ; .
|_x = x_0=q^(1)_0 = 0 q(x ) ^2n_p=0 q^(p)_0 ( x - x_0 ) ^p . expressing the wave function using wkb - approximation
, we can calculate the scattering matrix .
thus , we can get the formula for qnfs as , where @xmath46 is the @xmath47th order derivative of the potential v^(p)_0 .
|_x = x_0 , and the first and the second of @xmath48 are given by _ 1 & = & -30 ( n_tone + ) ^2_1 ^ 2 + 6_2 ( n_tone + ) ^2 - 7/2_1 ^ 2 + 3/2_2 , + & & + _ 2 & = & ( n_tone + ) ^3 ( -2820_1 ^ 4 + 1800 _ 1 ^ 2_2 - 280_1_3 - 68_2 ^ 2 + 20_4 ) + & & + ( n_tone + ) ( -1155_1 ^ 4 + 918_1 ^ 2_2 - 190_1_3 - 67_2 ^ 2 + 25_4 ) .
here , @xmath49 is defined as _
p ( ) ^+1 the parameter , @xmath50 , is called tone number of quasinormal modes .
this method is often called the @xmath43th - order - wkb - approximation .
this time , we used the 3rd - order wkb method .
so we need @xmath51 and @xmath52 .
it is known that in the case of @xmath53 this approximation is good .
so we focused on the case @xmath54 in this paper .
let us calculate qnfs of the fields in lovelock gravity semianalytically by using the wkb method . to use the wkb method , we need to know the potential function and hence the metric function @xmath21 .
that is , we need to solve the @xmath5th order algebraic equation ( [ algebraic_eq ] ) for @xmath12 in terms of @xmath30 .
unfortunately , it is impossible to solve the algebraic equation analytically for arbitrary dimensions .
thus , we need to invent a novel method adapted to lovelock gravity . to circumvent this difficulty , we take the following strategy . instead of solving eq .
@xmath55 , we change the variable from @xmath30 to @xmath17 using the relation r = .
thus , we are able to get the analytic representation of @xmath56 as f = 1 - r^2 = 1 - . [ f ] so , if we use this formulation , we will be able to calculate qnfs of the static and spherically symmetric black holes in lovelock gravity .
now , we check the range of @xmath17 .
we are considering the static spherically symmetric black hole solutions with the asymptotically flat region .
when @xmath30 goes to @xmath57 , @xmath21 should be 1 because of asymptotic flatness .
therefore , in the limit @xmath58 , we have 0 .
when @xmath30 approaches the horizon , @xmath21 must vanish .
hence , in this limit , we obtain ( ) ^2 = ^n+1 [ cond ]
. there will be the unique real solution @xmath59 of this equation because we know @xmath13 is the monotonic increasing function of @xmath17 .
to summarize , we obtained the relation between @xmath30 and @xmath17 as follow : \ { cccc r : & horizon & & + : & _ h & & 0 . .
in this region of @xmath17 , we must find @xmath60 which satisfy the equation , v^(1)_0 = 0 .
once we find it , we can use the wkb method and calculate qnfs of lovelock black holes .
+ @xmath61 , @xmath62 , @xmath63 , @xmath64 , @xmath65 + the effective potential of tensor perturbation in tensor field [ gp ] c ex.1 ) double peaks ex.2 ) negative region @xmath66 , @xmath67 , @xmath68 , @xmath69 , @xmath65 [ nrdp ] + the effective potential of scalar perturbation in tensor field in this subsection , we calculate the qnfs of lovelock black holes in seven and eight dimensions using the wkb method adapted to lovelock gravity . from now on , we rewrite the parameters as a_q\ { l 1 ( q=1 ) , + + ( q2 ) . .
we calculated the qnfs for various values of @xmath70 using the third order wkb method .
note that we fixed the @xmath3 parameter to zero and considered a massless scalar field ( @xmath71 ) .
we calculated qnfs only for the cases like in fig.[gp ] .
namely , the potential is positive definite in the range @xmath72 and has a single peak .
for some parameters , the effective potential has the multipeaks or the negative region ( fig.[nrdp ] ) .
we excluded these and the cases with ghost instability from the calculation . in fig.[sf ]
- fig.[sp ] , we plotted the various qnfs of various fields in lovelock black holes in seven and eight dimensions .
c , width=283 ] , width=283 ] + + + , width=283 ] , width=283 ] + + + , width=283 ] , width=283 ] [ sf ] in fig.[sf ] , we plotted qnfs of the scalar field for the masses @xmath73 in seven and eight dimensions . in the case of the scalar field
, we took the lowest angular parameter @xmath74 . in each panel , we plotted a curve by changing the coefficient @xmath75 from 0 to 1.5 with the interval of 0.05 for a fixed @xmath76 .
then , we repeated this by changing @xmath76 from 0 to 1.0 with the interval 0.1 .
the color of lines in the figures tells us the value of @xmath76 .
the most red color corresponds to the @xmath77 , and the most blue color corresponds to the @xmath78 .
the edge of the curve ( the point indicated by the arrow in the case of the red curve ) corresponds to @xmath79 , and the other edge of the same curve corresponds to @xmath80 .
so the qnfs of einstein gravity in seven and eight dimensions corresponds to the edge of the most red curve indicated by the arrow . in the case of the scalar field
, the imaginary part of qnfs seems to converge to a value as we decrease the mass . in other words ,
the gauss - bonnet term ceases to be relevant .
this trend can be seen clearly in eight dimensions rather than seven dimensions .
it is also interesting to observe that the pattern is rotating in clockwise .
c , width=283 ] , width=283 ] + + + , width=283 ] , width=283 ] + + + , width=283 ] , width=283 ] [ tp ] in fig.[tp ] , we plotted qnfs of the tensor type perturbations of the metric field for the masses @xmath81 in seven and eight dimensions .
we have chosen @xmath63 in the case of tensor type perturbations , which gives the positive definite effective potential with single peak . in each panel , as in the case of the scalar field , we plotted a curve by changing the coefficient @xmath75 from 0 to 1.5 with the interval of 0.05 for a fixed @xmath76 .
then , we repeated this by changing @xmath76 from 0 to 1.0 with the interval 0.1 .
the meaning of the arrow is the same as the case of the scalar field . in the case of the tensor type of metric perturbations
, we can see there appears a turning point in the curve . in this case , initially , the blue curves are always above the red ones irrespective of the mass of black holes . as we decrease the mass , the red curve becomes above the blue one for some gauss - bonnet parameters .
c , width=283 ] , width=283 ] + + + , width=283 ] , width=283 ] + + + , width=283 ] , width=283 ] [ vp ] c , width=283 ] , width=283 ] + + + , width=283 ] , width=283 ] + + + , width=283 ] , width=283 ] [ sp ] in fig.[vp ] , we plotted qnfs of the vector type perturbations of metric field for the masses @xmath73 in seven and eight dimensions .
we have chosen @xmath68 in the case of vector type perturbations , which gives the positive definite effective potential with single peak . in each panel , as in the case of the scalar field , we plotted a curve by changing the coefficient @xmath75 from 0 to 1.5 with the interval of 0.05 for a fixed @xmath76 .
then , we again repeated this by changing @xmath76 from 0 to 1.0 with the interval 0.1 .
the meaning of the arrow is the same as the previous cases .
in the case of vector type perturbations , there are no special features , which might be related to the fact that there is no instability in this sector .
however , in this case , the red curves are always above the blue curves . in fig.[sp ]
, we plotted qnfs of the scalar type perturbations of the metric field for the masses @xmath73 in seven and eight dimensions .
we have chosen @xmath68 in the case of scalar type perturbations , which gives the positive definite effective potential with single peak .
for the masses @xmath82 , as in the case of the scalar field , we plotted a curve by changing the coefficient @xmath75 with the interval of 0.05 for a fixed @xmath76 .
then , we repeated this by changing @xmath76 from 0 to 1.0 with the interval 0.1 .
however , for the masses @xmath83 , we repeated this by changing @xmath76 from 0 to 0.3 with the interval 0.1 in seven dimensions and from 0 to 0.2 with the same interval in eight dimensions .
the meaning of the arrow is the same as the previous one . in the fig.[sp ] , in the case of @xmath84 , we have only sparse plots . in the case of the scalar type of metric perturbations as well as the vector type perturbations , the trend is opposite to the tensor type of metric perturbations .
that is , the gauss - bonnet term decreases the absolute value of the imaginary part of qnfs .
admittedly , the wkb method is not appropriate for a small black hole for which we know there exists the instability .
we need other methods for these calculations .
we have studied quasinormal modes of black holes in lovelock gravity in seven and eight dimensions . since lovelock
black holes are defined by an algebraic equation , we needed to modify the wkb method .
thus , we proposed a novel wkb method adapted to lovelock gravity for the calculation of qnfs . as a demonstration , we calculated various qnfs of lovelock black holes in seven and eight dimensions .
we found remarkable features of qnfs of lovelock gravity which depend on the coefficients of the lovelock terms , the species of perturbations , and dimensions of the spacetime . in the case of the scalar field ,
when we increase the coefficients of the third order lovelock term the real part of qnfs increase , but the decay rate becomes small irrespective of the mass of the black hole .
for small black holes , the decay rate ceases to depend on the gauss - bonnet term .
this tendency can be seen clearly in eight dimensions than in seven dimensions .
the dependence on the gauss - bonnet coefficient changes for small black holes . indeed , there appears a turning point in each curve as the mass becomes small . in the case of tensor
type perturbations of the metric field , the tendency of the real part of qnfs is opposite to the scalar field . the qnfs of vector type perturbations of the metric field show no particular behavior .
the behavior of qnfs of the scalar type perturbations of the metric field is similar to the vector type .
however , available data are rather sparse , which indicates that the wkb method is not applicable to many models for this sector .
the features we found in this work are intriguing and worth verifying by using other methods .
in particular , for calculating qnfs for small black holes , we need to solve partial differential equations numerically using the time - domain method .
we leave this work for future study .
this work was supported by grants - in - aid for scientific research ( c ) no.25400251 and mext grant - in - aid for scientific research on innovative areas no.26104708 and mext grant - in - aid for scientific research on innovative areas `` cosmic acceleration '' ( no.15h05895 ) . | we study quasinormal modes of black holes in lovelock gravity .
we formulate the wkb method adapted to lovelock gravity for the calculation of quasinormal frequencies ( qnfs ) . as a demonstration ,
we calculate various qnfs of lovelock black holes in seven and eight dimensions .
we find that the qnfs show remarkable features depending on the coefficients of the lovelock terms , the species of perturbations , and spacetime dimensions . in the case of the scalar field ,
when we increase the coefficient of the third order lovelock term , the real part of qnfs increases , but the decay rate becomes small irrespective of the mass of the black hole . for small black holes ,
the decay rate ceases to depend on the gauss - bonnet term . in the case of tensor
type perturbations of the metric field , the tendency of the real part of qnfs is opposite to that of the scalar field . the qnfs of vector type perturbations of the metric show no particular behavior .
the behavior of qnfs of the scalar type perturbations of the metric field is similar to the vector type .
however , available data are rather sparse , which indicates that the wkb method is not applicable to many models for this sector . |
the usefulness of spheroidal functions in many branches of physics , like quantum mechanics , general relativity , signal processing , etc . , is well known and it does not need to be stressed . due to that usefulness
, the description of the spheroidal equation and of the main properties of its solutions deserves a chapter in handbooks of special functions like that by abramowitz and stegun ( * ? ? ?
21 ) , the best known one , or the _ nist digital library of mathematical functions _
30 ) , the most recent one . a review of the procedures used in the past century for obtaining the eigenvalues and eigenfunctions of the spheroidal wave equation can be found in a paper by li _
@xcite , where also an algorithm , implemented with the software package ` mathematica ` , is provided . in the present century , articles dealing with the solutions of the angular spheroidal wave equation have continued appearing . without aiming to be exhaustive ,
let us mention the papers by aquino _
et al . _
@xcite , falloon _ et al . _
@xcite , boyd @xcite , barrowes _ et al . _
@xcite , walter and soleski @xcite , abramov and kurochkin @xcite , kirby @xcite , karoui and moumni @xcite , gosse @xcite , tian @xcite , rokhlin and xiao @xcite , osipov and rokhlin @xcite , ogburn _ et al . _
@xcite and huang _ et al . _
@xcite , and the books by hogan and lakey @xcite , and by osipov , rokhlin and xiao @xcite .
different strategies have been used to solve the angular spheroidal wave equation .
the classical procedure starts with the angular spheroidal wave function written as a series of solutions of another similar differential equation , commonly the legendre one , with coefficients obeying a three term recurrence relation .
the resulting expansion becomes convergent only when such coefficients constitute a minimal solution of the recurrence relation .
the eigenvalue problem encountered in this way is solved either as a transcendental equation involving a continued fraction , or written in a matrix form .
procedures based on the direct solution of the angular spheroidal equation , without having recourse to comparison with other differential equations , have been less frequently used . the relaxation method proposed by caldwell @xcite and reproduced , as a worked example , in the _ numerical recipes _ ( * ? ? ?
17.4 ) , and the finite difference algorithm , described in the recently appeared paper by ogburn _ et al . _
@xcite , deserve to be quoted .
here we suggest to follow a procedure , based also on the direct treatment of the spheroidal equation , which benefits from an idea that can be found in a paper by skorokhodov and khristoforov @xcite dealing with the singularities of the eigenvalues @xmath0 considered as function of the ( complex ) prolateness parameter @xmath1 .
a shooting method is used .
but , instead of imposing the boundary conditions to a numerically obtained solution , algebraic regular solutions around the regular point @xmath2 or around the regular singular point @xmath3 are written .
smooth matching of both solutions , i. e. cancelation of their wronskian , at any point @xmath4 determines the eigenvalues . in our implementation of the procedure , we choose @xmath2 as matching point . a discomfort , when dealing with spheroidal wave functions , is the lack of universality of the notation used to represent them .
the _ digital library of mathematical functions _
30 ) provides information about the different notations found in the bibliography .
here we adopt , for the eigenvalues and eigenfunctions , the notation of the _ handbook of mathematical functions _
the same notation is used in ref .
@xcite , a paper whose results we will try to reproduce , for comparison , with the method here developed . in the next section , we recall the angular spheroidal equation and write its solutions in the form of power series expansions around the origin and around the singular point @xmath3 . the procedure for computing the eigenvalues
is presented in section 3 .
the results of its application in some typical cases are also given .
section 4 shows that normalized eigenfunctions can be trivially obtained .
some figures illustrate the procedure .
a few final comments are contained in section 5 .
the angular spheroidal wave function @xmath5 , defined in the interval @xmath6 , satisfies the differential equation ( * ? ? ?
21.6.2 ) @xmath7+\left(\lambda_{m , n}-c^2\eta^2-\frac{m^2}{1-\eta^2}\right)s_{m , n}(c,\eta)=0 \label{ii1}\ ] ] stemming from the separation of the wave equation in spheroidal coordinates , with separation constants @xmath8 and @xmath0 .
periodicity of the azimuthal part of the wave restricts the values of @xmath8 to the integers and , given the invariance of the differential equation in the reflection @xmath9 , only non - negative integer values of @xmath8 need to be considered .
the other separation constant , @xmath0 , commonly referred to as eigenvalue , must be such that @xmath5 becomes finite at the singular points @xmath10 .
their different values , for given @xmath8 and @xmath11 , are labeled by the integer @xmath12 . in most applications ,
the external parameter @xmath11 is real , positive in the case of prolate coordinates and negative for oblate ones .
there are , however , interesting cases corresponding to complex values of @xmath11 @xcite . instead of solving directly eq .
( [ ii1 ] ) , it is convenient to introduce the change of function @xmath13 and to solve the differential equation @xmath14 where @xmath15 is considered as the new eigenvalue .
two independent solutions about the ordinary point @xmath2 , valid in the interval @xmath16 , are @xmath17 with coefficients given by the recurrence relation @xmath18\,a_{k-2,\sigma}+c^2\,a_{k-4,\sigma}\ , .
\label{ii6}\end{aligned}\ ] ] obviously , @xmath19 and @xmath20 are respectively even and odd functions of @xmath21 .
solutions about the regular singular point @xmath3 can also be written . in terms of the variables
@xmath22 the differential equation ( [ ii3 ] ) turns into @xmath23u(t)=0\ , .
\label{ii8}\ ] ] the solution of this equation which makes @xmath24 to be regular at @xmath25 is , except for an arbitrary multiplicative constant , @xmath26 with coefficients given by @xmath27b_{j-1 } & & \nonumber\\ & & \hspace{-80pt } -\,2c^2\,b_{j-2}+c^2\,b_{j-3}\ , .
\label{ii10}\end{aligned}\ ] ] in terms of the variable @xmath21 , this regular solution , valid for @xmath28 , is @xmath29
the problem of finding the eigenvalues @xmath0 , for given @xmath11 and @xmath8 , reduces to require the regularity of @xmath30 at @xmath3 . ( the regularity at @xmath31 is then implied by the symmetry of @xmath32 . ) in the particular case of being @xmath33 , the problem can be solved algebraically .
the recurrence relation in ( [ ii6 ] ) reduces in this case to @xmath34\,a_{k-2,\sigma}\ , .
\label{iii1}\\\ ] ] obviously , the series in the right hand side of ( [ ii5 ] ) is divergent for @xmath3 unless the value of @xmath35 is such that one of the @xmath36 of even subindex , say @xmath37 ( with @xmath38 ) , becomes zero , in which case @xmath30 turns out to be a polynomial of degree @xmath39 .
this happens for @xmath40 by using the habitual notation @xmath41 for the degree of the polynomial , we obtain for the eigenvalues in the case of @xmath33 @xmath42 that is , in view of ( [ ii4 ] ) , @xmath43 for @xmath44 , a convenient way of guaranteeing the regularity of @xmath30 at @xmath3 is to require the cancelation of the wronskian @xmath45 of @xmath32 and @xmath46 ,
@xmath47(\eta)=-\left(\sum_{k=0}^\infty a_{k,\sigma}\,\eta^{k+\sigma}\right)\left(\sum_{j=0}^\infty jb_j\,(1-\eta)^{j-1}\right ) & & \nonumber \\ & & \hspace{-200pt}-\,\left(\sum_{k=0}^\infty ( k+\sigma ) a_{k,\sigma}\,\eta^{k-1+\sigma}\right)\left(\sum_{j=0}^\infty b_j\,(1-\eta)^j\right ) , \label{iii6}\end{aligned}\ ] ] at an arbitrarily chosen point of the interval @xmath16 . from the computational point of view ,
an interesting choice of @xmath21 seems to be @xmath48 , in which case @xmath49(1/2)=-\sum_{l=0}^\infty 2^{-l}\,(l+1)\left(\sum_{j=0}^{l+1-\sigma}a_{l+1-\sigma - j , \sigma}\,b_j\right)\ , .
\label{iii7}\ ] ] the set of coefficients @xmath50 and @xmath51 are solutions of the difference equations ( [ ii6 ] ) and ( [ ii10 ] ) , respectively . according to the perron - kreuser theorem on difference equations @xcite ,
@xmath52 that is , for any given @xmath53 , constants @xmath54 and @xmath55 can be found such that @xmath56 this makes evident that the series in the right hand side of eq .
( [ iii7 ] ) converges as fast as the geometric series @xmath57 .
we consider , however , that a better choice of the value of @xmath21 in eq .
( [ iii6 ] ) is @xmath2 . in this case , @xmath49(0)=-\,\delta_{\sigma,0}\,\left(\sum_{j=0}^\infty j\,b_j\right ) -
\delta_{\sigma,1}\,\left(\sum_{j=0}^\infty b_j\right)\ , .
\label{iii10}\ ] ] obviously , the cancelation of this wronskian occurs when either @xmath58 or @xmath46 vanish at the origin , as it occurs for respectively even or odd functions of @xmath21 .
needless to say , the right hand side of ( [ iii10 ] ) depends on the variable @xmath35 , introduced in ( [ ii4 ] ) , through the coefficients @xmath59 .
therefore , the problem of finding the eigenvalues of the angular spheroidal equations , for given @xmath11 and @xmath8 , reduces to the determination of the zeros of the function @xmath60(0)=\delta_{\sigma,0}\,\left(\sum_{j=0}^\infty j\,b_j(z)\right ) + \delta_{\sigma,1}\,\left(\sum_{j=0}^\infty b_j(z)\right)\ , , \label{iii11}\ ] ] where we have indicated the dependence of the @xmath59 on @xmath35 .
different procedures can be used in the determination of the zeros of @xmath61 .
a useful iterative method is the newton one . starting with an initial approximate value , @xmath62 , of a certain zero , repeated application of the algorithm @xmath63 allows one to get the value of the zero with the desired accuracy .
the extended newton method , which uses @xmath64^{1/2 } } { \mathcal{w}_\sigma^{\prime\prime}(z^{(i)})}\ , , \label{iii13}\ ] ] is even more efficient .
for the first and second derivatives of @xmath61 with respect to @xmath35 we have the expressions @xmath65 the first and second derivatives of the coefficients @xmath66 are easily obtained by means of the recurrence relations @xmath67b_{j-1}^\prime(z ) \nonumber \\ & & \hspace{-40pt } -\,2c^2\,b_{j-2}^\prime(z)+c^2\,b_{j-3}^\prime(z)-b_{j-1}(z)\ , , \label{iii16 } \\
b_0^{\prime\prime}(z)=0\,,\quad 2j(j+m)\,b_j^{\prime\prime}(z)&=&\left[(j-1)(j+2m)-z+c^2\right]b_{j-1}^{\prime\prime}(z ) \nonumber \\ & & \hspace{-40pt } -\,2c^2\,b_{j-2}^{\prime\prime}(z)+c^2\,b_{j-3}^{\prime\prime}(z)-2b_{j-1}^\prime(z)\ , , \label{iii17}\end{aligned}\ ] ] stemming from ( [ ii10 ] ) . from these difference equations , inequalities analogous to the second one in ( [ iii9 ]
) can be deduced .
such inequalities guarantee the convergence , as fast as the geometric series @xmath68 , of the series in the right hand sides of eqs . ( [ iii11 ] ) , ( [ iii14 ] ) , and ( [ iii15 ] ) . we have applied the procedure just described for obtaining the behaviour of the lowest eigenvalues of the spheroidal equation when the parameter @xmath11 varies in the interval @xmath69 $ ] and for values of @xmath700 , 1 , and 2 . the results are shown in figs .
( remember that @xmath71 . ) a glance at fig .
1 suggests a quasi - confluence of trajectories of the eigenvalues @xmath72 and @xmath73 for sufficiently large negative values of @xmath11 .
one may conjecture that , for larger negative values of @xmath11 other pairs of trajectories , those of @xmath74 and @xmath75 , present such quasi - confluence .
table 1 shows that this is the case , and that a similar phenomenon occurs for other values of @xmath8 .
.lowest eigenvalues of the ( oblate ) spheroidal equation for several values of @xmath8 and @xmath11 .
[ cols="<,>,>,>",options="header " , ] a comment concerning the values of the label @xmath12 of @xmath0 reported in table 3 is in order . for real or pure imaginary @xmath1 ,
i.e. for real @xmath11 , the eigenvalues @xmath0 for given @xmath8 are real and can be ordered by increasing value .
the label @xmath12 reflects that order . for complex @xmath11 , instead
, the values of @xmath76 become complex and such ordination is no more possible .
nevertheless , a label @xmath12 can be assigned to those complex values of @xmath77 , as done by skorokhodov and khristoforov . by keeping constant the real part of @xmath1 and continuously decreasing its imaginary part , @xmath76 describes , in the complex @xmath77-plane , a trajectory which intersects the real @xmath77-axis at a certain @xmath0 for @xmath78 .
this label @xmath12 can be attached to the whole trajectory described by @xmath76 as @xmath1 varies in the complex plane .
the paper by skorokhodov and khristoforov contains a very lucid discussion of those trajectories and shows that the branch points @xmath79 correspond to singular values of @xmath1 such that @xmath80 , with @xmath81 .
once the eigenvalues @xmath0 have been calculated , the corresponding eigenfunctions , in the interval @xmath82 , can be obtained immediately by means of the series expansion @xmath83 the coefficients @xmath59 being given by the recurrence relation ( [ ii10 ] ) with @xmath84 .
the normalization constant @xmath85 should be adjusted to the normalization scheme preferred .
a discussion of the different normalizations used in the literature can be found in the paper by kirby @xcite , where the advantage of the unit normalization @xmath86 is made evident . by choosing this normalization
, one has @xmath87^{-1/2}\ , , \label{iv3}\ ] ] where the asterisk indicates complex conjugation .
the constant phase @xmath88 in the right hand side of ( [ iv1 ] ) may be taken at will . in the case of real @xmath11 ,
it is natural to take @xmath89 . for complex @xmath11 ,
@xmath88 can be chosen in such a way that @xmath24 becomes real at @xmath2 , or at @xmath3 , or at any other point .
needless to say , @xmath90 , according to the even or odd nature of @xmath24 .
figures 5 and 6 show two examples of the application of the method to the computation of spheroidal angular wave functions in the case of real @xmath11 .
the first one is an even prolate angular wave function of parameters @xmath91 , @xmath92 and @xmath93 , and eigenvalue @xmath94 , a case considered in table iii of ref .
@xcite ( with the flammer @xcite normalization scheme ) .
the second one is an odd oblate angular wave function corresponding to the first of the cases considered in our table 2 . in both figures ,
the functions have been normalized according to eq .
( [ iv2 ] ) .
we have applied our procedure also in some cases of complex @xmath1 .
figure 7 shows the real and imaginary parts and the squared modulus of the wave function in a case considered by falloon _
@xcite , namely the first one in their table 2 .
the parameters are @xmath95 , @xmath92 and @xmath96 , and the eigenvalue , in our notation , is @xmath97 .
( we give here the eigenvalue with only 15 decimal digits , but our procedure is able to reproduce the 25 decimal digits given in ref .
@xcite and to obtain even more . ) finally , figures 8 , 9 and 10 correspond to the second of the cases considered in table 2 of ref .
@xcite , of parameters @xmath98 , @xmath92 , @xmath99 , and eigenvalue @xmath100 .
( notice the discrepancy , in the six last digits of both real and imaginary parts , with the value given in ref .
we have developed a rarely used method which allows to find the eigenvalues and eigenfunctions of the angular spheroidal equation .
instead of having recourse to comparison with other differential equations , the procedure deals with a direct solution , expressed in the form of a convergent series .
requiring it to be regular gives the eigenvalues , which appear as the zeros of a one variable function , @xmath61 .
this function and its derivatives @xmath101 and @xmath102 with respect to the variable @xmath35 can be computed , to the desired precision , by summing rapidly convergent series .
this fact makes possible the application of the extended newton method for the determination of the zeros of @xmath61 , i. e. , the eigenvalues of the spheroidal equation .
then , the computation of the corresponding eigenfunctions , conveniently normalized , becomes trivial . for the normalization , one benefits from the fact that the squared modulus of the wave function can be integrated algebraically .
the fact that , for given @xmath1 and @xmath8 , the eigenvalues are the zeros of an easily computable function , @xmath61 , makes possible to get an initial approximate location of all of them by a tabulation or a graphical representation of that function . repeated application of the extended newton method allows then to calculate the eigenvalues with the desired accuracy . we have shown the applicability of the method not only in the cases of prolate ( real @xmath1 ) and oblate ( imaginary @xmath1 ) spheroidal wave equations , but also when @xmath1 is complex .
the procedure provides in all cases a very precise determination of the eigenvalues .
this has allowed us to resolve the quasi - confluence of pairs of even - odd eigenvalues for large imaginary values of @xmath1 ( table 1 ) and of pairs of even - even or odd - odd eigenvalues for complex values of @xmath1 in the neighbourhood of branch points " ( table 3 ) .
the efficiency of the procedure proposed in this paper is subordinate to the capability of computing @xmath61 with sufficient accuracy .
the convergence of the series in the right hand side of eq .
( [ iii11 ] ) is guaranteed in all cases , since @xmath103 for all @xmath12 larger than a certain @xmath104 , and the series may replaced by a sum up to say @xmath105 .
nevertheless , for large values of @xmath1 and/or @xmath0 , the coefficients @xmath59 increase ( in modulus ) rapidly with @xmath106 before starting to decrease , and the adequate value of @xmath107 may become very large .
even worse , the values of the terms to be summed may cover so many orders of magnitude that the resulting sum is not reliable , unless many significant digits are carried along the computation .
this drawback is not outside other procedures .
however , the algorithms proposed by kirby @xcite and by ogburn _
@xcite , and procedures collected in refs .
@xcite , seem to be able to tackle the issue properly .
asymptotic methods @xcite have also been forwarded for the mentioned cases of large values of @xmath1 and/or @xmath0 .
the work has been supported by departamento de ciencia , tecnologa y universidad del gobierno de aragn ( project 226223/1 ) and ministerio de ciencia e innovacin ( project mtm2015 - 64166 ) b. e. barrowes , k. oneill , t. m. grzegorczyk and j. a. kong , on the asymptotic expansion of the spheroidal wave function and its eigenvalues for complex size parameter , _ stud .
_ * 113 * ( 2004 ) , 271301 .
z. huang , j. xiao , and j. p. boyd , adaptive radial basis function and hermite function pseudospectral methods for computing eigenvalues of the prolate spheroidal wave equation , _ j. comput
* 281 * ( 2015 ) , 269284 . | an algorithm for computing eigenvalues and eigenfunctions of the angular spheroidal wave equation , based on a known but scarcely used method , is developed . by requiring the regularity of the wave function , represented by its series expansion ,
the eigenvalues appear as the zeros of a one variable function easily computable .
the iterative extended newton method is suggested as especially suitable for determining those zeros .
the computation of the eigenfunctions is then immediate .
the usefulness of the method , applicable also in the case of complex values of the prolateness " parameter , is illustrated by comparing its results with those of procedures used by other authors .
msc [ 2010 ] 33e10 ( primary ) 33f05 , 34l16 , 65d20 ( secondary )
pacs 02.30.hq 02.30.gp 03.65.ge |
the first luminous objects in the universe formed from primordial gas , comprised solely of h and he , with only trace amounts of d an li . the relatively simple chemistry of metal - free gas , combined with tighly constrained cosmological parameters @xcite , has allowed the formation of the first stars to be simulated with extremely high precision , from the hierarchical growth of their host dark matter halos through to the point where the dense proto - stellar cores becomes optically thick @xcite . with the deaths of these stars came the creation of the first heavy elements .
core - collapse and pair - instability supernovae created metals in copious amounts @xcite and ejected them into the igm @xcite .
the presence of metals alters the dynamics of collapsing gas - clouds by increasing the number of available atomic and molecular transitions , allowing the gas to lose its internal energy more quickly than in case of no metals @xcite .
the introduction of metals adds a new level of complexity to the problem of simulating the formation and evolution of cosmic structure .
@xcite identified a minimal set of 21 chemical reactions necessary for accurately following the non - equilibrium evolution of a gas consisting solely of species of h and he , including h@xmath7
. showed that 33 total reactions were required when including d and li species to the gas .
@xcite performed one of the first numerical studies of collapsing gas - clouds to consider the contribution of metals .
their chemical network of h , he , c , and o included 50 atomic and molecular species and 478 reactions . while theirs was not a minimal model , the above examples illustrate the great expense associated with the expansion of chemical networks to include additional elements .
other works have studied the effect of metals on star - forming gas using similar methodologies to that of @xcite , e.g. , @xcite .
the complexity of the chemical networks used in these studies limited their treatment of gas evolution to one - zone , semi - analytical models . in the earliest work to incorporate metal cooling into three - dimensional hydrodynamic simulations to study metal - enriched star formation ,
@xcite used a small set of the most dominant atomic transitions of c , n , o , fe , si , and s , as decribed by @xcite .
their method also ignored the cooling from h@xmath7 , which was justified within their study by the assumption of a very large photo - dissociating uv background , but is otherwise an extremely important coolant in low - metallicity environments .
for high temperature gases , @xcite computed metal cooling functions that included 14 heavy elements over a range of metallicities , with solar abundance patterns .
these cooling functions are useful for simulating the igm and other hot , ionized environments , but a minimum temperature of 10@xmath8 k makes them inapplicable to studies of the cold , neutral gas associated with star - formation .
these cooling functions assume collisional equilibrium of the species and as such can not capture the important role of uv and x - ray radiation .
we introduce a new method for including the cooling from heavy elements in large - scale hydrodynamic simulations that is valid over a wide range of physical conditions , covers a great number of elemental species , and is fast enough to be used in large - scale numerical simulations .
we have utilized the established photoionization software , cloudy @xcite to construct large grids of metal cooling data .
we have developed a method to include both the cooling from heavy elements and the non - equilibrium cooling from h@xmath7 in hydrodynamic simulations .
this method has been used successfully in the numerical simulations of star formation performed by @xcite . in [ sec : method ] , we describe our method for creating the metal cooling data , including a new code to expedite the process .
we , then , present two implementations of the cooling method in the amr , hydrodynamic / n - body code , enzo @xcite . in [ sec : metals ] , we focus on the application of metals to low - temperature environments , identifying the dominant cooling mechanisms , and studying the possibility of fragmentation and thermal instability in metal - enriched gas .
finally , we end with a discussion in [ sec : discussion ] of the role played by the heavy elements in the formation of structure in the early universe .
at the current time , it is still too computationally expensive and memory intensive to follow the non - equilibrium chemistry for a large set of heavy elements in a three - dimensional hydrodynamic simulation .
the exact mass of the first massive stars is not known @xcite . also unknown are the exact yields of early supernovae @xcite .
similarly , in many astrophysical systems one might want to model computationally the exact metal distributions .
consequently , it is not clear a priori what level of sophistication of cooling model is needed to adequately capture the hydro and thermodynamic evolution of the gas under consideration .
note that uncertain grain physics also increases the potentially important parameter space . in our approach ,
we assume ionization equilibrium , which allows us to calculate , in advance , the cooling rate for a parcel of gas with a given density and temperature , with incident radiation of known spectral shape and intensity . for this problem , we find the photoionization code , cloudy @xcite , especially apt .
cloudy is conventionally used to model the transmitted spectrum from a cloud of gas with a given chemical composition , being irradiated by a specified source .
the code must calculate an equilibrium solution by balancing the incident heating with the radiative cooling from a full complement of atomic and molecular transitions , as well as continuum emission from dust .
the chemical network of cloudy covers all atomic species from h to zn , as well as a multitude of molecular species .
each elemental abundance can be specified individually , giving us the ability to model the cooling from a gas with any composition . since cloudy permits the use of virtually any input spectrum
, we are able to create cooling data that is suitable for any radiation environment . instead of allowing the code to cycle through temperatures
until converging on a thermodynamic equilibrium solution , we use the ` constant temperature ` command to fix the temperature externally , allowing us to utilize cloudy s sophisticated machinery to calculate cooling rates out of thermal equilibrium . in this manner
, we create a grid of heating and cooling values as a function of temperature , gas density , chemical composition , and incident spectrum .
the cooling rates presented in this work were created using version 07.02.01 of the cloudy software . to automate the process of data production and organization , we have created a code , called roco ( recursively organized cloudy output . )
roco uses a recursive algorithm to process user - specified loop parameters , making it possible to create data - grids of any dimension .
commands that are to be issued to cloudy are given to the roco code in either one of two formats - loop commands with a set of parameters through which the code will iterate , and constant commands that are to be issued with the same value during each iteration over the loop commands .
since most uses of cloudy involve the creation of large grids of models constructed by looping over a set of input parameters , the capabilities of roco give it the potential to be useful to a broader community of cloudy users than just those who would use it to create the cooling tables discussed here . to this end
, roco is structured in such a way that the post - cloudy data analysis routines can be easily interchanged to suit the needs of different users .
the code features an extra running mode that simply runs cloudy over the specified parameter - space with no further processing of the data , as well as a template designed to help users create new running modes suited to their specific needs .
roco also has the ability to run multiple instances of cloudy simultaneously , greatly reducing runtime .
the parallel feature works well on individual machines with multiple processors , as well as beowulf clusters using the mpi ( message passing interface ) framework .
a copy of the roco code will be made available upon request to the authors . in figure
[ fig : cooling1 ] , we display the resulting cooling function for gas with n@xmath9 = 1 @xmath3 at metallicities , from z = 0 ( metal - free ) to 10 z@xmath4 , over the temperature range , 50 @xmath0 t @xmath0 10@xmath1 k. for these cooling rates , we use the ` coronal equilibrium ` command in cloudy to simulate an environment free of radiation , where all ionization is collisional .
we also neglect the cooling from h@xmath7 , so as to better illustrate the cooling contribution from metals at temperatures less than 10@xmath8 k. we accomplish this by issuing the cloudy command , ` no h2 molecule ` .
we implement our metal cooling method in the eulerian adaptive mesh refinement hydrodynamic / n - body code , enzo@xcite .
when a simulation is initialized , enzo reads in the cloudy / roco data - grid , storing the heating and cooling values as functions of temperature , h number density , and any other parameters , such as spectral intensity , depending on the nature of the simulation .
the heating , @xmath10 , and cooling , @xmath11 , are stored with code units corresponding to [ ergs s@xmath12 @xmath13 . during the simulation , enzo stores the mass density and internal energy for each grid cell in the box . at each hydrodynamic time - step ,
the radiative cooling solver cools the gas by lowering the internal energy via a simple euler update , @xmath14 where @xmath15 denotes the internal energy of the grid cell with ( x , y , z ) coordinates , ( i , j , k ) , at the ( n+1)th time - step , @xmath16 is the cooling rate in code units corresponding to [ ergs s@xmath12 ] , and @xmath18 is the adopted time - step . for every hydrodynamic time - step , the code subcycles through equation [ eqn : cool ] , selecting from three possible time - steps , until one full hydrodynamic time - step has been completed .
the time - step , @xmath18 , in equation [ eqn : cool ] , adopts the minimum of the following three values : ( 1 ) half of the hydrodynamic time - step , ( 2 ) 10% of the cooling time , ( u/@xmath16 ) , or ( 3 ) the time remaining to have integrated over one full hydrodynamic time - step .
the euler update is the standard method used for updating the internal energy in all of the established cooling routines in the enzo code .
since the cooling rate is such a nonmonotonic function of temperature , this approach of using substeps with an explicit integration method has been found to yield the best combination of accuracy and speed @xcite . the internal energy and mass density for each grid cell
are converted to temperature and number density and the heating and cooling values are calculated by linearly interpolating over the cloudy / roco data - grid .
the change in internal energy from the cloudy / roco cooling rates is expressed as @xmath19 where n@xmath9 is the h number density .
we implement two distinct versions of the method described above . in the first and simplest version
, the cooling is calculated solely from the cloudy / roco data , as in equation [ eqn : dudt ] .
the total change in internal energy is @xmath20 when converting the internal energy to temperature , it is necessary to know the value of the mean molecular weight , @xmath21 . in this implementation ,
we assume @xmath21 to be a constant with the value 1.22 . for high temperatures , t @xmath22 10@xmath8 k
, this method is sufficient for providing accurate gas cooling within hydrodynamic simulations .
this implementation is not suitable , however , when t @xmath23 10@xmath8 k and the formation of h@xmath7 becomes important . disregarding formation on grain surfaces and three - body formation , h@xmath7 primarily forms through the following channels : + the h@xmath24 channel , @xmath25 and the h@xmath26 channel , @xmath27 when a significant electron fraction exists , these reactions proceed to form h@xmath7 very quickly , with the h@xmath24 channel typically dominating , except in the very high redshift universe ( z @xmath28 100 ) , where h@xmath24 is readily destroyed by the cmb @xcite .
recently , @xcite have suggested that formation of h@xmath26 via @xmath29 is responsible for more h@xmath7 than the h@xmath26 channel in equation [ eqn : h2plus ] . if , however , the ionization fraction is low , h@xmath7 forms very slowly , with equilibrium timescales that can exceed the current age of the universe .
the consequence is that h@xmath7 formation is so sensitive to the thermal history of the gas that h@xmath7 fractions can not be known without explicitly following the non - equilibrium chemistry during the simulation .
we find this to be the case when using cloudy to compute the cooling rate from h@xmath7
. in searching for ionization equilibrium , cloudy integrates over timescales that are unphysically long , leading to an overcalculation of the h@xmath7 fraction , producing cooling rates that are too high .
we solve this problem in our second implementation by first removing the h@xmath7 molecule from cloudy s chemical network with the ` no h2 molecule ` command .
we , then , run two cloudy / roco data - grids : one with the full set of elements and the other with h and he only .
we obtain a metals - only data - grid by subtracting the metal - free data - grid from the complete data - grid . using the established h / he network in enzo @xcite , we follow the non - quilibrium fractions of h , h@xmath30 , h@xmath24 , h@xmath7 , h@xmath26 , he , he@xmath30 , he@xmath31 , and e@xmath24 , and directly calculate the associated atomic @xcite and molecular cooling rates .
we provide the option to use the h@xmath7 cooling rates of @xcite , which are obsolete , but allow a means of comparison to older simulations .
we also include the cooling , or heating , from electrons scattering off the cmb as @xmath32 where t@xmath33 = 2.7 ( 1 + z ) k @xcite .
we prevent the metals from cooling the gas below the cmb temperature by subtracting the metal cooling rate at t = t@xmath33 , as in @xcite .
including the cmb explicitly in cosmological simulations would require adding an extra dimension in the cloudy / roco data - grid to account for the evolution of the cmb with redshift . while this is certainly possible
, interpolating over an extra dimension to calculate the cooling during the simulation would be significantly slower than the approximation described above . to test the validity of this approximation of the cmb temperature - floor ,
we create a cloudy / roco metals - only data - grid , explicitly including the cmb . at low densities
( n @xmath34 1 @xmath3 ) , the values of ( @xmath11 - @xmath10 ) from the data with the cmb included differ from the values of ( @xmath11 - @xmath35 ) from the data without the cmb by a factor of roughly 2 near t@xmath33 . at higher temperatures and densities ,
the two values are nearly identical .
the total rate of energy loss applied to the simulation gas in the second implementation is @xmath36 where @xmath37 is the total atomic and molecular cooling from the h / he network , and @xmath38 is the metals - only cooling rate taken from the cloudy / roco data in the manner described above .
the value of @xmath21 is calculated directly from the h / he species fractions , neglecting any addition from the metals . in low - metallicity gases ,
this approach is reasonable , as the increase in @xmath21 from the metals only reaches @xmath3410@xmath6 for z = 10@xmath39 z@xmath4 .
since the cloudy / roco data - grids also store the values of @xmath21 for each point , this can be added to the value calculated without the heavy elements when the metallicity is very high . in figure
[ fig : cooling2 ] , we display low - temperature cooling functions for gases with varying density and metallicity , constructed with the third implementation of the metal cooling method . to produce the data for figure [ fig : cooling2 ] , we set up an unphysical , two - dimensional grid in enzo that varies smoothly over density and temperature .
we iterate the reaction network for a time equivalent to that between z = 99 and 20 ( @xmath34160 myr ) , with hydrodynamics disabled , then compute the cooling with the third implementation of our metal cooling method , using the h@xmath7 cooling rates of . since the first stars are predicted to form at the centers of @xmath3410@xmath40 m@xmath4 dark matter halos at z @xmath34 20 @xcite , integrating the rate equations over this time interval places each of the species in the relative abundances in which they would be found during the epoch of first - star formation .
hence , figure [ fig : cooling2 ] provides a direct comparison of the cooling rate of the gas that formed the first and successive generations of stars .
the h@xmath7 fractions used to create the cooling rates for figure [ fig : cooling2 ] result from integrating the h / he chemical network for the period of time between z = 99 and 20 . as such ,
figure [ fig : cooling2 ] should not be used as a general cooling function , except in the context mentioned above .
however , figures [ fig : cocooln3_z-2][fig : cool_dust ] do not include the cooling from h@xmath7 and may , therefore , be used as general purpose cooling functions when the cooling from h@xmath7 is not needed .
much attention has been given recently to the role of the first heavy elements in transitioning from the singular , high - mass mode of star formation of the first stars to the mode producing stars with a salpeter initial mass function ( imf ) .
analytical studies by @xcite and @xcite have focused on the contributions of individual elements toward triggering fragmentation in star - forming clouds . @xcite
suggest c and o to be the dominant coolants in low - metallicity gas , in the presence of an h@xmath7 dissociating uv background created by the first stars @xcite . by calculating the cooling rate necessary to equate the cooling time to the free - fall time at n = 10@xmath8 @xmath3 and t = 200 k , the point where h@xmath7 cooling becomes inefficient @xcite , @xcite predict individual critical abundances of c and o to be [ c / h]@xmath41 @xmath42 -3.5 and [ o / h]@xmath41 @xmath42 -3.05 , where [ a / h ] = log@xmath43(n@xmath44/n@xmath9 ) - log@xmath43(n@xmath44/n@xmath9)@xmath4 .
@xcite consider the cooling from fe and si , in addition to c and o , and take into account the density dependence of metal cooling .
in doing so , they find that the critical abundance of each element varies with density , reaching a minimum at a critical density that is different in each case .
they also note that different elements dominate different density and temperature regimes . in figs .
[ fig : cocooln3_z-2][fig : fessicooln3_z0 ] , we plot the individual cooling contributions for number density , n@xmath9 = 10@xmath45 @xmath3 and metallicities , z = 10@xmath39 z@xmath4 and 1 z@xmath4 . in each case
, we create a set of cooling data with the full complement of elemental species , from h through zn , neglecting h@xmath7 .
we plot only the coolants whose contributions reach , at least , 10@xmath46 of the total cooling within the temperature range , 10 k @xmath0 t @xmath0 5000 k. since the cooling data were made assuming no incident ionizing radiation , as described in [ sec : cloudy ] , we observe the dominant c transitions to be from ci , instead of cii , as in @xcite and @xcite . for z = 10@xmath39 z@xmath4 ,
the most important coolants are the fine structure transtions at 369.7 @xmath21 m and 609.2 @xmath21 m from ci and at 63.2 @xmath21 m from oi . at higher temperatures ( t @xmath5 100 k ) , the siii transition at 34.8 @xmath21 m becomes important as well . at higher metallicities
, co replaces atomic c ( fig .
[ fig : cocooln3_z0 ] ) and emission from sii at 129.6 @xmath21 m becomes dominant ( fig .
[ fig : fessicooln3_z0 ] ) .
fe cooling is relatively unimportant up to n@xmath9 = 10@xmath47 @xmath3 , but is completely dominant for t @xmath5 200 k by n@xmath9 = 10@xmath48 @xmath3 , with oi strongest slightly below that temperature , and co the most important below 50 k ( fig .
[ fig : cooln9_z0 ] ) .
in addition to the elements covered by @xcite , we note that cooling from neutral s reaches roughly the 10% level at n@xmath9 = 10@xmath45 @xmath3 .
the [ si ] 25.2 @xmath21 m transition peaks at 40% of the total cooling at n@xmath9 = 10@xmath49 @xmath3 and t @xmath34 1000 k. the only other coolant that contributes at the level of at least 10@xmath46 of the total cooling is the 60.6 @xmath21 m transition of [ pii ] at n@xmath9 = 10@xmath47 @xmath3 .
the number of coolants that reach 10@xmath46 of the total grows quickly with density .
we observe 23 distinct coolants contributing at that level at n@xmath9 = 10@xmath40 @xmath3 , and 32 by 10@xmath48 @xmath3 .
if we lower the threshhold to 10@xmath50 , there are a total of 84 coolants at n@xmath9 = 10@xmath48 @xmath3 , illustrating the strength of cloudy and our cooling method .
recently , studies have suggested that dust cooling at high densities can trigger fragmentation for metallicities as low as 10@xmath50 @xmath51 @xcite .
@xcite have claimed that between 15 and 30% of the mass of the progenitor of a pair - instability supernova is converted in dust .
however , observations of the crab nebula @xcite and the cassiopeia a supernova remnant @xcite have returned little or no signs of dust , suggesting that type ii supernova may not actually be large dust producers . given the controversy surrounding the existence of dust grains in the formation environments of second - generation stars , we do not include the cooling from dust in the analysis , but leave it for a separate work .
to provide an example of the effect of dust on the cooling rate , we run a simple model including dust in cloudy .
we use a model designed to simulate the dust within the ism , using the cloudy command , ` grains ism ` .
the dust physics used in cloudy is described in detail by @xcite .
the ism dust model in cloudy consists of both graphite and silicates with sizes ranging from 5 @xmath52 10@xmath46 @xmath21 m to 0.25 @xmath21 m and a power - law size - distribution with a power - law index of -3.5 . for solar metallicity ,
the total grain abundances per h are 10@xmath53 for graphite and 10@xmath54 for silicates . in fig .
[ fig : cool_dust ] , we plot the cooling rate from metals at n@xmath9 = 10@xmath48 @xmath3 , with and without dust grains , for metallicities , z = 10@xmath50 z@xmath4 , 10@xmath6 z@xmath4 , and 10@xmath39 z@xmath4 . since it is inappropriate to think of the dust and gas - phase metal abundances as independent , we directly scale the dust abundances with the gas - phase metal abundancs . for number densities
lower than 10@xmath48 @xmath3 , the additional cooling from dust is almost negligible . at higher densities , dust becomes more important .
if dust is , in fact , produced in the supernovae of the first stars , it is likely to be as important as previous studies have claimed . in order to study the ability of a collapsing gas - cloud to fragment , we first identify the regions of density and temperature where the classical fragmentation criterion , t@xmath55 @xmath56 t@xmath57 @xcite , is met , with the dynamical time expressed as @xmath58 we limit this analysis to solar abundance patterns .
this represents the first step of an incremental approach to studying the general criteria that lead to fragmentation in collapsing clouds .
the use of solar abundance patterns will allow us to begin to quantify the chemical abundance required for fragmentation . in a following paper
, we will study nonsolar abundance patterns motivated by predicted yields of primordial supernovae . in this future work
, we will vary the abundances of individual elements , which will require the exploration of a much larger parameter - space .
we create a cloudy / roco data - grid with the following parameters : 50 k @xmath0 t @xmath0 1000 k with @xmath59log(t ) @xmath42 0.012 ( 100 points ) , 1 @xmath3 @xmath0 n@xmath9 @xmath0 10@xmath2 @xmath3 with @xmath59log(n@xmath9 ) = 0.1 , and 10@xmath50 z@xmath4 @xmath0 z @xmath0 10@xmath39 z@xmath4 with @xmath59log(z ) = 1 .
we , then , follow the same procedure used to produce fig .
[ fig : cooling2 ] , as described in
[ sec : enzo ] . in this section
, we omit the decrease in the cooling rate caused by compton heating on the cmb . in fig . [
fig : tdtc ] , we plot the log of the ratio of the dynamical time to the cooling time for each of the metallicities in the data - grid and for the metal - free case .
a cloud is able to fragment when log@xmath43(t@xmath57/t@xmath55 ) @xmath28 0 . as expected , there is no density at which metal - free gas can fragment for t @xmath56 200 k. as the metallicity increases , the value of log@xmath43(t@xmath57/t@xmath55 ) slowly increases , first in the low - temperature regime , where h@xmath7 cooling is inefficient , so even a small amount of metals has an effect . for gas with n@xmath9 = 10@xmath8 @xmath3 at t
= 200 k , the fragmentation criterion is nearly met by z = 10@xmath60 z@xmath4 and well satisfied one order of magnitude higher in metallicity .
once the metallicity reaches 10@xmath39 z@xmath4 , the entire parameter - space is fragmentable . at high densities , however , fragmentation will be curtailed as the cloud becomes optically thick to its own radiation @xcite . as was also reported by @xcite , the efficiency of the metal cooling peaks at n@xmath9 @xmath34 10@xmath40 @xmath3 , significantly lowering the critical metallicity required for fragmentation .
the addition of metals to a gas also has the potential to trigger thermal instabilities during cloud - collapse . as in @xcite @xcite
, we define a parcel of gas losing energy at a rate , l , to be thermally unstable if @xmath61 where l is expressed in terms of the cooling rate , @xmath11 , as @xmath62 the cooling rate , @xmath11 , can be locally approximated by a power - law in both temperature and density as @xmath63 the partial derivatives of eqn .
[ eqn : thermal ] become @xmath64 and @xmath65 the thermal instability criterion simplifies to @xmath66 in fig . [
fig : ctstability ] we plot the value of the instability parameter , ( @xmath67 - @xmath68 ) , for the same cooling data used for fig .
[ fig : tdtc ] . for metal - free gas ( fig .
[ fig : ctstability ] , top - left ) , the instability parameter is greater than 3 over nearly the entire parameter space , and always greater than 4 at high densities . @xcite and @xcite both arrive at the same conclusion using this analysis for metal - free gas , with the added assumption that the cooling function is independent of density . as the metallicity increases , a region of thermal instability forms at low density and temperature . when the metallicity reaches 10@xmath6 z@xmath4 ( fig .
[ fig : ctstability ] , middle - right ) , a second unstable region exists for 10@xmath45 @xmath3 @xmath23 n@xmath9 @xmath23 10@xmath40 @xmath3 , at a temperature of a few hundred k. the second unstable region coincides with the increase in cooling efficiency to its maximum value , illustrated in fig .
[ fig : tdtc ] .
fragmentation is more likely to occur when the gas is both thermally unstable , and can cool faster than the dynamical time .
we indicate the regions where both the fragmentation and thermal instability criteria are met in white in fig .
[ fig : doublestability ] . no doubly unstable realm exists for metallicities , z @xmath0 10@xmath60 z@xmath4 .
the cmb creates a temperature floor , below which gas can not cool radiatively .
we study the influence of the cmb on the evolution of star - forming gas by applying a cmb floor at z = 20 in the manner described in [ sec : enzo ] to the cooling data - grid used in
[ sec : thermal ] .
the cmb affects the cooling properties of the gas in two ways .
the first is by increasing the cooling time at temperatures near the cmb temperature to greater than the dynamical time so that the fragmentation criterion is no longer satisfied .
the second is by increasing the value of @xmath67 , from equation [ eqn : thermal2 ] , at low temperatures , making the gas thermally stable . in fig .
[ fig : doublestabilitycmb ] , we illustrate the influence of the cmb on the doubly unstable regions , shown previously .
the unstable region that existed in the low - density , low - temperature regime is completely eliminated .
there remains , however , a small area of instability for metallicities as low as 10@xmath6 z@xmath4 .
we have introduced a new method for including the radiative cooling from metals in large , three - dimensional hydrodynamic simulations .
in addition to its implementation in the amr code , enzo , this method has also been used by @xcite in numerical simulations with the smoothed particle hydrodynamics ( sph ) code , gadget @xcite .
our technique takes advantage of the extremely complex chemical reaction network of the preexisting radiative transfer code , cloudy , which includes a full elemental coverage from h to zn , along with a variety of molecular species and dust . with the singular assumption of ionization equilibrium for the heavy elements , we are able to precalculate cooling rates for gases with any chemical abundance in all manners of radiation environments over a temperature range of 10 to 10@xmath1 k. with cooling rates computed in advance , we eliminate the barrier that has classically prevented large chemical models from being incorporated into three - dimensional numerical simulations .
because our cooling scheme is valid over such a large range of density and temperature , and features so many coolants , it can be applied to a huge variety of astrophysical problems , such as the evolution of the ism and igm , normal star formation , planetary nebulae , accretion disks , and protoplanetary disks .
one advantage of the large chemical network of cloudy is that we are able to determine the dominant coolants from a complete sample of atomic species up to an atomic number of 30 .
fine - structure transitions of c and o are the greatest contributors to the cooling up to number densities of about 10@xmath40 @xmath3 , where fe cooling becomes significant and c is marginalized .
the importance of these three elements , along with si , in triggering the formation of the first low - mass stars has been studied in great detail by @xcite .
the cooling models we have constructed using solar abundance patterns reveal s cooling to be important for 10@xmath45 @xmath3 @xmath23 n@xmath9 @xmath23 10@xmath48 @xmath3 .
s is produced only slightly less than si in type ii @xcite and pair - instability supernovae @xcite , and should be taken into account when considering the metals responsible for the transition from population iii to population ii star formation .
the ability to specify individual abundances in our cooling method makes it straightforward to simulate the evolution of gas with non - solar abundance patterns .
in addition to elevating the cooling rate of a gas to satisfy the classical fragmentation criterion , metals also increase the potential for fragmentation by creating thermal instabilities .
we have identified regions in temperature and density in which both the classical fragmentation and thermal instability criteria are met to be the physical conditions most likely to see fragmentation occur .
we observe these doubly unstable regions to exist for metallicities as low as 10@xmath6 z@xmath4 .
if fragmentation can not occur outside these regions , then the fate of a star - forming gas - cloud will be determined by the path taken through density - temperature space as it collapses . if we consider the doubly unstable regions in fig .
[ fig : doublestability ] , appropriate for star - formation in current epoch , there will almost certainly be a period of double instability when z @xmath5 10@xmath46 z@xmath4 .
interestingly , the density - temperature tracks shown in fig . 1 of @xcite indicate that gas with z = 10@xmath6 z@xmath4 will pass right through the stable region that separates two instabilities .
@xcite also find that star formation at that metallicity only produces high - mass fragments . at high redshift ,
the cmb significantly reduces the size of the doubly unstable regions .
thermal instabilities , though , are extremely sensitive to the slope of the cooling rate as a function of density and temperature .
every element has distinct cooling properties , and will , therefore , produce different thermal instabilities . as such , the key to uncovering the nature of the first population ii stars will be in the determination of the mass function of their precursors .
in papers to follow , we will extend our study to gases with non - solar abundance patterns .
we will explore thermal and double instabilities created by individual elements , as well as abundance patterns produced in population iii supernovae .
future studies will also investigate the effects of background radiation on the evolution of star - forming gas .
the final word , however , will only come from three - dimensional , numerical simulations .
the simulations by @xcite , employing the methods described here with solar abundance patterns , have confirmed that fragmentation occurs for metallicities , z @xmath5 10@xmath46 z@xmath4 .
we intend to pair all future predictions made from analysis of thermal instabilities with full numerical simulations .
although dust physics has been implemented in cloudy , we currently treat only metals in the gas - phase in our analysis . in the future
, we will study the effects of dust cooling in more detail .
we will also couple the dust chemistry to the h / he chemical network in enzo , so as to properly model the formation of h@xmath7 on grain surfaces .
a great strength of this method is the use of the ever - expanding chemical network of cloudy .
as more physical processes are incorporated into the cloudy software , the utility of this method will increase as well .
one major constraint of this work , however , is that its validity is confined to the optically thin limit .
the approximations made here break down at opacities of order unity . for higher opacities
, our method provides a core module for flux - limited diffusion schemes .
complex geometries introduce problems of self - heating and shadowing , which will require full , three - dimensional radiative transfer .
based on observations made with the nasa / esa hubble space telescope , obtained from the data archive at the space telescope institute .
stsci is operated by the association of universities for research in astronomy , inc . under the nasa contract
nas 5 - 26555 .
this research was supported by hubble space telescope theory grant hst - ar-10978.01 , a grant from nasa s atp nng04gu99 g , and nsf career award ast-0239709 from the national science foundation .
we are grateful to an anonymous referee for sharing many insightful comments and suggestions .
we also thank brian oshea and mike norman for useful discussions and support .
ss also thanks kipac and stanford university for their hospitality .
t. , smith b. d. , eracleous m. , sigurdsson s. , 2006 , in laser interferometer space antenna : 6th international lisa symposium vol .
873 of american institute of physics conference series , electromagnetic signatures of massive black hole binaries .
pp 257263 g. , norman m. l. , 1997 , in chrisochoides n. , ed . , workshop on structured adaptive mech refinement grid methods i m a volumes in mathematics no .
117 , a hybrid amr application for cosmology and astrophysics .
springer - verlag a. , meynet g. , hirschi r. , 2005 , in barnes iii t. g. , bash f. n. , eds , cosmic abundances as records of stellar evolution and nucleosynthesis vol . 336 of astronomical society of the pacific conference series , chemical abundances and yields from massive stars .
pp 79+ b. w. , g. b. , bordner j. , norman m. l. , abel t. , harknes r. , kritsuk a. , 2004 , in plewa t. , linde t. , weirs v. g. , eds , adaptive mesh refinement - theory and applications vol . 41 of lecture notes in computational science and engineering , proceedings of the chicago workshop on adaptive mesh refinement | the addition of metals to any gas can significantly alter its evolution by increasing the rate of radiative cooling . in star - forming environments , enhanced cooling can potentially lead to fragmentation and the formation of low - mass stars , where metal - free gas - clouds have been shown not to fragment . adding metal cooling to numerical simulations has traditionally required a choice between speed and accuracy .
we introduce a method that uses the sophisticated chemical network of the photoionization software , cloudy , to include radiative cooling from a complete set of metals up to atomic number 30 ( zn ) that can be used with large - scale three - dimensional hydrodynamic simulations .
our method is valid over an extremely large temperature range ( 10 k @xmath0 t @xmath0 10@xmath1 k ) , up to hydrogen number densities of 10@xmath2 @xmath3 . at this density
, a sphere of 1 m@xmath4 has a radius of roughly 40 au .
we implement our method in the adaptive mesh refinement ( amr ) hydrodynamic / n - body code , enzo . using cooling rates generated with this method , we study the physical conditions that led to the transition from population iii to population ii star formation . while c , o , fe , and si have been previously shown to make the strongest contribution to the cooling in low - metallicity gas , we find that up to 40% of the metal cooling comes from fine - structure emission by s , when solar abundance patterns are present . at metallicities , z @xmath5 10@xmath6 z@xmath4 ,
regions of density and temperature exist where gas is both thermally unstable and has a cooling time less than its dynamical time .
we identify these doubly unstable regions as the most inducive to fragmentation . at high redshifts ,
the cosmic microwave background inhibits efficient cooling at low temperatures and , thus , reduces the size of the doubly unstable regions , making fragmentation more difficult .
[ firstpage ] stars : formation methods : numerical |
universality plays an important role in physics .
the interest to it is inspired by the striking similarity in behavior near the critical point among systems that are otherwise quite different in nature .
for example , various substances , which exhibit liquid - gas phase transition , near the critical point obey the universal law @xmath11 . here
@xmath12 denote the density of gas and critical density respectively , @xmath13 are temperature and critical temperature , @xmath14 is a constant and @xmath15 is the so - called critical exponent @xcite . amazingly , the value of @xmath16 is the same for many substances , which are completely different on the atomic level .
similar law with the same value of the critical exponent holds true for magnetization in ferromagnets as a function of temperature .
another example of universality is found in the ground state energy of the bose gas as a function of density . in the low density limit
it approaches an expression , which depends only on the scattering length but not on the overall form of pair interaction @xcite .
small quantum systems also exhibit universal features @xcite .
one example of universality in the two - particle case concerns the behavior of the energy depending on the coupling constant near the threshold .
suppose that @xmath8 is the energy of an isolated non - degenerate state of the hamiltonian @xmath17 in 3-dimensional space and @xmath18 for @xmath19 .
then universally for @xmath20 near @xmath5 one has @xmath21h.t . or @xmath22h.t .
depending on whether zero is an eigenvalue of @xmath23 or not , see @xcite . *
universal * in this context means that this behavior up to a constant is true for all short range interactions independently of their form .
`` h.t . ''
is the shorthand notation for `` higher terms '' and @xmath24h.t . for @xmath25
always implies that @xmath26 . if two particles are set into an n - dimensional space the scenario depends on the space dimension @xcite : for example , in 2-dimensional flatland the energy of the ground state energy can approach zero exponentially fast @xmath27h.t .
, and in 4 dimensions the ground state energy approaches zero very slow , namely , @xmath28h.t .. for a full account of possible scenarios see table i in @xcite .
another universality associated with 2-body system in 3-dimensional space relates to the wave function near the threshold . if the energy of a non - degenerate bound state near the threshold satisfies @xmath29h.t .
then the wave function of this bound state @xmath30 approaches spherically symmetric expression @xcite @xmath31 ) holds irrespectively of the form of the pair potential . in 3-particle systems
the notorious example of universality is the efimov effect .
efimov s striking and counterintuitive prediction @xcite was that just by tuning coupling constants of the short - range interactions in the 3-body system one can bind an infinite number of levels , even though the two - body subsystems bind none .
the infinitude of bound states was shown rigorously by yafaev in @xcite .
basing on the yafaev s method sobolev @xcite has proved that @xmath32 where @xmath33 is the number of bound states with the energy less than @xmath34 and @xmath35 is the universal positive constant , which depends only on masses .
let us remark that in physics@xcite it is generally conjectured that @xmath36 where @xmath37 is the energy of the @xmath38-th efimov level and @xmath39 . in physics this
is termed as universal scaling of efimov levels , see @xcite for the mathematical discussion of ( [ 1.14;10 ] ) .
efmiov s prediction was later confirmed experimentally in ultracold gases @xcite .
the so - called 4-body universality @xcite holds only approximately @xcite and the question of finite range corrections is still being debated @xcite . for further discussion it is useful to introduce the following mathematical notations . for an operator @xmath14 acting on a hilbert space @xmath40 , @xmath41 and @xmath42
denote the domain , the spectrum , and the essential spectrum of @xmath14 respectively @xcite .
@xmath43 means that @xmath44 for all @xmath45 , while @xmath46 means that there exists @xmath47 such that @xmath48 .
@xmath49 denotes the set of bounded linear operators on the hilbert space @xmath50 . for an interval @xmath51 the function
@xmath52 is such that @xmath53 if @xmath54 and @xmath55 otherwise .
recently a new type of universality in 3-body systems has been discovered in @xcite .
consider the hamiltonian of the 3particle system in @xmath56 @xmath57 where @xmath58 is the kinetic energy operator with the center of mass removed , @xmath2 is the coupling constant and none of the particle pairs has negative energy bound states .
all particles are supposed to have a finite mass . the pair
interactions @xmath59 are operators of multiplication by real @xmath60 and @xmath61 are particle position vectors . for pair potentials we require like in @xcite that @xmath62 < \infty , \label{restr}\ ] ] where @xmath63 is a fixed constant , and @xmath64 where @xmath65 are constants .
we shall assume that the interaction between the particles @xmath1 is tuned to make them have a zero energy resonance and no negative energy bound states @xcite .
this implies that in the absence of particle 3 the particles @xmath1 are `` almost '' bound , that is a bound state with negative energy appears if and only if the interaction @xmath66 is strengthened by a negligibly small amount . in mathematical terms
this can be expressed as follows @xmath67 let @xmath68 be the values of the coupling constants such that @xmath69 and @xmath70 are at critical coupling in the sense of def . 1 in @xcite .
this means that @xmath71 and @xmath72 for all @xmath73 and @xmath74 .
let us set @xmath75 .
\label{tildelambda}\ ] ] we shall always assume that the coupling constant in ( [ hami ] ) satisfies the inequality @xmath76 . in other words , the coupling constant takes the values for which the particle pairs @xmath3 and @xmath4 have neither zero energy resonances nor bound states with negative energy . under these conditions @xmath77 and @xmath78 has a finite number of bound states with negative energy as proved in @xcite . let @xmath5 be the value of the coupling constant such that @xmath79 but @xmath80 for all @xmath81 . in sec . 6 in @xcite
it is proved that @xmath82 . by the hvz theorem
( see @xcite , vol . 4 and @xcite ) for @xmath83 @xmath84 where @xmath85 is the ground state energy and @xmath86 . by theorems 1 , 3 in @xcite and
theorem 2 in @xcite @xmath87 for @xmath88 but zero is not an eigenvalue of @xmath89 .
moreover , @xmath90 for @xmath88 totally spreads , that is @xmath91 in ( [ 14.1;1 ] ) @xmath92 are jacobi coordinates in the 3-body problem , which are shown in fig .
[ fig:1 ] ( left ) .
they are defined as @xmath93 \nonumber , \end{gathered}\ ] ] where @xmath94 denote particle masses , @xmath95 and @xmath96 . in @xcite
it was proved that @xmath90 for @xmath97 approaches in norm a universal expression , namely , latexmath:[\[\label{3bh3 } \psi_\lambda \to \frac{1}{\sqrt 2 \pi^{3/2 } |\ln in the limit the wave function @xmath90 describes the state , in which average distances between all three particles go to infinity .
( this is partly the reason why the short range details of pair interactions becomes unimportant ) . by analogy with the 2-particle case
it is natural to assume that @xmath8 would exhibit universal behavior near @xmath5 . in this paper
we shall prove theorem [ th : main ] , which states that it is indeed so and universally one has @xmath9^{-1 }
+ $ ] h.t . , where @xmath99
it is important to stress that this result does not follow directly from ( [ 3bh3 ] ) , if one merely tries to substitute ( [ 3bh3 ] ) into the equation @xmath100 .
this is due to the error terms , which in spite of going to zero in norm may affect the resulting average , for a detailed explanation see remark [ newremark ] in the next section .
the obtained behavior of @xmath8 remarkably mimics that of the ground state energy of 2 particles in 4-dimensional space .
the experimental observation of this type of universality can possibly be obtained in ultracold gas mixtures , see @xcite for discussion .
when the pair interaction @xmath66 is not tuned there are 3 types of possible asymptotic behavior of @xmath8 , which are listed in theorem [ th : main1a ] .
points in the direction from the center of mass of the particles @xmath1 to particle @xmath101 .
@xmath102 points in the direction from the center of mass of the particles @xmath3 to particle @xmath103 .
the picture shows only directions of the vectors , the scales are set in order to ensure that @xmath104 and @xmath105 holds . ]
the hamiltonian @xmath78 in ( [ hami ] ) is self
adjoint on @xmath106 , where @xmath107 denotes the corresponding sobolev space @xcite .
the pair interaction between particles @xmath1 is tuned so that they have a zero energy resonance , that is eqs .
( [ 14.1;2])-([14.1;3 ] ) hold .
our aim in this paper is to prove [ th : main ] suppose that @xmath108 , then for @xmath88 one has @xmath109 where @xmath110 is a finite constant . before we proceed with the proof let us remark that 1 ) eq .
( [ main ] ) is universal , meaning that up to a constant it does not depend on the details of pair interaction ; 2 ) the function @xmath8 at @xmath111 can not be taylor expanded in powers of @xmath112 for any @xmath113 ; 3 ) the three - body ground state energy in the 3-dimensional case has the same behavior near @xmath5 as the 2-body ground state energy in the 4-dimensional case ( we do not have an explanation for this finding ) ; 4 ) the constant @xmath114 depends on pair interactions and can be expressed through zero energy solutions of birman - schwinger operators ; 5 ) the method of the proof is different from @xcite : the method in @xcite , which uses the low energy expansions of the birman - schwinger operator , is not applicable here .
the proof below hinges on theorem [ th : main2 ] in section [ sec:3 ] , whose technical proof is heavily based on the results and methods in @xcite . for @xmath115
there exists @xmath116 , @xmath117 such that @xmath118 , besides we can assume that @xmath119 because it is the ground state .
@xmath120 is smooth and monotone increasing on @xmath121 .
using perturbation theory @xcite we obtain @xmath122 by theorem [ th : main2 ] for @xmath20 close enough to @xmath5 there exists a constant @xmath123 such that @xmath124 for any given @xmath73 .
the last inequality can be equivalently rewritten as @xmath125 integrating ( [ 24.1:3 ] ) we obtain @xmath126 let us set @xmath127 where @xmath128 . from ( [ 24.1:4 ] ) we get @xmath129 \geq ( c_0 - \varepsilon ) .\ ] ] using that @xmath18 from ( [ 24.1:5 ] ) we conclude that @xmath130 . again substituting ( [ 24.1:6 ] ) into ( [ 24.1:5 ] ) we obtain @xmath131 \geq ( c_0 - \varepsilon)\ ] ]
the term in square brackets in the last inequality goes to zero for @xmath7 .
thus for @xmath20 close to @xmath5 we have @xmath132 now ( [ 24.1:9 ] ) follows from ( [ 24.1:6 ] ) , ( [ 24.1:8 ] ) since @xmath133 is arbitrarily small .
[ newremark ] we should explicitly warn against the direct `` physicist s approach '' , when one substitutes in ( [ 24.1:9 ] ) instead of @xmath90 the rhs of ( [ 3bh3 ] ) . eq .
( [ 3bh3 ] ) defines @xmath90 up to error terms , which go to zero in norm when the energy goes to zero .
there is otherwise no control of these error terms and one can not exclude the situation , where , for example , @xmath134 and @xmath135 equals the rhs of ( [ 3bh3 ] ) .
if one substitutes ( [ yasego ] ) into ( [ 24.1:9 ] ) one would find that the resulting @xmath8 would be very different from the form stated in theorem [ th : main ] .
theorem [ th : main ] considers the case when the particles @xmath1 have a zero energy resonance .
now let us consider a more general situation and assume that the 3-particle system is described by the hamiltonian ( [ hami ] ) , where for simplicity we require that @xmath136 are bounded and have a compact support .
we still require that @xmath137 , _
i.e. _ @xmath20 takes the values for which the subsystems @xmath3 and @xmath4 have no bound states with negative energy and no zero energy resonances .
however , we do not impose restrictions on the spectrum of the particles @xmath1 , which means that this pair determines the energy of the dissociation threshold @xmath138 , that is @xmath139 the critical coupling constant @xmath5 is the value of @xmath20 for which the 3-body bound state , whose energy lies below @xmath138 , is about to be formed . mathematically speaking @xmath140 by the methods similar to @xcite
one can prove that @xmath141 .
[ th : main1a ] one can distinguish 3 cases : ( a ) the pair @xmath1 has no negative energy bound states and no zero energy resonance ; ( b ) the pair @xmath1 has no negative energy bound states but has a zero energy resonance ; ( c ) the pair @xmath1 has at least one bound state with negative energy .
suppose that @xmath108 , then for @xmath88 in each case one has @xmath142^{-1}+ \mbox{\textnormal{h.t . } } \nonumber\\ & ( c ) \quad
\quad e(\lambda)-e_{thr } = - c ( \lambda-\lambda_{cr})^2 + \mbox{\textnormal{h.t . } } \nonumber\end{aligned}\ ] ] where @xmath143 is a finite constant . case ( b ) follows from theorem [ th : main ] and case ( c ) was proved in theorem 3.2 in @xcite . in case ( a )
let @xmath90 denote the eigenfunction of @xmath78 , which corresponds to the eigenvalue @xmath8 . as follows from the proof of theorem 2 in @xcite @xmath144 in norm , where @xmath145 is the eigenfunction corresponding to the zero eigenvalue of @xmath89 .
thus @xmath146 , where @xmath147 and , hence , @xmath148 .
the rest of the proof is trivial .
let @xmath149 be the ground state energy of the hamiltonian @xmath150 acting on @xmath151 , where @xmath66 is the operator of multiplication by @xmath152 and @xmath153 is the coupling constant . because the pair of particles @xmath1 has a zero energy resonance and no negative energy bound states the following expansion @xcite @xmath154 is true for @xmath155 .
the positive constant @xmath156 in this expansion can be expressed through the zero energy solution of the birman - schwinger operator , namely @xcite , @xmath157 where @xmath158 is the unique nonnegative and normalized solution of the equation @xmath159 below ) , the constant @xmath114 in ( [ main ] ) can also be expressed through zero energy solutions of certain birman - schwinger operators .
[ th : main2 ] suppose that the interaction between the particles @xmath1 is tuned to make them have a zero energy resonance and no negative energy bound states .
let @xmath90 for @xmath160 be the ground state wave function of @xmath78 defined in ( [ hami ] ) , which corresponds to the ground state energy @xmath8 .
then there exists @xmath161 such that @xmath162 ( the same constant @xmath114 is used in theorem [ th : main ] ) .
we shall extensively use the results from @xcite , therefore it is convenient to pass to sequences .
let @xmath163 be any sequence such that @xmath164 and @xmath165 .
instead of ( [ 1.24:152 ] ) it suffices to prove that @xmath166 where @xmath167 .
we need only to prove that the limit on the lhs of ( [ 1.24:15 ] ) exists and is positive .
note that all requirements r1-r3 in @xcite are satisfied and the sequence @xmath168 totally spreads ( see sec .
theorems 1,3 in @xcite for the proof ) .
before we proceed with the proof let us introduce additional notations .
let @xmath169 and @xmath170 denote the partial fourier transform and the projection operator , which act on @xmath171 as follows @xmath172 ( x , y ) = \phi_0 ( x ) \int f(x',y ) \phi_0 ( x ' ) d^3 x ' , \end{gathered}\ ] ] and where @xmath173 is defined in ( [ 1.16;1 ] ) .
for a shorter notation let us denote @xmath174 similar to eqs .
( 39)-(40 ) in @xcite we introduce the operator function @xmath175 where @xmath176 from the schrdinger equation for @xmath168 it follows that @xmath180 where @xmath181 for @xmath182 is a bounded operator on @xmath183 defined as @xmath184 like in sec .
ii in @xcite one proves that @xmath181 for @xmath185 has a norm limit @xmath186 , besides @xmath181 for @xmath187 is a positivity preserving self - adjoint operator .
let us show that the off - diagonal terms in ( [ blacblac ] ) are compact operators on @xmath188 . by the resolvent identity @xmath189 |v_{12}|^{\frac 12 } \bigl ( h_0 + v_{12 } + z\bigr)^{-1 } |v_{23}|^{\frac 12 } = \left [ |v_{13}|^{\frac 12 } \bigl ( h_0 + z\bigr)^{-1 } |v_{23}|^{\frac 12 } \right ] \nonumber\\ + \left[|v_{13}|^{\frac 12 } \bigl ( h_0 + z\bigr)^{-1 } |v_{12}|^{\frac 12 } \right ] \left[|v_{12}|^{\frac 12 } \bigl ( h_0 + z\bigr)^{-1 } |v_{23}|^{\frac 12 } \right ] \nonumber\\ - \left[|v_{13}|^{\frac 12 } \bigl ( h_0 + z\bigr)^{-1 } |v_{12}|^{\frac 12 } \right ] \left\{|v_{12}|^{\frac 12 } \bigl ( h_0 + v_{12}+z\bigr)^{-1 } |v_{12}|^{\frac 12}\right\ } \left[|v_{12}|^{\frac 12 } \bigl ( h_0 + z\bigr)^{-1 } |v_{23}|^{\frac 12 } \right ] .
\label{gogol}\end{gathered}\ ] ] on the rhs of ( [ gogol ] ) all operators in square brackets are hilbert - schmidt ( the result of this sort goes back at least to @xcite , see also @xcite for the proof in present notations ) . since the operator in curly brackets is bounded for @xmath182 the lhs of ( [ gogol ] ) is a compact operator .
thus by weyl s criterion @xcite @xmath190 , \ ] ] where @xmath191 was defined in ( [ tildelambda ] ) .
@xmath192 is a normalized eigenvector of @xmath193 corresponding to the eigenvalue @xmath194 . due to the location of the essential spectrum
@xmath195 equals the maximal eigenvalue of @xmath193 . because @xmath196 we conclude due to the positivity preserving property that @xmath197 ( see theorem xiii.43 in vol . 4 of @xcite ) .
therefore , due to the norm convergence @xmath198 is the maximal eigenvalue of @xmath199 , which is isolated and non - degenerate .
let @xmath200 with @xmath201 be the eigenvector of @xmath199 , which corresponds to @xmath202 . again by the norm convergence of @xmath193 we have that @xmath203 and @xmath204 in norm . to prove that @xmath205 note that by eq .
( 67 ) in @xcite @xmath206 we get @xmath207 } \bigl ( \sqrt{p_y^2 + k_n^2}\bigr ) \hat \varphi^{(4)}_n + \chi_{(\rho_0 , \infty ) } \bigl ( \sqrt{p_y^2 + k_n^2}\bigr ) \hat \varphi^{(4)}_n = \chi_{[0 , \rho_0 ] } \bigl ( \sqrt{p_y^2 + k_n^2}\bigr ) \hat \varphi^{(4)}_n + \mathcal{o}(1),\ ] ] where @xmath208 denotes the terms , which are uniformly norm - bounded . here
@xmath209 is a fixed cutoff parameter .
following lemma 11 in @xcite ( the value of @xmath210 is also defined there ) we can expand the operator in curly brackets in ( [ 29.12:1 ] ) .
this expansion gives ( see eqs .
( 73)-(74 ) in @xcite ) @xmath211 } \bigl ( \sqrt{p_y^2 + k_n^2}\bigr ) \hat \varphi^{(4)}_n = \chi_{[0 , \rho_0 ] } \bigl ( \sqrt{p_y^2 + k_n^2}\bigr ) a^{-1 } \mathbb{p}_0 |v_{12}|^{\frac 12 } \nonumber\\ \times \bigl(|p_y|^2 + k_n^2\bigr)^{-\frac 12}\xi_n ( p_y ) \bigl[-\delta_x + p_y^2 + k_n^2\bigr]^{-1}\widehat{|v_{13}|^{\frac 12 } } \hat \varphi^{(1)}_n + \mathcal{o}(1 ) , \end{gathered}\ ] ] where @xmath212 .
thus @xmath213 } ( |p_y| ) \bigl(|p_y|^2 + k_n^2\bigr)^{-\frac 12 } \xi_n ( p_y ) \bigl[-\delta_x + p_y^2 + k_n^2\bigr]^{-1}\widehat{|v_{13}|^{\frac 12}}\right\| + \mathcal{o}(1)\ ] ] it remains to prove that the operator norm on the right hand side ( rhs ) of ( [ 1.24:12 ] ) is uniformly bounded .
this can be trivially estimated through the hilbert - schmidt norm ( c. f. proof of lemma 9 in @xcite ) @xmath214 } ( |p_y| ) \bigl(|p_y|^2 + k_n^2\bigr)^{-\frac 12 } \xi_n ( |p_y| ) \bigl[-\delta_x + p_y^2 + k_n^2\bigr]^{-1}\widehat{|v_{13}|^{\frac 12}}\right\|^2 \nonumber \\ \leq
c \int_{|p_y|\leq \rho_0 } d^3 p_y \frac{\xi_n^2 ( |p_y| ) } { ( p_y^2 + k_n^2)^{3/2 } } , \label{1.24:14 } \end{gathered}\ ] ] where @xmath215 is a constant .
the integral on the rhs of ( [ 1.24:14 ] ) is clearly convergent and uniformly bounded .
in @xcite it is proved that @xmath220 where @xmath221 is the fourier image of @xmath222 defined in ( [ gn ] ) and @xmath223 ( [ 1.16;3 ] ) follows from eq .
( 90 ) in @xcite and the fact that the norm of the function on the rhs of that equation goes to one for @xmath224 ( see the text under eq . ( 90 ) in @xcite ) .
our aim is to show that the sequence @xmath225 converges and @xmath226 ; then ( [ 1.24:15 ] ) follows with @xmath227^{-2}$ ] .
note that due to the positivity preserving property of @xmath228^{-1}$ ] @xmath229 where @xmath230 are defined in eqs .
( 93 ) , ( 94 ) in @xcite .
it suffices to prove the convergence of the first term on the rhs of ( [ 1.24:17 ] ) ( the convergence of the second term is proved analogously ) . similar to @xcite let us introduce jacobi coordinates @xmath231 , which are pictured in fig . [ fig:1 ] ( right ) @xmath232^{\frac 12 } \left ( r_2 - \frac{m_1}{m_1+m_3 } r_1 - \frac{m_3}{m_1+m_3 } r_3\right ) \nonumber , \end{gathered}\ ] ] where @xmath233 .
the coordinates @xmath234 and @xmath235 are connected through the orthogonal linear transformation @xmath236 where @xmath237 are real and can be expressed through mass ratios in the system . for all @xmath238 we have @xmath239 } ( |\eta|)\phi_0 \mathfrak{x}_n \varphi_n^{(1)}\right\| _ 1 \nonumber\\ + m_n^{-1}\left\| \chi_{(r
, \infty ) } ( |\eta| ) \phi_0 |v_{12}|^{\frac 12 } \left[h_0 + k_n^2 \right]^{-1 } |v_{13}| \psi_n \right\| _ 1 , \label{1.24:21}\end{gathered}\ ] ] where we have defined @xmath240^{-1 } |v_{13}|^{\frac 12 } .\ ] ] the operators @xmath241 are norm - bounded and have a norm limit for @xmath224 , which we denote as @xmath242 ( for the proof see f. e. lemma 7 in @xcite ) .
thus by lemma [ lemma:1 ] @xmath243 in @xmath244 sense .
then @xmath245 } ( |\eta|)\phi_0 \mathfrak{x}_n \varphi_n^{(1)}$ ] converges to @xmath245 } ( |\eta|)\phi_0 \mathfrak{x}_0 \varphi_\infty^{(1)}$ ] in @xmath246 sense because by the cauchy - schwarz inequality @xmath247 } ( |\eta|)\phi_0 \bigl ( \mathfrak{x}_n \varphi_n^{(1 ) } - \mathfrak{x}_0 \varphi_\infty^{(1)}\bigr)\right\|_1 \leq \|\chi_{[0 , r ] } ( |\eta|)\phi_0 \| \left\|\mathfrak{x}_n \varphi_n^{(1 ) } - \mathfrak{x}_0 \varphi_\infty^{(1)}\right\| , \ ] ] where @xmath248 } ( |\eta|)\phi_0 \| $ ] is finite .
hence , the first term on the rhs of ( [ 1.24:21 ] ) converges for all @xmath238 .
now the convergence of the sequence on the left hand side ( lhs ) of ( [ 1.24:21 ] ) follows from lemmas [ lemma:2 ] , [ lemma:3 ] .
we have proved that the sequence on the lhs of ( [ 1.24:17 ] ) converges to @xmath249 .
it remains to show that @xmath250 .
this follows from the fact that @xmath251 and besides @xmath252 , so at least one of the terms on the rhs of ( [ 1.24:17 ] ) converges to a positive value .
[ remark:3 ] the proof of theorem [ th : main2 ] allows to express the constant @xmath114 in theorem [ th : main ] in terms of zero energy solutions of birman - schwinger operators , namely , @xmath253^{-2}$ ] , where @xmath254 and by definition @xmath255 is the operator norm limit @xmath256^{-1 } |v_{23}|^{\frac 12 } .\ ] ] the proof is partly based on that of lemma 4 in @xcite , and we need to introduce additional notations from @xcite .
@xmath259 denotes the partial fourier transform , which acts on @xmath260 as @xmath261 let us introduce the operator function @xmath262 where @xmath263 following eq .
( 30 ) in @xcite we define @xmath264^{-1 } v_{13 } \psi_n .\ ] ] then using eq . ( 98 ) in @xcite we obtain the inequality @xmath265^{-1 } |v_{13}| \psi_n \right\| _ 1 = m_n^{-1 } \lambda_n^{-1}\left\| \chi_{(r , \infty ) } ( |\eta| ) \phi_0 |v_{12}|^{\frac 12 } f_n \right\| _ 1 \nonumber\\ \leq m_n^{-1 } \lambda_n^{-1 } \sum_{i=1}^3\left\| \chi_{(r , \infty ) } ( |\eta| ) \phi_0 |v_{12}|^{\frac 12 } \tilde f_n^{(i ) } \right\| _ 1 , \label{1.20;23}\end{gathered}\ ] ] where @xmath266^{-1 } |v_{13}|^{1/2 } \tilde b_{13}(k_n ) \psi^{(i)}_n \label{xw17}\ ] ] and @xmath267^{-1 } |v_{12}| |\psi_n| , \label{psin1}\\ \psi^{(2)}_n : = \lambda_n |v_{13}|^{1/2 } \tilde b_{13}^{-1 } ( k_n ) \bigl [ h_0 + k_n^2\bigr]^{-1 } |v_{23}| |\psi_n| \label{psin2},\\ \psi^{(3)}_n : = \lambda_n |v_{13}|^{1/2 } \bigl [ h_0 + k_n^2\bigr]^{-1 } |v_{13}|^{1/2 } \bigl\{1 - \lambda_n |v_{13}|^{1/2 } \bigl[h_0 + k_n^2\bigr]^{-1 } |v_{13}|^{1/2 } \bigr\}^{-1 } \nonumber \\\times\tilde b_{13}^{-1 } ( k_n ) |v_{13}|^{1/2 } \bigl [ h_0 + k_n^2\bigr]^{-1 } \bigl(|v_{12}| + \lambda_n and ( [ psin1])-(([psin3 ] ) ) come from eqs . ( 43 ) , ( 44 ) , ( 48 ) and ( 49 ) in @xcite , where one has to take into account that all interaction potentials are nonpositive . by eq .
( 103 ) in @xcite @xmath268 where @xmath269 . using the exponential falloff of @xmath270 from ( [ 1.16;1 ] )
one can easily derive the inequality @xmath271 where @xmath272 are constants .
after substituting ( [ 20.1;15 ] ) and ( [ xw23 ] ) into ( [ 1.20;23 ] ) , interchanging the order of integration and applying the cauchy - shwarz inequality we get @xmath265^{-1 } |v_{13}| \psi_n \right\| _
1 \nonumber\\ \leq \frac { \tilde b_1 \lambda_n^{-1}}{2^{7/2 } \pi^{5/2 } } \sum_{i=1}^3 m_n^{-1 } \bigl\| \psi^{(i)}_n \bigr\| \left\ { \int d^3 \eta ' \int d^3 p_\zeta \ ; \bigl| v_{13}(\alpha ' \eta')\bigr| \ , { \tilde t}^{\ ; 2}_n ( p_\zeta ) \ , { \tilde j}^2 ( \eta ' , p_\zeta ) \right\}^{1/2 } , \end{gathered}\ ] ] where @xmath273 by lemma [ lemma:4 ] we only need to prove that @xmath274 for @xmath257 , where we defined @xmath275 let us split the last integral as @xmath276 , where @xmath277 clearly , we can write @xmath278 where @xmath279 is a constant .
for @xmath280 and @xmath281 we have @xmath282 , which gives the estimate @xmath283 where @xmath284 is a constant . substituting this estimate into ( [ 1.24:31 ] ) and using that @xmath285 we obtain @xmath286,\ ] ] where @xmath287 is a constant . substituting eq .
( 40 ) from @xcite we get @xmath288 ds + c_4 \int_1^\infty e^{-sr } \left[r^2 + \frac4{s^2 } \right]ds\ ] ] where @xmath289 is another constant .
let us set @xmath290 , where @xmath291 is finite and depends only on @xmath15 .
then @xmath292 ds = \hbox{o}(r)\ ] ] it is easy to see that the terms on the rhs of ( [ 1.24:34 ] ) vanish for @xmath257 .
let us estimate @xmath293 . from ( [ 1.24:37 ] )
we get @xmath294 where @xmath295 is a constant . substituting this into ( [ 1.24:32 ] )
we obtain @xmath296 where the integral in curly brackets goes to zero because @xmath285 and the second integral is uniformly bounded for all @xmath38 .
[ lemma:3 ] suppose that the sequence @xmath297 is such that @xmath298 , where @xmath299 depend on a parameter @xmath238 .
additionally assume that @xmath300 is convergent for all @xmath238 and @xmath301 for @xmath257
. then @xmath302 converges .
the proof is a trivial application of the cauchy convergence criterion . for any @xmath73 fix @xmath303 so that @xmath304 for all @xmath305 .
choose @xmath306 so that @xmath307 for all @xmath308 .
then @xmath309 for all @xmath310 .
we have @xmath314 , where @xmath315^{-1 } |v_{12}|^{1/2}\ ] ] are uniformly norm - bounded operators ( they are defined in the same way as in eq .
( 63 ) in @xcite ) .
thus by lemma [ lemma:1 ] @xmath316 . from definition of @xmath317
we have @xmath318 where @xmath319^{\frac{-3+\delta}4}\\ \mathfrak{d}_n:= |v_{12}|^\frac12 \tilde b_{13}^{-1 } ( k_n ) \bigl [ h_0 + k_n^2\bigr]^{\frac{-1-\delta}4}. \end{gathered}\ ] ] @xmath320 are bounded operators on @xmath188 . from ( [ 1.24:41 ] )
@xmath321 follows from @xmath320 being uniformly norm - bounded .
let us start with estimating the norm of @xmath322 .
let us define operator functions @xmath323 , which act on @xmath324 as follows @xmath325^{-1 } \bigl[-\delta_\eta + p_\zeta^2 + k_n^2\bigr]^{\frac{-3+\delta}4 } \nonumber \\ \times
\chi_{[0,1]}\bigl(|-i\nabla_\eta|\bigr ) h , \\ \mathfrak{t}^{(2)}_n ( p_\zeta ) h = |v_{13}(\alpha'\eta)|^{\frac 12}\left[\xi_n \bigl(m_{y\eta}(-i\nabla_\eta ) + m_{y\zeta}p_\zeta\bigr ) \right]^{-1 } \bigl[-\delta_\eta + p_\zeta^2 + k_n^2\bigr]^{\frac{-1-\delta}4 } \nonumber \\
\times\chi_{(1,\infty)}\bigl(|-i\nabla_\eta|\bigr ) h . \end{gathered}\ ] ] the operators like @xmath326}\bigl(|-i\nabla_\eta|\bigr)$ ] act in the sense described in chapter 4 in @xcite .
it is easy to see that @xmath327 ( \eta , p_\zeta ) = \mathfrak{t}^{(1)}_n ( p_\zeta ) \hat f ( \eta , p_\zeta ) + \mathfrak{t}^{(2)}_n ( p_\zeta ) \hat f ( \eta , p_\zeta ) , \ ] ] where @xmath328 .
thus @xmath329 the operator norms on the rhs of ( [ 1.24:44 ] ) can be bounded by the trace ideals norms , which in turn can be bounded by theorem 4.1 in @xcite .
@xmath330^{\frac 12},\ ] ] where @xmath331^{\frac{3-\delta}2}}\leq \int_{|s|\leq 1 } \frac{d^3
s}{\bigl| m_{y\eta } s + m_{y\zeta}p_\zeta\bigr|^{\frac \delta4 } |s|^{3-\delta } } \nonumber\\ + \int_{|s|\leq 1 } \frac{d^3 s}{|s|^{3-\delta } } \leq \left[\int_{|s|\leq 1 }
\frac{d^3 s}{|s|^{3-\delta/2}}\right]^{\frac{3-\delta}{3-\delta/2 } } \left[\int_{|s|\leq 1 } \frac{d^3 s}{\bigl| m_{y\eta } s + m_{y\zeta}p_\zeta\bigr|^{\frac 32 - \frac \delta4 } } \right]^{\frac{\delta}{6-\delta } } + \int_{|s|\leq 1 } \frac{d^3 s}{|s|^{3-\delta } } \nonumber\\ \leq \left[\int_{|s|\leq 1 } \frac{d^3 s}{|s|^{3-\delta/2}}\right]^{\frac{3-\delta}{3-\delta/2 } } \left[\int_{|s|\leq 1 } \frac{d^3 s}{m_{y\eta}| s|^{\frac 32 - \frac \delta4 } } \right]^{\frac{\delta}{6-\delta } } + \int_{|s|\leq 1 } \frac{d^3 s}{|s|^{3-\delta } } .
\label{1.24:46}\end{gathered}\ ] ] in ( [ 1.24:46 ] ) we have used hlders inequality .
the integrals on the rhs of ( [ 1.24:46 ] ) are convergent and independent of @xmath332 and @xmath38 , hence , @xmath333 .
similarly @xmath334^{\frac 13},\ ] ] where @xmath335 the integrals on the rhs of ( [ 1.24:48 ] ) obviously converge .
thus we find that @xmath336 and @xmath337 is uniformly bounded ( note that @xmath338 in ( [ 1.24:48 ] ) is bounded because @xmath339 ) .
now we pass to estimating @xmath340 and use the same method .
calculations similar to above ones give @xmath341^{\frac 12 } \nonumber\\ + \sup_{p_y } \left\| \bigl|v_{12}(\alpha x)\bigr|^{\frac 12 } \left[\tilde t_n \bigl ( m_{x\zeta } ( -i\nabla_x ) + m_{y\zeta } p_y \bigr)\right]^{-1}\right\| , \label{1.24:53}\end{gathered}\ ] ] where @xmath342^{\frac{2 - 2\delta}{3-\delta } } \left[\int_{|s|\leq 1 } \frac{d^3 s}{|s|^{3-\delta}}\right]^{\frac{1+\delta}{3-\delta } } + \int_{|s|\leq 1 } \frac{d^3 s}{|s|^{1+\delta } } \nonumber\\ \leq \left[\int_{|s|\leq 1 } \frac{d^3 s}{m_{x\zeta }
| s |^{3-\delta}}\right]^{\frac{2 - 2\delta}{3-\delta } } \left[\int_{|s|\leq 1 } \frac{d^3 s}{|s|^{3-\delta}}\right]^{\frac{1+\delta}{3-\delta } } + \int_{|s|\leq 1 } \frac{d^3 s}{|s|^{1+\delta } } .
\label{1.24:51}\end{gathered}\ ] ] the integrals on the rhs of ( [ 1.24:51 ] ) converge and it remains to estimate the operator norm in ( [ 1.24:53 ] ) .
@xmath343^{-1}\right\| \nonumber\\ \leq \left\| \bigl|v_{12}(\alpha x)\bigr|^{\frac 12 } \left[\tilde t_n \bigl ( m_{x\zeta } ( -i\nabla_x ) + m_{y\zeta } p_y \bigr)\right]^{-1 } \chi_{[0,1 ] } \left ( \bigl| m_{x\zeta } ( -i\nabla_x ) + m_{y\zeta } p_y \bigr| \right)\right\|_2 + \bigl\| v_{12}\bigr\|_\infty \nonumber\\ \leq \bigl\| v_{12}\bigr\|_\infty + ( 2\pi\alpha)^{-\frac 32 } \|v_{12}\|_1 \left[\sup_{p_y } \int_{|m_{x\zeta}s + m_{y\zeta}p_y| \leq 1 } \frac{d^3 s}{\bigl|m_{x\zeta}s + m_{y\zeta}p_y\bigr|^{2 - 2\delta}}\right]^{\frac 12 } \nonumber\\ \leq \bigl\| v_{12}\bigr\|_\infty + ( 2\pi\alpha m_{x\zeta})^{-\frac 32 } \|v_{12}\|_1 \left[\int_{|s| \leq 1 } \frac{d^3 s}{|s|^{2 - 2\delta } } \right]^{\frac 12 } , \end{gathered}\ ] ] where we have again used theorem 4.1 in @xcite .
thus we find that @xmath344 and , hence , @xmath345 .
note that because all potentials are nonpositive we have @xmath346 , \ ] ] where @xmath347^{-1 } |v_{13}|^{1/2 } , \\
q_n : = \bigl\{1 - \lambda_n |v_{13}|^{1/2 } \bigl[h_0 + k_n^2\bigr]^{-1 } |v_{13}|^{1/2 } \bigr\}^{-1 } .
\end{gathered}\ ] ] operators @xmath348 are defined in the same way in eqs .
( 62 ) and ( 46 ) in @xcite . in @xcite it was proved that @xmath349 and @xmath350 . thus @xmath351 as claimed .
g. teschl , _ mathematical methods in quantum mechanics ; with applications to schrdinger operators _ , lecture notes ( 2005 ) , http://www.mat.univie.ac.at/ gerald / ftp / book - schroe / index.html ; graduate studies in mathematics , vol . * 99 * , amer .
soc . , providence , 2009 . | we consider a pairwise interacting quantum 3-body system in 3-dimensional space with finite masses and the interaction term @xmath0 , where all pair potentials are assumed to be nonpositive .
the pair interaction of the particles @xmath1 is tuned to make them have a zero energy resonance and no negative energy bound states .
the coupling constant @xmath2 is allowed to take the values for which the particle pairs @xmath3 and @xmath4 have no bound states with negative energy .
let @xmath5 denote the critical value of the coupling constant such that @xmath6 for @xmath7 , where @xmath8 is the ground state energy of the 3-body system .
we prove the theorem , which states that near @xmath5 one has @xmath9^{-1}+$]h.t . , where @xmath10 is a constant and h.t .
stands for `` higher terms '' .
this behavior of the ground state energy is universal ( up to the value of the constant @xmath10 ) , meaning that it is independent of the form of pair interactions .
[ 1]#1 [ 1]#1 |
electromagnetic polarizabilities are fundamental properties of hadrons which are manifest in various processes involving two photons @xcite
. recent measurements @xcite of the electric , @xmath4 , and magnetic , @xmath5 , polarizabilities of the nucleon narrowed considerably experimental uncertainties in these observables , and were accompanied by a number of theoretical studies .
attempts to describe @xmath4 and @xmath5 were made in various approaches , ranging from chiral perturbation theory @xcite and dispersion relations @xcite to chiral soliton models @xcite .
earlier calculations of @xmath4 and @xmath5 are reviewed in ref .
@xcite . in this paper
we calculate the average electric polarizability of the proton and neutron , , using the solitonic approach @xcite to the nambu jona - lasinio model @xcite . this model has led to quite successful phenomenology @xcite . in particular , baryonic mass relations , magnetic moments , and various form factors
are reproduced well within expectations .
it is therefore challenging to see if we can also describe _ two - current observables _
, such as the polarizability , in the njl model
. there are two important features in our study which have not been considered in earlier works in soliton models : 1 ) the role of the dirac sea effects ( this can only be done in a model which has the dirac sea , such as the njl model ) , and 2 ) inclusion of rotational @xmath1-effects .
such effects were recently analyzed for the case of the axial coupling constant @xmath6 in ref .
@xcite and were more closely examined in refs .
@xcite . in the njl model
the only dynamical degrees of freedom are quarks , which occupy valence orbitals , as well as the dirac sea .
we find that the effects of the dirac sea in our calculation of @xmath4 are very well reproduced by the first two terms in the gradient expansion , where the first term is the so called `` seagull '' contribution to @xmath4 , discussed in many previous papers @xcite .
our analysis demonstrates explicitly that the inclusion of the sea - gull in the @xmath7-model or in the skyrme model does not violate gauge invariance , as has recently been claimed in ref .
this is readily seen from the full njl expression , which is purely _ dispersive _ in character ( _ i.e. _ involves no seagull terms ) and manifestly gauge invariant . the seagull term emerges from the full expression when the gradient expansion is performed ( see sec . [ se : leading ] ) .
the inclusion of rotational @xmath1 effects has very sound phenomenological consequences . as discussed extensively in ref .
@xcite , chiral soliton models , when treated at the leading-@xmath0 level , have problems in predicting electromagnetic polarizabilities correctly . the dominant contribution to the electric polarizability is obtained from pion tail effects , and is proportional to @xmath8 .
if one insists that the model fits the value of @xmath6 ( which one should ! ) , then the value of @xmath4 obtained at the leading @xmath0 level is roughly a factor of @xmath9 too large , as demonstrated in the @xmath7-model calculation of ref .
@xcite . in the njl model
( as well as in the skyrmion ) the value of @xmath6 obtained with various fits has always been notoriously too small at the leading-@xmath0 level @xcite , _
i.e. _ of the order @xmath10 . as first noticed in ref .
@xcite , the inclusion of rotational @xmath1 corrections in the njl model is capable of raising @xmath6 to a comfortable value of @xmath11 @xcite .
this is a big correction : it raises @xmath8 by a factor of @xmath12 .
we calculate analogous @xmath1 corrections for the electric polarizability , and find a sizeable contribution . as a result , and after including approximately additional corrections due to the @xmath3-@xmath2 mass splitting @xcite , we are able to obtain a number which is closer to experiment than in other studies in soliton models @xcite , but still too large . for the typical choice @xcite of the constituent quark mass
we obtain compared to the experimental value @xmath13 @xcite . in this paper
we do not analyze the splitting of the proton and neutron polarizabilities , since it involves a complicated problem of treating the translational zero mode @xcite .
also , the magnetic polarizability is not analyzed .
it was shown in ref .
@xcite that the large-@xmath0 approach is not a good starting point to describe this observable .
hence , we concentrate solely on the average proton and neutron electric polarizability , @xmath4 .
the outline of the paper is as follows : in sec . [ se : form ] we develop the necessary formalism of linear response theory for the njl model , and derive a basic expression for the electric polarizability . in sec .
[ se : leading ] we calculate the leading-@xmath0 contribution to @xmath4 , which comes from both valence and sea quark effects . in sec .
[ se : subleading ] we calculate the rotational @xmath1 corrections to @xmath4 .
the valence contribution is calculated exactly , while the sea part is estimated using gradient expansion .
additional corrections , due to @xmath2-@xmath3 mass splitting , are discussed in sec .
[ se : splitting ] .
[ se : conclusion ] contains our results and conclusions .
polarizabilities are defined as coefficients in the low - energy expansion of the compton amplitude @xcite .
it has been shown @xcite that the polarizability measured in compton scattering , @xmath14 , can be written as @xmath15 , where @xmath16 here @xmath17 is the nucleon state , and @xmath18 is any intermediate excited state connected by the electric dipole operator @xmath19 .
the first term in the expression for @xmath20 is the so called recoil correction , which involves the charge , @xmath21 , the mass , @xmath22 , and the mean squared charge radius of the particle , and the second term is the schwinger scattering term for a particle with anomalous magnetic moment @xmath23 .
our goal is to calculate the dispersive contribution @xmath4 in eq .
( [ alphas ] ) .
note that according to eq .
( [ alphas ] ) , this is equivalent to calculating the coefficient of the second - order energy shift of the nucleon in the constant electric field .
first , we very briefly review the njl model @xcite .
our expressions are written in the euclidean space - time . in the leading-@xmath0 treatment the effective action results from taking into account one - quark loops .
it contains the quark part , @xmath24 , the chirally invariant mesonic part , @xmath25 , and the chiral symmetry breaking part , @xmath26 , which provides the pion its mass @xmath27 @xmath28 the trace is over the quark spinors .
the subscript @xmath29 reminds of the njl cut - off ( in this paper we use the proper time regulator @xcite ) , @xmath30 is the current quark mass , and @xmath22 is the constituent quark mass .
the matrix @xmath31 represents the meson _ hedgehog _ profile in the non - linear representation @xmath32 where @xmath33 is the chiral angle , @xmath34 and @xmath35 are the sigma and pion hedgehog profiles with and , @xmath36 is the radial unit vector , and @xmath37 are the pauli isospin matrices .
the parameters of the model are fixed by reproducing the experimental values of the pion mass , @xmath38 , and the pion decay constant , @xmath39 , leaving the quark - meson coupling constant @xmath40 , or , equivalently , the constituent quark mass @xmath41 as the only free parameter @xcite .
solitonic solutions to the njl model are found by selfconsistently solving the coupled euler - lagrange equations for the profile function @xmath42 and the dirac spinors @xmath43 , which satisfy the dirac equation @xmath44 in the presence of the the electromagnetic field @xmath45 , the minimal substitution , with the quark charge , leads to the following gauged effective action @xmath46 the hedgehog solution brakes the spin , @xmath47 , and isospin , @xmath48 , symmetries of the lagrangian , preserving the grand spin @xmath49 symmetry . to restore good quantum numbers we use the semiclassical `` cranking '' projection scheme @xcite . introducing the collective rotation matrix @xmath50 in the usual way and transforming the quark fields and meson profiles to the isorotating frame @xcite we obtain @xmath51 it is convenient to introduce the notation @xmath52 , where @xmath53 is an element of a wigner @xmath54-matrix , @xmath55
then the quark electric charge in the isorotating frame can be written as @xmath56 the action ( [ cranked ] ) describes the system which is perturbed by _ small _ perturbations : electromagnetic , and cranking . from now on our method will follow very closely ref
. @xcite .
the idea is the following : we have found the soliton , and now we want to calculate its linear response to external perturbations .
the perturbation operators consist of products of two parts : _ intrinsic _ , which act on the internal structure of the state , and _ collective _ , which act on collective coordinates . for the cranking perturbation , @xmath57 , the collective operator is , and @xmath58 is the intrinsic operator .
similarly , the electromagnetic perturbation has a collective part ( vector @xmath59 in eq .
( [ charge ] ) ) and an intrinsic part . also , the nucleon state is a product of a collective and an intrinsic part @xcite .
formal perturbative expansion of eq .
( [ cranked ] ) to @xmath60-th order in small parameters involves expressions of the general form @xmath61 where @xmath62 is the baryon collective wave function , @xmath63 is the intrinsic hedgehog state , @xmath64 is the perturbation of the @xmath65 type , and @xmath66 is the quark feynman propagator in the background meson field , which in the spectral representation has the form @xmath67 where occ . denotes all occupied states , _ i.e. _ the valence as well as the dirac sea states and unocc . denotes the unoccupied positive energy states .
examples of application of eq .
( [ eq : kappan ] ) are given in ref .
it was shown in ref .
@xcite that the leading contribution to the electric polarizability of the nucleon scales as @xmath0 .
our first task is to obtain this leading contribution for @xmath4 in the njl model .
we denote it by @xmath68 .
since the cranking frequency @xmath69 , the leading-@xmath0 contribution is obtained by expanding the effective action ( [ cranked ] ) to second order in the electric field , and to zero order in @xmath70 . from eq .
( [ charge ] ) we notice , that for the leading @xmath0 contribution we have to keep only the isovector part of the charge , since the isoscalar part is one power of @xmath0 down . we now take @xmath71 , @xmath72 , and expand eq . (
[ action : gauged ] ) to second order in @xmath73 and zero order in @xmath70 .
we obtain the formula of the form ( [ eq : kappan ] ) @xmath74 where @xmath17 is the collective nucleon state , @xmath75 is a large time interval , over which the time integrations are carried , and @xmath76 and @xmath77 denote the eigenstates and eigenvalues of the dirac hamiltonian ( [ dirac : ham ] ) .
we can now split this formula into the valence and sea parts .
the sea contribution carries the cut - off function , and the vacuum contribution has to be subtracted . finally , using
, we get the following formulas for the leading @xmath0 valence and sea contributions to the electric polarizability of the nucleon @xmath78 @xmath79 where @xmath80 is the proper time regularization function @xmath81,\ ] ] with @xmath29 denoting the ( proper time ) cut - off parameter .
the value of @xmath29 is fit for any @xmath22 to reproduce the experimental value of the pion coupling constant @xmath39 using the amplitude of the weak pion decay @xcite .
the subtracted vacuum contribution in eq .
( [ sea ] ) has the same structure as the first term , but the free quark states and eigenvalues are used .
it may be immediately verified that both the valence and sea contributions to @xmath68 scale as @xmath0 @xcite . in the njl model ,
predictions depending on the cut - off are somewhat ambiguous .
as we shall see , however , in our case @xmath82 is dominated by the lowest term in the gradient expansion , which is independent of the cut - off .
the numerical results for the valence , @xmath83 , and dirac sea , @xmath82 , parts of @xmath68 are plotted in fig . 1 as a function of the constituent quark mass @xmath22
. both contributions decrease with increasing @xmath22 , the valence curve has much larger slope for smaller @xmath22 due to the fast increasing energy gap between the valence orbital and the positive continuum . for the optimum value @xcite of @xmath84 , we obtain .
the sea contributes to @xmath68 65% , and dominates over the valence part , which gives 35% .
since the dirac sea effects describe the physics of the pion cloud , this result is in qualitative agreement with the results of other models @xcite .
we have also performed the gradient expansion of @xmath82 up to first two terms following the scheme of refs .
this provides a link to the @xmath7-model @xcite as well as an independent check for our numerical procedure in the case of large size solitons .
the leading - order term in the gradient expansion of @xmath82 gives us exactly the _ seagull _ term of the @xmath7-model @xmath85 note that this seagull term arises in the njl model from _ purely dispersive _ dirac sea effects ( _ cf . _
( [ sea ] ) ) .
recently , the inclusion of the seagull term in soliton models such as the @xmath7-model or the skyrmion has been questioned @xcite .
the njl model provides a simple interpretation of how eq .
( [ seagull ] ) arises in the @xmath7-model , and clears the doubts of ref .
numerically the seagull term gives more than 80% of the total sea contribution .
looking at absolute numbers it may seem striking that our value of the seagull contribution , @xmath86 , and the @xmath7-model @xcite value , @xmath87 differ so drastically .
the explanation lies in the large difference in the values of the axial coupling constant at the leading @xmath0 level .
our model predicts @xmath88 , while in the @xmath7-model @xmath89 @xcite .
the asymptotic pion tail , which gives the dominant contribution in ( [ seagull ] ) , is proportional to @xmath90 and since the pion field enters quadratically in ( [ seagull ] ) the large effect follows .
we will come back to the question of @xmath6 in the next section .
it is interesting to investigate in some greater detail how well the gradient expansion works for the electric polarizability . in fig .
2 we compare the results of the gradient expansion with the full calculation of @xmath82 for large - size solitons parameterized by the profile function involving a size parameter @xmath91 @xmath92 with increasing size of the soliton the result of the gradient expansion for @xmath82 up to first two terms converges to the full result very well . at @xmath93 fm the difference is only 2% .
even the leading order of the gradient expansion , @xmath94 , shows good agreement with the full result , reproducing it up to 90% at @xmath93 fm .
now we turn to the problem of the next - to - leading order corrections in the @xmath1 expansion of the electric polarizability @xmath4 .
as already mentioned above , the value of @xmath6 in chiral soliton models is crucial for the results for @xmath4 .
the prediction for @xmath6 in the njl model ( as well as in the skyrme model ) at the leading-@xmath0 level , denoted by @xmath90 , turns out to be much too small , @xmath95 @xcite , and no parameter fits or extensions of the model were able to raise this value .
recently , ref .
@xcite proposed how to include rotational @xmath1 corrections to @xmath6 in the njl model .
a correct and detailed study of this effect has been presented in refs .
the nature of these corrections is very easy to understand , for instance from the point of view of the formalism of ref . @xcite . at the leading-@xmath0 level
the expression for @xmath6 involves a sum of matrix elements of the axial operator in the quark single - particle states @xcite
. however , in the presence of cranking the hedgehog state is perturbed , and we have a situation where one of the perturbations is the axial vector current , and the other one is the cranking . the effect discussed in refs .
@xcite is associated with the noncommutativity of the collective parts of the corresponding operators ( the cranking and the axial vector current operators in this case ) .
an analogous effect arises for the case of @xmath4 , as we discuss below . since the cranking operator scales as @xmath1 ( it involves the cranking frequency @xmath69 )
the correction is suppressed by one power of @xmath0 .
this correction to the axial vector coupling constant , denoted by @xmath96 , has the valence part and the sea part .
the numerical value for @xmath96 is about @xmath97 @xcite , and raises the total value of @xmath6 to a comfortable number . of course , the described @xmath1 corrections are not the only possible ones , but they seem to be particularly important . we have explained before that the small value of @xmath90 reduces the value of the seagull contribution to the polarizability , which is a desired effect . since roatational @xmath1 corrections raise significantly the value of @xmath6 , we have also to examine how analogous effects influence the electric polarizability .
the @xmath1 rotational corrections to @xmath4 are obtained by expanding the effective action ( [ cranked ] ) to second order in @xmath73 and first order in @xmath70 .
thus , we have formally the case of the third - order perturbation theory .
we get expression of the generic form ( [ eq : kappan ] ) , with all possible time - orderings of collective operators . the relevant collective matrix elements for the present case have the form @xmath98 where @xmath99 is the moment of inertia . finally ,
after straightforward algebra , we obtain for the valence contribution @xmath100 \nonumber , \end{aligned}\ ] ] and for the sea contribution @xmath101 - { \rm vac } .
\nonumber \end{aligned}\ ] ] in the above expressions notation @xmath102 means the level with @xmath103 , _ etc . _
recalling that @xmath104 , it can be verified that @xmath105 is of order @xmath106 , while the leading contribution to @xmath68 was of order @xmath0 .
the numerical value for @xmath107 is obtained directly from eq .
( [ omega1val ] ) .
it involves a double summation over quark levels , and computer time involved is similar to the calculation of , say , the moment of inertia .
our results are shown in fig . 1 .
the value of @xmath107 is about for @xmath108 , so the effect is about 60% of the leading-@xmath0 valence contribution @xmath83 .
the decrease of @xmath107 with increasing @xmath22 is similar to the behavior of the leading-@xmath0 valence contribution @xmath83 .
now we turn to the sea contribution @xmath109 .
this contribution in principle may involve a cut - off , similar to other observables calculated in the njl model .
however , since eq .
( [ omega1sea ] ) involves a computer - time - consuming triple sum over quark states , we approximate @xmath109 by the leading term in the gradient expansion . as a bonus
, we get the nice feature that the result is independent of the cut - off , exactly as in the case of the leading-@xmath0 contribution @xmath94 . in the present case
the gradient expansion technique involves some tricks @xcite , so we sketch the method in some greater detail .
the first step is to rewrite eq .
( [ omega1sea ] ) in a form involving unconstrained sums , _
i.e. _ sums over all quark levels .
this can be achieved using the identity true for any real numbers @xmath110 , @xmath111 and @xmath112 @xmath113 with help of eq .
( [ eq : ident ] ) we can replace the constrained sums in eq .
( [ omega1sea ] ) by sums over all indices . for instance @xmath114 where @xmath115 denotes the principal value integral . introducing the trace over the quark spinors
@xmath116 we can now rewrite eq .
( [ omega1sea ] ) as @xmath117 -{\rm vac}. \nonumber\end{aligned}\ ] ] standard gradient expansion techniques may now be used to expand the functional trace in eq .
( [ eq : gr0 ] ) @xcite . the lowest order term ( denoted by @xmath118 ) , contains no gradients , and we obtain @xmath119 where @xmath94 is given in eq .
( [ seagull ] ) . for @xmath108 the moment of inertia has the value @xmath120 , and we obtain for the proportionality factor in the last equality in eq .
( [ eq : grad1 ] ) the value @xmath121 .
the numerical results for @xmath122 are plotted in fig . 1
. the contribution increases with increasing constituent quark mass @xmath22 .
this is due to the decrease of the moment of inertia with @xmath22 . guided by our experience from gradient - expanding the leading-@xmath0 contribution ( sec .
[ se : leading ] ) , we hope that also in the present case the lowest term in the gradient expansion almost saturates the value of @xmath109 . at the end of this section
we would like to present an interesting observation .
the formula analogous to eq .
( [ eq : grad1 ] ) for the case of @xmath6 has the form @xcite @xmath123 note that the coefficient between the @xmath1 correction and the leading-@xmath0 term in eq .
( [ eq : gagrad1 ] ) is _ exactly twice _ the analogous coefficient in eq .
( [ eq : grad1 ] ) .
this is not a coincidence . in the chiral limit , @xmath124
, the sea contribution to @xmath4 is divergent as @xmath125 @xcite .
this divergent piece is proportional to @xmath8 , and this result is a general , model independent feature of the chiral limit .
therefore , one has @xmath126 , where @xmath127 is a numerical constant .
hence , if in a model one corrects @xmath6 by including @xmath1 terms , analogous correction should come out in the calculation of @xmath1 corrections to @xmath4 .
consider the formal case of solitons with a very large size @xmath91 . in that limit ,
@xmath6 consists only of the sea contribution , which in turn is saturated by the lowest term in the gradient expansion : @xmath128 . using eq .
( [ eq : gagrad1 ] ) and the general considerations discussed above , we obtain in the double limit of @xmath124 and @xmath129 the following result @xmath130 the first term in the last line of is our @xmath131 , the second term is @xmath132 , with exactly the same factor as in eq .
( [ eq : grad1 ] ) .
this shows that our method of including rotational @xmath1 corrections is consistent with chiral physics .
analogous consistency conditions have to hold for other observables which behave as @xmath133 , _
e.g. _ the isovector magnetic mean squared radius of the nucleon .
another digression concerns the structure of the rotational @xmath1 corrections .
collective operators which arise in evaluating corrections to @xmath6 or the isovector magnetic moment have the same structure as collective operators entering at leading-@xmath0 level .
formally , this is because @xmath134 = i \epsilon^{abd } d^1_{dc}$ ] .
thus , ratios of these observables ( also for non - diagonal matrix elements ) for various members of the flavor su(2 ) multiplet ( _ i.e. _ for @xmath3 and @xmath2 ) are preserved at the @xmath1 level , and obtain corrections starting at the @xmath135 level .
this complies to the consistency condition of ref .
before comparing our results to data , we note that there is another very important effect which has not been included , namely the @xmath3-@xmath2 mass splitting . as discussed in ref .
@xcite , in the limit @xmath136 our perturbative approach has the nucleon and the @xmath2 degenerate in mass . at the hadronic level , the seagull term results from diagrams with intermediate pion - nucleon or pion-@xmath2 states . because of the @xmath3-@xmath2 degeneracy , the contribution of the pion-@xmath2 states is strongly overestimated in the hedgehog treatment .
one may crudely include the effect of @xmath3-@xmath2 mass splitting by reducing the seagull contribution by about 30% @xcite .
a consistent treatment of this effect would require the use of a better collective quantization method , but this does not seem feasible for the njl model . as shown in section [ se : leading ] more than 80% of the sea contribution @xmath82 is carried by the seagull term .
therefore , as suggested in refs .
@xcite , we simply multiply the sea contribution by a factor of @xmath137 to mimic the effect of @xmath3-@xmath2 mass splitting . since @xmath122 is proportional to @xmath94
, we expect that it obtains analogous corrections from @xmath2-@xmath3 splitting as the leading-@xmath0 part
. therefore we also multiply it by a factor of @xmath137 before comparing to data .
our numerical results are collected in figs . 1,3 .
they are plotted as a function of the constituent quark mass @xmath22 .
figure 1 shows the sea and valence contributions to the leading-@xmath0 nucleon electric polarizability @xmath68 as well as to the rotational @xmath1 correction @xmath105 .
the total polarizability is presented on fig . 3 compared to the experimental value .
the dashed line corresponds to @xmath138 .
the full line shows the total @xmath4 , crudely corrected for the effects of the @xmath3-@xmath2 mass splitting by multiplying the sea contributions by a factor of @xmath137 ( sec .
[ se : splitting ] ) : @xmath139 . for values of @xmath22 in the physically relevant range 400450 mev
the changes of @xmath4 with @xmath22 are small . choosing for @xmath140 mev ( value which gives good agreement for other nucleon observables @xcite ) we obtain for the total nucleon electric polarizability the value @xmath141 , with @xmath142 , @xmath83 , @xmath143 , and @xmath107 contributing 36% , 29% , 17% , and 17% respectively .
our result is still too large compared to the experimental number @xmath13 @xcite .
keeping in mind the approximate nature of the model and techniques used , our model prediction is within expectations .
note , that in our calculation , after including the rotational @xmath1 effects , we have the correct value of @xmath6 .
this is not the case of other models @xcite , and apparent agreement of @xmath4 with experiment in these models results from having too low @xmath6 .
the main points of this paper can be summarized as follows : \1 ) dirac sea effects dominate the electric polarizability of the nucleon . in the gradient expansion
the sea effects are dominated by the seagull contribution .
this explains how this term arises in other models which do not have the explicit dirac sea .
\2 ) we have calculated rotational @xmath1 corrections , and found that they are important and of expected size .
\3 ) we have crudely estimated the @xmath3-@xmath2 splitting effects .
\4 ) with all above effects included , numerical predictions of the njl model are better than in other soliton models ( note that we have the correct value of @xmath6 ! ) but are still about a factor of @xmath144 too large compared to experiment .
this indicates that the njl soliton is too `` soft '' and other effects are still needed in order to describe the polarizability properly .
\5 ) calculation of the magnetic polarizability , or the proton - neutron splitting of the electric polarizability would require a much greater effort in the njl model than in the @xmath7-model calculation of ref .
this is because of the necessity to subtract zero - modes .
the authors are grateful to pavel pobylitsa for helpful comments .
this work has been supported in part by the polish state committee for scientific research , grants 2.0204.91.01 and 2.0091.91.01 , and by the bulgarian national science foundation , contract @xmath145 - 32
. 10 v. a. petrunkin , _ sov .
* 12 * ( 1981 ) 278 .
j. l. friar , in _ proc .
workshop on electron - nucleus scattering , eipc , marciana , italy _ , edited by a. fabrocini _ et al . _
( world scientific , singapore , 1989 ) .
f. federspiel _
_ , _ phys .
_ * 67 * ( 1991 ) 1511 .
a. zieger _
* b278 * ( 1992 ) 34 .
j. schmiedmayer , p. riehs , j. harvey , and n. hill , _ phys .
* 66 * ( 1991 ) 1015 .
v. bernard , n. kaiser , and u .- g .
meissner , _ phys .
* 67 * ( 1991 ) 1515 ; _ nucl
_ * b373 * ( 1992 ) 346 .
m. n. butler and m. j. savage , _ phys .
* b294 * ( 1992 ) 369 .
lett . _ * b304 * ( 1993 ) 29 .
m. chemtob , _ nucl .
phys . _ * a473 * ( 1987 ) 613 .
n. n. scoccola and w. weise , _ phys .
* b232 * ( 1989 ) 287 ; _ nucl .
_ * a517 * ( 1990 ) 495 .
w. broniowski , m. k. banerjee , and t. d. cohen , _ phys . lett .
_ * b283 * ( 1992 ) 22 .
w. broniowski and t. d. cohen , _ phys .
* d47 * ( 1993 ) 299 .
b. schwesinger , _ phys .
* b298 * ( 1993 ) 17 .
b. golli and r. sraka , _ phys .
_ * b312 * ( 1993 ) 24 .
g. m. g. dattoli and d. prosperi , _ lett .
nuovo cimento _ * 19 * ( 1977 ) 601 .
_ * b227 * ( 1989 ) 296 . , _ phys .
lett . _ * b215 * ( 1988 ) 577 .
y. nambu and g. jona - lasinio , _ phys .
rev . _ * 122 * ( 1961 ) 345 .
phys . _ * a524 * ( 1991 ) 719 .
, ruhr - univ .
bochum preprint no .
rub - tpii-56 - 93 ( 1993 ) .
m. wakamatsu and t. watabe , _ phys .
* b312 * ( 1993 ) 184 . , ruhr - univ .
bochum preprint no .
rub - tpii-53 - 93 ( 1993 ) .
a. blotz , m. praszaowicz , and k. goeke , _ phys .
* b317 * ( 1993 ) 195 .
a. i. lvov , univ .
mainz preprint no . mkph - t-93 - 6 ( 1993 ) . , _
z. phys . _ * a339 * ( 1991 ) 513 .
t. d. cohen and w. broniowski , _ phys .
lett . _ * b292 * ( 1992 ) 5 .
j. schwinger , _ phys .
* 82 * ( 1951 ) 664 .
d. i. dyakonov and v. y. petrov , _ jetp lett . _
* 43 * ( 1986 ) 57 .
g. s. adkins , c. r. nappi , and e. witten , _ nucl .
* b228 * ( 1983 ) 552 . t. d. cohen and w. broniowski , _ phys .
* d34 * ( 1986 ) 3472 .
w. broniowski and t. d. cohen , _ phys .
* d47 * ( 1993 ) 313 .
r. weiner and w. weise , _ phys .
lett . _ * b159 * ( 1985 ) 85 .
i. j. r. aitchison and c. m. frazer , _ phys .
_ * b146 * ( 1984 ) 63 ; _ phys .
_ * d31 * ( 1985 ) 2605 ; _ phys . rev . _ * d32 * ( 1985 ) 2190 . , _ phys
. lett . _ * b253 * ( 1991 ) 430 . , _ nucl .
phys . _ * a547 * ( 1992 ) 612 .
m. c. birse and m. k. banerjee , _ phys .
lett . _ * b136 * ( 1984 ) 284 ; _ phys
. rev . _ * d31 * ( 1985 ) 118 .
a. blotz , m. praszaowicz , and k. goeke , ruhr - univ .
bochum preprint no .
rub - tpii-41 - 93 ( 1993 ) . .
r. dashen and a. v. manohar , uc , san diego preprint no .
ucsd / pth 93 - 16 , 93 - 18 ( 1993 ) .
fig . 1 .
the electric polarizability of the nucleon , @xmath4 , as a function of the constituent quark mass @xmath22 .
leading @xmath0 dirac sea contribution @xmath82 ( dashed line ) ; leading @xmath0 valence contribution @xmath83 ( dash - dotted line ) ; rotational @xmath1 correction to the sea contribution @xmath122 ( dotted line ) and to the valence one @xmath107 ( dash - dot - dot ) .
the ratio of the gradient expanded electric nucleon polarizability to the full leading-@xmath0 @xmath146 for different soliton sizes @xmath91 .
the dash - dotted line represents the leading order in the gradient expansion @xmath94 and the dashed line the seagull plus the next term in the gradient expanded @xmath4 .
the total electric polarizability of the nucleon , @xmath4 , as a function of the constituent quark mass @xmath22 .
the dashed line shows the full @xmath147 and the full line @xmath148 , where the factor @xmath149 in front of @xmath150 crudely corrects for the overestimation of the @xmath2 contribution ( sec .
[ se : splitting ] ) . | the electric polarizability of the nucleon is calculated in the soliton approach to the nambu jona - lasinio model .
we analyze the leading-@xmath0 contributions , as well as the effects of rotational @xmath1 corrections and @xmath2-@xmath3 mass splitting .
our model prediction is substantially reduced compared to other soliton calculations , and is closer to the experimental value . #
10= 0=0 1= 1=1 0>1 # 1 / # 1 |
since the first experimental realizations @xcite of dilute - gas bose - einstein condensates , there has been much interest in the properties of quantum vortices in such systems , and the effect of thermal atoms on their dynamics and stability .
the description of the dynamics of vortices in the presence of thermal excitations provides a challenging test for dynamical theories of cold bosonic gases , and promises new insights into the role of thermal excitations in the dynamics of vortices in systems less amenable to such _ ab initio _
descriptions @xcite .
the stability of a vortex state in a dilute condensate was first investigated theoretically by rokhsar @xcite , who showed that a vortex is subject to decay in the presence of a nonrotating thermal cloud , providing a physical interpretation of a negative - energy excitation of the vortex previously found by dodd _
fedichev and shlyapnikov @xcite then put forward an analytic theory of the dynamics of vortex decay in the presence of a nonrotating thermal cloud , based on a two - fluid model of superfluidity .
@xcite subsequently observed experimentally a nonexponential decay in vortex survival probability , and the first evidence of an increase in the displacement of the vortex from the trap center during the decay , which was also observed in the experiments of anderson _ et al .
zhuravlev and co - workers @xcite provided an analytic description of the decay of vortex arrays , building on the work of @xcite , and including the rotational dynamics of a nonstationary thermal cloud on the basis of the theory of gury - odelin @xcite .
their work predicted two limiting regimes of relaxation dynamics , depending on the relative efficiencies of vortex - cloud and cloud - trap coupling : the so - called _ rotating trap _ limit , in which the vortex array and thermal cloud relax collectively as a single rigid body , and the _
static trap _
limit , in which the thermal cloud is quickly arrested by the trap anisotropy , and nonexponential decay @xcite of the array rotation follows .
abo - shaeer _ et al . _
@xcite performed experiments in the rotating trap regime , and observed the expected exponential decay and strong temperature dependence of the decay rate .
_ @xcite observed the decay of a single vortex experimentally , and found a much less severe dependence on temperature than that of @xcite .
they conjectured that the thermal cloud was rapidly arrested by residual trap anisotropy , leading to the static - cloud vortex - decay scenario of @xcite .
classical - field simulations performed by schmidt _ et al .
_ @xcite focused on the dynamics of a strongly nonequilibrium ` phase - imprinted ' vortex state .
@xcite presented an analytical description of the decay of a vortex in a nonrotating thermal cloud , obtained from the stochastic field theory of @xcite using a variational ansatz @xcite , which was subsequently extended to include the effects of cloud rotation by bradley and gardiner @xcite . in the present paper
we consider a condensate containing a vortex initially at thermal and rotational equilibrium with a rotating thermal cloud in a highly oblate trap , which is isotropic in the plane .
such a configuration is obtained as an ergodic classical - field equilibrium with fixed angular momentum on the order of @xmath0 per atom about the trap axis @xcite . due to the conservation of angular momentum , this rotating equilibrium configuration is stable , provided that the trapping potential remains invariant under rotations about its axis .
we then introduce an elliptical deformation of the trap which breaks this rotational symmetry , leading to the loss of angular momentum from the atomic field and thus the slowing of the rotating cloud , and consequently the decay of the vortex .
our simulations are the first to describe the arrest of the rotation of both the condensed and noncondensed components of the field following the introduction of a trap anisotropy , and our method describes the coupled relaxation dynamics of the two components implicitly . as predicted by @xcite , we find that the response of the condensed and noncondensed components of the field may be different depending on the relative efficiencies of vortex - cloud and cloud - trap coupling , where the latter depends on the ellipticity of the trap deformation . by varying the anisotropy over a range of values and monitoring the evolution of the condensed and noncondensed components of the field , we observe a rich phenomenology , ranging from an adiabatic steady state , to violently nonequilibrium dynamics , in which the rotation of the thermal cloud essentially decouples from that of the condensate .
this paper is organized as follows . in sec .
[ sec : formalism ] we discuss our classical - field formalism , and the parameters of the physical system we simulate . in sec .
[ sec : results ] we discuss the results of a representative simulation . in sec .
[ sec : anisotropy_dependence ] we discuss the dependence of the dynamics of the vortex and thermal field on the trap anisotropy . in sec .
[ sec : conclusions ] we summarize our findings and present our conclusions .
the general formalism of ( projected ) classical - field methods has recently been reviewed at length in @xcite , and so here we merely outline the particular method we use in this work . the system we represent
is described by the second - quantized , cold - collision ( @xmath1-wave ) hamiltonian @xmath2 where @xmath3 is the bosonic field operator , which satisfies @xmath4=\delta(\mathbf{x}-\mathbf{x}')$ ] , and @xmath5 is the interaction potential written in terms of scattering length @xmath6 and atomic mass @xmath7 .
the single - particle hamiltonian is of form @xmath8 where @xmath9 is a cylindrically symmetric trapping potential and @xmath10 is an additional anisotropic potential . in deriving the classical - field method
, we replace the field operator @xmath3 with the projected classical field @xmath11 .
the sum here is over eigenmodes of the cylindrically symmetric single - particle hamiltonian [ @xmath12 with eigenvalues satisfying @xmath13 , where @xmath14 is the _
cutoff energy_. the projected classical field is thus a vector in a low - energy subspace ( @xmath15 ) of the appropriate single - particle hilbert space . making this replacement
we obtain the classical hamiltonian @xmath16 defining the projector @xmath17 we can express the hamilton s equation for @xmath18 obtained from eq .
( [ eq : hcf ] ) as @xmath19 which is termed the projected gross - pitaevskii equation @xcite . in the present work we use the rotationally invariant ( gauss - laguerre ) projector introduced in @xcite . for a discussion of the evaluation of projections of the 2-body interaction and trap anisotropy operations in this basis
, we refer the reader to @xcite and @xcite respectively .
we simply note here that this method conserves both the energy and normalization of the field , and that the field angular momentum obeys the ehrenfest relation of the continuous field theory @xmath20 [ where the bar denotes spatial averaging : @xmath21 , which is a consequence of the relation @xmath22=0 $ ] .
we note finally that as our formalism describes only the low - energy region @xmath15 , it does not account for any population above the cutoff .
the results of any classical - field theory are in general cutoff dependent , and the cutoff dependence of the formalism used here was quantified in @xcite .
as noted in that work , ultimately this dependence can only be eliminated by the inclusion of above - cutoff physics , and currently no prescription for including such effects in general nonequilibrium scenarios is known . in practice , however , the low - energy region defined by @xmath23 , with @xmath24 the chemical potential ( see ref . @xcite and refs . therein ) is expected to include the most important dynamical mechanisms , the above - cutoff atoms influencing the dynamics only at a quantitative level . following @xcite , we choose physical parameters corresponding to @xmath25 atoms confined in a strongly oblate trap , with trapping frequencies @xmath26 rad / s .
the radial harmonic confinement defines the units of length ( @xmath27 ) and time [ one trap cycle ( cyc ) @xmath28 in which we will quote our results .
the @xmath1-wave scattering length is @xmath29 , placing our system in the quasi - two - dimensional regime ( oscillator length @xmath30 ) , with an effective two - dimensional ( 2d ) interaction parameter @xmath31 .
we are thus justified in representing our system by a 2d model , provided that we maintain thermodynamic parameters ( chemical potential @xmath24 and temperature @xmath32 ) such that @xmath33 @xcite at all times during the evolution .
the 2d representation is formalized in our classical - field method by choosing a cutoff @xmath34 , such that the low - energy space @xmath15 excludes all modes with excitation along the @xmath35-axis . in all our simulations
this energy cutoff ( projection operation ) is effected in an inertial ( laboratory ) frame , and the low - energy space consists of @xmath36 single - particle modes .
we form a finite - temperature initial state following the procedure of @xcite : we construct a ( nonequilibrium ) randomized classical - field configuration over the modes comprising the space @xmath15 , with chosen normalization , energy and angular momentum first integrals , and we evolve this state for some time ( @xmath37 trap cycles ) , so that the field has time to migrate to an equilibrium configuration .
following @xcite , we take as the starting point for forming these configurations the ground gross - pitaevskii ( gp ) eigenstate with @xmath38 atoms in a frame rotating at angular velocity @xmath39 .
this state has chemical potential @xmath40 and energy @xmath41 in an inertial frame , and angular momentum @xmath42 . to this state
we add energy and angular momentum , forming a configuration with @xmath43 , and @xmath44 .
we find that the corresponding equilibrium state is one in which a single vortex precesses very close to the trap axis , i.e. , with precession radius @xmath45 , where the healing length @xmath46 is estimated from the density of the ground state @xcite . in this configuration
the vortex is at equilibrium with the thermal component of the field , and the angular velocity @xmath47 of the condensate ( which is the same as the cloud s at equilibrium ) is close to that at the @xmath48 limit of the mechanically unstable @xcite precessing - vortex branch @xcite .
if additional angular momentum is added to this state , the condensate mode becomes unstable to the nucleation of a second vortex , i.e. , the single - vortex state we consider is essentially saturated with angular momentum for the given values of the other conserved integrals ( normalization and energy ) .
we find for this equilibrium initial state @xmath49 , @xmath50 , @xmath51 and condensate fraction @xmath52 .
having formed this initial state , we evolve field trajectories in the laboratory frame in the presence of trap anisotropies in the range @xmath53
. this procedure corresponds in each case to the sudden introduction of the anisotropy at time @xmath54 .
we evolve each of these trajectories with an adaptive integrator @xcite , with accuracy chosen such that the relative change in field normalization is @xmath55 per time step taken , until the vortex is expelled from the condensate and the total angular momentum of the field has been lost through its interaction with the anisotropy .
in this section we consider the results of a classical - field simulation with a particular choice of the trap anisotropy , @xmath56 .
the relaxation dynamics of the classical field in this case exhibit many features of interest , and serve as a point of comparison for simulations with weaker or stronger anisotropies .
position - space densities of the classical field at representative times are presented in fig .
[ fig : density_plots1](a - f ) .
initially the vortex precesses very close to the trap axis [ fig .
[ fig : density_plots1](a ) ] , and it remains close to the trap axis for some time .
[ fig : density_plots1](b ) shows the density at time @xmath57 cyc , in which the vortex core remains within @xmath58 of the trap axis . by @xmath59
cyc the vortex has begun to spiral out of the central density bulk which contains the condensate .
[ fig : density_plots1](c ) shows that by @xmath60 cyc the vortex has undergone significant radial displacement .
this increase in radial displacement continues [ fig .
[ fig : density_plots1](d ) ] , until the vortex core approaches the violently evolving condensate boundary [ fig .
[ fig : density_plots1](e ) ] . at @xmath61
cyc the vortex is lost into the peripheral thermal material , leaving the condensate vortex - free and essentially irrotational [ fig .
[ fig : density_plots1](f ) ] .
we now quantify the behavior of the vortex during its decay .
we observe that the vortex executes an essentially spiral - like motion as it decays , in agreement with the predictions of @xcite and the simulations of @xcite .
however , the vortex trajectory here is strongly stochastic , and the vortex orbits the trap axis many ( @xmath62 ) times during its decay , so we do not present the full vortex trajectory . to characterize the radial drift of the vortex , we track its location at a frequency of 25 samples per trap cycle , and average the resulting series over intervals of 100 samples ( @xmath63 cyc ) in order to smooth out rapid fluctuations due to thermal density fluctuations in the background field against which the vortex moves and uncertainties introduced by sampling the vortex location on a cartesian grid . in fig .
[ fig : vortex_rad_freq](a ) we present the smoothed vortex radius ( @xmath64 ) data .
( color online ) quantities characterizing the vortex motion .
( a ) vortex radial displacement and ( b ) angular velocity of the condensate orbital.,scaledwidth=45.0% ] it is clear from fig .
[ fig : vortex_rad_freq](a ) that there is an initial lag of @xmath65 cyc between the introduction of the trap anisotropy at @xmath54 cyc and the beginning of the upward trend in vortex displacement .
we note that the displacement radius does not increase monotonically , but exhibits large oscillations during its increase .
the period of these oscillations is often @xmath66 trap cycles , spanning many periods of the vortex orbit , in contrast to the trajectories presented in @xcite , in which the vortex generally precesses at a greater radius on each passage about the trap center . in fig .
[ fig : vortex_rad_freq](b ) we plot the vortex precession frequency . in practice , to evaluate this frequency , we follow the procedure of @xcite , forming the covariance matrix ( classical one - body density matrix ) @xmath67 where @xmath68 denotes a time average , in a frame rotating at angular velocity @xmath69 about the trap axis .
we vary @xmath69 such that the largest eigenvalue of @xmath70 is maximized .
the value of @xmath69 at which the maximum occurs is thus that of the rotating frame in which the coherent fraction of the classical field is most stationary , which forms a best estimate for the condensate angular frequency ( vortex precession frequency ) @xmath47 @xcite .
as we consider a nonequilibrium scenario , we construct eq .
( [ eq : density_matrix ] ) by averaging over short time periods @xcite , in each case calculating the average of 250 consecutive classical - field samples over a ten cycle period . fig .
[ fig : vortex_rad_freq](b ) shows that the vortex precession frequency determined in this manner increases as the vortex displacement increases , as is well known for a zero - temperature condensate @xcite and observed in the simulations of @xcite .
moreover , we observe that the oscillations in vortex radius @xmath71 at late times @xmath72 cyc are accompanied by oscillations in the condensate angular velocity .
the oscillations in the two are positively correlated , i.e. , the decreases in vortex radius are associated with periods of slowing of the condensate rotation during its otherwise steady increase with time .
the above definition of the condensate mode in terms of short - time covariance matrix eigenvectors allows us to resolve the dynamics of the condensed and noncondensed components of the field . as in @xcite
, we introduce the decomposition of the classical - field one - body density matrix in terms of its eigenvectors @xmath73 and corresponding eigenvalues @xmath74 ( indexed in order of decreasing eigenvalue ) , @xmath75 which separates it into condensed and noncondensed parts . as in @xcite
, we then define the averages of a single - body operator @xmath76 in the condensed and noncondensed components of the field by @xmath77 and @xmath78 respectively .
we use this decomposition to calculate the angular momentum of the condensate ( @xmath79 ) and thermal cloud ( @xmath80 ) .
these quantities are presented in fig .
[ fig : ang_mom_and_vel](a ) , along with the total angular momentum of the field ( @xmath81 ) .
( color online ) quantities characterizing the temporal evolution of the condensate and thermal cloud .
( a ) angular momenta of the condensate and thermal cloud , and total angular momentum of the field .
( b ) angular velocities of the condensate and thermal cloud.,scaledwidth=45.0% ] we observe that in the first few hundred trap cycles after the introduction of the trap anisotropy , the angular momentum of the thermal cloud [ solid line in fig .
[ fig : ang_mom_and_vel](a ) ] undergoes approximately exponential decay , reaching values close to zero by @xmath82 trap cycles .
by contrast , the angular momentum of the condensate ( dashed line ) is essentially unchanged during this period , which we identify with the initial period of quiescence of the vortex noted in sec .
[ subsec : vortex_trajectory ] . at @xmath83 cyc ,
the angular momentum of the condensate begins to decay , and after @xmath84 cyc the condensate is essentially irrotational , with some small value of angular momentum remaining in the form of surface excitations .
we note that the angular momentum of the thermal field component _ increases _ over the period @xmath85 $ ] cyc .
we understand this as follows : initially the angular momentum of the thermal cloud is lost due to its interaction with the trap anisotropy .
this makes the vortex thermodynamically unstable , and thus it begins to decay , liberating its angular momentum to the cloud .
the cloud loses its angular momentum to the trap at a rate proportional to its angular velocity ( frictional loss @xmath86 @xcite ) , which is initially small .
therefore , the cloud s angular momentum increases as it gains angular momentum from the vortex faster than it loses it to the trap .
as the cloud s rotation rate increases the rate of dissipation of @xmath80 increases , however , the precessing vortex condensate exhibits a thermodynamic anomaly : as the condensate loses angular momentum , its angular velocity _ increases _ @xcite .
the rate at which the angular momentum of the vortex is dissipated to the thermal cloud is proportional to their relative velocities @xcite , and so in the present case this transfer dominates the loss rate of angular momentum of the thermal cloud throughout the vortex decay . as shown in fig .
[ fig : ang_mom_and_vel](b ) , this leads to ` run - away ' spin - up of the vortex precession frequency , with the thermal - cloud rotation rate being driven up in sympathy .
closer inspection of the angular momenta @xmath87 and @xmath88 reveals negatively correlated oscillations in the angular momenta of the two components during the decay , as a result of the detailed nonlinear dynamics of angular momentum exchange and loss during the decay process .
we identify this oscillation in angular momentum transfer with the oscillations in vortex rotation rate discussed in sec .
[ subsec : vortex_trajectory ] . in fig .
[ fig : ang_mom_and_vel](b ) we plot the angular velocity of the condensate @xmath47 ( dots ) derived in the estimation of the condensate fraction ( sec .
[ subsec : vortex_trajectory ] ) , and the angular velocity of the thermal cloud @xmath89 ( line ) , where we assume the expectation value of the classical moment of inertia @xmath90 as an estimate of the cloud s true moment of inertia . in @xcite
we discussed the level of uncertainty inherent in this procedure , nevertheless it yields a clear qualitative description of the decay dynamics .
the oscillations in rotation rate of the thermal cloud are clearly visible here , and we note that they are _ positively _ correlated with the oscillations in angular velocity of the condensate , due to its anomalous rotational response .
we note that oscillatory behavior arises already in a linear analysis of the arrest of a rotating boltzmann gas @xcite by a trap anisotropy .
it is therefore not surprising that similar oscillations occur in the transfer of angular momentum from the condensate to the nonequilibrium thermal field , considering the complexity of their coupled dynamics ( cf .
@xcite ) , and the anomalous response of the vortical condensate .
finally , we note that the cloud angular momentum ( rotation rate ) reaches its peak when the vortex leaves the condensate ( @xmath84 cyc ) , after which it undergoes a second near - exponential decay phase , the angular momentum of the field decaying such that by @xmath91 cyc , only the thermodynamic fluctuations in @xmath80 exhibited by the finite - temperature field at rest in the anisotropic potential remain .
we now consider the heating of the atomic field during the arrest of its rotation .
as the system we evolve is hamiltonian , with a time - independent potential , the total energy of the classical field is a constant of the motion .
consequently , the trap anisotropy dissipates the angular momentum of the field by converting the rotational kinetic energy of the field into internal energy @xcite , and we therefore expect some heating of the field to occur , due to this redistribution of energy .
we can estimate the heating of the field as follows : the rotational energy of the gas is initially @xmath92 .
we compare this to the initial thermal energy , which should be reasonably well estimated by the energy added to the ground vortex state in forming the initial thermal state ( sec .
[ subsec : simulation_procedure ] ) , @xmath93 .
assuming a linear relationship between thermal energy and temperature ( which should be valid for the low temperatures we consider here , see @xcite ) , we might therefore expect an increase in the field temperature of @xmath94 during the arrest of the condensate rotation . in order to quantify the heating and its development during the field evolution
, we estimate the ( effective ) thermodynamic parameters of the field ( chemical potential @xmath24 and temperature @xmath32 ) using the procedure of @xcite . as in @xcite
, we allow for the differential rotation between the thermal cloud and the frame in which the classical - field cutoff is effected ( the laboratory frame , in the present case ) . we perform the fit to the wing of the ( time - averaged ) radial distribution @xmath95 of the classical field , i.e. , we fit over the radii range @xmath96 $ ] where @xmath97 marks the minimum of the hartree - fock effective potential @xmath98r^2 + 2u_\mathrm{2d } n(r)$ ] @xcite , and @xmath99 is the semiclassical turning - point of the low - energy space @xmath15 @xcite .
we fit to radial densities of the classical field averaged over @xmath100 cyc periods , and assume the values of @xmath101 calculated in sec .
[ subsec : rotational_dynamics ] .
the resulting estimates for @xmath24 and @xmath32 are presented in fig .
[ fig : cfrac_and_temp](a ) .
( color online ) heating of the field . evolution of ( a ) the effective chemical potential and temperature and ( b ) the condensate fraction.,scaledwidth=45.0% ] we find that the temperature ( dashed line ) is approximately constant during the first @xmath65 cycles of the field evolution , corresponding to the initial arrest of the thermal field component .
the initial angular momentum of the thermal component is small , comprising some @xmath102 of the total angular momentum of the field , and so any heating of the field due to the redistribution of the associated rotational energy is possibly too small to resolve above the uncertainty in the temperature estimates .
beginning at @xmath83 cyc , corresponding to the start of the vortex - decay phase , the temperature exhibits a steady , approximately linear increase .
we associate this increase with the conversion of rotational kinetic energy to thermal energy , due to the action of the trap anisotropy , and presumably also as a result of the scattering of excitations by the vortex which produces the frictional effect @xcite .
there is a final sharp rise in temperature at @xmath103 trap cycles .
this rise is perhaps due to the final decay of the vortex into excitations at the surface of the condensate @xcite .
we note however that the field , and in particular the condensate surface , undergoes strong fluctuations as the system passes through this transition to the vortex - free state , with , e.g. , multiple ( ghost ) vortices present at the condensate surface at times .
this can be viewed as the system essentially ` reversing ' through the surface - wave instability arising from the relative motion of condensate and thermal cloud , in which vortices spontaneously grow from surface excitations @xcite .
the thermodynamic parameters may therefore be ill - defined during this period .
after this period of strong surface fluctuations , the temperature levels off at @xmath104 , corresponding to heating of @xmath105 during the arrest .
we note that the chemical potential [ solid line in fig .
[ fig : cfrac_and_temp](a ) ] exhibits a slight downward trend beginning around @xmath83 cyc , falling by @xmath106 , though we note that this change is of the same order as the variation in estimates for @xmath24 obtained at late times @xmath107 cyc . in fig .
[ fig : cfrac_and_temp](b ) we plot the condensate fraction @xmath108 obtained from the procedure outline in sec .
[ subsec : vortex_trajectory ] .
the fluctuations in estimates of this quantity are large , as is expected given the short time scale ( @xmath100 cyc ) over which the appropriate averages are taken , and the nonequilibrium nature of the field .
however the condensate fraction estimates exhibit a clear downwards trend as time proceeds , and the condensate fraction drops by @xmath109 during the arrest .
we now consider the effect of varying the trap anisotropy @xmath110 on the behavior of the classical - field trajectories .
we intuitively expect the rate at which the trap dissipates the angular momentum of the thermal cloud to depend strongly on the magnitude of the anisotropy , and a quantitative model for this dependence in the case of a classical ( boltzmann ) gas was presented in @xcite . by contrast , the efficiency with which the thermal cloud extracts angular momentum from the condensate is dictated by the ( longitudinal ) mutual - friction coefficient , which depends on the temperature and chemical potential of the field , in addition to its microscopic properties @xcite .
we therefore expect , as discussed in the vortex - continuum analysis of @xcite , to explore different regimes of relaxation dynamics as we vary the trap anisotropy and consequently the relative strengths of vortex - cloud and cloud - trap friction . in this section
we consider the evolution of the classical field in the presence of anisotropies of different magnitudes , and observe the resulting differences in the behavior of the fields . in fig .
[ fig : adiabatic_and_not](a ) we plot the evolution of the angular momentum of the condensate and thermal cloud in the presence of a weak trap anisotropy ( @xmath111 ) .
we observe that after an initial decline over the first @xmath112 trap cycles , the cloud angular momentum , despite exhibiting large fluctuations , maintains a reasonably steady mean value which increases slowly over the period @xmath113 $ ] trap cycles . during this period the angular momentum of
the condensate drops by approximately a factor of @xmath114 , and large fluctuations in the distribution of angular momentum between the two components are visible .
this behavior is again apparent in the vortex precession frequency and cloud rotation rate presented in fig .
[ fig : adiabatic_and_not](b ) .
this shows that the cloud exhibits a steady - state nonequilibrium behavior , i.e. , its rotation rate remains somewhere between that of the precessing condensate mode , and that of the trap ( @xmath115 ) .
in this regime its angular momentum remains approximately constant , as the rate at which it gains angular momentum from the decaying vortex matches the rate at which it loses angular momentum to the trap .
we expect the angular velocity at which this balance occurs , and indeed whether such a regime occurs at all , to depend strongly on the relative strengths of the vortex - cloud and cloud - trap friction .
we note that the angular velocity of the cloud appears to slowly increase over time , as the condensate loses angular momentum and its angular velocity increases , shifting the cloud rotation rate at which the angular momentum transfer rates are balanced . at @xmath116
cyc the vortex s precession accelerates as it approaches the condensate boundary , and during the period @xmath117 $ ] cyc the adiabaticity of the vortex dissipation is lost as the field enters the critical regime associated with nucleation of the vortex at the surface @xcite . in this regime
the condensate surface is unstable and as noted in sec .
[ subsec : heating ] , multiple ` ghost ' vortices may be present simultaneously at the condensate boundary .
a large amount of angular momentum is transferred non - adiabatically to the thermal cloud during this period of criticality , while small surface - mode oscillations of the condensate persist until @xmath118 cyc .
the angular momentum is subsequently dissipated from the thermal cloud over a period of @xmath119 cyc , in an oscillatory but approximately exponential decay phase .
we now turn our attention to a scenario of vortex arrest due to the presence of a strong trap anisotropy ( @xmath120 ) .
the dynamics in this case are strongly nonequilibrium and the arrest of the field s rotation occurs on a shorter time scale than the cases already considered , and so here we form the density matrix eq .
( [ eq : density_matrix ] ) by averaging classical - field samples over a shorter period of two trap cycles . in this case
the angular momentum of the thermal cloud [ solid line in fig .
[ fig : adiabatic_and_not](c ) ] is rapidly depleted , and actually fluctuates below zero by @xmath121 trap cycles .
the condensate is slower to respond ; during this period the condensate angular momentum decreases by @xmath122 .
subsequently , the cloud angular momentum fluctuates strongly , and rises to @xmath123 ( close to its initial value ) , as the vortex is rapidly expelled from the condensate .
the angular momentum is then dissipated from the cloud over a period @xmath124 cyc , and thereafter
the angular momentum of the field fluctuates about zero .
the rapid expulsion of the vortex , and strong fluctuations of the cloud rotation , are again visible in the calculated angular velocities [ fig .
[ fig : adiabatic_and_not](d ) ] .
the behavior of the cloud angular momentum in this trajectory suggests that the cloud is overcritically damped by the trap anisotropy @xcite .
its response to the anisotropy appears unhindered by its coupling to the condensate via the vortex core , and it quickly yields the angular momentum it acquires from the vortex to the trap , despite its small angular velocity @xmath125 .
the relaxation process in this case is violently non - adiabatic , as evidenced by the strong fluctuations in the cloud angular momentum . indeed during the period @xmath126 $ ] cyc large , long - wavelength surface oscillations are visible in the field as the vortex precesses rapidly near the condensate boundary .
it appears that the strong trap anisotropy and rapid ` stirring ' motion of the vortex conspire to strongly perturb the condensate bulk in this regime . during this period
the measured condensate fraction is suppressed [ inset to fig . [
fig : adiabatic_and_not](d ) ] due to the strongly nonequilibrium behavior of the condensate , in which surface - wave excitations define frames of rotation distinct from that of the vortex .
the decomposition of the field into condensed and noncondensed components must therefore be viewed with some caution in this strongly nonequilibrium scenario .
nevertheless , it is clear that the effect of the trap on the cloud dominates the vortex - cloud coupling in this scenario , in stark contrast to the near - adiabatic decay scenario of sec .
[ subsec : weak_damping ] .
we now consider the dependence of the relaxation times on the strength of the trap anisotropy . as
the relaxation of both the condensate and the thermal cloud is generally nonexponential , we consider the times at which the condensate becomes irrotational , and at which the total angular momentum of the field is lost . due to the persistence of surface excitations which prevent the condensate angular momentum from reaching zero ,
we define the condensate stopping time @xmath127 as that at which the angular momentum of the condensate first drops below @xmath128 per particle , and the field stopping time as the first _ subsequent _ time @xmath129 at which the cloud angular momentum reaches zero .
these times are presented in fig .
[ fig : relaxation ] .
( color online ) relaxation of the field . circles ( plusses ) represent the times at which the vortex leaves the condensate ( total field angular momentum reaches zero ) .
inset : cloud spin - down times and the linear fit performed to extract their scaling with trap ellipticity.,scaledwidth=45.0% ] we observe that the condensate arrest times ( circles in fig . [ fig : relaxation ] ) vary by two orders of magnitude over the range of trap anisotropies we consider .
moreover , the total - field relaxation times ( plusses ) become increasingly longer than the vortex relaxation times as the trap anisotropy is weakened , causing slower dissipation of the cloud angular momentum even in the absence of the vortex .
we therefore consider the time @xmath130 over which the angular momentum of the cloud dissipates following the expulsion of the vortex .
although the angular momentum lost in this final damping phase varies over the range @xmath131 , precluding a precise analysis , the times @xmath132 provide a useful characterization of the dependence of the cloud relaxation on the trap anisotropy .
we perform a linear fit to @xmath133 as a function of @xmath110 in log - log space ( inset to fig . [
fig : relaxation ] ) , and find the scaling @xmath134 , in good agreement with the scaling @xmath135 for thermal - cloud spin - down predicted for weak anisotropies by gury - odelin @xcite .
we have carried out the first simulations of the arrest of a rotating bose - einstein condensate due to the presence of a trap anisotropy which includes the coupled nonequilibrium dynamics of the condensate and thermal cloud .
our method makes no assumptions of stationarity of a thermal bath , nor are the two components artificially given disparate rotational parameters .
rather , our approach describes the dynamical migration of an _ equilibrium _ rotating thermal state to a new , irrotational equilibrium , solely due to the action of the trapping potential .
we observe for all parameters we considered that the rotation rates of the condensate and thermal component are distinct during the decay .
the anomalous rotational response of the precessing - vortex condensate can lead to a counter - intuitive spin - up of the thermal component during the decay , and we observe nonequilibrium oscillations in transfer of angular momentum between the two components as the vortex responds to the dissipative effect of the thermal cloud , which is itself damped by the trapping potential anisotropy . for trap anisotropies that are weak , the thermal field settles to a nonequilibrium steady state , with rotation rate intermediate between that of the condensate and that of the ( static ) trap . in this scenario the angular momentum of the condensate
is slowly depleted while that of the thermal cloud remains nearly constant , until the vortex nears the condensate boundary and the linearity of the vortex decay breaks down . for stronger trap anisotropies ,
the angular momentum of the thermal cloud may be almost entirely depleted before the condensate responds .
we quantified the heating of the atomic field during the arrest , and found it to be commensurate with the conversion of rotational kinetic energy into thermal energy by the trap anisotropy .
we also considered the time scales over which the condensate angular momentum and total field angular momentum were dissipated , and found reasonable agreement between the scaling of the thermal cloud spin - down time with trap anisotropy and the predictions of a boltzmann gas model @xcite .
possible extensions of this work include the description of vortex - lattice arrest , and the effects of temperature on the decay dynamics .
in particular it would be interesting to study how the nonequilibrium rotational dynamics of the condensate approach those of a rigid body as the vortex density increases , and the effect of temperature on the cloud rotation in the steady - state scenario .
we wish to acknowledge valuable discussions with a. polkovnikov .
this work was supported by the new zealand foundation for research , science and technology under contract no .
nerf - uoox0703 and no . | we perform finite - temperature dynamical simulations of the arrest of a rotating bose - einstein condensate by a fixed trap anisotropy , using a hamiltonian classical - field method .
we consider a quasi - two - dimensional condensate containing a single vortex in equilibrium with a rotating thermal cloud .
introducing an elliptical deformation of the trapping potential leads to the loss of angular momentum from the system .
we identify the condensate and the complementary thermal component of the nonequilibrium field , and compare the evolution of their angular momenta and angular velocities . by varying the trap anisotropy
we alter the relative efficiencies of the vortex - cloud and cloud - trap coupling . for strong trap anisotropies
the angular momentum of the thermal cloud may be entirely depleted before the vortex begins to decay . for weak trap anisotropies ,
the thermal cloud exhibits a long - lived steady state in which it rotates at an intermediate angular velocity . |
centaurus a ( ngc 5128 ) is the nearest ( d = 3.5 mpc ; 1 @xmath117 pc , hui et al . 1993 ) example of a giant elliptical galaxy associated with a powerful radio source .
the large - scale radio morphology consists of twin radio lobes separated by @xmath1 5 degrees on the sky .
the compact ( @xmath1 milliarcsecond ) radio nucleus is variable and has a strong jet extending @xmath1 4 arcminutes towards the northeast lobe .
the spectacular optical appearance is that of a giant elliptical galaxy that appears enveloped in a nearly edge on , warped dust lane . there is also a series of faint optical shells .
the stellar population in the dominant elliptical structure is old , whilst that of the twisted dust lane is young , sporadically punctuated by hii regions , dust and gas ( graham 1979 ) .
the overall structure of cen a resembles that of a recent ( @xmath9 years , tubbs 1980 ) merger , between a spiral and a large elliptical galaxy .
the dust lane is the source of most ( 90 % ) of the far - infrared luminosity ( l@xmath10 l@xmath11 ) and is thought to be re - radiated starlight from young stars in the dusty disk ( joy et al .
1988 ) . in sect .
2 we describe the observations and data analysis . sect .
3 looks at the general fir properties and proceeds to model the hii regions and the pdrs in the dust lane .
4 summarises the results and presents our conclusions .
cen a was observed with the lws grating ( @xmath12 ) as part of the lws consortium s guaranteed time extragalactic programme . a full grating observation ( 43 - 196.7 @xmath0 m )
was taken of the nucleus at the centre of the dust lane and a series of line observations were taken at two positions in the se and nw regions of the dust lane .
a short 157 @xmath0 m line observation was taken off - source at position # 4 ( see table 1 ) to estimate the galactic emission near the source .
position # 1 was intended to provide a deeper integration coincident with position # 2 , but was accidently offset .
a series of half - second integration ramps were taken at each grating position with four samples per resolution element ( @xmath13 m @xmath14 m and @xmath15 m @xmath16 m ) .
the total integration time per resolution element and per pointing were : position # 1 88 s for the 52 @xmath0 m and 34 s for the 57 @xmath0 m ; position # 2 ( the centre ) , 30 s for the range 43196 @xmath0 m ; positions nw and se ( 2 point raster map ) 22 s for the the 63 @xmath0 m , 14 s for the 88 @xmath0 m , 12 s for the 122 @xmath0 m , 28 s for the 145 @xmath0 m and 12 s for the 158 @xmath0 m ; position # 4 12 s for the 158 @xmath0 m .
the data were processed with ral pipeline 7 and analysed using the lia and isap packages . the lws flux calibration and relative spectral response function ( rsrf )
were derived from observations of uranus ( swinyard et al . 1998 ) .
the full grating spectrum at the centre enabled us to estimate the relative flux uncertainty between individual detectors arising from uncertainties in the relative responsivity and the dark - current subtraction . the offsets between the detectors ( excluding detector sw1 ) was @xmath17 % .
the 88 @xmath0 m line on detectors sw5 and lw1 had a 15 % systematic uncertainty and the line on detectors lw3 and lw4 had a 10 % systematic uncertainty .
we therefore adopt a relative flux uncertainty of @xmath1 15% . because we only took spectra of individual lines at the nw and se positions there is no corresponding overlap in wavelength coverage at these positions
. one indicator of relative flux uncertainty is a discrete step down in flux , of @xmath1 25 % , at @xmath1 125 @xmath0 m at the se position .
the relative flux uncertainty is assumed to be @xmath18 25 % at these positions .
the absolute flux calibration w.r.t .
uranus for point like objects observed on axis is better than 15 % ( swinyard et al .
1998 ) . however , extended sources give rise either to channel fringes or to a spectrum that is not a smooth function of wavelength .
this is still a calibration issue .
for example , in fig .
2 , detectors sw5 , lw1 , lw2 have slopes that differ from those of their neighbours in the overlap region .
this may account for the continuum shape , which is discussed in sect .
the lws beam profile is known to be asymmetric and is still under investigation .
we therefore adopt a value for the fwhm of 70 at all wavelengths , believing that a more sophisticated treatment would not significantly affect our conclusions .
we also note that there is good cross calibration between the iso - lws results and the far - infrared imaging fabry - perot interferometer ( fifi ) ( madden et al .
1995 ) ; the peak fluxes agree to within @xmath1 10 % . .
observation log [ cols= " < , > , > , < " , ] the results for the three regions are consistent with each other , having a gas density , n @xmath1 10@xmath19 @xmath7 , and an incident far - uv field , g @xmath1 10@xmath20 . at the nw position , only the combination of the 63 @xmath0 m / 158 @xmath0 m ratio and the ( 63 @xmath0 m + 158 @xmath0 m ) /fir continuum ratio gives a meaningful solution for g and n. the 146 @xmath0 m line is clearly detected but with a very rippled baseline due to channel fringes .
the observed 146 @xmath0 m line flux would need to be reduced by @xmath1 60 % in order to obtain a consistent result with the 146 @xmath0 m / 63 @xmath0 m line ratio predicted by the pdr model .
the lws results for the nucleus confirm those previously derived from ir , submm and co observations .
the consistent set of derived pdr conditions for all three positions suggest that the observed fir emission in a 70 beam centred on the nucleus is dominated by star formation and not agn activity .
joy et al .
( 1988 ) mapped cen a at 50 and 100 @xmath0 m on the kao .
they concluded that the extended fir emission was from dust grains heated by massive young stars distributed throughout the dust lane , not the compact nucleus .
hawarden et al .
( 1993 ) mapped cen a at 800 @xmath0 m and 450 @xmath0 m with a resolution of @xmath110 .
they attribute the large scale 800 @xmath0 m emission to thermal emission from regions of star formation embedded in the dust lane .
they note that the h@xmath21 emission within a few arcseconds of the nucleus , observed by israel et al .
( 1990 ) , indicates that significant uv radiation from the nucleus does not reach large radii in the plane of the dust lane i.e. the nuclear contribution to exciting the extended gas and dust disk is small .
eckart et al . ( 1990 ) and wild et al .
( 1997 ) mapped cen a in @xmath22co j=1 - 0 , @xmath22co j=2 - 1 and @xmath23co j=1 - 0 .
all three maps have two peaks separated by @xmath1 90 centred on the nucleus .
it is interesting to note that our se position only clips the lowest contours of the co ( 1 - 0 ) and co ( 2 - 1 ) maps of wild et al .
( 1997 ) . in spite of this
the derived pdr parameters are consistent with those encompassing the bulk of the molecular emission .
there must be extended low level co ( 1 - 0 ) emission beyond the sensitivity limits of the wild et al .
( 1997 ) maps .
the lowest contour is 17.5 k kms@xmath24 , corresponding to m@xmath25 @xmath1 10@xmath26 m@xmath11 if the material filled the lws beam .
we present the first full fir spectrum from 43 - 196.7 @xmath0 m of cen a. we detect seven fine structure lines ( see table 2 ) , the strongest being those generated in pdrs . at the central position ,
the total flux in the far - infrared lines is @xmath1 1 % of the total fir luminosity ( l@xmath27 l@xmath11 for a distance of 3.5 mpc ) .
the line flux is @xmath10.4 % fir and the line flux is @xmath1 0.2 % fir .
these are typical values for starburst galaxies ( lord et al .
the 52 @xmath0 m / 88 @xmath0 m line intensity ratio is @xmath1 0.9 , which corresponds to an electron density , n@xmath28 100 @xmath7 ( rubin et al .
the _ thermal pressure _ of the ionized medium in the cen a dust lane is closer to that of starburst galaxies ( n@xmath29 250 @xmath7 in m82 ( colbert et al .
1999 ) and m83 ( stacey et al . 1999 ) ) than that of the milky way ( n@xmath29 3 @xmath7 ( pettuchowski & bennett 1993 ) ) .
the / line intensity ratio is @xmath1 1.6 , giving an abundance ratio n++/n+ @xmath1 0.3 , which corresponds to an effective temperature , t@xmath30 35500 k ( rubin et al . 1994 ) . assuming a coeval starburst , then the tip of the main sequence is headed by o8.5 stars , and the starburst is @xmath1 6 @xmath4 years old .
if the burst in cen a was triggered by the spiral - elliptical galaxy merger then its occurance was very recent .
alternatively , the merger triggered a series of bursts of star formation and we are witnessing the most recent activity .
we estimate that the n / o abundance ratio is @xmath1 0.2 in the hii regions in cen a. this value is consistent with the range of @xmath1 0.2 - 0.3 found for galactic hii regions ( rubin et al .
n / o is a measure of the chemical evolution and we expect it to increase with time ( c.f . the solar value of @xmath1 0.12 ) .
we estimate that @xmath1 10 % of the observed arises in the wim .
the cnm contributes very little ( @xmath31 % ) emission at our beam positions .
the bulk of the emission is from the pdrs .
we derive the average physical conditions for the pdrs in cen a for the first time .
there is active star formation throughout the dust lane and in regions beyond the bulk of the molecular material .
the fir emission in the 70 lws beam at the nucleus is dominated by emission from star formation rather than agn activity . on scales of @xmath1 1
kpc the average physical properties of the pdrs are modelled with a gas density , n @xmath1 10@xmath19 @xmath7 , an incident far - uv field , g @xmath1 10@xmath20 and a gas temperature of @xmath1 250 k.
many thanks to the dedicated efforts of the lws instrument team . the iso
spectral analysis package ( isap ) is a joint development by the lws and sws instrument teams and data centers .
contributing institutes are cesr , ias , ipac , mpe , qmw , ral and sron .
colbert j.w . , malkan m.a . ,
clegg p.e . , et al . , 1999 ,
apj 511 , 721 + eckart a. , cameron m. , rothermel h. , et al . , 1990 ,
apj 363 , 451 + graham j. , 1979 , apj 232 , 60 + hawarden t.g . , sandell g. , matthews h.e . , et al . , 1993 ,
mnras 260 , 844 + heiles c. , 1994 , apj 436 , 720 + hui x. , ford h.c . , ciardillo r. , et al . , 1993 ,
apj 414 , 463 + israel f.p . , van dishoeck e.f . , baas f. et al . , 1990 ,
a&a 227 , 342 + joy m. , lester d.f . , harvey p.m. , et al .
, 1988 , apj 326 , 662 + kaufman m.j .
, wolfire m.g .
, hollenbach d. , et al .
, 1999 , apj in press + kulkarni s.r . , heiles c. , 1987 , in hollenbach , d. , thronson jr , h.a .
interstellar processes .
reidel , dordrecht , p. 87
+ launay j.m . ,
roueff e. , 1977 , jphysb 10 , 879 + lester d.f . ,
dinnerstein h.l . ,
werner m.w . , et al . , 1987 ,
apj 320 , 573 + lord s.d .
, malhotra s. , lim t.l . , et al .
, 1996 , a&a 315 , l117 + madden s.c . ,
geis n. , genzel r. , et al , 1993 , apj 407 , 579 + madden s. , geis n. , townes c.h.et al . , 1995 , airbourne astronomy symposium on the galactic ecosystem , asp conf .
series 73 , 181 .
+ meyer d.m . , jura m. , cardelli j.a . , 1998 ,
apj 493 , 222 + petuchowski s.j .
, bennett c.l . , 1993 ,
apj 405 , 591 + rubin r.h . , 1985 , apjs 57 , 349 + rubin r.h . , simpson j.p . ,
erickson e.f .
, et al . , 1988 ,
apj 327 , 377 + rubin r.h . ,
simpson j.p . , lord s.d . , et al . , 1994 , apj 420 , 772 + sofia u.j . , cardelli j.a . , guerin k.p . ,
et al . , 1997 , apj 482 , l105 + stacey g.j . , swain m.r .
, bradford c.m . , et al . , 1999 , the universe as seen by iso , esa sp-427 p973 + swinyard b.m . , burgdorf m.j . ,
clegg p.e et al .
, 1998 , spie 3354 + tubbs a.d .
, 1980 , apj 241 , 969 + van gorkom j.h . , van der hulst j.m . ,
haschick a.d . , et al . , 1990 aj 99 , 1781 + wild w. , eckart a. , wilkind t. , 1997 , a&a 322 , 419 + wolfire m.g . , tielens a.g.g.m . , hollenbach d. , 1990 , apj 358 , 116 + wolfire m.g . , tielens a.g.g.m
, hollenbach d. , 1995 , apj 443 , 152 + | we present the first full fir spectrum of centaurus a ( ngc 5128 ) from 43 - 196.7 @xmath0 m .
the data was obtained with the iso long wavelength spectrometer ( lws ) .
we conclude that the fir emission in a 70 beam centred on the nucleus is dominated by star formation rather than agn activity .
the flux in the far - infrared lines is @xmath1 1 % of the total fir : the line flux is @xmath1 0.4 % fir and the line is @xmath1 0.2 % , with the remainder arising from , and lines .
these are typical values for starburst galaxies .
the ratio of the / line intensities from the hii regions in the dust lane corresponds to an effective temperature , t@xmath2 @xmath1 @xmath3 k , implying that the tip of the main sequence is headed by o8.5 stars and that the starburst is @xmath1 6 @xmath4 years old .
this suggests that the galaxy underwent either a recent merger or a merger which triggered a series of bursts .
the n / o abundance ratio is consistent with the range of @xmath1 0.2 - 0.3 found for galactic hii regions .
we estimate that @xmath5 5 % of the observed arises in the cold neutral medium ( cnm ) and that @xmath1 10 % arises in the warm ionized medium ( wim ) .
the main contributors to the emission are the pdrs , which are located throughout the dust lane and in regions beyond where the bulk of the molecular material lies . on scales of @xmath1 1
kpc the average physical properties of the pdrs are modelled with a gas density , n @xmath1 @xmath6 @xmath7 , an incident far - uv field , g @xmath1 @xmath8 times the local galactic field , and a gas temperature of @xmath1 250 k. 0h_0 |
the study of a germ of a vector fields in @xmath2 with two pairs of non - resonant imaginary eigenvalues reduces to the study of the special perturbations @xmath3 of the integrable quadratic foliation @xmath4 with first integral @xmath5 , e.g. ( * ? ? ?
* , section 4.6 ) .
equivalently , we may consider the following generalized lotka - volterra system associated to ( [ lv ] ) @xmath6\\ y ' = & y[x - p(1-x - y ) ] \end{aligned}\ ] ] and the perturbed foliation ( [ lvp ] ) is associated then to @xmath7\\ y ' = & y[x - p(1-x - y ) + \varepsilon_1 + \varepsilon_2 x^2 ] .
\end{aligned}\ ] ] the limit cycles of the perturbed system correspond to the zeros of the displacement map @xmath8 where @xmath9 is a continuous family of ovals ( closed orbits ) of the non - perturbed system ( [ glv ] ) .
the unicity of the limit cycle of ( [ lvp]),([glvp ] ) was shown first by ( * ? ? ?
* zoladek,1986 ) , who proved the monotonicity of the function @xmath10 on the maximal interval @xmath11 where the ovals of @xmath12 exist .
note that the system ( [ glvp ] ) has @xmath13 and @xmath14 as invariant lines . in this relation , recall that a plane quadratic vector field with an invariant line has a unique limit cycle ( if any ) , as explained by ( * ? ? ?
* coppel , 1989 ) , but see also ( * ? ? ?
* zegeling and kooij , 1994 ) .
the case of a more general quadratic perturbations of ( [ glv ] ) was studied also by ( * ? ? ?
* zoladek , 1994 ) and revised recetly in @xcite .
the commont point of the above mentioned papers is , that they use _ ad hoc _ methods based on apriori estimates .
the essential reason why all these estimates hold remains hidden . in 1985
van gils and horozov @xcite gave an overview of the varous methods , which have been used at this time to prove the uniqueness of the limit cycles for the perturbations of the generalized lotke - volterra system ( [ glv ] ) .
the central result of their paper is that in the particular case , in which @xmath15 is an integer , the functions @xmath16 satisfy a picard - fuchs equation of second order , whose coefficients are rational in @xmath17 .
the authors used then topological arguments as the rolle s theorem , to bound the zeros of @xmath18 in terms of the degrees of the coefficients of the picard - fuchs equation , from which the result of zoladek @xcite follows . more precisely , if @xmath19 and @xmath20 are the critical values of the lotka - volterra integral @xmath21 , then the abelian integral @xmath18 ( [ abelian ] ) allows an analytic continuation from @xmath22 to a small neighbourhood of @xmath23 .
this follows from the picard - lefschetz formula and the fact , that the oval @xmath24 represents a cycle vanishing at @xmath23 . at the other end of the interval , at @xmath25 ,
the function is not analytic , but has a logarithmic type of singularity , as it follows from a generalised picard - lefschetz formula @xcite .
namely @xmath26 for suitable functions @xmath27 , which are analytic in a neighbourhood of @xmath28 .
note that @xmath29 is an analytic continuation of @xmath18 and hence it is an abelian integral too .
it has therefore a similar logarithmic type singularity at @xmath30 .
following @xcite , denote by @xmath31 the wronskian @xmath32 the picard - lefschetz formula then implies , that @xmath33 as a function in @xmath34 allows an analytic continuation from the interval @xmath11 to a neighborhood of @xmath35 , and also that @xmath36 allows an analytic continuation from the interval @xmath11 to a covering of a punctured neighbourhood of @xmath37 , in which @xmath31 is analytic in @xmath38 . from this , the authors concluded that @xmath31 is in fact a rational function in @xmath38 , and even computed it explicitly ( * ? ? ?
* lemma 1 ) .
this conclusion , that @xmath31 is a rational function in @xmath17 is , however , wrong .
indeed , there exist functions , analytic on the universal covering of @xmath39 , with a branch on @xmath11 which are analytic at @xmath23 , analytic in @xmath17 at @xmath25 , but still not algebraic in @xmath34 .
to construct an example , consider a second order fuchs equation with singular points at @xmath40 and riemann scheme @xmath41 where @xmath42 .
let @xmath18 be a ( branch of a ) solution on @xmath11 , analyic in a small neighbourhood of @xmath30 .
for generic values of the characteristic exponent @xmath43 ( or @xmath44 ) our equation has no algebraic solutions
. therefore @xmath18 can not be analytic in a neighbourhood of @xmath25 too .
it is concluded that @xmath18 is analytic in @xmath17 in a neighbourhood of @xmath25 . clearly , this argument points out a gap , but does not disproof the result of @xcite .
the purpose of the present note is to study in more detail the monodromy representation of the lotke - volterra polynomial @xmath45 , in the case when @xmath46 is an integer .
the knowledge of this monodromy representation allows , according to @xcite , to compute the minimal degree of the differential equation satisfied by @xmath18 .
we shall show in this way , that in the first non trivial case @xmath47 , the abelian integral @xmath18 satisfies a linear differential equation of minimal degree four , even if the coefficients are supposed to be algebraic functions .
thus , the result of ( * ? ? ?
* lemma 1 ) is definitely wrong .
_ note also , that the the computation of @xcite , and in particular corollary 4 there , are also wrong .
we prove in fact a more general result about the attached lie group @xmath48 , which is the zarisky closure of the monodromy group of the polynomial @xmath49 .
namely , we show that in the case @xmath47 the group @xmath48 is isomorphic to the symplectic group @xmath1 , see theorem [ case2 ] .
as the standard representation of @xmath1 is irreducible , then according to corolary [ md ] we obtain _ the abelian integral ( [ abelian ] ) satisfies a fuchs type equation of minimal degree four , even if its coefficients are supposed to be algebraic functions in @xmath34 .
_ the paper is organized as follows . in the next short section
we give some background , concerning the reduction of the degree of picard - fuchs operators . in section [ section3 ]
we determine explicitely the monodromy operators , related to the two singular critical values of @xmath50 .
this ( long ) computation is contained in principal in @xcite , in the case of arbitrary integers @xmath51 and @xmath52 . in the particular case @xmath15 , part of these computations
simplify , and for this reason we give here an independent treatment .
a straightforward observation is , that the monodromy representation is reducible .
we have in fact a two - dimensional plane @xmath53 of zero - cycles on which the monodromy acts as identity , as well a complementary @xmath54-dimensional plane @xmath55 , invariant under the action of the monodromy group .
the nature of this sub - representation @xmath55 is studied in the last section [ section4 ] in the simplest non - trivial case @xmath47 . by taking the zarisky closure of the monodromy group , we find that the sub - representation @xmath56 coincides with the standard representation of the symplectic group @xmath1 .
this , combined with section [ section3 ] implies the claims about the degree of the picard - fuchs equation .
in this section we summarize , following @xcite , the necessary facts about the reduction of the degree of picard - fuchs differential operators . to a non - constant polynomial
@xmath57 $ ] we associate its monodromy representation @xmath58 where @xmath59 are the fibers of the fibration @xmath60 .
the image of the fundamental group @xmath61 is the monodromy group @xmath62 .
the zarisky closure @xmath63 of @xmath62 is a linear algebraic group , embedded in @xmath64 , @xmath65 .
it is nothing but the differential galois group of a generic picard - fuchs system , related to the fibration defined by @xmath66 . in the sequel , an important role
is played by the connected component @xmath67 of @xmath48 , containing the identity transformation .
let @xmath68 be a continuous family of cycles , and @xmath69 a polynomial one - form .
the abelian integral @xmath70 satisfies a picard - fuchs equation of minimal degree @xmath71 @xmath72 whose coefficients are rational functions in @xmath73 , @xmath74 .
let @xmath75 be a vector plane , invariant under @xmath48 .
then obviously @xmath76 .
let @xmath77 and @xmath78 the corresponding continuous family of cycles ( a locally constant section of the homology bundle ) .
the set of such @xmath79 with the property @xmath80 is an invariant sub - plane of @xmath81 which we denote by @xmath82 .
we conclude that @xmath83 .
consider now the following reduction problem .
: _ find a differential equation @xmath84 of minimal degree @xmath85 , such that @xmath86 are algebraic functions in @xmath73 .
_ the computation of the degree @xmath87 goes along the same lines as above , except that the lie group @xmath48 is replaced by @xmath67 .
namely , let @xmath88 be a plane , invariant under @xmath67 , that is to say a sub - representation of @xmath67 .
such a plane was called virtually invariant in @xcite .
it follows that @xmath89 .
let @xmath90 and @xmath78 the corresponding continuous family of cycles as above .
the set of such @xmath79 with the property @xmath80 is a virtually invariant sub - plane of @xmath91 which we denote by @xmath92 .
we conclude that @xmath93 .
if the representation of @xmath67 on @xmath91 is irreducible , then the minimal degree of the differential operator ( [ md ] ) equals @xmath94 .
see @xcite for proofs .
denote @xmath96 for some @xmath97 .
the polynomial @xmath66 has two critical values @xmath98 and
@xmath99 and non - isolated critical points along the lines @xmath100 and @xmath101 . for @xmath102 the algebraic curve @xmath103 is a smooth riemann surface and its fundamental group has @xmath104 generators , see fig.[fig1 ] . as @xmath105 has three punctures ( at infinity ) , then it is a genus @xmath46 algebraic curve .
we wish to describe the `` continuous variation '' of @xmath105 when @xmath73 varies along closed circuits in @xmath106 , or equivalently , the topology of the vibration @xmath107 the calculation of this geometric monodromy is in general a difficult task , see the survey of siersma @xcite . to begin with ,
we first localize @xmath66 at the singular points @xmath108 and @xmath109 and obtain a germ of analytic function with non - isolated critical points .
we study first their geometric monodromy , which is straightforward .
let @xmath111 be a germ of analytic function , such that @xmath112 defines a germ of a divisor with simple normal crossing and multiplicities @xmath113 and @xmath46 .
in appropriate coordinates in a suitable neighborhood of the origin we have @xmath114 , which will be assumed until the end of this section .
the _ marked fibration _ associated to @xmath66 is by definition the usual fibration @xmath115 whose fibers are the riemann surfaces ( topological cylinders ) @xmath116 with @xmath117 marked points corresponding to the intersection of @xmath118 with the two fixed lines @xmath119 and @xmath120 @xmath121 continuous deformation of @xmath73 induces an isotopy of the marked fibers @xmath118 .
a continous variation of @xmath73 along a closed circuit about the origin induces therefore a diffeomorphism ( geometric monodromy ) @xmath122 defined up to an isotopy , which permutes the marked points .
we wish to describe @xmath123 along the same lines , as in the case of an isolated singularity of morse type , @xmath124 , e.g. @xcite .
it is convenient to consider the fundamental groupoid @xmath125 of the pair @xmath126 @xcite , which replaces the common fundamental group @xmath127 in the case @xmath124 .
recall that this groupoid is the set of homotopy classes of loops @xmath128
\to f_t$ ] such that @xmath129 .
two loops @xmath130 are composable , if @xmath131 and in this case the homotopy class of @xmath132 is well defined . in the case when @xmath133 is a single point @xmath134 is the usual fundamental group of the fiber @xmath118 .
the fundamental groupoid @xmath135 is freely generated by @xmath117 loops , as shown on fig.[fig2 ] in the particular case @xmath136 .
the geometric monodromy @xmath123 defines a homomorphisms @xmath137 clearly , @xmath123 permutes cyclically the points @xmath138 and fixes @xmath139 .
define a loop @xmath140 connecting @xmath139 to a point in the set @xmath141 and then inductively @xmath142 [ paths ] the @xmath117 loops @xmath143 generate the groupoid @xmath125 , the closed loop @xmath144 generates the fundamental group @xmath145 .
( see fig .
[ fig2 ] ) .
the projection @xmath146 maps @xmath118 isomorphically to @xmath147 , and the images of the marked points are @xmath148 alternatively , in the ball @xmath149 , for sufficiently small @xmath150 the surface @xmath151 is projected under @xmath152 to the annulus shown on fig.[fig2 ] .
when @xmath73 makes one turn around the origin , in a clockwise direction , the marked points @xmath153 permute cyclically , and the corresponding paths @xmath154 connecting @xmath113 to @xmath153 are as on fig . [ fig2 ] . note that in the case @xmath124 we get the usual picard - lefschetz formula . for a further use ,
note the following [ cor1 ] it is always possible to choose the initial loop @xmath140 in such a way , that @xmath155 , @xmath156 , are non - intersecting , and non self - intersecting loops . in this case
the union @xmath157 is an embedded planar graph , which is a deformation retract of the marked cylinder @xmath118 .
consider finally the relative homology group @xmath158 which is isomorphic as a @xmath159-module to @xmath160 .
the generators of @xmath161 are represented by the paths @xmath143 defined in proposition [ paths ] .
we conclude that the corresponding equivalence classes of paths @xmath162 , [ \gamma_1 ] , \dots , [ \gamma_{p-1 } ] , [ \gamma_p ] = [ \alpha ] + [ \gamma_0 ] \in h_1(f_t , s_t ) .\ ] ] define a basis of the relative homology group .
the diffeomorphism @xmath163 induces a homomorphism ( picard - lefschetz monodromy operator ) @xmath164 note that @xmath165=[\alpha]$ ] and in the basis @xmath162 , [ \gamma_1 ] , \dots , [ \gamma_{p-1 } ] , [ \alpha]\ ] ] @xmath166 is represented by the matrix @xmath167 with characteristic polynomial @xmath168 our aim in this section is to construct explicit generators for the fundamental group of @xmath118 which allows a simple description of the action of the monodromy transformations of the fibration , defined by @xmath66 .
the generators will be presented in the form of an embedded graph , which is a deformation retract of @xmath118 ( see corollary [ cor1 ] ) .
namely , consider the real triangle , with vertices the singular points @xmath170 in @xmath171 .
let @xmath172 be a suitable tubular neighborhood of this triangle .
we suppose that @xmath173 is transversal to the complex lines @xmath174 and moreover @xmath175 is a deformation retract of @xmath176 .
it follows that for sufficiently small @xmath150 the border @xmath173 is transversal also to @xmath118 , and that @xmath177 is a deformation retract of @xmath118 .
this allows to localize our description of the fiber @xmath118 near the triangle with vertices @xmath170 , see fig .
[ fig4 ] .
consider three cross - sections ( complex discs ) transversal to the sides , dividing the triangle into three pieces , with corresponding tubular neighborhoods @xmath178 , where @xmath179 .
the cross sections intersect the fiber @xmath180 for sufficiently small non - zero @xmath150 in exactly @xmath46 , @xmath46 and @xmath113 points respectively , which will be the marked points from the section [ toy ] .
let @xmath181 be the tubular neighborhood , containing @xmath182 .
then @xmath183 has @xmath46 connected components ( topological cylinders ) which coincide with the fibers of a morse polynomial , and it is retracted to @xmath46 disjoint segments .
the fibers @xmath184 and @xmath185 are described as in the section [ toy ] , corollary [ cor1 ] .
we construct now a graph , embedded in @xmath177 which is a deformation retract of @xmath118 . for this purpose
we take together the corresponding graphs near the singular point @xmath170 , constructed in section [ toy ] .
the assembling of these graphs is shown on fig . 3 in the case @xmath136 , the general case of arbitrary @xmath46 being analogous .
it is easy to check ( by making use of a partion of the unity ) that the resulting graph is embedded in @xmath118 and is a deformation retract of @xmath118 .
the next step is to describe the action of the monodromy transformation @xmath187 , corresponding to the singular value @xmath186 , which once again easily follows from the preceding constructions .
more specifically , we shall describe the linear operator @xmath188 .
we begin by choosing a suitable basis of @xmath189 . in a suitable neighborhood of @xmath190
the fiber @xmath191 has @xmath46 connected components homeomorphic to cylinders , and each component carries a vanishing cycle denoted @xmath192 at @xmath108 and @xmath109 we have vanishing cycles denoted respectively @xmath193 and @xmath194 .
we have therefore @xmath195 vanishing cycles , additional @xmath46 cycles are constructed as follows @xmath196 where @xmath197 is represented by the oval of @xmath118 for @xmath198 . the @xmath104 cycles which we described are independent in @xmath189 .
the monodromy operator @xmath166 acts as follows @xmath199 and @xmath200 having described the monodromy of the fiber @xmath118 localized around the singular points of @xmath66 , everything is obvious , except @xmath201 . to compute @xmath202 we use the action of @xmath203 described in the relative homology of the local fibers . as noted in @xcite ( or see fig .
[ fig2 ] ) @xmath203 is a picard - lefschetz operator , and @xmath204 it is straigthtforward to check that the characteristic and the minimal polynomials of @xmath166 are equal respectively to @xmath205 the orbit of the cycle @xmath206 under the action of @xmath166 spans the @xmath54-dimensional vector space generated by @xmath207 the monodromy group is generated by two operators , @xmath209 and @xmath210 related to the singular values @xmath211 .
the value @xmath212 corresponds to a morse critical point and the operator @xmath209 is given by the picard - lefschetz formula .
its description amounts to compute the intersection form on @xmath189 .
[ intersections ] the intersection numbers of the @xmath104 cycles @xmath213 generating @xmath189 are as follows @xmath214 @xmath215 @xmath216 @xmath217 the first and the third lines are obvious . for the second , we choose an orientation on @xmath218 in such a way , that @xmath219 and then use the invariance of the intersection number under the action of the monodromy .
the only non - trivial fact is the fourth line .
we note that according to fig .
[ fig3 ] , the intersection number of the relative cycles @xmath220 , @xmath221 are well defined and equal @xmath222 . the fibration defined by @xmath66 in @xmath223 can be further continuously deformed in a way , which does not change the topology of the fibers and their monodromy .
namely , by such a deformation we may replace @xmath66 by @xmath224 and consider the linear projection @xmath225 parallel to the lines @xmath226 as in the section [ toy ] , the projection @xmath227 is an injectif local biholomorphism , and its image is shown on fig .
when @xmath73 makes one turn around the origin , the two marked points corresponding to the ends of the relative cycle @xmath228 turn in the same direction .
the result is the relative cycle @xmath229 shown on fig .
this already proves that @xmath230 and by invariance of the intersection form the numbers @xmath231 are all equal to the either @xmath232 or to @xmath233 .
we have finally @xmath234 and hence @xmath235 .
in this section we consider in detail the first non - trivial case @xmath47 , in which the fibers @xmath118 are genus two riemann surfaces with three punctures
. for definiteness , denote @xmath237 the monodromy operators associated to simple closed loops around @xmath212 or respectively @xmath99 .
the monodromy group @xmath238 of the polynomial @xmath236 is then the subgroup of @xmath239 generated by @xmath240 .
the smallest algebraic variety containing @xmath238 is an algebraic group , denoted @xmath48 .
it is the zarisky closure of @xmath238 .
we note that @xmath241 carries a ( degenerate ) intersection form @xmath69 of rank @xmath47 invariant under the action of @xmath240 .
it is easily verified , that @xmath238 and hence @xmath48 is isomorphic to a subgroup of the symplectic group @xmath1 .
we shall prove the following [ case2 ] the zarisky closure of the monodromy group of the polynomial @xmath236 is isomorphic to the symplectic group @xmath1 . to the end of the section we give the proof of this remarkable fact .
a basis of the first homology group @xmath189 will be chosen as in the preceding section @xmath242 where @xmath243 is a cycle vanishing at the unique morse critical point when @xmath73 tends to @xmath212 , @xmath218 are vanishing cycles at the singular points @xmath244 , and @xmath245 are cycles vanishing at @xmath182 .
the cycle @xmath246 is the image of @xmath206 under the action of the monodromy operator about the singular value @xmath186 , @xmath247 .
the cycles ( [ basis ] ) are represented by closed loops on the riemann surface @xmath118 , and by abuse of notation we denote these loops by the same letter .
the closed loops can be chosen in a way that their union is a deformation retract of @xmath118 , see fig.[fig1 ] . to compute @xmath209
we note that according to proposition [ intersections ] the sign of intersection indexes of the cycles of @xmath189 can be chosen as follows [ cols="^,^,^,^,^,^,^",options="header " , ]
let @xmath248 be an arbitrary non - constant polynomial .
the set @xmath249 of its atypical values is finite and therefore we can consider the associated monodromy representation of the fundamental group of @xmath250 on @xmath251 .
let @xmath48 be the zarisky closure of this group , with associated lie algebra @xmath252 .
what can be said about @xmath253 ?
cearly @xmath48 preserves the intersection form of the first homology group @xmath251 .
the subplane @xmath254 of zero - cycles is invariant , and @xmath255 acts on it trivially .
therefore the representation of this fundamental group on @xmath256 is well defined too , and @xmath81 carries an invariant non - degenerate intersection form .
we conclude that @xmath48 is a subgroup of the symplectic group @xmath257 , where @xmath46 is the genus of the riemann surface of @xmath258 .
it is well known that for generic @xmath66 ( e.g. morse plus polynomials ) we have @xmath259 .
according to theorem [ case2 ] this holds true also in the special lotke - volterra case @xmath260 , @xmath47 .
_ we conjecture that @xmath259 for every integer @xmath261 . _ on the other hand , if @xmath66 is a composite polynomial , @xmath262 , where @xmath263 is a polynomial mapping , we can not expect that @xmath259 . a simple example is @xmath264 for @xmath265 a polynomial of degree
at least three .
the natural involution @xmath266 induces a decomposition @xmath267 where @xmath268 are invariant under @xmath48
so @xmath269 . we note that by the ritt theorem @xcite a univariate polynomial @xmath270 $ ] is composite if and only if its monodromy group is imprimitive .
are there examples of non - composite bivariate polynomials @xmath248 , such that @xmath269 ? v. i. arnold , v. s. afrajmovich , yu . s. ilyashenko , and l. p. shilnikov . .
springer - verlag , berlin , 1999 .
reprint of the 1994 english edition from the series encyclopaedia of mathematical sciences _ dynamical systems .
v _ , encyclopaedia math . sci .
, 5 , springer , berlin , 1994 .
dirk siersma .
the vanishing topology of non isolated singularities . in _
new developments in singularity theory ( cambridge , 2000 ) _ , volume 21 of _ nato sci . ser
. ii math .
_ , pages 447472 .
kluwer acad .
publ . , dordrecht , 2001 .
stephan a. van gils and emil horozov .
uniqueness of limit cycles in planar vector fields which leave the axes invariant . in _
multiparameter bifurcation theory ( arcata , calif . , 1985 )
_ , volume 56 of _ contemp .
_ , pages 117129 .
soc . , providence , ri , 1986 . | we prove that the zarisky closure of the monodromy group of the polynomial @xmath0 is the symplectic group @xmath1 .
this shows that some previous results about this monodromy representation are wrong . |
the magnetocrystalline anisotropy energy ( mae ) is at the heart of magnetic properties of materials .
it is of crucial importance from the fundamental or technological point of views since it provides an energy scale for the stability of magnetic domains where for example magnetic information is stored . when the mae is large and favors an out - of - plane orientation of the magnetic moments , perpendicular magnetic recording or magneto - optical recording is possible ( see e.g. refs@xcite ) .
copt binary bulk alloy in the * l1@xmath0 * structure ( see fig.[cell_bulk ] ) is by now a classical example of a material exhibiting a large perpendicular mae , around 1 mev @xcite .
there has been a tremendous number of studies related to the magnetic properties of this alloy in its bulk phase , as nanoparticles or in nanostructures combining co and pt ( see e.g. refs .
@xcite ) . * alloy .
the primitive cell is also sketched using dashed lines . ] a large amount of work has been devoted to unveil the origin of the large perpendicular mae in binary bulk alloys , see e.g. refs@xcite .
the interplay of the tetragonality of the alloy , band filling , hybridization between the constituents affect certainly the magnitude of the mae .
for instance , tetragonality leads to the lifting of the degeneracy of the electrons by the tetragonal crystal field and produces thereby an additional contribution to the mae .
thus , and as expected from perturbation theory , the mae becomes proportional to @xmath1 instead of @xmath2 as found for cubic symmetry , where @xmath3 is the spin - orbit coupling constant .
indeed , in cubic bulk systems , the high symmetry allows only for a fourth - order anisotropy constant , and thus they are characterized by a small mae .
razee et al.@xcite argued however that the tetragonal distortion of copt , given by the axial ration c / a = 0.98 , contributes by only 15% of the mae .
it was then concluded that the compositional order of the alloy is an important ingredient for a large mae .
sakuma@xcite shows that by changing the axial ratio ( c / a ) defining the tetragonality of copt and fept alloys , the mae first smoothly decreases by increasing c / a till reaching a minimum at @xmath40.8 before a smooth increase in magnitude .
interestingly , except a small window of axial ratios ( @xmath5 ) , the mae favors an out - of - plane orientation of the magnetic moments .
the tetragonalization is then thought to provide an effect similar to the band filling@xcite . in the context of thin films ,
zhang et al.@xcite demonstrated with ab - initio simulations that for copt films terminated by co layers , a thickness of at least 9 monolayers exhibit a rather converged mae , with a bulk contribution of 1.36 mev favoring a perpendicular orientation of the magnetic moments and a counter - acting surface contribution of -0.76 mev favoring , interestingly , an in - plane orientation of the moments .
their interest in copt was motivated by the experimental demonstration of coercivity manipulation of * l1@xmath0 * fept and fepd thin films@xcite by external electric field .
their ab - initio simulations predicted a higher sensitivity of copt to electric field than that of fept films .
pustogowa et al.@xcite investigated from first - principles several components made of co and pt deposited on pt(100 ) and pt(111 ) surfaces .
they found that ordered superstructures of ( copt)@xmath6 deposited on both mentioned substrates are characterized by a perpendicular mae , which is heavily affected by chemical disorder in line with the analysis of razee et al.@xcite .
the goal of this manuscript is to present a systematic ab - initio investigation on the effect of reduced dimensionality on the magnetic properties of copt(100 ) films with a focus on their mae and by addressing the impact of the termination type of the films .
contrary to previous investigations@xcite , we consider not only co - terminated films but also pt - terminated films and several types of surface defects ( see fig.[cell_surf ] ) .
for instance , we found that decreasing the thickness of the films leads to a sign change of the surface mae .
pt covered thin films can boost the total perpendicular mae by a large amount stabilizing , thereby , more strongly the out - of - plane orientation of the moments .
molecular dynamics simulations demonstrated the likeliness of having pt on the surface of copt alloy@xcite and thus the pertinence of our predictions .
after a careful study of different defective terminations types ( stacking faults , anti - site defects ) , we provide the ingredient to increase the mae of the thin films .
if we label the co and pt layers by respectively a and b , the perfect stacking along the [ 001 ] direction is given for example by ababab for 6 layers .
possible stacking faults , which are planar defects , could be the sequence ababaa ( see figs.[cell_surf](c - d ) ) .
anti - site defects on the surface means that instead of having at the surface a pure layer a , or layer b , we have an alloy , for example , made of a and b. in our work , we considered an alloy of the type a@xmath7b@xmath8 in the surface layer instead of the perfect b layer of our example ( see figs.[cell_surf](e - f ) ) .
we simulate the thin films by adopting the slab approach with periodic boundary conditions in two directions while the periodic images in the third direction are separated by a sufficient amount of vacuum ( 15 ) to avoid interaction between neighboring supercells .
we have chosen to use symmetrical calculation cells with an odd number of planes to avoid the pulay stress .
some representative slabs are shown in fig.[cell_surf ] . here
it can be observed that for equiatomic * l1@xmath0 * type of alloys two different surfaces exist when the slabs are stacked along the [ 001 ] direction . in the perfect cases ,
the surface termination can be made of either purely co atoms or pt atoms .
the self - consistent calculations are carried out with the vienna ab initio simulation package ( vasp ) using a plane wave basis and the projector augmented wave ( paw ) approach @xcite . the exchange - correlation potential is used in the functional from of perdew , burke and ernzerhof ( pbe ) @xcite .
the cut - off energies for the plane waves is 478 ev .
the integration over the brillouin zone was based on finite temperature smearing ( methfessel - paxton method ) for the thin films while for the bulk case the tetrahedron method with blchl corrections has been used .
the k - points grids are @xmath9 for the bulk calculation , and @xmath10 for the ( 001 ) surface calculations .
the energy convergence criterion is set to @xmath11 ev while the geometrical atomic relaxations for the surfaces calculations were stopped when the forces were less than 0.01 ev / .
the mae is extracted from the difference between the total energies of the two configurations : out - of - plane versus in - plane orientations of the magnetic moments .
a positive value indicates a preference for the out - of - plane orientation of the magnetic moments .
copt alloy as function of the axial ratio c / a under constant volume .
two possible in - plane orientation of the magnetic moments are considered , [ 100 ] and [ 110 ] , but the obtained mae are very similar .
the closed circle represents the experimental value@xcite , which is well reproduced by our simulations .
other experimental values can be 50% larger , see e.g. ref .
to start our investigations , we revisited the bulk alloy phase by finding the tetragonal lattice structure minimizing its energy .
the optimal value of c / a ratio is equal to 0.984 with a lattice parameter value a of 3.80 , in good agreement with values available in the literature ( see e.g. @xcite ) .
the calculated magnetic moments ( @xmath12 = 1.91 @xmath13 , @xmath14 = 0.40 @xmath13 ) are also in line with previous works @xcite .
we also note the well - known emergence of an induced moment in pt , which is due to the hybridization of its 5d orbitals with the exchange splitted 3d orbitals of co. in order to calculate the mae , we considered two possible directions for the in - plane moments orientations , [ 110 ] and [ 001 ] , and we found a negligible difference in the obtained perpendicular mae . for the optimized structure ,
the mae reaches a value of 0.91 mev when the in - plane orientation of moments is along [ 100 ] and 0.97 mev for [ 110 ] as an in - plane orientation of the moments .
both values are close to the experimental values that are given around 1 mev @xcite . in fig.[mae_bulk ]
, we plot the bulk mae as function of the ratio c / a under the constant unit cell volume in a similar fashion then that of sakuma @xcite .
the obtained curve agrees well with the one published in the latter article . for ratios between 0.6 and 1.2 , the mae experiences one minimum and two maxima .
the largest in - plane mae is found for a ratio of 0.8 . as expected , if c / a = @xmath15 , i.e. c / a@xmath16 = 1 , the mae drops to zero since this corresponds to a primitive cell of the cubic b2 structure . as discussed by sakuma , upon tetragonalization , the electronic states of the atoms are shifted and the band filling changes , which affect the magnitude and sign of the mae . [ cols="^ " , ] another path for the analysis of the calculated maes is to use of the celebrated bruno s formula @xcite , which translates to the neglect of spin - flip contributions to the mae as given in eq.[mae_perturbation ] : @xmath17 } } - l^i_{\mathrm{[100]}}),\ ] ] where @xmath18 labels the different atoms , and @xmath19 being the orbital magnetic moment calculated when the spin magnetic moment points along the [ 100 ] or the [ 001 ] directions .
the essence of bruno s formula is to relate the mae to the orbital moment anisotropy ( oma ) , i.e. @xmath20 } } - l_{\mathrm{[001]}}$ ] , and leads to the conclusion that the orientation of the magnetic moments is favored when the orbital magnetic moment is maximized .
this formula is known to work reasonably well when the majority - spin states are occupied .
thus , its validity is probably limited to some of the co atoms discussed in this manuscript but certainly not for pt .
it is however instructive to analyze the results obtained with this well known formulation since it should correlate with the previous discussion . in fig.[anisotropy_orbital](a ) , the omas in the bulk of copt is plotted for co and pt as function of the c / a ratio in fashion similar to that used in fig.[mae_bulk ] .
one notices that the pt contribution counteracts the one of co. while the anisotropy of the co orbital moment increases with c / a , favoring thereby an out - of - plane orientation of the magnetic moment , the anisotropy of the pt orbital moment has an opposite slope and favors an in - plane orientation of the magnetic moment . when summing up the two curves , considering the spin - orbit coupling constant , @xmath3 , to be the same for co and pt , which is of course is not true since @xmath21 is one order of magnitude larger than @xmath22 , one recovers the shape of the curve obtained in fig,[mae_bulk ] , i.e. having a minimum of the curve at c / a = 0.8 .
similar to the bulk , the behavior of the co and pt oma in copt thin films counteract each other .
in general , the co oma favors an out - of - plane orientation of the moment contrary to the pt oma . in fig.[anisotropy_orbital](b ) , we plot the surface mae of the thin films characterized by @xmath23 as function of the pt concentration on the layer deposited on the co - terminated thin film . thus ,
in the case of one perfect pt overlayer the pt concentration is 100% , while the investigated anti - site corresponds to a pt concentration of 50% .
for the specific case of co stacking fault , the pt concentration is -100% .
the surface mae seems to increase with the pt concentration but not in a regular manner .
we plot on the same figure the average co oma per co atom and find that this quantity increases smoothly in magnitude with pt concentration .
in addition the contribution of the average pt oma per pt atom is shown in fig.[anisotropy_orbital](b ) .
the pt oma seems to correlate the irregular behavior of the surface mae .
interestingly , we find that thin films with large pt oma per pt atom compared to the co oma leads to an in - plane surface mae
. only the pt - terminated thin film , with a large perpendicular surface mae , is characterized by a large co oma .
we investigated from ab - initio the magnetic behavior of copt thin films as function of thickness considering different types of terminations : perfect co or pt layers or different types of defects : anti - site or stacking faults .
after this systematic study , we found that the mae is the largest when the thin films are terminated by a perfect pt overlayer .
surprisingly in the latter case , the mae can be 1000% times larger than the one of co - terminated thin films .
we also find that all types of investigated defects reduce dramatically the mae .
the surface mae experiences a sign change when increasing the thickness of several investigated films . except for the pt - terminated films
, the surface mae favors an in - plane orientation of the moments when the thickness @xmath24 is smaller then four .
we proceeded to an analysis of the electronic structure of the thin films with a careful comparison to the copt bulk case and related the behavior of the mae to the location of the different virtual bound states utilizing second order perturbation theory .
finally , the correlation between the mae and the oma is studied .
we are grateful to claude demangeat , vasile caciuc , julen ibanez - azpiroz and manuel dos santos dias for helpful discussions .
also we thank hongbin zhang for discussion and for providing us his data .
this work was supported by c. n. e. p. r. u project ( d 00520090041 ) of the algerian government , the hgf yig program vh - ng-717 ( functional nanoscale structure and probe simulation laboratory funsilab ) and the erc consolidator grant dynasore .
99 k. r. coffey , m. a. parker and j. k. howard , ieee trans .
magn . , * 31 * 2737 ( 1995 ) d. j. sellmyer , y. xu , m. yan and r. skomski , ieee trans . magn . * 41 * 560 ( 2005 ) p. eurin and j. pauleve , ieee trans .
magn . * 5 * , 216 ( 1969 ) v. v. maykov , a. ye yermakov , g. v. ivanov , v. i. khrabrov , and l. m. magat , phys . met .
metallogr . * 67 * , 76 ( 1989 ) a. ye .
yermakov and v. v. maykov phys . met .
metall . * 69 * 198 ( 1990 ) d. weller and a. moser , ieee trans .
magn . , * 35 * 4423 ( 1999 ) w. grange , i. galanakis , m. alouani , m. maret , j .-
kappler , and a. rogalev , phys .
b * 62 * , 1157 ( 2000 ) x. yang , c. yu . j. liu , t. klemmer , e. johns and d. weller , j. vac .
technol . ,
b * 22 * 31 ( 2004 ) a. perez , v. dupuis , j. tuaillon - combes , l. bardotti , b. prvel , e. bernstein , p. mlinon , l. favre , a. hannour and m. lamet , adv .
, * 7 * 475 ( 2005 ) l. szunyogh , p. weinberger and c. sommers , phys
b * 60 * 11910 ( 1999 ) p. entel and m. e. gruner , j. phys .
: condensed matter , * 21 * 064228 ( 2009 ) l. uba , s. uba , o. horpynyuk , v. n. antonov and a. n. yaresko j. appl
. phys . * 91 * 775 ( 2002 ) j. park , m. g. kim , y. jun , j. s. lee , w. lee and j. cheon , j. am .
soc * 126 * 9072 ( 2004 ) j. bartolom , l. m. garcia , f. bartolom , f. luis , f. petroff , c. deranlot , f. wilhelm and a. rogalev , j. magn .
mater . * 316 * e9 ( 2007 ) f. tournus , a. tamion , n. blanc , a. hannour , l. bardotti , b. prvel , p. ohresser , e. bonet , epicier and v. dupuis , phys .
b , * 77 * 144411 2008 ) d. alloyeau , c. ricolleau , c. mottet , t. oikawa , c. langlois , y. le bouar y , n. braidy and a. loiseau , nature mater . ,
* 8 * 940 ( 2009 ) n. blanc , f. tournus , v. dupuis and t. epicier , phys .
b , * 83 * 092403 ( 2011 ) s. ouazi , s. vlaic , s. rusponi , g. moulas , p. buluschek , k. halleux , s. bornemann , s. mankovsky , j. minr , j. b. staunton , h. ebert and h. brune , nat . commun * 3 * 2316 ( 2012 ) s. karoui , h. amara , b. legrand and f. ducastelle , j. phys . :
condensed matter * 25 * 056005 ( 2013 ) j. zemen , j. masek , j. kucera , j. a. mol , p. motloch and t. jungwirth , j. magn .
mater * 356 * 87 ( 2014 ) a. i. figueroa , j. bartolom , l. m. garca , f. bartolom , o. bunu , j. stankiewicz , l. ruiz , j. m. gonzlez - calbet , f. petroff , c. deranlot , s. pascarelli , p. bencok , n. b. brookes , f. wilhelm , a. smekhova and a. rogalev , phys .
b * 90 * 174421 ( 2014 ) bibitemtournus2008 f. tournus , a. tamion , n. blanc , a. hannour , l. bardotti , b. prvel , p. ohresser , e. bonet , t. epicier and v. dupuis , phys .
b * 77 * 144411 ( 2008 ) w. f. hu , h. k. yuan , h. chen , g. z. wang and g. l. zhang , phys .
a * 378 * 198 ( 2014 ) g. h. o. daalderop , p. j. kelly , and m. f. h. schuurmans , phys rev .
b * 44 * , 12054 ( 1991 ) a. sakuma , j. phys . soc .
japan * 63 * , 3053 ( 1994 ) i. v. solovyev , p. h. dederichs and i. mertig , phys .
b * 52 * 13419 ( 1995 ) p. oppeneer , j. magn
. magn . mater . * 188 * 275 ( 1998 ) s. s. a. razee , j. b. staunton , b. ginatempo , f. j. pinski , and e. bruno , phys .
lett . * 82 * , 5369 ( 2009 ) i. galanakis , m. alouani and h. dreyss , phys .
b * 62 * 6475 ( 2000 ) p. ravindran , a. kjekshus , h. fjellvag , p. james , l. nordstrm , b. johansson and o. eriksson , phys . rev .
b * 63 * , 144409 ( 2001 ) a. b. schick and o. n. mryasov , phy . rev .
b * 67 * 172407 ( 2003 ) g. kresse and j. furthmller , phys . rev .
b * 54 * 11169 ( 1996 ) g. kresse and d. joubert , phys . rev .
b * 59 * 1758 ( 1999 ) j. p.
perdew , k. burke and m. ernzerhof , phys .
* 77 * 3865 ( 1996 ) j. p.
perdew , j. a. chevary , s. h. vosko , k. a. jackson , m. r. pederson , d. j. singh and c. fiolhais , phys .
b * 46 * 6671 ( 1992 ) w. b. pearson , a handbook of lattice spacings and structures of metals and alloys ( oxford : pergamon ) ( 1964 ) m. c. cadeville and j. l. morn - lpez , phys
. rep . * 153 * 331 ( 1987 ) a. dannenberg , m. e. gruner , a. hucht and p. entel , phys .
b * 80 * 245438 ( 2009 ) v. dupuis , g. khadra , s. linas , a. hillion , l. gragnaniello , a. tamion , j. tuaillon - combes , l. bardotti , f. tournus , e. otero , p. ohresser , a. rogalev and f. wilhelm j. magn .
mater ( 2015 ) h. zhang , m. richter , k. koepernik , i. opahle , f. tasnadi , and h. eschrig , new j. phys . *
11 * , 043007 ( 2009 ) m. weisheit , s. fhler , a. marty , y. souche , c. poinsignon , d. givord , science * 315 * , 349 ( 2007 ) u. pustogowa , j. zabloudil , c. uiberacker , c. blaas , p. weinberger , l. szunyogh , c. sommers , phys .
b * 60 * , 414 ( 1999 ) | the impact of reduced dimensionality on the magnetic properties of the tetragonal * l1@xmath0 * copt alloy is investigated from ab - initio considering several kinds of surface defects . by exploring the dependence of the magnetocrystalline anisotropy energy ( mae ) on the thickness of copt thin films
, we demonstrate the crucial role of the chemical nature of the surface .
for instance , pt - terminated thin films exhibit huge maes which can be 1000% larger than those of co - terminated films .
besides the perfect thin films , we scrutinize the effect of defective surfaces such as stacking faults or anti - sites on the surface layers .
both types of defects reduce considerably the mae with respect to the one obtained for pt - terminated thin films .
a detailed analysis of the electronic structure of the thin films is provided with a careful comparison to the copt bulk case .
the behavior of the maes is then related to the location of the different virtual bound states utilizing second order perturbation theory . |
a quantum computer is a quantum system whose time evolution can be thought of as a computation , much in the same way as we think of the time evolution of a pocket calculator to be a computation . for our pourposes it will suffice to model the quantum system as a `` black box '' and focus our attention on two discrete observables out of a complete set , which we shall call the input and output register .
following the standard notation @xcite , we shall indicate the computation of a function @xmath0 as @xmath1 the first ket describing the state of the input register and the second the state of the output register .
kets are labelled according to the elements of @xmath2 and @xmath3 they represent .
one of the most powerful features of quantum computation is _ quantum parallelism_. the superposition principle of quantum mechanics allows us to prepare the computer in a coherent superposition of a set @xmath4 of input states . after a single run ,
all of the corresponding outputs @xmath5 appear in the final state , according to the time evolution @xmath6 unfortunately , this is no `` pay one , take n '' .
in fact , the result is an entangled state of the input and output registers and there is no single measurement allowing us to extract from it all the computed values of @xmath5 @xcite .
however , it may well be possible to distil from this final state some _ global property _ of the function , thus exploiting quantum parallelism .
one of the most famous examples was presented by d. deutsch , who showed that a single quantum computation may suffice to state whether a two valued function of a two valued variable is constant or not @xcite .
d. deutsch and r. jozsa later generalized this result @xcite showing that the problem of classifying a given function @xmath7 as `` not constant '' or `` not balanced '' can be solved in polynomial time by means of a quantum computer ( the time required by a classical solution is exponential ) .
also d. r. simon showed that the problem of determining if a function is invariant under a so called xor mask , while it is classically intractable , admits an efficient quantum solution @xcite .
all of the algorithms cited above ( apart from the last , for which simon also considered a fully probabilistic generalization ) are characterized by a variable running time and zero error probability .
they consist of a non classical computation like ( [ thestandardcomputation ] ) followed by a measurement of the final state of the computer , as a result of which either the correct answer is obtained or the relevant information is destroyed and an explicitly inconclusive result is returned . in the latter case one has to go through the whole procedure again , so that only an average running time for the algorithm can be estimated .
global properties of functions that can be determined by such an algorithm are said to be computable by quantum parallelism ( qpc).this definition was put forward by jozsa @xcite who also demonstrated that , at least in the case of two valued functions , the qpc properties that can be determined by means of a single computation are an exponentially small fraction of all the possible global properties . in this paper we tackle the general problem of stating whether a function @xmath8 is constant or not . we show that for @xmath9 or @xmath10 this property is not qpc , meaning that any measurement following a computation like ( [ thestandardcomputation ] ) has a finite probability of yielding a wrong result .
we therefore investigate the power of quantum parallelism in a fully probabilistic setting . assuming that the ( classical ) computation of @xmath11 on @xmath12 randomly sampled points yields a constant value
, we calculate the posterior probability that @xmath11 is actually constant .
we then compute the analogous probability for a quantum algorithm requiring the same number of computations of @xmath11 .
comparison of the two results shows that our quantum strategy allows making a better guess at the solution , its indications being more likely to be correct .
we shall now briefly recall the classical example put forward by d. deutsch @xcite , before confronting the problem of its generalization .
suppose we are given a function @xmath13 and we are interested to know whether @xmath11 is constant or not .
of course there are only four such functions ( i.e. four instances of the problem ) , namely @xmath14 if all we can use is a classical computer , there is only one way to do the job : we must compute _ both _ @xmath15 and @xmath16 and compare them to check if they are equal . on the contrary , since in this simple case the property `` @xmath11 is constant '' is qpc , a quantum computer gives us a fair chance of finding the solution at the cost of the single computation @xmath17 after the computation , the calculator halts with its input and output registers in one of four possible states , corresponding to the four possible functions : @xmath18 since the above states are linearly dependent , they can not be distinguished with certainty .
this means that no measurement can establish which function was actually computed , or , which is the same , it s impossible to extract from the final state _ both _ the values of @xmath15 and @xmath16 .
however , we need only discriminate @xmath19 and @xmath20 , the final states yielded by the constant functions , from @xmath21 and @xmath22 . this can actually be done by measuring on the final state of the two registers an observable with the following non degenerate eigenstates : @xmath23 these four states can be thought of as `` flags '' indicating the result of the computation and have been named according to their meaning .
this becomes clearer as soon as we rewrite the final states ( [ finalstates ] ) on the basis of the above eigenvectors : @xmath24 it is now evident that 1 .
projection along the eigenvector @xmath25 can only take place if the state of the computer is either @xmath19 or @xmath20 , i.e. when @xmath11 is constant ; 2 .
likewise , projection along the eigenvector @xmath26 can only occur if @xmath11 is not a constant function ; 3 . regardless of the final state @xmath27 of the two registers after the computation ,
the measurement can yield a @xmath28 state with probability @xmath29 ; 4 .
state @xmath30 is orthogonal to the four final states listed above , and therefore it should show up only as a consequence of noise induced errors .
note that , according to the definition of the qpc class @xcite , the quantum algorithm can either give us the correct answer or no answer at all : as long as everything works properly , we ll never get a wrong result .
this comes in handy when we are asked to solve a decision theoretic problem in which simply waiting has a much higher utility than taking a wrong action .
in this case we can discard the fail results and base our decisions upon the meaningful answers , which we know to be correct .
the straightforward generalization of deutsch s example would go as follows .
given a function @xmath31 and assuming we can perform the non classical computation @xmath32 we are asked to devise an observable @xmath33 on the joint state of the two registers such that , after a single measurement of @xmath33 , we can either 1 .
obtain a reliable indication that function @xmath11 is constant ; 2 .
obtain an equally reliable indication that @xmath11 is not constant , or finally 3 .
get an explicitly inconclusive result .
let @xmath34 be the hilbert space of the joint states of the input and output registers .
if @xmath35 is the basis of @xmath34 formed by the ( non degenerate ) eigenstates of @xmath33 , all that is needed would be the existence of two disjoint subsets @xmath36 such that 1 .
all the final states obtained from the computation of constant ( non constant ) functions have a non zero projection along @xmath37 ( along @xmath38 ) ; 2 .
the final states corresponding to non constant ( constant ) functions are orthogonal to @xmath37 ( to @xmath38 ) .
these two requirements are evidently fulfilled in the case of deutsch s example , as can be easily seen by taking @xmath39 and @xmath40 ( for further details see @xcite ) .
however , as soon as the domain and range of the function @xmath11 grow larger , requirements _ i. _ and _ ii . _ become incompatible .
what happens is that whenever @xmath9 or @xmath10 the computation of constant functions yields final states that are linearly dependent upon those obtainable from non constant functions .
this clearly forbids the existence of @xmath37 , since the final states coming from non constant functions can not be orthogonal to @xmath37 .
in other words , the global property `` @xmath11 is constant '' is no longer qpc in the general case . note that , as demonstrated by jozsa @xcite , this result is independent of the particular superposition used as the input state for the non classical computation ( [ thecomputation ] ) .
the fact that the investigated property of @xmath11 is not qpc compels us to work in a fully probabilistic setting in order to cope with the possibility of wrong results . preserving the general structure of the algorithm
as outlined at the beginning of the preceding section , we note that we can still devise an observable @xmath41 such that any `` '' ket has a large projection along a subset @xmath42 of the eigenstates of @xmath41 ; for example , we can arrange for @xmath43 to be the very space spanned by the `` constant '' vectors .
the problem is now that since `` non constant '' kets generally have a non zero projection along @xmath42 , measuring @xmath41 no longer ensures a clear cut distinction between constant and non constant functions .
however , since `` non constant '' final states do have some component along the orthocomplement of @xmath42 , measuring @xmath41 still gives some ( probabilistic ) information about the computed function @xmath11 .
we are left with two asymmetrical possibilities ( actually , as we shall see , a more convienient choice for @xmath42 also makes an explicitly inconclusive result possible ) : 1 . measuring @xmath41 yields an eigenvalue associated to the orthocomplement of @xmath43 in @xmath34 . since this can only happen
if the computed function @xmath11 is _ not _ constant , this is an exact solution to the problem .
2 . measuring @xmath41 projects the final state of the two registers onto a state in @xmath42 .
if the computed function were constant , this would be the only possibility ; unfortunately , as seen above , other functions may also yield the same result .
we have therefore obtained only a probabilistic indication about @xmath11 being constant .
it is now clear that the generalized algorithm is essentially similar to a classical probabilistic algorithm , in that its results are not necessarily correct .
nevertheless , as we shall see in the following sections , the posterior probability of the function actually being constant after a result of type _ b. _ is obtained turns out to be much larger for our quantum algorithm than for the classical `` sampling '' strategy ( see section [ evaluation ] ) . in the rest of this section
we shall deal with the choice of the observable , which constitutes the core of the algorithm .
we would now like to introduce a correspondence between the hilbert space @xmath34 of the two registers of the computer and the space @xmath44 of complex matrices with @xmath45 rows and @xmath46 columns .
let @xmath47 be the computational basis of @xmath34 , the first ket referring to the state of the input register and the second to that of the output register : @xmath48 we define the isomorphism @xmath49 by identifying @xmath50 with the @xmath51 matrix whose elements are @xmath52_{i , j}= \delta_{m , i}\delta_{n , j}.\ ] ] the isomorphism @xmath53 , which maps the elements of @xmath47 onto the canonical basis of @xmath54 , is then extended by linearity to the whole @xmath34 .
since the final state of the computer after the computation of function @xmath11 is @xmath55 the entries of the corresponding matrix @xmath56 turn out to be @xmath57 , so that @xmath58 somehow resembles the graph of @xmath11 drawn with the `` @xmath59 '' axis along the rows and the `` @xmath60 '' axis pointing down . instead of @xmath61 for the upper left element of matrix @xmath58 . ]
it is easy to check that the scalar product @xmath62 in @xmath54 is preserved by @xmath53 , i.e. @xmath63 for any two vectors @xmath64 , @xmath65 of @xmath34 ( we write @xmath66 for the scalar product in @xmath34 . )
we shall now construct the observable @xmath41 as specified at the beginning of this section .
an observable in @xmath67 is identified by its @xmath68 eigenstates that form an orthogonal basis , or , using the isomorphism @xmath53 , by @xmath68 orthogonal matrices in @xmath54 .
we propose to take the @xmath68 two parameter matrices @xmath69 with @xmath70 and @xmath71 whose entries are defined by @xmath72 we shall call the above matrices _ fourier transform matrices _ ( ftm ) .
we recall that given a matrix @xmath73 , its two dimensional discrete fourier transform @xmath74 is defined as @xmath75 therefore the components of @xmath76 on the ftm basis are the entries of its discrete fourier transform @xmath77 . we still have to decide which eigenvectors are to be taken as an indication of the function @xmath11 being constant .
in other words we have fixed the basis @xmath78 but have yet to choose the subset @xmath79 .
we take @xmath80 as composed by the @xmath81 matrices @xmath82 , with @xmath83 .
it is easy to check that @xmath84 spans the subspace of @xmath54 generated by the set of the @xmath45 matrices @xmath85 corresponding to the constant functions .
we have not included @xmath86 in @xmath87 because the projection probability of the computer s final state on @xmath86 is the same for all functions : @xmath88 therefore @xmath86 has the same role that state @xmath89 had in deutsch s example . in the following
we shall put @xmath90 and we shall speak equivalently of the matrix @xmath86 in @xmath54 or of the state @xmath89 in @xmath67 . likewise , since @xmath91 for all @xmath92 and every matrix @xmath58 , subset @xmath93 plays the role of the @xmath30 state in deutsch s example ( section [ theexample ] ) . the remaining ftm matrices constitute set @xmath38 : @xmath94 note that we did not put a prime on @xmath38 , since it does satisfy both conditions _ i. _ and _ ii . _
listed in section [ impossibility ] .
this accounts for the lack of symmetry we pointed out at the beginning of section [ generalization ] .
suppose we run the quantum algorithm @xmath12 times on the same function @xmath11 and we always get an indication that @xmath11 is constant ( a projection onto @xmath42 ) .
we need to gauge the reliability of this result , which we can do by computing the posterior probability @xmath95 ( @xmath96 for short ) that the function really is constant .
this quantity can also be used to compare the efficiency of the quantum algorithm against a conventional classical solution , since what we are looking for is a procedure giving the lowest probability of error in change for the same computational effort . to evaluate ( [ prob ] ) we use bayes theorem , that is @xmath97 where by @xmath98 we mean the joint probability that @xmath11 is constant _ and _ that @xmath12 runs of the algorithm yield a `` constant '' outcome , corresponding to the final state being projected along @xmath87 . by the product rule ,
this can be expressed as @xmath99 assuming a uniform probability distribution on all the possible functions , we have @xmath100 .
regardless of the computed function @xmath11 , fail results have a @xmath101 probability of showing up ( see equation [ prfail ] ) .
this leads to @xmath102 . as a consequence ( [ qnum ] ) becomes @xmath103 the denominator of ( [ bayes ] )
can be expanded over all the @xmath104 possible functions of type ( [ f ] ) : @xmath105 since we assumed the input functions to be uniformly distributed , we have @xmath106 the @xmath12 runs of the quantum algorithm are stochastically independent and that implies that the likelihoods @xmath107 appearing in ( [ total ] ) are simply given by @xmath108 ^k,\ ] ] where with @xmath109 we mean the likelihood of a single run , the probability of a projection onto @xmath80 when the function is @xmath11 .
so we can concentrate only on @xmath109 , which , with the help of the sum rule , can be expressed as @xmath110 here @xmath111 stays for event `` after the measure the computer s final state projects itself onto the subspace of all constant functions '' , @xmath112 for the projection onto the matrix @xmath113 , which represents the @xmath114-th constant function and is defined as @xmath115 , and fail for the projection onto @xmath116 .
we have used the fact that , thanks to the orthogonality relations , events @xmath117 are mutually exclusive and so are @xmath80 and fail and that @xmath118 and @xmath119 . as we have seen in ( [ prfail ] ) , @xmath120 . on the other hand @xmath121 where @xmath122 is the matrix related through isomorphism @xmath123 to the computer s final state when the function is @xmath11 ( note that most of the elements of @xmath58 are zero , since @xmath124 ) .
let us now compute explicitly the trace that appears in the r.h.s . of ( [ ter ] ) : @xmath125 equation ( [ ter ] ) then becomes @xmath126 equation ( [ trace ] ) contains the sum of the elements appearing in the @xmath127-th row of matrix ( [ af ] ) . since matrix @xmath58 has a sole one in any column , this sum is equivalent to the number of ones in the @xmath114-th row of @xmath58 .
this gives us an idea for a smart classification of all the possible functions appearing in ( [ total ] ) : we associate with every function an @xmath128uple @xmath129 , where @xmath130 is the number of rows of its corresponding matrix @xmath58 with @xmath131 ones and @xmath132 zeroes .
doing so we can replace the sum over @xmath133 appearing in ( [ total ] ) by a sum over the @xmath128uples @xmath129 , with conditions @xmath134 condition ( [ one ] ) expresses the requirement that the total number of ones in matrix @xmath58 is @xmath46 ( or , since each column contains a sole one , that @xmath58 has @xmath46 columns ) , while ( [ two ] ) is equivalent to the condition that @xmath58 has @xmath45 rows . in the following
, we shall indicate with @xmath135 the set of the @xmath128uples @xmath129 that satisfy equations ( [ zero])([two ] ) .
note that , since every @xmath128uple corresponds to more than one function , when summing over the @xmath128uples we must use the right combinatorial factors .
these , for a fixed @xmath128uple @xmath129 , are given by : @xmath136 the first term corresponding to column permutations and the second to row permutations .
we can now use equation ( [ finalpr ] ) together with this way of classifying the functions to evaluate the total likelihood @xmath137 that appears as the first term in equation ( [ key ] ) .
if @xmath138 stands for a function corresponding to the @xmath128uple @xmath129,@xmath139 consequently ( [ totlik ] ) becomes@xmath140 and ( [ indip ] ) becomes in turn @xmath141 ^k\text{.}\ ] ] now we can sum over all the possible @xmath128uples with conditions ( [ zero])([two ] ) and with the combinatorial factors ( [ comb ] ) , obtaining the expression of equation ( [ total ] ) in the quantum case:@xmath142 ^kc_{j_0,j_1,\ldots , j_n}\ ] ] ( n.b .
we have used the fact that @xmath143 , since we suppose a prior uniform probability distribution on all the functions ) .
finally , using also equation ( [ num ] ) we can express the posterior probability ( [ bayes ] ) in the quantum case as @xmath144^k c_{j_0,j_1,\ldots , j_n}}\text{.}\ ] ] in the following section we will derive the corresponding expression for the classical case .
there exists at least one obvious classical probabilistic algorithm that can be used to spot constant functions .
we can simply compute the value of @xmath145 on @xmath12 randomly chosen points of its domain and decide that @xmath11 is constant if its restriction to the sampled points is .
this procedure , which we shall call the `` sampling algorithm '' , evidently constitutes the best possible classical strategy to solve the problem , since it uses up all the information we can gain on @xmath11 by @xmath12 classical computations . in order to allow a direct comparison with the quantum algorithm
, we have to find out what the posterior probabilities @xmath146 are in this case .
starting again from bayes theorem , we can express the numerator of equation ( [ bayes ] ) as @xmath147 that by ( [ unif ] ) is equal to @xmath148 ( note that in the classical case @xmath149 , since no fail results exist ) .
we must now evaluate the denominator of bayes formula , namely equation ( [ total ] ) .
choosing the @xmath12 inputs at random actually turns out to be inessential as long as the functions are uniformly distributed : sampling the first @xmath12 points @xmath150 is just as good .
let us therefore divide all the possible functions into two classes .
the first is made up by those for which at least the first @xmath12 values are constant ; they are @xmath151 .
all the other functions belong to the second class . as a consequence , the likelihoods that appear in the r.h.s . of ( [ total ] )
are simply given by@xmath152 putting this expression in ( [ total ] ) , and recalling ( [ unif ] ) , we can rewrite ( [ bayes ] ) as @xmath153 this result is to be compared with equation ( [ quantum ] ) , which gives the corresponding posterior probability after @xmath12 runs of the quantum algorithm . in order to do so , formula ( [ quantum ] )
must evidently be evaluated by means of a ( classical ! ) computer . before listing the numerical results , however , we are going to discuss two special cases that can be solved analytically in the limit of large @xmath46 .
we shall now analyse the behaviour of our generalized quantum algorithm in the worst possible case , that is when the computed function has maximum probability of being mistaken for a constant function , even if it is not .
this occurs quite naturally for a matrix of the following kind : @xmath154 { \overbrace { \begin{array}[c]{cccccc } 1 & 1 & 1 & \ldots & 1 & 0 \\ 0 & 0 & 0 & \ldots & 0 & 1 \\ 0 & 0 & 0 & \ldots & 0 & 0 \\ \vdots & \vdots & \vdots & \ddots & \vdots & \vdots \\ 0 & 0 & 0 & \ldots & 0 & 0 \end{array } } ^{n } } \hspace{1 mm } \right ) \right\}\scriptstyle{m}\ ] ] representing a function @xmath155 that is constant on its whole domain but for one point .
the resulting probability of error is given by the squared modulus of the projection of @xmath156 on the space spanned by the set @xmath157 of the matrices associated to constant functions is so large that the probability of a fail result is negligible .
] , that is @xmath158 therefore @xmath159 tends to one in the limit of large @xmath46 . here again , in order to compensate for this we have to run the quantum algorithm several times , say @xmath12 ( classically , we would have to sample more and more points ) .
if we want to keep the probability of being `` cheated '' by an almost constant function @xmath155 as low as a given value @xmath160 , we evidently have to choose @xmath12 so that @xmath161 , that is @xmath162 as we would expect , @xmath12 does tend to infinity in the limit of large @xmath46 , meaning that exploring an even larger domain requires an infinite number of computations .
it is nevertheless interesting to study the ratio @xmath163 of the number of runs to the number of elements in the domain . in the limit of large @xmath46
, this becomes @xmath164 which is a constant independent on @xmath46 .
therefore , if we are required to perform the computation with a worst case error probability @xmath160 , we have to run our quantum computer a number of times which , in the limit of large @xmath46 , is a definite fraction @xmath165 of @xmath46 .
equation ( [ worstcaseerrorprobability ] ) can in this case be inverted to obtain @xmath160 as a function of @xmath165 , yielding @xmath166 .
coming now to the classical case , sampling a fraction @xmath165 of the points in the domain ( which requires @xmath167 computations ) entails having a probability @xmath168 of mistaking @xmath155 for @xmath169 .
in figure [ worstcasefigure ] we plotted the worst case probability of error against @xmath165 for both the quantum and the classical algorithm in the limit of large @xmath46 . in the quantum case
@xmath160 decreases more rapidly and stays well below the classical probability of error as long as @xmath165 is not too close to @xmath170 ( remember that the `` sampling '' algorithm is no longer probabilistic if we compute our function over its entire domain ! ) .
looking now at the best case , we find that there is again a single class of functions which is easily dealt with by both algorithms , that is one to one functions or permutations of the points in the domain ( this obviously requires @xmath46 to be equal to @xmath45 ) . using the classical sampling algorithm
, one can evidently be sure to distinguish an invertible function from a constant one with only two computations , since the former does not assume any value in its range more than once . in the quantum case ,
permutations are associated with matrices having exactly one `` 1 '' in each row and in each column .
such matrices turn out to be orthogonal to @xmath87 .
therefore , a measurement of the final state yielded by a permutation can either result in a fail or in projection along @xmath38 , which indicates that the function is not constant .
now fails can only be obtained with probability @xmath171 , which luckily vanishes as @xmath46 grows larger .
we conclude that , in the limit of large @xmath46 , the quantum algorithm is practically guaranteed to spot a one to one function at first sight , after a single computation , thus doubling the efficiency of the classical algorithm . by the way
, we note that if we only had to tell constant functions from permutations if our practical problem did nt require us to deal with non
invertible , non constant functions we would be back to the original situation of deutsch s example .
we can now see what was so special about the four functions considered by deutsch in his original example ( see equation [ thefourfunctions ] ) .
when both the domain and the range consist of two points only all non constant functions turn out to be one to one , so that all ambiguity is removed .
we are including , in figures [ fig8x2lin ] through [ fig24x24log ] , some comparative graphics of the posterior probabilities @xmath146 expressed by equations ( [ bayes2 ] ) and ( [ quantum ] ) versus the number @xmath12 of successful computations effected ( by successful computation we mean all computations barring fail results ) .
as our previous analysis suggested , the quantum algorithm turns out to be far more efficient than the classical `` sampling '' algorithm for small values of @xmath172 .
we emphasize that this result is entirely dependent upon the use of quantum parallelism .
this highly non classical feature of quantum computation apparently allows a quicker exploration of the domain of function @xmath11 , even in the case that the investigated property is _ not _ qpc .
the posterior probability @xmath173 we used for our numerical calculations is conditoned to a sequence of @xmath12 `` constant results '' of the quantum algorithm .
we have overlooked the possibility of obtaining one or more fail outcomes .
this is particularly significant when @xmath174 ( figures [ fig8x2lin ] and [ fig16x2log ] ) , because in such cases a fail result has a @xmath29 probability to show up .
this means that in order to obtain @xmath12 projections of the final state of the computer along @xmath42 one must expect to run the quantum computer @xmath175 times .
nervertheless , as the graphics show , the quantum strategy always turns out to be convenient , at least for small values of @xmath172 .
we finally note that as @xmath46 and @xmath45 grow larger ( see for instance figure [ fig24x24log ] ) the resulting posterior probabilities turn out to be so low that both the quantum and the classical algorithm are virtually useless .
this entirely depends on our assumption of an uniform distribution over functions @xmath11 , which is probably eccessively penalizing . in real
world situations , we can expect the quantum algorithm to be useful in any situation in which the `` sampling '' algorithm is successfully employed at the present day .
we are grateful to c. m. becchi for posing the question which led to this work .
we acknowledge the interest of g. castagnoli and the collaboration with elsag bailey ; we also thank a. ekert and c. macchiavello for interesting discussion .
special thanks to e. beltrametti for continuous help and advice .
2 a. ekert , r. jozsa : quantum computation and shor s factoring algorithm , _ reviews of modern physics _ * 68 * , 733754 ( 1996 ) r. jozsa : characterizing classes of functions computable by quantum parallelism , _ proc .
lond . _ a ( 1991 ) * 435 * , 563574 d. deutsch : quantum theory , the church turing principle and the universal quantum computer , _ proc .
lond . _ a * 400 * , 97117 ( 1985 ) d. deutsch , r. jozsa : rapid solution of problems by quantum computation , _ proc .
a ( 1992 ) * 439 * , 553558 d. r. simon : on the power of quantum computation , _ proceedings of the 35th annual ieee symposium on the foundations of computer science _ , 1994 , 116123 | quantum parallelism is the main feature of quantum computation . in 1985
d. deutsch showed that a single quantum computation may be sufficient to state whether a two valued function of a two valued variable is constant or not .
though the generalized problem with unconstrained domain and range size admits no deterministic quantum solution , a fully probabilistic quantum algorithm is presented in which quantum parallelism is harnessed to achieve a quicker exploration of the domain with respect to the classical `` sampling '' strategy . |
the formation of molecules through the _ harpoon mechanism _ @xcite occurs from the interaction of two fragments , one with a low ionization potential ( ip ) and another that has a large electron affinity ( ea ) .
the reactants approach each other and , at a certain distance , an electron from the fragment with low ip _
harpoons _ the fragment with large ea , giving rise to a rapid electron - transfer process that is triggered by the coulomb attraction exerted by the two fragments .
the interplay between the energy exchange due to the harpoon process and the coulomb attraction determines the distance at which the electron transfer takes place .
the harpoon mechanism was proposed by michael polanyi to explain the unusually large cross sections observed in the formation of alkali halides .
since the ip of the alkali metals is similar to the ea of halogens , the coulomb attraction needed to favor the electron - transfer process is usually small for these species and , therefore , the electron transfer occurs at large separation of the reactants . on the other hand ,
if the molecule under consideration is an alkali hydride the process takes place at short distances because the ea of hydrogen is much smaller than the ip of alkalis .
since the dissociation of ionic species in gas phase through the inverse harpoon mechanism usually involves the transition from ionic to neutral fragments in an electron - transfer process , the molecule is expected to change its electron sharing index ( esi ) commonly known as _ bond order_ from a low ( ionic ) value at equilibrium to a larger value ( when the electron transferred lies between the two fragments ) that should decay as the fragments separate from each other .
ponec and coworkers @xcite demonstrated that lih and beh show a maximum of the esi in the vicinity of the avoided crossing between two adiabatic states .
interestingly , the esi maximum was obtained using the quantum theory of atoms in molecules ( qtaim ) definition of an atom , but it could not be reproduced using a mulliken atomic partition . some of us have demonstrated @xcite that many atomic partitions fail to attain this condition and suggested that the ability to reproduce this feature could be used as a criterion to assess the goodness of atomic partitions @xcite .
thus far , the electron - transfer process has been also monitored using several quantities , including the dipole @xcite and higher - order moments @xcite , the topological properties of the electron density @xcite , the electron localization function ( elf ) @xcite and other descriptors of the electronic structure @xcite .
however , none of these methods has been able to actually follow the electron transferred as it jumps from one atom to the other .
this sudden jump is what differentiates the harpoon mechanism from a regular electron - reorganization reaction and , therefore , its detection would provide a means to distinguish these mechanisms . indeed , the three - dimensional pictures of the elf do not reveal a separate basin for the electron until the electron has been already completely transferred to the other fragment @xcite .
the goal of this work is analyzing the harpoon mechanism using a plethora of bonding descriptors to find which of these tools can be used to characterize the reaction . to this aim
, we will analyze the dissociation of three ionic compounds that present the electron transfer at different interatomic separations and other small molecules , whose formation does not take place through a harpooning reaction .
the analysis of these molecular systems shall shed some light on the abilities of some bonding tools , providing some interesting test cases that can be used in the development of new descriptors of the electronic structure of molecules .
in this section we briefly review several bonding descriptors that will be used to study the harpoon mechanism .
let us assume a given partition of the molecular space into atomic regions that will be labeled @xmath1 .
the integration of the electron density within an atomic region gives rise to the electron population : @xmath2 where @xmath3 is the electron density .
analogously , the pair density is needed to define pair populations , @xmath4 which enter the expression of the electron sharing indices ( esi ) or electron delocalization indices @xcite @xmath5=-2\text{cov}\left(n(a),n(b)\right)\ ] ] that are related to the covariance of the atomic populations .
the esi are a non - integer version of the classical concept of _ bond order _ @xcite that provides a real value in accord with the number of electron pairs shared by two fragments .
the elf requires the calculation of the laplacian of the same - spin pair density functions @xcite _ i.e. _ , @xmath6 where @xmath7 with @xmath8 .
the elf measures the extent of electron localization , giving values close to zero for regions with highly delocalized electrons and large values in nuclear regions , lone pairs and bonding electron pairs @xcite .
similarly , negative values of the laplacian of the electron density also determine molecular regions with localized electrons , sometimes providing an electronic picture similar to that of the elf @xcite .
the three - dimensional representations of the elf and the laplacian have been repeatedly used in the literature to characterize all sorts of chemical bonds @xcite .
information theory has been also used to characterize the bonding patterns and the electronic structure of molecular systems @xcite . among them ,
the most well known are shannon entropies and fisher information descriptors .
the shannon entropies are descriptors of the spread of the electronic density ( in position or momentum space ) in accord with the following expressions @xcite @xmath9 @xmath10 where @xmath11 is the position space electronic density normalized to unity and the @xmath12 is the momentum space electronic density also normalized to unity .
the sum of both entropies , @xmath13 , known as total shannon entropy has a physical bound imposed by the uncertainity principle @xcite .
the shannon entropy provides a _ global _ measure of the electron density spread , whereas the fisher information ( fi ) is defined as @xcite @xmath14 for the position space and @xmath15 for the momentum space .
although fi quantities provide a global number per molecule , unlikely shannon entropies , their value is largely affected by small local changes of the function as has been proved by lopez and coworkers @xcite . in this sense ,
the fi provides a _ local _ measure of the density sharpness .
the position space fi increases when the heterogeneity of the system leads to larger changes of the position space density , whereas the fi in the momentum space depends on large deviations of the momentum density , @xmath16 and can be obtained from the fourier
transform that connects the momentum space wave function with the position space one @xmath17 d{\bf r_1}d{\bf r_2} ...
d{\bf r_n } \nonumber\end{aligned}\ ] ] these information theory indicators have been used in the past to characterize the electronic structure of molecules and chemical reactions @xcite .
some authors @xcite have linked these concepts of information theory to particular chemical reactivity , attributing _ enhanced sensitivity to changes on the position and momentum densities along the chemical reaction paths _ @xcite . in this work
we will examine the description of the harpoon mechanism afforded by these indicators .
we study the bond dissociation mechanism of ten diatomic molecules ( bh , beh , co , f@xmath18 , lif , lih , h@xmath18 , he@xmath18 , n@xmath18 and o@xmath18 ) in their ground state and the first excited state of the same symmetry and spin multiplicity than the ground state for lih and lif . born - oppeheimer , full configuration interaction ( fci ) calculations of h@xmath18 , he@xmath18 , lih , bh and beh have been performed with a modified version of knowles and handy program @xcite , whereas gaussian 09 @xcite was used to obtain complete active space self - consistent field ( casscf ) energies .
namely , cas(6,6 ) , cas(10,6 ) , cas(8,6 ) , cas(6,6 ) and cas(6,6 ) single - point evaluations were performed for n@xmath18 , f@xmath18 , o@xmath18 , lif and co , respectively . the aug - cc - pvdz basis was employed in all cases .
the lif valence casscf calculation presents a wave function and energy discontinuity around the avoided crossing point @xcite .
the latter problems are avoided if fci is used or if , more conveniently , a state - average casscf calculation is performed @xcite .
the state - average cas computation produces a qualitatively correct result with an avoided crossing at about 4@xmath19 in contrast to the fci results that place the avoided crossing at 6 - 6.5@xmath19 @xcite . in this paper
we have performed a four - state average cas(6,6 ) with equal weights , which gives a qualitatively correct , discontinuity - free , description of the process that is sufficient for the purposes of this work .
a six - state average cas(6,6 ) calculation was also performed for co. the quantum - theory of atoms - in - molecules ( qtaim ) @xcite calculations have been performed with aimall package @xcite , and the topological fuzzy voronoi cells ( tfvc ) computations have been done using apost-3d software @xcite .
the tfvc is a three - dimensional atomic partition based on fuzzy atoms that produces results very similar to those obtained with qtaim at a much lower computational cost @xcite .
the pair densities ( 2-pd ) have been obtained with dmn @xcite program and inputed into esi-3d @xcite to produce the esi and atomic populations .
the elf analysis has been performed with a modified version of the topmod program @xcite using the hartree - fock like approximation of the 2-pd in the computation of the elf values .
this approximation is needed in order to preserve the antisymmetry property of the 2-pd that other more sophisticated approximations do not fulfill and provides a very accurate description of the elf even for stretched bonds @xcite .
the calculation of information theory quantities has been done with the rho - ops code developed in our group @xcite .
lih is the smallest ionic molecule that is formed through the harpoon mechanism .
the electron affinity of hydrogen is quite different from the ionization potential of lithium and , therefore , the electron transfer in lih occurs at shorter distances than in alkali halides . in this section
we will analyze two @xmath20 adiabatic states of lih , namely , x@xmath20 and a@xmath20 .
the former state dissociates into h(@xmath21)+li(@xmath21 ) and the latter into h(@xmath21)+li(@xmath22 ) .
these two states involve the crossing between three diabatic states : the ionic ground state and the two lowest - lying covalent states .
the x@xmath20 state results from the crossing between the first ionic and covalent diabatic states , showing an avoided crossing with the a@xmath20 curve at about 3@xmath19 .
the a@xmath20 state is characterized by two avoided crossings : the first one has just been described and the second avoided crossing at about 6@xmath19 results from the interaction between the ionic and the second covalent states . therefore , the x@xmath20 ground state is characterized by an ionic bonding at the equilibrium but its covalent character increases as the interatomic distance stretches .
the a@xmath20 state is even more interesting because as the bond stretches one expects two changes in its character : from rather covalent to ionic and back to covalent .
the presence of two crossings furnishes the a@xmath20 curve with a characteristic flat region that leads to huge anharmonic effects and anomalous spectroscopic constants @xcite .
lih lowest - lying states have been widely studied in the past @xcite due to the confluence of these peculiar features , which make them a formidable playground to test modern electronic structure methods .
the ground state electron harpooning process has been studied by means of chemical bonding descriptors @xcite but , to our knowledge , the excited state has never been explored from the chemical bonding viewpoint . ) and the two lowest - lying @xmath20 excited states ( a@xmath20 and b@xmath20 ) .
energy in a.u . ]
let us first describe the electron exchange between both atoms that occur upon stretching the lih bond in the two electronic states . to this aim
, the most straightforward analysis consists in checking the atomic populations along the energy curves .
[ fig : pops ] contains the change in the qtaim population of the most electropositive atom as the molecule dissociates into two atomic fragments ( only molecules that present electron reorganization have been included ) .
the total density transferred from hydrogen to lithium is about 0.9 electrons in the ground state , where the molecule passes through an avoided crossing that changes its bonding character from ionic to covalent .
conversely , the overall transfer in the a@xmath20 state is barely 0.2 electrons . as the bond stretches the electron passes from lithium to hydrogen until about 0.5 electrons are completely transferred .
this process takes places as the curve goes through the first avoided crossing and the bond becomes more ionic . at about 2@xmath19 stretching , the electron transfer is reversed and the lithium withdraws electron charge when passing by the second avoided crossing .
about 0.7 electrons are transferred in this second step and the bond evolves from ionic to rather covalent .
the covalent character of the molecule in the a@xmath20 state at geometries near the equilibrium is not as strong as one would expect in ground - state covalent molecules , however , the obtained esi values are larger than in pure ionic molecules and unprecedented of excited states stretched bonds ( _ vide infra _ ) . the plot of the dipole moment along the internuclear axis at different interatomic separations ( fig .
[ fig : dipolarlih ] ) can be also used to monitor the electron - transfer process . for the ground - state at equilibrium
@xmath23 a.u . and
this value increases as the bond stretches , peaking at 2.8@xmath19 and then decreasing to zero as the atoms tend to isolated neutral entities .
conversely , the a@xmath20 state shows a lower dipole moment at the equilibrium with opposite sign ( @xmath24 a.u . ) , an indication of the more covalent nature of this bond as compared to the x@xmath20 state .
the dissociation of the a@xmath20 state of lih involves the sign change of the dipole moment as it passes through the first avoided crossing ; reaches a maximum value around @xmath25 and then decreases to zero as it goes through the second avoided crossing . and the a@xmath20 states of lih as function of the interatomic distance . ] in fig .
[ fig : esis ] we find the esis along the dissociation curves for a series of diatomic molecules .
the ground - state lih esi has a low value ( 0.19 ) at the equilibrium that slightly decreases as the bond is stretched but soon enough increases as the covalent contribution plays a more important role in the description of the x@xmath20 state , reaching a maximum around @xmath26 and dying off at large interatomic separations .
the esi maximum is the point with the largest electron sharing between the atoms , _
i.e. _ , the point at which the electron is least attached to either fragment ( although , not loose enough to form a separate entity ) .
indeed , the esi peaks at the maximal transfer variation position , _
i.e. _ , the position at which the variation of electron population with the interatomic distance is largest ( see vertical lines in the inset plot of fig . [
fig : esis ] ) .
furthermore , at this point 0.5 electrons have been transferred from li to h when we compare their atomic populations with the ones of the isolated neutral atoms as already reported by pends and coworkers @xcite .
the esi can be thus used to locate the position at which the electron remains in _ no one s land _ , which one identifies with the avoided crossing region .
the analysis of the a@xmath20 state reveals a similar esi profile for interatomic distances from 1@xmath19 to @xmath27 than for x@xmath20 state . despite the esi values in this range
are similar , the nature of the bond is completely different
. let us analyze @xmath28 , which is the equilibrium geometry for the x@xmath20 state .
at this distance the ground state is mostly ionic ( low esi , atomic charges close to + 1 and -1 and the dipole moment vector pointing towards hydrogen ) , whereas the a@xmath20 state is _ covalent antibonding _ because the charges are virtually zero for each atom and the dipole moment vector is of the same magnitude but opposite sign .
the lih esi value at equilibrium for the a@xmath20 state ( @xmath29 ) is larger because the molecule becomes more covalent as it approaches the equilibrium distance in line with the more covalent character one expects from the excited state ( @xmath30(li , h)=0.19 vs. @xmath31(li , h)=0.30 ) . for bond lengths beyond 3@xmath19
the esi curves of the two states deviate from each other and the a@xmath20 state presents rather constant values ( ca .
0.5 ) up to @xmath25 ; from this point the esi decreases until it reaches zero beyond @xmath32 .
interestingly , the esi of the a@xmath20 state attains a second maximum around the position of the second avoided crossing . in the formation of lih in the a@xmath20 state
the two esi maxima correspond to points at which there is a maximal transfer variation ; around the first maximum , lithium passes electron density to hydrogen and the opposite occurs in the vicinity of the second one .
the latter is actually the point in the molecule formation at which 0.5 electrons have been transferred from li to h. at this point , one should ask whether other tools to characterize the electronic structure of molecules are capable of following the electron as it jumps from one atom to another .
the elf has been successfully used in the characterization of localized electrons as those occuring in covalent bonds , lone pairs @xcite or even isolated electrons in ionic compounds @xcite and it has been also used to analyze some electron - transfer processes @xcite .
however , the investigation of the elf for lih dissociation @xcite does not reveal a separate basin that can be identified with the electron transferred .
the process is characterized by a progressive elongation of hydrogen s valence basin that eventually splits and is absorbed by lithium s valence basin but , at no point , it becomes a separate , distinctive basin .
qtaim analysis along the two electronic states puts forward that there is no non - nuclear attractor of the density and no laplacian basin in the bonding region corresponding to the transferred electron .
indeed , depending on the interatomic separation , the electron is located in the lithium or the hydrogen atom but it is never found in a separate entity . therefore , a three - dimensional representation of these functions can not be used to pinpoint the electron position as it jumps between atoms . the electron jump process is difficult to identify by the isosurface representation but it is easily monitored by elf profiles along the internuclear axis .
[ fig : lihelf ] includes several profiles of the elf along the molecular axis of lih s x@xmath20 state at different interatomic distances .
a vertical dashed line marks the minimal density position between li and h in the molecular axis .
the latter is known in the qtaim framework as the bond critical point ( bcp ) , and in these plots the bcp indicates the atomic boundary along the internuclear axis .
these plots show that at about the avoided crossing region the valence basin of lithium emerges at one end of the molecule , whereas the hydrogen valence basin starts to split into two but it still belongs to the h atomic qtaim basin . at 3.3@xmath19
the hydrogen peak splits into two and the second peak carrying the transferred electron crosses the boundary between qtaim atomic basins afterwards . at 5@xmath19
the process is almost complete and the lithium valence basin is fully characterized ( see the two peaks around the lithium atom in the one - dimensional plot of fig .
[ fig : lihelf ] ) . at no point during the dissociation the ( one - dimensional ) maximum between li and h in fig .
[ fig : lihelf ] becomes a three - dimensional elf maximum that would give rise to a separate basin . in this sense ,
a one - dimensional elf representation is more useful than the usual isosurface picture in order to trace the _ electron motion _ in the reaction mechanism .
inspection of the a@xmath20 state elf profile along the molecular axis reveals the peak formation after the first avoided crossing and the peak splits and moves towards lithium until it has completely passed the atomic qtaim boundary after the second avoided crossing ( see fig .
[ fig : lihelfes ] ) .
state of lih along the internuclear axis for several values of the interatomic distance .
the zero is located at the center of mass and the vertical dashed line indicates the bond critical point .
the first avoided crossing occurs at about 3 . ]
state of lih along the internuclear axis for several values of the interatomic distance .
the zero is located at the center of mass and the vertical dashed line indicates the bond critical point .
the first avoided crossing occurs at about 3@xmath19 and the second at 6 . ] similarly , the molecular - axis scan of the laplacian of the electron density at different interatomic separations reveals the harpooning mechanism . in fig .
[ fig : lihlapl ] we represent the negative values of the laplacian of the electron density along the molecular axis at different interatomic distances ; we have used logarithmic scale to allow an easy identification of the peaks and dashed vertical lines to indicate the bcps . for the x@xmath20 state at 2@xmath19 ,
the peak corresponding to the hydrogen atom starts to expand , splitting into two separate peaks after 2.5 . by neglecting values of the laplacian below 10@xmath33 . ] then , the two peaks separate from each other and , eventually , the new peak separates completely from hydrogen peak and becomes the valence basin of lithium .
interestingly , the new peak containing the transferred electron starts crossing the qtaim atomic boundary at the first avoided crossing . for the a@xmath20 state ,
the peak between li and h shows at about @xmath34 .
as the distance elongates , the peak broadens and it begins to cross the qtaim atomic boundary after 4.0@xmath19 and it is completely transferred when the molecule passes the second avoided crossing ( see fig .
[ fig : lihlapl ] bottom picture ) .
therefore , according to the elf - qtaim description the electron transfer of the a@xmath20 state also takes place between the two avoided crossings .
, bottom a@xmath20 ) .
all molecules locate li atom in the z - axis at zero for comparison .
the vertical dashed lines in the inset plot indicate the bond critical points .
laplacian units are a.u.,title="fig : " ] , bottom a@xmath20 ) .
all molecules locate li atom in the z - axis at zero for comparison .
the vertical dashed lines in the inset plot indicate the bond critical points .
laplacian units are a.u.,title="fig : " ] in this section we analyze the formation process of other diatomic molecules in their ground state from the isolated neutral atoms .
first , we have chosen two additional molecules , beh and lif , which are also formed through an electron harpooning mechanism due to ionic and covalent diabatic state crossings .
the former consists of be , which has an ionization potential larger than li and , therefore , contributes to a larger energy difference with hydrogen s electron affinity than lih . as a result
, the electron jump takes places at quite short distances .
conversely , f has a larger electron affinity than h , giving a small ip - ea value for the atoms in lif .
the electron jump in lif thus occurs at large interatomic separations but the state - average casscf description of the process underestimates the distance at which the avoided crossing occurs ( _ vide supra _ ) .
notwithstanding , it does not affect the qualitative bonding picture of lif or the bond mutation process around the avoided crossing .
second , we analyze h@xmath18 , o@xmath18 , n@xmath18 , f@xmath18 , he@xmath18 , bh and co , which are mostly included for comparison .
these molecules present different bonding patterns from weakly bonded to covalent single- , double- , and triple - bonds , as well as a polar covalent bond .
an example of the latter is the co molecule and its formation also implies an important electron reorganization process in spite of the fact that the mechanism does not involve electron harpooning .
let us first analyze beh and lif .
fig [ fig : peslif ] shows the two lowest - lying @xmath20 states of lif : the x@xmath20 and a@xmath20 states .
the dipole moment of both states of lif coincides at a distance close to the avoided crossing ( ca .
@xmath35 ) ( see fig s1 in the supp .
the esi profile in fig [ fig : esis ] also shows a maximum around the avoided - crossing interatomic distance for both states .
the esi profile of lif s a@xmath20 state shows larger ( slightly more covalent ) values than the x@xmath20 state before the avoided crossing but the rest of the esi profile is very similar .
the avoided crossing point is not only the geometry at which the esi is maximum , but also the point at around which there is a maximal transfer variation and a net transfer of 0.5 electrons from li to f as compared to its isolated neutral counterparts .
the elf and the laplacian profiles along the internuclear axis show a separate peak ( corresponding to the transferred electron from f to li ) that emerges before the avoided crossing and it crosses the qtaim atomic boundary around the avoided crossing position ( see figs . s2 and s3 ) . ) and the lowest - lying a@xmath20 excited state .
energy in a.u . ] unlike lih and lif , beh s x@xmath36 ground state avoided crossing occurs at very short distances ( ca . 2.2@xmath19 @xcite ) .
the esi profile reveals a maximum at this position that is also the same position at which the transfer variation is maximal and 0.5 electrons have been transferred from h to be .
the elf and laplacian show features similar to those of lih and lif , with the peak corresponding to the electron carried from one atom to another crossing the qtaim atomic boundary around the avoided crossing ( see figs s4 and s5 ) .
bh in its x@xmath37 state presents a rather large esi value for an ionic species ( @xmath38 ) and , actually , the valence bond description assigns a covalent character to this bond @xcite .
the ground state energy curve does not present avoided crossings that rapidly could change the bond character from ionic to covalent or viceversa .
therefore , all evidences suggest a electron reorganization without harpoon mechanism .
notwithstanding , our results show otherwise .
the esi profile is characteristic of a harpooning mechanism taking place at very short distances ( ca .
a fact that is confirmed by elf and laplacian profiles along the molecular axis ( see figs s6 and s7 ) , which show that elf and laplacian hydrogen s peak splits into two and passes the atomic qtaim boundary from h to b at the vicinity of the esi maximum , where the molecule experiences its maximal transfer variation and 0.5 electrons have already passed from h to b. therefore , one is deemed to conclude that the formation of bh also takes places through a harpoon mechanism without the existence of an avoided crossing . according to qtaim population analysis ( see fig . [
fig : pops ] ) , carbon withdraws more than one electron from oxygen in the co molecule . unlike the lih or lif
, the electron - density transfer does not imply a large change in the dipole moment . although the shape of the dipole moment profile along the dissociation is similar in all electron - transfer reactions , the reaction in co ( see fig .
s8 ) takes places through very small values of the dipole moment ( max . 0.3 ,
-0.1 ) . the esi profile exhibits a smoothly decreasing profile that dies off after 4@xmath19 .
there is , therefore , no esi maximum at stretched distances but the maximal transfer variation coincides again with the point where 0.5 electrons have been transferred from c to o. the laplacian and the elf scans along the interatomic axis describe the dissociation process as the splitting of the bonding peak ( corresponding to the triple bond ) into two peaks ( see fig .
s9 and s10 in supp . info . ) .
the rest of the molecules are calculated to afford a comparison between the above - described electron reorganization processes and other reactions .
the esi profiles of f@xmath18 , o@xmath18 and n@xmath18 exhibit a sigmoidal shape that pends and coworkers identify with the breaking of covalent bonds , whereas the he@xmath18 esi dissociation curve decays exponentially , in line with noncovalent interactions @xcite .
the elf and laplacian scan along the molecular axis do not have much interest because they do not reveal additional features as compared to the isosurface representations discussed elsewhere @xcite .
finally , let us examine the profiles provided by five indicators defined in the framework of information theory .
[ fig : sr ] collects fi and shannon indicators ( see .
[ eq : sr]-[eq : fp ] ) at the dissociation limit minus its value at different interatomic separations ( @xmath39 and @xmath40 ) . since the molecules analyzed undergo different bond formation mechanisms one would expect some distinguishing features in the representation of these indicators along the dissociation curves .
@xmath41 shows that molecules that undergo an electron - transfer process show larger differences with respect to its value at infinite separation .
these molecules experience an internal electron reorganization at stretched distances that is expected to affect its @xmath42 . in this
regard , lif and lih experience a long - range electron exchange that explains why these molecules converge to @xmath43 at larger distances than the other molecules . on the other hand ,
it is not clear why the h@xmath18 profile lies very close to beh and bh curves or why covalent molecules show @xmath41 profiles very similar to he@xmath18 .
therefore , it does not seem to be an obvious connection between @xmath41 profiles and the bond formation mechanim in diatomics .
similar conclusions can be drawn from the inspection of @xmath40 profiles .
@xmath44 shows maximal values more or less close to the avoided crossings for lih , lif and beh , whereas @xmath44 decays exponentially for the other molecules except he@xmath18 , which shows a somewhat sigmoidal shape . in this sense
, the profile of @xmath44 could provide a distinctive shape for each type of mechanism , excluding h@xmath18 whose shape does not match the other covalent molecules . on the other hand , there does not seem to be a way to distinguish between single- , double- or triple - bond dissociations . as far as the magnitudes based on the momentum density ( @xmath45 and @xmath46 ) go , we can see that the molecules involving a electron - transfer harpoon mechanism present larger differences with respect to the value at infinite separation .
one more time , hydrogen molecule is an exception to this rule . ) ; ( b ) shannon entropy in the momentum space ( see eq .
[ eq : sp ] ) ; ( c ) fisher information in the position space ( see eq .
[ eq : fr ] ) ; ( d ) fisher information in the momentum space ( see eq .
[ eq : fp ] ) ; ( e ) total shannon entropy ( @xmath13 ) ; plots represent the values at the dissociation limit minus the value at different interatomic separations for the series of diatomic molecules taken into study .
information theory descriptors units are a.u . , title="fig : " ] ) ; ( b ) shannon entropy in the momentum space ( see eq .
[ eq : sp ] ) ; ( c ) fisher information in the position space ( see eq .
[ eq : fr ] ) ; ( d ) fisher information in the momentum space ( see eq .
[ eq : fp ] ) ; ( e ) total shannon entropy ( @xmath13 ) ; plots represent the values at the dissociation limit minus the value at different interatomic separations for the series of diatomic molecules taken into study .
information theory descriptors units are a.u . , title="fig : " ] ) ; ( b ) shannon entropy in the momentum space ( see eq .
[ eq : sp ] ) ; ( c ) fisher information in the position space ( see eq . [ eq : fr ] ) ; ( d ) fisher information in the momentum space ( see eq . [ eq : fp ] ) ; ( e ) total shannon entropy ( @xmath13 ) ; plots represent the values at the dissociation limit minus the value at different interatomic separations for the series of diatomic molecules taken into study .
information theory descriptors units are a.u . , title="fig : " ] ) ; ( b ) shannon entropy in the momentum space ( see eq .
[ eq : sp ] ) ; ( c ) fisher information in the position space ( see eq . [ eq : fr ] ) ;
( d ) fisher information in the momentum space ( see eq .
[ eq : fp ] ) ; ( e ) total shannon entropy ( @xmath13 ) ; plots represent the values at the dissociation limit minus the value at different interatomic separations for the series of diatomic molecules taken into study .
information theory descriptors units are a.u . , title="fig : " ] ) ; ( b ) shannon entropy in the momentum space ( see eq .
[ eq : sp ] ) ; ( c ) fisher information in the position space ( see eq .
[ eq : fr ] ) ; ( d ) fisher information in the momentum space ( see eq .
[ eq : fp ] ) ; ( e ) total shannon entropy ( @xmath13 ) ; plots represent the values at the dissociation limit minus the value at different interatomic separations for the series of diatomic molecules taken into study .
information theory descriptors units are a.u . ]
the lowest - lying states of lih have been widely used to develop and calibrate many different methods in quantum mechanics . in this paper
we have shown that the electron - transfer processes occurring in these two states are a difficult test for chemical bonding descriptors and can be used with a two - fold purpose : ( i ) to test new bonding descriptors on its ability to recognize the harpoon mechanism and ( ii ) to evaluate the goodness of new atomic partitions to reproduce the maximal esi value in several reactions . despite
the harpoon mechanism involves the electron transfer at large interatomic distances , this reaction can be mistaken by a regular electron reorganization process that does not involve a harpooning process .
indeed , in fig .
[ fig : pops ] one finds the population analysis of beh , which is formed by the harpoon mechanism at short interatomic distances .
the be population profile could be easily confused with a regular electron reorganization process as that taking place in co. therefore , the atomic population can be used to monitor the electron - density exchange between atoms but it can not be always used to discriminate the reaction mechanism . in all
studied mechanisms , the maximal transfer variation point along the bond formation occurs when about half electron has been transferred from one atom to another . if the process takes places through a harpoon mechanism , this point of the reaction path coincides with the avoided crossing ( if there is any ) .
thus far , the only bonding descriptor that had actually been shown to identify the harpoon mechanim from other processes is the esi @xcite , which shows a maximum around the avoided crossing ( see fig .
[ fig : esis ] ) .
the existence of a maximum of the esi does not guarantee the existence of an avoided crossing as proved by the bh molecule .
however , the esi provides a simple tool to identify the existence of the avoided crossing if there is such point in the potential energy curve .
some of us have shown in the past that atomic populations and multicenter esi are severely affected by the atomic partition @xcite but , in the big picture , atomic partitions do not have a large effect on the relative two - center esi values @xcite .
however , fig [ fig : esis ] shows that the presence of a maximum esi value in bonds formed by the harpoon mechanism is a very constrictive test for atomic partitions .
the tfvc is among the few atomic partitions that can pass such test ( see fig .
s11 ) , even though the esi values calculated with this partition fail to identify the bond in lih s a@xmath20 state as more covalent than the bond in the x@xmath20 state at short distances ( r=@xmath47 ) .
this work also shows that elf and laplacian plotted along the molecular axis at different interatomic distances permit to monitor the bond formation process better than its three - dimensional counterparts .
both tools describe the transferred electron as a one - dimensional peak that moves from one atom to another as the process takes places .
the position of the peak with respect to the qtaim boundary permits to follow this motion and shows that the peak passes from one atom to the other at the vicinity of the avoided crossing , where the esi value and the transfer variation are maximal .
global information theory measures based on the electron density seem to provide some distinctive features for different mechanisms of bond formation but they do not provide a convincing description that allows a clear - clut separation between the different processes . sometimes the descriptors provide alike profiles for different bonding mechanisms and other times they provide very disparate results for similar processes . in addition , the results obtained for the h@xmath18 dissociation are not trivial to interpret and suggest that previous successful results describing chemical reaction paths were , to some extent , fortuitous @xcite .
the information theory descriptors analyzed in the atomic regions within the molecule @xcite or the orbital entaglement measures @xcite ( which can be used to locate avoided crossings ) could actually provide some insights into the bonding mechanism of molecules but such study is beyond the scope of this work .
the authors thanks prof . bernard silvi for providing ref . @xcite .
this research has been funded by the mineco projects ctq2014 - 52525-p and ctq2014 - 54306-p , the basque country consolidated group project no .
it588 - 13 , the feder grant ungi10 - 4e-801 ( european fund for regional development ) , the generalitat de catalunya ( project number 2014sgr931 , xarxa de referncia en qumica terica i computacional , and icrea 2014 prize to m.s . ) and the predoctoral fellowship fpu13/00176
. excellent service by the centre de serveis cientfics i acadmics de catalunya ( cesca ) and technical and human support provided by sgi / izo - sgiker upv / ehu are gratefully acknowledged .
m. j. frisch , g. w. trucks , h. b. schlegel , g. e. scuseria , m. a. robb , j. r. cheeseman , g. scalmani , v. barone , b. mennucci , g. a. petersson , h. nakatsuji , m. caricato , x. li , h. p. hratchian , a. f. izmaylov , j. bloino , g. zheng , j. l. sonnenberg , m. hada , m. ehara , k. toyota , r. fukuda , j. hasegawa , m. ishida , t. nakajima , y. honda , o. kitao , h. nakai , t. vreven , j. a. montgomery , jr .
, j. e. peralta , f. ogliaro , m. bearpark , j. j. heyd , e. brothers , k. n. kudin , v. n. staroverov , r. kobayashi , j. normand , k. raghavachari , a. rendell , j. c. burant , s. s. iyengar , j. tomasi , m. cossi , n. rega , j. m. millam , m. klene , j. e. knox , j. b. cross , v. bakken , c. adamo , j. jaramillo , r. gomperts , r. e. stratmann , o. yazyev , a. j. austin , r. cammi , c. pomelli , j. w. ochterski , r. l. martin , k. morokuma , v. g. zakrzewski , g. a. voth , p. salvador , j. j. dannenberg , s. dapprich , a. d. daniels , . farkas , j. b. foresman , j. v. ortiz , j. cioslowski , and d. j. fox , gaussian 09 revision d.01 , gaussian inc .
wallingford ct 2009 . | the lowest - lying states of lih have been widely used to develop and calibrate many different methods in quantum mechanics . in this paper
we show that the electron - transfer processes occurring in these two states are a difficult test for chemical bonding descriptors and can be used to assess new bonding descriptors on its ability to recognize the harpoon mechanism . to this aim , we study the bond formation mechanism in a series of diatomic molecules . in all studied electron - reorganization mechanisms , the maximal electron - transfer variation point along the bond formation path occurs when about half electron has been transferred from one atom to another .
if the process takes places through a harpoon mechanism , this point of the reaction path coincides with the avoided crossing .
the electron sharing indices and one - dimensional plots of the electron localization function and the laplacian of the electron density along the molecular axis can be used to monitor the bond formation in diatomics and provide a distinction between the harpoon mechanism and a regular electron - reorganization process .
this paper is dedicated to andreas savin on the occasion of his @xmath0 birthday . ] |
this work is concerned with @xmath1 ym theory ( qcdn ) in 4d .
qcdn admits a large @xmath0 expansion @xcite . lattice work has shown that there is confinement at finite and infinite @xmath0 @xcite .
then , qcdn at @xmath2 ( qcd@xmath3 ) is similar to the starting point of the topological expansion ( te ) @xcite . in te one constructs iteratively an s - matrix from a set of postulated basic general properties .
another starting point for the te is provided by string theory . in both cases
one starts from a system consisting of an infinite set of stable particles interacting weakly at linear order . upon iteration
, other singularities build up .
the expansion is organized diagrammatically with an order given by the genus of a riemann surface .
the qcdn route is better founded than the string one .
we can safely assume that there exist wightman @xmath4-point functions of local gauge invariant observables that admit a single valued continuation to the extended tube @xmath5 @xcite for any @xmath0 .
these functions determine the leading nontrivial term in @xmath6 of any amplitude entering the s - matrix . from this off shell starting point one might be able to build a better founded qcdn string theory @xcite .
concretely , one would need explicit forms of a least some of the sets of entries of the s - matrix . despite quite a few papers which achieved high levels of popularity ,
there is not one non - perturbative physical number that has been analytically calculated , or , at least credibly estimated , in qcdn ( with or without a finite number of quarks ) at leading order in @xmath6 or @xmath7 .
nevertheless , interest in large @xmath0 does not seem to die out .
quite a few workers , me included , still are trying to get some new quantitative result in qcdn which rests on the simplification afforded by @xmath8 .
my idea has been to find a simple physical single scale observable whose behavior as a function of this scale showed a universal behavior at the crossover separating long from short scales .
large @xmath0 comes in to provide this universality by a large @xmath0 phase transition .
the universality then becomes a random - matrix type of universality .
the hope is to exploit it in order to match an effective string descriptions holding at large distances to perturbation theory holding at short distances .
for example , consider a circular wilson loop of radius @xmath9 .
for @xmath9 large effective string theory provides some universal information about the @xmath9 dependence , while at small @xmath9 perturbation theory applies ; the new ingredient is that random matrix universality would provide the means to connect these two dependencies .
the hope is that an approximate connection between the string tension and some standard perturbative scale would then be calculable .
the existence of the large @xmath0 phase transition is believable for the circular loop because it has been established numerically for square loops .
however , it would be preferable to consider smooth loops also on the lattice , and this leaves us with only polyakov loops winding around a compactified direction .
the length of this circle has to be bounded from below in order to stay in the confined phase .
the single eigenvalue density , @xmath10 , of a polyakov loop becomes uniform at @xmath2 on account of the well known @xmath11 symmetry .
this leaves us with @xmath12 , the connected correlation function of the @xmath10 s of two separated polyakov loops , as the simplest smooth observable on the lattice . in this paper ,
i focus on polyakov loops .
the outline of the papers is : sections [ vol_red ] and [ two_loop_a ] provide background material .
the concrete new results are in [ two_loop_b ] .
they consist of an evaluation of the single polyakov eigenvalue density connected two point correlation function under the assumption of second rank casimir dominance .
a formula for any @xmath0 ( taken as odd , for simplicity ) is provided , the large @xmath0 limit is taken and the validity of the latter is checked numerically .
next , a brief comparison with monte carlo data in four dimensional @xmath1 yang mills theory is carried out .
there are no large @xmath0 phase transitions .
incidentally it is noted that the result does not show universal features known to hold for large hermitian matrix ensembles .
section [ other ] contains ideas for future work . a short summary concludes the paper .
qcdn is a field theory , but geometrically the fundamental variables are not fields defined over @xmath13 , but rather fields defined over loops in @xmath14 .
this becomes particularly evident when one introduced a lattice uv cutoff : one can derive an infinite set of equations connecting various loop operators and the equations reflect the ordinary locality of four - space the field theoretical formulation rests on , without any of the collateral expenditures ( gauge fixing , faddeev - popov ghosts , gribov ambiguities ) associated with formulating the continuum theory in terms of gauge fields @xcite .
the loop equations self - truncate at infinite @xmath0 , feeding the hope that it ought to be easier to handle non - perturbative issues of qcdn at @xmath2 @xcite . taking the equations to the continuum
is hampered by the nonexistence of anything resembling a decent calculus in loop space .
one way to go around this obstacle is to try to guess a well defined solution directly in the continuum ( which obeys general symmetry / unitarity constraints ) and show that it satisfies a set of equations that can be viewed as a concrete realization of the formal continuum loop equations @xcite .
this has led to progress in string theory and even to a connection back to field theory , but not for qcdn @xcite .
as far as i know , we still do not have even one nontrivial example where the formal loop equations have been credibly defined in the continuum .
one consequence of the loop equation is that at @xmath2 the replacement of @xmath14 by @xmath15 , where the sides of the torus are all larger than the inverse deconfinement temperature , preserves a large subset of observables with no dependence on the actual finite length of these sides @xcite .
this is of some help in numerical work , but the saving is quite limited @xcite .
the term describing this phenomenon is `` reduction '' , on account of a reduction in the number of degrees of freedom as far as the four volume size goes .
reduction can be applied to any number of directions and one assumes that there is a hierarchy of scales associated with the preservation of the associated @xmath16 , @xmath17 @xcite .
the restriction on the sides of @xmath15 ensures that the global @xmath18 symmetry now present is not broken spontaneously by the @xmath2 limit .
the preservation of the consequence of this symmetry on expectation values of parallel transport round non - contractible loops is a necessary ingredient for reduction @xcite .
the equivalence between the @xmath14 and @xmath15 loops equations breaks down if @xmath18 is not obeyed by expectation values of winding loops at @xmath2 @xcite .
if the @xmath18 is preserved at @xmath2 , the spacings between momenta induced by the finite volume of @xmath15 get continuously filled in by the eigenvalue - sets of winding loops @xcite .
polyakov loops are the natural extra observables one has when considering @xmath19 .
we assume that the compact direction is large enough that the one @xmath11 is preserved at all @xmath0 . when we slice @xmath20 by three - spaces parallel , or orthogonal , to the compact direction we find two different transfer matrices .
they provide two hamiltonian pictures
. in one picture @xmath11 is just an extra global symmetry of the hilbert space and the system is at zero temperature . in the other ,
the system is at finite temperature . in either case ,
one has a hilbert space which can be chosen to transform irreducibly under the symmetries commuting with the hamiltonian .
the spaces and hamiltonians are different , providing different spectral representations of identical observables .
reduction applies only to non - winding loops ; it is of interest to see if any reduction - related simplifications hold at @xmath2 also for winding loops @xcite .
define the parallel transport winding loop operator by : @xmath21 the compact direction is 4 .
@xmath22 are the gauge fields given by traceless hermitian @xmath23 matrices and @xmath24 .
@xmath25 is path ordering .
polyakov loops are independent of @xmath26 : @xmath27 @xmath28 labels an irreducible representation of @xmath1 , @xmath29 is the character in @xmath28 and @xmath30 is the dimension of @xmath28 .
if the number of boxes in the young pattern corresponding to @xmath28 is @xmath31 , the @xmath0-ality of @xmath28 is given by @xmath32 .
under @xmath11 we have @xmath33 .
consider the case @xmath34 .
then @xmath35 and @xmath36 is generically nonzero . here
@xmath37 is conjugate to @xmath28 .
@xmath9 is a positive distance . denoting the length of the compact direction by @xmath38 ,
the two point function is a function of @xmath28 and the two length scales @xmath39 .
define ( formally ) @xmath40 .
we assume that the @xmath41-parameter in front of the @xmath42 $ ] parameter in the action is set to zero , so @xmath43 is real and positive .
the definition is only formal because after renormalization there will be an arbitrary term in @xmath44 of the form @xmath45 .
thus , @xmath46 is well defined up to an additive constant @xcite . since @xmath47 where @xmath48 is the deconfinement temperature
, the first reaction would be that @xmath49 can not be computed in perturbation theory .
nevertheless , to some extent , the quantity @xmath50 can . instead of the @xmath51 limit in eq .
( [ pot ] ) one takes an infinite uncompactified @xmath26 axis and replaces @xmath38 by @xmath52 and @xmath53 by the logarithm of a rectangular wilson loop of shape @xmath54 .
inspection of feynman diagrams shows that in the @xmath55 limit one has an expansion for @xmath56 up to two loop order in the coupling which ought to be useful for @xmath9 small enough . to go to higher loops one needs to include non - analytic terms in the coupling @xcite .
pnrqcd provides a prescription for how to do this , but i doubt that it is unique @xcite . in this expansion
the path integral over @xmath57 is expanded around @xmath58 , which breaks the @xmath11 .
this may not matter for the two point polyakov loop function at infinite @xmath38 .
whether it can be used at any finite @xmath38 is an open question , so long as @xmath47 .
if there were a credible perturbative regime in which the @xmath38-scale can somehow be removed from the problem , one would expect that @xmath12 might show some crossover as @xmath9 is varied .
then , the ingredients necessary for a large @xmath0 phase transition to develop are present .
the hope is that , for @xmath9 small enough , the eigenvalues of one polyakov loop would restrict the fluctuations of the eigenvalues of the other polyakov loop to such a degree that the periodicity in the angle differences would barely be felt .
for @xmath9 large this periodicity would get restored to full strength .
a separating crossover at finite @xmath0 would become a phase transition at infinite @xmath0 .
the intuition behind the focus on eigenvalues is that collectively their fluctuation explore the distance of parallel transport operators from unity . only when the compactness of the group is felt does one expect non - perturbative effects to become important .
compactness is felt only when parallel transport exceeds a certain distance from unity .
we shall see that the hope for a transition is not realized . beyond perturbation theory ,
@xmath56 is the ground state energy of the hamiltonian associated with evolution in the @xmath26 direction in the sector defined to transform under the local gauge group as @xmath28 at @xmath59 and @xmath37 at @xmath60 .
this ground state is @xmath61 degenerate .
when one computes the partition function viewing @xmath38 as the inverse temperature , the degeneracy of the ground state cancels the prefactors normalizing the polyakov loop operators .
there is no overall factor of @xmath62 in the physical piece of the free energy .
because of the representation content of the hilbert space which breaks translation invariance , there is no physical interpretation in this picture for plane waves propagating in the @xmath63 directions superposing ground state states . because the sources can move . at three loop order the expansion in powers of the strong - force coupling breaks down , but the non - analytic term responsible for this ( in this framework ) can be derived . ]
the dependence of @xmath64 on @xmath9 for @xmath65 starts with a constant ( the string tension of open strings with fixed endpoints transforming as @xmath28 and @xmath37 respectively ) and continues to subleading orders ; several terms in this expansion are universal .
i shall describe below in more detail this aspect in another asymptotic limit .
one can also consider the large @xmath9-separation fixed @xmath38 limit : @xmath66 confinement in this context means that @xmath67 , with @xmath68 depending only on the assumed non - zero @xmath0-ality of @xmath28 , @xmath4 ; @xmath69 .
@xmath70 gives the @xmath38-derivative of the @xmath9-derivative at @xmath71 of eigenvalues of the hamiltonian describing evolution in any one of the directions @xmath72 in the @xmath4-sector of the global @xmath11 .
now , it does make sense to superpose states and project on zero spatial momentum .
this gives the ground state energy in the @xmath4-winding sector .
one can look at several subleading terms in the large @xmath38 expansion . for this define @xmath73 @xmath74 assuming an effective string theory description one derives from symmetry principles alone that the coefficients @xmath75 are universal calculable finite numbers , independent of @xmath28 ( and consequentially of @xmath4 ) .
they are actually also independent of any other detail regarding the field theory , except the assumption of confinement and applicability of effective string theory @xcite . to summarize : at any finite @xmath0 , @xmath76 and @xmath77 with @xmath28 of nonzero @xmath0-ality have some universal coefficients in their asymptotic expansions in @xmath9 and @xmath38 respectively , which follow from the assumption of confinement and applicability of effective string theory .
numerical checks have yielded results consistent with this .
the @xmath78 limit provides no further simplification with respect to these properties . and the @xmath78 limit would turn the interactions off .
one would need the @xmath78 limit in order to justify the focus on the lowest genus surface relevant to the correlation under investigation .
the effective string theory on the other hand has already summed up all contributions from handles of the dream string theory , and the universal results provided by the cylinder are @xmath0-independent .
] there is a hope that the @xmath78 limit could provide a clear demarcation point for the domain in which the asymptotic expansion of long strings can make any sense at all , but the universal predictions of effective string theory are insensitive to this .
effective string theory can also be applied to the case where @xmath38 and @xmath9 are both taken to infinity ; a brief overview with references to original work can be found in @xcite .
this can be done also for ordinary contractible rectangular wilson loops in @xmath14 .
reduction applies in these cases .
there is no lower limit on @xmath38 or @xmath9 and therefore there is a perturbative regime .
renormalization to remove perimeter divergences is still required .
again , for individual irreducible representations large @xmath0 provides no additional constraints on the universal large @xmath39 results of effective string theory .
however , now there is an evident connection to a perturbative regime where both @xmath38 and @xmath9 are small .
this connection is smooth . only by looking at many representations
simultaneously does one detect a large @xmath0 transition : one finds a non - analyticity in the single eigenvalue distribution at a point which serves as a boundary between the perturbative and non - perturbative regime @xcite .
the boundary location depends on an arbitrary dimensional smearing parameter .
this parameter is a remnant of the need to eliminate the perimeter divergence .
it is defined in a manner independent of @xmath28 @xcite . only at infinite @xmath0
is there a well defined transition point . for any finite @xmath0
one has only a cross - over .
numerical and analytical considerations indicate that the large @xmath0 transition has a certain random - matrix model universality @xcite . at any finite @xmath0
the polyakov loop two point function has a discontinuity as @xmath38 is varied through @xmath79 separating the confined and deconfined phases . since one can compute @xmath56 in perturbation theory at @xmath80 , one may hope that a specifically transition as @xmath9 is varied takes place at any @xmath38 in the confined phase .
lattice checks of predictions of effective string theory have overall been successful , at times even surprisingly successful ; this has nothing to do with @xmath0 being large enough . just like effective lagrangians for massless pions work already on the lattice since these lagrangians can parametrize any reasonable uv behavior , effective string theory
should also hold directly on the lattice so long as one is in the rough phase . for long strings the lattice violation of @xmath81 invariance makes an impact determined by symmetries .
if we replace the hypercubic lattice by an @xmath82 lattice @xcite , the effective string theory predictions may apply even better .
so , the success of lattice checks of effective string theory might be `` explained '' by the relative unimportance of the proximity of the field theoretical continuum limit .
for example , qcdn with scalar matter fields in the adjoint has no continuum limit ; on the lattice it would confine and effective string theory would make some universal predictions in the rough phase of the loop under consideration . in general , strings are less sensitive to high mode cutoffs than field theories are @xcite . in four dimensions
there is an exception @xcite which has to do with the fluctuations in the extrinsic string curvature .
potentially significant deviations from effective string theory were observed in @xcite and independently in @xcite ; in @xcite these deviations were tentatively attributed to the corners of lattice loops because of the perturbation theory experience with the corner divergences in wilson loops .
the technical reason for the universality of the large @xmath0 transition in the case of contractible wilson loops has been second rank casimir dominance of the dependence phase transitions . ] on the representation @xmath28 .
one may then map the size parameters that are varied into an appropriate measure of separation in 2d ym .
so , we ask whether there is a large @xmath0 phase transition to be found in an analysis of polyakov loop correlators in ym on a 2d cylinder .
in general , it is known that there are large @xmath0 non - analyticities in 2d ym on a cylinder @xcite . here
we wish to see if they show up in the two point function of the single eigenvalue distributions associated with two separated polyakov loops . in the context of large size
hermitian matrices two point single eigenvalue correlation functions have been shown to have some universal properties @xcite .
a side result of the calculation below will be to check whether this universality extends to the simplest unitary matrices ensemble with a global @xmath11 symmetry . from experience with hermitian matrix models we expect smooth and strongly fluctuating contributions to the two point eigenvalue function of equal magnitude ; the oscillating piece has to be first separated out and only the remaining smooth piece can exhibit a universal large @xmath0 phase transition . in the subsequent equations the assumption that @xmath0 is odd is made implicitly .
the partition function of @xmath1 ym on a 2d cylinder connecting two loops , @xmath83 and @xmath84 , is given by @xcite : @xmath85 @xmath86 is the area in some area unit . let @xmath87 and @xmath88 be the single eigenvalue distributions associated with the @xmath23 unitary matrices @xmath89 respectively @xcite : @xmath90 the @xmath91 are the eigenvalues of @xmath92 and @xmath93 is the @xmath94-periodic delta - function .
the character expansion of @xmath95 is : @xmath96 \label{rhoone}\ ] ] the irreducible representation @xmath97 has a young pattern in the shape of a width - one hook , with @xmath98 rows and @xmath99 columns .
we wish to calculate the connected two point function @xmath100\ ] ] @xmath101 is the haar measure on @xmath1 . in order for a pair @xmath102 and @xmath103 to contribute we need that there be a singlet in their direct product .
as @xmath0 is odd , this will happen only when one pair is the conjugate of the other .
for odd @xmath0 there are no @xmath97 self - conjugate pairs .
for @xmath104 we have : @xmath105\ ] ] here @xcite , @xmath106 one can reorganize the sums to get @xmath107 the real part is taken using eq .
( [ rhoone ] ) with fixed small @xmath108 and subsequently setting @xmath109 .
@xmath110 @xmath111 contains the two - point connected correlations @xmath112 for @xmath113 . to obtain the large @xmath0 limit we write an integral representation @xcite , exploiting the quadratic nature of the @xmath114 dependence on @xmath115 : @xmath116 here ,
@xmath117 $ ] .
the @xmath118 integral can be done by saddle point with @xmath119 . to
leading order in @xmath0 we get @xmath120 we need to keep the factor @xmath121 in the denominator of the first fraction in the integrand in order to ensure its regularity for @xmath122 , because then the numerator divides evenly by the denominator .
carrying out the integral by saddle point we find two contributions , corresponding to the saddle points @xmath123 and @xmath124 .
the large @xmath0 limit is taken at finite non - zero @xmath125 and finite positive @xmath86 .
the limits @xmath126 and @xmath127 do not commute with the limit @xmath78 .
the final results for the leading large @xmath0 behavior comes out to be : @xmath128 here , @xmath129 .
the result is the sum of a smooth term and a rapidly oscillating one .
there are no large @xmath0 phase transitions .
the non - oscillating term is : @xmath130 it does not exhibit the universal structure seen in continuous chains of large hermitian matrix models @xcite .
most likely , the main difference is that for the unitary matrix ensemble we solved above there is no analogue of the nontrivial potential @xmath131 term in the action for the hermitian @xmath132 matrices . while the explicit form of @xmath133 is irrelevant , its mere presence is . to get a feel for the goodness of the large @xmath0 limit and also check whether its derivation was correct , we present some figures below . we see that the large @xmath0 approximation deteriorates when @xmath0 decreases , when @xmath134 is close to @xmath135 and when @xmath86 is small , but otherwise holds well
i now turn to 4d and report on some numerical simulation done in order to see qualitatively whether , overall , the data looks similar to the 2d case .
i only wish to confirm that also in the 4d case there are no signs of a large @xmath0 transition .
i do not aim here for anything quantitative and am content with low numerical precision .
the connected single - eigenvalue distribution for two polyakov loops in 4d will be a function of @xmath136 and @xmath137 similarly to the 2d case . for finite @xmath0
, there is no reason for this function to depend only on the angle difference .
the @xmath11 symmetry only provides invariance under simultaneous shifts of @xmath136 and @xmath137 by @xmath138 .
initial simulations were done collecting two dimensional histograms in the @xmath139 plane .
is was found that within practical numerical accuracy collapsing the histograms along constant @xmath140 lines did not loose any information .
this means that we may as well consider the following finite-@xmath0 definition of @xmath12 : @xmath141 @xmath12 depends only on the angle difference on account of the @xmath11 symmetry . in order to eliminate the uv divergences in 4d ,
the gauge field configurations were smeared . unlike in previous work @xcite , gauge fields along the direction separating the polyakov loops were left unsmeared ; smearing was only done for gauge fields tangent to all orthogonal three - spaces .
thus , the gauge fields entering the polyakov loop are smeared .
this is enough to remove the perimeter divergence , as can be seen from the formula of a massless propagator smeared in the above manner at infinite volume : @xmath142 one has then @xmath143 and the short distance singularity is regulated away .
this removes the perimeter divergence at the cost of a dependence on the new scale @xmath144 . at the level of feynman
diagrams it is obvious that this eliminates all uv perimeter divergences to any finite order , because the extra diagrams smearing introduces have a tree structure , reflecting the determinism of the smearing equation , which , in continuum notation reads : @xmath145 like in @xcite , @xmath146 is a coordinate along a new direction .
@xmath147 label directions orthogonal to the direction of separation between the two polyakov loops .
@xmath148 and @xmath149 are the covariant derivative and field strength respectively .
the three dimensional character of the smearing means that the quantities @xmath70 are @xmath146-independent on account of the limit @xmath65 which projects on the ground state of the relevant hamiltonian .
smearing only affects the ( regularized ) matrix element of the operator between the singlet @xmath11 ground state and the nontrivial @xmath11 ground state . in simulations we employed @xmath150 in lattice units , which is a moderate amount of smearing , found adequate in the study of contractible wilson loops @xcite .
the figures below show results from a lattice volume of @xmath151 with @xmath152 at inverse t hooft couplings , @xmath153 and separation @xmath154 in lattice units .
we see that at fixed @xmath155 the general behavior resembles the analytical results in 2d with @xmath86 increasing with @xmath9 .
one also sees a trend of increase in the difference from 2d as the angle difference increases .
only the angle difference range of @xmath156 is plotted on account of the symmetry under simultaneous sign switch of @xmath139 . as @xmath155 increases the effective @xmath86 decreases , as expected on account of asymptotic freedom .
the errors on the mc data are of the order of 10 percent but can not be reliably estimated .
each figure shows , in addition to raw data , a smoothed curve obtained by a cubic spline smoothing method .
the method of smoothing consists of a minimization of a weighted combination of some average of curve curvature and deviation from the data .
the smoothing procedure is quite ad - hoc , and only serves to produce curves to guide the eye .
i discovered no analogues to the large @xmath0 phase transition found for contractible wilson loops in @xmath14 in this case .
the reason must be that one can not make @xmath38 small .
if we try , we hit a discontinuity already at finite @xmath0 ; this discontinuity becomes quite spectacular at @xmath2 @xcite .
there seems to be no way to qualitatively distinguish between the finite @xmath0 discontinuity and the @xmath2 one .
so , we would like to work in a metastable phase where @xmath157 but @xmath11 is still a good symmetry . more mathematically
, we want to analytically continue in @xmath38 from the low temperature to the high temperature phase .
as pointed out in @xcite there is a way to define also a string description there .
the instability in the string theory occurs when some winding string states go tachyonic @xcite . and
@xmath158 are given by the right and left winding numbers .
the constraint on the closed string states that ties the left and right movers ends up leaving one extra integer labeling the string modes . for a small circumference ,
nontrivial winding modes have positive mass squares , overcoming the negative additive contribution reflecting the ordinary tachyon , but for a larger circumference some of the winding modes have negative mass square .
the smallest non - zero windings cross the tachyonic threshold at the hagedorn temperature . ]
numerically , one might try to go into the metastable phase by using quenching .
originally quenching was introduced as a device to maintain the global @xmath18 of @xmath15 and reduction to zero volume @xcite , but the idea was flawed @xcite .
the flaw was that it still left alive an annealed mechanism for breaking the @xmath18 to some proper subgroup . without respecting full @xmath18 symmetry
, reduction fails .
however , when a single direction is compactified , the preservation of the single @xmath11 would not suffer from this flaw . to be sure ,
we choose prime @xmath0 because then , unlike @xmath18 , @xmath11 has not proper subgroups . the point would be to check numerically whether the condensation of winding states , which would occur beyond the hagedorn temperature , is something that occurs in the analytically continued field theory at finite @xmath0 as a phase transition ( continuous or not ) . the alternative might be that there is no such ordinary phase transition , since @xmath38 is small enough to enter a field theoretic perturbative regime where string theory of any traditional sort is inapplicable . then , on the basis of analogy with the contractible wilson loops , one would guess that a large @xmath0 phase transition would develop in the two point single eigenvalue correlation function .
the investigation of this is left for the future . the large @xmath0 phase transition for contractible loops
is seen only when considering simultaneously many irreducible representations .
they may be viewed as coming from multiple windings of the boundary of the loop . since the loop is contractible ,
winding numbers are not conserved . in the polyakov case
they are , at least for windings between 0 and @xmath159 . the analytic continuation from @xmath160 to @xmath161 should provide a way to compute ( for small @xmath9 and @xmath38 ) @xmath53 from the two polyakov loop correlator directly and for arbitrary irreducible representations . the analytic continuation would amount to expanding around the one - loop unstable saddle point , given by @xmath162\ ] ] for odd @xmath0 , @xmath163 , where @xmath164 indicates a diagonal matrix with the listed elements on its diagonal .
this configuration is @xmath11 invariant , but unstable at one loop order .
there are two ingredients in 2d ym : one is the `` propagator '' defining the cylinder with fixed circular boundaries and the other is the `` vertex '' which sews together three boundaries @xcite .
this indicates that it would be of interest to study the connected correlation function of three polyakov loops .
the simplest example is to take 2 polyakov loops in the fundamental and a third in the irreducible representation made by combining two anti - fundamentals into a symmetric or anti - symmetric irreducible representation .
take @xmath165 and odd .
the three loops are positioned at distinct locations in @xmath166 . in euclidean space , using a different slicing this looks like a finite temperature setting for one among the many possible generalizations to large @xmath0 of @xmath167 baryons .
three infinitely heavy quarks are connected by a v - shaped string configuration .
one could go to minkowski space and endow these locations with zero masses , forcing them to evolve in time at the velocity of light .
for open strings in minkowski space such a situation was looked at in the context of cosmic strings . for @xmath167
this was considered in several papers , but the string tensions were taken to be equal @xcite . an effective string theory valid at large separations
would need to handle a case where one couples two strings of the same tension to one of a different tension .
it seems to me that string tension considerations would favor a v - shaped arrangement of `` fundamental '' strings .
it would be interesting to apply the methods of effective string theory to this setup . assuming the v - shape , in the field theory there would be a coupling associated with the vertex of the v. for large @xmath0 it would go as @xmath168 with a finite @xmath169 . back to fixed sources ,
the three point function of polyakov loops for the antisymmetric case is given by : @xmath170tr ( u_{p_3})\rangle ( r_a , r_b,\theta)\ ] ] here , @xmath171 is at @xmath172 , @xmath173 is at @xmath174 and @xmath175 is at @xmath176 with @xmath177 and @xmath178 . the new ingredient is the presence of corners . in the case of rectangular wilson loops , corners may change the rules of effective string theory , by exhibiting a field theoretical dependence on loop sides which is not exponentially suppressed even for asymptotically large loops . here , the same question can be addressed in a different set - up . on a hypercubic lattice only @xmath41 values which are multiples of @xmath179
. numerically there would be high noise problems , but it is worth a try .
one could then get back at our main theme , and consider the connected three point function of the eigenvalues of the three loops . to search for large @xmath0 phase transitions one would need to look at three point connected correlations @xmath180 , depending on two eigenvalue - angle differences at infinite @xmath0 .
it might be of interest to consider the problem of colliding two same direction wound polyakov loops in minkowski space .
the three point vertex would enter twice to produce a two to two particle scattering dominated by the exchange of the symmetric and antisymmetric long strings with masses above and below threshold .
the distribution of the excited string modes of the two separate outgoing strings might provide a thought experiment reminiscent to the bjorken model for high energy nucleus - nucleus collisions @xcite .
the correlations among single eigenvalue distributions associated with various polyakov loops has been studied for the simplest arrangement and found to provide no large-@xmath0 generated non - analyticities .
the results might be of some interest in random matrix theory .
one needs the interplay between different windings to get large @xmath0 phase transitions and also a perturbative regime .
one idea was to somehow analytically continue in @xmath38 to @xmath181 and follow the evolution of the single eigenvalue distribution of a polyakov loop as a function of @xmath38 .
the other was to construct arrangements involving mixtures of polyakov loops of different winding numbers .
i acknowledge partial support by the doe under grant number de - fg02 - 01er41165 .
i am grateful for support under the weston visiting scientist program at the weizmann institute in the faculty of physics .
i have benefited from communications with zohar komargodski and owe philipsen .
5 g. t hooft , nucl .
b72 ( 1974 ) 461 .
b. lucini , m. panero , phys .
526 ( 2013 ) 93 . g. f. chew , c. rosenzweig , phys .
41 ( 1978 ) 263 ; g. veneziano , nucl .
b74 ( 1974 ) 365 ; phys .
lett . 52b ( 1974 ) 220 .
pct , r. f. streater , a. s. wightman , `` spin and statistics , and all that '' , w. a. benjamin ( 1964 ) .
g. veneziano , il nuovo cimento , vol .
lvii a , n. 1 ( 1968 ) 190 .
m. makeenko , a. a. migdal , phys .
b 88 ( 1979 ) 135 ; d. foerster , phys . lett .
87b , 87 ( 1979 ) ; t. eguchi , phys .
87b , 91 ( 1979 ) ; d. weingarten , phys . lett .
87b , 97 ( 1979 ) .
a. m. polyakov , nucl .
b ( proc . suppl . ) 68 ( 1998 ) 1 .
j. maldacena , int .
jou . of th .
38 ( 1999 ) 1113 .
j. kiskis , r. narayanan , h. neuberger , phys .
b 574 ( 2003 ) 65 .
r. lohmayer , h. neuberger , jhep 08 ( 2012 ) 102 .
t. eguchi , h. kawai , phys .
48 ( 1982 ) 1063 .
g. bhanot , u. heller , h. neuberger , phys . lett .
b 113 ( 1982 ) 47 .
h. neuberger , phys .
b 119 ( 1982 ) 179 .
a. m. polyakov , nucl .
b 164 ( 1980 ) 171 .
o. jahn , o. philipsen , phys .
d. 70 ( 2004 ) 074504 .
t. appelquist , m. dine , i. j. muzinich , phys .
d. 17 ( 1977 ) 2074 a. pineda , progress in particle and nucl .
67 ( 2012 ) 735 ; o. aharaony , z. komargodski , arxiv:1302.6257v2 .
r. lohmayer , h. neuberger , phys .
( 2012 ) 061602 .
r. narayanan , h. neuberger , jhep 03 ( 2006 ) 064 ; r. lohmayer , h. neuberger , pos(lattice 2011 ) 249 .
r. lohmayer , h. neuberger , t. wettig , jhep 05 ( 2009 ) 107 ; h. neuberger , phys .
b 666 ( 2008 ) 106 .
h. neuberger , phys .
b199 ( 1987 ) 536 .
m. karliner , i. klebanov , l. susskind , int .
j. of mod .
phys . a vol . 3 ( 1988 ) 1981 .
a.gonzalez-arroyo , m. okawa , phys . lett . b718 ( 2013 ) 1524 .
d. j. gross , a. matytsin , nucl .
b 437 ( 1995 ) 541 . c. w. j. beenakker , rev . of mod . phys .
69 ( 1997 ) 731 ; b. eynard j. phys .
a : math . gen . 31
( 1998 ) 8081 . c. b. thorn , phys .
b99 ( 1981 ) 458 .
j. polchinski , phys .
68 ( 1991 ) 1267
. j. j. atick , e. witten , nucl .
b 310 ( 1988 ) 291 .
h. neuberger , m. sc .
thesis , `` bosonization in field theory in two space - time dimensions '' , rx-744 ( tel - aviv ) ( 1976 ) .
b. bringoltz , s. sharpe , phys .
d 78 ( 2008 ) 034507 .
e. j. copeland , t. w. b. kibble , d. a. steer , phys .
97 ( 2006 ) 021602 ; g. s. sharov , th . math . phys .
140 ( 2004 ) 256 and 1109 ; phys .
d79 ( 2009 ) 114025 ; g. t hooft , arxiv : hep - th/0408148
. j. d. bjorken , phys .
d 27 , ( 1983 ) 140 . | polyakov loop eigenvalues and their @xmath0-dependence are studied in 2 and 4 dimensional su(n ) ym theory .
the connected correlation function of the single eigenvalue distributions of two separated polyakov loops in 2d ym is calculated and is found to have a structure differing from the one of corresponding hermitian random matrix ensembles .
no large @xmath0 non - analyticities are found for two point functions in the confining regime .
suggestions are made for situations in which large-@xmath0 phase transitions involving polyakov loops might occur . |
observations of a sample of three late - type galaxies with low surface - brightness and the radio - weak edge - on galaxy ngc 5907 ( all with a low sfr ) revealed that they all have an unusually high thermal fraction and weak total and regular magnetic fields ( chyy et al .
2007 , dumke et al .
however , these objects still follow the total radio - fir correlation , extending it to the lowest values measured so far .
hence , these galaxies have a lower fraction of synchrotron emission than galaxies with higher sfr .
it is known that the thermal intensity is proportional to the sfr .
our findings fit to the equipartition model for the radio - fir correlation ( niklas & beck 1997 ) , according to which the nonthermal emission increases @xmath0 and the _ total _ magnetic field strength @xmath1 increases @xmath2 .
+ no similar simple relation exists for the _ regular _ magnetic field strength .
we integrated the polarization properties in 41 nearby spiral galaxies and found that ( independently of inclination effects ) the degree of polarization is lower ( @xmath3 ) for more luminous galaxies , in particular those for @xmath4 ( stil et al .
the radio - brightest galaxies are those with the highest sfr .
though a dynamo action needs star formation and supernova remnants as the driving force for velocities in vertical direction , we conclude from our observations that stronger star formation seems to reduce the magnetic field regularity . on kpc - scales , chyy ( 2008 ) analyzed the correlation between magnetic field regularity and sfr locally within one galaxy , ngc 4254 . while he found that the total and random field strength increase locally with sfr , the regular field strength is locally uncorrelated with sfr .
we determined the exponential scale heights of the total power emission at @xmath5 cm for four edge - on galaxies ( ngc 253 , ngc 891 , ngc 3628 , ngc 4565 ) for which we have combined interferometer and single - dish data ( vla and the 100-m effelsberg ) . in spite of their different intensities and extents of the radio emission , the vertical _ scale heights _ of the thin disk and the thick disk / halo are similar in this sample ( 300 pc and 1.8 kpc ) ( dumke & krause 1998 , heesen et al .
we stress that our sample includes the brightest halo observed so far , ngc 253 , with a very high sfr , as well as one of the weakest halos , ngc 4565 , with a small sfr . for ngc 253 heesen et al .
( this volume ) argued that the synchrotron lifetime ( which is @xmath6 ) mainly determines the vertical scale height of the synchrotron emission and estimated the cosmic ray bulk velocity to @xmath7 km / s .
as this is similar to the escape velocity , it shows the presence of a galactic wind in this galaxy . the fact that we observe similar averaged scaleheights at @xmath5 cm for
the four galaxies mentioned above imply that the galactic wind velocity is proportional to @xmath8 , and hence proportional to @xmath9 .
in a larger sample of 11 edge - on galaxies we found in all of them ( except the inner part of ngc 4631 , see krause 2009 ) mainly a disk - parallel magnetic field along the galactic midplane together with an x - shaped poloidal field in the halo .
our sample includes spiral galaxies of different hubble types and sfr , ranging from @xmath10 .
the disk - parallel magnetic field is the expected edge - on projection of the spiral magnetic field within the disk as observed in face - on galaxies .
it is generally thought to be generated by a mean - field @xmath11-dynamo for which the most easily excited field pattern is the axismmetric spiral ( ass ) field ( e.g. beck et al .
the poloidal part of the ass dynamo field alone , however , can not explain the observed x - shaped structures in edge - on galaxies as the field strength there seems to be comparable to that of the large - scale disk field .
model calculations of the mean - field @xmath12-dynamo for a disk surrounded by a spherical halo including a _ galactic wind _ ( brandenburg et al . 1993 ) simulated similar field configurations as the observed ones .
new mhd simulations are in progress ( see e.g. gressel et al . this volume , hanasz et al .
this volume ) which include a galactic wind implicitely .
a galactic wind can also solve the helicity problem of dynamo action ( e.g. sur et al .
2007 ) . hence
, a galactic wind may be essential for an effective dynamo action , and to explain the observed similar vertical scale heights and x - shaped magnetic field structure in edge - on galaxies . | from our radio observations of the magnetic field strength and large - scale pattern of spiral galaxies of different hubble types and star formation rates ( sfr ) we conclude that though a high sfr in the disk increases the total magnetic field strength in the disk and the halo the sfr does not change the global field configuration nor influence the global scale heights of the radio emission .
the similar scale heights indicate that the total magnetic field regulates the galactic wind velocities
. the galactic wind itself may be essential for an effective dynamo action . |
the ricci curvature plays a very important role in geometric analysis on riemannian manifolds .
ricci - flat manifolds are riemannian maniflods with ricci curvature vanishes . in physics , they represent vacuum solutions to the analogues of einstein s equations for riemannian manifolds of any dimension , with vanishing cosmological constant .
the important class of ricci - flat manifolds is calabi - yau manifolds .
this follows from yau s proof of the calabi conjecture , which implies that a compact @xmath4 manifold with a vanishing first real chern class has a @xmath4 metric in the same class with vanishing ricci curvature .
they are many works to find the calabi - yau manifolds .
yau conjectured that there are finitely many topological types of compact calabi - yau manifolds in each dimension .
this conjecture is still open . in this paper
, we will study this question on graphs .
first we will give a short history of the definition of ricci curvature in discrete setting .
the definition of the ricci curvature on metric spaces was first from the well - known bakry and emery notation .
bakry and emery@xcite found a way to define the `` lower ricci curvature bound '' through the heat semigroup @xmath5 on a metric measure space @xmath6 .
there are some recent works on giving a good notion for a metric measure space to have a `` lower ricci curvature bound '' , see @xcite , @xcite and @xcite .
those notations of ricci curvature work on so - called length spaces . in 2009 ,
ollivier @xcite gave a notion of coarse ricci curvature of markov chains valid on arbitrary metric spaces , such as graphs .
graphs and manifolds are quite different in their nature .
but they do share some similar properties through laplace operators , heat kernels , and random walks , etc .
many pioneering works were done by chung , yau , and their coauthors @xcite .
a graph @xmath7 is a pair of the vertex - set @xmath8 and the edge - set @xmath9 .
each edge is an unordered pair of two vertices . unless otherwise specified , we always assume a graph @xmath10 is simple ( no loops and no multi - edges ) and connected .
it may have infinite but countable number of vertices .
for each vertex @xmath11 , the degree @xmath12 is always bounded .
starting from a vertex @xmath13 we select a vertex @xmath14 in the neighborhood of @xmath13 at random and move to @xmath14 then we select a vertex @xmath15 in the neighborhood of @xmath14 at random and move to @xmath15 , etc .
the random sequence of vertices selected this way is a random walk on the graph .
ollivier @xcite s definition of the coarse ricci curvature of markov chains on metric space can be naturally defined over such graphs .
the first definition of ricci curvature on graphs was introduced by fan chung and yau in 1996 @xcite . in the course of obtaining a good log - sobolev inequality , they found the following definition of ricci curvature to be useful
: we say that a regular graph @xmath10 has a local @xmath16-frame at a vertex @xmath17 if there exist injective mappings @xmath18 from a neighborhood of @xmath17 into @xmath8 so that \(1 ) @xmath17 is adjacent to @xmath19 for @xmath20 ; \(2 ) @xmath21 if @xmath22 .
the graph @xmath10 is said to be ricci - flat at @xmath17 if there is a local @xmath16-frame in a neighborhood of @xmath17 so that for all @xmath23 , @xmath24 for a more general definition of ricci curvature , in @xcite , lin and yau give a generalization of lower ricci curvature bound in the framework of graphs in term the notation of bakry and emery . in our previous paper @xcite ,
the ricci curvature on graphs is defined based on ollivier s definition of ricci curvature for markov chains on graphs .
it is natural to define a _
ricci - flat graph _ to be a graph where ricci - curvature vanishes on every edge .
this definition does not require a graph to be regular ; which is an advantage over the chung - yau s definition .
the ricci flat graphs defined by chung and yau are not necessarily ricci - flat in the sense of our definition .
however , the ricci curvatures of those graphs are always non - negative . in the last section
, we constructed many `` ricci - flat '' graphs under both definitions .
a well - known bonnet - myers theorem on riemannian geometry said that if a complete riemannian manifolds with ricci curvature bounded below by a positive constant , then it is compact and has a finite fundamental group . in the paper of @xcite ,
we prove the first part result of bonnet - myers theorem on graphs with ricci curvature bounded below by a positive constant . in the paper of @xcite
, forman introduced the ricci curvature on cell complexes and also obtained the myers theorem on the fundamental group of the complexes . in this paper , we classified ricci flat graphs with large girth ( using our definition ) .
[ t1 ] suppose that @xmath10 is a ricci flat graph with girth @xmath25 .
then @xmath10 is one of the following graphs , 1 .
the infinite path , 2 .
cycle @xmath0 with @xmath1 , 3 . the dodecahedral graph , 4 . the petersen graph , 5 . the half - dodecahedral graph
. dodecahedral graph petersen graph half - dodecaheral graph the paper is organized as follows . in the section 2
, we will review the definitions and some facts about ricci curvature on graphs .
theorem [ t1 ] will be proved in section 3 . in the last section
, we will discuss general constructions using cartesian product , strong graph covering , and lattice graphs constructed by the root systems of simple lie algebras . combining these methods
, we constructed many ricci - flat graphs with girth 3 or 4 .
let @xmath7 be a simple graph .
a probability distribution ( over the vertex - set @xmath26 ) is a mapping @xmath27 $ ] satisfying @xmath28 .
suppose that two probability distributions @xmath29 and @xmath30 have finite support .
a coupling between @xmath29 and @xmath30 is a mapping @xmath31 $ ] with finite support so that @xmath32 the transportation distance between two probability distributions @xmath29 and @xmath30 is defined as follows .
@xmath33 where the infimum is taken over all couplings @xmath34 between @xmath29 and @xmath30 .
a function @xmath35 over @xmath10 is @xmath36-lipschitz if @xmath37 for all @xmath38 . by the duality theorem of a linear optimization problem
, the transportation distance can also be written as follows .
@xmath39,\ ] ] where the supremum is taken over all @xmath40-lipschitz function @xmath35 . for any vertex @xmath17 ,
let @xmath41 denote the set of neighbors of @xmath17 .
i.e , @xmath42 let @xmath43 .
for any @xmath44 $ ] and any vertex @xmath17 , the probability measure @xmath45 is defined as @xmath46 for any @xmath38 , we define @xmath47-ricci - curvature @xmath48 to be @xmath49 the ricci curvature @xmath50 is defined as @xmath51 let @xmath52 . then @xmath53 can be viewed as the ball of radius @xmath54 and centered at @xmath17 . under this setting , the ricci curvature capture the approximation of the transportation distance between two balls of radius @xmath54 : @xmath55 this is similar to the ricci - curvature in the differential manifolds @xmath6 of dimension @xmath56 ( see figures [ fig : gball ] and [ fig : mball ] ) .
let @xmath57 and @xmath11 be a unit tangent vector at @xmath17 .
let @xmath58 be a point on the geodesic issuing from @xmath11 , with @xmath59 small enough .
then , the average distance between two @xmath54-balls centered at @xmath17 and @xmath58 in @xmath6 is ( see @xcite ) : @xmath60 2 = 7.5 mm ( 7,2 ) ( 0,1)(1,0)6 ( 1,0)(0,1)2 ( 1,1 ) ( 2,1 ) ( 1,2 ) ( 1,0 ) ( 0,1 ) ( 6,1 ) ( 5,1 ) ( 6,1)(1,2)0.447 ( 6,1)(1,-2)0.447 ( 6.447,1.894 ) ( 6.447,0.106 ) ( 0.75,0,65)@xmath17 ( 5.75,0,65)@xmath58 ( 1.05,1.15)@xmath61 ( 1.1,1.75)@xmath62 ( 1.1,0.25)@xmath62 ( 0.1,0.65)@xmath62 ( 2.1,0.65)@xmath62 ( 6.2,0.65)@xmath61 ( 5.2,0.65)@xmath63 ( 6.547,1.75)@xmath63 ( 6.547,0.25)@xmath63 a graph @xmath10 is called _ ricci - flat _ if @xmath64 for any edge @xmath65 .
[ l1 ] suppose that an edge @xmath66 in a graph @xmath10 is not in any @xmath67 , @xmath68 , or @xmath69
. then @xmath70 .
[ c1 ] suppose that @xmath17 is a leaf - vertex ( i.e. , @xmath71 ) .
let @xmath58 be the only neighbor of @xmath17
. then @xmath72 .
[ l2 ] suppose that an edge @xmath66 in a graph @xmath10 is not in any @xmath67 or @xmath68
. then @xmath73 . *
proof : * define @xmath74 as follows .
@xmath75 since @xmath66 is not in any @xmath67 or @xmath68 , @xmath35 is a lipschitz function .
we have @xmath76\\ & = \alpha + 2(1-\alpha)\left(1-\frac{1}{d_y}\right)-\frac{1-\alpha}{d_x } \\ & = 2 - \alpha - ( 1-\alpha)\left(\frac{1}{d_x}+\frac{2}{d_y}\right).\end{aligned}\ ] ] we have @xmath77 the proof of lemma is finished .
@xmath78 [ l3 ]
suppose that an edge @xmath66 in a graph @xmath10 is not in any @xmath67 or @xmath68 .
without loss of generality , we assume @xmath79 . if @xmath64 , then one of the following statements holds ( see figure [ fig:3 ] ) .
1 . @xmath80 . in this case
, @xmath66 is not in any @xmath69 .
2 . @xmath81 . in this case
, @xmath66 is shared by two @xmath69s .
3 . @xmath82 and @xmath83 . in this case , let @xmath84 be the other neighbor of @xmath17 other than @xmath58 .
let @xmath85 be two neighbors of @xmath58 other than @xmath17 . then @xmath86 .
4 . @xmath82 and @xmath87 . in this case , let @xmath84 be the other neighbor of @xmath17 other than @xmath58 .
let @xmath88 be three neighbors of @xmath58 other than @xmath17 .
then at least two of @xmath88 have distance 2 from @xmath17 .
@xmath81 @xmath82 and @xmath83 @xmath82 and @xmath87 * proof : * by lemma [ l2 ] , we have @xmath89 since @xmath79 , we have @xmath90 . by corollary [ c1 ] , @xmath91 .
we have @xmath82 , or @xmath2 . if @xmath82 , then @xmath92 . if @xmath93 , then @xmath83 .
* case 1 : * @xmath80 and @xmath66 is in a @xmath69 , then @xmath72 .
contradiction !
* case 2 : * @xmath81 and @xmath66 is not shared by two @xmath69 s .
let @xmath94 be two other neighbors of @xmath17 and @xmath95 are two other neighbors of @xmath58 .
without loss of generality , we can assume @xmath96 and @xmath97 .
in this case , we can define a lipschitz function @xmath74 as follows , @xmath98 , @xmath99 , @xmath100 , and @xmath101 . using this function @xmath35
, we can show @xmath102 . contradiction .
* case 3 : * @xmath82 and @xmath83 .
let @xmath84 be the other neighbors of @xmath17 and @xmath95 are two other neighbors of @xmath58 .
if @xmath103 , then a similar calculation shows @xmath72 .
if @xmath104 , then @xmath105 . in both cases , we get contradiction .
* case 4 : * @xmath82 and @xmath87 .
let @xmath84 be the other neighbors of @xmath17 and @xmath106 are two other neighbors of @xmath58 . if @xmath107 ( @xmath108 ) have at least two @xmath2 s , then @xmath102 .
contradiction ! the proof of lemma is finished .
* proof of theorem [ t1 ] : * suppose @xmath109 . by lemma [ l1 ] , for any edge @xmath66
, we have @xmath110 the only integer solution is @xmath80 .
hence @xmath10 is @xmath111-regular graph .
since @xmath10 is connected , it is either an infinite path or a cycle @xmath0 .
now we consider the case @xmath112 . by lemma [ l3 ] ,
every vertex in @xmath10 has degree @xmath111 , @xmath2 , or @xmath3 .
we have the following three claims : * claim a : * @xmath10 has no vertex with degree @xmath3 .
suppose that @xmath58 is a vertex with degree @xmath3 .
by lemma [ l3 ] , all its neighbors must have degree @xmath111 .
let @xmath17 be one of its neighbors and @xmath84 be the other neighbor of @xmath17 other than @xmath58 .
the local structure is shown in figure [ fig:3 ] .
we observe that @xmath84 have degree at least @xmath2 . if @xmath113 , by lemma [ l3 ] , @xmath114 can not be shared by two @xmath69 ; contradiction !
if @xmath115 , then all its neighbors have degree @xmath111 .
now we have two vertices of degree @xmath111 in a @xmath69 that are adjacent to each other .
this is a contradiction to item 1 of lemma [ l3 ] .
thus , we proved claim a. * claim b : * any @xmath69 in @xmath10 can have at most one vertex with degree @xmath111 .
suppose that @xmath116 and @xmath11 are two vertices of degree @xmath111 located in one @xmath69 . by item 1 of lemma [ l3 ] ,
the vertices with degree @xmath111 are not adjacent to each other .
in particular , the other three vertices on this @xmath69 have degree @xmath2 .
let @xmath117 be two adjacent vertices of degree @xmath2 on a @xmath69 .
applying lemma [ l3 ] , there is another @xmath69 passing through @xmath66 .
the other vertices are labeled as shown in figure [ fig:4 ] . if @xmath118 , then @xmath119 is in two @xmath69 . as result , @xmath120 must be connected .
then @xmath121 is in two @xmath69s .
contradiction to item 3 of lemma [ l3 ] .
hence @xmath122 .
similarly , we have @xmath123 .
note that @xmath124 .
note @xmath125 is not in any @xmath69 .
claim b is proved .
* claim c : * suppose that @xmath66 is an edge of @xmath10 with @xmath82 and @xmath83 .
let @xmath126 and @xmath127 be the other two neighbors of @xmath58 besides @xmath17 .
then one of @xmath126 and @xmath127 has degree two while the other one has degree three . by lemma [ l3 ] ,
every vertex of degree @xmath111 is in a @xmath69 .
this @xmath69 must contain @xmath58 and one of @xmath126 and @xmath127 . without loss of generality , assume that @xmath126 is in this @xmath69 . by claim b ,
@xmath126 must have degree @xmath2 . by lemma [ l3 ] ,
@xmath128 is shared by two @xmath69s .
the second @xmath69 must pass through @xmath127 . since @xmath66 is only in one @xmath69
, there is no @xmath69 containing @xmath127 , @xmath58 , and @xmath17 .
thus , @xmath129 .
if @xmath10 has no vertex with degree @xmath111 , then @xmath10 is @xmath2-regular .
every edge is in two @xmath69 s .
@xmath10 can be embedded into a surface @xmath130 .
if @xmath130 is an oriented surface with genus @xmath131 , then euler formula gives @xmath132 here @xmath133 is the number of edges and @xmath35 is the number of faces . since @xmath10 is @xmath2-regular , we have @xmath134 . since every face is a @xmath69
, we have @xmath135 . thus @xmath136 we must @xmath137 .
hence @xmath138 , @xmath139 , and @xmath140 . in this case
, @xmath10 is the dodecahedral graph . if @xmath130 is a non - oriented surface with odd genus
@xmath131 , then euler formula gives @xmath141 here @xmath133 is the number of edges and @xmath35 is the number of faces .
since @xmath10 is @xmath2-regular , we have @xmath134 . since every face is a @xmath69
, we have @xmath135 . thus @xmath142 we must @xmath143 .
hence @xmath144 , @xmath145 , and @xmath146 . in this case
, @xmath10 is the petersen graph .
if @xmath10 contains a vertex @xmath116 with degree @xmath111 , then @xmath116 is in a @xmath69 .
the other four vertices have degree @xmath2 . by lemma 3 ,
this forces four @xmath69s as shown by figure [ fig:5 ] . by claim c
, we have @xmath147 . by claim b , we have @xmath81 .
now the graph continues to expand .
it finally results in the half - dodecahedral graph .
the proof of theorem is finished .
in this section , we will construct many ricci - flat graphs with girth @xmath2 or @xmath3 .
given two graphs @xmath10 and @xmath148 , the cartesian product ( denoted by @xmath149 ) is a graph over the vertex set @xmath150 , where two pairs @xmath151 and @xmath152 are connected if `` @xmath153 and @xmath154 '' or `` @xmath155 and @xmath156 '' . the following theorem is proved in @xcite .
* theorem ( see @xcite ) * _ suppose that @xmath10 is @xmath157-regular and @xmath148 is @xmath158-regular .
then the ricci curvature of @xmath149 is given by @xmath159 here @xmath160 , @xmath161 , @xmath155 , and @xmath154 .
_ another contruction of ricci - flat graphs is using a strong graph covering . a graph
@xmath10 is a _
strong covering graph _ of another graph @xmath148 if there is a surjective map @xmath162 satisfying for any edge @xmath163 , @xmath35 maps the induced graph of @xmath10 on @xmath164 to the induced graph of @xmath148 on @xmath165 bijectively .
note in the traditional definition of covering graph , it only requires @xmath35 maps the induced graph of @xmath10 on @xmath166 to @xmath148 on @xmath167 bijectively .
a strong covering graph @xmath148 is always a cover graph of @xmath148 , not vice versa .
observe that @xmath168 only depends on the induced graph on @xmath169 .
we have the following corollary .
many simple ricci - flat graphs can be constructed using cayley graphs .
suppose that @xmath10 is a group and @xmath130 is a subset of @xmath10 with @xmath170 .
the cayley graph @xmath171 is a graph with the vertex set of all elements in @xmath10 and where a pair @xmath172 forms an edge if @xmath173 .
[ cayleyinf ] let @xmath10 be a torsion - free abelian group and @xmath130 be a finite subset of @xmath10 satisfying @xmath174 .
let @xmath175 be the cayley graph .
suppose that @xmath148 has following property `` for any @xmath176 , @xmath177 , and @xmath178 , the graph distance @xmath179 . ''
then @xmath148 is ricci - flat .
* proof : * we first show @xmath180 for any edge @xmath66 .
we use addition `` + '' for the group operation of the abelian @xmath10 . note that @xmath148 is vertex - transitive ; thus it is regular with degree @xmath181 . for @xmath44
$ ] , we need construct a transportation @xmath34 that moves the probability distribution @xmath53 to @xmath182 as follows @xmath183 observe that @xmath34 simply moves every mass at vertex @xmath116 to @xmath184 . since @xmath185 , @xmath186 is always an edge .
the cost of this transportation is @xmath187 . by equations and
, we have @xmath188 .
thus , @xmath180
. now we show @xmath64
. we will prove it by contradiction .
suppose @xmath72 for some edge @xmath189 .
let @xmath190 and @xmath191 .
we have @xmath177 and @xmath192 .
since @xmath193 , there exists an @xmath194 such that @xmath195 choose an integer @xmath196 . consider a path @xmath197 .
for any @xmath198 , @xmath199 acts on @xmath148 as a graph homomorphism .
thus @xmath200 in particular , for @xmath201 , we have @xmath202 from equation , we get @xmath203 from the triangular inequality of the transportation distance , we have @xmath204 in the last step , we apply @xmath205 by our choice of @xmath206 . contradiction !
@xmath78 [ rootlattice ] let @xmath207 be a root system of type @xmath208 , @xmath209 , @xmath0 , @xmath210 , @xmath211 , @xmath212 , @xmath213 , or @xmath214 ( except @xmath215 ) .
let @xmath10 be the root lattice , which is viewed as the abelian additive group generated by the roots in @xmath207 .
let @xmath216 satisfying @xmath170 . then the cayley graph @xmath171 is ricci - flat . by the triangular inequality of the graph distance
, we have @xmath218 we only to verify @xmath219 .
we will compare the graph distance to the euclidean distance as @xmath10 is embedded as a root lattice .
there are two cases .
* case 1 : * the root system @xmath208 , @xmath210 , @xmath212 , @xmath213 , and @xmath214 have only one type of length , say @xmath220 .
for any @xmath217 and @xmath177 , for any @xmath221 , @xmath222 be the graph distance while @xmath59 be the euclidean distance .
then we have @xmath223 for any @xmath177 and integer @xmath206 , we have @xmath224 by theorem [ cayleyinf ] , @xmath171 is ricci - flat .
* case 2 : * now we consider the root system of type @xmath209 , @xmath0 , and @xmath211 .
those root systems have two types of lengths .
the ratio of the length of the longer root to one of the shorter root is @xmath225 .
in addition , the angle @xmath226 formed between a longer root @xmath116 and a shorter root @xmath11 is either @xmath227 , @xmath228 , or @xmath229 .
we have @xmath230 where @xmath231 .
let @xmath220 denote the length of the shorter root .
then the length of the longer root is @xmath232 .
equation implies for any shorter root @xmath11 and any root @xmath116 ( whether short or long ) : @xmath233 the rank of a root system is the dimension of the underlined euclidean space .
the only rank-1 root system is @xmath240 , which consists of two roots @xmath47 and @xmath241 .
the cayley graph generated by this root system is the infinite path .
the rank-2 root systems @xmath207 are @xmath242 , @xmath243 , @xmath244 , and @xmath215 .
since @xmath215 is excluded from theorem [ rootlattice ] , here we get three lattice graph @xmath245 .
( see figures [ fig : a1a1 ] , [ fig : a2 ] , [ fig : b2 ] , where the root systems are drawn in red . )
we did nt get more lattice graphs by taking a subset @xmath246 .
for example , for the root system @xmath247 and @xmath130 consisting of @xmath2 pairs of roots , we get a `` skewed '' drawing of the graph generated by @xmath242
. however , for the root system @xmath207 of rank more than 2 , we do get more ricciflat lattice graphs by taking proper subset @xmath248 . ; in 0,1 , ... , ; iin 0 , ... , ; ; ( , ) node
( + 1 , ) ( , ) ( + 0.5,+0.866 ) ( + 1 , ) ( + 0.5,+0.866 ) ; ; iin 0,1 , ... , ; ; ( , ) node ; ; ; ( , ) ( + 1 , ) ; ( , ) ( + 0.5,+0.866 ) ; ( , ) ( -0.5,+0.866 ) ; ( , ) ( -1 , ) ; ( , ) ( -0.5,-0.866 ) ; ( , ) ( + 0.5,-0.866 ) ; ( , ) node[fill = red ] ; ; in 0,1 , ... , in 0,1 , ... , ( , ) node ( + 1 , ) ( , ) ( , + 1 ) ( , ) ( + 1,+1 ) ( + 1 , ) ( , + 1 ) ; in 0,1 , ... , ( , + 1 ) node ( + 1,+1 ) ( + 1 , ) node ( + 1,+1 ) ; ( + 1,+1 ) node ; ( 4,4 ) ( 4,5 ) ; ( 4,4 ) ( 5,4 ) ; ( 4,4 ) ( 4,3 ) ; ( 4,4 ) ( 3,4 ) ; ( 4,4 ) ( 5,5 ) ; ( 4,4 ) ( 5,3 ) ; ( 4,4 ) ( 3,5 ) ; ( 4,4 )
( 3,3 ) ; ( 4,4 ) node[fill = red ] ; let @xmath175 be the lattice graph constructed in theorem [ rootlattice ] .
let @xmath249 be a subgroup of @xmath10 satisfying `` any two elements in @xmath249 has a graph - distance at least @xmath250 in @xmath148 . ''
define the quotient graph @xmath251 with vertex set @xmath252 and edge set @xmath253 if @xmath254 .
then @xmath251 is ricci - flat . | a graph is called _ ricci - flat _ if its ricci - curvatures vanish on all edges . here
we use the definition of ricci - cruvature on graphs given in [ lin - lu - yau , tohoku math . ,
2011 ] , which is a variation of [ ollivier , j. funct .
math . ,
2009 ] . in this paper
, we classified all ricci - flat connected graphs with girth at least five : they are the infinite path , cycle @xmath0 ( @xmath1 ) , the dodecahedral graph , the petersen graph , and the half - dodecahedral graph .
we also construct many ricci - flat graphs with girth @xmath2 or @xmath3 by using the root systems of simple lie algebras . |
we stress that the main purpose of this paper is to extend and complete the study of the notion of generalized riesz product associated to the rank one flows on @xmath0 formulated in the same manner as peyrire in @xcite .
the authors in @xcite mentioned that peyrire extended the notion of riesz product to the real line using a class of kernel functions .
furthermore , it is noted by peyrire in his pioneer paper @xcite that an alternative extension of the classical riesz products can be done using the bohr compactification of @xmath0 .
indeed , it is usual that the extension of some notions from the periodic setting to the almost periodic deals with the bohr compactification @xmath2 of @xmath0 ( or more generally , the bohr compactification of local abelian groups ) .
the bohr compactification plays in the almost periodic case the same role played by the torus @xmath3 in the periodic case as the domain of the fast scale variables .
as opposed to the torus , the bohr compactification is often a non - separable compact topological space and this lack of separability is a source of difficulties in trying to adapt the arguments from the periodic context to the almost periodic one .
peyrire @xcite mentioned this difficulty and introduced the riesz products on @xmath0 associated to some class of kernels .
+ notice that here we use the notion of separability used in hirotada - kakutani paper @xcite , that is , the topological group ( or space ) is called separable if it satisfies the second countability axiom of hausdorff which means that it has a countable basis . in that paper , hirotada and kakutani
established that the bohr compactification of a given locally compact abelian group has a countable basis if and only if the union of the spectrum of all almost periodic functions is countable .
+ our analysis here is also motivated by the recent growing interest in the problem of the flat polynomials suggested by a.a .
prikhodko @xcite in the context of @xmath0 .
it turns out that the main idea developed in @xcite does not seem well adapted to the context of our extension of generalized riesz products to the bohr compactification of @xmath0 .
this is due to the fact that the sequence of trigonometric polynomials constructed by a.a .
prikhodko is only locally @xmath4-flat and not @xmath4-flat in the usual sense ( see , for instance , remark [ rem - pri ] ) .
+ the paper is organized as follows . in section 2
we review some standard facts on the almost periodic functions including the ergodicity of the action of @xmath0 by translations on its bohr compactification . in section 3
we define the notion of the generalized riesz products on the bohr compactification of @xmath0 . in section 4
we summarize and extend the relevant material on the kakutani criterion and the bourgain criterion on the singularity of the generalized riesz products introduced in section 3 . in section 5
we state and prove the central limit theorem due to m. kac .
in section 6 we apply the central limit theorem of kac to prove our main result concerning the singularity of a large class of generalized riesz products on @xmath2 .
finally , in the appendix , we consider the problem of the flat polynomials on the bohr compactification of @xmath0 and we add a short note based on hirodata - kakutani paper @xcite on the bohr compactification compared with the stone - ech compactification .
the bohr compactification of @xmath0 is based on the theory of almost periodic functions initiated by h. bohr @xcite in connection with the celebrated @xmath5-function of riemann . in this section
we are going to recall the basic ingredients of this theory .
for the basic facts about almost periodic functions and generalizations of this concept the reader is referred to the classical presentation of bohr @xcite and besicovitch @xcite .
we point out that the theory of almost periodic functions can be extended to more general setting with applications in many context including the non - linear differential equations @xcite .
let @xmath6 be a bounded continuous function and @xmath7 ; we say that @xmath8 is an @xmath9-almost period for @xmath10 if latexmath:[\[\displaystyle \sup_{x \in \r}|f(x+\tau)-f(x)|\stackrel{\rm { def}}{= } @xmath10 is said to be almost periodic if for any @xmath12 the set of @xmath9-almost periods of @xmath10 is relatively dense , i.e. , there is @xmath13 such that any interval with length @xmath14 contains at least one @xmath9-almost period .
the space of all almost periodic functions is denoted by @xmath15 . from the above definition
we easily deduce that @xmath15 is a subspace of the space of bounded continuous functions on @xmath0 .
an important characterization of almost periodic functions is due to bochner and it can be stated as follows [ bochner ] a bounded function @xmath10 is almost periodic function if , and only if , the family of translates @xmath16 is relatively compact in the space bounded continuous functions on @xmath0 endowed with the sup - norm topology .
the proof of theorem [ bochner ] can be found in @xcite , @xcite or @xcite .
furthermore , we have the following fundamental theorem ( see for instance @xcite or @xcite ) .
a bounded continuous function @xmath10 is almost periodic function if , and only if , @xmath10 is uniformly approximated by finite linear combinations of functions in the set @xmath17 .
the space of all continuous functions on @xmath2 is denoted by @xmath18 .
@xmath18 is a commutative @xmath19-algebra under pointwise multiplication and addition .
below , we give an important topological characterization of the bohr compactification of @xmath0 due to gelfand , raikov and chilov .
they obtain this characterization as an application of their theory of commutative banach algebras . [
gelfand]the group @xmath0 , equipped with the usual addition operation , may be embedded as a dense subgroup of a compact abelian group @xmath2 in such way as to make @xmath15 the family of all restrictions functions @xmath20 to @xmath0 of functions @xmath21 .
the operator @xmath22 is an isometric @xmath23-isomorphism of @xmath18 onto @xmath15 .
moreover , the addition operation @xmath24 extends uniquely to the continuous group operation of @xmath2 , @xmath25 the group @xmath2 is called the bohr compactification of @xmath2 . for simplicity of notation , for any @xmath10 in @xmath15
, we use the same letter @xmath10 for its canonical extension to @xmath2 . as a consequence of theorem [ gelfand ] combined with the riesz representation theorem we have
the dual of the space @xmath15 is isometrically isomorphic to the space @xmath26 of all radon measures on the bohr compactification of @xmath0 .
the isomorphism @xmath27 is given by the formula @xmath28 we recall in the following the definition of the characters .
the characters play a important role in the abelian group and , by koopmann observation , in the spectral analysis of dynamical systems .
let @xmath29 be an abelian group and @xmath30 its identity element , then a character of @xmath29 is a complex valued function @xmath31 defined on @xmath29 such that @xmath32 and @xmath33 for all @xmath34 let us recall the following basic fact on compact abelian groups due to peter and weyl .
let @xmath29 be a compact abelian group , with @xmath35 its borel field and @xmath36 its haar measure .
then the set of continuous characters is fundamental both in @xmath37 and in @xmath38 the proof of peter - weyl theorem can be found in @xcite and for the thorough treatment we refer the reader to @xcite .
it is obvious that the continuous characters of @xmath2 are the functions @xmath39 .
in addition the orthogonality of two distinct characters can be checked directly .
indeed , @xmath40 for @xmath41 we denote by @xmath42 the asymptotic mean value of @xmath10 , given by @xmath43 as a consequence of the averaging properties of almost periodic functions we have the following [ mean ] for any @xmath41 we have @xmath44 where @xmath45 is the haar measure in @xmath2 , normalized to be a probability measure , and @xmath46 is the usual lebesgue measure in @xmath0
. moreover we have @xmath47 where @xmath48 is any bounded subset of @xmath0 with @xmath49 , @xmath50 is the lebesgue measure of @xmath48 .
following @xcite , for any @xmath51 , we introduce the notation @xmath52 that is , @xmath53 are the fourier coefficients of @xmath10 relative to orthonormal family @xmath54 ; the inner product is defined by @xmath55 + since the functions @xmath41 correspond to restrictions , @xmath56 , of continuous functions @xmath10 on @xmath2 , a natural question is whether it is possible to define a class of functions @xmath10 which correspond to restrictions " @xmath56 , of functions @xmath57 .
this motivates the following definition .
given @xmath58 the space @xmath59 , of besicovitch s generalized almost periodic functions on @xmath0 , consists of those functions @xmath60 for which there exists a sequence @xmath61 satisfying @xmath62 we denote @xmath59 simply by @xmath63 the space of generalized almost periodic functions @xmath64 was introduced by besicovitch , who also gave them a structural characterization .
we refer to @xcite for more details about functions in @xmath64 .
we immediately have @xmath65 for any @xmath66 and it is easy to see that any @xmath67 has the mean value property , that is , for any bounded measurable @xmath68 , with @xmath69 , we have @xmath70 where @xmath71 .
the space corresponds to @xmath72 in a way similar to the one in which the space @xmath15 corresponds to @xmath18 .
indeed , notice first that the definition of @xmath64 immediately gives that the asymptotic mean value @xmath73 of a function in @xmath64 is well defined ; moreover , any approximating sequence @xmath61 satisfying can be viewed as a cauchy sequence in @xmath72 and , hence , there exists @xmath74 such that @xmath75 converge to @xmath10 in @xmath72 . since @xmath76 is easily seen to be independent of the approximating sequence , in this way we may associate with each @xmath77 a well determined function @xmath74 which we may view as an _
`` extension '' _ of @xmath10 to @xmath2 .
notice that the map @xmath78 is a linear map and that the approximation procedure together with lemma [ mean ] show that @xmath79 as a consequence , the kernel of the map @xmath80 is made by the functions @xmath10 such that the asymptotic mean value of @xmath81 is @xmath82 .
the corresponding quotient space is denoted by @xmath83 .
for @xmath77 we set @xmath84 so that @xmath85 is a semi - norm on @xmath64 .
it is well known that @xmath64 is complete with respect to the semi - norm @xmath85 ( see for instance @xcite ) .
the space @xmath86 is endowed with the scalar product @xmath87 the second equality follows by with @xmath88 , implying that the scalar product is preserved under the map @xmath78 . finally , we define @xmath89 again , @xmath90 is a semi - norm on @xmath91 and the corresponding quotient space is denoted by @xmath92 .
we finish this paragraph by stating the following important fact on the properties of the map @xmath93 .
the proof of it is left to the reader .
the map @xmath93 is an isometric isomorphism between the banach spaces @xmath83 and @xmath72 for any @xmath94 $ ] furthermore , we point out that a suitable extension of lebesgue and fatou s convergence results is obtained in @xcite .
+ gelfand - raikov - chilov theorem [ gelfand ] allows us to define the bohr compactification of @xmath0 .
but , in the harmonic analysis , it is well known that the pontrygain theorem gives an alternative definition .
we briefly recall it here and we refer the reader to the large literature on the subject @xcite and for a deeper discussion on the relation between the almost periodic functions and the bohr compactification of @xmath0 to @xcite , @xcite .
+ let @xmath29 be locally compact abelian group and let @xmath95 be its dual , i.e. , @xmath95 is the set of the characters endowed with the topology inherited from @xmath29 .
let @xmath96 be @xmath95 with the discrete topology .
then @xmath97 is the bohr compactification of @xmath29 .
@xmath98 is a compact group such that @xmath29 is a dense subset of @xmath98 .
we end this section by stating and proving the classical result on the ergodicity of the action of @xmath0 by translation on @xmath2 .
we recall that the action of @xmath0 by translation is defined by @xmath99 where the extended addition is given by theorem [ gelfand ] .
clearly , the family @xmath100 is a flow acting on @xmath2 since @xmath101 and the haar measure is invariant under translation .
the action of @xmath0 on @xmath2 is ergodic , that is , for any borel set @xmath102 which is invariant under the translation action we have @xmath103 , where @xmath104 denotes the normalized haar measure of @xmath105 .
moreover @xmath106 let @xmath102 be an invariant borel set .
we have @xmath107 now , translations are strongly continuous on @xmath108 .
indeed , this is a standard consequence of the density of @xmath18 in @xmath109 , which follows from theorem [ gelfand ] , and the invariance of the haar measure .
therefore , the right - hand side is a continuous function of @xmath110 , and so the identity still holds with @xmath111 .
hence we get , using fubini theorem and the invariance of the haar measure , @xmath112 from which it follows that @xmath113 , as asserted .
it remains to show that @xmath114 .
first we observe that @xmath0 is a borel subset of @xmath2 , since it is the union of a countable family of compact sets , e.g. , the images of the intervals @xmath115,~~ k \in \n$ ] . since @xmath0 is invariant under the translation action we have @xmath116 .
but , for any @xmath117 , @xmath118 is also an invariant borel set and @xmath119 . by the invariance of the haar measure @xmath120 .
hence , we conclude that @xmath114 and the proof is achieved .
riesz products were discovered in 1918 by f. riesz @xcite to answer affirmatively a special question in the theory of fourier series , namely , whether there exists a continuous measure whose fourier coefficients do not vanish at infinity .
roughly speaking , the riesz products are a kind of measures on the circle constructed inductively .
the pioneer riesz product construction gives a concrete example .
since then , the riesz construction proved to be the source of powerful ideas that can be used to produce concrete counterexample of measures with a number of desired properties ( controllability of the convergence of the fourier coefficients being the goal of the original construction ) .
later , a. zygmund extended riesz construction and introduced what it is nowadays called classical riesz products @xcite . in 1975 ,
that riesz products appear as a spectral type of some dynamical systems was shown by f. ledrappier @xcite .
ten years later , m. queffelec @xcite , inspired by the work of coquet - kamae and mandes - france @xcite , showed that the specific generalized riesz products are the right tool to describe the spectrum of the class of dynamical systems arising from the substitution ( see @xcite and the references therein ) . in 1991 ,
b. host , j .- f .
mla , f. parreau in @xcite realized a large class of riesz products as the maximal spectral type of the unitary operator associated with a non - singular dynamical system and a cocycle .
finally , in the more general setting , j. bourgain established the connections between some class of generalized riesz products on the circle and the maximal spectral type of a class of maps called rank one maps @xcite .
one year later , an alternative proof is given by choksi - nadkarni using the host - mla - parreau argument @xcite,@xcite and at the same time a simple proof is obtained by klemes - reinhold @xcite using the standard fourier analysis argument .
+ in @xcite , el abdalaoui - lemaczyk - lesigne and ulcigrai proved that the generalized riesz products analogous to peyrire - riesz products can be realized as a spectral type of some class of rank one flows . + here our aim is to extend the notion of generalized riesz products to the bohr compactification of @xmath0 .
it turns out that such generalization can be done directly .
more precisely , in the study of the spectrum of some class of rank one flows , the following trigonometric polynomials on @xmath0 appears @xmath121 with @xmath122 is a sequence of positive integers greater than @xmath123 and @xmath124 is a sequence of positive real numbers with @xmath125 for any @xmath126 ; the sequence @xmath127 is defined inductively by @xmath128 for simplicity , for any @xmath129 we introduce the following sequence of real numbers @xmath130 and for any real number @xmath131 , we put @xmath132 let @xmath133 be a family of trigonometric polynomials given by .
then the weak limit of the sequence of probability measures on @xmath2 @xmath134 exists and is denoted by @xmath135 let @xmath136 and @xmath137 . by the definition of @xmath138 we have @xmath139 and it is obvious that @xmath140 therefore , @xmath141 is a probability measure on @xmath2 .
in addition , for any @xmath142 , we have @xmath143 hence @xmath144 consequently the limit @xmath145 of the sequence @xmath146 exists . now , since @xmath2 is compact and @xmath147 is a sequence of probability measure on @xmath2 we can extract a subsequence @xmath148 which converge weakly to some probability measure on @xmath2 .
we deduce that the limit of @xmath147 exists in the weak topology and this finishes the proof . the proof above is largely inspired by lemma 2.1 in @xcite ; it gives more , namely , the polynomials @xmath138 can be chosen with positive coefficients and satisfying @xmath149 we mention also that we have @xmath150 for any given sequence of positive integers @xmath151 .
+ we can now formulate our main result whose proof occupies all section 6 .
[ main - here ] let @xmath152 be a sequence of positive integers greater than @xmath123 and @xmath153 be a sequence of positive real numbers .
assume that there exists a sequence of positive integers @xmath154 such that for any positive integer @xmath155 , the numbers @xmath156 are rationally independent .
then the generalized riesz product @xmath157 where @xmath158 is singular with respect to the haar measure on @xmath2 .
the famous dichotomy theorem of kakutani has a rather long history . in his 1948
celebrated paper @xcite , kakutani established a purity law for infinite product measures . precisely , if @xmath159 and @xmath160 are a infinite product measures , where @xmath161 are probability measures such that @xmath162 is absolutely continuous with respect to @xmath163 , for each positive integer @xmath164 , then @xmath165 or @xmath166 according as @xmath167 converges or diverges .
there are a several proofs of kakutani criterion in literature ( see @xcite and the references given there ) . for a proof based on the hellinger integral
we refer the reader to @xcite .
kakutani s theorem was specialized to the gaussian measures on hilbert space with identical correlation operators in @xcite and it was extended to gaussian measures with non - identical correlation operators by segal @xcite , hajek @xcite , feldman @xcite and rozanov @xcite .
later , in 1979 , ritter in @xcite , @xcite generalized kakutani s theorem to a certain non - products measures with application to the classical riesz products . here ,
applying the bourgain methods , we obtain a new extension of kakutani s theorem to the class of generalized riesz products on the bohr compactification of @xmath0 .
indeed , we show that the independence along subsequence suffices to prove the singularity .
nevertheless , our strategy is similar to the strategy of @xcite and it is based on the extension of bourgain methods to the generalized riesz products on the bohr compactification of @xmath0 combined with the central limit tools introduced in @xcite .
moreover , having in mind applications beyond the context of this paper , we shall state and prove a guenais sufficient condition on the @xmath4 flatness of the polynomials which implies the existence of generalized riesz products on @xmath2 with haar component .
we recall that the generalized riesz products @xmath168 is given by @xmath169 where @xmath170 [ bourg - cri]the following are equivalent 1 .
@xmath168 is singular with respect to haar measure . 2 .
@xmath171 _ the proof of @xmath2 version of bourgain criterion is based on the following lemma .
[ cs - b ] the following are equivalent 1 .
@xmath172 2 .
_ @xmath173 _
the proof is a simple application of cauchy - schwarz inequality .
consider @xmath174 and @xmath175 .
denote @xmath176 and @xmath177 its complement in @xmath178 .
let @xmath179 be two real numbers and define a probability measure on @xmath0 by @xmath180}(t)}{b - a}dt , { \textrm{~where~ } } dt { \textrm{~is~the~lebesgue~measure}}.\ ] ] then we have @xmath181 by letting @xmath182 goes to infinity , we get @xmath183 the last equality follows from .
[ proof of theorem [ bourg - cri ] ] assume that ( i ) holds . to prove that @xmath168 is singular , it suffices to show that for any @xmath184 , there is a borel set @xmath185 with @xmath186 and @xmath187 .
let @xmath188 .
fix @xmath189 such that for any @xmath190 , we have @xmath191 . the set
@xmath192 satisfies : @xmath193 and since @xmath194 is open set , it follows from the portmanteau theorem that we have @xmath195 for the converse , given @xmath188 , there exists a continuous function @xmath196 on @xmath2 such that : @xmath197 let @xmath198 . by cauchy - schwarz inequality
, we have @xmath199 since @xmath168 is the weak limit of @xmath200 , we have @xmath201 thus , @xmath202 .
since @xmath203 is arbitrary , we get @xmath204 and this completes the proof . from now on , let @xmath205 be a sequence of positive integers for which @xmath206 are linearly independent over the rationals and let us fix some subsequence @xmath207 of @xmath205 , @xmath208 and @xmath209 with @xmath210 . put @xmath211
we define the degree of any trigonometric polynomial @xmath10 by @xmath212 and we denote @xmath213 . from equations and ,
we have @xmath214 since @xmath215 , telescoping we get @xmath216 in the same spirit as above it is easy to see the following lemma .
the proof of it in the case of the torus is given in @xcite and @xcite .
[ inegalite1 ] with the above notations we have @xmath217 the following proposition is a simple extension of proposition 2.4 in @xcite . [ haar ]
we have @xmath218 the sequence of probability measures @xmath219 converges weakly to the haar measure . from the proposition [ haar ] and lemma [ inegalite1 ]
we deduce the following [ inegalite2]with the above notations we have @xmath220 now , in the following lemma , we state a sufficient condition for the existence of an absolutely continuous component with respect to the haar measure for the given generalized riesz product . in the case of @xmath221 action , the lemma is due to mlanie guenais @xcite , and the proof is similar .
if @xmath222 , then @xmath168 admits an absolutely continuous component .
we denote by @xmath223 the norm in @xmath224 . for all functions @xmath225 and @xmath226 in @xmath227 , by cauchy - schwarz inequality we have @xmath228 and by assumption , @xmath229 hence @xmath230 and the infinite product @xmath231 is convergent : @xmath232 let @xmath233 be a positive integers and take @xmath234 and @xmath235 , + then @xmath236 and @xmath237 ; hence by we have @xmath238 using the fact that @xmath239 , we obtain by induction @xmath240 from combined with we deduce that for large enough @xmath241 @xmath242 hence the sequence @xmath243 does not go to zero in @xmath4-norm .
it follows from bourgain criterion that the generalized riesz product @xmath244 is not purely singular .
as @xmath245 has only countably many zeros , we conclude that @xmath168 admits also an absolutely continuous component with respect to the haar measure .
the kac central limit theorem in the context of the bohr compactification of @xmath0 is stated and proved in @xcite . for sake of completeness
we prove it here using the standard probability arguments .
the real numbers @xmath246 are called rationally independent if they are linearly independent over @xmath221 , i.e. for all @xmath247 @xmath248 [ kac - clt ] let @xmath249 be rationally independent . then , the functions @xmath250 are stochastically independent with respect to the haar measure of the bohr compactification of @xmath0 .
it is sufficient to show that for any positive integer @xmath251 and for a given positive integers @xmath252 , we have @xmath253 write @xmath254 and recall that @xmath255 hence @xmath256 whence @xmath257 because of linear independence , @xmath258 can be zero only if @xmath259 , for any @xmath260 and thus it follows that @xmath261 we conclude that @xmath262 and this finish the proof of the theorem . in the following we recall the classical well known multidimensional central limit theorem in probability theory @xcite stated in the following forms [ mclt ] let @xmath263 be a triangle array of random variables vectors in @xmath264 and put @xmath265 suppose that 1 . the random variables @xmath266 are square integrable .
2 . for each @xmath267 , the @xmath268 , are independents .
3 . for any @xmath269 , @xmath270 4 .
@xmath271 5 . for each @xmath272 , 1 .
@xmath273 2 .
@xmath274 then @xmath275 is a hermitian non - negative definite matrix and the sequence of random vectors @xmath276 converges in distribution to the complex gaussian measure @xmath277 on @xmath264 .
we also need the following important and classical fact from probability theory connected to the notion of the uniform integrability .
the sequence @xmath278 of random variables is said to be uniformly integrable if and only if @xmath279 it is well - known that if @xmath280 for some @xmath9 positive , then @xmath281 are uniformly integrable .
+ let us mention that the convergence in distribution or probability does not in general imply that the moments converge ( even if they exist ) . the useful condition to ensure the convergence of the moments is the uniform integrability .
indeed , we have [ chung]if the sequence of random variables @xmath281 converges in distribution to some random variable @xmath282 and for some @xmath283 , @xmath284 , then for each @xmath285 , @xmath286 for the proof of theorem [ chung ] we refer the reader to @xcite or @xcite . + now let us state and prove the kac central limit theorem .
let @xmath287 be a sequence of rationally independent real numbers .
then the functions @xmath288 , are stochastically independent under the haar measure of the bohr compactification of @xmath0 and converge in distribution to the complex gaussian measure @xmath289 on @xmath290 . by theorem [ kac - clt ] the functions @xmath288 , are stochastically independent under the haar measure of @xmath2 and it is straightforward to verify that the hypotheses of central limit theorem [ mclt ] are satisfied .
we conclude that the sequence @xmath291 converges in distribution to the complex gaussian measure @xmath289 on @xmath290 .
using the analogous lemma of fjer s lemma @xcite combined with the clt methods introduced in @xcite , we shall give a direct proof of the singularity of a large class of generalized riesz products on @xmath2 .
therefore , our strategy is slightly different from the strategy of the proofs given by many authors in the case of the torus @xcite , @xcite , @xcite , @xcite .
indeed , the crucial argument in their proofs is to estimate the following quantity @xmath292 precisely , they showed that the weak limit point of the sequence @xmath293 is bounded below by a positive constant and it is well - known that this implies the singularity of the generalized riesz products ( see for instance @xcite or @xcite ) .
+ let us start our proof by proving the following lemma analogous to fjer s lemma @xcite [ fejer ] with the above notations we have @xmath294 by our assumption the sequence @xmath295 is rationally independent .
hence , by kac theorem [ kac - clt ] , for @xmath296 , the function @xmath226 and @xmath297 are stochastically independent .
this allows us to write @xmath298 which proves the lemma . applying lemma [ fejer ]
, we proceed to construct inductively the sequence @xmath299 such that , for any @xmath300 , we have @xmath301 indeed , by our assumption combined with kac clt [ kac - clt ] , it follows that @xmath302 converges in distribution to the complex gaussian measure @xmath289 on @xmath290 .
but , according to , @xmath302 is uniformly integrable .
hence , from theorem [ chung ] , we get @xmath303 we remind that the density of the standard complex normal distribution @xmath289 is given by @xmath304 now assume that we have already construct @xmath305 and apply with @xmath306 combined with lemma [ fejer ] to get @xmath307 such that @xmath308 put @xmath309
. therefore the inequality holds and by letting @xmath310 we conclude that @xmath311 which yields by bourgain criterion that @xmath168 is singular with respect to haar measure and completes the proof .
the fundamental argument in the proof above is based on lemma [ fejer ] and therefore strongly depended on the assumption that along subsequence the positive real numbers @xmath312 are linearly independent over the rationals .
we argue that in the general case , one may use the methods of bourgain @xcite , klemes - reinhold @xcite , klemes @xcite and el abdalaoui @xcite to establish the singularity of a large class of generalized riesz products on @xmath2 .
in particular , the case when @xmath313 is bounded . in the forthcoming paper , we will show how to extend a classical results from the torus and real line setting to the generalized riesz products on the bohr compactification of @xmath0 .
we are concerned here with the flat polynomials issue in @xmath2 .
first , we recall briefly the relevant fact on the flatness problem in the torus @xmath315
. + the problem of flatness go back to littlewood in his 1968 famous paper @xcite . in that paper , littlewood introduce two class of complex polynomials @xmath316 and @xmath317 where @xmath318 is a positive integer .
the class @xmath316 is a class of those polynomials @xmath319 that are _ unimodular _ , that is , @xmath320 , for @xmath321 .
@xmath317 is the subclass of @xmath316 with real coefficients , i.e. , @xmath322 , for @xmath321 .
the polynomials @xmath225 in @xmath317 are nowadays called littlewood polynomials . by parseval s formula @xmath323
therefore , for all @xmath324 , @xmath325 in @xcite littlewood raised the problem of the existence of a sequence @xmath326 of unimodular polynomials such that @xmath327 such sequence of unimodular polynomials are called ultraflat .
precisely , the usual definition of ultraflatness is given as follows let @xmath328 be a sequence of positive integers and @xmath329 a sequence of positive real numbers tending to @xmath82 , we say that a sequence @xmath330 of unimodular polynomials is @xmath329-ultraflat if @xmath331 the problem of existence of the ultraflat polynomials was solved affirmatively by kahane in his 1980 paper @xcite . precisely , kahane proved that there exists a sequence of @xmath332-ultraflat with @xmath333 but as noted by queffelec and saffari @xcite kahane proof is some kind of miracle .
this is due to the fact that kahane work is inspired by krner paper @xcite and krner paper is based on byrnes paper @xcite .
but , in 1996 j. benedetto and his student hui - chuan wu discovered that theorem 2 in byrnes paper @xcite was erroneous and as a consequence invalidated krner main result .
fortunately kahane proof was independent of theorem 2 in byrnes paper .
+ the problem of the existence of the ultraflat littlewood polynomials is unsettled to this date and as pointed by erdlyi @xcite it is a common belief that there is no ultraflat sequence of littlewood polynomials . as a consequence no long barker sequences exist ( see @xcite and the references given there ) .
+ one more important class of polynomials is the class of polynomials with coefficients @xmath334 called a class of idempotent polynomials for obvious reasons of being convolution idempotents and denoted by @xmath335~~:~~ p(x)=\sum_{0}^{n-1}a_k x^k , a_k \in \{0,1\}\},~~n \in \n^*.\ ] ] a subclass of @xmath336 with constant term @xmath123 is called a class of newman polynomials and denoted by @xmath337 . another extremal open problem in the class of littewood polynomials or in the class of newman polynomials is the problem of @xmath4-flatness . in the same way as the ultraflatness
, the @xmath4-flatness is defined as follows let @xmath328 be a sequence of positive integers and @xmath329 a sequence of positive real numbers tending to @xmath82 , we say that a sequence @xmath330 of polynomials is @xmath338-flat if @xmath339 the problem of the existence of @xmath4-flat sequence of unimodular polynomials was solved by newman in @xcite .
later , m. guenais inspired by newman work constructed @xcite a sequence of @xmath4-flat littlewood polynomials on @xmath340 where @xmath341 is an increasing sequence of prime numbers . as a consequence she established a existence of countable group action with simple spectrum and haar component
subsequently , el abdalaoui and lemaczyk @xcite proved that the sequence of littlewood polynomials constructed by guenais is ultraflat .
before , j. bourgain in his 1993 paper @xcite conjectured that the supremum of the @xmath4 norm by @xmath342 norm over all idempotent polynomials on the circle must be strictly less than one . precisely , he make the following conjecture @xcite @xmath343 using the bourgain ideas , m. guenais @xcite connected the problem of the existence of @xmath4-flat sequence of littlewood polynomials or newman polynomials and the banach problem on whether there exist a dynamical system with simple lebesgue spectrum .
ulam in his book @xcite stated the banach problem in the following form does there exist a square integrable function @xmath344 and a measure preserving transformation @xmath345 , @xmath346 , such that the sequence of functions @xmath347 forms a complete orthogonal set in hilbert space ?
let us formulate the bourgain conjecture in context of bohr compactification of @xmath0 .
for that , let @xmath348 be the subspace of trigonometric polynomials on @xmath0 and @xmath349 be an increasing sequence of real numbers .
put @xmath350 then , we can state the bourgain conjecture in context of bohr compactification of @xmath0 in the following form for any increasing sequence @xmath349 of real numbers , we have @xmath351 where @xmath352 for the sequence @xmath349 of rationally independent numbers , we are able to prove the following proposition .
let @xmath353 be a sequence of real numbers such that for any @xmath354 , the real numbers @xmath355 are rationally independent .
then , we have @xmath356 since , for any @xmath354 , the real numbers @xmath355 are rationally independent , then we can apply the kac clt theorem [ kac - clt ] .
but , the functions @xmath357 are in @xmath109 we deduce that they are uniformly integrable . hence , by theorem [ chung ] , we conclude that @xmath358 this finish the proof of the proposition .
the discussion above allows us to formulate the following questions does there exist a sequence of @xmath4-flat littlewood polynomials or of the @xmath4-flat newman polynomials on @xmath2 ? subsequently does there exist a the ultraflat littlewood or newman polynomials on @xmath2 ? [ rem - pri]we remind here that in @xcite , a. a. prikhodko constructed a sequence of trigonometric polynomials @xmath133 which is locally @xmath4-flat , that is , for any @xmath359 and @xmath360 there exists a positive integer @xmath361 such that for any @xmath362 , we have @xmath363 the polynomials @xmath364 are given by @xmath365 where , @xmath366 @xmath367 is a sequence of positive integers with @xmath368 and @xmath369 , @xmath370 is a sequence of rational numbers which goes to @xmath82 as @xmath318 goes to @xmath371 . + but , this sequence is not @xmath4-flat on the bohr compactification of @xmath0 . indeed
, one may show that we have @xmath372 in addition , it is proved in @xcite that this sequence is not @xmath4-flat on @xmath0 .
let @xmath29 be a locally abelian compact group and denoted by @xmath373 its topology .
let @xmath95 be a group of the characters on @xmath29 , that is , the continuous homomorphism from @xmath29 to the torus @xmath374 put @xmath375 the space of all functions from @xmath95 to @xmath315 equipped with the product topology ( i.e. ; the pointwise convergence topology ) .
therefore , there is a canonical injective homomorphism from @xmath29 to @xmath376 given by @xmath377 @xmath30 is called a dual homomorphism . by abuse of notation
we denote by the same letter @xmath29 the image of @xmath29 under @xmath30 .
therefore , @xmath29 is equipped with the bohr topology denoted by @xmath378 and the topology inherited from @xmath29 . hence @xmath379 . by taking the closure of @xmath29 with respect to the bohr topology we get the bohr compactification of @xmath29 and we denoted it by @xmath380 .
thus , by construction , @xmath380 is compact .
+ let us state the following useful lemma . by the definition of the product topology on @xmath376 , a sequence of functions
@xmath385 converge to some function @xmath10 if and only if , for any @xmath383 , @xmath386 by taking @xmath387 and @xmath388 we get @xmath389 which means that for any @xmath383 , @xmath390 and the proof of the lemma is complete .
we deduce from the lemma the following crucial fact about the separability of the bohr compactification of @xmath29 .
we recall that the topological space is separable if it is contains a countable , dense subset . in the case of @xmath0 ,
the lemma [ cvg ] say that the sequence of real numbers @xmath391 converge in the sense of the bohr topology to @xmath110 if and only if , for any @xmath142 , @xmath392 but since the characters on @xmath0 are lipschitz we deduce easy that the bohr topology is contained in the usual topology and @xmath226 is dense in the usual topology and bohr topology .
+ nevertheless , by the hirodata - kakutani theorem , @xmath0 equipped with the bohr topology does nt have a countable basis and it is often that the bohr topology does nt have a countable basis .
precisely , hirodata - kakutani theorem asserts we recall that the fourier coefficient of @xmath10 on @xmath393 is given by @xmath394 where @xmath36 is the haar measure on @xmath29 and @xmath393 is in the spectrum of @xmath10 if the fourier coefficients of @xmath10 on @xmath31 is not zero the author wishes to express his gratitude to m. lemaczyk , j - p .
thouvenot , e. lesigne , a. bouziad , j - m .
strelcyn and a. a. prikhodko for fruitful discussions on the subject .
it is a pleasure for him to acknowledge the warm hospitality of the poncelet french - russian mathematical laboratory in moscow where a part of this work has been done .
p. billingsley , _ convergence of probability measures , _ second edition .
wiley series in probability and statistics : probability and statistics . a wiley - interscience publication .
john wiley & sons , inc .
, new york , 1999 .
borwein and m. j. mossinghoff , _ barker sequences and flat polynomials , _ ( english summary ) number theory and polynomials , 71 - 88 , london math . soc .
lecture note ser .
, 352 , cambridge univ .
press , cambridge , 2008 .
g. brown and w. moran , _ products of random variables and kakutani s criterion for orthogonality of product measures , _ j. london math .
( 2 ) 10 ( 1975 ) , part 4 , 401 - 405 .
j. bourgain , _ on the spectral type of ornstein class one transformations , _ isr .
j. math . , 84 ( 1993 ) , 53 - 63 . j. s. byrnes,_on polynomials with coefficients of modulus one , _ bull .
london math .
, 9 ( 1977 ) , no . 2 , 171 - 176 . j. r. choksi and m. g. nadkarni , _ the maximal spectral type of rank one transformation , _ can .
math . bull .
, 37 ( 1 ) ( 1994 ) , 29 - 36 .
k. l. chung , _ a course in probability theory , _ third edition . academic press , inc . , san diego , ca , 2001 . j. coquet , t. kamae and m. mends france , _ sur la mesure spectrale de certaines suites arithmtiques , _ ( french ) bull .
france 105 ( 1977 ) , no . 4 , 369 - 384 .
c. corduneanu , _ almost periodic functions , _ with the collaboration of n. gheorghiu and v. barbu . translated from the romanian by gitta bernstein and eugene tomer .
interscience tracts in pure and applied mathematics , no .
[ john wiley & sons ] , new york - london - sydney , 1968 .
d. dacunha - castelle and m. duflo , _ probabilits et statistiques .
tome 2 , _ ( french ) [ probability and statistics .
vol . 2 ] , problmes temps mobile [ movable - time problems ] , collection mathmatiques appliques pour la matrise [ collection of applied mathematics for the master s degree ] , masson , paris , 1983 .
n. dunford and j. t. schwartz , _ linear operators .
i and ii , _ interscience publishers , inc . , new york , 1958 , 1963 .
t. erdlyi , _ how far is an ultraflat sequence of unimodular polynomials from being conjugate - reciprocal ? , _ michigan math
. j. , 49 ( 2001 ) , no . 2 , 259 - 264 .
s. a. grigoryan and v. t. tonev , _ shift - invariant uniform algebras on groups , _ instytut matematyczny polskiej akademii nauk .
monografie matematyczne ( new series ) [ mathematics institute of the polish academy of sciences .
mathematical monographs ( new series ) ] , 68 .
birkhuser verlag , basel , 2006 .
m. kac , _ statistical independence in probability , analysis and number theory .
_ the carus mathematical monographs , no .
12 published by the mathematical association of america . distributed by john wiley and sons , inc .
, new york , ( 1959 ) . v.
p. khavin and n. k. nikolki , _ commutative harmonic analysis . i. general survey .
classical aspects , _
( russian ) translation by d. khavinson and s. v. kislyakov , encyclopaedia of mathematical sciences , 15 .
springer - verlag , berlin , 1991 .
i. e. segal , _ distributions in hilbert space and canonical systems of operators , _ trans .
88 1958 12 - 41 .
s. m. ulam , _ problems in modern mathematics , _ science editions john wiley & sons , inc .
, new york 1964 .
a. zygmund , _ trigonometric series vol .
i _ , second ed .
, cambridge univ . press ,
cambridge , 1959 . | we study a class of generalized riesz products connected to the spectral type of some class of rank one flows on @xmath0 . applying a central limit theorem of kac
, we exhibit a large class of singular generalized riesz products on the bohr compactification of @xmath0 .
moreover , we discuss the problem of the flat polynomials in this setting .
+ @xmath1dedicated to professors jean - paul thouvenot and bernard host .
+ ( 2010 ) primary : 42a05 , 42a55 ; secondary : 11l03 , 42a61 .
+ generalized riesz products , almost periodic functions , bohr compactification , mean value , besicovitch space , kac central limit theorem , kakutani criterion . + |
the mechanism that produces the prompt gamma - ray emission in gamma ray burst ( grbs ) is still uncertain .
the non - thermal character of the spectrum points out towards inverse compton ( ic ) and synchrotron as the two natural candidates .
the latter become , somehow , the standard " process but the former remained always a serious alternative ( * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * and others ) .
the observations of numerous bursts with low energy spectral slopes that are inconsistent with synchrotron @xcite provided additional motivation to consider ic .
recently @xcite have shown further inconsistency with the overall synchrotron model and suggested that synchrotron self - compton ( ssc ) can resolve some of these problems .
the recent observations of a naked eye optical flash from grb080319b @xcite that coincided in time with the prompt @xmath2ray emission provided further motivation to consider ic as the source of the prompt @xmath0-rays . among the different models that appeared so far @xcite , several favor models in which the prompt @xmath0-ray emission is ic of the optical flash and there have been suggestions that this is generic to many grbs . motivated by these ideas we examine , here , the possibility that ic is the source of the prompt soft @xmath0-ray emission in grbs .
this requires a soft component at the ir - uv range that serves as the seed for the ic process .
the flux of these seed photons is constrained by observations ( or upper limits ) of the prompt optical emission .
grb 990123 @xcite and grb 080319b @xcite are rare exceptions with very strong optical emission , @xmath3 and @xmath4 mag respectively .
however most bursts are much dimer optically with observations or upper limits around 14 mag @xcite .
this should be compared with fluxes of @xmath5 in soft gamma rays for a modest burst .
what is important , in this work is the flux ratio @xmath6 which is typically larger than 0.1 during the peak soft gamma emission @xcite .
the basic problem of the ic model can be explained simply .
if the low energy seed emission is in the optical , while the observed soft @xmath0-ray spectrum is the first ic component , then second ic scatterings would create a tev component .
upper limits or observations of the prompt optical signal show that the @xmath7 parameter , i.e. the ratio between the energy in the first ic component to that in the low energy seed photons is very large , typically greater than thousands .
theory would then show that the second ic component in the tev range would carry an even larger amount of energy , again by a factor of @xmath8 , producing an energy crisis " for this model , and possibly violating upper limits from egret ( energetic gamma - ray experiment telescope ) @xcite .
this problem is generic and it does not depend on the specific details of the overall model .
the above analysis is oversimplified and two factors may alleviate the energy catastrophe .
first , the frequency of the seed photons may differ from that where upper limits exist , allowing larger seed flux and reducing the lower limits on @xmath7 .
second , the klein - nishina ( kn ) suppression , which does not affect the first scattering , may affect the second , resulting in a lower @xmath7 parameter for the second scattering than the first one . in this article
, we explore the parameter space to see weather there exist a regime where a combination of these two factors allows for less energy in the second ic component ( typically in the tev range ) than in the @xmath0-rays .
we find that possible solutions are limited to a very small region in the parameters space in which the seed photons are in the ir , the bulk lorentz factor is very low @xmath9 ) and the electrons lorentz factor is very large ( @xmath10 .
however , this solution implies a healthy emission in the ir , while self absorption limits it .
therefore , when taking self - absorption into account , this solution is ruled out as well . a second possible solution exists if the seed photons are in the uv .
this solution requires a very low electrons lorentz factor @xmath11 , and a seed photon flux that carries comparable energy to the observed prompt @xmath0-rays .
furthermore , prompt x - ray observations limit the high energy tail of the uv component and practically rule out this model .
we take the lorentz factor of the electrons and the bulk lorentz factor as free parameters and we estimate what is the second ic fluence ( at tev or multi gev ) given the observed prompt gamma - ray flux and the limits on the prompt optical band .
most of our analysis is insensitive to the size of the source , which appears only in the final section when we estimate the self absorption flux . in our numerical examples we use very conservative parameters .
for example we use r magnitude of 11.2 as an upper limit on the optical flux , while many limits are much stronger and the @xmath0-ray flux we take , @xmath12 , is quite modest .
similarly we use conservative rather rather than canonical " values for the spectral slopes .
consider electrons that move with a bulk lorentz factor @xmath13 while in the bulk ( or fluid ) rest frame they have a typical lorentz factor @xmath14 in a random direction .
we examine ic scattering of seed photons with a peak frequency @xmath15 and a peak flux @xmath16 ( both measured at the observer s rest frame ) .
we assume that the seed photons are roughly isotropic in the fluid s frame .
this would be the case if the seed photons are produced by synchrotron radiation in the bulk , or any other mechanism local to the moving fluid .
we will consider external ic , in which the seed photons are produced by an external source elsewhere .
for simplicity we assume that all the photons have the same energy and all the electrons have the same lorentz factor .
the energy and flux of the scattered photons are : @xmath17 and @xmath18 where @xmath19 and @xmath20 are the compton parameter and the optical depth in the thomson scattering regime .
note that the unknown optical depth , @xmath20 , is introduced here in the definition of @xmath7 but it is not used elsewhere in the paper .
our analysis is independent of this unknown factor .
the factor , @xmath21 corresponds to the correction that arises if the scattering is in the kn region : @xmath22 the expression given in eq .
[ fkn ] is approximate .
again this approximation is sufficient for our purpose .
we consider now the possibility that the prompt gamma - rays arise due to ic scattering of a lower energy component .
we use now the observed gamma - ray flux , @xmath23 , and its peak energy , @xmath24 and the upper limits ( or detections ) of prompt optical emission , @xmath25 at @xmath26 to set limits on the ic process .
the peak flux of the low energy component , @xmath27 , is at @xmath28 which is not necessarily at the observed frequency @xmath26 . given an upper limit on the prompt optical flux , @xmath29 at @xmath30 ( or on the flux at any other frequency ) , we can set a limit on @xmath27 if the optical frequency is in the same spectral region as @xmath28 , the peak frequency of the lower spectral component of slope @xmath31 : @xmath32 the equality here and elsewhere holds when @xmath29 corresponds to a detection and an inequality corresponds to an upper limit .
there are two possibilities , either @xmath33 which we call the uv solution " or @xmath34 which we call the ir solution " .
since by definition , the seed photon energy peaks at @xmath28 , we must have @xmath35 in the uv solution and @xmath36 in the ir solution .
moreover , since the spectrum around @xmath28 is up - scattered to create the familiar band spectrum @xcite around @xmath24 , we can expect @xmath37 for the ir solution and @xmath38 for the uv solution . as the first ic scattering results in soft @xmath0-rays , it is clearly away from the kn regime and we obtain , using eqs .
( [ gamma],[flux],[limit ] ) a limit on the compton parameter @xmath39 , in the first compton scattering : @xmath40 using this limit we turn now to the second order ic component .
this process will produce photons in the gev - tev range .
as the scattered photon is energetic , it might be in the kn regime and we have : @xmath41 and @xmath42 @xmath43 is the ratio of energy emitted in the high energy ( tev ) band and in lower energy gamma - rays ( see fig .
1 ) . as a conservative numerical example we will use the following typical parameters : @xmath44 , @xmath45 , leading to a ratio of @xmath46 . this optical flux corresponds to r magnitude 11.2 , which is a very conservative upper limit to the prompt optical emission of most grbs while the prompt gamma - ray flux is moderate .
we use @xmath47hz and @xmath48 [ both energies are larger by a factor of @xmath49 than the observed frequencies , r band and 250kev ] .
thus @xmath50 .
we will use @xmath51 and @xmath52 for the canonical values of @xmath53 and @xmath26 .
we find : @xmath54\ ] ] and @xmath55 \label{eh}\ ] ] the essence of the ic problem is the very large value of @xmath43 , which arises from the fact that the energy released in prompt gamma - rays is at least a factor of 1500 larger than the energy released in prompt optical emission ( see eq .
the large values of @xmath43 implies that the energy emitted in the tev range exceeds the observed soft @xmath0-rays by several orders of magnitude .
as a function of @xmath56 for @xmath57 ( from top to bottom ) and @xmath58 ( from top to bottom ) for @xmath59 and @xmath60 for @xmath61 .
parameters used in this figure are : @xmath62 , @xmath47hz and @xmath63 .
the breaks in the lines appear at @xmath64 when we change from negative to positive @xmath31 and at the frequency , that depends on @xmath65 , where the kn correction begins .
[ fig : y ] ] fig .
[ fig : y ] depicts @xmath43 as a function of @xmath28 for different values of @xmath65 and for different spectral indices .
@xmath43 peaks when @xmath66 .
this is expected as in this case the observed limits on the lower energy flux are strongest .
if @xmath28 increases or decreases more energy can be hidden " in the lower energy component and the corresponding @xmath39 and @xmath43 will be smaller . because of a similar reason @xmath43 decreases when @xmath67 increases .
we find two possible regimes for ic solutions that are not over producing a high energy ( tev ) component .
the uv solution requires @xmath68 and @xmath69 .
the electrons lorentz factor in the uv solution satisfies @xmath70 .
the second compton scattering is not in the kn regime since @xmath71 and correspondingly @xmath21 is small .
since kn suppression is negligible @xmath72 and the total energy , given by @xmath73 , is at least @xmath74 .
uv solutions with @xmath75 are therefore also wasteful as they require a large ( @xmath76 ) low energy component .
a second problem arises , for this solution , with the spectral shape .
the observed low energy spectral index ( in the x - ray band ) is typically close to zero , while this solution requires a steeply rising flux from @xmath26 to @xmath28 .
note that in fig .
[ fig : y ] we show conservatively curves for @xmath77 even though the `` canonical '' value is 0 .
moreover , unless there is a pair loading ( that is if there is one electron per proton ) , then the low @xmath78 required for the uv solution implies that the protons carry significantly more energy than the electrons by at least a factor of @xmath79 .
thus this solution is a very inefficient .
the analysis above is based on the optical limits but for the modest values of @xmath80 needed for the uv solution , @xmath28 , the peak flux frequency of the seed photons becomes large ( eq . [ gamma ] ) and
@xmath27 is now limited by prompt soft x - ray observations in additional to the optical limits . for the discussion below
, we use @xmath81 and @xmath82 as the low energy and high energy spectral indices , respectively . as stated before ,
the canonical values are @xmath83 and @xmath84 @xcite .
one can estimate the x - ray flux at @xmath85 kev directly from the observations at this energy or using the flux at @xmath86kev and the low energy spectral slope @xmath81 .
recalling that the ic does not change the spectral slope , we use the same indices both around @xmath24 and around @xmath28 .
threfore : @xmath87 using eq .
[ flux ] we obtain : @xmath88 since the uv solution is not in the kn regime we have @xmath89 . if we take the typical spectral indices below and above @xmath24 to be @xmath83 and @xmath84 respectively @xcite , and we impose the condition @xmath90 ( where the total energy required is minimized to @xmath91 ) , we find that @xmath92 or @xmath93 - thus the whole uv regime is ruled out .
this condition depends strongly on the spectral indices : @xmath81 and @xmath82 .
clearly if @xmath82 is smaller ( a steeper drop on the high energy side ) @xmath28 can be larger and y is smaller is large for grb 080319b , which might be an ic burst with a uv solution . ] .
the limits are depicted in fig .
[ fig : uv ] for several values of the spectral indices .
one can see that the available x - ray data rules out the uv solution for most of the phase space . as a function of @xmath94 for @xmath95 ( blue , red and green ) and @xmath96 ( solid , dotted and dashed ) for @xmath97kev and @xmath98 .
the corresponding @xmath80 range is from 158 at @xmath99 to 5 at @xmath100 .
[ fig : uv ] ] the ir solution holds for @xmath101hz and @xmath102 .
it requires a large electron s lorentz factor @xmath103 and a relatively low bulk lorentz factor @xmath104 .
the solution is deep in the kn regime and the kn suppression is very significant .
it allows for a large amplification between the ir and the soft @xmath0-rays and no amplification between the low energy gamma and the tev emission .
a solution is possible in a small region of the parameter space if the high energy spectrum is steep ( @xmath105 ) - this increases the allowed flux at @xmath28 . such a spectrum above the peak frequency , though steeper than the canonical @xmath106 , is not rare in the observations of prompt @xmath0-ray bursts . . for three spectral indexes @xmath58 ( from bottom to top ) for @xmath59 and @xmath60 for @xmath61(from bottom to top ) .
parameters used are : @xmath62 , @xmath47hz and @xmath98 .
the @xmath80 axis corresponds to values of @xmath28 ranging from @xmath107hz=0.2kev for @xmath108 to @xmath109hz for @xmath110 .
[ fig : gamma ] ] to demonstrate the severity of the constraint we plot ( fig .
[ fig : gamma ] ) the allowed region " in the ( @xmath111 ) phase space for which @xmath112 .
it is remarkable to note that the expected parameter region for internal shocks @xmath113 , @xmath114 is deep inside the ruled out region .
the parameter expected for external shocks @xmath115 and @xmath114 are allowed with seed photon wavelength in the cm range . however , as we show in 4 , self absorption limits the amount of energy in such low frequency seed photons , ruling out this solution . for low values of @xmath80
the whole @xmath65 range is seemingly allowed
. however , this only happens at @xmath116 for @xmath117 respectively , and therefore conflicts with the soft x - ray observations .
in cases when @xmath118 most of the electrons energy is emitted as very high energy ( tev ) gamma - rays .
when the scattering is in the kn regime , that is if eq .
[ fkn ] holds , the scattered photon , that has an energy of almost @xmath119 ( in the fluid s rest frame ) and can therefore produce a pair when it encounters a typical low energy @xmath0-ray photon with energy @xmath120 ( in this frame ) . more specifically , for a head on collision between a photon with energy @xmath121 and an electron with a lorentz factor @xmath80 , the energies of the electron and the photon after the collision are : @xmath122 and @xmath123 the resulting photon has now enough energy to collide with a photon with energy @xmath124 and produce two electrons with a lorentz factor : @xmath125 as the optical depth for pair creation is huge all the scattered photons will create pairs with typical energy of @xmath126 . as a result we will have colder electrons and positrons with a ratio 2:1 in higher ( @xmath127 ) and lower ( @xmath128 ) energies .
these colder electrons and pairs will inverse compton scatter more photons and will produce a second generation of cooler pairs with @xmath129 .
the process will continue until pair creation will stop .
this will happen when @xmath130 .
this situation was considered numerically by @xcite , and most recently @xcite .
if the physical conditions , like magnetic field and total number of particles are fixed @xmath28 , @xmath24 and @xmath131 as well as the corresponding fluxes will vary as a result of the changing electron energy distribution due to the created pairs .
these variations will be very significant because of the strong dependence ( 2nd and 4th powers ) of the first two on @xmath80 .
the dynamical evolution of such a system is interesting by itself .
however , we are interested , here , in the final steady state in which @xmath24 and @xmath23 are fixed as the observed quantities . in this case
we can search for the physical parameters that exists in such a steady state .
we can express @xmath65 in terms of @xmath78 and @xmath24 using the pair creation threshold criteria ( eq . [ fkn ] ) and we can express @xmath28 in terms of @xmath53 and @xmath80 ( using eq .
[ gamma ] ) . given these expressions
we can estimate the steady state @xmath43 as a function of @xmath80 . as a function of @xmath80 for a situation in which pair avalanche leads to @xmath132 . shown are curves for three different values of @xmath133 ( from top to bottom ) for @xmath59 and @xmath60 for @xmath61 .
the parameters used are : @xmath62 , @xmath47hz and @xmath63 .
the @xmath80 axis corresponds to values of @xmath28 ranging from @xmath134hz=0.2kev for @xmath108 to @xmath135hz for @xmath110 .
[ fig : pairs ] ] fig .
[ fig : pairs ] depicts the resulting @xmath43 values as a function of @xmath80 for different values of @xmath31 .
the uv solution for @xmath68 and with rather low values of @xmath80 and @xmath65 is possible .
however this solutions suffers from the problems discussed earlier .
it seems that if we impose the pair creation threshold conditions the ir solution is ruled out with very high @xmath43 values ( for any reasonable @xmath31 ) .
however , as discussed earlier , there is a region in the parameter space for the ir solution for which @xmath136 . in this case only a small fraction of the energy goes into the high energy photons and it is possible ( depending on time scales ) that most of the electrons cool down rapidly before pair avalanche arises .
a natural source of the seed photons is synchrotron emission by the same electrons that produce the ic emission . assuming that this source is indeed synchrotron we can proceed and estimate the strength of the magnetic filed and the size of the emitting region . we can then check if these values are reasonable within given grb models
however , we choose a more general approach and ask whether the large seed flux needed is limited by self absorption . to the needed seed flux as function of @xmath56 for three values of @xmath137 ( from top to bottom ) and
three different values of @xmath31 : @xmath60 ( from top to bottom ) for @xmath61 .
parameters used in this figure are : @xmath62 , @xmath47hz and @xmath63 .
[ fig : fsa ] ] self absorption limits the flux at @xmath28 to be below the black body flux , @xmath138 , for a local temperature @xmath139 : @xmath140 where @xmath141 is the radius of the source and @xmath142 is the luminosity distance for z=1 . in the following examples we use conservatively @xmath143 cm as the emission radius of the prompt emission .
[ fig : fsa ] depicts a comparison of this limiting flux , @xmath138 with the needed flux @xmath144 . for @xmath145 , @xmath146 .
this implies that the electrons that produce the ic emission can not produce the lower energy seed photons .
the ratio @xmath147 decreases with increasing @xmath65 .
it also decreases when @xmath148 increases .
so in most of the region where @xmath112 ( see fig . [
fig : y ] ) the seed flux is insufficient !
the combined limits on the @xmath149 parameter space from self absorption with @xmath150 are shown in fig .
[ fig : fsa2 ] .
only an extremely small region around @xmath151 ( corresponding to @xmath152hz ) and @xmath153 is allowed .
this used a conservative over estimate for the emission radius @xmath154 .
if we use the variability time scale @xmath155sec , with @xmath156 and the low values of @xmath65 obtained , @xmath141 will be much smaller , invalidating even this solution .
the self absorption limit rules out also the region in the parameter space that corresponds to external shocks .
this solution requires a very low seed frequency which would have implied a very small self - absorption limit .
parameter space .
the limit on the left ( decreasing curve ) corresponds to the condition @xmath157 .
the limit on the right ( increasing curve ) corresponds to @xmath150
. also marked is @xmath158 , which is considered as a minimal value for the bulk lorentz factor to resolve the compactness problem @xcite .
the limits are shown for @xmath159 .
( on the right side around @xmath160 shown are the corresponding curves for @xmath161 . ) .
the @xmath80 range from 1000 to 5000 corresponds to @xmath162hz to @xmath163hz .
parameters used in this figure are : @xmath62 , @xmath47hz and @xmath63 . for @xmath164 an extremely small region around @xmath151 ( corresponding to @xmath165hz ) and @xmath153
is allowed .
[ fig : fsa2 ] ]
for a typical grb , ic has to amplify the total energy of a low energy seed photon flux by a factor of @xmath166 to produce the observed prompt gamma - ray flux .
the same relativistic electrons will , however , continue and upscatter the gamma - ray flux to very high energies in the tev range . in many cases
this second generation ic will be in the klein - nishina regime ( that is the photon s energy will be larger than the electrons rest mass , in the electron s rest frame ) .
this will suppress somewhat the efficiency of conversion of gamma - rays to very high energy gamma - rays , however it wo nt stop it altogether .
our analysis focused on the case that the low energy seed photons are produced within the moving region that includes the ic scattering relativistic electrons
. such will be the case , for example , in synchrotron self - compton .
related considerations , that will be published elsewhere , apply when the seed photons are external and constrain ic processes in this case as well .
the analysis is also limited to the important implicit assumption that the emitting region is homogenous .
it is possible that very strong inhomogeneities could change this picture .
we have shown that , under quite general conservative assumptions , if ic produces the prompt mev photons , then a second scattering will over produce a very high ( gev - tev ) prompt component that will carry significantly more energy than the prompt gamma - rays themselves . on the theoretical front such a component will cause an energy crisis " for most current progenitor models . from an observational point of view , this component is possibly already ruled out by egret upper limits@xcite .
fermi should very soon put much stronger limit to ( or verify ) this possibility .
for example , a burst with isotropic energy @xmath167erg , locating at @xmath168 , would produce @xmath169 photons detected by fermi .
one may not over produce a high energy component if the seed photons are in the uv regime .
however , in this case , the needed seed photon energy should be equal or larger than the observed prompt gamma - ray energy .
downwards extrapolation of the x - ray observations put strong limits on this solution and probably rule it out as well .
moreover this uv solution requires pair loading to be efficient .
we thank r. mochkovitch , e. nakar , j. poutanen , p. kumar and x. f. wu for helpful discussions .
the research was partially supported by the isf center for excellence in high energy astrophysics , by an israel - france collaboration grant , by an irg grant , by an atp grant and by the schwartzman chair .
akerlof c. , et al .
, 1999 , nature , 398 , 400 ando s. , nakar e. , sari r. , 2008 , apj , 689 , 1150 ( arxiv:0807.0012 ) band , d. , et al . , 1993 , astrophys .
j. 413 , 281 bloom j. s. , et al . , 2008 ,
apj in press ( arxiv:0803.3215 ) cohen e. , katz j. i. , piran t. , sari r. , preece r. d. , band d. l. , 1997 , apj , 488 , 330 coppi p. s. , 1992 , mnras , 258 , 657 delia v. , et al .
, 2008 , apjl submitted ( arxiv:0804.2141 ) fan y. z. , piran t. , 2008 , front .
phys . china .
, 3 , 306 ( arxiv:0805.2221 ) ghisellini , g. , lazzati , d. , celotti , a. , & rees , m. j. 2000 , mnras , 316 , l45 gonzalez sanchez gonzalez sanchez , m. m. 2005 , ph.d .
thesis kobayashi s. , zhang b. , mszros p. , burrows d. , 2007 , apj , 655 , 391 kumar p. , mcmahon e. , 2008 , mnras , 384 , 33 kumar p. , panaitescu a. , 2008 , mnras , 391 , l19 ( arxiv:0805.0144 ) lithwick y. , sari r. , 2001 , apj , 555 , 540 peer a. , waxman e. , 2005 , apj , 628 , 857 preece r. d. , briggs m. s. , mallozzi r. s. , pendleton g. n. , paciesas w. s. , band d. l. , 1998 , apj , 506 , l23 preece r. d. , briggs m. s. , giblin t. w. , mallozzi r. s. , pendleton g. n. , paciesas w. s. , band d. l. , 2002 , apj , 581 , 1248 racusin j. l. , et al .
, 2008 , nature , 455 , 183 ( arxiv:0805.1557 ) shemi a. , 1994 , mnras , 269 , 1112 shaviv nir j. , dar a. , 1995 , mnras , 277 , 287 sari r. , narayan r. , piran t. , 1996 , apj , 473 , 204 sari r. , piran t. , 1997 , mnras , 287 , 110 stern b. e. , begelman m. c. , sikora m. , svensson r. , 1995 , mnras , 272 , 291 stern b. e. , poutanen j. , 2004 , mnras , 352 , l35 vurm i. , poutanen j. , 2008 , arxiv:0807.2540 waxman e. , 1997 , apj , 485 , l5 yost s. a. , et al
. , 2007 , apj , 669 , 1107 yu y. w. , wang x. y. , dai z. g. , 2008 , apj in press ( arxiv:0806.2010 ) zou y. c. , piran t. , sari r. , 2008 , apjl in press ( arxiv:0812.0318 ) | inverse compton ( ic ) scattering is one of two viable mechanisms that can produce the prompt non - thermal soft gamma - ray emission in gamma - ray bursts .
ic requires low energy seed photons and a population of relativistic electrons that upscatter them .
the same electrons will upscatter the gamma - ray photons to even higher energies in the tev range . using the current upper limits on the prompt optical emission
we show that under general conservative assumption the ic mechanism suffers from an energy crisis " .
namely , ic will over - produce a very high energy component that would carry much more energy than the observed prompt gamma - rays , or alternatively it will require a low energy seed that is more energetic than the prompt @xmath0-rays .
our analysis is general and it makes no assumptions on the specific mechanism that produces the relativistic electrons population . gamma rays : bursts@xmath1ism : jets and outflows radiation mechanisms : nonthermal |
for years , the protostellar collapse stage had been an impenetrable mystery , until walker et al .
( 1986 ) observed line profiles indicative of collapse in iras 16293 - 2422 .
menten et al .
( 1987 ) disputed this interpretation and claimed that the asymmetric line profiles of walker et al . were caused by rotation , but zhou ( 1995 ) later modeled iras 16293 - 2422 as collapse with rotation .
zhou et al . ( 1993 ) observed b335 , a slowly rotating source , and modeled its blue - peaked profiles as inside - out collapse ( shu 1977 ) .
andr et al .
( 1993 ) extended the tripartite taxonomy of young stellar objects to include class 0 objects ( very embedded sources , such as b335 and iras 16293 ) .
andr and montmerle ( 1994 ) found that the class 0 sources were more embedded than class i sources and inferred that they had not yet accreted most of their mass .
spectral line surveys of class 0 sources ( gregersen et al . 1997 , mardones et al .
1997 ) , found nearly a third to half of class 0 objects displayed asymmetries in optically thick lines like those seen in b335 and iras 16293 - 2422 .
however , the earliest phase of the collapse process , the transition between the quasi - static core formation and the beginning of infall onto a central object , is poorly understood .
beichman et al . ( 1986 ) examined the iras data for 95 cloud cores previously surveyed by myers et al .
( 1983 ) , myers and benson ( 1983 ) and benson ( 1983 ) in @xmath2co , c@xmath3o and nh@xmath4 and found that half had iras sources , which they deduced as arising from protostars .
ward - thompson et al .
( 1994 ) observed 17 cores from beichman et al . that have no iras sources .
they detected 12 of these cores in the submillimeter and used maps to study the density profiles of 5 cores .
since these objects lacked iras sources , it is believed that protostars have not yet formed . from statistical arguments about the lifetime of these cores and the fact that the observed density profiles are similar to those predicted by ambipolar diffusion models , ward - thompson et al . identified these starless cores as in the ambipolar diffusion phase and pre - protostellar .
this stage precedes the class 0 phase and is sometimes referred to as the pre - protostellar core stage .
we observed the objects surveyed by ward - thompson et al . using the hco@xmath0 @xmath1 line , a line that readily displays an asymmetry indicative of protostellar collapse to see if an early collapse phase could be found .
lee et al .
( 1999 ) have completed a similar survey using cs and lines .
we observed the 17 starless cores listed in table 1 in the hco@xmath0 @xmath1 line with the 10.4-m telescope of the caltech submillimeter observatory ( cso ) at mauna kea , hawaii in march 1995 , december 1995 , june 1996 , july 1998 , december 1998 and july 1999 .
we used an sis receiver ( kooi et al .
1992 ) with an acousto - optic spectrometer with 1024 channels and a bandwidth of 49.5 mhz as the backend .
the frequency resolution ranged from slightly less than 3 channels , 0.15 at 267 ghz , for the 1995 observations to closer to 2 channels , 0.12 at 267 ghz , for the 1998 observations .
the antenna temperature , @xmath5 , was obtained from chopper - wheel calibration .
information about the observed lines is listed in table 2 .
planets were used as calibration sources for calculating the main beam efficiency .
data from separate runs were resampled to the resolution of the run with the worst frequency resolution before averaging .
a linear baseline was removed before scans were averaged .
line properties are listed in table 3 . for lines without two clearly distinguished peaks , @xmath5 , the peak temperature , @xmath6 , the line centroid , and @xmath7v , the line width , were found by fitting a single gaussian to the line profile . for lines with two clearly distinguished peaks , we list two values of @xmath8 and @xmath6 , one for each peak , and we give one value for the line width , which is the full width across the spectrum at the temperature where the weaker peak falls to half power . we observed 17 sources in this survey .
all of the sources were observed in the hco@xmath0 @xmath1 line .
six sources were also observed in the h@xmath2co@xmath0 @xmath1 line .
six sources showed a blue asymmetry in the hco@xmath0 @xmath1 line ( figure 1 ) .
eight sources showed symmetric lines ( figure 2 ) and three sources were not detected .
the spectra in figures 1 and 2 are from the central position except that of l1689b which is from ( @xmath915 , 15 ) , which we chose because it was the strongest position .
wang ( 1994 ) observed absorption in the h@xmath10co 6 cm line against the cosmic microwave background radiation similar to that observed in b335 .
kuiper et al . (
1996 ) posited that this core is quasi - static or slowly contracting and that the outer envelope is growing .
they also concluded that this core could collapse within the next 5 @xmath11 10@xmath12 years .
wolkovitch et al .
( 1997 ) determined that this core was extremely quiescent based on its narrow ccs line widths .
the hco@xmath0 @xmath1 line ( figure 2 ) shows no asymmetry and is at the same velocity as the n@xmath10h@xmath0 and c@xmath4h@xmath10 lines observed by benson et al.(1998 ) .
the hco@xmath0 @xmath1 spectrum ( figure 2 ) shows one component at 7.5 with a possible second component at 9 .
the hco@xmath0 @xmath1 line ( figure 2 ) is too weak to detect an asymmetry .
we observe a slight blue shoulder in hco@xmath0 @xmath1 ( figure 2 ) .
caselli et al . (
1995 ) observed the hyperfine components of the n@xmath10h@xmath0 @xmath13 line and found that a single excitation temperature could not fit the spectra , an anomaly usually seen in starless cores
. the hco@xmath0 @xmath1 line ( figure 2 ) is symmetric and is at the same velocity as the n@xmath10h@xmath0 and c@xmath4h@xmath10 lines ( benson et al .
1998 ) .
myers et al .
( 1996 ) modeled the h@xmath10co @xmath14 line as arising from infall .
tafalla et al .
( 1998 ) found that their cs @xmath15 observations could be modeled as arising from inward motions , but those inward motions are not consistent with the predictions of the shu ( 1977 ) inside - out collapse model .
williams et al .
( 1999 ) observed similar infall speeds in n@xmath10h@xmath0 @xmath13 on scales of 10 to those observed by tafalla et al .
ohashi et al . ( 1999 ) mapped this core in ccs @xmath16 and observed both infall and rotational motion .
ciolek and basu ( 2000 ) have modeled the observations of tafalla et al . and williams et al . in the context of ambipolar diffusion .
the hco@xmath0 @xmath1 spectrum is blue - peaked with the h@xmath2co@xmath0 @xmath1 line peaking in the dip ( figure 1 ) .
the n@xmath10h@xmath0 peaks between the h@xmath2co@xmath0 line and the hco@xmath0 peak while the c@xmath4h@xmath10 peaks on the h@xmath2co@xmath0 line ( benson et al . ) .
the hco@xmath0 @xmath1 line is symmetric ( figure 2 ) .
the n@xmath10h@xmath0 line of benson et al . has the same peak velocity as the hco@xmath0 line .
there is a hint of a blue shift relative to the velocity , but not enough to warrant inclusion as an infall candidate .
the hco@xmath0 @xmath1 spectra displays a symmetric line ( figure 2 ) .
the n@xmath10h@xmath0 and c@xmath4h@xmath10 lines ( benson et al . )
peak at the blue edge of our hco@xmath0 line .
because those lines were observed 115 from our position , we disregard this source in the statistical discussion in 4.2 .
fulkerson and clark ( 1984 ) used the h@xmath10co 6 cm line to model the density distribution as an inverse square law . the cs @xmath15 line observed by snell et al .
( 1982 ) is similar in velocity and shape to our hco@xmath0 @xmath1 spectra ( figure 1 ) .
the hco@xmath0 @xmath1 line is self - absorbed with the blue peak slightly stronger .
the h@xmath2co@xmath0 , n@xmath10h@xmath0 and c@xmath4h@xmath10 lines ( benson et al . )
peak in the self - absorption dip .
the hco@xmath0 @xmath1 line is symmetric with a faint blue shoulder ( figure 2 ) .
the peak velocity corresponds to that of the optically thin lines of benson et al .
we observe a broad , blue - skewed line in hco@xmath0 @xmath1 ( figure 1 ) .
the spectra for the hco@xmath0 @xmath1 line is blue skewed with the h@xmath2co@xmath0 @xmath1 line peaking to the red of the peak velocity of the hco@xmath0 line ( figure 1 ) .
the hco@xmath0 @xmath1 line is strongly blue - skewed and the h@xmath2co@xmath0 line peaks in the middle of the hco@xmath0 line ( figure 1 ) .
the n@xmath10h@xmath0 and c@xmath4h@xmath10 lines ( benson et al . ) are at the velocity of the red edge of the blue peak .
hong et al .
( 1991 ) have mapped this core in @xmath17co and @xmath2co @xmath13 .
the hco@xmath0 @xmath1 line is slightly blue - skewed ( figure 1 ) .
the n@xmath10h@xmath0 and c@xmath4h@xmath10 lines of benson et al .
peak at the red edge of the blue peak .
the hco@xmath0 @xmath1 line was not detected down to a limit of 0.05 k in l1495d , l1521a and l1517d .
six of these cores , l1544 , l1689a , l1689b , l183 , l63 and b133 , have line profiles with blue asymmetry . a blue asymmetry in an optically thick line like hco@xmath0
@xmath1 can be caused by protostellar collapse ( leung and brown 1977 , zhou & evans 1994 ) .
we observed h@xmath2co@xmath0 @xmath1 in all of these sources to find the rest velocity and we also used the n@xmath10h@xmath0 and c@xmath4h@xmath10 observations of benson et al . to provide optically thin line velocities . if optically thin lines have the same velocity as the dip of the double - peaked line , the dip is self - absorption from the ambient cloud and not caused by two separate velocity components blended together .
for such narrow lines as seen in these sources , the rest frequencies of lines become an issue .
the frequency of the hco@xmath0 @xmath1 line is known to within an uncertainty of 0.01 mhz , but the uncertainty of the h@xmath2co@xmath0 @xmath1 is potentially large . the standard value in the jpl data base ( pickett et al .
1998 , http://spec.jpl.nasa.gov ) is 260.255478 ghz .
it is calculated from measurements of the two lower lines ( woods et al .
1976 , bogey et al .
1981 ) , making it impossible to estimate uncertainties in the @xmath1 line frequency . comparing the velocities of sources done in h@xmath2co@xmath0 here and elsewhere ( gregersen et al .
1997 , 2000 ) with line velocities of the same sources in n@xmath10h@xmath0 , on average the h@xmath2co@xmath0 line is @xmath18 to the red of the n@xmath10h@xmath0 line . since it is unlikely that the h@xmath2co@xmath0 would be consistently red - shifted from the n@xmath10h@xmath0 , we have shifted our h@xmath2co@xmath0 spectra by 0.16 to the blue and propose a frequency of [email protected] ghz for the h@xmath2co@xmath0 @xmath1 line , 0.139 mhz higher than the standard value . for four of these sources , l1544 , b133 , l183 and l1689b ,
the h@xmath2co@xmath0 line does peak in the self - absorption dip , one of the conditions that must be met before a core can be called a candidate for protostellar collapse . in l63 ,
the h@xmath2co@xmath0 coincides with the dip but the n@xmath10h@xmath0 lies between the dip and the red peak .
we did not detect h@xmath2co@xmath0 in l1689a , so we can not make a claim about its quality as a collapse candidate .
how do the conclusions from compare to those of lee et al .
( 1999 ) who observed cs and ?
they did not observe b133 or l1689a , but they found l1544 , l1689b and l183 ( their position for l183b is closest to the core we call l183 ) to be infall candidates .
there are slight differences between the cs and spectra .
l1689b has a blue peak with a red shoulder in and double - peaked profile with a strong blue peak in cs .
l1544 has a strong blue peak in but two peaks of equal strength in cs .
l183 has similarly shaped spectra in both lines .
lee et al .
( 1999 ) observed a red peaked profile in l63 , but their position differs substantially from ours .
in general , the different tracers agree reasonably well .
optically thick lines in collapsing sources show a double - peaked line profile with the blue peak stronger than the red peak .
there are two ways of quantifying the asymmetry of the line .
one could use the ratio of the strengths of the two peaks or the asymmetry parameter ( mardones et al .
1997 ) , @xmath20 , where @xmath21 is the velocity of the peak of the optically thick line , @xmath22 is that of the optically thin line and @xmath23 is the linewidth of the optically thin line .
we list the results in table 4 .
we do not list the results for all the objects where hco@xmath0 was observed but only those objects for which we were able to establish a rest velocity from h@xmath2co@xmath0 , n@xmath10h@xmath0 or c@xmath4h@xmath10 .
l134a is listed in parentheses because its spectrum was observed 115 from our spectrum .
all the other sources for which we used the line to determine the asymmetry were observed within 40 of our position . in table 5 , we list the number of blue , symmetric , and red sources as determined by the asymmetry parameter and visual inspection of the line profiles .
following the classification of mardones et al .
sources with asymmetry @xmath24 @xmath25 are blue , while those with asymmetry @xmath26 0.25 are red .
the excess , which characterizes how blue is a sample of objects , is the number of blue sources minus the number of red sources divided by the total number of sources .
the excess as determined from the asymmetry parameter leaving aside l134a is 0.66 , higher than those derived for samples of class 0 and i sources , 0.25 and 0.39 respectively , using the @xmath1 line ( gregersen et al .
because the number of sources in our sample is small , it is unclear whether the larger excess is significant , but it appears to be at least as large as in the later stages .
lee et al . found the mean @xmath27 of starless cores to be @xmath280.04 not so different from the mean @xmath27 for class 0 sources , @xmath290.10 ( mardones et al . ) .
for the hco@xmath0 observations of all three categories , the mean @xmath27 using the h@xmath2co@xmath0 @xmath1 line frequency proposed in the previous section are @xmath300.13 , @xmath310.09 and @xmath320.08 for starless cores , class 0 and class i objects , respectively .
so on the basis of line asymmetries , these starless cores can not be distinguished from an older , protostellar population .
zhou ( 1992 ) modeled cs lines in cores that have not yet formed a protostar but are evolving toward the singular isothermal sphere , the starting point of the shu ( 1977 ) collapse model .
he found cs lines with no asymmetry , narrow line widths that increased after collapse began and peak line temperatures that increase until collapse begins .
gregersen et al .
( 1997 ) did evolutionary models of hco@xmath0 @xmath1 in collapsing clouds and found that such line parameters as peak temperature , line width , and blue - red asymmetry increase with time to a maximum value and then decline .
therefore , if we observed sources that were just beginning to collapse , these sources would have hco@xmath0 @xmath1 lines that have slight asymmetry , small peak temperatures and narrow line widths . if we extend these results backward to t = 0 , to the beginning of collapse , we should expect similar line parameters to those of zhou ( 1992 ) . also , since the luminosity of the protostar does not rise immediately , there is some time after collapse begins but before a source would have been detected by iras , so the early stages of the evolutionary models of gregersen et al .
could be applied to pre - protostellar " objects . in these models ,
important processes like cloud chemistry and protostellar heating that increase the line width and peak temperatures are glossed over or simplified , so the resulting evolutionary tracks " are best seen as rough sketches of line parameter evolution . in figure 3 , we plot peak temperature and linewidth versus the blue - red ratio for the two abundance distributions modeled in gregersen et al . the evolutionary tracks " in both plots go roughly from the lower left corner to the upper right .
we see that the starless cores are congregated toward the lower ( i.e. younger " ) half of the plot and the class 0 and i sources which have iras sources extend further to the upper ends of the two plots . however , there are indications that some of these sources have already begun to collapse in a different way than the shu model predicts .
several of the starless cores have blue / red ratios that are too large for early collapse and more like the extreme blue - red ratios seen in more evolved class 0 sources .
in fact , b133 displays the most extreme ratio between the blue and red peak of any source in any class .
the blue - red ratios for this and other sources are too high to arise in a shu model . mardones ( 1998 )
modeled several alternative velocity fields and found that a model in which the entire cloud collapses produces the highest blue - red ratio .
these large blue - red ratios are further discussed in the next section .
we have partially mapped five cores that showed asymmetric line profiles to observe the extent of the asymmetry .
the blue - skewed profiles typically stretch over a large area for such young cores .
for example , in l1689b ( figure 4 ) , the self - absorption or blue - skewed lines stretches over roughly 0.04 by 0.03 parsecs .
there are some red profiles in the southeast area of the map . for l63 ( figure 5 )
, the blue skewness is seen in an area 0.07 by 0.03 parsecs .
we show these two cores because these are the largest maps we have done with good signal - to - noise .
tafalla et al .
( 1998 ) have commented on this large extent of the infall signature in l1544 ( map in figure 6 ) .
they showed that if l1544 was undergoing inside - out collapse , it would have produced a protostar easily visible to iras and that such an early , large - scale infall does not fit with most theories of protostellar collapse .
these observations of extended infall signatures in combined with the extreme blue - red ratios mentioned in the last section strengthen the case for extended inward motions .
myers and lazarian ( 1998 ) have proposed that turbulent dissipation in a cloud produces an overall implosion starting from the core exterior as an explanation for extended infall signatures .
alternatively , li ( 1999 ) has found evidence for extended inward motions at speeds up to half the sound speed in calculations of core formation in weakly magnetized clouds .
similar speeds also appear in the model of ciolek and basu ( 2000 ) .
the question of whether to call these cores starless or pre - protostellar is a thorny one .
while these cores have no iras sources and so can all be called starless , the term pre - protostellar implies that the future evolution of these objects can be definitely predicted .
however , we can use the submillimeter observations of ward - thompson et al .
( 1994 ) to divide the sample into those cores detected by ward - thompson et al . at one or more wavelengths and those not detected at all ( table 6 ) .
of the 12 cores in the first group , 6 display blue asymmetry while the other six have no asymmetry .
the six cores that display blue asymmetry are the strongest in the submillimeter continuum . of the 5 cores
ward - thompson et al . did not detect , 2 have symmetric profiles while we did not detect the other 3 in hco@xmath0 .
the 2 detections of these cores were among the weakest lines observed .
we also include the nh@xmath4 observations of benson and myers ( 1989 ) as a marker of dense core evolution . of the 12 cores with submillimeter emission , all
were observed in nh@xmath4 . of the 5 others ,
only one was observed .
five of the cores in the first group have since been mapped in the submillimeter continuum ( shirley et al .
the four most centrally condensed cores show blue asymmetry .
the least centrally condensed core , l1512 , has no asymmetry in hco@xmath0 .
there is a clear distinction between those cores with submillimeter continuum and those without .
the first group has stronger line emission that often shows what could be the beginning of infall .
the group with relatively strong submillimeter continuum , hco@xmath0 , and nh@xmath4 emission are the likeliest to be pre - protostellar in nature , consistent with the suggestion by ward - thompson et al .
( 1994 ) .
the cores with weak submillimeter continuum could eventually form stars but were undetected simply because of lower column density .
these cores may be in an even earlier evolution stage in which cores are just beginning to form , or these cores may be forming stars of lower mass .
if these cores are weakly centrally concentrated , they would not have a sufficient excitation temperature gradient for infall asymmetry .
further observations would be needed to resolve the nature of these objects .
we have observed 17 starless cores in hco@xmath0 @xmath1 in search of the spectral signature of pre - protostellar collapse .
these objects do seem to be a younger population than the class 0 sources based on their narrow line widths and weak peak temperatures .
we have observed blue asymmetric line profiles in 6 of these cores , and we suggest l1544 , l1689b , b133 and l183 as good protostellar collapse candidates and l1689a and l63 as worthy of further observations in the h@xmath2co@xmath0 @xmath1 line .
the blue excess of this sample is as prominent as in samples of older " class 0 and i sources , suggesting that infalling protostars are not exclusively to be found among the class 0 sources .
a population of likely pre - protostellar cores can be distinguished by their strong submillimeter continuum and hco@xmath0 and nh@xmath4 spectral line emission .
however , recently tafalla et al .
( 1998 ) have found that l1544 can not be described as inside - out collapse .
if infall is happening in that core , it can not be explained by any current infall model , suggesting that further study of these cores can tell us about the very beginning of the collapse process .
we would like to thank tommy greathouse , wenbin li , byron mattingly and yancy shirley for their help with observations .
we thank the referee for helpful suggestions , one of which led us to reconsider the rest frequency for the h@xmath2co@xmath33 line .
this work was supported by nsf grant ast-9317567 and nasa grant nag5 - 7203 .
andr , p. & montmerle , t. 1994 , , 420 , 837 andr , p. , ward - thompson , d. , & barsony , m. 1993 , , 406 , 122 beichman , c. a. , myers , p. c. , emerson , j. p. , harris , s. , mathieu , r. , benson , p. j. , & jennings , r. e. 1986 , , 307 , 337 benson , p. j. 1983 , phd thesis , mit benson , p. j. , caselli , p. , & myers , p. c. 1998 , , 506 , 743 bogey , m. , demuynck , c. , & destombes , j. l. 1981 , mol .
43 , 1043 caselli , p. , myers , p. c. , & thaddeus , p. 1995
, , 455 , l77 ciolek , g. e. & basu , s. 2000 , , in press ( astro - ph/9909429 ) fulkerson , s. a. & clark , f. o. 1984 , , 287 , 723 gregersen , e. m. , evans , n. j. , ii , zhou , s. , & choi , m. 1997 , , 484 , 256 gregersen , e. m. , evans , n. j. , ii , mardones , d. , & myers , p. c. 2000 , apj , in press ( astro - ph/9912175 ) hong , s. s. , kim , h. g. , park , s. h. , park , y. s. , & imaoka , k. 1991 , jkas , 24 , 71 kooi , j. w. , chan , m. , phillips , t. g. , bumble , b. , & leduc , h. g. 1992 , ieee transactions on microwave theory and techniques , 40 , 812 kuiper , t. b. h. , langer , w. d. , & velusamy , t. 1996 , , 468 , 761 lee , c. w. , myers , p. c. , tafalla , m. 1999 , , 526 , 788 leung , c. m. & brown .
r. l. 1977 , , 214 , l73 li , z. y. 1999 , , in press ( astro - ph/9907358 ) mardones , d. 1998 , phd thesis , harvard university mardones , d. , myers , p. c. , tafalla , m. , wilner , d. j. , bachiller , r. , & garay , g. 1997 , 489 , 719 menten , k. m. , serabyn , e. , gusten , r. , & wilson , t. l. 1987 , , 177 , l57 myers , p. c. & benson , p. j. 1983 , , 266 , 309 myers , p. c. & lazarian , a. 1998 , , 507 , l157 myers , p. c. , linke , c. , & benson , p. j. 1983 , , 264 , 517 myers , p. c. , mardones , d. , tafalla , m. , williams , j. p. , & wilner , d. j. 1996 , , 465 , l133 ohashi , n. , lee , s. w. , wilner , d. j. , & hayashi , m. 1999 , , 518 , l41 pickett , h. m. , poynter , r. l. , cohen , e. a. , delitsky , m. l. , pearson , j. c. , & muller , h. s. p. 1998
, j. quant .
spectrosc . &
transfer , 60 , 883 shirley , y. l. , evans , n. j. , ii , rawlings , j. m. c. , & gregersen , e. m. , 2000 , submitted to apj shu , f. h. 1977 , , 214 , 488 snell , r. , langer , w. d. , & frerking , m. a. 1982 , , 255 , 149 tafalla , m. , mardones , d. , myers , p. c. , caselli , p. , bachiller , r. , & benson , p. j. 1998 , , 504 , 900 walker , c. k. , lada , c. j. , young , e. t. , maloney , p. r. , & wilking , b. a. 1986 , , 309 , l47 wang , y. 1994 .
phd thesis , university of texas at austin ward - thompson , d. , scott , p. f. , hills , r. e. , & andr , p. 1994
, , 268 , 276 williams , j. p. , myers , p. c. , wilner , d. j. , & di francesco , j. 1999 , , 513 , l61 wolkovitch , d. , langer , w. d. , goldsmith , p. f. , & heyer , m. 1997 , , 477 , 241 woods , r. c. , saykally , r. j. , dixon , t. a. , szanto , p. g. , & anderson , t. 1976 , 31st symposium on molecular spectroscopy , columbus , ohio zhou , s. 1992 , , 394 , 204 zhou , s. 1995 , , 442 , 685 zhou , s. & evans , n. j. , ii 1994 , in clouds , cores , and low mass stars , ed .
d. p. clemens & r. barvainis , san francisco : astron .
pacific , 183 zhou , s. , evans , n. j. , ii , kompe , c. , & walmsley , c. m. 1993 , , 404 , 232 lllllllll l1498 & 04:07:50.0 & 25:02:31 & ( -1200,0 ) & 140 & 700@xmath1980 & 120@xmath1918 & 35@xmath196 & [email protected] & 04:11:15.5 & 28:07:20 & ( -900,0 ) & 140 & - & @xmath24135 & @xmath2444 & @xmath2430 l1506 & 04:15:30.3 & 25:13:22 & ( -900,0 ) & 140 & - & @xmath2466 & @xmath2448 & -l1521a & 04:23:38.4 & 26:09:27 & ( -900,0 ) & 140 & - & @xmath24110 & @xmath2480 & -l1517c & 04:51:35.9 & 30:30:00 & ( -900,0 ) & 140 & 836@xmath19160 & 100@xmath1920 & @xmath2483 & @xmath247.5l1517a & 04:51:54.8 & 30:28:53 & ( -900,0 ) & 140 & 1280@xmath19330 & 105@xmath1918 & @xmath2460 & @xmath245.4l1517d & 04:52:36.5 & 30:34:02 & ( -900,0 ) & 140 & @xmath244500 & @xmath24120 & @xmath24130 & -l1512 & 05:00:54.4 & 32:39:37 & ( -900,0 ) & 140 & @xmath246000 & 107@xmath1921 & 45@xmath199 & @xmath2416l1544 & 05:01:13.1 & 25:06:36 & ( -900,0 ) & 140 & 1300@xmath19240 & 450@xmath1958 & 193@xmath1930 & 46@xmath194l1582a & 05:29:14.6 & 12:28:08 & ( -900,0 ) & 140 & @xmath241240 & 160@xmath1927 & @xmath2454 & @xmath2430l134a & 15:51:05.6 & -04:26:10 & ( -900,0 ) & 150 & - & @xmath2460 & - & @xmath24163 l183 & 15:51:32.7 & -02:42:19 & ( 600,0 ) & 150 & @xmath241500 & 269@xmath1930 & 108@xmath1926 & @xmath24134l1696a & 16:25:30.0 & -24:13:22 & ( 0,-1200 ) & 125 & 800@xmath19160 & 105@xmath1918 & 62@xmath1912 & @xmath2458l1689a & 16:29:10.5 & -24:57:22 & ( 0,-900 ) & 125 & 2200@xmath19300 & 290@xmath1945 & @xmath24102 & 54@xmath1915 l1689b & 16:31:47.0 & -24:31:45 & ( 0,-900 ) & 125 & @xmath243000 & 362@xmath1940 & 140@xmath1934 & 134@xmath1911 l63 & 16:47:19.4 & -18:01:16 & ( 0,-900 ) & 125 & 1600@xmath19200 & 367@xmath1923 & @xmath2493 & @xmath2496 b133 & 19:03:27.3 & -06:57:00 & ( 0,-600 ) & 400 & @xmath241800 & 341@xmath1963 & @xmath24120 & @xmath2456 lllll l1498 & hco@xmath0 @xmath1 & [email protected] & [email protected] & [email protected] l1495d & hco@xmath0 @xmath1 & @xmath240.05 & & l1506 & hco@xmath0 @xmath1 & [email protected] & [email protected] & [email protected] l1521a & hco@xmath0 @xmath1 & @xmath240.05 & & l1517c & hco@xmath0 @xmath1 & [email protected] & [email protected] & [email protected] l1517a & hco@xmath0 @xmath1 & [email protected] & [email protected] & [email protected] l1517d & hco@xmath0 @xmath1 & @xmath240.05 & & l1512 & hco@xmath0 @xmath1 & [email protected] & [email protected] & [email protected] l1544 & h@xmath2co@xmath0 @xmath1 & [email protected] & [email protected] & [email protected] - & hco@xmath0 @xmath1 & [email protected] & [email protected] & [email protected] - & - & [email protected] & [email protected] & - l1582a & hco@xmath0 @xmath1 & [email protected] & [email protected] & [email protected] l134a & hco@xmath0 @xmath1 & [email protected] & [email protected] & [email protected] l183 & h@xmath2co@xmath0 @xmath1 & [email protected] & [email protected] & [email protected] - & hco@xmath0 @xmath1 & [email protected] & [email protected] & [email protected] - & & [email protected] & [email protected] & l1696a & hco@xmath0 @xmath1 & [email protected] & [email protected] & [email protected] l1689a & h@xmath2co@xmath0 @xmath1 & @xmath240.03 & & & hco@xmath0 @xmath1 & [email protected] & [email protected] & [email protected] l1689b & h@xmath2co@xmath0 @xmath1 & [email protected] & [email protected] & [email protected] & hco@xmath0 @xmath1 & [email protected] & [email protected] & [email protected] l63 & h@xmath2co@xmath0 @xmath1 & [email protected] & [email protected] & [email protected] & hco@xmath0 @xmath1 & [email protected] & [email protected] & [email protected] b133 & h@xmath2co@xmath0 @xmath1 & [email protected] & [email protected] & [email protected] & hco@xmath0 @xmath1 & [email protected] & [email protected] & [email protected] lll l1498 & & [email protected] l1512 & & [email protected] l1544 & [email protected] & [email protected] l1582a & & [email protected] l134a & & ( [email protected] ) l183 & [email protected] & [email protected] l1696a & & [email protected] l1689a & [email protected] & l1689b & [email protected] & [email protected] l63 & [email protected] & [email protected] b133 & [email protected] & [email protected] lllll l1498 & y & y & y & n l1517c & y & y & y & n l1517a & y & y & y & n l1512 & y & y & y & n l1544 & y & y & y & y l1582a & y & y & y & n l183 & y & y & y & y l1696a & y & y & y & n l1689a & y & y & y & y l1689b & y & y & y & y l63 & y & y & y & y b133 & y & y & y & y l1495d & n & n & n & n l1506 & n & y & n & n l1521a & n & n & n & n l1517d & n & n & n & n l134a & n & y & y & n | we have observed the hco@xmath0 @xmath1 line toward 17 starless cores selected from the list of ward - thompson et al .
( 1994 ) .
six of these cores have line asymmetries indicative of collapse .
the excess of blue - skewed profiles over red - skewed profiles is at least as large as that found in samples of class 0 and early class i sources .
the observed line profiles have the same narrow linewidths and small peak temperatures predicted for young sources in evolutionary models , but the blue / red ratios , like those of older sources , are higher than models predict .
the infall signature also occurs over large scales , suggesting that these cores have overall inward motions .
we have divided these starless cores into two groups based on the continuum photometry of ward - thompson et al . and our hco@xmath0 data .
we find stronger hco@xmath0 emission among the cores detected in the submillimeter and all the blue - skewed line profiles are in this group , supporting the suggestion of ward - thompson et al . that these are the pre - protostellar cores .
= 2 ed at min . , |
in this paper we study the applicability of nekhoroshev s theorem @xcite ( see also @xcite ) to the central motion .
the main point is that nekhoroshev s theorem applies to perturbations of integrable systems whose hamiltonian is a _ steep _ function of the actions .
even if such a property is known to be generic , it is very difficult ( and not at all explicit ) to verify it .
here we prove that , under an explicit condition on the potential ( see eq . ) , the hamiltonian of the central motion is a quasi - convex function of the actions and thus it is steep , so that nekhoroshev s theorem applies .
actually , the form of nekhoroshev s theorem used here is not the original one , but that for degenerate systems proved by fass in @xcite .
this is due to the fact that the hamiltonian of the central motion is a function of two actions only , namely , the modulus of the total angular momentum and the action of the effective one dimensional hamiltonian describing the motion of the radial variable . on the one hand ,
as pointed out in @xcite , this feature creates some problems for the proof of nekhoroshev s theorem , but these problems were solved in @xcite . on the other hand , degeneracy reduces the difficulty for the verification of steepness or of the strongest property of being quasi - convex , since , in the two - dimensional case , quasi - convexity is generic and equivalent to the nonvanishing of the arnold determinant @xmath0 a property that it is not too hard to verify . indeed , since is an analytic function of the actions , it is enough to show that it is different from zero at one point in order to ensure that it is different from zero almost everywhere . here
we explicitly compute the expansion of @xmath1 at a circular orbit and we show that , provided the central potential @xmath2 does not satisfy identically a fourth order differential equation that we explicitly write , the hamiltonian @xmath1 is quasi - convex on an open dense domain ( whose complementary is possibly empty ) .
the rest of the paper is organized as follows : in sect .
[ statement ] we introduce the central motion problem and state the main results .
[ proof ] contains all the proofs . in the appendix
we prove that in the two dimensional case quasi - convexity is equivalent to arnold isoenergetic nondegeneracy condition .
_ acknowledgements .
_ we thank f. fass for a detailed discussion on action angle variables in the central motion problem , m. guzzo , l. niederman and g. pinzari for pointing to our attention some relevant references and a. maspero for interesting discussions .
we first recall the structure of the action angle variables for the central motion . introducing polar coordinates
, the hamiltonian takes the form @xmath3 and the actions on which @xmath4 depends are @xmath5 and the action @xmath6 of the effective one dimensional hamiltonian system @xmath7 by construction @xmath4 turns out to be a function of the two actions only .
we still write @xmath8 according to fass s theorem , if @xmath4 depends on @xmath9 in a steep way , then nekhoroshev s estimate applies .
we recall that steepness is actually implied by quasi - convexity , the property that we are now going to verify . a function @xmath4 of the actions
is said to be _ quasi - convex _ at a point @xmath10 if the system @xmath11 admits only trivial solutions . here
we denoted by @xmath12 the second differential of @xmath4 at @xmath10 applied to the two vectors @xmath13 . to define the set @xmath14 in which the actions vary we first assume that there exists an interval @xmath15 such that , for @xmath16 one has @xmath17 then we define @xmath18 and , in order to fix ideas , we assume that @xmath19 . then for @xmath20 , the effective potential @xmath21 has a non degenerate minimum at some @xmath22 .
then , there exists a curve @xmath23 such that for @xmath24 , all the orbits of the hamiltonian are periodic .
correspondingly , their action @xmath25 vary in some interval @xmath26 .
thus , the domain @xmath27 of the actions @xmath28 has the form @xmath29 we remark that @xmath27 is simply connected , a property that will play an important role in the following .
our main result is the following .
[ teorema1 ] consider the hamiltonian @xmath30 with @xmath2 analytic on @xmath31 .
assume that there exists a value @xmath32 of the radius such that the following fourth order equation @xmath33 is _ not _ satisfied .
then , there exists a subset @xmath34 of the action space , with the property that its intersection with any compact set is composed by at most a finite number of lines , and such that @xmath4 restricted to @xmath35 is quasi - convex .
the fourth order equation can be rewritten as a second order ordinary differential equation in terms of the variable @xmath36 , namely , @xmath37 it is interesting to see what are the homogeneous potentials which do not give rise to steep hamiltonians .
thus take @xmath38 with @xmath39 , then the assumptions of theorem [ teorema1 ] are fulfilled if @xmath40 thus the excluded cases are the kepler and the harmonic potentials .
we also remark that the equation has also the solution @xmath41 , which however is excluded by the second of .
fix a positive parameter @xmath42 and denote by @xmath43 the open ball of radius @xmath42 and center @xmath28 and define @xmath44 we now consider a small perturbation @xmath45 of @xmath4 , with @xmath46 a function of the original cartesian coordinates @xmath47 in @xmath48 which is analytic .
[ teorema2 ] fix a positive small parameter @xmath42 and consider the hamiltonian @xmath49 .
then , for every compact set @xmath50 , there exist positive constants @xmath51 such that if the initial value @xmath52 of the actions fulfills @xmath53 and @xmath54 one has @xmath55 for all times @xmath56 satisfying @xmath57 the main dynamical consequence is that , as in the central motion , for any initial datum as above , there exist @xmath58 such that @xmath59 for the times . actually , one can weaker the analyticity requirement on @xmath46 , since it would be enough to have that it is analytic on the set of @xmath60 and @xmath61 s such that the action is close to @xmath52 .
the theorem holds also if one couples the system to an other system with a dynamic taking place over a much faster time scale . for example , this occurs in the case of a soliton interacting with radiation in the nls equation as in @xcite .
as anticipated above , in the case of _ two _ actions , quasi - convexity is equivalent to the nonvanishing of the arnold determinant @xmath62 ( cf . ) . in order to compute @xmath62
, we have to compute @xmath63 . to do this we proceed as follows . first ,
as explained in sect .
[ statement ] , it is easy to introduce the action @xmath64 which coincides with the modulus of the total angular momentum .
then , @xmath6 is the action of the effective one dimensional system in which @xmath64 plays the role of a parameter . in order to have an explicit formula for the hamiltonian as a function of the actions , we work at circular orbits .
precisely , we exploit the remark that in one dimensional systems birkhoff normal form converges in a neighborhood of a nondegenerate minimum . indeed , birkhoff normal form allows to construct an analytic canonical transformation which , in a neighborhood of the critical point , conjugates the hamiltonian @xmath4 to a function of the form @xmath65 which moreover is explicitly constructed as a power series in @xmath66 .
thus , one can define the first action @xmath6 by @xmath67 .
remark that , since , as a function of @xmath68 is analytic in a whole _ complex _
neighborhood of zero , then @xmath63 is analytic for @xmath6 in a whole neighborhood of zero .
then , from uniqueness of the actions in one dimensional systems , one has that the expression one gets is actually the expression of @xmath4 as a function of the actions as defined in sect .
[ statement ] .
it also follows that @xmath69 as a function of @xmath6 extends to a complex analytic function in whole neighborhood of zero , and such a function can be computed using the expression of @xmath4 obtained from .
we use this remark in order to compute the arnold determinant at the equilibrium point @xmath70 of the effective one dimensional system described by @xmath4 , which coincides with a circular orbit of the original system .
this can be done by computing explicitly the second order taylor expansion of @xmath4 which coincides with the fourth order birkhoff normal form of @xmath71 at @xmath70 .
actually this was already done in @xcite getting @xmath72 where @xmath73 fix a point @xmath74 and let @xmath75 be the corresponding critical point of the effective potential . inserting into the arnold determinant @xmath62 , the first and second derivatives of the hamiltonian evaluated at the point @xmath76
, one gets @xmath77 where we denoted by @xmath78 thus , @xmath79 is equivalent to @xmath80 isolating the fourth order derivative of @xmath81 one gets @xmath82 using that @xmath70 is a critical point of @xmath83 , one can express @xmath74 in terms of @xmath70 , namely , @xmath84 and computing @xmath85 one can rewrite the equation in terms of the radius @xmath70 , obtaining the equation . thus ,
if there exists @xmath70 such that is not satisfied , then at this point @xmath86 .
this concludes the proof of the theorem .
@xmath87 * proof of theorem [ teorema2 ] * from the previous result , one gets that the hamiltonian is quasi - convex on the subset @xmath88 with uniform bounds on the quasi - convexity constants .
thus , if we choose an initial datum in such a set and a sufficiently small parameter @xmath89 , then fass s version of the nekhoroshev theorem for degenerate hamiltonians applies ( see @xcite for details ) .
we remark that , in the case of quasi - convex hamiltonians , one gets optimal nekhoroshev s exponents which for two dimensional systems are @xmath90 . @xmath87
in this appendix we show that , in the two dimensional case , quasi - convexity is equivalent to the nonvanishing of the arnold determinant @xmath62 .
we start by recalling the arnold condition .
let @xmath4 be a complete integrable hamiltonian with @xmath91 degrees of freedom and frequency @xmath92 .
then , @xmath4 is said to satisfy the _
arnold condition _ at @xmath93 if the following map @xmath94 has maximal rank at @xmath95 .
thus , we conclude that , in the case @xmath105 , the arnold condition is equivalent to the request of the second differential @xmath106 to be different from zero on the hyperplane generated by the vector @xmath107 normal to the gradient , namely , quasi - convexity . | in this paper we prove a nekhoroshev type theorem for perturbations of hamiltonians describing a particle subject to the force due to a central potential . precisely , we prove that under an explicit condition on the potential , the hamiltonian of the central motion is quasi - convex .
thus , when it is perturbed , two actions ( the modulus of the total angular momentum and the action of the reduced radial system ) are approximately conserved for times which are exponentially long with the inverse of the perturbation parameter . |
let @xmath1 be a continuous linear operator defined on a separable banach space @xmath2 .
let us say that @xmath1 is cyclic if @xmath3 such that @xmath4 is dense in @xmath2 .
on the other hand , we said that @xmath1 is locally quasinilpotent at @xmath3 if @xmath5 the notion of local quasinilpotence was introduced in @xcite to obtain nontrivial invariant subspaces for positive operators .
positive results about @xmath0-tuples of operators with a nontrivial common invariant subspace have been obtained in @xcite . in this article , we extend the results of abramovich _
et al _ @xcite to the case of not - necessarily commuting @xmath0-tuples of continuous linear operators . to extend these results
it will be essential to introduce the notion of joint local quasinilpotence for @xmath0-tuples of operators ( not necessarily commuting ) .
the results complement the results obtained in @xcite .
the main result of this paper appears in 3 and is stated as follows : let @xmath6 be a @xmath0-tuple of continuous linear operators .
if @xmath1 is positive with respect to a cone @xmath7 and @xmath8 exists such that @xmath1 is joint locally quasinilpotent at @xmath9 , then there exists a common nontrivial invariant subspace for @xmath6 .
moreover , using this theorem we deduce new results on nontrivial common invariant subspaces for @xmath0-tuples of operators ( see theorem 3.2 , corollary 3.3 ) .
we will conclude this article with a section including open problems and further directions .
it is easy to see that if @xmath10 are @xmath0 commuting operators and they are locally quasinilpotent at @xmath11 , then the compositions @xmath12 for all @xmath13 and all @xmath14 , are locally quasinilpotent at @xmath9 .
in fact the intersection of the sets @xmath15 is a common invariant manifold .
however if @xmath16 are not commuting , the problem becomes more complicated .
let @xmath17 be two operators on the hilbert space @xmath18 defined in the following way : @xmath19 where @xmath20 is the canonical basis of @xmath18 .
the operators @xmath21 and @xmath22 are locally quasinilpotent at @xmath23 for each @xmath24 , since @xmath25 for each @xmath26 , and therefore @xmath27 .
on the other hand , @xmath28 , hence @xmath29 and therefore @xmath21 and @xmath22 are locally quasinilpotent at @xmath23 for each @xmath24
. however , @xmath30 and @xmath31 are not locally quasinilpotent at @xmath23 for each @xmath24 . indeed , since @xmath32 , @xmath33 on the other hand , we know @xmath34 and hence @xmath35 .
let @xmath36 be an @xmath0-tuple of continuous linear operators on a banach space @xmath2 not necessarily commuting .
let us denote by @xmath37 the collection of all possible products of @xmath38 elements in @xmath1 .
@xmath39 let @xmath36 be an @xmath0-tuple of continuous linear operators on a banach space @xmath2 not necessarily commuting .
then , we will say that @xmath1 is uniform joint locally quasinilpotent at @xmath11 if @xmath40 the notion of uniform joint local quasinilpotence is closely related with the joint spectral radius defined by rota and strang @xcite .
we can get more information about spectral theory in several variables in @xcite .
although the results of this article are formulated under the hypothesis of uniform joint local quasinilpotence , nevertheless , sometimes it will be possible to replace the above - mentioned hypothesis by the following weaker property .
@xmath39 let @xmath36 be an @xmath0-tuple of continuous linear operators on a banach space @xmath2 .
then we will say that @xmath1 is joint locally quasinilpotent at @xmath11 if @xmath41 for all @xmath42 .
the difference between the concept of uniform joint local quasinilpotence and joint local quasinilpotence is the lack of uniformity .
next we see some properties of both concepts .
@xmath39 let @xmath43 be an @xmath0-tuple of continuous linear operators on a banach space @xmath2 and let us suppose that @xmath1 is uniform joint locally quasinilpotent at @xmath44 .
then for all polynomial @xmath45 of @xmath46 variables , such that @xmath47 we have that @xmath48 where @xmath49 , that is , the operator @xmath50 is locally quasinilpotent at @xmath51 .
fix @xmath52 .
let us suppose that @xmath53 is the number of summands of the polynomial @xmath45 , and let us denote by @xmath54 the maximum of the modulus of the coefficients of @xmath45 . then , since @xmath55 is uniform joint locally quasinilpotent at @xmath51 , there exists @xmath56 such that @xmath57 for all @xmath58 .
now , taking into account that the polynomial @xmath45 has no independent term , for all @xmath58 , @xmath59 which proves the desired result .
@xmath60 in fact the condition on the polynomial @xmath61 is a necessary condition in the proof of proposition 2.3 .
indeed , let @xmath62 be the forward shift defined on the sequences space @xmath63 by @xmath64 , for all @xmath65 .
it is easy to see that the operator @xmath66 is not locally quasinilpotent at @xmath67 ( where @xmath68 denotes the identity operator ) .
definitions 2.1 and 2.2 are the natural extensions of the notion of local quasinilpotence for @xmath0-tuples of continuous linear operators . in fact , let us denote @xmath69 and let us denote by @xmath70 the set of all uniform joint locally quasinilpotent vectors for @xmath71 .
then , we have the following result . @xmath39 let @xmath36 be an @xmath0-tuple of continuous linear operators on a banach space @xmath2 , then the sets @xmath70 and @xmath72 are common invariant manifolds by @xmath73 .
clearly , @xmath74 implies that @xmath75 .
now let @xmath76 , and fix @xmath52 .
then , there exists some @xmath77 such that @xmath78 and @xmath79 for each @xmath80 and each @xmath81 .
therefore , @xmath82 for all @xmath83 .
therefore @xmath84 and so @xmath72 is a vector manifold .
fix @xmath85 and let @xmath86 . then @xmath87 for each @xmath88 and for each @xmath89 .
therefore @xmath72 is a common invariant manifold for @xmath90 .
similar proof follows for the set @xmath70 , and hence we omit it .
@xmath60 the above propositions show that if @xmath91 ( @xmath92 respectively ) and @xmath93 ( @xmath94 respectively ) , then @xmath95 ( @xmath96 respectively ) is a common nontrivial closed invariant subspace for @xmath16 .
as far as the invariant subspace problem is concerned , we need only consider the two extreme cases @xmath97 and @xmath98 .
let @xmath2 be a banach space with a schauder basis @xmath99 .
we say that @xmath6 is positive with respect to the cone @xmath100 if @xmath101 for all @xmath102 .
let us see the main result of this paper .
let @xmath6 be an @xmath0-tuple of continuous linear operators on a banach space with a schauder basis such that @xmath6 is positive with respect to the cone @xmath7 , and let us suppose that @xmath103 exists such that @xmath36 is joint locally quasinilpotent at @xmath104 .
then there exists a common nontrivial invariant subspace for @xmath105 .
let @xmath99 be a schauder basis of the banach space @xmath2 and let @xmath106 be the sequence of coefficient functionals associated with the basis @xmath99 .
assume that @xmath6 is joint locally quasinilpotent at some @xmath107 , i.e. , @xmath108 with @xmath109 for all @xmath110 .
let us suppose that @xmath111 for all @xmath112 .
then @xmath113 is a common nontrivial invariant subspace for each @xmath73 .
thus , we can suppose that @xmath114 for some @xmath115 . by an appropriate scaling of @xmath104
, we can assume that @xmath116 and @xmath117 for some @xmath118 and some @xmath115 .
now let @xmath119 denote the continuous projection onto the vector subspace generated by @xmath120 defined by @xmath121 .
clearly , @xmath122 holds for each @xmath123 .
we claim that @xmath124 for each @xmath125 . to see this , fix @xmath125 and let @xmath126 for some nonnegative scalar @xmath127 .
since @xmath128 is a positive operator and the composition of positive operators is a positive operator , it follows that @xmath129 let us observe that the following inequality is not true because the norm is not monotone @xmath130 . however , if we use the fact that @xmath131 is a positive linear functional , we conclude that @xmath132 consequently , @xmath133 , and so @xmath134 from definition 2.2 we know @xmath135 .
thus we deduce that @xmath136 , and condition ( [ uno ] ) must be true .
now let us consider the linear subspace @xmath137 of @xmath2 generated by the set @xmath138 clearly , @xmath137 is invariant for each @xmath139 and , since @xmath140 for some @xmath141 , we see that @xmath142 . from ( [ uno ] )
it follows that @xmath143 , therefore @xmath144 for each @xmath145 .
this implies that @xmath146 for each @xmath147 , and consequently @xmath146 for all @xmath148 , that is , @xmath149 .
the latter shows that @xmath150 is a common nontrivial closed invariant subspace for the operators @xmath151 , and the proof is complete.@xmath60 let @xmath73 be joint locally quasinilpotent operators at @xmath8
. then we can add arbitrary weights to each matrix representing the operators @xmath73 and still guarantee the existence of a common nontrivial closed invariant subspace .
@xmath39 1 .
first , let us observe that if @xmath71 is joint locally quasinilpotent at @xmath51 it is possible to obtain a closed invariant subspace @xmath62 ( nontrivial ) invariant also for every positive operator @xmath152 such that @xmath153 for all @xmath154 .
indeed , the above proof follows considering the closed subspace generated by @xmath155 \hskip -1.25pc & \quad\ , \forall\ j\in \mathbb{n } , a\ \hbox{positive}\ at_{i}=t_{i}a ( \forall i)\}.\end{aligned}\ ] ] 2 . on the other hand ,
let us mention that the subspace guaranteed in the above theorem is in fact an invariant nontrivial ideal . in the following theorem ,
positivity is with respect to the cone generated by the schauder basis of the banach space .
let @xmath2 be a banach space with a schauder basis .
assume that the matrix @xmath156 defines a continuous operator @xmath157 for all @xmath158 , such that the @xmath0-tuple @xmath36 is joint locally quasinilpotent at a nonzero positive vector .
let @xmath159 be @xmath0-double sequences of complex numbers .
if the weighted matrices @xmath160 define continuous operators @xmath161 , then @xmath162 have a common nontrivial closed invariant subspace .
let @xmath99 be a schauder basis of the banach space @xmath2 , and let @xmath106 be the sequence of functional coefficients associated with the basis @xmath99 .
assume that the operators @xmath16 satisfy @xmath163 for some positive nonzero vector @xmath104 .
an appropriate scaling of @xmath104 shows that there exists @xmath164 satisfying @xmath165 . let us suppose that @xmath166 for all @xmath89 , then an easy argument shows that @xmath167 for all @xmath89 , and @xmath168 is a nontrivial closed invariant subspace ( here we assume that @xmath169 for all @xmath89 ) .
thus , we can suppose that @xmath170 for some @xmath115 .
now , let us denote by @xmath119 , the positive projection defined by @xmath171 .
arguing as in the proof of theorem 3.1 , we can establish that @xmath172 for each @xmath14 , where @xmath173 for all @xmath110 .
in particular , we have @xmath174 for each @xmath145 .
consequently , for each @xmath175 and for each positive operator @xmath176 satisfying @xmath177 , we have @xmath178 next , let us consider the vector subspace @xmath137 generated by the set @xmath179 clearly @xmath137 is invariant for each operator @xmath180 satisfying @xmath181 . also , from ( [ dos ] ) it follows that @xmath182 for all @xmath148 , that is , @xmath149 .
the latter shows that @xmath150 is a nontrivial closed vector subspace of @xmath2 .
let @xmath183 now be the operators defined by @xmath184 since @xmath185 satisfy @xmath186 for all @xmath89 , it follows that @xmath150 is invariant for each one of the operators @xmath185 .
therefore , the vector subspace @xmath150 is invariant under the operators @xmath187 however , the sequence of operators @xmath188 converges in the strong operator topology to @xmath189 .
therefore , @xmath190 and , thus , the operators @xmath191 , have a common nontrivial closed invariant subspace . @xmath60
@xmath39 let @xmath2 be a banach space with a schauder basis .
assume that the positive matrices @xmath156 define continuous operators on @xmath2 , which are joint locally quasinilpotent at a nonzero positive vector .
if the continuous operators @xmath192 are defined by the matrices @xmath193 satisfying @xmath194 whenever @xmath195 , then the operators have a common nontrivial closed invariant subspace .
the notion of uniform joint local quasinilpotence is used extensively in @xcite to obtain common nontrivial invariant subspaces .
both concepts , joint local quasinilpotence and uniform joint local quasinilpotence , play an important role in the search of common nontrivial invariant subspaces .
it would be interesting to know something more on the sets @xmath196 and @xmath197 .
our conjecture is that both sets are equal in majority of the cases . on the other hand
, it would be interesting to extend the results of theorems 3.1 and 3.3 for the case of @xmath0-tuples of positive operators defined on a hausdorff topological vector space , where the partial order is defined by means of a markushevish basis .
the authors would like to thank the referee for his comments and suggestions that have improved this article . | in this article we obtain some positive results about the existence of a common nontrivial invariant subspace for @xmath0-tuples of not necessarily commuting operators on banach spaces with a schauder basis .
the concept of joint quasinilpotence plays a basic role .
our results complement recent work by kosiek @xcite and ptak @xcite .
[ ssection ] [ ssection ] |
in its first year , the _ hubble space telescope _ ( hst ) discovered that a majority of all baryons in the current universe are not in galaxies , but instead remain in the intergalactic medium ( igm ) . in subsequent years , the uv spectrographs aboard hst and the _ far ultraviolet spectroscopic explorer _ ( fuse ) have continued these investigations of the multiphase igm , using sensitive uv tracers of diffuse gas : the lyman series of h i ( at 1215.67 , at 1025.72 , etc ) and the o vi doublet ( 1031.926 , 1037.617 ) .
these hst and fuse studies have led to a preliminary `` baryon census '' of the `` warm '' ( photoionized ) and `` warm - hot '' ( collisionally ionized ) igm .
with spectrographs aboard the _ chandra _ and _ xmm / newton _ x - ray telescopes , astronomers are beginning to search for even more highly ionized gas through resonance absorption lines of o vii , o viii , n vii , and ne ix . unlike virtually all other astronomical objects
, the ly@xmath16 absorption systems were first discovered at great distances ( @xmath17 ) owing to their cosmological redshifts and the near - uv atmospheric cutoff . only with the advent of hst
have nearby examples been found .
the first low-@xmath15 absorbers were seen in the spectrum of 3c 273 at @xmath18 ( bahcall et al .
1991 ; morris et al .
while the number of absorbers was significantly less than the line density at high-@xmath15 , the `` local forest '' contains far more absorbers than expected from extrapolating the ground - based data ( bahcall 1993 and subsequent qso absorption - line key project papers by jannuzi 1998 and weymann 1998 ) . although the absorbers at @xmath17 are sufficiently abundant to account for nearly all the baryons ( rauch 1997 ; schaye 2001 ) , their substantial numbers at @xmath3 imply that @xmath230% of all baryons remain in these photoionized clouds locally ( penton , stocke , & shull 2000a , penton , shull , & stocke 2000b , 2004 , hereafter denoted papers i , ii , and iv ) . numerical simulations ( fig . 1 ) of the evolving igm ( cen & ostriker 1999 ; dav et al .
1999 , 2001 ) explain not only the general features of the number density evolution , but also many detailed properties , including distributions in column density ( ) and doppler @xmath19-value ( papers ii and iv ; dav & tripp 2001 ; ricotti , gnedin , & shull 2000 ) , and their relationship to galaxies ( dav 1999 ; impey , petry , & flint 1999 ; penton , stocke , & shull 2002 , hereafter denoted paper iii ) .
any accounting of the present - day distribution of baryons must include an accurate census of these absorbers and the associated mass , inferred from their ionized fractions , column densities , and physical extents .
moderate - resolution uv spectroscopy of bright quasars , qsos , blazars , and seyfert galaxies has provided a substantial database of low-@xmath15 absorbers . at the same time , several ground - based galaxy surveys ( morris 1993 ; lanzetta 1995 ; chen 1998 ; tripp , lu , & savage 1998 ; rao & turnshek 1998 , 2000 ; impey , petry , & flint 1999 ; nestor 2002 ; bowen 2002 ; paper iii ; bowen & pettini 2003 ; stocke 2005 , hereafter denoted paper v ) have probed the relationship between absorbers and galaxies , filaments of galaxies , and voids .
using nearby examples of the phenomenon , these authors sought to identify the galaxy types responsible for the absorption and thus assist in interpreting the wealth of information ( number densities , metallicities , ionization states , line widths ) of absorption systems at high-@xmath15 .
these efforts have been somewhat successful , although the results in most column - density regimes remain controversial ( see conference proceedings edited by mulchaey & stocke 2002 ) . in this review ,
we describe the various hst qso absorption line surveys that have been undertaken ( 2 ) , review our present knowledge of the baryon content of the igm ( 3 ) , and describe the emerging , but still controversial , evidence for the relationship between the various column densities of absorbers and galaxies ( 4 ) .
the last information has come largely from studying low-@xmath15 absorbers discovered with hst .
we conclude ( 5 ) with a brief prospectus on low-@xmath15 igm studies facilitated by the _
cosmic origins spectrograph _ ( cos ) , a new instrument that may be installed on hst in the coming years .
the hst with its uv spectrographs ( fos , ghrs , stis ) conducted several important surveys of the igm , which provided basic data for studying the bulk of local baryons . owing to its modest spectral resolution ( 200 - 300 ) , the
_ faint object spectrograph _ ( fos ) used for the initial qso absorption - line key project ( bahcall 1993 ) primarily detected high column density absorbers with equivalent widths @xmath20 m . the key project provided examples of the various types of absorbers : damped ( dla ) absorbers , lyman - limit / strong mg ii absorbers , weak high - ionization ( c iv ) and low - ionization ( mg ii ) metal - line absorbers , and -only absorbers ( bahcall 1996 ; jannuzi 1998 ) .
even though the broad uv wavelength coverage ( g130h , g190h , g270h gratings ) of the key project spectra allowed the discovery of many absorbers at @xmath21 , the detection efficiency of low redshift ( @xmath22 ) absorbers was reduced by lower than expected far - uv sensitivity of the fos digicon .
the fos key project survey firmly established the column density distribution , @xmath23 ( ) , for high- absorbers and @xmath24 , the number density of absorbers per unit redshift . above limiting equivalent width ,
@xmath25 m ( @xmath7 @xmath1 ) , weymann ( 1998 ) found @xmath26 over a substantial redshift pathlength ( @xmath27 30 ) .
as we discuss below , the line density increases substantially to lower columns , reaching @xmath5 for @xmath6 2 ( paper iv ) . the absorber number density ( weymann 1998 ) shows a dramatic break from rapid evolution ( @xmath28 ) to almost no evolution ( @xmath29 ) .
this observation was explained by cosmological hydrodynamical simulations ( dav 1999 ) , in which the rapid evolution in @xmath24 is controlled by the response of photoionized gas to evolving qso populations in a hierarchical distribution of large - scale structure . at high redshift ,
the number of absorbers decreased rapidly with time , owing to a nearly constant extragalactic ionizing flux from qsos in an expanding universe .
[ the cosmological expansion decreases recombinations , thus decreasing the neutral hydrogen column density .
] at @xmath30 , the ionizing flux began a rapid decline with the rapidly decreasing qso numbers and/or luminosities , thereby decreasing the ionized fraction in the absorbers .
as a consequence of reduced photoionization , along with large - scale structure formation , the @xmath24 evolution slowed rapidly . although the key project discovered only one dla ( jannuzi 1998 ) , rao & turnshek ( 1998 ) used fos to increase the number of low-@xmath15 dlas by targeting strong mg ii absorbers found by ground - based telescopes .
bowen ( 1996 ) used archival fos spectra to search for strong and metal - line absorption from known foreground galaxies . see bowen ,
pettini , & blades ( 2002 ) for a stis extension of this successful project .
the _ goddard high - resolution spectrograph _ ( ghrs ) was used to co - discover ( morris 1991 ) the absorbers in the 3c 273 sightline .
however , few qsos were sufficiently bright to observe in moderate - length integrations .
tripp , lu , & savage ( 1998 ) and impey , petry , & flint ( 1999 ) observed a few uv - bright agn targets with the low - resolution far - uv ghrs grating , obtaining observations only slightly more sensitive than the fos key project spectra . in the mid-1990s ,
the colorado group conducted a moderate - resolution ( @xmath2 20 ) survey of 15 sightlines with the ghrs / g160 m first - order grating ( papers i and ii ) , followed by a comparable survey of 15 additional targets with the stis first - order , medium - resolution grating g140 m ( paper iv ) . in these surveys , the key strategy was to select only the brightest targets ( see fig .
2 ) , while observing only a modest pathlength per target ( @xmath31 10,000 with ghrs and 20,000 with stis ) .
the number density of absorbers rises steeply with decreasing equivalent width .
after subtracting intrinsic absorbers ( near the redshift of the qso ) we found 187 intervening low- absorbers at @xmath32 significance over a total unobscured pathlength @xmath33 . a careful analysis , accounting for s / n sensitivity bias and overlap with galactic interstellar lines , gives a line density @xmath5 for @xmath6 @xmath1 , or one absorber every @xmath34 mpc .
our more sensitive ghrs / stis survey provides an important extension of the key project to lower- absorbers . as shown in figure 2 ,
the very best spectra in our survey reach equivalent - width limits of @xmath35 10 - 20 m ( @xmath36 ) , corresponding to @xmath37 2 , comparable to the best spectra obtained with keck / hires or vlt / uves .
the cosmic evolution of the ly@xmath16 forest absorbers shown in figure 3 represents the historical record of a large fraction of the baryons during the emergence of the modern universe at @xmath30 .
the @xmath24 evolution of the high- absorbers exhibits a substantial break in slope to a much slower evolution at @xmath381.5 , as expected from the physical effects acting on intergalactic absorbers in an expanding universe .
the low- data point at @xmath39 from our survey is lower than that of the key project .
we suspect this results from our survey s higher spectral resolution , which affects the counting and -bin assignments of weak absorbers .
the break in the @xmath24 evolution of the higher- absorbers is therefore not as dramatic as the key project data suggest .
a more gradual decline in slope is also more consistent with the simulations in a @xmath40cdm universe model ( dav et al .
the physical effects that account for the break in the high - column @xmath24 should create a similar break in the evolution of the lower- absorbers , although this is not yet evident in figure 3 .
new stis spectra obtained by b. jannuzi and collaborators should determine whether the expected break is present , and whether the numerical simulations have used the correct physics to account for these absorbers .
recently , several investigators have used stis in its medium - resolution ( @xmath41 ) echelle modes ( e140 m and e230 m ) to obtain long integrations on bright agn targets .
many of these spectra have been analyzed for interesting individual absorbers , including a growing number of redshifted o vi absorbers : h1821 + 643 ( tripp 2000 , 2001 ) ; 3c 273 ( sembach 2001 ; tripp 2002 ) ; pg 0953 + 415 ( savage 2002 ) ; pks 2155 - 304 ( shull 2003 ) ; pg 1259 + 593 ( richter 2004 ) ; pg 1116 + 215 ( sembach 2004 ) ; and pg 1211 + 143 ( tumlinson 2004 ) . these spectra have the potential to be hst s best resource for low-@xmath15 absorber studies . as of july 2004 , 35 bright agn have been observed in stis - echelle mode , providing in some cases the most sensitive detection limits for , as well as significant pathlengths at @xmath42 for potential discovery of o vi absorbers . in the absence of the _ cosmic origins spectrograph
_ , the stis medium - resolution echelle gratings continue to provide high - quality qso spectra for igm studies .
however , obtaining the s / n needed to make sound inferences about the igm absorbers requires increasingly long integration times on fainter targets .
at high redshift ( @xmath43 ) the number density of absorbers is so large that estimates of their total baryonic masses , after substantial ionization corrections for photoionization equilibrium , suggest that the entire baryon content of the universe ( @xmath44 , as measured by the cosmic microwave background anisotropies or cosmic [ d / h ] ratio ) can be accomodated within the warm ( @xmath45 k ) photoionized igm ( rauch 1997 ; schaye 2001 ) .
this result is consistent with numerical simulations of large - scale structure formation ( cen & ostriker 1999 ; dav 1999 , 2001 ) , which also predict that by @xmath39 the baryons are approximately evenly divided among several components as follows : * * stars and gas in or near galaxies ; 30% predicted*. salucci & persic ( 1999 ) and fukugita ( 2004 ) account for only @xmath46% in this reservoir . *
* very hot ( @xmath47 k ) intracluster and intragroup gas ; @xmath4810% predicted*. fukugita ( 2004 ) estimates only @xmath49% of the baryons in this component , but the gas in groups may not be fully accounted for . *
* warm ( @xmath50 k ) photoionized gas ; 3040% predicted in forest absorbers*. in our paper iv , we find @xmath51% of the baryons in local absorbers . * * warm - hot intergalactic medium ( whim ) at @xmath52 k ; 2040% predicted*. a series of observations with hst and fuse ( tripp , savage , & jenkins 2000 ; savage 2002 ) estimate that 5 - 10% of the baryons have been detected ( at 10% assumed o / h metallicity ) through o vi absorption .
recent _ chandra _ observations of o vii absorption ( fang 2002 ; nicastro 2004 ) suggest even larger percentages , although with enormous statistical uncertainties and a few unconfirmed observations .
thus , we submit that the current - epoch baryon census remains an unsolved problem , with large and uncertain corrections for ionization state ( h i , o vi , and o vii ) and an overall uncertainty over the appropriate metallicity ( for o vi , o vii , o viii in whim ) . these problems can be addressed by hst and fuse in the coming years , and by future high - throughput x - ray spectroscopic missions such as _ constellation - x _ or _
xeus_. in paper ii , we described a method for estimating the baryon content of the local absorbers , based upon the observed column density distribution . in the context of optically - thin , photoionized clouds , the assumptions of this simplified model
are : ( 1 ) spherically symmetric absorbers ; ( 2 ) an isothermal density profile ; ( 3 ) absorber sizes of @xmath53 kpc , based on qso pairs experiments conducted at somewhat higher redshifts ( dinshaw 1997 , but see also rosenberg 2003 ) ; and ( 4 ) a value and slope for the extragalactic ionizing flux ( shull 1999 ) using space densities and ionizing spectra of local seyfert galaxies ( telfer 2002 ) .
however , numerical simulations show that absorbers are not at all spherical , especially at lower column densities .
cen & simcoe ( 1997 ) used numerical simulations at @xmath15=24 to show that lower column density ( @xmath54 2 ) absorbers have large physical sizes , consistent with or somewhat larger than @xmath53 kpc ( equivalent spherical radius ) .
we have integrated this simplified model over our observed column density distribution , from 12.5 @xmath48 log @xmath55 , the range where the above assumptions are most valid . in paper
ii we obtained a baryon fraction @xmath56% , with specific dependences on measurable parameters as follows : @xmath57^{1/2 } h_{70}^{-1 } \ ; .\ ] ] here , @xmath58 is the characteristic h i column density in the forest in units of @xmath59 2 , @xmath60 is the extragalactic ionizing specific intensity at 1 ryd in units of @xmath61 erg @xmath1 s@xmath62 hz@xmath62 sr@xmath62 , @xmath63 is the characteristic absorber radius in units of @xmath64 kpc , and @xmath65 is the mean power - law spectral index ( @xmath66 ) for the metagalactic ionizing radiation , which presumably comes from qsos ( shull 1999 ; telfer 2002 ) .
schaye ( 2001 ) developed a methodology for estimating the baryon content that makes somewhat different assumptions .
as with the method of our paper ii , this model assumes photoionization of optically - thin absorbers , but it also assumes gravitationally - bound clouds whose observed column densities are equal to their characteristic column densities over a jeans length .
this estimate has the following specific dependences : @xmath67 where @xmath68 is the h i photoionization rate in units of @xmath69 s@xmath62 , @xmath70 is the igm temperature in units of @xmath50 k , and @xmath71 is the h i column density distribution . in this model ,
the absorber baryon content depends on the minimum column density of h i absorbers as @xmath72 or @xmath73 for @xmath74 ( see below ) .
the above expression assumes that the absorbers contain the universal ratio of baryons to dark matter , with no bias . here
we employ both of these methods to estimate the baryon content of the local absorbers , based on our latest column density distribution from paper iv , which fitted power - law slopes of @xmath74 and @xmath75 on the lower and upper sides , respectively , of a break at @xmath76 2 . from our enlarged sample ,
we are confident that the @xmath77 slope extends down to at least log = 12.5 , so that we can reliably integrate this distribution from log = 12.5 to 16.0 .
we recognize that , at or near our adopted lower limit ( which corresponds approximately to an overdensity @xmath78 at the current epoch ) , some of the assumptions of both methods may break down .
also , at @xmath79 2 , absorbers become optically thick in the lyman continuum and can hide additional gas mass . therefore , neither method can hope to derive an extremely accurate measurement for the baryon content of the forest , even if the column density distribution and other absorber properties ( size , ionizing flux , temperature ) are known to some precision .
we quote the resulting baryon densities as percentages of the total baryon density from measurements of the cosmic microwave background anisotropy ( @xmath80 , spergel 2003 ) .
a slightly smaller total baryon density has recently been obtained by measurement of d / h in a high-@xmath15 qso absorber ( @xmath81 , kirkman 2003 ) .
the baryon fraction in the igm is dominated by the lower column density absorbers ( 22% out of 29% total baryons in absorbers at @xmath82 2 ) , owing to the steep slope of the column density distribution .
the schaye ( 2001 ) prescription yields a divergent baryon fraction at low , owing to an inverse dependence of size on column density , which must break down well above the limit employed here . on the other hand , at @xmath12 10@xmath83 @xmath1 , the 100 kpc size assumed in paper ii probably becomes too large for many absorbers at higher column densities ( cen & simcoe 1997 ; tripp 2002 ) .
thus , the most conservative , and we believe most accurate approach , is to use our paper ii method at low columns and the schaye ( 2001 ) method at high columns .
this technique yields values of 22@xmath842% below @xmath76 2 , and @xmath85% above that column density , for a total baryon fraction in the low-@xmath15 photoionized igm of @xmath86% ( paper iv ) . because the baryon fraction is dominated by the lowest column density absorbers , our result is relatively insensitive to the assumed @xmath19-values for individual absorbers . here , we have used @xmath87 , which is comparable to the median @xmath88 found by dav & tripp ( 2001 ) using medium - resolution echelle spectra . in order to make further progress in the baryon census of warm igm gas ,
it is important to : ( 1 ) estimate absorber sizes and shapes ( and thus equivalent cloud radii ) as a function of column density . in the absence of observations of many qso pairs
, this is best done using numerical simulations ( the cen & simcoe [ 1997 ] analysis , but for simulations at @xmath39 ) ; ( 2 ) estimate the percentage of absorbers that are collisionally ionized rather than photoionized .
this will require observations of o vi with fuse or hst to address the `` double - counting '' issue ; see 3.2 ; ( 3 ) observe a much longer pathlength for absorbers , since we find that the bias introduced by cosmic variance is still significant for the observed redshift pathlength , @xmath89 ( paper iv ) ; ( 4 ) determine more accurate h i column densities from fuse observations of higher lyman lines and a curve - of - growth analysis ( e.g. , shull 2000 ; sembach 2001 ) for saturated absorbers ; and ( 5 ) observe a few very bright targets over very long integration times ( @xmath90 orbits ) to push the detection limits below @xmath91 2 ; very weak absorbers can still contribute to the baryon budget .
the search for the whim gas has now begun , using sensitive uv resonance absorption lines : primarily the o vi lithium - like ( @xmath92 ) doublet ( 1031.926 , 1037.617 ) but also the li - like doublet of ne viii ( 770.409 , 780.324 ) see savage ( 2004 ) .
the o vi lines can be observed with hst / stis for igm absorbers with @xmath42 , while the fuse spectrographs can observe o vi essentially down to @xmath93 .
the k - shell x - ray absorption lines of o vii ( 21.602 ) , o viii ( 18.97 ) , n vii ( 24.782 ) , ne ix ( 13.447 ) , and a few other highly ionized species can ( barely ) be observed with _
chandra _ and _ xmm / newton _ , primarily toward x - ray flaring blazars . as mentioned in 2 , there now exists a substantial number ( 35 to our knowledge ) of stis medium - resolution echelle spectra of qsos .
however , fewer than half of these targets have sufficient redshift ( @xmath42 ) and continuum signal - to - noise ratio ( s / n @xmath94 per resolution element ) to facilitate a sensitive search for o vi absorption .
thus far , only a modest pathlength has been searched for o vi absorbers with hst ( tripp , savage , & jenkins 2000 ; savage 2002 ) . these spectra yield measured baryon fractions of 510% ( assuming 10%
o / h metallicity ) , consistent with expectations from simulations ( gnedin & ostriker 1997 ; cen 2001 ; fang & bryan 2001 ) .
we caution that these estimates involve large uncertainties , including untested assumptions of cloud sizes and shapes , metallicities , ionization equilibrium , and multi - phase structure .
the multiphase nature of igm absorbers leads directly to the classic census problem of double counting " . in the igm context
, this means that the photoionized h i and o vi absorbers could count the same baryons twice .
recent surveys down to 50 m equivalent width give h i line densities approximately ten times those of the o vi absorbers : @xmath95 ( paper iv ) and @xmath96 ( savage 2002 ) .
many igm absorbers ( shull 2003 ) as well as galactic high velocity clouds ( collins 2004 ) clearly contain gas at vastly different temperatures , in order to explain the observed range of ionization stages . deriving the metallicities of such complex systems requires subtle ionization modeling : collisional vs. photoionization , time - dependent ionization , range of temperatures , inhomogeneities in density and velocity .
theoretical considerations of the shocked igm ( dav 1999 ) suggest that collisional ionization dominates the stronger o vi systems , although some weak o vi absorbers have been modeled successfully using photoionization equilibrium ( tripp 2001 ) .
the problem with these photoionization models for high ions is that they require a large photoionization parameter ( @xmath97 ) to produce sufficient o vi . at fixed ionizing radiation intensity , @xmath98 , this requires very low hydrogen densities , @xmath99 @xmath100 . as a result , photoionization models with high-@xmath101 and low-@xmath102 are physically implausible , since they produce a low neutral fraction , @xmath103 and unrealistically large absorption pathlengths , @xmath104 here , we have scaled the low - redshift parameters to h i photoionization rate @xmath105 , a temperature @xmath106 , and a typical strong h i absorber with @xmath107 .
shull ( 1999 ) find that @xmath108 0.30.7 at @xmath39 ; the ionization rate can also be written @xmath109 $ ] , where @xmath60 is the specific intensity at 1 ryd and @xmath65 is the mean spectral index of the ionizing radiation .
thus , some of the photoionized baryons in the o vi absorbers may have been accounted for by the h i calculation in the previous section .
however , the hotter , collisionally ionized o vi should represent a distinct thermal phase from the h i. a substantial survey of o vi absorbers may reveal systematics among the o vi absorbers that will allow us to separate the collisionally - ionized absorbers from the photoionized absorbers and thus unravel the double - counting problem .
ongoing stis echelle spectroscopy by tripp , savage , howk , and others will go a long way towards answering these questions and better quantifying the `` o vi forest '' statistics . equally important to this effort is ongoing fuse spectroscopy of lower - redshift o vi with complementary data from _
hubble_. to date , we have found 114 absorbers ( @xmath110 m ) with acceptable o vi data , and roughly half ( 57 systems ) are detected in o vi at @xmath111 significance ( danforth & shull 2004 ) .
this combined hst / fuse database can address important questions concerning the extent to which metals are spread away from galaxies into the igm . while much data exist and more is scheduled to be obtained , these are still early days . if the study of the o vi absorbers is still in its infancy , the detection of the remainder of the whim gas is not yet `` out of the womb '' .
simulations predict that this hotter whim ( 10@xmath112 k ) contains @xmath12 2040% of the total local baryons .
nicastro ( 2003 ) claims that whim gas at @xmath113 k has been detected at @xmath114 along several sightlines , through absorption lines of o vi with fuse and of c vi , o vii , o viii , n vii , and ne ix with _
chandra _ and/or _
xmm / newton_. however , it is possible that this gas is not intergalactic at all , but instead resides in smaller physical regions in the halo of the milky way ( sembach 2002 ; shull 2003 ; fox 2004 ; collins 2004 ) . while two groups have claimed whim detections at @xmath115 in the sightlines of pks 2155 - 304 ( fang 2002 ) and h1821 + 643 with _
( mathur 2003 ) , neither of these results has been confirmed .
the pks 2155 - 304 detection has been refuted by _
xmm / newton _ observations ( rasmussen 2003 ; cagnoni 2004 ) .
recently , two new whim detections were obtained ( nicastro 2004 ) toward the blazar mrk 421 at @xmath116 , while it flared in x - rays .
the claimed o vii absorbers lie at @xmath117 ( where a absorber was identified by shull , stocke , & penton 1996 ) and @xmath118 ( where we see no or o vi absorbers to @xmath119 limits of 30 m and 20 m , respectively ) .
the @xmath118 absorber was detected in three x - ray lines ( n vi , n vii , o vii ) at @xmath120 significance each , while the @xmath117 absorber was seen in only o vii at @xmath121 significance ( but not in o viii ) .
the o viii / o vii is consistent with whim at temperatures log @xmath122 5.76.5 . over the small observed pathlength , @xmath123
, these @xmath124 detections yield an o vii line density , @xmath125 , of 24 times that of o vi , with large uncertainty . the o vii columns ( @xmath126 @xmath1 ) are @xmath127 times those typically seen in o vi .
therefore , the baryon fraction in the o vii - bearing gas could be anywhere from 2070% , with enormous uncertainties arising from cosmic variance of this one short sightline . a possible one - line whim detection in the _ chandra _ spectrum of 3c 120 ( mckernan 2003 ) yields a similar o vii line density and baryon fraction .
obviously , the statistics of these studies are poor , and the model assumptions substantial .
sadly , the prospects of further observations with _ chandra _ are remote ; when mrk 421 was observed it was the brightest extragalactic source in the sky ( the brightest extragalactic soft x - ray sources tend to be blazars ) . because the velocity resolution of _ chandra s _ low energy transmission grating is only @xmath2 1000 , one worries about the proximity ( 900 ) of the @xmath118 absorber to the blazar redshift ( @xmath116 ) .
thus , the @xmath118 absorber might be outflowing gas along a jet and not intervening whim at all . taking this observational program to the level of fuse ( or even hst )
may have to await the _ constellation - x _ grating spectrographs .
while the baryon census and its evolution are ample reasons for studying the local forest in detail , it is also only at low redshift that the locations of absorbers and galaxies can be compared accurately .
this allows the relationship between these `` clouds '' and galaxies to be determined to some degree of certainty .
however , the degree to which absorbers correlate with individual galaxies has been controversial .
lanzetta ( 1995 ) and chen ( 1998 , 2001 ) argue that the absorbers are the very extended halos of individual galaxies , while others ( morris 1993 ; stocke 1995 ; impey , petry , & flint 1999 ; dav 1999 ; papers iii and v ) argue that the absorbers are related to galaxies only through their common association with large - scale gaseous filaments , arising from overdensities in the high - redshift universe .
much of the difference between these results primarily reflects the column density range of the absorbers studied in each case ; at some level both sides to this argument are correct ( see conference papers in mulchaey & stocke 2002 ) . in this section
, we review the current evidence for the relationship between galaxies and absorbers as a function of h i column density . at the highest column densities ( @xmath128 2 )
exhibits broad damping wings that allow an accurate determination of h i column density .
this has proved valuable for studying the evolution of the chemical abundance of galaxies with cosmic time , since even quite weak metal lines are detectable in these systems with high - quality spectra ( e.g. , pettini 1999 ; prochaska & wolfe 2000 , 2002 ) .
dlas have also been used to measure the evolution of the amount of neutral gas in galaxies ( boissier 2003 ) .
however , clues about the types of galaxies probed by the dlas are largely circumstantial at high redshift .
thick disks of massive galaxies are suggested by the kinematics ( wolfe & prochaska 2000 ) , but are hardly demanded by the evidence .
surveys of dlas with hst ( e.g. , rao & turnshek 2000 ; turnshek 2002 ) and their subsequent imaging with large - aperture ground - based telescopes ( nestor 2002 ; rao 2003 ) have shown convincingly that no single galaxy population dominates the dlas .
surprisingly , there is a large contribution to dlas made by dwarf and low surface brightness ( lsb ) galaxies ( see bowen , tripp , & jenkins 2001 for an excellent example ) in contrast to the wolfe & prochaska ( 2000 ) kinematic analysis . until recently , an apparent contradiction existed between the cross - sectional areas on the sky covered by high column density h i at @xmath114 as measured either by dlas or by 21 cm emission maps ( rao & briggs 1993 ) .
however , new `` blind '' 21 cm surveys ( zwaan 2001 ; rosenberg & schneider 2003 ) show that the 21-cm cross - sectional area now agrees with the dla measurement at low-@xmath15 .
these new surveys also show that the galaxy populations detected by these two different methods now agree .
rosenberg & schneider ( 2003 ) find that dla galaxies span a wide range of total h i masses ( two - thirds are between @xmath129 ) and luminosities 0.011 @xmath10 .
thus , the chemical abundance data from dlas at high-@xmath15 must be interpreted with these new hst @xmath130 ground - based results in mind ; that is , the chemical abundance evolution refers to h i disks and amorphous h i clumps in a wide variety of galaxies .
this certainly explains the broad spread in dla metallicities at any one redshift .
in the range @xmath11 2 , the lyman limit system ( lls ) absorbers constitute a substantial number ( @xmath131 13 per unit redshift ) of intervening systems in the spectra of high-@xmath15 qsos .
as shown by steidel ( 1995 , 1998 ) , strong ( @xmath132 ) mg ii absorbers at @xmath133 are invariably also llss , although a few dlas are also present in such samples .
physically , this can be understood because h i and mg ii have similar ionization potentials , so that once hydrogen becomes optically thin , mg ii ionizes rapidly to mg iii . in an important pre - hst result ,
steidel ( 1995 , 1998 ) found that , of 58 llss in his sample , 55 could be identified with nearby ( @xmath134 kpc offsets ) , bright ( @xmath135 ) galaxies at the same redshift ( see also bergeron & boiss 1991 ) .
these results have been strengthened by obtaining ground - based spectra of the underlying stellar kinematics of the associated galaxies and hst imaging and uv spectra of the c iv absorptions in these systems ( churchill 2000 ; steidel 2002 ) .
these results suggest that the llss are almost exclusively the bound gaseous halos of luminous galaxies , and that these halos share the kinematics of the underlying stellar disk .
recently , hst observations of the quasar 3c 232 probed the halo of the modest starburst galaxy ngc 3067 .
the spectra show a complex lls with three absorption components at small radial velocity differences ( @xmath136 , @xmath137 and @xmath138 ) from the nucleus of ngc 3067 , in ions ranging from na i and ca ii detected optically ( stocke 1991 ) to mg i , mg ii , fe ii and mn ii in the near - uv ( tumlinson 1999 ) to si iv and c iv in the far - uv . surprisingly , all these species share the same three velocities ( keeney et al .
2005 ) , which are gravitationally bound to ngc 3067 ; two of the systems , including the strongest at @xmath139 2 , are infalling .
the 3c 232 sightline has an offset of @xmath140 kpc from the nucleus of ngc 3067 ; the sightline clearly penetrates extra - planar h i , seen in 21 cm emission with the vla ( carilli & van gorkom 1992 ) .
the h i column densities , kinematics , metallicity , spin temperature , and inferred size of these clouds are similar to high velocity clouds in the milky way .
thus , despite a modest starburst ( @xmath141 yr@xmath62 ) in this @xmath142 galaxy , even the high ionization gas does not appear to be escaping .
in summary , both the detailed hst and ground - based observations of this nearby lls and the cumulative data from the steidel lls sample are consistent with llss being the bound , gaseous halos of luminous galaxies .
both the hst qso absorption line key project and our ghrs / stis survey of papers i iv have found large numbers of intermediate column density ( @xmath143 2 ) absorbers at low redshift .
available hst or fuse spectra are not sensitive enough to detect metals ( c ii / iii / iv , si ii / iii / iv , o vi ) in these absorbers , except in a few cases at @xmath144 @xmath1 .
for example , metallicities have been measured to 110% of solar values in strong absorbers in the sightlines toward 3c 273 ( sembach 2001 ; tripp 2002 ) , pks 2155 - 304 ( shull 1998 , 2003 ) , pg 1259 + 593 ( richter 2004 ) , and pg 1211 + 143 ( tumlinson 2004 ) . in an ongoing survey ,
danforth & shull ( 2004 ) have found that @xmath250% of all absorbers contain both h i and o vi in the range for which high - quality fuse spectra exist .
the multiphase ratio " ,
n(h i)/n(o vi ) , varies substantially , from @xmath145 to greater than 100 ( shull 2003 ; danforth & shull 2004 ) , probably indicating a wide range in shock velocities , and possibly o / h abundances , in these multiphase systems . given these statistics , it is conceivable that a large fraction of @xmath39 absorbers at @xmath146 2 contain metals .
the o vi non - detections are either at lower ionization ( lower shock velocity ) than allows significant columns of o vi , or at slightly lower metallicity than the fuse sensitivity limits can detect .
since most or all absorbers at @xmath43 in this column density range contain c iv and/or o vi absorption lines at metallicities @xmath147 = -1 $ ] to @xmath148 ( e.g. , schaye 2000 ) , it is not surprising that many low column density , low-@xmath15 absorbers would contain some metals .
thus , some relationship with nearby galaxies is expected .
figure 4 shows the galaxies surrounding the sightline to pks 2155 - 304 , coincident with a cluster of strong absorbers , in which c iv or si iii have not been detected to a limit of @xmath23% solar metallicity ( there are two weak o vi absorbers detected by fuse ) .
lanzetta ( 1995 ) and chen ( 1998 , 2001 ) obtained deep images and multi - object spectroscopy for a large number of galaxies in the fields of key project absorbers .
lanzetta ( 2002 ) summarizes the evidence for a @xmath230% success rate in matching redshifts of absorbers and luminous galaxies within @xmath149 kpc of the sightline .
they then extrapolate to the conclusion that all fos - discovered absorbers can be associated with very extended galaxy halos .
chen ( 1998 ) found that , with the exception of a weak dependence on galaxy luminosity , there is no nearest - galaxy property that correlates with absorber properties . on the other hand , using absorbers at somewhat lower column densities , morris ( 1993 ) ,
tripp , lu , & savage ( 1998 ) , impey , petry , & flint ( 1999 ) and our papers iii and v found no convincing statistical evidence that these ghrs- and stis - discovered absorbers are associated with individual galaxies down to @xmath150 luminosities .
however , over 3/4 of these weak absorbers are found in galaxy large - scale structure filaments ( paper iii ) , with bright galaxies several hundred kpc away . stocke ( 2002 ) summarizes the evidence in favor of this position . at the extremes of the column density range
are obvious examples that both of these positions are correct for some absorbers ; at the high- end of this range , llss are so close to the nearby bright galaxy that an association of some sort seems inevitable .
on the other hand , an increasing percentage of lower column density absorbers are found in galaxy voids ( papers iii and v ) , more than @xmath151 mpc from the nearest bright or faint galaxy ( mclin 2001 ) . at intermediate ,
the absorbers adjacent to gas - rich galaxies do not have velocities consistent with rotation curves of the nearby galaxy .
this suggests no kinematic link to the nearby galaxy ( ct 2002 ; putman 2004 ) , in contrast to the steidel ( 2002 ) kinematic analysis of llss and their associated galaxies that often shows a relationship between absorber metal - line velocities and the kinematics of the underlying stellar disk .
recently , the discovery ( stocke 2004 ) of a dwarf ( @xmath152 , @xmath153 ) post - starburst galaxy @xmath154 kpc from a weak - metal line absorber ( figure 5 ) with @xmath155 2 suggests a solution compatible with all the data presented above .
not only do the absorber and galaxy redshifts match to within their combined errors , but the absorber metallicity ( 6% solar ) and the mean stellar metallicity of the galaxy ( @xmath210% solar ) approximately match .
further , the absorber has an overabundance of silicon to carbon indicative of recent supernova type ii enrichment .
the galaxy is a pure disk system whose optical spectrum has both strong balmer and metal absorption lines , and no evidence for dust or gas ( no emission lines and @xmath156 ; van gorkom 1993 ) . from ratios of lick absorption - line indices
, we estimate that the mean stellar age in this galaxy is @xmath157 gyrs . taken together
, this information provides a consistent picture in which a massive ( @xmath158 yr@xmath62 ) starburst @xmath159 gyrs ago created enough supernovae to blow away the remaining gas from this galaxy into the igm .
because the dwarf is quite low mass , this wind can easily escape from the galaxy and move to @xmath160 kpc at the required 2030 to create the metal - line absorber that we see between us and 3c 273 . because this is the nearest weak metal - line system known
, many other similar absorbers could arise from starburst winds produced by dwarf galaxies too faint to be detected by the above surveys ( at @xmath161 , a galaxy like this dwarf would have @xmath162 ) .
this explains why lanzetta ( 1995 ) found bright galaxies @xmath163 kpc away that do not have properties that correlate with absorber properties .
these absorbers were produced not by the bright galaxy , but by a dwarf or dwarfs that accompany the bright galaxy .
this also explains the correlation between absorber locations and large - scale structure in the samples of paper iii and impey , petry & flint ( 1999 ) .
because dwarf galaxies are so numerous ( 13 mpc@xmath164 ) , if each dwarf galaxy had at least one massive starburst that ejected most or all of the gas from the galaxy to a distance of @xmath160 kpc , this process would be sufficient to create several hundred weak metal - line absorbers per unit redshift .
this number is comparable to the line density of all metal - line absorbers at @xmath165 . while this dwarf galaxy would have been much more luminous ( @xmath166 ) when it was `` starbursting ''
, the present - day absence of gas means that it will no longer form stars and will eventually fade to the approximate luminosity of a local group dwarf spheroidal .
thus , the dwarf in figure 5 is the expected intermediate stage between the `` faint blue galaxies '' seen at @xmath167 and the present - day dwarf spheroids ( babul & rees 1992 ) .
that such `` cheshire cat '' galaxies might be responsible for a large portion of the forest was suggested by salpeter ( 1993 ) and charlton ( 1995 ) .
however attractive this solution appears , based upon this one example , an equally plausible solution consistent with all available data is that large - scale filaments of galaxies are enriched with metals throughout the filament to approximately the same metallicity . in this picture ,
metal enrichment is not due to any one galaxy .
indeed , in some environments , this almost has to be the case
. in the virgo cluster , studied through qso absorption line probes by impey , petry & flint ( 1999 ) , no nearby dwarf galaxies were found , despite a galaxy redshift survey complete to @xmath168 . in some cases ( the @xmath169
h i @xmath130 o vi absorber toward 3c 273 , tripp 2002 ; and the 1666 metal - line absorber toward rxj 1230.8 + 0115 , rosenberg 2003 ) weak metal - line absorbers have no nearby ( @xmath170 kpc ) galaxies , despite a considerable redshift survey effort . while our galaxy surveying continues in these directions , these two hypotheses may not be easy to discriminate .
however , a statistical study of the distances that metal - enriched gas extends away from galaxies may help to select the better model .
many of the lowest- absorbers could be metal - free , a hypothesis consistent with all available data at high and low redshift . although hst does not have the sensitivity to address this question directly , there is considerable circumstantial evidence in favor of this hypothesis .
while @xmath280% of all low column density absorbers are found in galaxy filaments , the remaining 20% and an increasing percentage of absorbers with decreasing column densities , are found @xmath171 mpc from the nearest bright galaxy .
a deep survey of the few nearest examples of `` void absorbers '' ( mclin et al . 2002 ) found no faint galaxies ( @xmath172 to @xmath173 , depending on sightline ) within several hundred kpc of these absorbers .
thus , there is no plausible nearby source for metals in these absorbers . in paper iv
, we presented a two - point correlation function ( tpcf ) in absorber velocity separation that showed excess power at @xmath174 but only for the higher column density absorbers in our sample ( @xmath175 m ) . the lower column density absorbers ( @xmath176 ) appear to be randomly distributed in space with no excess power at any @xmath177 ( paper iv ) .
searches for metals in the highest column density `` void absorbers '' are capable of setting metallicity limits at @xmath147 \approx -2 $ ] ( b. keeney 2004 , private communication ) , but some high-@xmath15 absorbers contain metals at even lower metallicities .
therefore , even deeper hst observations of very bright targets are required to test this interesting hypothesis .
when hst servicing mission 4 was still scheduled , the installation of the _ cosmic origins spectrograph _ ( cos ) on hst was eagerly awaited by astronomers who study the local igm . with 1020 times
the far - uv throughput of stis at comparable ( 15 ) resolution , cos would revolutionize low-@xmath15 absorption work by allowing us to obtain high - resolution , high signal - to - noise ratio spectra of 16@xmath178
@xmath179 magnitude qsos in a few orbits .
stis studies of the igm require ever - increasing numbers of orbits to address new science goals .
below is a sampling of the questions that either stis or cos can address ; those that truly require cos are noted with asterisks ( * ) : 1 . * baryons in the whim .
* how many are there ? where are they relative to galaxies ?
what are the ionization processes that create the observed absorption - line signatures ? 2 . * galaxy halo sizes , shapes , metallicities and kinematics.*@xmath180 a few of the brightest examples can be attempted with stis , but to provide gaseous halo properties as a function of galaxy type and luminosity is clearly a cos project .
* low column density absorbers.*@xmath180 how many baryons do these absorbing clouds contain ?
do they contain metals and at what metallicity ?
this is a particularly interesting question for those absorbers in voids , which are the most obvious locations to search for primordial hydrogen clouds , if they exist .
their metals provide an important `` fossil record '' of very early star formation .
* extent of metal transport away from galaxies.*@xmath180 how far away from star formation sites do metals spread in the igm ?
what types of galaxies are responsible for enriching the igm in metals ? can details of nucleosynthesis be found in the igm to understand early star formation history and evolution ?
* accurate h i column densities .
* fuse observations of higher lyman lines can be used to obtain accurate h i column densities by curve - of - growth techniques .
these are also essential for accurate metallicities .
therefore , this goal is best carried out while fuse and hst are both still operational .
* physical properties of igm absorbers . * the column density distribution , @xmath19-value distribution ( temperature ) , detailed relationship to galaxies and other properties are still affected by cosmic variance .
obtaining 1020 high signal - to - noise stis medium - resolution echelle spectra or a similar number of high - quality cos spectra of @xmath181 qsos would be sufficient to address concerns about cosmic variance for absorbers , but not for the rarer o vi absorbers . for o
vi , fuse spectroscopy of already detected absorption lines appears to be a viable route to obtaining a sufficient sample .
one could also use hst / stis data on qso targets at @xmath182 ( to shift o vi @xmath183 or @xmath184 into the hst / stis band ) .
this project also means observing fainter targets for many more orbits .
* size of igm absorbers in galaxy filaments and voids.*@xmath180 qso pairs available to perform these observations are too faint to observe adequately with stis .
as nearly everyone has noticed over the past decade , the igm has attracted considerable interest from the fields of cosmology , galaxy formation , and galactic chemical evolution .
even at low redshift , a substantial fraction ( 3060% ) of the baryons reside in the igm , where they influence the mass infall and chemical history of gas in galaxies , including the milky way . if this
cosmic web " of multi - phase ( photoionized , collisionally ionized ) gas truly represents a significant baryon reservoir ( shull 2003 ; sembach 2003 ) , one of the best scientific legacies from _ hubble _ ( and fuse ) will be an accurate characterization of its features .
+ * acknowledgements * financial support for the local ly@xmath16 forest work at the university of colorado comes from grants provided through hst go programs 6593 , 8182 , 8125 , 8571 , 9221 , 9506 , 9520 , and 9778 from the fuse project ( nasa contract nas5 - 32985 , grant nag5 - 13004 ) , and theoretical grants from nasa / ltsa ( nag5 - 7262 ) and nsf ( ast02 - 06042 ) .
we thank ray weymann for the faint - galaxy redshift survey work and his inspirational leadership in this field .
we thank mark giroux for ionization modeling and insightful comments .
we thank our other collaborators , brian keeney , jessica rosenberg , jason tumlinson , john hibbard , charles danforth , and jacqueline van gorkom for valuable contributions and discussions , and f. nicastro and b. keeney for permission to quote their results before publication .
babul , a. , & rees , m. j. 1992 , , 255 , 346 bahcall , j. n. , 1991 , , 377 , l5 bahcall , j. n. , 1993 , , 87 , 1 bahcall , j. n. , 1996 , , 457 , 19 bechtold , j. , 1994 , , 437 , l38 bergeron , j. , & boiss , p. 1991
, , 243 , 344 boissier , s. , proux , c. , & pettini , m. 2003 , , 338 , 131 bowen , d. v. , blades , j. c. , & pettini , m. 1996 , , 464 , 141 bowen , d. v. , & pettini , m. 2003 , in the igm / galaxy connection , eds . j. l. rosenberg & m. e. putman ( kluwer : dordrecht ) , 133 bowen , d. v. , pettini , m. , & blades , j. c. 2002 , 580 , 169 bowen , d. v. , tripp , t. m. , & jenkins , e. b. 2001 , , 121 , 1456 cagnoni , i. , nicastro , f. , maraschi , l. , treves , a. , & tavecchio , 2004 , , 603 , 449 carilli , c. l. , & van gorkom , j. h. 1992 , , 399 , 373 cen , r. , & ostriker , j. p. 1999
, , 519 , l109 cen , r. , & simcoe , r. a. 1997 , 483 , 8 cen , r. , tripp , t. m. , ostriker , j. p. , & jenkins , e. b. 2001 , , 559 , l5 charlton , j. c. 1995 , in qso absorption lines , ed .
g. meylan ( berlin : springer ) , 405 chen , h .- w . , 1998 , , 498 , 77 chen , h .- w . , 2001 , , 559 , 654 churchill , c. w. , 2000 , , 543 , 577 collins , j. a. , shull , j. m. , & giroux , m. l 2004 , , 605 , 216 ct , s. , 2002 , in the dynamics , structure , & history of galaxies , ed
g. s. da costa & h. jerjen , ( asp : provo ) , vol .
254 , 353 danforth , c. , & shull , j. m. 2004 , , in preparation ( also baas , vol .
36 , 679 , meeting 204 , # 15.03 ) dav , r. , hernquist , l. , katz , n. , & weinberg , d. h. 1999 , , 511 , 521 dav , r. , 2001 , , 552 , 473 dav , r. , & tripp , t. m. 2001 , , 553 , 528 dinshaw , n. , weymann , r. j. , impey , c. d. , foltz , c. b. , morris , s. l. , & ake , t. 1997 , , 491 , 45 fang , t .- t . , & bryan , g. l. 2001 , , 561 , l31 fang , t .-
t . , canizares , c. , 2002 , , 572 , l127 fox , a. j. , et al .
2004 , 602 , 738 fukugita , m. 2004 , in dark matter in galaxies , iau symposium 220 , sydney , 21 - 25 july , 2003 , astro - ph/0312517 gnedin , n. y. , & ostriker , j. p. 1997 , , 486 , 581 impey , c. d. , petry , c. e. , & flint , k. p. 1999 , , 524 , 536 janknecht , e. , baade , r. , reimers , d. 2002 , , 391 , 11 jannuzi , b. t. , 1998 , , 118 , 1 kim , t .- s . , 1997 , , 114 , 1 kim , t .- s . , cristiani , s. , & dodorico , s. 2001 , , 373 , 757 keeney , b. a. , stocke , j. t. , momjian , e. , carilli , c. l. , & tumlinson , j. 2005 , in preparation kirkman , d. , tytler , d. , suzuki , n. , omeara , j. m. , & lubin , d. 2003 , , 149 , 1 lanzetta , k. m. , 2002 , in extragalactic gas at low redshift , eds . j. mulchaey & j. stocke , ( provo : asp ) , vol .
254 , 78 lanzetta , k. m. , bowen , d. v. , tytler , d. , & webb , j. k. 1995 , , 442 , 538 lu , l. , wolfe , a. , & turnshek , d. 1991 , , 367 , 19 mathur , s. weinberg , d. h. , & chen , x. 2003 , in the igm / galaxy connection , eds . j. l. rosenberg & m. e. putman ( kluwer : dordrecht ) , 103 mckernan , b. , yaqoob , t. , mushotzky , r. , george , i. m. , & turner , t. j. 2003 , , 598 , l83 mclin , k. m. , stocke , j. t. , weymann , r. j. , penton , s. v. , & shull , j. m. 2002 , , 574 , l115 morris , s. l. , weymann , r. j. , savage , b. d. , & gilliland , r. l. 1991 , , 377 , l21 morris , s. l. , 1993 , , 419 , 524 mulchaey , j. s. , & stocke , j.t . 2002 , in extragalactic gas at low redshift ( provo : asp vol 254 ) nestor , d. b. , rao , s. m. , 2002 , in extragalactic gas at low redshift , ed
. j. mulchaey & j. stocke ( provo : asp ) , vol .
254 , 34 nicastro , f. 2003 in the igm / galaxy connection , eds . j. l. rosenberg & m. e. putman ( kluwer : dordrecht ) , 277 nicastro , f. , 2004 , , submitted penton , s. v. , stocke , j. t. , & shull , j. m. 2000a , , 130 , 121 ( paper i ) penton , s. v. , shull , j. m. , & stocke , j. t. 2000b , , 544 , 150 ( paper ii ) penton , s. v. , stocke , j. t. , & shull , j. m. 2002 , , 565 , 720 ( paper iii ) penton , s. v. , stocke , j. t. , & shull , j. m. 2004 , , 152 , 29 ( paper iv ) pettini , m. , ellison , s. l. , steidel , c. c. , & bowen , d. v. 1999 , , 510 , 576 prochaska , j. x. , & wolfe , a. m. 2000 , , 533 , l5 prochaska , j. x. , & wolfe , a. m. 2002 , , 566 , 68 putman , m. e. , rosenberg , j. l , ryan - weber , e. v. , & stocke , j. t. 2004 , , 204 , 7902 rao , s. m. , & briggs , f. 1993 , , 419 , 515 rao , s. m. , & turnshek , d. a. 1998 , , 500 , l11 rao , s. m. , & turnshek , d. a. 2000 , , 130 , 1 rao , s. m. , et al .
2003 , , 595 , 94 rasmussen , a. , kahn , s. m. , & paerels , f. 2003 , in the igm / galaxy connection , eds .
j. l. rosenberg & m. e. putman ( kluwer : dordrecht ) , 109 rauch , m. , et al .
1997 , , 489 , 7 richter , p. , savage , b. d. , tripp , t. m. , & sembach , k. r. 2004 , , 153 , 165 ricotti , m. , gnedin , n. y. , & shull , j. m. 2000 , , 534 , 41 rosenberg , j. l. , & schneider , s. e. 2003 , , 585 , 256 rosenberg , j. l. , ganguly , r. , giroux , m. l. , & stocke , j. t. 2003 , , 591 , 677 salpeter , e. e. 1993 , , 106 , 1265 salucci , p. , & persic , m. 1999 , , 309 , 923 savage , b. d. , sembach , k. r , tripp , t. m , & richter , p. 2002
, , 564 , 631 savage , b. d. , lehner , n. , wakker , b. p. , & sembach , k. r , & tripp , t. m. 2004 , , submitted savaglio , s. , 1999 , , 515 , l5 schaye , j. 2001 , , 559 , 507 schaye , j. , rauch , m. , sargent , w. l. w. , & kim , t - s .
2000 , , 541 , l1 sembach , k. r. , howk , j. c. , savage , b. d. , shull , j. m. , & oegerle , w. r. 2001 , , 561 , 573 sembach , k. r. 2002 , in extragalactic gas at low redshift , ed .
j. mulchaey & j. stocke ( provo : asp ) , vol 254 , 283 sembach , k. r. , tripp , t. m. , savage , b. d. , & richter , p. 2004
, , submitted shull , j. m. , stocke , j. t. , & penton , s. 1996 , , 111 , 72 shull , j. m. , 1998 , , 116 , 2094 shull , j. m. , 1999 , , 118 , 1450 shull , j. m. , 2000 , , 538 , l13 shull , j. m. , 2003 , , 594 , l107 shull , j. m. 2003 in the igm / galaxy connection , eds . j. l. rosenberg & m. e. putman ( kluwer : dordrecht ) , 1 spergel , d. n. , et al .
2003 , , 148 , 175 steidel , c. c. 1995 , in qso absorption lines , ed .
g. meylan ( garching : springer ) , 139 steidel , c. c. 1998 , in galaxy halos , ed .
d. zaritsky ( provo : asp ) , vol .
136 , 167 steidel , c. c. , et al .
2002 , , 570 , 526 stocke , j. t. 2002 in extragalactic gas at low redshift , ed .
j. mulchaey & j. stocke ( provo : asp ) , vol 254 , 98 stocke , j. t. , case , j. , donahue , m. , shull , j. m. , & snow , t. p. 1991
, , 374 , 72 stocke , j. t. , shull , j. m. , penton , s. v. , donahue , m. , & carilli , c. 1995 , , 451 , 24 stocke , j. t. , keeney , b. a. , mclin , k. m. , rosenberg , j. l. , weymann , r. j. & giroux , m. l. 2004 , , 609 , 94 stocke , j. t. , penton , s. v. , mclin , k. m. , shull , j. m. , & weymann , r. j. 2005 , in preparation telfer , r. , zheng , w. , kriss , g. a. , & davidsen , a. f. 2002 , , 565 , 773 tripp , t. m. , lu , l. , & savage , b. d. 1998 , , 508 , 200 tripp , t. m. , savage , b. d. , & jenkins , e. b. 2000 , , 534 , l1 tripp , t. m. , 2001 , , 563 , 724 tripp , t. m. , 2002 , , 575 , 697 tumlinson , j. , giroux , m. l. , shull , j. m. , & stocke , j. t. 1999 , , 118 , 2148 tumlinson , j. , shull , j. m. , giroux , m.l . , & stocke , j. t. 2004 , , submitted turnshek , d. a. , rao , s. m. , & nestor , d. b. 2002 , in extragalactic gas at low redshift , ed .
j. mulchaey & j. stocke ( provo : asp ) , vol .
254 , 42 van gorkom , j. h. , bahcall , j. n. , jannuzi , b. t. , & schneider , d. p. 1993
, , 106 , 2213 weymann , r. j. , 1998 , , 506 , 1 wolfe , a. m. , & prochaska , j. x. 2003 , , 545 , 603 zwaan , m. a. , briggs , f. , & sprayberry , d. 2001 , , 327 , 124 | in this review , we describe our surveys of low column density ( ) absorbers ( @xmath0 @xmath1 ) , which show that the warm photoionized igm contains @xmath230% of all baryons at @xmath3 .
this fraction is consistent with cosmological hydrodynamical simulations , which also predict that an additional 2040% of the baryons reside in much hotter 10@xmath4 k gas , the warm - hot igm ( whim ) .
the observed line density of absorbers , @xmath5 for @xmath6 2 , is dominated by low- systems that exhibit slower redshift evolution than those with @xmath7 2 .
hst / fuse surveys of o vi absorbers , together with recent detections of o vii with _
chandra _ and _ xmm / newton _ , suggest that anywhere from 2070% ( with large errors ) of the baryons could reside in the whim , for an assumed abundance o / h @xmath8 10% solar .
we also review the relationship between the various types of absorbers and galaxies . at the highest column densities , @xmath9 @xmath1 , the damped ( dla ) systems
are often identified with gas - rich disks of galaxies over a large range in luminosities ( 0.031 @xmath10 ) and morphologies .
lyman - limit systems ( @xmath11 @xmath1 ) appear to be associated with bound bright ( @xmath12 0.10.3 @xmath10 ) galaxy halos .
the absorbers with @xmath13 @xmath1 are associated with filaments of large - scale structure in the galaxy distribution , although some may arise in unbound winds from dwarf galaxies .
our discovery that @xmath14% of low-@xmath15 absorbers reside in galaxy voids suggests that a substantial fraction of baryons may be entirely unrelated to galaxies . in the future
, hst can play a crucial role in a precise accounting of the local baryons and the distribution of heavy elements in the igm .
these studies will be especially effective if nasa finds a way to install the _ cosmic origins spectrograph _
( cos ) on _ hubble _ , allowing an order - of - magnitude improvement in throughput and a comparable increase in our ability to study the igm . |
the aim to identify the symmetry of the superconducting phase in sr@xmath0ruo@xmath1 has stimulated numerous experiments since more than a decade @xcite .
most of them provide strong evidence that the pairing state has the so - called chiral @xmath2-wave symmetry .
this state is the analog of the a - phase of superfluid @xmath3he , which is a spin triplet state breaking time reversal symmetry with an angular moment along the fourfold @xmath4-axis of the tetragonal crystal lattice of sr@xmath0ruo@xmath1 .
the full gap function is a @xmath5-matrix in spin space , @xmath6 , and can be represented by the @xmath7-vector in the case of spin - triplet pairing : @xmath8 with @xmath9 as the gap magnitude .
squid interferometer experiments probing the internal phase structure of the cooper pairs are consistent with odd - parity ( @xmath2-wave ) pairing @xcite .
nmr knight shift measurements are compatible with equal - spin spin triplet pairing with the spin axis lying in the basal plane of the tetragonal crystal lattice , i.e. @xmath10 @xcite .
muon spin relaxation studies show enhanced internal magnetism in the superconducting phase suggesting a state with broken time reversal symmetry @xcite .
similarly , recent kerr effect studies imply the presence of an orbital magnetic moment pointing along the @xmath11-axis @xcite . on the other hand
, the search for the spontaneous magnetization at the surface of samples , as expected for a chiral @xmath2-wave state , by highly sensitive scanning probes has only given negative results so far @xcite .
a further extraordinary property of the chiral @xmath2-wave state is its two - fold degeneracy which can lead to domains , which are distinguished by the orientation of the orbital angular momentum of the cooper pairs , parallel or antiparallel to the c - axis .
early on the question of the formation of domains of the two states ( angular momentum up and down ) has been discussed , but no experimental indications of domains had been reported , until recently .
kidwingira et al investigated carefully the interference pattern of a josephson junction between sr@xmath0ruo@xmath1 and the conventional superconductor pb in a magnetic field @xcite .
they interpreted these patterns as the result of domain walls intersecting the extended josephson junction and so giving rise to a spatial variation of the josephson phase along the junction .
it has been speculated in the past that domain walls intersecting the interface between a chiral @xmath2-wave and conventional superconductor could influence the josephson effect , as the josephson current - phase relation could be different for the two types of domains @xcite .
such a property could lead to an intrinsically inhomogeneous junction and would alter the interference pattern .
this situation bears some similarity with the @xmath12 asymmetric interface in oriented films of high-@xmath13 superconductors where faceting of the boundary introduces a random switch of josephson phase by @xmath14 along the interface @xcite .
this type of boundaries display unusual interference patterns of the critical current in a magnetic field and generate spontaneous flux pattern on the interface .
unlike in the standard case the maximum of the critical current is here usually not located at zero field .
this finding could eventually provide indirect evidence for the presence of domain walls in the @xmath2-wave superconductor .
the discussion of this problem requires a detailed analysis of the domain wall structure . for interpretation of their results
kidwingira et al introduced several types of domain walls which could give rise to strong variations of the josephson phase @xcite . while they are possible on a topological level , not all the proposed domain walls are energetically stable . indeed in the most simple approach assuming full rotation symmetry for the fermi surface around the @xmath15-axis of the @xmath2-wave superconductor none of the stable domain walls give rise to any shift of the josephson phase and the presence of domain walls could go basically unnoticed in the interference experiment @xcite . however , as we will show below a more careful analysis taking more general conditions of the electronic spectrum into account leads to stable and metastable domain walls which can give rise to non - trivial intrinsic phase patterns in a josephson junction and reproduce the experimentally observed anomalies in the interference pattern .
our study on the josephson effect involves both a conventional @xmath16-wave and a chiral @xmath2-wave superconductor , representing the experimental arrangement of pb coupled to sr@xmath0ruo@xmath1 @xcite .
we first introduce here the basic order parameters and their corresponding ginzburg - landau theories which will be used later to discuss the structure of domain walls in the chiral @xmath2-wave state and their influence on the josephson effect in configurations , as shown in fig .
[ jj2 ] .
the conventional superconductor is described by a scalar order parameter @xmath17 for the spin - singlet pairing state of highest possible symmetry ( @xmath16-wave " pairing state ) .
the ginzburg - landau free energy functional has the standard form , @xmath18 = \int_{v_s } d^3r \biggr [ a_s(t ) \vert \psi\vert^2+b_s\vert\psi\vert^4+k_s\vert{\mbox{\boldmath$d$}}\psi\vert^2 + \frac{{{\mbox{\boldmath$b$}}}^2}{8\pi}\biggl ] \ ; , \label{fs}\ ] ] with @xmath19 as the gauge - invariant derivative and @xmath20 ( @xmath21 is the vector potential with the magnetic field @xmath22 ) , @xmath23 is the magnetic flux quantum , @xmath24 , @xmath25 and @xmath26 are parameters @xcite .
the chiral @xmath2-wave phase requires a two - component order parameter @xmath27 with @xmath28 which belongs to the irreducible representation @xmath29 of the tetragonal point group @xmath30 of sr@xmath0ruo@xmath1 .
the free energy functional has then the following general form @xmath31 = \displaystyle{\int_{v_p } } d^3r \biggr [ a_p(t ) |\bm{\eta}|^2 + b_1 | \bm{\eta}|^4 \nonumber\\ \\ \qquad + \dfrac{b_2}{2 } ( { { \eta_x}^*}^2 { \eta_y}^2 + { \eta_x}^2 { { \eta_y}^*}^2 ) + b_3 \vert \eta_x \vert^2 \vert \eta_y \vert^2 \nonumber\\ \\ \qquad + k_1 \bigr(\vert d_x \eta_x \vert^2 + \vert d_y \eta_y \vert^2\bigl ) + k_2 \bigr(\vert d_x \eta_y \vert^2 + \vert d_y \eta_x \vert^2\bigl ) \nonumber\\ \\ \qquad + \bigr\ { k_3 ( d_x\eta_x)^ * ( d_y\eta_y ) + k_4 ( d_x\eta_y)^ * ( d_y\eta_x)+c.c . \bigl\ } \nonumber\\ \\ \qquad + k_5 ( \vert d_z \eta_x \vert^2 + \vert
d_z \eta_y \vert^2 ) + \dfrac{{{\mbox{\boldmath$b$}}}^2}{8\pi } \biggl ] \;. \label{fp } \end{array}\ ] ] the coefficients @xmath32 and @xmath33 are material - dependent parameters and @xmath34 @xcite . in order to stabilize the chiral @xmath2-wave state in the bulk the coefficients have to satisfy the relations : @xmath35 , @xmath36 and @xmath37 . minimizing the free energy functional with respect to @xmath38 in the homogeneous case
, we obtain the uniform phase=(-4b_1+b_2-b_3)\eta_0 ^ 4=a_p\eta_0 ^ 2=-h_c^2/8\pi$ ] , which defines the critical magnetic field @xmath39 . ]
@xmath40 such that the corresponding gap function is @xmath41 the two states with opposite relative sign are degenerate and violate time reversal symmetry , as the operation of time reversal @xmath42 yields @xmath43 .
these two states can form domains . in the following
we will first analyze the structure of domain walls between such domains of opposite chirality and their influence on the josephson effect in geometries as shown in fig.[jj2 ] .
-axis between the two degenerate states : @xmath44 .
the dimensions of the junction are @xmath45 . ]
before addressing the influence of the domain wall on the josephson effect it is necessary to analyze here the structure of a domain wall in detail . our aim is to show which types of domain walls are energetically favorable .
it will turn out that the anisotropy of the fermi surface plays an important role and with this also the orientation of the domain wall . for a convenient discussion it will be advantageous to change the order parameter representation of the @xmath2-wave superconductor and to formulate the ginzburg - landau functional also in rotated coordinate frames . in a first state
we express now the free energy functional in a new form which simplifies the rotation of the coordinate frame .
we then always define the domain wall as the @xmath15-@xmath46-plane and the spatial variation of the order parameter occurs only along the corresponding @xmath47-axis perpendicular to this plane . for simplicity we keep the @xmath15-axis as fixed along the crystaline @xmath15-axis .
+ we introduce the parametrization @xcite , @xmath48 which inserted into the free energy leads to @xmath49 = \displaystyle \int_{v_p } d^3r \biggr [ 2a_p ( |\eta_+|^2+|\eta_-|^2 ) \nonumber \\ \\ \qquad+ b\bigr\ { |\eta_+|^4+|\eta_-|^4 + 4\vert\eta_+\vert^2\vert\eta_-\vert^2 \nonumber \\
\\ \qquad + \nu ( { { \eta_+}^*}^2 { \eta_-}^2 + { \eta_+}^2 { { \eta_-}^*}^2)\bigl\ } \nonumber \\ \\
\qquad+ k \bigr\ { \vert { \mbox{\boldmath$d$}}\eta_+ \vert^2 + \vert { \mbox{\boldmath$d$}}\eta_- \vert^2 -\dfrac{1}{2 } \bigr ( \nu ( d_- \eta_+)^ * ( d_+ \eta_- ) \nonumber \\ \\ \qquad + ( d_+ \eta_+)^ * ( d_- \eta_-)+c.c.\bigl ) \bigl\ } + \dfrac{1}{8\pi } ( { { \mbox{\boldmath$\nabla$}}}\times { { \mbox{\boldmath$a$}}})^2\biggl ] \ ; .
\end{array } \label{fpm}\ ] ] in a weak - coupling approach the parameter @xmath50 is a measure for the anisotropy of the fermi surface ( @xmath51 for a cylindrical fermi surface and @xmath52 for a square - shaped fermi surface ) where @xmath53 are the components of the fermi velocity and @xmath54 defines the average over the fermi surface . the coefficients in eq.([fp ] ) satisfy the following relations @xmath55 where @xmath56 and @xmath57 are material dependent parameters .
this form of the free energy functional allows us now to deal easily with the rotation of the reference frame around the @xmath15-axis : @xmath58 with the angle @xmath59 relative to the original @xmath46-axis in the tetragonal crystal . for the new coordinate frame ,
the order parameter and the gradients are transformed as @xmath60 and @xmath61 .
when we express ( [ fpm ] ) in the new coordinates , we have only to modify the two terms containing the parameter @xmath62 by phase factors , @xmath63 in this way we can use the angle @xmath59 to define the coordinate frame . in the following we will omit the primes and always assume that we describe the domain wall in the corresponding frame .
we now turn to the structure of the domain wall which will depend qualitatively on the choice of parameters in the free energy .
we choose here a variational approach to discuss behavior of the order parameter around the domain wall with the following ansatz which is most useful to eventually obtain the key information relevant for the josephson effect , @xmath64 with the boundary conditions within a given reference frame , @xmath65 where we restrict the spatial dependence of the order parameter to the function @xmath66 and use the relative phase @xmath67 as a further ( variational ) parameter . in this way the domain wall appears as an interface between the two superconducting domains which is additionally characterized by a phase difference @xmath68 , similar to a josephson junction or weak link @xcite . for the later discussion of the josephson effect with an @xmath16-wave superconductor
, it will be useful to return to the order parameter components @xmath69 which are expressed as @xmath70 the chosen boundary conditions in ( [ bc ] ) correspond to the situation , @xmath71 for which , in the parametrization @xmath72 , the phase shifts of the order parameters are given by @xmath73 here @xmath74 plays the role of the total phase difference of the order parameter and also determines the current flow through the domain wall analogous to a josephson junction .
with this variational ansatz we rewrite the free energy , @xmath75 \ ; , \nonumber \\
\end{array}\ ] ] where we substitute @xmath76 using ( [ eta0 ] ) and ( [ bk ] ) .
assuming homogeneity along the @xmath15- and @xmath46-axis , we take @xmath77 and @xmath78 with @xmath79 , the free energy can be written as @xmath80 \ ; , \end{array } \label{faxy}\ ] ] where @xmath81 defines the coherence length .
additionally we introduced @xmath82 which are the only coefficients depending on the anisotropy parameter @xmath83 , the phase @xmath68 and the angle @xmath59 of the domain wall orientation relative to the crystalline main axis ( crystal @xmath46-axis ) .
certain symmetries become immediately obvious here .
the free energy is invariant under the rotation @xmath84 and the operation @xmath85 leaves the coefficients unchanged .
it is also clear that the phase @xmath68 is closely linked to the orientation of the domain wall .
a further aspect of symmetry is connected with the operation @xmath86 , leading to @xmath87 , @xmath88 and @xmath89 .
the variational minimization of the free energy functional with respect to @xmath90 , @xmath91 and @xmath92 leads to the corresponding three equations : @xmath93\ ; , \nonumber\\ \\
\kappa^2\xi_0 ^ 2\gamma ~\partial_y^2 a_x - \gamma ( 1-c_+ \sin 2\chi ) a_x = s_- \gamma a_y\sin 2\chi \nonumber\\ \\ \qquad - c_- \partial_y\chi \ ; , \nonumber\\ \\ a_y = -\dfrac{s_+ \partial_y\chi + s_- \gamma a_x\sin 2\chi } { \gamma(1+c_+ \sin 2\chi)}\;. \end{array } \label{min - equ}\ ] ] for a concise notation we introduce the ginzburg - landau parameter @xmath94 with @xmath95^{1/2}$ ] as the london penetration length ( we take @xmath96 ) .
the third equation leads to @xmath97 under @xmath98 . as a consequence
we observe that the free energy is invariant under the operation @xmath99 , which will be important again later in the analysis of the josephson effect .
these equations will now be solved numerically , although we will also present in [ aa ] an approximate analytical solution . for this purpose
it is useful to turn to the dimensionless variables , measuring lengths in units of @xmath100 , @xmath101 , and using for the vector potential @xmath102 .
the free energy per unit area of the domain wall can then be written as @xmath103 \;. \end{array } \label{fdw}\ ] ] the numerical results for the spatial dependence for the order parameter and the vector potential derived from @xmath90 , @xmath104 and @xmath105 minimizing @xmath106 are shown in [ sol_eta_phi ] . for rotations of the domain wall by the angles @xmath107 ( solid line ) , @xmath108 ( dashed line ) and @xmath109 ( dotted line ) .
the anisotropy parameter is chosen @xmath110 . ] : @xmath111 , as a function of the angle of the domain wall and for @xmath110 . ] . ] in fig .
[ sola ] we show the domain wall energy per unit area as a function of @xmath68 and @xmath59 .
we do not use here @xmath68 as a variational parameter but search for the variational local minimum for given @xmath68 .
for this calculations we choose the anisotropy parameter to be negative , @xmath112 .
note that the result for @xmath113 follows immediately from the symmetry relation , @xmath114 .
for @xmath115 we observe two degenerate minima of @xmath116 at @xmath117 corresponding to the stable domain wall configuration for given @xmath59 and @xmath83 . for any other angle @xmath59 this degeneracy
is lifted leading to a stable and metastable minimum , which are located at @xmath118 and @xmath119 , respectively , if @xmath120 .
the energy of the stable and metastable domain wall configuration depends on the angle @xmath59 , as can be seen in fig .
[ fmin_nu05 ] where we plot @xmath121 for @xmath110 ( the solid line marks the stable and the dashed line the metastable states ) .
we find that a minimum of this energy occurs at an angle @xmath122 away from @xmath120 .
analogously by symmetry such a minimum is found at @xmath123 for @xmath113 . from this
we conclude that there are special orientations for the domain wall which are energetically favorable and depend on the anisotropy properties of the superconductor .
these special orientations need not to lie along symmetry axes or symmetry planes .
note that also the angle with respect to the @xmath15-axis is important in this respect , since domain walls parallel to @xmath46-@xmath47-plane are probably most stable .
however , here we consider only the case of domain walls parallel to the @xmath15-axis , as they are generally most important for the modification of the josephson effect . as a reference we consider also the case of an isotropic fermi surface ( @xmath51 ) which
naturally does not show any dependence on the angle @xmath59 .
we find that the most stable domain wall state corresponds to the phase @xmath124 which agrees with the result obtained from a corresponding microscopic model calculation based on a quasi - classical approach @xcite . as a function of the anisotropy of the fermi surface @xmath62 , for a domain wall parallel to the crystal @xmath46-axis , i.e. @xmath107 .
the solid lines correspond to stable states while the dashed line denotes metastable values @xmath125 . ] in fig .
[ dtheta0 ] we show the behavior of @xmath126 as a function of @xmath83 for the angle @xmath127 .
there are two obvious regions ; for @xmath128 the minimum corresponds to @xmath129 and for @xmath130 we find two degenerate values .
in addition metastable states ( indicated as dashed lines ) appear at @xmath125 for @xmath131 . in view of our later discussion of the josephson junctions intersected by domain walls we now consider the situation of a domain wall ending at the surface of the superconductor and being pinned at a defect somewhere in the bulk ( for illustration we simplify our model introducing a columnar pinning center along the @xmath15-axis ) .
we now search for the possible domain wall configurations .
the domain wall has to compromise between being as short as possible and as close as possible to the orientation ( angle @xmath132 ) minimizing its wall energy whereby , in principle , both stable and metastable ( local minimum ) situation can play a role .
we restrict ourselves here to the case of @xmath133 which is relevant for sr@xmath0ruo@xmath1 @xcite .
let us start with the situation that the surface normal vector @xmath134 is assumed to be along a crystal main axis in basal plane , say the @xmath46-axis .
the columnar defect along the @xmath15-axis is located at a distance @xmath135 from the surface ( fig .
[ dwjump ] ) . the domain wall energy per unit length in @xmath15-direction for a phase @xmath68 and an angle @xmath59 relative to the @xmath46-axis
is given by @xmath136 with @xmath137 defined in eq.([fdw ] ) .
this is a symmetric function under @xmath138 and plotting @xmath139 we find two stable situations with @xmath140 and @xmath141 .
the two configurations are depicted in fig.[dwjump ] .
we show in fig .
[ fminc_nu05 ] the corresponding domain wall energies of the stable and metastable branch given by the solid line and the dashed line respectively .
the lowest energy contributions correspond to the two rather shallow minima of the stable branch .
if the surface normal vector does not lie along a main axis ( but within the @xmath46-@xmath47-plane ) then the situation becomes more complex .
while in the previous case two degenerate stable domain wall states were found , under general conditions only two local minima of the domain wall energy exist which are not degenerate . in fig .
[ fmin_nu05m_bar02 ] we show the domain wall energy for an angle @xmath142 of the normal vector @xmath134 relative to the @xmath143-axis . because the domain walls for given orientation possesses phases @xmath68 for stable and metastable energy minima , there are two branches of fig .
[ fmin_nu05m_bar02 ] which correspond to possible domain wall configuration at the surface .
these configurations are neither degenerate nor symmetric unlike for the case of @xmath144 .
nevertheless , they represent domain wall states which are , in principle , accessible depending on the history of the system as there are two local energy minima in fig .
[ fmin_nu05m_bar02 ] , a stable ( global minimum ) and a metastable ( local minimum ) . , for @xmath110 . ]
, when the surface normal vector makes an angle @xmath145 with the crystal @xmath46-axis ( @xmath146 ) . ]
the symmetry aspects of the josephson coupling between a conventional @xmath16-wave and an odd - parity superconductor have been discussed many years ago @xcite . considering the standard lowest - order josephson coupling we are confronted with the apparent two - fold obstacle that there is a mismatch in the spin configuration ( singlet versus triplet ) as well as in parity of the orbital part of the two pair wave functions .
it has been shown , however , that spin - orbit coupling and the reduced inversion symmetry at the interface are sufficient to yield a finite coupling .
the lowest - order josephson coupling for an interface with normal vector @xmath147 and the @xmath16-wave order parameter @xmath148 and the @xmath2-wave order parameter @xmath149 has then the form @xcite , @xmath150 if the @xmath7-vector is parallel to @xmath15-axis as for the chiral @xmath2-wave state . here
@xmath151 denotes the coupling strength and the integral runs over the interface . unlike the coupling between two singlet - superconductors , which can be estimated through the experimentally determined normal state tunneling conductance , in our case the difference in the pseudo - spinors yields spin - active ( spin - flip ) tunneling processes which depend on the spin - orbit coupling of the two materials . for both pb and sr@xmath0ruo@xmath1 spin - orbit coupling
is not small , such that it can be expected that the josephson coupling is comparable to an ordinary , although a reliable estimate is not easy due to the complex band structure . using @xmath152 as a boundary term in the ginzburg - landau equations we can derive the following josephson current - phase relation of the junction : @xmath153 \;. \end{array}\ ] ] we assume @xmath154 such that the coupling reduces to @xmath155 with @xmath156 as the phase difference between the order parameters of the p - wave and the s - wave superconductor . in this geometry
only the @xmath157-component of the @xmath158-wave side order parameter contributes to the josephson coupling .
this is the order parameter component of the pairing state whose nodes point towards the interface .
only this component can combine with the spin to conserve the total angular momentum of the cooper pair in the tunneling through the interface .
we want to calculate the josephson critical current assuming that several domain walls intersect the josephson junction from the sr@xmath0ruo@xmath1 side . in the following we model the effect of the domain wall as a step - like phase shift at the point of intersection .
this approach is justified , if we consider the josephson coupling as weak such that the currents are small and the josephson penetration depth @xmath159 defined below is longer than the extension of the junction and even much longer than the width of the domain walls @xcite .
analogous to ordinary josephson junctions we find also in junctions between an @xmath16- and a chiral @xmath2-wave superconductor a relation between the derivative of the phase @xmath156 with respect to the coordinate along the contact and the magnetic field threading perpendicularly .
thus with the assumption that the junction normal vector is @xmath160 , we examine the variation of the phase @xmath161 and magnetic field @xmath162 along the @xmath163-direction and obtain , @xmath164 where the effective width of the josephson contact ( parallel to @xmath147 ) @xmath165 includes the real width @xmath166 and the london penetration depths on both sides , @xmath167 ( for pb @xmath168 and for sr@xmath0ruo@xmath1 @xmath169 ) .
in contrast to the standard case we add here a contribution @xmath170 which corresponds to the intrinsic phase variation on the @xmath158-side which , for example , can be induced by a domain wall intersecting the josephson contact , as we will discuss below .
however , also faceting of the interface can introduce such contributions .
the spatial variation of the phase @xmath171 obeys the extended sine - gordon equation , @xmath172 with josephson penetration depth @xmath173 .
assuming now that @xmath159 is larger than the extension @xmath174 of the josephson junction along the @xmath47-direction , we find that , for a uniform external field @xmath175 , @xmath161 is given approximately by @xmath176 where @xmath177 and @xmath178 is an integration constant .
the total current is then @xmath179 with @xmath180 the extension of the junction along the @xmath15-direction .
this allows us now to discuss the effect of domain wall on the interference pattern in the maximal josephson current obtained by maximizing @xmath181 with respect to @xmath182 . before addressing the interference pattern
it is necessary to determine the change of the phase @xmath184 as we pass through a domain wall along the interface of the josephson junction .
this phase enters eq .
( [ phase - grad ] ) through the relation @xmath185 on the side of the @xmath2-wave superconductor .
consider now a single domain wall which reaches the junction interface at a @xmath186-angle , i.e. @xmath120 . the phase shift @xmath187 through the domain wall can be determined by integrating along the @xmath163-axis , @xmath188 where @xmath189 has been calculated in eq.([min - equ ] ) .
the plus ( minus ) sign of the phase @xmath68 reflects the two possible situations for the domain wall : moving along the @xmath47-axis we cross a domain wall from @xmath190 ( @xmath191-sign ) or vice versa ( @xmath192-sign ) .
since the two situations are related through the transformation @xmath193 it is clear that the phase shift @xmath194 is identical in absolute magnitude for the two cases .
in general , this phase difference depends also on the angle of intersection @xmath195 , the angle between interface normal vector @xmath134 ( in @xmath46-@xmath47-plane ) and domain wall .
a rotation by an angle @xmath195 involves a phase shift @xmath196 for @xmath197 .
first we consider the case @xmath198 for which @xmath199 , if @xmath134 is parallel to the domain wall .
rotating @xmath134 by the angle @xmath200 , we obtain @xmath201 .
thus the phase shift through the domain wall is then given by @xmath202 where @xmath203 and @xmath204 are solutions of eq.([min - equ ] ) for @xmath205 .
for an interface with @xmath206 the most stable states are characterized by the two orientations of the domain wall @xmath207 , according to the previous discussion . by symmetry , we know that the free energy is invariant under the transformation @xmath208 . changing from @xmath209 to @xmath210 and keeping the same boundary conditions ( @xmath211 ) , the phase shift simply changes sign @xmath212 similarly , going from the configuration @xmath213 to @xmath214 ( keeping the same angle @xmath209 ) , i.e. under a time reversal transformation , only the sign of the phase shift is changed @xmath215 where we have introduced the notation @xmath216 and @xmath217 .
we note that this is in agreement with the fact that the later transformation is equivalent to an inversion of the boundary conditions done by the transformation @xmath218 , which causes a sign change for the phase difference . therefore , to each domain wall intersecting the josephson junction
we can associate a phase difference @xmath219 ( resp .
@xmath220 ) corresponding to the geometry @xmath221 ( resp .
@xmath222 ) which depends on the orientation of the domain wall . for orientation of the interface normal vector different from @xmath206
, we compare the two lowest energy configurations of the domain wall ( stable and metastable ) state . using the above calculation scheme we can show that the phase differences are in general different for the two states , yielding @xmath223 and @xmath224 . under time
reversal operation , changing between the two cases @xmath225 and @xmath226 , we obtain the opposite sign of the phase shifts again .
note that , in general , the two phase shifts , @xmath223 and @xmath224 are comparable in magnitude with those cases where @xmath134 lies along a symmetry axis .
note , however , that there is an energy difference between the two domain wall states , if they are not degenerate .
this energy difference depends on the depth over which the domain wall is influenced by the change of configuration at the interface .
we extend now our model to a josephson junction which is intersected by several domain walls .
first consider the case of one domain wall reaching the interface at the position @xmath227 , which yields the josephson phase , @xmath228 here @xmath229 denotes the phase shift at the domain wall as calculated in the previous section .
the index @xmath230 labels the two types of domain wall state ( stable or metastable minima ) .
it is admissible to use a step function @xmath231 to describe the spatial change of the phase , since the extension of the domain wall is small compared to the length of the josephson junction and the josephson penetration depth @xmath159 .
this leads to a piecewise constant phase shift .
the generalization to an array of domain walls is straightforward . for @xmath232 successive domain walls , the josephson phase difference across the junction
is given by @xmath233 where @xmath171 depends on the configuration @xmath234 of all domain walls .
we will now show that the distortion of the josephson interference pattern through @xmath171 depends on these configurations .
the total josephson current which traverses the junction is given by eq.([current-1 ] ) .
using our piece - wise constant approximation for @xmath184 we obtain readily the following expression , @xmath235 with @xmath236 , @xmath237 , @xmath238 and @xmath239 ( @xmath240 being the interface area ) , and @xmath241 .
the josephson critical current , i.e. the maximal supercurrent possible for this configuration , is obtained by the maximization of @xmath242 with respect to @xmath182 .
this condition reads @xmath243 leading to @xmath244 with @xmath245 the josephson current can be written as @xmath246 and finally we find @xmath247 in fig.[imax_06 m ] and fig.[imax_06m2 ] we show two simulations for interference pattern of the josephson critical current in a magnetic field , using two random domain wall configurations ( indicated in the inserted panels ) where we assumed @xmath248 for the number of intersecting domain walls .
the randomness occurs in the positions of domain walls as well as in the sequence of @xmath249 .
the deviation from the standard fraunhofer pattern is obvious .
however , the maximum of the josephson current lies rather close to @xmath250 .
this feature has to do with the fact that , in the case of generic random configurations , the phase shift does not vary much overall , i.e. @xmath251 . stronger deviations can be observed , if we bias the domain wall configuration in a way as to have a larger net shift , e.g. by assuming for all domain walls the same sign of @xmath187 . .
the anisotropy parameter is chosen @xmath252 @xcite . ]
@xcite . ] an important feature supporting the idea that domain walls are involved in producing the irregular interference pattern is the observation of hysteresis effects by kidwingira _
et al _ , when the external magnetic field was cycled between positive and negative maximal fields @xcite .
they argued that the applied magnetic field induces a rearrangement of the domain walls in the sample , and they substantiated their claim by simulations looking at the effect of shifted domain wall positions . since the domain walls are pinned at defects of the sample large shifts in positions are rather unlikely .
hysteresis effects are rather easily discussed within our simplified model in eq.([model-1 ] , [ model-2 ] , [ model-3 ] ) .
the free energy of the junction in a magnetic field can be approximated by @xmath253 we neglect the change of the intersection points @xmath254 for different configurations @xmath255 and examine the condition to minimize the free energy .
moreover we assume that the interface is smooth and that we have the same two essentially degenerate configurations for all domain walls : @xmath256 and @xmath257 .
thus , @xmath134 is assumed to lie very close to a high - symmetry axis of sr@xmath0ruo@xmath1 , say @xmath258 . for values
@xmath259 the free energy in eq.([fen ] ) can be lowered by choosing @xmath260 , because under this condition @xmath261 note that there are special values of @xmath262 where the full polarization of @xmath187 may not be the best choice . however , considering a sweep of the field to maximal value @xmath263 and back would favor eq.([polar - phi ] ) .
the same argument can be used for negative fluxes , driving the domain walls to adopt @xmath264 . note that for metastable domain wall states a polarization would only be possible , if the energy expense of the metastable configurations is sufficiently small .
we consider now geometries for which this is true .
starting at zero field , the domain walls shall be essentially randomly configured .
such a situation would lead to a critical current pattern as shown in fig .
[ imax_06m2 ] .
if the magnetic field is increased up to a sufficiently high value , then the domain walls would likely polarize after a certain waiting time .
when we decrease now the field to measure the interference pattern , we observe the interference pattern modified by the polarized domain walls .
after reaching a sufficiently large negative field value , the domain walls polarize in the opposite way .
therefore , tuning the field back towards positive values we find an altered interference pattern of the critical current .
simulation results taking the two ( polarized ) domain wall configurations into account are shown in fig.[imax_06mrl ] .
the most striking feature is the shift of the maximum of the critical current .
these curves are obtained assuming that changes of domain wall configurations during the field sweep , when the critical currents is measured , can be neglected .
reorganizations of the domain wall configurations during the field sweep would most likely lead to discontinuities in the critical current .
this kind of behavior is also observed in a set of measurements by kidwingira et al ( fig.3b in @xcite ) where the field sweeping range is restricted to rather small fields only .
also in this case a hysteretic behavior of critical current ( dependence on the field sweep direction ) was observed , though less pronounced .
it is not unlikely that some part of the noise on these data can be interpreted as an effect due to the domain walls . turning back to the hysteresis effect with polarized domain wall configurations
, we might use the shift of the position of the maximal critical current in order to estimate the density of domain walls .
we denote the flux of the maximal current as @xmath265 which can be determined approximately by the condition @xmath266 assuming for the last equality a completely polarized domain wall configuration where all domain walls contribute the same phase shift @xmath187 .
first , we conclude that the maximum lies at @xmath267 for a sweep down from the positive field side ( opposite for the opposite sweep direction ) in accordance with experimental findings @xcite . the number of domain walls intersecting the josephson junction is then given by @xmath268 kidwingira et al find in a measurement of hysteresis effect the magnetic field @xmath269 ( fig.3a in @xcite ) which yields a flux @xmath270 .
for this estimate of @xmath265 we took @xmath271 , @xmath272 , @xmath273 and @xmath274 .
( [ deltaphi_definition ] ) with @xmath252 we find @xmath194 in the range of @xmath275 - @xmath276 depending on the orientation of the domain wall with respect to the interface . from this
we obtain @xmath277 - @xmath278 which leads to a mean distance between domain walls of @xmath279 - @xmath280 .
kidwingira et al also report the presence of peculiar noise in the time dependence of the voltage drop of the junction in a constant current slightly above the critical current @xcite .
the question arises whether this feature can also be attributed to the dynamics of domain walls .
as we have seen above , the modification of the domain wall configuration @xmath281 changes the critical current .
the current - voltage characteristics of a josephson junction for currents @xmath181 immediately above the critical current @xmath282 is very non - linear .
the standard textbook form of the current - voltage relation of a josephson junction gives usually a good approximation of the general behavior @xmath283 for @xmath284 .
for a given current @xmath181 the time dependence of the critical current , @xmath285 leads to @xmath286 which can be large for @xmath181 close to @xmath287 .
the reported noise feature ( fig.3d of @xcite ) suggest that the critical current fluctuates between two values , which could correspond to two domain wall configurations ( see .
fig.[iv ] ) . and @xmath288 which for the same applied current lead to different values of the voltage on the junction , @xmath289 and @xmath290 , respectively .
a time sequence of changing between the two critical current values ( by change of domain wall configurations ) gives rise to steplike noise on the voltage signal . ]
our theoretical discussion of the interference pattern did not include the faceting of the interface between the @xmath16-wave superconductor and sr@xmath0ruo@xmath1 .
the effect of faceting is two - fold .
first , a faceted surface can lead to additional stable and metastable domain wall configurations as domain walls can be pinned at surface inhomogeneities .
thus , they add to the number of possible junction states .
second , faceting corresponds to a varying interface normal vector .
it is easy to see that the modulation of the inplane normal vector angle @xmath59 by @xmath291 corresponds directly to the josephson phase variation @xmath292 ( @xmath293 ) , if the length scale of the faceting is larger than the coherence length of sr@xmath0ruo@xmath1 ( @xmath294 ) . as a random phase modulation of this kind with @xmath295
does not lead to a significant shift of the maximal critical current in the interference pattern and no hysteretic effect is possible , faceting alone can not be responsible for the features observed . together with a spatial variation of the josephson coupling the variation of the phase due to faceting can contribute to the static random structure of the interference pattern .
fine structures in the interference pattern ( variations of the critical current on small field changes ) are usually due to variation on long length scales along the junction and are also known for conventional josephson contacts . note that the effect of faceting in @xmath296-wave high - temperature superconductors has a stronger impact on the interference pattern , since the phase jumps between 0 and @xmath297 can lead to phase wandering along the junction such that @xmath298 and produces a field shift for the maximal josephson current , as observed in experiment @xcite .
for completeness we address here a further interesting feature connected with a domain wall intersecting the interface .
this time , however , we change to limit of @xmath299 being much shorter than the extension of the junction . in order to describe the behavior of the phase @xmath161
, we have then to solve the sine - gordon equation @xmath300 the solution here has a kink shape which is slightly modified by the last term of the right - hand side .
this term acts like a source term in the vicinity of the domain wall . without explicitly writing down the solution
it is clear that within the length scale of @xmath299 from the domain wall @xmath171 approaches a constant value . as discussed earlier the derivative @xmath301 corresponds to a local magnetic field , such that here a well - localized magnetic flux line appears at the position where the domain wall meets the interface .
the enclosed magnetic flux can be easily determined by @xmath302dy\nonumber\\ & = & \int_{-l/2}^{l/2}\frac{\phi_0}{2\pi}\left(\partial_y\varphi(y)-\partial_y\phi(y)\right)dy\nonumber\\ & \approx&\frac{\phi_0}{2\pi}n2\pi-\frac{\phi_0}{2\pi}\delta\phi \;,\end{aligned}\ ] ] where @xmath303 is a positive integer .
we finally see that the minimal possible fluxes is given by @xmath304 since @xmath305 , we encounter here fractional vortices which are generally a sign of broken time reversal symmetry @xcite .
the detection of such well - localized fluxes could be used to detect the position of domain walls .
motivated by the experiments of kidwingira et al . on josephson junctions between sr@xmath0ruo@xmath1 and the conventional superconductor pb @xcite , we studied effect of domain walls on the josephson interference effect in a magnetic field assuming that sr@xmath0ruo@xmath1 is a chiral @xmath2-wave superconductor . for this purpose
we analyzed the domain wall structure and showed that its internal phase structure is crucial for the josephson effect , if domain walls intersect the josephson junction .
the anisotropy of the electronic band structure plays an important role , as it influences the phase shift of the superconducting order parameter between the two types of chiral domains .
in addition , it determines the energetically most favorable orientation of the domain wall .
the theory presented here is able to explain the most important features reported by kidwingira et al @xcite .
( 1 ) the interference pattern in a single junction deviates from the the standard fraunhofer pattern not only through irregularities , but also shows a distinct asymmetry between positive and negative magnetic fields and the maximum of the critical current can be shifted away from zero field .
this can be attributed to the phase shifts introduced by domain walls intersecting the interface between sr@xmath0ruo@xmath1 and the conventional superconductor .
( 2 ) kidwingira et al report strong differences between the interference pattern of different samples .
this may be explained by the fact that the samples were prepared in different ways , with possibly also different normal vector directions as well as different degrees of faceting .
( 3 ) cycling the field continuously covering a positive and negative field range , a hysteretic behavior in the interference pattern appears which can be understood as a field - driven motion of the domain walls , whereby the anisotropy of the domain wall energy could play an important role , as our model shows . allowing for the change of domain wall configurations in time we can also understand the noise effect seen in the current voltage measurements for currents above the critical current .
the question arises whether it would be possible to test the domain wall scenario by generating a single - domain superconducting phase in sr@xmath0ruo@xmath1 .
this would be most straightforwardly realized by field - cooling .
unfortunately , this procedure might lead to vortex trapping in both superconductors which would jeopardize a clear outcome . domains naturally arise in the zero - field cooling process , as superconductivity nucleates in the sample with some extent of inhomogeneity , since the chiral @xmath2-wave state is rather susceptible to disorder effects . in this way
both chiral states can emerge in the sample in different regions and the domain walls appearing between them are eventually pinned in the sample , making it difficult to anneal the sample to a domain wall free phase .
it is difficult to estimate a priori the density of domain walls .
however , our analysis suggests that domain walls intersecting the josephson junctions in the experiment may be separated by a few micrometers which is considerably larger than the extension of the domain walls .
this estimate may be interesting in the context of recent measurements by kirtley et al aiming at the observation of spontaneous magnetic flux at surfaces of sr@xmath0ruo@xmath1 , as expected for chiral @xmath2-wave states @xcite .
since these experiments gave a negative result for the presence of such fluxes , it was speculated that the dense population of chiral domains would cancel out the signal for the scanning squid microscope used .
our estimate of domain wall density corresponds roughly to the spatial resolution of the squid microscope .
thus it remains unclear whether domain walls really could explain the absence of a positive signal from spontaneous flux .
extensive theoretical studies on this issue point to an important puzzle in this context @xcite . eventually , we would like to note that this experiment does not decide between even- or odd - parity pairing .
while the experimental situation satisfies the selection rules for the coupling between the chiral @xmath2-wave and an @xmath16-wave superconductor , the effects observed in the josephson interference experiments would be similar for a chiral @xmath296-wave superconductor such as the @xmath306-wave state belonging to @xmath307-representation of the tetragonal point group @xmath30 , for which , however , additional complications appear due to the nodal gap structure . at the present stage ,
however , we do not have any experimental signature in ref.@xcite , which would rule out the @xmath308-state .
we are grateful to d.f .
agterberg , y. asano , y. maeno , m. matsumoto , t.m .
rice , y. tanaka and d. van harlingen for many enlightening discussions .
this work has been financially supported by the swiss nationalfonds and the nccr manep .
we start from the free energy functional ( [ faxy ] ) where we neglect the terms in @xmath91 .
varying with respect to @xmath90 and @xmath92 gives , respectively , @xmath309 \;,\\ \label{eqa } a_y&=&-\frac{s_+\partial_y\chi}{\gamma(1+c_+\sin2\chi ) } \;.\end{aligned}\ ] ] following the argument of ref.@xcite we neglect the contribution of the third term of ( [ eqchi ] ) and the @xmath310 term in the denominator of @xmath92 in ( [ eqa ] ) .
this leads to the simplified equation for @xmath311 , @xmath312 where @xmath313 $ ] .
this has the following kink solution @xmath314 which we use to determine @xmath92 . inserting @xmath90 and @xmath189 into the free energy functional we derive an analytical form of the variational domain wall energy per unit area , @xmath315 \biggr\ } \;.\end{aligned}\ ] ] in fig .
[ functionala ] we show the domain wall energy as a function of @xmath74 for the different angles @xmath316 , with the anisotropy parameter chosen @xmath317 .
we see that those results are qualitatively and even quantitatively close to the numerical results plotted in fig.[sola ] . in fig .
[ functional0 ] we plot the free energy density at @xmath51 . here
some short - comings of the above approximation becomes obvious , since there a two minima of the energy in contrast to the single one of the numerical result shown in fig .
this discrepancy originates from neglecting the vector potential @xmath91 and the corresponding terms in the free energy . as a function of the phase difference
@xmath318 when the angle of the domain wall with respect to the crystal @xmath46-axis are : @xmath319 and @xmath109 . ]
we show here the numerical solutions of the structure of the domain wall in the most stable state for @xmath252 and compare with the analytical approximation .
we plot in fig .
[ chian ] the function @xmath320 where the solid line is the numerical solution and the dot - dashed line the approximative solution in eq .
( [ chi - approx ] ) .
the deviation is rather small . in fig .
[ a_xy06 m ] we show the two components of the vector potential , the solid line for the numerical and the dot - dashed line for the approximate solution . here
the discrepancy is larger .
[ eta06 m ] and [ phi06 m ] depict the modulus of both order parameter components , @xmath321 and @xmath322 and the phase of the order parameter passing through the domain wall ( we choose @xmath323 ) . obviously , the two phase differences follow the relation @xmath324 and @xmath325
. 18 f. kidwingira , j.d . strand , d.j .
van harlingen and y. maeno , science * 314 * ( 2006 ) , 1267 .
mackenzie and y. maeno , rev .
* 75 * ( 2003 ) , 657 .
y. maeno , t.m .
rice and m. sigrist , physics today * 54 * ( 2001 ) , 42 . k.d .
nelson , z.q .
mao , y. maeno and y. liu , science * 306 * ( 2004 ) , 1151 .
k. ishida et al . , nature ( london ) * 396 * ( 1998 ) , 658 .
g. luke et al , nature ( london ) * 394 * ( 1998 ) , 558 .
j. xia , y. maeno , p.t .
beyersdorf , m.m .
fejer , and a. kapitulnik : phys .
* 97 * ( 2006 ) , 167002 .
t. tamegai , k. yamazaki , m. tokunaga , z. mao , and y. maeno : physica c * 388 * ( 2003 ) , 499 .
j. r. kirtley , c. kallin , c. w. hicks , e .- a .
kim , y. liu , k. a. moler , y. maeno , and k. d. nelson : phys .
b * 76 * ( 2007 ) , 014526 .
y. asano , y. tanaka , m. sigrist and s. kashiwaya , phys .
b * 71 * ( 2005 ) , 214501 .
j. mannhart , h. hilgenkamp , b. mayer , c. gerber , j.r .
kirtley , k.a .
moler and m. sigrist , phys .
77 ( 1996 ) , 2782 .
p. g. de gennes , _ superconductivity of metals and alloys _ , westview press ( 1999 ) .
m. sigrist and k. ueda , rev .
* 63 * ( 1991 ) , 239 , and references therein .
m. sigrist and d. agterberg , prog .
* 102 * ( 1999 ) , 965 .
m. matsumoto and m. sigrist , j. phys .
jpn * 68 * ( 1999 ) , 994 .
d. f. agterberg , phys .
b * 58 * ( 1998 ) , 14484 .
r. heeb and d. f. agterberg , phys .
b * 59 * ( 1999 ) , 7076 . v. b. geshkenbein and a. i. larkin , jetp lett .
* 43 * ( 1986 ) , 395 .
a. bouhon , diploma thesis ( 2007 ) .
m. sigrist , d.b .
bailey and r.b .
laughlin , phys .
* 74 * ( 1995 ) , 3249 .
ashby and c. kallin , phys . rev .
b 790 ( 2009 ) , 224509 . | a detailed theoretical interpretation of the josephson interference experiment between sr@xmath0ruo@xmath1 and pb reported by kidwingira _ et al _ @xcite is given . assuming chiral p - wave pairing symmetry a ginzburg - landau theory
is derived in order to investigate the structure of domain walls between chiral domains .
it turns out that anisotropy effects of the fermi surface and the orientation of the domain walls are essential for their internal structure . introducing a simple model for a josephson junction the effect of domain walls intersecting the interface between sr@xmath0ruo@xmath1 and pb
is discussed .
it is shown that characteristic deviations of the fraunhofer interference pattern for the critical josephson current as a function of the magnetic field occurs in qualitative agreement with the experimental finding .
moreover the model is able also to account for peculiar hysteresis effects observed in the experiment . |
we collected and analyzed more than 4.5 million time - stamped emails from students at a globally top - ranked mba program , focusing specifically on the relationship between students evolving communication networks and their subsequent career outcomes .
this data is available in the form of email logs recorded and stored by the university , along with registrar data on each student before and during their matriculation in the program .
included in the dataset is a record of each email sent by an mba student between fall 2006 and spring 2008 .
the record includes the date and time at which the email was sent and received and the ( anonymous ) numeric ids of the sender and receiver of the message .
academic records ( gmat scores , grades , extra - curricular activities , prior work experience and job titles ) , and demographic data ( age , race and nationality ) , were merged with the network data to connect email transmissions with personal characteristics .
there are approximately 11.5 million e - mails and 4.5 million student - student e - mails in the data .
an important characteristics of the data is its randomized design .
students are randomly assigned to sections within the school , minimizing selection effects .
since observations began when students first met each other , we eliminated the left censoring that typically occurs when network data are captured after ties have already been formed .
in order to understand the causal relationship between network effects and job rank , we used coarsened exact matching ( cem ) ( see @xcite ) to construct a reduced , matched sample ( matching students on all characteristics possible , i.e. , age , gpa , industry experience ) . then on the matched data , we further examine the significance of network effects on students job rank . additionally , in order to demonstrate that our observations can not be explained by a random network process
, we compare real observations to the null model where the degree sequence of network is preserved and links are placed completely randomly .
three important findings emerge from the data .
first , a student s likelihood of securing a coveted , high paying job after graduation is strongly associated with the type of network they develop during their time in the program .
students with a higher network degree , greater network centrality , and more balanced communication among alters tend to have the highest post - graduation salaries , even when matching students along several covariates ( figure [ fig_1 ] ) .
second , and somewhat surprisingly , we find that the structural characteristics of a student s ego network emerge as early as one month into the program and remain remarkably stable thereafter ( figures [ fig_1 ] [ fig_2 ] ) .
the above findings suggest that , when exposed to a new social system , actors initial network configurations may provide `` early warning signals of success '' a finding that has important practical and policy - level implications .
lastly , we observe robust patterns of `` rich - club '' behavior at the level of the global network , where highly central students are more densely connected among themselves than students of a lower degree @xcite @xcite . while the students who comprise the rich - club networks were more likely to have a higher salary post - graduation , they were also less likely to have the highest rank on other objective indicators of ability , such as gmat scores and gpa ( figure [ fig_3 ] ) .
we interpret these findings as evidence of a trade - off between the development of human capital and social capital during an mba program : on the one hand , students can choose to invest in building technical skills and domain - specific knowledge to enhance their career prospects ; on the other hand , they can choose to invest in building their social capital by developing new ties and fostering a robust community of peers . though these choices are certainly not mutually exclusive , the economic benefits of the former appear to significantly outweigh the benefits of the latter .
despite the fact that business schools advertise their role in fostering a valuable , life - long network ( haas mba 2013 ) , we find considerable differences in the types of networks that students actually develop - differences that are strongly linked to their future job placement and , ultimately , their access to the inner circles of the managerial elite .
our findings have important implications for both mba students and the firms that hire them . from the standpoint of a student
, the data suggest that important resources exist within an emerging mba network that ultimately improve one s attractiveness to employers . while we can only speculate about the mechanism underling this pattern of results i.e. specific networks may foster the development of particular social skills or , instead
, they may provide access to employer networks outside of the program itself we nevertheless find that it quite literally `` pays '' to develop one s social network in the early stages of a program . from the standpoint of prospective recruiters , the subtle signals that allow one to reliably infer a student s network structure @xcite may provide valuable insight into the qualities that the applicant will bring with them to their new job .
michael useem .
classwide rationality in the politics of managers and directors of large corporations in the united states and great britain .
administrative science quarterly vol .
27 , no . 2 ( jun . , 1982 ) , pp . | the `` business elite '' constitutes a small but strikingly influential subset of the population , oftentimes affecting important societal outcomes such as the consolidation of political power @xcite , the adoption of corporate governance practices , and the stability of national economies more broadly .
research has shown that this exclusive community often resembles a densely structured network , where elites exchange privileged access to capital , market information , and political clout in an attempt to preserve their economic interests and maintain the status quo @xcite .
while there is general awareness that connections among the business elite arise because `` elites attend the same schools , belong to the same clubs , and in general are in the same place at the same time '' , surprisingly little is known about the network dynamics that emerge within these formative settings .
here we analyze a unique dataset of all mba students at a top 5 mba program .
students were randomly assigned to their first classes ; friendship among students prior to coming into the program was rare ; and the network data email transmissions among students were collected for the year 2006 when students almost entirely used the school s email server to communicate , thereby providing an excellent proxy for their networks . after matching students on all available characteristics ( e.g. , age , grade scores , industry experience , etc . ) i.e. creating `` twin pairs '' we find that the distinguishing characteristics between students who do well in job placement and those who do not is their network .
further , we find that the network differences between the successful and unsuccessful students develops within the first month of class and persists thereafter , suggesting a network imprinting that is persistent .
finally , we find that these effects are pronounced for students who are at the extreme ends of the distribution on other measures of success students with the best expected job placement do particularly poorly without the right network ( `` descenders '' ) , whereas students with worst expected job placement pull themselves to the top of the placement hierarchy ( `` ascenders '' ) with the right network . |
the question whether and how a given function can be expressed approximately by polynomials are of great importance in theory as well as in practice . for example , by definition , an explicit , finite formula is unavailable for transcendental functions , and instead , an appropriate polynomial approximation is chosen to replace the function . because polynomials , particularly the ones of low order , are easy to manipulate , this approach provides computational speed with minimal penalty in accuracy . a natural candidate for polynomial approximation
is a truncated taylor expansion , typically at the midpoint of the interval where the approximation is most accurate .
taylor s theorem and the weierstrass approximation theorem @xcite asserts the possibility of local approximation of an arbitrary function @xmath0
. moreover , the approximation accuracy improves as the degree of the polynomial increases .
however , this improvement comes at the expense of complexity and computational speed .
this expense can be substantially reduced if the function can be approximated to the same accuracy with a lower degree polynomial . here
, we show analytically that an arbitrary function can be approximated via legendre polynomials using _ non - uniformly _ spaced points on an interval as the input , and that at least for some functions , approximation with legendre polynomials yields a substantially higher accuracy and faster convergence compared to taylor expansion of the same order ( i.e. , with the same number of non zero coefficients ) .
we further demonstrate the improvement in accuracy over taylor expansion numerically , using the sine , exponential , and entropy functions .
consider the problem of estimating the instantaneous slope of the curve mapping the the output of the function @xmath0 to @xmath1 $ ] .
the formula for the slope in a linear regression @xmath2 for uniformly spaced continuous points @xmath3 over the interval @xmath4 $ ] is given by @xmath5 where @xmath6 denotes the expectation of @xmath3 . because @xmath3 is uniform in
@xmath4 $ ] , the denominator of the equation [ eq : slope ] can be written as @xmath7}=\dfrac{(b - a)^{2}}{12}\label{eq : denom}\ ] ] where @xmath8}$ ] is the variance in the interval @xmath4 $ ] .
the numerator of the equation [ eq : slope ] can be written as @xmath9\right)\label{eq : intp0 } \\ & = & \dfrac{1}{b - a}\left[\intop_{a}^{b}y(x)\left(x-\dfrac{b+a}{2}\right)dx\right]\label{eq : intp1}\end{aligned}\ ] ] which can be solved using integration by parts : @xmath10 where @xmath11 . plugging equations [ eq : intpp0 ] and [ eq : intpp1 ] into equation [ eq : intp0 ] , equation [ eq : intp1 ]
can be written as @xmath12 & = & \frac{1}{b - a}\left[\dfrac{b^{2}y(b)-a^{2}y(a)}{2}-\intop_{a}^{b}\dfrac{x^{2}}{2}y'(x)dx\right.\nonumber \\ & & \left.-\dfrac{b+a}{2}\left(by(b)-ay(a)-\intop_{a}^{b}xy'(x)dx\right)\right]\nonumber \\ & = & \frac{1}{b - a}\left[-\dfrac{ab}{2}\left(y(b)-y(a)\right)+\dfrac{b+a}{2}\intop_{a}^{b}xy'(x)dx-\intop_{a}^{b}\dfrac{x^{2}}{2}y'(x)dx\right]\nonumber \\ & = & \frac{1}{b - a}\left[-\dfrac{ab}{2}\intop_{a}^{b}y'(x)dx+\dfrac{b+a}{2}\intop_{a}^{b}xy'(x)dx-\intop_{a}^{b}\dfrac{x^{2}}{2}y'(x)dx\right]\nonumber \end{aligned}\ ] ] @xmath13\nonumber \\ & = & \dfrac{1}{b - a}\left[\intop_{a}^{b}y'(x)\left(-\dfrac{ab}{2}-\dfrac{1}{2}\left(x-\dfrac{b+a}{2}\right)^{2}+\dfrac{(a+b)^{2}}{8}\right)dx\right]\nonumber \\ & = & \dfrac{1}{b - a}\left[\intop_{a}^{b}y'(x)\left(-\dfrac{1}{2}\left(x-\dfrac{b+a}{2}\right)^{2}+\dfrac{(b - a)^{2}}{8}\right)dx\right]\label{eq : intpp2}\end{aligned}\ ] ] finally , combining equations [ eq : intpp2 ] and [ eq : denom ] , equation [ eq : slope ] can be rewritten as @xmath14\label{eq : slope-2}\ ] ] which is just an average with respect to a quadratic kernel that is centered at the midpoint of the interval @xmath4 $ ] and zero at the ends .
equation [ eq : slope-2 ] allows estimation of the instantaneous slope over not just the points that are uniformly spaced , but all points in the interval @xmath4 $ ] .
this result for estimation of the slope is far more general .
it provides a least squares polynomial approximation to an arbitrary function on the interval @xmath4 $ ] . to see this , consider the shifted legendre polynomials of order @xmath15 , defined by rodrigues formula @xcite : @xmath16}(x ) & = & \dfrac{1}{2^{n}n!}\dfrac{d^{n}}{dx^{n}}\left(1-\left[\dfrac{2x - a - b}{b - a}\right]^2\right)^{n}\nonumber \\ & = & \dfrac{1}{n!(b - a)^{n}}\dfrac{d^{n}}{dx^{n}}\left[\left(b - x\right)(x - a)\right]^{n } \label{eq : rodrigues}\end{aligned}\ ] ] which are orthogonal functions with respect to the @xmath17 inner product @xmath18},p_{m,[b , a]}>=\dfrac{b - a}{2n+1}\delta_{nm}\label{eq : innerprod}\ ] ] where @xmath19 denotes the kronecker delta , equal to @xmath20 if @xmath21 and to @xmath22 otherwise .
furthermore , legendre polynomials of order @xmath20 to @xmath15 are the same as the orthogonal polynomials obtained by the gram - schmidt process on the polynomials @xmath23 with respect to the inner product given by equation [ eq : innerprod ] up to a constant multiplication factor @xcite .
therefore , by adding the basis functions @xmath24 , we obtain the @xmath25order polynomial fit to an arbitrary function @xmath26 on the interval @xmath4 $ ] as @xmath27}>}{<p_{n,[b , a]},p_{n,[b , a]}>}p_{n,[b , a]}\label{eq : polyfit}\ ] ] by completing the sum of squares , the expected integrated squared residual error for equation [ eq : polyfit ] can be written as @xmath28}>^{2}}{\left\|p_{n,[b , a]}\right\|^{2}}\right)\label{eq : isre}\ ] ] the term @xmath29}>}{<p_{n,[b , a]},p_{n,[b , a]}>}$ ] in equation [ eq : polyfit ] has a simple and telling interpretation .
note that @xmath30 } > & = & \dfrac{1}{n!(b - a)^{n}}\intop_{a}^{b}y\dfrac{d^{n}}{dx^{n}}\left[\left(b - x\right)(x - a)\right]^{n}dx\label{eq : legendre_approx } \\ & = & \dfrac{(-1)^{n}}{n!(b - a)^{n}}\intop_{a}^{b}\dfrac{d^{n}y}{dx^{n}}\left[\left(b - x\right)(x - a)\right]^{n}dx\label{eq : legendre}\end{aligned}\ ] ] which follows from equation [ eq : rodrigues ] .
the integral in equation [ eq : legendre ] can be solved using integration by parts , and none of the boundary terms appear in the solution because @xmath31 for @xmath32 and @xmath33 . moreover , @xmath34 } > = \dfrac{(b - a)^{n}(-1)^{n}n!}{(2n+1)!}e_{\beta(n+1,n+1;a , b)}\left[\dfrac{d^{n}y}{dx^{n}}\right]\ ] ] where @xmath35 is a beta probability distribution with @xmath36 degrees of freedom , shifted and scaled so that it lies on the interval @xmath37 $ ] , and @xmath38 is the expected value with respect to this distribution .
therefore , the contribution of each of the @xmath39order legendre polynomial to the residual error for equation [ eq : polyfit ] ( shown in equation [ eq : isre ] ) can be rewritten as @xmath40}>^{2}}{\left\|p_{n,[b , a]}\right\|^{2}}=\left[\dfrac{(b - a)^{2n-1}}{(2n+1)!{2n \choose n}}\right]\left(e_{\beta(n+1,n+1;a , b)}\left[\dfrac{d^{n}y}{dx^{n}}\right]\right)^{2}\label{eq : legendre-2}\ ] ] equation [ eq : legendre-2 ] is analogous to the taylor expansion , and uses a scaled value of the expected value of the @xmath25 derivative in the expansion .
note that the distribution @xmath41 is a parabola raised to the @xmath25 power and scaled to integrate to 1 ,
so it rapidly approximates @xmath42 . the mean and the variance of the distribution @xmath43 is given by @xmath44 and the entire series ( equation [ eq : polyfit ] ) takes the form @xmath45}{n!{2n \choose n}}(b - a)^{n-1}\right]p_{n,[b , a]}\label{eq : legendre_series}\ ] ] the series in equation [ eq : legendre_series ] converges much more rapidly than the taylor series because the coefficient @xmath46 converges to zero much more rapidly than the analogous term @xmath47 in taylor series .
furthermore , if @xmath48 for arbitrary constants @xmath49 and @xmath50 , it is easy to show ( by stirlings formula and by the nature of legendre functions ) that the squared norm of the @xmath25 term is bounded above by @xmath51^{2}!}^{2}\approxeq\dfrac{k^{2}\alpha^{2n}e^{2n}(b - a)^{2n-1}}{16^{n}n^{4n}}=\dfrac{k^{2}}{(b - a)}\left(\dfrac{\alpha^{2}e^{2}(b - a)^{2}}{16n^{4}}\right)^{n}\label{eq : conv_bound}\ ] ] note that the multiplier of the numerator , but not that of the denominator , of equation [ eq : conv_bound ] is bounded .
let @xmath52 denote the first term of the series in equation [ eq : legendre_series ] .
if we choose its order , @xmath53 , such that @xmath54 , then the squared norm of each successive term @xmath55 has a norm @xmath56 for @xmath57 .
this shows that the sum of squares of the terms after the @xmath58 term is at most twice the size of the previous term , and the series is dominated by a geometric sequence with a decay constant @xmath59 . because one can start with any arbitrary @xmath60
, the upper bound of the absolute value of @xmath61 term ( equation [ eq : conv_bound ] ) can be reduced arbitrarily by simply increasing the number of terms , @xmath15 .
demonstrating convergence in absolute value is also simple because the legendre polynomials are normalized so that they are bounded by one .
the bound , equivalent to that in equation [ eq : conv_bound ] , for the maximum of the @xmath25 term is given as : @xmath62!}\approxeq\dfrac{k\alpha^{n}(b - a)^{n-1}e^{n}}{\sqrt{2}(4)^{n}n^{2n}}=\dfrac{k}{\sqrt{2}}\left(\dfrac{\alpha(b - a)e}{(4)n^{2}}\right)^{n}\ ] ] this bound also indicates that for the same @xmath63 ( as above ) , the maximum norm of each successive term is again less than @xmath64 .
similar to above , for large @xmath63 , the absolute value of the terms @xmath65 is less than twice the absolute value of the term @xmath63 , which can be arbitrarily small .
thus , the legendre polynomial expansion converges at a very rapid rate .
there is no simple computable remainder ( corresponding to the one in taylor series expansion ) in legendre polynomial approximation , but only bounds ( see above ) .
therefore , we demonstrate the improvement in accuracy via an empirical approach using three representative transcendental functions : the sine function ( @xmath66 ) , the exponential function ( @xmath67 ) , and the entropy of a bernoulli trial ( @xmath68 ) in the interval @xmath69 $ ] .
note that all three functions satisfy the boundedness condition @xmath48 above .
for comparison , we approximate the functions with taylor and legendre polynomials using at most six non - zero coefficients ( figure [ fig : tayvslagfit ] ) .
because the sine function has odd symmetry about @xmath70 and the entropy function @xmath71 is odd , the expansions of these function involve , respectively , polynomial degrees of 11 and 10 of even and odd order .
we estimate the signal - to - noise ratio ( snr , in decibels ) for approximations as @xmath72 to quantify the accuracy in approximation ( table [ tab : impr ] ) .
.signal - to - noise ratios ( in decibels ) for the taylor ( t ) and legendre ( l ) polynomial approximations .
note the higher snr for the legendre polynomial approximation .
the accuracy of legendre polynomial approximation is significantly greater than that of taylor approximation of equal order . [
tab : impr ] [ cols="^,^,^,^,^,^,^ " , ] consistent with the theoretical considerations in above , these results show a rapid convergence of legendre polynomials for increasing number of polynomial coefficients , and an improved accuracy compared to taylor polynomials of the same order .
note that using six coefficients , even the least improvement ( in the case of the exponential function ) is 21.59 decibels ( which amounts to an average error for the legendre approximation 0.087 times the error for the taylor approximation ) .
the approximation for the sine function leads to an improvement of 63.18 decibels ( that is , the error the legendre approximation is 0.0007 times as that for the taylor approximation on average ) .
note in figure [ fig : tayvslagfit ] that while the taylor polynomial approximation has maximum accuracy at the midpoint , legendre polynomial approximation distributes the error more uniformly throughout the entire interval . as a result ,
the squared error of is smaller and the snr is larger in case of legendre polynomials .
our analytical and numerical results show that legendre polynomials can substantially improve the speed and accuracy of function approximation compared to taylor polynomials of the same order .
the fast convergence of legendre polynomials was noted in a prior study @xcite , but the geometric convergence in norm has not been shown analytically before .
the geometric convergence rate is consistent with the general result of srivastava @xcite on the relation of the generalized rate of growth of an entire function to the rate of uniform convergence of a polynomial approximation on an arbitrary infinite compact set . and
, it should be noted that the geometric convergence is not possible in general for @xmath73 approximants if @xmath74 @xcite .
thus , approximation using legendre polynomials can provide significant performance improvements in practical applications .
we also showed that legendre polynomials has the additional advantage that an arbitrary function can be approximated using _ non - uniformly _ spaced points on a given interval .
importantly , its accuracy of approximation is substantially higher than that of the taylor expansion with the same order of polynomials , with a uniform error distribution across the entire interval .
therefore , legendre expansion , instead of taylor expansion , should be used when global accuracy is important .
k. weierstrass , ber die analytische darstellbarkeit sogenannter willklicher functionen reeller argumente , sitzungsberichte der kniglich preuischen akademie der wissenschaften zu berlin 11 ( 1885 ) 633 . | we describe an expansion of legendre polynomials , analogous to the taylor expansion , to approximate arbitrary functions .
we show that the polynomial coefficients in legendre expansion , thus , the whole series , converge to zero much more rapidly compared to the taylor expansion of the same order . furthermore , using numerical analysis with sixth - order polynomial expansion , we demonstrate that the legendre polynomial approximation yields an error at least an order of magnitude smaller than the analogous taylor series approximation .
this strongly suggests that legendre expansions , instead of taylor expansions , should be used when global accuracy is important . numerical approximation , least squares , legendre polynomial 41a10 , 65d15 |
during the last few years amorphous semiconductors have been used in the manufacture of solar cells , batteries and phototransistors as well as in some steps of technological processing of very large scale integration ( vlsi ) microelectronics circuits @xcite .
thermally induced structural and electrical effects are known to be important in inducing the memory phenomenon in semiconducting chalcogenide glasses @xcite-@xcite .
alzewel et al . , @xcite have studied the electrical conductivity of powdered antimony chalcogenides .
antimony alloys have recieved great attention in the past due to semiconductivity of sb concentration .
it was reported by mott and devis @xcite that the effect of sb in se is even greater than in te in promoting crystallisation .
recently , study has been made on thin and thick films of the system @xmath1 ( @xmath2 35 , 40 and 45 ) @xcite .
crystalline has been prepared and studied by several workers @xcite .
antimony triselenide ( band gap @xmath3 1.2 ev ) , @xcite a p - type semiconductor having an orthorhombic @xmath4 structure with unit cell dimensions @xmath5 @xcite is used in preparation of recording optical laser disc and as an insulating material in mis solar cells .
the preparation and study of amorphous - can throw light on the possible applications of this material in that form .
measurements of electrical properties of a - are reported in this paper over a wide temperature and frequency range .
results indicate that electrical properties can be explained on the basis of widely accepted modified - correlated barrier hopping ( m - cbh ) model for chalcogenide glasses .
ingots of glassy material were prepared by melt quench method .
the 99.99% pure elements were sealed in an evacuated ( @xmath6 torr ) quartz ampoule ( 12 mm diameter and 160 mm long ) .
the ampoule was heated in a muffle furnace at the reaction temperature ( 630 @xmath7c for 24 hr ) .
the composition ( sb : se:1:1 ) of the glasses was determined by edax and amorphous nature was verified by x - ray diffraction pattern obtained using by x - ray diffractometer ( philips pw 1130/60 ) .
the pellets of 0.677 cm diameter were prepared in the hydraulic press at a pressure of @xmath8 .
the measurements were carried out in a specially designed vacuum cell in the temperature range 233 to 313 k and frequencies from 6 hz to 100 khz .
a sophisticated computer controlled ac - impedance system ( model 5206 , eg & g , parc usa ) was used for the measurements @xcite .
polished samples with a thickness 0.04 - 0.10 cm and area 0.36 @xmath9 were put in contact with aquadag ( a conducting emulsion ) in a sandwich configuration .
the sample was kept in vacuum in a copper sample holder ensuring no temperature gradient between the electrods and the sample .
the effect of temperature is studied by using a closed cycle refrigerator ( model f-70 , julabo ) was used to obtain different temperatures , which can maintain constant temperature within @xmath10 1 k for all the measurements .
ohmic contacts were confirmed through linear i - v characterstics in the voltage range .
all electrical measurements of real and imaginary components of impedance parameters ( @xmath11 and @xmath12 ) and real and imaginary component of admittance parameters ( @xmath13 and @xmath14 ) were made over a wide range of temperature ( 233 to 313 k ) and frequency ( 6 hz to 100 khz ) .
figure [ f2 ] shows the plot of @xmath15 versus 1000/t which was obtained using the expression _
dc = c^ ( -e /kt ) . [ e1 ]
the plots in fig .
[ f2 ] are found to be linear over the temperature range studied . in the above expression @xmath16 and @xmath17
is the temperature coefficient of the band gap . from the figure ,
the activation energy @xmath0e and @xmath18 for the sample are found to be 0.43 ev and @xmath19 respectively .
it has also been observed that @xmath0e is constant for different pellets ( thickness = 0.04 cm to 0.08 cm ) with a statistical error of 0.3% .
the variation of @xmath20 versus @xmath21 ( fig .
[ f3 ] ) indicates that the impedance is sensitive to frequency at lower temperatures and gradually becomes independent of frequency with increase in temperature .
this type of behaviour indicates dominance of band conduction at higher temperatures .
the sensitivity of to temperature is shown in fig .
the figure gives the nyquist plots ( @xmath11 versus @xmath12 ) for at different temperatures over the frequency range studied .
the semicircular nyquist plots indicate dipolar nature of the sample , which may be due to hopping back and forth of bipolarons between charged defects states ( d@xmath22 and d@xmath23 ) reported to be present in chalcogenides @xcite .
the size of the nyquist plots reduces with rise in temperature and regain their original size when temperature is increased . the nyquist plots for form a perfect arc of a semicircle with its centre lying considerably below the abscissa ( i.e. , distribution parameter @xmath24 ) .
figure [ f4 ] also gives the value of @xmath25 calculated at different temperatures .
in dielectric materials , the finite value of the distribution parameter @xmath25 and a depressed arc are typical for a dipolar system involving multirelaxation processes @xcite-@xcite .
but the dependence of @xmath25 on temperature indicates that the mean position of the dipole is varying with variation in temperature .
the dipolar nature of is further confirmed by investigating the variation of capacitance with temperature at different frequencies .
it is clear from fig .
[ f5 ] that in the lower temperature range the value of c is nearly constant and it increases with increase in temperature .
however the rate of change of capacitance ( i.e. , dc / dt ) is higher for lower frequencies and decreases with increasing temperature .
these results can be explained by the fact that thermally assisted hopping results in increasing the capacitance of the material . in other words , in the lower temperature range , the dipoles remain frozen and attain rotational freedom when the temperature is increased .
thus the relaxation effects are confirmed by the rate of increase in capacitance with frequency .
the inversion of nyquist plots in the admittance plane are shown in the fig .
it is clear from the figure that the angle of inclination is less than @xmath26 .
this type of behaviour has been explained on the basis of r - c network model ( fig .
[ f7 ] ) , such that total impedance is given by @xcite y ( ) & = & b ( i)^n + ic _ + g_0 + & = & g_0 + b \ { ( s/2 ) + i ( s/2 ) } ^n + ic _ [ e2 ] moreover , this angle ( shown in fig . [ f6 ] ) is also temperature dependent . at lower temperatures
( below 273 k ) it is almost constant , while after a certain temperature ( 276 - 278 k ) it shows a sudden decrease , which implies a change in the conduction mechanism .
the complex impedance plots show the dipolar multi - relaxation nature of a- .
this behaviour is explained by the r - c network model ( fig .
however , this does not further elucidate the nature of conduction mechanism in this sample
. therefore a.c .
conductivity @xmath27 at different temperatures is calculated .
it is found that dispersion in the temperature dependence of @xmath28 curves increases at lower frequencies while it merges at higher frequencies ( fig . [ f8 ] ) . from the expression _ ac ( )
= _ total ( ) -_dc [ e3 ] the a.c .
conductivity is obtained , where @xmath29 is given by eqn.[e1 ] and its behaviour is shown in the fig .
the @xmath30 is measured directly from lock - in - analyser ( section [ exp ] ) .
the @xmath27 dominates over @xmath29 at lower temperatures , while @xmath29 dominates at higher temperatures ( fig .
[ f8 ] ) .
the modified correlated barrier hopping ( m - cbh ) model explains the experimental results reported in this paper .
it states that bipolaron hopping between the charged defects states d@xmath22 and d@xmath23 is responsible for the a.c .
conductivity in these semiconductors .
dipolar multi - relaxation behaviour of this sample is already indicated from the nyquist plots described in the last subsection .
so the charged defect states are expected to be present in a- . according to cbh model a.c .
conductivity is given by @xcite _
ac = [ e4 ] where @xmath31 is the number of polarons involved in the hopping process , @xmath32 is proportional to the square of the concentration of the states and @xmath33 is the dielectric constant .
@xmath34 is the hopping distance for the conduction @xmath35 and is given as r _
= [ e5 ] where @xmath36 is the maximum barrier height and @xmath37 is the characteristic relaxation time for the material .
figure [ f8 ] shows the variation of a.c .
conductivity with temperature at different frequencies .
it is evident from the figure that @xmath38 is very sensitive to temperature in the higher temperature regime .
the low temperature a.c .
conduction can be explained by considering bipolaron hopping between d@xmath22 and d@xmath23 centers whereas the higher temperature behaviour is due to thermally activated single polaron hopping . at higher temperatures
a number of thermally generated d@xmath39 centers are produced with a temperature dependent concentration @xcite n_0 = n_t ( - u_eff /2 kt ) [ e6 ] where @xmath40 is the concentration of d@xmath22 or d@xmath23 centers at @xmath41 .
the defect concentration factor @xmath42 in eqn .
[ e4 ] is replaced by : n n_p = n^2_t /2 & ( for bipolaron hopping ) + n n_p = n^2_t /2 ( - u_eff /2kt ) & ( for single polaron hopping ) [ e7 ] thus for single bipolaron hopping this factor is thermally activated and hence @xmath27 is also activated . according to cbh model @xcite b = w_m -w_1 + w_2 [ e8 ] where @xmath43 is optical band gap , @xmath44 is the maximum band width and behaviour with frequency @xmath45 is as _ ac ( ) = a ^s [ e9 ] here [ @xcite ] , s = = 1 - 6kt / w [ e10 ] while @xmath46 for bipolaron hopping and @xmath47 for single polaron hopping .
therefore , @xmath44 is estimated from the @xmath48 values at lower temperatures
. values of @xmath49 , @xmath50 and @xmath51 were estimated from the values of @xmath48 at higher temperatures .
these parameters are then fed to cbh model to fit the a.c .
conductivity data calculated from the experiments .
the behaviour of eqn.[e10 ] ( @xmath52 ) can be seen in the fig .
this indicates that @xmath53 follows @xmath54 for lower temperatures and @xmath50 for higher temperatures .
figure [ f9 ] also shows a sudden change in conduction mechanism at a certain temperature ( 276 - 278k ) .
the similar behaviour is noticed in the admittance plots ( fig .
the different parameters used for cbh model calculations are as following further , the effect of temperature and frequency on the hopping length @xmath55 , which is a measure of the effective length of a dipole can be studied using cbh model .
the variation of @xmath34 with frequency at different temperature is shown in the fig .
the figure indicates that @xmath56 is more sensitive to frequency in the high temperature regime and the sensitivity decreases with decreasing temperature .
thus at low temperature , @xmath34 is constant with frequency , which implies a constant value of capacitance which is found to be true from the fig .
nyquist plots or complex impedance studies confirm the dipolar nature of and multi - relaxation behaviour as seen in most chalcogenide glasses .
these plots also indicate that the capacitive nature dominates over the resistive nature of the sample at lower temperature ( @xmath57 k ) , while the resistive nature is dominant at higher temperatures .
the conduction mechanism has a sudden change from bipolaron to single polaron hopping ( 276 - 278 k ) . the temperature and frequency dependence of a.c .
conductivity is well explained by the m - cbh model .
the contribution to a.c .
conductivity from single polaron hopping is dominant at higher temperatures .
the authors are grateful to university grant commission for providing funds under cosist and cas programmes for purchase of measurement system which was essential for investigations reported in this paper .
a c adams in _ plasma deposited thin films _ edited by j mort and f jansen ( crc , boca raton , 1986 ) .
a d pearson , j. non - cryst .
solids * 2 * , 1 ( 1970 ) .
d l eaton , j. am .
caram soc .
* 477 * , 554 ( 1964 ) .
h j stocker , appl .
. lett . * 15 * , 55 ( 1969 ) .
b g bagley & h e bar , j. non - cryst .
solids * 2 * , 155 ( 1970 ) .
h fritzsche & s r ovshinsky , j. non - cryst .
solids * 2 * , 148 ( 1970 ) .
k alzewl , m m sekkina and z m hanafi , z. phys .
chem * 94 * , 235 ( 1975 ) .
k alzewl , m m sekkina and z m hanafi , egypt .
* 12 * , 51 ( 1981 ) .
mott n f & davis e a in _ electronic processes in noncrystalline materials _ ( clarendon , oxford , 1971 ) p. 371 .
m fadel , k m sharaf & n a hegab , indian j. pure & appl
* 34 * , 810 - 814 ( 1996 ) .
b r chakraborty , b ray , r bhattacharya and a k dutta , j. phys . chem .
solids * 41 * , 913 ( 1980 ) .
f kosek , j tulka and l stourac , czech .
j. phys . *
b28 * , 325 ( 1978 ) .
c wood , z hurych and j c shaffer , j. non . cryst .
solid , 209 ( 1972 ) .
w e pearson in _ handbook of lattice spacing and crystal structure of metals _
( pergamon press , new york , 1964 ) p. 843 .
vohra a and srivastva k k , philos . mag . * 61 * , 201 ( 1990 ) .
shukla r , khurana p and srivastva k k , j. mat .
sci : mat in electr ; * 3 * , 132 ( 1992 ) .
navdeep goyal , parmana j. phys .
* 40 * , 97 ( 1993 ) . a k jonscher in _ dielectric relaxation in solids _
( chelsea dielectric press , london , 1983 ) p. 90 .
s r elliott , phil . mag . * 36 * , 1291 ( 1977 ) .
s r elliott , phil . mag . * b36 * , 135 ( 1977 ) .
s r elliott , adv . phys . * 36 * , 135 ( 1989 ) .
k shimakawa , phil . mag . * 46 * , 123 ( 1982 ) . | this paper reports conduction mechanism in a - over a wide range of temperature ( 238k to 338k ) and frequency ( 5hz to 100khz ) . the d.c . conductivity measured as a function of temperature shows semiconducting behaviour with activation energy @xmath0e= 0.42 ev .
thermally induced changes in the electrical and dielectric properties of a - have been examined .
the a.c .
conductivity in the material has been explained using modified cbh model .
the band conduction and single polaron hopping is dominant above room temperature .
however , in the lower temperature range the bipolaron hopping dominates .
* keywords . * chalcogenides glasses ; amorphous semiconductors ; a.c . conductivity ; electrical properties .
+ * pac nos 72 - 20 ; 71 - 55 * |
this work has been funded in canada by nserc ( grant sapin 386432 ) , cfi - lof and orf - sif ( project 24536 ) , and by the france - canada research fund ( project `` listening to scintillating fractures '' ) .
prof . c. dujardin of lyon , france , kindly provided access to his equipment to measure x - ray emission spectra .
we thank m. chapellier for stimulating discussion , and a. b. mcdonald for comments on our manuscript .
32ifxundefined [ 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://arxiv.org/abs/1101.5205 [ ( ) ] , link:\doibase 10.1016/s0927 - 6505(02)00111 - 1 [ * * , ( ) ] link:\doibase 10.1016/j.physleta.2006.03.059 [ * * , ( ) ] arxiv : hep - ex/0011064 [ * * , ( ) ] , link:\doibase 10.1140/epjc / s10052 - 012 - 1971 - 8 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1016/s0022 - 2313(00)00230 - 1 [ * * , ( ) ] link:\doibase 10.1088/0022 - 3727/17/1/016 [ * * , ( ) ] link:\doibase 10.1063/1.339650 [ * * , ( ) ] link:\doibase 10.1007/bf00555916 [ * * , ( ) ] link:\doibase 10.1038/nature07378 [ * * , ( ) ] http://arxiv.org/abs/astro-ph/0106200 [ ( ) ] , http://repository.dl.itc.u-tokyo.ac.jp/dspace/handle/2261/12045 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.79.3202 [ * * , ( ) ] link:\doibase 10.1007/s10704 - 006 - 0051 - 1 [ * * , ( ) ] @noop _ _ , ed .
( , ) @noop _ _ , ed .
( , ) @noop * * , ( ) link:\doibase 10.1016/j.nima.2005.01.303 [ * * , ( ) ] link:\doibase 10.1016/0266 - 3538(87)90086 - 8 [ * * , ( ) ] link:\doibase 10.1103/physrevb.52.9224 [ * * , ( ) ] link:\doibase 10.1016/s0022 - 2313(01)00412 - 4 [ * * , ( ) ] link:\doibase 10.1063/1.1662183 [ * * , ( ) ] link:\doibase 10.1006/adnd.1993.1013 [ * * , ( ) ] link:\doibase 10.1016/0022 - 2313(85)90114 - 0 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.68.1244 [ * * , ( ) ] @noop _ _ , ed .
( , ) link:\doibase 10.1007/bf00726989 [ * * , ( ) ] \doibase doi:10.1063/1.1735524 [ * * , ( ) ] http://arxiv.org/abs/0708.4315 [ ( ) ] link:\doibase 10.1016/j.astropartphys.2012.05.005 [ * * , ( ) ] link:\doibase 10.1209/0295 - 5075/99/28003 [ * * , ( ) ] | prompted by intriguing events observed in certain particle - physics searches for rare events , we study light and acoustic emission simultaneously in some inorganic scintillators subject to mechanical stress .
we observe mechanoluminescence in , and , in various mechanical configurations at room temperature and ambient pressure .
we analyze the temporal and amplitude correlations between the light emission and the acoustic emission during fracture .
a novel application of the precise energy calibration of provided by radioactive sources allows us to deduce that the fraction of elastic energy converted to light is at least @xmath0 .
rare - event searches in particle physics , like those looking for particle dark matter @xcite , employ draconian measures to reduce all forms of background that could hide their signal .
the main background is usually of radioactive nature , however other forms are possible .
for instance , the first , calorimetric phase of the cresst experiment measured phonons in sapphire crystals cooled to around @xmath1 to detect particle interactions @xcite .
initially , thermal contraction of the holders caused fractures in the crystals , limiting sensitivity of the experiment by three orders of magnitude in rate above the radioactive background until the problem was identified @xcite .
certain more recent experiments use a coupled measurement of phonons and scintillation to determine the nature of the interacting particles and reject the radioactive background @xcite .
however , mechanoluminescence , i.e. the emission of light by a solid subject to mechanical stress , occurs in many materials @xcite , including plastic scintillators @xcite .
in addition , correlated acoustic emission ( ae ) and visible - photon emission have been observed during mechanical deformation of colored alkali halides @xcite .
other types of emission , including electrons @xcite , positive ions @xcite , and x - rays @xcite , have also been studied in various materials under mechanical stress .
nevertheless , there is an opinion that fracture events in scintillation - phonon detectors would produce little or no light ( e.g. @xcite ) .
prompted by recent intriguing events observed by the cresst ii dark matter search using cryogenic scintillation - phonon detectors @xcite , we study mechanoluminescence as a form of photon - producing background in such devices . we also argue that photon emission can help characterize the rupture dynamics in a manner analogous to ae @xcite , and can provide complementary information about the fracture energy .
to investigate these two points , we have carried out what we believe are the first experiments to correlate the acoustic and light emission from common scintillators under mechanical stress , and to quantify the conversion of elastic energy into light .
our studies involve the measurement of acoustic phonons and luminescence as a piece of scintillating crystal is stressed to the point of rupture .
the inorganic scintillators studied here are widely used in particle detection @xcite : bismuth germanate ( , a.k.a .
bgo ) , zinc tungstate ( ) and cadmium tungstate ( ) . except where specified
, all samples were kindly supplied by crystal manufacturing lab ltd . , novosibirsk .
we have tested two loading geometries , at room temperature and ambient pressure .
initially , @xmath2 rectangular prisms were indented by a steel bead driven by a manual screw , causing the material to break in multiple fragments . in order to better control the rupture process
, we subsequently adopted the double cleavage drilled compression ( dcdc ) geometry @xcite ( fig .
[ fig_fig_1 ] ) .
these samples were @xmath3 rectangular prisms , polished to optical quality , with a @xmath4 diameter circular hole drilled perpendicularly in the middle of the @xmath5 face . under compression along the long axis ,
cracks formed reproducibly on either side of the hole in the plane parallel to the @xmath6 face . unlike bgo , tungstates have a cleavage plane @xcite ; orientation of the crystals was chosen so that this plane was parallel to the dcdc fracture direction .
the dcdc samples were pressed against a backstop by a pushrod that was driven by an actuator ( controlled manually in early experiments , and by computer in later ones ) .
this geometry provides some control of crack velocity via the applied compressive stress , up to a critical length at which point the sample cleaves abruptly @xcite . a force gauge in the pushrod measured the load imposed by displacements of the actuator .
the acoustic activity was recorded by two piezo - electric pinducers in contact with opposite sides of the crystal via ultrasound gel .
light emission was recorded by a standard bialkali photomultiplier ( pm ) covering the crystal .
the output of the pm was integrated with a time constant ranging from @xmath7 to @xmath8 depending on the sample to produce a signal that could be digitized more slowly than the raw pm output .
all four channels ( force , 2 ae , pm ) were digitized at a rate of @xmath9 samples per second in a continuous stream .
the setpoint of the computer - controlled actuator was also recorded .
the fracture procedure involves advancing the actuator by small steps ( down to @xmath10 ) .
each step is followed by a waiting time of at least @xmath11 .
the whole procedure lasts @xmath12@xmath13 during which we continuously monitor ae and pm activity .
figure [ fig_mainresult ] shows typical observations for the tungstate crystals .
spurts of activity on the ae and pm channels are correlated to drops in the force .
these abrupt force relaxations are a direct consequence of crack growth in the samples .
there is also a regime in which the force decreases slowly , and during which other processes , such as rearrangements and dislocations , produce ae and pm signals @xcite .
similar correlations have been observed for other materials , such as colored @xcite , composites @xcite , and silica glass @xcite .
indentation of bgo samples ( from fibercryst , france ) also produces correlated ae and light signals .
these correlations can be analyzed more quantitatively by looking at the statistical properties of individual acoustic and light events .
in this analysis , individual events have been determined on each channel by requiring that the signal surpass a threshold , and that it be separated from the previous event by at least a minimum time .
figure [ fig_wait_extra ] shows that the distributions of waiting times between events are similar for ae and light over two decades . for light events , it is possible to obtain a much smaller minimum separation time ( @xmath14 ) than for ae ( @xmath15 ) , since the former pulses are shorter than the latter . to quantify the amount of light emitted and to investigate the emission mechanism , we next compared the spectra of light emitted by bgo ( sample from fibercryst ) during fracture and during scintillation stimulated by x - rays . to avoid afterglow from the high dose of x - rays ,
the fracture spectrum was measured first , and then one of the fragments was measured while being irradiated by x - rays .
the measurements of the light spectra were carried out with a monochromator and a ccd camera . in the x - ray excitation configuration , a generator provided x - rays with a broad spectrum up to a few tens of . over the recorded range of wavelengths , both resulting spectra ( fig .
[ fig_spectra ] ) show a very similar shape ( as is the case for many materials @xcite ) , indicating a common underlying luminescence mechanism , at least for the last stages of light emission .
the likeness of our spectra to previous work on bgo indicates this last stage involves @xmath16 electronic transitions of the luminescence centers @xcite .
this measurement does not determine if the fractures excite the luminescence centers directly .
it can not be excluded that the crystal is indirectly excited by particles produced by fracto - emission @xcite interacting in the scintillator .
for instance , if x - rays or electrons are produced , the bgo sample would be quite efficient at detecting them itself ( attenuation length @xmath17 for @xmath18 x - rays @xcite ) , as opposed to needing an external x - ray detector @xcite .
alternatively , any uv light ( potentially from arcs in the atmosphere near the fracture ) might be re - absorbed in the scintillator since bgo absorbs wavelengths below @xmath19 @xcite .
we then calibrated the light channel for a dcdc bgo sample using radioactive @xmath20-ray sources as we would calibrate a regular scintillator used for particle detection .
calibration was carried out in the main dcdc setup and with the standard daq ( fig .
[ fig_fig_1 ] ) .
the analysis involves extracting individual events as described earlier , and then building the distribution of the event integrals .
the integral of an event , like its amplitude , is a proxy for the energy deposited by the @xmath20 , so we can identify the peaks in the distribution corresponding to the energies of the radioactive sources and of backscattering @xcite , as shown in fig .
[ fig_calibr ] . over the energy range available with these sources ,
the crystal and readout behave in a linear manner with respect to the energy deposited in the scintillator , and we extrapolate this calibration to higher energies .
calibrations before and after the fracture are compatible to within @xmath21 . from the standpoint of the scintillation mechanism , @xmath20-rays and x - rays can interact in inorganic scintillators via the photoelectric effect or through compton scattering , creating a primary electron - hole pair which eventually transfers a portion of the deposited energy to the luminescence centers ( for bgo ) . for a given light signal ,
calibrations provide the equivalent amount of energy deposited by a @xmath20-ray .
this can be converted to the actual energy of the emitted light by knowing the light yield of the scintillator ( @xmath22 for @xmath20-rays in bgo @xcite ) , and the energy of individual scintillation photons ( @xmath23 c.f .
fig [ fig_spectra ] ) .
the calibrated bgo crystal was then fractured in the main dcdc setup , using the protocol described earlier .
we focus on the last drop in force , occurring when the crack splits the sample in two .
figure [ fig_bgobreak ] shows that the overall behavior is similar to that observed for the tungstates .
the photon channel displays a wide distribution of event amplitudes .
they in fact reach up to the saturation level of the integrator , equivalent to @xmath20-ray energy deposits of @xmath24 a testimony to the amount of light emitted .
there is also a hint of increased photonic activity before the main fracture occurs .
the correlations are better quantified in fig .
[ fig_corr_extra ] illustrating the cumulative number of events on the acoustic channels and those on the photon channel .
minimum separation time required between photon events was @xmath8 , and @xmath25 between ae events .
on all three channels , the cumulative number of events rises steadily with a slight increase in slope until the fracture , as determined by the drop in the force .
this drop marks a sharp increase in the slope of cumulative events that eventually peters out .
when normalized to the total number of events on each channel , all curves are very similar .
lastly , event amplitudes for all channels are power - law distributed ( fig . [ fig_hist_spectra ] ) . for both ae channels ,
the exponents are @xmath26 ( errors are one standard deviation ) . assuming the ae energy is proportional to the amplitude squared
, the power - law distribution of the energy would have as exponent @xmath27 .
this is compatible with the exponent obtained for the light channel , @xmath28 , for which amplitude is proportional to energy .
a more precise measurement might check if the absolute value of the ae exponent is in fact larger than that of the light channel , an indication of dissipation in the acoustic channel @xcite .
we can also use our technique to estimate certain contributions to the energy budget of the rupture process in bgo .
elastic energy is converted into broken bonds , phonons ( that are either measured as acoustic emission or decay into heat ) and light .
for the emitted light , the calibrations with @xmath20-ray sources establish that the total amount of emitted light is equivalent to that caused by @xmath20-ray energy deposits of at least @xmath29 , in other words a total energy converted to light greater than @xmath30 . only a lower bound
is set because of integrator saturation mentioned previously .
eighty - five percent of this light is emitted in a @xmath31 window centered on the main fracture .
we next estimate the drop in elastic energy of the crystal during the main fracture . neglecting the hole in the dcdc geometry ,
the elastic energy in the crystal of section @xmath32 ( perpendicular to the force @xmath33 ) and of volume @xmath34 is @xmath35 , where @xmath36 is young s modulus . applying a standard acoustic measurement of longitudinal and shear waves @xcite to bgo
, we find @xmath37 , somewhat above the range of published values ( up to @xmath38 @xcite and references therein ) .
assuming the loaded volume and surface vary little , the drop in force of @xmath39 is equivalent to a drop in elastic energy of @xmath40 .
since the surface created during the fracture is roughly @xmath41 , this amounts to a surface creation energy of the order of @xmath42 , consistent with the scales typically found for crystals ( e.g. @xmath43 for , @xmath44 for @xcite ) .
furthermore , the conversion efficiency of elastic to luminous energy is therefore at least @xmath45
. a fuller understanding of the energy budget requires knowledge of the energy transformed into phonons , or heat .
the standard room - temperature technique to study brittle fracture only involves measurement of acoustic phonons using acoustic emission .
we are not aware of any work quantifying the fraction of energy going into acoustic phonons , or calibrating the energy scale of acoustic phonons .
the harder - to - implement technique of cryogenic calorimetry would allow measurement of all the phonons and would therefore provide more insight into the energy budget @xcite .
we have evidenced correlated mechanical - stress - induced emission of photons and of phonons in several common inorganic scintillators used for particle detection ( , and ) , at room temperature and in a normal atmosphere .
at least for bgo , both emissions share similar distributions of waiting times over several decades . also for bgo ,
the two energy distributions are similar , and the wavelength of fractoluminescence is the same as that of scintillation . in a mechanically - stressed scintillation - only detector , the power - law distribution of mechanoluminescence energies means that spurious low - energy events may mimick dark matter ones .
it seems reasonable to assume our results apply to other tungstates , and it would be interesting to study the light - phonon correlations of individual fracture events in cresst ii - like conditions of vacuum and low - temperature ( other background - based explanations of the intriguing cresst ii events have been proposed @xcite ) .
in addition , the re - absorption of products of fracto - emission could affect other types of solid - state particle detectors ( e.g. ionization or ionization - phonon ones @xcite ) . from the standpoint of fracture physics , in contrast to previous studies by acoustic emission or by spectrally - resolved luminescence ( e.g. @xcite ) , our use of the light channel offers a precisely calibrated energy scale .
this allows us to quantify the amount of elastic energy converted into light , and is a step towards a better fundamental understanding of fracture mechanisms . |
dynamics and dissipative properties of nuclear fermi liquid depend in many aspects on the dynamic distortion of the fermi surface in the momentum space .
it is well - known that taking into account this distortion allows the description of a new class of phenomena , most famous of which are giant multipole resonances .
furthermore , scattering of particles from the distorted fermi surface leads to relaxation of collective motion and gives rise to fluid viscosity @xcite .
we point out that the development of instability in nuclear processes like binary fission or multifragmentation in hi - reactions also depends on dynamic effects such as the dynamic fermi - surface distortion . a convenient way to introduce the fermi - liquid effects into the nuclear many body problem is to start from the kinetic equation for the phase space distribution function @xmath0 .
the kinetic equation can then be transformed identically to a set ( infinite ) of equations for the moments of @xmath0 in @xmath1-space , namely , local single - particle density @xmath2 , the velocity field @xmath3 , the pressure tensor @xmath4 , etc . , see @xcite .
in case of small variations of the particle density , @xmath5 , the first order moment of the kinetic equation has the form of euler - navier - stokes equation and is given by @xcite @xmath6 the internal energy density @xmath7 in eq .
( [ eiler ] ) contains both kinetic @xmath8 and potential @xmath9 energy densities : @xmath10 .
the pressure tensor @xmath11 depends on the fermi - surface distortion effect . in general case
, tensor @xmath11 also includes the viscosity tensor which is derived by the collision integral .
( [ eiler ] ) is not closed because it contains the pressure tensor @xmath4 given by the second order moment of the distribution function @xmath0 .
we will follow the nuclear fluid dynamic approach @xcite and take into account dynamic fermi - surface distortions up to the multipolarity of @xmath12 . assuming a periodic in time eigenvibrations with the eigenfrequency @xmath13 , where @xmath14 and @xmath15 are real , and a separable form of the velocity field , @xmath16 , with @xmath17 , eq .
( [ eiler ] ) is reduced to the equation of motion for the macroscopic variable @xmath18 with the following secular equation @xmath19 here , @xmath20 and @xmath21 are the mass coefficient and the stiffness coefficient in the traditional liquid drop model ( ldm ) @xcite respectively . the additional contribution from @xmath22 to the stiffness coefficient and the dissipative term @xmath23
depend on the relaxation time @xmath24 and are given by @xmath25 and @xmath26 where @xmath27 is the equilibrium pressure of the fermi gas .
the additional contribution from @xmath22 to the stiffness coefficient in eq .
( [ 6.7 ] ) is absent in the ldm , i.e. in the liquid drop limit @xmath28 , and represents the influence of the dynamic fermi - surface distortion on the conservative forces in the fermi system .
finally , the dissipative term @xmath23 appears due to the interparticle scattering from the distorted fermi surface . in general , both , @xmath29 and @xmath30 depend implicitly on the temperature , @xmath31 , via the dependence of the relaxation time @xmath24 and of @xmath27 on @xmath31 . in cold nuclei , in the zero - sound limit @xmath32 , the main contribution to the stiffness coefficient in eq .
( [ 6.7 ] ) is due to the fermi - surface distortion effect given by @xmath33 . in fig .
1 , this effect is shown in a transparent way for isoscalar quadrupole excitations . as it is seen from this figure
, the fermi - surface distortion effect leads to a significant upward shift of the energy of vibrational states to locate it in the region of the quadrupole giant resonance ( solid line ) .
we will discuss the fermi - surface distortion effect in more detail for the case of isoscalar giant monopole resonances ( isgmr ) .
this particular case is important for understanding the nature of nuclear incompressibility .
we will consider below the model for a fermi - liquid drop having a sharp surface of the equilibrium radius @xmath34 and the bulk density @xmath35 .
the particle density variation @xmath5 is then given by @xmath36 where the unknown functions @xmath37 and @xmath38 are related to each other by the particle number conservation and the bulk density parameter @xmath37 is found from the equation of motion derived by eq .
( [ eiler ] ) .
namely , @xmath39 here @xmath40 is the _ static _
incompressibility @xmath41 the additional contribution @xmath42 to the incompressibility @xmath43 in eq .
( [ eq2 ] ) is due to the _ dynamic _ fermi - surface distortion effect @xcite .
the value of @xmath42 depends on the landau scattering amplitude @xmath44 . in nuclear case ,
@xmath45 , one has @xmath46 .
an essential property of a finite liquid drop having a free surface is that the motion of the surface should be consistent with the motion of the liquid inside the drop .
this can be achieved by imposing a boundary condition for the compensation of the compressional pressure @xmath47 at the liquid surface by the pressure generated by the surface tension forces @xmath48 . finally , the eigenenergies in eq .
( [ eq2 ] ) are given by @xmath49 where @xmath50 are derived from the following boundary condition : @xmath51 here the coefficients @xmath52 and @xmath53 are related to the surface tension and the fermi - surface distortion respectively and are given by @xmath54 in the general case of fermi - liquid drop with @xmath55 , the eigenfrequency @xmath56 given in eq .
( [ omn ] ) is renormalized due to two contributions associated with the fermi - surface distortion : 1 ) the direct change of the sound velocity , i.e. in eq .
( [ omn ] ) @xmath43 appears instead of @xmath40 ; 2 ) the change of the roots @xmath57 of the secular equation ( [ sec ] ) due to additional contribution from @xmath55 in eq .
( [ sec ] ) .
these two effects work in opposite directions : @xmath43 increases @xmath56 while @xmath53 in eq .
( [ sec ] ) decreases it .
both values of @xmath43 and @xmath57 in eq .
( [ omn ] ) depend on the landau scattering amplitude @xmath44 . in the limit of zero- to first sound transition at @xmath58
, one has @xmath59 and the solution of eqs .
( [ sec ] ) and ( [ omn ] ) is close do the one obtained in ldm .
however , for realistic nuclear forces @xmath45 , the general expressions ( [ sec])-([omn ] ) with @xmath55 have to be used . in fig .
2 we plotted the eigenenergy , @xmath60 , of the lowest monopole mode @xmath61 given by eqs .
( [ omn ] ) and ( [ sec ] ) as a function of mass number @xmath62 .
the dashed line shows the result for @xmath60 from eq .
( [ omn ] ) where @xmath43 was replaced by @xmath63 and boundary condition ( [ sec ] ) with @xmath64 was used .
the solid line shows the result of fldm calculations .
we adopted the following expression for @xmath40 @xcite @xmath65 with @xmath66 and @xmath67 and for @xmath68 and @xmath69 .
so , in spite of strong renormalization of nuclear incompressibility due to the fermi - surface distortion effect , the energy of the giant monopole resonance @xmath60 in the fermi - liquid drop is located close to the one in the traditional ldm @xcite where the fermi - surface distortion effects are absent .
the compensation of the fermi - surface distortion effect in nuclear incompressibility @xmath43 due to the consistent account of the same effect in the boundary condition ( [ sec ] ) is essential for the lowest mode @xmath70 only .
the above mentioned compensation is rather small for the highest modes with @xmath71 . in particular
, the eigenenergy of the double ( overtone ) giant monopole resonance @xmath72 ( in contrast to @xmath60 ) has a significant upward shift with respect to @xmath73 given by the traditional liquid drop model .
the relaxation of nuclear collective motion toward thermal equilibrium has been described in great detail within the framework of the kinetic theory , taking into account the collision integral @xmath74 @xcite .
the collision integral @xmath74 depends on the equilibrium distribution function , @xmath75 . in case of a finite fermi system ,
the equilibrium distribution function , @xmath75 , contains both the diffusive layer and oscillations in @xmath76-space which are caused by particle reflections at the potential surface .
both of them depend on the distance @xmath77 .
this fact also leads to the @xmath77-dependence of the collisional relaxation time , @xmath78 , given by @xmath79 the @xmath80-dependence of @xmath81 is due to the memory effects in the collision integral @xcite .
the evaluation of the collisional relaxation time @xmath78 in a finite fermi system requires additional accuracy .
the diffusive layer in the equilibrium distribution function in momentum space can lead to a spurious effect in non - zero gain- and loss fluxes of the probability in the ground state of the fermi system .
however , it can be shown , see ref .
@xcite , that the general condition for the disappearance of both above mentioned fluxes is reached due to the occurrence of quantum oscillations in the equilibrium distribution function @xmath75 in momentum space .
the @xmath77-dependent diffusive tail of the distribution function in momentum space leads to an increase of the collisional damping of the collective motion in the surface region of a nucleus and thus to an increase of the isoscalar giant quadrupole resonance ( gqr ) width . however , this increase is strongly reduced due to the above mentioned oscillations of the equilibrium distribution functions appearing in the collision integral . as a result
, the collisional width of the isoscalar gqr does not exceed 30 - 50% of the experimental value and agrees with the estimates of the width where the sharp thomas - fermi distribution function is used @xcite . in the case of heated nuclei ,
both temperature and memory effects have to be taken into account in the collision integral . for frequencies small
compared to the fermi energy , one finds @xcite : @xmath82/ \tilde{\tau}_{2 } , \label{3.20}\ ] ] which is valid for @xmath83 and when sharp thomas - fermi distribution for @xmath75 is assumed .
the magnitude of @xmath84 can be given in the following general form at low temperatures @xmath85 @xcite @xmath86 where the quantity @xmath87 is for the isoscalar mode and @xmath88 is for the isovector mode : @xmath89 @xcite .
in general , there are several contributions to the relaxation time @xmath24 .
the main contribution @xmath90 arises from the above discussed interparticle collisions and the other from collisions of nucleons with the moving nuclear surface ( one - body dissipation with relaxation time @xmath91 )
. we will also include the contribution from particle emission with relaxation time @xmath92 .
thus , @xmath93 introduction of the one - body dissipation reflects the peculiarities of our consideration , namely , restricting fermi - surface distortions to the multipolarity @xmath94 does not allow us to take into consideration landau damping ( fragmentation width ) .
the total width , @xmath15 , of a collective state can be derived from eq .
( [ gamma ] ) with @xmath90 replaced by the total relaxation time @xmath24 from eq .
( [ 3.12 ] ) . assuming sharp nuclear surface , the total width @xmath15 is reduced to the following form , see ref .
@xcite , @xmath95 where @xmath96 and @xmath97 is the first sound velocity in nuclear matter .
we want to note that the total intrinsic width given by eq .
( [ 3.10 ] ) has a bell - shaped form as a function of @xmath98 .
the width @xmath99 is peaked around @xmath100 and the maximum value of @xmath15 is @xmath101 .
it is easy to see that @xmath102 represents the crossing point of both curves @xmath103 ( zero sound regime ) and @xmath104 ( first sound regime ) . due to this fact
the condition @xmath105 can be used as the condition for the transition from long to short relaxation time regimes .
the magnitude of the intrinsic width @xmath99 decreases when the parameter @xmath106 exceeds @xmath102 .
we have @xmath107 for the isovector gdr at the realistic value of @xmath108 .
such a value of @xmath106 can be reached at the temperature @xmath109 , see eq .
( [ 3.20 ] ) . if the relation ( [ 4.4 ] ) with @xmath110 and @xmath111 is used as the condition for transition between different regimes of sound wave propagation then the transition between both regimes occurs at the higher temperature @xmath112 . in fig .
3 we show the intrinsic width of the giant dipole resonance ( gdr ) in the nucleus @xmath113_sn_. as it is seen from fig .
3 , the expression ( [ 3.10 ] ) ( solid line ) leads to a smoother behaviour of the total intrinsic width with increasing excitation energy as compared to the prediction of the zero - sound regime given by the following condition : @xmath114 with @xmath115 and @xmath116 ( dashed line ) .
our calculation of the intrinsic width @xmath15 for the isovector gdr in the hot nucleus @xmath113_sn _ confirms a saturation effect in the energy dependence of @xmath15 .
however , we observe a systematic deviation of the evaluated width with respect to experimental values . this deviation can be reduced by varying the parameter @xmath117 in eq .
( [ 3.10 ] ) .
an increase in the parameter @xmath117 improves agreement with the experiment ( see dotted line in fig .
in the vicinity of the equilibrium state of the nuclear fermi - liquid drop the stiffness coefficients for particle density and surface changes are positive and the nucleus is stable with respect to the corresponding distortions .
with decreasing bulk density or increasing internal excitation energy ( temperature ) the liquid drop reaches regions of mechanical or thermodynamical instabilities with respect to the small particle density and experiences shape fluctuations and separation into liquid and gas phases .
let us consider small density fluctuations @xmath118 starting from eq .
( [ eiler ] ) . we will use a simple saturated equation of state @xmath119 = { \cal e}_{kin}[\rho ] + { \cal e}_{pot}[\rho ] $ ] with @xmath120 = { \hbar^2\over 2\,m}\ , \left[{3\over 5}\,\big({3\pi^2\over 2}\big)^{2/3 } \,\rho^{5/3 } + { 1\over 4}\,\eta \,{(\vec{\nabla}\rho)^2\over \rho}\right],\,\,\,\ ] ] @xmath121 @xmath122 = t_0\,\rho^2 + t_3\,\rho^3 + t_s\,(\vec{\nabla}\rho)^2 + \epsilon_c[\rho ] .\ ] ] the potential energy @xmath123 $ ] has a skyrme - type form including the surface term @xmath124 and the coulomb energy density , @xmath125 $ ] .
we will again assume a sharp surface behaviour of the equilibrium particle density .
assuming also @xmath126 and taking into account continuity equation and eq .
( [ epot ] ) , the equation of motion ( [ eiler ] ) can be reduced to the following form ( we consider the isoscalar mode ) : @xmath127 where @xmath128 , @xmath129 and @xmath40 is the incompressibility coefficient @xmath130 let us consider volume instability regime , @xmath131 , and introduce the growth rate @xmath132 ( @xmath15 is real , @xmath133 ) . using eq .
( [ 0eq3 ] ) , one obtains @xmath134 where @xmath135 the quantity @xmath136 in eq .
( [ 0disp ] ) appears due to the fermi surface distortion effect .
equation ( [ 0disp ] ) is valid for arbitrary relaxation times @xmath24 and thus describes both rare and frequent collision limits as well as intermediate cases . from it
one can obtain the leading order terms in the different limits mentioned . _
( i ) frequent collision regime : @xmath137 .
high temperatures . _
+ the contribution from the dynamic distortion of the fermi surface , @xmath138 , can be neglected in this case and we have from eqs .
( [ 0disp ] ) and ( [ gamma ] ) , @xmath139 where @xmath140 is the classical friction coefficient . in the limit of small friction , we obtain @xmath141 the amplitude of the density oscillations with a certain multipolarity @xmath142 , @xmath143 , grows exponentially if @xmath133 . the expression ( [ 0disp2 ] ) determines two characteristic values of the wave number @xmath144 , namely , @xmath145 and @xmath146 where the growth rate reaches the maximum of magnitude @xmath147 and @xmath15 goes to zero ( excluding @xmath148 ) respectively , i.e. , @xmath149 both @xmath145 and @xmath146 are given by @xmath150 for a saturated nuclear liquid , one has @xmath151 and @xmath152 .
thus , the critical value @xmath146 from eq .
( [ crit ] ) increases with a decrease of the bulk density @xmath35 , see eqs .
( [ k1 ] ) and ( [ crit ] ) .
the presence of friction decreases the critical value @xmath146 , i.e. reduces the instability .
the existence of the critical wave number @xmath146 for an unstable mode is a feature of a finite system .
the growth rate @xmath15 depends on the multipolarity @xmath142 of the nuclear density distortion and on the position of @xmath153 in the interval @xmath154 @xcite .
the value of @xmath153 is determined from the boundary condition for the pressure on free nuclear surface . for given @xmath34 the value of @xmath153 increases with @xmath142 at @xmath155 .
that means that if @xmath156 instability increases with @xmath142 and the nucleus becomes more unstable with respect to internal clusterization into small pieces ( high multipolarity or multifragmentation regimes ) rather than to binary fission ( low multipolarity regime ) .
in contrast , the binary fission is preferable if @xmath157 .
the critical value @xmath146 and the value @xmath145 are given by @xmath160 thus , the distortion of the fermi - surface leads to a decrease of the critical value @xmath146 i.e. the fermi - liquid drop becomes more stable with respect to volume density fluctuations due to dynamic fermi - surface distortion effects .
abrikosov and i.m .
khalatnikov , _ rep .
_ * 22 * , 329 ( 1959 ) .
v.m . kolomietz and h.h.k .
tang , _ phys .
scripta _ * 24 * , 915 ( 1981 ) .
kolomietz , _ local density approach for atomic and nuclear physics _ ( naukova dumka , kiev , 1990 ) ( in russian ) .
g. holzwarth and g. eckart , _ nucl .
_ * a325 * , 1 ( 1979 ) .
kolomietz , v.a .
plujko and s. shlomo , _ phys .
c _ * 52 * , 2480 ( 1995 ) .
a. bohr and b.r .
mottelson , _ nuclear structure , volume ii _
( benjamin , new york , 1975 ) .
kolomietz , a.g .
magner and v.a .
z. phys . _ * a345 * , 131 , 137 ( 1993 ) .
blaizot and b. grammaticos , _ nucl .
_ , 115 ( 1981 ) . k. sato and s. yoshida , _ phys .
c _ * 52 * , 837 ( 1995 ) .
kolomietz , v.a .
plujko and s.shlomo , _ phys .
c _ * 54 * , 3014 ( 1996 ) .
kolomietz , s.v.lukyanov , v.a .
plujko and s. shlomo , _ phys .
c _ * 58 * , # 1 ( 1998 ) .
k. ando , a. ikeda and g. holzwarth , _
z. phys .
_ * a310 * , 223 ( 1983 ) .
c.j.pethick and d.g.ravenhall , _ nucl .
phys . _ * a471 * , 19c ( 1987 ) .
d.kiderlen , v.m.kolomietz and s.shlomo , _ nucl .
phys . _ * a608 * , 32 ( 1996 ) .
bertrand , _ nucl .
. _ * a354 * , 129c ( 1981 ) . j. j. gaardhoje , _ ann . rev .
nucl . part .
_ * 42 * , 483 ( 1992 ) . | the kinetic theory of fermi liquid is applied to finite nuclei .
the nuclear collective dynamics is treated in terms of the observable variables : particle density , current density , pressure etc .
the relaxation processes and the development of instabilities in nuclear fermi - liquid drop are strongly influenced by the fermi - surface distortion effects . |
in the field of atom interferometry , the improving sensitivity of inertial sensors @xcite is paving the way for many new applications in geophysics , navigation and tests of fundamental physics . most of these experiments are based on raman transitions @xcite to realize beamsplitters and mirrors , which manipulate the atomic wave - packets . among others ,
this technique has the advantage of an internal state labelling of the exit ports of the interferometer @xcite , enable an efficient detection methods .
moreover , the atoms spend most of the time in free fall , with very small and calculable interactions with the environment .
the inertial forces are then determined by the relative displacement of the atomic sample with respect to the equiphases of the laser beams , which realise a very accurate and stable ruler .
this makes this technique suitable for high precision measurements , as required for instance for inertial sensors and for the determination of fundamental constants @xcite . a limit to the accuracy and
the long term stability of these sensors comes from wave - front distortions of the laser beams .
this wave - front distortion shift appears directly on the signal of an interferometer when the atoms experience different wave - fronts at each raman pulse .
this effect thus depends on the actual trajectories of the atoms , so that a precise control of the initial position , velocity and temperature of the atomic clouds is required @xcite .
a convenient technique to reduce this bias is to minimize the number of optical components in the shaping of the two raman laser beams and by implementing them in a retro - reflected geometry @xcite . indeed ,
as long as the two beams travel together , wave - front aberrations are identical for the two beams and thus have no influence on their phase difference .
this geometry also provides an efficient way to use the * k * reversal technique , which allows to diffract the atomic wavepackets in one or the opposite direction and thus to separate effects of many major systematic errors such as gradients of magnetic fields or light shifts @xcite .
the main drawback of this geometry arises from the presence of off - resonant raman transitions , which induce a light shift on the resonant raman transition and thus a phase shift of the atom interferometer . in the following ,
we investigate this effect called two photon light shift ( tpls ) @xcite .
we first show that the tpls arises from several off - resonant transitions and evaluate each contribution .
we then derive the impact onto the phase of an atom interferometer and use our gravimeter and gyroscope - accelerometer for quantitative comparisons .
in particular we measure the systematic shifts and we investigate the influence on the long term stability .
the study demonstrates that the precise control of experimental parameters , in particular the raman laser intensities and polarisations , is needed to reduce the influence of this effect for such interferometers .
) , cross the experiment and are reflected by a mirror , crossing twice a second quater - wave plate .
the wave - plates are set in such a way that counter - propagating raman transitions are allowed but co - propagating raman transitions are forbidden.,width=321 ] the two experiments are using different alkali - metal atoms : @xmath0rb in the case of the gravimeter and @xmath1cs in the case of the gyroscope . as hyperfine structures , transition selection rules and raman laser setups are similar ( see figure [ schema ] ) , their results can be compared easily .
the raman transitions couple the two hyperfine ground states of the alkaline atom ( labelled @xmath2 and @xmath3 ) via an intermediate state ( labelled @xmath4 ) and two lasers with frequencies ( labelled @xmath5 and @xmath6 ) detuned by @xmath7 on the red of the @xmath8 line . during the interferometer sequence
, a bias magnetic field is applied along the direction of propagation of the raman laser beam to lift the degeneracy of the magnetic sublevel manifold .
the two raman lasers are overlapped with orthogonal linear polarisations and delivered within the same polarisation maintaining optical fiber to the vacuum chamber .
after the fiber , the raman beams pass through a quarter - wave plate to convert the initial linear polarisations into circular polarisations , noted @xmath9 for the raman laser at the frequency @xmath5 and @xmath10 for the orthogonal polarisation at @xmath6 .
these beams are then retro - reflected through a quarter - wave plate to rotate the polarisation of each beam into its orthogonal polarisation ( @xmath11 , @xmath12 ) . for @xmath13 to @xmath13 transitions ,
there are two pairs of beams ( @xmath14 and @xmath15 ) , which can drive counter - propagating raman transitions with effective wave - vectors @xmath16 .
then , the ground state @xmath17 is coupled with the excited state @xmath18 by the pair of raman laser ( @xmath19 ) and to the excited state @xmath20 with the pair of raman laser ( @xmath21 ) .
we use the doppler effect to lift the degeneracy between the two resonance conditions . indeed , if the atoms have a velocity in the direction of propagation of the raman lasers , the doppler shifts are of opposite sign for the two counter - propagating transitions .
the resonance condition for each of these couplings is @xmath22 , where @xmath23 is the hyperfine transition frequency , @xmath24 is the recoil energy and @xmath25 the doppler shift due to the atomic velocity @xmath26 in the reference frame of the apparatus .
consequently , the detuning between the two resonances is @xmath27 , therefore we can discriminate between the two transitions when the doppler shift is large enough compared to the linewidth of the raman transition .
this linewidth is characterised by the effective rabi frequency @xmath28 , which depends on the product of the two raman lasers intensities and inversely to the raman detuning @xmath7 @xcite . in this first part
, we use the gyroscope - accelerometer experiment described in detail in @xcite .
the experiment has been performed with a cloud of cold caesium atoms ( 1.2 @xmath29 ) prepared initially in the @xmath30 state .
the atoms are launched at 2.4 m.s@xmath31 with an angle of @xmath32 with respect to the vertical direction .
the raman lasers are implemented in the horizontal plane with a @xmath33 angle with the normal of the atomic flight direction .
thus , to select only one raman transition for the interferometer , the frequency difference between the two raman lasers can be tuned to be resonant with either the @xmath34 or the @xmath35 transition . ) .
the data have been recorded with laser parameters corresponding to a @xmath36 pulse of 135 @xmath37s duration .
lines ( 1,1 ) correspond to the two counter - propagating transitions , line 2 to the copropagating transition between the two @xmath13 states and lines 3 to the copropagating magnetic sensitive transitions.,width=302 ] figure [ spectre ] shows the transition probability as a function of the detuning of the raman transition with respect to the hyperfine transition frequency .
one can identify the two velocity selective counter - propagation transitions ( labelled 1 and 1 ) , whose widths reflect the velocity distribution of the atomic cloud .
in addition to the counter - propagating transitions , we observe transitions due to residual co - propagating raman coupling , also detuned from resonance by a doppler shift ( lines 2 , 3 and 3 ) . when the frequency difference of the raman lasers is tuned to be resonant with one of the counter - propagating transitions , the second counter - propagating transition and the co - propagating ones induce a light shift ( tpls ) on the selected raman transition used for the interferometer .
the tpls is the differential shift between the two atomic levels corresponding to the atomic states @xmath17 and @xmath38 involved in the atomic interferometer .
the energy of the state @xmath17 is shifted of @xmath39 by the off - resonant @xmath40 transition detuned by @xmath41 ( eq . [ epg ] ) , while the energy of the state @xmath38 is shifted of @xmath42 by the off - resonant @xmath43 transition detuned by @xmath44 ( eq . [ epe ] ) .
the two levels are shifted in opposite directions as illustrated in figure [ couplages ] , here for the case of a raman transition resonant with @xmath45 . and
@xmath46 involved in the atomic interferometer .
the two states are coupled together through the selected raman transition ( + * * k * * in this particular case ) . but each state is also coupled to an other one , through an off - resonant raman transition of opposite wave - vector.,width=321 ] tpls corrections are calculated from fourth order perturbation theory . in the case of our system , the level shifts @xmath39 and @xmath42 for a @xmath47 interferometer are given by : @xmath48 where @xmath28 is the effective rabi frequency corresponding to counter - propagating raman transitions .
thus , the shift of the resonance condition depends on the sign of direction of the selected raman laser pair ( i.e. @xmath47 ) , quadratically on the rabi frequency and inversely to the doppler detuning : @xmath49 & = \frac{\omega_\mathrm{eff}^2}{\pm 8 \omega_\mathrm{d}}+\frac{\omega_\mathrm{eff}^2}{4 ( \pm 2 \omega_\mathrm{d}+4 \omega_\mathrm{r } ) } \end{array}\end{aligned}\ ] ] the frequency shift is measured from fits of the spectrum lines , as displayed in figure [ spectre ] , with different raman intensities ( @xmath50 is proportional to the product of the intensity of the two lasers ) .
note that we discriminate it from the shifts independent of @xmath51 , like quadratic zeeman effect or ac stark shift , by alternating measurements @xmath34 and @xmath35 , leading to a differential determination of the effect .
the difference in the resonance condition @xmath52 depends only on the tpls and the doppler effect .
the doppler effect does not depend on the rabi frequency and can be determined by extrapolating @xmath53 to @xmath54 .
the results of these measurements are displayed in figure [ tpls ] as a function of @xmath50 .
the curve clearly shows the quadratic dependence of the frequency shift with @xmath28 . for this experimental configuration
the doppler shift was 85 khz , and the value of @xmath55 for the largest @xmath28 ( @xmath56 khz at the center of the beam ) is 2.1 khz in good agreement with the expected 2.4 khz .
the laser beams have a gaussian shape , with a 15 mm waist ( radius at @xmath57 in intensity ) . in order to evaluate the phase shift ,
we first measure the tpls for different atomic positions by scanning the resonance along their trajectory .
as the raman beam are horizontal , the doppler shift is constant for the three pulses and the tpls varies only with the raman laser intensity .
figure [ tplsposition ] displays the measured tpls , proportional to the laser intensity , which follows exactly the gaussian profile of the laser beams . in an ideal experiment , with perfect circular polarization and a raman detuning @xmath7 large compared to the hyperfine structure of the intermediate state ( 201 mhz in the case of the caesium atom ) , co -
propagating transitions are forbidden . in a real experiment , with imperfect polarization and/or finite raman detuning , co - propagating transitions are slightly allowed , and will lead to additional tpls .
imperfect polarization leads to a residual combination of @xmath58 and @xmath59 in the co - propagating beams and allows coupling between @xmath60 states .
as the momentum exchanged @xmath61 , the doppler and recoil effect are negligible and the resonance condition is @xmath62 ( line 2 of figure [ spectre ] ) . the rabi frequency corresponding to the transition @xmath63 is determined experimentally using the residual co - propagating transition probability , @xmath64 at full raman laser power .
it gives @xmath65 , which can be explained by an error of linear polarization of one of the raman laser of 2% in power .
the detuning of this transition , compared to the two counter - propagating transitions , depends on the doppler and recoil shift : @xmath66 for @xmath67 khz and @xmath68 khz we find an effect due to this coupling smaller than 100 hz . the second source of residual co - propagating transitions stems from the coupling of @xmath70 by the co - propagating raman laser pairs ( @xmath71,@xmath72 ) and ( @xmath73,@xmath74 ) .
because of the hyperfine splitting in the intermediate state , there are two paths for the raman transition .
both transitions interfere destructively when the detuning compared to the intermediate state @xmath7 is larger than the hyperfine splitting of the intermediate state , and so in this case the transition strength is zero .
however , in our experimental set up , with @xmath75 mhz and the @xmath76 mhz the ratio between the rabi frequency of counter - propagating transitions @xmath28 and the rabi frequency of the co - propagating @xmath77 transition @xmath78 is 6.1 , leading to a transition probability of 6.4 @xmath79 in good agreement with the experimental value ( see figure [ spectre ] ) .
these transition resonance conditions depend on the magnetic field amplitude as @xmath80 , where @xmath81 khz / g for caesium . with a calculation similar to the one used to obtain eq.[nuls2cp ] ,
we deduce the two photon light shifts @xmath82 induced by the magnetically sensitive transitions for the @xmath47 case to be : @xmath83 the first term in eq .
[ nuls2b ] is due to the coupling with @xmath84 whereas the second term is induced by the coupling with @xmath85 .
it is clear from eq .
[ nuls2b ] that a residual magnetic contribution appears in the half difference and creates a magnetic sensitivity in addition to the standard quadratic zeeman effect .
this contribution to the two photon light shift is measured by changing the bias field in the raman interaction zone .
using the differential method previously described , we show in figure [ tpls_b ] the variation of the total tpls with the bias field .
the resonance around @xmath86 mg corresponds to the case where a magnetic co - propagating transition and the counter - propagating transition are resonant simultaneously .
previous measurements have been performed with a 31 mg magnetic field bias .
in the following we will consider interferometers constituted of three raman pulses in a @xmath87 sequence . if the tpls is constant during the interferometer it is equivalent to a fixed frequency shift of the raman transition . in that case
, it is well known that no phase shift is introduced in the interferometer . on the contrary
, a fluctuation of the tpls during the interferometer sequence leads to a phase shift given by : @xmath88 where @xmath89 is the sensitivity function of the atom interferometer , defined in @xcite . in the case where the interaction pulses are short enough that one can neglect the variation of the tpls during the pulses , and that the area of first and last pulse fulfils the @xmath90 condition ,
the two - photon interferometer phase shift can be approximated by : @xmath91 where @xmath92 and @xmath93 are the tpls and the rabi frequencies of the i - th pulse , respectively .
one might notice that the frequency shift during the @xmath36 pulse does not contribute to the interferometer phase shift .
moreover , as all components of the tpls , counter - propagating and co - propagating terms , increase as the square of the rabi frequencies , the interferometer phase shift scales linearly with the raman laser power . in the limit where the co - propagating transitions
are negligible ( perfect polarization and very large raman detuning ) and the dominant source of tpls is due to the counter - propagating transition , the phase shift of the interferometer can be simplified as : @xmath94 where @xmath95 is the doppler shift for the i - th pulse .
more generally , the interferometer phase shift can be calculated when the @xmath97 pulse condition is no longer fulfilled .
this appears when the rabi frequency drifts due to changes in power or polarization of the raman lasers .
generalising the formalism of the sensitivity function to the case where @xmath96 allows deriving the interferometer phase shift : @xmath98 & & -\frac{\delta\omega^{(3)}_\mathrm{tpls}}{\omega^{(3)}_\mathrm{eff}}\tan\left(\frac{\omega^{(3)}_\mathrm{eff}\tau^{(3)}}{2}\right ) \end{array}\end{aligned}\ ] ] usually , the rabi frequencies and pulse durations can be taken equal for the first and the last pulses ; the expression of the interferometer shift is then : @xmath99 as before , for a dominant counter - propagating transition , the previous expression can be simplified to : @xmath100 an other aspect of the influence of the tpls on the atomic phase shift concerns the stability of the experiment versus the experimental parameters fluctuations , in particular the raman laser power . as
rabi frequency fluctuations are small in relative values ( typically smaller than 10 % ) , we can develop eq .
[ tppsgeneral3 ] to first order in @xmath101 close to the usual conditions @xmath90 and find : @xmath102 similar calculations may be derived to extract the dependance on the duration of the pulse or the doppler detuning .
as the stability of the latter parameters is much better controlled in cold atom interferometers , no measurable influence on the short term stability of the interferometer is expected .
we first consider the case of the gravimeter developed at syrte and described in detail in @xcite . in this compact experimental set - up ,
cold @xmath103 atoms are trapped in a 3d mot in 60 ms and further cooled during a brief optical molasses phase before being released by switching off the cooling lasers . during their free fall over a few centimetres ,
the interferometer is created by driving the raman laser in the vertical direction , with pulses separated by free evolution times of @xmath104 ms .
the first raman pulse occurs 17 ms after releasing the atoms .
the doppler shift at the first pulse is thus relatively small , about 400 khz , and gets large , about 3 mhz , for the last pulse . with a rabi frequency of 40 khz
, this leads to a tpls of about 22 mrad for counter - propagating transition . following the method of paragraph [ copro ]
, we find a 11 mrad for co - propagating ones .
this corresponds to a large shift for the gravity measurement of about @xmath105 . to measure the bias on the atomic interferometer phase due to tpls
, we exploit its dependence with the rabi frequency .
the principle of this measurement is based on a differential method , where one performs an alternating sequence of measurements of the interferometer phase with two different rabi frequencies @xmath28 and @xmath106 , but keeping the areas of the pulses constant by changing the duration of the pulses @xmath107 .
the rabi frequency is modulated with the power of the raman lasers . in practice ,
the differential measurement is performed by alternating sequences of measurements with four different configurations : ( @xmath108 ) , ( @xmath109 ) , ( @xmath110 ) and ( @xmath111 ) . after averaging for 5 minutes ,
we extract the difference of the tpls between the two rabi frequencies with an uncertainty below 1 mrad .
this measurement was repeated for various @xmath106 , keeping @xmath28 fixed .
the phase differences are displayed in figure [ shiftgravi ] in according to the ratio of the rabi frequency @xmath112 .
the results clearly demonstrate the linear dependence of the phase shift with the rabi frequency .
the fit of the data allows to extract a 32 mrad shift , in very good agreement with the expected value ( 33 mrad ) , deduced from eq .
deviations from the linear behavior and discrepancies ( up to @xmath113 ) between different measurements that correspond to the same ratio can not be explained by uncontrolled fluctuations in the rabi frequencies , as the simultaneous monitoring of the laser intensities showed stability at the per cent level during the course of the measurements .
we demonstrated that these fluctuations were correlated with changes of the polarization of the raman beams , which modulate the contribution to the phase shift of the undesired co - propagating transitions .
the line corresponds to the calculated shift.,width=321 ] we perform a complementary measurement by changing the duration of the first and last pulse simultaneously , while keeping @xmath28 constant .
the acquisition is performed with a similar differential method than previously , but now alternating between different pulse durations and a pulse duration of 6 @xmath114 , when the @xmath97 pulse condition is fulfilled .
the data are then shifted by the bias deduced from previous measurements with @xmath115 @xmath114 ( result of figure [ shiftgravi ] ) and displayed in figure [ shiftvstau ] .
using eq .
[ tppsgeneral2 ] , the fit of the data gives a rabi frequency of 39 khz , in very good agreement with the expected value ( 42 khz ) .
a small deviation appears for long pulse durations when the areas of each of the two pulses is close to @xmath36 . in the case of our gravimeter , where cold atoms are dropped from rest
, the tpls is very large , almost two orders of magnitude above the pursued accuracy . in principle
, this effect can be measured accurately by alternating measurements with different rabi frequencies .
but , it seems desirable to decrease the effect by operating with lower rabi frequencies , the drawback being an increased velocity selectivity of the raman pulses . a more stringent velocity selection , or smaller temperatures ,
are then required in order to preserve a good fringe contrast . in the case of a fountain gravimeter , where the atoms are launched upwards at a few m / s ,
the doppler shift at the first and last pulse is much larger , considerably reducing this effect . for the parameters of the stanford gravimeter @xcite ( long pulse duration of 80 @xmath37s and time between pulses of 160 ms )
, we find a phase shift of 0.8 mrad , which corresponds to @xmath116 . in the case of the gyro - accelerometer , the mean velocity of the wave - packet is not collinear with the effective raman wave vector .
consequently , the atomic phase shift measured with the interferometer is sensitive to the rotation rate * @xmath117 * in addition to the acceleration * a*. as in the case of the gravimeter , the atomic phase is shifted by the tpls and by other systematics ( labelled @xmath118 ) , for instance the phase shift induced by the laser wave - front distortions @xcite .
the total phase shift is expressed by : @xmath119 our gyroscope - accelerometer uses a double interferometer with two atomic clouds following the same trajectory but with opposite directions to discriminate between acceleration and rotation phase shifts . moreover ,
our experiment is designed to measure different axis of rotation and acceleration according to direction of propagation of the raman laser beams .
we will illustrate the impact of the tpls on the interferometer in the configuration where the raman lasers are directed along the vertical direction .
acquisitions have been recorded for a total interaction time between the first and the last pulse of 40 ms . to remove drifts from other sources ,
the phase shift is measured in a differential way with respect to the maximum rabi frequency available at the location of the @xmath36/2 pulses ( 33 khz ) .
the line shows the calculated shift for acceleration signal.,width=321 ] the measurement is realized in the same way than for the gravimeter experiment by comparing atomic phases with high and low rabi frequency , changing the raman laser power but keeping the pulse duration constant to 7.6 @xmath37s . in order to enhance the tpls signal
we decrease the time between pulses to 20 ms .
indeed , the doppler effect is reduced and the available laser power on the side of the gaussian laser profile is increased .
the first pulse occurs 15 ms before the apogee and the third 25 ms after the apogee , corresponding respectively to a doppler shift of about @xmath120344 khz and @xmath121574 khz .
the rabi frequency for each pulse is approximately 33 khz , then the tpls expected from eq.[tpps ] is about 38 mrad for each interferometer .
as this shift is similar for the two interferometers when the two atomic clouds perfectly overlap and experiment the same tpls , it bias the acceleration signal only . figure [ shiftgyro40 ] displays the variation of the acceleration and rotation signals with the rabi frequency . acceleration shift ( squares )
varies in a good agreement with expected shift ( continuous curve ) calculated from eq .
[ tppsgeneral ] .
rotation shift ( circles ) shows no dependance on the rabi frequency and illustrates that the rejection from the acceleration signal is efficient .
nevertheless , fluctuations from expected behaviour are clearly resolved and repeatable .
we attribute these deviations to wave - front distortions of the raman laser beams . indeed ,
when the atomic trajectories do no perfectly overlap , a residual bias appears on the rotation due to unperfected cancellation .
this bias depends on the details of the wave - front distortions weighted by the actual atomic cloud distributions , and is modified when the rabi frequency is changed . in the usual conditions , with interrogation time of 80 ms , the acceleration shift is reduced to about 12 mrad thanks to the increase of the doppler shift and reduction of the rabi frequency on the side of the gaussian raman beams .
we finally estimate the impact of the raman laser power fluctuations on the stability of the rotation signal in usual conditions ( interrogation time of 80 ms ) .
we performed a complementary measurement by recording the interferometer signal and the raman laser power in the same time when applying a modulation of the laser powers of 10@xmath79 .
the modulation is applied by attenuating the radio - frequency signal sends to the acousto - optic modulator used to generate the raman pulses .
the power of the lasers was recorded during the third pulse thanks to a photodiode measuring the intensity on the edge of the laser beams .
we found a small dependance of the rotation signal on the power fluctuation of @xmath122 rad.s@xmath31/@xmath79 , which can limit the long term stability of the gyroscope .
as no dependance was expected , we attribute it again to non perfect superimposition of the two atomic clouds trajectories , leading to a different rabi frequencies experienced by the two clouds .
we have shown that the use of retro - reflected raman lasers in atom interferometers induces off - resonant raman transitions , which have to be taken into account in order to achieve best accuracy and stability of interferometers .
we have first quantitatively evaluated the effect on the resonance condition for each off - resonant line : the other counter - propagating transition , which can not be avoided in the retro - reflected design , and co - propagating transitions arising mainly from imperfections in the polarisation .
then , we have measured the impact of this two photon light shift on the phase of two atom interferometers : a gravimeter and a gyroscope .
in particular , we show that this shift is an important source of systematic errors for acceleration measurements .
nevertheless , it can be measured accurately by modulating the raman laser power and/or the pulse durations .
our study has also shown that it can impact the stability of the two sensors if the polarization and/or the power of the raman lasers fluctuate .
the tpls appears as a drawback of using retro - reflected raman laser beams .
but , as it can be well controlled , it does not reduce the benefit from this geometry , whose key advantage is to drastically limit the bias due to wave - front aberrations , which is larger and more difficult to extrapolate to zero .
this study can be extended to other two photon transitions ( like bragg transitions ) and to other possible polarization configurations ( when using linear instead of circular polarizations ) , when using the doppler effect to select the transition .
if the signal is generated from the subtraction of the phase shifts of two independent atomic clouds , e.g. gradiometers or gyroscopes , perfect common mode rejection , is required to suppress this effect . in our case this means a perfect overlap of atomic trajectories .
finally , the two photon light shift can be drastically reduced by increasing the doppler effect and/or using colder atoms , allowing to reduce the rabi frequency during the raman pulses . by contrast , for set - up with intrinsic small doppler effect , as for space applications @xcite
, this effect becomes extremely large and has to be taken into account in the design of the experiment .
we express our gratitude to f. biraben for pointing out this effect and to p. cheinet for earlier contributions to the gravimeter experiment .
we would like to thank the institut francilien pour la recherche sur les atomes froids ( ifraf ) , the european union ( finaqs contract ) , the dlgation gnrale pour larmement ( dga ) and the centre national dtudes spatiales ( cnes ) for financial supports .
b. c. , j. l.g .
and t. l. thank dga for supporting their work .
30 a. peters , k. y. chung , s. chu , metrologia * 38 * 25 ( 2001 ) .
t. l. gustavson , a. landragin , m. a. kasevich , class .
quantum grav .
* 17 * 1 - 14 ( 2000 ) .
b. canuel , f. leduc , d. holleville , a. gauguet , j. fils , a. virdis , a. clairon , n. dimarcq , ch.j .
bord , a. landragin , and p. bouyer , phys .
lett . * 97 * , 010402 ( 2006 ) .
j. m. mcguirk , g. t. foster , j. b. fixler , m. j. snadden , m. a. kasevich , phys.rev .
a * 65 * 033608 ( 2002 ) .
m. kasevich , d.s .
weiss , e.riis , k. moler , s. kasapi and s. chu , phys .
lett . * 66 * 2297 ( 1991 ) .
bord , phys .
a * 140 * , 10 ( 1989 ) .
j. b.fixler , g. t. foster , j. m. mcguirk and m. a. kasevich , science magazine * vol . 315 .
5808 * , 74 ( 2007 ) .
g. lamporesi , a. bertoldi , l. cacciapuoti , m. prevedelli , and g. m. tino , phys .
lett . * 100 * , 050801 ( 2008 ) .
a. wicht , j.m .
hensley , e. sarajlic and s. chu , physica scripta * t102 * , 82 ( 2002 ) .
p. clad , e. de mirandes , m. cadoret , s. guellati - khlifa , c. schwob , f. nez , l. julien and f. biraben , phys .
lett . * 96 * , 033001 ( 2006 ) .
g. genevs _
ieee trans . on instr . and meas .
* 54 * , 850 ( 2005 ) .
j. fils , f. leduc , p. bouyer , d. holleville , n. dimarcq , a. clairon and a. landragin , eur.phys .
j. d * 36*,257 - 260 ( 2005 ) .
a. landragin and f. pereira dos santos , in _ proc . of the enrico fermi international school of physics _
* 168 * ( 2007 ) , in press , arxiv : 0808.3837 p. cheinet , f. pereira dos santos , t. petelski , j. le gout , j. kim , k. t. therkildsen , a. clairon and a. landragin appl .
b * 84 * , 643 ( 2006 ) .
d. s. weiss , b. c. young and s. chu , appl .
b * 59 * , 217 ( 1994 ) .
p. clad , e. de mirandes , m. cadoret , s. guellati - khlifa , c. schwob , f. nez , l. julien and f. biraben , phys .
a * 74 * , 052109 ( 2006 ) .
p. cheinet , b. canuel , f. pereira dos santos , a. gauguet , f. leduc , a. landragin , ieee trans . on instr . and meas .
* 57 * , 1141 ( 2008 ) .
j. le gout , t. e. mehlstubler , j. kim , s. merlet , a. clairon , a. landragin and f. pereira dos santos , appl .
b * 92 * , 133 - 144 ( 2008 ) .
see for instance apll .
b * 84 * , number 4 , special issue : `` quantum mechanics for space application : from quantum optics to atom optics and general relativity '' , p. wolf _
_ , accepted for publication in experimental astronomy , arxiv : 0711.0304v5 , w. ertmer _
submitted for publication in experimental astronomy , and references therein . | we study the influence of off - resonant two photon transitions on high precision measurements with atom interferometers based on stimulated raman transitions .
these resonances induce a two photon light shift on the resonant raman condition .
the impact of this effect is investigated in two highly sensitive experiments : a gravimeter and a gyroscope - accelerometer .
we show that it can lead to significant systematic phase shifts , which have to be taken into account in order to achieve best performances in term of accuracy and stability . |
in the field of hydrogeology , many interesting concepts are related to finding the lag between two time series .
for example , it is often hypothesized that for a seepage lake there is a significant time lag between net precipitation ( precipitation minus water loss through evaporation and runoff ) and the water levels over time , while such a lag for a drainage lake is often nonexistent or insignificant .
seepage lakes are hydraulically isolated from surface water features and primarily fed by groundwater and direct precipitation .
drainage lakes are typically connected to a network of streams and rivers ( wisconsin department of natural resources , 2009 ) .
another example , which is our motivating example , is the relationship between precipitation and water levels of a shallow well in an unconfined aquifer versus water levels in a relatively deeper well in a semi - confined aquifer .
this relationship is particularly important to water resource managers and groundwater modelers who need to accurately quantify groundwater recharge into aquifers , for developing water - supply - plans for sustainable use of aquifers .
groundwater recharge , defined as entry of water into the saturated zone , is influenced by a wide variety of factors including vegetation , topography , geology , climate , and soils ( dripps , 2003 , dripps , hunt and anderson 2006 ) .
groundwater recharge , which is a small percentage of the precipitation that eventually reaches the water table , is one of the most difficult parameters to quantify .
this is because processes such as evaporation , transpiration and infiltration through unsaturated subsurface must first be estimated to determine the amount of water lost after a rainfall event .
often times , groundwater models are developed by estimating the groundwater recharge using empirical relationships or as a percentage of precipitation .
it is a common practice to use groundwater recharge as a calibration parameter , meaning the recharge value that provides the best calibration to the model is selected as representative for the watershed simulated . for temporal simulations , the lag time between a rainfall event and groundwater recharge into deeper aquifers are often ignored .
although the underlying hydrogeological theory supports the existence of above time lags between time series , evidence based on empirical data for such lags have been typically assessed using visual inspection ( e.g. westoff _ et al _ , 2010 in a different hydrogeological context ) or cross - correlations ( levanon _ et al _ , 2016 ) in hydrogeological literature .
cross - correlation method is essentially a parametric method , where certain parameters has to be estimated under the transfer - function - model framework and certain assumptions ( such as joint bivariate stationarity of the two time series ) has to be met ( see chapter 14 , wei 2006 ) .
also diagnostic checking for model adequacy ( such as whether the noise series and the input series are independent - see again chapter 14 , wei 2006 for the definition of the noise series and input series referred to ) has to be done before cross - correlograms are plotted , although such checking are rarely done in practice . in this paper , we propose a non - parametric method to quantify the time lag using a simple adaptation of the visibility graph algorithm ( vga ) , which is an algorithm that converts a time series into a graph and was developed by physicists and seen mainly only within the physics literature so far ( lacasa , 2008 , lacasa and luque , 2010 , nunez _ et al _ 2012 ) . the method that we propose may be summarized as follows . in the proposed method
, we consider one of the time series ( e.g. water levels observed in a well ) as a reference time series and create time shifted copies of the other time series of interest ( e.g. precipitation ) .
we then use vga to convert all the time series ( original , copies and the reference ) to graphs and their corresponding adjacency matrices , and compare the copies of the latter time series with that of the reference .
the ` distance measure ' that is used for the comparisons is the usual @xmath0 metric distance ( based on the frobenius norm ) between two matrices .
we identify the copy of the latter time series for which this distance is minimized compared to the reference , and we define the time shift corresponding to this copy as the time lag between the orginal two time series .
more details about vga and our adaptation to the time lag problem is provided in the next section using mathematical notation . in section 3
we present results from simulations conducted to essentially identify an appropriate sample size and also to assess the performance of the method when values are missing .
section 4 illustrates the application of the proposed method to real hydrogeologic datasets , where we also present a strategy to assess the uncertainty related to the lag estimated . finally in the last section
, we make our concluding remarks .
let us denote the two hydrogeological time series that we are interested in , namely precipitation and water levels , by @xmath1 and @xmath2 ( or simply @xmath3 and @xmath4 ) , respectively .
in order to find the time lag between the two time series , as a first step we fix one of the series , say @xmath4 , and obtain time - shifted copies of the other series , @xmath5 the key step in our methodology is the conversion of all the above time series into graphs based on the visibility graph algorithm .
graphs are mathematical constructs that are used to study relationships among various objects . in graph models
the objects of interest are modeled as nodes or vertices and the relationships among the objects are modeled using edges or lines connecting the vertices .
+ etc .
denote the time points as well as the corresponding nodes in the visibility graph.,width=624,height=360 ] visibility graph algorithm ( lacasa , 2008 , lacasa and luque , 2010 , nunez _ et al _ 2012 ) is a relatively novel method that extends usefulness of the techniques and focus of mathematical graph theory to characterize time series .
it has been shown that the visibility graph inherits several properties of the time series , and its study reveals nontrivial information about the time series itself .
figure 1 top panel illustrates how the visibility algorithm works .
the time series plotted in the upper panel is an approximate sine series ; specifically , a sine series with gaussian white noise added .
the values at 24 time points are plotted as vertical bars .
one may imagine these vertical bars as , for example , buildings along a straight line in a city landscape ( i.e. a city block ) .
each node in the associated visibility graph ( shown in the bottom panel ) corresponds to each time point in the series .
so , the graph in figure 1 has 24 nodes .
we draw a link or an edge between a pair of nodes , say @xmath6 and @xmath7 , if the visual line of sight from the top of the building ( vertical bar ) situated at @xmath6 towards the top of the building / bar at @xmath7 is not blocked by any intermediate buildings - that is , if we were to draw a line from the top of the vertical bar at @xmath6 to the top of the vertical bar at @xmath7 , it should not intersect any intermediate vertical bars . visibility lines corresponding to the edges in the graph are plotted as dotted lines in the figure in the upper panel . for example , there is no edge between @xmath8 and @xmath9 since the line of sight ( not shown ) between the top points of the vertical bars at these two time points is blocked by the vertical bar at @xmath10 . on the other hand , there is an edge between @xmath11 and @xmath10 since the corresponding visibility line ( shown as a dotted line ) does not intersect the vertical bar at @xmath8 .
+ more formally , the following visibility criteria can be established : two arbitrary data values ( @xmath12 , @xmath13 ) and ( @xmath14 ) will have visibility , and consequently will become two connected nodes of the associated graph , if any other data ( @xmath15 ) placed between them fulfills : @xmath16 this simple intuitive idea has been proven useful practically because of certain nice features exhibited by the graphs generated by this algorithm .
first of all they are connected , since each node is connected to at least its neighbors .
secondly , there is no directionality between the edges , so that the graph obtained is undirected . in addition , the visibility graph is invariant under rescaling of the horizontal and vertical axes and under horizontal and vertical translations . in other words ,
the graph is invariant under affine transformations of the original time series data .
+ in mathematical notation any graph with @xmath17 nodes could be represented by its @xmath18 adjacency matrix @xmath19 which consists of @xmath20 s and @xmath21 s .
the @xmath22 element of @xmath19 is @xmath21 if there is an edge connecting the @xmath23 and the @xmath24 node , @xmath20 otherwise .
two graphs , @xmath25 and @xmath26 , can be be compared by the metric `` distance '' , @xmath27 between their corresponding adjacency matrices , @xmath28 and @xmath28 . here ,
@xmath29 , called the frobenius norm of a matrix , is the square root of the sum of the squares of the elements of the matrix ; that is , the square root of the trace of the product of the matrix with itself . + our proposed method to assess the time lag between the two hydrogeological time series @xmath3 and @xmath4 using the visibility graph approach is as follows : convert the @xmath4 time series into a visibility graph and obtain its corresponding adjacency matrix , @xmath30 .
consider time - shifted copies of the @xmath3 time series , @xmath31 , each shifted in time by a lag from the set @xmath32 .
convert these time - shifted copies of @xmath3 into their visibility graphs and obtain the corresponding adjacency matrices @xmath33 .
we determine the copy @xmath34 for which the frobenius norm @xmath35 is minimized .
the time lag between the two original hydrogeological series is then taken as @xmath36 .
+ we further illustrate our method using the plots in figure 2 . the time series in the top panel
, @xmath37 is an approximately a series of values based on the sine function obtained using the following r codes : + `
n ` @xmath38 ` 50 ` + ` ts.a ` @xmath38 ` 100*sin(2*pi*(80/1000)*n ) + rnorm(n , 0 , 25 ) ` + the time series , @xmath39 , plotted in the middle panel of figure 2 is derived from @xmath37 as follows : + ` ts.b ` @xmath38 ` ( 1/3)*c(ts.a[3:n ] , ts.a[1:2 ] ) + rnorm(n , 0 , 5 ) ` + that is , @xmath39 is derived by shifting @xmath37 to the left by two units , by reducing the amplitude to one - third that of @xmath37 , and adding some gaussian noise . in other words ,
@xmath37 and @xmath39 have roughly the same shape although their amplitudes are different and one is shifted by two time units relative to the other as seen in the figure .
one may think of @xmath37 and @xmath39 as two time series one affecting the other ( since , @xmath39 is shifted to the left , physically we would think of @xmath39 affecting @xmath37 ) ; e.g. @xmath39 as precipitation and @xmath37 as water levels .
physically , water levels and precipitation never take negative values ; so , if one really wants to think of @xmath37 and @xmath39 as water levels and precipitation , one could think of them as mean - subtracted and scaled appropriately .
we considered time - shifted copies of @xmath39 with time - shifts from the following set : @xmath40 vga was applied and adjacency matrices for the corresponding graphs were obtained .
distance - measure based on the frobenius norm for the time - shifted copies of @xmath39 compared to the reference @xmath37 , are plotted in the bottom panel of figure 2 .
the distance - measure is minimized at 2 , which was the lag that we set _ a priori_. thus , in this illustrative example , the lag was correctly identified by the method that we proposed .
we conducted monte carlo simulations to assess the performance of the vga - based method as we varied some of the parameters of the two time series @xmath37 and @xmath39 considered in the previous section .
the parameters that we considered were _ a _ ) the ratio of the amplitudes between the two simulated series @xmath37 and @xmath39 , _ b _ ) the variance for the noise term `` rnorm(n , 0 , * ) ` ' in the series @xmath37 ( indicated by @xmath41 ) and _ c _ ) the variance for the noise term `` rnorm(n , 0 , * ) ` ' in the series @xmath39 . for each simulation scenario considered in this section ( that is , for each set of the above parameters ) , 1000 pairs of @xmath37 and @xmath39 were generated , and for each pair time lag was assessed based on the proposed method and compared with the lag that was set _ a priori_. the performance of the method was assessed based on the percentage of times that the _ a priori _ lag was correctly identified .
the _ a priori _ lags that we considered for each scenario were @xmath42 and @xmath43 ; we assumed that in typical examples from physical sciences , @xmath44 will be a small lag and @xmath43 will be a very large lag .
the reason for considering the ratio of amplitudes was that even if two physical time series ( especially , hydrogeological time series ) are roughly of the same shape with only a lag between them , their amplitudes ( i.e. roughly their sizes ) are often vastly different . for @xmath37 and @xmath39 used in the introductory illustrative example , the ratio of their amplitudes was @xmath45 .
one of the questions that was addressed in our simulations was whether our method was still good if we changed this ratio drastically , e.g. to @xmath46 .
another question that we thought should be addressed is that whether the proposed method works only for nice periodic time series such as the ` sine series ' .
increasing the variance for the noise term in @xmath37 makes it less like a sine series. finally , increasing the variance of the noise term in @xmath39 makes the shape of @xmath39 quite different from that of @xmath37 , and by doing so in our simulations we also addressed the performance of the method in such scenarios .
.performance of the method when the ratio of amplitudes between @xmath37 and @xmath39 was @xmath45 , the noise term in @xmath37 was @xmath47 and the noise term for @xmath39 was @xmath48 .
[ cols="<,^,^,^,^",options="header " , ] the performance of the proposed method with both locf and mean imputation was near perfect when only @xmath49 of the values
( that is , 9 out of 180 ) were missing .
this was true regardless of whether the values were missing for only one time series or for both , and also true across all _ a priori _ set lags , 2 , 5 , 10 and 15 .
when @xmath50 of the values ( that is , 18 out of 180 ) were missing for only one time series , the method did very well under both locf and mean imputation for all lags .
when @xmath50 of the values were missing for both time series , the performance was still very good when the lags were large ( 10 or 15 ) ; when the lags were small ( 2 or 5 ) , the performance with both imputation methods was still good but not as good as when the lags were large .
for example , when @xmath50 values were missing and when the lag was 2 , the performance with locf was @xmath51 and @xmath52 , respectively , depending on whether the values were missing for only one time series or both ; the corresponding values for lag 10 , on the other hand , were even better : @xmath53 and @xmath51 . with @xmath54 missing values ( 27 out of 180 ) , the performance was still good ( that is , in the range @xmath55 ) with locf and mean imputation , for lags 2 , 5 , and 10 , irrespective of whether it was missing for only one or for both time series ( although , of course , if it was missing only for one time series , it was better ) . however , when the _ a priori _ set lag was 15 , the performance with locf was weak ( @xmath56 ) , when @xmath54 values were missing for both time series ; it was still good ( @xmath57 ) with locf when only one time series had @xmath54 missing values , and with mean imputation also ( @xmath52 and @xmath53 ) .
with @xmath58 missing values the method worked well under both types of imputations and for all lags , only when one time series had missing values .
when both time series had @xmath58 missing values , the performance of locf was not good with small lags ( @xmath56 for lag 2 and @xmath59 for lag 5 ) and got worse for larger lags ( @xmath60 for lag 10 and @xmath61 for lag 15 ) .
the performance with mean imputation was slightly better ( @xmath62 and @xmath63 , for lags 2 , 5 , 10 and 15 , respectively ) but still not quite up to the mark . in summary , based on the above simulation results , we consider it quite safe to use the proposed method in conjunction with either of the imputation methods if it is only @xmath49 values missing for only one time series or for both .
with @xmath64-@xmath54 values missing , the imputation methods give good results only if it is missing for one time series ; if it is missing for both , then it is not very safe to say that imputations will work , but still reasonably safe . with about @xmath58 of the values missing for both time series ,
it is definitely not recommended to use the proposed method with either of the imputations although it may be somewhat acceptable if it is missing for only one time series .
also , in general , we observed that the performance with mean imputation was slightly better except for one or two scenarios .
if the statistical practitioner has a preference of one method over the other , it may still be recommended to use both for the proposed method , at least as a sensitivity analysis .
finally , we emphasize again the point made in the beginning of the section , that if large chunks of data are missing at a stretch then the imputation methods wo nt work ; in such cases , it is better to focus the analysis on other chunks of data with no or very sparse missing values .
in this section , we present the results from an application of the proposed method on real hydrogeological time series
. in southwest florida , two of the shallow aquifers that are tapped for water supply are the unconfined water table aquifer and the semi - confined lower tamiami aquifer .
these aquifers are considered as sources of limited supply and regulated by the south florida water management district ( sfwmd , 2015a ) .
water table aquifer is generally less than 50 feet thick and consists of less permeable unconsolidated sediments at the upper portion and relatively permeable limestone at the basal portion .
the water table aquifer is hydraulically separated from the lower tamiami aquifer by about 15 to 30 feet of confining beds that consists of low permeable strata ( bonita springs marl member and caloosahatchee clay member ) .
the top of lower tamiami aquifer is generally between 60 and 80 feet below land surface .
this aquifer extends to 100 to 150 feet below grade and generally consists of sandy , biogenic limestone and calcareous sandstone ( sfwmd , 2015b ) .
this aquifer is one of the primary sources of public water supply in southwest florida . to understand the lag responses of rainfall in these shallow aquifers are important for water management . for this study , in order to determine the lag responses within these two aquifers due to rainfall events , we utilized daily water level data recorded in the water table aquifer and lower tamiami aquifer .
it is relevant to note that data collected in shorter frequencies ( e.g. hourly ) are ideal for `` lag '' related studies .
hourly water level data was available from the water table aquifer well as well as the lower tamiami aquifer in the study area ; however , precipitation data was available only on a daily basis . in order to have both water level data and precipitation in the same time interval , we averaged the water levels to a daily average .
the daily - averaged data was used solely for illustration of the statistical technique presented in this paper .
data was available from july 1st , 2010 till june 29th 2016 , but water level data were missing for the following time intervals : january 5th , 2011 to april 8th , 2012 , july 10th , 2012 to october 1st , 2012 , april 2nd , 2013 to september 30th , 2013 , and finally between april 2nd , 2014 to june 29th , 2014 .
complete data was available between june 30th , 2014 and june 29th , 2016 ( 731 days ) ; we analyzed this data for our illustration since this was the largest available time period with no missing data .
the water level and precipitation data that were analyzed are graphically presented on figure 3 .
vga method was applied to all the times series plotted in figure 3 , and frobenius distance between the corresponding pairs of adjacency matrices are plotted in figure 4 . for both unconfined water table aquifer and
semi - confined lower tamiami aquifer , the frobenius distance is minimized at lag 2 .
this makes hydrogeological sense since , although one aquifer is a bit deeper than the other , considering the difference in total depths between the wells is roughly only about 40 feet , the water level response in the relatively deeper well may be observed in a matter of hours . in figure 4 , for both plots , we note that although the frobenius distnace for lag 1 is not the minimum it is close to the minimum compared to that of the other lags .
thus , in order to check whether the minimum was attained at lag 2 just due to chance , we need to quantify the uncertainty regarding the estimate . since naive resampling strategies like bootstrap would create independent samples ( that is , with autocorrelations near zero ) , we used the following subsampling strategy that would preserve the original autocorrelations .
we set a time - window - size , say 100 , and use the consecutive data points for water levels and precipitation in that window , and apply the proposed method to this sub - sampled data , as we did for the original data .
first we place the left end of the window on the first time point of the original data , conduct the analysis , find the lag , and then move the window one unit to the right and repeat the analysis .
we continue this process iteratively until the right end of the window touches the final time point for the original data . thus
, with a window - size of 100 , and with a time series of length 731 , we will have 631 iterations , and for each iteration , we will have a lag between the pair of time series under consideration , so that we will have 631 lags at the end .
the histograms for the 631 lags obtained using this iterative process are plotted in the top panel of figure 5 .
the highest frequency for the lower tamiami aquifer is at lag 2 consistent with our finding for the original time series with 731 points .
however , interestingly , the highest frequency for the water table aquifer was at lag 1 , although the frequency for lag 2 is almost as high .
now the question arises whether this reversal was due to the size of the window ( 100 ) , which is quite smaller than the length of the original series ( 731 ) .
so , we repeated the analysis with a window - size of 365 ( that is , 366 iterations ) .
the results of the second iterative analysis are shown in the bottom panel of figure 5 ; in this case , the highest frequency for both aquifers is at lag 2 .
we repeated the analysis using a window size of 50 ( 681 iterations ) and 25 ( 706 iterations ) ; in these analyses , only lag 2 appeared for all windows , so that the histogram will look like a single bar at 2 ( of height 50 or 25 , respectively ) , and no bars at any other lags .
since this is simple enough to convey without a histogram , we did nt plot the histograms for window - sizes 50 and 25 . based on these results , and on hydrogeological sense
, we would conclude that on the average , the water levels rise and fall 2 days after a corresponding fluctuation in precipitation , for both the aquifers .
the analysis presented in this section suggests that it is critical to quantify the uncertainty prior to making conclusions and that the selected window size can influence the conclusions .
quantifying time lags between two time series data , where one affects the other , is important for modelers of many physical phenomena , especially in hydrogeology .
we propose an approach based on a simple extension of the visibility graph algorithm .
we conducted simulations to assess the performance of the proposed method under different scenarios , and determined that the method worked well under reasonable settings .
based on simulations we were also able to recommend sample size necessary to conduct the proposed analysis , and the maximum percentage of missing values under which the method will still work reasonably well with imputations .
we also illustrated the method by applying it real data of water levels from aquifers and precipitation levels , and emphasized the importance of quantifying the uncertainty related to estimate of the lag .
25 dripps , w.r . ( 2003 ) , the spatial and temporal variability of ground water recharge . ph.d .
, department of geology and geophysics , university of wisconsin madison .
dripps , w.r . , hunt , r.j . and anderson , m.p .
( 2006 ) , estimating recharge rates with analytic element models and parameter estimation .
ground water , 44 : 4755 .
lacasa , l. , luque , b. , ballesteros , f. , luque , j. , and nuno , j.c .
( 2008 ) from time series to complex networks : the visibilty graph .
usa 105 , 13 , 4972 - 4975 .
lacasa , l. , luque .
b. ( 2010 ) mapping time series to networks : a brief overview of visibility algorithms .
computer science research and technology vol 3 ( edited by nova publisher ) , isbn : 978 - 1 - 61122 - 074 - 2 .
levanon , e. , shalev , e. , yechieli y , gvirtzman h. ( 2016 ) fluctuations of fresh - saline water interface and of water table induced by sea tides in unconfined aquifers .
advances in water resources , volume 96 , 34 - 42 nunez , a. , lacasa , l. , luque , b. ( 2012 ) visibility algorithms : a short review .
graph theory ( edited by intech ) , isbn 979 - 953 - 307 - 303 - 2 .
south florida water management district ( sfwmd ) ( 2015a ) , water use permit applicant s handbook , part b. south florida water management district ( sfwmd ) ( 2015b ) , tech pub ws-35 , hydrogeologic unit mapping update for the lower west coast water supply planning area wei , w.w.s . , ( 2006 ) time series analysis : univariate and multivariate methods ( 2nd edition ) .
westoff , m.c . ,
bogaard , t.a . ,
savenije , h.h.g .
( 2010 ) quantifying the effect of in - stream rock clasts on the retardation of heat along a stream advances in water resources , volume 33 , issue 11 , 1417 - 1425 wisconsin department of natural resources , pub - fh-800 2009 ,
wisconsin lakes " | estimating the time lag between two hydrogeologic time series ( e.g. precipitation and water levels in an aquifer ) is of significance for a hydrogeologist - modeler . in this paper , we present a method to quantify such lags by adapting the visibility graph algorithm , which converts time series into a mathematical graph .
we present simulation results to assess the performance of the method .
we also illustrate the utility of our approach using a real world hydrogeologic dataset .
time series , visibility graph algorithm , hydrogeology , aquifer water level , precipitation |
optical nano - mechanical systems that couple optical degree of freedom to the mechanical motion of a cantilever have been subject of increasing investigation @xcite . in these optomechanical systems coupling
is obtained via radiation pressure inside a cavity @xcite , or via quantum dots @xcite or ions @xcite .
recently , it is made possible to couple mechanical resonators with the ensembles of atoms , where , the interaction is mediated by the field inside the cavity which couples the mechanical resonators to the internal degrees of freedom of the atoms @xcite , or to motional degrees of the freedom of the atoms @xcite causing effects , ( _ e.g _ , cooling of the mechanical resonator via bath of atoms @xcite ) . in quantum meteorology
, various targets , such as , measurement of displacement with larger accuracy @xcite and the detection of gravitational waves @xcite are set as milestone achievements .
engineering entanglement in nano - mechanical systems is useful in quantum technologies @xcite .
the possibility of entangling the electromagnetic field with motional degree of freedom of mechanical systems have been explored in various approaches @xcite . in this paper
, we consider a hybrid optomechanical system which consists of a bose - einstein condensate ( bec ) trapped inside a fabry - perot cavity with a vibrating end mirror ( nano - mechanical resonators ) driven by a single mode optical field .
the intracavity field mediates the interaction of nanomechanical resonator with collective oscillations of the atomic density .
hence , the motional degrees of the nano - mechanical resonator indirectly couples to motional degrees of the freedom of the bose - einstein condensates via optical field inside the cavity .
therefore , intracavity field acts as nonlinear spring between collective atomic density and nano - mechanical resonator .
we show that ( i ) the mechanical vibration of the nanomechanical resonator is entangled to the motional degree of the freedom of the bec in the steady state .
( ii ) furthermore , we analyze the steady - state entanglement as a function of temperature , coupling strength between bec and field , moving mirror and field , and power of input driving laser field . in section 2 ,
we model the system and explain its interaction with the environment . in section 3 , we calculate the quantum langevin equations for our system and solve the dynamics . in section 3 ,
we quantify the steady - state entanglement between mechanical resonator and bose - einstein condensate .
later , in section 5 , we explain the possibility to experimentally measure the generating entanglement .
finally , in section 6 , we provide concluding remarks .
we consider a fabry - perot cavity with a moving end mirror driven by a single - mode optical field of frequency @xmath0 , and bec of @xmath1-two level atoms are trapped inside the fabry - perot cavity @xcite .
the condensate atoms placed in the cavity observe one - dimensional optical lattice , formed by the oppositely propagated electromagnetic field inside the cavity .
we consider that the atom - field detuning @xmath2 is very large , spontaneous emission is negligible , and , as a consequence we adiabatically eliminate the internal excited state dynamics of the atoms .
in addition , we also consider that the atomic densities are low enough that one can neglect the two - body interactions . in a weakly interacting regime
, the recent experiment @xcite suggests that only the first two symmetric momentum side modes are excited with momentum @xmath3k , where k is the wave number of the field .
moreover , we assume that the vibrating end mirror of the optical cavity of length @xmath4 is performing harmonic oscillations with frequency @xmath5 along the cavity axis ( x - axis ) .
[ [ section ] ] the hamiltonian of the hybrid optomechanical system formed by the bec , the intracavity field , and the vibrating end mirror of the cavity ( nanomechanical resonator ) in the rotating frame at the laser frequency @xmath6 is given by @xcite @xmath7 where , @xmath8 , @xmath9 and @xmath10 . here
@xmath11 ) is the annihilation ( creation ) operator of the single - mode optical field , @xmath12 , where @xmath13 is the optical lattice barrier depth per photon and represents the atomic back action on the field @xcite , and @xmath14 shows the driving of the cavity field with amplitude @xmath15 related to the laser power @xmath16 by @xmath17 , where @xmath18 is the decay rate of the photons in to the associated outgoing modes .
moreover , @xmath19 is the annihilation ( creation ) operator of the bec mode with frequency @xmath20 , where @xmath21 is the recoiled frequency .
in addition , @xmath22 and @xmath23 are frequency and annihilation ( creation ) operator of the nano - mechanical resonator , respectively .
we assume that the frequency of the nanomechanical resonator is less than the free spectral range , ( _ i.e _ , @xmath24 ( v is the speed of light ) .
therefore , scattering of photons into others modes , except the driven mode , is neglected @xcite . here
, @xmath25 accounts for the interaction between nano - mechanical resonator and intracavity field and is given by @xmath26 here , the coupling strength between nano - mechanical resonator and light radiation pressure is defined by @xmath27 , where @xmath28 is the zero - point motion of the mechanical mode of mass @xmath29 . moreover ,
the interaction of the bec with intracavity mode is described by the hamiltonian @xmath30 given by @xmath31 where , @xmath32 is described the strength of interaction between bec mode and intracavity field .
[ [ section-1 ] ] in order to describe the complete dynamics of the system we include the dissipation effects .
in addition to the dynamics described by the hamiltonian in eq.([ham ] ) , the system is exposed to the random noise forces due to quantum fluctuations of the radiation field and fluctuation of the phononic heat bath associated to the mechanical resonator .
we neglect the thermal effects of the atomic cloud and assume that the vacuum noise associated with the cavity field is markovian in nature with decay rate @xmath18 and noise operator @xmath33 of the input field which obeys the following correlation functions , @xmath34 here , @xmath35^{-1}$ ] is the equilibrium occupation number of the optical oscillator . for optical frequency
@xmath36 we consider @xmath37 and set @xmath38 .
[ [ section-2 ] ] the motion of the nano - mechanical resonator is affected due to thermal bath is brownian and non - morkovian in nature @xcite .
the quantum effects on mechanical resonator are only observed in the limit of very high mechanical quality factor @xmath39 , the brownian noise operator can be modeled as markovian with the decay rate of the mechanical mode is @xmath40 .
therefore , the noise operator @xmath41 can be characterized as @xmath42 where , @xmath43^{-1}$ ] is the equilibrium thermal occupation number of the mechanical resonator .
in order to describe the complete dynamics of the subsystems involved in this problem , an adequate choice is to use the formalism of the quantum langevin equations .
therefore , the heisenberg - langevin equation of motion for the intracavity mode , mechanical mode and bosonic field operator can be written as @xmath44 where _ dot _ denotes the time derivatives and for simplicity we omit the _ hat _ symbol from the operators .
these are the nonlinear quantum langevin equations and dynamics is complicated . in the following we linearized the operators around the steady state values , @xmath45 , @xmath46 , @xmath47 . here
, we assume that the fluctuation operators @xmath48 , @xmath49 and @xmath50 have zero mean .
the steady state value of the intracavity mode is @xmath51 , where the total effective detuning is @xmath52 for the sake of simplicity we assume that the field is real positive and this can be achieved by adjusting the phase of the laser field .
similarly , the steady state value of the bec and mechanical resonator modes are @xmath53\left<\mathrm{c^\dag c}\,\right>_{\mathrm{ss}}$ ] and @xmath54\left<\mathrm{c^\dag c}\,\right>_{\mathrm{ss}}$ ] respectively .
[ [ section-3 ] ] we linearize the langevin equations of motion given in eq.([langeq ] ) , and assume that pump field is intense and keep terms only up to first order in the fluctuation operators .
we rewrite each heisenberg operator in eq.([langeq ] ) as a sum of steady state value and fluctuation operator with zero mean value .
therefore , the linear set of equations are , @xmath55 the linearized quantum langevin equations show that the fluctuations of mechanical resonator and bec are now coupled to the cavity field quadrature fluctuations by the effective couplings @xmath56 and @xmath57 , which can be made very large by increasing the amplitude @xmath58 of the intracavity field .
linearized quantum langevin equations ( [ linrz ] ) and their corresponding hermitian conjugate form a system of six first - order coupled operator equations , for which the ruth - hurwitz criteria implies that the system will be stable only if the following stability condition is satisfied @xmath59 in the following we consider only red - detuning regime ( @xmath60 ) and from now on we assume that the above stability condition is satisfied . the quantum langevin equations ( [ linrz ] ) are linear in creation and annihilation operators .
we transform to the quadratures , _ i.e _ , @xmath61 .
the system of linearized equations of motion can be written in compact matrix form as @xmath62 , where , @xmath63 is the vector of the quadrature fluctuations , and in superscript @xmath64 describes transpose of matrix .
furthermore , @xmath65 is the vector corresponding to noises , whereas @xmath66 is drift matrix .
since the quantum noises are white in nature and the dynamics is linearized , hence the state of the system will be a zero mean gaussian state , and therefore completely determined by the covariance matrix @xmath67 . in order to find the steady state covariance matrix @xmath68 , we solve the linearized quantum langevin equations as did in @xcite .
in steady state the covariance matrix fulfills the lyapunove equation @xmath69 eq.([lymp ] ) is the linear matrix equation and can straight forwardly be solved .
however , the general exact expression is too cumbersome and not reported here .
one can extract all the information about the steady state of the system from the correlation matrix .
we compute the entanglement between mechanical mode and the bec mode in steady - state by tracing out the cavity mode .
the steady state entanglement is determined by computing the logarithmic negativity @xmath70 from the corresponding covariance matrix @xmath68 .
we consider an example where the length of the cavity @xmath71 and laser input power @xmath72 @xcite .
the effective mass and resonance frequency of the mechanical resonator are @xmath73 and @xmath74 .
the decay rate of the mechanical resonator is @xmath75 and cavity finesse @xmath76 . for these values of the parameters ,
the coupling @xmath77 between mechanical resonator is in the order of @xmath78 .
moreover , the interaction of the optical field with bose - einstein condensate is kept small , so that bogoliubove mode expansion becomes possible .
[ [ section-4 ] ] measurement of the entanglement between the mechanical resonator and bose - einstein condensate , requires as to compute @xmath79 , which is obtained by tracing out the cavity mode , _
i.e _ , removing the rows and columns of @xmath68 which correspond to the cavity mode .
the reduce state is still gaussian and fully characterized by @xmath80 matrix @xmath81 . in order to measure the entanglement between mechanical mirror and intracavity filed ,
we consider the logarithmic negativity @xmath82 . in the case of continuous variable ( cv )
, @xmath70 can be defined as @xcite @xmath83\,,\label{55}\ ] ] where , @xmath84^{1/2}\}^{1/2}$ ] , with @xmath85 , is the smallest symplectic eigenvalue .
however , the second eigenvalue @xmath86^{1/2}\}^{1/2}>>1/2 $ ] at any value of the parameters .
therefore , it has no effect on the non - separability of the state @xcite .
moreover , the correlation matrix @xmath87 in @xmath88 block form can be written as @xmath87=@xmath89 .
it is clear from eq.([55 ] ) that the @xmath70 is the decreasing function of @xmath90 and it quantifies how much two gaussian states are entangled .
the gaussian state gets entangled only if @xmath91 , and it is simon s necessary and sufficient entanglement non - positive partial transpose criterion of the gaussian states @xcite , and this condition can also be written as @xmath92 .
[ [ section-5 ] ] in fig.[ent ] we show that the entanglement between nano - mechanical resonator and bose - einstein condensate quantify by logarithmic negativity @xmath70 as a function of the coupling of the mechanical resonator and bose - einstein condensate with field in the cavity . in fig.[e=2 ] , we have @xmath93 for @xmath94 and @xmath95 . in fig.[e=3 ] , we find that the @xmath96 for @xmath94 and @xmath97 .
+ [ [ section-6 ] ] in fig.[pzet ] , we also study the robustness of the steady state entanglement of the nano - mechanical resonator to the bose - einstein condensate with respect to several parameters in system .
fig.[p ] shows the dependence of @xmath70 on the input laser power @xmath16 and normalized detuning @xmath98 for @xmath99 .
fig.[z ] shows the dependence of @xmath70 on optomechanical coupling rate @xmath77 and normalized detuning @xmath98 for @xmath99 .
moreover , in fig.[e ] we plot the @xmath70 as function of normalized detuning @xmath98 for two different values of the mass m of the mechanical resonator , _ i.e _ , black solid and red dashed lines for @xmath100 and @xmath101 respectively .
one can easily observe that @xmath70 exists only within a finite interval of values of the @xmath102 around @xmath103 .
now , we investigate the effects of temperature on the entanglement .
we consider @xmath70 of the evolved system at a fixed value of detuning . at high temperature ,
thermal fluctuation always suppress the entanglement of the coupling systems .
therefore , the logarithmic negativity is the decreasing function of temperature . in fig.[t ] , we plot the logarithmic negativity as a function of temperature at a fixed value of the detuning @xmath104 .
the black solid line refers to mass @xmath100 and red dashed line refers to mass @xmath101 .
fig.[t ] shows that @xmath70 monotonically decreases with temperature .
moreover , it is noted that @xmath70 is also very sensitive to the mass of mechanical resonator and quickly decay for large value of the mass .
[ [ section-7 ] ] latter , we find the effective coupling between mechanical resonator and bose - einstein condensate .
we consider the regime in which the dynamics of the cavity mode remains unperturbed due to the motion of the mechanical mode and bec mode , and slowly mediates the interaction between the two .
therefore , we chose the regime where the cavity mode can be eliminated adiabatically .
the hamiltonian corresponding to the linearized quantum langevin equations for the fluctuations operator @xmath105 and @xmath106 is given by @xmath107 the corresponding parameteric regime for the fast cavity mode dynamics is , @xmath108 or @xmath109 . due to the second condition , cavity - mediated coherent dynamics is destroyed .
therefore , we only consider the regime where @xmath102 takes large values , and the fluctuation quadratures of the cavity mode adiabatically follow the dynamics of positions fluctuations of the mechanical resonator and bec mode .
we further assume that both the mechanical resonator and bec modes to be on resonance , _ i.e _
, @xmath110 . on eliminating the photon degree of freedom adiabatically
we get the following effective hamiltonian @xmath111 where , @xmath112 , @xmath113 and @xmath114 .
it is noted that the effective interaction @xmath115 between mechanical resonator and bose - einstein condensate is increased as the coupling of mechanical oscillator and bec is increased with the intracavity field .
the effective interaction between a mechanical resonator and bose - einstein condensate can be described via the hamiltonian @xmath116 , from eq.([eff ] ) @xmath117 where @xmath118 is the effective coupling strength between mechanical mode and bec .
this hamiltonian is analogous to the interaction of two optical fields in a nonlinear medium generating the 2-modes squeezed state @xcite . the first term in eq.([squeez ] )
corresponds to an energy exchange take place between mechanical and atomic modes . the second term in eq.([squeez ] ) accounts for the down - conversion interaction , describes the creation and annihilation of the atomic and mechanical modes phonons in pairs , which corresponds to entangling the atomic and mechanical modes .
for @xmath119 or greater , the mechanical resonator and bose - einstein condensate are entangled , and the parameters are chosen such that this condition is fulfilled .
for experimental realization of the generated entanglement , one has to measure several quadrature correlations @xcite as have been experimentally measured for the entanglement of the two optical modes @xcite . however , in our case we consider another fabry perot cavity adjacent to first one and is driven by a weak laser field , a scheme of this kind has been discussed in @xcite . in addition , we assume that the movable mirror is perfect reflector at both sides so there is entanglement between the optical modes of the two cavities .
the equation of motion of the annihilation operator @xmath120 of the optical mode of the second cavity similar to the linearized version of eq.([linrz ] ) is @xmath121 where @xmath122 , @xmath123 and @xmath124 are the cavity decay rate , effective detuning and the input noise of the second cavity mode respectively .
in addition , @xmath125 and @xmath126 are the effective coupling rates of the second cavity mode to the mechanical resonator and bose - einstein condensate respectively .
moreover , we assume @xmath127 and @xmath128 , @xmath125 , @xmath126 . therefore , in the rotating frame at @xmath129 , eq.([c2 ] ) for slow variables @xmath130 can be rewritten as @xmath131\nonumber\\ & & + \sqrt{2\,\kappa_1 } \tilde{\mathrm{c}}_{\mathrm{in1}}. \label{slow}\end{aligned}\ ] ] here , we assume @xmath132 and neglect the fast oscillating terms at frequency @xmath133 . on eliminating the cavity mode adiabatically , eq.([slow ] ) can be written as @xmath134 + \sqrt{\dfrac{2}{\kappa_1}}\tilde{\mathrm{c}}_{\mathrm{in1}}.\ ] ] according to the stranded input - output relation @xcite , @xmath135 , the output field is given by @xmath136 + \tilde{\mathrm{c}}_{\mathrm{in1}}. \label{homo}\ ] ] from eq.([homo ] ) we can measure the @xmath137 by homodyning the output of the second cavity mode .
therefore , all the entries of the correlation matrix can be determined by measuring the correlation between the output of the two cavities , hence from the correlation matrix logarithmic negativity can be calculated numerically .
in conclusion , we discuss quantum correlation in a system which consists of bose - einstein condensates trapped inside a fabry - perot cavity with a moving end mirror driven by single mode optical field .
we describe a scheme to generate the steady state entanglement of the motional degree of freedom of the nano - mechanical resonator ( end mirror of the cavity ) and collective oscillation of the atomic density of the bose - einstein condensate .
moreover , it is observed that the entanglement generated between bose - einstein condensate and nano - mechanical resonator is very sensitive with respect to temperature and persist only upto @xmath138 and @xmath139 for mass @xmath140 and @xmath141 , respectively .
the generated entanglement is measured by considering the second optical cavity driven by a weak laser field .
t. j. kippenberg and k. j. vahala , science * 321 * , 1172 ( 2008 ) .
v. b. braginsky , _ measurement of weak forces in physics experiments _ ( university of chicago press , chicago , 1977 ) .
s. mancini , v. giovannetti , d. vitali and p. tombesi , phys .
* 88 * , 120401 ( 2002 ) .
j. zhang , k. peng , and s. l. braunstein , phys .
a * 68 * , 013808 ( 2003 ) .
l. tian and p. zoller , phys .
lett . * 93 * , 266403 ( 2004 ) .
a. naik , o. buu , m. d. lahaye , a. d. armour , a. a. clerk , m. p. blencowe and k. c. schwab , nature ( london ) * 443 * , 193 ( 2006 ) .
h. ian , z. r. gong , y. x. liu , c. p. sun , and f. nori , phys .
a * 78 * , 013824 ( 2008 ) . c. genes , d. vitali , and p. tombesi , phys .
a * 77 * , 050307(r ) ( 2008 ) ; k. j@xmath142hne , c. genes , k. hammerer , m. wallquist , e. s. polzik and p. zoller , phys .
a * 79 * , 063819 ( 2009 ) .
d. meiser and p. meystre , phys .
a * 73 * , 033417 ( 2006 ) . c. genes , h. ritsch , and d. vitali , phys .
rev . a * 80 * , 061803(r ) ( 2009 ) .
d. rugar , r. budakian , h. j. mamin and b. w. chui , nature ( london ) * 430 * , 329 ( 2004 ) .
v. braginsky and s. p. vyatchanin , phys .
a * 293 * , 228 ( 2002 ) .
m. a nielsen and i. l. chuang , _ quantum computation and quantum information _
( cambridge university press , cambridge , 2000 ) . m. paternostro , d. vitali , s. gigan , m. s. kim , c. brukner , j. eisert , and m. aspelmeyer , phys .
lett . * 99 * , 250401 ( 2007 ) .
d. vitali , s. gigan , a. ferreira , h. r. b@xmath143hm , p. tombesi , a. guerreiro , v. vedral , a. zeilinger , and m. aspelmeyer , phys .
* 98 * , 030405 ( 2007 ) .
j. esteve , c. gross , a. weller , s. giovanazzi and m. k. oberthaler , nature ( london ) * 455 * , 1216 ( 2008 ) ; ferdinand brennecke , tobias donner , stephan ritter , thomas bourdel , michael khl and tilman esslinger , ibid , * 450 * , 268 ( 2007 ) .
f. brennecke , s. ritter , t. donner , t. esslinger , science * 322 * , 235 ( 2008 ) .
d. nagy , p. domokos , a. vukics , and h. ritsch , eur .
j. d * 55 * , 659 ( 2009 ) . c. k. law , phys .
a * 51 * , 2537 ( 1995 ) . c. maschler and h. ritsch , phys .
* 95 * , 260401 ( 2005 ) .
g. adesso , a. serafini , and f. illuminati , phys .
a * 70 * , 022318 ( 2004 ) .
r. simon , phys .
lett . * 84 * , 2726 ( 2000 ) .
l. m. duan , g. giedke , j. i. cirac , and p. zoller , phys .
lett . * 84 * , 2722 ( 2000 ) . z.
y. ou , s. f. pereira , h. j. kimble , and k. c. peng , phys . rev .
* 68 * , 3663 ( 1992 ) .
j. laurat , g. keller , j. a. oliveira - huguenin , c. fabre , t. coudreau , a. serafini , g. adesso , and f. illuminati , j. opt .
b : quantum semiclass . opt .
* 7 * , s577 ( 2005 ) . c. w. gardiner and p. zoller,_quantum
( berlin : springer - verlag 2000 ) . | we analyze the steady - state entanglement between bose - einstein condensate trapped inside an optical cavity with a moving end mirror ( nanomechanical resonator ) driven by a single mode laser .
the quantized laser field mediates the interaction between the bose - einstein condensate and nanomechanical resonator . in particular , we study the influence of temperature on the entanglement of the coupled system , and note that the steady - state entanglement is fragile with respect to temperature . |
delmotte @xcite proved that the transition density of the simple random walk on a graph satisfies gaussian bounds and the parabolic harnack inequality holds if all the balls have regular volume growth and satisfy a poincar inequality .
barlow @xcite relaxed these conditions by imposing them only on all _ large enough _ balls , and showed that they imply large time gaussian bounds and the elliptic harnack inequality for large enough balls .
later , barlow and hambly @xcite proved that the parabolic harnack inequality also follows from barlow s conditions .
barlow @xcite verified these conditions for the supercritical cluster of bernoulli percolation on @xmath0 , which lead to the almost sure gaussian heat kernel bounds and parabolic harnack inequality . by using stationarity and heat kernel bounds ,
the quenched invariance principle was proved in @xcite , which lead to many further results about supercritical bernoulli percolation , including the local central limit theorem @xcite and the fact that the dimension of harmonic functions of at most linear growth is @xmath2 @xcite .
the independence property of bernoulli percolation was essential in verifying barlow s conditions , and up to now it has been the only example of percolation model for which the conditions were verified . on the other hand ,
once the conditions are verified , the derivation of all the further results uses rather robust methods and allows for extension to other stationary percolation models .
the aim of this paper is to develop an approach to verifying barlow s conditions for infinite clusters of percolation models , which on the one hand , applies to supercritical bernoulli percolation , but on the other , does not rely on independence and extends beyond models which are in any stochastic relation with bernoulli percolation . motivating examples for us are random interlacements , vacant set of random interlacements , and the level sets of the gaussian free field @xcite . in all these models , the spatial correlations decay only polynomially with distance , and classical peierls - type arguments do not apply . a unified framework to study percolation models with strong correlations
was proposed in @xcite , within which the shape theorem for balls @xcite and the quenched invariance principle @xcite were proved . in this paper
we prove that barlow s conditions are satisfied by infinite percolation clusters in the general setting of @xcite .
in particular , all the above mentioned properties of supercritical bernoulli percolation extend to all the models satisfying assumptions from @xcite , which include supercritical bernoulli percolation , random interlacements at every level in any dimension @xmath3 , the vacant set of random interlacements and the level sets of the gaussian free field in the regime of local uniqueness .
let @xmath4 be an infinite connected graph with the vertex set @xmath5 and the edge set @xmath6 . for @xmath7 , define the weights @xmath8 and extend @xmath9 to the measure on @xmath6 and @xmath10 to the measure on @xmath5 . for functions
@xmath11 and @xmath12 , let @xmath13 and @xmath14 , and define @xmath15 by @xmath16 for @xmath17 .
let @xmath18 be the graph distance on @xmath4 , and define @xmath19 .
we assume that @xmath20 for all @xmath21 and @xmath22 .
in particular , this implies that the maximal degree in @xmath4 is bounded by @xmath23 .
we say that a graph @xmath4 satisfies the _ volume regularity _ and the _ poincar inequality _ if for all @xmath21 and @xmath24 , @xmath25 and , respectively , @xmath26 , with some constants @xmath27 and @xmath28 .
graphs satisfying these conditions are very well understood .
delmotte proved in @xcite the equivalence of such conditions to gaussian bounds on the transition density of the simple random walk and to the parabolic harnack inequality for solution to the corresponding heat equation , extending results of grigoryan @xcite and saloff - coste @xcite for manifolds . under the same assumptions
, he also obtained in @xcite explicit bounds on the dimension of harmonic functions on @xmath4 of at most polynomial growth .
results of this flavor are classical in geometric analysis , with seminal ideas going back to the work of de giorgi @xcite , nash @xcite , and moser @xcite on the regularity of solutions of uniformly elliptic second order equations in divergence form .
the main focus of this paper is on random graphs , and more specifically on random subgraphs of @xmath0 , @xmath29 .
because of local defects in such graphs caused by randomness , it is too restrictive to expect that various properties ( e.g. , poincar inequality , gaussian bounds , or harnack inequality ) should hold globally .
an illustrative example is the infinite cluster @xmath30 of supercritical bernoulli percolation @xcite defined as follows . for @xmath31
$ ] , remove vertices of @xmath0 independently with probability @xmath32 .
the graph induced by the retained vertices almost surely contains an infinite connected component ( which is unique ) if @xmath33 , and contains only finite components if @xmath34 .
it is easy to see that for any @xmath35 with probability @xmath36 , @xmath30 contains copies of any finite connected subgraph of @xmath0 , and thus , none of the above global properties can hold .
barlow @xcite proposed the following relaxed assumption which takes into account possible exceptional behavior on microscopic scales .
[ def : vgb ] ( ( * ? ? ?
* definition 1.7 ) ) let @xmath37 , @xmath38 , and @xmath39 be fixed constants . for @xmath22 integer and @xmath21
, we say that @xmath40 is @xmath41-_good _ if @xmath42 and the weak poincar inequality @xmath43 holds for all @xmath44 .
we say @xmath45 is @xmath41-_very good _ if there exists @xmath46 such that @xmath47 is @xmath41-good whenever @xmath48 , and @xmath49 .
[ rem : wpi : minimum ] for any finite @xmath50 and @xmath51 , the minimum @xmath52 is attained by the value @xmath53 .
for a very good ball , the conditions of volume growth and poincar inequality are allowed to fail on microscopic scales .
thus , if all large enough balls are very good , the graph can still have rather irregular local behavior . despite that , on large enough scales it looks as if it was regular on all scales , as the following results from @xcite illustrate .
let @xmath54 and @xmath55 be the discrete and continuous time simple random walks on @xmath4 .
@xmath56 is a markov chain with transition probabilities @xmath57 , and @xmath58 is the markov process with generator @xmath59 . in words ,
the walker @xmath56 ( resp . , @xmath58 ) waits a unit time ( resp .
, an exponential time with mean @xmath36 ) at each vertex @xmath60 , and then jumps to a uniformly chosen neighbor of @xmath60 in @xmath4 . for @xmath21 , we denote by @xmath61 ( resp . ,
@xmath62 ) the law of @xmath56 ( resp . , @xmath58 ) started from @xmath60 .
the transition density of @xmath56 ( resp . , @xmath58 ) with respect to @xmath10 is denoted by @xmath63}{\mu_y}$ ] ( resp . , @xmath64}{\mu_y}$ ] ) .
the first implications of definition [ def : vgb ] are large time gaussian bounds for @xmath65 and @xmath66 .
[ thm : hk : vgb ] ( ( * ? ? ? * theorem 5.7(a ) ) and ( * ? ? ?
* theorem 2.2 ) ) let @xmath21 .
if there exists @xmath67 such that @xmath45 is @xmath41-very good with @xmath68 for each @xmath69 , then there exist constants @xmath70 such that for all @xmath71 and @xmath72 , @xmath73 @xmath74 where @xmath75 stands for either @xmath65 or @xmath76 .
the next result gives an elliptic harnack inequality .
[ thm : ehi : vgb ] ( ( * ? ? ? * theorem 5.11 ) ) there exists @xmath77 such that for any @xmath21 and @xmath78 , if @xmath79 is @xmath41-very good with @xmath80 , then for any @xmath81 , and @xmath82 nonnegative and harmonic in @xmath83 , @xmath84 in fact , more general parabolic harnack inequality also takes place .
( for the definition of parabolic harnack inequality see , e.g. , ( * ? ? ?
* section 3 ) . )
[ thm : phi : vgb ] ( ( * ? ? ? * theorem 3.1 ) ) there exists @xmath85 such that for any @xmath21 , @xmath78 , and @xmath86 , if @xmath87 is @xmath41-very good with @xmath88 , then for any @xmath89 , the parabolic harnack inequality ( in both discrete and continuous time settings ) holds with constant @xmath90 for @xmath91\times\ballg(y , r)$ ] .
in particular , the elliptic harnack inequality also holds .
next result is about the dimension of the space of harmonic functions on @xmath4 with at most polynomial growth .
[ thm : vgb : hf : poligrowth ] ( ( * ? ? ?
* theorem 4 ) ) let @xmath21 .
if there exists @xmath67 such that @xmath45 is @xmath41-very good for each @xmath69 , then for any positive @xmath92 , the space of harmonic functions @xmath93 with @xmath94 is finite dimensional , and the bound on the dimension only depends on @xmath92 , @xmath95 , @xmath23 , @xmath37 , @xmath38 , and @xmath96 . the notion of very good balls is most useful in studying random subgraphs of @xmath0 . up to now , it was only applied to the unique infinite connected component of supercritical bernoulli percolation , see @xcite .
barlow ( * ? ? ?
* section 2 ) showed that on an event of probability @xmath36 , for every vertex of the infinite cluster , all large enough balls centered at it are very good .
thus , all the above results are immediately transfered into the almost sure statements for all vertices of the infinite cluster . despite the conditions of definition [ def : vgb ] are rather general , their validity up to now has only been shown for the independent percolation .
the reason is that most of the analysis developed for percolation is tied very sensitively with the independence property of bernoulli percolation .
one usually first reduces combinatorial complexity of patterns by a coarse graining , and then balances the complexity out by exponential bounds coming from the independence , see , e.g. , ( * ? ? ?
* section 2 ) .
the main purpose of this paper is to develop an approach to verifying properties of definition [ def : vgb ] for random graphs which does not rely on independence or any comparison with bernoulli percolation , and , as a result , extending the known results about bernoulli percolation to models with strong correlations .
our primal motivation comes from percolation models with strong correlations , such as random interlacements , vacant set of random interlacements , or the level sets of the gaussian free field , see , e.g. , @xcite .
[ rem : vgb ] * the lower bound of theorem [ thm : hk : vgb ] can be slightly generalized by following the proof of ( * ? ? ?
* theorem 5.7(a ) ) .
let @xmath97 $ ] and @xmath98 .
if there exists @xmath67 such that @xmath45 is @xmath41-very good with @xmath99 for each @xmath69 , then for all @xmath100 , @xmath101 the constants @xmath102 and @xmath103 are the same as in , in particular , they do not depend on @xmath104 and @xmath105 . for @xmath106 and @xmath107
, we recover .
( there is a small typo in the statements of ( * ? ? ? * theorem 5.7(a ) ) and ( * ? ? ?
* theorem 2.2 ) :
@xmath108 should be replaced by @xmath109 . ) + indeed , the proof of ( * ? ? ?
* theorem 5.7(a ) ) is reduced to verifying assumptions of ( * ? ? ?
* theorem 5.3 ) for some choice of @xmath110 .
the original choice of barlow is @xmath111 , and it implies . by restricting the choice of @xmath112 as above ,
one notices that all the conditions of ( * ? ? ?
* theorem 5.3 ) are satisfied by @xmath113 , implying . * in order to prove the lower bound of for the same range of @xmath114 s as in the upper bound , one needs to impose a stronger assumption on the regularity of the balls @xmath45 ( see , for instance , ( * ? ? ?
* definition 5.4 ) of the exceedingly good ball and ( * ? ? ? * theorem 5.7(b ) ) ) .
the recent result of ( * ? ? ?
* theorem 1.10 ) states that the volume doubling property and the poincar inequality satisfied by large enough balls are equivalent to certain partial gaussian bounds ( and also to the parabolic harnack inequality in large balls ) .
* under the assumptions of theorem [ thm : phi : vgb ] , various estimates of the heat kernels for the processes @xmath56 and @xmath58 killed on exiting from a box are given in ( * ? ? ?
* theorem 2.1 ) .
* theorem [ thm : vgb : hf : poligrowth ] holds under much weaker assumptions , although reminiscent of the ones of definition [ def : vgb ] ( see ( * ? ? ?
* theorem 4 ) ) .
roughly speaking , one assumes that the conditions from definition [ def : vgb ] hold with @xmath112 only sublinear in @xmath110 , i.e. , a volume growth condition and the weak poincar inequality should hold only for macroscopic subballs of @xmath45 .
we consider the measurable space @xmath115 , @xmath1 , equipped with the sigma - algebra @xmath116 generated by the coordinate maps @xmath117 . for any @xmath118 ,
we denote the induced subset of @xmath0 by @xmath119 we view @xmath120 as a subgraph of @xmath0 in which the edges are drawn between any two vertices of @xmath120 within @xmath121-distance @xmath36 from each other , where the @xmath121 and @xmath122 norms of @xmath123 are defined in the usual way by @xmath124 and @xmath125 , respectively . for @xmath126 and @xmath127
, we denote by @xmath128 the closed @xmath129-ball in @xmath0 with radius @xmath130 and center at @xmath60 .
[ def : setr ] for @xmath131 $ ] , we denote by @xmath132 , the set of vertices of @xmath120 which are in connected components of @xmath120 of @xmath121-diameter @xmath133 . in particular
, @xmath134 is the subset of vertices of @xmath120 which are in infinite connected components of @xmath120 .
on @xmath135 we consider a family of probability measures @xmath136 with @xmath137 , satisfying the following assumptions and from @xcite .
parameters @xmath95 , @xmath138 , and @xmath139 are considered fixed throughout the paper , and dependence of various constants on them is omitted .
an event @xmath140 is called _ increasing _
( respectively , _ decreasing _ ) , if for all @xmath141 and @xmath142 with @xmath143 ( respectively , @xmath144 ) for all @xmath145 , one has @xmath146 .
* _ ( ergodicity ) _ for each @xmath147 , every lattice shift is measure preserving and ergodic on @xmath148 . * _ ( monotonicity ) _ for any @xmath149 with @xmath150 , and any increasing event @xmath151 , @xmath152 \leq \mathbb p^{u'}[g]$ ] . * _ ( decoupling ) _ let @xmath153 be an integer and @xmath154
. for @xmath155 , let @xmath156 be decreasing events , and @xmath157 increasing events .
there exist @xmath158 and @xmath159 such that for any integer @xmath160 and @xmath161 satisfying @xmath162 if @xmath163 , then @xmath164 \leq \mathbb p^{\widehat u}\left[a_1\right ] \cdot \mathbb p^{\widehat u}\left[a_2\right ] + e^{-\funcp(l ) } , \ ] ] and @xmath165 \leq \mathbb p^u\left[b_1\right ] \cdot \mathbb p^u\left[b_2\right ] + e^{-\funcp(l ) } , \ ] ] where @xmath166 is a real valued function satisfying @xmath167 for all @xmath168 . * _ ( local uniqueness ) _ there exists a function @xmath169 such that for each @xmath147 , @xmath170 and for all @xmath147 and @xmath78 , the following inequalities are satisfied : @xmath171 \geq 1 - e^{-\funcs(u , r ) } , \ ] ] and @xmath172 \geq 1 - e^{-\funcs(u , r ) } .\ ] ] * _ ( continuity ) _ let @xmath173 $ ] .
the function @xmath174 is positive and continuous on @xmath175 . * the use of assumptions , , and will not be explicit in this paper .
they are only used to prove likeliness of certain patterns in @xmath134 produced by a multi - scale renormalization , see .
( of course , they are also used in already known results of theorems [ thm : chd ] and [ thm : qip ] ) .
roughly speaking , we use repeatedly on multiple scales for a convergent sequence of parameters @xmath176 and use and to establish convergence of iterations . *
if the family @xmath177 , @xmath147 , satisfies , then a union bound argument gives that for any @xmath147 , @xmath177-a.s .
, the set @xmath134 is non - empty and connected , and there exist constants @xmath178 such that for all @xmath78 , @xmath179 \geq 1 - c_1 \cdot e^{-c_2\cdot(\log r)^{1+\consts } } .\ ] ] here we briefly list some motivating examples ( already announced earlier in the paper ) of families of probability measures satisfying assumptions and .
all these examples were considered in details in @xcite , and we refer the interested reader to ( * ? ? ?
* section 2 ) for the proofs and further details . *
bernoulli percolation with parameter @xmath180 $ ] corresponds to the product measure @xmath177 with @xmath181 = 1 - \mathbb p^u[\atom(x ) = 0 ] = u$ ] .
the family @xmath177 , @xmath147 , satisfies assumptions and for any @xmath1 and @xmath182 , see @xcite .
* random interlacements at level @xmath183 is the random subgraph of @xmath0 , @xmath3 , corresponding to the measure @xmath177 defined by the equations
@xmath184 = e^{-u\cdot \mathrm{cap}(k)},\qquad \text{for all finite } k\subset\z^d,\ ] ] where @xmath185 is the discrete capacity .
it follows from @xcite that the family @xmath177 , @xmath147 , satisfies assumptions and for any @xmath137 .
curiously , for any @xmath183 , @xmath120 is @xmath177-almost surely connected @xcite , i.e. , @xmath186 . * vacant set of random interlacements at level @xmath183 is the complement of the random interlacements at level @xmath187 in @xmath0 .
it corresponds to the measure @xmath177 defined by the equations @xmath188
= e^{-u\cdot \mathrm{cap}(k)},\qquad \text{for all finite } k\subset\z^d.\ ] ] unlike random interlacements , the vacant set undergoes a percolation phase transition in @xmath187 @xcite .
if @xmath189 then @xmath177-almost surely @xmath134 is non - empty and connected , and if @xmath190 , @xmath134 is @xmath177-almost surely empty .
it is known that the family @xmath191 , @xmath147 , satisfies assumptions for any @xmath137 @xcite , for any @xmath192 @xcite , and for some @xmath192 @xcite . *
the gaussian free field on @xmath0 , @xmath193 , is a centered gaussian field with covariances given by the green function of the simple random walk on @xmath0 .
the excursion set above level @xmath194 is the random subset of @xmath0 where the fields exceeds @xmath93 .
let @xmath195 be the measure on @xmath196 for which @xmath120 has the law of the excursion set above level @xmath93 .
the model exhibits a non - trivial percolation phase transition @xcite .
if @xmath197 then @xmath195-almost surely @xmath134 is non - empty and connected , and if @xmath198 , @xmath134 is @xmath195-almost surely empty .
it was proved in @xcite that the family @xmath199 , @xmath200 , satisfies assumptions and for any @xmath137 , and for some @xmath137 .
the last three examples are particularly interesting , since they have _ polynomial decay of spatial correlations _ and can not be studied by comparison with bernoulli percolation on any scale . in particular ,
many of the methods developed for bernoulli percolation do not apply . as we see from the examples , assumptions and are satisfied by all the @xmath201 models through their _ whole _
. however , assumption is currently verified for the whole range of interesting parameters only in the cases of bernoulli percolation and random interlacements , and only for a non - empty subset of interesting parameters in the last two examples .
we call all the parameters @xmath187 for which @xmath177 satisfies the regime of _ local uniqueness _
( since under , there is a unique giant cluster in each large box ) .
it is a challenging open problem to verify if the regime of local uniqueness coincides with the supercritical phase for the vacant set of random interlacements and the level sets of the gaussian free field .
a positive answer to this question will imply that all the results of this paper hold unconditionally also for the last two considered examples through their whole supercritical phases .
below we recall some results from @xcite about the large scale behavior of graph distances in @xmath134 and the quenched invariance principle for the simple random walk on @xmath134 .
both results are formulated in the form suitable for our applications .
[ thm : chd ] ( ( * ? ? ?
* theorem 1.3 ) ) let @xmath1 and @xmath202 .
assume that the family of measures @xmath177 , @xmath147 , satisfies assumptions and .
let @xmath147 .
there exist @xmath203 with @xmath204 = 1 $ ] , constants @xmath205 , @xmath206 , and @xmath207 all dependent on @xmath187 and @xmath208 , and random variables @xmath209 , @xmath210 , such that for all @xmath211 and @xmath212 , * @xmath213 , * for all @xmath214 and @xmath215 , @xmath216 * for all @xmath217 and @xmath22 , @xmath218\leq c_{\scriptscriptstyle { \ref{thm : chd}}}\cdot e^{-c_{\scriptscriptstyle { \ref{thm : chd}}}\cdot ( \log r)^{1 + \consts}},\ ] ] where @xmath219 is defined in . for @xmath220 ,
let @xmath221 $ ] be the space of continuous functions from @xmath222 $ ] to @xmath223 , and @xmath224 the borel sigma - algebra on it .
let @xmath225 [ thm : qip ] ( ( * ? ? ?
* theorem 1.1 , lemma a.1 , and section 5 ) ) let @xmath1 .
assume that the family of measures @xmath177 , @xmath147 , satisfies assumptions and .
let @xmath147 and @xmath220 .
there exist @xmath226 with @xmath227 = 1 $ ] and a non - degenerate matrix @xmath228 , such that for all @xmath229 , * there exists @xmath230 such that @xmath231 is harmonic on @xmath232 , and @xmath233 , * the law of @xmath234 on @xmath235,\mathcal w_t)$ ] converges weakly ( as @xmath236 ) to the law of brownian motion with zero drift and covariance matrix @xmath237 .
in addition , if reflections and rotations of @xmath0 by @xmath238 preserve @xmath177 , then the limiting brownian motion isotropic , i.e. , @xmath239 with @xmath240 .
[ rem : qip ] ( * ? ? ?
* theorem 1.1 ) is stated for the ( `` blind '' ) random walk which jumps to a neighbor with probability @xmath241 and stays put with probability @xmath242 . since the blind walk and the simple random walk are time changes of each other , the invariance principle for one process implies the one for the other ( see , for instance , ( * ? ? ?
* lemma 6.4 ) ) .
the main contribution of this paper is theorem [ thm : vgb : main ] , where we prove that under the assumptions and , all large enough balls in @xmath134 are very good in the sense of definition [ def : vgb ] .
this result has many immediate applications , including gaussian heat kernel bounds , harnack inequalities , and finiteness of the dimension of harmonic functions on @xmath134 with prescribed polynomial growth , see theorems [ thm : hk : vgb ] , [ thm : phi : vgb ] , [ thm : ehi : vgb ] , [ thm : vgb : hf : poligrowth ] .
in fact , all the results from @xcite can be easily translated from bernoulli percolation to our setting , since ( as also pointed out by the authors ) their proofs only rely on ( some combinations of ) stationarity , gaussian heat kernel bounds , and the invariance principle . among such results
are estimates on the gradient of the heat kernel ( theorem [ thm : hk : grad ] ) and on the green function ( theorem [ thm : gf : bounds ] ) , which will be deduced from the heat kernel bounds by replicating the proofs of ( * ? ? ?
* theorem 6 ) and ( * ? ? ?
* theorem 1.2(a ) ) , the fact that the dimension of at most linear harmonic functions on @xmath134 is @xmath2 ( theorem [ thm : hf : d+1 ] ) , the local central limit theorem ( theorem [ thm : localclt ] ) , and the asymptotic for the green function ( theorem [ thm : gf : asymp ] ) , which we derive from the heat kernel bounds and the quenched invariance principle by mimicking the proofs of ( * ? ? ?
* theorem 5 ) , ( * ? ? ?
* theorem 1.1 ) , and ( * ? ? ?
* theorem 1.2(b , c ) ) .
we begin by stating the main result of this paper .
[ thm : vgb : main ] let @xmath1 and @xmath243 .
assume that the family of measures @xmath177 , @xmath147 , satisfies assumptions and .
let @xmath147 .
there exist @xmath244 with @xmath245 = 1 $ ] , constants @xmath37 , @xmath38 , @xmath96 , @xmath246 , and @xmath247 all dependent on @xmath187 and @xmath248 , and random variables @xmath249 , @xmath210 , such that for all @xmath250 and @xmath212 , * @xmath251 , * for all @xmath252 , @xmath253 is @xmath41-very good with @xmath254 , * for all @xmath217 and @xmath22 , @xmath255\leq c_{\scriptscriptstyle { \ref{thm : vgb : main}}}\cdot e^{-c_{\scriptscriptstyle { \ref{thm : vgb : main}}}\cdot ( \log r)^{1 + \consts}},\ ] ] where @xmath219 is defined in .
theorem [ thm : vgb : main ] will immediately follow from a certain isoperimetric inequality , see definition [ def : vrb ] , claim [ cl : rb - gb ] , and proposition [ prop : verygoodbox ] . this isoperimetric inequality is more than enough to imply the weak poincar inequality that we need .
in fact , as we learned from a discussion with jean - dominique deuschel , it implies stronger sobolev inequalities , and may be useful in situations beyond the goals of this paper ( see , e.g. , ( * ? ? ?
* section 3 ) ) .
[ cor : vgb : main ] theorem [ thm : vgb : main ] immediately implies that all the results of theorems [ thm : hk : vgb ] , [ thm : phi : vgb ] , [ thm : ehi : vgb ] , and [ thm : vgb : hf : poligrowth ] hold almost surely for @xmath256 . since the constants @xmath37 , @xmath38 , and @xmath96 in the statement of theorem [ thm : vgb : main ] are deterministic , all the constants in theorems [ thm : hk : vgb ] , [ thm : phi : vgb ] , [ thm : ehi : vgb ] , and [ thm : vgb : hf : poligrowth ] are also deterministic .
combining corollary [ cor : vgb : main ] with theorem [ thm : chd ] and remark [ rem : vgb](1 ) , we notice that the quenched heat kernel bounds of theorem [ thm : hk : vgb ] hold almost surely for @xmath257 with @xmath258 replaced by @xmath259 in , , and . since we will use the quenched heat kernel bounds often in the paper , we give a precise statement here . [ thm : hk : dzd ] let @xmath1 .
assume that the family of measures @xmath177 , @xmath147 , satisfies assumptions and .
let @xmath147 and @xmath260 .
there exist @xmath261 with @xmath262 = 1 $ ] , constants @xmath178 , @xmath263 , and @xmath264 , and random variables @xmath265 , @xmath210 , such that for all @xmath266 and @xmath212 , * @xmath267 , * for all @xmath268 and @xmath269 , @xmath270 @xmath271 where @xmath75 stands for either @xmath65 or @xmath76 , and @xmath272 for either @xmath273 or @xmath259 , * for all @xmath217 and @xmath22 , @xmath274\leq c_{\scriptscriptstyle { \ref{thm : hk : dzd}}}\cdot e^{-c_{\scriptscriptstyle { \ref{thm : hk : dzd}}}\cdot ( \log r)^{1 + \consts}},\ ] ] where @xmath219 is defined in . in the applications of theorem
[ thm : hk : dzd ] in this paper , _ we always take _ @xmath275 ( the original choice of barlow ) and omit the dependence on @xmath105 from the notation . for instance , we will always write @xmath276 meaning @xmath277 . any other choice of @xmath105 would also do .
it is well known that the parabolic harnack inequality of theorem [ thm : phi : vgb ] implies hlder continuity of caloric functions ( e.g. , @xmath65 and @xmath66 ) , see ( * ? ? ?
* proposition 3.2 ) , in particular , by corollary [ cor : vgb : main ] this is true almost surely for @xmath257 .
the next result is a sharp bound on the discrete gradient of the heat kernel , proved in ( * ? ? ? * theorem 6 ) for supercritical bernoulli percolation using an elegant entropy argument .
[ thm : hk : grad ] let @xmath1 .
assume that the family of measures @xmath177 , @xmath147 , satisfies assumptions and .
let @xmath147 .
there exist constants @xmath178 , such that for all @xmath278 and @xmath279 , @xmath280 \leq \frac{c_1}{n^{d+1}}\cdot e^{-c_2\cdot \frac{\dist_{\z^d}(x , y)^2}{n}}.\ ] ] the heat kernel bounds of theorem [ thm : hk : dzd ] imply also the following quenched estimates on the green function @xmath281 for almost all @xmath257 .
it is proved in ( * ? ? ?
* theorem 1.2 ) for supercritical bernoulli percolation , but extension to our setting is rather straightforward .
[ thm : gf : bounds ] let @xmath3 .
assume that the family of measures @xmath177 , @xmath147 , satisfies assumptions and .
let @xmath147 .
there exist constants @xmath282 such that for all @xmath283 and distinct @xmath284 , if @xmath285 , then @xmath286 the remaining results are derived from the gaussian heat kernel bounds and the quenched invariance principle . in the setting of supercritical bernoulli percolation ,
all of them were obtained in @xcite , but all the proofs extend directly to our setting .
we begin with results about harmonic functions on @xmath134 .
it is well known that theorems [ thm : vgb : main ] and theorem [ thm : ehi : vgb ] imply the almost sure liouville property for positive harmonic functions on @xmath134 .
the absence of non - constant sublinear harmonic functions on @xmath134 is even known assuming just stationary of @xmath120 ( see ( * ? ? ?
* theorem 3 and discussion below ) ) .
in particular , it implies the uniqueness of the function @xmath287 in theorem [ thm : qip](a ) .
the following result about the dimension of at most linear harmonic functions is classical on @xmath0 .
it was extended to supercritical bernoulli percolation on @xmath0 in ( * ? ? ?
* theorem 5 ) .
[ thm : hf : d+1 ] let @xmath1 .
assume that the family of measures @xmath177 , @xmath147 , satisfies assumptions and .
let @xmath147 .
there exist @xmath288 with @xmath289 = 1 $ ] such that for all @xmath290 , the dimension of the vector space of harmonic functions on @xmath232 with at most linear growth equals @xmath2 . since the parabolic harnack inequality for solutions to the heat equation on @xmath134 implies hlder continuity of @xmath66 and @xmath65 , it is possible to replace the weak convergence of theorem [ thm : qip ] by pointwise convergence .
* theorems 4.5 and 4.6 ) give general sufficient conditions for the local central limit theorem on general graphs .
they were verified in ( * ? ? ?
* theorem 1.1 ) for supercritical bernoulli percolation .
theorems [ thm : qip ] and [ thm : hk : dzd ] allow to check these conditions in our setting leading to the following ( same as for bernoulli percolation ) result . for @xmath291 , @xmath292 , the gaussian heat kernel with covariance matrix @xmath237
is defined as @xmath293 where @xmath294 is the transpose of @xmath60 .
[ thm : localclt ] let @xmath1 .
assume that the family of measures @xmath177 , @xmath147 , satisfies assumptions and .
let @xmath147 , @xmath295 $ ] , and @xmath220 .
there exist @xmath296 with @xmath297 = 1 $ ] , and a non - degenerate covariance matrix @xmath228 such that for all @xmath298 , @xmath299 where @xmath300 stands for @xmath301 or @xmath302 , @xmath303 is @xmath36 if @xmath304 and @xmath305 otherwise , and @xmath306 is the closest point in @xmath134 to @xmath307 .
theorems [ thm : hk : dzd ] and [ thm : localclt ] imply the following asymptotic for the green function , extending results of ( * ? ? ?
* theorem 1.2(b , c ) ) to our setting . for a covariance matrix @xmath237 ,
let @xmath308 be the green function of a brownian motion with covariance matrix @xmath237 .
in particular , if @xmath239 , then @xmath309 for all @xmath310 , where @xmath311 stands for the euclidean norm on @xmath223 .
[ thm : gf : asymp ] let @xmath3 .
assume that the family of measures @xmath177 , @xmath147 , satisfies assumptions and .
let @xmath147 , @xmath312 and @xmath237 as in theorem [ thm : localclt ] , and @xmath313
. there exist @xmath314 with @xmath315 = 1 $ ] and a proper random variable @xmath316 , such that for all @xmath317 , * for all @xmath212 with @xmath318 , @xmath319 * for all @xmath320 , @xmath321 = \frac{\mathrm{g}_\sigma(y)}{m}$ ] .
[ rem : results ] * let us emphasize that our method does not allow to replace @xmath322 in by @xmath323 from . in particular , even if @xmath323 growth polynomially with @xmath110 , we are not able to improve the bound in to stretched exponential . in the case of independent bernoulli percolation , it is known from ( * ? ? ?
* section 2 ) that the result of theorem [ thm : vgb : main ] holds with a stretched exponential bound in .
* the fact that the right hand side of decays faster than any polynomial will be crucially used in the proofs of theorems [ thm : hk : grad ] , [ thm : hf : d+1 ] , and [ thm : gf : asymp ] .
quenched bounds on the diagonal @xmath324 under the assumptions and were obtained in @xcite ( see remarks 1.3 ( 4 ) and ( 5 ) there ) for all @xmath325 , although without any control on the tail of @xmath326 . * in the case of supercritical bernoulli percolation , barlow showed in ( * ? ? ?
* theorem 1 ) that the bound holds for all @xmath327 .
the step `` from @xmath260 to @xmath328 '' is highly nontrivial and follows from the fact that very good boxes on _ microscopic _ scales are dense , see ( * ? ? ?
* definition 5.4 and theorem 5.7(b ) ) .
we do not know if such property can be deduced from the assumptions and or proved for any of the specific models considered in section [ sec : examples ] ( except for bernoulli percolation ) .
our renormalization does not exclude the possibility of dense mesoscopic traps in @xmath134 , but we do not have a counterexample either . for comparison ,
let us mention that the heat kernel bounds and were obtained in @xcite for the random conductance model with i.i.d .
weights , where it is also stated in ( * ? ? ?
* remark 3.4 ) and ( * ? ? ?
* remark 4.12 ) that the lower bound for times comparable with @xmath329 can likely be obtained by adapting barlow s proof , but omitted there because of a considerable amount of extra work and few applications .
* the first proofs of the quenched invariance principle for random walk on the infinite cluster of bernoulli percolation @xcite relied significantly on the quenched upper bound on the heat kernel .
it was then observed in @xcite that it is sufficient to control only the diagonal of the heat kernel ( proved for bernoulli percolation in @xcite ) .
this observation was essential in proving the quenched invariance principle for percolation models satisfying and in @xcite , where the desired upper bound on the diagonal of the heat kernel was obtained by means of an isoperimetric inequality ( see ( * ? ? ? * theorem 1.2 ) ) .
theorem [ thm : hk : dzd ] allows now to prove the quenched invariance principle of @xcite by following the original path , for instance , by a direct adaptation of the proof of ( * ? ? ?
* theorem 1.1 ) . *
our proof of theorem [ thm : localclt ] follows the approach of @xcite in the setting of supercritical bernoulli percolation , namely , it is deduced from the quenched invariance principle , parabolic harnack inequality , and the upper bound on the heat kernel .
if we replace in @xmath330 by @xmath331 for any fixed @xmath332 , then it is not necessary to assume the upper bound on the heat kernel , see ( * ? ? ?
* theorem 1 ) . *
a new approach to limit theorems and harnack inequalities for the elliptic random conductance model under assumptions on moments of the weights and their reciprocals has been recently developed in @xcite .
it relies on moser s iteration and new weighted sobolev and poincar inequalities , and is applicable on general graphs satisfying globally conditions of regular volume growth and an isoperimetric inequality ( see ( * ? ? ?
* assumption 1.1 ) ) .
we will comment more on these conditions in remark [ rem : ads ] .
the method of @xcite was recently used in @xcite to prove the quenched invariance principle for the random conductance model on the infinite cluster of supercritical bernoulli percolation under the same assumptions on moments of the weights as in @xcite .
theorem [ thm : chd ] is enough to control the volume growth , thus we only discuss here the weak poincar inequality .
a finite subset @xmath333 of @xmath5 satisfies the ( strong ) poincar inequality @xmath334 , if for any function @xmath51 , @xmath335 .
the well known sufficient condition for @xmath334 is the following isoperimetric inequality for subsets of @xmath333 ( see , e.g. , ( * ? ? ?
* proposition 3.3.10 ) or ( * ? ? ? * lemma 3.3.7 ) ) : there exists @xmath336 such that for all @xmath337 with @xmath338 , the number of edges between @xmath339 and @xmath340 is at least @xmath341 .
thus , if the ball @xmath47 is contained in a subset @xmath342 of @xmath5 such that @xmath343 and the above isoperimetric inequality holds for subsets of @xmath342 , then it is easy to see that the weak poincar inequality with constants @xmath344 and @xmath345 holds for @xmath47 ( see claim [ cl : rb - gb ] ) . in the case
@xmath346 , the natural choice is to take @xmath342 to be the cluster of @xmath347 in @xmath348 , which turns out to be also the largest cluster in @xmath349 ( here and below , we implicitly assume that @xmath350 is large enough ) . in the setting of bernoulli percolation
, it is known that subsets of @xmath342 satisfy the above isoperimetric inequality ( see ( * ? ? ?
* proposition 2.11 ) ) . in our setting ,
theorem [ thm : chd ] implies that @xmath351 , thus we only need to prove the isoperimetric inequality . the first isoperimetric inequality for subsets of @xmath342 was proved in ( * ?
* theorem 1.2 ) .
it states that for any @xmath352 with @xmath353 , the number of edges between @xmath339 and @xmath354 is at least @xmath355 ( thus , also at least @xmath356 ) .
note the key difference , the edges are taken between @xmath339 and @xmath354 , not just between @xmath339 and @xmath357 .
the above isoperimetric inequality implies certain nash - type inequalities sufficient to prove a diffusive upper bound on the heat kernel ( see ( * ? ? ?
* theorem 2 ) , ( * ? ? ?
* proposition 6.1 ) , ( * ? ? ?
* lemma 3.2 ) , ( * ? ? ? * ( a.4 ) ) ) , but it is too weak to imply the poincar inequality ( see , e.g. , ( * ? ? ?
* sections 3.2 and 3.3 ) for an overview of the two isoperimetric inequalities and their relation to various functional inequalities ) .
let us also mention that in the setting of bernoulli percolation , the `` weak '' isoperimetric inequality admits a simple proof ( ( * ? ? ?
* theorem a.1 ) ) , but the proof of the `` strong '' one is significantly more involved ( ( * ? ? ? * proposition 2.11 ) ) .
after all said , we have to admit that we are not able to prove the strong isoperimetric inequality for subsets of @xmath342 , and do not know if it holds in our setting .
nevertheless , we can rescue the situation by proving that a certain _ enlarged _ set @xmath358 , obtained from @xmath342 by adding to it all vertices from @xmath134 to which it is _ locally _ connected , satisfies the desired strong isoperimetric inequality ( see proposition [ prop : verygoodbox ] , theorem [ thm : isop : cemax ] , and corollaries [ cor : isop : cemax ] and [ cor : tildec : existence ] ) . the general outline of the proof of our isoperimetric inequality for @xmath359 is similar to the one of the proof of the weak isoperimetric inequality for @xmath342 in @xcite , but we have to modify renormalization and coarse graining of subsets of @xmath359 and rework some arguments to get good control of the boundary and the volume of subsets of @xmath359 in terms of the boundary and the volume of the corresponding coarse grainings . for instance , it is crucial for us ( but not for @xcite ) that the coarse graining of a big set ( say , of size @xmath360 ) should not be too big ( see , e.g. , the proof of claim [ cl : isopmain2 ] ) .
we partition the lattice @xmath0 into large boxes of equal size . for each configuration @xmath361 , we subdivide all the boxes into _ good _ and _ bad_. restriction of @xmath120 to a good box contains a unique largest _ in volume _ cluster , and the largest clusters in two adjacent good boxes are connected in @xmath120 in the union of the two boxes . traditionally in the study of bernoulli percolation , the good boxes are defined to contain a unique cluster of large diameter . in our case ,
the existence of several clusters of large diameter in good boxes is not excluded .
the reason to work with volumes is that the existence of a unique giant cluster in a box can be expressed as an intersection of two events , an increasing ( existence of cluster with big volume ) and decreasing ( smallness of the total volume of large clusters ) .
assumption gives us control of correlations between monotone events , which is sufficient to set up two multi - scale renormalization schemes with scales @xmath362 ( one for increasing and one for decreasing events ) and conclude that bad boxes tend to organize in blobs on multiple scales , so that the majority of boxes of size @xmath362 contain at most @xmath305 blobs of diameter bigger than @xmath363 each , but even their diameters are much smaller than the actual scale @xmath362 . by removing two boxes of size @xmath364 containing the biggest blobs of an @xmath362-box , then by removing from each of the remaining @xmath363-boxes two boxes of size @xmath365 containing its biggest blobs , and so on , we end up with a subset of good boxes , which is a dense in @xmath0 , locally well connected , and well structured coarse graining of @xmath134 .
similar renormalization has been used in @xcite . by reworking some arguments from @xcite
, we prove that large subsets of the restriction of the coarse graining to any large box satisfy a @xmath95-dimensional isoperimetric inequality , if the scales @xmath362 grow sufficiently fast ( theorem [ thm : isop : pl ] ) .
we deduce from it the desired isoperimetric inequality for large subsets @xmath339 of @xmath359 ( theorem [ thm : isop : cemax ] ) as follows . if @xmath339 is spread out in @xmath359 , then it has large boundary , otherwise , we associate with it a set of those good boxes from the coarse graining , the unique largest cluster of which is entirely contained in @xmath339 .
it turns out that the boundary and the volume of the resulting set are comparable with those of @xmath339 .
moreover , if @xmath366 , then the volume of its coarse graining is also only a fraction of the total volume of the coarse graining of @xmath359 .
the isoperimetric inequality then follows from the one for subsets of the coarse graining . in section [ sec :
perforatedlattices ] we define perforated sublattices of @xmath0 and state an isoperimetric inequality for subsets of perforations . the main definition there is , and the main result is theorem [ thm : isop : pl ] .
the proof of theorem [ thm : isop : pl ] is given in section [ sec : pl : isop : proof ] . in section [ sec : propertiesofclusters ] we define a coarse graining of @xmath134 and study certain extensions of largest clusters of @xmath134 in boxes ( definition [ def : cemax ] ) .
particularly , we prove that they satisfy the desired isoperimetric inequality ( theorem [ thm : isop : cemax ] ) and the volume growth ( corollary [ cor : chemdist ] ) . in section [ sec :
proof : main ] we introduce the notions of _ regular _ and _ very regular _ balls , so that a ( very ) regular ball is always ( very ) good , and use it to prove the main result of the paper . in fact , in proposition [ prop : verygoodbox ] we prove that large balls are very likely to be very regular , which is stronger than theorem [ thm : vgb : main ] . in section [ sec : proofs ] , we sketch the proofs of theorems [ thm : hk : grad ] [ thm : gf : asymp ] .
finally , let us make a convention about constants .
as already said , we omit from the notation dependence of constants on @xmath138 , @xmath139 , and @xmath95 .
we usually also omit the dependence on @xmath367 , @xmath368 , and @xmath219 .
dependence on other parameters is reflected in the notation , for example , as @xmath369 .
sometimes we use @xmath344 , @xmath345 , @xmath370 , etc . , to denote `` intermediate '' constants , their values may change from line to line , and even within a line .
in this section we define lattices perforated on multiple scales and study their isoperimetric properties .
informally , for a sequence of scales @xmath371 , we define a perforation of the box @xmath372 by removing small rectangular regions of @xmath363-boxes from it , then removing small rectangular regions of @xmath373-boxes from each of the remaining @xmath363-boxes , and so on down to scale @xmath374 .
the precise definition is given in .
such perforated lattices will be used in section [ sec : propertiesofclusters ] as coarse approximations of largest connected components of @xmath120 in boxes .
the main result of this section is an isoperimetric inequality for subsets of perforations , see theorem [ thm : isop : pl ] .
the rules for perforation ( the shape and location of removed regions ) are determined by certain cascading events , which we define first , see and definition [ def : dinbad ] .
the recursive construction of the perforated lattice is given in section [ sec : perforatedlattice : construction ] , where the main definition is .
let @xmath375 , @xmath376 be sequences of positive integers such that @xmath377 and @xmath371 , for @xmath378 . to each @xmath362
we associate the rescaled lattice @xmath379 with edges between any pair of ( @xmath121-)nearest neighbor vertices of @xmath380 .
let @xmath381 be a family of events from some sigma - algebra .
for each @xmath382 , @xmath383 , define recursively the events @xmath384 by @xmath385 and @xmath386 the events in also depend on the scales @xmath387 and @xmath388 , but we omit this dependence from the notation , since these sequences will be properly chosen and fixed later .
[ def : dinbad ] given sequences @xmath389 , @xmath376 , as above , and two families of events @xmath390 and @xmath391 , we say that for @xmath376 , @xmath383 is @xmath392-_bad _
@xmath392-_good _ ) , if the event @xmath393 occurs ( resp . , does not occur ) .
good vertices give rise to certain geometrical structures on @xmath0 ( perforated lattices ) , which we define in the next subsection . the choice of the families @xmath390 and @xmath391 throughout the paper is either irrelevant for the result ( as in sections [ sec : perforatedlattices ] and [ sec : pl : isop : proof ] ) or fixed ( as in section [ sec : localevents ] ) .
thus , from now on we write @xmath394-bad ( resp . ,
@xmath394-good ) instead of @xmath392-_bad _ ( resp . , @xmath392-_good _ ) ,
hopefully without causing any confusions .
definition [ def : dinbad ] can be naturally generalized to @xmath92 families of events @xmath395 , for any fixed @xmath92 , and all the results of sections [ sec : perforatedlattices ] and [ sec : pl : isop : proof ] still hold ( with suitable changes of constants ) . for our applications
, it suffices to consider only two families of events ( see section [ sec : localevents ] ) .
thus , for simplicity of notation , we restrict to this special case . throughout this subsection , we fix sequences @xmath389 , @xmath376 , such that @xmath396 and @xmath387 is divisible by @xmath388 for all @xmath394 .
we also fix two local families of events @xmath390 and @xmath391 , and integers @xmath397 and @xmath398 .
recall definition [ def : dinbad ] of @xmath394-good vertices in @xmath380 . for @xmath210 , define @xmath399 and write @xmath400 for @xmath401 .
we also fix @xmath402 and assume that @xmath403 our aim is to construct a subset of @xmath404-good vertices in the lattice box @xmath405 by recursively perforating it on scales @xmath406 .
we use definition [ def : dinbad ] to determine the rules of perforation on each scale .
we first recursively define certain subsets of @xmath407-good vertices in @xmath408 for @xmath409 , see and .
let @xmath410 by , all @xmath411 are @xmath412-good .
assume that @xmath413 is defined for some @xmath409 so that all @xmath414 are @xmath407-good .
by definition [ def : dinbad ] , for each @xmath414 , there exist @xmath415 such that the boxes @xmath416 and @xmath417 are contained in @xmath418 , and all the vertices in @xmath419 are @xmath420-good .
if the choice is not unique , we choose the pair arbitrarily .
all the results below hold for any allowed choice of @xmath421 and @xmath422 . to save notation , we will not mention it in the statements . define @xmath423 to be * the empty set , if all the vertices in @xmath424 are @xmath420-good , or * @xmath425 if @xmath426 , or * a box @xmath427 in @xmath428 , with @xmath429 , which contains @xmath425 . possible outcomes ( b ) and ( c ) of @xmath430 are illustrated on figure [ fig : rzi ] .
[ rem : proprzi ] by construction , the set @xmath430 is the disjoint union of @xmath404 , @xmath305 , or @xmath431 boxes @xmath432 with @xmath433 . to complete the construction ,
let @xmath434 note that all @xmath435 are @xmath420-good .
now that the sets @xmath436 , are constructed by and , we define the multiscale perforations of @xmath405 by @xmath437 see figure [ fig : qks0 ] for an illustration . by construction , * for all @xmath438 , @xmath439 , * all the vertices of @xmath440 are @xmath404-good .
we will view the sets @xmath441 as subgraphs of @xmath442 with edges drawn between any two vertices of the set which are at @xmath121 distance @xmath374 from each other .
the next lemma summarizes some basic properties of @xmath440 s , which are immediate from the construction .
[ l : rect ] let @xmath1 , @xmath398 , and @xmath397 . for any choice of scales
@xmath389 , @xmath376 , such that @xmath396 and @xmath387 is divisible by @xmath388 for all @xmath394 , and for any admissible choice of @xmath443 or @xmath444 in the construction of @xmath440 , * @xmath440 is connected in @xmath442 , * @xmath445 . for a graph @xmath4 and a subset @xmath339 of @xmath4 ,
the boundary of @xmath339 in @xmath4 is the subset of edges of @xmath4 , @xmath6 , defined as @xmath446 the next theorem states that under assumption and some assumptions on @xmath387 and @xmath388 ( basically that @xmath447 is sufficiently small ) , there exist @xmath448 such that for all large enough @xmath449 with @xmath450 , @xmath451 .
[ thm : isop : pl ]
let @xmath1 .
let @xmath387 and @xmath388 , @xmath376 , be integer sequences such that for all @xmath394 , @xmath396 , @xmath387 is divisible by @xmath388 , and @xmath452 then for any integers @xmath397 , @xmath453 , and @xmath398 , @xmath402 , and two families of events @xmath390 and @xmath391 , if all the vertices in @xmath454 are @xmath412-good , then any @xmath455 with @xmath456 satisfies @xmath457 [ rem : isop : pl ] in the setting of theorem [ thm : isop : pl ] , if @xmath455 satisfies @xmath458 for some @xmath459 , then @xmath460 this easily follows from theorem [ thm : isop : pl ] by passing , if necessary , to the complement of @xmath461 in @xmath440 , see , for instance , remark [ rem : isop ] .
we postpone the proof of theorem [ thm : isop : pl ] to section [ sec : pl : isop : proof ] .
in fact , in two dimensions , we are able to prove the analogue of theorem [ thm : isop : pl ] for _ all _ subsets @xmath455 with @xmath462 , see lemma [ l : isopineqg:2d ] .
we believe that also in any dimension @xmath3 , the isoperimetric inequality of theorem [ thm : isop : pl ] holds for all subsets @xmath455 with @xmath462 , but can not prove it .
theorem [ thm : isop : pl ] follows immediately from a more general isoperimetric inequality in theorem [ thm : isopframe ] .
in this section we study properties of the largest subset of @xmath463 ( where @xmath400 is defined in ) .
we first define two families of events such that the corresponding perforated lattices defined in serve as a `` skeleton '' of the largest subset of @xmath463 . then , we provide sufficient conditions for the uniqueness of the largest subset of @xmath463 ( lemma [ l : cmax : uniqueness ] ) . to avoid problems , which may be caused by roughness of the boundary of the largest subset of @xmath463 , we
_ enlarge _ it by adding to it all the points of @xmath120 which are _ locally _ connected to it ( definition [ def : cemax ] ) . for the enlarged set we prove under some general conditions ( definition [ def : mainevent ] ) that its subsets satisfy an isoperimetric inequality ( theorem [ thm : isop : cemax ] and corollary [ cor : isop : cemax ] ) . under the same condition
we prove that the graph distance is controlled by that on @xmath0 ( lemma [ l : chemdist ] ) , large enough balls have regular volume growth ( corollary [ cor : chemdist ] ) and have local extensions satisfying an isoperimetric inequality ( corollary [ cor : tildec : existence ] ) . recall definition [ def : setr ] of @xmath132 .
consider an ordered pair of real numbers @xmath464 two families of events @xmath465 and @xmath466 are defined as follows . *
the complement of @xmath467 is the event that for each @xmath468 with @xmath469 , the set @xmath470 contains a connected component @xmath471 with at least @xmath472 vertices such that for all @xmath468 with @xmath469 , @xmath471 and @xmath473 are connected in @xmath474 . *
the event @xmath475 occurs if @xmath476 .
note that @xmath467 are decreasing and @xmath475 increasing events . from now on we fix these two local families , and say that @xmath383 is @xmath394-bad / @xmath394-good , if it is @xmath394-bad / @xmath394-good for the two local families @xmath477 and @xmath478 in the sense of definition [ def : dinbad ] . in particular , @xmath479 is @xmath404-good if both @xmath467 and @xmath475 do not occur , see figure [ fig:0good ] .
the following lemma is immediate from the definition of @xmath404-good vertex and the conditions on @xmath480 .
( see , e.g. , ( * ? ? ?
* lemma 5.2 ) for a similar result . )
[ l : fromg0tozd ] let @xmath453 and @xmath480 as in . * for any @xmath404-good vertex @xmath479 , connected component @xmath473 in @xmath481 with at least @xmath472 vertices is defined uniquely .
* for any @xmath404-good @xmath482 with @xmath483 , ( uniquely chosen ) @xmath473 and @xmath471 are connected in the graph @xmath484 .
[ def : cmax ] let @xmath485 be an increasing sequence of scales . for @xmath486 and @xmath22 , let @xmath487 be the largest connected component in @xmath488 ( with ties broken arbitrarily ) , and write @xmath489 .
fix @xmath480 as in and two families of events @xmath477 and @xmath478 as in section [ sec : localevents ] .
[ l : cmax : uniqueness ] let @xmath387 and @xmath388 be integer sequences such that for all @xmath394 , @xmath387 is divisible by @xmath388 , @xmath490 , and @xmath491 let @xmath453 , @xmath398 , and @xmath397 integers , @xmath402 . if all the vertices in @xmath454 are @xmath412-good , then @xmath492 is uniquely defined and @xmath493 . since all vertices in @xmath494 are @xmath412-good , we can define the perforation @xmath495 by . by definition , all the vertices of @xmath495 are @xmath404-good , and by lemma [ l : rect ] ,
@xmath495 is connected in @xmath442 . by lemma [ l : fromg0tozd ] , for any @xmath496
, there is a uniquely defined connected subset @xmath473 of @xmath497 with at least @xmath472 vertices . since @xmath495 is connected in @xmath442 , by lemma [ l : fromg0tozd ] , the set @xmath498 is contained in a connected component of @xmath499 and @xmath500 where the second inequality follows from lemma [ l : rect ] . on the other hand , since for any @xmath404-good vertex @xmath60 , the set @xmath501 contains at most @xmath502 vertices from @xmath503 , @xmath504 where the second inequality follows from the inequality @xmath505 and lemma [ l : rect ] , and the third inequality follows from the assumption .
we have shown that the connected component of @xmath499 which contains @xmath498 has volume @xmath506 . in particular
, it is the unique largest in volume connected component of @xmath499 .
moreover , by , its volume is @xmath507 which proves .
[ cor : cmax : uniqueness ] from the proof of lemma [ l : cmax : uniqueness ] , if the conditions of lemma [ l : cmax : uniqueness ] are satisfied , then @xmath508 in particular , for any @xmath509 and @xmath510 such that @xmath511 , @xmath512 . in this section
we prove an isoperimetric inequality for subsets of a certain extension @xmath513 of @xmath514 obtained by adding to @xmath514 all the vertices to which it is locally connected .
[ def : cemax ] let @xmath515 be the set of vertices @xmath516 such that for some @xmath517 , @xmath518 is connected to @xmath347 in @xmath519 , and define @xmath520 mind that @xmath521 is contained in @xmath522 , but it is different from the largest cluster of @xmath523 .
we study isoperimetric properties of @xmath513 for configurations from the following event . [
def : mainevent ] let @xmath480 be as in , @xmath398 and @xmath397 integers , @xmath402 .
the event @xmath524 occurs if * all the vertices in @xmath525 are @xmath412-good , * any @xmath526 with @xmath527 are connected in @xmath528 .
we write @xmath529 for @xmath530 .
here is the main result of this section .
[ thm : isop : cemax ] let @xmath480 be as in .
assume that the sequences @xmath387 and @xmath388 satisfy the conditions of theorem [ thm : isop : pl ] and @xmath531 let @xmath453 , @xmath398 , and @xmath397 integers , @xmath402 .
if @xmath532 occurs , then @xmath492 is uniquely defined and there exists @xmath533 such that for any @xmath534 with @xmath535 , @xmath536 in the applications , we will not use directly the result of theorem [ thm : isop : cemax ] , but only the following corollary , which estimates from below the size of the boundary of _ any _ subset of @xmath537 with volume @xmath538 .
[ cor : isop : cemax ] let @xmath480 be as in and @xmath539 $ ] .
assume that the sequences @xmath387 and @xmath388 satisfy the conditions of theorem [ thm : isop : cemax ] .
assume that @xmath540 if @xmath532 occurs , then for any @xmath534 with @xmath541 , @xmath542 in particular , if @xmath543 , then @xmath544 .
if @xmath545 , then we apply theorem [ thm : isop : cemax ] , @xmath546 if @xmath547 , then we use the trivial bound @xmath548 . by the assumption on @xmath104 , @xmath549 which implies , using the assumption on @xmath550 , that @xmath551 . thus , in this case , @xmath552 the proof of corollary is complete .
the proof of theorem [ thm : isop : cemax ] is subdivided into several claims . in claim [ cl :
cemax : localconnected ] we prove that @xmath553 is locally connected and in claims [ cl : isopmain ] and [ cl : isopmain2 ] we reduce the isoperimetric problem for subsets of @xmath553 to the one for subsets of a perforated lattice .
[ cl : cemax : localconnected ] any @xmath554 with @xmath527 are connected in @xmath555 .
fix @xmath554 with @xmath527 , and take @xmath556 such that @xmath60 and @xmath294 are connected in @xmath557 , @xmath347 and @xmath518 are connected in @xmath558 . by the triangle inequality , @xmath559 .
since all the vertices in @xmath494 are @xmath412-good , there exist @xmath560 such that @xmath561 and @xmath562 . by the definitions of @xmath529 and @xmath553
, @xmath563 is connected to @xmath294 in @xmath564 and @xmath565 is connected to @xmath518 in @xmath558 . by the triangle inequality , @xmath566 .
let @xmath567 be a box in @xmath400 which contains both @xmath563 and @xmath565 , where @xmath568 .
since all the vertices in @xmath569 are @xmath412-good , the perforation @xmath570 of @xmath567 is connected in @xmath442 by lemma [ l : rect ] .
thus , by lemma [ l : fromg0tozd ] , the set @xmath571 is contained in a connected component of @xmath572 .
in particular , @xmath563 and @xmath565 are connected in @xmath572 . by and
the fact that implies , the set @xmath571 is contained in @xmath573 .
therefore , @xmath563 is connected to @xmath565 in @xmath574 .
we conclude that @xmath60 is connected to @xmath347 in @xmath555 .
let @xmath575 since all the vertices in @xmath576 are @xmath412-good , we can define its perforation @xmath577 as in . by definition
, @xmath577 is a subset of @xmath404-good vertices in @xmath578 , and by lemma [ l : rect ] , @xmath577 is connected in @xmath442 . by the fact that implies , lemma [ l : fromg0tozd ] , , and the definition of @xmath553 ,
@xmath579 the next two claims allow to reduce the isoperimetric problem for subsets of @xmath553 to the isoperimetric problem for subsets of @xmath577 .
the crucual step for the proof is the following definition of @xmath580 and @xmath581 . for @xmath582 ,
let @xmath580 be the set of all @xmath583 such that @xmath584 , and @xmath581 the set of @xmath585 such that there exists @xmath586 with @xmath527 .
[ cl : isopmain ] @xmath587 and @xmath588 and @xmath589 such that @xmath483 , @xmath584 and @xmath590 . by lemma [
l : fromg0tozd ] and , @xmath473 and @xmath471 are connected in @xmath591 .
each path in @xmath553 connecting @xmath473 and @xmath592 contains an edge from @xmath593 .
this implies that latexmath:[\[\label{eq : isopmain : boundary:1 } next , by the definition of @xmath581 , for any @xmath595 , there exists @xmath586 such that @xmath527 . by claim [ cl : cemax : localconnected ] , @xmath60 and @xmath347 are connected in @xmath555 .
in particular , the ball @xmath596 contains an edge from @xmath593 . since every edge from @xmath593 is within @xmath122 distance @xmath597 from at most @xmath598 vertices of @xmath581 , @xmath599 . by the definition of @xmath553
, there exists @xmath600 such that @xmath601 by the definition of @xmath581 and , for any @xmath602 , @xmath603 .
thus , @xmath604 . by lemma [ l : rect ] and , @xmath605 thus , @xmath606 and we conclude that @xmath607 which implies .
let @xmath608 be the isoperimetric constant from theorem [ thm : isop : pl ] : @xmath609 [ cl : isopmain2 ] let @xmath610 . then @xmath611 if @xmath612 , then trivially holds .
thus , we assume that @xmath613 .
we will deduce from theorem [ thm : isop : pl ] . by ,
@xmath614 since @xmath545 , we obtain that @xmath615 . since @xmath616 , for all @xmath617 , @xmath618 .
thus , @xmath619 .
since also all the vertices in @xmath576 are @xmath412-good , we obtain as in that @xmath620 where the last inequality follows from lemma [ l : rect ] .
thus , @xmath621 . by theorem [ thm : isop : pl ] and remark [ rem : isop : pl ] , @xmath622 completing the proof of .
we are now ready to prove theorem [ thm : isop : cemax ] .
it easily follows from claims [ cl : isopmain ] and [ cl : isopmain2 ] . by , , and , @xmath623 on the one hand , if @xmath624 , then @xmath625 on the other hand , if @xmath626 , then by , @xmath627 , and @xmath628 the proof of theorem [ thm : isop : cemax ] is complete with @xmath629 .
with a more careful analysis and assuming that theorem [ thm : isop : pl ] holds for all subsets of size at least @xmath630 ( see remark [ rem : isopframe : conditions ] ) , condition on @xmath339 in theorem [ thm : isop : cemax ] can be relaxed to @xmath631 . assuming that theorem [ thm : isop : pl ] holds for all subsets ( see remark [ rem : isopframe : conditions ] ) , condition on @xmath339 in theorem [ thm : isop : cemax ] can be relaxed to @xmath632 . since for our purposes the current statement of theorem
[ thm : isop : cemax ] suffices , we do not prove the stronger statement here . in this section
we study the graph distances @xmath633 in @xmath120 between vertices of @xmath492 for configurations in @xmath532 . as consequences
, we prove that large enough balls centered at vertices of @xmath492 have regular volume growth ( corollary [ cor : chemdist ] ) and allow for local extensions which satisfy an isoperimetric inequality ( corollary [ cor : tildec : existence ] ) .
these results will be used in section [ sec : proof : main ] to prove our main result .
[ l : chemdist ] let @xmath1 and @xmath480 as in .
let @xmath387 and @xmath388 , @xmath376 , be integer sequences such that for all @xmath394 , @xmath634 and @xmath635 .
let @xmath453 , @xmath398 , and @xmath397 integers , @xmath402 .
there exists @xmath636 such that if @xmath532 occurs , then for all @xmath637 , @xmath638 let @xmath639 be such that @xmath640 and @xmath641 . by ( * ? ? ?
* lemma 5.3 ) ( applied to sequences @xmath387 and @xmath642 ) , there exist @xmath643 and @xmath644 which are connected by a nearest neighbor path of @xmath404-good vertices @xmath645 in @xmath440 , where @xmath646 .
let @xmath647 be an arbitrary vertex in @xmath648 .
( recall the definition of @xmath649 from lemma [ l : fromg0tozd ] . ) by lemma [ l : fromg0tozd ] , for all @xmath650 , @xmath647 is connected to @xmath651 in @xmath652 .
therefore , any vertices @xmath653 and @xmath654 are connected by a nearest neighbor path in @xmath655 .
any such path consists of at most @xmath656 vertices . by corollary [ cor : cmax : uniqueness ] ,
@xmath657 and @xmath658 .
thus , by the definition of @xmath532 , @xmath347 is connected to @xmath659 in @xmath519 and @xmath518 is connected to @xmath660 in @xmath661 . by putting all the arguments together
, we obtain that @xmath347 is connected to @xmath518 by a nearest neighbor path in @xmath120 of at most @xmath662 vertices .
since @xmath663 , the result follows .
[ cor : chemdist ] in the setup of lemmas [ l : cmax : uniqueness ] and [ l : chemdist ] , there exists @xmath664 such that for any @xmath665 and @xmath666 , @xmath667 let @xmath668 .
there exists @xmath669 such that @xmath670 .
since @xmath532 occurs , we can define the perforation @xmath440 of @xmath671 as in .
consider also the perforation @xmath672 of @xmath673 . by , @xmath674 since also @xmath675 , lemma [ l : chemdist ] implies that @xmath676 by applying lemma [ l : rect ] to @xmath677 and using the fact that @xmath678 , we conclude from the above inclusion that @xmath679 since @xmath680 , the result follows with @xmath681 . [ cor : tildec : existence ] let @xmath539 $ ] . in the setup of theorem [ thm : isop : cemax ] and lemma [ l : chemdist ] , if @xmath682 occurs with @xmath683 , then for all @xmath684 and @xmath666 , there exists @xmath685 such that @xmath686 and for all @xmath687 with @xmath688 , @xmath689 in particular , if @xmath543 , then @xmath690 .
let @xmath691 .
( note that @xmath692 . ) for @xmath666 , let @xmath693 be such that @xmath694 .
since @xmath695 occurs , by corollary [ cor : cmax : uniqueness ] , @xmath696 . by lemma [ l : chemdist ] , for @xmath697 , @xmath698 by corollary [ cor : isop : cemax ] , since @xmath699 , for any @xmath700 with @xmath701 , @xmath702 the proof is complete by taking @xmath703 .
in this section we collect together the deterministic results that large enough balls have regular volume growth ( corollary [ cor : chemdist ] ) and allow for local extensions satisfying an isoperimetric inequality ( corollary [ cor : tildec : existence ] ) to deduce theorem [ thm : vgb : main ] .
in fact , the result that we prove here is stronger . in definition [ def : vrb ]
we introduce the notions of regular and very regular balls , so that ( very ) regular ball is always ( very ) good ( see claim [ cl : rb - gb ] ) , and then prove in proposition [ prop : verygoodbox ] that large balls are likely to be very regular .
the main result is an immediate consequence of proposition [ prop : verygoodbox ] .
the following definition will only be used for the special choice of @xmath704 , see claim [ cl : rb - gb ] .
nevertheless , we choose to work with the more general definition involving arbitrary @xmath705 $ ] , since smaller @xmath105 s give better isoperimetric inequalities , and could be used to prove stronger functional inequalities than the poincar inequality , as we learned from jean - dominique deuschel ( see , e.g. , ( * ? ? ?
* section 3.2 ) ) .
[ def : vrb ] let @xmath37 , @xmath38 , and @xmath39 be fixed constants .
let @xmath539 $ ] . for @xmath22 integer and @xmath21
, we say that @xmath40 is @xmath706-_regular _ if @xmath707 and there exists a set @xmath708 such that @xmath709 and for any @xmath710 with @xmath711 , @xmath712 we say @xmath45 is @xmath706-_very regular _ if there exists @xmath46 such that @xmath47 is @xmath41-regular whenever @xmath48 , and @xmath49 . in the special case @xmath704 , we omit @xmath105 from the notation and call @xmath713-(very ) regular ball simply @xmath41-(very ) regular .
[ cl : rb - gb ] if @xmath40 is @xmath41-_regular _ , then it is @xmath41-_good_. by ( * ? ?
* proposition 3.3.10 ) and remark [ rem : wpi : minimum ] , @xmath714 thus , again by remark [ rem : wpi : minimum ] , @xmath715 theorem [ thm : vgb : main ] is immediate from claim [ cl : rb - gb ] and the following proposition , in which one needs to take @xmath543 .
[ prop : verygoodbox ] let @xmath1 , @xmath147 , and @xmath243 .
let @xmath539 $ ] .
assume that the family of measures @xmath177 , @xmath147 , satisfies assumptions and .
there exist constants @xmath37 , @xmath38 , and @xmath96 , @xmath716 and @xmath717 depending on @xmath187 , @xmath248 , and @xmath105 , such that for all @xmath78 , @xmath718 \geq 1 - c_{\scriptscriptstyle \ref{prop : verygoodbox}}\cdot e^{-c_{\scriptscriptstyle \ref{prop : verygoodbox}}(\log r)^{1+\consts}}.\ ] ] we first make a specific choice of various parameters .
fix @xmath147 .
we take @xmath719 where @xmath720 is defined in .
it is easy to see that @xmath721 and @xmath722 satisfy assumptions .
we fix this choice of @xmath723 throughout the proof .
next we choose the scales for renormalization . for positive integers @xmath724 , @xmath725 , and @xmath374 ,
we take @xmath726 where @xmath367 is defined in . by ( * ? ? ?
* lemmas 4.2 and 4.4 ) , under the assumptions and , there exist @xmath727 and @xmath728 such that for all @xmath729 , @xmath730 , and @xmath376 , @xmath731 \leq 2\cdot 2^{-2^n } .\ ] ] we choose @xmath729 so that the scales @xmath387 and @xmath388 defined in satisfy the conditions of lemma [ l : cmax : uniqueness ] , theorem [ thm : isop : cemax ] , and lemma [ l : chemdist ] , and choose @xmath730 .
thus , is also satisfied .
next we choose @xmath412 and @xmath104 .
fix @xmath78 . without loss of generality
, we can assume that @xmath732 let @xmath733 with this choice of @xmath412 , let @xmath734 , @xmath735 such that @xmath736 , and @xmath683 .
we begin with the proof .
if the event @xmath737 occurs , then @xmath738 . therefore , for all @xmath739 and @xmath740 , by corollaries [ cor : chemdist ] and [ cor : tildec : existence ] , the ball @xmath741 is @xmath742-regular .
thus , @xmath743 \text { is $ ( c_{\scriptscriptstyle \ref{cor : chemdist}},\frac{64^{\epsilon d}}{\gamma_{\scriptscriptstyle \ref{thm : isop : cemax}}^2},8c_{\scriptscriptstyle \ref{l : chemdist}},\epsilon)$-very regular with $ n_{\balls(0,r)}\leq r^\vgb$. } \end{array}\ ] ] let @xmath744 by , it suffices to prove that there exist constants @xmath745 and @xmath746 such that for all @xmath78 , @xmath747 \geq 1 - ce^{-c(\log r)^{1+\consts}}.\ ] ] by definition [ def : mainevent ] , , and , there exists @xmath748 such that @xmath749 \leq ( 5k+4)^d\cdot 2\cdot 2^{-2^s } + ( 5kl_s)^d\cdot c\cdot e^{-\funcs(u,2l_s ) } ~.\ ] ] thus , it remains to show that for our choice of all the parameters , the right hand side of the above display is at most @xmath750 .
let @xmath751 . by and the choice of @xmath412 , for all @xmath752 , @xmath753 which implies that @xmath754 by and , there exists a constant @xmath755 such that
for all @xmath752 , @xmath756 using , , and , we deduce that there exist @xmath757 and @xmath758 such that for all @xmath759 , @xmath760 by the choice of @xmath104 , @xmath761 .
therefore , there exist @xmath762 and @xmath763 such that for all @xmath764 , @xmath765\leq c''e^{-c''(\log r)^{1+\consts } } .\ ] ] since @xmath766 = \eta(u)>0 $ ] , implies .
the proof is complete .
the events @xmath467 and @xmath475 slightly differ from the corresponding events @xmath767 and @xmath768 in @xcite , but only minor modifications are needed to adapt ( * ? ? ?
* lemmas 4.2 and 4.4 ) to our setting .
there is room for flexibility in the choice of @xmath480 .
for instance , if @xmath769 is chosen so that @xmath770 , then @xmath771 and @xmath772 satisfy , and remains true for this choice of @xmath480 by monotonicity .
[ rem : ads ] as we already mentioned in remark [ rem : results](6 ) , a new approach to the random conductance model on general graphs satisfying some regularity assumptions has been recently developed in @xcite .
the main assumption on graphs there is ( * ? ? ?
* assumption 1.1 ) , which is reminiscent of definition [ def : vrb ] , but stronger .
the main difference is that we do not require that an isoperimetric inequality is satisfied by subsets of a ball , but by those of a local extension of the ball .
in fact , we do not know how to show ( and if it is true ) that subsets of balls satisfy the desired isoperimetric inequality of ( * ? ? ? * assumption 1.1 ) in our setting .
it would be very interesting to see if the machinery developed in @xcite can be applied to graphs with all large balls being very regular in the sense of definition [ def : vrb ] .
the rough outline of the proof is the following .
we first prove the isoperimetric inequality for all subsets of perforated lattices in two dimensions , see lemma [ l : isopineqg:2d ] . in dimensions
@xmath3 , we proceed in two steps .
we first consider only macroscopic subsets @xmath461 of the perforated lattice , i.e. , those with the volume comparable with the volume of the perforated lattice . by applying a selection lemma ,
see lemma [ l : selection ] , we identify a large number of disjoint two dimensional slices in the ambient box which on the one hand have a small non - empty intersection with @xmath461 , and on the other , all together contain a positive fraction of the volume of @xmath461 .
we estimate the boundary of @xmath461 in each of the slices using the two dimensional result , and conclude by estimating the boundary of @xmath461 in the perforation by the sum of the boundaries of @xmath461 in each of the slices .
finally , we treat the general case by constructing a suitable coarse graining of @xmath461 from mesoscopic boxes in which @xmath461 has positive density .
the restriction of the boundary of @xmath461 to such boxes is estimated by using the result from the first case .
both isoperimetric inequalities in @xmath3 are stated in theorem [ thm : isopframe ] .
we begin with a number of auxiliary ingredients for the proof : ( a ) some general facts about isoperimetric inequalities ( section [ sec : isop : generalfacts ] ) and ( b ) a combinatorial selection lemma ( section [ sec : selection ] ) . here
we collect some isoperimetric inequalities that we will frequently use .
[ l : boundaries ] let @xmath1 , @xmath773 integers with @xmath774 , and @xmath344 a positive real such that @xmath775 . then , for any subset @xmath339 of @xmath776 with @xmath777 , @xmath778 the proof is similar to that of ( * ? ? ? * proposition 2.2 ) .
let @xmath779 be the projection of @xmath0 onto the @xmath780 dimensional sublattice of vertices with @xmath407th coordinate equal to @xmath404 .
let @xmath781 , @xmath782 be a coordinate corresponding to @xmath783 with the maximal size , and @xmath784 .
let @xmath785 , i.e. , the projection of those @xmath782-columns that contain vertices from both @xmath339 and @xmath786 .
note that @xmath787 and @xmath788 . also note that @xmath789 . by the loomis - whitney inequality , @xmath790 .
thus , @xmath791 and @xmath792 .
[ rem : isop ] let @xmath4 be a finite graph , and assume that for all @xmath793 with @xmath794 , @xmath795 . then for any @xmath796 with @xmath797 , @xmath798 .
thus , any such @xmath799 also satisfies an isoperimetric inequality , but possibly with a smaller constant . the aim of this section is to prove the following combinatorial lemma .
its corollaries [ cor : selection1 ] and [ cor : selection2 ] together with the two dimensional isoperimetric inequality of lemma [ l : isopineqg:2d ] will be crucially used in the proof of the isoperimetric inequality for macroscopic subsets of perforated lattices in any dimension @xmath3 in theorem [ thm : isopframe ] .
[ l : selection ] let @xmath800 , and for @xmath1 , let @xmath801 let @xmath802 be positive integers . then , for any subset @xmath339 of @xmath803 satisfying @xmath804 there exist @xmath805 , disjoint two dimensional subrectangles of @xmath806 such that @xmath807 and for all @xmath808 , @xmath809 [ cor : selection1 ] note that @xmath810 .
thus , if we take @xmath811 , then @xmath812 , and lemma [ l : selection ] implies that for any @xmath813 with @xmath814 , there exist disjoint two dimensional rectangles @xmath815 such that @xmath816 and @xmath817 .
[ cor : selection2 ] if @xmath818 , and @xmath819 for some @xmath820 , then at least @xmath821 of the @xmath822 s contain at least @xmath823 vertices from @xmath339 . indeed ,
if such a choice did not exist , then we would have @xmath824 the proof is by induction on @xmath95 .
for @xmath825 the statement is obvious .
we assume that @xmath3 .
consider all two dimensional slices of the form @xmath826 , @xmath827 .
if among them there exist slices @xmath815 such that @xmath828 and for all @xmath407 , @xmath829 , then we are done .
thus , assume the contrary .
let @xmath830 be the subset of those slices that contain @xmath831 vertices from @xmath339 , and @xmath832 the rest . by definition , @xmath833 , and by assumption , @xmath834 .
consider @xmath780 dimensional rectangles @xmath835 and consider separately their intersections with @xmath830 and @xmath832 .
first , consider intersections with @xmath830 .
each of the rectangles from @xmath836 intersects @xmath837 in at most @xmath838 vertices . since @xmath839 ,
the number of rectangles @xmath840 with @xmath841 is at least @xmath842 . indeed , if not , then at least @xmath843 of rectangles from @xmath836 contain @xmath844 vertices from @xmath339 , and @xmath845 which is a contradiction .
next , consider intersections with @xmath832 . since @xmath846 ,
the number of rectangles @xmath840 with @xmath847 is at least @xmath848 . indeed , if not , then for at least @xmath843 of them , @xmath849 , and @xmath850 which is a contradiction . therefore , we can choose @xmath851 such that for each @xmath852 ,
latexmath:[\[|a\cap m_i|\geq \frac{|a|}{3\cdot r_1 } , \quad \quad in particular , for each @xmath852 , @xmath854 and @xmath855 if @xmath856 , then @xmath857 are disjoint two dimensional rectangles satisfying all the requirements of the lemma .
if @xmath858 , consider the sets @xmath859 , @xmath852 .
they satisfy assumption of the lemma with @xmath95 replaced by @xmath860 .
therefore , there exist disjoint two dimensional rectangles @xmath861 in @xmath857 such that for all @xmath862 , @xmath863 and @xmath864 it is now easy to conclude that the two dimensional rectangles @xmath865 satisfy all the requirements of the lemma .
indeed , they are disjoint , @xmath866 and for each @xmath407 and @xmath438 , @xmath867 the proof is complete . the main goal of this section is to prove the following lemma .
it immediately implies theorem [ thm : isop : pl ] in the case @xmath825 , but actually gives an isoperimetric inequality which holds for _ all _ @xmath868 with @xmath869 .
[ l : isopineqg:2d ] let @xmath825 .
let @xmath387 and @xmath388 , @xmath376 , be integer sequences such that for all @xmath394 , @xmath396 , @xmath387 is divisible by @xmath388 , and @xmath870 then for any integers @xmath397 , @xmath453 , and @xmath398 , @xmath402 , and two families of events @xmath390 and @xmath391 , if all the vertices in @xmath454 are @xmath412-good , then for any @xmath455 such that @xmath871 , @xmath872 * assumptions and the constant @xmath873 in the result of lemma [ l : isopineqg:2d ] are not optimal for our proof , but rather chosen to simplify calculations .
* we believe that an analogue of lemma [ l : isopineqg:2d ] holds for all @xmath1 , but can not prove it .
there is only one place in the proof where the assumption @xmath825 is used , see remark [ rem : extensiond>2 ] .
fix @xmath397 and @xmath398 integers , @xmath402 .
recall the definition of @xmath874 from , and write @xmath875 for @xmath874 throughout the proof .
note that @xmath876 and for all @xmath407 , @xmath877 .
let @xmath461 be a subset of @xmath878 such that @xmath879 .
we need to prove that @xmath880 .
first of all , without loss of generality we can assume that both @xmath461 and @xmath881 are connected in @xmath442 .
( for the proof of this claim , see page 112 in ( * ? ? ?
* section 3.1 ) . )
let @xmath882 be all the connected components ( in @xmath442 ) of @xmath883 , of which @xmath884 is the unique component intersecting @xmath878 , and @xmath885 s are the `` holes '' in @xmath886 completely surrounded by @xmath461 .
( see figure [ fig : mathcala ] . )
the boundary of @xmath461 in @xmath878 does not contain any edges adjacent to @xmath885 s .
it is convenient to absorb all the holes @xmath885 s into @xmath461 to get the set @xmath887 with the same boundary in @xmath878 , but with an important feature that its exterior vertex boundary in @xmath886 is @xmath888-connected .
more precisely , let @xmath889 then , ( a ) @xmath890 , ( b ) @xmath891 , ( c ) @xmath887 is connected in @xmath442 , ( d ) @xmath892 ( in particular , connected in @xmath442 ) , and ( e ) for any @xmath893 ( the exterior vertex boundary of @xmath887 in @xmath886 ) there exist @xmath894 such that @xmath895 for all @xmath92 ( i.e. , @xmath896 is @xmath888-connected ) .
properties ( a - d ) are immediate from the definition of @xmath887 , and property ( e ) follows from ( * ? ? ?
* lemma 2.1(ii ) ) and the facts that @xmath887 and @xmath897 are connected in @xmath442 . by properties ( a - b ) of @xmath887 , it suffices to prove that @xmath898 by lemma [ l : rect ] and the first part of , @xmath899 .
thus , by lemma [ l : boundaries ] , @xmath900 into the sets @xmath901 of edges with one end vertex in @xmath887 and the other in @xmath902 and comparing the cardinality of @xmath901 s with that of @xmath903 . if @xmath903 is very large ( macroscopic ) , then all @xmath901 are negligibly small in comparison to @xmath903 .
it is more delicate to estimate the size of @xmath901 s if @xmath903 is small , as the contribution of some @xmath901 s to the boundary @xmath903 may be quite significant . in this case , we will introduce a suitable scale on which @xmath903 is large , and view @xmath887 as a disjoint union of subsets of boxes on the new scale .
let @xmath904 then , @xmath905 s are disjoint and for any @xmath906 , @xmath907 the scale @xmath908 is the correct scale to study @xmath909 . as we will see below in , the intersection of @xmath909 with @xmath910 , @xmath911 ( holes of size significantly smaller than @xmath908 ) , is negligible in comparison to @xmath909 .
in particular , it will be enough to conclude in the case @xmath912 , see .
if @xmath913 , then @xmath909 is small and may have a significant intersection with @xmath914 .
an additional argument will be used to deal with this case , see below .
we begin with an estimation of the part of @xmath909 adjacent to `` small holes '' .
[ cl : relationdelta ] for all @xmath915 , latexmath:[\[\label{eq : relationdelta } by the definition of @xmath875 s , the set @xmath902 can be expressed as the disjoint union of boxes @xmath917 , for some @xmath918 , such that every box @xmath919 is within @xmath122 distance @xmath920 from at most @xmath921 @xmath922 s .
( by remark [ rem : proprzi ] , each @xmath920-box contains at most @xmath201 @xmath919 s , and it is adjacent to at most @xmath923 other @xmath920-boxes , hence @xmath924 . ) to estimate the size of @xmath901 , we consider two cases : ( a ) @xmath909 is adjacent to few @xmath919 s , in which case @xmath901 is very small , ( b ) @xmath909 is adjacent to many @xmath919 s , in which case many of the @xmath919 s will be well - separated and @xmath887 will be spread out .
to handle this case we will use the fact that the exterior vertex boundary of @xmath887 is @xmath888-connected , thus the majority of edges in @xmath909 will be `` in between '' @xmath919 s .
( see figure [ fig : deltai ] . )
let @xmath925 be the total number of those @xmath919 s which are adjacent ( in @xmath442 ) to @xmath887 .
since for each @xmath438 , @xmath926 , it follows that @xmath927 .
we consider separately the cases @xmath928 and @xmath929 . if @xmath928 , then @xmath930 where the last inequality follows from the definition of @xmath114 and the fact that @xmath911 .
if @xmath931 , then @xmath887 is adjacent to at least @xmath932 of @xmath919 s which are pairwise at @xmath122 distance at least @xmath920 from each other .
recall from property ( e ) of @xmath887 that @xmath896 is the exterior vertex boundary of @xmath887 , which is @xmath888-connected .
since @xmath896 intersects each of the @xmath933 well separated @xmath919 s , the intersections of @xmath896 with @xmath934-neighborhoods of the @xmath919 s are disjoint sets of vertices of cardinality @xmath935 each .
therefore , @xmath936 , and we obtain that @xmath937 where the last inequality follows from the fact that each vertex of @xmath896 is adjacent to at most @xmath201 edges from @xmath909 . combining and we get .
if the boundary @xmath909 is macroscopic , namely , if @xmath912 , then the intersection of @xmath909 with any hole is negligible , and claim [ cl : relationdelta ] immediately implies . indeed , by and , @xmath938 and follows from and the second part of
. in the rest of the proof we consider the case of small @xmath909 , namely @xmath913 . in this case , @xmath939 as already mentioned , this case is more delicate , since @xmath909 may have large intersection with big holes , for instance , @xmath940 is generally not negligible in comparison to @xmath909 .
we first consider the case when @xmath909 is still relatively big in comparison to the boundary of holes in @xmath941 .
assume that @xmath942 .
in this case we will show that @xmath943 , is sufficient for .
indeed , by , @xmath944 and follows from and the second part of . to estimate the size of @xmath940
, we proceed as in the proof of .
the set @xmath914 can be expressed as a disjoint union of boxes @xmath945 , for some @xmath946 , such that every box is within @xmath122 distance @xmath947 from at most @xmath921 of the boxes .
since @xmath948 and the exterior vertex boundary of @xmath887 is @xmath888-connected , the set @xmath887 can be adjacent ( in @xmath442 ) to at most @xmath921 such boxes ( in fact , to at most @xmath949 ) , which implies that @xmath950 . to estimate @xmath951 from below
, we view @xmath887 as a disjoint union of subsets @xmath952 of @xmath947-boxes , and estimate from below the relative boundary of each @xmath952 in the corresponding box . by definition
, @xmath953 is the disjoint union of boxes @xmath954 , @xmath955 . let @xmath952 be the restriction of @xmath887 to the box @xmath956 . by and , for every @xmath438 , @xmath957 by applying lemma [ l : boundaries ] in each of @xmath954 ,
@xmath958 the combination of and gives .
it remains to consider the case @xmath959 . in this case
@xmath909 is comparable to the boundary of holes in @xmath941 .
we will show that @xmath960 , is sufficient for .
indeed , by , @xmath961 and follows from and the second part of .
since @xmath909 is comparable to the boundary of holes in @xmath941 , this time we will look at @xmath887 on the scale @xmath962 . by lemma [ l : rect ] and the assumption that @xmath963 is divisible by @xmath964 , @xmath965 can be expressed as a disjoint union of boxes @xmath966 , @xmath967 . let @xmath952 be the restriction of @xmath887 to the box @xmath966 .
we will compare the boundary @xmath909 to the relative boundary of @xmath952 s in the respective boxes .
if for all @xmath438 , @xmath968 , then by lemma [ l : boundaries ] applied in each of @xmath969 , @xmath970 since the sets @xmath971 are disjoint subsets of @xmath972 , @xmath973 which implies . on the other hand ,
if @xmath974 for at least one @xmath438 , then there exists @xmath975 such that * @xmath976 and * @xmath977 .
indeed , if none of @xmath978 s satisfies the two requirements , then there exist @xmath979 and @xmath980 such that @xmath981 , @xmath982 and @xmath983 .
then , @xmath984 satisfies the two requirements for some @xmath985 .
( if @xmath964 is divisible by @xmath305 , then one can take @xmath986 . ) by applying lemma [ l : boundaries ] to @xmath987 , @xmath988 where the last inequality follows from the assumption on @xmath989 .
this inequality completes the proof of .
to summarize , the desired relation between @xmath990 and @xmath909 follows from the three inequalities ( the boundary @xmath909 is macroscopic ) , ( the boundary @xmath909 is small , but much bigger than the boundaries of holes on the given scale ) , and ( the boundary @xmath909 is small and comparable to the boundaries of holes on the given scale ) .
the proof of lemma [ l : isopineqg:2d ] is complete .
[ rem : extensiond>2 ] the only step in the proof of lemma [ l : isopineqg:2d ] that uses ( crucially ! ) the assumption @xmath825 is the derivation of .
more precisely , the fact that the boundary of a set is well approximated by simple paths . in higher dimensions
this is clearly not the case ( the dimension of the boundary is generally bigger than the dimension of a simple path ) , and the above argument breaks down .
see figure [ fig : deltai ] . in this section we prove the following theorem , which includes theorem [ thm : isop : pl ] as a special case .
[ thm : isopframe ] let @xmath1 , @xmath336 .
let @xmath387 and @xmath388 , @xmath376 , be integer sequences satisfying assumptions of lemma [ l : isopineqg:2d ] and such that @xmath991 then for any integers @xmath397 , @xmath453 , and @xmath398 , @xmath402 , and two families of events @xmath390 and @xmath391 , if all the vertices in @xmath454 are @xmath412-good , then any @xmath455 with @xmath992 satisfies @xmath993 fix @xmath397 and @xmath398 integers , @xmath402 , and assume that all the vertices in @xmath454 are @xmath412-good .
take @xmath994 such that @xmath995 .
we consider separately the cases @xmath996 and @xmath997 .
in fact , we will use the result for the first case to prove the result for the second . in the first case , we use corollaries [ cor : selection1 ] and [ cor : selection2 ] to the selection lemma from section [ sec : selection ] to identify a large number of disjoint two dimensional slices in the ambient box @xmath998 which on the one hand have a small non - empty intersection with @xmath461 , and on the other , all together contain a positive fraction of the volume of @xmath461 .
we estimate the boundary of @xmath461 in each of the slices using the two dimensional isoperimetric inequality of lemma [ l : isopineqg:2d ] .
since the slices are pairwise disjoint , we can estimate the boundary of @xmath461 by the sum of the boundaries of @xmath461 in each of the slices . in the second case
, we consider a coarse graining of @xmath461 by densely occupied @xmath999-boxes .
if the number of densely occupied @xmath999-boxes is small , then @xmath461 is scattered in @xmath998 and has big boundary . if , on the other hand , the number of densely occupied @xmath999-boxes is big , then the set of such boxes has large boundary ( the poorly occupied boxes adjacent to some densely occupied ones ) .
each pair of adjacent densely and poorly occupied @xmath999-boxes are contained in a @xmath1000-box .
vertices from @xmath461 occupy a non - trivial fraction of vertices in this @xmath1000-box .
thus , we can estimate the boundary of @xmath461 restricted to this box using the first part of the theorem . by summing over all pairs of adjacent densely and poorly occupied @xmath999-boxes
we obtain a desired lower bound on the size of the boundary of @xmath461 .
we first consider the case @xmath996 . by corollaries
[ cor : selection1 ] and [ cor : selection2 ] , there exist @xmath1001 two dimensional subrectangles @xmath822 in @xmath998 ( see figure [ fig : slice ] ) such that for all @xmath407 , @xmath1002 ( applied to the perforation @xmath1003 of @xmath822 ) and the first part of , @xmath1004 , which implies that for all @xmath407 , @xmath1005 we apply the two dimensional isoperimetric inequality of lemma [ l : isopineqg:2d ] and remark [ rem : isop ] to each of the sets @xmath1006 in @xmath1003 , and obtain that for all @xmath407 , @xmath1007 since all @xmath1008 are disjoint subsets of @xmath1009 , @xmath1010 this completes the proof of theorem [ thm : isopframe ] for sets with @xmath996 .
next , we consider the case @xmath997 .
let @xmath1011 be the set of bottom - left corners of @xmath999-boxes which contain a vertex from @xmath461 .
note that @xmath1012 .
we also define the subset @xmath1013 of @xmath1014 corresponding to the densely occupied boxes , @xmath1015 we consider separately the cases when @xmath1016 and @xmath1017 .
we first consider the case @xmath1016 , i.e. , the number of densely occupied boxes is large .
since @xmath1018 , @xmath1019 by applying lemma [ l : boundaries ] to @xmath1020 , we get @xmath1021 . take any pair @xmath1022 and @xmath1023 from @xmath1024 .
note that @xmath1025 in @xmath1026 containing both @xmath1027 and @xmath1028 , where @xmath1029 .
note that @xmath461 occupies a non - trivial fraction of vertices in @xmath1030 .
more precisely , @xmath1031 moreover , all the vertices in @xmath1032 are @xmath412-good .
we are in a position to apply the first part of the theorem to @xmath1033 in @xmath1034 . combining the upper bound on @xmath1035 with the lower bound on the volume of the perforation @xmath1036 given by lemma [ l : rect ] and the second part of the assumption
, we obtain that @xmath1037 therefore , by the first part of the theorem ( with @xmath1038 ) applied to the subset @xmath1033 of @xmath1039 and remark [ rem : isop ] , @xmath1040 this inequality gives us an estimate on the part of the boundary @xmath1009 contained in @xmath1030 for each @xmath1041 such that the cube @xmath1030 contains an overcrowded and undercrowded adjacent @xmath999-boxes @xmath1042 and @xmath1043 with @xmath1022 and @xmath1023 . by
, the total number of such @xmath1044 s is @xmath1045 where the factor @xmath1046 counts for possible overcounting , since every cube @xmath1034 , @xmath1047 , contains at most @xmath1048 pairs @xmath1049 with @xmath1050 .
moreover , every edge from @xmath1009 belongs to at most @xmath431 cubes @xmath1034 , @xmath1047 .
thus , @xmath1051 by putting all the estimates together , we obtain that @xmath1052 .
it remains to consider the case @xmath1017 . in this case , @xmath461 is scattered in @xmath1026 , and should have big boundary
. indeed , for each @xmath1053 , @xmath1054 by the lower bound on the volume of the perforation @xmath1055 given in lemma [ l : rect ] and the second part of , @xmath1056 thus , @xmath1027 contains vertices from both @xmath461 and @xmath1057 .
by lemma [ l : rect ] , @xmath1055 is connected in @xmath442 , thus it contains an edge from @xmath1058 . since all @xmath1027 , @xmath1053 are disjoint
, we conclude that @xmath1059 where the last inequality follows from the case assumption . the proof of theorem [ thm : isopframe ] in the case @xmath997 is complete by and .
[ rem : isopframe : conditions ] we believe that theorem [ thm : isopframe ] holds for all @xmath461 with @xmath1060 . with a more involved proof , we can relax the assumption @xmath997 of theorem [ thm : isopframe ] to @xmath1061 . since this does not give us the result for all @xmath461 , and the current statement of theorem [ thm : isopframe ] suffices for the applications in this paper
, we do not include this proof here .
in this section we give proof sketches of theorems [ thm : hk : grad ] , [ thm : gf : bounds ] , [ thm : hf : d+1 ] , [ thm : localclt ] , and [ thm : gf : asymp ] .
their proofs are straightforward adaptations of main results in @xcite from bernoulli percolation to our setup .
the proof is essentially the same as that of ( * ? ? ?
* theorem 6 ) .
the only minor care that is required comes from the fact that the bound is not stretched exponential .
since this fact is used several times , we provide a general outline of the proof . as in the proof of (
* theorem 6 ) , by stationarity and the ergodicity of @xmath134 with respect to the shift by @xmath1062 ( see , e.g. , ( * ? ? ?
* theorem 3.1 ) ) , it suffices to prove that @xmath1063\leq \frac{c}{n^{d+1}}\cdot e^{-c\frac{\dist_{\z^d}(0,x)^2}{n}},\ ] ] where @xmath344 and @xmath370 only depend on @xmath95 and @xmath187 . if @xmath1064 , where @xmath219 is defined in , then by the general upper bound on the heat kernel ( see , e.g. , ( * ? ? ?
* ( 1.5 ) ) ) , @xmath1063\leq c\cdot e^{-c\frac{\dist_{\z^d}(0,x)^2}{n } } \leq \frac{c'}{n^{d+1}}\cdot e^{-c'\frac{\dist_{\z^d}(0,x)^2}{n}}.\ ] ] thus , we can assume that @xmath1065 .
let @xmath1066 . by , @xmath1067\leq \mathbb p^u\left[n(\omega)\geq n\right]\\ \leq cn^d\cdot e^{-c \cdot ( \log n)^{1+\consts } } \leq \frac{c'}{n^{d+1}}\cdot e^{-c ' \cdot ( \log n)^{1+\consts } } \leq \frac{c'}{n^{d+1}}\cdot e^{-c'\frac{\dist_{\z^d}(0,x)^2}{n}}.\end{gathered}\ ] ] it remains to bound @xmath1068 $ ] .
as in ( * ? ? ?
* section 2 ) , define the quenched entropy of the simple random walk on @xmath134 by @xmath1069 , where @xmath1070 and @xmath1071 for @xmath292 , and the mean entropy by @xmath1072 $ ] . by a general argument in the proof of (
* theorem 6 ) , the heat kernel upper bound implies that @xmath1073 \leq \left(h_n - h_{n-1}\right)\cdot \frac{c}{n^d}\cdot e^{-c\frac{\dist_{\z^d}(0,x)^2}{n}}.\ ] ] the proof of ( * ? ? ?
* theorem 6 ) is completed by showing in ( * ? ? ?
* lemma 20 ) that @xmath1074 .
thus , in order to finish the proof of theorem [ thm : hk : grad ] , it suffices to prove that @xmath1074 in our setting too .
this is a simple consequence of theorem [ thm : hk : dzd ] . indeed , by writing @xmath1075 as the sums over @xmath60 with @xmath1076 and @xmath1077 , applying and to the summands in the first sum , and showing smallness of the second sum by using , for instance ,
the general upper bound on the heat kernel ( see , e.g. , ( * ? ? ?
* ( 1.5 ) ) ) , we prove that for all @xmath1078 , @xmath1079 . for @xmath1080
, we use the crude bound @xmath1081 ( see the proof below ( * ? ? ?
* ( 25 ) ) ) . by integrating @xmath1075 and using
, we get that @xmath1082 , which implies that @xmath1083 for some @xmath344 . since @xmath1084 is decreasing by ( * ?
* corollary 10 ) , we conclude that @xmath1085 , finishing the proof of theorem [ thm : hk : grad ] .
the proof of theorem [ thm : gf : bounds ] is literally the same as the proof of ( * ? ? ?
* theorem 1.2(a ) ) . for the upper bound ,
one splits the green function into the integrals over @xmath1086 $ ] and @xmath1087 . using general bounds on the heat kernel ( see ( * ? ? ?
* ( 6.4 ) and ( 6.5 ) ) ) , one shows that the first integral is @xmath1088 , and by , the second integral is bounded by @xmath1089 . for the lower bound , one estimates the green function from below by the integral of heat kernel over @xmath1090 , applies , and arrives at the desired bound .
the proof of theorem [ thm : hf : d+1 ] is identical to the one of ( * ? ? ?
* theorem 5 ) .
the constant functions and the projections of @xmath1091 ( see theorem [ thm : qip](a ) ) on coordinates of @xmath0 are independent harmonic functions with at most linear growth . thus , the dimension of such functions is at least @xmath1092 .
it remains to show that the above functions form a basis .
let @xmath93 be a harmonic function @xmath93 on @xmath134 with at most linear growth and @xmath1093 , and assume that it is extended on @xmath223 ( see above ( * ? ? ?
* proposition 19 ) ) . by theorem [ thm : vgb : main ] and the upper bound on the heat kernel , the proof of ( * ? ? ?
* proposition 19 ) goes through without any changes in our setting , implying that the sequence @xmath1094 is uniformly bounded and equicontinuous on compacts .
thus , there exists a sequence @xmath1095 such that @xmath1096 converges uniformly on compact sets to a continuous function @xmath1097 . by using the quenched invariance principle of theorem [ thm : qip ] ,
one obtains by repeating the proof of ( * ? ? ?
* theorem 5 ) that @xmath1097 is harmonic in @xmath223 .
since @xmath1097 has at most linear growth and @xmath1098 , it is linear .
therefore , the function @xmath1099 is harmonic on @xmath134 and for every @xmath313 and all large enough @xmath92 , @xmath1100 for all @xmath1101 . by , @xmath1102\leq
\varepsilon n_k^2 $ ] for all large @xmath92 .
the proof of ( * ? ? ? * theorem 5 ) is finished by applying ( * ? ? ?
* corollary 21 ) which states that @xmath1103 must be constant .
the proof of ( * ? ? ?
* corollary 21 ) is rather general and only uses the fact that the mean entropy @xmath1104 ( see the proof of theorem [ thm : hk : grad ] ) satisfies @xmath1105 .
we already proved this bound in the proof of theorem [ thm : hk : grad ] .
thus , ( * ? ? ?
* corollary 21 ) holds in our setting , and we conclude that @xmath1103 must be constant .
the proof is complete .
theorem [ thm : localclt ] was proved in the case of supercritical bernoulli percolation in ( * ? ? ?
* theorem 1.1 ) by first providing general assumptions ( * ? ? ?
* assumption 4.4 ) for the local limit theorem on infinite subgraphs of @xmath0 ( see ( * ? ? ?
* theorems 4.5 and 4.6 ) ) , and then verifying these assumptions for the infinite cluster of bernoulli percolation .
* assumption 4.4 ) is tailored for random subgraphs of @xmath0 with laws invariant under reflections with respect to coordinate axes and rotations by @xmath1106 .
these assumptions only simplify the expression for the heat kernel of the limiting brownian motion , and can be naturally extended to the case without such symmetries .
we only consider the case of discrete time random walk ( the continuous time case is the same ) .
as in ( * ? ? ?
* theorem 4.5 ) , to prove theorem [ thm : localclt ] it suffices to show that there exist an event @xmath1107 with @xmath1108=1 $ ] , positive constants @xmath1109 , @xmath1110 , and @xmath1111 , and a covariance matrix @xmath237 , such that for all @xmath1112 , * for any @xmath320 and @xmath24 , as @xmath236 , @xmath1113^d)\right]$ ] converges to @xmath1114^d}k_{\sigma , t}(y ' ) dy'$ ] uniformly over compact subsets of @xmath1115 ( @xmath1116 is as in ) , * there exists @xmath1117 such that for all @xmath1118 and @xmath1119 , @xmath1120 , * for each @xmath1121 , there exists @xmath1122 such that the parabolic harnack inequality holds with constant @xmath1111 in @xmath91\times\ballz_{\set_\infty}(y , r)$ ] for all @xmath1123 , * for @xmath1124^d\text { such that $ \set_\infty\cap(y+[-r',r']^d)=\emptyset$}\}$ ] , the ratio @xmath1125 tends to @xmath404 as @xmath1126 , * for any @xmath210 and @xmath24 , @xmath1127^d))}{(2\sqrt{n}r)^d } = \mathbb e^u[\mu_0\cdot\mathds{1}_{0\in\set_\infty}]$ ] , * for each @xmath210 and @xmath24 , there exists @xmath1128 such that for all @xmath1129 , and @xmath1130^d)$ ] , @xmath1131 , * for @xmath210 and @xmath1132 as in ( c ) , @xmath1133 .
* follows from theorem [ thm : qip ] , * follows from , * follows from theorems [ thm : phi : vgb ] and [ thm : vgb : main ] , * follows from stationarity , , and the borel - cantelli lemma , * follows from a spatial ergodic theorem ( * ? ? ? * theorem 2.8 in chapter 6 ) , since the sequence of boxes @xmath1134^d)_{n\geq 1}$ ] is regular in the sense of ( * ? ? ?
* definition 2.4 in chapter 6 ) ( see ( * ? ? ?
* lemma 5.1 ) ) , * follows from theorem [ thm : chd ] , * follows from , theorem [ thm : phi : vgb ] , and the borel - cantelli lemma .
statement ( a ) follows from theorem [ thm : localclt ] and by repeating the proof of ( * ? ? ?
* theorem 1.2(b ) ) without any changes .
for the statement ( c ) we use bounds ( * ? ? ?
* ( 6.30 ) and ( 6.31 ) ) and , to get @xmath1135 \leq \mathbb e^u\left[g_{\set_\infty}(0,x)~\big|~0\in\set_\infty\right]\\ \leq \frac{(1+\varepsilon)\mathrm{g}_\sigma(x)}{m}+\frac{c'\mathbb p^u\left[m > |x|~\big|~0\in\set_\infty\right]}{|x|^{d-2 } }
\\ + c'\left(\mathbb e^u\left[g_{\set_\infty}(0,x)^2~\big|~0\in\set_\infty\right]\right)^{\frac 12}\cdot e^{-c'(\log |x|)^{1+\consts}},\end{gathered}\ ] ] where @xmath1136 is defined in the statement of theorem [ thm : gf : asymp ] . as in (
* ( 6.17 ) ) , by , @xmath1137 combining this bound with , we obtain that @xmath1138<c''$ ] .
let @xmath1139 .
since @xmath1140 , by taking limits @xmath1141 and then @xmath1142 , we compete the proof of statement ( c ) .
the author thanks takashi kumagai and alain - sol sznitman for encouragements to write this paper and for valuable comments , jean - dominique deuschel for sending him the master thesis of tuan anh nguyen @xcite and for a discussion about isoperimetric and sobolev inequalities , and ji ern for a suggestion to use general densities @xmath721 and @xmath722 in section [ sec : propertiesofclusters ] .
special thanks go to the anonymous referee for helpful suggestions and comments .
a. drewitz , b. rth and a. sapozhnikov ( 2012 ) local percolative properties of the vacant set of random interlacements with small intensity .
_ to appear in the ann .
henri poincar_. arxiv:1206.6635 . | for a general class of percolation models with long - range correlations on @xmath0 , @xmath1 , introduced in @xcite , we establish regularity conditions of barlow @xcite that mesoscopic subballs of all large enough balls in the unique infinite percolation cluster have regular volume growth and satisfy a weak poincar inequality . as immediate corollaries ,
we deduce quenched heat kernel bounds , parabolic harnack inequality , and finiteness of the dimension of harmonic functions with at most polynomial growth .
heat kernel bounds and the quenched invariance principle of @xcite allow to extend various other known results about bernoulli percolation by mimicking their proofs , for instance , the local central limit theorem of @xcite or the result of @xcite that the dimension of at most linear harmonic functions on the infinite cluster is @xmath2 . in terms of specific models ,
all these results are new for random interlacements at every level in any dimension @xmath3 , as well as for the vacant set of random interlacements @xcite and the level sets of the gaussian free field @xcite in the regime of the so - called local uniqueness ( which is believed to coincide with the whole supercritical regime for these models ) . |
the experimental exploration of bose - einstein condensates ( becs ) in many different systems such as alkali metal gases @xcite , hydrogen @xcite , meta - stable helium @xcite , and molecules @xcite has led to a surge of interest in manipulating ultracold atomic samples under very different circumstances .
one of the initial motivations for such development was and is the prospect of creating a continuous and coherent atomic beam @xcite , which is the atomic analogy of the optical laser beam @xcite . among the major challenges in achieving a continuous atom laser
is how to overcome the difficulty due to the restrictive cooling conditions for continuously condensing the atomic gas .
spatial separation of the evaporative cooling from the destructive laser cooling may play a role in alleviating these challenges @xcite .
recently , a continuous bec source was achieved by periodically replenishing a reservoir with condensates @xcite . there , optical tweezers were used to transport sodium condensates from where they were created to the reservoir .
such procedure allows one to continuously replenish the reservoir which contains on average more than @xmath2 atoms . bringing a new condensate to the reservoir
each time implies the merging of condensates with different relative phases since each condensates is created with a different phase .
the merging of two condensates with different phases poses a limitation on the coherence of the process .
recently , interference effects in the merging of 30 uncorrelated bose - einstein condensates released from a one - dimensional optical lattice have been discussed in @xcite , whereas coherent splitting of becs by deforming an optical single - well into a double - well potential for a trapped atom interferometer have been addressed in refs .
very recently , schumm _ et al . _
@xcite has demonstrated a coherent beam splitter on an atom chip by splitting the condensate in a double - well potential and merging it again .
they have demonstrated phase preservation in this process , even in the case when the split becs are far enough apart to inhibit tunnel coupling . in this paper
, we analyze the axial splitting of a very much elongated cigar - shape condensate into two condensates @xcite and their subsequent merging along the axial direction .
our analysis includes finite - temperature effects . in other words , phase fluctuations arising due to temperature
are taken into account during the whole process : splitting and merging .
we observe that as long as the process of splitting and merging is adiabatic enough , both the split and the merged condensates , even at relatively large temperatures , do survive the effects of these fluctuations .
low - dimensional quantum gases exhibit very fascinating properties and have attracted a lot of interest , both theoretically and experimentally @xcite .
it is known that low - dimensional [ one- ( 1d ) and two- ( 2d ) dimensional ] quantum gases differ qualitatively from their 3d counterparts @xcite . from a theoretical point of view , the use of a mean - field theory to describe a low - dimensional quantum gas is severely restricted . a widely used criterion to apply a mean - field approach is to demand that the average distance between particles , d , is clearly smaller than the correlation length of the condensate @xmath3 where @xmath4 , @xmath5 , and @xmath6 denote the mass , the interaction coupling , and the density , respectively . in three dimensions
, the above condition leads to @xmath7 and is well satisfied for small densities , and the description of the system at @xmath8 with a mean - field gross - pitaevskii equation is fully satisfactory . in the one - dimensional case
, however , this ratio behaves as @xmath9 and this fact changes drastically the range of applicability of a mean - field treatment . a careful analysis of the validity of a mean - field treatment in the 1d case @xcite leads to the following condition : @xmath10 where @xmath11 is the number of condensed atoms , @xmath12 and @xmath13 are the axial and radial oscillator lengths , respectively , and @xmath14 and @xmath15 are the angular frequencies in the corresponding directions .
thus , in 1d , contrary to the 3d case , when the density decreases the gas gradually becomes strongly correlated , acquires a fermionic character , and enters into the so - called tonks - girardeau regime @xcite .
experimental demonstration of a tonks gas has been recently achieved @xcite .
the possibility of generating low - dimensional bosonic gases raises the question of the effects of quantum fluctuations . in an untrapped 1d bose system
these fluctuations destroy finite- as well as zero - temperature condensation .
for trapped bose gases , the situation is quite different : for noninteracting bosons in such a trap the finite spacing between the lowest and the next energy level allows for the occurrence of 1d bose - einstein condensation even at finite - temperatures as stipulated in refs . @xcite .
in such a case the absence of gapless excitations indicates that the bec will not be destroyed immediately as interactions between bosons are turned on . in 1d trapping geometries ,
long - wavelength density and phase fluctuations lead to a new intermediate state between a condensate and a noncondensed system , which is commonly referred to as a _ quasicondensate_. in quasicondensates , the phase of the condensate is only coherent over a finite distance that is smaller than the system size . in other words , the phase coherence length is smaller than the axial size of the sample . to understand the nature of quasicondensates at finite - temperature
, one has to analyze the behavior of the single particle correlation function by calculating the fluctuations of phase and density as has been done by petrov _
et al . _
there it is shown that for temperatures below the degeneracy temperature , the condensate s phase indeed fluctuates , but fluctuations of the density are still highly suppressed .
this character of thermal fluctuations is also present in highly elongated 3d gases @xcite , and has been recently observed experimentally in @xcite .
quasi 1d geometries can be accessible in magnetic traps with a cylindrically symmetric harmonic potential along the axial direction that have transverse frequencies @xmath16 much larger than the axial one @xmath14 .
in such configurations the resulting condensate looks like a cigar aligned along the symmetry or z axis . with current technology , condensates with aspect ratio @xmath17 as large as 1000 are achievable . in this paper
, we study the process of splitting and subsequent merging of an elongated condensate confined in a 1d geometry both at @xmath8 , i.e. , when the condensate has a well - defined phase , and at finite - temperature ( @xmath18 ) , in the quasicondensate regime . for finite @xmath19
, we analyze the process of splitting and merging for a wide range of temperatures , i.e. , @xmath20 where @xmath21 ( @xmath22 being the chemical potential ) corresponds to the characteristic temperature above which a true - condensate turns into a quasicondensate in which phase fluctuations begin to play a role . on the other hand @xmath23 is the 1d degeneracy temperature @xcite .
the transition , or crossover , between the different regimes for the 1d degenerate interacting bosonic gas , i.e. , true - condensate , quasi - condensate , and tonks gas is smooth .
thus , in the regime of quasi - condensation density fluctuations are relatively suppressed while phase fluctuations are enhanced .
by keeping all parameters fixed and reducing the number of atoms , phase fluctuations become more and more pronounced , mean - field theory fails , and the gas enters into the strongly correlated regime or tonks gas @xcite . in our case , the splitting of the condensate is achieved by means of a double - well potential grown adiabatically on top of a trapping harmonic potential . by adiabatically switching off the double - well potential ,
we merge these condensates into a single one ( merged bec ) .
we would like to stress that we use the mean - field gross - pitaevskii equation ( gpe ) throughout this paper .
it is worth pointing out that although the gross - pitaevskii equation @xcite describes properly coherent evolution of the atomic mean - field at @xmath8 , it can also be used to solve time evolution at finite - temperature in a relatively straightforward manner .
it might look like that the mean - field method of the gpe allows to make statements only about first order coherence .
but , as it is well known , a closely inspection reveals that the gpe contains as such classical bogoliubov - de gennes ( bdg ) equations , i.e. equations describing small fluctuations around a given solution of the gpe .
since we are here interested in the study of relatively high temperatures @xmath24 , classical description of fluctuations is fully appropriate .
it turns out that one can simulate temperature effects by adding fluctuations to the ground state solution of the gpe at @xmath8 in a way which mimics thermal fluctuations . at low temperatures @xmath25 , where @xmath26 @xcite , this can be generally done by identifying phonon ( quasi - particle ) modes , i.e. eigensolutions of the bdg equations .
the fluctuations are expressed thus as sums over the quasi - particles with amplitudes taken from monte carlo sampling and corresponding to the thermal ( boltzmann ) populations of the quasi - particle modes .
such a method is used in this paper , with the additional simplification that for quasi-1d situations only the phase fluctuations are relevant .
note , that once we add the fluctuations initially at @xmath27 , and as long as they remain small in the course of evolution , they will propagate in time with a very good approximation as appropriate solutions of the time dependent bdg equations .
note also , that our approach allows in principle to obtain information about all correlation functions : either in the form of an average over monte carlo realizations of the initial fluctuations , or in the form of time averages , due to the ( expected ) ergodic character of the evolution . in some cases ,
even averaging over the initial data is not necessary : the results for different realizations are so similar that looking for few single cases allows one to draw conclusions about `` coherence '' . by coherence
we mean here , that the splitting process of an initial 1d bec into two spatially separated 1d becs occurs with a well - defined relative phase between them .
their subsequent merging into a single merged bec , when the process is fully coherent , should result in a single condensate with a well - defined relative phase with respect to the initial one .
if this is the case , there is a perfect overlap between the densities of the initial and merged condensates .
it is worth stressing that our approach is a simplified version of the classical field methods used by several groups @xcite . in their approach
some emphasis was put on the explanation of the temperature concept , and a certain model of finite - temperature effects has been studied .
particularly interesting is here the possibility of extracting higher order correlations from a single shot measurements @xcite , which in our case corresponds to a single realizations of the initial fluctuations .
here we consider @xmath28rb condensate with @xmath29 atoms confined in an harmonic trap with frequencies @xmath30 hz and @xmath31 hz . for such parameters , the system is in the 1d thomas - fermi regime ( @xmath32 ) along the axial direction .
the 1d chemical potential is given by @xmath33 where @xmath34 is the effective 1d coupling strength @xcite .
for our parameters , transverse excitations are suppressed ( @xmath35 ) , and the dynamics of such a bose gas can be described by the usual mean - field gpe in 1d @xmath36 where @xmath37 is the mean - field order parameter , or in other words the condensate wave function .
the potential term @xmath38 includes both the magnetic trap and the double - well potential as described below : @xmath39 the potential used to split the condensate into two spatially separated condensates , @xmath40 is switched on and off adiabatically by means of a time - dependent function @xmath41 the maximum depth of this potential is @xmath42 in terms of the recoil energy @xmath43 and @xmath44 to achieve spatial separation of the split condensates one has to require that the distance between the two wells is , at least , of the order of the thomas - fermi radius ( @xmath45 ) this radius is given by @xmath46 so that @xmath47 for @xmath48 atoms , the thomas - fermi radius is @xmath49 m .
this is in agreement with the experimental results in refs .
@xcite .
the time dependent function @xmath50 controls the switching on and off of the double - well potential and hence the overall splitting and merging process of the condensates .
we define this function as @xmath51 in this equation ( [ tevo - tmer ] ) , @xmath52 is the time duration required for evolving the 1d bec in real time before the splitting process begins .
it extends for 10ms .
just immediately after , the splitting process begins and continues for a time interval @xmath53 . at the end of @xmath53 , the function @xmath50 attains a maximum value of unity and remains constant for a time interval @xmath54 . during this time interval ,
two spatially separated 1d becs are created . at the beginning of the time span @xmath55 , merging of the condensates starts by switching off the double - well potential .
this process continues until the two 1d becs merge together .
a complete merging is only possible when @xmath50 becomes finally zero , i.e. , when the optical potential is completely switched off . in this case
the atoms remain only under the influence of the trapping potential . to ensure coherence during the process of the switching on and off of the optical potential , the raising of the double - well has to be slow enough to avoid excitations and to allow for quantum tunneling between the two wells .
notice that the relevant time scale of excitations is given by the inverse of the frequency of the trap , in our case @xmath56 ms .
if @xmath57 we expect coherent splitting and merging . on the other hand , for @xmath58 of the order of @xmath59 or less , the process creates more and more excitations that cause incoherence .
the faster the double - well potential switches on and off the stronger the excitations are . to check our claims
we have carried out further numerical simulations for several values of @xmath53 and @xmath55 around @xmath59 . in general , we have observed that a coherent process , at @xmath8 , is achieved for switching times of 400 ms ( or larger ) .
the full time dependency of @xmath50 necessary for the coherent splitting and merging of the condensates based on eq .
( [ tevo - tmer ] ) is illustrated in fig .
[ splitparam ] . .
for the whole process , we fixed a time evolution of @xmath60 ms , a splitting time of @xmath61 ms , a time @xmath62 ms in which @xmath50 remains constant ( unity ) , and a time of merging @xmath63 ms and an additional time of @xmath64 ms for allowing the merged condensate to make its final evolution into a single condensate . ]
we calculate first the ground state of the system by evolving the gpe ( [ 1d - gpe - t0 ] ) with @xmath65 in imaginary time .
after the ground state has been found , we numerically solve eq .
( [ 1d - gpe - t0 ] ) with the pulse profile given by the time dependent - function @xmath50 .
the results of our simulations at temperature @xmath8 , are summarized in fig .
[ merg - denst0 ] .
there we display the condensates density at three different times .
first at @xmath66 the initial condensate ( initial 1d bec ) has a characteristic thomas - fermi profile with @xmath67 m .
then at @xmath68 ms ( corresponding to the ramping up of the double - well potential in 400 ms and keeping it constant during additional 400 ms ) , two spatially well separated identical condensates ( split becs ) appeared , centered at @xmath69 , each of them with a number of atoms @xmath70 . finally , we display the density of the condensate at @xmath71 ms .
the merged condensate ( merged 1d bec ) has exactly the same profile as the initial one ( initial 1d bec ) and thus they can not be distinguished in the figure , asserting that the process of splitting and merging is fully coherent . .
after coherently splitting an initial 1d condensate ( initial bec ) into two spatially separated ones ( split becs ) , by adiabatically switching on the double - well potential given in fig .
[ pot ] , centered at @xmath72 and @xmath73 , the optical potential is again switched off adiabatically .
this leads to a coherent merging of the split condensates into a single one ( merged 1d bec ) . the merged 1d bec overlaps on top of the initial 1d bec which is an indication of a completely coherent merging process . ]
coherence may be a prerequisite for further applications such as in atom interferometry and quantum - information processing @xcite . before we proceed to the case of finite - temperature
, it may be relevant to address the case of a nonadiabatic splitting and merging process , i.e. , when the double - well potential switches on and off too fast .
we have carried out simulations for switching times as short as 20 ms . in such cases
the splitting of the condensate ( even at @xmath8 ) becomes completely incoherent , there is no trace of phase preservation , and the condensate is destroyed , as is shown in fig .
[ incoherence_split_merg ] . .
the initial bec is split incoherently by considering a very small splitting time ( @xmath53 ) into two picked spatially separated ones ( incoherent split ) by nonadiabatically switching on the double - well potential ( fig .
[ pot ] ) .
the optical potential is again switched off nonadiabatically which leads to a completely incoherent pattern ( incoherent merging ) due to larger excitations . ] for shorter splitting times , @xmath74 ms , it is not even possible to split the initial bec into two well spatially separated condensates .
when it comes to the time of merging ( @xmath75 ) , a similar behavior is observed if @xmath76 .
so far we have considered only the case of @xmath8 where the initial condensate has a well - defined phase . in this section , following exactly the same approach as in the previous section , we investigate the effects that finite - temperature might have on the overall coherence during the splitting and merging process .
fluctuations of phase and density of a bec are the result of thermal excitations , and appear usually at finite - temperature @xcite . in such a case
the system includes some noncondensed or thermal particles and hence the total bose field operator can be expressed as @xmath77 where @xmath78 describes the thermal depletion part . for a bec in 3d trapping geometries , fluctuations of density and phase
are only important in a narrow temperature range near the bec transition temperature @xmath0 @xcite . for 1d systems , however , phase fluctuations are present at temperatures far below the degeneracy temperature .
phase fluctuations can be studied by solving the bogoliubov - de gennes equations describing elementary excitations .
writing the quantum field operator as @xmath79 $ ] where @xmath80 denotes the condensate density at @xmath81 [ @xmath82 the phase and density operators take , respectively , the following forms @xcite : @xmath83 \label{phas - fluc}\ ] ] and @xmath84 where @xmath85 is the annihilation ( creation ) operator of the excitations with quantum number @xmath86 and energy @xmath87 , and @xmath88 , where @xmath89 and @xmath90 denote the excitation functions determined by the bdg equations .
more explicitly , the functions @xmath91 in a 1d thomas - fermi regime and at finite - temperature take the form : @xmath92\right)^{\pm 1/2 } \nonumber \\ & & \times p_j\left(z / r_{\rm tf}\right ) \label{wavefuns}\end{aligned}\ ] ] where @xmath93 are legendre polynomials .
the phase coherence length , in terms of the thomas - fermi radius @xmath94 , is expressed as @xmath95 and characterizes the maximal distance between two phase - correlated points in the condensate .
phase fluctuations increase for large trap aspect ratios and small @xmath11 @xcite .
temperature is included at the level of the gpe ( [ 1d - gpe - t0 ] ) by calculating first the density at @xmath8 in the presence of the magnetic trap only , and then mimicking finite - temperature effects via the phase operator of eq .
( [ phas - fluc ] ) . in other words
a phase is imprinted on the condensate wave function at this stage . to this aim
, we calculate the bose occupation @xmath96 modes in the bogoliubov approximation for fixed temperatures replacing the operators @xmath97 and @xmath98 by random complex variables @xmath99 and @xmath100 , respectively such that @xmath101 @xcite .
although the gpe ( [ 1d - gpe - t0 ] ) in this limit remains valid , the bdg equations become modified by the integration over the transverse profile of the condensate @xcite . in effect
the mode functions @xmath102 are given by jacobi polynomials , whereas @xmath103 are given by a slightly different expression than in the pure 1d case .
we stress , however , that for the regimes of temperatures we consider , there will be no qualitative and practically no quantitative difference between the pure- and quasi-1d results .
the dependence of the magnitude of the phase fluctuations of the condensate on the temperature is shown in fig .
[ phas ] . even though we present here results calculated for a fixed number of condensate atoms ( @xmath11 ) , it is numerically verifiable that the magnitude of the phase fluctuations is inversely proportional to the square root of the number of atoms and hence to the peak density of the condensate as described by eq .
( [ phas - fluc ] ) . ) .
( b ) with the rise in temperature ( @xmath104 ) , the phase fluctuations begin to be enhanced .
( c ) still at an intermediate but relatively high temperature ( @xmath105 ) , the magnitude of phase fluctuations increases .
( d ) finally at @xmath106 , which is near the critical point , stronger phase fluctuations are displayed . ] as can be seen by inspecting the different plots ( a)-(d ) in fig .
[ phas ] , the phase fluctuations get more and more enhanced with the increase of temperature .
having seen the enhancement of phase fluctuations with temperature , we proceed now to analyze the coherence properties of splitting and merging of a 1d bec in the presence of these phase fluctuations for temperatures in the range of @xmath107 .
our results for finite - temperature are summarized in fig .
[ denstemp ] .
= ( i ) 0.01 , ( ii ) 0.1 , ( iii ) 0.4 and ( iv ) 0.8 .
each one in turn consists three plots with in it : curve a , the initial 1d bec ( red solid line ) ; curve b the split 1d becs ( green dashed line ) ; and curve c , the merged 1d bec ( blue dotted line ) . ] using exactly the same approach that we employed for the adiabatic case at @xmath8 but now including temperature we study again the splitting and merging process for the same parameters . in this case , the density fluctuations , which are highly suppressed in elongated 3d condensates , are very pronounced in the 1d density after time evolution , when the temperature increases from a very small value ( @xmath108 ) to a value near the critical point for condensation ( @xmath106 ) . at high temperatures ,
the density fluctuations get more and more enhanced .
however , notice that the effect of fluctuations is very similar on the single and on the merged condensates .
on the other hand , fluctuations on the split condensates remain relatively small due to fact that the densities in the double wells are higher as shown in all plots of fig .
[ denstemp ] .
this is in confirmation of the prediction of eq .
( [ phas - fluc ] ) .
the presence of fluctuations on the density profile is a consequence of the bdg equations present in the gpe equation . in spite of such fluctuations , the split and merged condensates present almost the same thomas - fermi density profiles for any temperature @xmath25 .
from these observations , we conclude that in spite of the initial phase fluctuations in the quasicondensate regime at finite - temperature , there is a preservation of `` phase coherence length '' during the splitting and merging process , if adiabaticity is satisfied .
summarizing , we have discussed coherence effects in the splitting and merging of a 1d @xmath28rb bec .
this is done by creating two spatially separated condensates from an initial 1d condensate by deforming the trapping potential into a double - well potential .
we have analyzed the case of zero temperature as well as finite - temperature in the so - called quasi - condensate regime , where the phase coherence length is smaller than the size of the system . at zero temperature and for a process adiabatic enough where the splitting and merging times are much larger than the characteristic time of the system given by the inverse of the trap frequency @xmath109 , a coherent splitting followed by a coherent merging is achieved and there is a constant relative phase between the initial and the final merged condensates . on the contrary
if the splitting and merging times are not larger than the relevant time scale , the split condensates acquire a random relative phases and merging is no longer possible . in this case
, the system acquires large density and phase fluctuations on a length scale comparable with the coherence length @xmath110 , and a description based on the gpe becomes invalid . in the case of finite - temperature
our results show that even in the presence of strong phase fluctuations , if the process of splitting and merging is carried out fully adiabatically , the condensate preserves the thomas - fermi density profile and there is phase coherence length preservation . in such cases ,
the merged condensate is a quasi - condensate with the same initial density profile as the initial condensate and with the same phase coherence length , although the `` overall '' phase of the quasi - condensate is not preserved .
this situation occurs even at temperatures very near the critical value for condensation as long as the trapping potential remains in place .
our results may have a useful implication for manipulating 1d becs at zero as well as finite - temperatures such as in atom lasers , interferometry , and solitons .
in particular , our results agree qualitatively well with the recent measurements in ref . @xcite .
a. m. wishes to thank the deutscher akademisher austausch dienst ( daad ) for financial support .
we thank v. ahufinger , j. arlt , h. kreutzmann , j. schmiedmeyer , and m. k. oberthaler for discussions .
we acknowledge support from deutsche forschungsgemeinschaft ( sfb 407 , spp1116 , 432 pol , and gk 282 ) , from the esf programme qudedis , from the eu ip programme scala , and the spanish grant no .
mcyt bfm-2002 - 02588 .
99 m. h. anderson , j. r. ensher , m. r. matthews , c. e. wieman , and e. a. cornell , science * 269 * , 198 ( 1995 ) . k. b. davis , m .- o .
mewes , m. r. andrews , n. j. van druten , d. s. durfee , d. m. kurn , and w. ketterle , phys .
lett . * 75 * , 3969 ( 1995 ) . c. c. bradley
, c. a. sackett , j. j. tollett , and r. g. hulet , phys .
* 75 * , 1687 ( 1995 ) .
d. g. fried , t. c. killian , l. willmann , d. landhuis , s. c. moss , d. kleppner , and t. j. greytak , phys .
lett . * 81 * , 3811 ( 1998 ) .
a. robert , o. sirjean , a. browaeys , j. poupard , s. nowak , d. boiron , c. i. westbrook , and a. aspect , science * 292 * , 461 ( 2001 ) .
s. jochim , m. bartenstein , a. altmeyer , g. hendl , s. riedl , c. chin , j. h. denschlag , and r. grimm , science * 302 * , 2101 ( 2003 ) .
m. w. zwierlein , c. a. stan , c. h. schunck , s. m. f. raupach , s. gupta , z. hadzibabic , and w. ketterle , phys .
lett . * 91 * , 250401 ( 2003 ) .
m. greiner , c. a. regal , and d. s. jin , nature ( london ) * 426 * , 537 ( 2003 ) .
m. khl , t. w. hansch , and t. esslinger , phys .
a. * 65 * , 021606(r ) ( 2002 ) .
w. h. pinkse , t. fischer , p. maunz , and g. rempe , nature ( london ) * 404 * , 365 ( 2000 ) .
a. p. chikkatur , y. shin , a. e. leanhardt , d. kielpinski , e. tsikata , t. l. gustavson , d. e. pritchard , and w. ketterle , science * 296 * , 2193 ( 2002 ) . t. h. maiman , nature ( london ) * 187 * , 493 , ( 1960 ) .
w. yi and l .- m .
duan , phys .
a * 72 * , 043607 ( 2005 ) .
z. hadzibabic , s. stock , b. battelier , v. bretin , and j. dalibard , phys .
lett . * 93 * , 180403 ( 2004 ) .
y. shin , m. saba , t. a. pasquini , w. ketterle , d. e. pritchard , and a. e. leanhardt , phys .
* 92 * , 050405 ( 2004 ) .
l. pezz , l. a. collins , a. smerzi , g. p. berman , and a. r. bishop , phys .
a * 72 * , 043612 ( 2005 ) .
t. schumm , s. hofferberth , l. m. andersson , s. wildermuth , s. groth , i. bar - joseph , j. schmiedmayer , and p. kruger , nature * 1 * , 57 ( 2005 ) . j. javanainen and m. wilkens , phys .
lett . * 78 * , 4675 ( 1997 ) .
a. j. leggett and f. sols , phys .
lett . * 81 * , 1344 ( 1998 ) .
n. j. van druten and w. ketterle , phys .
rev a * 54 * , 656 ( 1996 ) .
d. s. petrov , g. v. shlyapnikov , and j. t. m. walraven , phys .
lett . * 85 * , 3745 ( 2000 ) .
h. moritz , t. stferle , m. khl , and t. esslinger , phys .
* 91 * , 250402 ( 2003 ) . t. stferle , h. moritz , c. schori , m. khl , and t. esslinger , phys .
* 92 * , 130403 ( 2004 ) .
b. paredes , a. widera , v. murg , o. mandel , s. flling , i. cirac , g. v. shlyapnikov , and t. w. hnsch , nature ( london ) * 429 * , 277 ( 2004 ) . t. kinoshita , t. wenger , and d. s. weiss , science * 305 * , 1125 ( 2004 ) .
a. grlitz , j. m. vogels , a. e. leanhardt , c. raman , t. l. gustavson , j. r. abo - shaeer , a. p. chikkatur , s. gupta , s. inouye , t. rosenband , and w. ketterle , phys .
* 87 * , 130402 ( 2001 ) .
d. s. petrov , m. holzmann , and g. v. shlyapnikov , phys .
lett . * 84 * , 2551 ( 2000 ) . yu .
kagan , v. a. kashurnikov , a. v. krasavin , n. v. prokofev , and b. v. svistunov , phys .
a * 61 * , 43608 ( 2000 ) .
a. i. safonov , s. a. vasilyev , i. s. yasnikov , i. i. lukashevich , and s. jaakkola , phys .
lett . * 81 * , 4545 ( 1998 ) .
lev pitaevskii and s. strigari , bose - einstein condensation ( oxford science publications , london , 2003 ) .
m. girardeau , j. math . phys .
* 1 * , 516 ( 1960 ) . c. menotti and s. stringari , phys
a * 66 * , 043610 ( 2002 ) .
k. v. kheruntsyan , d. m. gangardt , p. d. drummond , and g. v. shlyapnikov , phys .
* 91 * , 040403 ( 2003 ) .
p. pedri and l. santos , phys .
* 91 * , 110401 ( 2003 ) .
d. s. petrov , g. v. shlyapnikov , and j. t. m. walraven , phys .
lett . * 87 * , 050404 ( 2001 ) .
s. dettmer , d. hellweg , p. ryytty , j. j. arlt , w. ertmer , k. sengstock , d. s. petrov , g. v. shlyapnikov , h. kreutzmann , l. santos , and m. lewenstein , phys .
87 * , 160406 ( 2001 ) .
h. kreutzmann , a. sanpera , l. santos , m. lewenstein , d. hellweg , l. cacciapuoti , m. kottke , t. schulte , k. sengstock , j. j. arlt , and w. ertmer , appl .
b : lasers opt . *
76 * , 165 ( 2003 ) .
d. hellweg , s. dettmer , p. ryytty , j. j. arlt , w. ertmer , k. sengstock , d. s. petrov , g. v. shlyapnikov , h. kreutzmann , l. santos , and m. lewenstein , appl .
b : lasers opt . *
73 * , 781 ( 2001 ) .
s. richard , f. gerbier , j. h. thywissen , m. hugbart , p. bouyer , and a. aspect , phys .
rev . lett . * 91 * , 010405 ( 2003 ) . f. gerbier , j. h. thywissen , s. richard , m. hugbart , p. bouyer , and a. aspect , phys .
a * 67 * , 051602(r ) ( 2003 ) .
d. kadio , m. gajda , and k. rzaewski , phys .
a * 72 * , 013607 ( 2005 ) .
f. dalfovo and s. giorgini , rev .
71 * , 463 ( 1999 ) .
m. j. davis , s. a. morgan , and k. burnett , phys .
lett . * 87 * , 160402 ( 2001 )
. m. j. davis , s. a. morgan , and k. burnett , phys .
a * 66 * , 053618 ( 2002 ) .
m. j. davis , r. j. ballagh , and k. burnett , j. phys .
b * 34 * , 4487 ( 2001 )
. m. j. davis and s. a. morgan , phys .
a * 68 * , 053615 ( 2003 ) .
m. brewczyk , p. borowski , m. gajda , and k. rzaewski , j. phys .
b * 37 * , 2725 ( 2004 ) .
k. gral , m. gajda , and k. rzaewski , opt .
express * 8 * , 92 ( 2001 ) .
k. gral , m. gajda , and k. rzaewski , phys .
a * 66 * , 051602(r ) ( 2002 ) .
h. schmidt , k. gral , f. floegel , m. gajda , and k. rzaewski , j. opt .
b : quantum semiclassical opt . * 5 * , s96 ( 2003 ) .
r. bach and k. rzaewski , phys .
a * 70 * , 063622 ( 2004 ) .
r. bach and k. rzaewski , phys .
* 92 * , 200401 ( 2004 ) .
e. altman , e. demler , and m. d. lukin , phys .
a * 70 * , 013603 ( 2004 ) .
y. shin , c. sanner , g.b .
jo , t.a .
pasquini , m. saba , w. ketterle , d.e .
pritchard , m. vengalattore , and m. prentiss , phys .
a * 72 * , 021604(r ) ( 2005 ) .
m. olshanii , phys .
lett . * 81 * , 938 ( 1998 ) .
s. i. shevchenko , sov .
phys . * 18 * , 223 ( 1992 ) . | we analyze coherence effects during the splitting of a quasi one - dimensional condensate into two spatially separated ones and their subsequent merging into a single condensate .
our analysis takes into account finite - temperature effects , where phase fluctuations play an important role .
we show that , at zero - temperature , the two split condensates can be merged into a single one with a negligible phase difference . by increasing temperature to a finite value below the critical point for condensation ( @xmath0 ) , i.e. , @xmath1 ,
a considerable enhancement of phase and density fluctuations appears during the process of splitting and merging .
our results show that if the process of splitting and merging is sufficiently adiabatic , the whole process is quite insensitive to phase fluctuations and even at high temperatures , a single condensate can be produced . |
quantum entanglement @xcite describes a scenario where the quantum states of two objects separated in space are strongly correlated .
these correlations can be exploited in emerging technologies such as quantum computing , should one be able to spatially separate the entangled objects without destroying the correlations . in a broader context , quantum entanglement could prove to be of practical importance in the fields of spintronics @xcite and information cryptography @xcite .
it also holds a considerable interest from a purely fundamental physics point of view , prompting some of the more philosophically inclined discussions related to quantum theory and causality .
superconductors have been proposed as natural sources for entangled electrons @xcite , as cooper pairs consist of two electrons that are both spin and momentum - entangled .
the cooper pair can be spatially deformed by means of the crossed andreev reflection ( car ) process in superconducting heterostructures . in this scenario ,
an electron and hole excitation are two separate metallic leads are coupled by means of andreev scattering processes at two spatially distinct interfaces .
unfortunately , the signatures of car are often completely masked by a competing process known as elastic co - tunneling ( ct ) which occur in the same type of heterostructures .
in fact , the conductances stemming from ct and car may cancel each other completely @xcite , thus necessitating the usage of noise - measurements to find fingerprints of the car process in such superconducting heterostructures .
recently , graphene @xcite has been studied as a possible arena for car - processes . in ref .
@xcite , it was shown how a three - terminal graphene sheet containing @xmath0-doped , @xmath1-doped , and superconducting regions could be constructed to produce perfect car for one particular resonant bias voltage .
also , the signatures of the car process in the noise - correlations of a similar device were studied in ref .
however , the role played by the spin degree of freedom in graphene devices probing non - local transport has not been addressed so far .
this is a crucial point since it might be possible to manipulate the spin - properties of the system to interact with the spin - singlet symmetry of the cooper pair in a fashion favoring car . in this paper ,
we show that precisely such an opportunity exists it is possible to obtain a spin - switch effect between virtually perfect car and perfect ct in a superconducting graphene spin valve .
in contrast to ref .
@xcite , this effect is seen for all bias voltages in the low - energy regime rather than just at one particular applied voltage difference .
the key observation is that the possibility of tuning the local fermi - level to values equivalent to a weak , magnetic exchange splitting in graphene renders both the usual andreev reflection process and ct impossible .
in contrast , this opportunity does not exist in conventional conductors where the fermi energy is large and of order @xmath2(ev ) .
we show that graphene spin valves provide a possibility for a unique combination of non - local andreev reflection and spin - dependent klein tunneling @xcite . our model
is shown in fig .
[ fig : model ] , where ferromagnetism and superconductivity are assumed to be induced by means of the proximity effect @xcite to leads with the desired properties .
a similar setup was considered in ref .
@xcite , where the magnetoresistance of the system was studied .
we organize this work as follows . in sec .
[ sec : theory ] , we establish the theoretical framework which will be used to obtain the results . in sec .
[ sec : results ] , we present our main findings for the non - local conductance in the graphene superconducting spin - valve with a belonging discussion of them . finally , we summarize in sec . [
sec : summary ] .
we consider a ballistic , two - dimensional graphene structure as shown in fig .
[ fig : model ] . in the left ferromagnetic region @xmath3 ,
the exchange field is @xmath4 , while it is @xmath5 in the right ferromagnetic region @xmath6 . in the superconducting region @xmath7
, the order parameter is taken to be constant with a real gauge @xmath8 . to proceed analytically
, we make the usual approximation of a step - function behavior at the interfaces for all energy scales , i.e. the chemical potentials @xmath9 , the exchange field @xmath10 , and superconducting gap @xmath11 .
this assumption is expected to be good when there is a substantial fermi - vector mismatch between the f and s regions , as in the present case . to make contact with the experimentally relevant situation
, we assume a heavily doped s region satisfying @xmath12 .
we use the dirac - bogoliubov de gennes equations first employed in ref .
@xcite . for quasiparticles with spin @xmath13 ,
one obtains in an f@xmath14s graphene junction : @xcite @xmath15 where @xmath16 \hat{1}\end{aligned}\ ] ] and @xmath17 denotes a @xmath18 matrix . here ,
we have made use of the valley degeneracy and @xmath19 is the momentum vector in the graphene plane while @xmath20 is the vector of pauli matrices in the pseudospin space representing the two a , b sublattices of graphene hexagonal structure .
the superconducting order parameter @xmath21 couples electron- and hole - excitations in the two valleys ( @xmath22 ) located at the two inequivalent corners of the hexagonal brillouin zone .
the @xmath23 spinor describes the electron - like part of the total wavefunction @xmath24 and in this case reads @xmath25 while @xmath26 . here
, @xmath27 denotes the transpose while @xmath28 is the time - reversal operator . from eq .
( [ eq : bdg ] ) , one may now construct the quasiparticle wavefunctions that participate in the scattering processes @xcite .
we consider positive excitation energies @xmath29 with incoming electrons of @xmath0-type , i.e. from the conduction band @xmath30 ( we set @xmath31 from now on ) .
the incoming electron from the left ferromagnet may either be reflected normally or andreev - reflection ( ar ) . in the latter process ,
it tunnels into the superconductor with another electron situated at @xmath32 , leaving behind a hole excitation with energy @xmath33 .
the scattering coefficients for these two processes are @xmath34 and @xmath35 , respectively , and the total wavefunction may thus be written as : @xmath36 where we have defined the wavevectors @xmath37 we have omitted a common factor @xmath38 for all wavefunctions . similarly , assuming that the charge carriers in the right ferromagnetic region are also of the @xmath0-type , we obtain : @xmath39 it should be noted that the ar hole is generated in the conduction band if @xmath40 ( retro - ar ) , whereas it is generated in the valence band otherwise ( specular - ar ) . the @xmath41 sign above refers to parallell / antiparallell ( p / ap ) magnetization configuration .
we assume that the superconducting region is heavily doped , @xmath42 , which causes the propagating quasiparticles to travel along the @xmath43-axis since the scattering angle in the superconductor satisfies @xmath44 .
we obtain the following wavefunction ( @xmath45 ) : @xmath46 where @xmath47 while @xmath48 for subgap energies @xmath49 and @xmath50 for supergap energies @xmath51 .
it is important to consider carefully the scattering angles in the problem .
since we assume translational invariance in the @xmath52-direction , the @xmath52-component of the momentum is conserved .
this gives us @xmath53 it is clear that the angle of transmission for the electrons in the right ferromagnet is equal to the angle of incidence when the magnetizations are p , i.e. @xmath54 .
also , one infers that there exists a critical angle above which the scattered waves become evanescent , i.e. decaying exponentially .
this may be seen by observing that the scattering angles exceed @xmath55 ( thus becoming imaginary ) above a certain angle of incidence @xmath56 .
for instance , the ar wave in the left ferromagnetic region becomes evanescent for angles of incidence @xmath57 , where the critical angle @xmath58 is obtained by setting @xmath59 in the equation @xmath60 expressing conservation of momentum perpendicular to the interface .
one finds that : @xmath61|.\end{aligned}\ ] ] thus , ar waves in the regime @xmath62 do not contribute to any transport of charge .
a similar argument can be made for the transmitted electron wave - function in the right ferromagnetic region , corresponding to the ct process , where the critical angle for this process becomes @xmath63|.\end{aligned}\ ] ] in the p configuration , the ct process thus always contributes to the transport of charge .
finally , the contribution to transport of charge from car comes from the hole - wave function in the right ferromagnetic region , which becomes evanescent for angles of incidence above the critical angle @xmath64|.\end{aligned}\ ] ] in the p configuration , this criteria is the same as the vanishing of local ar expressed by eq .
( [ eq : criticalar ] ) .
intuitively , one might expect that the most interesting phenomena occur when the exchange field @xmath10 is comparable in magnitude to the chemical potential @xmath65 . if @xmath66 , the effect of the exchange field should be minor and the ar is never specular .
in contrast , the situation becomes quite fascinating when we consider the case @xmath67 under the assumption of a doped situation @xmath68 .
first of all , the incoming quasiparticles from the left ferromagnetic region are completely dominated by the majority spin carriers @xmath69 , since the density of states ( dos ) for @xmath70 electrons vanishes at the fermi level . since @xmath67 , the ar process is suppressed for all incoming waves as @xmath71 .
we now show how the fate of the cross - conductance in the right ferromagnetic region depends crucially on whether the magnetization configuration is p or ap . in the p configuration , we see that @xmath72 , which means that the transport is purely governed by the ct process . in the ap configuration , we see that @xmath73 , which means that the transport is mediated purely by the car process .
this suggests a remarkable spin - switch effect by reversing the direction of the field in the right ferromagnet , one obtains an abrupt change from pure ct to pure car processes mediating the transport of charge . in each case
, there is no local ar in the left ferromagnetic region . in the standard metallic case ,
the distinct signatures for the ct and car contributions are masked by each other , and it becomes necessary to resort to noise - measurements in order to say something about the contribution from each process . in the present scenario ,
we have showed how it is possible to separate the two contributions directly by a simple spin - switch effect which is commonly employed in experimental work on f@xmath14s heterostructures .
( color online ) plot of the conductance for ct processes @xmath74 versus bias voltage in the upper panel and versus length of the s region in the lower panel . here
, we consider the p alignment and @xmath67 such that @xmath75 .
, width=302 ] let us now evaluate the conductance in the p and ap configuration quantitatively by using @xmath76 where we have introduced @xmath77 as the spin-@xmath13 normal - state conductance that takes into account the valley degeneracy , in addition to @xmath78 the density of states is determined by @xmath79 where @xmath80 is the width of the junction .
the expression for @xmath81 is obtained by replacing @xmath82 with @xmath83 in eq .
( [ eq : conductance ] ) .
since we here consider the case @xmath67 and @xmath84 , the formulas for the @xmath85 and @xmath81 may be simplified since @xmath86 .
also , since the dos vanishes for minority spins for the injected electrons , only @xmath69 contributes for incoming electrons .
the crucial point here is that in the p alignment , @xmath87 and @xmath88 such that @xmath89 and @xmath90 such that @xmath91 and @xmath92 . assuming a value of @xmath93 mev for the proximity - induced gap , this corresponds to an exchange splitting of @xmath94 mev in the f regions and a doping level @xmath95 mev in the s region , which should be experimentally feasible @xcite and well within the range of the validity for the linear dispersion relation in graphene . in fig .
[ fig : p ] , we plot the cross - conductance @xmath96 in the p alignment both as a function of bias voltage and width of the s region .
the same thing is done for @xmath97 in the ap alignment in fig .
[ fig : ap ] . in both cases ,
the magnitude of the conductance varies strongly when considering different widths @xmath98 due to the fast oscillations which pertain to the formation of resonant transmission levels inside the superconductor .
also , it is seen that while the ct process is favored for short junctions @xmath99 , the car process is suppressed in this regime in favor of normal reflection . upon increasing the junction width , the ct conductance drops while the car conductance peaks at widths @xmath100 .
the remarkable aspect is that it is possible to switch between these two scenarios of exclusive ct and exclusive car simply by reversing the direction of magnetization in one of the ferromagnetic layers .
( color online ) plot of the conductance for car processes @xmath101 versus bias voltage in the upper panel and versus length of the s region in the lower panel . here , we consider the ap alignment and @xmath67 such that @xmath102 .
, width=302 ] in order to obtain analytical results , we have assumed that the coulomb interaction and charge inhomogeneities may be neglected .
it would be challenging to obtain a truly homogeneous chemical potential in a graphene sheet , and electron - hole puddles appear to be an intrinsic feature of graphene sheets @xcite .
moreover , it has been speculated that such charge inhomogeneities may play an important role with regard to limiting the transport characteristics of graphene @xcite near the dirac points .
however , for our purposes this is actually beneficial
it is precisely the suppression of charge and spin transport at fermi level for the andreev reflection and co - tunneling process which renders possible the spin - switch effect .
therefore , we do not expect that the inclusion of charge inhomogeneities should alter our results qualitatively .
finally , we note that since the spin of the charge - carriers in each of the non - superconducting graphene sheets are practically speaking fixed due to the vanishing dos for minority spins , the spin - switch effect for car and ec predicted in this paper can not be directly related to entanglement .
nevertheless , it constitutes a clear non - local signal for quantum transport which can be probed experimentally , and should be helpful in identifying clear - signatures of the mesoscopic car phenomenon .
to summarize , we have considered non - local quantum transport in a graphene superconducting spin - valve .
we have shown how one may create a spin - switch effect between perfect elastic co - tunneling and perfect crossed andreev - reflection for all applied bias voltages by reversing the magnetization direction in one of the ferromagnetic layers .
the basic mechanism behind this effect is that the local fermi - level in graphene may be tuned so that the fermi surface for minority spins reduces to a single point in the presence of a weak , magnetic exchange splitting .
this is very distinct from the equivalent spin valve structures in conventional metallic systems , where noise - measurements are required to clearly distinguish between these processes .
h. b. heersche , p. jarillo - herrero , j. b. oostinga , l. m. k. vandersypen , and a. f. morpurgo , nature ( london ) * 446 * , 56 ( 2006 ) ; a. shailos , w. nativel , a. kasumov , c. collet , m. ferrier , s. gueron , r. deblock , and h. bouchiat , europhys . lett . *
79 * , 57008 ( 2007 ) . | we consider the non - local quantum transport properties of a graphene superconducting spin - valve .
it is shown that one may create a spin - switch effect between perfect elastic co - tunneling ( ct ) and perfect crossed andreev - reflection ( car ) for all bias voltages in the low - energy regime by reversing the magnetization direction in one of the ferromagnetic layers .
this opportunity arises due the possibility of tuning the local fermi - level in graphene to values equivalent to a weak , magnetic exchange splitting , thus reducing the fermi surface for minority spins to a single point and rendering graphene to be half - metallic .
such an effect is not attainable in a conventional metallic spin - valve setup , where the contributions from ct and car tend to cancel each other and noise - measurements are necessary to distinguish these processes . |
estimating a density function using a set of initial data points in order to find probability information is a very significant tool in statistics@xcite .
the method of kernel density estimation ( kde)@xcite is now standard in many analysis and applications . furthermore , this idea has been applied in multiple fields ( archaeology @xcite , economy @xcite , etc ) .
the author of this article is particularly interested in constructing perception of security ( pos ) hotspots using ( kde ) methods to analyze real data registered by security experts in bogot @xcite .
nowadays a wide variety of methods are available to find density functions ( kde ) @xcite,@xcite .
the method of kde via difussion is of particular interest for this document ; a recent article @xcite develops a systematic method for ( kde ) using the diffusion equation , also they propose a more general equation to solve some biases for data estimation .
however in their analysis , it is only considered the normalization ( conservation of mass ) of the density function via neumann boundary conditions , the mean of the sample data is not considered , thus inducing a change of an important initial parameter from the discrete data sample . in this article
, we propose a new set of boundary conditions for the diffusion equation that maintain the initial mean and mass of the the discrete data sample in the density estimation process .
a complete study of this framework is performed using the finite element method ( fem ) to solve the one - dimensional diffusion equation for different boundary conditions .
we show the induced error on the final density when the mean is not conserved .
we also show how this one - dimensional model can be used to simulate a ( pos ) in a busy avenue of a city .
lastly the new boundary conditions are presented for the two - dimensional diffusion equation for future applications in two dimensional domains .
as it was first noted in @xcite and expanded in @xcite , solving the diffusion equation with a discrete data sample @xmath0 as initial condition ( [ eq2 ] ) give an estimate of a continuous probability density function .
then by solving the diffusion equation @xcite , - = 0 a < x < b , t>0 , [ eq1 ] + u(x,0)=_i=1^n(x - b_i ) , x , b_i , [ eq2 ] with appropriate boundary conditions and then finding the best @xmath1 ( bandwidth ) for the initial data sample one obtains a continuous estimation of the experimental density . in this article we do not consider algorithms for bandwidth selection , we consider only the conservation of the mean .
for more information on the bandwidth selection see @xcite .
this one - dimensional toy problem is nevertheless of interest in applications for constructing ( pos ) .
for instance we can model an avenue as a one dimensional domain where predictions of the most dangerous places in a selected zone can be accomplished . in the following sections we present the non - conservation of the mean for the neumann boundary conditions for problem ( [ eq1 ] ) .
we also propose new boundary conditions .
for the derivations we assume that the functions are sufficiently smooth in order for the theorems of vector analysis to hold
. moreover the following derivations can be done for a more general diffusion equation with a variable diffusion coefficient @xmath2 .
if we consider the neumann or natural boundary conditions on the problem ( [ eq1 ] ) , we have @xmath3 as is widely known , the total mass is conserved over time , see section [ mass - conv ] , however the mean of the initial condition is , in general , not conserved .
indeed , we have @xmath4_{a}^{b}- \left[u(x , t)\right]_{a}^{b}\\ & = u(a , t ) - u(b , t).\end{aligned}\ ] ] where we used ( [ eq1 ] ) , ( [ eq3 ] ) and integration by parts .
hence the mean is generally not conserved , it depends on the values of @xmath5 at the boundary in a time @xmath1 .
we propose the following boundary conditions for ( [ eq1 ] ) , @xmath6 note that this boundary conditions are non - local , we need to evaluate in both boundary points at the same time . now we show that both the mean and the mass are conserved over time using this boundary conditions .
consider first the conservation of the total mass .
we have , @xmath7_{a}^{b } = \frac{\partial u(x , t)}{\partial x}\big|_{a}-\frac{\partial u(x , t)}{\partial x}\big|_{b}=0.\end{aligned}\ ] ] where we used ( [ eq1 ] ) , ( [ eq4 ] ) and integration by parts .
this shows that the total mass is conserved .
consider now the conservation of the mean .
we have , @xmath4_{a}^{b}- \left[u(x , t)\right]_{a}^{b}\\ & = ( b - a)\frac{\partial u(x , t)}{\partial x}\big|_{b } -u(b , t ) + u(a , t)\\ & = 0.\end{aligned}\ ] ] again ( [ eq1 ] ) , ( [ eq4 ] ) and integration by parts were used to obtain the desired result .
this shows that the boundary conditions ( [ eq4 ] ) for problem ( [ eq1 ] ) conserve both mean and mass .
now we proceed to make some numerical simulations using fem to show the consequences of the application of this boundary conditions in the process of estimation a probability density for a data sample ( [ eq2 ] ) .
now the problem ( [ eq1]),([eq4 ] ) is written in a weak formulation @xcite in order to apply the finite element method to the problem .
now for all @xmath8 we have , @xmath9 we solve this weak formulation using fem with low order elements in theinterval @xmath10=[0,10]$ ] , where the number of elements is @xmath11
. then problem ( [ eq5]),([eq2]),([eq4 ] ) yields the problem in the discretised space @xmath12 .
find @xmath13 , such thatfor all @xmath14 : _ a^b v(x)dx + _ a^b dx = ( v(b)-v(a))|_b , [ eq6 ] + u(x,0)=_i=1^n(x - b_i ) , x , b_i , [ eq7 ] + where we represent delta measures by the closest base element of the finite element approximation . note that ( [ eq7 ] ) contains a normalization factor , since now the elements integral are not one ( since they are not delta measures ) .
now we use the galerkin method of mean weighted residuals for the spatial part of the problem choosing low order elements @xmath15 .
this formulation can be found in @xcite . for our numerical studies
we solve the temporal part of the problem ( element coefficients ) using the implicit - euler galerkin discretization @xcite , thus the problem is reduced to solve a linear system iteratively for every timestep @xmath16 . in order to implement the previous formulation numerically ,
we use ` python ` to do all the calculations for the simulation .
the code is available publicly in @xcite .
there we start by generating a list of @xmath17 uniformly distributed points in theinterval [ 0,10 ] .
this points are located in the closest interval of the spatial fem partition @xmath18 .
the histogram of this points , figure [ histogram ] can be seen for instance as the number of times a certain criminal act was informed in a zone from the avenue .
see figure [ fig : fig1 ] .
.47 $ ] .,title="fig : " ] .47 $ ] .,title="fig : " ] if we represent this data as an initial condition we obtain the figure [ initial - condition ] . where we plotted alternatively each consecutive fem basis function red and black .
now we solve numerically the problem using the implicit - euler galerkin discretization for the problem , , and we evolve the solution until time @xmath19 using either neumann boundary conditions , see figure [ neumman - sol ] and mean conserving boundary conditions , see figure [ mean - sol ] .
the solution for the mean conserving boundary condition is positive for this numerical experiment , see figure [ mean - sol ] , this fact is currently being explored for future analytical studies .
as the figure [ fig : fig2 ] shows , the solutions are similar and therefore we can see that for this example the new boundary condition does not generate a noticeable change on the generation of the continuous density distribution .
nevertheless we present the plots of change of mass @xmath20 , figures [ mass - conv - neu ] , [ mass - conv - mean ] and change of mean @xmath21 , figures [ mean - conv - neu ] , [ mean - conv - mean ] for both neumann and mean conserving boundary conditions . .47 .,title="fig : " ] .47 .,title="fig : " ] .47 and @xmath22 for the density estimation with neumann boundary conditions for @xmath23$].,title="fig : " ] .47 and @xmath22 for the density estimation with neumann boundary conditions for @xmath23$].,title="fig : " ] .47 and @xmath22 for the density estimation with mean conserving boundary conditions for @xmath23$].,title="fig : " ] .47 and @xmath22 for the density estimation with mean conserving boundary conditions for @xmath23$].,title="fig : " ] figures [ mean - conv - neu ] and [ mean - conv - mean ] present the real difference in the evolution of the density .
we effectively see that the mean conserving boundary conditions conserve the mean in the density estimation process . on the other hand if we where to have an initial condition that is biased to one of the boundaries , the differences of the estimated densities by both boundary conditions would differ significantly .
however there is no evidence to think that this phenomena occurs in real avenues . for the numerical experiment presented here we can see that the mean for the neumann boundary conditions has changed about 0.4% in @xmath24 .
this change is small , in fact , for an avenue of 10 km , the change in mean would be about 40 m. we conclude that for this numerical experiment for the process of density estimation ( when the data has not change to much due to the smoothing process ) the neumann boundary condition provide a very fast ( since they are easy to implement ) and accurate way to estimate a continuous probability density
. nevertheless the mean of the sample is not preserved exactly , on the other hand , the mean conserving boundary condition , apart from being also easily implementable , is accurate and do preserve the mean of the sample .
we now present the problem for the diffusion equation @xcite in two dimensions , @xmath25 again we want the conservation of mass and mean in the time evolution of the density .
consider first the conservation of the total mass .
we have , @xmath26 where @xmath27 , and @xmath28 denotes the outward normal unit vector to @xmath29 . to deduce this relation we used ( [ eq9 ] ) , and the first green identity @xcite .
consider now the conservation of the mean .
we have , @xmath30 where @xmath31 , assuming cartesian unit vectors . again
( [ eq9 ] ) and the first green s identity were used to obtain the desired result .
then the conditions that we have to impose on @xmath32 in order to conserve mean and mass are : @xmath33 the advantage of two dimensional domains is that we are not restricted to impose only two conditions for the boundary(mean and mass conservation ) . for these domains we can in principle conserve additional higher moments of the density distribution that are meaningful for the particular problem .
applications on two dimensional domains are of special interest for the author since a two dimensional map of the city can generate really robust results in the field of perception of security(pos ) .
the proposed mean conserving boundary conditions were shown to effectively maintain the mean of the initial data sample over the continuous density estimation process .
this was also confirmed by the numerical simulation of the estimation process where we used a list of uniformly distributed points in the interval [ 0,10 ] as an initial condition .
the numerical experiments presented here show that even though neumann boundary conditions do not conserve the mean over time , they are accurate enough to maintain the mean in a very restricted interval before the over - smoothing of the density estimation process .
we showed the application and some of the consequences of both the idea of ( kde ) and the new boundary conditions to avenues in a city .
the consequences of implementing the diffusion equation with the proposed boundary conditions in companion of more special initial conditions and in 2d domains remains to be analyzed .
i would like to express my gratitude to juan galvis , whose guidance was essential for the completion of this manuscript . i also want to thank francisco a. gmez and zdravko botev , whose comments were really appreciated for the analysis of the results . 1d inhomogenious diffusion equation solution with fem
patarroyo k. available online . | we propose boundary conditions for the diffusion equation that maintain the initial mean and the total mass of a discrete data sample in the density estimation process .
a complete study of this framework with numerical experiments using the finite element method is presented for the one dimensional diffusion equation , some possible applications of this results are presented as well .
we also comment on a similar methodology for the two - dimensional diffusion equation for future applications in two - dimensional domains . |
recently there has been a measurement of the form factors in the semi - leptonic decay @xmath11 @xcite , where a fit to the data was done using the model of krner and kramer @xcite .
the model , constructed within the framework of hqet , treats both the charm quark and the strange quark as heavy and considers @xmath12 expansion keeping @xmath13 .
even though the model has only one input parameter , use of @xmath12 expansion is questionable and the omission of @xmath14 corrections is also not justified . for a heavy to light transition , for example of the type @xmath15 ,
the use of hqet in the limit @xmath16 allows one to express all the form factors in terms of only two form factors @xcite .
semileptonic decays of @xmath17 have been studied in this limit @xcite . in ref .
@xcite we constructed a model for the form - factors which takes into account @xmath18 corrections for heavy - to - light transitions of the @xmath19 type baryon ( light degrees in spin 0 state ) . in this work we use the form factors calculated in ref .
@xcite to study the decays @xmath20 and @xmath21 .
the paper is organized in the following way . in the next section we present the formulae for the calculation of asymmetries and decay rates for the semi - leptonic decays @xmath5 and @xmath6 while in the third section we present our results .
here we present the formalism for the semi - leptonic decays of @xmath5 and @xmath6 .
the hadronic part of the amplitude is the matrix elements of the weak quark current between baryonic states that is parametrized in terms of form factors .
we define the six vector and axial vector form factors through the following equations @xmath22 u{_b{_q } } ( p , s ) \nonumber \\
\left < b'(p ' , s ' ) \mid \bar{q } \
\gamma^\mu \gamma^5 \ q \mid b_q ( p , s ) \right > = \bar{u}_{b ' } ( p ' , s ' ) \left [ g_1 \gamma^\mu - i \frac{g_2}{m_{b_q}}\sigma^{\mu\nu } q_\nu + \frac{g_3}{m_{b_c } } q^\mu \right ] \gamma^5 u{_b{_q}}(p , s)\\end{aligned}\ ] ] where @xmath23 is the four momentum transfer , @xmath24 is the baryon with a heavy quark and @xmath25 is the light baryon .
we have shown in ref .
@xcite that we can write the form factors in the above equation in terms of two form factors @xmath26 and @xmath27 provided some assumptions are made about the @xmath14 corrections . for both @xmath0 and @xmath1 semi - leptonic decays we will work in the approximation that the lepton mass is negligible .
the decay @xmath28 proceeds via @xmath29 followed by @xmath30 .
following @xcite we define the helicity amplitudes which are given by @xmath31 where @xmath32 are the polarizations of the daughter baryon and the w - boson respectively . in terms of the form
factors the helicity amplitudes are given by @xmath33v(q^2 ) \nonumber\\ h^{a}_{1/2,0 } & = & a_{+}\left [ -(m_1-m_2)g_1+\frac{q^2}{m_{b_c}}g_2\right ] a(q^2 ) \nonumber\\ h^{v}_{1/2,1 } & = & \sqrt{2q_{-}}\left [ -f_1 - \frac{(m_1+m_2)}{m_{b_{c } } } f_2 \right ] v(q^2 ) \nonumber\\ h^{a}_{1/2,1 } & = & \sqrt{2q_{+}}\left [ g_1 - \frac{(m_1-m_2)}{m_{b_c}}g_2 \right ] a(q^2)\end{aligned}\ ] ] where @xmath34 where @xmath35 are the parent and daughter baryon masses and @xmath36 are the appropriate pole masses .
the decay @xmath28 is analysed by looking at the two sided cascade decay @xmath37 + w[\rightarrow l \nu_l]$ ]
. the normalized four - fold joint angular distribution for the decay @xmath38 + w[\rightarrow l \nu_l]$ ] is given , following the notation in ref .
@xcite , as @xmath39\\end{aligned}\ ] ] where @xmath40\nonumber\\ t_2 & = & \frac{3}{4}\sin^2\theta\left[|h_{1/2 \ , 0}|^2 ( 1+\alpha_{\lambda}\cos\theta_{\lambda } ) + |h_{-1/2 \ , 0}|^2 ( 1-\alpha_{\lambda}\cos\theta_{\lambda})\right]\nonumber\\ t_3 & = & \mp\frac{3}{2\sqrt{2}}\alpha_{\lambda}\cos\chi\sin\theta\sin\theta_{\lambda }
\left[(1\pm\cos\theta)re(h_{-1/2 \ , 0}h^*_{1/2 \ , 1 } ) + ( 1\mp\cos\theta)re(h_{-1/2 \ , 0}h^*_{-1/2 \ , -1})\right ] \\end{aligned}\ ] ] where @xmath41 , the upper and lower signs in the above equation hold for the @xmath42 and @xmath43 leptonic final states respectively and @xmath44 is the ckm element for the @xmath45 transition .
the polar angles are defined in ref .
@xcite . by integrating over two of the angles one can look at the following distributions @xmath46 the first distribution in the above equation gives the polar angle distribution for the cascade decay @xmath47 .
the second distribution is the polar angle distribution for the deacy @xmath48 , while the third distribution is the azimuthal angle distribution .
the asymmetries are given by the expressions below @xcite and will depend only on the ratio @xmath49 in our model for the form factors @xmath50 for unpolarized @xmath0 . for polarized
@xmath0 one has similar decay distributions and we refer the reader to ref .
@xcite for the relevant details .
we give here the expression for the asymmetries @xmath51 all the asymmetries have limiting values as @xmath52 .
all the polar asymmetries vanish at this limit while the azimuthal asymmetryies @xmath53 and @xmath54 . at the @xmath55 point
however the limiting value of the asymmetries @xmath56 , @xmath57 and @xmath58 depend on the dynamics of the semi - leptonic decay and we will comment on this issue further in the next section . the other asymmetries @xmath59 and @xmath60 tends to zero while @xmath61 tend to @xmath62 in the @xmath55 point .
these asymmetries were studied for different representative values of the ratio @xmath63 @xcite which is related to the ratio @xmath64 via @xmath65 .
the fit , performed by ref .
@xcite , to the semileptonic decay data on @xmath66 uses the korner kramer(kk ) model to extract @xmath67 and the asymmetry @xmath56 @xcite .
since , as already mentioned in the introduction , the kk model does not include @xmath14 corrections and the use of @xmath12 expansion is questionable , a more correct approach would be therefore to do a fit to the semi - leptonic data including @xmath14 corrections to extract the ratio @xmath68 and hence the asymmetries . the formula for the calculation of the absolute decay rate for the semi - leptonic decay of baryons
is given in ref .
in our analysis we have studied the semi - leptonic decays for @xmath69 and @xmath70 . in ref . @xcite
we found the @xmath14 expansion to be valid for @xmath71 and from the study of non - leptonic two body charmed baryon decays @xmath72 was obtained ( we assumed @xmath73 ) . in our analysis
we found that the @xmath18 corrections depend on the value of @xmath74 .
as far as the @xmath18 corrections to the form factors are concerned for the @xmath0 decay the @xmath14 corrections to the form factors are @xmath75 for @xmath76 . for @xmath1
decays the @xmath77 corrections are small and are less than @xmath78 for @xmath79 .
fig.1 shows @xmath80 versus @xmath81 for different @xmath82 values .
we note that for @xmath83 the asymmetry @xmath56 is not very sensitive to @xmath74 .
this feature is common to most of the calculated integrated asymmetries . in fig.2
we show @xmath84 versus @xmath74 . in the kk model or in general with @xmath85 , @xmath86 for all @xmath74 .
fig.2 shows the effect of including @xmath14 corrections . fig.3 and
fig.4 show the predictions for the integrated asymmetries for different @xmath74 for unpolarised and polarised @xmath0 and except for @xmath60 we see that the integrated asymmetries are insensitive to r in the range @xmath83 .
the @xmath14 corrections to the integrated asymmetries again depend on the value of r and can be as large as @xmath87 .
the decay rates depend on both @xmath74 and @xmath88 .
so instead of calculating the individual decay rates for @xmath5 and @xmath6 , to reduce the uncertainties from these sources we calculate the ratio of the decay rates which will depend only on @xmath74 as the factor @xmath88 cancels .
fig.5 shows this ratio as a function of @xmath74 with and without @xmath18 corrections with a monopole and dipole form for the form factors .
we see that the @xmath18 corrections are small for the dipole form factors .
the ratio above will receive corrections of the order @xmath89 and higher where @xmath90 is the c or b quark .
an estimate of these corrections will depend on the value of @xmath74 which can be extracted by performing a fit to the @xmath11 data .
the final result can be written as @xmath91|\frac{v_{ub}}{v_{cs}}|^{2}\\end{aligned}\ ] ] where @xmath92 is the ratio with @xmath93 and @xmath94 represents the corrections to the ratio due to @xmath95 corrections to the form factors .
since we do not know @xmath74 we can make an estimate of the ratio above by using the value of @xmath74 extracted from non - leptonic charmed baryon decays with dipole form factors @xcite i.e @xmath96 .
we write @xmath97 where @xmath98 is the derivative of @xmath99 at @xmath100 , and where @xmath101 may also include experimental uncertainties in extracting @xmath74 .
so an estimate of the ratio defined in eqn.(9 ) can be made through the following approximation @xmath102|\frac{v_{ub}}{v_{cs}}|^{2}\\end{aligned}\ ] ] for monopole and dipole form factors we then have @xmath103|\frac{v_{ub}}{v_{cs}}|^{2}\\end{aligned}\ ] ] and @xmath104|\frac{v_{ub}}{v_{cs}}|^{2}\\end{aligned}\ ] ] naively we can set @xmath105 and so we see that @xmath106 for the dipole form factors and @xmath107 is negligible for the monopole form factors .
measurement of the ratio @xmath108 can be used to extract @xmath109 .
clearly the largest uncertainty in @xmath108 comes from the assumed @xmath81 dependence of the form factors and to reliably extract @xmath109 using @xmath108 the @xmath81 dependence of the form factors must be measured . our model for the form factors at @xmath110 can be tested by studying the decay distribution in @xmath111 . with enough statistical power such
a study could determine the @xmath81 dependence of the form factors .
so if our model for the form factors at @xmath110 is tested successfully by doing a fit to the @xmath111 decay distribution , and if @xmath74 and information on the @xmath81 dependence of form factors extracted , we can use the ratio in eqn.(9 ) to determine @xmath109 with a small error .
it is interesting to compare this method of extracting @xmath109 and hence @xmath112 with other methods of extracting @xmath112 .
one method is to use semi - leptonic inclusive @xmath113 decays and study the decay distribution near the end point of the spectrum in the narrow window @xmath114 gev @xcite .
however in this region the @xmath115 transition is subject to large qcd radiative corrections as well as large non - perturbative corrections which are entangled with each other and theoretical calculations are model dependent .
another way to extract @xmath112 is through the study of exclusive @xmath113 decays @xcite and this approach is similar to our approach with the difference that we are working with baryons instead of mesons .
the calculations of form factors in exclusive decays of @xmath113 mesons are done in specific models , like the quark model , or with chiral lagrangian with heavy meson . in ref .
@xcite it is suggested that @xmath116 may be extracted from the ratio of the lepton distribution in @xmath117 and @xmath118 with an theoretical uncertainty of 10 - 20 @xmath119 .
the relevant form factors are calculated up to order @xmath18 in the soft pion limit with the additional assumption that only intermediate states degenerate with the ground state meson contribute and so the calculations of the form factors are model dependent . in ref .
@xcite the ratio of the lepton distribution in @xmath120 and @xmath121 at maximum @xmath81 is used for an extraction of @xmath116 .
the @xmath18 corrections are calculated in quark model and the theoretical uncertainty in the measurement of @xmath116 is small . however at maximum @xmath81 the decay rates vanish , actual measurements are performed away from maximum @xmath81 where additional form factors contribute adding to the theoretical uncertainty in the measurement of @xmath116 .
another approach to measuring @xmath116 , which is the mesonic counterpart of our method , is given in ref .
( @xcite ) where experimental data on @xmath122 semi - leptonic decay are used to calculate the @xmath123 semileptonic decay .
the @xmath81 dependence of the form factors is assumed to be of the monopole type , no @xmath18 corrections are considered and the extraction of @xmath116 is found to depend on the value of @xmath124 and @xmath125 whose values are not known very accurately . since most approaches to extracting @xmath116 from @xmath113 meson decays are model dependent and can involve significant theoretical uncertainties , extraction of @xmath116 using heavy baryon decays is interesting as it is possible that theoretical uncertainties in heavy baryon decays might be small .
the predictions for the decay rates for @xmath126 and @xmath127 are subject to significant @xmath18 corrections and may not be reliable .
nevertheless we give the predictions for the decay rates with @xmath128 and @xmath129@xcite @xmath130 the particle data group gives @xmath131 .
measurement of @xmath132 is likely to be made at lep in the future .
the numbers for the individual decay rates are calculated using @xmath133 @xcite , the pole masses @xmath134 gev and @xmath135 gev@xcite and the quark masses @xmath136 gev and @xmath137 gev@xcite in conclusion we have studied the semi - leptonic decays of charmed and bottom baryons involving transition of a heavy to light quark based on a model for form factors that includes @xmath18 corrections .
we also suggest a method for determining @xmath138 . * the asymmetry @xmath80 versus @xmath81 for some representative values of the form factor ratio @xmath74 . * the asymmetry @xmath84 versus @xmath74 .
the @xmath139 prediction for @xmath84 is @xmath62 for all @xmath74 .
this figure therefore shows the effect of @xmath14 corrections .
* the average asymmetries as functions of @xmath74 when the initial heavy baryon is unpolarised . * the average asymmetries as functions of @xmath74 when the initial heavy baryon is polarised . *
the ratio of the decay rates of @xmath140 versus @xmath74 .
@xmath92 and @xmath108 represents the ratio without and with @xmath18 corrections .
monopole and dipole form for the form factors are used for the @xmath81 dependence of the form factors . | we calculate the semi - leptonic decays of @xmath0 and @xmath1 baryons which involve transitions of a heavy quark to a light quark . in a previous work we used data on the cabibbo
favoured non - leptonic decays of charmed baryons @xmath2 and @xmath3 to obtain information on the form factors in the @xmath4 transition .
this form factor information is used to study the semi - leptonic decay @xmath5 . using su(3 ) symmetry and hqet it
is shown how a knowledge of form factors in @xmath5 can be applied to @xmath6 decay and used to measure @xmath7 . plus
1pt
@=12 -0.5 in 0.0 in 0.0 in 8.5 in 6.5 in -.375 in = 0 uh-511 - 825 - 95 + april 1995 * semi - leptonic decays of @xmath8 and @xmath9 baryons involving heavy to light transitions and the determination of @xmath10 + * + .5 in .1 in .25 in |
the term sitnikov problem " appeared originally in the context of studies of oscillatory solutions in the restricted three body problem .
these studies were initiated by sitnikov @xcite ; they stimulated the application of symbolic dynamics in celestial mechanics @xcite .
we recall that sitnikov considered the case when two primaries have equal masses and rotate around their barycenter @xmath0 , while the infinitesimal third body moves along a straight line normal to the plane defined by the motion of the primaries and passing through @xmath0 ( usually the motions of the third body perpendicularly to the plane of the primaries are called vertical " ; below we will follow this tradition ) .
sitnikov concentrated his attention on phenomena taking place when the primaries move in elliptic orbits .
more bibliography on `` elliptic '' sitnikov problem can be found , for example , in @xcite .
if the primaries move in circular orbits , then the vertical motions are integrable .
the corresponding quadratures were presented at the beginning of the xx century by pavanini @xcite and macmillan @xcite - much before the start of sitnikov s studies .
relatively simple formulae for the vertical motions , written in terms of jacobi elliptic functions , can be found in @xcite . since the integrability of third body motion is something extraordinary within the restricted three body problem , many specialists investigated the properties of vertical motions in the case of primaries moving on circular orbit .
very often the term
circular sitnikov problem " is applied to describe this field of research .
taking into account its popularity , we will use it too . nevertheless ,
some authors prefer terms like pavanini problem " or macmillan problem " , which are probably more correct from the historical point of view . depending on the initial values , three types of vertical motions
are possible in the circular sitnikov problem : the hyperbolic escape ( i.e. , the escape of the third body with non - zero velocity at infinity ) , the parabolic escape ( i.e. , the escape of the third body with zero velocity as the limit at infinity ) and , finally , the periodic motion , in which third body goes away up to a distance @xmath1 from the plane defined by primaries and then returns to it .
the first stability analysis of the periodic vertical motions in the circular sitnikov problem was undertaken by perdios and markellos @xcite , but they drew the wrong conclusion that vertical motions are always unstable ( perdios and markellos only analyzed the vertical motions with the initial conditions such that @xmath2 ; as it was established lately it is not enough to put any hypothesis about the stability properties of the motions with larger values of @xmath1 ) .
the mistake was pointed out in @xcite , where the alternation of stability and instability of vertical motions were found numerically in the case of continuous monotone variation of their amplitude @xmath1 .
lately the existence of such an alternation was confirmed by the results of computations presented in @xcite and @xcite .
taking into account their numerical results , the authors of @xcite proposed the hypothesis that the lengths of stability and instability intervals have finite limits as @xmath1 increases .
this hypothesis was formulated on the basis of computations in which @xmath1 did not exceed the value @xmath3 .
our numerical investigations demonstrate that the rapidly decreasing difference of the stability intervals at @xmath4 is a manifestation of a local maximum of their lengths ; if @xmath1 is increased further , then the lengths of the stability and instability intervals tend to zero .
there is one more important property of vertical motions , which can be observed only for @xmath5 : the intervals of complex saddle " instability , when all eigenvalues of the monodromy matrix are complex and do not lie on the unit circle . according to our computations
first such an interval begins at @xmath6 , its length is @xmath7 .
it means the erroneous of the statement in @xcite ( p. 113 ) , that the stability indexes of the vertical motions in circular sitnikov problem are always real ( this statement was based on the results of numerical studies in which the amplitude of the motion @xmath1 was smaller @xmath8 ; as one can see it was not enough for such a general conclusion ) . to conclude our short review on previous investigations of vertical motions stability in circular sitnikov problem
we would like to mention the generalization of this problem for systems of four and more bodies @xcite .
numerical results presented in @xcite demonstrate that in the generalized problem the absence of stability / instability alternation in the family of vertical motions persists .
the aim of our paper is to study the stability property of the periodic vertical motions at large values of the oscillation amplitude " @xmath1 , both numerically and analytically .
a special attention will be given to the phenomenon of infinite alternation of stability and instability in this family .
in fact , the infinite alternation of stability and instability in the one - parameter family of periodic solutions is rather typical for hamiltonian systems , although the general investigation was carried out only for 2dof systems @xcite .
different examples can be found in @xcite .
nevertheless , an important difference exists between the circular sitnikov problem and other systems in which the alternation of stability and instability was established earlier . in the circular sitnikov problem the discussed family of periodic solutions possesses as a limit unbounded aperiodic motions - parabolic escapes , while in previously considered systems the corresponding families and their aperiodic limits were bounded @xcite . due to this difference
, the alternation of stability and instability in the circular sitnikov problem can not be studied in the same way as it was done in @xcite ( one could try to compactify the phase space by means of certain changes of variables , but we were unable to find any reduction to what was investigated already ) .
this paper is organized as follows . in sect .
2 some general properties of the vertical motions are discussed . in sect .
3 we present the linearized motion equations used in our studies of the vertical motions stability .
the results of the numerical investigation of the stability are reported in sect .
4 . in sect .
5 we prepare for the analytical investigation : the approximate expression for the monodromy matrix is derived here . using this expression , some important stability properties of vertical periodic solutions with large amplitudes @xmath1
are established in sect . 6
( in particular , the asymptotic formulae for the intervals of stability and instability are obtained ) . in sect .
7 we discuss briefly the vertical motions in the generalized circular sitnikov problem with four and more bodies .
some concluding remarks can be found in sect .
we consider the restricted , circular , three - body problem with primaries having equal masses , say @xmath9 .
let @xmath10 be a synodic ( rotating ) reference frame with the origin at the barycenter @xmath0 ; the masses @xmath11 and @xmath12 are arranged on the axis @xmath13 , while the axis @xmath14 is directed along the rotation axis of the system .
the coordinates of the infinitesimal third body in the synodic reference frame will be used as generalized variables : @xmath15 below we assume that all variables are dimensionless .
the equations of motion of the third body can be written in hamiltonian form with hamiltonian function @xcite @xmath16 here @xmath17 and @xmath18 denote the distance between the third body and the corresponding primary , while @xmath19 are the momenta conjugated to @xmath20 .
the phase space @xmath21 possesses a manifold @xmath22 which is invariant with respect to the phase flow .
the phase trajectories lying on @xmath23 correspond to vertical motions with the third body staing always on the axis @xmath14 .
consequently , the vertical motions are governed by a reduced 1dof system with hamiltonian @xmath24 the phase portrait of the system with the hamiltonian ( 1 ) is shown in fig .
it is remarkable that the separatrices ( the borders between trajectories representing periodic motions and hyperbolic escapes ) intersect at infinity .
-1.0 cm 2.cm .
thick lines denote the separatrices ( @xmath25 ) . , title="fig:",width=453 ] -10 cm the periodic solutions associated to the system with hamiltonian @xmath26 form a one - parameter family @xmath27 where as parameter @xmath1 one can choose the amplitude " of the periodic motion ( i.e. , @xmath28 ) or the absolute value of @xmath29 at the passage trough the barycenter @xmath0 or the value @xmath30 of the hamiltonian @xmath26 in this periodic motion .
the first variant is the most convenient for us , therefore @xmath1 in ( 2 ) will denote the amplitude " of the periodic motion . for definiteness
we assume that @xmath31 there exist explicit expressions for the solutions ( 2 ) in terms of jacobi elliptic functions @xcite .
since they are not used in the forthcoming analysis , we do not rewrite them here , except for the formula about the period of vertical motion : @xmath32 .
\eqno(3)\ ] ] here @xmath33 is the complete elliptic integral of the second kind , @xmath34 is the heuman lambda function , while the value of the modulus @xmath35 is given by the formula @xmath36 where @xmath37 for motions with large amplitudes ( @xmath38 ) the following approximate formula can be used in place of ( 3 ) : @xmath39 as it was mentioned before , the separatrices @xmath40 , representing the parabolic escapes , can be interpreted as a formal limit for periodic motions at @xmath41 .
the parabolic escapes obey the approximate law @xmath42 formulae ( 4 ) and ( 5 ) are easily obtained if one suitably relates the properties of vertical motions with the properties of rectilinear motions of a particle in a newtonian field .
our efforts are concentrated on the analysis of the vertical motions stability with respect to horizontal " perturbations , due to which the third body leaves the axis @xmath14 . under the linear approximation ,
the behavior of the variables @xmath43 in the perturbed motion is described by the linear hamiltonian system of equations with periodic coefficients : @xmath44 here @xmath45 @xmath46 the symbol @xmath47 is used to denote the identity matrix of the @xmath35-th order .
the function @xmath48 depends periodically on time with a period @xmath49 , where @xmath50 denotes the period of the particular vertical motion whose stability is investigated .
as it is known , the restricted circular three - body problem admits several types of symmetry ( for example , they are used for the numerical construction of 3d periodic solutions @xcite ) .
the consequence of these symmetries is the following property of the variational equations ( 6 ) : if @xmath51 is a solution of ( 6 ) , then these equations admit the solution @xmath52 where @xmath53 it the @xmath54-diagonal matrix , @xmath55 . according to floquet theory , in order to draw a conclusion about the stability or instability of the solutions of ( 6 ) , one should analyze the spectral properties of the monodromy matrix @xmath56 , where @xmath57 denotes the normal fundamental matrix corresponding to the system ( 6 ) ( i.e. , the matrix solution of ( 6 ) with the initial condition @xmath58 ) .
the normal fundamental matrix corresponding to the linear hamiltonian system ( 6 ) is a symplectic one , i.e. @xmath59 it is also worthwhile to mention some other properties of this matrix : @xmath60 @xmath61 the first two equalities in ( 8) are elementary , while the last one is a consequence of the symmetry property ( 7 ) . using the relation ( 8) one easily obtains @xmath62 the characteristic equation of the system ( 6 ) @xmath63 is reciprocal and it can be written as @xmath64 where @xmath65 the quantities @xmath66 in the last formula are the elements of the monodromy matrix @xmath67 .
it is also possible to rewrite the characteristic equation ( 9 ) as the product @xmath68 the coefficients @xmath69 in ( 10 ) are the roots ( real or complex ) of the quadratic equation : @xmath70 often enough the quantities @xmath69 are called the stability indices @xcite . the periodic vertical motion is stable whenever @xmath71 ( i.e. , when @xmath69 are real and their absolute values are smaller than 1 ) . in the case
@xmath72,\quad \{b_1,b_2\}\lefteqn{\subset}{\;\,/ } \ ; i\ ] ] an additional investigation is needed to draw a conclusion about stability or instability . in all other cases
the instability takes place .
we recall that in @xcite the alternation of the stability and instability in the family of periodic vertical motions ( 2 ) was discovered . later on ,
more accurate results were published in @xcite : the length of the first 35 intervals of stability and of the first 34 intervals of instability was calculated . in @xcite
also an attempt was undertaken to establish certain regularity in the variation of these quantities : the existence of non - zero limits for the intervals lengths was proposed . in fig.2 and fig.3
we present the results of some calculations , when the first 700 intervals of stability and instability are considered .
the graph in fig .
2 shows that for the first 30 intervals of stability the length of the intervals increases and only afterwards the decrease of the length takes place .
the hypothesis formulated in @xcite was based on the wrong interpretation of the small variation of the intervals length in vicinity of the maximum . in fig .
3 the length of the instability intervals decreases monotonically and it does not follow the empirical law derived in @xcite ( according to this law , the length of the instability intervals has the limit @xmath73 ; evidently , it is not so ) .
our results allow us to propose the following approximate formulae to characterize the behavior of the stability and instability intervals length in fig . 2 and fig .
3 : @xmath74 more precisely , these formulae are valid for the periodic vertical motions with amplitude @xmath1 smaller the critical value @xmath75 the reason of such a restriction and the situation for @xmath76 will be revealed a little bit later .
-0.0 cm 2.cm , title="fig:",width=453 ] -9.0 cm -0.0 cm 2.2 cm ( the first interval is not presented : if it was shown in the same scale with all subsequent intervals , it would have been difficult to understand the behaviour of the graph for large @xmath77),title="fig:",width=453 ] -9.cm it is also useful to discuss here in what way the length of the stability intervals @xmath78 and the length of the instability intervals @xmath79 depend on the amplitude of the vertical oscillations . under the same restriction @xmath80
we obtain from our numerical investigations @xmath81 _ remark_. if one needs a rigorous definition about the meaning of the quantity @xmath1 in the last formulae , one could interpret it as the boundary value between two successive intervals of stability and instability . in fig .
4 the behavior of the coefficients @xmath69 appearing in the characteristic equations ( 10 ) is shown . fig .
4a , 4b and 4c allow us to compare the properties of these coefficients , when the parameter @xmath1 varies in different intervals .
all graphs demonstrate the approximate periodicity , their period with respect to the parameter
@xmath1 corresponds to an increase of period of vertical oscillations @xmath82 of about @xmath83 .
it is important to point out the small gaps in the fig .
4c : for the corresponding value of the parameter @xmath1 ( i.e. , when @xmath1 belongs to the intervals where the graphs are not defined ) the stability indices @xmath69 have complex values and the so - called complex saddle " instability of the vertical motion takes place .
the enlarged fragments of the graphs in the vicinity of the gaps are given in fig .
-1.0 cm 2.cm and @xmath84 appearing in the characteristic equation ( 10),title="fig:",width=453 ] -10.0 cm 2.cm and @xmath84 appearing in the characteristic equation ( 10),title="fig:",width=453 ] -10.0 cm 2.cm and @xmath84 appearing in the characteristic equation ( 10),title="fig:",width=453 ] -10.0 cm -5.0
cm 2.cm -9.0 cm 1.6 cm -10.0 cm as it follows from our calculations , the first interval of `` complex saddle '' instability begins at @xmath85 .
since such a value of vertical motion amplitude is large enough , it provides us with an explanation why this kind of instability of vertical motions in the circular sitnikov problem was not recognized in previous studies where relatively small values of @xmath1 were considered . increasing further the parameter @xmath1(i.e .
, for @xmath86 ) , we observe a stability / instability alternation of more complicated type : wide " interval of instability - narrow " interval of stability - narrow " interval of complex saddle `` instability - ' ' wide " interval of stability - wide " interval of instability - ... an analog of the formulae ( 11 ) can be constructed in the case @xmath76 , but we prefer to present in sect .
6 several asymptotics written in a more convenient way .
finally it worth while to mention that the runge - kutta - fehlberg method of 7 - 8 order with variable step was used to integrate numerically the variation equations ( 6 ) .
the accuracy of the integration procedure ( the local tolerance ) was taken @xmath87 .
since the period of vertical oscillations increases proportionally @xmath88 the variation equations should be integrated over relatively large time intervals : if we take for example @xmath89 then the value of half - period @xmath90 . to check the influence of the round - off errors some computations
were done both with double and quadruple precision arithmetic .
in this section an approximate expression for the monodromy matrix @xmath67 is derived .
it will be used to discuss the phenomena described in sec
. 4 ( the alternation of stability and instability , the decrease of stability and instability intervals by increasing the parameter @xmath1 , etc ) .
we assume that the amplitude @xmath1 of the periodic solution ( 2 ) is so large , that we can define an auxiliary quantity @xmath91 such that @xmath92 to start with we write down the monodromy matrix @xmath56 as the product of three fundamental matrices : @xmath93 where @xmath94 and @xmath95 are the instants at which the third body is at distance @xmath91 from the barycenter @xmath0 in the periodic vertical motion ( 2 ) ( at @xmath96 the third body moves away from barycenter , at @xmath97 it approaches the barycenter ) . _ approximate expression for the matrix @xmath98_. if the condition ( 12 ) is satisfied the phase point @xmath99 moves on the manifold @xmath23 in close vicinity of the separatrix @xmath100 at @xmath101 $ ] . within such time interval
the difference between @xmath102 and @xmath103 is small enough . neglecting this difference
, we replace @xmath102 in ( 6 ) by @xmath103 ; as a consequence , the normal matrix solution @xmath104 of the obtained system provides us the suitable approximation for @xmath105 at @xmath101 $ ] .
the behavior of @xmath104 at @xmath106 is described by the remarkable asymptotic formula : @xmath107 here @xmath108 @xmath109
@xmath110 the derivation of the formula ( 14 ) is based on some simple ideas .
let us take @xmath111 and write down @xmath104 as the product @xmath112 where @xmath113 is the moment of time when the third body is at distance @xmath114 from the barycenter @xmath0 in the motion corresponding to the parabolic escape @xmath115 .
as next step , we modify the equations ( 6 ) to find the approximate expression for @xmath116 at @xmath117 . since at @xmath117
the third body is far enough from the primaries @xmath11 and @xmath12 , it looks natural to replace @xmath118 by @xmath103 in the right parts of the first two equations in system ( 6 ) and to neglect the small term @xmath119 .
the system ( 6 ) takes the form @xmath120 with @xmath121 now it is worthwhile to make the following remark .
let us consider the rectilinear parabolic escape of the material point in the field of an attracting center . under a proper choice of units ,
the distance between the attracting center and the point varies as @xmath122 if the asymptotics ( 5 ) is used for @xmath103 in the equations ( 16 ) , then these equations coincide with the motion equations of the above mentioned material point , linearized in the vicinity of the solution ( 17 ) and written in the reference frame uniformly rotating around the line of the escape . taking this into account
, we implement in ( 16 ) the change of variables @xmath123 where @xmath124 this change of variables can be interpreted as the transfer from the synodic reference frame @xmath125 to the sidereal ( fixed ) reference frame @xmath126 ( @xmath127 ) . as a result the linearized equations of motion split into two independent subsystems @xmath128 it is not difficult to find partial solutions to the system ( 18 ) @xmath129 and @xmath130 here and below the dots are used for derivatives with respect to time .
four independent partial solutions allow us to write down the normal fundamental matrix in terms of the variables @xmath131 : @xmath132 coming back to the initial variables , we get @xmath133 substituting ( 19 ) into ( 15 ) we obtain the expression for the normal fundamental matrix @xmath104 as the product of three matrices with only one of them depending on time : @xmath134 here @xmath135 the formula ( 20 ) can be used to compute the elements of the matrix @xmath104 at @xmath136 .
asymptotically their values should not depend on the choice of @xmath114 .
it means that the following limit exists : @xmath137 substituting @xmath138 instead of @xmath139 into ( 20 ) we arrive to the formula ( 14 ) .
the fundamental matrix @xmath104 was introduced in such a way that it provides the vertical motions satisfying ( 12 ) with a `` universal '' ( i.e. , independent on @xmath1 ) approximation @xmath140 at @xmath141 $ ] . using the relation ( 14 )
, we finally obtain @xmath142 _ approximate expression for the matrix @xmath143_. since at @xmath144 $ ] the third body is far enough from the primaries , we neglect again the difference between their gravity field and the gravity field of the attracting center placed at the barycenter @xmath0 . to obtain the expression for @xmath143 within such an approximation
we need to integrate the system @xmath145 where @xmath146 describes the motion in the newtonian field along the segment @xmath147 $ ] on the axis @xmath14 .
it is supposed that the maximum distance @xmath1 from the body to the attracting center is achieved at @xmath148 . in this case
@xmath149 at @xmath144 $ ] .
of course the motion along a segment corresponds to the singular impact orbit @xcite , but it is used here to approximate the regular vertical motion on the time interval were the singularities are absent . the change of variables @xmath150 where @xmath151 allows us to rewrite the equations ( 22 ) in the more simple form : @xmath152 it is easy to check that the system ( 24 ) admits the following partial solutions : @xmath153 and @xmath154 to compute @xmath155 in ( 25 ) and ( 26 ) the energy integral can be used . in the case of the motion along the segment
@xmath147 $ ] in the newtonian field , this integral takes the form @xmath156 and consequently @xmath157 taking into account ( 25),(26 ) we write down the fundamental matrix for the system ( 24 ) as @xmath158 and then ( taking into account the relation ( 23 ) ) we write the matrix for the system ( 22 ) @xmath159 using the expression ( 27 ) we find @xmath160 where we denote by @xmath161 @xmath162 the final step is based on the last formula in ( 8) , namely @xmath163 @xmath164 here @xmath165 for completeness we should add the following formula : @xmath166 _ approximate expression for the matrix @xmath167_. using again the relations ( 8) we get @xmath168 then the substitution of the previously obtained expression for @xmath98 ( the formula ( 21 ) ) into the right part of ( 29 ) provides us with the desired approximate formula for @xmath167 . _
finalizing the construction of the approximate formula for the monodromy matrix @xmath67_. the substitution of the approximate expressions for @xmath167 , @xmath143 , @xmath98 into ( 13 ) yields @xmath169 simplifying ( 30 ) the extraordinary simple result can be obtained @xmath170 this formula allow us to investigate analytically the stability properties of the periodic vertical motions in the case @xmath38 .
as it follows from ( 31 ) the coefficients of the monodromy matrix @xmath67 and , respectively , the coefficients of the characteristic equation ( 9 ) are @xmath171-periodic functions of the semiperiod of the vertical motion @xmath172 . to illustrate this we present in fig.6 the graphs of the coefficients @xmath69 ( only the real values ) as @xmath172 varies in the interval @xmath173 $ ] , where @xmath174 is a large enough integer number . using the formula ( 3 ) , which defines the dependence of @xmath172 on the amplitude @xmath1 , it is not difficult to prove that in terms of @xmath1 the lengths of the stability and instability intervals decrease proportionally to @xmath175 as @xmath176 .
1.cm computed on the base of the approximate formula for the monodromy matrix @xmath67 .
only real values are shown.,title="fig:",width=529 ] -13 cm taking into account the approximate expression for the monodromy matrix @xmath67 , we describe in more details the repeating pattern of stable and unstable intervals mentioned at the end of sec .
this pattern consists of four intervals appearing in the following order as @xmath1 increases : _ wide " interval of instability .
_ both coefficients @xmath69 are real , but one of them has absolute value greater than 1 ( `` saddle - center '' instability ) .
the asymptotic length of the interval in terms of the amplitude of the motion is about @xmath177 , while the variation of the semiperiod @xmath172 equals to about @xmath178 . _ narrow " interval of stability . _
the coefficients @xmath69 are real and belong to the interval @xmath179 .
the approximate length is @xmath180 ; the variation of the semiperiod equals to about @xmath181 . _ narrow " interval of instability . _ the coefficients @xmath69 are complex ( `` complex saddle '' instability ) .
the approximate length is @xmath182 ; the variation of the semiperiod equals to about @xmath183 .
_ wide " interval of stability . _ the coefficients @xmath69 are real and belong to the interval @xmath179 again .
the approximate length is @xmath184 ; the variation of the semiperiod equals to about @xmath185 . to conclude
, we recall that before the first appearance of the interval of complex saddle `` instability at @xmath85 , a more simple pattern with only two intervals was observed .
the ' ' transient " asymptotics for the length of the stability intervals in the case @xmath186 can be obtained by adding of the lengths of the narrow " instability interval and both stability intervals in the final pattern .
it yields @xmath187 which is in good agreement with the corresponding numerical result presented in sec .
the investigation of the generalized circular sitnikov problem with four and more bodies revealed that in contrast to the case of the three body problem there is no alternation of stability / instability in the family of vertical motions @xcite .
for simplicity we limit our consideration to the case of the restricted four body problem .
it is assumed that three primaries of equal mass rotate around the barycenter @xmath0 in circular orbit with the radius @xmath188 @xcite . under the linear approximation
the stability analysis of the fourth body periodic vertical motion @xmath189 is reduced to the study of the spectral properties of the monodromy matrix associated to the system of linear differential equations with periodic coefficients @xmath190 where @xmath191 @xmath192 it is remarkable that equations ( 32 ) possess a circular symmetry : for any real @xmath193 they are invariant with respect to transformations of the form @xmath194 while the non - linearized equations of motion of the fourth body in the synodic reference frame admit only the rotational symmetry of the 3rd order . the possibility for the linearized equations of motion to have a larger group of symmetries in comparison to the original non - linear system
was pointed out by v.i .
arnold @xcite(sec .
in particular , in the case of the initial system rotational symmetry of @xmath77-th order ( @xmath195 ) the linearized equations always have circular symmetry .
this is the reason why the stability analysis of the vertical motions , based on the linearized equations , yields similar results for the sitnikov problem with four and more bodies and for the particle dynamics in the gravity field of the circular ring @xcite ( numerically it was shown in @xcite ) .
it is convenient to rewrite the equations of motion ( 32 ) in a sidereal ( fixed ) reference frame by means of the transformation of variables @xmath196 where @xmath197 after that the equations of motion split into two identical independent subsystems : @xmath198 let @xmath199 denote the normal fundamental matrix for the system ( 33 ) in the case when @xmath200 is replaced by @xmath201 , which corresponds to the parabolic escape . using the same technique as in sect .
5 we obtain the asymptotic formula @xmath202 where @xmath203 applying the main ideas of sect . 5 , we establish the following property of the monodromy matrix @xmath204 associated to ( 33 ) : at @xmath205 the matrix @xmath206 , where the constant matrix @xmath207 .
the eigenvalues of the matrix @xmath208 are the asymptotic limits for multiplicators ( of multiplicity @xmath209 ) of the system ( 33 ) : @xmath210 finally , it is not difficult to derive the asymptotic formulae the for multiplicators of the original system ( 32 ) : @xmath211 on the complex plane @xmath212 are placed in the small vicinity of the circles with radii @xmath213 and @xmath214 .
consequently in the circular sitnikov problem with four bodies the periodic vertical motions with large amplitudes are always unstable .
finally we would like to note the another opportunity to introduce the generalized circular sitnikov problem with @xmath77 bodies using the appropriate straight line solution of the problem of @xmath215 bodies @xcite .
if in such a solution @xmath215 primaries are arranged symmetrically with respect to the barycenter then the infinitesimal @xmath77th body can move periodically along an axis around which the rotation of the primaries takes place ( naturally , the proposed generalization is possible for odd @xmath77 only ) .
likely this family of periodic motions exhibits the alternation of stability and instability .
the combination of numerical and analytical approaches provided us with the opportunity to correct , clarify and extend some previously known results related to the circular sitnikov problem ( mainly about the stability of vertical motions ) .
for the first time under the scope of this problem the possibility of the `` complex saddle '' instability was revealed within the family of vertical motions . for our theoretical constructions it was essential that the phase trajectories corresponding to the solution under consideration have lengthy parts in the vicinity of the peculiar separatrices of the problem - the parabolic escapes to infinity .
often enough it is possible to introduce a suitable auxiliary mapping in the vicinity of the separatrix in order to study the local properties of the phase flow .
it would be very interesting to develop similar for the circular sitnikov problem .. 1 in * acknowledgements .
* the author would like to express his gratitude to a.i .
neishtadt and a.b .
batkhin for useful discussions during the accomplishment of this work .
also the author thanks a.celletti for reading the manuscript and suggesting many improvements .
neishtadt , a.i . ,
sidorenko , v.v . ,
investigation of the stability of long - periodic planar motions of a satellite in a circular orbit .
_ kosmicheskie issledovaniya _ , 2000 , * 38 * , 307 - 321 ( in russian .
english trans . : _ cosmic research _ , 2000 , * 38 * , 289 - 303 ) . | this paper is devoted to the special case of the restricted circular three - body problem , when the two primaries are of equal mass , while the third body of negligible mass performs oscillations along a straight line perpendicular to the plane of the primaries ( so called periodic vertical motions ) .
the main goal of the paper is to study the stability of these periodic motions in the linear approximation .
a special attention is given to the alternation of stability and instability within the family of periodic vertical motions , whenever their amplitude is varied in a continuous monotone manner . |
electron diffraction patterns exhibiting octagonal , decagonal , dodecagonal , and icosahedral point symmetry are found in various alloys .
the most well - known pattern is the icosahedral phase in @xmath3-@xmath4 alloys , which is observed when these materials are cooled at a rapid rate such that their constituent atoms do not have adequate time to form a crystal lattice .
these structures are referred to as quasicrystals @xcite . in principle , quasicrystals are characterized as atomic structures that present long - range quasiperiodic translational and long - range orientational order .
they can exhibit rotational symmetries otherwise forbidden to crystals . in the last decade , quasicrystals have attracted significant attention , mostly because of their stronger magnetic properties and enhanced elasticity at higher temperatures , compared with the traditional crystals . a most intriguing research topic about quasicrystals is to determine whether its intrinsic complicated structure can result in a change of the universality class compared with its counterpart periodic structure . to this end , potts model @xcite offers a simple and feasible way to study quasicrystals from this perspective , as it contains both first- and second - order phase transitions .
however , given the lack of periodicity of these quasiperiodic lattices , only numerical approaches can be performed .
previous monte carlo studies on the ferromagnetic potts model for quasiperiodic lattices @xcite have revealed that both the systems belong to the same universality class , despite the critical temperature of the quasiperiodic lattices being higher than that of the square lattices .
however , given the great variety of existing quasiperiodic lattices , this query has not been solved completely . consequently , this necessitates extensive computational research for accurately estimating the static critical exponents in these lattices . to the best of our knowledge ,
studies concerning the @xmath2 potts model on quasiperiodic lattices have been rarely reported in the literature .
the present study investigates the critical behavior of the ferromagnetic @xmath5-potts model on quasiperiodic decagonal lattices ( qdl ) to accurately estimate the infinite qdl critical temperature and critical exponents for each case .
an interesting example of a natural structure which presents a decagonal symmetry is the @xmath6 quasicrystal found in the khatyrka meteorite @xcite .
the quasiperiodic lattices analyzed in this study were generated using the strip projection method @xcite with spins placed in the vertices of the rhombi that constitute the qdl ( fig .
[ fig:1 ] ) .
periodic boundary conditions were applied on these lattices to avoid the boundary effects caused by the finite size .
this paper is organized as follows .
section [ sec : proc ] briefly describes the strip projection method adopted for generating the qdl and periodic boundary conditions used in the simulations . details of the potts model and monte carlo simulation approach are described in section [ sec : mms ] . in section [ sec : f ] , a succinct description of the finite - size scaling ( fss ) relations used in the study is presented . in section [ sec : r ] , we present the results for @xmath1 and @xmath2 potts model and compare them with previous results on quasi - periodic lattices . in section [ sec : c ] , we conclude by summarizing the results and providing recommendations for further research . a periodic approximation of the qdl generates by the strip projection method .
the lattice is shown inside a square projection window .
the periodic boundary conditions are imposed at lattice sites closer to the projection window . ]
ccccc & & & & + & @xmath7 & @xmath8 & @xmath9 & @xmath10 + + & & & & + @xmath11&@xmath12&@xmath13&@xmath14&@xmath15 + @xmath16&@xmath17&@xmath18&@xmath19&@xmath20
the strip projection method is a powerful technique for constructing periodic and non - periodic lattices .
the methodology can be summarized as follows .
first , starting from a regular lattice @xmath21 whose unit cell , @xmath22 , is spanned by the @xmath23 vectors @xmath24 , we can resolve @xmath25 into two mutually orthogonal subspaces , namely , @xmath26 and @xmath27 , of dimensions @xmath28 and @xmath29 , respectively , i.e. , @xmath30 .
second , we define a `` strip '' @xmath31 as a set of all the points whose positions are obtained by adding any vector in @xmath26 to any vector in @xmath22 , i.e. , @xmath32 .
the required lattice , @xmath33 , is the projection in @xmath26 of all the points in @xmath34 that are included in the strip , i.e. , @xmath35 .
the requirement that any point @xmath36 lies in the strip is equivalent to the condition that the projection of @xmath37 in @xmath27 lies within the projection of @xmath22 in @xmath27 .
this equivalence can be mathematically expressed as @xmath38 where @xmath39 and @xmath40 , accordingly , the lattice can be defined as follows : @xmath41 one way to describe the projection of the points @xmath42 given by @xmath43 ( where the @xmath44 s are integers ) onto @xmath26 and @xmath27 is to choose an orthogonal basis @xmath45 in @xmath26 and an orthogonal basis @xmath46 in @xmath27 .
together they form a new basis @xmath47 of @xmath25 . assuming @xmath48 , the relationship between the two basis can be given by a rigid rotational operation . by defining a rotation matrix @xmath49 , it is possible to determine the projection matrices using the following equations @xmath50 where @xmath51 .
the rotation matrix @xmath49 can be split into an @xmath52 submatrix @xmath53 and @xmath54 submatrix @xmath55 : @xmath56 to generate the decagonal quasiperiodic lattice , the points in the finite region of a 5d hypercubic lattice ( @xmath57 ) are projected onto an 2d subspace ( @xmath26 ) only if these points are projected inside a rhombic icosahedron , which in this case is the `` strip '' .
the resulting quasiperiodic lattice is obtained through the standard rotation matrix : @xmath58 where @xmath59 and @xmath60 .
the decagonal quasiperiodic lattice consists of two types of building blocks , usually represented by a fat rhombus with an acute angle of @xmath61 and a thin rhombus with an acute angle of @xmath62 , arranged according to specific matching rules .
however , the quasiperiodic lattices are not suitable for monte carlo simulation with periodic boundary conditions .
a more suitable approach is to construct a periodic approximation of these lattices .
this can be realized by simply replacing the golden number @xmath63 in the sub - matrix @xmath55 of eq .
( [ eq:5 ] ) by a rational number @xmath64 , where @xmath65 and @xmath66 are successive terms in the fibonacci sequence @xmath67 fig .
[ fig:1 ] shows the periodic approximation of the qdl inside a square projection window .
the periodic boundary conditions are imposed at lattice sites closer to the square projection window . for finite lattices , the number of nearest neighbors at a given site range from @xmath68 to @xmath69 with a mean coordination number equal to @xmath70 .
the mean coordination number is expected to be lower than @xmath71 , given the existence of a small fraction ( @xmath72 1.0% ) of sites on the boundary with a coordination number lower than @xmath68 in the finite lattices is analyzed in this study .
to study the critical behavior in the qdl , we updated our lattices using the wolff algorithm @xcite . for a fixed temperature , we define a monte carlo step ( mcs ) per spin by accumulating the flip times of all the spins and then dividing them by the total spin number .
the hamiltonian of the _ q_-states ferromagnetic potts model ( @xmath73 ) can be written as @xmath74 where @xmath75 is the kronecker delta function , and the sum runs over all nearest neighbors of @xmath76 .
we also define the order parameter @xmath77 as @xmath78 [ ! r]0.49 [ !
r]0.49 ccccc & & & & + & @xmath7 & @xmath8 & @xmath9 & @xmath10 + + & & & & + @xmath79&@xmath80&@xmath81&@xmath82&@xmath83 + @xmath84&@xmath85&@xmath86&@xmath87&@xmath88 where @xmath89 is the maximum number of spins in the same state and @xmath90 is the total number of spins . once the critical region is established , we apply the single histogram method @xcite along with fss analysis to obtain accurate estimates of the critical temperature and critical exponents .
system sizes up to @xmath91 are used in these simulations with @xmath92 mcs per spin performed at a single temperature @xmath93 , where @xmath94 configurations are discarded for thermalization . for each system size
considered , we calculated the average of @xmath95 independent realizations to obtain reliable estimates of the statistical errors .
the static thermodynamics quantities such as specific heat , magnetic susceptibility , logarithmic derivatives of the order parameter , and binder s fourth - order cumulants @xcite are then calculated inside the critical region .
depending on the analysis of the location of the maximum values of these quantities and their magnitudes , one can estimate the infinite qdl critical temperature and critical exponents , respectively .
for @xmath1 , we perform simulations at the temperatures @xmath96 and @xmath97 .
the probability distribution obtained at each @xmath98 was reweighted from @xmath99 to @xmath100 for @xmath101 and from @xmath99 to @xmath102 for @xmath103 and @xmath104 . for @xmath2 , we perform simulations at the temperatures @xmath105 and @xmath106 . here again , the probability distributions were reweighted from @xmath107 to @xmath108 for the first five smaller lattices and from @xmath109 to @xmath110 for the remaining lattices .
the specific heat can be calculated from the fluctuations of the @xmath111 measurements [ !
r]0.49 [ !
r]0.49 @xmath112 similarly , from the fluctuations of @xmath77 , we can calculate the magnetic susceptibility @xmath113 and the fourth - order magnetization cumulant @xmath114 we can also calculate the logarithmic derivative of @xmath23-power of @xmath77 , i.e , @xmath115
according to the finite - size scaling theory @xcite , the free energy of a system of linear dimensional @xmath116 is described by the scaling ansatz @xmath117 where @xmath118 ( @xmath119 is the infinite qdl critical temperature ) and @xmath120 is the magnetic field . the leading critical exponents
@xmath121 , @xmath122 , @xmath123 and @xmath7 define the universality class of the system . considering zero - field regime , the derivatives of eq .
( [ eq:16 ] ) yield important scaling equations , i.e. , @xmath124 where @xmath125 , @xmath126 and @xmath127 are scaling functions , and @xmath128 is the temperature scaling variable .
in addition , the critical temperature scales as @xmath129 where @xmath130 is a constant and @xmath131 is the effective transition temperature for the qdl of linear size @xmath116 .
this effective temperature can be obtained by the location of the peaks of the above quantities : @xmath132 , @xmath133 , @xmath134 and @xmath135 . due to the presence in two - dimensional @xmath2 potts model of a marginal operator @xcite , which is absent in any other two - dimensional potts model , the leading power - law scaling behavior of this model
is modified by multiplicative logarithms .
so eqs . ( [ eq:16]-[eq:20 ] )
must be modified to allow for these logarithmic corrections .
the free energy scaling relation is suitably modified @xcite by @xmath136 where @xmath137 and @xmath138 . moreover , considering zero - field regime , the derivatives of eq.([eq:21 ] ) yield suitable scaling equations for 4-state potts model : @xmath139 where @xmath125 , @xmath126 and @xmath127 are scaling functions , and @xmath128 is the temperature scaling variable .
correspondingly , the critical temperature scales as @xmath140 in the above equations , the leading critical exponents @xcite for the @xmath2 potts model on 2d periodic lattices are given by @xmath141 and the logarithmic - correction exponents @xcite are given by @xmath142
taking the slope of the log - log plot of the maximum values of the quantities @xmath143 , @xmath144 and @xmath133 versus @xmath116 , three estimates are obtained for @xmath145 .
[ fig:4 ] shows the log - log plot of these quantities for data simulated at @xmath11 .
we obtained @xmath146 for @xmath133 , @xmath147 for @xmath148 , and @xmath149 for @xmath144 . by combining these results ,
we get @xmath150 .
similar analysis has been performed for @xmath151 , which yielded @xmath152 for @xmath133 , @xmath153 for @xmath148 and @xmath154 for @xmath144 .
the effective range of @xmath7 is found to be @xmath17 by combining the above results .
these estimates are in good agreement with the exact result based on the 2d periodic lattice ( with @xmath155 ) and in reasonable agreement with the estimates that have previously been obtained for other quasiperiodic systems @xcite . after obtaining an estimate for @xmath7 ,
the infinite qdl critical temperature is computed by plotting the size dependence of the location of the peaks of @xmath148 and @xmath144 .
[ fig:5 ] shows the finite - size scaling of the effective transition temperatures at @xmath96 .
we obtained @xmath156 for @xmath148 and @xmath144 . with the simulations performed at the temperature @xmath96 , and using the corresponding @xmath7 estimated at this temperature
, we obtained @xmath157 for @xmath148 and @xmath144 .
these values of @xmath119 are higher than the exact value on the 2d periodic lattices @xcite , given by @xmath158 .
using eqs .
( [ eq:17]-[eq:19 ] ) for the size dependence of the maximum values of @xmath77 , @xmath134 and @xmath135 , we can estimate @xmath159 , @xmath160 and @xmath161 , respectively . fig .
[ fig:6 ] shows the log - log plot of @xmath77 ( measured at the temperature with maximum value of @xmath162 ) versus the linear size of the system @xmath116 .
the slopes of the linear fit to the data obtained by simulating at @xmath16 and @xmath11 are @xmath163 and @xmath164 , respectively .
similarly , fig .
[ fig:7 ] shows the log - log plot of the maximum value of @xmath134 versus the linear size of the system .
particularly , in this plot , we have inserted correction - to - scaling terms @xcite to improve the fit quality of the data by scaling as @xmath165 where the proper correlation amplitudes @xmath166 and @xmath167 and the nonuniversal correction - to - scaling exponent @xmath168 are chosen in order to minimize the @xmath169 of the fit .
the slopes of the linear fit to the data obtained by simulating at @xmath16 and @xmath11 are @xmath170 and @xmath171 , respectively . similarly , fig .
( [ fig:8 ] ) shows a log - log plot of the maximum values of @xmath135 versus @xmath116 .
the estimated values for @xmath161 at @xmath16 and @xmath11 are @xmath172 and @xmath173 , respectively . in these figures ,
the error bars are purely statistical and are estimated according to 100 different trial runs for each data point .
the estimates of the ratios of the critical exponents and the average value for @xmath7 at each simulated temperature @xmath93 are summarized in table [ tab:1a ] . from table
[ tab:1a ] , by multiplying the values of the ratios of the exponents at each simulated temperature @xmath93 by its respective value of @xmath7 , we obtain @xmath174 , @xmath175 and @xmath176 at @xmath16 , and @xmath177 , @xmath178 and @xmath179 at @xmath11 . on the 2d periodic lattice ,
the exact values for the @xmath1 potts model of the critical exponents are @xmath180 , @xmath181 , @xmath182 and @xmath183 .
the final estimates of the infinite qdl critical temperature and the critical exponents for the @xmath1 potts model on qdl are summarized and compared with the exact values on the 2d periodic lattices in table [ tab:1 ] .
ccccccc & & & & & & + & & @xmath184 & @xmath7 & @xmath121 & @xmath122 & @xmath123 + + & & & & & & + & @xmath11 & @xmath185 & @xmath186 & @xmath187 & @xmath188 & @xmath189 + & @xmath16 & @xmath190 & @xmath191 & @xmath192 & @xmath193 & @xmath194 + & 2d periodic lattice & @xmath195 & @xmath196 & @xmath197 & @xmath198 & @xmath199 + & & & & & & + & @xmath79 & @xmath200 & @xmath201 & @xmath202 & @xmath203 & @xmath204 + & @xmath84 & @xmath205 & @xmath206 & @xmath207 & @xmath208 & @xmath209 + & 2d periodic lattice & @xmath210 & @xmath211 & @xmath212 & @xmath213 & @xmath214 + fig .
[ fig:11 ] shows the log - log plot of the size dependence of the maximum values of the quantities @xmath143 , @xmath144 and @xmath133 . taking the slope of the quantities shown in fig .
[ fig:11 ] , we obtain @xmath215 for @xmath133 , @xmath216 for @xmath148 and @xmath217 for @xmath144 . by combining these results
, we find @xmath218 .
similarly at @xmath219 , we obtain @xmath220 for @xmath133 , @xmath221 for @xmath148 and @xmath222 for @xmath144 . taking an average of the above results , we get @xmath223 .
as one can see , these average values for @xmath7 are in reasonable agreement with the exact result on 2d periodic lattices , and especially for @xmath148 and @xmath144 , we have a very good convergence to this exact value .
[ fig:12 ] shows the finite - size scaling of the effective transition temperatures . from eq .
( [ eq:25 ] ) and locating the peaks of the quantities @xmath148 and @xmath144 , we find @xmath224 .
similar analysis has been done for the temperature @xmath106 with @xmath7 estimated at this temperature , which yielded @xmath225 . following the case @xmath1 , these values
are also higher than the exact value on 2d periodic lattices , given by @xmath226 .
using eqs .
( [ eq:22]-[eq:24 ] ) and taking the exact exponents from eqs .
[ eq:26 ] and [ eq:27 ] in the logarithmic - correction terms , we can estimate @xmath159 , @xmath160 and @xmath161 . fig .
[ fig:13 ] shows the log - log plot of @xmath77 ( measured at the temperature with maximum value of @xmath162 ) versus linear size of the system .
the slopes of the linear fit to the data obtained by simulating at @xmath84 and @xmath79 are @xmath227 and @xmath228 , respectively .
[ fig:14 ] presents a log - log plot of the maximum values of @xmath134 versus @xmath116 .
the slopes of the linear fit to the data obtained by simulating at @xmath84 and @xmath79 were @xmath229 and @xmath230 , respectively .
similarly , fig .
[ fig:15 ] shows the log - log plot of peaks of @xmath135 versus @xmath116 .
our estimates are @xmath231 at @xmath84 and @xmath232 at @xmath79 .
the estimates of the ratios of the critical exponents and the average value for @xmath7 at each simulated temperature @xmath93 are summarized in table [ tab:2a ] . from table
[ tab:2a ] , by multiplying the values of the ratios of exponents at each simulated temperature @xmath93 by its respective value of @xmath7 , we obtain @xmath233 , @xmath234 and @xmath235 at @xmath84 , and @xmath236 , @xmath237 and @xmath238 at @xmath79 .
the exact values of the critical exponents for the @xmath2 potts model on 2d periodic lattice are @xmath239 , @xmath240 , @xmath241 and @xmath242 .
the final estimates of the infinite qdl critical temperature and critical exponents for the @xmath1 potts model on qdl are summarized and compared with the exact values on the 2d periodic lattices in table [ tab:1 ] .
we performed monte carlo simulations of the @xmath0-potts model on qdl to estimate the infinite critical temperature and the leading critical exponents for both @xmath1 and @xmath2 states .
our analysis reveals that for both @xmath1 and @xmath2 states , the infinite lattice critical temperature is higher than that of the square lattice , which can be attributed to the different geometric structure between the two models . for the @xmath1 potts model , the leading critical exponents @xmath7 , @xmath122 and @xmath123 are , within the error precision , in good agreement with the corresponding values for the 2d periodic lattices , whereas for the @xmath2 potts model , all the critical exponents are found to be very close to the exact values on the 2d periodic lattices .
this provides strong evidence to support the claim that @xmath1 and @xmath2 potts model on quasiperiodic lattices belong to the same universality class as those on 2d periodic lattices .
future work will involve numerical studies on 3d quasiperiodic lattices so that the icosahedral phase found in alloys such as @xmath3-@xmath243 , @xmath3-@xmath4 and @xmath3-@xmath244 can also be better investigated . | in this study , we performed monte carlo simulations of the @xmath0-potts model on quasiperiodic decagonal lattices ( qdl ) to assess the critical behavior of these systems . using the single histogram technique in conjunction with the finite - size scaling analysis
, we estimate the infinite lattice critical temperatures and the leading critical exponents for @xmath1 and @xmath2 states .
our estimates for the critical exponents on qdl are in good agreement with the exact values on 2d periodic lattices , supporting the claim that both the @xmath1 and @xmath2 potts model on quasiperiodic lattices belong to the same universality class as those on 2d periodic lattices . |
99 bohigas o 1991 _ random matrix theories and chaotic dynamics _ , in giannoni m j , voros a and zinn - justin j ( eds ) , _ proceedings of the 1989 les houches summer school on chaos and quantum physics _ , pages 88 - 199 .
( amsterdam : elsevier ) . | the spectral statistics of the circular billiard with a point - scatterer is investigated . in the semiclassical limit
, the spectrum is demonstrated to be composed of two uncorrelated level sequences .
the first corresponds to states for which the scatterer is located in the classically forbidden region and its energy levels are not affected by the scatterer in the semiclassical limit while the second sequence contains the levels which are affected by the point - scatterer .
the nearest neighbor spacing distribution which results from the superposition of these sequences is calculated analytically within some approximation and good agreement with the distribution that was computed numerically is found .
classical dynamics may be illuminating for the understanding of the corresponding quantum mechanical systems .
one of the most studied aspects of the relation between classical and quantum mechanics is the connection between the spectral statistics of the quantum system and the dynamical properties of its classical counterpart .
classically integrable systems typically exhibit poisson - like spectral statistics @xcite while classically chaotic systems exhibit spectral statistics of random matrix ensembles @xcite .
the spectral statistics of integrable and chaotic systems are universal , that is , they do not depend on specific details of the system but rather on the type of motion and its symmetries .
there are systems which are intermediate between integrable and chaotic ones and their spectral properties are not known to be universal .
such systems are of experimental relevance .
the spectral statistics of mixed systems , for which the phase space is composed of both integrable and chaotic regions , were studied by berry and robnik @xcite .
the spectrum can be viewed as a superposition of uncorrelated level sequences , corresponding to the various regions , which are either chaotic or integrable .
the nearest neighbor spacing distribution ( nnsd ) of such a superposition of sequences was calculated in @xcite .
the resulting statistics are , in some sense , intermediate between those of integrable and chaotic systems .
other types of systems with intermediate statistics include pseudointegrable systems and integrable systems with flux lines or point - scatterers @xcite .
the spectral statistics of billiards with flux lines @xcite and of some pseudointegrable billiards @xcite were recently studied .
a possible route towards an understanding of the spectral statistics of these systems is based on their classical periodic orbits .
it is possible to compute the correlation function of the energy levels from these orbits by using trace formulae @xcite . for the billiards with flux lines or for pseudointegrable billiards ,
the orbits include not only the periodic orbits but also diffracting orbits which are built from segments that start and end at some singularity of the system .
while the contributions from periodic orbits were easily calculated , those from diffracting orbits turn out to be much more involved .
the spectral statistics of these systems , that can be obtained numerically , appear to be intermediate between those of integrable and chaotic systems .
in particular , the nnsd show level repulsion at small spacings and an exponential falloff at large spacings . since the contributions of diffracting orbits to the spectral statistics of pseudointegrable systems and of billiards with flux lines is far from being understood , it is of interest to study simpler systems which exhibit intermediate statistics . a class of such systems is given by integrable systems with a point - scatterers .
a point - scatterer is the self adjoint extension of a `` @xmath0-function potential '' in two or three dimensions @xcite .
the spectral statistics of an integrable system with such a point - scatterer was first studied by eba @xcite .
it is a rectangular billiard with the perturbation at its center .
this `` eba billiard '' also exhibits intermediate statistics which differs from that of psedointegrable systems .
integrable systems with point - scatterers are much easier to study analytically compared to integrable systems with flux lines or to pseudointegrable systems .
the contributions of diffracting orbits to the correlation function of the energy levels were recently calculated for the rectangular billiard with a point - scatterer @xcite .
exact results for the nnsd were also obtained @xcite .
one of the intriguing features of the spectral statistics of the rectangular billiard with a point - scatterer ( and dirichlet boundary conditions ) is its dependence on the location of the scatterer . if the coordinates of the scatterer ( divided by the sides of the rectangle ) are rational numbers @xmath1 ( where @xmath2 ) then the spectral statistics depend in a non trivial way on @xmath3 @xcite .
in contrast , for typical locations , the spectral statistics seem to be location independent .
the cause for this dependence on location is that many wavefunctions vanish at rational values of the coordinates . at these locations
there are many degeneracies in the lengths of the diffracting orbits ( including repetitions ) .
since such dependence on location is atypical , it is of interest to study the dependence of the spectral statistics on the location of the perturbation in other systems .
for instance , it may be possible that for typical systems the location of the point scatterer affects the spectral statistics only smoothly ( and does not depend on the rationality of the coordinates of the scatterer ) .
the system that is studied in this work is the circle billiard perturbed by a point - scatterer and the dependence of the spectral statistics on its location is studied .
the ( two dimensional ) circle billiard with radius @xmath4 is described by the schrdinger equation @xmath5 with dirichlet boundary conditions @xmath6 .
( the units where @xmath7 are used in most of this work . )
the hamiltonian with the point - scatterer is the self - adjoint extension of a hamiltonian where one point , say @xmath8 , is removed from its domain .
it can be considered as the self - adjoint extension of a hamiltonian with a @xmath0-function potential at @xmath8 . given the eigenvalues ( and eigenfunctions ) of the unperturbed system @xmath9 ( and @xmath10 ) ,
namely the system in absence of the @xmath0-scatterer , the eigenvalues of the system with the point - scatterer are given by the roots of @xcite @xmath11 where @xmath12 and @xmath13 are two parameters .
for a more complete discussion regarding this equation , the roles of the parameters as well as the method of its numerical solution see , for example , @xcite .
this equation turns out to be very convenient for numerical solution since every root is located between two eigenvalues of the unperturbed system .
the ( unnormalized ) eigenfunctions of the circle billiard are @xmath14 where @xmath15 are bessel functions of the first kind and the angular momentum @xmath16 is any nonnegative integer .
the energy levels @xmath17 are determined by the boundary condition @xmath18 .
it is obvious that all the energy levels with @xmath19 are doubly degenerate . as a result
there is a linear combination of the two degenerate wavefunctions which vanishes at @xmath8 and an orthogonal linear combination which does not vanish at @xmath8 .
the perturbation breaks this degeneracy . the linear combination which vanishes is also an eigenfunction of the hamiltonian _ with _ the point - scatterer and
thus @xmath20 is an eigenvalue of the perturbed problem .
therefore , half of the spectrum is unchanged by the perturbation . to avoid this trivial part of the spectrum we choose to work with the non vanishing
linear combinations ( which are eigenfunctions of the unperturbed hamiltonian ) and only the half of the spectrum which is affected by the perturbation will be considered in this work . for convenience ,
the location of the perturbation is chosen at @xmath21 and therefore the eigenfunctions of the unperturbed hamiltonian which do not vanish there are @xmath22 the spectrum is determined by substituting these eigenfunctions and the corresponding energies @xmath17 in equation ( [ solvept ] ) .
we are interested in the dependence of the spectral statistics on the location of the scatterer .
this dependence can be easily understood in terms of the properties of the wavefunctions of the unperturbed system .
the quantum numbers @xmath23 correspond to a state with an angular momentum @xmath24 and an energy @xmath25 . in the semiclassical limit where @xmath26 , but @xmath27 and @xmath28 are kept fixed ,
the values of the wavefunctions are small in the classically forbidden region @xmath29 .
therefore , one can divide the states into two groups .
the first consists of states for which the point - scatterer is located in the classically forbidden region .
as will be demonstrated , the eigenvalues of these states change only slightly due to the perturbation ( and do not change at all in the semiclassical limit ) .
the second group includes the states for which the perturbation is located in the classically allowed region .
these states will be ( strongly ) affected by the perturbation .
this separation of the spectrum into a superposition of two sequences is the cause for the dependence of the spectral statistics on the location of the scatterer .
this separation into ( semiclassically ) affected and unaffected states is justified in the following . to demonstrate that some of the eigenvalues are almost unchanged by the perturbation one should solve equation ( [ solvept ] ) and show that for ( exponentially ) small eigenfunctions the corresponding eigenvalues are almost unaffected . for simplicity , instead of equation ( [ solvept ] ) it is sufficient to consider the finite sum @xmath30 where @xmath31 is assumed to be slowly varying function of @xmath32 .
we denote @xmath33 . to further simplify the argument
let us assume that @xmath34 and @xmath35 are of order unity while for @xmath36 the wavefunctions on the scatterer @xmath37 are all small .
the solutions of ( [ finitept ] ) are to a good approximation given by the solutions of @xmath38 for @xmath39 and by @xmath40 for @xmath36 .
this is true since substituting a solution of the form @xmath41 in ( [ finitept ] ) leads to @xmath42 expanding with respect to @xmath43 and then solving to the leading order in @xmath44 results in @xmath45 when @xmath46 , as is the case in the semiclassical limit , @xmath43 also approaches @xmath47 . for the other eigenenergies one can substitute @xmath48 ( with @xmath36 ) and find that in the leading order @xmath49 which also vanish when @xmath46 .
note that we have just found @xmath50 ( approximate ) solutions which are all the solutions between @xmath51 to @xmath52 .
it is not hard to generalize this calculation for more wavefunctions which are of order unity . in the semiclassical limit
, the resulting spectrum always consist of such affected and unaffected components . in this limit
the values of the wavefunctions at the scatterer are exponentially small if it is located in the classically forbidden region and the corresponding eigenvalues can be treated as unchanged by the perturbation for any semiclassical consideration . consider a circle billiard with the point - scatterer at @xmath53 .
the spectral statistics of its energy levels in an energy window of width @xmath54 around @xmath55 are studied in what follows .
assume that @xmath55 is very large compared to @xmath54 , i.e. that all the levels in the window have similar energies which are high enough to be considered semiclassical . a natural question to ask is how many of these levels are affected by the point - scatterer and how many are unaffected by it .
as was argued , the semiclassically unaffected levels are those for which the classical turning point , @xmath56 , satisfies @xmath57 .
equivalently , for a given energy @xmath55 , for a state to be unaffected , the angular momentum @xmath28 should satisfy @xmath58 ( for the estimate of @xmath59 , the values of the energies are approximated by @xmath55 ) .
the fraction of such states was calculated in @xcite where it was used to determine how many levels are poorly approximated in the wkb method .
this fraction is @xmath60.\ ] ] it is clear that when @xmath61 then @xmath62 , as expected , while for @xmath63 , @xmath64 .
as one approaches the semiclassical limit these states are less and less affected .
the predictions of equation ( [ xunaff ] ) can be checked numerically .
the number of unaffected levels was calculated for two energy windows as a function of the location of the perturbation .
the results are presented in figure [ unfig ] . , width=453,height=377 ] the radius of the billiard was chosen so that the mean level spacing is @xmath65 .
therefore both energy windows contain @xmath66 levels .
a level was counted as unaffected if its difference from an energy level of the unperturbed system was less than @xmath67 of the mean level spacing .
this criterion is somewhat arbitrary , since in the semiclassical limit the difference can be taken to be arbitrarily small .
figure [ unfig ] indicates that equation ( [ xunaff ] ) correctly describes the number of unaffected levels .
there is a slight deviation which is smaller for the levels from the higher energy window .
this deviation is caused by the fact that for any finite energy the wavefunctions do not vanish at the turning point @xmath56 but rather exhibit an airy - like structure ( in the radial direction ) near the turning point .
this means that for states with @xmath57 , for which @xmath56 is close to @xmath68 , @xmath69 might not be small at a finite ( but large ) energy .
the deviations are expected to vanish in the semiclassical limit as indicated by figure [ unfig ] .
the spectrum of the circle billiard with a point - scatterer can therefore be viewed as composed of two uncorrelated components .
one is unaffected by the point - scatterer and its relative fraction is @xmath70 while the other is affected and its relative fraction is @xmath71 .
the nnsd of a spectrum which is composed of several uncorrelated level sequences was computed by berry and robnik @xcite and is applied to the circle billiard with the point - scatterer in what follows .
the unaffected spectrum consists of many levels with different angular momentum quantum numbers and thus its statistics are poissonian @xcite . since the density of levels in this sequence is @xmath70 , and the radius was chosen so that the total level density is unity , its nnsd is @xmath72 the second level sequence contains the levels which are influenced by the point - scatterer and their density is @xmath71 .
the exact form of its nnsd is unknown and an exact computation of this nnsd is complicated and beyond of the scope of this letter . instead , following experience with other systems @xcite
, we will _ assume _ that the nnsd can be _ approximated _ by a semi - poisson distribution , that is , @xmath73 this distribution exhibits level repulsion at small spacings and exponentially small probability to find large spacings . in these works
it was found numerically to describe the distribution of spacings reasonably well but there is no analytical justification for its use .
note that even if the semi - poisson distribution is an approximation for the nnsd it may not approximate other spectral measures well . for instance , the form factor , which is the fourier transform of the energy - energy correlation function , satisfies @xmath74 for the semi - poisson distribution @xcite while for a billiard with point - scatterer one expects to find @xmath75 @xcite . in particular , for the rectangular billiard with a point - scatterer , the nnsd was computed analytically under some assumptions in @xcite and was found
_ not _ to be given by the semi - poisson or the poisson distributions .
the nth neighbor spacing distributions were also calculated there and found to be those of the poisson distribution at large spacings .
following @xcite the nnsd of the circle billiard with the point - scatterer , obtained by superposing the two sequences , is given by @xmath76 e^{-(2-x_{un})s}.\ ] ] when the perturbation is at the center , @xmath62 , and @xmath77 approaches the poisson distribution .
alternatively , when the perturbation is near the boundary , and @xmath64 , the nnsd approaches the semi - poisson distribution .
note that the distribution ( [ totalps ] ) does not exhibit complete level repulsion since its value at @xmath78 , @xmath79 , does not vanish and it is a manifestation of the existence of an infinite class of states that are unaffected by the perturbation . the nnsd of ( [ totalps ] ) is compared to numerical results in figure [ pofs ] . , ( c ) @xmath80 and ( d ) @xmath81 .
[ pofs],width=453,height=377 ] the nnsd was computed using the levels @xmath82-@xmath83 for three locations of the point - scatterer as well as for the circle billiard without the perturbation .
it is clear that the agreement is very good .
the main features of the distribution are captured by the simple argument leading to ( [ totalps ] ) .
there are slight deviations from the predictions of equation ( [ totalps ] ) , mainly at large @xmath84 .
these can be attributed to the fact that the nnsd of the affected spectrum differs from the semi - poisson distribution .
the results presented in figures [ unfig ] and [ pofs ] suggest that the spectrum of the circle with a point - scatterer consists of a superposition of two uncorrelated level sequences .
the relative densities of these sequences are determined by the way the classical tori of the integrable system are projected into coordinate space .
states for which the perturbation is in the classically allowed region are affected while states for which the perturbation is in the classically forbidden region are nearly unaffected .
we expect this behavior to be typical of systems where the scatterer affects only a fraction of the tori of the otherwise classically integrable systems .
this differs from the rectangular billiard where all tori are affected by the scatterer .
another important difference compared to the rectangle billiard results of the different nature of the wavefunctions . for the rectangular billiard
there are infinite classes of wavefunctions that have common zeros at rational points . consequently ,
if the scatterer is placed at such a location the wavefunctions are not affected , and the distribution depends strongly on the rationality of the location of the scatterer . for the circle billiard studied in the present work , on the other hand , there is one class of eigenfunctions that vanish on the scatterer since they are antisymmetric in @xmath85 .
these are not considered in the present work
. the symmetric eigenfunctions always satisfy @xmath86 . to obtain many functions , symmetric in @xmath85 , that vanish at the same location
is equivalent to finding many bessel functions which satisfy @xmath87 and @xmath88 for integer @xmath16 , @xmath89 and for @xmath90
. finding an infinite number of such solutions , for the same @xmath91 ( corresponding to the same location of the perturbation ) , is unlikely .
however , since any bessel function of large argument is asymptotically given by a cosine one can find states with close zeros , that is where @xmath92 and @xmath93 are zeros of @xmath15 , while @xmath94 is a zero of @xmath95 and @xmath96 is close to a zero of @xmath95 . in this case when the scatterer is at a zero of one of these states , the square of the wave function of the other state is much smaller there than its average value .
many such close zeros should exist to affect the spectral statistics .
this question is beyond the scope of the present letter and is left for future research .
our numerical results are not sensitive enough to resolve this issue .
the numerical results are used here just to verify that the mean dependence on the location of the scatterer is given by equation ( [ totalps ] ) .
we believe that the behavior of the circle billiard rather than that of the rectangular billiard is typical of integrable systems perturbed by a localized potential . in summary ,
the spectral statistics of the circle billiard , perturbed by a point - scatterer are intermediate between those of the poisson distribution , characteristic of integrable systems , and of the semi - poisson distribution .
the spectrum was shown to be composed of two uncorrelated components .
the first contains energy levels which are nearly unaffected by the perturbation , since the point - scatterer is in a classically forbidden region where the wavefunctions are exponentially small .
the relative fraction of such states was computed analytically and found to depend smoothly on the location of the point - scatterer .
the second contribution is from states which are affected by the perturbation .
the exact statistics of this level sequence are complicated but can be approximated by the semi - poisson statistics .
the nearest neighbor spacing distribution results of combination of the two and is a manifestation of the berry - robnik statistics .
other integrable systems should also exhibit qualitatively similar behavior when a localized perturbation is added to them .
this research was supported in part by the us - israel binational science foundation ( bsf ) and by the minerva center of nonlinear physics of complex systems . |
astronomers combine telescopic observations of cosmic populations in the effort to understand astrophysical conditions and processes throughout the universe .
telescopes at all wavebands of light @xmath0 and recently telescopes for neutrinos and gravitational waves @xmath0 are pointed at a myriad targets to characterize properties of planets , stars , the milky way galaxy , other galaxies , material between the stars and galaxies , and the universe as a whole . in an increasing proportion of studies ,
a dataset of considerable size is collected .
this might be zero - dimensional photometry , one - dimensional spectra or time series , two - dimensional images , three - dimensional hyperspectral or video images .
the targets may be a single cosmic target , a small sample of targets sharing common properties , or a large heterogeneous sample of targets .
this last class emerges from sensitive wide - field astronomical surveys that have growing importance at all wavebands of astronomy .
once the telescope observations are complete , the astronomer faces the task of data analysis . according to r. a. fisher ( 1922 ) , the brilliant founder of much of 20th century statistics , this is the task of statistics : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ in order to arrive at a distinct formulation of statistical problems , it is necessary to define the task which the statistician sets himself : briefly , and in its more concrete form , the object of statistical methods is the reduction of data .
a quantity of data , which usually by its mere bulk is incapable of entering the mind , is to be replaced by relatively few quantities which shall adequately represent the whole , or which , in other words , shall contain as much as possible , ideally the whole , of the relevant information contained in the original data .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ for the early stages of data analysis , i suggest , the astronomer is quite proficient . the ccd image must be flat - fielded with bias removed and photometrically calibrated to standard stars .
the spectrum must be extracted after removal of cosmic ray streaks , and drizzled onto a fixed wavelength grid after fitting a polynomial derived from calibration observations of atomic line standards .
the interferometric visibilities must be fourier transformed into an image or datacube with computationally intensive restoration procedures to treat incomplete coverage in the fourier plane .
ancillary information from the instrument is collected and used to improve calibration , point spread functions , and registration to a fixed grid on the sky .
the astronomer must then engage in fisher s data reduction , transforming terabytes or petabytes of data with kilobytes of digestible information in the form of tables and figures for communication to the wider scientific community .
this data analysis and reduction is then followed by the more intellectually challenging stage of science analysis
. this can start with _ prima facie _ interpretation of the results , but often proceeds with comparing the findings to mathematical models
. these might be simple heuristic models , such as a power law relationship between two variables , or more complex nonlinear and multivariate models derived from astrophysics .
astronomy is unusual in the intensity of this last step .
many fields that collect and interpret data @xmath0 social sciences such as economics , biological sciences such as ecology or genomics , earth sciences such as meteorology or seismology @xmath0 do not have physical models equivalent to elliptical orbits based on newtonian mechanics or spectral lines based on atomic physics .
nonlinear regression thus plays an important role in linking astronomical data to astrophysical models and , if the astrophysical theory or past observational studies constrain parameters in advance of a particular study , then bayesian modeling to estimate posteriors that update priors is used . in the past few years , astronomers have become among the most active scientific practitioners of bayesian inference .
the astronomer and astrophysicist needs a very wide suite of statistical methods . in observational cosmology ,
for example , table 1 shows an association between statistical topics and astronomical problems .
once a statistical analysis is performed , the scientist should ask questions such as : how reliable are my results ( often using [email protected] , the 3-sigma criterion ) ?
have i used the most effective and reliable statistical procedures for achieving my results ?
do my interpretations depend on heuristic or uncertain models ?
do my conclusions depend on my chosen analysis path ? * statistical fields for research in cosmology * [ cols= " < , < " , ] the role of computation is easily discerned .
the telescopic data are typically passed through a complex instrument - specific software pipeline , and then are subject to flexible analysis for the specific purposes of the research project .
virtually all astronomers are competent programmers in general purpose languages like python , c or idl as well as discipline - specific software environments developed over many years . for many problems
, analysis is performed on desktop computers with a single cpu and sub - terabyte local storage .
but increasingly , the dataset has high volume , arrives with high velocity , and exhibits nontrivial variety of structure .... the three v s of big data .
but even with moderate - sized datasets , the computational burden may be heavy due to , for example , fitting high - dimensional models , bootstrap - type simulations for model evaluation , or markov chain monte carlo procedures for bayesian model computations .
for both big data and computationally intensive situations , the desktop hardware is insufficient and the astronomer must move into a parallel computation on multi - processor computers , sometimes equipped with high speed graphical processing units .
these supercomputers might reside at the local institution , at the observatory responsible for the telescope , or on the commercial cloud .
despite these strong needs , astronomers are ill - prepared for community - wide use of advanced statistical and computational methods .
the research literature demonstrates a wide range of methodological skills . in many papers ,
the analysis is confined to a narrow suite of familiar statistical methods such as least - squares regression ( legendre 1805 ) , chi - squared test ( pearson 1901 ) , kolmogorov - smirnov goodness - of - fit test ( kolmogorov , 1933 ) , principal components analysis ( hotelling 1936 ) .
even these traditional methods can be misused .
for example , the astronomers ` minimum chi - squared ' regression method has ill - determined degrees of freedom if it is based on arbitrarily binned data , according to an important theorem ( chernoff & lehmann 1954 ) .
the kolmogorov - smirnov statistic is less sensitive than the anderson - darling statistic that gives a tail - weighted cramer - von mises test ( stephens 1986 ) .
aperiodic autocorrelated behaviors in time series , such as long - memory @xmath2 variability , is difficult to estimate in a reliable fashion
. the weak methodology in many studies can be directly attributable to the lack of formal training in statistics .
astronomers usually take zero or one course in statistics during their university and graduate education , in contrast to extensive training in physics and associated mathematics .
education is similarly weak in computational methods .
a survey of @xmath31100 astronomers worldwide ( momcheva & tollerud 2015 ) shows that 90% write software but only 8% receive substantial training in software development and 43% receive no training .
preferred languages are python ( 67% ) and idl ( 44% ) , but only 6% ( 3% in usa ) prefer r , the premier software environment for advanced statistical methods . momcheva and tollerun remark that " considering the wide - spread use of r in other scientific fields , its popularity among astronomers is strikingly low ? .
for this reason , the r language is briefly reviewed below ( 4 ) .
the result of weak training is a widespread ignorance of both fundamental principles and of the myriad recent advances in modern statistical and computational methodology .
these fields are huge , so even if an astronomer is familiar with contemporary methods in one area , they may be unfamiliar with methodology in other areas .
fortunately , a growing fraction of the astronomical research community has recognized these deficiencies , and considerably activity in methodology has emerged in recent years .
papers in the astronomical literature with the keywords `` methods : statistical '' or `` methods : numerical '' have more than doubled in the past decade to @xmath31000 studies annually , nearly 10% of the total literature .
nearly 2000 astronomers , many of them graduate students , have attended short ( 1 - 5 day ) professional development tutorials in astrostatistics and astroinformatics in a dozen countries .
cross - disciplinary research collaborations are active in several distinguished universities , particularly in the united states .
cross - disciplinary conferences in astrostatistics and astroinformatics , ranging from topical workshops to sessions within large congresses to week - long symposia , have been held with increasing frequency .
several scholarly societies have taken note of the growing importance of advanced methodology for astronomy , particularly in light of the growth of big data surveys .
these include the international astrostatistics association affiliated with the international statistical institute , commission on astroinformatics and astrostatistics within the international astronomical union , interest group in astrostatistics within the american statistical association , working group in astroinformatics and astrostatistics within the american astronomical society , and the astrominer task force within the institute of electrical and electronics engineers .
the large synoptic survey telescope , the u.s . national project , has a cross - disciplinary informatics and statistics science collaboration .
new resources are available for students and researchers .
textbooks that focus on statistical analysis of astronomical data are available with software implementations in mathematics , python and r ( gregory 2005 , wall & jenkins 2012 , feigelson & babu 2012 , ivezi et al .
2014 ) . a large collection of recent papers , meetings , jobs , blogs and other on - line resources is available from the astrostatistics and astroinformatics portal ( asaip.psu.edu ) . up to forty
gatherings worldwide occur annually related to these growing fields .
there is little practical utility for advanced statistical and computational methodology if software implementations are not available .
the astronomical community commits considerable resources to software specific to particular instruments and science goals , but these incorporate mostly familiar and simple methods .
the more sophisticated methods are often focused on particular science projects and are the responsibility of individual scientists , rather than the larger software teams committed to observatory data analysis .
it is in this later science analysis phase , illustrated in table 1 , rather than earlier pipelined data reduction phase that new software capabilities are needed . often the problems are not unique to a particular study , but embody analysis goals seen elsewhere in astronomy and other scientific fields .
thus , general utility statistical packages can be quite useful for the science phase in astronomical studies . in the era of mainframe computers
, academic statisticians did not promulgate code but rather relegated the task to commercial software systems like sas ( statistical analysis system ) and spss ( statistical package for the social sciences ) . during the 1980s when the c language was developed at at&t labs , john chambers at the labs created the s statistical package written in c. this became the commercial s - plus package . in the 1990s as the internet emerged , new zealand statisticians ross ihaka and robert gentleman rewrote the s language as an open source system called r. the r core development team of a dozen statistical computing experts around the world formed , and the system was released to the public under a gnu general public license .
this included a procedure for users to contribute specialized packages within the comprehensive r analysis network ( cran ) .
unexpectedly , the contributed cran packages proliferated at an enormous rate , growing exponentially from 2001 through 2011 .
today 5 - 6 new cran packages are submitted every day , some with one function ( subroutine ) and others with hundreds .
major fields with collections of disciplinary cran packages include genomics and econometrics but packages serve fields as diverse as actuarial science , archeology , clinical trials , earthquake , entomology , industrial quality control , internet streams .
most packages implement advanced statistical methods without reference to particular disciplines . in late-2016 , r has about 10,000 cran packages with perhaps 150,000 functions including statistical operations , datasets and infrastructure .
about 20 packages are specifically oriented towards astronomy including fits format input / output and a translation of much of the idl astronomy library .
r and python are the principal languages in the business fields of data science and analytics , and r dominates posts on software forums like stackoverflow and talk stats .
thus , r has quickly grown to be the world s premier statistical computing environment that implements a large fraction of modern statistics .
it is larger than sas , the most comprehensive of the commercial statistical packages , and is used by at least 2 million individuals .
astronomers have little difficulty learning the r language , as its syntax is quite similar to idl and matlab .
it is a compact scripting language where a statistical operation , which may be very complicated and require extensive computation , is called in a single line .
interactive r provides a graphical user interface with several available integrated development environments .
r has a byte code compiler similar to idl , python and matlab .
all of these languages give rapid computation for vector / matrix inputs but are slower for more complex program structures like loops .
fortran , c or c++ code can be easily incorporated into r scripts for high performance computing .
important to astronomy , two - way communication exists between r and python ; one can wrap an r / cran function in a python code , or wrap a python program within an r script .
the main difficulty in using r is its size and diversity .
astronomers need sufficient education in methodology to identify what procedure within r / cran is needed for a particular application .
the _ cran task views _ give brief roadmaps ro r and cran capabilities in broad areas of statistics like bayesian inference , machine learning or time series analysis .
for example , cran currently has 34 packages for wavelet analysis , some disciplinary specific and others general toolboxes .
some individual cran packages are very impressive .
the _ spatstat _
package gives over a thousand functions to analyze spatial point processes in 2 or 3 dimensions for datasets like galaxy redshift surveys .
important packages are described in books such as the _ spatstat _ volume by baddeley et al .
( 2015 ) or the dozens of cookbooks in springer s _ use r ! _ book series .
dozens of other books give introductions to basic statistical analysis with r or , like feigelson & babu ( 2012 ) and hilbe et al .
( 2017 ) for astronomers , are guides for particular disciplines . while r was not respected for high performance computing in the past , dozens of recent cran packages link r to various modern computational hardware structures . with the @xmath4 function
, loops can be easily distributed among local cores .
cran packages treat parallel supercomputers with message passing interface , parallel virtual machine , open multi - processor , apache hadoop and other frameworks .
other packages facilitate use of large datasets out of local memory ( including commercial cloud computing ) , and gpu computing .
astronomy played a critical role in the foundation of modern statistics with the development of least squares regression in the early 19th century , just as it was at the foundation of modern physics with celestial mechanics of the 18 - 19th centuries .
but during the 20th century , astronomy and statistics parted ways .
the former developed astrophysics based on electromagnetism , quantum mechanics and other branches of physics , while the latter moved to serve human needs in industry , government , social and biological sciences . by the late 20th century , the situation was poor : astronomers were unaware of important advances in methodology while statisticians were unaware of research issues in contemporary astronomy .
but the field of astrostatistics , both in learning established sophisticated methods and in pushing the envelope for new methodology inspired by astronomical research problems , exhibits strong growth during the 21st century .
astronomy did not play a central role in the development of computing machines and languages .
the innovations of charles babbage and ada lovelace , alan turing , claude shannon , john von neumann , bill gates , ken thompson and dennis ritchie ( developers of unix ) , and richard stallman ( gnu pioneer ) had little link to astronomy or other physical sciences .
only david stern , the developer of the interactive data language since the 1970s , was motivated by astronomical data problems .
rather , astronomers have become customers of computer hardware and software products and methodologies intended for business , government , engineering , and other purposes .
while the majority of astronomers use widely distributed standard hardware and software , a growing segment of the community has acquired skills in parallel processing , gpu computing , and machine learning .
astroinformatics is a very new speciality that is also quickly emerging today .
the landscape of these fields is thus changing rapidly .
the situation today feels somewhat disorganized as there is no natural center for these developments and they are widely dispersed geographically .
although scholarly societies show strong interest in these new fields , funding agencies are only beginning to be responsive to their needs .
the pattern has been to fund ` software ' development as an engineering effort closely linked to astronomical instrumentation without much study of optimal statistical and computational ` methods ' underlying the software . however , the external worlds of statistics , applied mathematics and computer science are so vast , and change so rapidly , that this informal approach can no confidently give rise to high quality procedures and outcomes .
cross - disciplinary experts should be hired within , or should serve as outside consultants to , astronomical software teams .
cross - disciplinary research groups in astrostatistics and astroinformatics should be nurtured at universities and institutes , just as astrochemistry and instrumentation groups with specialized skills emerged during the late 20th century .
the impetus for quickly incorporating astrostatistics and astroinformatics into the astronomical research community is propelled by the emergence of superb new high - throughput telescopes culminating with the optical - infrared band large synoptic survey telescope and radio band square kilometre array expected during the 2020s .
advanced methodology has always been an opportunity for improving science analysis and results .
but in the big data era , it is essential for reaching the scientific potential of the new instruments .
ivezi , z. , connolly , a. j. , vanderplas , j. & gray , a. ( 2014 ) _ statistics , data mining , and machine learning in astronomy : a practical python guide for the analysis of survey data _ , princeton univ . press | the history and current status of the cross - disciplinary fields of astrostatistics and astroinformatics are reviewed .
astronomers need a wide range of statistical methods for both data reduction and science analysis . with the proliferation of high - throughput telescopes
, efficient large scale computational methods are also becoming essential .
however , astronomers receive only weak training in these fields during their formal education .
interest in the fields is rapidly growing with conferences organized by scholarly societies , textbooks and tutorial workshops , and research studies pushing the frontiers of methodology .
r , the premier language of statistical computing , can provide an important software environment for the incorporation of advanced statistical and computational methodology into the astronomical community .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ * an aphorism * _ _ .
+ the scientist collects data in order to understand natural phenomena .
+ the statistician helps the scientist acquire understanding from the data .
+ the computer scientist helps the scientist perform the needed calculations ( * ) .
+ the individual proficient at all these tasks is a data scientist + _ ( * ) help needed only for big data _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ |
20 p. blaha , k. schwarz , g. k. h. madsen , d. kvasnicka and j. luitz , _ an augmented plane wave plus local orbitals program for calculating crystal properties _ , vienna univ . of technology , austria ( 2001 )
isbn 3 - 950131 - 1 - 2 | we have undertaken a study of diluted magnetic semiconductors @xmath0 and @xmath1 with @xmath2 , using the all electron linearized augmented plane wave method ( lapw ) for different configurations of mn as well as cr .
we study four possible configurations of the impurity in the wurtzite gan structure to predict energetically most favorable structure within the 32 atom supercell and conclude that the near - neighbor configuration has the lowest energy .
we have also analyzed the ferro - magnetic as well as anti - ferromagnetic configurations of the impurity atoms .
the density of states as well as bandstructure indicate half metallic state for all the systems . @xmath3 has also been estimated for the above systems .
introduction gallium nitride is one of the most promising materials among the diluted magnetic semiconductor ( dms ) material for application in spintronics . by doping transition metal ( tm ) atoms , mn or cr , local magnetic moment
are introduced in semiconductor which mediate ferromagnetically .
( ga , cr)n based dms was predicted to show high @xmath3 @xcite for high enough concentration of cr and further hashimoto _
et al . _
@xcite observed that ( ga , cr)n based dms grown by ecr molecular beam epitaxy showed @xmath3 above @xmath4k . cr@xmath5-implanted gan , studied by photoluminescence and superconducting quantum interference device ( squid ) reveal that the implanted cr@xmath5 incorporates substitutionally at ga site and the ferromagnetic order is retained upto @xmath6k @xcite .
takeuch _
et al . _
@xcite . have reported a systematic study of changes in the occupied and unoccupied n - partial density of states ( dos ) and confirm the wurtzite n @xmath7 dos and substitutional doping of cr into ga sites using sxes and xas .
recently , ferromagnetism above @xmath8k was reported in cr - gan thin films @xcite .
theoretically it was predicted that the ferromagnetic ( fm ) interaction in ( ga , mn)n may be retained upto room temperature @xcite .
the initial reports of high @xmath3 in ( ga , mn)n were followed by controversial results where the reported @xmath3 varied between @xmath9k - @xmath10k @xcite .
zajac and coworkers observed mn ions in ga@xmath11mn@xmath12n ( @xmath13 ) crystals coupled anti - ferromagnetically ( afm ) @xcite .
electronic structure and magnetic properties of zinc blende ga@xmath11mn@xmath12n for several values of @xmath14 with varied spatial distribution of dopant atoms to understand the magnetic interaction for explanation of fm - afm competition is discussed by uspenskii _
et al . _
@xcite where the calculations were done using the tight binding lmto method in the local spin density approximation .
sanyal and mirbt @xcite have studied mn doped gaas and gan dms using the _ ab - initio _ plane wave code ( vasp ) within density functional theory ( dft ) .
they have determined the interatomic exchange interactions by substituting mn in various positions in the unit cell and have attributed the origin of ferromagnetism in ( ga , mn)n to double - exchange mechanism involving the hopping of mn@xmath15 electrons .
raebiger _
et al . _
@xcite used the full potential linearized augmented plane wave ( fp - lapw ) method to investigate the interplay between clustering and exchange coupling in magnetic semiconductor ga@xmath11mn@xmath12as .
they have studied all possible arrangements of the two mn atoms on ga sublattice for @xmath16 and found that clustering of mn atoms at near neighbour ga sites is energetically preferred .
our analysis of the wurtzite gan doped with mn or cr is motivated by the latter study .
method and computational details we have employed the spin - polarized linearized augmented plane wave method ( fp - lapw ) as implemented in the wien2k package @xcite with the generalized gradient approximation ( gga ) for the exchange - correlation potential proposed by perdew , burke and ernzerof ( pbe96 ) @xcite .
this is state - of - the - art electronic structure method , which does not use any shape approximation for the potential , to solve the kohn - sham type of equations self - consistently .
gan normally occurs in the wurtzite structure with lattice constants @xmath17 and @xmath18 , giving @xmath19 ratio of 1.62 .
each ga is tetrahedrally bonded to n atoms at an average distance of @xmath20 and each n in turn is surrounded by four ga neighbors .
the calculations for dms were performed within a 32 atom supercell , constructed from @xmath21 standard unit cell of wurtzite structure wherein the dopant is substituted at various cation sites , since it has been shown that the formation energy for interstitial mn doping is higher than substitutional doping @xcite .
the supercell approach is used to restrict the dopant concentration to a small value , which is of interest for studying magnetic properties of the system , without altering the original underlying lattice structure .
our interest was in observing the changes in the electronic structure of the dms with respect to the possible different geometries of the dopants within the host semiconductor .
self - consistent electronic structure calculations were performed using the apw + local orbitals ( lo ) basis set for the valence and semi - core electrons with @xmath22 , @xmath23 and total energy convergence of @xmath24 the muffin - tin radii for ga , mn and cr were kept at @xmath25 and that for n at @xmath26 .
spin - polarized calculations were carried out to observe the effect of spin - splitting and to calculate the on - site magnetic moment at tm site .
we have studied wurtzite gan doped with one tm atom impurity , which is @xmath27 doping and two identical tm atoms in the @xmath28 atom unit cell amounting to @xmath29 doping . to simulate different surroundings for the transition metal ( tm ) atoms we have spanned certain geometries within the @xmath28 atom unit cell wherein the distance between the dopants is varied . in the case of the single impurity substitution ,
the nearest distance between two tm atoms is @xmath30 in plane and @xmath31 along the @xmath32-axes .
we have studied four different geometries of two tm atom substitutions at @xmath33 , near neighbor ( nn ) , @xmath34 , @xmath35 and @xmath36 separations .
when two near neighbor ( nn ) ga atoms are substituted by tm atoms the in - plane tm - tm atoms distance is @xmath33 and along the @xmath32-axes it is @xmath31 . for the second case
the out - of - plane distance is @xmath34 and the in - plane distance between the dopants is @xmath37 . in the third case ,
two ga atoms lying one above the other , along the @xmath38-axes , separated by a distance of @xmath35 are substituted by tm atoms and the in - plane tm atoms are at @xmath30 .
the last case is such that the in - plane separation ( @xmath30 ) and out of plane separation ( @xmath39 ) between the tm atoms is comparable . for estimating the magnetically favorable system , the spins of the dopants are aligned along the same direction , corresponding to the fm configuration , and aligned in the opposite directions corresponding to afm configuration .
the self consistency was achieved on a mesh of @xmath40 k - points .
structural relaxation for the tm site and the nn n sites was carried out to observe changes in the bond lengths between tm and the first shell of n atoms .
very small change ( @xmath41 ) was observed in the bond lengths and no significant changes were seen in the band structure in agreement with the earlier reported results @xcite .
thus the calculations reported here are for systems without allowing any relaxation .
results in the wurtzite gan semiconductor , each ga ( n ) is tetrahedrally bonded to 4 n ( ga ) atoms .
pure gan is a direct band gap semiconductor with top of the valence band consisting of n _ p_-states and the bottom of the conduction band having ga _
sp_-character .
the ga _
d _ levels are deep and do not take part in the bonding .
thus they are treated as core states .
the band gap of gan , which is underestimated by density functional theory within the approximation used for the exchange correlation energy functional , is @xmath42ev .
the experimentally determined band gap of undoped
gan is @xmath43ev .
das _
et al . _
@xcite have shown that , for mn atoms to couple ferromagnetically , they need to be kept apart by more than the critical distance of @xmath44 . similar calculations on clusters of ( gan)cr indicate that the critical cr - cr distance is @xmath45 @xcite . in all our calculations the distance between the dopants was greater than the corresponding critical distances .
mn doped systems localized magnetic moments are introduced within the gan system by substituting the cations with tm impurity atom(s ) .
mn atom with @xmath46 and @xmath47 electrons in the valence region replaces ga atom with valency @xmath48 . on substitution mn atoms contributing five @xmath49 levels per atom
are thus expected to contribute to the observed magnetic moment . 1.0 in 2.0 in 0.2 in 2.0 in + 0.02
cm + 2.1 in ( a ) 0.2 in 1.5 in since three of the valence electrons from mn go into compensating the three electron states of substituted ga , one hole per mn is introduced into the system .
figure [ mnet2 ] shows the mn - projected @xmath50 and @xmath51 majority spin electronic structure for @xmath52 .
the mn-@xmath49 states lie at the top of the valence band and cross the @xmath53 in some places .
these are split into @xmath51 and @xmath50 states , the @xmath50 level is two thirds filled and the @xmath51 is almost occupied .
the minority spin levels are empty and lie above the @xmath53 indicating @xmath54 spin - polarized states .
the mn induced states lie in the gap region of gan .
the top of the valence band in gan is composed of the n-@xmath55 levels and the unique properties , particularly the half metallic state of dms , thus arise from the tm _
d _ and host _ p _ interactions that couple the two subsystems .
mn - projected @xmath49-dos in @xmath56 ( a ) with mn - mn distance equal to @xmath33 and @xmath35 and ( b ) @xmath57 and @xmath37 .
the upper and lower panels represent the majority and minority spin dos respectively.,title="fig:",height=321 ] + ( a ) + mn - projected @xmath49-dos in @xmath56 ( a ) with mn - mn distance equal to @xmath33 and @xmath35 and ( b ) @xmath57 and @xmath37 .
the upper and lower panels represent the majority and minority spin dos respectively.,title="fig:",height=321 ] + ( b ) in order to understand the variation of exchange interaction among the tm impurity with the distance between the tm atoms , the concentration of mn atoms was increased to @xmath58 , equivalent to introducing @xmath59 mn atoms in the supercell .
self consistent calculations were carried out for two different magnetic configurations of mn electrons in which the electrons are parallel or antiparallel corresponding to fm or afm configuration .
for all the geometries of the dopants , systems , as described in section ii , with fm configuration of the mn atoms were found to have lower energy .
since the mn-@xmath49 levels are responsible for observed half metallic behavior , a comparison of mn-@xmath49 dos in various geometries is shown in figure [ mndos ] , ( a ) for separations @xmath33 and @xmath60 and ( b ) for @xmath57 and @xmath37 .
here the half metallic state is evident in all the cases .
the tm - tm distance of @xmath37 corresponds to single tm doping ( @xmath61 ) .
the mn @xmath62dos is broad for substitution at nn distance . in all the other cases
the band is split indicating that the majority spin @xmath49-bands of the two mn atoms at nn overlap . on increasing mn - mn distance , @xmath49-band splitting takes place implying a reduction in the interaction between the tm atoms .
it may be noted that the minority spin conduction band overlaps with the majority spin band for nn substitution .
a gap of @xmath63 is present between the minority spin conduction band and majority spin band for mn - mn distance greater than nn . the minority spin valence band
as well as conduction band is far apart from the @xmath53 thus retaining the highly spin polarized state also seen in the single mn doped @xmath64 system .
the down spin gap is @xmath65 for nn configuration and increases to @xmath66 at larger separations .
on increasing the distance between the mn atoms , splitting of d - level increases .
this is consistent with the observation that in single impurity doping , tm - tm atom distance is @xmath57 and splitting of the mn-@xmath49 band is larger as seen in figure [ mndos](b ) .
the magnetic moment at mn - site does not depend on the distance between the dopant atoms and has a value @xmath67 for all the geometries as indicated in table [ table1 ] . @llll system & total@xmath68 & dopant@xmath68 & n@xmath68 + @xmath52 & 4.00 & 3.33 & 0.001 + @xmath69 & 8.00 & 3.34 & 0.001 + mn - mn = @xmath33 & & & + @xmath69 & 8.00 & 3.34 & 0.005 + mn - mn = @xmath34 & & & + @xmath69 & 8.00 & 3.33 & 0.006 + mn - mn
= @xmath35 & & & + @xmath69 & 8.00 & 3.34 & 0.005 + mn - mn = @xmath57 & & & + the presence of localized moment influences the near neighbor n atoms within the gan system , such that dos of the nn n - atoms around the impurity atom becomes as shown in figure [ ndos](a ) . due to the @xmath70 interaction , induced states are seen on n atoms .
the magnitude of the induced states is maximum at nn n atoms and decreases as the distance from the tm atom increases .
this again indicates a localized nature of the tm states .
figure [ ndos](b ) shows the spin charge density ( scd ) in a plane containing three nn n atoms for single mn doping .
the plot shows that the scd on the n atoms lying above the mn atom is negative whereas it is positive for the n atom lying below the mn atom .
average magnetic moment on the nn n atoms is positive as shown in table [ table1 ] for all the different geometries .
( a ) variation in partial - dos at 3 n - sites in @xmath52 .
the solid lines : nn - n along @xmath71axis .
dotted lines :
nn - n in plane .
dashed - dotted lines : next nn - n ( b ) scd plot in a plane of three nn n atoms.,title="fig:",height=245 ] + ( a ) + 0.2 cm ( a ) variation in partial - dos at 3 n - sites in @xmath52 . the solid lines : nn - n along @xmath71axis .
dotted lines : nn - n in plane .
dashed - dotted lines : next nn - n ( b ) scd plot in a plane of three nn n atoms.,title="fig:",width=188,height=188 ] ( a ) variation in partial - dos at 3 n - sites in @xmath52 . the solid lines : nn - n along @xmath71axis .
dotted lines :
nn - n in plane .
dashed - dotted lines : next nn - n ( b ) scd plot in a plane of three nn n atoms.,title="fig:",width=75,height=188 ] + ( b ) cr doped systems electronic structure calculation for substitutional doping of cr in the gan system was also done and is analyzed in a similar fashion . for each cr
doped in the @xmath28 atom supercell , equivalent to @xmath72 doping , there are five spin up @xmath49-states which are introduced in the gan band gap .
since the cr atom has @xmath73 valence electrons , only three of the electron states out of the five @xmath74 levels are occupied , thus creating two hole states per cr substitution .
single cr doping into the @xmath28 atom supercell at cation site results in the cr-@xmath49 levels appearing in the band gap of the semiconductor host as seen in figure [ cret2 ] . 1.0 in 2.0 in 0.2 in 2.0 in 0.02 cm + 2.1 in ( a ) 0.2 in 1.5 in the cr d@xmath75 levels split ( figure [ cret2](a ) ) , out of which two energy levels lie below the fermi level ( @xmath76 ) and are occupied .
the third level which is @xmath77ev above is unoccupied .
one of the d@xmath78 level is occupied and the other lies just above the @xmath76 as seen in figure [ cret2](b ) .
however , the hybridization of @xmath50 and @xmath51 majority spin states is negligible and these levels are well separated as opposed to mn doped case . as in the case of mn doping ,
the cr minority spin @xmath49-states are above the @xmath76 and so the impurity states at the @xmath76 are @xmath54 spin polarized .
cr - projected @xmath49-dos in @xmath79 for cr - cr separation of ( a ) @xmath33 and @xmath35 and ( b ) @xmath57 and @xmath37.,title="fig:",height=321 ] + ( a ) + cr - projected @xmath49-dos in @xmath79 for cr - cr separation of ( a ) @xmath33 and @xmath35 and ( b ) @xmath57 and @xmath37.,title="fig:",height=321 ] + ( b ) in the @xmath79 system , figure [ cr2 ] shows that even though the cr atoms are substituted at nn sites there is a gap seen between the split cr-@xmath49 levels , unlike in the nn mn - doped system . the minority spin conduction band overlaps with the majority spin band for nn cr case , as in nn mn case . for all the geometries of the dopant atoms the system is half metallic .
scd plot in a plane of three nn n atoms around the tm atom in @xmath80.,title="fig:",width=188,height=188 ] scd plot in a plane of three nn n atoms around the tm atom in @xmath80.,title="fig:",width=75,height=188 ] the band gap for the minority spin , in case of two cr substitution at @xmath33 is @xmath81ev which is larger than the corresponding mn case which has a gap of @xmath82ev . when the tm - tm distance is @xmath35 , in the cr case the gap is @xmath83ev whereas for mn substitution it is @xmath84ev .
the lowest energy configuration for cr substitution occurs for cr - cr distance of @xmath33 and for all the geometries studied the fm configuration of tm atoms has lower energy compared to afm configuration .
the magnetic moments at the cr site in various geometries is as shown in table [ tab2 ] and it is seen that the magnetic moment does not show much variation depending on the distance , indicating that the direct interaction between the cr atoms is minimal . @llll system & total@xmath68 & dopant@xmath68 & n@xmath68 + @xmath85 & 3.00 & 2.47 & -0.025 + @xmath86 & 6.00 & 2.48 & -0.031 + cr - cr = @xmath33 & & & + @xmath86 & 6.00 & 2.47 & -0.022 + cr - cr = @xmath34 & & & + @xmath86 & 6.00 & 2.48 & -0.025 + cr - cr = @xmath35 & & & + @xmath86 & 6.00 & 2.48 & -0.026 + cr - cr = @xmath57 & & & + magnetic moments observed at the various dopant sites from our calculation are shown in the table [ tab2 ] .
the total magnetic moment per unit cell per mn atom is @xmath87 and the average magnetic moment on the nn n atoms in case of mn - doping is parallel to the mn - moment .
this can be understood as penetration of the spin - polarized mn states to the neighboring host which does not have any of its own states in the gap region .
the magnetic moment per unit cell per cr atom is @xmath88 .
the average magnetic moment on the nn n atom is anti - parallel to cr - moment .
the difference in the orientation of the average magnetic moment on nn n atoms of mn and cr is due to the difference in the @xmath55 dos of the nn n along the z axis ( figure not shown here ) compared to the nn n atoms lying in the xy plane above the tm atoms .
there is not much variation of magnetic moment with increase in cr - cr distance and shows a similar trend as mn doped systems .
the magnitude of average nn - n magnetic moment is greater in case of cr substitution , which contribute one less electron to the hybridized valcen band .
scd in figure [ crscd ] on all of the nn - n atoms of the single cr ( only 3 nn n atoms shown in figure [ crscd ] ) doped system is negative thus showing that the tm atom and the nn - n are anti - ferromagnetically coupled .
estimation of @xmath3 we have predicted the @xmath3 for the dms based on ga@xmath89mn@xmath90n@xmath89 and ga@xmath89cr@xmath90n@xmath89 considering the mean field approximation .
@xmath91 for mn and cr doping .
inset shows the mean field @xmath3 variation with distance between dopants.,height=321 ] figure [ evsd ] shows the @xmath91 for the mn / cr doped systems , where @xmath92 is the total energy for the antiferromagnetic ( afm ) configuration and @xmath93 is the total energy for the ferromagnetic ( fm ) configuration
. observed variation @xmath94 vs distance for mn substituted dms agrees with the one reported by sanyal @xcite .
@xmath94 is a measure of the exchange interaction in the system .
highest @xmath94 is seen for the case where the tm atoms are substituted as near neighbors , signifying larger overlap of the magnetic impurity orbitals . for mn doping at nn @xmath95ev ,
this compares well with the value calculated for dimer substitution by uspenskii _ et . al _
@xcite which compared qualitatively with the high @xmath96k measured @xcite . as for the identical cr case @xmath97 also compares well with the observed @xmath3 but is lower than the @xmath8k observed by liu and co - workers @xcite . from figure [ evsd ]
it also emerges that the exchange interaction decreases sharply as the distance between the tm atoms increases .
thus the exchange interaction is short range and could be interpreted as the double exchange mechanism .
summary and conclusions we have analyzed the electronic structure of gan doped with tm mn and cr with @xmath72 and @xmath58 doping for various possible geometries to replicate the situation where the tm atoms would appear either to cluster or be separated .
the self consistent fp - lapw calculations predict half metallic state for @xmath72 as well as @xmath58 doping . comparing the total energies of the fm and afm configurations for @xmath58 doping ,
the fm state is found to be lower in energy and is predicted to be the preferred state .
on - site magnetic moment at the tm site shows insignificant variation with distance between the dopants .
the near neighbor n atoms contribute to the states in energy gap of the semiconductor due to the influence of the tm atoms .
average magnetic moment at nn n site is parallel to the mn magnetic moment where as it is anti - parallel to the cr atoms .
we observe that both the systems with nn substitution of mn / cr atom would show high @xmath3 .
the energy gap between the minority spin band in mn is @xmath98 lower than in cr doped system and we think this could be an important factor in determining a suitable system .
but since the magnetic moment at mn site is higher than cr , it would be of interest to study mixed systems of mn and cr to incorporate the salient features of both tm atoms . |
the nuclear many - body problem of @xmath0 interacting nucleons can be solved exactly only in very specific cases or for very small particle numbers .
this is due to the large number of degrees of freedom involved in such a complex system .
let us for instance consider particles interacting through n hamiltonian written as @xmath1 then the exact ground state energy can be written as @xmath2 where @xmath3 , @xmath4 , ... denote the one- , two- , ... body density matrices that contain all the information on the one- , two- ... body degrees of freedom respectively .
a natural way to reduce the complexity of this problem is to assume that at a given level , the @xmath5body ( and higher - order ) density matrices becomes a functional of the lower - order ones .
this is what is done for instance in the hartree - fock ( hf ) approximation where all @xmath6-body density matrices ( with @xmath7 ) become a functional of @xmath8 .
unfortunately , the hf theory applied to the nuclear many - body problem in terms of the vacuum hamiltonian is a poor approximation and many - body theories beyond hf are necessary .
the introduction of energy density functional ( edf ) approaches in the 70 s was a major breakthrough ( see for instance @xcite for a recent review ) . in its simplest form ,
the edf formalism starts with an energy postulated as a functional of @xmath8 , the latter being built out of a slater determinant .
then the ground state energy is obtained by minimizing the energy with respect to @xmath8 , i.e. @xmath9 parameters are generally adjusted on specific experimental observations and therefore encompass directly many - body correlations .
current edf uses a generalization of eq .
( [ eq : simpleedf ] ) obtained by considering quasi - particle vacua as trial states . by making explicit use of symmetry breaking , such a functional called hereafter single - reference ( sr- ) edf is able to account for static correlation associated with pairing and deformation .
actual sr - edf takes the form : @xmath10 where @xmath11 denotes the anomalous density . to restore symmetries and/or incorporate dynamical correlations , guided by the generator
coordinate method ( gcm ) , a second level of edf implementation , namely multi - reference ( mr- ) edf is introduced .
recently , difficulties with the formulation and implementation of have been encountered in mr - edf .
a minimal solution has been proposed in ref . @xcite . besides these problems , the authors of ref . @xcite
have pointed out the absence of a rigorous theoretical framework for the mr edf approach . at the heart of the problem
is the possibility to break symmetries in functional theories and then restore them using configuration mixing .
this issue needs to be thoroughly addressed in the future . in this context , it is interesting to see if extensions of the functional used at the sr - edf level can grasp part of the effects that for standard functionals require the mr level .
it is worth realizing that , in the canonical basis for which @xmath12 , we have @xmath13 = \frac{1}{4 } \sum_{i , j } \bar v^{\kappa \kappa}_{i\bar i j \bar j } \sqrt{n_i ( 1-n_i ) } \sqrt{n_j ( 1-n_j ) } , \label{}\end{aligned}\ ] ] and therefore , the energy can be regarded as a functional of natural orbitals @xmath14 and occupation numbers @xmath15 . as a matter of fact , for electronic systems , gilbert has generalized the kohn - sham theory and shown that the exact energy of a system can be obtained by minimizing such a functional @xcite leading to the so - called density matrix functional theory ( dmft ) . the possibility to consider occupation numbers as building blocks of the nuclear energy functional
has recently been discussed in ref .
two levels of theory can be developed along the line of gilbert s idea ( i ) either , functionals in the strict gilbert framework can be designed . in that case , since the density identify with the exact density at the minimum , it should respect all symmetries of the bare hamiltonian .
( ii ) or we exploit the concept of symmetry breaking . in the latter case , similarly to the sr - edf , strictly speaking we can not anymore rely on the theorem , but we may gain better physical insight with relatively simple functionals .
the descriptive power of dmft is illustrated here in the two - level lipkin model @xcite . in this model , the hartree - fock ( hf ) theory fails to reproduce the ground state energy whereas configuration mixing like generator coordinate method ( gcm ) provides a suitable tool @xcite .
therefore , the two - level lipkin model is perfectly suited both to illustrate that dmft could be a valuable tool and to provide an example of a functional for system with a `` shape '' like phase - transition . in this model ,
one considers @xmath0 particles distributed in two n - fold degenerated shells separated by an energy @xmath16 .
the associated hamiltonian is given by @xmath17 where @xmath18 denotes the interaction strength while @xmath19 , @xmath20 are the quasi - spin operators defined as @xmath21 , @xmath22 and @xmath23 .
@xmath24 and @xmath25 are creation operators associated with the upper and lower levels respectively . due to the specific form of the lipkin hamiltonian , @xmath8 simply writes in the natural basis as @xmath26 . introducing the angle @xmath27 between the state @xmath28 and @xmath29 , leads to the following mean - field functional @xcite @xmath30 where @xmath31 .
this expression is easily obtained by generalizing the hartree - fock case ( recovered here if @xmath32 ) .
the main challenge of the method is to obtain an accurate expression for @xmath33 . to get the functional , clearly identified cases from which properties of the functional could be inferred have been used@xcite , namely the @xmath34 case and the large @xmath0 limit . in the two - particles case ,
the correlation energy can be analytically obtained and reads @xmath35 a simple extension of the @xmath34 case for larger number of particles is to assume that each pair contributes independently from the others leading to @xmath36 { \cal e}^{^{{n=2}}}_{\rm cor}$ ] .
however , such a simple assumption leads to a wrong scaling behavior in the large @xmath0 limit .
indeed , in this case , @xmath37 as @xmath0 tends to infinity while a @xmath38 scaling is expected @xcite .
to obtain the correct limit , a semi - empirical factor @xmath39 can be introduced such that @xmath40 with @xmath41 .
the value @xmath42 has been retained using a fitting procedure .
examples of results obtained by minimizing the functional given by eqs .
( [ eq : emflipkin ] ) and ( [ eq : coreta ] ) are shown in fig .
[ fig : chi ] for different particle numbers and interaction strengths . in all cases , a very good agreement ,
much better than the hf case is found . for @xmath43 to @xmath44 resp .
from top to bottom . in each case , the corresponding hf ( dashed line ) and dmft ( filled circle ) minimum energy are shown .
the dmft calculation is performed using the mean - field and correlation energy resp . given by eq .
( [ eq : emflipkin ] ) and eq .
( [ eq : coreta ] ) with @xmath45 ( adapted from @xcite ) . ]
the lipkin example suggests that dmft can be a valuable tool for describing ground state of a many - body system when symmetry breaking plays a significant role .
the functional designed here is exact only in the @xmath34 .
note that the functional proposed here breaks signature symmetry and therefore enters into the level ( ii ) of functional discussed in the introduction .
the lipkin model is however rather schematic and can not be used as a guidance for realistic situations .
the possibility to design a new accurate functional for nuclei remains a challenging problem .
the author thanks m. assi , b. avez , m. bender , t. duguet , c. simenel , o. sorlin and p. van isacker for enlightening discussions at different stages of this work and t. papenbrock for useful remarks on the scaling behavior in the lipkin model .
d. lacroix , t. duguet , and m. bender , phys . rev . *
c79 * , 044318 ( 2009 ) .
m. bender , t. duguet , and d. lacroix , phys . rev . *
c79 * , 044319 ( 2009 ) .
t. duguet , m. bender , k. bennaceur , d. lacroix , and t. lesinski , phys . rev . *
c79 * , 044320(2009 ) . | the possibility to use functionals of occupation numbers and natural orbitals for interacting fermions is discussed as an alternative to multi - reference energy density functional method . an illustration based on the two - level lipkin model
is discussed .
address = ganil , cea and in2p3 , bote postale 5027 , 14076 caen cedex , france |
let @xmath2 be a finite group .
factor equivalence of finitely generated @xmath3-free @xmath4$]-modules is an equivalence relation that is a weakening of local isomorphism .
it has been used e.g. in @xcite among many other works to derive restrictions on the galois module structure of rings of integers of number fields and of their units in terms of other arithmetic invariants .
more recently , a set of rational numbers has been attached to any finitely generated @xmath4$]-module , called regulator constants @xcite , with the property that if two modules are locally isomorphic , then they have the same regulator constants .
these invariants have been used in @xcite and in @xcite to investigate the galois module structure of integral units of number fields , of higher @xmath0-groups of rings of integers , and of mordell - weil groups of elliptic curves over number fields .
it is quite natural to ask whether there is a connection between the two approaches to galois modules and whether the results of one can be interpreted in terms of the other .
it turns out that there is indeed a strong connection , which we shall investigate here .
we will begin in the next section by recalling the definitions of factorisability , of factor equivalence , and of regulator constants .
we will then establish some purely algebraic results that link factor equivalence and regulator constants . in [ sec : results ] we will revisit the relevant results of @xcite on galois module structures and will use the link established in [ sec : conn ] to compare them to each other . finally , in
[ sec : k ] we will use the results of [ sec : conn ] to prove a factorisability result on @xmath0-groups of rings of integers that is a direct analogue of ( * ? ? ?
* theorem 5.2 ) . throughout the paper , whenever there will be mention of a group @xmath2 , we will always assume it to be finite .
all @xmath4$]-modules will be assumed to be finitely generated and all representations will be finite - dimensional .
this work is partially funded by a research fellowship from the royal commission for the exhibition of 1851 .
part of this work was done , while i was a member of the mathematics department at postech , korea .
it is a great pleasure to thank both institutions for financial support , and to thank postech for a friendly and supportive working environment .
i am also grateful to an anonymous referee for carefully reading the manuscript and for helpful suggestions .
we will begin by recalling the definition of factorisability and of factor equivalence , and by discussing slight reformulations .
this concept first appears in @xcite and plays a prominent rle e.g. in the works of frhlich .
[ def : factorisability ] let @xmath2 be a group ( always assumed to be finite ) , and let @xmath5 be an abelian group , written multiplicatively .
a function @xmath6 on the set of subgroups @xmath7 of @xmath2 with values in @xmath5 is _ factorisable _ if there exists an injection of abelian groups @xmath8 and a function @xmath9 on the irreducible characters of @xmath2 with values in @xmath10 , with the property that @xmath11\rangle}\ ] ] for all @xmath12 , where @xmath13 denotes the set of irreducible characters of @xmath2 , and @xmath14 denotes the usual inner product of characters .
the definition one often sees in connection with galois module structures is a special case of this : @xmath5 is usually taken to be the multiplicative group of fractional ideals of the ring of integers @xmath15 of some number field @xmath16 , and @xmath10 is required to be the ideal group of @xmath17 for some finite galois extension @xmath18 with galois group @xmath2 , with @xmath19 being the natural map @xmath20 .
let us introduce convenient representation theoretic language to concisely rephrase the above definition .
the _ burnside ring _
@xmath21 of a group @xmath2 is the free abelian group on isomorphism classes @xmath22 $ ] of finite @xmath2-sets , modulo the subgroup generated by elements of the form @xmath23+[t ] - [ s\sqcup t],\ ] ] and with multiplication defined by @xmath23\cdot [ t ] = [ s\times t].\ ] ] the _ representation ring _
@xmath24 of a group @xmath2 over the field @xmath25 is the free abelian group on isomorphism classes @xmath26 $ ] of finite dimensional @xmath25-representations of @xmath2 , modulo the subgroup generated by elements of the form @xmath27+[\tau ] - [ \rho\oplus \tau],\ ] ] and with multiplication defined by @xmath27\cdot [ \tau ] = [ \rho\otimes \tau].\ ] ] in the case that @xmath28 , which will be the main case of interest , we will omit the subscript and simply refer to the representation ring @xmath29 of @xmath2 .
there is a natural map @xmath30 that sends a @xmath2-set @xmath5 to the permutation representation @xmath31 $ ] .
denote its kernel by @xmath32 . by artin
s induction theorem , this map always has a finite cokernel @xmath33 of exponent dividing @xmath34 .
moreover , @xmath33 is known to be trivial in many special cases , e.g. if @xmath2 is nilpotent , or a symmetric group .
the cokernel @xmath33 is important when strengthenings of the notion of factorisability are considered , such as @xmath35-factorisability , but will not be important for us .
it follows immediately from definition [ def : factorisability ] and from standard representation theory that for @xmath36 to be factorisable , it has to be constant on conjugacy classes of subgroups .
there is a bijection between conjugacy classes of subgroups of @xmath2 and isomorphism classes of transitive @xmath2-sets , which assigns to @xmath12 the set of cosets @xmath37 with left @xmath2-action by multiplication , and to a @xmath2-set @xmath1 the conjugacy class of any point stabiliser @xmath38 , @xmath39 .
an arbitrary @xmath2-set is a disjoint union of transitive @xmath2-sets , and so an element of @xmath21 can be identified with a formal @xmath3-linear combination of conjugacy classes of subgroups of @xmath2 .
so if @xmath36 is a factorisable function , then it can be thought of as a function on conjugacy classes of subgroups of @xmath2 , equivalently on transitive @xmath2-sets , and then extended linearly to yield a group homomorphism @xmath40 . [
prop : factorisability ] let @xmath41 be a group homomorphism , where @xmath5 is an abelian group .
the following are equivalent : 1 .
@xmath36 is factorisable in the sense of definition [ def : factorisability ] .
2 . there exists an injection @xmath42 of abelian groups such that the composition @xmath43 factors through the natural map @xmath44 .
3 . there exists an injection @xmath45 such that @xmath46 factors through the natural map @xmath30 , i.e. there is a homomorphism @xmath47 that makes the following diagram ( whose first row is exact ) commute : @xmath48 & k(g ) \ar[r ] & b(g ) \ar[r ] \ar[d]_f & r(g)\ar[d]^{g ' } \ar[r ] & c(g)\ar[r ] & 0\\ & & x \ar@{^{(}->}[r]_{\iota ' } & y ' . & & } \ ] ] 4 .
the homomorphism @xmath36 is trivial on @xmath49 .
the condition ( 2 ) is just a reformulation of ( 1 ) .
suppose that condition ( 2 ) is satisfied , and let us deduce ( 3 ) .
let @xmath50 be the map @xmath51 whose existence is postulated by ( 2 ) .
define @xmath52 to be the subgroup of @xmath10 generated by @xmath53 and by @xmath54 , define @xmath55 to be the restriction of @xmath50 to @xmath29 , followed by the inclusion @xmath56 , and @xmath57 to be @xmath19 , followed by the inclusion @xmath58 .
then @xmath52 , @xmath57 , @xmath55 satisfy ( 3 ) .
a brief diagram chase shows that ( 3 ) implies ( 4 ) : since @xmath57 , is an injection , @xmath59 .
so for the diagram in ( 3 ) to commute , we must have @xmath60 .
incidentally , exactly the same proof shows also that ( 2 ) implies ( 4 ) .
finally , the implication ( 4 ) @xmath61(2),(3 ) follows from two standard facts about abelian groups : * any abelian group can be embedded into a divisible abelian group , * and any homomorphism from a subgroup @xmath62 of an abelian group @xmath63 to a divisible group @xmath64 extends to a homomorphism from @xmath63 to @xmath64 .
since @xmath36 is trivial on @xmath32 , it induces a homomorphism from @xmath65 , which is canonically identified with a subgroup of @xmath66 .
now , embed @xmath5 into a divisible group @xmath10 , and extend @xmath67 to a homomorphism @xmath68 . 1 .
it follows from the last part of the proof that if @xmath5 is divisible , then @xmath52 can be taken to be equal to @xmath5 in proposition [ prop : factorisability ] . also , if @xmath33 is trivial , then @xmath69 , and again @xmath52 can be taken to be equal to @xmath5 .
if @xmath5 is the group of fractional ideals of a number field @xmath16 , and if @xmath36 vanishes on @xmath32 , then @xmath52 can always be taken to be the group of fractional ideals of a suitable galois extension @xmath18 , so this is not an additional restriction .
indeed , a sufficient condition on @xmath52 is that elements of @xmath65 that are @xmath70-divisible in @xmath29 are mapped under @xmath36 to elements of @xmath5 that become @xmath70-divisible in @xmath52 .
so if @xmath5 is the group of fractional ideals of a number field @xmath16 , this condition translates into relative ramification indices of some integral ideals of @xmath0 being divisible by some integers , and some elements of @xmath16 having certain @xmath70-th roots in @xmath0 . in @xcite ,
the word `` representation - theoretic '' has been used in place of `` factorisable '' . let @xmath2 be a group , and let @xmath71 , @xmath72 be two @xmath3-free @xmath4$]-modules such that there is an isomorphism of @xmath73$]-modules @xmath74 .
fix an embedding @xmath75 of @xmath2-modules with finite cokernel . then @xmath71 and @xmath72 are said to be _ factor equivalent _ , written @xmath76 , if the function @xmath77 $ ] is factorisable .
the notion of factor equivalence is independent of the choice of the embedding @xmath78 , and defines an equivalence relation on the set of @xmath3-free @xmath4$]-modules .
if @xmath79 for some prime @xmath80 , then @xmath78 can be chosen to have a cokernel of order coprime to @xmath80 .
indeed , @xmath79 if and only if @xmath81 ( @xcite , see also @xcite ) , and an isomorphism @xmath82 gives rise to an embedding @xmath78 with cokernel of order coprime to @xmath80 , by composing it with multiplication by an integer to clear denominators .
it follows that two modules that are locally isomorphic at all primes @xmath80 are factor equivalent .
the above definition is the one usually appearing in the literature , but it will be convenient for us to follow @xcite in defining factor equivalence for @xmath4$]-modules that are not necessarily @xmath3-free : let @xmath2 be a group , and let @xmath71 , @xmath72 be two @xmath4$]-modules such that there is an isomorphism of @xmath73$]-modules @xmath83 .
fix a map @xmath84 of @xmath2-modules with finite kernel and cokernel .
then @xmath71 and @xmath72 are said to be _ factor equivalent _ if the function @xmath85\cdot|\ker(i)^h|^{-1}$ ] is factorisable .
again , this notion is independent of the choice of the map @xmath78 , and defines an equivalence relation on the set of @xmath4$]-modules that weakens the relation of lying in the same genus ( where @xmath71 and @xmath72 are said to lie in the same genus if @xmath79 for all primes @xmath80 ) .
we continue to denote by @xmath2 an arbitrary ( finite ) group .
we also continue to use the identification between conjugacy classes of subgroups of @xmath2 and isomorphism classes of transitive @xmath2-sets . under this identification , a general element of @xmath21
will be written as @xmath86 with the sum running over mutually non - conjugate subgroups , and with @xmath87 .
an element of @xmath32 is such a linear combination with the property that the virtual permutation representation @xmath88^{\oplus n_h}$ ] is 0 .
alternatively , more down to earth , if we write @xmath89 as @xmath90 with all @xmath91 , @xmath92 non - negative , then @xmath89 is in @xmath32 if and only if the permutation representations @xmath93^{\oplus n_i}$ ] and @xmath94^{\oplus n_j'}$ ] are isomorphic .
an element @xmath95 of @xmath32 is called a _
brauer relation_. the following invariants of @xmath4$]-modules were introduced in @xcite and used e.g. in @xcite to investigate galois module structures , as we shall review in the next section : let @xmath2 be a group and @xmath71 a @xmath4$]module .
let @xmath96 be a bilinear @xmath2invariant pairing that is non
degenerate on @xmath97 .
let @xmath98 be a brauer relation .
the regulator constant of @xmath71 with respect to @xmath89 is defined by @xmath99 here and elsewhere , the abbreviation @xmath100 refers to the @xmath3-torsion subgroup .
this is independent of the choice of pairing ( * ? ? ?
* theorem 2.17 ) . as a consequence ,
@xmath101 is always a rational number , since the pairing can always be chosen to be @xmath102-valued .
it is also immediate that @xmath103 , so given a @xmath4$]-module , it suffices to compute the regulator constants with respect to a basis of @xmath32 . in other words ,
this construction assigns to each @xmath4$]-module essentially a finite set of rational numbers , one for each element of a fixed basis of @xmath32 .
one can show that if @xmath71 , @xmath72 are two @xmath4$]-modules such that @xmath104 , then for all @xmath105 the @xmath80-parts of @xmath106 and @xmath107 are the same .
so , like factor equivalence , regulator constants provide invariants of a @xmath4$]-module that , taken together , are coarser than the genus .
let @xmath71 , @xmath72 be two @xmath4$]-modules with the property that @xmath74 , let @xmath84 be a map of @xmath2-modules with finite kernel and cokernel .
fix a @xmath108-valued bilinear pairing @xmath14 on @xmath72 that is non - degenerate on @xmath109 .
the following immediate observation is crucial for linking regulator constants with the notion of factorisability : @xmath110 ^ 2\cdot \det\left(\langle\cdot,\cdot\rangle\big|n/\tors\right)\\ & = & \frac{[n : i(m)]^2}{|\ker i|^2}\cdot\frac{|m_{\tors}|^2}{|n_{\tors}|^2}\cdot \det\left(\langle\cdot,\cdot\rangle\big|n/\tors\right).\end{aligned}\ ] ] we deduce [ lem ] let @xmath71 , @xmath72 be two @xmath4$]-modules such that @xmath74 , let @xmath95 be a brauer relation . then @xmath111}{|\ker(i|_m^h)|}\cdot\frac{|m_{\tors}^h|}{|n_{\tors}^h|}\right)^{2n_h}\cdot\cc_\theta(n)\ ] ] for any map @xmath84 of @xmath2-modules with finite kernel and cokernel . by combining this with proposition [ prop : factorisability ] ,
we obtain [ cor ] two @xmath4$]-modules @xmath71 and @xmath72 with the property that @xmath112 are factor equivalent if and only if @xmath113 for all brauer relations @xmath114 . in particular ,
if @xmath71 and @xmath72 are @xmath3-free and satisfy @xmath112 , then they are factor equivalent if and only if @xmath115 for all @xmath105 .
we shall now show by way of several examples how lemma [ lem ] and corollary [ cor ] link known results on galois module structures with each other . throughout this section ,
let @xmath18 be a finite galois extension of number fields with galois group @xmath2 .
the ring of integers @xmath17 , and its unit group @xmath116 are both @xmath4$]-modules . more generally , if @xmath1 is any @xmath2-stable set of places of @xmath0 that contains the archimedean places , then the group of @xmath1-units @xmath117 of @xmath0 is a @xmath4$]-module .
it is a long standing and fascinating problem to determine the @xmath2-module structure of these groups , e.g. by comparing it to other well - known @xmath2-modules or by linking it to other arithmetic invariants .
a starting point is the observation that @xmath118^{\oplus [ k:\q]}$ ] as @xmath73$]-modules .
also , by dirichlet s unit theorem , @xmath119 , where @xmath120\rightarrow \z\right),\ ] ] with the map being the augmentation map that sends each @xmath121 to 1 .
it is therefore natural to compare the galois module @xmath17 to @xmath4^{\oplus [ k:\q]}$ ] and @xmath117 to @xmath122 .
it had been known since e. noether that @xmath17 lies in the same genus as @xmath4^{\oplus [ k:\q]}$ ] if and only if @xmath18 is at most tamely ramified .
the following is therefore particularly interesting in the wildly ramified case : we always have that @xmath17 is factor equivalent to @xmath4^{\oplus[k:\q]}$ ]
. we will now give a very short proof of this result in terms of regulator constants .
first , note that by corollary [ cor ] the statement is equivalent to the claim that for any @xmath105 , @xmath123^{\oplus[k:\q]})$ ] . since regulator constants are multiplicative in direct sums of modules ( ( * ? ? ?
* corollary 2.18 ) ) , and since @xmath124)=1 $ ] for all @xmath105 ( ( * ? ? ?
* example 2.19 ) ) , we have reduced the proof of the theorem to showing that @xmath125 for all @xmath105 .
if we choose the pairing on @xmath17 defined by @xmath126 with the sum running over all embeddings @xmath127 , then the determinants on @xmath128 , @xmath12 , appearing in the definition of regulator constants are nothing but the absolute discriminants @xmath129 .
the fact that these vanish in brauer relations follows immediately from the conductor - discriminant formula .
as we have mentioned above , it is natural to compare @xmath117 with @xmath122 , since they span isomorphic @xmath73$]-modules . for @xmath12 ,
let @xmath130 denote the set of places of @xmath131 below those in @xmath1 , and let @xmath132 denote the @xmath1-class number of @xmath131 .
[ thm : mult ] fix an embedding @xmath133 of @xmath2-modules with finite cokernel . for @xmath134 , let @xmath135 be its residue field degree in @xmath136 , define @xmath137 then the function @xmath138\frac{n(h)}{h_s(k^h)l(h)}\ ] ] is factorisable . as in the additive case
, we want to understand and to reprove this theorem in terms of regulator constants .
more specifically , we will show it to be equivalent to [ thm : propmult ] for @xmath139 , let @xmath140 be the decomposition group of a prime @xmath141 above @xmath142 ( well - defined up to conjugacy ) . for any brauer relation @xmath143
, we have @xmath144 ) } \prod_h \left(\frac{w(k^h)}{h_s(k^h)}\right)^{2n_h},\ ] ] where @xmath145 denotes the number of roots of unity in @xmath131 , i.e. the size of the torsion subgroup of @xmath146 . note that since @xmath122 is torsion free and @xmath147 is injective , proposition [ prop : factorisability ] and lemma [ lem ] imply that theorem [ thm : mult ]
is equivalent to the following statement : for any brauer relation @xmath148 , @xmath149 the equivalence of theorems [ thm : mult ] and [ thm : propmult ] will therefore be established if we show that @xmath150 ) } \prod_h\left(\frac{l(h)}{n(h)}\right)^{2n_h}.\ ] ] this is just a linear algebra computation that we will not carry out in full detail , since it is a combination of the computations of @xcite and @xcite .
indeed , it is shown in @xcite that under the embedding @xmath151\hookrightarrow \z[s ] , \;\;\;\fp\mapsto \sum_{\fq\in s,\fq | \fp}f_\fp\fq\end{aligned}\ ] ] we have @xmath152 = \frac{n(h)}{l(h)}$ ] .
so , instead of computing @xmath153 for a suitable choice of pairing @xmath14 on @xmath122 , we may compute @xmath154 where @xmath155 is identified with a submodule of @xmath122 as in ( [ eq : embedding ] ) . to do that
, we note that for any @xmath12 , @xmath155 is generated by @xmath156 , @xmath157 for any fixed @xmath158 , and that there is a natural @xmath2-invariant non - degenerate pairing on @xmath122 that makes the canonical basis of @xmath159 $ ] orthonormal .
it is now a straightforward computation , which has essentially been carried out in @xcite , to show that the quantity ( [ eq : regconst ] ) is equal to @xmath160)},\ ] ] as required .
as another illustration of the connection we have established , we will give an easy proof of an analogue of ( * ? ? ?
* theorem 5.2 ) for higher @xmath0-groups of rings of integers
. the main ingredient will be the compatibility of lichtenbaum s conjecture on leading coefficients of dedekind zeta functions at negative integers with artin formalism , as proved in @xcite .
let @xmath161 be an integer .
let @xmath162 , respectively @xmath163 denote the set of real embeddings , respectively of representatives from each pair of complex conjugate embeddings of a number field @xmath35 , and denote their cardinalities by @xmath164 , respectively @xmath165 .
denote @xmath166 by @xmath167 .
it is shown in @xcite that the ranks of the higher @xmath0-groups or rings of integers are as follows : @xmath168 let @xmath18 be a finite galois extension with galois group @xmath2 , and let @xmath169 denote the set of real places of @xmath16 that become complex in @xmath0 . for @xmath170 , let @xmath171 denote the non - trivial one - dimensional @xmath102-representation of the decomposition group @xmath140 , which has order 2 . by artin
s induction theorem , a rational representation of a finite group is determined by the dimensions of the fixed subrepresentations under all subgroups of @xmath2 .
it therefore follows that we have the following isomorphisms of galois modules : @xmath172\nonumber\\ & \cong & \bigoplus_{\fp\in s_\infty(k)}\q[g / d_\fp]\;\;\;\text{if $ n$ is odd , and}\label{eq : kqodd}\\ k_{2n-1}(\co_k)\otimes \q & \cong & \bigoplus_{\fp \in s_r(k / k)}\ind_{g / d_\fp}\epsilon_{\fp}\oplus \bigoplus_{\fp\in s_2(k)}\q[g]\nonumber\\ & \cong & \bigoplus_{\fp \in s_r(k / k)}\q[g]\big/\q[g / d_\fp ] \oplus \bigoplus_{\fp\in s_2(k)}\q[g]\;\;\;\text{if $ n$ is even.}\label{eq : kqeven}\end{aligned}\ ] ] we are thus led to compare , using the machine of factorisability , the galois module structure of @xmath173 with @xmath174 $ ] when @xmath70 is odd , and with @xmath175 % = \bigoplus_{\fp \in
s_r(k / k ) } \z[g]\big/\z[g / d_\fp]\oplus \bigoplus_{\fp\in s_2(k)}\z[g]\ ] ] when @xmath70 is even . here and elsewhere
, we write @xmath176 interchangeably for the rational representation and for the unique ( up to isomorphism ) @xmath3-free @xmath177$]-module inside it .
let @xmath18 be a finite galois extension of number fields with galois group @xmath2 , let @xmath161 be an integer . then the function @xmath178}{|k_{2n-2}(\co_{k^h})|}\ ] ] is factorisable at all odd primes , where @xmath179\cong \bigoplus_{\fp\in s_\infty(k)}\z[g / d_\fp]\;\;\;\;\text{if $ n$ is odd , and}\\ m & = & \bigoplus_{\fp \in s_r(k / k)}\ind_{g / d_\fp}\left(\epsilon_\fp\right ) \oplus \bigoplus_{\fp\in s_2(k)}\z[g ] \;\;\;\text{if $ n$ is even},\end{aligned}\ ] ] and where @xmath180 is any inclusion of @xmath2-modules .
proposition [ prop : factorisability ] and lemma [ lem ] imply that the assertion of the theorem is equivalent to the claim that for any brauer relation @xmath148 , @xmath181^{2n_h}}{|k_{2n-2}(\co_{k^h})|^{2n_h}}\\ & = _ { 2 ' } & \frac{\cc_{\theta}(m)}{\cc_{\theta}(k_{2n-1}(\co))}\cdot \prod_h \left(\frac{|k_{2n-1}(\co_{k})^h_{\tors}|}{|k_{2n-2}(\co_{k^h})|}\right)^{2n_h } , \end{aligned}\ ] ] where @xmath182 means that the two sides have the same @xmath80-adic valuation for all odd primes @xmath80 .
now , for any odd prime @xmath80 and any subgroup @xmath12 , we have @xmath183 this is a consequence of the quillen lichtenbaum conjecture ( see e.g. ( * ? ? ?
* proposition 2.9 and the discussion preceding it ) ) , which is known to follow from the bloch kato conjecture , which in turn is now a theorem of rost , voevodsky , and weibel @xcite .
moreover , it follows from @xcite ( see ( * ? ? ?
* equation ( 2.6 ) ) ) that
@xmath184 putting this together , we see that the assertion of the theorem is equivalent to the claim that @xmath185 for all brauer relations @xmath89 . but @xmath186 ( not just up to powers of 2 ) by ( * ? ? ?
* corollary 2.18 and proposition 2.45 ( 2 ) ) , and because cyclic groups have no non - trivial brauer relations . | we compare two approaches to the study of galois module structures : on the one hand factor equivalence , a technique that has been used by frhlich and others to investigate the galois module structure of rings of integers of number fields and of their unit groups , and on the other hand regulator constants , a set of invariants attached to integral group representations by dokchitser and dokchitser , and used by the author , among others , to study galois module structures .
we show that the two approaches are in fact closely related , and interpret results arising from these two approaches in terms of each other .
we then use this comparison to derive a factorisability result on higher @xmath0-groups of rings of integers , which is a direct analogue of a theorem of de smit on @xmath1-units . |
the virtual state wavefunctions are essential for the construction of the multi - indexed laguerre and jacobi polynomials @xcite .
they are polynomial type solutions of one - dimensional schrdinger equations for shape - invariant potentials @xcite .
they are characterised as having negative energies ( the groundstate has zero energy ) , no zeros in the physical domain and that they and their reciprocals are square non - integrable . by dropping the condition of no zeros and the reciprocals
are required to be square - integrable at both boundaries , _ pseudo virtual state wavefunctions _ are obtained . in most cases ,
the virtual and pseudo virtual state wavefunctions are obtained from the eigenfunctions by twisting the parameter(s ) based on the discrete symmetries of the hamiltonian @xcite .
starting from a shape - invariant potential , a darboux transformation @xcite in terms of a nodeless pseudo virtual state wavefunction @xmath3 with energy @xmath4 produces a solvable system with an extra eigenstate below the original groundstate with energy @xmath4 and eigenfunction @xmath5 .
this method of generating a solvable system by `` adding an eigenstate '' below the groundstate is known for many years , starting from the simplest harmonic oscillator potential examples @xcite and followed by many authors @xcite@xcite .
as remarked by adler @xcite for the harmonic oscillator case and generalised by the present authors @xcite for other potentials , such a system can be derived by special types of krein - adler transformations .
that is , the krein - adler transformation for a system with negatively shifted parameters in which the created state will be the groundstate .
the transformation use all the eigenstates between the new and the original groundstates . in this paper
we present straightforward generalisation of the above result for various shape - invariant potentials listed in section [ sec : exa ] ; coulomb potential with the centrifugal barrier ( c ) , kepler problem in spherical space ( k ) , morse potential ( m ) , soliton potential ( s ) , rosen - morse potential ( rm ) , hyperbolic symmetric top @xmath6 ( hst ) , kepler problem in hyperbolic space ( kh ) , hyperbolic darboux - pschl - teller potential ( hdpt ) , on top of the well - known harmonic oscillator ( h ) , the radial oscillator ( l ) and the darboux - pschl - teller potential ( j ) .
they are divided into two groups according to the eigenfunction patterns in [ sec:2group ] .
we mainly follow infeld - hull @xcite for the naming of potentials .
a darboux - crum transformation in terms of multiple pseudo virtual state wavefunctions is equivalent to a certain krein - adler transformation deleting multiple eigenstates with shifted parameters .
in contrast to the use of genuine virtual state wavefunctions @xcite , not all choices of the multiple pseudo virtual states would generate singularity free systems .
the singularity free conditions of the obtained system are supplied by the known ones for the krein - adler transformations @xcite . underlying the above equivalence are infinitely many polynomial wronskian identities relating wronskians of polynomials with twisted parameters to those of shifted parameters .
these identities imply the equality of the deformed potentials with the twisted and shifted parameters .
this in turn guarantees the equivalence of all the other eigenstate wavefunctions .
we present the polynomial wronskian identities for group a ; the harmonic oscillator ( h ) , the radial oscillator ( l ) and the darboux - pschl - teller potential ( j ) and some others . for group
b , the identities take slightly different forms ; determinants of various polynomials with twisted and shifted parameters .
the infinitely many polynomial wronskian identities are the consequences of the fundamental wronskian ( determinant ) identity as demonstrated in section [ sec : main ] .
this paper is organised as follows .
the essence of darboux - crum transformations for the schrdinger equation in one dimension is recapitulated in
[ sec : genstr ] .
the definitions of virtual states and pseudo virtual states are given in
[ sec : vir ] . in section [ sec : exa ] two groups of eigenfunction patterns are introduced in [ sec:2group ] and related wronskian expressions are explored in
[ sec : wro ] .
the details of the eleven examples of shape - invariant systems are provided in [ sec : h][sec : hdpt ] .
section [ sec : main ] is the main part of the paper .
we demonstrate the equivalence of the darboux - crum transformations in terms of multiple pseudo virtual states to krein - adler transformations in terms of multiple eigenstates with shifted parameters .
the underlying polynomial wronskian identities are proven with their more general determinant identities .
the final section is for a summary and comments .
darboux transformations in general @xcite apply to generic second order differential equations of schrdinger form @xmath7 without further structures of quantum mechanics , _ e.g. _ the boundary conditions , self - adjointness of @xmath8 , hilbert space , etc . in the next subsection ,
we summarise the formulas of multiple darboux transformations , which are purely algebraic .
let @xmath9 ( @xmath10 ) be distinct solutions of the original schrdinger equation : @xmath11 to be called _ seed _ solutions . by picking up one of the above seed solutions ,
say @xmath12 , we form new functions with the above solution @xmath13 and the rest of @xmath14 ( @xmath15 ) : @xmath16}(x){\stackrel{\text{def}}{=}}\frac{\text{w}[\varphi_1,\psi](x)}{\varphi_1(x ) } = \frac{\varphi_1(x)\partial_x\psi(x ) -\partial_x\varphi_1(x)\psi(x)}{\varphi_1(x)},\quad \varphi^{[1]}_{1,k}(x){\stackrel{\text{def}}{=}}\frac{\text{w}[\varphi_1,\varphi_k](x)}{\varphi_1(x)}.\ ] ] it is elementary to show that @xmath17}(x)$ ] , @xmath18 and @xmath19}_{1,k}(x)$ ] are solutions of a new schrdinger equation of a deformed hamiltonian @xmath20}$ ] @xmath21}=-\frac{d^2}{dx^2}+u^{[1]}(x),\quad u^{[1]}(x){\stackrel{\text{def}}{=}}u(x)-2\partial_x^2\log\bigl|\varphi_1(x)\bigr| , \label{1pot}\ ] ] with the same energies @xmath22 , @xmath23 and @xmath24 : @xmath25}\psi^{[1]}(x)&=\mathcal{e}\psi^{[1]}(x),\quad \mathcal{h}^{[1]}\varphi^{-1}_1(x)=\tilde{\mathcal{e}}_1\varphi^{-1}_1(x ) , \label{newschr}\\ \mathcal{h}^{[1]}\varphi^{[1]}_{1,k}(x ) & = \tilde{\mathcal{e}}_k\varphi^{[1]}_{1,k}(x)\ \ ( k\neq 1 ) . \label{newschr2}\end{aligned}\ ] ] by repeating the above darboux transformation @xmath26-times , we obtain new functions @xmath27}(x)&{\stackrel{\text{def}}{=}}\frac{\text{w}[\varphi_1,\varphi_2,\ldots,\varphi_m,\psi](x ) } { \text{w}[\varphi_1,\varphi_2,\ldots,\varphi_m](x ) } , \label{psim}\\ \breve{\varphi}^{[m]}_j(x)&{\stackrel{\text{def}}{=}}\frac{\text{w}[\varphi_1,\varphi_2,\ldots,\breve{\varphi}_j,\ldots , \varphi_m](x ) } { \text{w}[\varphi_1,\varphi_2,\ldots,\varphi_m](x)}\ \ ( j=1,2,\ldots , m ) , \label{varphim}\end{aligned}\ ] ] which satisfy an @xmath26-th deformed schrdinger equation with the energies @xmath22 and @xmath28 @xcite : @xmath29}=-\frac{d^2}{dx^2}+u^{[m]}(x),\quad u^{[m]}(x){\stackrel{\text{def}}{=}}u(x)-2\partial_x^2\log\bigl| \text{w}[\varphi_1,\varphi_2,\ldots,\varphi_m](x)\bigr| , \label{mpot}\\ & \mathcal{h}^{[m]}\psi^{[m]}(x)=\mathcal{e}\psi^{[m]}(x),\quad \mathcal{h}^{[m]}\breve{\varphi}^{[m]}_j(x ) = \tilde{\mathcal{e}}_j\breve{\varphi}^{[m]}_j(x)\ \ ( j=1,2,\ldots , m ) . \label{mschr}\end{aligned}\ ] ] here @xmath30(x)$ ] is a wronskian @xmath31(x){\stackrel{\text{def}}{=}}\det\bigl(\frac{d^{j-1}f_k(x)}{dx^{j-1}}\bigr)_{1\le j , k\le n}.\ ] ] for @xmath32 , we set @xmath33(x)=1 $ ] and @xmath34(x)$ ] means that @xmath35 is excluded from the wronskian . in deriving the determinant formulas and use is made of the properties of the wronskian @xmath36(x ) = g(x)^n\text{w}[f_1,f_2,\ldots , f_n](x ) , \label{wformula1}\\ & \text{w}\bigl[\text{w}[f_1,f_2,\ldots , f_n , g ] , \text{w}[f_1,f_2,\ldots , f_n , h]\,\bigr](x){\nonumber\\}&=\text{w}[f_1,f_2,\ldots , f_n](x)\cdot \text{w}[f_1,f_2,\ldots , f_n , g , h](x ) \qquad(n\geq 0 ) .
\label{wformula2}\end{aligned}\ ] ] another useful property of the wronskian is that it is invariant when the derivative @xmath37 is replaced by an arbitrary ` covariant derivative ' @xmath38 with an arbitrary smooth function @xmath39 : @xmath40(x)= \det\bigl(d_{j-1}\cdots d_2d_1f_k(x)\bigr)_{1\le j , k\le n } , \label{covw}\ ] ] with @xmath41 . under the change of variable @xmath42 , the wronskian behaves @xmath43(x ) = \bigl(\frac{d\eta(x)}{dx}\bigr)^{\frac12n(n-1 ) } \text{w}[f_1,f_2,\ldots , f_n]\bigl(\eta(x)\bigr ) .
\label{wformula3}\ ] ] obviously the potential @xmath44}(x)$ ] is independent of the order of the seed solutions .
the zeros of the seed solution @xmath12 ( the wronskian @xmath45(x)$ ] ) induce singularities of the potential @xmath46}(x)$ ] in ( @xmath44}(x)$ ] in ) .
we apply darboux transformations for various extensions of exactly solvable potentials @xmath47 in one dimensional quantum mechanics defined in an interval @xmath48 .
we assume that the potential is smooth in the interval and the system has a finite ( or an infinite ) number of discrete eigenstates with a vanishing groundstate energy : @xmath49 hereafter we consider real solutions only and use the term _ eigenstates _ in their strict sense , _
i.e. _ they only apply to those with square integrable wavefunctions and correspondingly the _
eigenvalues_. the zero groundstate energy condition can always be achieved by adjusting the constant part of the potential @xmath50 .
then the hamiltonian is positive semi - definite and it has a simple factorised form expressed in terms of the groundstate wavefunction @xmath51 which has no node in @xmath52 : @xmath53 for all the examples of solvable potentials to be discussed in this paper , the main part of the discrete eigenfunctions @xmath54 are _ polynomials _ @xmath55 , _ in a certain function _
@xmath56 , which is called the _ sinusoidal coordinate
the original systems to be extended usually contain some parameter(s ) , @xmath57 and the parameter dependence is denoted by @xmath58 , @xmath59 , @xmath60 , @xmath61 , @xmath62 etc .
we consider the _ shape - invariant _
@xcite original systems only , which are characterised by the condition : @xmath63 or equivalently @xmath64 here @xmath65 is the shift of the parameters . it should be remarked that the pair of hamiltonians , @xmath66 and @xmath67 together with the corresponding darboux - crum transformations constitute the basic ingredients of the supersymmetric quantum mechanics @xcite .
the series of multiple darboux - crum transformations can also be formulated in the language of supersymmetric quantum mechanics .
we believe the formulation in [ sec : genstr ] is simpler and more direct than that of susy quantum mechanics .
the explicit extensions depend on the choices of the seed solutions @xmath9 ( @xmath68 ) .
the obvious choices are a subset of the discrete eigenfunctions @xmath69
( @xmath70 ) , in which @xmath71 is a subset of the index set of the total discrete eigenfunctions . by using those from the groundstate on @xmath72 , @xmath73 , , @xmath74 ( @xmath75 ) successively , crum @xcite has derived an essentially iso - spectral extension @xmath29}\phi_n^{[m]}(x)=\mathcal{e}_n\phi_n^{[m]}(x)\quad ( n = m , m+1,\ldots),\\ & \phi_n^{[m]}(x){\stackrel{\text{def}}{=}}\frac{\text{w}[\phi_0,\phi_1,\ldots,\phi_{m-1},\phi_n](x ) } { \text{w}[\phi_0,\phi_1,\ldots,\phi_{m-1}](x)},\quad ( \phi_m^{[m]},\phi_n^{[m ] } ) = \prod_{j=0}^{m-1}(\mathcal{e}_n-\mathcal{e}_{j})\cdot h_n\delta_{m\,n},\\ & u^{[m]}(x){\stackrel{\text{def}}{=}}u(x)-2\partial_x^2\log \bigl|\text{w}[\phi_0,\phi_1,\ldots,\phi_{m-1}](x)\bigr|,\end{aligned}\ ] ] in which the potential @xmath76}(x)$ ] ( @xmath77 ) at each step is non - singular . shape - invariance means simply @xmath78}(x;{\boldsymbol}{\lambda})&=u(x;{\boldsymbol}{\lambda}+s{\boldsymbol}{\delta } ) + \mathcal{e}_s({\boldsymbol}{\lambda})\quad(s=1,2,\ldots , m ) , \label{crumshape}\\ \phi^{[s]}_{s+n}(x;{\boldsymbol}{\lambda})&\propto \phi_n(x;{\boldsymbol}{\lambda}+s{\boldsymbol}{\delta})\quad ( s=1,2,\ldots , m;\ \ n=0,1,\ldots , ) . \label{crumshape2}\end{aligned}\ ] ] by allowing gaps in the final spectrum , krein and adler @xcite have generalised crum s results for @xmath79 ( @xmath80 : mutually distinct ) : @xmath29}\phi_n^{[m]}(x)=\mathcal{e}_n\phi_n^{[m]}(x)\quad ( n\notin\mathcal{d } ) , \label{kaeq}\\ & \phi_n^{[m]}(x){\stackrel{\text{def}}{=}}\frac{\text{w}[\phi_{d_1},\phi_{d_2},\ldots,\phi_{d_m},\phi_n](x ) } { \text{w}[\phi_{d_1},\phi_{d_2},\ldots,\phi_{d_m}](x)},\quad ( \phi_m^{[m]},\phi_n^{[m ] } ) = \prod_{j=1}^m(\mathcal{e}_n-\mathcal{e}_{d_j})\cdot h_n\delta_{m\,n } , \label{mnorm}\\ & u^{[m]}(x){\stackrel{\text{def}}{=}}u(x)-2\partial_x^2\log \bigl|\text{w}[\phi_{d_1},\phi_{d_2},\ldots,\phi_{d_m}](x)\bigr| .
\label{kau}\end{aligned}\ ] ] the potential @xmath44}(x)$ ] is non - singular when the set @xmath71 , which specifies the gaps , @xmath81}(x)\equiv0 $ ] ( @xmath82 ) , satisfies the conditions @xcite : @xmath83 these simply mean that the gaps are even numbers of consecutive levels .
these extensions are well - known . in the next subsection , we consider seed functions which are not eigenfunctions . now
let us consider extensions in terms of _ non - eigen seed functions_. for simplicity sake , we first consider the cases of _ exactly iso - spectral extensions _
( _ deformations _ ) .
the seed functions @xmath84 ( @xmath10 ) satisfying the following conditions are called _ virtual state wavefunctions _ : 1 . no zeros in @xmath48 , _
i.e. _ @xmath85 or @xmath86 in @xmath48 .
2 . negative energy , @xmath87 .
@xmath88 is also a polynomial type solution , like the original eigenfunctions , see .
square non - integrability , @xmath89 . 5
. reciprocal square non - integrability , @xmath90 .
of course these conditions are not totally independent .
the negative energy condition is necessary for the positivity of the norm as seen from the norm formula , since a similar formula is valid for the virtual state wavefunction cases when @xmath91 is replaced by @xmath28 . when the first condition is dropped and the reciprocal is required to be square integrable at both boundaries , @xmath92 , @xmath93 , @xmath94 , see , such seed functions are called _ pseudo
virtual state wavefunctions_. when the system is extended in terms of a pseudo virtual state wavefunction @xmath88 , the new hamiltonian @xmath20}$ ] has an extra _ eigenstate _
@xmath95 with the eigenvalue @xmath28 , _ if the new potential is non - singular_. the extra state is below the original groundstate and @xmath20}$ ] is no longer iso - spectral with @xmath8 .
this is a consequence of .
its non - singularity is not guaranteed , either . when extended in terms of @xmath26 pseudo virtual state wavefunctions @xmath84
( @xmath10 ) , the resulting hamiltonian @xmath96}$ ] has @xmath26 additional _ eigenstates _ @xmath97}(x)$ ] , _ if the potential @xmath44}(x)$ ] is non - singular_. they are all below the original groundstate .
since @xmath88 is finite in @xmath48 , the non - square integrability can only be caused by the boundaries .
thus the virtual state wavefunctions belong to either of the following type @xmath98 and @xmath6 and the pseudo virtual state wavefunctions belong to type @xmath99 : @xmath100 an appropriate modification is needed when @xmath101 and/or @xmath102 .
hereafter we denote the pseudo state wavefunctions by @xmath103 , which are the main ingredients of this paper .
the darboux - crum transformations in terms of type @xmath98 and @xmath6 virtual state wavefunctions have been applied to achieve exactly iso - spectral deformations of the radial oscillator potential and the darboux - pschl - teller potentials @xcite , which generate the multi - indexed laguerre and jacobi polynomials . in this paper
we show that the darboux - crum transformations in terms of a set of @xmath79 pseudo virtual state wavefunctions @xmath29}\phi_n^{[m]}(x)=\mathcal{e}_n\phi_n^{[m]}(x ) , \label{vdarb}\\ & \phi_n^{[m]}(x){\stackrel{\text{def}}{=}}\frac{\text{w}[\tilde{\phi}_{d_1},\tilde{\phi}_{d_2},\ldots , \tilde{\phi}_{d_m},\phi_n](x ) } { \text{w}[\tilde{\phi}_{d_1},\tilde{\phi}_{d_2},\ldots , \tilde{\phi}_{d_m}](x)},\quad ( \phi_m^{[m]},\phi_n^{[m ] } ) = \prod_{j=1}^m(\mathcal{e}_n-\tilde{\mathcal{e}}_{d_j})\cdot h_n\delta_{m\,n } , \label{mnorm2}\\ & u^{[m]}(x){\stackrel{\text{def}}{=}}u(x)-2\partial_x^2\log \bigl|\text{w}[\tilde{\phi}_{d_1},\tilde{\phi}_{d_2},\ldots , \tilde{\phi}_{d_m}](x)\bigr| , \label{vpot}\end{aligned}\ ] ] are equivalent to a system generated by krein - adler transformations specified by a certain set @xmath104 of eigenfunctions with shifted parameters for various systems with shape - invariant potentials , which are listed in the subsequent section .
in contrast to the extensions in terms of the virtual states , these extensions in terms of pseudo virtual states are not _ iso - spectral _ and the obtained systems are not _ shape - invariant_. as seen in each example , the virtual and pseudo virtual wavefunctions are generated from the eigenstate wavefunctions through twisting of parameters based on the _ discrete symmetries _ of the original hamiltonian .
here we provide the essence of shape - invariant and thus exactly solvable one - dimensional quantum mechanical systems .
the first five examples [ sec : h][sec : k ] have infinitely many discrete eigenstates , whereas the rest [ sec : m][sec : hdpt ] has finitely many eigenstates .
they are divided into two groups of eigenfunction patterns .
the eigenfunctions @xmath105 and the pseudo virtual state wavefunctions @xmath103 @xmath106 are listed in detail for reference purposes . here
@xmath107 is the index set of the pseudo virtual state wavefunctions , which is specified for each example .
the type @xmath98 and @xmath6 virtual state wavefunctions for systems with finitely many eigenstates will be discussed in a separate paper @xcite .
the basic tools of shape - invariant systems are the forward shift and backward shift relations : @xmath108 for the parameters of shape - invariant transformation @xmath109 , we use the symbol @xmath110 to denote an increasing ( @xmath111 ) parameter and the symbol @xmath112 for a decreasing ( @xmath113 ) parameter and @xmath114 for an unchanging ( @xmath115 ) parameter , except for the darboux - pschl - teller potentials ( j ) in which @xmath110 and @xmath112 are both increasing parameters . throughout this paper
we assume that the parameters @xmath110 and/or @xmath112 take _ generic values _ , that is not integers or half odd integers . in the next subsection , we introduce two groups of the eigenfunction patterns of the eleven examples of shape - invariant systems , [ sec : h][sec : hdpt ]
. then the basics of the wronskians for the two groups are discussed in [ sec : wro ] . the explicit formulas of the eigenvalues , eigenfunctions , pseudo virtual state wavefunctions and the discrete symmetries of the eleven shape - invariant systems are provided in
[ sec : h][sec : hdpt ] for reference purposes . among them , we also report type @xmath6 virtual state wavefunctions for two potentials ( c ) [ sec : c ] and ( kh ) [ sec : kh ] .
these have been reported in @xcite and @xcite , respectively in connection with the @xmath116 rational extensions .
we stress that multi - indexed orthogonal polynomials can be constructed for these two potentials in exactly the same way as in @xcite . for another type of virtual state wavefunctions for the potentials with finitely many discrete eigenstates [ sec : m][sec : hdpt ] , see a subsequent publication @xcite . the sections [ sec : h][sec : hdpt ] could be skipped in the first reading . in all the eleven hamiltonian systems
[ sec : h][sec : hdpt ] the energy formula of the pseudo virtual state is related to that of the ` eigenstate ' with a negative ` degree ' : @xmath117 these hamiltonian systems are divided into two groups of eigenfunction patterns : @xmath118 \phi_0(x;{\boldsymbol}{\lambda}+n{\boldsymbol}{\delta})&:\text{group b } \end{array}\right .. \label{grab}\ ] ] seven potentials ( h ) , ( l ) , ( j ) , ( m ) , ( s ) , ( hst ) and ( hdpt ) belong to group a and four potentials ( c ) , ( k ) , ( rm ) and ( kh ) belong to group b. the group a has a simple structure . the ` @xmath119 ' dependence is carried only by the degree of the polynomials ( except for ( m ) ) .
they all satisfy the closure relations @xcite and @xmath56 is called a sinusoidal coordinate .
the groundstate wavefunctions of this group satisfy @xmath120 the group b has a more complicated structure .
the ` @xmath119 ' dependence appears also in the other factor and in the parameter of the polynomials @xmath121 , @xmath122 .
the pseudo virtual state wavefunctions are obtained from the eigenstate wavefunctions by twisting @xmath123 ( and @xmath124 for ( h ) and ( l ) ) : @xmath125 \phi_0\bigl(x;\mathfrak{t}({\boldsymbol}{\lambda})+\text{v}{\boldsymbol}{\delta}\bigr ) & : \text{group b } \end{array}\right.,\quad \xi_{\text{v}}(\eta;{\boldsymbol}{\lambda } ) = p_{\text{v}}\bigl(\eta;\mathfrak{t}({\boldsymbol}{\lambda})\bigr ) , \label{tphi}\end{aligned}\ ] ] with a slight modification for ( h ) and ( l ) : @xmath126 the twist operation @xmath127 satisfies @xmath128 ( @xmath129 ) . corresponding to the forward and backward shift relations of the eigenfunctions and the energy factorisation formula , those for the pseudo virtual state wavefunction read : @xmath130 based on the eigenfunction patterns , the wronskians of the eigenfunctions and the pseudo virtual state wavefunctions for the set @xmath131
@xmath132(x;{\boldsymbol}{\lambda}),\quad \text{w}[\tilde{\phi}_{d_1},\tilde{\phi}_{d_2},\ldots , \tilde{\phi}_{d_m}](x;{\boldsymbol}{\lambda})\ ] ] are reduced to determinant formulas of the polynomials .
since the latter is obtained from the former by twisting , we present derivations of the former . for group
a , the reduction to the wronskian of the polynomials is achieved by the wronskian formulas and ( @xmath133 ) : @xmath134(x;{\boldsymbol}{\lambda } ) = \phi_0(x;{\boldsymbol}{\lambda})^m \text{w}[\check{p}_{d_1},\check{p}_{d_2},\ldots,\check{p}_{d_m } ] ( x;{\boldsymbol}{\lambda}){\nonumber\\}&=\phi_0(x;{\boldsymbol}{\lambda})^m \bigl(\frac{d\eta(x)}{dx}\bigr)^{\frac12m(m-1 ) } \text{w}[p_{d_1},p_{d_2},\ldots , p_{d_m}]\bigl(\eta(x);{\boldsymbol}{\lambda}\bigr).\end{aligned}\ ] ] for group b , the derivative operator @xmath135 in the wronskian is replaced by ` covariant derivatives ' @xmath136 and use is made of the forward shift relation to obtain @xmath137 we have @xmath134(x;{\boldsymbol}{\lambda } ) = \det\bigl(\prod_{i=0}^{j-2}f_{d_k - i}({\boldsymbol}{\lambda}+i{\boldsymbol}{\delta})\cdot \phi_{d_k - j+1}\bigl(x;{\boldsymbol}{\lambda}+(j-1){\boldsymbol}{\delta}\bigr ) \bigr)_{1\leq j , k\leq m}{\nonumber\\}&=\prod_{k=1}^m\phi_0(x;{\boldsymbol}{\lambda}+d_k{\boldsymbol}{\delta})\cdot \det\bigl(\prod_{i=0}^{j-2}f_{d_k - i}({\boldsymbol}{\lambda}+i{\boldsymbol}{\delta})\cdot p_{d_k - j+1}\bigl(\eta(x);{\boldsymbol}{\lambda}+(j-1){\boldsymbol}{\delta}\bigr ) \bigr)_{1\leq j , k\leq m}.\end{aligned}\ ] ] this method is applicable to group a , too , in which @xmath138 is replaced by @xmath139 let us summarise the results : @xmath134(x;{\boldsymbol}{\lambda } ) = \bar{a}_{\mathcal{d}}(x;{\boldsymbol}{\lambda } ) \bar{\xi}_{\mathcal{d}}\bigl(\eta(x);{\boldsymbol}{\lambda}\bigr ) , \label{wphi = axi}\\ & \bar{a}_{\mathcal{d}}(x;{\boldsymbol}{\lambda}){\stackrel{\text{def}}{=}}\left\ { \begin{array}{ll } \phi_0(x;{\boldsymbol}{\lambda})^m \bigl({c_{\text{\tiny$\mathcal{f}$}}}^{-1}\frac{d\eta(x)}{dx}\bigr)^{\frac12m(m-1 ) } & : \text{group a}\\[4pt ] \prod_{k=1}^m\phi_0(x;{\boldsymbol}{\lambda}+d_k{\boldsymbol}{\delta } ) & : \text{group b } \end{array}\right.,\\ & \bar{\xi}_{\mathcal{d}}(\eta;{\boldsymbol}{\lambda}){\stackrel{\text{def}}{=}}\left\ { \begin{array}{ll } { c_{\text{\tiny$\mathcal{f}$}}}^{\,\frac12m(m-1 ) } \text{w}[p_{d_1},p_{d_2},\ldots , p_{d_m}](\eta;{\boldsymbol}{\lambda } ) & : \text{group a}\\[4pt ] \det\bigl(\bar{x}^{\mathcal{d}}_{j , k}(\eta;{\boldsymbol}{\lambda } ) \bigr)_{1\leq j , k\leq m } & : \text{group b } \end{array}\right.,\\ & \bar{x}^{\mathcal{d}}_{j , k}(\eta;{\boldsymbol}{\lambda}){\stackrel{\text{def}}{=}}\prod_{i=0}^{j-2}f_{d_k - i}({\boldsymbol}{\lambda}+i{\boldsymbol}{\delta})\cdot p_{d_k - j+1}\bigl(\eta;{\boldsymbol}{\lambda}+(j-1){\boldsymbol}{\delta}\bigr).\end{aligned}\ ] ] the pseudo virtual state wavefunction wronskian is simply obtained by the twisting : @xmath140(x;{\boldsymbol}{\lambda } ) = a_{\mathcal{d}}(x;{\boldsymbol}{\lambda } ) \xi_{\mathcal{d}}\bigl(\eta(x);{\boldsymbol}{\lambda}\bigr ) , \label{wtphi = axi}\\ & a_{\mathcal{d}}(x;{\boldsymbol}{\lambda}){\stackrel{\text{def}}{=}}\left\ { \begin{array}{ll } \tilde{\phi}_{0(0)}(x;{\boldsymbol}{\lambda})^m \bigl({c_{\text{\tiny$\mathcal{f}$}}}^{-1}\frac{d\eta(x)}{dx}\bigr)^{\frac12m(m-1 ) } & : \text{group a}\\[4pt ] \prod_{k=1}^m\phi_0\bigl(x;\mathfrak{t}({\boldsymbol}{\lambda})+d_k{\boldsymbol}{\delta}\bigr ) & : \text{group b } \end{array}\right.,\\ & \xi_{\mathcal{d}}(\eta;{\boldsymbol}{\lambda}){\stackrel{\text{def}}{=}}\left\ { \begin{array}{ll } { c_{\text{\tiny$\mathcal{f}$}}}^{\,\frac12m(m-1 ) } \text{w}[\xi_{d_1},\xi_{d_2},\ldots,\xi_{d_m}](\eta;{\boldsymbol}{\lambda } ) & : \text{group a}\\[4pt ] \det\bigl(x^{\mathcal{d}}_{j , k}(\eta;{\boldsymbol}{\lambda})\bigr)_{1\leq j , k\leq m } & : \text{group b } \end{array}\right.,\\ & x^{\mathcal{d}}_{j , k}(\eta;{\boldsymbol}{\lambda}){\stackrel{\text{def}}{=}}\prod_{i=0}^{j-2}f_{d_k - i}\bigl(\mathfrak{t}({\boldsymbol}{\lambda } ) + i{\boldsymbol}{\delta}\bigr)\cdot \xi_{d_k - j+1}\bigl(\eta;{\boldsymbol}{\lambda}-(j-1){\boldsymbol}{\delta}\bigr).\end{aligned}\ ] ] ( note that the expressions of @xmath141 and @xmath142 for group b are valid for group a , too . )
their relations are @xmath143 with a slight modification for ( h ) and ( l ) : @xmath144 both @xmath145 and @xmath146 are polynomials in @xmath147 and their degrees are generically @xmath148 : @xmath149 : strictly speaking , the notation @xmath71 of the polynomials @xmath141 and @xmath142 represents an ordered set . by changing
the order of @xmath150 s , @xmath141 and @xmath142 may change sign . on the other hand the functions @xmath151 and @xmath152 are invariant under the permutations of @xmath150 s . in order to avoid excessive appearance of @xmath153 signs in the general formulas involving @xmath141 and @xmath142 , etc , we adopt the following convention . the formulas in [ sec : main ] involving the wronskians and determinants of various polynomials depending on @xmath71 , @xmath104 and other sets are understood to be true up to a multiplicative constant @xmath154 coming from the multi - linearity of the determinants .
this does not affect the main propositions , _
e.g. _ in proposition [ polywron ] .
the well known system of the harmonic oscillator has infinitely many eigenstates : @xmath155 here @xmath156 is the hermite polynomial .
the system satisfies the _ closure relation _ introduced in @xcite .
the pseudo virtual state wavefunctions are obtained by the following discrete symmetry : @xmath157 the radial oscillator potential has also infinitely many discrete eigenstates in the specified parameter range : @xmath158 here @xmath159 is the laguerre polynomial .
the system satisfies the closure relation @xcite .
there are three types of discrete symmetries : @xmath160 and the pseudo virtual states are generated by type @xmath99 : @xmath161 the type @xmath98 and @xmath6 virtual states are obtained by using type @xmath98 and @xmath6 discrete symmetries @xcite .
the dpt potential has also infinitely many discrete eigenstates in the specified parameter range : @xmath162 here @xmath163 is the jacobi polynomial .
the system satisfies the closure relation @xcite .
there are three types of discrete symmetries : @xmath164 and the pseudo virtual states are generated by type @xmath99 : @xmath165 the type @xmath98 and @xmath6 virtual states are obtained by using type @xmath98 and @xmath6 discrete symmetries @xcite .
the hamiltonian @xmath8 has also the ` left - right ' mirror symmetry @xmath166 , @xmath167 .
the system has also infinitely many discrete eigenstates in the specified parameter range : @xmath168 the discrete symmetry and the pseudo virtual state wavefunctions are : @xmath169 for @xmath170 ( @xmath171 ) , the discrete symmetry generates the type @xmath6 virtual states @xmath172 @xcite .
this system can be obtained from the kepler problem in spherical space
[ sec : k ] in a certain limiting procedure .
the system has also infinitely many discrete eigenstates in the specified parameter range : @xmath173 the discrete symmetry and the pseudo virtual state wavefunctions are : @xmath174 the coulomb potential plus the centrifugal barrier [ sec : c ] is obtained by the following limit : @xmath175 together with the eigenfunctions .
the system has finitely many discrete eigenstates @xmath176'$ ] in the specified parameter range ( @xmath177'$ ] denotes the greatest integer not exceeding and not equal to @xmath178 ) : @xmath179 the system satisfies the closure relation @xcite .
the discrete symmetry and the pseudo virtual state wavefunctions are : @xmath180 this system can be obtained from the hyperbolic darboux - pschl - teller potential [ sec : hdpt ] in a certain limiting procedure .
the system has finitely many discrete eigenstates @xmath176'$ ] in the specified parameter range : @xmath181 the system satisfies the closure relation @xcite .
one can rewrite @xmath182 as @xmath183)_{[\frac{n+1}{2}]}}{(h - n+\frac12)_{[\frac{n+1}{2 } ] } } i^np_n^{(-h-\frac12,-h-\frac12)}(i\eta),\ ] ] where @xmath177 $ ] denotes the greatest integer not exceeding @xmath178 . the discrete symmetry and the pseudo virtual state wavefunctions are : @xmath184 this system can be obtained from the rosen - morse potential [ sec : rm ] by taking @xmath185 limit .
this potential is also called rosen - morse @xmath6 potential .
the system has finitely many discrete eigenstates @xmath186'$ ] in the specified parameter range : @xmath187 by taking the limit @xmath185 , the soliton potential [ sec : s ] is obtained .
based on the symmetry @xmath188 , @xmath189 , positive @xmath114 is selected .
the discrete symmetry and the pseudo virtual state wavefunctions are : @xmath190 the system has finitely many discrete eigenstates @xmath176'$ ] in the specified parameter range : @xmath191 the system satisfies the closure relation @xcite .
the discrete symmetry and the pseudo virtual state wavefunctions are : @xmath192 this potential is also called eckart potential .
it has finitely many discrete eigenstates @xmath193'$ ] in the specified parameter range : @xmath194 the discrete symmetry and the pseudo virtual state wavefunctions are : @xmath195 for @xmath170 ( @xmath171 ) , the discrete symmetry generates the type @xmath6 virtual states @xmath172 @xcite .
it has finitely many discrete eigenstates @xmath196'$ ] in the specified parameter range : @xmath197 the eigenvalues can be also expressed as @xmath198 .
the system satisfies the closure relation @xcite .
three types of discrete symmetries are : @xmath199 the pseudo virtual state wavefunctions are generated by type @xmath99 : @xmath200 the type @xmath98 and @xmath6 virtual states are obtained by using type @xmath98 and @xmath6 discrete symmetries .
the morse potential [ sec : m ] is obtained by the following limit : @xmath201 together with the eigenfunctions .
let us introduce appropriate symbols and notation for stating the main results .
let @xmath79 ( @xmath80 ) be a set of distinct non - negative integers .
we introduce an integer @xmath202 and fix it to be not less than the maximum of @xmath71 : @xmath203 let us define a set of distinct non - negative integers @xmath204 together with the shifted parameters @xmath205 : @xmath206 starting from the well - defined original system , the system after darboux - crum transformations in terms of a set of pseudo virtual state wavefunctions @xmath71 is described by the hamiltonian @xmath207 @xmath208(x;{\boldsymbol}{\lambda})\bigr| .
\label{hdc}\end{aligned}\ ] ] general theory presented in
[ sec : darb - crum ] states that , if the hamiltonian @xmath207 is non - singular , the eigenstates are given by @xmath209 and @xmath210 : @xmath211(x;{\boldsymbol}{\lambda } ) } { \text{w}[\tilde{\phi}_{d_1},\tilde{\phi}_{d_2},\ldots , \tilde{\phi}_{d_m}](x;{\boldsymbol}{\lambda } ) } \quad(n=0,1,\ldots , n_{\text{max}}({\boldsymbol}{\lambda})\ \text{or}\ \infty),{\nonumber\\}&\breve{\phi}^{\text{dc}}_j(x)= \frac{\text{w}[\tilde{\phi}_{d_1},\tilde{\phi}_{d_2},\ldots , \breve{\tilde{\phi}}_{d_j},\ldots,\tilde{\phi}_{d_m}](x;{\boldsymbol}{\lambda } ) } { \text{w}[\tilde{\phi}_{d_1},\tilde{\phi}_{d_2},\ldots , \tilde{\phi}_{d_m}](x;{\boldsymbol}{\lambda } ) } \quad(j=1,2,\ldots , m),{\nonumber\\}&\mathcal{h}^{\text{dc}}\phi^{\text{dc}}_n(x ) = \mathcal{e}_n({\boldsymbol}{\lambda})\phi^{\text{dc}}_n(x),\quad \mathcal{h}^{\text{dc}}\breve{\phi}^{\text{dc}}_j(x ) = \mathcal{e}_{-d_j-1}({\boldsymbol}{\lambda})\breve{\phi}^{\text{dc}}_j(x ) .
\label{dcdiffeq}\end{aligned}\ ] ] the system after krein - adler transformations in terms of @xmath104 with shifted parameters @xmath205 is described by the hamiltonian @xmath212 @xmath213(x;\bar{{\boldsymbol}{\lambda}})\bigr| .
\label{hka}\end{aligned}\ ] ] here we assume that the original system with the shifted parameters @xmath205 has square integrable eigenstates , etc .
if the krein - adler conditions are fulfilled , eigenstates are given by @xmath214 and @xmath215 : @xmath216 ( x;\bar{{\boldsymbol}{\lambda } } ) } { \text{w}[\phi_0,\phi_1,\ldots,\breve{\phi}_{\bar{d}_1},\ldots , \breve{\phi}_{\bar{d}_m},\ldots,\phi_n](x;\bar{{\boldsymbol}{\lambda } } ) } \quad(n=0,1,\ldots , n_{\text{max}}({\boldsymbol}{\lambda})\ \text{or}\ \infty),{\nonumber\\}&\breve{\phi}^{\text{ka}}_j(x)= \frac{\text{w}[\phi_0,\phi_1,\ldots,\breve{\phi}_{\bar{d}_1 } , \ldots,\phi_{\bar{d}_j},\ldots,\breve{\phi}_{\bar{d}_m},\ldots,\phi_n ] ( x;\bar{{\boldsymbol}{\lambda } } ) } { \text{w}[\phi_0,\phi_1,\ldots,\breve{\phi}_{\bar{d}_1},\ldots , \breve{\phi}_{\bar{d}_m},\ldots,\phi_n](x;\bar{{\boldsymbol}{\lambda } } ) } \quad(j=1,2,\ldots , m),{\nonumber\\}&\mathcal{h}^{\text{ka}}\phi^{\text{ka}}_n(x ) = \mathcal{e}_{n+1+n}(\bar{{\boldsymbol}{\lambda}})\phi^{\text{ka}}_n(x),\quad \mathcal{h}^{\text{ka}}\breve{\phi}^{\text{ka}}_j(x ) = \mathcal{e}_{\bar{d}_j}(\bar{{\boldsymbol}{\lambda}})\breve{\phi}^{\text{ka}}_j(x ) .
\label{kadiffeq}\end{aligned}\ ] ] note that @xmath217 in [ sec : m][sec : hdpt ] satisfies @xmath218 the differential equations @xmath219 and ( with @xmath220 ) hold irrespective of the non - singularity of @xmath207 and @xmath212 .
our main results hold for any one of the shape - invariant quantum mechanical systems in [ sec : h][sec : hdpt ] . [
main ] the two systems with @xmath221 and @xmath222 are equivalent .
to be more specific , the equality of the potentials and the eigenfunctions read : @xmath223 the singularity free conditions of the potential are @xmath224 for @xmath116 , @xmath225 , @xmath226 , the above conditions are satisfied by even @xmath227 , @xmath228 .
in other words , the pseudo virtual state wavefunctions @xmath229 for even @xmath230 are nodeless .
the above equalities ( up to multiplicative factors )
( with @xmath220 ) are algebraic and they hold irrespective of the non - singularity conditions . the following energy formulas satisfied by the eleven systems @xmath231 and the illustration ( figure 1 ) would be helpful to understand the proposition .
the form of the energy curve is that of the dpt potential but the situation is similar in all other ten potentials . if the equality of the potentials is shown , then , under the condition , the hamiltonians @xmath232 are non - singular and it implies and . therefore the remaining task is to show .
the basic ingredients of the proposition [ main ] are the two wronskians @xmath233 ( x;{\boldsymbol}{\lambda}),\quad \text{w}[\phi_0,{\phi}_{1},\ldots,\breve{\phi}_{\bar{d}_1},\ldots , \breve{\phi}_{\bar{d}_m},\ldots,\phi_n](x;\bar{{\boldsymbol}{\lambda}}),\ ] ] which determine the potentials and eventually all the eigenfunctions . by using the wronskian formulas and they are expressed by the determinants of polynomials : @xmath140(x;{\boldsymbol}{\lambda } ) = a_{\mathcal{d}}(x;{\boldsymbol}{\lambda } ) \xi_{\mathcal{d}}\bigl(\eta(x);{\boldsymbol}{\lambda}\bigr ) , \label{eigwron1}\\ & \text{w}[\phi_0,{\phi}_{1},\ldots,\breve{\phi}_{\bar{d}_1},\ldots , \breve{\phi}_{\bar{d}_m},\ldots,\phi_n](x;\bar{{\boldsymbol}{\lambda } } ) = \bar{a}_{\bar{\mathcal{d}}}(x;\bar{{\boldsymbol}{\lambda } } ) \bar{\xi}_{\bar{\mathcal{d}}}\bigl(\eta(x);\bar{{\boldsymbol}{\lambda}}\bigr ) .
\label{eigwron2}\end{aligned}\ ] ] these two polynomials @xmath146 and @xmath234 have the same degree @xmath235 .
an important relation between @xmath236 and @xmath237 is that its ratio is independent of @xmath26 and @xmath150 in each of the eleven potentials [ sec : h][sec : hdpt ] , which is shown by straightforward calculation .
we set the ratio as @xmath238 : @xmath239 from this independence and the obvious relation @xmath240 , the ratio @xmath241 can be expressed as @xmath242 it is easy to show the following identity for any of the eleven systems : @xmath243 by this identity , the equality of the potentials reduces to @xmath244 since both polynomials are of the same degree @xmath148 , this means that the two polynomials are proportional .
we state this as [ polywron ] two polynomials characterising @xmath221 and @xmath222 , @xmath146 and @xmath234 , are proportional : @xmath245 in particular , it means _
polynomial wronskian identities _
@xmath246(\eta;{\boldsymbol}{\lambda } ) \propto \text{w}[p_0,p_{1},\ldots,\breve{p}_{\bar{d}_1},\ldots , \breve{p}_{\bar{d}_m},\ldots , p_n](\eta;\bar{{\boldsymbol}{\lambda } } ) , \label{genwronide}\ ] ] for all the systems in group a. recall that @xmath247 ( with a slight modification for ( h ) and ( l ) , ) . for the simplest case of @xmath116 , @xmath248 ( @xmath249 @xmath250 , @xmath251 ) ,
the wronskian identities are reported as ( a.22 ) in @xcite for the three types of the classical orthogonal polynomials , the hermite ( h ) , laguerre ( l ) and jacobi ( j ) . to the best of our knowledge , the general polynomial wronskian identities have not been reported before .
the proof of proposition [ polywron ] is done by induction in @xmath26 .
: in the first step we prove for @xmath116 , @xmath252 : @xmath253 recall that the differential equations and hold , and @xmath254(x;\bar{{\boldsymbol}{\lambda } } ) = \bar{a}_{\{0,1,\ldots , n\}}(x;\bar{{\boldsymbol}{\lambda } } ) \bar{\xi}_{\{0,1,\ldots , n\}}(\eta;\bar{{\boldsymbol}{\lambda}}),{\nonumber\\}&\text{w}[\phi_0,\phi_1,\ldots,\breve{\phi}_{\bar{\text{v}}},\ldots,\phi_n ] ( x;\bar{{\boldsymbol}{\lambda } } ) = \bar{a}_{\{0,1,\ldots,\breve{\bar{\text{v}}},\ldots , n\ } } ( x;\bar{{\boldsymbol}{\lambda } } ) \bar{\xi}_{\{0,1,\ldots,\breve{\bar{\text{v}}},\ldots , n\ } } ( \eta;\bar{{\boldsymbol}{\lambda } } ) .
\label{m1eq2}\end{aligned}\ ] ] by substituting into the differential equation @xmath255 , it becomes @xmath256 where @xmath257 . by substituting into the differential equation @xmath258 and using @xmath259 ( see ) ,
we obtain @xmath260 where @xmath261 , @xmath262 and @xmath263 . by using and ,
we find @xmath264 .
the above equation is rewritten as @xmath265 from and , we have @xmath266 thus @xmath257 and @xmath261 satisfy the same differential equation .
since both of them are polynomials of degree @xmath267 in @xmath147 , they should be proportional for generic parameters @xcite . :
assume that holds till @xmath26 ( @xmath268 ) , we will show that it also holds for @xmath269 . before presenting a general proof
, we illustrate the outline by taking the simple case of group a .
we have shown @xmath116 case : @xmath270 ( \eta;\bar{{\boldsymbol}{\lambda}}).\ ] ] by using the algebraic wronskian identity , we have @xmath271(\eta;{\boldsymbol}{\lambda } ) & \propto\text{w}\bigl [ \text{w}[p_0,\ldots,\breve{p}_{\bar{d}_1},\ldots , p_n ] , \text{w}[p_0,\ldots,\breve{p}_{\bar{d}_2},\ldots , p_n ] \bigr](\eta;\bar{{\boldsymbol}{\lambda}})\\ & = \pm \text{w}[p_0,\ldots,\breve{p}_{\bar{d}_1},\ldots,\breve{p}_{\bar{d}_2 } , \ldots , p_n](\eta;\bar{{\boldsymbol}{\lambda}})\cdot \text{w}[p_0,p_1,\ldots , p_n](\eta;\bar{{\boldsymbol}{\lambda}})\\ & \propto \text{w}[p_0,\ldots,\breve{p}_{\bar{d}_1},\ldots,\breve{p}_{\bar{d}_2 } , \ldots , p_n](\eta;\bar{{\boldsymbol}{\lambda}}),\end{aligned}\ ] ] where we have used @xmath272(\eta;\bar{{\boldsymbol}{\lambda}})=\text{constant}$ ] ( see ) .
this is the @xmath273 result . for @xmath274 , we use the @xmath273 results to obtain @xmath275(\eta;{\boldsymbol}{\lambda})\\ & = \text{w}\bigl [ \text{w}[\xi_{d_1},\xi_{d_2}],\text{w}[\xi_{d_1},\xi_{d_3}]\bigr ] \bigl(\eta;{\boldsymbol}{\lambda})\\ & \propto\text{w}\bigl [ \text{w}[p_0,\ldots,\breve{p}_{\bar{d}_1},\ldots,\breve{p}_{\bar{d}_2 } , \ldots , p_n ] , \text{w}[p_0,\ldots,\breve{p}_{\bar{d}_1},\ldots,\breve{p}_{\bar{d}_3 } , \ldots , p_n]\bigl](\eta;\bar{{\boldsymbol}{\lambda}})\\ & = \pm \text{w}[p_0,\ldots,\breve{p}_{\bar{d}_1},\ldots,\breve{p}_{\bar{d}_2 } , \ldots,\breve{p}_{\bar{d}_3},\ldots , p_n](\eta;\bar{{\boldsymbol}{\lambda}})\cdot \text{w}[p_0,\ldots,\breve{p}_{\bar{d}_1},\ldots , p_n ] ( \eta;\bar{{\boldsymbol}{\lambda}})\\ & \propto \text{w}[p_0,\ldots,\breve{p}_{\bar{d}_1},\ldots,\breve{p}_{\bar{d}_2 } , \ldots,\breve{p}_{\bar{d}_3},\ldots , p_n](\eta;\bar{{\boldsymbol}{\lambda } } ) \cdot\xi_{d_1}(\eta;{\boldsymbol}{\lambda}).\end{aligned}\ ] ] this establishes the identities for @xmath274 .
higher @xmath26 identities follow in a similar way .
let us present a general proof .
is equivalent to @xmath276(x;{\boldsymbol}{\lambda } ) } { a_{\mathcal{d}}(x;{\boldsymbol}{\lambda } ) } \propto \frac{\text{w}[\phi_0,{\phi}_{1},\ldots,\breve{\phi}_{\bar{d}_1},\ldots , \breve{\phi}_{\bar{d}_m},\ldots,\phi_n](x;\bar{{\boldsymbol}{\lambda } } ) } { \bar{a}_{\bar{\mathcal{d}}}(x;\bar{{\boldsymbol}{\lambda}})},\ ] ] which implies @xmath277(x;{\boldsymbol}{\lambda } ) \propto f(x , n,{\boldsymbol}{\lambda } ) \text{w}[\phi_0,{\phi}_{1},\ldots,\breve{\phi}_{\bar{d}_1},\ldots , \breve{\phi}_{\bar{d}_m},\ldots,\phi_n](x;\bar{{\boldsymbol}{\lambda}}).\ ] ] assume that holds till @xmath26 ( @xmath268 ) . by using the wronskian identity
, we obtain @xmath278 ( x;{\boldsymbol}{\lambda } ) \cdot\text{w}[\tilde{\phi}_{d_1},\ldots,\tilde{\phi}_{d_{m-1 } } , \tilde{\phi}_{d_m},\tilde{\phi}_{d_{m+1}}](x;{\boldsymbol}{\lambda})\\ & = \text{w}\bigl [ \text{w}[\tilde{\phi}_{d_1},\ldots,\tilde{\phi}_{d_{m-1 } } , \tilde{\phi}_{d_m } ] , \text{w}[\tilde{\phi}_{d_1},\ldots,\tilde{\phi}_{d_{m-1 } } , \tilde{\phi}_{d_{m+1}}]\bigr](x;{\boldsymbol}{\lambda})\\ & \propto\text{w}\bigl[f(x , n,{\boldsymbol}{\lambda } ) \text{w}[\phi_0,\ldots,\breve{\phi}_{\bar{d}_1},\ldots , \breve{\phi}_{\bar{d}_{m-1}},\ldots,\breve{\phi}_{\bar{d}_m},\ldots , \phi_n],\\ & \phantom{\propto\text{w}\bigl [ } \
f(x , n,{\boldsymbol}{\lambda } ) \text{w}[\phi_0,\ldots,\breve{\phi}_{\bar{d}_1},\ldots , \breve{\phi}_{\bar{d}_{m-1}},\ldots,\breve{\phi}_{\bar{d}_{m+1}},\ldots , \phi_n]\bigr](x;\bar{{\boldsymbol}{\lambda}})\\ & = f(x , n,{\boldsymbol}{\lambda})^2\text{w}\bigl [ \text{w}[\phi_0,\ldots,\breve{\phi}_{\bar{d}_1},\ldots , \breve{\phi}_{\bar{d}_{m-1}},\ldots,\breve{\phi}_{\bar{d}_m},\ldots , \phi_n],\\ & \phantom{=f(x , n,{\boldsymbol}{\lambda})^2\text{w}\bigl [ } \ \text{w}[\phi_0,\ldots,\breve{\phi}_{\bar{d}_1},\ldots , \breve{\phi}_{\bar{d}_{m-1}},\ldots,\breve{\phi}_{\bar{d}_{m+1}},\ldots , \phi_n]\bigr](x;\bar{{\boldsymbol}{\lambda}})\\ & = \pm f(x , n,{\boldsymbol}{\lambda})^2 \text{w}[\phi_0,\ldots,\breve{\phi}_{\bar{d}_1},\ldots , \breve{\phi}_{\bar{d}_{m-1}},\ldots , \breve{\phi}_{\bar{d}_m},\ldots , \breve{\phi}_{\bar{d}_{m+1}},\ldots,\phi_n](x;\bar{{\boldsymbol}{\lambda}})\\ & \phantom{=f(x , n,{\boldsymbol}{\lambda})^2 } \times\text{w}[\phi_0,\ldots,\breve{\phi}_{\bar{d}_1},\ldots , \breve{\phi}_{\bar{d}_{m-1}},\ldots,\phi_n](x;\bar{{\boldsymbol}{\lambda}})\\ & \propto f(x , n,{\boldsymbol}{\lambda } ) \text{w}[\phi_0,\ldots,\breve{\phi}_{\bar{d}_1},\ldots , \breve{\phi}_{\bar{d}_{m+1}},\ldots,\phi_n](x;\bar{{\boldsymbol}{\lambda}})\cdot \text{w}[\tilde{\phi}_{d_1},\ldots,\tilde{\phi}_{d_{m-1}}](x;{\boldsymbol}{\lambda}).\end{aligned}\ ] ] this leads to @xmath279 ( x;{\boldsymbol}{\lambda } ) } { a_{\{d_1,\ldots , d_{m+1}\}}(x;{\boldsymbol}{\lambda})}{\nonumber\\}&\propto \frac{f(x , n,{\boldsymbol}{\lambda } ) \bar{a}_{\{0,\ldots,\bar{d}_1,\ldots,\bar{d}_{m+1},\ldots , n\ } } ( x;\bar{{\boldsymbol}{\lambda } } ) } { a_{\{d_1,\ldots , d_{m+1}\}}(x;{\boldsymbol}{\lambda } ) } \frac{\text{w}[\phi_0,\ldots,\breve{\phi}_{\bar{d}_1},\ldots , \breve{\phi}_{\bar{d}_{m+1}},\ldots,\phi_n](x;\bar{{\boldsymbol}{\lambda } } ) } { \bar{a}_{\{0,\ldots,\bar{d}_1,\ldots,\bar{d}_{m+1},\ldots , n\ } } ( x;\bar{{\boldsymbol}{\lambda}})}{\nonumber\\}&=\frac{\text{w}[\phi_0,\ldots,\breve{\phi}_{\bar{d}_1},\ldots , \breve{\phi}_{\bar{d}_{m+1}},\ldots,\phi_n](x;\bar{{\boldsymbol}{\lambda } } ) } { \bar{a}_{\{0,\ldots,\bar{d}_1,\ldots,\bar{d}_{m+1},\ldots , n\ } } ( x;\bar{{\boldsymbol}{\lambda}})},\end{aligned}\ ] ] which means @xmath280 this concludes the induction proof of and the proof of propositions [ main ] and [ polywron ] is completed . in order to obtain the rational forms ( the ratios of polynomials ) of the eigenfunctions in proposition
[ main ] , we need the wronskian expressions of the numerators : @xmath281(x;{\boldsymbol}{\lambda } ) = { a}_{\mathcal{d},n}(x;{\boldsymbol}{\lambda } ) p_{\mathcal{d},n}\bigl(\eta(x);{\boldsymbol}{\lambda}\bigr),{\nonumber\\}&a_{\mathcal{d},n}(x;{\boldsymbol}{\lambda})=\left\ { \begin{array}{ll } \tilde{\phi}_{0(0)}(x;{\boldsymbol}{\lambda})^m \phi_0(x;{\boldsymbol}{\lambda } ) \bigl({c_{\text{\tiny$\mathcal{f}$}}}^{-1}\frac{d\eta(x)}{dx}\bigr)^{\frac12m(m-1)-m } & : \text{group a}\\[4pt ] \prod_{k=1}^m\phi_0\bigl(x;\mathfrak{t}({\boldsymbol}{\lambda})+d_k{\boldsymbol}{\delta}\bigr ) \cdot\phi_0(x;{\boldsymbol}{\lambda}+n{\boldsymbol}{\delta } ) & : \text{group b } \end{array}\right .. \end{aligned}\ ] ] here @xmath282 is a polynomial in @xmath147 and its degree is generically @xmath283 .
this leads to @xmath284 ( x;{\boldsymbol}{\lambda } ) } { \text{w}[\tilde{\phi}_{d_1},\ldots,\tilde{\phi}_{d_m}](x;{\boldsymbol}{\lambda } ) } = \frac{p_{\mathcal{d},n}\bigl(\eta(x);{\boldsymbol}{\lambda}\bigr ) } { \xi_{\mathcal{d}}\bigl(\eta(x);{\boldsymbol}{\lambda}\bigr ) } \times\left\ { \begin{array}{ll } \phi_0(x;{\boldsymbol}{\lambda}-m{\boldsymbol}{\delta } ) & : \text{group a}\\[2pt ] \phi_0(x;{\boldsymbol}{\lambda}+n{\boldsymbol}{\delta } ) & : \text{group b } \end{array}\right .. \ ] ] eq .
leads to @xmath285(x;{\boldsymbol}{\lambda } ) } { \text{w}[\phi_{d_1},\ldots,\phi_{d_m}](x;{\boldsymbol}{\lambda } ) } = \frac{\bar{p}_{\mathcal{d},n}\bigl(\eta(x);{\boldsymbol}{\lambda}\bigr ) } { \bar{\xi}_{\mathcal{d}}\bigl(\eta(x);{\boldsymbol}{\lambda}\bigr ) } \times\left\ { \begin{array}{ll } \phi_0(x;{\boldsymbol}{\lambda}+m{\boldsymbol}{\delta } ) & : \text{group a}\\[2pt ] \phi_0(x;{\boldsymbol}{\lambda}+n{\boldsymbol}{\delta } ) & : \text{group b } \end{array}\right.,{\nonumber\\}&\bar{p}_{\mathcal{d},n}(\eta;{\boldsymbol}{\lambda}){\stackrel{\text{def}}{=}}\bar{\xi}_{\{d_1,\ldots , d_m , n\}}(\eta;{\boldsymbol}{\lambda } ) \ \ : \
\text{degree $ = \ell_{\{d_1,\ldots , d_m , n\}}=~\ell_{\mathcal{d}}-m+n$}.\end{aligned}\ ] ] in the present case we have @xmath286 ( x;\bar{{\boldsymbol}{\lambda } } ) } { \text{w}[\phi_{e_1},\ldots,\phi_{e_{n+1-m}}](x;\bar{{\boldsymbol}{\lambda } } ) } = \frac{\bar{p}_{\bar{\mathcal{d}},n+1+n } \bigl(\eta(x);\bar{{\boldsymbol}{\lambda}}\bigr ) } { \bar{\xi}_{\bar{\mathcal{d}}}\bigl(\eta(x);\bar{{\boldsymbol}{\lambda}}\bigr ) } \times\left\ { \begin{array}{ll } \phi_0(x;{\boldsymbol}{\lambda}-m{\boldsymbol}{\delta } ) & : \text{group a}\\[2pt ] \phi_0(x;{\boldsymbol}{\lambda}+n{\boldsymbol}{\delta } ) & : \text{group b } \end{array}\right.,\ ] ] where the degree of @xmath287 is @xmath288 .
therefore implies @xmath289 similarly implies @xmath290 let us introduce @xmath291 based on @xmath131 : @xmath292 the newly introduced polynomials at @xmath32 are related to the old ones : @xmath293 based on these formulas , the following relations follow : @xmath294(x;{\boldsymbol}{\lambda } ) } { \text{w}[\phi_{d_1},\ldots,\phi_{d_m}](x;{\boldsymbol}{\lambda})}\biggr| \quad(\min_j{d_j}\geq 2),{\nonumber\\}&\quad\bigl(\partial_x\bar{f}_{\mathcal{d}}(x;{\boldsymbol}{\lambda})\bigr)^2 -\partial_x^2\bar{f}_{\mathcal{d}}(x;{\boldsymbol}{\lambda } ) = \bigl(\partial_x\bar{f}_{\mathcal{d}_- } ( x;{\boldsymbol}{\lambda}+{\boldsymbol}{\delta})\bigr)^2 + \partial_x^2\bar{f}_{\mathcal{d}_-}(x;{\boldsymbol}{\lambda}+{\boldsymbol}{\delta } ) + \mathcal{e}_1({\boldsymbol}{\lambda } ) , \label{fdsi-}\\ & f_{\mathcal{d}}(x;{\boldsymbol}{\lambda } ) { \stackrel{\text{def}}{=}}\log\biggl| \frac{\text{w}[\tilde{\phi}_{d_1},\ldots,\tilde{\phi}_{d_m},\phi_0 ] ( x;{\boldsymbol}{\lambda } ) } { \text{w}[\tilde{\phi}_{d_1},\ldots,\tilde{\phi}_{d_m}](x;{\boldsymbol}{\lambda } ) } \biggr|,{\nonumber\\}&\quad\bigl(\partial_xf_{\mathcal{d}}(x;{\boldsymbol}{\lambda})\bigr)^2 -\partial_x^2f_{\mathcal{d}}(x;{\boldsymbol}{\lambda } ) = \bigl(\partial_xf_{\mathcal{d}_+}(x;{\boldsymbol}{\lambda}+{\boldsymbol}{\delta})\bigr)^2 + \partial_x^2f_{\mathcal{d}_+}(x;{\boldsymbol}{\lambda}+{\boldsymbol}{\delta } ) + \mathcal{e}_1({\boldsymbol}{\lambda } ) , \label{fdsi+}\end{aligned}\ ] ] which have the same forms as the shape - invariance relation but they do not mean the shape - invariance . the @xmath116 case for ( m ) , ( rm ) and ( kh ) was presented as ` enlarged ' shape - invariance in @xcite .
in the rest of this section we provide the proofs of . for , we note that the forward shift relation can be written as @xmath295 by using this , we obtain @xmath296(x;{\boldsymbol}{\lambda}){\nonumber\\}&=\phi_0(x;{\boldsymbol}{\lambda})^{m+1 } \text{w}\bigl[\frac{\phi_{d_1}}{\phi_0},\ldots,\frac{\phi_{d_m}}{\phi_0},1 \bigr](x;{\boldsymbol}{\lambda}){\nonumber\\}&=\phi_0(x;{\boldsymbol}{\lambda})^{m+1}(-1)^m \text{w}\bigl[\frac{d}{dx}\frac{\phi_{d_1}}{\phi_0},\ldots , \frac{d}{dx}\frac{\phi_{d_m}}{\phi_0}\bigr](x;{\boldsymbol}{\lambda}){\nonumber\\}&=\phi_0(x;{\boldsymbol}{\lambda})^{m+1}(-1)^m \text{w}\bigl[f_{d_1}({\boldsymbol}{\lambda } ) \frac{\phi_{d_1 - 1}(x;{\boldsymbol}{\lambda}+{\boldsymbol}{\delta})}{\phi_0(x;{\boldsymbol}{\lambda } ) } , \ldots , f_{d_m}({\boldsymbol}{\lambda } ) \frac{\phi_{d_m-1}(x;{\boldsymbol}{\lambda}+{\boldsymbol}{\delta})}{\phi_0(x;{\boldsymbol}{\lambda } ) } \bigr](x){\nonumber\\}&=(-1)^m\prod_{j=1}^mf_{d_j}({\boldsymbol}{\lambda})\cdot \phi_0(x;{\boldsymbol}{\lambda } ) \text{w}[\phi_{d_1 - 1},\ldots,\phi_{d_m-1}](x;{\boldsymbol}{\lambda}+{\boldsymbol}{\delta}).\end{aligned}\ ] ] this can be rewritten as @xmath297 and the fact that its right hand side is a constant @xmath298 can be verified for any of the systems listed in [ sec : h][sec : hdpt ] .
this concludes the proof of .
we remark that is @xmath299 , which implies @xmath300 for , we note that the forward shift relation for @xmath172 can be rewritten as @xmath301 by using this , we obtain @xmath302 ( x;{\boldsymbol}{\lambda}){\nonumber\\}&=\phi_0(x;{\boldsymbol}{\lambda})^{m+1 } \text{w}\bigl[\frac{\tilde{\phi}_{d_1}}{\phi_0},\ldots , \frac{\tilde{\phi}_{d_m}}{\phi_0},1\bigr](x;{\boldsymbol}{\lambda}){\nonumber\\}&=\phi_0(x;{\boldsymbol}{\lambda})^{m+1}(-1)^m \text{w}\bigl[\frac{d}{dx}\frac{\tilde{\phi}_{d_1}}{\phi_0},\ldots , \frac{d}{dx}\frac{\tilde{\phi}_{d_m}}{\phi_0}\bigr](x;{\boldsymbol}{\lambda}){\nonumber\\}&=\phi_0(x;{\boldsymbol}{\lambda})^{m+1}(-1)^m \text{w}\bigl[-\epsilon\,b_{d_1}(-{\boldsymbol}{\lambda } ) \frac{\tilde{\phi}_{d_1 + 1}(x;{\boldsymbol}{\lambda}+{\boldsymbol}{\delta } ) } { \phi_0(x;{\boldsymbol}{\lambda } ) } , \ldots,-\epsilon\,b_{d_m}(-{\boldsymbol}{\lambda } ) \frac{\tilde{\phi}_{d_m+1}(x;{\boldsymbol}{\lambda}+{\boldsymbol}{\delta } ) } { \phi_0(x;{\boldsymbol}{\lambda})}\bigr](x){\nonumber\\}&=\prod_{j=1}^m\epsilon\,b_{d_j}(-{\boldsymbol}{\lambda})\cdot \phi_0(x;{\boldsymbol}{\lambda } ) \text{w}[\tilde{\phi}_{d_1 + 1},\ldots,\tilde{\phi}_{d_m+1 } ] ( x;{\boldsymbol}{\lambda}+{\boldsymbol}{\delta}).\end{aligned}\ ] ] this can be rewritten as @xmath303 and the fact that its right hand side is a constant @xmath304 can be verified for any of the systems listed in [ sec : h][sec : hdpt ] .
this concludes the proof of .
in the context of rational extensions of solvable potentials , the concept of the pseudo virtual state wavefunctions is introduced .
they are obtained by relaxing two conditions of the virtual state wavefunctions ; the reciprocals are square - integrable at both boundaries and they need not be nodeless .
a darboux - crum transformation in terms of a pseudo virtual state wavefunction will produce a new eigenstate below the original ground state .
the same system can be derived by a special type of krein - adler transformation with negatively shifted parameters .
the main results of the paper is the equivalence of darboux - crum transformation in terms of multiple pseudo virtual states and krein - adler transformations in terms of multiple eigenstates with negatively shifted parameters .
this is based on polynomial wronskian identities , which are generalisations of those reported by the present authors @xcite a few years ago .
the equivalence holds for most of the known shape - invariant potentials consisting of eleven explicit examples having finite as well as infinite discrete eigenstates .
the type @xmath6 virtual state wavefunctions have been obtained by the discrete symmetries for two potentials ( c ) [ sec : c ] and ( kh ) [ sec : kh ] .
they have been used in the context of @xmath116 rational extensions of these potentials in @xcite and @xcite .
multi - indexed extensions of these two potentials can be constructed in exactly the same way as in @xcite . in a separate publication @xcite
, we will discuss rational extensions in terms of genuine virtual state wavefunctions for shape - invariant potentials having finitely many discrete eigenstates .
they have different features from those having infinitely many eigenstates , which have been explored in connection with the multi - indexed orthogonal polynomials @xcite .
the multi - indexed laguerre and jacobi orthogonal polynomials are labeled by the multi - index @xmath71 , but different multi - index sets may give the same multi - indexed polynomials , e.g. eqs.(50)(51 ) in @xcite . the proposition [ polywron ]
gives its generalisation . by applying the twist based on the type @xmath6 discrete symmetry to
, the l.h.s becomes the denominator polynomial with multiple type @xmath98 virtual state deletion and the r.h.s .
becomes that of type @xmath6 . after completing the manuscript
, we became aware of a recent work @xcite , which discusses some rational extensions of the harmonic oscillator .
they correspond to some special examples of the equivalence for the harmonic oscillator ( h ) and @xmath273 .
r.s . is supported in part by grant - in - aid for scientific research from the ministry of education , culture , sports , science and technology ( mext ) , no.22540186 .
y.grandati , `` solvable rational extensions of the isotonic oscillator , '' ann
* 326 * ( 2011 ) 2074 - 2090 , arxiv:1101.0055[math - ph ] ; solvable rational extensions of the morse and kepler - coulomb potentials , " j. math .
* 52 * ( 2011 ) 103505 ( 12pp ) , arxiv:1103.5023[math - ph ] ; `` multistep dbt and regular rational extensions of the isotonic oscillator , '' ann
* 327 * ( 2012 ) 2411 - 2431 , arxiv:1108.4503[math - ph ] ; `` new rational extensions of solvable potentials with finite bound state spectrum , '' phys .
a376 * ( 2012 ) 2866 - 2872 , arxiv:1203.4149[math - ph ] .
m.g.krein , on continuous analogue of a formula of christoffel from the theory of orthogonal polynomials , " ( russian ) doklady acad .
cccp , * 113 * ( 1957 ) 970 - 973 ; v..adler , `` a modification of crum s method , '' theor
* 101 * ( 1994 ) 1381 - 1386 .
s.odake and r.sasaki , `` unified theory of annihilation - creation operators for solvable ( ` discrete ' ) quantum mechanics , '' j. math . phys . * 47 * ( 2006 ) 102102 ( 33pp ) , arxiv : quant - ph/0605215 ; exact solution in the heisenberg picture and annihilation - creation operators , " phys
. lett . * b641 * ( 2006 ) 112 - 117 , arxiv : quant - ph/0605221 .
i.marquette and c.quesne , two - step rational extensions of the harmonic oscillator : exceptional orthogonal polynomials and ladder operators , " j. phys .
* a46 * ( 2013 ) 155201 ( 14pp ) , arxiv:1212.3474[math - ph ] . | for eleven examples of one - dimensional quantum mechanics with shape - invariant potentials , the darboux - crum transformations in terms of multiple _ pseudo virtual state wavefunctions _ are shown to be equivalent to krein - adler transformations deleting multiple eigenstates with _ shifted parameters_. these are based upon infinitely many polynomial wronskian identities of classical orthogonal polynomials , _
i.e. _ the hermite , laguerre and jacobi polynomials , which constitute the main part of the eigenfunctions of various quantum mechanical systems with shape - invariant potentials .
yukawa institute kyoto + dpsu-12 - 3 + yitp-12 - 85 + * krein - adler transformations for shape - invariant potentials and pseudo virtual states + + * * satoru odake@xmath0 and ryu sasaki@xmath1 * @xmath2 department of physics , shinshu university , + matsumoto 390 - 8621 , japan + @xmath1 yukawa institute for theoretical physics , + kyoto university , kyoto 606 - 8502 , japan + e - mail : [email protected] |
luminous infrared galaxies ( lirgs ) and ultra - luminous infrared galaxies ( ulirgs ) are known to be huge molecular gas reservoirs [ @xmath6 ] .
the results of @xmath0co(10 ) and ( 21 ) observations reveal that this molecular gas is usually concentrated at the center of the galaxies ( e.g. @xcite ) , and in general it forms a compact ( radius @xmath7 kpc ) rotating disk or ring ( e.g. , @xcite ) .
the molecular gas within the disk or ring often shows large turbulent motions , and the molecular line profiles exhibit large velocity widths of up to 1000 km s@xmath8 .
the physical properties , namely the density and the temperature , of the molecular gas in lirgs / ulirgs are still being studied .
single - dish hcn(10 ) observations ( @xcite ; @xcite , ) revealed that most of their molecular gas is supposed to be dense [ @xmath9 cm @xmath10 .
single - dish observations of higher excitation @xmath0co and @xmath2co emissions suggest there is large amount of warm molecular gas ( @xcite ) . for the denser and warmer molecular gas tracers ( e.g. hcn , @xmath2co ) ,
however , the spatial distributions in lirgs / ulirgs are still unknown , although the physical properties of gas are keys to understand the activity of lirgs / ulirgs ( starburst in the nuclear region and/or agn ) . moreover , the spatial variations of the molecular gas density and the temperature in the circumnuclear regions of lirgs / ulirgs are still unclear . for such studies ,
high angular resolution observations of molecular lines with radio interferometers are needed , but only a few observations have been performed so far ( e.g. @xcite ) .
ngc 6240 ( @xmath11 , @xmath12 ( b1950 ) ; @xmath13 mpc , if we adopt @xmath14 km s@xmath8 mpc@xmath8 ; = 475 pc ) is one of the most well - known nearby lirgs , with an infrared luminosity of @xmath15 @xcite .
this galaxy is a merger , with clearly visible disturbed morphology and tidal tails in optical images .
ngc 6240 also has double nuclei detected in a wide range of wavelengths from x - rays to radio ( @xcite ) .
the apparent separation of the nuclei is ( 710 pc ) in the 15 ghz radio continuum map @xcite .
each of the nuclei are accompanied by an intense massive star - forming region ( e.g. @xcite ) , and both of them also host an agn ( @xcite ) .
interferometric @xmath0co(10 ) and ( 21 ) observations have been carried out by many groups ( @xcite ) .
these observations show that about half of the molecular gas is concentrated in a rotating thick disk - like structure ( @xmath16 pc diameter ) between the double nuclei , but the motion of the gas is highly disturbed and shows large velocity dispersions .
the total mass of the molecular gas within a radius of @xmath17 pc is estimated to be 24@xmath18 , which constitutes about half of the dynamical mass @xcite .
intense hcn(10 ) emission has been detected with the iram 30 m telescope ( @xcite ) .
the obtained hcn/@xmath0co(10 ) intensity ratio is 0.15 , which is a typical value for starburst galaxies , and suggests the presence of large amounts of dense molecular gas . in this paper
, we present the results of interferometric observations of molecular emission lines toward the central region of ngc 6240 .
the main goal of this study is to reveal the location and distribution of dense gas and to constrain the physical properties of the molecular gas .
we describe the observational parameters and data - reduction procedures in section 2 , and the observational results of the @xmath0co , hcn , hco@xmath1 , and @xmath2co emission lines and continuum emission in section 3 . in this section
, we also refer to the observed molecular line intensity ratios and those characteristics . with these data ,
the physical properties of the molecular gas are discussed in section 4 , and the evolutionary scenario of molecular gas and star formation according to the merging process is described in section 5 . finally , we summarize our study in section 6 .
aperture synthesis observations toward ngc 6240 were carried out with the nobeyama millimeter array ( nma ) and the rainbow interferometer at the nobeyama radio observatory ( nro ) between 1995 november 1996 march , and 2000 january 2003 april .
the nma consists of six 10 m - antennas equipped with cooled dsb sis receivers .
the rainbow interferometer is a 7-element array combining the nma with the nro 45 m telescope , providing higher spatial resolution and sensitivity than those of the nma .
we observed ngc 6240 in the following emission lines : @xmath0co(10 ) , hcn(10 ) , hco@xmath1(10 ) , and @xmath2co(10 ) .
except for the @xmath0co observations , the backend we used was the ultra - wide - band correlator ( uwbc ; @xcite ) . for the @xmath0co observations ,
the old - fx correlator was employed ( @xcite ) .
the quasars b1548 + 015 , b1656 + 053 , and b1655 + 077 were used for phase and amplitude reference calibrations ; 3c 279 , 3c 345 , and 3c 454.3 were used for bandpass calibration .
uranus , neptune , and mars were used as absolute flux - scale calibrations for the amplitude calibrator .
the uncertainties in the absolute flux scale were estimated to be @xmath19% for each observation .
the hcn and hco@xmath1 emissions were observed simultaneously within the same correlator passband .
the uv - data were calibrated with the uvproc - ii software package developed at nro ( @xcite ) , and then imaged with natural uv weighting , and cleaned , with the nrao aips package .
the detailed observational parameters are summarized in table [ obslog ] .
in this section we describe the distributions and kinematical properties of the observed emission lines and calculate the intensity ratios .
our hco@xmath1 and @xmath2co interferometric observations are the first ones on ngc 6240 .
the observational results of continuum emission at 87 ghz and 108 ghz are also presented .
table [ obsres ] summarizes the observational results .
figure [ intmaps]a shows integrated intensity map of the @xmath0co(10 ) emission . (
no primary beam correction has been applied to the maps presented in this paper . )
the distribution of the @xmath0co emission is more extended than the size of the synthesized beam , suggesting that it is mildly resolved .
the peak of the @xmath0co emission is located between the double nuclei defined by the 15 ghz continuum peaks @xcite , not on either of them .
figure [ cochmap ] shows the @xmath0co velocity channel maps , and the @xmath0co position
velocity map sliced along the double nuclei ( the position angle is from north to east ) is presented in figure [ copvmap ] .
the velocity profile of the @xmath0co line at the emission peak is presented in figure [ lineprofs ] , together with those of the hcn , hco@xmath1 , and @xmath2co emissions .
these plots reveal that the velocity width of the @xmath0co line is extremely large .
the full width at half maximum is 420 km s@xmath8 , and the full width at zero intensity ( fwzi ) is @xmath20 600 km s@xmath8 [ the recent interferometric observations @xcite show that the fwzi of @xmath0co emission is about 900 km s@xmath8 ] .
the observed total integrated intensity of @xmath0co is 168.3 jy km s@xmath8 , which is 71% of the single dish flux obtained by the iram 30 m telescope ( @xcite).co data . in order to obtain the continuum - free @xmath0co intensity
, the continuum emission contribution is removed numerically .
the way we employed is as follows . from the velocity integrated map within 767 km s@xmath8 ( figure [ intmaps]a )
, we primarily obtained the total emission flux ( @xmath0co + continuum ) of 176.9 jy km s@xmath8 .
a 112 ghz continuum flux of 11.2 mjy was estimated from the spectrum assuming a single power - law ( @xmath22 ) between cm - wave and mm - wave ( see also figure [ contsed ] ) . thus the total continuum emission contribution in figure [ intmaps]a is estimated to be 8.59 jy km s@xmath8 (= 11.2 mjy @xmath23 767 km s@xmath8 ) .
we then subtracted this from the primarily flux , and obtained the continuum - free and velocity- and spatially - integrated @xmath0co line flux of 168.3 jy km s@xmath8 . ] a slight velocity gradient along the double nuclei is seen in the position velocity map ( see figure [ copvmap ] ) .
the trend of the velocity gradient is the same as that in the @xmath0co(21 ) map by @xcite ; the redshifted velocity components are in the north - east , and the blueshifted velocity ones are in the south - west
. these results are consistent with the recent @xmath0co(10 ) observations by bryant and scoville ( 1999 ) .
figure [ intmaps]b shows the hcn(10 ) integrated intensity map .
the continuum emission , which was detected by averaging line - free channels with an effective bandwidth of 376 mhz , was subtracted from the visibility data before creating the hcn synthesized image .
since the hcn emission is unresolved with our @xmath24 beam , it must be more spatially compact than the @xmath0co distribution .
the emission peak is located between the double nuclei , within of the peak position of the @xmath0co(10 ) emission .
it should be noted that the hcn emission peak ( and also @xmath0co ) is not coincident with the current star - forming regions traced by near - infrared ( nir ) emission lines ; the star - forming regions are associated with both of the double nuclei ( e.g. @xcite ) . we will focus on this issue in section 5 .
the detected total flux of hcn emission is 14.1 jy km s@xmath8 , or @xmath25% of single - dish flux obtained by solomon , downes , and radford ( 1992 ) .
the observed flux also agrees very well with the interferometric data by @xcite .
figure [ intmaps]c shows the integrated intensity map of hco@xmath1(10 ) emission .
the procedure of continuum subtraction is the same as that of the hcn data , because both lines are observed simultaneously in the same passband .
this is the first detection of the hco@xmath1 line from ngc 6240 . like the @xmath0co and hcn emissions
, the hco@xmath1 emission is peaked between the double nuclei .
the spatial distribution of the hco@xmath1 emission is also compact and resembles that of the hcn emission .
no systematic velocity gradient is evident in either the hcn or the hco@xmath1 data - cubes .
this is probably due to the fact that we have not resolved their spatial distributions .
both the hcn and hco@xmath1 emission - line profiles ( figure [ lineprofs ] ) agree reasonably well with that of @xmath0co emission , taking into account their noise levels ( 20 mjy beam@xmath8 per channel for @xmath0co , and 4 mjy beam@xmath8 for hcn and hco@xmath26 ) .
an integrated intensity map of @xmath2co(10 ) emission is presented in figure [ intmaps]d , and emission is detected at the 4.1@xmath27 level .
this is the first image ever of @xmath2co emission from ngc 6240 .
the continuum emission was subtracted from the visibility data before creating the synthesized image .
the flux of the @xmath2co emission is much smaller than those of @xmath0co and hcn .
the obtained total flux is about 60% of the single dish flux obtained by @xcite .
the synthesized beam size of the @xmath2co map ( @xmath23 ) is larger than those of @xmath0co and hcn , and the emission dose not seem to be resolved .
the peak of @xmath2co also coincides with those of @xmath0co , hcn , and hco@xmath1 . in this subsection
, we briefly comment on the emission line intensity ratios , in brightness temperature scale , ) and brightness temperature ( @xmath28 ) is @xmath29 , where @xmath30 is the observed wavelength , @xmath31 is the boltzmann constant , and @xmath32 is the solid angle of the region in which the line ratio is calculated .
] for the molecular gas concentration between the double nuclei .
taking into account the relatively low spatial resolution of the @xmath2co map , the line ratio is calculated using the integrated fluxes within the region of @xmath23 ( 2.2 kpc @xmath23 1.7 kpc ) , which corresponds to the synthesized beam size of the @xmath2co map .
this area both contains the region between the double nuclei and the double nuclei , themselves .
however , the emissions are almost concentrated between the double nuclei , and the peaks of the observed emission lines coincide with each other .
thus , the line ratios within the @xmath23 region are representative of those between the double nuclei .
the error of each ratio is calculated considering the absolute flux calibration uncertainty for each observation ( 10% ) and the noise level of the emission line maps , except for hco@xmath1/hcn . for the hco@xmath1/hcn ratio , only the noise level of each emission line map is considered , because the hcn and hco@xmath1 line is observed simultaneously within the same passband of the correlator .
we obtained a high hcn/@xmath0co ratio of @xmath33 at the region between the double nuclei .
this contrasts with the total integrated intensity ratio of hcn/@xmath0co = @xmath34 , which agrees with the single - dish values reported by solomon , downes , and radford ( 1992 ) .
the hcn/@xmath0co ratio of 0.25 is higher than the mean ( @xmath35 ) for lirgs and ulirgs , and one of the highest value among them @xcite .
gao and solomon ( 2004a ) show that a large hcn/@xmath0co ( @xmath36 ) is likely to come from regions where most of ( @xmath37 50% ) molecular gas is high density [ @xmath38 @xmath4 based on lvg calculations .
we obtain a hco@xmath1/hcn ratio of @xmath39 .
the typical hco@xmath1/hcn ratio for nearby starburst galaxies and seyfert galaxies with nuclear starbursts is 0.3 2 ( @xcite ) .
therefore , the value in ngc 6240 is within the range for that of starburst galaxies , but larger than the mean value ( @xmath7 )
. the high hco@xmath1/hcn ratio would be due to the existence of shocked dense gas , of which the envelope is exposed to the intense ionization flux of the supernovae ( @xcite , and references therein ) .
the hco@xmath1/hcn ratio of this galaxy will be discussed in a forthcoming paper together with those in the other lirgs and mergers .
the @xmath0co/@xmath2co ratio is @xmath40 at the region between the double nuclei ( 2.2 kpc @xmath23 1.7 kpc ) .
@xcite found that the mean @xmath0co/@xmath2co ratio is @xmath41 in their survey of infrared - bright galaxies .
the ratio in ngc 6240 is thus significantly larger than the mean for starburst galaxies .
some galaxies ( e.g. , ngc 1614 , ic 694 ) show a high ratio of @xmath0co/@xmath2co@xmath42 , and @xcite suggested that such a high ratio could be reproduced by warm turbulent molecular gas ( @xmath43 k ) .
they also maintain that a high @xmath0co/@xmath2co ratio also implies that the optical depth of @xmath0co(10 ) must be moderate [ @xmath44(@xmath0co)@xmath45 and that of @xmath2co(10 ) is very small [ @xmath44(@xmath2co)@xmath46 .
figure [ contmap ] shows both the 87 ghz ( 3.5 mm ) and 108 ghz ( 2.8 mm ) continuum images obtained using the line - free channels of the visibility data in hcn@xmath47hco@xmath1 and @xmath2co observations , respectively . unlike the molecular lines , the continuum emission peak is almost located at the southern nucleus . since the hcn and hco@xmath1 maps and 87 ghz continuum map were created from identical uv - data ,
the difference between the peak positions is real . in the 87 ghz continuum map
, the emission extends more than the size of the synthesized beam ( @xmath23 , which corresponds to 1.7 [email protected] kpc ) .
figure [ contsed ] shows the spectrum of ngc 6240 in the region surrounding the double nuclei ( within @xmath23 , which corresponds to 5.1 [email protected] kpc ) .
the total continuum flux densities at 87 ghz and 108 ghz are 16.6 mjy and 10.8 mjy , respectively .
these values can be reproduced well by assuming a single power - law spectrum ( @xmath48 ) interpolated between the cm - wave continuum emission @xcite and 1.3 mm ( 228 ghz ) emission @xcite .
the best - fit spectral index is @xmath49 , which is within the range of the spectral indices in both synchrotron radiation from agn and that from supernova remnants ( snrs ) ( e.g. , @xcite ) .
therefore , most of our observed mm - wave continuum emission can be explained by synchrotron radiation from an agn , snrs , or both , come mostly from the bright southern nuclei , but there are surely some contributions from the northern nucleus .
this conclusion supports the discussion by @xcite . thermal free free emission from star - forming regions
are also expected to contribute some degree to the mm - wave continuum emission .
based on @xcite and @xcite , the relation of the thermal free free continuum flux ( @xmath50 ) and far - infrared luminosity ( @xmath51 ) is expressed as follows : @xmath52 where @xmath53 is the electron temperature . even if we assume all of the fir luminosity from ngc 6240 , @xmath54 @xcite , comes from the nuclear region , then the estimated free
free emission flux from h ii regions at @xmath55 k is at most @xmath56 mjy at 100 ghz .
this flux value is smaller than the observed flux at 87 ghz and 108 ghz .
moreover , as can be seen in figure [ contsed ] , the spectral index between 100 ghz and 228 ghz is significantly smaller than @xmath57 , which is a typical index for optically thin radio free
free emission .
therefore , the free - free emission contribution on the 100 ghz continuum flux must be less than 7 mjy .
intensity ratios between molecular emission lines are powerful tools to probe the physical properties of molecular gas . in this section
we compare the observed hcn/@xmath2co intensity ratio with radiative transfer model calculations , to try to constrain the physical properties of the molecular gas between the double nuclei of ngc 6240 .
we have compared the observed hcn/@xmath2co ratio with the outcome of radiative transfer model calculations performed by @xcite based on the large velocity gradient ( lvg ) approximation @xcite .
the intense emission of the hcn line and high hcn/@xmath0co ratio ( 0.25 ) imply that most of the molecular gas is dense between the double nuclei @xcite .
hcn emission traces molecular gas at high density [ @xmath58 @xmath4 , and @xmath2co emission is more excited at higher density gas than @xmath0co due to its low optical depth .
thus , the hcn/@xmath2co ratio reflects the physical properties of dense molecular gas better than hcn/@xmath0co ratio @xcite .
figure [ lvgreslt ] shows the contour maps of the intensity ratio , hcn/@xmath2co , as a function of @xmath59/(@xmath60 ) and @xmath61 , where @xmath59 is the fractional abundance of @xmath2co to h@xmath62 and @xmath60 is the velocity gradient in km s@xmath8 pc@xmath8 . the kinetic temperature ( @xmath63 ) in each map
is shown in the top - right corner . in the lvg calculation ,
the single - component model was employed , and a constant relative abundance of [ @xmath2co]/[hcn ] = 50 was adopted @xcite . the full description of the lvg calculation is presented in @xcite . as can be seen in figure [ lvgreslt ] , the observed high ratio , hcn/@xmath2co = @xmath64 , can be reproduced in dense and warm molecular clouds . in the upper two panels showing kinetic temperatures higher than 50 k , the ratio can only be reproduced by dense gas conditions , @xmath65 @xmath66 .
the optical depth of @xmath2co emission under the dense and warm conditions is very small ( 0.01 0.1 ) in the panels of @xmath67 k and 100 k , as suggested by @xcite . to explain the high hcn/@xmath2co ratio under a lower gas kinetic temperature ( @xmath68 k ) , extremely small values of @xmath59/(@xmath60 ) [ @xmath69 ( km s@xmath8 pc@xmath8)@xmath8 ] would be required
this would happen if the velocity gradient is extremely large ( @xmath70 km s@xmath8 pc@xmath8 ) , or the @xmath2co abundance is very low [ @xmath71 .
if a @xmath2co abundance of @xmath72 ( @xcite ) is adopted , a velocity gradient of @xmath70 km s@xmath8 pc@xmath8 is needed to meet the extremely small @xmath59/(@xmath60 ) value . as we show now , this velocity gradient is , however , too large and unrealistic .
we made a simple estimate of the velocity gradient from the @xmath0co emission data .
@xcite reported that , at the peak of @xmath0co(21 ) emission , the line width at zero intensity ( fwzi ) is about 1000 km s@xmath8 in a @xmath23 (= 330pc @xmath23 240pc ) beam . assuming that the @xmath0co emission originated from a single molecular clump having the same size as the observed beam , the velocity gradient is nominally estimated to be @xmath73 km s@xmath8pc@xmath8 .
even if the size of molecular clump were 50 pc ( typical for galactic giant molecular clouds ) , the estimated velocity gradient would be @xmath74 km s@xmath8 pc@xmath8 , which is still quite smaller than the value above .
therefore , the extremely large velocity gradient as @xmath70 km s@xmath8 pc@xmath8 is not realistic .
we now discuss the possibility that the @xmath2co abundance is quite low . in the calculations for figure [ lvgreslt ] , a constant relative abundance of [
@xmath2co]/[hcn]=50 was used .
the hcn/@xmath2co ratio hardly depends on the hcn abundance [ @xmath75/(@xmath60 ) ] because of the large optical depth of hcn ( see figure 6 of @xcite ) . on the other hand ,
a low @xmath2co abundance can reproduce high hcn/@xmath2co ratios .
the `` standard abundance '' @xmath72 is obtained based on observations of molecular clouds in the inner galaxy ( @xcite ) .
although the @xmath2co abundance at the centers of other galaxies are still not very well constrained , there appears to be little difference between the galaxy and others ( e.g. @xcite , , ) .
the galaxy galaxy merging may cause a low @xmath2co abundance in the galactic center . the molecular gas in the disk of merging progenitors
are predicted to converge into the centers of each of the original galaxies @xcite . in the galaxy ,
the @xmath0c/@xmath2c abundance ratio becomes larger toward the outer disk than in the galactic center , but the difference is only a factor of @xmath76 ( @xcite , and references therein ; @xcite ) .
the @xmath0co/@xmath2co abundance ratio will not largely differ from that of the @xmath0c/@xmath2c ( e.g. @xcite ) .
thus , the low @xmath2co abundance might somewhat contribute to the high hcn/@xmath2co ratio , but the extremely low @xmath2co abundance [ @xmath71 can not be realized .
selective dissociation of the @xmath2co molecule might play a role , but the abundance of @xmath2co is supposed to be affected little @xcite . from the above considerations , it seems to be quite reasonable to conclude that the molecular gas is dense [ @xmath77 @xmath4 and warm ( @xmath78 k ) between the double nuclei of ngc 6240 .
our results are also consistent with the fact that the analysis of the spectral energy distributions from the infrared to the sub - mm suggests a significant amount of warm ( @xmath79 k ) dust in ngc 6240 ( @xcite ) . to confirm the high - temperature condition of the molecular gas
positively , observations of higher excitation @xmath0co and @xmath2co emission should be performed .
it is expected that the @xmath0co/@xmath2co ratio in @xmath8021 and 32 show values smaller than that of @xmath81=10 ( cf .
@xcite ) . here , we make a brief comparison of physical properties of molecular gas with those of other starburst galaxies .
the observational results of higher excitation @xmath0co emission lines ( @xmath8076/65 ) combined with lvg calculations suggest that , for the central region ( @xmath82 pc ) , the physical properties of molecular gas is @xmath83 @xmath66 and @xmath84 k in ngc 253 @xcite and m 82 @xcite .
@xcite obtained interferometric maps of @xmath0co , @xmath2co , and hcn ( 10 ) toward the lirg merger arp 299 ( ic 694+ngc 3690 ) in ( 460 1090 pc ) resolutions . at the nucleus of ic 694 ,
the ratio @xmath0co/@xmath2co is 60 , the ratio hcn/@xmath0co is 0.11 , and the ratio hcn/@xmath2co is 6.7 .
these ratios are similar to those of ngc 6240 .
single - dish @xmath2co(21 ) observations toward ic 694 have also been performed by @xcite ; they concluded that the molecular gas consists of a dense [ @xmath85 @xmath4 and warm ( @xmath78 k ) medium based on a high @xmath2co(21)/(10 ) ratio ( @xmath86 ) .
these previous results imply that the estimated properties of the molecular gas in ngc 6240 are common for starburst and merging galaxies .
as we have shown in previous sections , ngc 6240 has a large amount of dense and warm molecular gas in the central region .
the dense molecular gas distribution , however , does not coincide with that of the massive star - forming regions .
figure [ hcnonfeii ] shows our hcn map superimposed on the distribution of nir [ fe ii ] ( left panel ) and h@xmath62 ( right panel ) emissions .
the figure clearly shows that the hcn peak agrees well with the h@xmath62 peak rather than the [ fe ii ] peak .
the [ fe ii ] emission is supposed to be generated in fast shocks from supernova remnants @xcite , and is thus supposed to follow the current massive star forming regions . on the other hand ,
the h@xmath62 emission arise from hot molecular hydrogen excited by large - scale shock rather than uv fluorescence or x - ray heating ( @xcite ) .
the dense gas traced by the hcn and hco@xmath1 emissions are concentrated between the double nuclei ( see also figure [ intmaps ] ) , whereas the massive star - forming regions traced by nir [ fe ii ] , pa@xmath87 , and br@xmath88 emission lines are associated with both of the nuclei ( @xcite ) .
usually the dense molecular gas is located at current massive star - forming regions in the central region of starburst galaxies and seyferts @xcite .
this is in contrast with what we have observed in ngc 6240 .
we provide a possible illustration to explain this discrepancy in the next subsection . here , we propose a possible scenario to explain the evolution of the molecular gas and star formation in ngc 6240 .
the origin of the unusual discrepancy between the dense gas distribution and the star - forming regions is also illustrated within it .
this scenario is based on our observational results together with those of others and results from numerical simulations of galaxy galaxy merging .
a brief summary of the scenario is as follows : 1 ) during a galaxy galaxy interaction , dense molecular gas is formed and intense star formation occurs in the circumnuclear regions of each of the merger progenitor galaxies , 2 ) an almost head - on collision or very close encounter of the nuclei cause the merging of the two dense gas concentrations , 3 ) the star - forming nuclei can still remain separate while the single dense gas concentration is left behind between them ( this is the present view of ngc 6240 ) , 4 ) a new and intense star formation will begin at some period in the dense molecular gas concentration .
we describe this scenario in more detail following the order of events . at the early merging stages , the molecular gas pre - existing in the merger progenitor galaxies accumulate toward their respective galactic centers approaching each other .
the gas accumulation probably results in the formation of dense molecular gas and nuclear starburst detected by the nir observations ( e.g. @xcite ) .
n - body simulations of galaxy galaxy merging show that gas inflow , driven by gravitational torques , will cause condensations of gas clouds around the nuclei of both original galaxies @xcite .
numerical simulations also predict that the gas accumulation in each of the nuclei probably lead to considerable amount of dense molecular gas formation , which is able to cause star - formation activities within it .
for example , in the merging galaxy arp 299 , most of the molecular gas is concentrated toward the nuclei @xcite , and the nuclei harbor intense star formation activity @xcite .
in the course of the merging process , the progenitor nuclei accompanied by dense gas concentration may suffer a nearly head - on collision or very close encounter . in the case that the impact parameter was smaller than the gas concentrations ( @xmath89 kpc ) , although the nuclei and associated stellar systems can remain gravitationally bounded , the dense molecular gas would be stripped off from the nuclei and form a single molecular gas condensation ( e.g. @xcite ) .
these events ultimately lead the nuclei to be deficient in dense molecular gas , while almost terminating the nuclear star formation .
we apply this to ngc 6240 , and try to make a rough estimate of the time - scale of the collision / encounter using the relative velocity and distance between the double nuclei .
the projected relative velocity is measured to be @xmath90 km s@xmath8 from the stellar absorption and @xmath91 km s@xmath8 from the br@xmath88 emission @xcite ; thus , the actual relative velocity is assumed to be on the order of 100 km s@xmath8 . from the 15 ghz continuum image @xcite ,
the projected distance of the double nuclei is 710 pc .
if our viewing angle is , then the actual separation of the nuclei is about 1 kpc .
if we adopt 100 km s@xmath8 for the relative velocity and 1 kpc for the distance , it takes about @xmath92 yr for the double nuclei to move from the pericenter to the present positions .
@xcite estimated that the actual orbital velocity and separation between the double nuclei are 155 km s@xmath8 and 1.4 kpc , respectively .
they are very close to our assumed values .
this time - scale is consistent with the age of the nuclear star formation .
the latest star - formation activity is estimated to be suspended for 11.5 @xmath93 yr ago from the results of the stellar population synthesis based on the optical and nir spectroscopic observations ( @xcite ) .
after collision / encounter , the double nuclei leave each other from the pericenter , and the dense molecular gas concentration remains between the double nuclei .
some recently formed massive stars exist around the nuclei , but the major part of nuclear star - formation activities have already faded out because of the deficiency of dense molecular gas .
this is the present view of ngc 6240 .
the molecular gas concentration shows a velocity gradient within it ; the north side is redshifted with respect to the south ( see figure [ copvmap ] ) .
this is consistent with the orbital motions of the double nuclei @xcite .
@xcite concluded that the velocity gradient originates from the rotational motion of the gas concentration . in a perfect head - on collision ,
the merged gas is expected to have no systematic velocity structure .
but , in the case of an offset head - on collision or a very close encounter , the orbital angular momentum of dense gas in the pre - collision phase will remain as the rotational motion of the merged gas concentration . in our scenario ,
the observed velocity gradient can be explained as such rotational motion .
the molecular gas between the double nuclei is not only dense , but also warm .
there is no sign of intense massive star formation or prominent agn activity between the double nuclei , and thus ultraviolet radiation from them should not be a principal heating source .
one probable heating source is dynamical energy sources , that is to say , turbulence and shocks .
the large velocity width of the molecular emission lines and complicated distributions of molecular gas imply the existence of large turbulence within the dense gas concentrations @xcite .
extraordinary bright shock excited nir h@xmath62 emission has been detected in this object ( @xcite ) , and the h@xmath62 peak agrees well with that of the dense gas concentration ( see figure [ hcnonfeii ] ) .
these facts imply that the large turbulence and the intense shocks causes a high molecular gas temperature .
a brand - new starburst will begin in the dense gas concentration in the near future .
@xcite estimated the dissipation time - scale for random velocities within the gas concentration to be @xmath94 yr , and suggest that ngc 6240 may experience the next major starburst within the gas concentration .
some external effects on the dense gas might also proceed star - formation . the `` starburst driven starburst '' proposed by @xcite is one of the possible mechanisms .
dense gas clumps can collapse under high external pressure caused by the starburst driven super - wind from the double nuclei @xcite , and massive star - formation will be induced in them .
star formation within the dense gas concentration may have already commenced , but it is very young ( @xmath95 yr ) and suffering from large interstellar absorption because of huge amounts of interstellar medium .
there exists a hint that the starburst activity should not be older than several @xmath96 yr , if the starburst has already begun .
there is no evident sign of strong supernova explosions , because of the weakness of cm - wave radio continuum emission and nir [ fe ] emission between the double nuclei @xcite .
mid - infrared ( mir ) spectroscopic observations with the infrared space observatory ( iso ) revealed that the star - forming regions in ngc 6240 suffer from large interstellar extinction .
the extinction toward the star - forming regions , estimated from the emission line flux ratio between mir [ ne ii ] and nir br@xmath88 , is @xmath9715 20 mag ( corresponding to @xmath98 1.5 2 mag ) in the simplified screen assumption ( @xcite ) .
this estimation is larger ( 10 15 mag in @xmath99 ) than the extinction derived from optical and nir spectroscopic observations ( @xcite ) .
therefore , the spatial distributions seen even at the nir wavelength ( [ fe ii ] , pa@xmath87 , and br@xmath88 emission lines ) may not totally trace star - forming regions deeply obscured by the interstellar medium .
although the location of an obscured star - forming region has not been identified yet because the iso aperture ( @xmath100 ) covers the whole nuclear region of ngc 6240 , a young star - forming region may hide in the dense molecular gas concentration between the double nuclei .
we have carried out high spatial - resolution observations of molecular lines toward the infrared luminous merger ngc 6240 using the nma and the rainbow interferometers .
the @xmath0co emission peak lies between the double nuclei of this galaxy ; our results confirm the previous interferometric @xmath0co(10/21 ) observations .
our new high - resolution hcn and hco@xmath1 observations reveal that the emissions are also peaked between the double nuclei , and are not spatially resolved with our beam size of .
the ratio hcn/@xmath0co is as high as 0.25 in the region between the double nuclei
. the @xmath2co emission has also been observed , and it turns out to be much weaker than others .
the comparison between the high ratio hcn/@xmath2co = 5.9 and the lvg calculations suggests that the molecular gas is substantially dense and warm .
the estimated molecular hydrogen density and kinetic temperature are @xmath3 @xmath66 and @xmath78 k , respectively .
these physical properties are similar to those seen at the centers of the nearby starburst galaxies and infrared luminous mergers .
the observed characteristics that most of the dense and warm gas in ngc 6240 is located between the starburst nuclei is very unusual among the infrared luminous mergers observed so far .
we propose a scenario to explain the evolution of dense molecular gas and starbursts , and to resolve the discrepancy between their locations in ngc 6240 . in our scenario ,
a head - on collision or very close encounter of the star - forming nuclei caused the merging of dense molecular gas associated with both nuclei .
at the present moment , the double star - forming nuclei are still apart from each other , but a dense and warm molecular gas concentration has been left between them . in the near future , a new starburst is expected to commence within the dense gas concentration .
the authors wish to express their deep gratitude to dr .
satoki matsushita for kindly providing us with his lvg calculation results .
we are grateful to prof .
paul van der werf for readily furnishing us with his high - quality near - infrared images of ngc 6240 .
we are indebted to dr .
baltasar vila - vilar for carefully reading the manuscript and valuable comments .
we would like to thank dr .
dennis downes , the referee , for his helpful comments .
we thank the nro staff for the operation of the telescopes and assisting us in our observations .
was financially supported by the japan society for the promotion of science ( jsps ) .
k.n . was also partially supported by a grant - in - aid for scientific research from the ministry of education , culture , sports , science and technology ( mext ) no .
15037212 .
was financially supported by jsps grant - in - aid for scientific research ( b ) no.14403001 and mext grant - in - aid for scientific research on priority areas no.15071202 .
the nobeyama radio observatory is a branch of the national astronomical observatory of japan , the national institutes of natural sciences ( nins ) .
httemeister , s. , & aalto , s. 2001 , in asp conf .
ser . 249 .
, the central kiloparsec of starbursts and agn : the la palma connection , ed . j. h. knapen , j. e. beckman , i. shlosman , & t. j. mahoney ( san francisco : asp ) , 619 kohno , k. , matsushita , s. , vila - vilar , b. , okumura , s. k. , shibatsuka , t. , okiura , m. , ishizuki , s. , & kawabe , r. 2001 , in asp conf . ser . 249 . , the central kiloparsec of starbursts and agn : the la palma connection , ed .
j. h. knapen , j. e. beckman , i. shlosman , & t. j. mahoney ( san francisco : asp ) , 672 okumura , s. k. , kawabe , r. , ishiguro , m. , kasuga , t. , morita , k .-
ishizuki , s. 1991 , in iau symp .
146 , dynamics of galaxies and their molecular cloud distributions , ed .
f. combes & f. casoli ( dordrecht : kluwer ) , 425 lcccc & @xmath0co & hcn & hco@xmath1 & @xmath2co + array configurations & ab , c , d & rb , ab , c & rb , ab , c & c , d + observed frequency ( ghz ) & 112.52 & 86.52 & 87.06 & 107.57 + year of observation & 19951996 & 20002002 & 20002002 & 20012003 + correlator & old - fx & uwbc & uwbc & uwbc + frequency resolution ( mhz ) & 0.3125 & 8 & 8 & 8 + bandwidth ( mhz ) & 320 & 1024 & 1024 & 1024 + synthesized beam(@xmath101 ) & @xmath102 & @xmath103 & @xmath103 & @xmath104 + lcccc & @xmath0co & hcn & hco@xmath1 & @xmath2co + on - source integration time ( hour ) & 9.7 & 17.0 & 17.0 & 27.1 + velocity resolution of data cube ( km s@xmath8 ) & 21.3 & 55.4 & 55.1 & 178.4 + typical noise level ( mjy beam@xmath8 ) & 20 & 4 & 4 & 2 + typical t@xmath105 in ssb ( k ) & 760 & 330 & 330 & 400 + & & & & + half intensity size ( @xmath101 ) & @xmath106 & @xmath103 & @xmath107 & @xmath104 + line width ( fwhm ; km s@xmath8 ) & 420 & 390 & 420 & 450 + total flux ( jy km s@xmath8 ) & 168.3 & 14.1 & 20.9 & 3.7 + nma / iram 30 m flux ratio & 0.71 & 0.67 & & 0.58 + luminosity ( k km s@xmath8 pc@xmath108 ) & @xmath109 & @xmath110 & @xmath111 & @xmath112 + ratio to @xmath0co ( total ) & & 0.14 & 0.21 & 0.024 + ratio to @xmath0co ( @xmath113@xmath23 ) & & 0.25 & 0.37 & 0.047 + ratio to hcn ( total ) & 7.1 & & 1.5 & 0.17 + continuum flux ( mjy ) & & 16.6 & 16.6 & 10.8 + | high spatial resolution observations of the @xmath0co(10 ) , hcn(10 ) , hco@xmath1(10 ) , and @xmath2co(10 ) molecular lines toward the luminous infrared merger ngc 6240 have been performed using the nobeyama millimeter array and the rainbow interferometer .
all of the observed molecular emission lines are concentrated in the region between the double nuclei of the galaxy .
however , the distributions of both hcn and hco@xmath1 emissions are more compact compared with that of @xmath0co , and they are not coincident with the star - forming regions .
the hcn/@xmath0co line intensity ratio is 0.25 ; this suggests that most of the molecular gas between the double nuclei is dense .
a comparison of the observed high hcn/@xmath2co intensity ratio , 5.9 , with large velocity gradient calculations suggests that the molecular gas is dense [ @xmath3 @xmath4 and warm ( @xmath550 k ) . the observed structure in ngc 6240
may be explained by time evolution of the molecular gas and star formation , which was induced by an almost head - on collision or very close encounter of the two galactic nuclei accompanied with the dense gas and star - forming regions . |
our universe is assumed to be filled with dark energy because it can explain the accelerated expansion of the universe , which is strongly supported by many cosmological observations@xcite .
dark energy is an exotic energy component with negative pressure and constitutes about @xmath0 of present total cosmic energy .
the leading interpretation of such a dark energy is the cosmological constant with an equation of state @xmath1 ( for a classic review see @xcite , for a recent nice review see @xcite , and for a recent discussion see @xcite ) .
the energy density of this dark energy is associated with quantum vacuum @xcite .
although this explanation is consistent with observational data , it is plagued with the so - called coincidence problem namely , why are the vacuum and matter energy densities of precisely the same order today ? " .
therefore the dynamical scalar fields , such as quintessence @xcite , k - essence @xcite and phantom field @xcite , are proposed as possible alternatives of dark energy .
comparing with other dynamical scalar fields , the phantom field model is more interesting because that it has a negative kinetic energy and the super negative equation of state @xmath2 .
although the null energy condition is violated , this dark energy model is not ruled out by recent precise observational data involving cmb , hubble space telescope and type ia supernova @xcite .
the dynamical evolution of the phantom field in the cosmology has been investigated in the last years @xcite .
it shows that the energy density increases with the time and approaches to infinity in a finite time @xcite .
this implies in the standard einstein cosmology the flat universe dominated by phantom energy will blow up incessantly and arrive at a future singularity finally named big rip which has such a strong exclusive force that anything in the universe including the large galaxies will be torn up . recently ,
many efforts have been working to avoid the big rip @xcite .
it has argued that this future singularity could be vanished in the universe if one considers the effects from loop quantum gravity @xcite .
the presence of negative kinetic energy results in many exotic properties of phantom field in the black hole spacetime .
e. babichev _
et al _ @xcite considered the phantom energy accretion of black hole and find that the mass of the black hole is decreased . this can be explained by that the kinetic energy of the phantom field is negative which yields the super negative equation of state @xmath2 .
the decrease of mass of black hole in the phantom energy accretion will lead to the ratio between charge and mass of black hole could be larger than @xmath3 ( @xmath4 ) and there may exists a naked singularity @xcite , which implies that the cosmological censorship is violated .
the negative kinetic energy also yields that the dynamical evolution of phantom scalar perturbation possesses some unique characteristics in the black hole spacetime @xcite .
one of is that it grows with an exponential rate in the late - time evolution rather than decays as the usual scalar perturbations .
these new results will excite more efforts to be devoted to the study of phantom energy in the background of a black hole . in this paper
we will focus on the hawking radiation of the phantom scalar particles in the kerr black hole spacetime and see what effect of the negative kinetic energy on the power and angular momentum emission spectra of the hawking radiation .
in the curve spacetime , the action of the phantom scalar field with the negative kinetic energy term is @xmath5.\end{aligned}\ ] ] here we take metric signature @xmath6 and the potential @xmath7 , where @xmath8 is the mass of the scalar field . varying the action with respect to @xmath9 , we obtain the klein - gordon equation for a phantom scalar field in the curve spacetime @xmath10 the presence of negative kinetic energy leads to the sign of the mass term @xmath11 is negative in the wave equation , which will yield the peculiar properties of hawking radiation of the phantom scalar particle in the black hole spacetime .
the well - known kerr metric in the boyer - lindquist coordinate is @xmath12 with @xmath13 where @xmath14 is the mass and @xmath15 is the angular momentum of the black hole . equation ( [ we ] ) is separable in terms of the spheroidal harmonics @xmath16 .
the angular and the radial functions @xmath17 , @xmath18 obey to @xmath19 + \bigg[(\omega^2+\mu^2)a^2\cos^2{\theta}-\frac{m^2}{\sin^2{\theta}}+\lambda\bigg]s(\theta)=0,\label{angd}\end{aligned}\ ] ] and @xmath20 + \bigg\{\frac{[(r^2+a^2)\omega - ma]^2}{\delta}+\mu^2r^2-e_{lm}\bigg\}r(r)=0,\label{radial}\end{aligned}\ ] ] respectively .
where @xmath21 is the eigenvalues and the function @xmath22 . in order to calculate the absorption probability @xmath23 and the luminosity of hawking radiation for a phantom scalar particle
, we must solve the radial equation ( [ radial ] ) above .
following the standard matching techniques @xcite , we can create a smooth analytical solution of the equation ( [ radial ] ) in the low - energy and low - angular momentum limit .
near the horizon ( @xmath24 ) regime and at infinity , it has the form @xmath25 respectively . unlike the usual scalar particle , we find that for the phantom particle with an arbitrary value of @xmath8 the solution above denotes an incoming and an outgoing spherical waves at large distance from the black hole . from this solution
, we can calculate the absorption probability @xmath26 where @xmath27 is the hawking temperature of kerr black hole .
these equations can be integrated numerically . here
we present the numerical results about the absorption probability @xmath23 and the hawking radiation of a phantom scalar field in the background of a kerr black hole . in fig.(1 ) , we fix @xmath28 and examine the dependence of the absorption probability of phantom scalar particle on its mass @xmath8 for the first partial waves ( @xmath29 ) in the background of a kerr black hole . comparing with the usual scalar particle , the absorption probability of the phantom increase rather than decrease as the mass @xmath8 increases .
this is not surprising because that in the wave equation of phantom field ( [ we ] ) the sign of the mass term is negative get rise to the absorption probability has the form @xmath30 in the low - energy and low - angular momentum limit . for other values of @xmath31
, we also find that the absorption probability of phantom scalar particle increase with the mass @xmath8 . of a scalar particle propagating in the kerr black hole with its mass @xmath8 ( the left panel is for a usual scalar particle and the right panel
is for a phantom one ) for fixed @xmath32 , @xmath33 and @xmath28 .
we set @xmath34.,title="fig:",width=302 ] of a scalar particle propagating in the kerr black hole with its mass @xmath8 ( the left panel is for a usual scalar particle and the right panel is for a phantom one ) for fixed @xmath32 , @xmath33 and @xmath28 .
we set @xmath34.,title="fig:",width=302 ] [ fig1 ] ( the left panel is for a usual scalar particle and the right panel is for a phantom one ) in the kerr black hole for fixed @xmath35 , @xmath36 and @xmath28 .
we set @xmath34.,title="fig:",width=302 ] ( the left panel is for a usual scalar particle and the right panel is for a phantom one ) in the kerr black hole for fixed @xmath35 , @xmath36 and @xmath28 .
we set @xmath34.,title="fig:",width=302 ] [ fig2 ] of scalar particles propagating with @xmath8 in the kerr black hole , for @xmath37 and different @xmath15 .
we set @xmath34 .
the left panel for a usual scalar particle and the right panel for a phantom particle.,title="fig:",width=302 ] of scalar particles propagating with @xmath8 in the kerr black hole , for @xmath37 and different @xmath15 .
we set @xmath34 .
the left panel for a usual scalar particle and the right panel for a phantom particle.,title="fig:",width=302 ] [ fig3 ] in the kerr black hole .
we set @xmath34.,width=302 ] [ fig4 ] in fig.(2 ) , we plot the absorption probability of phantom particle for the mode ( @xmath38 ) and find that the super - radiance possesses some peculiar properties in this case . first , for phantom particle the range of @xmath39 for the super - radiance to happen is independent of @xmath8 , which means that the phantom scalar particle can be radiated away without any constraint of the particle s energy . this can be explained by that the wave - function of phantom particle at the spatial infinity has the form @xmath40 .
it is different from that of the usual scalar particles .
it is well known that the usual scalar particle with the energy @xmath41 can not be radiated away .
these particles are is forced to be trapped in the system and the amplified wave can accumulate in the potential well , which will yield black hole bomb in which the mass @xmath8 of a usual scalar particle plays the role of a natural mirror .
while in the super - radiance of the phantom scalar particle , it can not yield black hole bomb since all of phantom scalar particles can be radiated away and their wave functions are not in the bounded - state .
second , the magnitude of the super - radiance of phantom particle increases with the mass @xmath8 . while for usual scalar particle , it is decreases .
the new properties of super - radiance imply that the phantom field will enhance hawking radiation of the black hole . in the kerr black hole .
we set @xmath34.,title="fig:",width=302 ] in the kerr black hole .
we set @xmath34.,title="fig:",width=302 ] [ fig5 ] in the kerr black hole .
we set @xmath34.,title="fig:",width=302 ] in the kerr black hole .
we set @xmath34.,title="fig:",width=302 ] [ fig6 ] let us now to study the luminosity of the hawking radiation of phantom particles in the kerr black hole spacetimes .
we first focus on the first partial waves ( @xmath32 , @xmath33 ) , which plays a dominant role in the greybody factor .
the luminosity of the hawking radiation for the phantom scalar particle can be expressed as @xmath42 here the low bound of the integral is zero rather than @xmath8 since the emission of the phantom scalar particle is not constrained by the value of the particle mass @xmath8 . in generally , this integral can not be computed analytically . in figure ( 3 ) , we present some numerical results . for the fixed @xmath8 , the hawking radiation of the phantom scalar particles decreases as the angular momentum a increases .
this is the same as that of the usual scalar particles in the kerr black hole .
but for the fixed @xmath15 , with the mass @xmath8 increase the hawking radiation of black hole increases for the phantom scalar field and decreases for the usual scalar field . therefore the presence of negative kinetic leads to that the emission of phantom field increases and black hole evaporates more quickly .
figure ( 4 ) also tell us that with the increase of the mass @xmath8 the difference between the emission of phantom and usual scalar particle increase rapidly .
when @xmath43 , the luminosity of hawking radiation of phantom field is almost 1.5 multiples of that of the usual scalar one for all @xmath15 .
thus for the larger @xmath8 , the hawking radiation of black hole is dominated by phantom scalar field . in figs.(5 ) and ( 6 ) , we summed up to @xmath44 modes in calculating the energy and angular momentum emission rates .
we find that the mass @xmath8 of phantom scalar particle enhance the the power and angular momentum emission spectra .
our results are consistent with that obtained in the phantom energy accretion of black hole where the black hole mass is decreased .
this leads to increase of the temperature of the black hole and enhance the luminosity of the hawking radiation .
in this paper , we have studied the greybody factor and hawking radiation for a phantom scalar field in the background of kerr black hole in the low - energy and low - angular momentum approximation .
we have found that the absorption probability and hawking radiation contain the imprint of the phantom scalar field .
the effects of @xmath8 on the super - radiance and the luminosity of the hawking radiation for phantom scalar field are different from that for the usual scalar field since phantom scalar particle possesses a negative kinetic energy .
our results may help us to detect whether our universe is filled with phantom field or not .
it would be of interest to generalize our study to other black hole spacetimes , such as stringy black holes etc .
work in this direction will be reported in the future .
this work was partially supported by the national natural science foundation of china under grant no.10875041 and the construct program of key disciplines in hunan province .
j. l. jing s work was partially supported by the national natural science foundation of china under grant no.10675045 , no.10875040 and no.10935013 ; 973 program grant no .
2010cb833004 and the hunan provincial natural science foundation of china under grant no.08jj3010 . | = 0.6 cm * abstract * we study the absorption probability and hawking radiation spectra of a phantom scalar field in the kerr black hole spacetime .
we find that the presence of the negative kinetic energy terms modifies the standard results in the greybody factor , super - radiance and hawking radiation . comparing with the usual scalar particle , the phantom scalar emission is enhanced in the black hole spacetime .
= 0.8 cm |
the purpose of this paper is to sketch a classification of birational transformations based on various notions of `` complexity '' . in previous papers @xcite
an analysis , based on the examination of the successive ( bi)rational expressions corresponding to the iteration of some given birational mappings , has been performed .
when one considers the degree @xmath0 of the numerators ( or denominators ) of the corresponding successive rational expressions for the @xmath1-th iterate , the growth of this degree is ( generically ) exponential with @xmath1 : @xmath2 .
@xmath3 has been called the `` growth complexity '' @xcite and it is closely related to the arnold complexity @xcite . a semi - numerical analysis , enabling to compute these growth complexities @xmath4 for these birational transformations , has been introduced in @xcite .
it has been seen , on particular sets of birational transformations @xcite , that these `` growth complexities '' correspond to a remarkable spectrum of _ algebraic _ values @xcite .
these `` growth complexities '' , summing up the ( asymptotic ) evolution of the _ degree _ of the successive iterates , amount to seeing these mappings as mappings of ( two ) _ complex _ variables .
however , when one considers the phase portrait of these mappings , one also gets some `` hint '' of the `` complexity '' of these mappings seen as mappings of ( two ) _ real _ variables . in the following ,
we will consider a one - parameter dependent birational mapping of two variables . on this very example , it will be seen , considering phase portraits corresponding to various values of the parameter , that these `` real complexities '' vary for the different ( positive ) values of the parameter .
two universal ( or `` topological '' ) measures of the complexities were found to identify @xcite , namely the ( asymptotic of the ) _ arnold complexity_@xcite ( or growth complexity ) and the ( exponential of the ) _ topological entropy _ @xcite .
the topological entropy is associated with the exponential growth @xmath5 of the number of fixed points ( real or complex ) of the @xmath6-th iterate of the mapping : looking at various phase portraits , corresponding to different values of the parameter ( see below ) , it is tempting to define , in an equivalent way , a `` real topological entropy '' associated with the exponential growth @xmath7 of the number of _ real fixed points only _ of the @xmath6-th iterate of the mapping .
this notion of `` real topological entropy '' would actually correspond to the `` visual complexity '' as seen on the phase portrait of the mapping .
such a concept , corresponding to the evaluation of the real complexity @xmath8 of the mapping seen as a mapping bearing on _ real _ variables , would be less universal : it would have only `` some '' of the remarkable topological universal properties of the topological entropy . similarly , it is also tempting to slightly modify the definition of the _ arnold complexity _ @xcite .
the arnold complexity @xcite , which corresponds ( at least for the mappings of two variables ) to the degree growth complexity @xcite , is defined as the number of ( real or complex ) intersections of a given ( generic and complex ) line with its @xmath6-th iterate : it is straightforward to similarly define a notion of `` real arnold complexity '' describing the number of _ real _ intersections of a given ( generic ) _ real _ line with its @xmath6-th iterate .
this real - analysis concept is , at first sight , also very well - suited to describe the `` real complexity '' of the mapping as it can be seen in the phase portrait ( see fig.[f : fig2 ] below ) .
recalling the identification , seen on this one - parameter family of birational mappings , between the ( asymptotic of the ) arnold complexity and the ( exponential of the ) topological entropy @xcite , it is natural to wonder if this identification also works for their `` real '' partners , or if , as the common wisdom could suggest , real analysis is `` far less universal '' , depending on a lot of details and , thus , requires a `` whole bunch '' of `` complexities '' ( lyapounov dimensions , ... ) to be described properly .
in order to see the previous identification even more clearly , one can also slightly modify the definition of the dynamical zeta function , introducing a `` real dynamical zeta function '' associated with the counting of the real cycles only , and , similarly , one can also introduce some `` real arnold complexity '' generating functions .
we will show that several of these two `` real '' generating functions have the same singularities .
furthermore we will actually conjecture several _ simple rational expressions _ for them , yielding , again , _ algebraic values _ for the ( exponential of the ) `` real topological entropy '' .
in particular , when the parameter of our family of birational transformations becomes large , we will get an interesting non trivial rational expression .
these rational results for real mappings can not be simply understood by any `` obvious '' markov s partition , or symbolic dynamics hyperbolic interpretation .
a _ one - parameter _ family of birational mappings of _ two _ ( complex ) variables has been introduced in previous papers @xcite ( see definition ( 3 ) in @xcite ) .
this mapping actually originates from a lattice statistical mechanics framework that will not be detailed here @xcite . in the following
, we will use the extreme simplicity of this mapping of _ two _ ( complex ) variables to first compare two quite universal ( topological ) notions of `` complexity '' namely the growth complexity @xmath9 , which measures the exponential growth of the _ degree _ of the successive rational expressions encountered in an iteration ( a notion which coincides with the ( asymptotic of the ) _ arnold complexity _ is proportional ( for plane maps ) to @xmath0 , the degree of the @xmath10-th iteration of the birational mapping which behaves like @xmath2 .
this `` degree notion '' was also introduced by a. p. veselov in exact correspondence with the general arnold definition @xcite .
note that the concept of arnold complexity _ is not _ restricted to two - dimensional maps . ]
@xcite ) , and the ( exponential of the ) _ topological entropy _ @xcite . in section ( [ bsome ] ) , we will go a step further and compare , more particularly , the notion of _ `` real '' arnold complexity _ versus the notion of _ `` real '' topological entropy_. these two notions will be seen to be suitable to describe the properties of the mapping seen as a mapping of _ real _ variables .
let us consider the following birational transformation ( see ( 3 ) in @xcite ) of two ( _ complex _ ) variables @xmath11 , depending on one parameter @xmath12 : @xmath13 in spite of its simplicity , this birational mapping can , however , have quite different behaviors according to the actual values of the parameter @xmath12 . for example , for @xmath14 , as well as @xmath15 , 1/2 , 1/3 or 1 , the mapping becomes _ integrable _ , whereas it is not @xcite for all other values of @xmath16 .
let us now compare , in the following , two notions of `` complexity '' ( _ arnold complexity versus topological entropy _ ) according to various values of @xmath17 .
the ( growth ) complexity @xmath19 , which measures the exponential growth of the degrees of the successive rational expressions one encounters in the iteration of the birational transformation ( [ yz ] ) , can be obtained by evaluating the degrees of the numerators , or equivalently of the denominators , of the successive ( bi)rational expressions obtained in the iteration process .
one can actually build a _ semi - numerical method _
@xcite to get the value of the complexity growth @xmath18 for any value of the parameter @xmath12 .
the idea is to iterate , with the birational transformation ( [ yz ] ) , a generic _ rational _ initial point @xmath20 and to follow the magnitude of the successive numerators , or denominators , of the iterates . during the first few steps some accidental simplifications may occur , but , after this transient regime , the integer denominators ( for instance ) grow like @xmath21 where @xmath22 is the number of iterations .
typically , a best fit of the logarithm of the numerator as a linear function of @xmath22 , between @xmath23 and @xmath24 , gives the value of @xmath25 within an accuracy of @xmath26 .
let us remark that an _ integrable mapping _ yields a _ polynomial growth _ of the calculations @xcite : the value of the complexity @xmath18 has to be numerically very close to @xmath27 .
fig.[f : fig1 ] shows the values of the complexity growth @xmath4 as a function of the parameter @xmath12 .
one remarks on fig.[f : fig1 ] that all the values of @xmath28 ( except a zero measure set ) give a growth complexity @xmath29 .
the calculations have been performed using an infinite - precision c - library @xcite .
this semi - numerical analysis @xcite clearly indicates , that , beyond the known integrable values @xcite of @xmath30 , namely @xmath31 , two sets of values @xmath32 and @xmath33 are singled out .
this suggests that the growth complexity @xmath4 takes lower values than the generic one on two infinite sequences of values of @xmath34 , namely @xmath35 and @xmath36 for @xmath22 and @xmath37 integers such that @xmath38 and @xmath39 and @xmath40 odd .
one can revisit all these results using the stability of the factorization schemes , performing exact formal ( maple ) calculations .
for instance , one can consider , for various values of @xmath41 , the degrees of the successive rational expressions one encounters when performing the successive iterates , and build various generating functions matrices @xcite .
these generating functions , denoted @xmath42 in @xcite , are deduced from the existence of remarkable stable _ factorization schemes _
these results are in complete agreement with the one given here for the mapping of two variables ( [ yz ] ) ( see for instance equations ( 11 ) , ( 12 ) , ( 13 ) and ( 14 ) in @xcite ) . ] corresponding to these successive degrees .
in particular , having singled out a set of values of @xmath28 , one can revisit these various values , to see how the generic growth complexity @xmath43 gets modified , and deduce the degree generating functions , and the associated complexity @xmath4 , in each case .
let us denote by @xmath44 the degree generating function , corresponding , for some given value of the parameter @xmath12 , to the degree of ( for instance ) the numerator of the @xmath45 component of the successive rational expressions obtained in the iteration process of transformation ( [ yz ] ) . at this step ,
it is worth recalling , again , the notion of arnold complexity @xcite which corresponds to iterate a given ( complex ) line and count the number @xmath46 of intersections of this @xmath1-th iterate with the initial line .
it is straightforward to see that these arnold complexity numbers @xmath47 are closely linked to these successive degrees ( see for instance @xcite ) .
actually , if one considers the iteration of the @xmath48 line ) , ( [ rac ] ) ) . ] , the generating function of the @xmath47 s _ `` almost '' identifies _
( often up to a simple @xmath49 factor ) with the degree generating functions @xmath50 .
the `` arnold '' generating functions @xmath51 , and the degree generating functions @xmath50 , read ( up to order fifteen for the `` arnold '' generating functions and order ten , or eleven , for the degree generating functions ) : @xmath52 where the expression for @xmath53 is valid for @xmath28 generic and the expressions for @xmath54 are valid for @xmath55 , the @xmath56 for @xmath57 with @xmath58 odd .
one also has for various integrable value of @xmath59 : @xmath60 these various exact generating functions are in agreement with the previous semi - numerical calculations .
in particular , the first expression in ( [ firsteq ] ) yields an algebraic value for @xmath61 in agreement with the generic value of the complexity @xmath43 of fig.[f : fig1 ] ( and fig.1 in @xcite ) .
it is well known that the periodic orbits ( cycles ) of a mapping @xmath62 strongly `` encode '' dynamical systems @xcite .
the fixed points of the @xmath1-th power of the mapping being the cycles of the mapping itself , their proliferation with @xmath1 provides a `` measure '' of chaos @xcite . to keep track of this number of cycles
, one can introduce the fixed points generating function @xmath63 where @xmath64 is the number of fixed points of @xmath65 , _ real or complex_. this quantity only depends on the number of fixed points , and _ not _ on their particular localization . in this respect , @xmath66 is a _ topologically invariant quantity_. the same information can also be coded in the so - called function , by artin and mazur @xcite . ] _ dynamical zeta function _
@xmath67 @xcite related to the generating function @xmath66 by @xmath68 . the dynamical zeta function is defined as follows @xcite : @xmath69 the topological entropy @xmath70 is : @xmath71 if the dynamical zeta function is _ rational _ , @xmath72 will be the inverse of the pole of smallest modulus of @xmath73 or @xmath74 . if the dynamical zeta function can be interpreted as the ratio of two characteristic polynomials of two linear operators and @xmath75 . ]
@xmath76 and @xmath75 , namely @xmath77 , then the number of fixed points @xmath78 can be expressed from @xmath79 . in this case , the poles of a rational dynamical zeta function are related to the ( inverse of the zeroes of the ) characteristic polynomial of the linear operator @xmath80 only . since the number of fixed points remains unchanged under _ topological conjugacy _
( see smale @xcite for this notion ) , the dynamical zeta function is also a _ topologically invariant function _ , invariant under a large set of transformations , and does not depend on a specific choice of variables .
such invariances were also noticed for the growth complexity @xmath18 .
it is thus tempting to make a connection between the _ rationality _ of the complexity generating function previously given , and a possible _ rationality _ of the dynamical zeta function .
we will also compare the singularities of these two sets of generating functions , namely the growth complexity @xmath18 and @xmath81 , the ( exponential of the ) topological entropy .
.1 cm * some results for the dynamical zeta function :* let us now get the expansion of the dynamical zeta function of the mapping @xmath82 , for generic values of @xmath34 .
we can first concentrate on the specific , close to the 1/2 value where the mapping is integrable @xcite , namely @xmath83 , has been analyzed in some detail in @xcite .
for this value @xmath84 , the enumeration of the number of fixed points , @xmath22-cycles and the actual status of these fixed points ( elliptic , hyperbolic , points ... ) are given in @xcite .
] , but arbitrary , value @xmath85 . of course
, there is nothing particular with this specific @xmath86 value : the same calculations have been performed for many other _
generic _ values of @xmath87 yielding the same number of ( complex ) fixed points and , thus , the same dynamical zeta function .
the total number of fixed points of @xmath88 , for @xmath89 running from @xmath90 to @xmath91 , yields , up to order fourteen , the following expansion for the generating function @xmath66 of the number of fixed points : @xmath92 this expansion coincides with the one of the _ rational _ function , up to some algebraic values of @xmath30 corresponding to cycle - fusion mechanism see ( [ previous ] ) and ( [ moreprevious ] ) below and see @xcite . ] : @xmath93 which corresponds to a very _ simple ) , a clipped bernoulli shift with the `` pruning rule '' to forbid substring @xmath94 ( that is 1 must be always followed by 0 ) in any sequence of 0 and 1 . however , constructing the markovian partitions ( if any ) , yielding this simple pruning rule for the symbolic dynamics , remains to be done . ]
rational _ expression for the dynamical zeta function : @xmath95 an alternative way of writing the dynamical zeta functions relies on the decomposition of the fixed points into _ irreducible cycles _ : @xmath96 for generic values of @xmath34 , one gets the following numbers of irreducible cycles : @xmath97@xmath98 it has been conjectured in @xcite that _ the simple rational expression _ ( [ conjec ] ) _ is the actual expression of the dynamical zeta function for any generic value of @xmath34 _ ( up to some algebraic values of @xmath12 , see below and in section ( [ creal ] ) ) .
similar calculations have been performed for the other values of @xmath99 that have been singled out in the semi - numerical analysis @xcite . for the non generic values of @xmath34 , @xmath100 with @xmath101 , we have obtained expansions compatible with the following _ rational _ expression : @xmath102 for the other non generic values , namely @xmath103 with @xmath104 odd , the expansions are not large enough to conjecture a single formula valid for any @xmath40 . for @xmath105 ( namely @xmath106 )
one actually gets a dynamical zeta function given by ( [ zetamtexte ] ) for @xmath107 and this might also be the case for @xmath108 . for @xmath109
the expansions are in agreement with a @xmath110 singularity . comparing the various rational expressions in ( [ firsteq ] ) corresponding to generic , and non - generic , values of @xmath12 , with ( [ conjec ] ) , and ( [ zetamtexte ] ) , respectively
, one sees that the singularities ( poles ) of the dynamical zeta function happen to _ coincide _ with the poles of the generating functions of the growth complexity @xmath4 , for all the values of @xmath34 .
in particular , the growth complexity @xmath111 and @xmath81 , the exponential of the topological entropy , are always _
equal_. let us just mention , here , that the modification of the number of fixed points , from the `` generic '' values of @xmath28 to the particular values ( @xmath112 , @xmath113 ) , corresponds to the _ disappearance of some cycles _ which become singular points ( indetermination of the form @xmath114 ) .
these mechanisms will be detailed in @xcite .
actually the `` non - generic '' values of @xmath115 , like @xmath116 , correspond to such a `` disappearance of cycles '' mechanism which modifies the denominator of the rational generating functions , and , thus , the topological entropy and the growth complexity @xmath18 .
in contrast , there actually exists for @xmath117 , _ other singled - out values _ of @xmath115 , like @xmath118 for instance , which correspond to _ fusion of cycles _ ( see section ( [ creal ] ) ) : in the @xmath119 limit , the order three cycle tends to coincide with the order one cycle , which amounts to multiplying the dynamical zeta function ( [ conjec ] ) by @xmath120 .
such `` fusion - cycle '' mechanism _ does not modify _ the denominator of the rational functions , and thus , the topological entropy , or the growth complexity @xmath4 , _ remain unchanged_. * to sum up :* considering a ( very simple ) one parameter - dependent birational mapping of _ only two ( complex ) variables _ , we have deduced an exact identification between the ( asymptotic of the ) arnold complexity , that is the growth complexity @xmath19 , and the ( exponential of the ) topological entropy for _ all _ the various @xmath12 cases ( generic or not ) .
this identification can be understood heuristically @xcite . as a byproduct
, one finds that these two complexities correspond , in this very example , to _ simple algebraic numbers_. .3 cm * a canonical degree generating function :* this identification result is not completely surprising : the dynamical zeta function is a quite `` universal '' function , invariant under a large set of _ topological conjugaisons _ @xcite , and the concept of arnold complexity ( or the degree growth complexity @xmath4 ) also has the same `` large '' set of ( topological and projective ) invariances @xcite .
actually , as far as degree generating functions are concerned , it is natural to introduce , instead of some generating functions of the degrees of the numerator of the @xmath45 component of the @xmath1-th iterate , a more `` canonical '' degree generating function @xmath121 associated with the birational mapping ( [ yz ] ) written in a _ homogeneous way _ ( see the bi - polynomial mapping ( 4 ) in @xcite ) .
iterating ( [ yz ] ) , written in a homogeneous way , and , factoring out at each iteration step the gcd s , one gets a new degree generating function @xmath121 well - suited , at first sight , to describe such large ( topological and projective ) invariances .
a simple calculation shows that this ( projectively well - suited ) degree generating function reads ( for generic @xmath12 ) : @xmath122 for the @xmath123 particular values , and for the two integrable values , @xmath124 and @xmath125 , one gets respectively : @xmath126 since the expansions for the infinite set of values of the form @xmath127 for @xmath128 , can only be performed up to order eleven ( or twelve ) , it is difficult to `` guess '' any expression valid for any @xmath129 like ( [ 1surmhom ] ) .
recalling the results ( [ firsteq ] ) given in section ( [ generdegree ] ) for the degree growth generating functions , one may suspect that , among these @xmath127 for @xmath128 values , one should make a distinction between @xmath130 on one side , and @xmath131 on the other side .
in fact , up to order eleven , all our calculations for various @xmath127 values for @xmath128 ( @xmath132 ) are in agreement with a general equality between @xmath133 and @xmath134 .
more details are available in appendix a. one gets simpler expressions for the integrable values @xmath135 and @xmath136 : @xmath137 one remarks that @xmath121 verifies the simple functional equation @xmath138 , for @xmath139 , and @xmath140 for @xmath141 . for @xmath142
we have not written any dynamical zeta function @xmath143 because , for such an integrable birational mapping , there exist , at ( almost ) any order @xmath144 of iteration , an _ infinite number _ of fixed points of order @xmath144 ( all the points of some elliptic curves @xcite ) and , therefore , our previous `` simple '' definition ( [ zeta ] ) for the dynamical zeta function is not valid anymore .
.3 cm .1 cm * a possible universal relation .
* one can imagine many simple relations between the `` canonical '' degree generating function , @xmath145 , and the dynamical zeta function , @xmath146 .
for instance , for generic @xmath30 , one gets ( among many ... ) the relation @xmath147 , however this relation is not anymore valid for @xmath116 .
one would like to find a `` true universal '' relation between @xmath148 and @xmath149 , that is a relation _ independent _ of @xmath17 ( generic or non - generic ) .
in order to achieve this goal one may imagine to barter @xmath150 , and @xmath149 , for _ projectively well - suited generating functions _ taking into account the point at @xmath151 , namely a dynamical zeta function taking into account the fixed point at @xmath152 ( see ( 52 ) in @xcite ) , @xmath153 , and @xmath154 defined as follows : @xmath155 one verifies immediately that the relation : @xmath156 is _ actually verified for generic _ values of @xmath12 , as well as , for the _ non - generic values _ of the form @xmath157 , and also some _ non - generic values _ of the form @xmath158 ( see appendix a ) .
a similar relation for the two - parameters family of birational transformations depicted in @xcite will be detailed elsewhere .
relation ( [ universal ] ) should give some hint for a true mathematical proof of the relation between arnold complexity and topological entropy .
cm * remark :* recalling the `` arnold '' generating functions @xmath51 , ( see ( [ firsteq ] ) ) , which identifies `` most of the time '' ( namely @xmath17 generic , @xmath159 , @xmath160 for @xmath161 ) with the new well - suited generating function @xmath162 , up to a simple multiplicative factor @xmath163 , one can rewrite , for @xmath17 generic and @xmath159 , relation ( [ universal ] ) as : @xmath164
as far as the growth complexity @xmath25 is concerned , the generic values of @xmath28 ( that is the values different from the previous @xmath112 , @xmath113 singled - out values ) are all on the same `` complexity footing '' ( see the previous fig.[f : fig1 ] ) .
this is clearly confirmed by the exponential _ growth of the computing time _ during the iteration process , which seems to be similar for all these values ( and clearly _ smaller _ for the @xmath165 , @xmath113 particular values ) .
it is however worth noticing that these generic values , which are all on the same @xmath19-footing , clearly yield phase portraits which are quite different and , obviously , correspond to _ drastically different _ `` visual complexities '' of the phase portrait of the mapping .
this `` visual complexity '' corresponds to the ( exponential ) growth of the number of ( real ) fixed points of the mapping seen as a mapping _
bearing on two real variables_. the previous definitions of the dynamical zeta functions @xmath166 and of the generating function @xmath167 counting the number of fixed points , can be straightforwardly modified to describe the counting of _ real fixed points _ : @xmath168 where the number of _ real _ fixed points @xmath169 grow exponentially with the number @xmath170 of iterates , like @xmath171 . a quick examination of various phase portrait for various `` generic values '' of the parameter @xmath17 seems to indicate quite clearly that this `` real topological entropy '' @xmath172 _ varies with _
@xmath30 , in contrast with the `` usual '' topological entropy @xmath173 .
an obvious inequality is : @xmath174 .
let us give here various phase portraits corresponding to different ( generic except the first one ) values of @xmath12 .
note the different scale for the frame of these various phase portraits .
for most of the phase portraits below around 300 orbits of length 1000 , starting from randomly chosen points inside the frame , have been generated ( only points inside the frame are shown ) : on these various phase portraits one sees that one gets , near the integrable value @xmath175 , quite different phase portraits which seem , however , to have roughly the same number of ( real ) fixed points ( see fig.[f : fig4 ] and fig.[f : fig5 ] ) . on these various phase portraits
one also sees , quite clearly , that the number of fixed points seems to _ decrease _ when @xmath17 crosses the integrable value @xmath176 and @xmath177 , going , for instance , from @xmath178 to @xmath179 ( see fig.[f : fig6 ] and fig.[f : fig7 ] ) or , from @xmath180 to @xmath181 ( see fig.[f : fig8 ] and fig.[f : fig9 ] ) .
these results will be revisited in a forthcoming section ( see section ( [ rac ] ) ) .
of course , exactly on the integrable value @xmath177 , the phase diagram corresponds to a ( simple ) foliation of the two - dimensional parameter space in _ ( rational ) curves _
( linear pencil of rational curves , see @xcite ) : @xmath182 where @xmath183 denotes some constant . for the other integrable values
one also has either a linear pencil of rational curves , namely @xmath184 for @xmath185 , as well as : @xmath186 for @xmath187 , or a linear pencil of _ elliptic curves _ for @xmath124 , namely : @xmath188 and also : @xmath189 for @xmath125 .
one also remarks that @xmath190 , which corresponds to the generic @xmath191 growth complexity , also yields a _ remarkably regular _ phase portrait , `` visually '' similar to a foliation of the two - dimensional parameter space in curves , suggesting a `` real topological complexity '' @xmath8 very close , or even equal to @xmath90 .
this fact will also be revisited in the next section ( see section ( [ rac ] ) ) . in order to describe , less qualitatively , the `` real topological complexity '' @xmath8 as a function of the parameter @xmath34 , we have calculated in section ( [ bsome ] ) , the first ( ten , eleven or even twelve ) coefficients of the expansions of @xmath192 , and of the `` real dynamical function '' @xmath193 , for various values of @xmath12 .
let us now try to understand why ( and how ) @xmath8 varies as a function of @xmath194 , and why some other values of @xmath12 , like @xmath190 , different from the previous @xmath195 and @xmath196 non - generic values , seem to play a special role .
the method to get the fixed points of the @xmath1-th iterate of @xmath11 has been detailed in previous papers @xcite .
let us just mention here that , due to the symmetries of this mapping , there exist two singled out lines , namely @xmath48 and @xmath197 , playing a key role in classifying all these fixed points @xcite .
the fixed points of @xmath198 belong to @xmath6-cycles ( if @xmath199 is a fixed point of @xmath198 then @xmath200 is also a fixed point of @xmath198 , for @xmath201 ) .
it can be seen that _ there always exist _ a fixed point in these @xmath1-cycles which belongs , either to line @xmath202 , or to the @xmath197 line .
we call the fixed points , corresponding to line @xmath197 , the `` p - type '' points and the ones , corresponding to line @xmath202 , the `` q - type '' points @xcite . for @xmath203 , other @xmath1-cycles with no points lying on these two lines do occur :
other remarkable sets occur like @xmath204 ( see @xcite ) .
we call the fixed points , corresponding this `` remaining '' set of points , the points of the `` r - type '' @xcite .
one can use these localization properties to get , very quickly , a subset of all the fixed points , namely , for instance , the `` q - type '' fixed points ( one representant in the @xmath6-cycle belongs to line @xmath205 ) .
these calculations can be performed quite efficiently since one can eliminate the @xmath206 variable ( @xmath48 ) and , thus , reduce the calculations to looking for the roots ( real or not ) of an @xmath34-dependent polynomial in this remaining @xmath45 variable .
one gets , for the first values of @xmath1 , the following polynomial expressions relating @xmath45 and @xmath34 : @xmath207 it is easy to see that the number of real roots @xmath45 , of one of these @xmath208 conditions , varies with @xmath30 by intervals .
the changes of this number of real roots take place at _ algebraic _ values of @xmath30 ( resultant of @xmath209 in @xmath45 ) .
the details of the calculations , and a description of these polynomials , will be given elsewhere @xcite .
the number of the fixed points of the `` @xmath210-type '' ( see @xcite ) is , thus , a function of @xmath34 constant by interval , the limits of the intervals corresponding to some algebraic values ( resultants deduced from the @xmath211 s by eliminating @xmath45 ) . for illustration ,
let us just plot here the real roots @xmath212 , as a function of @xmath12 , for @xmath213 : let us give , for order @xmath214 , a few examples of these algebraic `` threshold '' real values of @xmath30 corresponding to the real roots of such `` @xmath210-type '' polynomials ( [ qtype ] ) : @xmath215 the roots of the first polynomial are of the form ( [ previous ] ) .
the `` threshold '' values of @xmath30 are thus given by two real roots @xmath216 and @xmath217 .
the second polynomial only gives two real roots @xmath218 and @xmath219 .
similar calculations can be performed for the fixed points of the `` @xmath220-type''(see @xcite ) corresponding to the line @xmath221 ( see @xcite ) .
one can also get the real roots @xmath45 of @xmath222 as a function of @xmath34 .
the algebraic values of @xmath12 , occurring in this case , are , again , the two roots of the first polynomial in ( [ q10 ] ) together with the only two real roots @xmath223 and @xmath224 of polynomial : @xmath225 as well as two real roots @xmath226 and @xmath227 of a polynomial of degree 24 in @xmath34 that will not be written here . the last set of points of the `` @xmath228-type '' ( see @xcite ) , which corresponds to fixed points that are neither of the `` @xmath229-type '' , nor of the `` @xmath210-type '' , give the following real roots @xmath230 , @xmath231 , and @xmath232 , corresponding to polynomial @xmath233 . on all these algebraic values of @xmath12
, one can see a variation of the total number of fixed points ( `` p - type '' , `` q - type '' and `` r - type '' ) of order @xmath214 ( see insertion in fig.[f : figq10 ] ) .
these values are in fact particular examples of families of algebraic @xmath12 values .
the simplest family of singled out _ algebraic _ values of @xmath34 , corresponds to the fusion on an @xmath1-cycle with the @xmath234-cycle , and reads : @xmath235 for any integer @xmath1 ( with @xmath236 , @xmath237 not a divisor of @xmath1 ) .
other cycle - fusion mechanisms take place yielding new families of algebraic values for @xmath12 .
for instance the coalescence of the @xmath238-cycles in the @xmath239-cycle , and the coalescence of the @xmath240-cycles in the @xmath241-cycle , yield respectively ( with some constraints on the integer @xmath237 that will not be detailed here ) : @xmath242 .3 cm * status of the fixed points :* the fixed point of @xmath243 , which is elliptic for @xmath244 , becomes _ hyperbolic _ for @xmath245 . for three iterations ( @xmath246 )
one finds that , moving through the @xmath247 value , also changes the status of these fixed points from elliptic to hyperbolic .
in fact , the algebraic values , like the ones depicted in ( [ previous ] ) or ( [ moreprevious ] ) , _ also occur _ in such changes of status from elliptic to hyperbolic ( see @xcite ) .
therefore , the number of elliptic fixed points , or hyperbolic fixed points , is not as `` universal '' as the total number of ( complex ) fixed points , however it has `` some universality '' : for a given value of @xmath6 , the number of elliptic ( resp .
hyperbolic ) fixed points depends on @xmath34 also by
_ intervals _ ( staircase function ) .
this has , again , to be compared with the dependence of the growth complexity @xmath4 , seen as a function of @xmath34 , depicted in fig.[f : fig1 ] .
the fact that the number of hyperbolic versus elliptic fixed points , as well as the number of _ real versus non - real _ fixed points , is modified when @xmath34 goes through the _ same set _ of values , like ( [ previous ] ) or ( [ moreprevious ] ) , seems to indicate that a modification of the number of _ real fixed points is not independent of the actual status of these points _
( hyperbolic versus elliptic ) .
this phenomenon , in fact , corresponds to a quite involved , and interesting , _ structure _ @xcite that will be sketched in @xcite .
let us recall some results corresponding to @xmath249 , in particular the product decomposition ( [ weil ] ) of the dynamical zeta function @xcite . in @xcite ,
the number of irreducible cycles @xmath250 ( see ( [ weil ] ) ) , as well as the number of irreducible cycles corresponding to hyperbolic points , elliptic points , real points is detailed ( see table i in @xcite ) .
these results ( and further calculations ) enable to write , for @xmath249 , the `` real dynamical function '' @xmath193 as the following product : @xmath251 yielding the following expansion for @xmath252 and @xmath253 : @xmath254 the number of _ real _ @xmath22-th cycles of the @xmath229-type , @xmath210-type and @xmath228-type , denoted @xmath255 , @xmath256 , and @xmath257 respectively , are given in table [ latable ] in appendix a. for the @xmath258 s one can not reduce , in contrast with the @xmath229-type or @xmath210-type analysis , the calculations to a only one variable : one is obliged to perform a first resultant calculation where one eliminates one of the two variables and another resultant calculation where one eliminates the other one , and check back , in the cartesian product of these possible solutions , the solutions which are actually fixed points . in order to get integer values that can be trusted
, one needs to perform these ( maple ) calculations with more than 2000 digits for order twelve , but then one faces severe memory limitations in the formal calculations .
we have been able to find integer values for the @xmath257 s for orders larger than twelve , however it is clear that these integers are just lower bounds of the true integers ( not enough precision does not enable to discriminate between fixed points that are very close ) . therefore we prefer not to give these integers here , and put a `` star '' in table [ latable ] , as well as in the forthcoming tables given in appendix a , when we encounter these computer limitations . the total number , @xmath259 , of _ real _ cycles of the @xmath229-type , @xmath260-type and @xmath210-type actually corresponds to the exponents in the product decomposition ( [ eps52 ] ) for the `` real dynamical zeta function '' .
unfortunately , these series are not large enough to `` guess '' any possible ( and simple , like ( [ conjec ] ) ) rational expression for @xmath261 , if any ... series ( [ eps52expand ] ) , however , give a first `` rough estimate '' for the `` real topological complexity '' @xmath262 : @xmath263 or may be @xmath264 , _ clearly smaller _ than the exact algebraic value for @xmath81 corresponding to ( [ conjec ] ) : @xmath265 .
let us consider other values of @xmath12 .
@xmath266 for @xmath267 , one finds out that _ all the fixed points seem to be real _ and , thus , one can conjecture for @xmath267 ( but @xmath268 ) : @xmath269 the number of cycles of the @xmath229-type , @xmath210-type and @xmath228-type is given order by order in table [ latable2 ] in appendix a. these successive integer values for the total number of irreducible real cycles , @xmath259 , yield : @xmath270 @xmath266 for @xmath271 , one has , at every order of iteration , up to order twelve , a _ only one real _ fixed point ( the fixed point of order one but , of course , many complex fixed points ) yielding : @xmath272 and , for @xmath273 very close to @xmath274 , the expressions of @xmath275 and @xmath276 can not be distinguished ( at the orders where we have been able to perform these fixed points calculations ) from : @xmath277 which just correspond to add an additional 3-cycle
. expressions ( [ eq15 ] ) are in agreement with the phase portrait of fig.[f : fig10 ] for @xmath278 .
this indicates that , seen as a mapping of two _ real variables _ , the mapping `` looks like '' an integrable mapping : the `` real topological complexity '' @xmath8 seems to be exactly equal to @xmath90 for @xmath190 ( and @xmath279 for @xmath280 ) .
the `` real topological entropy '' @xmath281 seems to be exactly zero for @xmath282 and is , thus , drastically different from the generic `` usual '' topological entropy @xmath283 .
the possible foliation of the two - dimensional space in ( transcendental ) curves is discussed elsewhere @xcite .
for other values of @xmath115 the series are not large enough to `` guess '' a rational expression ( if any ... ) for the `` real dynamical zeta function '' @xmath276 .
@xmath266 miscellaneous examples are given in appendix b. in particular the number of cycles of the @xmath229-type , @xmath210-type and @xmath228-type is given in table [ latable311 ] for @xmath284 , yielding the following expansion for the real dynamical zeta function : @xmath285 clearly yielding a value for @xmath286 close to one ( may be @xmath287 ) significantly smaller than @xmath288 .
this result has to be compared with the equivalent one for @xmath289 or for @xmath290 : @xmath291 yielding a larger value for @xmath286 : @xmath292 .
this expansion is actually compatible with the following simple rational expression and for its logarithmic derivative @xmath293 : @xmath294 note that all the coefficients of the expansion of the rational expression ( [ compat21sur25 ] ) and of its logarithmic derivative @xmath293 are positive ( in contrast with pade approximation ( [ simpledeuxsurtrois ] ) given in appendix c for @xmath295 which is ruled out because coefficient @xmath296 of its expansion is negative ) .
if this simple rational expression is actually the exact expression for the real dynamical zeta function @xmath297 this would yield the following algebraic value for @xmath298 : @xmath299 . for @xmath300
, one gets the same product decomposition , at least up to order ten .
the number of @xmath22-th cycles of the @xmath229-type , @xmath210-type and @xmath228-type for @xmath300 are given in appendix b. one thus sees that @xmath286 decreases when @xmath301 crosses the @xmath302 value .
@xmath303 for @xmath304 we have obtained ( see appendix b ) : @xmath305 the `` non - generic '' values @xmath306 and @xmath196 require a special and careful analysis ) , disappear on these very values stricto sensu : they become singular .
one has to verify carefully that all the points , obtained in such calculations , are fixed points and not singular points . ] .
however , similarly to what was seen for @xmath307 , one clearly verifies on all these `` real dynamical zeta function '' @xmath308 that the coefficients in these expansions are continuous in @xmath17 near these points except on these very values of @xmath12 where one gets smaller integers and thus smaller values for @xmath309 ( the limit on the left and on the right of @xmath309 are equal and larger than @xmath309 on these very `` non - generic '' values ) . the numbers of irreducible _ real _ @xmath22-cycles of the @xmath229-type , @xmath210-type and @xmath228-type are given in appendix b for miscellaneous values of @xmath17 : @xmath310 , @xmath311 , @xmath312 , @xmath313 .
we also give , in appendix c , for various values of @xmath301 ( @xmath314 , @xmath315 , @xmath316 , @xmath317 , @xmath318 , @xmath319 , @xmath320 ) , the product decomposition and expansions for @xmath321 up to order 11 .
similar calculations of the expansions of @xmath322 and @xmath323 , for many other values of the parameter @xmath12 , have been performed and will be detailed elsewhere .
all these results confirm that @xmath324 _ varies with _
@xmath30 when @xmath30 is positive , while @xmath8 is constant ( except @xmath325 ) when @xmath326 .
when @xmath30 is positive , the estimates of @xmath8 are in agreement with the `` visual complexity '' as seen on the phase portraits ( see the previous section ) .
in particular one finds that @xmath8 roughly decreases as a function of @xmath34 in the intervals @xmath327 $ ] and @xmath328 $ ] , and increases in the interval @xmath329 $ ] , ( with a sharp decrease near @xmath330 and @xmath331 ) , that @xmath8 is close or very close to one when @xmath17 belongs to an interval @xmath332 $ ] , that @xmath8 grows monotonically with @xmath17 for @xmath333 to reach some asymptotic value in the @xmath334 limit .
it will be seen , in the next subsection , that the `` real topological complexity '' @xmath335 , in the @xmath334 limit , tends to a value @xmath336 clearly different , again , from the generic `` topological complexity '' @xmath337 .
recalling the large number of rational expressions , obtained for the dynamical zeta functions @xcite and the degree generating functions @xcite , one may have a rationality `` prejudice '' for these `` real dynamical zeta functions '' @xmath323 , calculated for a _ given _ value of @xmath34 .
however , the occurrence of any symbolic dynamic , and associated markov partition , is _ far from being natural in this real analysis framework _ @xcite . if one bets on the rationality of the real dynamical zeta function @xmath252 ( see ( [ defreal ] ) ) , it must , however , be clear that @xmath323 depends on @xmath28 in a very complicated way ( piecewise continuous functions , devil s staircase ? ... ) .
if , for some given value of the parameter @xmath12 , the partial dynamical zeta function @xmath323 actually corresponds to a rational expression , one should , in fact , have an _ infinite set _ of such rational expressions associated with the infinite number of steps ( intervals ) ) , up to a ( zero measure ) set of values of @xmath34 ( see fig.[f : fig1 ] ) . ] in @xmath34 ) in the `` devil s staircase '' .
the actual location of these `` steps '' , that is , the limits of these intervals in @xmath34 , corresponds to an infinite number of values of @xmath34 like ( [ previous ] ) or ( [ moreprevious ] ) ( and others @xcite ) . for a given @xmath34 , the calculations of the first terms of the expansion of the `` real dynamical zeta function '' @xmath252 do not rule out , at the order for which we have been able to perform these calculations ( ten , eleven ) , _ rational expressions _ ( see for instance @xcite ) .
the number of _ real _ @xmath22-th cycles of the @xmath229-type , @xmath210-type and @xmath228-type , for @xmath338 , is given in table [ latable32 ] in appendix b. one remarks that , at order eleven , the number of irreducible _ real _ cycles , and therefore the expansion of the `` real dynamical zeta function '' are the same for @xmath339 for @xmath340 , @xmath341 one has a product expansion for the dynamical zeta function identical , up to order eleven , to the product expansion corresponding to the @xmath115 large limit ( see ( [ expanrealarnold1inf ] ) below ) .
these expansions are , however , different at order twelve , see appendix b. .3 cm @xmath303 * @xmath342 large . * for @xmath28 large enough one
gets the following cycle product decomposition : @xmath343 corresponding to the number of _ real _ @xmath229-type , @xmath210-type and @xmath228-type @xmath344 cycles for @xmath345 given in table [ latable3 ] in appendix b. note that one gets the same table ( up to order 16 ) for @xmath346 .
one finds out easily that these results , for the `` real dynamical zeta function '' @xmath347 , are ( up to order twelve ) actually in _ perfect agreement _ with ( the expansion of ) the rational expression : @xmath348 yielding an algebraic value for @xmath8 : @xmath349 .
if one `` believes '' in some symbolic dynamic coding interpretation , or in the existence of a linear transfer operator , matrix @xmath350 , such that the denominator of ( [ zetarealinfty ] ) , @xmath351 , can be written as @xmath352 , one finds that a possible choice for this transition matrix is : @xmath353\end{aligned}\ ] ] in contrast with a markov s transition matrix the previous matrix is not such that the sums of the entries in each row , or column , are equal .
let us recall , again , the identification between @xmath81 , the ( exponential of the ) topological entropy , and @xmath19 , the ( asymptotic of the ) arnold complexity @xcite .
similarly to the topological entropy , the arnold complexity can be `` adapted '' to define a `` _ real _ arnold complexity '' .
the arnold complexity counts the number of intersections between a fixed ( complex projective ) line and its @xmath1-th iterate @xcite : let us now count , here , the number of _ real _ points which are the intersections between a _ real fixed line _ and its @xmath1-th iterate .
with this restriction to _ real _ points we have lost `` most of the universality properties '' of the `` usual '' ( complex ) arnold complexity . for various values of @xmath30 , we have calculated the number of intersections of various ( real ) lines with their @xmath1-th iterates .
in contrast with the `` usual '' arnold complexity @xcite , which does not depend on the ( complex ) line one iterates ( topological invariance @xcite ) , it is clear that the number of real intersections depends on the chosen line , but one can expect that the _ asymptotic behavior _ of these numbers for @xmath6 large enough , will not depend too much of the actual choice of the ( real ) line one iterates . actually , we have discovered on this very example , that this seems to be the case ( except for some non generic lines ) .
furthermore , the real line @xmath354 , which is known to play a particular role for mapping ( [ yz ] ) ( see section ( [ creal ] ) ) , is very well - suited to perform these numbers of intersections calculations : for this particular line the successive numbers of intersections are extremely regular , thus enabling to better estimate this asymptotic behavior @xmath355 of the `` real arnold complexity '' , but of course similar calculations can be performed with an arbitrary ( generic ) line .
@xmath356 can be seen as the equivalent , for _ real mappings _ , of the growth complexity @xmath25 ( see section ( [ seminum ] ) and @xcite ) .
let us try to get @xmath357 for various values of @xmath12 .
similarly to the semi - numerical method detailed in section ( [ seminum ] ) , we have developed a c - program using again the multiprecision library gmp @xcite , counting the number of ( real ) intersections of the @xmath358 real line with its @xmath1-th iterate . this program does not calculate the precise location of the intersection points : it is based on the
_ sturm s theorem _ has no multiple roots , one can build a finite series of polynomials corresponding to the successive euclidean division of @xmath359 by its first derivative @xmath360 .
see for instance @xcite for more details on the sturm sequences and sturm s theorem . ] .
all these calculations have been cross - checked by a ( maple ) program calculating these numbers of intersections using the sturm procedure in maple , even the interval @xmath361\ , -\infty\ , , \ , \,+\infty\ , [ $ ] .
the procedure sturm uses _
sturm s theorem _ to return the number of real roots of polynomial p in the interval [ a , b ] .
the first argument of this sturm procedure is a _ sturm sequence _ for p , which can be obtained with another procedure , the procedure sturmseq which returns the sturm sequence as a list of polynomials and replaces multiple roots by single roots . ] .
the results of these calculations are given in fig.[f : fig11 ] .
let us denote by @xmath362 the number of ( real ) intersections for the @xmath1-th iterate .
in order to estimate `` real growth complexity '' @xmath363 we have plotted @xmath364 , for various values of the number of iterations ( @xmath365 ) , as a function of @xmath12 , in the range @xmath366 $ ] where @xmath367 has a quite `` rich '' behavior .
this behavior should be compared with the `` universal '' behavior of fig .
1 . on fig.[f : fig11 ] , one remarks that the singled out values @xmath368 , as well as @xmath369 for @xmath370 , seem , again , to play a special role in the large @xmath1 limit .
of course , recalling the results of section ( [ creal ] ) , it is clear that @xmath371 is a staircase function of @xmath41 , for @xmath1 finite , the limits of each interval corresponding to _ algebraic values _ ( like ( [ previous ] ) ) sketched in sections ( [ bsome ] ) , ( [ creal ] ) .
these algebraic values form an
_ infinite set _ of values which accumulate everywhere in the @xmath372 $ ] interval .
what is the limit of these functions @xmath373 when @xmath6 gets large : a devil s staircase or a ( piecewise ) continuous function ?
the `` shape '' of @xmath374 , as a function of @xmath375 , is `` monotonic enough '' ( see fig.[f : fig12 ] below ) in different intervals , namely @xmath376 , and in the intervals of @xmath87 roughly given by : @xmath377 $ ] , @xmath378 $ ] , @xmath379 $ ] , @xmath380 $ ] and @xmath381 $ ] , that one may expect that the infinite accumulation of these algebraic values ( like ( [ previous ] ) ) could yield a perfectly continuous function @xmath382 ( except on the non - generic values @xmath383 and @xmath384 ) and not a devil s staircase - like function .
this question remains open at the present moment .
when @xmath34 varies from @xmath385 to @xmath386 the behavior of the `` real growth complexity '' @xmath387 , as a function of @xmath194 , is not as `` rich '' as in the interval @xmath388 $ ] depicted in fig.[f : fig11 ] .
one finds that @xmath367 is close , or extremely close to @xmath90 , in a quite large interval @xmath380 $ ] and that it increases monotonically with @xmath12 in the @xmath389 $ ] interval to reach some asymptotic value in the @xmath390 limit .
in fact , a logarithmic scale in @xmath12 is better suited to describe @xmath367 as a function of @xmath391 .
fig.[f : fig12 ] represents @xmath387 , more precisely @xmath392 , as a function of @xmath393 .
for @xmath394 we know that @xmath367 will be exactly equal to @xmath234 ( integrable cases @xcite ) . on these points
( represented by squares on fig.[f : fig12 ] ) , as well as on the @xmath116 and @xmath384 non - generic points , @xmath363 is _ not continuous _ as a function of @xmath12 .
we have not represented these other non generic points .
they have to be calculated separately .
a first estimate of @xmath387 , for @xmath34 large , is @xmath395 .
we are now ready to compare the `` real topological entropy '' and the `` real arnold complexity '' for different values of @xmath375 , and see if the identification , between @xmath396 , ( the exponential of ) the topological entropy , and @xmath19 , characterizing the ( asymptotic behavior of the ) usual arnold complexity , also holds for their `` real partners '' namely @xmath397 and @xmath398 .
actually , one finds that this identification ( which is obviously true for @xmath399 ) also holds for @xmath190 and give numerical results for various values of @xmath28 ( for which we have estimated the `` real '' topological entropy @xmath400 ( see section ( [ bsome ] ) ) ) , _ quite compatible with this identification_. in particular , for @xmath342 large , we do see that these two `` real complexities '' _ give extremely close results _ , namely @xmath401 versus @xmath402 .
similarly to the introduction of the `` real dynamical zeta functions '' @xmath347 , one can introduce the generating function of the previous `` real arnold complexities '' @xmath403 : @xmath404 recalling the large number of rational expressions , obtained for the dynamical zeta functions @xcite and the degree generating functions @xcite , one may have , again , a rationality `` prejudice '' for these `` real arnold complexity generating functions '' .
let us try to see if the expansions of these generating functions @xmath405 could coincide , for some given values of @xmath12 , with the expansion of some ( hopefully simple ) rational expressions .
in order to compare more carefully @xmath8 and @xmath398 , and find some possible non trivial rational expressions for @xmath406 , let us give , in the following , miscellaneous expansions of @xmath405 for various values of @xmath12 .
in fact , we have not only calculated the real arnold complexities @xmath407 , @xmath408 and @xmath409 required to plot fig.[f : fig11 ] and fig.[f : fig12 ] , but actually obtained all the coefficients up to order 13 for 2000 values of @xmath34 , and up to order 15 for 200 values of @xmath34 .
we thus have the expansion of @xmath405 up to order 13 ( resp .
15 ) for several thousands of values of @xmath12 .
let us first give the expansion of @xmath405 for @xmath410 for which the expansion of the real dynamical zeta function has been given previously ( see ( [ eps52 ] ) and ( [ eps52expand ] ) ) .
one gets the following expansion : @xmath411 this series yields a first rough approximation of @xmath367 corresponding to @xmath412 , clearly smaller compared to the generic complexity @xmath413 , and in good enough agreement with the estimation of @xmath8 one can deduce from ( [ eps52expand ] ) , namely @xmath414 .
of course these two series are to short to see if an identity like @xmath415 really holds . considering @xmath397 as a function of @xmath30 , it is clear that the general shape of this `` curve '' looks extremely similar to the curve corresponding to @xmath367 as a function of @xmath28 ( see fig.[f : fig11 ] and fig.[f : fig12 ] ) : it is also constant for @xmath28 negative , gets to close to 1 around @xmath416 , grows monotonically for @xmath417 and tends to a non - trivial asymptotic value @xmath418 .
therefore , in order to get some hint on the relevance of a possible @xmath415 identity , it is necessary to see if this relation holds for various values of @xmath17 for which @xmath309 , and @xmath419 , can be calculated exactly or for which very good approximations of can be obtained , namely @xmath420 , all the integrable values , or @xmath421 and its neighborhood ... for _ any negative value _ of @xmath12 ( except @xmath422 see below ( [ expanrealarnold3 ] ) ) the expansion of the `` real arnold complexity '' generating function @xmath423 is equal , up to order 15 , to : @xmath424 in agreement with ( [ zetarealepsneg ] ) .
for @xmath425 the generating function for the `` real arnold complexity '' generating function @xmath423 is equal , up to order 15 , to the simple _ rational _ expression : @xmath426 which is in perfect agreement with the result for the `` real dynamical zeta function '' @xmath427 ( see ( [ eq15 ] ) ) . for @xmath273 very close to @xmath274
one gets : @xmath428 again in good agreement with ( [ eq16 ] ) .
.3 cm * integrable values for @xmath12 and around . * for @xmath429 the generating function for the `` real arnold complexity '' , @xmath430 ,
is equal , up to order 38 , to the expansion of the rational expression : @xmath431 to be compared with @xmath432 given in ( [ corner ] ) .
for @xmath433 the calculations corresponding to the generating function for the `` real arnold complexity '' are , in contrast , quite trivial yielding : @xmath434 for @xmath435 the generating function for the `` real arnold complexity '' is equal , up to order 15 , to the rational expression : @xmath436 for @xmath437 the generating function for the `` real arnold complexity '' is equal , up to order 15 , to the rational expression : @xmath438 all these results have to be compared with the generating functions ( [ corner ] ) .
.3 cm * non - generic values for @xmath12 and around . *
the non - generic values of @xmath28 require a special attention .
for instance for @xmath439 one obtains the following expansion : @xmath440 and for @xmath441 one gets : @xmath442 since @xmath443 is a non - generic value ( it is of the form @xmath112 ) , the previous expansion ( [ expanrealarnold17 ] ) can be compared with the ones corresponding to values very close to @xmath444 but not equal , for instance @xmath445 , and @xmath446 : @xmath447 and : @xmath448 closer to @xmath443 one obtains for @xmath449 , and @xmath450 , the following expansions : @xmath451 and : @xmath452 similar expansions , corresponding to values close to the non - generic value @xmath304 , are given in appendix d. all these results show that , similarly to the situation for the customary topological entropy , or the growth complexity @xmath4 ( see fig.[f : fig1 ] ) , @xmath367 is continuous as a function of @xmath34 _ near _ the non - generic values of @xmath453 , however exactly on these very non - generic values @xmath367 takes _ smaller _ values ( continuous function up to a zero measure set ) .
.3 cm * remark :* it is natural to compare the expansion corresponding to @xmath454 with the one corresponding to @xmath455 , since @xmath456 and @xmath454 have the _ same topological entropy _ ( growth complexity @xmath4 ) associated with @xmath457 for @xmath105 ( see relation ( [ zetamtexte ] ) ) .
one gets for @xmath454 : @xmath458 and for @xmath455 : @xmath459 these two expansions _ do not seem to yield the same value for _ @xmath398 ( @xmath460 and @xmath461 ) _ though they share the same growth complexity _ @xmath18 .
.3 cm the expansions of @xmath405 near @xmath106 are given in appendix d. .3 cm * miscellaneous values of @xmath12 .
* for most of the values of @xmath17 the expansions are not long enough to `` guess '' rational expressions ( if any ... ) .
one can however get some estimates of @xmath398 that can be compared with @xmath309 .
@xmath462 for @xmath463 one gets the following results : @xmath464 this expansions seem to yield the following estimated value for @xmath398 ( @xmath465 to be compared with ( [ 21sur25 ] ) ) .
in fact , one remarks that this expansion is _ actually compatible _ with the expansion of the rational expression : @xmath466 one does remark that the rational expression ( [ interesting2 ] ) has actually the _ same singularity _ that the rational expression ( [ compat21sur25 ] ) suggested for the real dynamical zeta function @xmath467 .
all the coefficients of the expansion of ( [ interesting2 ] ) are positive ( in contrast with ( [ simpledeuxsurtrois ] ) given in appendix c which is ruled out because coefficient @xmath468 of its expansion is negative ) . if this simple rational expression is actually the exact expression for @xmath469 this would yield the following algebraic value for @xmath470 : @xmath471 . .2
cm @xmath462 for @xmath472 one gets the following results : @xmath473 @xmath462 near @xmath474 ( for instance for @xmath475 or @xmath476 ) one gets : @xmath477 @xmath462 some results for @xmath28 larger than 3 ( again obtained with 6000 digits ) are given in appendix d ( @xmath478 , @xmath479 , @xmath480 , @xmath481 , @xmath482 , @xmath483 ) .
these series indicate that an estimated value for @xmath367 could correspond to @xmath367 very close to 1 for @xmath484 , and quite close to 1 for @xmath485 .
the examination of fig.[f : fig12 ] shows that @xmath363 goes to some non - trivial limit , @xmath486 , in the large @xmath342 limit .
let us examine the expansion of @xmath487 for various increasing values of @xmath12 , in order to study this @xmath488 limit .
the expansions of @xmath405 for @xmath489 , @xmath490 , @xmath491 are given in appendix d. .3 cm @xmath492 for @xmath493 the expansion of @xmath405 reads : @xmath494 @xmath492 for @xmath495 : @xmath496 for @xmath28 large the expansion of @xmath497 , the generating function for the `` real arnold complexity '' , is equal , up to order 15 , to ( for instance for @xmath498 ) : @xmath499 which actually coincides with the expansion of the simple rational expression : @xmath500 this last result has to be compared with the equivalent one for the `` real dynamical zeta function '' @xmath501 ( see ( [ zetarealinfty ] ) in section ( [ seek ] ) ) .
these two non - trivial rational results , for @xmath34 large , are in _ perfect agreement _ , yielding the same _ algebraic value _ for the two `` real complexities '' @xmath502 and @xmath503 , namely @xmath504 .
.3 cm all the results displayed in this section seem to show that the identification between @xmath502 and @xmath503 actually holds .
.3 cm * remark :* recalling the `` universal '' relation ( [ universal ] ) , or more precisely ( [ universal2 ] ) , which gives ( for @xmath17 generic and for @xmath159 , @xmath160 for @xmath161 ) a ratio @xmath505 equal to @xmath506 , one can look at the `` real ratio '' @xmath507 . of course for @xmath508
this `` real ratio '' is also equal to @xmath506 , however in the @xmath509 limit it tends to be equal to @xmath510
. therefore one should not expect any simple `` universal '' relation like ( [ universal ] ) between @xmath511 and @xmath487 .
these various arnold complexity generating functions @xmath487 were associated with the iteration of the ( real or complex ) line @xmath202 .
one could introduce a generating function for each line ( or fixed curve ) one iterates .
the corresponding series become slightly more difficult to extrapolate but give similar results in particular the asymptotic values for @xmath398 .
the sensitivity of the previous analysis , according to the chosen curve one iterates , will be discussed elsewhere
. it would be interesting to see if the `` real dynamical zeta functions '' @xmath511 , or the `` real degree generating functions '' @xmath512 , could also be _ rational _ expressions , for _ other _ values of @xmath12 , or , even , if these `` real generating functions '' could be _ rational _ expressions for _ any given _ value of @xmath12 . in this last case
there should be an infinite number of such _ rational _ expressions : it is clear that they could not all be `` simple '' like ( [ zetarealinfty ] ) or ( [ expanrealarnold1 ] ) .
the results presented here seem to be _ in agreement with , again , an identification between _ @xmath398 , the ( asymptotic ) `` _ real arnold complexity _ '' , and @xmath513 , the ( exponential of the ) _ real topological entropy_. in contrast with the `` universal '' behavior of the `` usual '' arnold complexity , or topological entropy , displayed in fig . 1 , @xmath367 and @xmath514 are quite involved functions of the parameter @xmath17 the birational transformations depend on ( see fig .
11 and fig .
we have , however , obtained some remarkable _ rational expressions for the real dynamical zeta function _ @xmath515 and for a `` real arnold complexity '' generating function @xmath512 .
in particular we have obtained in particular two non - trivial _ rational _ expressions ( [ zetarealinfty ] ) and ( [ expanrealarnold1 ] ) , ( yielding algebraic values for @xmath502 and @xmath503 ) . there is no simple down - to - earth markov partition , symbolic dynamics , or hyperbolic systems interpretation of these rational results .
let us consider here various non - generic values of the form @xmath113 ( with @xmath516 , @xmath37 odd ) .
@xmath303 for @xmath106 ( corresponding to @xmath107 ) the homogeneous generating function defined in section ( [ dynzeta ] ) , reads : @xmath517 which is compatible with the expansion of the rational expression : @xmath518 recalling a possible rational expression for the corresponding dynamical zeta function @xcite : @xmath519 one immediately verifies that the `` universal '' relation ( [ universal ] ) actually holds .
@xmath303 for @xmath520 ( corresponding to @xmath521 ) the homogeneous generating function defined in section ( [ dynzeta ] ) , reads : @xmath522 this could be the expansion , up to order twelve , of the simple rational expression : @xmath523 these results should be compared with the expansion of the dynamical zeta function .
unfortunately , here , the series for the dynamical zeta function are not sufficiently large to allow any `` safe conjecture '' .
a possible exact expression does not seem to be equal to @xmath524 , but could be @xcite : @xmath525 the `` universal '' relation ( [ universal ] ) is verified with ( [ gh2sur3could ] ) together with @xmath526 , _ but not with _ ( [ gh2sur3could ] ) together with ( [ zet2sur3sec ] ) .
one can however imagine that the `` universal '' relation ( [ universal ] ) could be slightly modified on some of these @xmath113 values ( @xmath527 ) .
for instance , ( [ gh2sur3could ] ) and ( [ zet2sur3sec ] ) verify ( up to order twelve ) the simple relation : @xmath528 these calculations need to be revisited . @xmath303 for @xmath529 ( corresponding to @xmath530 ) the homogeneous generating function defined in section ( [ dynzeta ] ) , reads : @xmath531 this could be the expansion of : @xmath532 @xmath303 for @xmath533 ( corresponding to @xmath534 ) the homogeneous generating function reads : @xmath535 this series is not large enough
. it could be the expansion of the simple expression : @xmath536
let us just give the number of _ real _ @xmath22-th cycles of the @xmath229-type , @xmath210-type and @xmath228-type for miscellaneous values of @xmath34 in increasing order
. for @xmath539 ( and @xmath540 ) one gets : .number of _ real _ @xmath22-th cycles of the @xmath229-type , @xmath210-type and @xmath228-type for @xmath539 .
[ latable2 ] [ cols="^,^,^,^,^,^,^,^,^,^,^,^,^,^,^,^,^,^,^",options="header " , ]
let us just give some additional expansions for @xmath321 for increasing values of @xmath12 . for @xmath314 ,
one obtains the following expansions for @xmath321 : @xmath541 yielding the following `` rough '' approximation for @xmath286 : @xmath542 . for the `` non - generic '' value @xmath543
, @xmath544 reads : @xmath545 yielding the following `` rough '' approximation for @xmath286 : @xmath546 . for @xmath543 the previous @xmath547 s and @xmath258 s
are equal to zero up to order twelve .
the exponents in ( [ unsur5 ] ) are thus the @xmath548 s . for @xmath549 , @xmath316 , @xmath550 , @xmath551 , @xmath552 , @xmath553 , @xmath554 , one obtains , respectively , the following expansions for @xmath321 : @xmath555 yielding the approximation for @xmath286 : @xmath556 , @xmath557 for @xmath316 , yielding the following rough approximation for @xmath286 : @xmath558 , @xmath559 for @xmath560 , yielding @xmath561 . for @xmath520 ( that is @xmath562 for @xmath521 )
the real dynamical zeta function reads : @xmath563 yielding @xmath564 .
let us note that one must be careful converting systematically a series to a rational function ( pade approximation ) .
up to order twelve , expansion ( [ deuxsurtrois ] ) is in agreement with the expansion of the following simple rational expression : @xmath565 which is reminiscent of the exact expression ( [ zetarealinfty ] ) .
however , one easily finds that the coefficients of @xmath566 in ( [ simpledeuxsurtrois ] ) becomes negative ( the coefficients grow like @xmath567 ) .
expression ( [ simpledeuxsurtrois ] ) _ can not _ be the exact expression of a ( real ) dynamical zeta function . for @xmath568
, the real dynamical zeta function reads : @xmath569 yielding : @xmath570 , for the non - generic value @xmath533 ( that is @xmath562 for @xmath571 ) the real dynamical zeta function reads : @xmath572 yielding @xmath573 .
finally , for @xmath574 , the real dynamical zeta function reads : @xmath575 yielding : @xmath576 .
let us give here a few expansions for the `` real arnold complexity '' generating functions @xmath512 .
let us first give the expansion of @xmath487 corresponding to @xmath577 in order to compare it with ( [ deuxsurtrois ] ) and ( [ simpledeuxsurtrois ] ) : @xmath578 yielding the following estimation for @xmath579 to be compared with @xmath580 from ( [ deuxsurtrois ] ) .
the coefficients of the expansions of @xmath581 , and @xmath582 , are very close . up to order ten , the ratio @xmath583 coincides with the expansion of : @xmath584 the expansion of @xmath487 corresponding to the non - generic value @xmath585 ( that is @xmath113 for @xmath534 ) reads : @xmath586 the expansions of @xmath587 and @xmath588 are again very close .
expansion ( [ expanrealarnoldsimpletroissurquatre ] ) yields the following estimation for @xmath589 to be compared with @xmath590 from ( [ troissurquatre ] ) .
n. abarenkova , j - c .
angls dauriac , s. boukraa , s. hassani and j - m .
maillard , _ from yang - baxter equations to dynamical zeta functions for birational transformations _ , to be published by world scientific n. abarenkova , j - c .
angls dauriac , s. boukraa and j - m .
maillard , _ elliptic curves from finite order recursions or non - involutive permutations for discrete dynamical systems and lattice statistical mechanics_. the european physical journal * b 5 * ( 1998 ) pp .
647661 .
angls dauriac , s. boukraa , and j - m .
maillard , _ functional relations in lattice statistical mechanics , enumerative combinatorics and discrete dynamical systems _ , to be published in annals of combinatorics . | we consider a family of birational transformations of two variables , depending on one parameter , for which simple rational expressions with integer coefficients , for the exact expression of the dynamical zeta function , have been conjectured . together with an equality between the ( asymptotic of the ) arnold complexity and the ( exponential of the ) topological entropy .
this identification takes place for the birational mapping seen as a mapping bearing on two complex variables ( acting in a complex projective space ) .
we revisit this identification between these two quite `` universal complexities '' by considering now the mapping as a mapping bearing on two _ real _ variables .
the definitions of the two previous `` topological '' complexities ( arnold complexity and topological entropy ) are modified according to this real - variables point of view .
most of the `` universality '' is lost
. however , the results presented here are , again , in agreement with an identification between the ( asymptotic of some ) `` real arnold complexity '' and the ( exponential of some ) `` real topological entropy '' .
a detailed analysis of this `` real arnold complexity '' as a function of the parameter of this family of birational transformations of two variables is given .
one can also slightly modify the definition of the dynamical zeta function , introducing a `` real dynamical zeta function '' associated with the counting of the real cycles only .
similarly one can also introduce some `` real arnold complexity '' generating functions .
we show that several of these two `` real '' generating functions seem to have the same singularities .
furthermore we actually conjecture several simple rational expressions for them , yielding again algebraic values for the ( exponential of the ) `` real topological entropy '' .
in particular , when the parameter of our family of birational transformations becomes large , we obtain two interesting compatible non trivial rational expressions .
these rational results for real mappings can not be understood by any obvious markov s partition , or symbolic dynamics hyperbolic systems interpretation .
.2 cm .2 cm pacs numbers : 05.45.+b , 03.20 , 46.10 , 47.52.+j , 05.50.+q , 02.90.+p .2 cm .2 cm * key words :* arnold complexity , topological entropy , discrete dynamical systems of real variables , birational mappings , cremona transformations , rational dynamical zeta functions , complex mappings versus real mappings . |
let @xmath0 be a locally finite covering of the plane @xmath6 by closed unit discs .
the doubly covered region of @xmath6 by @xmath0 consists of the sets of points @xmath7 that are contained in at least two elements of @xmath0 .
a few years ago gabor fejes - tth posed the following question : if @xmath8 are two points at distance @xmath3 apart and contained in the doubly covered region of @xmath6 , what is the length of the shortest path @xmath9 joining @xmath1 and @xmath2 that is completely contained in the doubly covered region ? in a sense , this is the dual of a problem by laszlo fejes tth about the length of a path avoiding a packing of discs @xcite .
gabor fejes - tth conjectured that when the centres of the circles in @xmath0 form a unit square lattice , the length of @xmath9 is maximal .
for any two points @xmath1 and @xmath2 in this example , @xmath10
. a general upper bound of @xmath11 for @xmath12 is not difficult to obtain .
baggett and bezdek proved in @xcite that when the centres of the circles of @xmath0 form a lattice , then the unit square lattice is indeed the extreme case .
in this short note we give an upper bound for any locally finite covering .
@xmath13 this is still closer to @xmath14 than to @xmath15 , but we hope that our methods may help others to continue improving this bound .
let @xmath16 be the segment joining @xmath1 and @xmath2 and @xmath17 be a minimal sub - cover of @xmath16 .
suppose that @xmath16 is horizontal and that @xmath1 is to the left of @xmath2 .
the elements of @xmath18 can be ordered as @xmath19 such that if @xmath20 then the centre @xmath21 of @xmath22 is to the left of the centre @xmath23 of @xmath24 .
we may assume that @xmath1 is the leftmost point of the intersection of the line @xmath25 with @xmath26 and that @xmath2 is the rightmost point of the intersection of @xmath25 with @xmath27 .
if this were not the case then we may extend the length of @xmath16 by at most @xmath28 and move each one of @xmath1 and @xmath2 through paths of length at most @xmath29 so that they end up in this way .
this contributes a term @xmath30 to the length of the curve we find with respect to the original length of the segment @xmath16 . since the family @xmath0 is locally finite , every point in the boundary of a circle @xmath31
is doubly covered . for @xmath31 , define @xmath32 as the closure of a connected component of @xmath33 that does not contain the centre of @xmath34 . since @xmath18 is a minimal covering of @xmath16 then @xmath35 if and only if @xmath36 .
let @xmath37 and @xmath38 be the points of intersection of the boundary of @xmath22 with @xmath16 such that @xmath37 is to the left of @xmath38 and let @xmath39 be the midpoint of the segment @xmath40 for @xmath41 , @xmath42 and @xmath43 .
note that @xmath44 . .
] now we construct the path @xmath9 with the algorithm below , an example is shown in figure [ fig : gamma ] .
the path starts at @xmath45 . assuming the path has been constructed up to @xmath39 , let @xmath46 be the largest integer such that the sets @xmath47 are all on the same side of @xmath16 . without loss of generality
we assume they are all above @xmath16 .
the path then continues vertically upwards until it reaches the boundary of @xmath48 .
from here it continues towards the right while staying contained in the boundary of @xmath49 until it is vertically aligned with @xmath50 .
finally it goes vertically downwards until it reaches @xmath50 .
this is repeated until the path ends at @xmath51 .
it is easy to see that @xmath9 is well defined and completely contained in the doubly covered region , however it is not so easy to directly bound its length .
this path could be shortened by taking diagonal lines instead of vertical ones , but these two paths coincide in the extreme case . in order to bound the length of this path , for every @xmath52
we construct a new path @xmath53 .
the paths @xmath53 may not be contained in the doubly covered region but they satisfy @xmath54 .
assume for simplicity that @xmath55 is above @xmath16 , then the path @xmath53 starts at @xmath56 , goes upwards until it intersects the boundary of @xmath55 at a point @xmath57 , then goes to the right staying contained in the boundary of @xmath55 until it is vertically aligned with @xmath39 at the point @xmath58 , and finally goes downwards until it reaches @xmath39 .
this path is shown in figure [ fig : gammai ] .
it is not difficult to see that @xmath54 . and
@xmath59 . ]
now it is enough to prove the following .
[ lem : lemma ] @xmath60 we may assume that @xmath61 is the diameter of @xmath22 , otherwise let @xmath32 be a circle with diameter @xmath61 and @xmath59 be the curve defined on @xmath32 in the same way that @xmath53 is defined on @xmath22 .
this new curve clearly has larger length than @xmath53 ( see figure [ fig : gammai ] ) .
let @xmath62 and @xmath63 , then @xmath64 recall also that @xmath65 .
if we fix the value of @xmath66 , then by using lagrange multipliers we obtain that @xmath67 is maximum when @xmath68 , @xmath69 or @xmath70 .
since the cases @xmath68 and @xmath69 are symmetrical , we have essentially two cases . now we only need to determine the maximum of @xmath71 as a function of @xmath72 .
this occurs when @xmath73 and corresponds to the case @xmath70 .
this gives @xmath74 which proves the lemma .
the method we use only considers discs that intersect the segment @xmath25 , we construct a path contained in the boundary of these discs and in their doubly covered region . considering only these circles it is impossible to obtain the bound fejes - tth conjectured .
below we construct an example considering only these circles such that @xmath75 .
let @xmath26 and @xmath76 be intersecting circles and assume @xmath77 and @xmath78 .
if we only allow the path @xmath9 to be in the intersection of the circles and their boundary , then there are only @xmath28 possible choices for @xmath9 .
this is a simplified version of our problem , as the boundary of a circle in @xmath0 is always doubly covered .
but because we only consider two circles , this new problem can be solved precisely to obtain @xmath79 .
here the number @xmath80 is an approximation obtained as the maximum of a function involving trigonometric functions .
the example that gives this value can be extended to families with more circles and have arbitrarily large @xmath3 .
this is done by using several translated copies of this example ( see fig .
[ fig : example ] ) . .
] thus , it is essential to consider elements of @xmath81 further away from the segment @xmath25 .
however an improvement on our algorithm would surely lower our bound .
to obtain our bound we consider only @xmath82 discs at a time , this is precisely what is done in lemma [ lem : lemma ] .
a this bound could potentially be lowered if we were to consider @xmath83 or more discs instead .
the algorithm would have to be modified or changed to take advantage of this .
we tried this approach together with alexey garber , however we were unable to obtain any substantial improvements .
a problem we encountered was that the functions to be minimised became extremely complicated . | given a covering of the plane by closed unit discs @xmath0 and two points @xmath1 and @xmath2 in the region doubly covered by @xmath0 , what is the length of the shortest path connecting them that stays within the doubly covered region ?
this is a problem of g. fejes - tth and he conjectured that if the distance between @xmath1 and @xmath2 is @xmath3 , then the length of this path is at most @xmath4 . in this paper
we give a bound of @xmath5 . |
the origin of cooperation is a fundamental problem in evolutionary biology .
cooperation is essential in the functioning of almost every known biological system ( hamilton ( 1964a ) , hamilton ( 1964b ) , trivers ( 1971 ) , dugatkin ( 1997 ) ) .
for example , according to eigen & schuster ( 1979 ) , michod ( 1983 ) , and maynard smith & szathmry ( 1995 ) , early replicating molecules may have cooperated to form larger entities which could encode more information .
also , the transition from free - living single - cell protists to multicellular organisms seems to have depended on cooperation ( maynard smith & szathmry ( 1995 ) , buss ( 1987 ) ) .
it is however , difficult to explain why individuals should cooperate . in the traditional prisoner s dilemma model of cooperation
, defecting individuals always have a higher fitness than cooperators .
cooperation is not an evolutionary stable strategy , because it can be invaded by defectors .
hence , the emergence of cooperation is generally assumed to require repeated play ( with memory ) and strategies such as tit for tat , or `` tags '' ( axelrod ( 1984 ) , guttman ( 1996 ) , lindgren & nordahl ( 1994 ) , miller ( 1996 ) ) .
the work of nowak & may ( 1992 ) showed that placing ensembles of cooperators and defectors on a lattice generates changing spatial patterns , in which both cooperators and defectors persist indefinitely .
the introduction of spatial structure changes the picture from the mean - field result in which defection always wins to a final state with both cooperators and defectors present .
similar results were obtained by epstein ( 1998 ) , who introduced the demographic prisoner s dilemma , in which the individuals have a fixed strategy ( which is their phenotype ) , but are placed in a spatially structured lattice world .
epstein ( 1998 ) found that regions of cooperation persisted in this spatial model .
the studies of nakamaru et al .
( 1997 ) , iwasa et al .
( 1998 ) , nakamaru et al .
( 1998 ) , and irwin and taylor ( 2001 ) showed that spatially structured models , such as the lattice model , produce the clumping of the cooperative players , and then enables them to invade a population of defectors , but the spatial structure also encourages the evolution of spiteful behaviour .
these models consider the invasiveness and stability of fully developed , highly cooperative interactions .
the gradual evolution of cooperation from an initially selfish state represents a more plausible evolutionary scenario .
it is then more natural to consider models in which several degrees of cooperation are possible ( doebeli & knowlton ( 1998 ) , roberts & sherratt ( 1998 ) , wahl & nowak ( 1999a ) , wahl & nowak ( 1999b ) , szab & hauert ( 2002a ) , szab & hauert ( 2002b ) , killingback & doebeli ( 2002 ) ) .
when we take into account the possibility of variable levels of cooperation , we can study the crucial issue of how cooperation can gradually evolve from a non - cooperative initial state .
roberts & sherratt ( 1998 ) considered a `` raise - the - stakes '' strategy for the iterated pd , and showed that it invades and is stable against a number of alternative strategies .
doebeli & knowlton ( 1998 ) considered interspecific symbiosis in the context of iterated asymmetric pd , and concluded that such interactions could increase in extent and frequency if the populations are spatially structured . in this model , strategies with very low levels of cooperation can gradually evolve to much more cooperative strategies .
the end result is a high degree of mutualism between pairs of interacting individuals that belong to different species .
killingback et al .
( 1999 ) extended the classical prisoner s dilemma , introducing a model of cooperation which is based on the concept of investment , and develops further the ideas of doebeli & knowlton ( 1998 ) .
this evolutionary game is called continuous prisoner s dilemma ( cpd ) .
killingback et al .
( 1999 ) showed that intraspecific cooperation easily evolves from very low levels , and is sustained , with fluctuations , at relatively high levels , when the game is played in spatially structured populations .
killingback et al .
( 1999 ) assume that individuals play against their immediate neighbours , and also compare their payoffs to those of the same individual neighbours .
it is important to know how robust are the results obtained by killingback et al .
( 1999 ) when these assumptions are relaxed , i.e. when individuals are allowed to play against more distant neighbours ( than their nearest ones ) , and then compare their payoffs to those of a different group of neighbours , which may be larger or smaller than the ones included in the first interaction neighbourhood . also
, killingback et al .
( 1999 ) conjecture that clustering is the mechanism that allows the establishment and maintenance of a cooperative state . to investigate the validity of this hypothesis
, we studied the behaviour of the cpd game on different topologies , such as networks with different clustering properties .
the continuous prisoner s dilemma ( cpd ) game between two individuals is based on the assumption that each of them makes an investment ( which can take any non - negative real value ) .
making an investment @xmath0 has the effect of reducing the fitness of the individual who makes it by `` the cost '' @xmath1 and increasing the fitness of the beneficiary by `` the benefit '' @xmath2 .
so , if two individuals 1 and 2 , play against each other and make investments @xmath3 and @xmath4 , the payoff of 1 is @xmath5 and that of 2 is @xmath6 .
possible benefit and cost functions are shown in fig . [
fig : fig1 ] .
cost and benefit functions of this type are typical of what might be expected in a real biological situation , such as those discussed by hart & hart ( 1992 ) and wilkinson ( 1984 ) . the common feature of the functions @xmath1 and
@xmath2 is that @xmath7 for a range of the argument @xmath8 ( for @xmath9 the cost is higher than the benefit , so it does not make sense for a player to invest an amount higher than @xmath10 ) . limiting the investment levels to just a pair of investment amounts would bring us back to the standard prisoner s dilemma . in the absence of an additional structure ,
i.e. in the mean - field approximation , or the well - mixed system , the zero - investment strategy is again the winning one .
starting at any level , investments will gradually evolve to zero , since the defectors will benefit from the investment of the cooperators without bearing the costs . using adaptive dynamics , one can see that the zero investment strategy is the evolutionary stable strategy .
killingback et al . ( 1999 ) introduced spatial structure into the model , following the general approach of spatial evolutionary game theory ( axelrod ( 1984 ) , nowak & may ( 1992 ) , killingback & doebeli ( 1996 ) ) .
the individuals are placed in the cells of a 2d square lattice .
each individual makes an investment , and interacts with the individuals within a moore neighbourhood of hers .
a moore neighbourhood of , e.g. size one , includes a cell s eight immediate neighbours : north , northeast , east , southeast , south , southwest , west , northwest , i.e. a `` ring '' of `` thickness '' one neighbour in all directions ; that of size 2 will include all individuals within a `` ring '' of `` thickness '' two neighbours in all directions , etc . in this paper
we will refer to the thickness of the interaction neighbourhood as the * neighbourhood parameter*. the individual will get payoffs as prescribed by the rules of the game , when playing against each of the neighbours within her interaction neighbourhood .
her fitness is then the sum of the payoffs she gets from playing against all her neighbours .
( in the process , the fitness of the neighbours is calculated too , as the sum of the payoffs they get from playing against the individuals within their own interaction neighbourhoods . )
she then compares her payoff to those of the individuals within her learning neighbourhood , whose size ( thickness ) may be the same or different from that of the interaction neighbourhood . in traditional spatial evolutionary game theory , this distinction between the individual s interaction and learning neighbourhood is not made , silently assuming that they are the same . in this paper , we will refer to the thickness of the learning neighbourhood as the * dispersal parameter*. our focal individual then identifies the neighbour with the highest fitness , and adopts that neighbour s strategy , by changing the amount of her investment to that of the neighbour s . at this stage the investment level of the player can be mutated at a fixed probability .
this corresponds to an economic scenario in which the agents learn ( with occasional errors ) from their more successful partners , or to an evolutionary scenario in which the more successful phenotypes replace the less successful ones ( with mutations ) . every few generations the average investment per individual is calculated .
killingback et al .
( 1999 ) simulated the cpd game in a square lattice with neighbourhood sizes for both interacting and learning ( i.e. both the neighbourhood and dispersal parameters ) set to one .
they observed that the average investment increases from an almost zero starting value , to a significant final level of investment .
its average value for a cost function @xmath11 and benefit function @xmath12 , in a @xmath13 lattice , with periodic boundary conditions , is around @xmath14 .
( the probability of mutations is set to @xmath15 , and they follow a gaussian curve with centre at the individual s investment value , and width 10 percent of its peak . ) the average investment amount is maintained close to @xmath14 by the dynamics of the spatial system . it is smaller than the maximum investment @xmath10 , but much larger than the initial maximum investment amount of 0.0001 .
the key factor that causes a state with considerable average investment levels to be established is the fact that higher investing individuals benefit from clustering together ( killingback et al .
( 1999 ) ) .
this result agrees with those of nakamaru et al .
( 1997 ) , iwasa et al .
( 1998 ) , nakamaru et al .
( 1998 ) , and irwin and taylor ( 2001 ) .
we introduced `` real - time '' evolution as follows : every individual gets her own `` clock '' which tells her when to `` wake up '' , look around , play against the neighbours within the interaction neighbourhood , compare her payoff to that of the neighbours within the learning neighbourhood , and revise her strategy .
the `` wake up '' times are generated as @xmath16 where @xmath17 is a random number with uniform distribution in @xmath18 $ ] , which guarantees that the events are independent ( gibson & bruck ( 2000 ) ) .
( this corresponds to asynchronous updating , but with a different distribution of `` wake up '' events . )
we use the same lattice size , cost and benefit functions , and mutation rate as killingback et al .
( 1999 ) . after a sufficiently long time
( which is of the order of 2 0003 000 generations ) , the ensemble again reaches a state with a considerable degree of cooperativity , and the average investment level is about the same as that obtained by killingback et al .
a typical evolutionary outcome is shown in fig .
[ fig:11 ] ( top graph ) . here
we can take a moment to discuss the role of mutations .
the top graph in fig .
[ fig:11 ] corresponds to a very high mutation rate of 5% , with gaussian distribution ; the width of the gaussian is set at 10% of its height . in fig .
[ fig:11 ] , in the bottom graph , we show a time series obtained for a mutation rate of 0.5% , width of the gaussian is 1% of its height .
it is clear that the mutation rate does not determine the mean value of the average investment in the steady state reached by the system ; it rather influences how fast the steady state is reached ( occasional `` kicks '' by mutations help the average investment grow quickly ) , and the magnitude of fluctuations in steady state .
one general question is : what amount of spatial structure do we need ( i.e. how big can neighbourhood sizes become ) before we reach the mean - field ( all defection ) situation ? keeping a moore type of neighbourhood , with the neighbourhood and dispersal parameters equal to each - other ( for the moment ) , we consider the cases where an individual plays against everybody within a neighbourhood of radius 2 ( 24 individuals ) , 3 ( 48 individuals ) , 4 ( 80 individuals ) , and so on .
the rules of the game , the cost function , benefit function , mutation rate , and width of the gaussian mutation curve , are kept fixed and equal to those used by killingback et al .
( 1999 ) . for each case
, we averaged over 500 realisations of the system , and report results obtained on a @xmath13 lattice , unless otherwise stated . for neighbourhood and dispersal parameters equal to 2 ( i.e. a neighbourhood that includes 24 individuals , as opposed to 8 for neighbourhood parameter 1 ) ,
the final average investment value is very close to that obtained when those parameters are 1 .
when we increase the neighbourhood and dispersal parameters to 3 , the final average investment drops to around 0.85 , and it drops even further to about 0.6 for parameters values equal to 4 . for larger lattice sizes we get similar values of average investment in the steady state .
some typical runs are shown in fig .
[ fig : nn ] for lattice size @xmath13 .
the picture changes qualitatively when the neighbourhood and dispersal parameters become 5 .
even though we start with a relatively large initial investment , the final average investment is extremely small ( about 0.04 , as opposed to 0.6 for a size four neighbourhood ) , and can occasionally drop to zero under the influence of even moderate fluctuations .
this is shown in fig .
[ fig:55 ] .
the gradual drop in the value of average investment as we increase the neighbourhood size is typical for nonequilibrium phase transitions . for larger lattice sizes we get similar results for the average investment , but , as expected ,
the fluctuations are lower .
we can then state that at this neighbourhood size we have reached the mean - field limit , i.e. the picture we would obtain in absence of a spatial structure altogether ( which for the cpd corresponds to a state of pure defection , as discussed above ) .
the value of the neighbourhood and dispersal parameters for which the mean - field limit is reached was found to be independent of the lattice size , at least for reasonably large lattices ( the meaning of `` reasonably large '' will become clear in the next paragraph ) .
the immediate neighbourhood comprises about 100 individuals , while the extended one ( the neighbours of the neighbours ) comprises about 300 .
furthermore , the results were independent of the choice of the boundary conditions ; they hold also when periodic boundary conditions are not used ( and the last row cells have only five neighbours , as opposed to eight the other cells have , for a size one neighbourhood , and similarly for other neighbourhood sizes ) .
let us now examine the role of the ratio of the neighbourhood size to lattice size . in fig .
[ fig:44sm ] we show a time series of average investment for the case when both the neighbourhood and dispersal parameters are 4 , and the lattice size is @xmath19 .
it is important to note that for smaller lattices the size of fluctuations can become comparable to the average investment amount , even when the latter is considerably above zero , and often a mutation is enough to bring the average investment amount to zero .
the average investment does come `` dangerously '' close to zero a few times in fig .
[ fig:44sm ] , and we observed it drop to zero in many other runs not shown here .
so , if the lattice is too small , the fluctuations are enormous , and one might be misled by the result of a particular simulation .
( for nonequilibrium phase transitions like this one , physicists scale the average investment with the lattice size , but this goes beyond the purpose of the present paper . ) the @xmath13 lattice size we worked with ( most of the time ) is large enough to keep the fluctuations moderate , but also small enough to allow our simulations to run within reasonable times .
we can conclude , based on our simulations , that the neighbourhood size for which we achieve the mean - field limit is about 5 , and does not depend on the lattice size ( at least for reasonably large lattices ) , contrary to the expectations of killingback et al .
the larger neighbourhood sizes `` expose ''
the players that sit close to the borders of the square to a larger number of defector `` outsiders '' , whose easily earned large payoffs might `` tempt '' our players .
the numbers of these `` tempters '' only depend on the neighbourhood size .
however , the lattice size influences a very important factor , the size of the fluctuations , which can make or break the fate of the cooperative state .
the above results were obtained when the size of the interaction neighbourhood was the same as that of the learning neighbourhood .
it is interesting to investigate what happens when this requirement is relaxed , and the neighbourhood and dispersal parameters are allowed to be different from each - other . for this purpose
, we simulated a situation in which our agents play against individuals within a certain neighbourhood , but compare their payoffs to those of individuals within a neighbourhood that is different from their interaction neighbourhood .
we observe that , if the neighbourhood parameter differs from the dispersal parameter by one or more , then cooperation is no longer sustained .
furthermore , the final state is defection , independent of whether the neighbourhood parameter is larger than the dispersal one , or vice - versa .
it appears as if , when the neighbourhood parameter is smaller than the dispersal parameter , the defectors outside the interaction neighbourhood , but within the learning neighbourhood , manage to `` tempt '' the cooperators . while in the case when the neighbourhood parameter is larger than the dispersal one , there are simply too many defectors in the interaction neighbourhood , taking advantage of the cooperators . in the economic context , too little or too much information favours defection , and kills the incentive to cooperate . in the real life
, however , there is no rigid separation of neighbourhoods from one - another .
individuals interact not only within a strictly defined neighbourhood , but rather occasionally include other individuals in their sphere of relations .
this motivated us to determine what fractional difference between the neighbourhood and dispersal parameters is required for cooperation to be no longer sustained .
this can be studied in our model by making the neighbourhood and dispersal parameters `` continuous '' . to do this we constructed a `` fractional size '' neighbourhood in the following way : suppose the parameter has the form @xmath20 with @xmath21 the integer part and @xmath22 the fractional part .
then for the @xmath23-th neighbour we generate a random number with uniform distribution in @xmath18 $ ] .
if the random number is smaller than @xmath22 , the individual plays against this particular neighbour , otherwise the next event in the queue is generated .
if the neighbour is at the corner , the random number is compared to @xmath24 ( i.e. the fractional area of the rectangle ) .
the remaining rules of the game are unchanged .
we find in this situation that cooperation only persists for differences between parameters up to about 0.5 , and breaks down when the difference between the neighbourhood and dispersal parameters is 0.5 or larger .
we have looked at both the case when the neighbourhood parameter is larger or smaller than the dispersal parameter .
the results are the same : as soon as the difference between the two parameters becomes larger than 0.5 , cooperation vanishes .
some results with different neighbourhood and dispersal parameters are shown in fig .
[ fig : c20 ] . here
we can see that the average investment has already dropped to zero , when the difference between the neighbourhood and dispersal parameters reaches 0.6 , and is very small , when that difference is 0.5 .
there is no reason why every agent should have the same parameters , so we may treat them as `` local '' , i.e. every individual has her own neighbourhood and dispersal parameters , @xmath25 and @xmath26 , which generally are not integers .
when her turn comes , she plays against individuals within a neighbourhood of radius @xmath25 and then compares herself to the individuals within the neighbourhood of radius @xmath26 . however , there are cases when she meets a neighbour @xmath27 , whose dispersal parameter is smaller than hers ( i.e. he does not `` see '' her , even though she `` sees '' him ) . in such cases
she could include him into her learning neighbourhood , comparing her payoff to his , and then adopting his investment amount , if he is doing better than her .
this would correspond to the economics scenario , when the focal player can learn from players who do not see her .
otherwise she could leave that player who is doing better than her , but can not see her , out of her learning neighbourhood , even though she sees him .
this would correspond to the evolutionary scenario , where a player can only take over a cell he does see .
a natural question is : do the two scenarios lead to different results ?
we consider both possibilities : first when she does compare her payoff to that of the neighbour with a smaller dispersal parameter ( and , if he is doing better , adopts his strategy , i.e. investment amount ) , and second when she does not .
for that we had both the neighbourhood and dispersal parameters initialized randomly between the values 1 and 4 ( which are the interesting values , since for larger parameters there is no sustainable cooperation ) .
the neighbourhood and dispersal parameters are treated as evolutionary phenotypes : the parameters of the neighbour that is doing better are adopted , together with the value of the investment amount . both parameters , as well as the investment amount ,
are set to mutate at a rate of 0.01 , following a gaussian curve with a width 10 % of its peak . in both scenarios ,
something interesting happens with the neighbourhood and dispersal parameters : in fig .
[ fig : pars ] ( a ) we plot their initial values , which are uniformly distributed within a square . at the end of the simulation , we find their values have converged closer to the diagonal @xmath28 ( see fig .
[ fig : pars ] ( b ) ) .
the convergence of the neighbourhood and dispersal parameters toward each - other corresponds to the development of a cooperative state , after which the average investment continues to sustain relatively high values , almost as high as those initially reported by killingback et al .
( 1999 ) , but the fluctuations are now larger .
a typical time series for the average investment in the evolutionary scenario is shown in fig .
[ fig : evol ] .
based on the observed dynamics of the spatial cpd on regular lattices , killingback et al .
( 1999 ) formulated the hypothesis that the mechanism that facilitates and maintains cooperation in worlds of this particular topology is their _ clustering _ properties .
it is not easy to study this hypothesis within the regular lattices topology , since , in order to verify or refute it , we would have to obtain results on other topologies , different from the regular lattices . furthermore , for many systems of interacting individuals , the regular lattices do not provide a fully adequate model for the observed topological properties .
these properties include large clustering , and short average path length between any two nodes ( individuals , in our context ) .
watts and strogatz first introduced `` small world networks '' ( watts & strogatz ( 1998 ) , watts ( 1998 ) ) , which are clustered structures .
a spatial structure is called clustered , if the existence of links between nodes a and b , and b and c , makes it very probable that there be a link between a and c ( sociologists call this `` transitivity '' ) .
the regular lattices are highly clustered structures .
they also exhibit a long path length , i.e. the number of intermediate nodes between any two nodes a and b is usually large .
the random graphs of erds and rnyi ( bollobs ( 1985 ) ) , in which a node links to any other with a constant probability , have short path lengths between nodes , but exhibit no clustering , i.e. are very distributed structures .
a small world network has the two properties of high clustering and short average path length between two nodes , which increases with the network size @xmath29 as the logarithm of @xmath29 ( watts & strogatz ( 1998 ) , amaral et al.(2000 ) ) .
social networks ( i.e. networks in which humans constitute the nodes , and acquaintances the links ) are examples of small world networks ( strogatz ( 2001 ) , albert & barabsi ( 2002 ) , dorogovtsev & mendes ( 2002 ) ) .
a special class of small world networks are the scale - free ones , in which the degree ( i.e. the number of the links a node has ) distribution obeys a power law ( hence it is independent of network size ) .
many algorithms for generating scale - free networks have been proposed .
this kind of topology is also observed in some social networks ( strogatz ( 2001 ) , albert & barabsi ( 2002 ) , dorogovtsev & mendes ( 2002 ) , newman ( 2001 ) , barabsi et al .
( 2002 ) , ebel et al .
( 2002a ) ) .
davidsen , ebel , and bornholdt ( davidsen et al.(2002 ) , ebel et al.(2002b ) ) introduced an algorithm that produces small world networks which follow closely the characteristics of social and acquaintance networks .
their basic assumption is that the mechanism which generates these networks is that people are introduced to one - another by a common acquaintance , so - called _ transitive linking_. at each time step two processes take place : @xmath30 a randomly chosen individual introduces two ( randomly picked ) acquaintances of hers to one - another .
if this individual has less than two acquaintances , she introduces herself to a randomly picked individual .
@xmath31 one randomly chosen individual leaves the network with probability @xmath32 .
all her links are disconnected , and she is replaced by a new individual with one randomly picked acquaintance .
this probability @xmath32 , which is called the death probability , is typically very small .
the size of the network @xmath29 remains constant in the process .
the finite lifetime of links brings the network to a steady state , in which the degree distribution is exponential for larger death rates , and becomes power law ( with a cutoff ) as the death rate decreases . the hypothesis that clustering in lattices is the key factor that makes possible the establishment and maintenance of cooperation can be verified in networks of different topologies , since some of them exhibit large clustering , and some do not . to do this
, we study the evolution of average investment in social networks , and also in random networks , for comparison .
the two topologies were chosen , since they represent extremes when clustering properties in networks are considered : the random networks are distributed structures with very little clustering , while the social networks exhibit a high clustering degree .
also , in these two structures we can tune the connectivity ( defined as the ratio of the actual number of links in the network to the total number of all possible links ) : for the random networks it is equal to the probability of linking between two nodes , while for the social networks model it can be varied by varying the death probability .
the game proceeds as follows : `` wake - up '' times are generated for each node as @xmath33 , where @xmath17 is a random number with uniform distribution in @xmath18 $ ] .
the picked node then plays against the nodes it can `` see '' , i.e. it is linked to .
recall that in a network each node can have different degrees , i.e. number of immediate neighbours .
this means that what we previously called neighbourhood parameter differs from node to node , and is equal to the degree of the focal node ( player ) .
this focal player then calculates her payoff from playing against her immediate neighbours , according to the rules of the cpd game
. these neighbours also play against their immediate neighbours , and calculate their respective payoffs .
the picked ( focal ) node then compares her playoff to that of the neighbour that is doing best , among her immediate neighbours .
the neighbours she compares herself with ( learns from ) are the same ones she played against ( interacted with ) , i.e. albeit different from one node to another , the neighbourhood and dispersal parameters for one node are equal to each - other ( and the degree of that node ) .
the focal node then adopts the investment amount of the neighbour that is doing best as her new investment amount , and then the next event is generated .
the process is iterated , and the average investment is calculated every few `` generations '' .
this process is repeated for different network sizes and averaged over 100 different configurations for each network size . for any
fully connected ( saturated ) network , with connectivity equal to one , the final state will be of pure defection , since this is a well - mixed system , in which `` everybody sees ( or connects to ) everybody '' .
as the connectivity decreases , for each network size there is a certain threshold value , for which the final state ceases to be that of pure defection , but rather a very small average investment . as the connectivity decreases even further , the steady state average investment increases .
the variation of average investment with the connectivity for a social network with @xmath34 nodes is shown in fig .
[ fig : avinvscc ] .
the transition from the cooperative network to the non - cooperative one is what physicists call a critical transition , i.e. the value of the average investment decays slowly , as opposed to exhibiting a sharp drop at the crossover connectivity value . for a random network of the same size ( @xmath34 ) ,
the same transition is present , but the connectivity for which the transition happens is @xmath35 .
this critical connectivity is much smaller than @xmath36 we obtained for a social network of the same size .
since the fundamental difference between social and random networks is in their clustering properties , it is natural to check the dependence of the average degree ( number of neighbours ) per node at the transition point on the network size for both topologies . the results are shown in fig .
[ fig : criad ] . for both topologies ,
the critical average degree remains essentially constant as the network size is varied .
it is about 150 for the social network model , and about 20 for the random network .
this surprising result is a very strong argument in favour of the hypothesis that clustering is the key mechanism which leads to cooperation being first established and then maintained .
indeed , social networks , with their pronounced tendency to cluster , maintain the cooperative state for a much larger average connectivity ( and node degree ) than the maximally distributed random networks .
we have studied the behaviour of the continuous prisoner s dilemma model on different spatial structures , including two - dimensional lattices and networks of different topologies .
the continuous prisoner s dilemma is an evolutionary game which is a natural extension of the standard prisoner s dilemma . in the continuous prisoner
s dilemma each individual makes an investment in favour of herself and the neighbours , and the investment takes on any value , not just two discrete values . in spatial structures like the ones we have considered , and for small neighbourhood sizes / connectivities
the investment levels evolve by getting higher and stabilizing around a given value in a steady state .
the cooperation develops and is maintained relatively easy , which leads us to believe that the cooperation is not such a difficult evolutionary paradox .
we have tested this in the case of the square lattice , by increasing the neighbourhood sizes , and found that the neighbourhood size for which the steady state is pure defection is about 100 neighbours . as long as the number of the neighbours the individual plays against
( which we call the neighbourhood parameter ) and that of the neighbours the individual compares herself to ( called the dispersal parameter ) remains the same , the cooperation is quite robust .
the mean field neighbourhood size limit does not depend on the size of the lattice , as long as the lattice is reasonably large . since the interaction neighbourhood and the learning neighbourhood
do not have to coincide with one - another , we studied the game when the neighbourhood and dispersal parameters are different . in this case , we found that cooperation only develops for differences between them of about 0.5 or smaller .
too little information favours defection , but apparently so does too much information .
something interesting happens when we allow the neighbourhood and dispersal parameters to mutate , and at the same time , treat them as phenotypes : they are adopted , together with the investment amount of the individual that is doing better .
the values of the neighbourhood and dispersal parameters converge toward one - another , i.e. their differences decrease .
cooperation persists , and the average investment fluctuates around relatively high levels in the steady state .
we verified the hypothesis that clustering is the factor that facilitates and maintains high average investment values by considering players in networks of different topologies . in all of them
, a transition is observed between the cooperative steady state and the purely defective one , as the connectivity of the network increases . for both social networks and random ones ,
the average degree at the transition point is practically independent of the network size .
however , the average degree at transition for social networks is about 150 , while that for random networks is much lower ( about 20 ) .
the two topologies differ in their degree of clustering .
the social networks topology , with its pronounced tendency to cluster , allows the establishment and maintenance of a cooperative state , something the distributed random networks topology do not .
however , when the connectivity of the network exceeds a certain value ( i.e. too many individuals `` see '' one - another and communicate ) , cooperation can not develop or be maintained , and the steady state strategy is pure defection .
the cpd model deepens our understanding as to how cooperation has evolved gradually from lower to higher levels in many spatially structured systems , where the entities involved can vary from replicating molecules to whole organisms ( wilson ( 1980 ) , michod ( 1983 ) , buss ( 1987 ) , maynard smith & szathmry ( 1995 ) , dugatkin ( 1997 ) ) : the clustered topology helps cooperation develop and survive .
it is very encouraging that cooperation is quite robust in such systems . on the other hand , in the economic world
, globalization brings about an increase in the connectivities of social and economic networks , and cooperation may consequently be less likely to persist .
we thank birger bergersen for helpful and interesting discussions and suggestions . | the prisoner s dilemma , a 2-person game in which the players can either cooperate or defect , is a common paradigm for studying the evolution of cooperation . in real situations cooperation
is almost never all or nothing .
this observation is the motivation for the continuous prisoner s dilemma , in which individuals exhibit variable degrees of cooperation .
it is known that in the presence of spatial structure , when individuals `` play against '' ( i.e. interact with ) their neighbours , and `` compare to '' ( `` learn from '' ) them , cooperative investments can evolve to considerable levels . here
we examine the effect of increasing the neighbourhood size : we find that the mean - field limit of no cooperation is reached for a critical neighbourhood size of about five neighbours on each side in a moore neighbourhood , which does not depend on the size of the spatial lattice .
we also find the related result that in a network of players , the critical average degree ( number of neighbours ) of nodes for which defection is the final state does not depend on network size , but only on the network topology .
this critical average degree is considerably ( about ten times ) higher for clustered ( social ) networks , than for distributed random networks .
this result strengthens the argument that clustering is the mechanism which makes the development and maintenance of the cooperation possible . in the lattice topology , it is observed that when the neighbourhood sizes for `` interacting '' and `` learning '' differ by more than 0.5 , cooperation is not sustainable , even for neighbourhood sizes that are below the mean - field limit of defection .
we also study the evolution of neighbourhood sizes , as well as investment level . here
we observe that the series of the interaction and learning neighbourhoods converge , and a final cooperative state with considerable levels of average investment is achieved .
prisoner s dilemma , continuous , spatial , lattice , network |
mixing of different baryonic states in nuclear systems remains a topic rather exciting but being far from sophisticated understanding .
a long time ago , it was speculated that a nucleus is not generally a pure nucleonic system . due to @xmath11 and @xmath12 transitions
, a nuclear wave function incorporates some admixture of states with one ( or even more ) @xmath13 baryon(s ) @xcite . strictly speaking , also heavier baryons should be taken into account .
admixtures of @xmath14 states in @xmath15 hypernuclei probably are more important for hypernuclear dynamics .
the @xmath16 mass difference ( 80 mev ) is rather less than the @xmath17 difference ( 300 mev ) .
moreover , pionic exchange between @xmath15 and @xmath18 necessarily gives rise to virtual @xmath14 because of zero @xmath15 isospin .
it was suggested a long ago @xcite that the @xmath19 coupling is responsible for the so - called @xmath20he overbinding problem , which was confirmed recently in a consistent study of @xmath21-shell hypernuclei @xcite .
probably , the @xmath19 coupling plays the crucial role in binding of hypertriton @xcite .
the coupling contributes greatly to the @xmath15 binding in nuclear matter @xcite . however , there are no direct unambiguous measurements of the baryonic admixtures so far .
various processes are considered to detect @xmath13 admixtures in ordinary nuclei , but discrimination between @xmath13 components preexisting in a nucleus and @xmath22 baryons produced in a reaction is a difficult problem ( for some recent attempts , see @xcite ) . for contributions of @xmath19 mixing to @xmath15 hypernuclear binding energies , alternative
dynamical pictures without explicit @xmath23 admixtures usually exist .
for instance , when a single channel @xmath24 interaction is described in terms of correlated two - pion exchange , it is probable that the dynamics of virtual intermediate @xmath14 s is included implicitly .
also effective @xmath25 force can mimic effects of the @xmath26 coupling without explicit @xmath14 degrees of freedom .
several other implications of the @xmath19 coupling providing in principle ways to identify the @xmath14 admixtures , have been discussed .
it was suggested that measurement of hypernuclear magnetic moments is promising for this aim @xcite .
the probability of the rare @xmath27 channel of hypernuclear weak decay is sensitive to @xmath28 admixture @xcite .
the coupling can lead in some cases to the reverse order of spin doublet levels @xcite .
also production of neutron - rich @xmath15 hypernuclei from @xmath29 and @xmath30 reactions can proceed via @xmath31 admixture as a doorway state @xcite .
relevant data are not available so far . moreover , in all of these cases , some background effects [ as meson charge exchange in the @xmath27 decay as well as in the @xmath29 and @xmath30 reactions ] occur , which can hinder detection of @xmath32 admixture .
the @xmath33 coupling in double - strangeness hypernuclei is of particular interest , since the relevant mass difference ( about 25 mev ) is lowest among all possible known baryonic couplings .
however , experimental information on double - strangeness systems is rather scarce so far , and no definite knowledge of the coupling exists yet .
theoretically , the @xmath33 coupling in @xmath34 hypernuclei hypernucleus '' in this context means a state of an @xmath35 hypernucleus with dominant @xmath34 component . ]
have been considered by several groups @xcite . mostly , hypernuclei observed experimentally have been studied , namely , @xmath36he @xcite , @xmath37be , @xmath38b @xcite .
with meson - exchange coupling potentials , probabilities of @xmath39 admixtures less than 1% were obtained . contributions of the coupling to the binding energies are as small as several tenths of mev except the case of extremely strong @xmath34 attraction providing @xmath34 bound state @xcite , when the contribution can reach several mev .
much larger coupling has been obtained @xcite within a quark model predicting free bound @xmath40 dibaryon . in this case , not only @xmath41 , but also @xmath42 component is of a great weight ( more than 10% ) . myint and
akaishi @xcite argued that the @xmath43 coupling is considerably enhanced in five - baryon hypernucleus @xmath0h .
a proton , appearing from the @xmath44 transition , can be bound rather strongly in the @xmath45 particle .
thus , the difference between the thresholds of channels @xmath46 and @xmath47 is reduced to 8 mev from 29 mev for free @xmath34 and @xmath48 pairs . myint and
akaishi @xcite obtained 1% for the @xmath49 admixture probability and 0.5 mev for the binding excess appearing due to the coupling .
these values are larger than those typically obtained by other authors for other hypernuclei , but still small to provide more or less unambiguous signature of the coupling . in the studies performed in the 1990s , nijmegen hard - core model d ( nhc - d ) @xcite has been used popularly as a standard meson - theoretical model for @xmath35 interactions .
the reason was that this model is compatible with strong @xmath34 attraction ( @xmath505 mev ) supported by earlier data on @xmath51 hypernuclei @xcite .
this strong @xmath34 attraction of nhc - d is due to its specific feature that only the scalar singlet is taken into account . in the cases of the other nijmegen models incorporating the whole scalar nonet , the meson - exchange parts of the @xmath52 interactions are much weaker than those of nhc - d . in the case of the hard - core model f ( nhc - f )
@xcite , for instance , the strength of that part is about a half of nhc - d . of course ,
in these models the hard - core radii can be treated as adjustable parameters to reproduce any strength of @xmath51 interactions .
then , it is difficult to discriminate between strong and weak meson - exchange attraction compensated by small and large hard - core radii , respectively , in @xmath51 single - channel treatment .
on the other hand , the @xmath33 coupling of nhc - d is relatively weak because of small contributions of strange mesons , which leads to small @xmath39 admixtures appeared in structure calculations of double-@xmath9 hypernuclei @xcite . in the case of nhc - f , for instance , the @xmath33 coupling has been known to be stronger by about two times than that of nhc - d . as shown later ,
the strength of the @xmath33 coupling of nhc - d is the weakest among the various nijmegen models due to the above - mentioned specific feature .
since the discovery of nagara event identified uniquely as @xmath53he @xcite in 2001 , @xmath34 attraction is believed to be rather less ( @xmath54 mev ) . on the basis of this new datum , hiyama _ et al .
_ have performed the systematical analysis for light double-@xmath9 hypernuclei @xcite .
the strength of obtained @xmath51 interaction is very similar to the meson - exchange part of nhc - f .
even if nhc - d is used , it is possible to reproduce @xmath55 mev by taking a larger value of its hard - core radius appropriately .
however , weak @xmath51 attraction consistent with @xmath56 mev can be obtained more plausibly by the other nijmegen models .
therefore , it is likely that their stronger @xmath57 coupling interactions are more realistic and the mixing in @xmath34 hypernuclei is more significant than it was speculated earlier from nhc - d .
very recently , myint _
et al . _
@xcite performed a new study of the five - baryon @xmath34 hypernuclei ( @xmath0h as well as @xmath0he ) and showed that the coupling effect in @xmath58he exceeds considerably that in @xmath0h . since @xmath59 is lighter than @xmath60 by 6.4 mev , the threshold difference of channels @xmath61 and @xmath62 is as small as 2.4 mev .
so the @xmath39 admixture is probably greatest in @xmath0he , and remarkable difference between @xmath0he and @xmath58h binding energies appears .
similar suggestions have been made also in @xcite . in this paper
, we study properties of @xmath0h and @xmath58he with diagonal @xmath34 and coupling @xmath63 @xmath6-matrix interactions deduced from various meson - exchange models . while
myint _ _ et al.__@xcite employed a single - channel approach , in which the coupling is involved effectively in the diagonal potential , we solve the relevant two - channel problem explicitly .
particularly , we incorporate the diagonal @xmath64 potential into the calculation and emphasize that effect of this potential is rather important .
we investigate also the role of the other inputs ( @xmath65z and @xmath34 potentials ) . in sec .
[ s2 ] we describe our model , choice of phenomenological hyperon - core potentials , and also @xmath6-matrix derivation of the diagonal @xmath4 and coupling @xmath5 effective interaction .
section [ s3 ] contains presentation and discussion of our results . also comparison of our approach with that of myint _
_ et al.__@xcite is outlined .
section [ s4 ] is devoted to concluding remarks and some outlook .
accurate five - body calculations for hypernuclei @xmath66h and @xmath0he are available so far only in the single - channel approach without baryonic mixings @xcite . in @xcite ,
three - body models were utilized with free two - baryon @xmath35 potentials ( more precisely , their phase - equivalents ) .
since we also do not solve the five - body problem directly , we use effective ( @xmath6-matrix ) two - body @xmath67 and @xmath63 interactions instead of free - space ones . here
, we adopt the hartree - fock ( hf ) description , which is not only simpler , but also is suitable for the @xmath6-matrix formalism defined in a single - particle model space . because density distributions of the lightest nuclei are known to be described poorly in the hf approximation
, we do not apply the hf description for all the five baryons .
instead , we treat the hypernuclei as @xmath68 and @xmath69 in the @xmath34 and @xmath39 channels , respectively .
the nuclear cores are treated as inert , and hyperon - nucleus interactions are described fully by @xmath65z and @xmath70he potentials .
this hf approach is similar to that in @xcite , but it is extended to incorporate the second channel .
we adopt the wave function of the five - baryon hypernucleus as @xmath71 , where @xmath72 , @xmath73 , @xmath74 , and @xmath75 are the three - nucleon core wave function and single - particle wave functions of @xmath9 , nucleon ( @xmath18 ) , and @xmath76 , respectively .
the @xmath18 and @xmath76 components appear from the @xmath77 transition . for brevity ,
spin parts are not shown explicitly in the wave function above .
we restrict ourselves to the single @xmath74 state corresponding to the ground state @xmath78 particle and also to the @xmath79 states of the hyperons . since all the baryons are in @xmath21 states , the kinetic energy term containing scalar product of gradients acting on @xmath80 and @xmath81 @xcite vanishes identically . excluding c.m .
motion in the usual way and neglecting at that instance the difference between c.m .
positions in the two channels , we come to equations for single - particle wave functions . for the radial wave functions , the equations to be solved are as follows : @xmath82 here @xmath83 and @xmath84 are the radial wave functions of @xmath15 and @xmath85 , respectively .
the threshold - energy difference @xmath86 is taken as 8.4 ( 2.4 ) mev for @xmath0h ( @xmath0he )
. then , @xmath87 and @xmath88 are the diagonal potentials , whereas @xmath89 is the nondiagonal one , where the nucleon radial wave function @xmath90 is calculated in the same potential as in @xcite .
potentials @xmath91 , @xmath88 ,
@xmath92 and @xmath93 correspond to @xmath9-core , @xmath8 , @xmath4 diagonal and @xmath94 coupling interactions , respectively . if @xmath72 and @xmath74 are normalized to unity , the normalizing condition for @xmath95 and @xmath96 is @xmath97 .
since @xmath98 and @xmath99 depend on @xmath83 , the system must be solved self - consistently as usually in hf approaches . in ( [ e2 ] ) , only one state ( @xmath100 ) is kept .
generally , it is an important problem how to take account of other states beyond the @xmath79 shell .
it was shown that the @xmath63 coupling contributes to the @xmath101he binding energy sizably ( though not so much for realistic interactions ) @xcite , despite that all the nucleonic @xmath79 states are occupied . in our approach , highly - excited components are renormalized into the effective @xmath51 interactions in some approximate way , as commented in the end of subsec .
[ s2c ] .
our main purpose in this work is to calculate the quantity @xmath102 defined as @xmath103 where @xmath104 and @xmath105 denote binding energies of @xmath9 and a pair of @xmath9 s in a nuclear core , respectively . in order to obtain @xmath106 values of @xmath107h , @xmath107he , and @xmath108he
, we need @xmath109h , @xmath109he , and @xmath110he potentials . for @xmath9-core interactions
, we use phenomenological prescriptions . for @xmath65z potential @xmath111
we use the isle - type form @xcite @xmath112 with @xmath113 fm , @xmath114 fm .
we fit the strength to binding energies of @xmath115h and @xmath116he in the ground and the first excited states , and then average the potential over the singlet and triplet states .
the obtained values are @xmath117 mev ( @xmath116h ) and @xmath118 mev ( @xmath119he ) .
quantity @xmath120 ( 1.25 mev in @xmath115h and 1.53 mev in @xmath115he ) in ( [ e3 ] ) is calculated in this averaged potential . for the @xmath121he potential ,
fitting is performed to @xmath122he ) , which gives @xmath123 mev . to address sensitivity of the results to the shape of the hyperon - nucleus potentials ,
we examine also another force .
namely , @xmath111 is taken as one - range gaussian ( org):@xmath124 with @xmath125 fm and @xmath126 mev for @xmath20he @xcite .
the fitting procedure for the potential strengths is the same as described above .
we obtain @xmath127 @xmath128 mev for @xmath115h ( @xmath115he ) .
it is well known that the effect of the @xmath19 coupling is especially important just in the @xmath115h and @xmath115he hypernuclei .
thus , this effect may be expected to be important also in the five - body @xmath4 hypernuclei . in our model ,
the @xmath19 coupling is incorporated effectively at the mean - field level in the @xmath9-nucleus potentials , since they are determined so as to reproduce the experimental @xmath9 binding energies in @xmath115h , @xmath115he , and @xmath20he .
it should be noted that some direct interplay between two types of baryonic mixing is also possible , which is beyond the mean - field treatment .
for instance , a @xmath129 hyperon appeared from the @xmath19 coupling can participate in further couplings ( like @xmath130 ) .
however , since the @xmath129 admixtures are not larger than 2% @xcite , we omit such effects . in our two - channel model , important roles are played by @xmath76-core interactions . particularly , we study effects of the @xmath64 potential @xmath131 for @xmath76 admixtures in @xmath132h and @xmath132he . unfortunately ,
too little is known about @xmath39-nucleus interaction .
the @xmath76-core attractions comparable to @xmath9-core ones were deduced by analyzing the compilation of the emulsion events of `` @xmath76 hypernuclei '' in the past @xcite , although none of individual events was identified unambiguously .
the recent data on @xmath133c(@xmath134 reaction @xcite give some evidence that the @xmath39-nucleus interaction , being attractive , is weaker by about half than the corresponding @xmath15-nucleus interaction .
we consider that this information is most reliable in the present stage .
we adopt here a simple folding model to obtain the strength of @xmath41 interaction compatible with @xmath39 well depth @xcite of 14 mev in @xmath135b .
the @xmath136 potential is deduced by folding of the obtained @xmath41 interaction with the density of @xmath78 , and then @xmath76 binding energies in the @xmath64 systems are calculated .
we find no bound state of @xmath137he , while @xmath138 hyperon is bound by @xmath45 particle with @xmath139h)@xmath1400.5 mev ( depending on details of the folding procedure ) .
this value is much greater than the corresponding coulomb energy ( about 0.1 mev ) .
the rms radius ( about 6 fm ) also confirms that it is a nuclear ( or , maybe , `` hybrid coulomb - assisted '' @xcite ) state .
we consider also the @xmath39 well depth in @xmath135b of 24 mev according to the earlier analysis @xcite . in this case ,
bound @xmath137he appears with @xmath1411.1 mev . the much stronger @xmath142 potential suggested by filikhin and gal @xcite gives @xmath143 mev and thus it is probably incompatible with the data @xcite . for radial dependence of @xmath64 potential @xmath88
, we adopt the isle - type form ( 4 ) as well .
we fit strength @xmath144 to @xmath145h)@xmath146 mev ( potential xa1 ) . for comparison
, we examine the filikhin and gal @xcite choice ( @xmath147he)@xmath148 mev , potential xa2 ) . at
last , we test also zero @xmath64 potential xa0 ( to be consistent with calculations @xcite , we switched off also coulomb @xmath149 interaction in the last case ) . for consistency with the @xmath9-core potentials ,
we prepare also the @xmath64 potentials in the org form with @xmath150 fm , which corresponds to folding of a @xmath41 org - type potential with gaussian @xmath45 density with reasonable range parameters .
potential xa3 is fitted to the same value @xmath151h)@xmath146 mev as potential xa1 .
the strongest org - type @xmath64 potential xa4 is fitted to @xmath147h)@xmath152 mev , which is compatible with @xmath39 well depth of 24 mev @xcite in @xmath135b , being rather less than the potential by filikhin and gal @xcite predicts .
the @xmath64 potentials used are presented in table [ t1 ] .
[ cols="^,^,^,^,^,^,^,^ " , ] in fig .
[ f1](a ) , @xmath102 values obtained from the full calculation with various isle - type @xmath8 potentials are shown as functions of volume integral @xmath153 . it is seen that the coupling effect is anyway meaningful and may be rather high .
even with moderate @xmath64 potential xa1 , full @xmath102 is more than twice as large as the single - channel value for strong coupling interactions . for the nhc - f model
, the difference @xmath154he)@xmath155h ) is about 0.4 mev . for the strongest @xmath142 potential xa2 , @xmath156he ) can reach 2.3 mev ( remind that @xmath157he)@xmath158 mev ) . even for zero @xmath64 potential xa0 , @xmath154he ) can exceed @xmath159he ) as has been pointed out in @xcite . the @xmath39 admixture probability @xmath160 is presented in fig . [ f1](b ) .
its dependence on the volume integral follows closely the corresponding curves for @xmath102 .
most of the values in fig.[f1](b ) exceeds considerably @xmath161 obtained in @xmath162 hypernuclei with meson - exchange models earlier . in the extreme case of the strong coupling and strong @xmath64 potential
, @xmath161 reaches 14% .
the same value has been obtained in @xmath37be @xcite in a quark model with strong dibaryonic mixing .
it is seen that @xmath102 and @xmath161 are smooth functions of the volume integral . at relatively weak couplings ,
@xmath102 and @xmath161 are nearly quadratic in the volume integral according to the lowest order of perturbation theory .
generally the coupling effect depends on the coupling potential strength ( volume integral ) rather stronger than on its shape .
the coupling contributions to @xmath102 s in our calculations with no @xmath8 potential are systematically smaller than those obtained by myint _
et al . _
the numerical comparison between the two approaches can be exemplified for the nsc97e model .
the @xmath6-matrix interaction derived from their simplified version ( 97e ) is similar qualitatively to that from the original nsc97e , as mentioned before .
the obtained values of @xmath106 for @xmath163h and @xmath163he are 0.96 mev and 1.28 mev , respectively @xcite .
these values are considerably larger than our obtained values not only with xa0 ( 0.79 and 0.94 mev ) , but also with xa1 ( see table [ t3 ] ) .
certainly , our approach differs from theirs by many features . here , it is worthwhile to comment critically the method applied by myint _
et al . _ to reduce the two - channel problem to the single - channel one .
for this aim , the effective @xmath4 diagonal potential @xmath164 is used incorporating the coupling due to the second term ( for brevity , we do not write down separately the terms corresponding to the @xmath165 and @xmath166 channels ) . essential approximation of myint _
et al . _
is replacement of the full energetical denominator by average constant @xmath167 , which is adjusted to @xmath4 scattering properties calculated in a meson - exchange model .
the contribution of the second term of ( [ e8 ] ) to a @xmath4 hypernuclear energy can be easily expressed as @xmath168 where @xmath169 is the wave function of the first ( @xmath4 ) channel solved with @xmath170 , and @xmath171 represents the complete set of states of the second ( @xmath41 ) channel . on the other hand , the accurate expression for the second - order perturbative contribution , considering the pauli suppression effect on nucleon occupied states ,
is given as [ e101 ] @xmath172 where @xmath173 and @xmath174 correspond to the unperturbed ( uncoupled ) state and @xmath175 denotes pauli forbidden @xmath41 states .
the summation in the first term in ( [ e11 ] ) runs over all the @xmath176 states and corresponds to ( [ e9 ] ) .
the second term comes from the pauli correction , corresponding to @xmath177 in @xcite .
since the @xmath76 admixtures are typically not so high , the lowest order of perturbation is appropriate at least qualitatively ; the difference between @xmath169 and @xmath173 is probably small . to reduce the first term of ( [ e11 ] ) to ( [ e9 ] ) , one should assume some averaged value for the denominator .
clearly , this averaged value is defined by the @xmath76 hypernuclear spectrum and bears no relation to the two - baryon c.m . averaged energy @xmath167 adjusted so as to simulate the @xmath4 scattering parameters . on the other hand , myint _
et al . _
@xcite use the energy denominators conforming to ( [ e101 ] ) in the calculation of many - body corrections ( @xmath178 and @xmath179 in their notations ) .
conceptually , in ( [ e11 ] ) @xmath167 is used in the first term , but not in the second term .
it leads to the contradiction pointed out by filikhin _
et al . _
@xcite : the correction introduced to exclude the pauli forbidden state , if calculated properly , is greater than the whole coupling effect .
this contradiction is considered to be originated from the inconsistent treatments for the two terms in eq .
( [ e11 ] ) , which are obtained only by reforming the single term ( [ e10 ] ) .
namely , their treatment does not accord to the clear - cut expression ( [ e101 ] ) and the obtained results are considered to be questionable .
we stress also the role of the @xmath64 potential omitted in @xcite .
naturally , the stronger is @xmath64 attraction , the greater is the coupling .
substitution of zero @xmath64 potential xa0 by even relatively weak attractive xa1 gives an energy gain up to 0.2 mev in @xmath0h and 0.3 mev in @xmath0he .
for the xa2 potential , the gain can reach 0.5 mev ( @xmath0h ) and 1.1 mev ( @xmath0he ) .
the @xmath64 potential is found to amplify the effect , making it observable more simply . on the other hand ,
it is clear that reliable quantitative extraction of the coupling strength from future data is improbable until one deduces reliable @xmath64 potential .
qualitatively , the effect of the @xmath64 potential can be described as follows . introducing @xmath64 attraction ,
one essentially moves the unmixed @xmath180z state closer to the unmixed @xmath0z state .
naturally , the smaller is the energy difference , the greater is the mixing .
seemingly , similar effect can be provided by changing of the @xmath181 diagonal potential .
but for the @xmath182 potential , this is not the case .
if the diagonal @xmath162 interaction increases ( the @xmath183z level moves down and away from the @xmath180z one ) , the coupling can even grow .
the reason is that the weaker is @xmath34 attraction , the more extended is @xmath15 spatial distribution .
therefore , overlap of the @xmath15 wave function with the nucleonic one becomes poorer ( note that the nucleon is bound in the @xmath45 particle rather strongly ) and the coupling strength is reduced .
this factor can overcome the decrease of the initial energy difference .
some schematic numerical illustrations have been presented in @xcite . here
we demonstrate another example of the importance of the hyperonic spatial distributions .
we repeat the calculations using org - type @xmath184 and @xmath88 .
the @xmath102 values are shown in fig .
[ f2](a ) and @xmath39 probabilities are presented in fig .
[ f2](b ) .
it is seen that the coupling effect increases considerably with respect to the isle - type potential form . in the extreme cases , @xmath185he )
reaches huge values , almost five times as @xmath156he ) from the single - channel calculation ( remind that `` extreme '' org - type @xmath142 potential xa4 is weaker than xa2 of the isle - type ) .
the @xmath39 admixture probabilities are larger too , though to a less extent .
radial wave functions @xmath186 and @xmath187 for @xmath0h with the nsc89 potential and the xa1 and xa3 potentials .
thick lines correspond to the isle - type potentials , thin ones are for the org - type potentials .
the dashed line shows proton wave function @xmath188.,width=321 ] the reason for this enhancement is similar to that discussed above .
the org - type @xmath9-core potential gives much more concentrated hyperon spatial distributions than the isle - type potential does .
it is seen from fig .
[ f3 ] , where the hyperonic radial wave functions in @xmath0h are shown in comparison with the proton wave function .
therefore , the overlap becomes larger , and the mixing matrix element @xmath99 in ( [ e1 ] ) and ( [ e2 ] ) increases .
it is interesting that replacement of the strongest org - type xa4 potential by still stronger isle - type potential xa2 does not enhance further the coupling since the shapes of the @xmath65z and @xmath189 potentials become inconsistent .
usually , the isle - type potentials are treated as more realistic for light hypernuclei .
however , it is instructive that there exists such a high sensitivity of the coupling effect to the shape of the @xmath15-core and @xmath64 potentials .
though the coupling effect in the calculation with the org - type potentials is probably overestimated , it is still possible that the true effect is somewhat larger than that obtained with the isle - type potentials .
we studied properties of mirror hypernuclei @xmath0h and @xmath58he in the two - channel approach with @xmath6-matrix effective interactions .
the @xmath33 coupling is particularly efficient in these hypernuclei owing to the small energy - differences between the @xmath51 and @xmath41 states .
most importantly , the coupling in @xmath0he is substantially larger than in @xmath0h .
the @xmath33 coupling leads to considerable difference of the binding energies of @xmath0he and @xmath58h .
this difference hardly can be explained by any other reasons and can be a clear signature of the baryonic mixing .
possibly , it is the most unambiguous signature of baryonic mixing among various suggestions considered so far for ordinary nuclei as well as hypernuclei . essentially , the effect analyzed here is a charge symmetry breaking effect of an unusual ( many - body ) nature .
namely , the @xmath34 interaction in the mirror @xmath34 hypernuclei appears to be different due to electromagnetic mass difference of @xmath60 and @xmath59 , since different nucleonic states in the mirror cores are occupied .
another charge symmetry breaking mechanism for free @xmath24 interaction , originating from the mass difference of @xmath28 and @xmath190 , has been suggested by dalitz and von hippel @xcite ( for further studies , see @xcite ) . by contrast , our mechanism is essentially many - body , existing only due to the nuclear environment .
our analysis demonstrates that detailed knowledge of the @xmath142 and @xmath65z potentials is needed in order to extract the @xmath33 coupling strength quantitatively from future data on @xmath0h and @xmath0he . whereas the @xmath9-nucleus interaction
, though being far from complete understanding , is recognized to a large extent , @xmath39-nucleus interaction is known quite poorly at present .
our consideration exhibits that generally there are no separate branches of @xmath34 and @xmath39 hypernuclear dynamics , but rather the unified field of @xmath35 hypernuclei .
we showed the example when not only the @xmath33 coupling interaction , but also the @xmath39-nucleus diagonal potential , is essential for @xmath34 hypernuclear properties . on the other hand ,
the same @xmath33 potential is responsible for conversion widths of @xmath39 hypernuclei whereas branching ratios of conversion channels depend substantially on the @xmath162 potential @xcite . due to the small mass difference of the @xmath41 and @xmath162
pairs , the @xmath41 admixtures in @xmath34 hypernuclei are not only important , but rather its importance is drastically different in different hypernuclei ( and , probably , in different states ) . evidently , the @xmath41 mixing in the five - baryon @xmath34 hypernuclei is considerably larger than in @xmath191he .
it was suggested that the @xmath33 coupling effect in @xmath192h is also large @xcite .
it should be noted , however , that baryonic spatial distributions in @xmath192h are expectedly rather extended .
we showed here that it is an unfavorable factor for the coupling .
this problem is very interesting in view of the question whether @xmath193h is bound or not , which was answered oppositely in two recent studies @xcite basing on comprehensive four - body ( however , single - channel ) calculations .
lastly , some comments on the possible observation of the five - body @xmath162 hypernuclei are in order
. the @xmath0h hypernucleus can be produced both from @xmath60 capture reactions @xcite or the @xmath194li@xmath195 reaction @xcite and detected by characteristic @xmath196 emission @xcite though detection by the @xmath197 weak decay suggested in @xcite is problematic in view of recent calculations @xcite predicting very small branching ratios for the @xmath14 emission channel .
on the other hand , production rates of @xmath0he from @xmath138 capture by various nuclei are much smaller @xcite .
possibly , @xmath0he can be produced with a detectable probability from @xmath60 capture by lithium .
however , the detection of @xmath58he is rather difficult since its @xmath196 decay is expectedly rather improbable .
the technique to detect @xmath0he is , therefore , an open problem . | properties of hypernuclei @xmath0h and @xmath1he are studied in a two - channel approach with explicit treatment of coupling of channels @xmath2 and @xmath3 .
diagonal @xmath4 and coupling @xmath5 interactions are derived within @xmath6-matrix procedure from nijmegen meson - exchange models .
bond energy @xmath7 in @xmath0he exceeds significantly that in @xmath0h due to the channel coupling .
diagonal @xmath8 attraction amplifies the effect , which is sensitive also to @xmath9-core interaction .
the difference of the @xmath10 values can be an unambiguous signature of the @xmath5 coupling in @xmath4 hypernuclei . however , improved knowledge of the hyperon - nucleus potentials is needed for quantitative extraction of the coupling strength from future data on the @xmath4 hypernuclear binding energies . |
ab - initio electronic structure theory for bulk materials has experienced tremendous advances in many areas such as density functional theory @xcite , improved post - dft @xcite and , e.g. , van der waals functionals,@xcite as well as highly accurate quantum chemical @xcite and monte - carlo approaches@xcite .
however , many problems are still out of reach of an advanced theoretical description due to their size : the accurate description of , for example , defect structures requires both a highly accurate treatment of the local defect region , as well as the treatment of a large number of atoms of the environment@xcite .
it is often challenging for a single method to meet both requirements .
embedding is therefore a suitable strategy to evercome this hurdle .
its underlying idea is to treat the local structure or , more generally , the subsystem of interest by a high - level method while treating the environment with the help of a numerically less demanding lower level method . consistently combining different electronic structure methods within the same calculation
is both the advantage and the challenge of the embedding approach @xcite .
several embedding schemes have been proposed @xcite , relying either on some form of a local embedding potential @xmath0 that mediates the interaction between the subsystem referred in the following as the cluster and the surrounding environment .
more elaborate operator - based approaches @xcite introduce a nonlocal embedding @xmath1 .
typically , subsystems are treated in the presence of @xmath0 [ or @xmath2 using a high - level method , while the entire system is handled by density functional theory ( dft ) .
the individual subsystem densities are then added to obtain an approximation for the total density of the entire system . while conceptually simpler , local embedding potentials feature the distinct disadvantage that no set of mutually orthogonal orbitals of the entire system exists
. consequently , evaluation of the total energy becomes challenging : in particular the kinetic energy needs to be approximated .
huang et al .
@xcite used an optimized effective potential method to recover the kinetic energy given a total electron density .
conversely , fornace et al . presented an embedded mean - field theory @xcite partitioning the one - particle density matrix of the system based on its basis functions .
a single hamiltonian then describes the entire system , avoiding any issues with evaluating the kinetic energy for cluster and environment separately .
additionally , this approach , by construction , allows for direct charge exchange between the cluster of interest and the environment .
however , a direct extension to plane - wave basis sets used in periodic solid state computations seems challenging .
goodpaster et al .
@xcite have presented a scheme relying on projection operators to ensure mutual orthogonality of orbitals belonging to different subsystems . in the present article ,
we present an alternative strategy to generate and maintain mutually orthogonal orbitals for the subsystems throughout the calculation .
we determine wannier - like orbitals localized within the cluster by performing unitary rotations within the subspace of fully occupied kohn - sham orbitals while the orthogonal complement of remaining orbitals resides within the environment @xcite . during the optimization cycle for the cluster involving an advanced functional , the environment orbitals remain frozen and thus orthogonality
is preserved .
this approach avoids the inaccuracies associated with approximating the kinetic energy . in the present paper
we demonstrate the power of our embedding scheme in a proof - of - principle calculation adressing two problems for which standard kohn - sham dft is known to be inadequate : defects in silicon and polarons in titania .
we use the following hierarchy of methods : the cluster is treated by the ( expensive ) hybrid functional pbeh while the environment is treated only by the pbe functional .
we show that this embedding scheme implemented in the vienna ab initio simulation package ( ` vasp ` ) is robust and efficient .
we emphasize that the present embedding scheme is not limited to hybrid - dft in dft embeddings .
future extensions will adress the treatment of the cluster by rpa or quantum chemistry approaches .
we partition a system into two parts : a cluster of interest @xmath3 with atomic sites @xmath4 , ( @xmath5 ) with @xmath6 the number of atomic sites included in the cluster , and the surounding environment @xmath7 , containing @xmath8 atomic sites @xmath9 , ( @xmath10 ) . in a first step , the entire system ( @xmath11 )
is solved using a single , comparatively cheap exchange - correlation functional , e.g. , pbe @xcite , @xmath12 yielding kohn - sham orbitals @xmath13 with orbital energies @xmath14 and the density matrix @xmath15 with occupation numbers @xmath16 $ ] , where the index @xmath17 goes over all orbitals and physical spin . note that we we have not included @xmath18-point sampling in the present ansatz , since it is not straightforward to treat the transformations at different @xmath18-points independently .
we aim to find a unitary rotation within the subspace of fully occupied orbitals ( @xmath19 , @xmath20 ) that yields a set of orbitals aligned with the atomic orbitals @xmath21 localized around the atomic sites of the cluster .
the index @xmath22 of the atomic orbitals includes both the site index as well as radial and angular momentum quantum numbers . to this end
we apply to the orbital overlap matrix w , @xmath23 a singular value decomposition according to @xmath24 with @xmath25 .
the unitary matrix @xmath26 represents the rotation in the space of the @xmath27 occupied orbitals that optimally aligns @xmath28 of these orbitals with the atomic orbitals @xmath29 keeping the remaining @xmath30 orbitals orthogonal . the singular values @xmath31 provide a measure for the degree of overlap between the orbitals @xmath29 and the rotated orbitals @xmath32 , @xmath33 orbitals with indices @xmath34 outside the space of occupied orbitals are unaffected by the rotation , @xmath35 . using the rotated @xmath32 , we can thus partition the occupied space into orbitals which have an overlap with the @xmath29 , and those who do not .
ideally , if the kohn sham orbitals are well covered by the atomic wavefunctions , we expect the singular values @xmath31 to be close to 1 .
after the orbital rotations , a subset @xmath36 of those @xmath37 with @xmath38 can now be optimized based on a more expensive exchange - correlation ( xc ) functional @xmath39 , e.g. , a hybrid functional @xcite , while freezing the orthogonal complement of @xmath40 environment orbitals @xmath41 . in general
, the number of orbitals @xmath42 used in the embedding procedure may be smaller than the number of atomic basis functions @xmath28 : in principle , one may choose any subset of the localized orbitals @xmath43 .
in practice , we sort the rotated orbitals by their singular values @xmath31 , and choose the @xmath42 orbitals corresponding to the largest @xmath31 , where @xmath42 is chosen according to the number of orbitals of interest within the cluster .
a typical cut - off will be @xmath44 .
we find that our results do not strongly depend on @xmath42 , as long as the number of optimized orbitals is sufficiently large as to properly describe the local bonding .
we will discuss the choice of @xmath42 in more detail in the results section below .
note that after the orbital rotation , the @xmath32 are no longer eigenvectors of the kohn - sham hamiltonian @xmath45 .
the diagonals of the hamiltonian are given by the expectation values @xmath46 related to the original eigenvalues through the invariance of the trace @xmath47 under unitary rotations .
in the case of fractional occupations of some of the orbitals in @xmath48 , the above considerations remain valid in the subspace of the @xmath49 fully occupied orbitals . all orbitals with fractional occupation
are assigned to cluster @xmath3 , even if they are not well localized .
however , extension of the present approach to quantum - chemistry based correlated wavefunction approaches may face difficulties for such delocalized metallic states .
taking into consideration that different exchange - correlation functionals will be employed for the cluster @xmath3 and the environment @xmath7 we write the energy functional for the entire system @xmath11 as @xmath50 + e_{{\mathrm{xc}}}[\gamma_a,\gamma_b]\ ] ] where @xmath51 is the density , and @xmath52 $ ] the hartree energy @xmath53 = \frac 12 \sum_{i , j=1}^{n_{\mathrm{tot}}}f_if_j{\left < i\;j \right|}\mathcal{c}{\left| i\ ; j \right>},\ ] ] with @xmath54 denoting the coulomb operator . the mixed exchange - correlation
functional @xmath55 $ ] containing both the lower ( @xmath7 ) and higher ( @xmath3 ) level functionals can be written as @xmath56 = e_{xc}^a[\gamma_a ] + e_{{\mathrm{xc}}}^b[\gamma_b ] + e^{\mathrm{int}}_{{\mathrm{xc}}}[\gamma_a,\gamma_b].\ ] ] to use eq .
( [ eq : xcab ] ) in practice within our embedding approach , the contribution due to the interaction between the two subsystems , @xmath57 , should be approximated by the lower - level functional ( @xmath7 ) applied also to the environment , i.e. , @xmath58 = e_{\mathrm{xc}}^{\mathrm{int},b}[\gamma_a,\gamma_b]$ ] with @xmath59 = e_{\mathrm{xc}}^{b}[\gamma_{\mathrm{tot } } ] - e_{\mathrm{xc}}^b[\gamma_a ] - e_{\mathrm{xc}}^b[\gamma_b]\ ] ] this allows for very expensive functionals to be used in the cluster @xmath3 , including rpa or quantum chemistry approaches .
the drawback is that the error introduced in such a mixed approach is difficult to quantify a priori .
for the special case of hybrid functionals chosen in the present work as the `` higher - level '' functional @xmath3 the interaction term @xmath57 can be much more accurately approximated by the full hybrid functional itself @xmath60 & = & e_{{\mathrm{xc}}}^a[\gamma_a+\gamma_b ] - e_{{\mathrm{xc}}}^a[\gamma_a ] - e_{{\mathrm{xc}}}^b[\rho_b ] \end{aligned}\ ] ] as the most expensive summation @xmath61 is constant and hence not relevant for the optimization of orbitals in @xmath3 .
the only approximation here is that the orbitals in @xmath7 are kept frozen .
even when optimizing large supercells , only a subset @xmath62 of the full @xmath63 orbital pairings needs to be calculated to evaluate the relevant exchange contribution @xmath64 , greatly reducing the numerical effort .
the orbitals in @xmath3 can now be efficiently optimized minimizing the energy functional of eq .
( [ eq : etotal ] ) , while the orbitals in @xmath7 are kept frozen . consequently , during a single optimization , any change in the electronic structure of @xmath3 due to
a more accurate xc functional can not lead to a redistribution of charge in @xmath7 .
the embedding can now be made self - consistent by alternating between subsystems @xmath3 and @xmath7 in freeze - and - thaw cycles : after the initial solution of the entire system using the lower - level functional , an orbital rotation is performed to partition into orbital sets @xmath3 and @xmath7 .
then , starting with @xmath3 , alternatingly one of the subsystems is optimized while the other one is kept frozen .
each cycle that optimizes the orbitals @xmath3 uses the higher - level xc functional @xmath3 while each cycle that optimizes the orbitals in @xmath7 employs the lower - level functional @xmath7 .
for the example of polarons in titania discussed below , we find rapid convergence after about six freeze - and - thaw cycles @xcite .
since the interaction of the cluster region ( e.g. defect ) with its periodic image needs to be minimized , conventional defect modeling is hampered by the requirement of large supercells .
if the bulk material could well be described by conventional xc functionals , and only the defect structure requires more advanced techniques , conventional techniques still require an expensive evaluation of the entire exchange contribution . the embedding procedure outlined above
is ideally suited to significantly reduce computational effort while retaining high accuracy .
we have implemented the embedding scheme outlined above in the vienna ab initio simulation package ( ` vasp ` ) using the projector augmented wave method of blchl in the implementation of kresse and joubert@xcite .
usage is simple : in a first step , a conventional dft calculation of a system is performed . in a second step , the localized atomic basis functions @xmath29 are defined .
our implementation currently supports the paw basis functions ( the pseudo partial waves ) and standard spherical harmonics ( including hybrid orbitals such as @xmath65 ) with the radial dependence taken from suitably scaled hydrogen functions . `
vasp ` then starts an embedded calculation , performs the orbital rotation and optimizes the set of @xmath66 orbitals localized on the cluster @xmath3 while the remaining fully occupied kohn - sham orbitals of the environment @xmath7 are frozen ( for spin - polarized calculations the two spin components are handled independently in terms of the rotation and the number of optimized orbitals ) . during the freeze - and - thaw cycles , no further localization procedure according to eq .
( [ eq : rotation ] ) is required , and orbital sets @xmath3 and @xmath7 are interchanged . to enforce orthogonality between the currently optimized orbitals and the frozen environment we use the frozen orbitals as projector .
we do not currently support @xmath18-point sampling in the embedding calculation , since the orbital rotations at different @xmath18-points are not independent .
likewise , forces are not currently implemented in our formalism .
the geometries used in this work were taken from ref .
for the defects in silicon and where relaxed using the hse functional similar to ref . for the polarons in titania
. to compare final energies calculated from the functional eq .
( [ eq : etotal ] ) , we need to evaluate the full exchange energy with all orbitals once after self - consistently converging the orbitals of @xmath3 .
we benchmark our embedding approximation against a fully self consistent optimization of all orbitals using the hybrid functional .
additionally , we compare against evaluating the hybrid functional with orbitals obtained from using a conventional functional ( pbe ) .
obviously , both the embedded and the pbe orbitals are , by construction , not self - consistent with respect to the hybrid functional . in comparison with a full self - consistent optimization , a single evaluation step using the full hybrid functional with non - self consistent orbitals takes a small amount of time , while substantially improving the accuracy : errors in an inexact evaluation of the interaction between subsystems in eq .
( [ eq : excint ] ) are eliminated .
we denote corresponding energies by an asterisk ( @xmath67 ) in the following .
ccrrrr + defect & hybrid & & + h & 3.00 & 2.99 & ( -17 ) & 3.01 & ( 10 ) + x & 3.01 & 3.01 & ( 1 ) & 3.04 & ( 27 ) + c@xmath68 & 3.05 & 3.03 & ( -19 ) & 3.06 & ( 18 ) + t & 3.77 & 3.34 & ( -423 ) & 3.75 & ( -20 ) + vjt & 4.14 & 4.52 & ( 377 ) & 4.19 & ( 44 ) + v & 4.23 & @xmath694.83 & ( 599 ) & @xmath69 4.26 & ( 25 ) + as a first practical test of our new algorithm , we consider point defect structures in silicon @xcite .
we use a 64 atom supercell , and neglect @xmath18-point sampling both in the full hybrid benchmark and in the embedding calculations . due to the localized nature of the covalent bonds
involved , conventional kohn - sham dft fails to correctly reproduce experimental observations . by using hybrid functionals or even more advanced rpa formulations
@xcite , these problems are mitigated .
however , comparison with more accurate correlation functionals such as rpa and experiment show that currently available methods yield a wide range of predictions depending on the employed functional @xcite , highlighting the necessity to move towards higher level correlated wavefunction approaches . due to the large supercells required to avoid interaction of the defect sites with their periodic images , embedding the orbitals close to
the defect site seems desirable .
as benchmark for our embedding scheme , we consider defect formation energies of a set of common interstitial defects and vacancies .
we aim to reproduce the energetics of full hybrid functional calculations based on pbeh by a cheaper embedding calculation in which only a few ( six to ten ) orbitals localized in the immediate vicinity of the defect ( taken to be the @xmath3 orbitals of the cluster ) are treated using the hybrid functional , while the remaining 118 - 122 orbitals ( taken to be the @xmath7 orbitals ) are only treated by pbe .
we also compare our results to purely dft - based predictions . for all si defect calculations , we choose as atomic orbitals @xmath29 the paw pseudo - partial waves of the si atoms at and directly adjacent to the defect site , resulting in @xmath70 ( vacancies ) or 20 atomic basis functions ( 1 @xmath71 + 3 @xmath72 per atom ) for most cases .
increasing the number of basis functions per atom increases the overall overlap of the occupied kohn - sham orbitals with the defect site at the cost of a larger @xmath28 , and thus a larger overlap matrix @xmath73 of eq .
( [ eq : w ] ) .
consequently , the number of singular values @xmath31 increases . to obtain a set of orbitals well localized at the defect site , we choose all orbitals with singular values @xmath44 as embedded orbitals .
this procedure yields a number of selected embedded orbitals @xmath42 from 9 ( x defect ) to 16 ( t - defect ) , in line with the number of si - si bonds one would expect for the respective defect sites .
for example , each of the two defects atoms of the dumpbell defect ( x ) interacts strongly with four close neighbors in the surrounding lattice and with the other atom in the dumpbell , yielding a total of nine covalent bonds .
indeed , we find nine singular values substantially larger than 0.5 for this defect .
the paw basis functions we choose yield a set of orbitals with a bimodal distribution : a significant number of orbitals with @xmath74 , well seperated from delocalized orbitals with small overlap @xmath75 with the defect site .
the threshold of 0.5 is therefore a good compromise between choosing all @xmath28 possible orbitals ( which will include orbitals with very small singular values ) and too few orbitals that will not allow for reasonable optimization .
note , however , that care must be taken to check that there are no singular values close to 0.5 , to avoid arbitrarily including ( or discarding ) orbitals upon small fluctuations in @xmath31 . as mentioned in sec .
ii , calculating the defect formation energies from the total energies of eqs .
( [ eq : estart])-([eq : excint ] ) for the defects and the defect - free ( bulk ) system proves challenging . for the embedded case , vacancies and interstitials change the number of orbitals , and
make a comparison of absolute energies problematic .
we therefore compare the predictions by different methods for the formation energies by evaluating the same hybrid energy functional with the help of all orbitals ( i.e. , not just the ones localized at the defect ) generated by the different methods [ denoted by an asterisk ( * ) ] .
our results for various defect structures are summarized in tab .
[ tab : resultsdefects ] .
overall , we find excellent agreement between the hybrid functional benchmark and our embedding approach . for simple , non - metallic defects such as the dumbbell configuration ( x ) , the hexagonal hollow ( h ) and a lower - symmetry variant ( c@xmath68 ) we find that both the embedding , as well as the evaluation of the hybrid functional with the low - level dft orbitals produces good agreement with benchmark calculations ( see second column of tab . [ tab : resultsdefects ] ) .
by contrast , the metallic tetragonal site is badly described by dft : it features one interstitial si atom coordinated to its four nearest neighbors , so that the local coordination of the interstitial is identical to the other si atoms .
this position is unique insofar that the highest occupied orbital is threefold degenerate ( @xmath76 symmetry ) but only occupied by two electrons .
this degeneracy is preserved in dft yielding three fractionally occupied orbitals with occupation numbers @xmath77 .
consequently , the evaluation of the hybrid energy functional based on these orbitals fails to yield reasonable formation energies .
by contrast , the embedding method locally breaks the degeneracy as does the full hybrid calculation , leading to good agreement of the embedding results with the benchmark ( see t , vjt and v lines in tab .
[ tab : resultsdefects ] ) .
our results compare poorly with experimental data : one important reason is the interaction of periodic images of the defects to the supercell size .
we therefore consider a larger supercell of 512 atoms , still with a single defect .
we note that on our hardware , the full hybrid calculations takes ten times as long as the embedded one , with a relative error of 0.4% in total energy .
we find a substantial change in results for the larger cell ( compare tab .
[ tab : resultsdefectslarge ] ) , that now fit well to experimental results for the h defect . to achieve better agreement also for vacancies ( v )
requires a more accurate treatment of electronic correlation ( e.g. , rpa ) or inclusion of van der waals contributions @xcite .
.same comparison as in tab .
[ tab : resultsdefects ] for an eight times larger supercell containing 512 atoms .
rightmost column shows the range of available experimental data taken from [ ] . [ cols="<,^,>,>,>,>,^ " , ] as a second demonstration of our method , we consider the formation of polarons in titania .
we consider a @xmath78 supercell with 24 ti and 28 o atoms in the rutile structure . in an accurate hybrid functional description , an additional electron localizes , distorting the lattice and forming a small polaron .
the distortion decreases the energy compared to a delocalized charge .
a full hybrid functional calculation yields a decrease in energy by 514 mev for the distorted geometry .
we will use this value in the following as benchmark for our embedded description of the polaron .
concerning the atomic basis functions @xmath29 used for the initial localization , we typically use the 1 @xmath71 and 5 @xmath79 orbitals of the ti atom centered at the small polaron deformation , as well as all @xmath71 and @xmath72 orbitals of the six nearest neighbor oxygen atoms , yielding a total of @xmath80 orbitals .
geometries for the distorted structure were relaxed using hse calculations .
our results for the small polaron formation energy are summarized in tab .
[ tab : selfcons ] .
conventional density functional theory invoking a pbe functional is not capable of reproducing small polaron formation , predicting even a negative energy gain ( i.e. energy costs ) of -355 mev to form the polaron . inserting the dft orbitals in the hybrid energy functional leads to a correction of the sign .
however the energy gain is underestimated by a factor of two ( 270 mev ) , see tab .
[ tab : selfcons ] .
a single - cycle embedding calculation yields a slightly larger error predicting 190 mev .
the origin of this error is obvious : while the hybrid functional tries to localize the charge in the cluster region @xmath3 , the surrounding region @xmath7 can not react to the substantial change in the electrostatics , since all @xmath7 orbitals are frozen .
subsequent freeze - and - thaw cycles rapidly improve the result : we alternate between optimizing the two sets of orbitals @xmath3 and @xmath7 , one with the expensive hybrid , the other with pure gga ( pbe ) .
we find convergence in about seven iterations , quite independent of the number of embedded orbitals @xmath42 ( tab .
[ tab : selfcons ] ) . as minimum requirement for @xmath42 , the ti atom at the center of the distortion and the surrounding oxygen atoms need to be treated accurately , which is already achieved with as few as six orbitals ( tab .
[ tab : selfcons ] ) .
note that only choosing the central ti atom as atomic basis , @xmath81 , yields a smaller polaron energy than chosing the six orbitals with the highest singular values from the @xmath80 localized orbitals including also the closest oxygen atoms .
the reason is that in the latter case , the response of the surrounding shell of oxygen atoms is - to some degree - also treated by the hybrid functional .
however , further increasing the number of localized orbitals @xmath28 by , e.g. , also including a shell of neighboring ti atoms does not result in a stronger overlap of @xmath3 orbitals on the central ti atom ( note that the localization procedure does not distinguish between the different atomic basis functions @xmath29 ) .
consequently , such a large @xmath28 would require a comparatively large @xmath42 to ensure that orbitals close to the central site are included in the embedded calculation .
otherwise accuracy is lost .
indeed , we find a better agreement with the benchmark for @xmath82 than for @xmath83 , @xmath84 ( see tab .
[ tab : selfcons ] ) .
since the numerical effort of the embedding calculation scales with a power of @xmath42 , in practice a small @xmath28 that allows for @xmath85 is preferable .
it is intstructive to follow the charge density variations along the freeze - and - thaw cycles [ fig .
[ fig : convergence ] ] .
additional charge is localized in the @xmath3 cycles using the hybrid funcional in the cluster ( top row in fig .
[ fig : convergence ] ) .
the density spreads out again and the environment relaxes in the @xmath7 cycles when the orbitals @xmath7 of the environment are optimized using the dft functional .
however , the magnitude of these changes quickly decreases with the iteration number and yields a well - converged density ( and well - converged energy ) within 7 iterations . the full hybrid and the converged embedded unpaired spin densities closely match ( fig .
[ fig : methodcomp ] ) ( b , c ) .
by contrast , the dft density does not show a strong localization of the surplus electron at all ( fig .
[ fig : methodcomp ] ) ( a ) . indeed , projecting the converged polaron orbital ( i.e.
, the occupied majority spin kohn - sham orbital with the highest energy ) onto the central ti atom of the distortion yields quite small values for the overlap ( 0.39 ) for dft , while the full hybrid ( 0.69 ) and embedded calculations ( 0.65 ) agree quite well .
this underlines that despite the correct sign for the energy gain when using the dft orbitals in the hybrid energy functional , the dft description of the charge density is qualitatively deficient .
we have demonstrated a new embedding framework based on a suitable rotation in the subspace of fully occupied kohn - sham orbitals . using a projection on local basis functions ,
a set of orbitals may be localized at a site of interest , for example a defect .
subsequently , these localized orbitals inside the cluster can now be optimized based on a more expensive exchange - correlation functional , such as a hybrid functional involving the exact evaluation of fock exchange . since exchange interactions within the frozen environment
are neglected , the computation time is drastically reduced .
the response of the environment to the charge rearrangement in the cluster can be self - consistently included by freeze - thaw cycles in which alternatingly the orbitals in the embedded cluster or in the environment are optimized .
we have implemented our ansatz in the popular ` vasp ` software package .
as proof of principle , we have applied our method to two problems of interest : a set of defects in bulk silicon , and small polarons in bulk titania .
we find excellent agreement with ( much more expensive ) benchmark bulk hybrid calculations .
w. kohn and l. j. sham , phys . rev . *
140 * , 1133 ( 1965 ) .
k. burke , j. chem .
phys . * 136 * , 150901 ( 2012 ) .
a. j. cohen , p. mori - snchez , and w. yang , chem . rev . *
112 * , 289 - 320 ( 2012 ) .
a. pribram - jones , d. a. gross , and k. burke , ann .
phys . chem . *
66 * , 283 ( 2015 ) .
p. nozires , and d. pines , phys . rev . *
111 * , 442 ( 1958 ) .
d. langreth , and j. perdew , phys .
b * 15 * , 2884 ( 1977 ) .
t. miyake , f. aryasetiawan , t. kotani , m. van schilfgaarde , m. usuda , and k. terakura , phys .
b * 66 * , 245103 ( 2002 ) .
m. fuchs , y .- m .
niquet , x. gonze , and k. burke , j. chem . phys . * 122 * , 094116 ( 2005 ) f. furche , j. chem . phys . * 129 * , 114105 ( 2008 ) .
j. paier , b. g. janesko , t. m. henderson , g. e. scuseria , a. grneis , and g. kresse , j. chem . phys . * 132 * , 094103 ( 2010 ) .
j. paier , x. ren , p. rinke , g. e. scuseria , a. grneis , g. kresse , and m. scheffler , new j. phys .
* 14 * , 043002 ( 2012 ) .
a. tkatchenko , r. a. distasio jr . , r. car , and m. scheffler , phys .
lett . * 108 * , 236402 ( 2012 ) .
m. marsman , a. grneis , j. paier , and g. kresse , j. chem . phys . * 130 * , 184103 ( 2009 ) .
a. grneis , m. marsman , and g. kresse , j. chem . phys . * 133 * , 074107 ( 2010 ) .
a. grneis , g. h. booth , m. marsman , j. spencer , a. alavi , and g. kresse , j. chem .
* 7 * , 2780 ( 2011 ) . g. booth , a. grneis , g. kresse , and a. alavi , nature * 493 * , 365 - 370 ( 2013 )
. f. libisch , c. huang , p. liao , m. pavone , and e. a. carter , phys . rev
. lett . * 109 * , 198303 ( 2012 ) .
t. dresselhaus , j. neugebauer , s. knecht , s. keller , y. ma , and m. reiher , j. chem . phys . * 142 * , 044111 ( 2015 )
. j. nafziger , q. wu , and a. wasserman , j. chem .
* 135 * , 234101 ( 2011 ) .
j. d. goodpaster , n. ananth , f. r. manby , and t. f. miller , j. chem . phys . * 1133 * , 084103 ( 2010 ) . j. d. goodpaster , t. a. barnes , and t. f. miller , j. chem .
phys . * 134 * , 164108 ( 2011 ) .
j. d. goodpaster , t. a. barnes , f. r. manby , and t. f. miller , j. chem . phys . * 137 * , 224113 ( 2012 ) . t. a. wesolowski and a. warshel , j. phys .
chem . * 97 * , 8050 ( 1993 ) .
s. fux , c. r. jacob , j. neugebauer , l. visscher , and m. reiher , j. chem .
phys . * 132 * , 164101 ( 2010 ) .
f. libisch , c. huang , and e. a. carter , acc .
47 * , 2768 ( 2014 ) . c. huang , m. pavone , and e. a. carter , j. chem
* 134 * , 154110 ( 2011 ) .
g. knizia , and g. k .- l .
chan , phys .
* 109 * , 186404 ( 2012 ) .
f. r. manby , m. stella , j. d. goodpaster , and t. f. miller , j. chem .
theory comput .
* 8 * , 2564 ( 2012 ) .
m. e. fornace , j. lee , k. miyamoto , f. r. manby , and t. f. miller , iii , j. chem .
theory comput .
* 11 * , 568 - 580 ( 2016 ) j. p. perdew , k. burke , and m. ernzerhof , phys
. rev . lett . * 77 * , 3865 ( 1996 ) .
a. d. becke , j. chem . phys . * 98 * , 1372 ( 1993 ) .
j. p. perdew , m. ernzerhof , and k. burke , j. chem . phys . * 105 * , 9982 ( 1996 ) . p. e. blchl , phys .
b * 50 * , 17953 ( 1994 ) .
g. kresse and d. joubert , phys .
b * 59 * , 1758 ( 1999 ) .
m. kaltak , j. klimes , and g. kresse , phys .
b * 90 * , 054115 ( 2014 ) .
a. janotti , c. franchini , j. b. varley , g. kresse , c. g. van de walle , phys .
* 7 * , 199203 ( 2013 ) .
w. gao and a. tkatchenko , phys .
* 111 * , 045501 ( 2013 ) g. kresse , and j. furthmller , comp
. mat .
science * 6 * , 15 ( 1996 ) .
d. riedel et al . , phys .
109 * , 226402 ( 2012 ) .
s. dannefaer , p. mascher , and d. kerr , phys .
lett . * 56 * , 2195 ( 1986 ) .
p. m. fahey , p. b. griffin , and j. d. plummer , rev .
phys . * 61 * , 289 ( 1989 ) .
y. shimizu , m. uematsu , and k. m. itoh , phys .
lett . * 98 * , 095901 ( 2007 ) h. bracht et al . , phys .
* 91 * , 245502 ( 2003 ) .
h. bracht , e. e. haller , and r. clark - phelps , phys .
rev . lett . * 81 * , 393 ( 1998 ) .
a. ural , p. b. griffin , and j. d. plummer , j. appl .
phys . * 85 * , 6440 ( 1999 ) .
a. ural , p. b. griffin , and j. d. plummer , phys .
* 83 * , 3454 ( 1999 ) h. bracht , j. f. pedersen , n. sangenberg , a. n. larsen , e. e. haller , g. lulli , and m. posselt , phys . rev
. lett . * 91 * , 245502 ( 2003 ) .
v. ranki , and k. saarinen , phys .
lett . * 93 * , 255502 ( 2004 ) . | we present an embedding approach for semiconductors and insulators based on orbital rotations in the space of occupied kohn - sham orbitals .
we have implemented our approach in the popular vasp software package .
we demonstrate its power for defect structures in silicon and polaron formation in titania , two challenging cases for conventional kohn - sham density functional theory . |
in order to study the effects of the tensor and spin orbit interactions in nuclei we use a simple interaction v = v@xmath5 + x v@xmath6 + y v@xmath7 where c @xmath8 central , s.o . @xmath8 spin - orbit and t @xmath8 tensor . for x=1 , y=1 we select v so as to be close to a realistic g matrix like bonn a. we can turn the spin orbit interaction off ( on ) by setting x=0 ( x=1 ) .
likewise we can turn the tensor interaction off ( on ) by setting y=0(1 ) .
this allows us to study behaviors as a function of x and/or of y. this interaction @xcite is a modification and correction of a previous interaction @xcite .
this change does not affect calculations purely in the p shell but there are some changes when core excitations are included .
to avoid confusion we call the current interaction v(2005 ) and the previous one v(1991),i.e after the year of publication .
the details and reasons for the changes are given in ref .
@xcite .
our main thesis will be that the tensor interaction given by a bare g matrix is too strong in the isospin t=0 channel .
by simply making it weaker we can correlate a lot of data and be rid of a lot of anomalies .
we conclude by presenting simple arguments by gerry brown to justify using a weaker tensor interaction in the valence space .
care must be taken in that alternate explanations could give the same result as a weaker tensor interaction e.g. a stronger spin - orbit interaction .
the topics we discuss are : 1 . the quadrupole moment of the j=1@xmath9 state of @xmath0li .
2 . the t=1 t=0 energy splitting of 0@xmath1 states in @xmath20 .
the near vanishing of the gamow - teller matrix element for a=14 @xmath3o(j=0 t=1 ) @xmath10
@xmath3n(j=1 t=0 ) .
4 . the effect of the tensor interaction on single particle energies in open shell nuclei - @xmath4c and @xmath3n . some of these results have been discussed in previously @xcite- @xcite .
the nucleus @xmath0li is often described in cluster models as a deuteron plus an alpha particle . however , the quadrupole moment of the deuteron is positive @xmath12 e @xmath13 whereas that of @xmath0li is negative , @xmath14 e @xmath13 . that the deuteron has a quadrupole moment leads to it having a j=1@xmath11 ground state and hence the isospin must be t=0 . without a tensor interaction the quadrupole moment of the deuteron
would be zero .
.static quadrupole moments e @xmath13 for various model spaces and tensor interaction strengths ( y ) using the bare electric charges @xmath15 and @xmath16 .
all calculations are done with the full spin - orbit strength x=1 . [
cols="^,^,^,^ " , ] the second order contribution to the central interaction is negative definite and for the tensor it is positive definite .
hence whether there is destructive or constructive interference depends on the sign of the first order term as given above .
we start from a moszkowski - scott cutoff radius of 1 fermi corresponding to x of 0.7 .
they argued that the part of the attraction up to the cut - off radius of 1 fermi cancels out the short - range repulsion whose range is about 0.4 fermi .
@xcite the first order central even is negative so adding the second order tensor contributions makes it more negative .
this is reasonable from the existence of the t=0 even channel bound state for two nucleons , namely the deuteron .
the bare central interaction in this channel is not deep enough to support a bound state so the tensor interaction has to contribute .
we can see this in figure [ fig : vce ] the first order tensor even interaction is negative ( when the -3 factor is included ) so the combination of first and second order terms must be less negative or weaker .
the sign of the first order tensor interaction is determined phenomenologically by the positive sign of the quadrupole moment of the deuteron .
also the sign is consistent with the one pion exchange potential .
this supports all the conclusions of the previous sections where we see repeatedly that the bare g matrix tensor interaction is too strong in the t=0 channel and needs weakening . for the odd channels we have first in figure [ fig : vco ] the central odd potential . here
the attractive contribution of the second order term pulls an initially repulsive first order term down sufficiently so that it is slightly attractive . for the tensor odd interaction in [ fig : vto ] ,
the inclusion of the second order term again makes the total tensor portion less attractive .
we emphasize that the main point of this work is to show that there are clear experimental signatures that require that in the spin triplet channel renormalization relative to a bare g matrix are required .
this is especially true for even l states where not only is the effective central part of the interaction made deeper but also the effective tensor interaction is weakened ( screening effects ) .
relative to the use of only a first order tensor interaction , the combined first and second order tensor interaction helps to explain the the smaller energy splitting of t=1 and t=0 0@xmath23 states in @xmath2o and the vanishing of the gamow - teller matrix element in the @xmath3c beta decay .
also the anti - spin orbit effects in open shell nuclei like @xmath4c are reduced although they are still substantial .
we still have a problem with @xmath0li .
although we have shown that one needs the tensor interaction to get a negative quadrupole moment we get it to be increasingly negative with increasing model space .
in closing we note that the shell model works very well in the p shell as noted by the many works of cohen and kurath @xcite .
one purpose here is to understand why it works so well .
we see that although the explicit configurations involving higher shells are not present in most calculations , their implicit presence is of crucial importance for the success of the model .
we have adopted a low - tech approach to illustrate this point .
for more trustworthy quantitative results , the high powered ab initio shell model methods or other equivalent methods need to be employed .
nevertheless we feel that the more qualitative methods used here are of considerable value in providing insight into the physics behind these more complex approaches . | we show several examples were the tensor interaction of the lowest order g matrix in a nucleus is too strong .
the examples include the quadrupole moment of @xmath0li , the isosplitting of the lowest 0@xmath1 states in @xmath2o , the near vanishing gamow - teller matrix element in the weak decay of the j=0 t=1 state of @xmath3o to the j=1 t=0 ground state of @xmath3n , and the magnitude of the deformation of @xmath4c .
it would appear that we could get better results by decreasing the tensor interaction strength by about a factor of two .
we then examine the simple estimates of gerry brown concerning second order tensor effects .
we note that for the triplet even channel the combination of first and second order tensor does indeed yield an effective weaker tensor interaction and helps to get better agreement with experiment . |
we appreciate the assistance of the technical staff and operators at the 88-inch cyclotron .
this work was supported by the director , office of science , office of nuclear physics , u.s .
department of energy under contract no .
de - ac02 - 05ch11231f .
part of this work was performed under the auspices of the u.s .
department of energy lawrence livermore national laboratory under contract de - ac52 - 07na27344 .
sm fed by ground and isomer state decays of @xmath0eu .
the minimum chi - squared fits with eq .
[ eq : oscdec ] ( red ) and exponential decay ( blue ) are shown with the data .
the 768 kev data shows decay of both the 2.4 s ground state and the 1.22 m ( @xmath63 ) state of @xmath0eu.,width=340 ] when the data are fit with just exponential decay , without an oscillating term , assuming the known lifetimes of the ground and isomeric states in @xmath0eu .
the fourier transform is taken of residuals ( data minus fit ) to search for a frequency peak not accounted for by an exponential decay term .
no peak is resolved at either 0.14 hz ( 7 seconds ) or 0.2 hz ( 5 seconds ) in either data set.,width=340 ] nd k@xmath3 x - rays .
the number of counts in a window surrounding the k@xmath3 peak is plotted as a function of time after production of @xmath0pm .
the best fit exponential decay curve is also shown . ,
width=340 ] for the first 40 seconds of decay , as examined in ref .
the best fit to eq .
[ eq : oscdec ] is shown ( red ) , finding @xmath48 and @xmath64 s , as is a curve ( green ) calculated with an oscillation amplitude @xmath65 and @xmath66 s , suggested by ref . | we have searched for time modulation of the electron capture decay probability of @xmath0pm in an attempt to confirm a recent claim from a group at the gesellschaft fr schwerionenforschung ( gsi ) .
we produced @xmath0pm via the @xmath1sn(@xmath2na , 5n)@xmath0pm reaction at the berkeley 88-inch cyclotron with a bombardment time short compared to the reported modulation period .
isotope selection by the berkeley gas - filled separator is followed by implantation and a long period of monitoring the @xmath0nd k@xmath3 x - rays from the daughter .
the decay time spectrum of the x - rays is well - described by a simple exponential and the measured half - life of 40.68(53 ) seconds is consistent with the accepted value .
we observed no oscillatory modulation at the proposed frequency at a level 31 times smaller than that reported by litvinov _
et al . _
( phys .
lett .
b 664 ( 2008 ) 162 ) . a literature search for previous experiments that might have been sensitive to the reported modulation uncovered another example in @xmath0eu electron - capture decay .
a reanalysis of the published data shows no oscillatory behavior .
a recent paper reported observation of an oscillating decay rate of the isotopes @xmath4pr and @xmath0pm when electron capture decays were measured using highly charged ions with a schottky mass spectroscopy technique @xcite .
the authors concluded that the decay activity oscillated in time with a period of about 7 seconds and a relative amplitude of 20% for both isotopes , and attributed the oscillating behavior to interference between neutrino mass eigenstates in the two - body kinematics for electron capture decay of the hydrogen - like ions .
if confirmed , this effect is surprising and might offer a new avenue for studying neutrino mixing .
references @xcite suggest quantitative explanations for the observations of ref .
@xcite in terms of two - species neutrino mixing .
other authors argue strongly in refs .
@xcite that associating the claimed decay rate modulation with neutrino oscillation is inconsistent with well - established principles of quantum mechanics .
we do not wish to enter the debate about the consistency of the experimental claims with the theoretical treatment of neutrino oscillations and quantum mechanics in the present paper .
instead we focus on the experimental issues and describe our attempt to confirm the findings of the gsi group .
the experiment at gsi exploited several unique features of the heavy ion synchrotron and storage ring facility @xcite .
the measurement was made on hydrogen - like ions of @xmath0pm and @xmath4pr produced in flight and then mass separated . after a short period of stochastic cooling , one or two ions at a time are captured , and the cyclotron frequency is measured as the ions coast in the ring .
the relevant parameter is the time to electron capture following ion injection . with this elegant technique , the daughter ions ( which have the same charge but slightly different masses )
are also trapped and the time of electron capture is determined as a discontinuous change of the cyclotron frequency .
the authors find that their decay data are well - described by @xmath5 .
% \end{equation}\end{aligned}\ ] ] according to the model in refs .
@xcite , the oscillation period is @xmath6 , where @xmath7 is the mass of the daughter nucleus , @xmath8 is the lorentz factor of the moving stored ions ( @xmath9 for the stored ions in the gsi experiments ) , and @xmath10 is the squared mass difference between the two participating neutrino mass eigenstates .
we have considered the possibility that the oscillating decay rate effect might have been missed in previous experiments studying electron capture or even ordinary beta decay using more traditional radiation detectors and implanted sources of neutral atoms .
we searched the literature for decays with relatively high electron capture probability and with half - lives within a range suggested by the model in @xcite ( @xmath11 ) .
the modulation period depends linearly upon the mass of the decaying nucleus , and in nuclei around a=150 , the period ( at rest in the lab frame ) is about 5 seconds .
there are numerous examples of measured decays in this mass region that have characteristics similar to @xmath0pm and @xmath4pr . in most experiments to measure decay properties , a typical procedure is to bombard the target for a time comparable to the lifetime of the isotope of interest .
this timing optimizes the data collection rate , but reduces sensitivity to a time - modulated decay rate because the modulation for nuclei produced at different times would destructively interfere .
an oscillating effect would be most clearly seen when decay time and modulation period are very similar . on the other hand
, the effect would be difficult to see for lifetimes much shorter than the modulation period .
for @xmath4pr and @xmath0pm , typical long bombardment and production times would preclude the observation of a 5 to 10 second modulated decay time spectrum .
we found one earlier experiment which did have the necessary short time bombardment followed by a long counting period : a study of the isomeric decay of @xmath0eu to states in @xmath0sm @xcite . in that work , the @xmath0eu
was prepared using short irradiations from 0.1 to 2.0 s , which would preserve a 5 to 10 second oscillation , but with a diminution of the modulation amplitude from the suggested 20% to between 19.9% and 15% ( depending on the irradiation time ) .
the decays of the ground @xmath12 ( @xmath13 s ) and @xmath14 isomer ( @xmath15 m ) states were observed by subsequent gamma decay of excited states of @xmath0sm populated uniquely by ground or isomer decays .
figure 1 from ref .
@xcite shows the decay of the gamma activity produced by decays of @xmath0eu@xmath16 , with 1 second time bins , extending to 50 seconds after bombardment .
the electron capture probability for the 1.22 m @xmath14 isomer decay is 17% . the data from figure 1 of ref .
@xcite show no obvious oscillations
. a fit of that data ( shown in this paper in fig .
[ fig:142eudatfit ] ) using eq .
[ eq : oscdec ] finds a minimum of @xmath17 with an oscillation period of 3.540(53 ) seconds , an amplitude of @xmath18 , and a phase @xmath19 rad for the 1023 kev data . for the 768 kev data ,
we find a minimum @xmath20 with an oscillation period of 4.854(73 ) seconds , @xmath21 , and @xmath22 rad .
when performing fits to the data using eq .
[ eq : oscdec ] we restrict the phase to @xmath23 and ensure @xmath24
. fits to simple exponential decay find @xmath25 for the 1023 kev data and @xmath26 for the 768 kev data .
the fourier transform power spectra in fig .
[ fig:142eufft ] are calculated for the residuals to simple exponential fits .
the fourier spectra show no well - resolved peaks corresponding to 5 second oscillation ( or 7 second oscillation , under the assumption that the oscillation period is not proportional to the lorentz factor @xmath8 ) .
our @xmath0eu analysis would limit an oscillatory term to be a factor of about 3 smaller than that reported in ref .
@xcite , if we assume that our uncertainty in @xmath27 ( @xmath28 ) represents our sensitivity to an oscillating term , and if we correct for the electron capture branching ratio and account for the small reduction in an oscillation amplitude from the ( up to ) 2 s bombardment time .
our best - fit period and phase parameters disagree strongly with the results of @xcite .
the electron capture branching ratio also depends on the ionization state of the parent atom , as shown in refs .
@xcite , a distinction between the experiments in ref .
@xcite and ref .
@xcite which used stopped , presumably neutral atoms .
there is no reason _ a priori _ to expect that an electron capture decay rate oscillation would have the same amplitude in eu compared to pm and pr .
although this result would seem to disfavor the time modulation of electron capture decay , the uncertainty in the possible oscillation amplitude is large and the statistical power of this data is limited . despite the negative evidence from @xmath0eu and because of the importance of the gsi claim , we performed a test in @xmath0pm , one of the two isotopes reported to show a positive effect .
we studied the electron capture and beta decay rate from @xmath0pm , using a source of stopped ions in a foil . a thin target (
400 @xmath29g/@xmath30 @xmath1sn backed with carbon ) was bombarded with 95 mev @xmath2na@xmath31 ( average beam intensity 100 pna ) from the 88-inch cyclotron at lawrence berkeley national laboratory .
this beam energy was selected to produce predominantly @xmath0pm , and calculations with pace-2 @xcite indicate that other @xmath32 isotopes have production rates about a factor of 9 lower than @xmath0pm .
the reaction products moved through the berkeley gas - filled separator ( bgs ) @xcite , which separated the @xmath0pm from the beam and other products by their different magnetic rigidities .
the @xmath0pm stopped in a 25 @xmath29 m thick aluminum foil at the focal plane of the bgs .
an intrinsic germanium
clover " detector @xcite is located just outside a 2 mm thick aluminum vacuum window and counted the x - ray and gamma - ray emissions from the stopped @xmath0pm . after a short bombardment time of 0.5 s ,
the primary beam was shut off , and events from the germanium detector were recorded for 300 s before repeating the beam / count cycle .
approximately 190 bombardment cycles were performed over a 32 hour run . with the cyclotron beam off ,
the germanium detector recorded roughly 300 events per second above the 20 kev threshold .
figure [ fig : pmgloenergy ] is the energy spectrum from the clover germanium detector , summed over all beam bombardment cycles and counting time bins .
events during the beam bombardment are shown separately from the spectrum measured after the beam has been shut off .
events showing energy deposition in multiple leaves " of the clover detector have been excluded ( which reduces the compton background under the x - ray portion of the spectrum ) .
k@xmath33 and k@xmath34 from the daughter @xmath0nd are not resolved at 37.36 and 36.85 kev , but k@xmath3 is distinguished from k@xmath35 at about 42.5 kev in the beam off " spectrum .
the peaks at 433 , 381 , 241 , 208 , and 43 kev are cascaded from a 67 @xmath29s ( @xmath36 ) state populated in the production reaction @xcite , and confirm that @xmath0pm was indeed produced and well - distinguished from other isobars .
these peaks disappear in the beam off " spectrum , confirming that the cyclotron beam was turned off , and confirming that the production time of the @xmath0pm was short compared to a 5 to 10 second oscillation period .
the events were sorted into time - binned energy spectra for every 0.5 s. to measure the electron capture decays of @xmath0pm , we measured the rate of detected k - shell x - rays ( k@xmath33 , k@xmath34 ) .
electron capture decay strongly favors k - shell x - ray emission compared to positron decay .
the electron capture branching ratio for neutral @xmath0pm is 22% , of which about 85% is k - shell capture @xcite .
the x - ray fluorescence yield for a k - shell vacancy ( k@xmath3 plus k@xmath35 ) for nd is 91% @xcite .
positron emission by @xmath0pm can in principle produce k - shell vacancy and hence k - shell x - rays via shakeoff and direct collision .
however , these processes strongly favor outer shell vacancy production , particularly in high z atoms .
the k - shell vacancy probability after @xmath37 decay should be very small about @xmath38 for @xmath0pm @xcite . a very conservative upper limit on the amount of k - shell x - rays arising from @xmath37 emission would be 1.5% , assuming that all the auger k electron yield from @xmath0pm decay @xcite was attributable to @xmath37 .
a detected k - shell x - ray therefore has a greater than 92% probability to have been created by electron capture decay , and a less than 8% probability to have come from positron emission .
detection of a k - shell x - ray therefore strongly selects electron capture decays .
figure [ fig : pmkadecay ] shows the time decay of the nd k@xmath3 x - rays for the entire 300 s counting time along with a best - fit exponential decay curve .
for each 0.5 s time bin after the beam bombardment , we histogram the number of counts in a 5 kev window surrounding the k@xmath3 peak .
no oscillation is apparent in fig .
[ fig : pmkadecay ] .
we performed @xmath39 minimizing fits of the decay data to the function in eq .
[ eq : oscdec ] to search for a resolved oscillatory time dependence .
fixing @xmath40 and allowing @xmath41 , @xmath42 , and @xmath43 to vary for a simple exponential decay gave @xmath44 with 594 degrees of freedom . allowing @xmath41 , @xmath42 , @xmath43 , @xmath27 , @xmath45 , and @xmath46 to vary in the fits , we find a minimum @xmath47 , with @xmath48 , @xmath49 , and an oscillation period of 3.178(36 ) s. to evaluate the statistical significance of the fits using eq .
[ eq : oscdec ] compared to a simple exponential , we performed identical fits of monte - carlo generated data which had similar statistical power .
the monte - carlo generated data contained only exponential decay with the same time - bin structure as the experimental data , with no oscillating terms .
repeated trials of the fit procedure to eq .
[ eq : oscdec ] on several hundred randomized exponential decay data sets found an average oscillation amplitude @xmath50 by 2.5 standard deviations .
moreover , the @xmath39 parameter of the fits using oscillation terms improved , according to a fisher f - test , at a confidence level of 5% . comparing the fits of the real data using a fisher f - test
, we find a confidence level of 29.5% for the `` null hypothesis '' that the oscillating terms are real .
that is , the hypothesis that there _ is _ an oscillating term in the real @xmath0pm decay data would only be justified at a 29.5% confidence level , compared to the 5% confidence level generated by the fits to simulated , non - oscillatory data , and compared to the 0.6% confidence claimed in @xcite ) .
this supports the conclusion that the fits to our data from @xmath0pm do not include statistically significant oscillations .
the fisher f - test confidence levels are misleading in this instance .
fourier transforms of the residuals to a simple exponential fit to the data are shown in fig .
[ fig : fftresid ] , and show no statistically significant peaks at 1/(5 s ) ( or 1/(7 s ) , assuming that the oscillation period does not depend on @xmath8 ) . in all fit cases ,
we find @xmath51 s , which agrees with the accepted lifetime of @xmath0pm ( 40.5(5 ) s @xcite ) .
figure [ fig : pmcompare ] shows the first 40 seconds of our decay data , roughly the time period considered in @xcite for @xmath0pm , along with a calculated decay curve with @xmath52 , @xmath53 s , and @xmath54 rad found in @xcite .
these parameters are inconsistent with the data .
we conclude that any oscillation of 5 ( or 7 ) seconds , if present , must have an amplitude smaller than that found in ref .
@xcite ( @xmath55 ) by about a factor of 31 , using the uncertainty in the oscillation amplitude fit parameter as our sensitivity .
our beam bombardment time of 0.5 s would reduce our sensitivity to an oscillation amplitude , but a simple calculation ( integrating oscillating exponential decay and production ) suggests that this would reduce our sensitivity by 2% ( i.e. if @xmath52 , this experiment would have measured @xmath56 ) .
our best - fit oscillation amplitude is not statistically significant , and in any case has an amplitude a factor of 16 smaller than that proposed in @xcite , with a much different period and phase . as a cross - check on our data , we measured the decay of 511 kev positron annihilation activity . reference @xcite suggests that the decay rate for electron capture events will oscillate while the @xmath37 decay rate might not , since the three - body phase space density for positron emission has a large neutrino momentum distribution , which would average out any coherence between final neutrino mass states
. reference @xcite concludes that the positron decay of hydrogen - like ions will oscillate on a time scale too fast to have been observed in our data .
the 511 kev decay data was fit to a simple exponential decay , finding a half - life of @xmath57 s with @xmath58/596 . fitting this data to eq .
[ eq : oscdec ] we find a minimum @xmath59 for @xmath60 , @xmath61 and period @xmath62 s. the confidence level for this fit compared to simple exponential decay is 2.8% , suggesting that the oscillatory fit parameters improve the fit , more so than in the electron capture data set analyzed here , and slightly more so than in our monte - carlo simulated data which containted no oscillations .
this is contrary to the expectation under the neutrino mixing hypothesis that there should be no oscillation in the positron decay activity .
this result supports the conclusion that our oscillation fit parameters in the electron capture decay data are not statistically significant .
it might be argued that our experiment using neutral atoms would be insensitive to the proposed neutrino oscillation effect , since the participation of the remaining atomic electrons could provide a decoherence of the neutrino momentum states in the larger phase space of the final atomic states after the decay .
however , the usual description of electron capture decay suggests that the neutrino and recoil nucleus momenta are determined largely by the structure of the weak interaction hamiltonian . in the case where multiple atomic electrons are present ,
there can in principle be interference between contributions from different atomic electronic states . in calculating the total decay rate , or the ratio of k - shell to l - shell electronic capture , or the ratio of electron capture to positron decay ,
the neutrino ( and recoil nucleus ) momentum is determined by an overlap integral of the atomic electrons wave functions , summed over electron states @xcite .
but these calculations , particularly in high - z nuclei such as nd , are dominated by the k - shell contribution , and multiple electron terms represent corrections to the total decay rate of only a few percent @xcite . moreover , our experiment detected k - shell x - rays , meaning that the captured electron was indeed a k - shell electron with a similar wavefunction to the the hydrogenic ions investigated in @xcite .
the subsequent atomic de - excitation processes do not greatly influence the generation of the neutrino or recoil momenta .
if multiple electron effects destroy the coherence of the mixed neutrinos momenta in the final state , this would be apparent in data from the gsi group comparing the decay time spectrum of hydrogen - like and helium - like stored ions , similar to data reported in @xcite .
a further desirable confirmation of the data from the gsi group would be to examine the @xmath37 decays of the hydrogen - like ions , which should show no oscillations on the timescales available for examination , according to ref .
@xcite . to summarize
, no convincing oscillation was observed in the decay time spectrum of electron capture decays of @xmath0pm ( or in @xmath0eu using published data ) when dressed with their full complement of electrons and at rest in a solid metal matrix .
any 5 second oscillation not resolved in this experiment must have an amplitude a factor of 31 times smaller than that reported in ref .
@xcite .
the proposed oscillating decay rate could in principle be attributable to the truly two - body nature of the final state in the hydrogen - like decays observed in @xcite , although this would require an unconventional explanation with respect to electron capture decay in neutral atoms . under this hypothesis
, data on the decay of helium - like stored ions would show much different oscillation behavior . |
the recent discovery of the transiting extrasolar planetary system wasp-18 ( * ? ? ?
* hereafter h09 ) lights the way towards understanding the tidal interactions between giant planets and their parent stars .
wasp-18b is one of the shortest - period ( @xmath5d ) and most massive ( @xmath6@xmath1 ) extrasolar planets known .
these properties make it an unparallelled indicator of the tidal dissipation parameters @xcite for the star and the planet @xcite . a value similar to that observed for solar system bodies ( @xmath7@xmath8 ; @xcite ) would cause the orbital period of wasp-18 to decrease at a sufficient rate for the effect to be observable within ten years ( h09 ) .
in this work we present high - precision follow - up photometry of wasp-18 , obtained using telescope - defocussing techniques @xcite which give a scatter of only 0.47 to 0.83 mmag per observation .
these are analysed to yield improved physical properties of the wasp-18 system , with careful attention paid to statistical and systematic errors .
the quality of the light curve is a critical factor in measurement of the physical properties of transiting planets @xcite . in the case of
wasp-18 the systematic errors arising from the use of theoretical stellar models are also important , and are a limiting factor in the understanding of this system .
[ cols="<,^,^,^,^,^,<,^,^,^",options="header " , ] the physical properties of a transiting planetary system can not in general be calculated purely from observed quantities . the most common way to overcome
this difficulty is to impose predictions from theoretical stellar evolutionary models onto the parent star .
we have used tabulated predictions from three sources : _ claret _
@xcite and _ cambridge _ @xcite .
this allows the assessment of the systematic errors caused by using stellar theory .
we began with the parameters measured from the light curve and the observed velocity amplitude of the parent star , @xmath10ms@xmath11 ( h09 ) .
these were augmented by an estimate of the velocity amplitude of the _ planet _ , @xmath12 , to calculate preliminary physical properties of the system .
we then interpolated within one of the grids of theoretical predictions to find the expected radius and @xmath13 of the star for the preliminary mass and the measured metal abundance ( @xmath14}}= 0.00 \pm 0.09 $ ] ; h09 ) .
@xmath12 was then iteratively refined to minimise the difference between the model - predicted radius and @xmath13 , and the calculated radius and measured @xmath13 ( @xmath15k ; h09 ) .
this was done for a range of ages for the star , and the best overall fit retained as the optimal solution .
finally , the above process was repeated whilst varying every input parameter by its uncertainty to build up a complete error budget for each output parameter @xcite .
a detailed description of this process can be found in @xcite .
table[tab : absdimall ] shows the results of these analyses .
the physical properties calculated using the _ claret _ and _ y@xmath9 _ sets of stellar models are in excellent agreement , but those from the _ cambridge _ models are slightly discrepant .
this causes systematic errors of 4% in the stellar mass and 2% in the planetary mass , both of similar size to the corresponding statistical errors .
the quality of our results is therefore limited by our theoretical understanding of the parent star .
our final results are in good agreement with those of h09 ( table[tab : absdimall ] ) , but incorporate a more comprehensive set of uncertainties .
these results give the equilibrium temperature of the planet to be one of the highest for the known planets : @xmath16 where @xmath17 is the bond albedo and @xmath18 is the heat redistribution factor . this equilibrium temperature , and the closeness to its parent star , make wasp-18b a good target for the detection of thermal emission and reflected light .
we have presented high - quality observations of five consecutive transits by the newly - discovered planet wasp-18b , which has one of the shortest orbital periods of all known transiting extrasolar planetary systems ( teps ) .
our defocussed - photometry approach yielded scatters of between 0.47 and 0.83 mmag per point in the final light curves .
these data were analysed using the jktebop code , which was modified to include the spectroscopically derived orbital eccentricity in a statistically correct way .
the light curve parameters were then combined with the predictions of theoretical stellar evolutionary models to determine the physical properties of the planet and its host star .
a significant source of uncertainty in our results stems from the use of theoretical models to constrain the physical properties of the star .
further uncertainty comes from observed @xmath13 and @xmath19 $ ] , for which improved values are warranted .
however , the systematic error from the use of stellar theory is an important uncertainty in the masses of the star and planet .
this is due to our fundamentally incomplete understanding of the structure and evolution of low - mass stars . as with many other transiting systems ( e.g. wasp-4 ;
@xcite ) , our understanding of the planet is limited by our lack of understanding of the parent star .
we confirm and refine the physical properties of wasp18 found by h09 .
wasp-18b is a very massive planet in an extremely short - period and _ eccentric _ orbit , which is a clear indicator that the tidal effects in planetary systems are weaker than expected ( see h09 ) .
long - term follow - up studies of wasp-18 will add progressively stricter constraints on the orbital decay of the planet and thus the strength of these tidal effects .
we now split the full sample of known ( i.e. published ) teps into two classes according to planetary mass .
the mass distribution of transiting planets shows a dearth of objects with masses in the interval 2.03.1@xmath1 .
there are nine planets more massive than this and 46 less massive .
seven of the nine high - mass teps have eccentric orbits ( hat - p-2 , @xcite ; hd17156 , @xcite ; hd80606 , @xcite ; wasp-10 , @xcite ; wasp-14 , @xcite ; wasp-18 ; xo-3 , @xcite ) , and the existing radial velocity observations of the remaining two can not rule out an eccentricity of @xmath20 or lower ( corot - exo-2 , @xcite ; ogle - tr - l9 , @xcite ) . by comparison ,
only four of the 46 low - mass teps have a _ significant _ @xcite orbital eccentricity measurement .
these numbers imply that the more massive teps are a different population to the less massive ones ; fisher s exact test @xcite returns a probability lower than @xmath21 of the null hypothesis ( although this does not account for our freedom to choose the dividing line between the two classes ) .
this indicates that the two types of teps have a different internal structure , formation mechanism , or evolution , a suggestion which is supported by observations of misalignment between the spin and orbital axes of @xmath22@xmath1 teps @xcite .
there is , however , a bias at work here .
the more massive teps cause a larger radial velocity signal in their parent star ( @xmath23 ) , so a given set of radial velocity measurements can detect smaller eccentricities ( see also * ? ? ?
the eccentricity of the wasp-18 system is in fact below the detection limit of existing observations of most teps .
we therefore advocate the acquisition of additional velocity data for the known low - mass teps , in order to equalise the eccentricity detection limits between the two classes of teps .
these observations would allow acceptance or rejection of the hypothesis that more massive teps represent a fundamentally different planet population to their lower - mass brethren .
the observations presented in this work will be made available at the cds ( http://cdsweb.u-strasbg.fr/ ) and at http://www.astro.keele.ac.uk/@xmath24jkt/. the operation of the danish 1.54 m telescope was financed by the danish natural science research council ( fnu ) .
we thank dr .
j. eldridge for calculating the _ cambridge _ set of stellar models used in this work .
jsouthworth and dra acknowledge financial support from stfc in the form of postdoctoral research assistant positions .
astronomical research at the armagh observatory is funded by the northern ireland department of culture , arts and leisure ( dcal ) .
dr ( boursier fria ) , ff and jsurdej acknowledge support from the communaut franaise de belgique actions de recherche concertes
acadmie wallonie - europe .
the following internet - based resources were used in research for this paper : the eso digitized sky survey ; the nasa astrophysics data system ; the simbad database operated at cds , strasbourg , france ; and the ar@xmath25iv scientific paper preprint service operated by cornell university . | we present high - precision photometry of five consecutive transits of wasp-18 , an extrasolar planetary system with one of the shortest orbital periods known . through the use of telescope defocussing we achieve a photometric precision of 0.470.83 mmag per observation over complete transit events .
the data are analysed using the jktebop code and three different sets of stellar evolutionary models .
we find the mass and radius of the planet to be @xmath0@xmath1 and @xmath2@xmath3 ( statistical and systematic errors ) respectively .
the systematic errors in the orbital separation and the stellar and planetary masses , arising from the use of theoretical predictions , are of a similar size to the statistical errors and set a limit on our understanding of the wasp-18 system .
we point out that seven of the nine known massive transiting planets ( @xmath4@xmath1 ) have eccentric orbits , whereas significant orbital eccentricity has been detected for only four of the 46 less massive planets .
this may indicate that there are two different populations of transiting planets , but could also be explained by observational biases .
further radial velocity observations of low - mass planets will make it possible to choose between these two scenarios . |
weyl semimetal is a new topological state of three - dimensional ( 3d ) quantum matters @xcite , different from the 3d topological insulators ( ti ) @xcite .
it can be characterized by weyl nodes ( at fermi level ) in the bulk and fermi arcs on the surfaces @xcite . around the weyl nodes ,
the low energy physics is given as 3d two - component weyl fermions @xcite , @xmath2 ( where @xmath3 is pauli matrix and @xmath4 is crystal moment ) , which carries chirality , left- or right - handed defined by the sign of velocity @xmath5 .
weyl nodes are stable topological objects as long as @xmath4 is well defined , and can be viewed as effective magnetic monopoles in the 3d momentum space @xcite . to get weyl semimetal , either time - reversal ( tr ) or inversion symmetry needs to be broken @xcite .
otherwise , there will be double degeneracy for all @xmath4 . in the case with both tr and inversion symmetries ,
however , we may expect a 3d dirac semimetal state described as four - component dirac fermions , @xmath6 , which can be viewed as two copies of distinct weyl fermions .
unfortunately , this expectation is generally not true , because two weyl nodes with opposite chirality may annihilate each other if they overlap in momentum space , and open up a gap in general .
therefore , additional symmetry is required to protect the 3d dirac semimetal @xcite state and to prohibit the possible mass term , unless it is at the phase boundary between ti and normal insulators @xcite , a subtle situation hard to be controlled .
the symmetry protected 3d dirac semimetal has been discussed @xcite for systems with spin - orbit coupling ( soc ) , focusing on special @xmath4 points with four - dimensional - irreducible - representation ( fdir ) @xcite , which usually appears at the brillouin zone ( bz ) boundary with non - symmorphic double space groups . in general , this fdir requirement is too strong , and we may expect much wider compound choices by considering two doubly - degenerate bands with distinct 2d representations and unavoidable band crossing ( protected by crystalline symmetry ) . in such case , we may get 3d dirac points along the high - symmetry lines rather than high symmetry points at the bz boundary .
this scenario of dirac semimetal has been suggested in our earlier studies on na@xmath0bi @xcite , which is unfortunately not stable in air . in this paper , we show that a well known compound cd@xmath0as@xmath1 is a symmetry - protected 3d dirac semimetal with a single pair of dirac points in the bulk and non - trivial fermi arcs on the surface .
it can be driven into a topological insulator , a weyl semi - metal , or a quantum spin hall ( qsh ) insulator with gap more than 100mev .
it can also support sizable linear quantum magnetoresistance ( mr ) even up to room temperature .
the nice aspect of cd@xmath0as@xmath1 is the high carrier mobility up to 1.5 m@xmath7v@xmath8s@xmath8 at room temperature and 8.0 m@xmath7v@xmath8s@xmath8 at 4 k , reported about 50 years ago @xcite .
this makes it a promising candidate for future transport studies .
we will start from the structure and methods in sec .
[ methodology ] , present the main results in sec .
[ result ] , and finally conclude in sec . [ conclusion ] .
among the ii@xmath0-v@xmath1-types narrow gap semiconductors , cd@xmath0as@xmath1 has drawn crucial attention , because it was believed to have inverted band structure , @xcite whereas all others cd@xmath0p@xmath1 , zn@xmath0as@xmath1 and zn@xmath0p@xmath1 have normal band ordering . in contrast to other inverted band compounds ( like hgte , hgse , and @xmath9-sn ) , cd@xmath0as@xmath1 belongs to tetragonal symmetry , and is the representative of this group , which has the splitted valence band top at @xmath4=0 .
the crystal structure of cd@xmath0as@xmath1 is complicated , and can be related to tetragonally - distorted anti - fluorite structure with 1/4 cd site vacancy . if the distribution of these vacancies is random , one may treat it by virtual crystal approximation ( vca ) for simplicity @xcite .
however , those vacancies are in fact ordered even at room temperature , leading to a tetragonal structure with @xmath10 ( @xmath11 ) symmetry ( 40 atoms per unit cell , called structure i hereafter ) , or a body centered tetragonal structure with @xmath12 ( @xmath13 ) symmetry ( 80 atoms per unit cell , called structure ii hereafter ) , with the later structure more favored @xcite .
this vacancy ordering and very large cell of cd@xmath0as@xmath1 cause historically serious problems for theoretical studies , and there is no existing first - principles calculations up to now .
we report here the first band structure calculations of cd@xmath0as@xmath1 with its true structures and with soc included .
we perform the first - principles band - structure calculations within the density functional formalism as implemented in vasp @xcite , and use the all - electron projector augmented wave ( paw ) @xcite basis sets with the generalized gradient approximation ( gga ) of perdew , burke and ernzerhof ( pbe ) @xcite for the exchange correlation potential .
the hamiltonian contains the scalar relativistic corrections , and the spin - orbit coupling is taken into account by the second variation method @xcite .
the cutoff energy for the plane wave expansion was 500 ev and a * k*-point mesh of @xmath14 and @xmath15 are used for the bulk calculations of structure i and ii , respectively . for the convenience of our later discussions for the effective low energy physics
, here we briefly introduce our modified second - order 8-band kane model @xcite for typical semiconductors .
we start from the standard 4-band second - order kane model @xcite for the case of without soc , and then introduce additional terms to take into account the particular tetragonal symmetry of cd@xmath0as@xmath1 . in the @xmath16 approximation , considering the low energy @xmath17 , @xmath18 , @xmath19 , @xmath20 states ( as basis ) around @xmath21 , the modified 4-band kane model is given as , @xmath22 comparing with the strandard isotropic 4-band kane model @xcite , here we consider the anisotropic tetragonal symmetry , and introduce the paramter @xmath23 for the crystal - field splitting of @xmath24 orbitals .
the other parameter @xmath25 is introduced to describe the breaking of inversion symmetry for structure ii , and it should be zero for structure i. then our modified 8-band kane model can be obtained by adding the soc term as , @xmath26 where * i * is the @xmath27 identity matrix and @xmath28 denotes the strength of soc .
( color online ) the calculated band structures for cd@xmath0as@xmath1 with crystal structure i ( a ) and ii ( b ) , as well as the definition of high symmetric points in each brillouin zone ( bz ) .
the projected surface bz for structure ii is also shown .
the representation of selected bands at @xmath21 and along high symmetric @xmath4 path are indicated .
solid dots in band structures indicate the projected @xmath29 bands . ]
similar to most of the semiconductors with anti - fluorite or zinc - blende structures , the low energy electronic properties of cd@xmath0as@xmath1 are mostly determined by the cd-@xmath30 states ( conduction bands ) and the as-@xmath31 states ( valence bands ) , as shown in fig .
[ ldaband ] . however , there are two distinct features : ( 1 ) band - inversion around @xmath21 point with the @xmath29-state ( red solid cycle ) lower than the @xmath32-states , which is an important sign of non - trivial topology ; ( 2 ) semi - metallic with band crossing along the @xmath21-z direction
. this band - crossing is unavoidable , because the two bands belong to different ( @xmath33 and @xmath34 ) representations respectively , as distinguished by @xmath35 rotational symmetry around @xmath36 axis .
the different representation prohibits hybridization between them , resulting in the protected band - crossing .
furthermore , the crossing points should locate exactly at the fermi level due to charge neutrality requirement , resulting in a 3d dirac semimetal state with fermi surface consisting of a single pair of fermi points ( two symmetric points along @xmath36 related by tr ) .
both the structure i and ii share the above common features but with one important difference : structure i has inversion center but structure ii does nt
. as had been attempted with perturbation method @xcite , vacancy ordering and bz folding play important role for the band - inversion , in contrast to the cases of hgte or ag@xmath1te
@xcite , where it was driven by the the shallow @xmath25 states . to prove this , we performed calculations for cd@xmath0as@xmath1 in hypothetic anti - fluorite structure without vacancy ( using vca ) and in cubic anti - fluorite structure with one cd vacancy , keeping the same lattice constant .
we found that the former ( later ) has normal ( inverted ) band ordering at @xmath21 . at the bz boundary x point of hypothetic anti - fluorite structure without vacancy , there exist shallow @xmath29 and @xmath32 states , which are folded onto the @xmath21 point of real structure .
therefore , the hybridization among the states with the same representation will push the states away from each other , i.e , make the lowest @xmath29-state further lower and highest @xmath32-state higher , resulting in the band inversion at @xmath21 .
the band - inversion calculated from gga is about 0.7 ev for both structure i and ii . considering the possible underestimation of gga for the @xmath29-@xmath32 gap
, we have improved the calculations by hse method @xcite , and still found the band - inversion around 0.3 ev , being consistent with most of the existing experimental evidence , such as the optical and transport measurements @xcite .
the calculated band structures of cd@xmath0as@xmath1 can be well fitted by using our modified 8-band kane model presented in previous section , and the obtained parameters are listed in table.i .
.[paramts]the fitted parameters for the 8-band model .
[ cols="^,^,^,^,^",options="header " , ] [ tab : chern111 ] the atomic cd-@xmath30 and as-@xmath31 states with soc can be written as the states with definite angular momentum @xmath37 and @xmath38 , i. e. , @xmath39 , @xmath40 , @xmath41 , @xmath42 . in the tetragonal crystal symmetry , however , the total angular momentum @xmath37 is no longer a good quantum number , and the valence @xmath32-states have complete splitting at @xmath21 .
the heavy - hole @xmath32-state @xmath43 and the conduction @xmath29-state @xmath44 remain to be the eigenstates at @xmath21 , while the light - hole states @xmath41 can mix with the split - off state @xmath42 , forming two new eigenstates which are irrelevant to the low energy physics here .
the band inversion nature of cd@xmath0as@xmath1 around the @xmath21 can be caught by considering only the minimal basis set of @xmath45 , @xmath46 , @xmath47 and @xmath48 states . to describe the 3d dirac fermion , an effective low energy hamiltonian @xmath49
can therefore be obtained by downfolding the 8-band model into the subspace spanned by the 4 mimimal basis .
the resulting @xmath50 reads , @xmath51 where @xmath52 , @xmath53 , and @xmath54 with parameters @xmath55 , @xmath56 , @xmath57 @xmath580 to reproduce band inversion . under the tetragonal symmetry
, the leading - order term of @xmath59 has to take the high - order form of @xmath60 , which can be neglected if we only consider the expansion up to @xmath61 .
the terms containing @xmath62 describe the breaking of inversion symmetry , which should be zero for structure i. in such case , the energy dispersion is @xmath63 , having a pair of four - fold degenerate zero energy dirac points at @xmath64=(0 , 0 , @xmath65 ) .
around the neighborhood of each dirac point , we can further expand the hamiltonian up to @xmath61 .
the resulting hamiltonian is nothing but the one for 3d massless dirac fermions , which has anisotropic linear dispersion @xmath66 ( where @xmath67 is the deviation of momentum @xmath4 from the dirac point @xmath68 ) .
the block diagonal form allows us to decouple the 4@xmath694 matrix into two 2@xmath692 matrices , which are weyl fermions with degenerate energy but opposite chirality . for structure
ii , which has no inversion symmetry , the non - vanishing d term will modify the in - plane velocity from @xmath70 to @xmath71 , while keep the weyl nodes degenerate ( because the @xmath72 symmetry along the gamma - z axis contains only 2d irreducible representations for its double space group ) .
the resulting dirac semimetal state has 4-fold degenerate dirac points , but with splitting of in - plane band dispersions away from dirac points ( fig .
[ diracband ] ) .
this is a new type of 3d dirac semimetal state , in contrast to other examples @xcite .
( color online ) band dispersions and band - splitting in the plane passing through dirac point ( 0,0,@xmath73 ) and perpendicular to @xmath21-z for structure ii .
the k - points are indicated in cartesian coordinates . @xmath74 and @xmath75 are around 0.1 and 0.032 @xmath8 , respectively . ]
a stable weyl semimetal state with two weyl nodes separated in momentum space can be introduced either by lowering the crystal symmetry from @xmath72 to @xmath35 , or by breaking the tr symmetry @xcite .
in particular , due to the large @xmath76-factor ( around 30@xmath7740 ) of cd@xmath0as@xmath1 @xcite , an exchange - splitting of @xmath77 2mev can be introduced by moderate 1 t magnetic field ( if we neglect the orbital effects ) . on the other hand ,
if the 4-fold rotational symmetry is broken , a linear leading order term of @xmath78 , will be introduced in the effective hamiltonian @xmath79 . in such case
, two weyl nodes will be coupled together , resulting in a massive dirac fermions with gap opening .
we have checked the topological invariant @xmath80 number for this resulting insulating state , and found it is odd , confirming that it is a 3d topological insulator @xcite ( due to the inverted band structure ) .
the non - trivial topology and the 3d dirac cones in cd@xmath0as@xmath1 suggest the presence of non - trivial surface states . for this purpose
, we transform the eight - band model into a tight - binding model on a tetragonal lattice by introducing the substitutions : @xmath81 here , @xmath82 refers to @xmath83 , @xmath84 and @xmath36 , while @xmath85 and @xmath86 are the tetragonal lattice constants , which are taken as 3.0 and 5.0 , respectively .
this approximation is valid in the vicinity of the @xmath21 point .
we use an iterative method to obtain the surface green s function of the semi - infinite system @xcite .
the imaginary part of the surface green s function is the local density of states ( ldos ) at the surface . since gga underestimate the @xmath29-@xmath32 band gap by about 0.4 ev as discussed above , we have artificially lifted the on - site energy of @xmath29-state ( as given in table .
[ paramts ] ) by 0.4 ev in the surface state calculations .
the obtained ldos on semi - infinite ( 001 ) and ( 110 ) surfaces of the structure ii are presented in fig .
[ surfband ] . for ( 001 )
surface , the surface projection of continuous bulk states superposes the non - trivial surface states , and its fermi surface is just a point ( fig .
[ surfband](b ) ) . for ( 110 ) surface , however , the non - trivial surface states are clearly visible .
its fermi surface is composed of two half - circle fermi arcs , touching at the singularity points ( @xmath87 , @xmath88 ) where the surface projection of bulk dirac points exist .
( color online ) the calculated surface states ( left panels ) and corresponding fermi surface ( right panels ) of structure ii cd@xmath0as@xmath1 for its ( 001 ) ( upper panels ) and ( 110 ) ( lower panels ) surface . ]
the existence of 3d dirac cones in the bulk also implies that we can expect the qsh effect if we reduce the dimensionality and form the quantum well structure of cd@xmath0as@xmath1 .
this provides an alternative compound choice to the existing experiments @xcite , which all require extreme conditions up to now .
the band structure of @xmath89-oriented cd@xmath0as@xmath1 thin film of different thickness can be understood as the bulk bands in different @xmath90-fixed planes .
the 2d z@xmath1 number can be nonzero only for limited regions where the band inversion happens when going from @xmath21 to z. in the quantum - well structure , those low energy states around @xmath21 should be further quantized into subbands , whose energy levels change as a function of film thickness ( fig .
[ wellband](a ) ) . when the thickness of the film is thin enough
, the band inversion in the bulk band structure will be removed entirely by the finite size effect . with the increment of the film thickness
, finite size effect is getting weaker and the band inversion among these subbands restores subsequently , which leads to jumps in the z@xmath1 number .
then , depending on the number of band inversions associated with the sub - bands , the system should cross over between trivial and nontrivial 2d insulators oscillatorily as a function of thickness @xcite ( fig . [ wellband](b ) ) .
comparing with hgte / cdte quantum well @xcite , the first critical thickness of cd@xmath0as@xmath1 is much thinner ( 2.9 nm vs. 6.3 nm ) , suggesting possibly larger gap . to be a concrete example
, 5 nm thick cd@xmath0as@xmath1 film is a good qsh insulator with gap more than 100mev .
given the known high mobility of cd@xmath0as@xmath1 , it is therefore a good candidate for the qsh measurement .
( color online ) the thickness dependence of ( a ) sub - band energies and ( b ) z@xmath1 number of @xmath89-oriented quantum well of cd@xmath0as@xmath1 . ]
finally we propose that crystalline cd@xmath0as@xmath1 with 3d dirac cone is an ideal system to check abrikosov s proposal of quantum mr @xcite , and may support sizable linear mr even up to room temperature .
quantum effects become noticeable when the individual landau levels are distinct : @xmath91@xmath92@xmath93 .
dirac system has linear energy spectrum , and its cyclotron frequency @xmath94 should follow the square root rule ( @xmath94=@xmath95 ) , in contrast with non - relativistic system where @xmath94 is linear in field @xmath96 . using calculated velocity @xmath5=@xmath97cm / s for cd@xmath0as@xmath1
, this then leads to @xmath98280k for @xmath96=10 t .
the linear quantum mr can be estimated @xcite as @xmath99=@xmath100=@xmath101 ( where @xmath102=1/@xmath103 is used , @xmath104 is carrier density , @xmath105 is mobility , and @xmath106 is the density of scattering centers ) . taking the experimental values
@xmath1071.5m@xmath7v@xmath8s@xmath8 at room temperature , and assuming @xmath104 and @xmath106 are in the same order , mr ratio can reach 50% per 1 t field . if the scattering is not phonon - mediated , we expect that @xmath106 is not sensitive to temperature , then mr is mostly determined by @xmath104 and @xmath105 , which should lead to even enhanced mr at lower temperature .
in summary , based on the first - principles calculations and effective model analysis , we have shown that the known compound cd@xmath0as@xmath1 is a symmetry - protected topological semimetal with a pair of 3d dirac points ( located at the fermi level ) in the bulk and fermi arcs on the surfaces . it can be driven into various topologically distinct phases , such as topological insulator and weyl semimetal state by symmetry breakings .
in addition , due to its unique 3d dirac cone type electronic structure , we can expect the qsh effect in its quantum well structure and the sizable linear quantum mr even up to the room temperature
. it will be of particular interest , in such a 3d topological semimetal , to see whether the superconductivity can be achived by carrier doping or not , because such superconducting state if obtained may be related to the topological superconductivity @xcite .
the first - principles calculations are performed by using the computers in the nation supercomputer center in tianjin of china .
we acknowledge the supports from nsf of china , the 973 program of china ( no .
2011cba00108 and 2013cbp21700 ) .
see , w. zdanowicz , and l. zdanowicz , ann .
sci . * 5 * , 301 ( 1975 ) , and references therein .
m. j. aubin , l. g. caron , and j .-
jay - gerin , phys .
b * 15 * , 3872 ( 1977 ) ; l. g. caron , j .-
jay - gerin , and m. j. aubin , phys . rev .
b * 15 * , 3879 ( 1977 ) .
b. dowgiallo - plenkiewicz , p. plenkiewicz , phys .
( b ) * 94 * , k57 ( 1979 ) ; g. chuiko , n. don , et al , moldavian journal of the physical sciences * 2 * n1.88 ( 2003 ) .
g. kresse and j. hafner , phys .
b * 48 * , 13115 ( 1993 ) ; g. kresse and j. furthmller , comput .
* 6 * , 15 ( 1993 ) p. e. blchl , phys .
b @xmath108 , 17953 ( 1994 ) ; g. kresse and d. joubert , phys .
b @xmath109 , 1758 ( 1999 ) j. p.
perdew , k. burke and m. ernzerhof , phys .
@xmath110 , 3865 ( 1996 ) d. d. koelling and b. n. harmon , j. phys .
c : solid state phys .
@xmath111 , 3107 ( 1977 ) lok c. lew yan voon and m. willatzen , _ the k@xmath112p method : electronic properties of semiconductors _
( springer , berlin , 2009 ) . see pp .
59 - 60 for the definition of parameters and four - bands kane model .
m. knig , s. wiedmann , c. brune , a. roth , h. buhmann , l. w. molenkamp , x .-
qi , s .- c .
zhang , science * 318 * , 766 ( 2007 ) . c. x. liu , t. l. hughes , x. l. qi , k. wang , s .- c .
zhang , phys .
lett . * 100 * , 236601 ( 2008 ) ; i. knez , r. r. du , and g. sullivan , phys .
lett . * 107 * , 136603 ( 2011 ) . c. x. liu , h. j. zhang , b. h. yan , x. l. qi , t. frauenheim , x. dai , z. fang , and s. c. zhang , phys . rev .
b * 81 * , 041307 ( 2010 ) . | based on the first - principles calculations , we recover the silent topological nature of cd@xmath0as@xmath1 , a well known semiconductor with high carrier mobility . we find that it is a symmetry - protected topological semimetal with a single pair of three - dimensional ( 3d ) dirac points in the bulk and non - trivial fermi arcs on the surfaces .
it can be driven into a topological insulator and a weyl semi - metal state by symmetry breaking , or into a quantum spin hall insulator with gap more than 100mev by reducing dimensionality .
we propose that the 3d dirac cones in the bulk of cd@xmath0as@xmath1 can support sizable linear quantum magnetoresistance even up to room temperature . |
the hyperbolic dehn filling theory was established by w. thurston in his original note @xcite .
besides the general theory , he extensively analyzed the space of dehn fillings of the figure eight knot complement in terms of the decomposition by two ideal tetrahedra .
the analysis leaded us to extremely fascinated research activities on dehn fillings in the last 25 years .
one of unsolved problems in his analysis was to determine the global shape of the dehn filling space .
since then , there have been many deep researches such as @xcite discussing this problem in fact , however it is still mysterious .
even , though w. neumann and a. reid @xcite have determined the true shape for the whitehead link complement with one cusp component being complete , there seems to be no concrete examples of `` one - cusped '' hyperbolic manifolds or orbifolds for which we see the true shape of the dehn filling space . in this paper , we construct hopefully the first such example of an orbifold , regretfully rather than a manifold .
the example we present here is topologically @xmath0-covered as an orbifold by the complement of a seven component link in the connected sum of three copies of @xmath1 . based on the study of one - circle packings on complex affine tori by the second author in @xcite
, we construct the example , which we will denote by @xmath2 , and its all possible dehn filling deformations in the following three sections . then using careful analysis of deformations of such tori in @xcite again
, we determine the global shape of the dehn filling space of @xmath2 in subsequent two sections .
we start with the _ hexagonal packing _
@xmath3 on the complex plane @xmath4 by equi - radii circles , see figure [ fig : hexagonalpacking ] .
the set of euclidean translations which leave @xmath3 invariant forms a group @xmath5 acting freely on @xmath4 .
it is isomorphic to @xmath6 and the quotient is a particular euclidean torus admitting a cyclic symmetry of order @xmath7 , which we call the _
hexagonal torus_. @xmath3 then descends to what we call a one - circle packing on the hexagonal torus .
( 100,100 ) ( 0,0 ) let us regard the complex plane @xmath4 as the boundary of the upper half space model @xmath8 of the hyperbolic 3-space .
@xmath3 together with the dual packing @xmath9 which consists of circumscribed circles of interstices in @xmath3 fills @xmath4 .
each member of both @xmath3 and @xmath9 bounds a hemisphere in @xmath8 and such a hemisphere bounds a hemiball facing @xmath4 . cutting off those hemiballs from @xmath8
, we obtain a region @xmath10 with ideal polygonal boundary in @xmath8 .
the intersections of @xmath3 and @xmath9 correspond to ideal vertices of @xmath10 with rectangular section .
the group @xmath5 acts properly discontinuously on @xmath8 by the poincar extensions and the region @xmath10 is invariant under the action . hence taking quotient , we obtain a hyperbolic 3-manifold @xmath2 with ideal polygonal boundary and a cusp .
the polygonal boundary consists of two ideal triangles and one ideal hexagon , where they intersect orthogonally .
take the double of @xmath2 along triangular faces , and then take again the double of the result along the boundary which now consists of two ideal hexagons .
this double doubling construction gives us a hyperbolic 3-manifold @xmath11 with seven cusps among which 4 are old and 3 new .
it is obvious by the construction that @xmath11 admits a @xmath12-symmetry generated by two reflections associated to the doubling .
@xmath2 is the quotient of @xmath11 by this symmetry and therefore we regard @xmath2 as an orbifold from now on , and study its dehn filling deformations as an orbifold .
topologically @xmath2 is a solid torus with a core loop and three points on the boundary removed . to see the effect on the first double , stretch the removed vertices on @xmath13 to arcs appropriately so that they become circles after doubling .
then by first doubling , we obtain a genus 3 handle body with two core loops and three loops on the boundary removed .
this is depicted in figure [ fig : firstdouble ] , where the shaded disks represent triangular faces along which we took the double .
( 141,123 ) ( 0,0 ) a framed link representation of @xmath11 , the double of figure [ fig : firstdouble ] , can be immediately described by figure [ fig : framedlink ] .
( 145,110 ) ( 0,0 ) ( 45,72)@xmath14 ( 81,72)@xmath14 ( 117,72)@xmath14
we will construct dehn filling deformations of @xmath2 using deformations of the hexagonal packing on @xmath4 with the same combinatorial type .
the hexagonal packing descends to a one - circle packing on the hexagonal torus .
conversely , a one - circle packing on the torus combinatorially equivalent to one on the hexagonal torus is universal covered by a packing on the plane equivalent to the hexagonal packing .
however since the hexagonal packing is rigid in euclidean geometry , there are no other one - circle packings on euclidean tori than one on the hexagonal torus . hence to get deformations , we will work with one - circle packings not on euclidean but rather on complex affine tori .
this section is to review what has been known about complex affine tori and circle packings on them according to @xcite .
a _ complex affine structure _ on the torus is by definition a collection of local charts on @xmath4 whose transition function is a restriction of a complex affine transformation , so in particular it defines a complex structure .
the uniformization assigning the conformal class to each affine torus defines a map of the space @xmath15 of marked complex affine tori to the teichmller space @xmath16 of the torus .
it is a natural complex line bundle projection , where the fiber is parameterized by explicit description of developing maps of affine structures under some normalization as we will see below .
@xmath16 is identified with the upper half plane @xmath17 in @xmath4 and hence @xmath15 is homeomorphic to the product @xmath18 .
we call the first factor a _ teichmller parameter _ @xmath19 and the second an _ affine parameter _ @xmath20 .
complex affine tori sharing the same @xmath21 are conformally equivalent .
we denote by @xmath22 the marked complex affine torus associated to @xmath23 .
also we denote by @xmath24 specified simple closed curves generating the fundamental group of the torus for marking .
the zero section of the line bundle @xmath25 precisely consists of euclidean structures . in this case
, the developing map @xmath26 is a homeomorphism , and the images of @xmath27 and @xmath28 by the holonomy @xmath29 are translations defined for each @xmath30 by @xmath31 after appropriate normalization .
( 250,120 ) ( 0,20 ) ( 13,0)images of @xmath27 and @xmath28 ( 150,20 ) ( 145,0)image of a circle packing in affine but non - euclidean case , identifying the universal cover @xmath32 with @xmath33 by the uniformization , the developing map is described by an exponential map @xmath34 under the identification of @xmath15 with @xmath18 . here
@xmath21 and @xmath35 are the teichmller and affine parameters of the complex affine torus in question .
the images of @xmath36 by the holonomy @xmath29 are obviously similarities @xmath37 which fix the origin .
( 250,120 ) ( 0,20 ) ( 13,0)images of @xmath27 and @xmath28 ( 150,20 ) ( 145,0)image of a circle packing since a complex affine transformation is a similarity , it sends a circle on @xmath4 to a circle .
hence we can still discuss about circle packings on complex affine tori .
the second author showed in @xcite that the set of complex affine tori which admit a one - circle packing forms a real two - dimensional moduli space @xmath38 in @xmath18 . in constructing the moduli space , the marking for the packing
was specified so that the edges of the nerve of the packing are homotopic to @xmath27 , @xmath28 and @xmath39 .
the moduli space @xmath38 contains only one euclidean torus associated to @xmath40 , which is the hexagonal one .
moreover , the restriction of the uniformization @xmath25 to @xmath38 is onto and two to one except at the hexagonal torus . the affine parameters of a generic fiber consist of nonzero @xmath35 and its negative @xmath41 .
to each pair @xmath42 on the moduli space @xmath43 , we construct a dehn filling deformation of @xmath2 as an orbifold based on the construction of @xmath2 . to see this , we start with a complex affine torus @xmath44 packed by one circle combinatorially equivalent to the one - circle packing on @xmath45 .
the developing map of @xmath44 is an exponential map and @xmath46-covers @xmath47 .
thus if we let @xmath48 be the universal cover of @xmath49 , it canonically lifts to a homeomorphism @xmath50 . by mapping a circle packing on @xmath32 to @xmath48
, we obtain a circle packing @xmath51 on @xmath48 combinatorially equivalent to the hexagonal packing on @xmath4 .
also the holonomy lifts uniquely to an injective homomorphism of @xmath52 to the group of transformations of @xmath48 .
we denote its image by @xmath53 .
then we let @xmath54 denote the geodesic in @xmath8 joining @xmath14 and @xmath55 .
the universal cover @xmath56 of @xmath57 will be a substitution of @xmath8 in the euclidean case .
in fact , cutting off hemiballs in @xmath56 bounded by the members of a deformed hexagonal packing @xmath51 and its dual @xmath58 on @xmath48 , we obtain a region @xmath59 with ideal polygonal boundary in @xmath56 .
also since the action of @xmath53 on @xmath48 extends to a properly discontinuous action on @xmath56 which leaves @xmath59 invariant , by taking a quotient we obtain a hyperbolic 3-manifold with ideal polygonal boundary and incomplete end near @xmath54 .
we denote its completion by @xmath60 .
it carries a dehn surgery type singularity by definition , see @xcite , and we thus obtain to each @xmath42 in @xmath61 a dehn filling deformation of @xmath62 .
on the other hand , let @xmath63 be a dehn filled deformation of @xmath2 as an orbifold and @xmath64 a manifold obtained from @xmath63 by removing the points attached by the completion .
@xmath64 is homeomorphic to the torus times @xmath65 with three points on the boundary removed .
the boundary of @xmath64 consists of two ideal triangles and one ideal hexagon , where they intersect orthogonally .
choose a developing map @xmath66 of @xmath64 so that a neighborhood of an incomplete end winds @xmath54 in @xmath8 . by this normalization , the holonomy of @xmath64 maps @xmath67 to transformations fixing @xmath14 and @xmath55 , the image of the holonomy contains an element of infinite order by the definition of the dehn filling deformation .
@xmath64 is not complete , but not so wild .
let @xmath68 be an equidistant neighborhood of an incomplete end and @xmath69 its complement .
it is obvious that any point in @xmath64 admits the unique shortest path to @xmath70 .
thus @xmath68 is developed to an equidistant neighborhood of @xmath54 and @xmath69 its complement .
in particular , the image of the developing map @xmath66 misses @xmath54 , and @xmath66 is in fact a map of the universal cover @xmath71 into @xmath57 .
then @xmath66 lifts to a map @xmath72 of the universal cover @xmath71 to @xmath56 , the universal cover of @xmath57 .
accordingly , the holonomy of @xmath64 lifts to a homomorphism to the group of transformations of @xmath56 .
we denote its image by @xmath73 .
let @xmath74 be the preimage of the boundary of the equidistant neighborhood of @xmath54 in @xmath56 .
@xmath73 acts properly discontinuously and freely on @xmath74 by the definition of the dehn filling deformation , and @xmath74 bijectively corresponds to the preimage of @xmath74 by @xmath72 .
this shows that @xmath72 is injective at least on a neighborhood of the preimage of @xmath74 by @xmath72 .
we will show that @xmath72 is globally injective .
suppose we have two points in @xmath71 mapped by @xmath72 to a point @xmath75 .
@xmath76 admits the unique shortest path to the point @xmath77 on @xmath74 .
since the preimage of @xmath74 by @xmath72 bijectively corresponds to the preimage of @xmath70 in @xmath71 , the preimage of @xmath77 consists of a single point @xmath78 .
hence two points in question lie on the orthogonal path to the preimage of @xmath70 through @xmath78 on the same side with the same distance to @xmath78 .
thus they must be the same .
then the ideal hexagonal faces on the boundary of @xmath71 in the developed image define a packing on @xmath48 , the universal cover of @xmath47 , combinatorially equivalent to the hexagonal one .
moreover the packing is invariant under the induced action of @xmath73 on @xmath48 .
hence it descends to a one - circle packing on some affine torus @xmath22 parameterized by some pair @xmath79 , and @xmath63 is isometric to @xmath60 .
following @xcite , we review how to find a pair @xmath42 lying on the moduli space @xmath61 .
if @xmath42 lies on @xmath61 , a simple observation shows that @xmath21 and @xmath35 must satisfy the identities , @xmath80 moreover , the inequalities @xmath81 must be satisfied by obvious reason .
let @xmath82 be a smooth function defined by @xmath83 where @xmath84 is the largest possible region for @xmath35 to vary in @xmath38 .
the identity ( [ eq : conditionforadmittingpacking ] ) shows that @xmath35 , @xmath85 and @xmath86 lie on the same level set @xmath87 .
when @xmath88 , @xmath89 consists of a single point corresponding to the hexagonal torus by simple computation .
for each @xmath90 , @xmath91 is topologically a simple loop around the origin , has symmetries about the real and imaginary axis , and bounds a strictly convex interior . for @xmath92 ,
the origin @xmath14 , @xmath35 , @xmath85 and @xmath86 form a parallelogram by obvious reason in plane geometry .
conversely , if @xmath93 is given and let @xmath91 be the level set on which @xmath35 lies , then there are unique two points @xmath94 on @xmath91 so that @xmath95 lie in counterclockwise order on @xmath91 , and they together with the origin form a parallelogram .
then we obtain a pair @xmath42 in the moduli space @xmath61 by letting @xmath96 .
hence @xmath97 determines @xmath21 . in view of this fact
, we identify the moduli space @xmath61 with the region @xmath98 .
in the level set @xmath91 for @xmath90 , there are twelve special points depicted in figure [ fig : twelvepointsonlevelset ] which give rise to pairs on @xmath61 rather obviously . here , @xmath99 and @xmath100 are intersections with the real axis .
similarly @xmath101 , @xmath102 lie on the imaginary axis .
@xmath103 for @xmath104 are points such that @xmath105 , @xmath106 , @xmath107 and @xmath108 are the bisectors of the segments @xmath109 , @xmath110 , @xmath111 and @xmath112 where @xmath14 is the origin in @xmath113 .
three points @xmath114 , @xmath115 , @xmath116 together with the origin @xmath14 form a parallelogram and give rise to a pair on @xmath61 .
as @xmath117 decreases from @xmath118 toward @xmath14 , an ellipselike closed curve @xmath91 shrinks down to a point @xmath89 , however the order of @xmath103 s on the level set never changes .
hence the trace of @xmath119 defines a locus @xmath120 in the region @xmath121 , and @xmath121 is divided by these loci into twelve regions denoted by @xmath122 , @xmath123 , @xmath124 as in figure [ fig : twelveregions ] .
( 140,70 ) ( 0,-7 ) ( 0,0 ) , title="fig : " ] ( 123,35)(0,0)[lb]@xmath99 ( 119,53)(0,0)[l]@xmath125 ( 98,63)(0,0)[lb]@xmath126 ( 77,65)(0,0)[b]@xmath101 ( 49,63)(0,0)[rb]@xmath127 ( 25,53)(0,0)[r]@xmath128 ( 20,30)(0,0)[r]@xmath100 ( 25,18)(0,0)[r]@xmath129 ( 49,7)(0,0)[rt]@xmath130 ( 70,5)(0,0)[rt]@xmath102 ( 91,7)(0,0)[lt]@xmath131 ( 119,21)(0,0)[l]@xmath132 ( 69,34)(0,0)[tr]@xmath14 ( 140,70 ) ( 0,-7 ) ( 0,0 ) , title="fig : " ] ( 144,35)(0,0)[l]@xmath133 ( 144,49)(0,0)[l]@xmath134 ( 144,63)(0,0)[l]@xmath135 ( 72,69)(0,0)[l]@xmath136 ( -4,63)(0,0)[r]@xmath137 ( -4,49)(0,0)[r]@xmath138 ( -4,35)(0,0)[r]@xmath139 ( -4,21)(0,0)[r]@xmath140 ( -4,7)(0,0)[r]@xmath141 ( 70,2)(0,0)[r]@xmath142 ( 144,7)(0,0)[l]@xmath143 ( 144,21)(0,0)[l]@xmath144 ( 105,42)(0,0)[]@xmath122 ( 98,56)(0,0)[t]@xmath145 ( 77,60)(0,0)[t]@xmath146 ( 63,60)(0,0)[t]@xmath147 ( 42,56)(0,0)[t]@xmath148 ( 35,42)(0,0)[]@xmath149 ( 35,28)(0,0)[]@xmath150 ( 42,14)(0,0)[b]@xmath151 ( 63,11)(0,0)[b]@xmath152 ( 79,11)(0,0)[b]@xmath153 ( 98,14)(0,0)[b]@xmath154 ( 105,28)(0,0)[]@xmath124 to each pair @xmath42 on the moduli space @xmath61 , we choose four points @xmath155 which form a parallelogram . choosing the origin and the third and fourth vertices @xmath156 , and adding the negative of the second one at the end , we obtain a next parallelogram spanned by @xmath157 .
doing the same shift twice , we obtain parallelograms spanned by @xmath158 and @xmath159 .
the last one is symmetric to the original parallelogram about the origin .
this shows that the moduli space @xmath160 has a rotational symmetry of order @xmath7 . on the other hand
, the complex conjugation gives an orientation reversing involutive symmetry .
thus by these all together , the moduli space @xmath161 admits a dihedral symmetry of order @xmath162 .
regarding @xmath36 as a meridian - longitude pair for @xmath2 , we discuss the hyperbolic dehn filling coefficients of @xmath60 .
recall that @xmath163 is a similarity defined by @xmath164 and @xmath165 by @xmath166 for nonzero @xmath35 .
then the dehn filling coefficients @xmath167 for @xmath60 is by definition , see @xcite , a continuous solution of the equation @xmath168 under the normalization that @xmath169 go to the identity when the structure approaches the complete one .
more precisely for our choice of @xmath35 and @xmath21 , the dehn filling coefficient @xmath170 of @xmath60 is a solution of the linear equation @xmath171 note that a pair @xmath42 corresponds to @xmath172 and conversely a pair @xmath173 to @xmath174 .
we define @xmath173 for @xmath175 as @xmath176 . to each finite dehn filling coefficient @xmath170
, we assign a slopelike number @xmath177 , where the slope is @xmath178 by definition .
though @xmath42 corresponds to coordinates @xmath172 , the unique @xmath179 is assigned for them . dividing ( [ eq : coefficient ] ) on both sides by @xmath180 and taking the real parts , we obtain an obvious identity , @xmath181 which shows how to read off @xmath179 with respect to the dehn filling coefficient of @xmath60 by an elementary plane geometry in @xmath4 as follows .
@xmath35 and @xmath85 lie on the same level set of @xmath182 in @xmath98 .
here we draw the straight line through @xmath35 and @xmath85 .
then the intersection of the line with the imaginary axis is the point which divides the segment connecting @xmath35 and @xmath85 in @xmath183 to @xmath179 . for our convenience ,
we define a function @xmath184 by assigning @xmath177 to each @xmath185 .
it is easy to see that the level set of @xmath186 at @xmath187 coincide with the union of @xmath133 and @xmath139 , the union of @xmath134 and @xmath140 , @xmath123 , the union of @xmath138 and @xmath144 . when @xmath35 moves along the level set of @xmath182 in counterclockwise direction , @xmath85 does the same .
hence @xmath186 is strictly increasing along a counterclockwise move on the level set of @xmath182 except a gap at @xmath188 .
( 140,70 ) ( 0,-15 ) ( 0,10 ) ( 60,85)@xmath189 ( 145,72)@xmath190 ( 145,58)@xmath191 ( 145,44)@xmath192 ( 145,30)@xmath193 ( 145,16)@xmath194 ( 60,0)@xmath189 ( -28,72)@xmath194 ( -32,58)@xmath193 ( -34,44)@xmath192 ( -28,30)@xmath191 ( -37,16)@xmath190 choose a point @xmath35 in the region @xmath146 , then @xmath179 assigned lies in @xmath195 . when @xmath35 moves along the level set of @xmath186 toward the boundary of @xmath121 , @xmath85 must move along the level set of @xmath186 in @xmath148 . since @xmath196 goes to @xmath197 and @xmath198
remains zero , @xmath199 must go to @xmath55 .
the limit of the curve @xmath135 is @xmath200 , and this gives @xmath201 .
on the other hand , by ( [ eq : conditionforadmittingpacking ] ) , @xmath202 .
since @xmath203 is bounded and @xmath204 is asymptotically @xmath205 , @xmath206 converges to @xmath14 and the limit of @xmath207 is @xmath208 .
thus , @xmath209 , and we have @xmath210 this shows that @xmath211 and @xmath212 . by the dihedral symmetry of the moduli space
, we can calculate the limit for other @xmath179 .
the image of the dehn filling coefficients @xmath173 parameterized by the moduli space @xmath213 then covers the shaded region @xmath214 in figure [ fig : dehn surgery space ] , and the boundary consists of a hexagon as follows : @xmath215 ( 115,110 ) ( 0,-10 ) ( 0,0 ) ( 2,122)@xmath192 ( 13,120 ) ( 50,115)@xmath191 ( 105,105)@xmath190 ( 105,59)@xmath189 ( 105,25)@xmath194 ( 105,0)@xmath216 ( 105,51)@xmath217 ( 50,105)@xmath218 ( 54,82)@xmath219 ( 80,55)@xmath219 ( 37,18)@xmath220 ( 11,44)@xmath220
we check how the degeneration occurs when they approach the boundary .
we first think of the deformations along the path defined by the level set of @xmath186 at @xmath189
. then the deformations are realized by hyperbolic cone - manifolds and the meridian loop @xmath27 bounds a disk hitting the cone singularity .
see @xcite for details about cone - manifolds .
since @xmath221 the cone angle converges to @xmath208 and the length of the singular locus diverges .
in fact , the dehn filled manifold splits along an euclidean sub cone - manifold , a disk with two corner of angle @xmath222 and a cone point of cone angle @xmath208 , at the limit .
the same result holds when @xmath191 and @xmath193 because of the dihedral symmetry . more generally , choose coprime integers @xmath223 and @xmath224 and think of the deformations of @xmath2 along the path on the level set of @xmath186 at @xmath225
. then again the deformations are realized by cone - manifolds and a loop homotopic to @xmath226 bounds a disk hitting the cone singularity . for @xmath227 , the level set of @xmath186 lies in @xmath228 . when the deformations approach the boundary along this path , @xmath229 , @xmath230 , @xmath231 and @xmath232 .
the calculation similar to one in the previous section shows that @xmath233 and hence the cone angle converges to @xmath234 . to see the length of the singular locus
, we choose integers @xmath235 and @xmath117 such that @xmath236 . then , @xmath237 represents a curve which intersects @xmath238 once and hence the translation factor of @xmath239 , or equivalently @xmath240 represents the length of the singular locus . making use of the property @xmath241 , we have @xmath242 which implies that the length of the singular locus diverges and a corresponding cone - manifold splits as before .
we can see similar facts for other rational @xmath179 by the dihedral symmetry .
in fact , the cone angle approaches @xmath243 if @xmath244 and @xmath245 if @xmath246 . in either case , the length of the singular locus diverges .
if we relax the complex affine structure modeled on @xmath4 to the complex projective structure modeled on the riemann sphere , the relaxation defines the involution @xmath247 on @xmath18 which sends @xmath248 to @xmath249 . the moduli space @xmath250 of complex projective tori admitting a one - circle packing is the quotient of @xmath38 by @xmath247 .
this space is studied in @xcite by the cross ratio parameters and represented as a semi algebraic set by @xmath251 see the appendix in @xcite . assigning the holonomy representation to each projective structure
, we obtain a map @xmath252 of @xmath250 to the space @xmath253 of representations of the fundamental group of the torus in @xmath254 up to conjugation .
it is a regular map with respect to the algebraic structures on the source and target , and easily seen to be a proper injective homeomorphism by concrete computation in @xcite .
this in particular implies also that the boundary we described here is the true one since holonomy representations diverge at the boundary .
99 d. cooper , c. hodgson and s. kerckhoff : _ three - dimensional orbifolds and cone - manifolds _ , msj memoirs , * 5 * ( 2000 ) . j.
dowty : _ orthlength and hyperbolic dehn surgery _
thesis , melbourne univ . ,
r. c. gunning : _ affine and projective structures on riemann surfaces _ , riemann surfaces and related topics : ann . of math .
, * 97 * ( 1981 ) , 225 - 244 . c. hodgson : _ degeneration and regeneration of geometric structures on three - manifolds _ , ph.d .
thesis , princeton univ . ,
1986 . s. kojima , s. mizushima and s. p. tan : _ circle packings on surfaces with projective structures _ , preprint .
available at http://www.arxiv.org/abs/math.gt/0111214 .
s. mizushima : _ circle packings on complex affine tori _ , osaka j.math . , * 37 * ( 2000 ) , 873 - 881 . w. d. neumann and a. reid : _ arithmetic of hyperbolic manifolds _ , topology 90 , ed . by b. apanasov
et al , de gruyter , ( 1992 ) , 273 - 310 .
thurston : the geometry and topology of 3-manifolds , lecture notes , princeton univ .
, 1977/78 . | we construct the first example of a `` one - cusped '' hyperbolic 3-orbifold for which we see the true shape of the space of hyperbolic dehn fillings . |
in the study of both atomic and optical physics problems , often analyzed in the realm of nonlinear schrdinger ( nls ) type equations @xcite , the study of double well potentials has a prominent position .
such potentials can be straightforwardly realized in atomic bose - einstein condensates ( becs ) through the combination of a parabolic ( harmonic ) trap with a periodic potential .
their experimental realization and subsequent study in becs with self - repulsive nonlinearity has led to numerous interesting observations including tunneling and josephson oscillations for small numbers of atoms in the condensate , and macroscopic quantum self - trapped states for large atom number @xcite and symmetry - breaking dynamical instabilities @xcite .
these experimental developments have been accompanied by a larger array of theoretical studies on issues such as finite - mode reductions and symmetry - breaking bifurcations @xcite , quantum effects @xcite , and nonlinear variants of the potentials @xcite .
similar features have also emerged in nonlinear optical settings including the formation of asymmetric states in dual - core fibers @xcite , self - guided laser beams in kerr media @xcite , and optically - induced dual - core waveguiding structures in photorefractive crystals @xcite . on the other hand ,
a theme that has also been progressively becoming of increasing importance within both of these areas of physics is that of long range interactions . in the atomic context ,
the experimental realization of becs of magnetically polarized @xmath0cr atoms @xcite ( see recent review @xcite and for a study of double well effects @xcite ) , as well as the study of dipolar molecules hetmol , and atoms in which electric moments are induced by a strong external field @xcite have been at the center of the effort to appreciate the role of long range effects . on the other hand , in nonlinear optics , where nonlocal effects have been argued to be relevant for some time now @xcite ,
numerous striking predictions and observations have arisen in the setting of thermal nonlocal media @xcite . among them
, we single out the existence of stable vortex rings @xcite the experimental realization of elliptically shaped spatial solitons @xcite and the observation of potentially pairwise attracting ( instead of repelling as in the standard local cubic media ) dark solitons @xcite .
another very important large class of systems displaying a nonlocal nonlinearity consists of materials with a quadratic nonlinearity . in @xcite
, it has been shown that , in fact , the quadratic nonlinearity is inherently nonlocal .
this implies that nonlocality can be used explain the beautiful x - wave @xcite observations and even the different regimes of soliton pulse compression in quadratic materials @xcite .
it is interesting to note that in these quadratic media , not only does the prototypical ingredient of ( effective ) nonlocality arise , but it is also possible for a competition of this feature with the cubic nonlinearity to emerge as is discussed in @xcite . our aim in the present work is to expand on the framework of studies of double well potentials in the presence of nonlocal nonlinear interactions by considering cubic - quintic models .
part of the motivation for doing so consists of the fundamental relevance of the cubic - quintic nls .
the latter is a model that has been used in a variety of physical settings .
these include the light propagation in optical media such as non - kerr crystals @xcite , chalcogenide glasses @xcite , organic materials @xcite , colloids @xcite , dye solutions @xcite , and ferroelectrics @xcite .
it has also been predicted that this type of nonlinearity may be synthesized by means of a cascading mechanism @xcite .
an additional part of the motivation stems from an interesting set of observations that were made in an earlier work featuring competing _
nonlinearities , one of which was a cubic local and another was a cubic nonlocal one ; see @xcite and the discussion therein . in that work , it was found that for repulsive nonlocal cubic interactions and attractive local ones , it was possible to tune the prefactors determining the competition so as to produce not only a symmetry breaking , but also a symmetry - restoring bifurcation .
more recently , a similar conclusion in a local cubic - quintic double well potential was reached in @xcite . here
, we present a framework where the competition of cubic and quintic terms can be systematically quantified .
in addition , to address the problem from a broader perspective , we consider fully nonlocal interactions both for the cubic and the quintic terms , rendering the local case a straightforward special - case scenario of our study .
the specific setup we consider here is partially of interest to the field of cold gases e.g. in the case of repulsive quintic ( but local ) interactions and attractive cubic nonlocal ones .
this exactly corresponds to the model of the dipolar tonks - girardeau gas with the dipole moments polarized along the axis , considered earlier in @xcite .
the difference here is that in this setting the quintic interaction can not be made nonlocal ( although the relevant mathematical norm form description and physical phenomenology will be essentially the same as presented herein ) .
a setup more precisely in tune with the considerations given below arises in the field of nonlinear optics and , more particularly , in the case of thermal optical nonlinearity @xcite but when the heating is provided by the resonant absorption by dopants , in which case the absorption may be saturable . in the appendix
, we justify more precisely this connection to the specific model analyzed in what follows .
we start our presentation of the theoretical analysis of section ii by developing a two - mode reduction of the system with both the cubic and the quintic terms .
we systematically examine all the relevant terms and offer a prescription for assessing the dominant contributions to the resulting dynamics of the left and the right well . following an amplitude - phase decomposition and examining the variables associated with the population imbalance of the two wells , and their relative phase , we construct the hamiltonian normal form of the two - mode reduction of the cubic - quintic double well system .
we then _ explicitly illustrate _ how the bifurcation analysis of this normal form encapsulates not only the symmetry breaking but _ also _ the symmetry restoring .
we argue that this cubic - quintic realization is the prototypical one where both of these effects can be observed and analytically quantified .
subsequently , in section iii , we proceed to test the relevant predictions by means of a computational bifurcation analysis , as well as through direct numerical simulations ( in order to monitor the predicted dynamical instabilities ) .
we find very good agreement with the symmetry breaking predictions of the model and even a quite fair agreement with the symmetry restoring ones ( which arise in a highly nonlinear regime and are hence less amenable to a two - mode analysis ) .
we also quantify the disparity of the analytical predictions and numerical results for large values of the nonlocality range parameter .
finally , section iv contains our conclusions and some directions for future study .
as indicated above , our fundamental model will be the 1d nls equation in the presence of two nonlocal terms , namely the cubic and quintic ones : @xmath1 with @xmath2 and the linear operator will be of the standard schrdinger type @xmath3 this encompasses the double - well potential of the form : @xmath4 with @xmath5 being the normalized strength of the parabolic trap and it is @xmath6 in a quasi-1d situation in becs ( here the effective trap frequency is the ratio of the longitudinal trap strength along the condensate over the one of the tightly confined transverse directions ) . in our study
we consider a typical experimentally relevant value of @xmath7 , while the generally tunable ( see e.g. @xcite ) parameters of the laser beam forming the light defect are chosen to be @xmath8 and @xmath9 ( which we have found to be fairly typical values representative of the phenomenology to be analyzed below ) . for the kernels @xmath10 , @xmath11 we will focus our considerations on either the gaussian @xmath12 or the exponential @xmath13 while the latter is more specifically relevant to the thermal nonlocal ( optical ) media and to quadratic nonlinear materials @xcite
, we also use the former due to the mathematical simplicity of its kernel . in any event
, our results will not be significantly different qualitatively between the two cases , although obviously the quantitative details will not be the same .
the key parameter here is the range of the nonlocal interaction parametrized by @xmath14 .
notice that both kernels in the limit of @xmath15 tend to a genuinely local interaction ( i.e. , @xmath16 ) .
we now develop the two - mode approximation in order to obtain a decomposition ( or more accurately a galerkin truncation ) of the solution @xmath17 over the minimal basis of fundamental states .
more specifically , we use an orthonormal basis composed by the wave functions @xmath18 , where @xmath19 and @xmath20 ( fig . 1 ) are the ground state and the first excited state , respectively , corresponding to the first two eigenvalues of @xmath21 that are @xmath22 and @xmath23 for our choice of potential parameters above . notice that these two eigenfunctions represent modes with support predominantly on the left and right well , respectively .
the eigenfunctions @xmath24 and the rotated basis employed herein of @xmath25 are both shown in fig . [ basis ] .
the two - mode approximation is then defined as @xmath26 where @xmath27 and @xmath28 are complex time - dependent amplitudes and the approximation consists of the truncation of the higher modes within the expansion . before substituting into the initial gross - pitaevskii ( gp ) equation
, we notice that the action of the linear operator @xmath21 on our basis elements is as follows : @xmath29 where @xmath30 and @xmath31 are linear combinations of the two eigevalues of @xmath21 respectively to the solutions @xmath19 , @xmath20 .
subsequently , substitution of our ansatz of eq .
( [ twomode ] ) in the full nonlinear problem of eq .
( [ eq1 ] ) yields : @xmath32\int r(x - x')\phi_l(x')\phi_r(x')dx'\\ + \delta|c_l|^4(c_l\phi_l + c_r\phi_r)\int r_2(x - x')\phi_l^4(x')dx ' + \delta|c_r|^4(c_l\phi_l + c_r\phi_r)\int r_2(x - x')\phi_r^4(x')dx'\\ + \delta\left[(4|c_l|^4|c_r|^4c_l + c_l^3{c_r^*}^2 + c_l^*|c_l|^2c_r^2)\phi_l + ( 4|c_l|^4|c_r|^4c_r + c_r^3{c_l^*}^2 + c_r^*|c_r|^2c_l^2)\phi_r \right ] \cdot\\ \cdot \int r_2(x - x')\phi_l^2(x')\phi_r^2(x')dx'\\ + 2\delta\left[(|c_l|^2c_l^2c_r^ * + |c_l|^4c_r)\phi_l + ( |c_l|^2|c_r|^2c_l + |c_l|^2c_r^2c_l^*)\phi_r\right]\int r_2(x - x')\phi_l^3(x')\phi_r(x')dx'\\ + 2\delta\left[(c_l^2|c_r|^2c_r^
* + |c_l|^2|c_r|^2c_r)\phi_l + ( |c_r|^4c_l + c_r^2|c_r|^2c_l^*)\phi_r \right]\int r_2(x - x')\phi_r^3(x')\phi_l(x')dx'.\end{aligned}\ ] ] in order to project the above equation onto the states @xmath25 we multiply with the respective function ( notice that the eigenfunctions are real due to the hermitian nature of the operator @xmath21 ) and integrate .
this involves the following integrals which will play a fundamental role in our considerations below : @xmath33 from the first nonlocal term , as well as @xmath34 from the second nonlocal term .
some alternatives that are derived if we interchange the variables @xmath35 and @xmath36 or swap @xmath37 and @xmath38 can also be equivalently considered .
a numerical study of the first four intergrals was already conducted in @xcite , where it was found that typically the integrals @xmath39 can be considered as negligible in comparison to @xmath40 which is the dominant term . on the other hand
, @xmath41 is close to @xmath39 for near - local interactions ( i.e. , for small values of @xmath14 ) , but becomes comparable to @xmath40 as the latter descreases for wide nonlocal interaction ranges ( i.e. , for large @xmath14 ) .
the criterion that we use to determine whether @xmath41 is negligible or not was @xmath42 where @xmath43 .
this yields that @xmath41 remains significant until ( i.e. , down to ) a critical value @xmath44 and @xmath45 for the gaussian and exponential kernels , respectively .
the dependence of the relevant overlap integrals on the range of the interaction @xmath14 is shown in fig .
[ overlap_int ] . taking into regard the second nonlocal term ( which for simplicity we have assumed to share the same range parameter as the first )
, we can see from fig .
[ overlap_int2 ] that the integrals @xmath46 are always negligible but @xmath47 appears to be a nontrivial competing term .
this is to a certain degree intuitively anticipated , as this represents the dominant term associated with the quintic interaction .
adapting the same criterion as in @xcite ( namely @xmath48 ) , we incorporate the relevant @xmath49 for @xmath50 , @xmath51 for the gaussian and exponential kernel , respectively . according to this
we may distinguish three cases : * the terms @xmath40 and @xmath49 are considered for @xmath52 ( for the gaussian kernel ) ; * the term with prefactor @xmath41 is added when @xmath53 .
* for @xmath54 , @xmath49 is omitted and only @xmath40 , @xmath41 are taken into account .
for the first case , the projection of the equation onto the states @xmath25 yields @xmath55 and by introducing madelung representation of action - angle or amplitude - phase decomposition ( @xmath56 ) , we obtain @xmath57 where we have defined the relative phase @xmath58 and the respective equations for @xmath59 and @xmath60 can be obtained by exchanging @xmath37 and @xmath38 and using @xmath61 instead of @xmath62 .
focusing now on the steady solutions ( satisfying @xmath63 ) , we need to enforce @xmath64 or @xmath65 for non - zero amplitudes .
this leads us to symmetric and antisymmetric ( equal or opposite amplitudes ) _ pairs _ of solutions , namely for @xmath66 we have the symmetric ( only positive ones among the ) solutions @xmath67 with @xmath68 , for @xmath69 ( @xmath70 for @xmath71 ) .
also , for @xmath72 , we have ( only the positive amplitude ones among ) the antisymmetric solutions @xmath73 with @xmath74 for @xmath69 ( resp .
@xmath75 for @xmath76 ) . for the asymmetric solutions one has to solve the polynomial @xmath77 which can more conveniently be written as a function of the norm of the solutions ( representing the atom number in becs and the optical intensity in optics ) .
thus , introducing @xmath78 ( @xmath79 ) yields the quartic polynomial @xmath80 for the second case ( @xmath81 ) the integrals @xmath40 , @xmath41 , @xmath49 are taken into account and the projection equations onto the states @xmath25 , read : @xmath82 here , the amplitude - phase decomposition yields @xmath83 we can , once again , obtain the set of stationary solutions as follows . when @xmath66 ( symmetric case ) the solutions will be ( the positive amplitude ones among ) @xmath84 for @xmath85 for @xmath69 ( @xmath86 for @xmath87 ) and when @xmath88 ( antisymmetric case ) the solutions are ( the positive amplitude ones among ) @xmath89 and exist for @xmath90 for @xmath69 ( @xmath91 for @xmath87 ) .
the asymmetric solutions now , directly in norm expression , will be given by the polynomial @xmath92 with @xmath93 here standing for @xmath94 .
in the third case , when @xmath54 , the effect of the quintic terms is deemed to be negligible and the situation reverts to the analysis of @xcite and is hence omitted here . in order to derive a more convenient form of the system so that we can proceed to the analysis of the spontaneous symmetry breaking ( ssb ) bifurcation , we introduce the population imbalance between the two wells , @xmath95 where @xmath96 and @xmath97 .
together with the relative phase between the two wells @xmath58 , this forms a set of conjugate variables , in which we obtain the dynamical system : @xmath98 this can be written in the hamiltonian form @xmath99 with the hamiltonian function @xmath100 note that @xmath93 stands either for @xmath40 ( @xmath101 ) or for @xmath94 ( @xmath102 ) .
the system possesses the stationary solutions ( critical points ) @xmath103 and @xmath104 with @xmath105 , @xmath106 , @xmath107 that correspond to the symmetric and antisymmetric solutions , identified above .
furthermore , the stationary solutions representing the asymmetric branches are given by : @xmath108 these branches emerge and merge as bifurcations from and to the symmetric or antisymmetric solutions and they exist for those values of @xmath78 for which @xmath109 .
taking @xmath110 , we get that @xmath111 by substituting @xmath112 or @xmath113 we get the same four possible expressions for @xmath78 as a function of @xmath14 that are displayed in fig.@xmath114 and we denote them with @xmath115 , @xmath116 , @xmath117 and @xmath118 ( the subscripts @xmath119 and @xmath120 correspond to the ( - ) signs in the left and right expressions of eq .
( [ ns ] ) , respectively , while the subscripts @xmath121 and @xmath122 to the ( + ) signs ) .
one can then see that when @xmath112 and demanding that @xmath123 , one gets that @xmath78 should either lie in the area outside the curves @xmath115 and @xmath116 or in the area inside the curves @xmath117 and @xmath118 . in the case of @xmath124 and @xmath71 ,
the role of the symmetric and anti - symmetric branches gets exchanged in as far as the bifurcation of the asymmetric branch is concerned ( see also below ) .
importantly , it can be observed in fig . [ crit_points ]
that @xmath115 is always negative , hence it is omitted for the principal case considered herein , namely @xmath125 and @xmath126 .
on the one hand , the _
critical _ conclusion of our analysis is that for @xmath127 , the system is predicted to have for the anti - symmetric branch both a symmetry breaking bifurcation ( at @xmath128 ) and a symmetry restoring one that eliminates the asymmetric branch ( at @xmath129 ) . on the other hand
, the right panel suggests that @xmath117 , @xmath118 coincide @xmath130 , beyond which there is only a single ( symmetry breaking ) bifurcation .
however , as will be discussed below , for large interaction range @xmath14 this prediction seems to have some discrepancy from what actually happens as we will see that in fact , we observe a symmetry restoring bifurcation while we do not observe a bifurcation at all in the symmetric branch . to the best of our knowledge
, this is the first example of an analytical prediction of the existence of a symmetry restoring bifurcation , a feature that is unique to the analysis of the normal form of the bifurcation for the cubic - quintic case ( and can not be predicted e.g. in the purely cubic case two - mode analysis of @xcite ) .
the new critical points appear or disappear as a pitchfork bifurcation that emerges from the antisymmetric solutions for @xmath88 respectively . from the symmetric solution , in this case of @xmath125 and @xmath126 ,
only a single bifurcation arises at @xmath131 . for the opposite case ( to the one principally considered herein ) of @xmath124 and @xmath71 ,
i.e. , for a focusing cubic nonlinearity , the bifurcations emerge from the symmetric branch , while for @xmath125 , i.e. , for a defocusing cubic term , then the relevant symmetry breakings arose from the anti - symmetric branch .
thus , in this case , we expect an asymmetric branch to bifurcate and break the symmetry at @xmath128 , while it returns to the parent symmetric branch restoring the symmetry at @xmath132 . on the other hand , for the anti - symmetric waveform with a focusing cubic nonlinearity ,
only a single bifurcation arises at @xmath133 .
we provide further details of each of these bifurcations and their comparison with the full numerics of the underlying nls model in the next section . from the system of eqs .
( [ system ] ) , one can reduce the dynamical evolution to a single second - order ode : @xmath134 which can also be rewritten in the `` position - momentum '' variables as : @xmath135 this renders the system amenable to the phase plane representation of the form shown in fig .
[ phase_plane ] . here
we observe that there is a stationary solution @xmath136 which is a fixed point of the center type .
however , for the cases when @xmath112 , for @xmath78 crossing the critical point @xmath116 in the case of the symmetric branch and for @xmath137 $ ] in the case of the anti - symmetric branch , there appear two more critical points at @xmath138 and @xmath139 , representing the asymmetric solutions . the point @xmath140 is a fixed point of center type before the bifurcation occurs , but past the relevant critical number of atoms ( or optical intensity ) , it becomes a saddle as the two new ( asymmetric ) fixed points that appear are of center type .
[ phase_plane ] shows the phase space of the full system , as well as the vicinity of the critical points for the gaussian kernel with @xmath141 , @xmath142 and @xmath143 .
it is worth mentioning at this point that there are no further changes in the stability of the critical points ( and thus in the corresponding stationary solutions ) for other values of @xmath78 except for those reported above . for the sake of simplicity
we illustrate this below for the antisymmetric solution bifurcation as a preamble towards the corresponding numerical results of the next section .
the antisymmetric solution corresponds the critical point @xmath144 where the jacobian of the linearized version of ( [ system ] ) is @xmath145 and its eigenvalues satisfy @xmath146 for the case where @xmath147 the graph of @xmath148 versus @xmath78 ( illustrated in the left panel of fig .
[ eigen ] ) shows clearly that the two initially ( i.e. , close to the linear limit ) purely imaginary eigenvalues become real at @xmath149 , so that the center type equilibrium becomes a saddle until @xmath150 where it turns back to its initial state , restoring the stability on the antisymmetric branch ( symmetry - restoring bifurcation ) with no other changes in between ( or after that ) . for the asymmetric solution that corresponds to the point @xmath151 , where @xmath152 and @xmath88 , the jacobian becomes @xmath153 and for its eigenvalues we obtain @xmath154 again for @xmath147 the graph of @xmath148 versus @xmath78 ( illustrated in the right panel of fig . [ eigen ] ) shows that the eigenvalues are always purely imaginary which corresponds to an equilibrium of the center type .
one can observe here that the values of @xmath78 where the eigenvalues of the asymmetric branch touch " the @xmath35-axis ( @xmath155 and @xmath156 ) coincide with the values where the bifurcation occurs ( fig . [ crit_points]-right panel )
therefore the critical points @xmath157 cease to exist before @xmath158 and after @xmath159 . for the values of @xmath78 within this interval ,
no further change of stability is observed .
as it is made clear in the next section , these stability results are in excellent agreement with the corresponding numerical ones .
additionally , it will be come transparent therein that additional turning points in the @xmath78 vs. @xmath160 bifurcation diagram do _ not _ correspond to any instabilities in complete agreement with the recent analysis of @xcite .
we now turn to the examination of our analysis against the results of numerical bifurcation analysis ( and in the next subsection also compare them to direct numerical simulations ) .
we focus here on the case where @xmath125 , @xmath69 , as we are especially interested in the case of _ competing _ interactions ; we will briefly also touch upon the case of @xmath124 and @xmath71 . in our numerical computations ,
the stationary solutions are obtained by using a fixed - point newton - raphson iteration for a finite difference decomposition of the relevant boundary value problem , with a choice of the grid spacing of @xmath161 and employing a parametric ( and wherever needed a pseudo - arclength ) continuation of the solutions with respect to the chemical potential parameter @xmath160 ( in optics this is the so - called propagation constant ) .
the linear stability is analyzed by considering the standard linearization around the stationary solutions @xmath162 in the form @xmath163 this yields the eigenvalue problem @xmath164 where the operators are defined as @xmath165\phi + \\
+ s\int_{-\infty}^{+\infty}k(x - x')\psi_0(x)\psi_0^*(x')\phi(x')dx ' + 2\delta\int_{-\infty}^{+\infty}k(x - x')\psi_0(x)\psi_0(x'){\psi_0^*}^2(x')\phi(x')dx'\end{aligned}\ ] ] and @xmath166 for any real function @xmath167 .
instability is guaranteed by the existence of any eigenvalues @xmath168 of the linearized operator with @xmath169 in the sense that perturbations along the corresponding eigendirections will deviate exponentially from the corresponding fixed point .
recall that this is also the case for all eigenvalues of our hamiltonian system , since when @xmath168 is an eigenvalue , so are @xmath170 , @xmath171 and @xmath172 . in the case where all eigenvalues are found to be purely imaginary ,
then the solution is found to be marginally stable . in our specific case of competing interactions ,
we comment on the following .
the positive value ( @xmath173 ) denotes the repulsive behavior of the cubic nonlocal term while the negative one @xmath69 leads to attractive behavior of the quintic nonlocal nonlinearity .
as we examine the bifurcation problem of nonlinear states from the corresponding linear eigenstates , we expect that for lower values of @xmath78 ( i.e. , weaker nonlinearities ) , the former repulsive term should be dominant , while for larger values of @xmath78 ( i.e. , stronger nonlinearities ) , it is anticipated that the latter attractive term will take over .
this is accurately reflected in the numerical bifurcation diagrams that we now show in figs .
[ stat1]-[stat3 ] , for three ( distinct by roughly an order of magnitude in each case ) values of the range @xmath14 . the first value of @xmath174 in fig . [ stat1 ]
is supposed to reflect the local case ( since the range of interaction is much smaller than any other intrinsic length scale in the system ) . here
the agreement with the two - mode approximation is very good quantitatively for low @xmath78 and very good qualitatively ( and even good quantitatively for some features such as chemical potentials of critical points ) for large @xmath78 .
the quality of these types of agreements is found to be preserved for an intermediate interaction range of @xmath141 in fig .
[ stat2 ] .
however , when the interaction range becomes sufficiently large that it competes ( or overcomes ) the length scale of the potential wells , then fundamental disparities are expected to be found and that is the very conclusion of fig .
[ stat3 ] for @xmath175 .
[ stat ] + in the first case where @xmath147 , the symmetric and antisymmetric branches of nonlinear states emanate from @xmath176 and @xmath177 , as expected , respectively ( @xmath178 and @xmath179 ) , both of them being dynamically stable , for sufficiently small values of @xmath78 .
the rightward bending of the branches for small @xmath78 confirms the dominance of the self - repulsive part of the ( cubic ) interactions for small @xmath78 , as indicated above .
the antisymmetric branch ( top right panel of fig .
[ stat1 ] and see also the zoom of the bottom panel of the figure ) is destabilized and the theoretically predicted asymmetric branch emerges .
the numerical value of the chemical potential for the bifurcation point is found to be @xmath180 , whereas the corresponding analytical one is @xmath181 , confirming the quantitative nature of the agreement with the two - mode approximation . for larger @xmath78
, we observe that the asymmetric solution has two apparent turning points ( where the sign of @xmath182 changes , but in fact its stability does not change - which agrees with the theoretical result presented in the previous section ) , before it reaches the anti - symmetric branch at the numerically computed value @xmath183 where we observe the symmetry restoring effect , which , in fact , re - stabilizes the anti - symmetric branch . in our theoretical analysis , we observe the same qualitative behavior and the symmetry restoring occurs at @xmath184 , in reasonable agreement with the full numerical results .
two additional observations should be made here .
on the one hand , since the symmetry restoring occurs at much larger values of @xmath78 , the relevant agreement is expected to be less adequate quantitatively than for the symmetry breaking occurring at lower @xmath78 .
this is because a two - mode expansion is less appropriate of a reduction at such higher nonlinearities . on the other hand
, it can indeed be observed that while the overall trend of the two curves is the same ( and even critical / turning points in terms of their chemical potential are rather accurately captured ) , this agreement is not adequate quantitatively e.g. for critical values of @xmath78 ( or for detailed quantitative matching of the curves for large @xmath78 ) .
for the symmetric solution of the top left panel of fig .
[ stat1 ] , we can observe that it is increasing monotonically until @xmath185 where it sustains a pitchfork bifurcation leading to the emergence of an asymmetric branch and also a subsequent turning point .
the symmetric branch becomes unstable thereafter and the asymmetric emerging state is the stable daughter branch .
notice that the theoretical analysis is once again quantitatively accurate for small @xmath78 and the agreement becomes more qualitative for higher @xmath78 s .
the critical point for the emergence of the asymmetric branch is predicted for @xmath186 in reasonable agreement with the full numerical result . for the case of @xmath187
the effects are similar to those in the previous case .
the symmetry breaking of the antisymmetric branch ( top right , as well as zoom in of the bottom panel of fig .
[ stat2 ] ) occurs now at @xmath188 according to the numerical results and at @xmath189 in the two - mode approximation , again attesting to its validity for small @xmath78 . after following a similar trajectory with the case @xmath147
, the asymmetric solution merges back to the antisymmetric one at @xmath190 ( numerical value ) or at @xmath191 ( analytical value ) with the antisymmetric branch again regaining its stability past the symmetry restoring bifurcation .
the symmetric solution ( top left panel of fig . [ stat2 ] ) again increases monotonically until it sustains a symmetry breaking bifurcation of its own at @xmath192 .
the two - mode approximation predicts this bifurcation to arise at @xmath193 .
+ next , in fig .
[ stat3 ] , we increase the interaction range , roughly , another order of magnitude by setting @xmath194 . here , as may be intuitively expected given that the interaction range is wider than the wells of the potential , the results are quite different .
for small values of @xmath160 ( and thus atom number @xmath78 or optical power ) we have a quite satisfactory agreement ( even quantititative ) with the two mode approximation , as may be expected . as a demonstration of that ,
we note that the symmetry breaking of the antisymmetric branch occurs in our analysis at @xmath195 , while numerically it is found to take place at @xmath196 . on the other hand , due to the predicted earlier collision of the critical points @xmath117 and @xmath118
, there is no symmetry restoring taking place in our normal form reduction .
nevertheless , we observe that such a restoring , in fact , still takes place in the full numerical bifurcation diagram .
furthermore , in this case , we have not been able to detect a symmetry - breaking bifurcation in the case of the symmetric branch , even though such a bifurcation is predicted within the reduction .
this illustrates that for such large values of @xmath14 , even the qualitative agreement previously associated with the large @xmath78 case dynamics should not be expected to be present . finally ,
we examine also one case where we switch the signs of the nonlocal terms to @xmath197 , so now the cubic term is the one that behaves attractively while the quintic one behaves repulsively .
this is illustrated in fig .
[ stat3a ] .
the interaction range @xmath14 is selected here to be @xmath121 and here we see that the same phenomenology appears in a region where the cheminal potential varies from @xmath198 to @xmath199 , thus attaining negative values . as earlier ,
both states emanate for the same values of @xmath160 and as we decrease its value we observe the symmetry breaking at @xmath200 ( both for numerical and analytical ) this time on the symmetric state which becomes unstable . as we further decrease the chemical potential to negative values of @xmath160 , the symmetry restoring of the asymmetric state towards its parent symmetric branch occurs at @xmath201 ( numerical value ) .
the analytical prediction for this critical point is @xmath202 .
hence , once again we observe a good qualitative agreement for larger @xmath78 ( although once again slight quantitative disparities exist between the overall curves and the critical points in terms of @xmath78 ) .
a look at the antisymmetric branch now shows us that a bifurcation occurs at the point where the solution changes slope ( @xmath182 ) , precisely at @xmath203 ( numerical ) and is theoretically predicted to arise at @xmath204 ( analytical ) with the antisymmetric branch becoming unstable past this critical point .
once again the zoom of the bottom panel confirms the quantitative nature of the analytical - numerical agreement for small values of @xmath78 , which retains its qualitative value even for larger @xmath78 .
+ + finally , we briefly turn to the dynamics of the system , in order to observe the implications of the dynamical instability due to the symmetry breaking .
the relevant evolution of the unstable solutions for @xmath205 and @xmath206 , in the case of @xmath141 ( recall that @xmath125 and @xmath126 ) are shown in fig .
[ evolution ] . in both cases
, it can be seen that the weak perturbation added on top of the exact numerical solution in the initial conditions has a projection along the unstable eigenmode .
this projection , for sufficiently long times ( about @xmath207 in the left panel and about @xmath208 in the right panel ) , gets amplified and eventually leads to a visible ( i.e. , of order unity ) symmetry breaking in the profile of the state . while the space - time evolution of the density ( in the atomic case ; optical intensity in the optical case ) is shown in fig .
[ evolution ] , an interesting alternative way to visualize the instability was proposed recently by @xcite . in the latter work ,
the pde dynamics was , in fact , projected to the phase plane of the two - mode approximation and visualized therein .
an example of such a visualization for the case of @xmath205 can be seen in fig .
[ pde_phase_plane ] . from
both the phase plane curves and the profiles illustrated underneath of the solution at different times , we can extract some interesting conclusions .
in particular , in the one degree of freedom reduction of our theoretical analysis , the trajectory occurs over iso - contours of the energy .
hence , the kind of phase plane picture shown in fig . [ pde_phase_plane ] would only be possible by `` conglomerating '' many distinct orbits .
however , it is important to appreciate that the pde has infinitely many degrees of freedom . in that capacity
, it is possible for the `` subspace '' of our two - mode approximation to _ dissipate _ energy towards ( or possibly regain energy from ) higher energy states ( of the point spectrum of the system ) . in so doing
, it appears as if the system visits further and further inward trajectories of lower energy , because indeed the excess energy has been imparted to other degrees of freedom .
this yields a clear illustration of how the subspace of our two - modes is a _ closed system _ for the ode reduction , but instead is an _
open system _ for the full pde evolutionary dynamics .
in the present work , we examined double well potentials in the presence of nonlocal interactions both in the cubic and in the quintic part of the nonlinearity .
we attempted to address such settings by means of a two - mode decomposition that has the notable advantage that nonlocality is not substantially different to handle therein , as the nonlocal kernels merely contribute to relevant overlap integrals that need some systematic book - keeping , but are otherwise not considerably harder than is the locally nonlinear case .
there are some particularly important attributes of the quintic case that we were able to extract via a normal form reduction and phase plane visualization ( under suitable circumstances of `` competition '' e.g. for a defocusing cubic but focusing quintic nonlinearity ) .
one such is that contrary to the purely cubic case , the reduction is able to predict not only a symmetry breaking bifurcation , but _ also _ a symmetry restoring one ( at least for a suitable interval of range parameters for the interaction kernel ) .
another unusual characteristic is that symmetry breaking bifurcations are encountered _ both _ for the symmetric and the antisymmetric branch , again differently than is the case for the cubic nonlinearity in the double well setting .
these features were tested against numerical bifurcation results and good agreement was found where appropriate ( e.g. low atom numbers and a suitable range of the interaction range ) . disparities arising for high @xmath78 and large @xmath14
were systematically explained . finally , the instability dynamics was visualized not only by space - time density evolution plots but also by offering its projection to the phase plane of the double well theoretical reduction and assessing the similarities and differences therein of the ode approximation and full pde result .
there are numerous possibilities for the extension of the present results to more elaborate contexts . on the one hand , even in the one - dimensional setting
, one could envision a study of different interaction ranges between the cubic and quintic terms ( or , for that matter , combinations of local and nonlocal nonlinearities within the cubic and/or quintic terms ) . on the other hand , extensions to one dimensional settings with more wells
would bring along a richer phenomenology ( in that setting the three - well local case has been studied @xcite and was recently revisited in @xcite ) , while in higher dimensional settings such as 2d , four well settings in a square configuration @xcite or other configurations exploiting the geometry of the system would be interesting to study .
pgk gratefully acknowledges support from the national science foundation under grants dms-0806762 and cmmi-1000337 , as well as by the alexander von humboldt foundation through a research fellowship , the alexander s. onassis public benefit foundation ( grant rzg 003/2010 - 2011 ) and the binational science foundation ( grant 2010239 ) .
the work of pat was partially supported by the state scholarships foundation in greece .
vmr gratefully acknowledges support from research council of auth ( grant 87872 ) .
this research has been co - financed by the european union ( european social fund - esf ) and greek national funds through the operational program `` education and lifelong learning '' of the national strategic reference framework ( nsrf ) - research funding program : thales .
investing in knowledge society through the european social fund .
the standard 1d model of the thermal optical nonlinearity is based on the following system ( see , e.g. , refs .
@xcite):@xmath209where @xmath210 is the squared correlation length of the nonlocal nonlinearity , the real field @xmath211 is a local perturbation of the refraction index , and @xmath212 is the coefficient of the optical absorption which leads to heating of the medium , so that @xmath213 is the local source in the effective heat - conductivity equation ( [ m ] ) .
if the heating is provided by the resonant absorption by dopants , the absorption may be saturable .
the saturation may be described , in the simplest approximation , by the following modification of eq .
( [ m]):@xmath214 s. raghavan , a. smerzi , s. fantoni , and s. r. shenoy , phys .
a * 59 * , 620 ( 1999 ) ; s. raghavan , a. smerzi , and v. m. kenkre , phys . rev .
a * 60 * , r1787 ( 1999 ) ; a. smerzi and s. raghavan , phys . rev .
a * 61 * , 063601 ( 2000 ) . c. par and m. florjaczyk , phys .
a * 41 * , 6287 ( 1990 ) ; a. i. maimistov , kvant . elektron .
* 18 * , 758 ( 1991 ) [ in russian ; english translation : sov .
j. quantum electron . * 21 * , 687 ; w. snyder , d.
j. mitchell , l. poladian , d. r. rowland , and y. chen , j. opt .
b * 8 * , 2102 ( 1991 ) ; p. l. chu , b. a. malomed , and g. d. peng , j. opt .
b * 10 * , 1379 ( 1993 ) ; n. akhmediev , and a. ankiewicz , phys .
lett . * 70 * , 2395 ( 1993 ) ; b. a. malomed , i. skinner , p. l. chu , and g. d. peng , phys .
e * 53 * , 4084 ( 1996 ) .
a. griesmaier , j. werner , s. hensler , j. stuhler , and t. pfau , phys . rev
. lett . * 94 * , 160401 ( 2005 ) ; j. stuhler , a. griesmaier , t. koch , m. fattori , t. pfau , s. giovanazzi , p. pedri , and l. santos , _
ibid_. * 95 * , 150406 ( 2005 ) ; j. werner , a. griesmaier , s. hensler , j. stuhler , and t. pfau , _ ibid_. * 94 * , 183201 ( 2005 ) ; a. griesmaier , j. stuhler , t. koch , m. fattori , t. pfau , and s. giovanazzi , _ ibid_. * 97 * , 250402 ( 2006 ) ; a. griesmaier , j. phys .
. phys . * 40 * , r91 ( 2007 ) ; t. lahaye , t. koch , b. frhlich , m. fattori , j. metz , a. griesmaier , s. giovanazzi , and t. pfau , nature ( london ) * 448 * , 672 ( 2007 ) .
t. khler , k. gral , and p. s. julienne , rev .
phys . * 78 * , 1311 ( 2006 ) ; j. sage , s. sainis , t. bergeman , and d. demille , phys .
lett . * 94 * , 203001 ( 2005 ) ; c. ospelkaus , l. humbert , p. ernst , k. sengstock , and k. bongs , _ ibid_. * 97 * , 120402 ( 2006 ) ; j. deiglmayr , a. grochola , m. repp , k. mrtlbauer , c. glck , j. lange , o. dulieu , r. wester , and m. weidemller , _ ibid_. * 101 * , 133004 ( 2008 ) ; f. lang , k. winkler , c. strauss , r. grimm , and j. h. denschlag , _ ibid_. * 101 * , 133005 ( 2008 ) . m. marinescu and l. you , phys . rev . lett . * 81 * , 4596 ( 1998 ) ; s. giovanazzi , d. odell , and g. kurizki , phys . rev
. lett . * 88 * , 130402 ( 2002 ) ; i. e. mazets , d. h. j. odell , g. kurizki , n. davidson , and w. p. schleich , j. phys .
b 37 , s155 ( 2004 ) ; r. lw , r. gati , j. stuhler and t. pfau , europhys . lett . * 71 * , 214 ( 2005 ) .
w. krlikowski , o. bang , j. j. rasmussen , and j. wyller , phys .
e * 64 * , 016612 ( 2001 ) ; o. bang , w. krlikowski , j. wyller and j. j. rasmussen , phys .
e * 66 * , 046619 ( 2002 ) ; j. wyller , w. krlikowski , o. bang and j. j. rasmussen , phys .
e * 66 * , 066615 ( 2002 ) .
b. l. lawrence and g. i. stegeman .
two - dimensional bright spatial solitons stable over limited intensities and ring formation in polydiacetylene para - toluene sulfonate . _ optics letters
_ , * 23 * , 8 ( 1998 ) 591593 .
f. smektala , c. quemard , v. couderc , and a. barthlmy , j. non - cryst .
solids 274 , 232 ( 2000 ) ; g. boudebs , s. cherukulappurath , h. leblond , j. troles , f. smektala , and f. sanchez , opt . commun .
219 , 427 ( 2003 ) . | in the present work , we examine the combined effects of cubic and quintic terms of the long range type in the dynamics of a double well potential . employing a two - mode approximation ,
we systematically develop two cubic - quintic ordinary differential equations and assess the contributions of the long - range interactions in each of the relevant prefactors , gauging how to simplify the ensuing dynamical system .
finally , we obtain a reduced canonical description for the conjugate variables of relative population imbalance and relative phase between the two wells and proceed to a dynamical systems analysis of the resulting pair of ordinary differential equations . while in the case of cubic and quintic interactions of the same kind ( e.g. both attractive or both repulsive ) , only a symmetry breaking bifurcation can be identified , a remarkable effect that emerges e.g. in the setting of repulsive cubic but attractive quintic interactions
is a `` symmetry restoring '' bifurcation .
namely , in addition to the supercritical pitchfork that leads to a spontaneous symmetry breaking of the anti - symmetric state , there is a subcritical pitchfork that eventually reunites the asymmetric daughter branch with the anti - symmetric parent one .
the relevant bifurcations , the stability of the branches and their dynamical implications are examined both in the reduced ( ode ) and in the full ( pde ) setting .
the model is argued to be of physical relevance , especially so in the context of optical thermal media . |
synchronizability is one of the currently leading problems in the fast - growing field of complex networks @xcite .
a number of studies have been devoted to scrutinize which network topologies are more prone to sustain a stable globally - synchronized state of generic oscillators defined at each of its nodes @xcite .
this question is of broad interest since many complex systems in fields ranging from physics , biology , computer science , or physiology , can be seen as networks of coupled oscillators , whose functionality depends crucially on the network ability to maintain a synchronous oscillation pattern .
in addition , it has been shown that networks with good synchronizability are also `` good '' for ( i ) fast random walk spreading and therefore for efficient communication @xcite , ( ii ) searchability in the presence of congestion @xcite , ( iii ) robustness in the absence of privileged hubs @xcite , ( iv ) performance of neural networks @xcite , ( v ) generating consensus in social networks , etc .
another related and important problem that has received a lot of attention , but that we will not study here , is the dynamics _ towards _ synchronized states ( see for example @xcite ) . in general terms
, we can say that the degree of synchronizability is high when all the different nodes in a given network can `` talk easily '' to each other , or information packets can travel efficiently from any starting node to any target one .
it was first observed that adding some extra links to an otherwise regular lattice in such a way that a small - world topology @xcite is generated , enhances synchronizability @xcite .
this was attributed to the fact that the node - to - node average distance diminishes as extra links are added .
afterwards , heterogeneity in the degree distribution was shown to hinder synchronization in networks of symmetrically coupled oscillators , leading to the so called `` paradox of heterogeneity '' as heterogeneity is known to reduce in average the node - to - node distance but still it suppresses synchronizability .
the effect of other topological features as betweenness centrality , correlation in the degree distribution and clustering has been also analyzed .
for example , it has been shown that the presence of weighted links ( rather than uniform ones ) and asymmetric couplings do enhance further the degree of synchronizability @xcite , but here we focus on un - weighted and un - directed links . certainly , the main breakthrough was made by barahona and pecora @xcite who , in a series of papers , established a criterion based on spectral theory to determine the stability of synchronized states under very general conditions .
their main contribution is to link graph spectral properties with network dynamical properties . in particular , they considered the laplacian matrix , encoding the network topology , and showed that the degree of synchronizability ( understood as the range of stability of the synchronous state ) is controlled by the ratio between its largest eigenvalue ( @xmath0 ) and the smallest non - trivial one ( @xmath1 ) , i.e. @xmath2 , where @xmath3 is the total number of nodes @xcite .
the smaller @xmath4 the better the synchronizability .
note that , as the range of variability of @xmath5 is quite limited ( it is directly related to the maximum connectivity @xcite ) , minimizing @xmath4 is almost equivalent to maximizing the denominator @xmath6 ( i.e the _ spectral gap _ ) when the degree distribution is kept fixed .
it is worth noticing that , even if the eigenratio @xmath4 can be related to ( or bounded by ) topological properties such as the ones cited above ( average path length , betweenness centrality , etc . )
, none of these provides with a full characterization of a given network and therefore they are not useful to determine _ strict _ criteria for synchronizability @xcite .
nevertheless , they can be very helpful as long as they give easy criteria to determine in a _ rough _ way synchronizability properties , without having to resort to lengthly eigenvalue calculations .
in a couple of recent papers , we tackled the problem of finding the optimally synchronizable topology , given a fixed number of nodes and edges linking them , and assuming symmetric and un - weighted links @xcite .
the strategy we followed was to implement a simulated annealing algorithm @xcite with a cost - function given by @xmath4 ; starting with a random topology with @xmath3 nodes and @xmath7 links , random rewirings that decrease the value of @xmath4 are accepted with larger probability than those increasing @xmath4 ( for more details see @xcite ) , until eventually a stationary ( optimal or close to optimal ) network is generated . employing this optimization algorithm
, we identified the family of `` optimal network topologies '' which we called _ entangled networks_. the main topological trait of entangled networks is the absence of bottlenecks and hubs ; all sites are very much alike and the links form very intricate structures , which lead typically to ( i ) the absence of a well - defined community structure , ( ii ) poor modularity , and ( iii ) large shortest - loops . in this way , every single site is close to any other one owing to the existence of a very `` democratic '' or entangled structure in which properties such as site - to - site distance , betweenness , and minimum - loop - size are very homogeneously distributed ( see @xcite ) .
entangled networks were identified as _ ramanujan graphs _ and they have been related to other interesting concepts in graph theory as _ expanders _ @xcite and cage - graphs ( see @xcite and references therein ) .
these are used profusely in computer science and are under current intense study in the mathematical literature . for example
, expanders and ramanujan graphs are very useful in the design of efficient communication networks , construction of error - correcting codes , or de - randomization of random algorithms @xcite .
these applications greatly amplify the relevance of entangled networks in different contexts . despite of their mathematical beauty and excellent performance in network - design , entangled topologies
are not easily found in biological , social , or any other `` real - life '' networks .
an exception are some food - webs , for which topologies very similar to entangled ones have been reported @xcite . as argued in @xcite the rarity in nature of entangled networks comes from the fact that they emerge out of a _ global _
optimization process not easily fulfilled by means of any dynamical simple mechanism in growing networks where , usually , only _
information is available . instead
, real complex networks in very different contexts have been shown to exhibit , rather generically , scale - free degree distributions .
these are much more heterogeneous than entangled topologies . keeping this in mind , in this paper we explore the question of ( global ) optimization of synchronizability within the realm of scale - free networks with a fixed degree - distribution . in particular , constraining our optimization algorithm to preserve a scale - free architecture , we are able to find the optimally synchronizable networks and study the emergence of non - trivial degree - degree correlations .
this study is related to previous works by sorrentino , di bernardo and others @xcite , who argued that disassortative networks ( in which nodes with similar degrees tend to be _ not _ connected among themselves ) @xcite are more synchronizable that assortative ones ( where nodes with similar degrees tend to be connected ) .
our study differs from previous ones in that ( i ) we derive a rigorous lower - bound for @xmath4 in terms of a parameter measuring degree - degree correlations and ( ii ) we explicitly design optimal networks with a given degree - distribution and , by doing so , we verify that even if it is true that more disassortative networks typically exhibit better synchronizability , this is _ not _ always the case .
finally , we also face the question of which are the _ pessimal _ networks for synchronization purposes .
actually , in some applications , synchronization ( or consensus , or complete homogenization ) are not desirable properties .
this might be the case , for example , in neural networks for which global synchronization implies epileptic - like activity @xcite .
the question of how topology can hinder such states is both pertinent and relevant and also , it can give further insight on the key structural features of synchronization . with this goal in mind
, we revert the optimization algorithm , and define an inverse optimization process just by maximizing @xmath4 ( rather than minimizing it ) , we analyze the topology of the resulting pessimal ( or optimally un - synchronizable ) networks .
in this section we derive an upper bound for the spectral graph in terms of the correlation coefficient @xmath8 .
this coefficient was introduced by newman in @xcite to quantify the tendency of nodes with similar degrees , @xmath9 , to be connected between themselves . in particular , calling @xmath3 the number of nodes and @xmath7 the total number of links ( @xmath10 ) the correlation coefficient @xmath8 can be computed as ( see @xcite for more details ) : @xmath11 ^ 2 } { l^{-1}\sum_{i \sim j}\frac12(k_i^2+k_j^2 ) - [ l^{-1}\sum_{i \sim j}\frac12(k_i+k_j)]^2 } \quad.\ ] ] where @xmath12 stands for the sum over links ( i.e. over all nodes i and j connected by a link ; every link is counted only once ) .
this parameter takes positive ( negative ) values for assortative ( disassortative ) configurations . defining the laplacian matrix as @xmath13
if nodes @xmath14 and @xmath15 are connected ( disconnected ) , and @xmath16 , we can obtain an upper bound for @xmath4 by recalling that the first non - trivial laplacian eigenvalue @xmath1 can be expressed as @xcite : @xmath17 where @xmath18 is is the set of all possible non - constant vectors ( in the space in which the laplacian operator acts ) .
taking @xmath19 , which is one possible vector out of the set @xmath18 , we obtain : @xmath20 where @xmath21 is defined as : @xmath22 a different selection of the vector would lead to a different inequality .
the advantage of our choice is that the obtained bound can be related to the correlation coefficient @xmath8 , even if it is not guaranteed that it provides a tight bound .
the different terms in the numerator and denominator of equation ( [ r ] ) can be written as @xmath23 by substituting these expressions in eq .
( [ r ] ) and rearranging them we readily obtain : @xmath24 and finally @xmath25 which provides a rigorous upper bound for the spectral gap in terms of @xmath8 .
observe that the more negative the value of @xmath8 ( i.e. the more disassortative the network ) the larger the upper bound for @xmath1 and , therefore , @xmath4 is allowed to take smaller values , and the corresponding network can be more synchronizable . summing up , the inequality eq.([ineq ] )
establishes that , as a rule of thumb , disassortative networks are more prone to have stable synchronized states than assortative ones , in agreement with previous results @xcite . as a word of caution ,
let us underline that this does _ not _ imply that , given a fixed degree distribution , any disassortative network is better synchronizable than any assortative one , as we will illustrate in the following section . a similar result to ours has been recently derived @xcite . in particular , upper and lower bounds for @xmath1 were obtained in terms of a parameter @xmath26 quantifying the degree - degree correlation ( @xmath26 is a simplified version of the more detailed one , @xmath8 , defined by eq.([r ] ) ) .
these upper and lower bounds were derived elaborating upon known bounds for the spectral gap in terms of the cheeger constant @xcite . in order to obtain them , the authors implicitly assume that , for a fixed degree distribution , the cheeger constant is an uni - parametric function of @xmath26 .
however , given a fixed @xmath26 , as this parameter does not specify completely the graph topology @xcite , different graphs with different cheeger constants can be constructed .
therefore , the derivation of the bounds in @xcite involves some type of mean - field - like approximation , while the upper bound here has been obtained in a rigorous way .
in this section we describe the optimization algorithm suitable for finding the network topology which extremizes the stability range of a global synchronous state in networks subject to a topological constraint : a fixed degree distribution .
in particular , we apply this method to networks with scale - free topology and analyze the degree - degree correlations of the resulting ( extremized ) graphs .
the algorithm is a modified simulated - annealing @xcite aimed at minimizing a cost function @xmath27 , where @xmath2 .
a detailed description of the algorithm can be found either in @xcite or in @xcite .
it yields networks for which the synchronizability is close to extremal ( i.e. maximum or minimum ) , depending on the selected cost function @xmath28 .
in particular , setting @xmath29 one gets networks with _ optimal _
synchronizability while choosing @xmath30 the optimization procedure yields what we call _ pessimal _ networks .
the simulated - annealing rewiring process starts from networks generated using the configuration model @xcite ; in particular , it starts from connected networks with @xmath3 nodes and @xmath7 links , such that their degree distribution sample a power law @xmath31 , @xmath32 , with trivial ( random ) degree - degree correlation between neighboring nodes ( see fig.[red ] ) .
all the results in what follows correspond to @xmath33 . the graphs emerging out of the @xmath4 and @xmath34 minimization processes , starting from the network in fig .
[ red ] , are depicted in fig.[opt - pes ] .
naked eye inspection reveals the enormous differences between optimal and pessimal topologies . while the optimal ones resemble very much the very intricate and as - homogeneous - as - possible topology of entangled networks , pessimal topologies are as chain - like as possible , with two non - linear `` heads '' at both extremes , necessary to preserve the scale - free topology constraint .
note that large values of @xmath4 imply small values of @xmath1 and therefore , following the criterion for graph ( bi)partitioning described for instance in @xcite , pessimal graphs have to be easily divisible into two parts by cutting an as - small - as - possible number of links .
this is , indeed , achieved in an optimal way for linear ( chain - like ) topologies .
let us remark , that different initial conditions with the same scale - free distribution , lead to outputs indistinguishable statistically from the ones in fig.[opt - pes ] , rendering robust the previous results . to put these observations under a more quantitative basis , we measure degree correlations using ( i ) the average degree @xmath35 of the neighbors of a node with degree @xmath9 , and ( ii ) the correlation coefficient @xmath8 given by eq.([r ] ) .
fig.[pkk ] shows @xmath35 , averaged over @xmath36 different realizations , for initially uncorrelated networks ( see , as an example , fig .
[ red ] ) as well as for the final optimal ( fig .
[ opt - pes ] , left ) and pessimal ( fig .
[ opt - pes ] , right ) networks .
it reveals that optimally - synchronizable scale - free networks tend to display disassortative mixing ( high - degree nodes tend to be connected with low - degree ones ) while , on the contrary , pessimal scale - free networks tend to be assortative .
this result agrees with the tendency predicted by the bound on the spectral gap in eq .
( [ ineq ] ) as well as with previous results @xcite . actually , one could have anticipated these conclusions knowing that a network with good synchronization properties is also able to efficiently communicate any two nodes @xcite . in this sense ,
disassortative mixing , in which low connected nodes are preferentially linked to hubs which act as information distributors , seems most efficient . on the other hand ,
pessimally synchronizable networks resulting from the minimization of @xmath34 ( or , equivalently , maximization of @xmath4 ) , tend to exhibit assortative mixing , i.e. @xmath35 grows with @xmath9 , at least up to a finite - size cutoff @xmath37 , as shown in fig .
the origin of such a cutoff is evident after realizing that the probability of having large hubs connected to other very large hubs must go to zero since the total number of links present in the system is finite .
obviously , the cutoff grows with system size and diverges asymptotically .
the highly assortative chain - like topology of pessimal networks can be understood by the necessity of hampering the efficient communication between any two nodes in the system .
this is achieved by maximizing the distance between any two hubs by interposing between them a linear chain of poorly - connected nodes .
let us underline that the above observation on the effect of disassortative ( assortative ) mixing does not necessarily imply that maximizing disassortativity ( assortativity ) leads to optimal ( pessimal ) synchronizability .
this is illustrated in figs .
[ evol].a - b ( figs .
[ evol].d - e ) , which plot respectively the time evolution of the eigenratio @xmath4 and the correlation coefficient @xmath8 during the optimization processes .
the figure regarding optimization shows that the eigenratio @xmath4 is not a monotonic function of @xmath8 .
this fact is made explicit in the insets to figs .
[ evol].a - b : the asymptotic minimun value of @xmath4 does not correspond to the network for which @xmath8 is minimum ( obtained in the example shown around @xmath38 steps ) . this points out that , despite being a good indicator of synchronizability , disassortativity can not be regarded as an unique topological measure of the stability of the synchronous state @xcite .
moreover , further correlations apart from the observed assortativity / disassortativity are built up during the optimization process . in fig .
[ evol].c we plot the time evolution of the shortest - loop average length @xmath39 , defined as the average over all nodes of the shortest loop passing through each node ; it shows that @xmath39 grows during the minimization of @xmath4 , but again it exhibits a maximum before reaching its optimal - topology value .
the tendency towards forming large loops for optimally synchronizable networks was reported before for entangled networks , where the smallest loops tend to be as large as possible @xcite .
at this point we want to emphasize that the approach we have undertaken here is a constructive one , as opposed to that in @xcite , where different random networks with predefined degree distribution and correlations are explored to analyze how the externally - imposed assortative or disassortative correlation affects the eigenratio @xmath4 of the resulting networks .
results here complement those in @xcite .
we have studied the problem of network synchronizability , which is directly related to many other important problems as efficient communication , searchability in the presence of congestion , many computer - science tasks , etc . while generically , the optimal networks for synchronizability , assuming un - weighted and un - directed links , are super - homogeneous , _ entangled _ topologies , in which all nodes look very much alike , in this work we have investigated the nature of optimal scale - free networks . the final goal is to analyze how are the degree - degree correlations of optimal scale - free networks . for that
, we have used the standard spectral approach consisting in relating the degree of synchronizability to the laplacian matrix eigenratio @xmath4 . in a first part of this work ,
we have derived a rigorous lower bound for @xmath4 in terms of the correlation coefficient @xmath8 ( as defined by eq.([r ] ) ) , which is a measure of the degree - degree correlations .
this lower bound turns out to be proportional @xmath40 , hence , showing that the more negative @xmath8 ( i.e. the more disassortative the network ) the smaller the lower bound and , therefore , the smaller values @xmath4 is allowed to take , and the better the synchronizability of the resulting network . in the second part
, we have explicitly constructed optimal networks ( with a fixed number of nodes , links , and a given scale - free degree distribution ) by employing a recently introduced simulated - annealing algorithm @xcite .
we find that optimal networks tend to be disassortative , as found already in previous studies @xcite , and in agreement with the expectations from the previously found lower bound .
however , as there is not a one - to - one correspondence between @xmath4 and the correlation coefficient @xmath8 , more disassortative networks do not always synchronize better .
actually , we have illustrated how during the optimization process , at some point , the degree of assortativity increases ( i.e. @xmath8 increases ) as the network becomes more and more synchronizable .
the emerging optimal networks exhibit also a tendency to have large loops and a rather intricate structure ( as occurs for entangled networks ) . finally , we have reverted the optimization process and , by minimizing @xmath34 , we have found what we call pessimally synchronizable networks .
these topologies are characterized by a long string ended by two `` heads '' of nodes with degrees larger than @xmath41 ( required to preserve the scale - free degree distribution ) and are , therefore , highly assortative .
contrarily to the case above , loops are very short .
these topologies are the worst possible ones ( compatible with the imposed scale - free degree distribution ) if the task is to synchronize the network .
but , on the contrary , they constitute the best choice if the goal is to avoid synchronization ( or , equivalently , avoid communicability , searchability , homogenization ) , which might be important for some applications .
for instance , in order to maximize the average time that a random infection ( or random walk ) takes to reach an arbitrary target node , this is the type of network to design .
we acknowledge financial support from the spanish ministerio de educacin y ciencia ( fis2005 - 00791 ) and junta de andaluca ( fqm-165 ) .
10 s. h. strogatz , nature * 410 * , 268 ( 2001 ) .
+ a. l. barabsi and r. albert , rev . mod
. phys . * 74 * , 47 ( 2002 ) .
+ s. n. dorogovtsev and j. f. f. mendes , _ evolution of networks : from biological nets to the internet and www _ , oxford univ . press ( 2003 ) .
+ m. e. j. newman , siam review * 45 * , 167 ( 2003 ) .
+ s. boccaletti , v. latora , y. moreno , m. chavez , and d .- u .
hwang , phys .
rep . * 424 * , 175 ( 2006 ) .
m. barahona and l. m. pecora , phys .
89 * , 054101 ( 2002 ) .
+ see also , l. m. pecora and t. l. carroll , phys .
lett . * 64 * , 821 ( 1990 ) ; ibid , * 80 * , 2109 ( 1998 ) .
+ x. f. wang and g. chen , int .
j. bifurcation chaos appl .
* 12 * , 187 ( 2002 ) .
a. e. motter , c. zhou , and j. kurths , phys .
rev e * 71 * , 016116 ( 2005 ) ; aip conference proceedings * 776 * , 201 ( 2005 ) ; europhys . lett . * 69 * , 334 ( 2005 ) . + t. nishikawa and a.
e. motter , phys .
e * 73 * , 065106 ( 2006 ) .
+ a. e. motter , new j. phys .
* 9 * , 182 ( 2007 ) .
hwang , m. chavez , a. amann , and s. boccaletti , phys .
94 * , 138701 ( 2005 ) .
+ m. chavez , d .- u .
hwang , a. amann , h. g. e. hentschel , and s. boccaletti , phys .
lett . * 94 * , 218701 ( 2005 ) .
+ m. chavez , d .- u .
hwang , and s. boccaletti , eur .
j. special topics * 146 * , 129 ( 2007 ) .
j. gmez - gardees , y. moreno , and a. arenas , phys .
* 98 * , 034101 ( 2007 ) .
+ j. gmez - gardees , y moreno , and a arenas , phys .
e * 75 * , 066106 ( 2007 ) .
+ p. n. graw and m. menzinger , phys .
rev . * 72 * , 015101 ( 2005 ) .
+ e. oh , k. rho , h. hong , and b. khang , phys .
e * 72 * , 047101 ( 2005 ) .
note that the laplacian eigenvalues @xmath42
satisfy @xmath43 , where @xmath44 is the largest degree in the graph @xcite .
b. bollobs , _ extremal graph theory _ academic press , new york .
+ w. tutte , _ graph theory as i have known it _ , oxford u. press , new york , ( 1998 ) .
+ f. chung , _ spectral graph theory _
, number 92 in cbms region conference series in mathematics .
am . math . soc . 1997 .
m. di bernardo , f. garofalo , and f. sorrentino , proceedings of the 44th ieee conference on decision and control , pp . 4616 ( 2005 ) .
+ f. sorrentino , m. di bernardo , g. huerta , cellar , and s. boccaletti , physica d * 224 * , 123 ( 2006 ) .
+ see also , b. wang , h. tang , t. zhou , and z. xiu , arxiv : cond - mat/0512079 .
m. girvan , m. e. j. newman , proc .
usa , * 99 * , 7821 - 7826 ( 2002 ) .
m. e. j. newman , m. girvan , phys .
e * 69 * , 026113 ( 2004 ) .
see also , l. donetti and m. a. muoz , j. stat .
: theor . exp .
( 2004 ) p10012 ; l. donetti and m. a. muoz , in `` _ modeling cooperative behavior in the social sciences _ '' , aip conf . proc .
779 , 104 ( 2005 ) . | by employing a recently introduced optimization algorithm we explicitely design optimally synchronizable ( unweighted ) networks for any given scale - free degree distribution .
we explore how the optimization process affects degree - degree correlations and observe a generic tendency towards disassortativity .
still , we show that there is not a one - to - one correspondence between synchronizability and disassortativity . on the other hand
, we study the nature of optimally un - synchronizable networks , that is , networks whose topology minimizes the range of stability of the synchronous state
. the resulting `` pessimal networks '' turn out to have a highly assortative string - like structure .
we also derive a rigorous lower bound for the laplacian eigenvalue ratio controlling synchronizability , which helps understanding the impact of degree correlations on network synchronizability .
_ keywords _ : article preparation , iop journals |
linear regression for interval - valued data has been attracting increasing interests among researchers .
see @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , for a partial list of references .
however , issues such as interpretability and computational feasibility still remain .
especially , a commonly accepted mathematical foundation is largely underdeveloped , compared to its demand of applications . by proposing our new model
, we continue to build up the theoretical framework that deeply understands the existing models and facilitates future developments . in the statistics literature ,
the interval - valued data analysis is most often studied under the framework of random sets , which includes random intervals as the special ( one - dimensional ) case .
the probability - based theory for random sets has developed since the publication of the seminal book of @xcite .
see @xcite for a relatively complete monograph . to facilitate the presentation of our results
, we briefly introduce the basic notations and definitions in the random set theory .
let @xmath1 be a probability space .
denote by @xmath2 or @xmath3 the collection of all non - empty compact subsets of @xmath4 . in the space @xmath3 , a linear structure is defined by minkowski addition and scalar multiplication , i.e. , @xmath5 @xmath6 and @xmath7 .
a natural metric for the space @xmath3 is the hausdorff metric @xmath8 , which is defined as @xmath9 where @xmath10 denotes the euclidean metric .
a random compact set is a borel measurable function @xmath11 , @xmath3 being equipped with the borel @xmath12-algebra induced by the hausdorff metric . for each @xmath13 ,
the function defined on the unit sphere @xmath14 : @xmath15 is called the support function of x. if @xmath16 is convex almost surely , then @xmath17 is called a random compact convex set .
( see @xcite , p.21 , p.102 . )
the collection of all compact convex subsets of @xmath4 is denoted by @xmath18 or @xmath19 .
when @xmath20 , the corresponding @xmath19 contains all the non - empty bounded closed intervals in @xmath0 .
a measurable function @xmath21 is called a random interval .
much of the random sets theory has focused on compact convex sets .
let @xmath22 be the space of support functions of all non - empty compact convex subsets in @xmath19 .
then , @xmath22 is a banach space equipped with the @xmath23 metric @xmath24^{\frac{1}{2}},\ ] ] where @xmath25 is the normalized lebesgue measure on @xmath14 . according to the embedding theorems ( see @xcite , @xcite ) ,
@xmath19 can be embedded isometrically into the banach space @xmath26 of continuous functions on @xmath14 , and @xmath22 is the image of @xmath27 into @xmath26 .
therefore , @xmath28 , @xmath29 , defines a metric on @xmath27 .
particularly , let @xmath30=[x^c - x^r , x^c+x^r]\ ] ] be an bounded closed interval with center @xmath31 and radius @xmath32 , or lower bound @xmath33 and upper bound @xmath34 , respectively .
then , the @xmath35-metric of @xmath36 is @xmath37 and the @xmath35-distance between two intervals @xmath36 and @xmath38 is @xmath39^{\frac{1}{2}}\\ & = & \left[\left(x^c - y^c\right)^2+\left(x^r - y^r\right)^2\right]^{\frac{1}{2}}.\end{aligned}\ ] ] existing literature on linear regression for interval - valued data mainly falls into two categories . in the first
, separate linear regression models are fitted to the center and range ( or the lower and upper bounds ) , respectively , treating the intervals essentially as bivariate vectors .
examples belonging to this category include the center method by @xcite , the minmax method by @xcite , the ( constrained ) center and range method by @xcite , and the model m by @xcite .
these methods aim at building up model flexibility and predicting capability , but without taking the interval as a whole .
consequently , their geometric interpretations are prone to different degrees of ambiguity .
take the constrained center and range method ( ccrm ) for example . adopting the notations in @xcite ,
it is specified as @xmath40 where @xmath41 and @xmath42 .
it follows that @xmath43 ^ 2+\left[\beta_1^r\left(x_i^r - x_j^r\right)\right]^2.\end{aligned}\ ] ] because @xmath44 in general , a constant change in @xmath45 does not result in a constant change in @xmath46 .
in fact , a constant change in any metric of @xmath36 as an interval does not lead to a constant change in the same metric of @xmath38 .
this essentially means that the model is not linear in intervals . in the second category ,
special care is given to the fact that the interval is a non - separable geometric unit , and their linear relationship is studied in the framework of random sets .
investigation in this category began with @xcite developing a least squares fitting of compact set - valued data and considering the interval - valued input and output as a special case .
precisely , he gave analytical solutions to the real - valued numbers @xmath47 and @xmath48 under different circumstances such that @xmath49 is minimized on the data .
the pioneer idea of @xcite was further studied in @xcite , where the @xmath35-metric was extended to a more general metric called @xmath50-metric originally proposed by @xcite .
the advantage of the @xmath50-metric lies in the flexibility to assign weights to the radius and midpoints in calculating the distance between intervals .
so far the literature had been focusing on finding the affine transformation @xmath51 that best fits the data , but the data are not assumed to fulfill such a transformation . a probabilistic model along this direction kept missing until @xcite , and simultaneously @xcite , proposed the same simple linear regression model for the first time .
the model essentially takes on the form of @xmath52 with @xmath53 and @xmath54 , c\in\mathbb{r}$ ] .
this can be written equivalently as @xmath55 it leads to the following equation that clearly shows linearity in @xmath19 : @xmath56 some advances have been made regarding this model and the associated estimators .
@xcite derived least squares estimators for the model parameters and examined them from a theoretical perspective . @xcite
established a test of linear independence for interval - valued data
. however , many problems still remain open such as biases and asymptotic distributions , as anticipated in @xcite .
this paper presents a continuous development addressing some issues and open problems in the direction of model ( [ mod - gil ] ) .
first , we relax the restriction of model ( [ mod - gil ] ) that the hukuhara difference @xmath57 must exist ( see @xcite ) and generalize the univariate model to the multiple case .
we also give analytical least squares ( ls ) solutions to the model parameters .
second , we show that our model and ls estimation together accommodate a decomposition of the sums of squares in @xmath19 analogous to that of the classical linear regression .
third , we derive explicit formulas of the ls estimates for the univariate model , which exist with probability going to one .
the ls estimates are further shown to be asymptotically unbiased .
a simulation study is carried out to validate our theoretical findings , as well as compare our model to ccrm .
finally , we apply our model to a climate data set to illustrate the applicability of our model .
the rest of the paper is organized as follows : section 2 formally introduces our model and the associated ls estimators .
then , the sums of squares and coefficient of determination in @xmath19 are defined and discussed .
section 3 presents the theoretical properties of the ls estimates for the univariate model .
the simulation study is reported in section 4 , and the real data application is presented in section 5 .
we give concluding remarks in section 6 . technical proofs and useful lemmas
are deferred to the appendices .
we consider an extension of model ( [ mod - gil ] ) to the form @xmath58 where @xmath59=[-c , c]$ ] , @xmath60 .
it is equivalently expressed as @xmath61 this leads to the following center - radius specification @xmath62 where @xmath63 , @xmath64 , and the signs
@xmath65 " correspond to the two cases in ( [ mod - cases ] ) . define @xmath66 our model
is specified as @xmath67 where @xmath68 , @xmath69 $ ] , @xmath70 , and @xmath71 . to model the outcome intervals @xmath72 $ ] by @xmath73 interval - valued predictors
@xmath74 $ ] , @xmath75 ; @xmath76 , we consider the multivariate extension of ( [ mod-1 ] ) : @xmath77 which leads to the following center - radius specification @xmath78 where @xmath68 , @xmath69 $ ] , @xmath79 , and @xmath80 .
we have assumed @xmath81 and @xmath82 are independent in this paper to simplify the presentation .
the model that includes a covariance between @xmath81 and @xmath82 can be implemented without much extra difficulty .
least squares method is widely used in the literature to estimate the interval - valued regression coefficients ( @xcite , @xcite , @xcite ) .
it minimizes @xmath83 on the data with respect to the parameters .
denote @xmath84 then the sum of squared @xmath35-distance between @xmath85 and @xmath86 is written as @xmath87\\ & = & \sum_{i=1}^{n}\left[\left(b+\sum_{j=1}^{p}a_jx_{j , i}^c - y_i^c\right)^2+\left(\sum_{j=1}^{p}\left|a_j\right|x_{j , i}^r+\mu - y_i^r\right)^2\right].\end{aligned}\ ] ] therefore , the lse of @xmath88 is defined as @xmath89 let @xmath90 be the sample covariances of the centers and radii of @xmath91 and @xmath92 , respectively .
especially , when @xmath93 , we denote by @xmath94 and @xmath95 the corresponding sample variances .
in addition , define @xmath96 as the sample covariances of the centers and radii of @xmath91 and @xmath38 , respectively .
then , the minimization problem ( [ def - ls ] ) is solved in the following proposition .
[ prop : ls_solu ] the least squares estimates of the regression coefficients @xmath97 , if they exist , are solution of the equation system : @xmath98 and then , @xmath99 are given by @xmath100 the variance of a compact convex random set @xmath36 in @xmath4 is defined via its support function as @xmath102 where the expectation is defined by aumann integral ( see @xcite , @xcite ) as @xmath103 see @xcite . for the case
@xmath20 , it is shown by straightforward calculations that @xmath104,\\ & & \text{var}(x)=\text{var}\left(x^c\right)+\text{var}\left(x^r\right).\end{aligned}\ ] ] this leads us to define the sums of squares in @xmath105 to measure the variability of interval - valued data .
a definition of the coefficient of determination @xmath101 in @xmath105 follows immediately , which produces a measure of goodness - of - fit .
the total sum of squares ( sst ) in @xmath27 is defined as @xmath106.\ ] ] the explained sum of squares ( sse ) in @xmath27 is defined as @xmath107.\ ] ] the residual sum of squares ( ssr ) in @xmath27 is defined as @xmath108.\ ] ] the coefficient of determination ( @xmath101 ) in @xmath27 is defined as @xmath109 where @xmath110 and @xmath111 are defined in ( [ def : sst ] ) and ( [ def : ssr ] ) , respectively .
analogous to the classical theory of linear regression , our model ( [ mmod-1**])-([mmod-2 * * ] ) together with the ls estimates ( [ def - ls ] ) accommodates the partition of @xmath110 into @xmath112 and @xmath111 . as a result , the coefficient of determination ( @xmath101 )
can also be calculated as the ratio of @xmath112 and @xmath110 .
the partition has a series of important implications of the underlying model , one of which being that the residual @xmath113/@xmath114 and the predictor @xmath115 are empirically uncorrelated in @xmath116 . [ thm : ss ] assume model ( [ mmod-1**])-([mmod-2 * * ] ) .
let @xmath117 and @xmath118 in ( [ exp - c])-([exp - r ] ) be calculated according to the ls estimates @xmath119 in ( [ def - ls ] ) .
then , @xmath120 it follows that the coefficient of determination in @xmath19 is equivalent to @xmath121 it is possible to get negative values of @xmath122 by its definition ( [ exp - r ] ) .
theorem [ thm : pred - adjust ] gives an upper bound of the probability of this unfortunate event . if the model largely explains the variability of @xmath123 , @xmath124 should be very small and so is this bound .
then , the rare cases of negative @xmath122 can be rounded up to 0 since @xmath118 is nonnegative . otherwise , if most of the variability of @xmath123 lies in the random error , the probability of getting negative predicts may not be ignorable , but it is essentially due to the insufficiency of the model and a different model should be pursued anyway .
[ thm : pred - adjust ] consider model ( [ mmod-1**])-([mmod-2 * * ] ) .
let @xmath86 be defined in ( [ exp - c])-([exp - r ] ) .
then , @xmath125
in this section , we study the theoretical properties of the lse for the univariate model ( [ mod-1**])-([mod-2 * * ] ) . applying proposition [ prop : ls_solu ] to the case @xmath126
, we obtain the two sets of half - space solutions , corresponding to @xmath127 and @xmath128 , respectively , as follows : @xmath129 and @xmath130 the final formula for the ls estimates falls in three categories . in the first
, there is one and only one set of existing solution , which is defined as the lse . in the second ,
both sets of solutions exist , and the lse is the one that minimizes @xmath131 . in the third situation , neither solution exists , but this only happens with probability going to @xmath132 .
we conclude these findings in the following theorem .
[ thm : ls_solu ] assume model ( [ mod-1**])-([mod-2 * * ] ) .
let @xmath133 be the least squares solution defined in ( [ def - ls ] ) .
if @xmath134 , then there exists one and only one half - space solution .
more specifically , + * i. * if in addition @xmath135 , then the ls solution is given by @xmath136 * ii .
* if instead @xmath137 , then the ls solution is given by @xmath138 otherwise , @xmath139 , and then either both of the half - space solutions exist , or neither one exists .
in particular , + * iii . *
if in addition @xmath140 , then both of the half - space solutions exist , and @xmath141 * iv . *
if instead @xmath142 , then the ls solution does not exist , but this happens with probability converging to 0 . unlike the classical linear regression , ls estimates for the model ( [ mod-1**])-([mod-2 * * ] )
are biased .
we calculate the biases explicitly in proposition [ prop : ls_exp ] , which are shown to converge to zero as the sample size increases to infinity
. therefore , the ls estimates are asymptotically unbiased .
[ prop : ls_exp ] let @xmath143 be the least squares solution in theorem [ thm : ls_solu ] . then ,
@xmath144,\end{aligned}\ ] ] @xmath145.\end{aligned}\ ] ] [ thm : ls_consist ] consider model ( [ mod-1**])-([mod-2 * * ] ) .
assume @xmath146 and @xmath147 .
then , the least squares solution @xmath143 in theorem [ thm : ls_solu ] is asymptotically unbiased , i.e. @xmath148 as @xmath149 .
we carry out a systematic simulation study to examine the empirical performance of the least squares method proposed in this paper .
first , we consider the following three models : + * model 1 : @xmath150 , @xmath151 , @xmath152 , @xmath153 , @xmath154 ; * model 2 : @xmath155 , @xmath151 , @xmath152 , @xmath153 , @xmath156 ; * model 3 : @xmath150 , @xmath151 , @xmath157 , @xmath153 , @xmath154 ; where data show a positive correlation , a negative correlation , and a positive correlation with a negative @xmath25 , respectively .
a simulated dataset from each model is shown in figure [ fig : sim - data ] , along with its fitted regression line . + to investigate the asymptotic behavior of the ls estimates , we repeat the process of data generation and parameter estimation 1000 times independently using sample size @xmath158 for all the three models
. the resulting 1000 independent sets of parameter estimates for each model / sample size are evaluated by their mean absolute error ( mae ) and mean error ( me ) .
the numerical results are summarized in table [ tab : sim ] .
consistent with proposition [ prop : ls_exp ] , @xmath159 tends to underestimate @xmath47 when @xmath160 and overestimate @xmath47 when @xmath128 .
this bias also causes a positive and negative bias in @xmath161 , when @xmath160 and @xmath128 , respectively .
similarly , a positive bias in @xmath162 is induced by the negative bias in @xmath163 .
all the biases diminish to 0 as the sample size increases to infinity , which confirms our finding in theorems [ thm : ls_consist ] .
.evaluation of parameter estimation [ cols="^ , > , > , > , > , > , > , > " , ] [ tab : sim - com ]
in this section , we apply our model to analyze the average temperature data for large us cities , which are provided by national oceanic and atmospheric administration ( noaa ) and are publicly available .
the three data sets we obtained specifically are average temperatures for 51 large us cities in january , april , and july .
each observation contains the averages of minimum and maximum temperatures based on weather data collected from 1981 to 2010 by the noaa national climatic data center of the united states .
july in general is the hottest month in the us . by this analysis
, we aim to predict the summer ( july ) temperatures by those in the winter ( january ) and spring ( april ) .
figure [ fig : real - data ] plots the july temperatures versus those in january and april , respectively .
the parameters are estimated according to ( [ eqn : lse-1])-([eqn : lse-3 ] ) as @xmath164 denote by @xmath165 , @xmath166 , and @xmath167 , the average temperatures in a us city in january , april , and july , respectively .
the prediction for @xmath167 based on @xmath165 and @xmath166 is given by @xmath168 the three sums of squares are calculated to be @xmath169 therefore , the coefficient of determination is @xmath170 finally , the variance parameters can be estimated as @xmath171 thus , by theorem [ thm : pred - adjust ] , an upper bound of @xmath172 on average is estimated to be @xmath173 which is very small and reasonably ignorable .
we calculate @xmath174 for the entire sample and all of them are well above @xmath132 .
so , for this data , although @xmath175 and it is possible to get negative predicted radius , it in fact never happens because the model has captured most of the variability .
the empirical distributions of residuals are shown in figure [ fig : real - residual ] .
both distributions are centered at 0 , with the center residual having a slightly bigger tail .
we have rigorously studied linear regression for interval - valued data in the metric space @xmath176 .
the new model we introduces generalizes previous models in the literature so that the hukuhara difference @xmath177 needs not exist .
analogous to the classical linear regression , our model together with the ls estimation leads to a partition of the total sum of squares ( ssr ) into the explained sum of squares ( sse ) and the residual sum of squares ( ssr ) in @xmath176 , which implies that the residual is uncorrelated with the linear predictor in @xmath176 .
in addition , we have carried out theoretical investigations into the least squares estimation for the univariate model .
it is shown that the ls estimates in @xmath176 are biased but the biases reduce to zero as the sample size tends to infinity .
therefore , a bias - correction technique for small sample estimation could be a good future topic .
the simulation study confirms our theoretical findings and shows that the least squares estimators perform satisfactorily well for moderate sample sizes . 99 artstein z , vitale , ra . a strong law of large numbers for random compact sets .
annals of probability . 1975;5:879882 .
aumann rj .
integrals of set - valued functions . j. math .
1965;12:112 .
billard l , diday e. regression analysis for interval - valued data . in : data
analysis , classification and related methods , proceedings of the seventh conference of the international federation of classification societies ( ifcs00 ) .
springer , belgium ; 2000.p.369374 .
billard l , diday e. symbolic regression analysis . in : classification , clustering and data
analysis , proceedings of the eighth conference of the international federation of classification societies ( ifcs02 ) .
springer , poland ; 2002.p.281288 .
billard l. dependencies and variation components of symbolic interval - valued data . in : selected contributions in data analysis and classification .
springer , berlin heidelberg ; 2007.p.312 .
blanco - fernndez a , corral n , gonzlez - rodrguez g. estimation of a flexible simple linear model for interval data based on set arithmetic .
computational statistics & data analysis .
2011;55:25682578 .
blanco - fernndez a , colubi a , gonzlez - rodrguez g. confidence sets in a linear regression model for interval data . journal of statistical planning and inference .
2012;142:13201329 .
carvalho fat , lima neto ea , tenorio , cp . a new method to fit a linear regression model for interval - valued data .
lecture notes in computer sciences .
2004;3238 : 295306 .
cattaneo megv , wiencierz a. likelihood - based imprecise regression . international journal of approximate reasoning .
2012;53:11371154 .
diamond p. least squares fitting of compact set - valued data . j. math .
1990;147:531544 .
gil ma , lopez mt , lubiano ma , and montenegro m. regression and correlation analyses of a linear relation between random intervals .
2001;10,1:183201 .
gil ma , lubiano ma , montenegro m , lopez mt .
least squares fitting of an affine function and strength of association for interval - valued data . metrika .
2002;56 : 97111 .
gil ma , gonzlez - rodrguez g , colubi a , montenegro m. testing linear independence in linear models with interval - valued data .
computational statistics & data analysis .
2007;51:30023015 .
gonzlez - rodrguez g , blanco a , corral n , colubi a. least squares estimation of linear regression models for convex compact random sets .
advances in data analysis and classification . 2007;1:6781 . hrmander h. sur la fonction dappui des ensembles convexes dans un espace localement convexe .
arkiv fr mat .
1954;3:181186 .
hukuhara m. integration des applications mesurables do nt la valeur est un compact convexe .
funkcialaj ekvacioj .
1967;10:205223 .
kendall dg .
foundations of a theory of random sets . in : harding ef and kendall dg ( ed ) stochastic geometry .
john wiley & sons , new york ; 1974 .
krner r. a variance of compact convex random sets .
institut fr stochastik , bernhard - von - cotta - str .
2 09599 freiberg ; 1995 .
krner r. on the variance of fuzzy random variables .
fuzzy sets and systems . 1997;92:8393 .
krner r , nther w. linear regression with random fuzzy variables : extended classical estimates , best linear estimates , least squares estimates .
information sciences .
1998;109 : 95118 .
lyashenko nn . limit theorem for sums of independent compact random subsets of euclidean space .
journal of soviet mathematics .
1982;20:21872196 .
lyashenko nn . statistics of random compacts in euclidean space .
journal of soviet mathematics .
1983;21:7692 .
manski cf , tamer t. inference on regressions with interval data on a regressor or outcome .
2002;70:519546 .
matheron g. random sets and integral geometry .
john wiley & sons , new york ; 1975 .
molchanov i. theory of random sets .
springer , london ; 2005 .
lima neto ea , carvalho fat .
centre and range method for fitting a linear regression model to symbolic interval data .
computational statistics & data analysis . 2008 ; 52:15001515 .
lima neto ea , carvalho fat . constrained linear regression models for symbolic interval - valued variables .
computational statistics & data analysis . 2010;54:333347 .
radstrm h. an embedding theorem for spaces of convex sets .
differentiating @xmath131 with respect to @xmath25 , @xmath48 , and @xmath178 , @xmath76 , respectively , and setting the derivatives to zero , we get @xmath179 equations ( [ lse-1])-([lse-2 ] ) yield @xmath180 equations ( [ eqn : lse-1 ] ) are obtained by plugging ( [ lse-4])-([lse-5 ] ) into ( [ lse-3 ] ) , and equations ( [ eqn : lse-2])-([eqn : lse-3 ] ) follow from ( [ lse-4])-([lse-5 ] ) .
this completes the proof . according to definitions ( [ def : sst])-([def : ssr ] ) , @xmath181\nonumber\\ & = & sse+ssr+2\sum_{i=1}^{n}\left[\left(y_i^c-\hat{y}_i^c\right)\left(\hat{y}_i^c-\overline{y^c}\right ) + \left(y_i^r-\hat{y}_i^r\right)\left(\hat{y}_i^r-\overline{y^r}\right)\right]\nonumber\\ &
= & sse+ssr+2\sum_{i=1}^{n}\left[\left(y_i^c-\hat{y}_i^c\right)\hat{y}_i^c+\left(y_i^r-\hat{y}_i^r\right)\hat{y}_i^r\right].\label{ss : eqn-1}\end{aligned}\ ] ] the last equation is due to ( [ lse-1])-([lse-2 ] ) .
further in view of ( [ exp - c])-([exp - r ] ) and ( [ lse-3 ] ) , we have @xmath182\\ & = & \sum_{i=1}^{n}\left[\left(y_i^c-\hat{y}_i^c\right)\sum_{j=1}^{p}a_jx_{j , i}^c+\left(y_i^r-\hat{y}_i^r\right)\sum_{j=1}^{p}|a_j|x_{j , i}^r\right]\\ & = & \sum_{j=1}^{p}a_j\sum_{i=1}^{n}\left[\left(y_i^c-\hat{y}_i^c\right)x_{j , i}^c+\left(y_i^r-\hat{y}_i^r\right)sgn(a_j)x_{j , i}^r\right]\\ & = & 0.\end{aligned}\ ] ] this together with ( [ ss : eqn-1 ] ) completes the proof .
notice that @xmath183 an application of markov s inequality completes the proof .
parts * i * , * ii * and * iii * are obvious from proposition [ prop : ls_solu ] .
part * iv * follows from lemma [ lem : cov_r ] in appendix ii .
we prove the cases @xmath127 and @xmath128 separately .
to simplify notations , we will use @xmath184 throughout the proof , but the expectation should be interpreted as being conditioned on @xmath36 . + * case i : @xmath127*. + from lemma [ lem : cov_est ] , we have @xmath185 + \sum_{i < j}(x_i^r - x_j^r ) \left [ ( y_i^r - y_j^r)-a(x_i^r - x_j^r ) \right]}{\sum_{i
< j}(x_i^c - x_j^c)^2+\sum_{i < j}(x_i^r - x_j^r)^2}\\ = & \frac{\sum_{i < j}(x_i^c - x_j^c)(\lambda_i-\lambda_j)+\sum_{i < j}(x_i^r - x_j^r)(\eta_i-\eta_j ) } { \sum_{i < j}(x_i^c - x_j^c)^2+\sum_{i < j}(x_i^r - x_j^r)^2}.\\\end{aligned}\ ] ] this immediately yields @xmath186 similarly , @xmath187 } { \sum_{i < j}(x_i^c - x_j^c)^2+\sum_{i < j}(x_i^r - x_j^r)^2},\ ] ] and consequently , @xmath188 notice now @xmath189\\ & & -\left[\int_{\{\hat{a}=a^-\}}(a^+-a ) \mathrm{d}\mathbb{p}+\int_{\{\hat{a}=a^-\}}(a^--a ) \mathrm{d}\mathbb{p } \right]\nonumber\\ & = & e\left(a^+-a\right)-\int_{\{\hat{a}=a^-\}}(a^+-a^- ) \mathrm{d}\mathbb{p}\nonumber\\ & = & -e(a^+-a^-)i_{\{\hat{a}=a^-\}}\label{eqn-3}.\end{aligned}\ ] ] here , equation ( [ eqn-3 ] ) is due to ( [ eqn-1 ] ) . recall that @xmath190 } { \sum_{i < j}(x_i^c - x_j^c)^2+\sum_{i < j}(x_i^r - x_j^r)^2},\label{a+-a-}\end{aligned}\ ] ] since @xmath191 .
therefore , @xmath192 } { \sum_{i < j}(x_i^c - x_j^c)^2+\sum_{i < j}(x_i^r - x_j^r)^2}\right\}i_{\{\hat{a}=a^-\}}\nonumber\\ & = & -\frac{2\sum_{i < j}\left[|a|(x_i^r - x_j^r)^2p(\hat{a}=a^-)+(x_i^r - x_j^r)e(\eta_i-\eta_j)i_{\{\hat{a}=a^-\}}\right ] } { \sum_{i < j}(x_i^c - x_j^c)^2+\sum_{i < j}(x_i^r - x_j^r)^2}\nonumber\\ & = & -\frac{2\sum_{i < j}(x_i^r - x_j^r)^2p(\hat{a}=a^-)}{\sum_{i < j}(x_i^c - x_j^c)^2+\sum_{i < j}(x_i^r - x_j^r)^2}\nonumber\\ & = & -\frac{2as^2(x^r)}{s^2(x^c)+s^2(x^r)}p(\hat{a}=a^-).\label{rst-1}\end{aligned}\ ] ] similar to the preceding arguments , @xmath193 recall again that @xmath194}{s^2\left(x^c\right)+s^2\left(x^r\right)}.\label{a++a-}\end{aligned}\ ] ] it follows that @xmath195 * case ii : @xmath128 * + in this case , we have @xmath196}{s^2(x^c)+s^2(x^r)},\\ & a^--a = \frac{\sum_{i < j}(x_i^c - x_j^c)(\lambda_i-\lambda_j)-\sum_{i < j}(x_i^r - x_j^r)(\eta_i-\eta_j)}{s^2(x^c)+s^2(x^r)}.\\\end{aligned}\ ] ] these imply @xmath197 similar to the case of @xmath127 , we obtain @xmath198 these , together with ( [ a+-a- ] ) and ( [ a++a- ] ) , imply , @xmath199 the desired result follows from ( [ rst-1 ] ) , ( [ rst-2 ] ) , ( [ rst-3 ] ) and ( [ rst-4 ] ) . from ( [ b+ ] ) and ( [ b- ] ) , @xmath200 similarly , from ( [ mu+ ] ) and ( [ mu- ] ) , @xmath201 hence , the desired result follows by proposition [ prop : ls_exp ] and lemma [ lem : sign - consist ] in the appendix .
[ lem : cov_r ] assume model ( [ mod-1**])-([mod-2 * * ] ) and @xmath202 . then @xmath203 . consequently , @xmath204 with probability converging to 1 . according to ( [ mod-2 * * ] ) , @xmath205-e\left(x^r\right)e\left(|a|x^r+\eta_1\right)\nonumber\\ & = & |a|e\left(x^r\right)^2+\mu e\left(x^r\right)-|a|\left[e\left(x^r\right)\right]^2-\mu e\left(x^r\right)\nonumber\\ & = & |a|\text{var}\left(x^r\right)\nonumber\\ & \geq & 0,\label{cov_true}\end{aligned}\ ] ] provided that @xmath202 . by the slln , @xmath206 ( [ cov_sample ] ) together with ( [ cov_true ] ) completes the proof . to prove , @xmath209\\ & = n\sum_{i=1}^nx_i^vy_i^v-(\sum_{i=1}^nx_i^v)(\sum_{i=1}^ny_i^v)=n^2s\left(x^v , y^v\right).\\\end{aligned}\ ] ] follows by replacing @xmath210 with @xmath211 and @xmath210 with @xmath212 in the above calculations .
[ lem : sign - consist ] assume model ( [ mod-1**])-([mod-2 * * ] ) .
assume in addition that @xmath146 and @xmath147 .
let @xmath133 be the least squares solution defined in ( [ def - ls ] ) .
then @xmath213 as @xmath149 .
we prove the case @xmath127 only .
the case @xmath128 can be proved similarly . under the assumption that @xmath127 , @xmath214 and consequently , @xmath215 . according to theorem [ thm : ls_solu ] , the only other circumstance under which @xmath216 is when @xmath217 and @xmath218 simultaneously .
it is therefore sufficient to show that @xmath219 notice @xmath220\\ & & + \frac{1}{n}\sum_{i=1}^{n}\left[\left(a^+x_i^r+\mu - y_i^r\right)^2-\left(a^-x_i^r+\mu - y_i^r\right)^2\right]\\ : = & & \frac{1}{n}\left(i+ii\right).\end{aligned}\ ] ] the first term @xmath221\\ & = & \sum_{i=1}^{n}\left[\left(a^{+}-a\right)^2\left(x_i^c-\overline{x^c}\right)^2+\left(\lambda_i-\overline{\lambda}\right)^2 -2\left(a^{+}-a\right)\left(x_i^c-\overline{x^c}\right)\left(\lambda_i-\overline{\lambda}\right)\right]\\ & & -\sum_{i=1}^{n}\left[\left(a^{-}-a\right)^2\left(x_i^c-\overline{x^c}\right)^2+\left(\lambda_i-\overline{\lambda}\right)^2 -2\left(a^{-}-a\right)\left(x_i^c-\overline{x^c}\right)\left(\lambda_i-\overline{\lambda}\right)\right]\\ & = & \left[\left(a^{+}-a\right)^2-\left(a^{-}-a\right)^2\right]\sum_{i=1}^{n}\left(x_i^c-\overline{x^c}\right)^2\\ & & -2\left(a^{+}-a^{-}\right)\sum_{i=1}^{n } \left(x_i^c-\overline{x^c}\right)\left(\lambda_i-\overline{\lambda}\right)\\ & = & \left(a^{+}-a^{-}\right)\left[\left(a^{+}+a^{-}-2a\right)\sum_{i=1}^{n}\left(x_i^c-\overline{x^c}\right)^2 -2\sum_{i=1}^{n}\left(x_i^c-\overline{x^c}\right)\left(\lambda_i-\overline{\lambda}\right)\right].\end{aligned}\ ] ] from this , and the assumption that @xmath217 , we see that @xmath222 is equivalent to @xmath223 on the other hand , @xmath224\sum_{i=1}^{n}\left(x_i^c-\overline{x^c}\right)^2 -\sum_{i=1}^{n}\left(x_i^c-\overline{x^c}\right)\left(\lambda_i-\overline{\lambda}\right)\\ & & = \left[\frac{\sum_{i < j}\left(x_i^c - x_j^c\right)\left(\lambda_i-\lambda_j\right)}{\sum_{i < j}\left(x_i^c - x_j^c\right)^2+\sum_{i < j}\left(x_i^r - x_j^r\right)^2 } -a\frac{s^2\left(x^r\right)}{s^2\left(x^c\right)+s^2\left(x^r\right)}\right]\sum_{i=1}^{n}\left(x_i^c-\overline{x^c}\right)^2\\ & & -\sum_{i=1}^{n}\left(x_i^c-\overline{x^c}\right)\left(\lambda_i-\overline{\lambda}\right)\\ & & = \frac{\sum_{i=1}^{n}\left(x_i^c-\overline{x^c}\right)^2}{\sum_{i < j}\left(x_i^c - x_j^c\right)^2+\sum_{i < j}\left(x_i^r - x_j^r\right)^2 } \sum_{i < j}\left(x_i^c - x_j^c\right)\left(\lambda_i-\lambda_j\right)\\ & & -\sum_{i=1}^{n}\left(x_i^c-\overline{x^c}\right)\left(\lambda_i-\overline{\lambda}\right ) -a\frac{s^2\left(x^r\right)}{s^2\left(x^c\right)+s^2\left(x^r\right)}\sum_{i=1}^{n}\left(x_i^c-\overline{x^c}\right)^2\\ & & = \frac{\sum_{i=1}^{n}\left(x_i^c-\overline{x^c}\right)^2}{\sum_{i < j}\left(x_i^c - x_j^c\right)^2+\sum_{i < j}\left(x_i^r - x_j^r\right)^2 } \left[n\sum_{i=1}^{n}\left(x_i^c-\overline{x^c}\right)\left(\lambda_i-\overline{\lambda}\right)\right]\\ & & -\sum_{i=1}^{n}\left(x_i^c-\overline{x^c}\right)\left(\lambda_i-\overline{\lambda}\right ) -a\frac{s^2\left(x^r\right)}{s^2\left(x^c\right)+s^2\left(x^r\right)}\sum_{i=1}^{n}\left(x_i^c-\overline{x^c}\right)^2\\ & & = \sum_{i=1}^{n}\left(x_i^c-\overline{x^c}\right)\left(\lambda_i-\overline{\lambda}\right ) \left[\frac{s^2\left(x^c\right)}{s^2\left(x^c\right)+s^2\left(x^r\right)}-1\right]\\ & & -a\frac{s^2\left(x^r\right)}{s^2\left(x^c\right)+s^2\left(x^r\right)}\sum_{i=1}^{n}\left(x_i^c-\overline{x^c}\right)^2\\ & & = -\frac{s^2\left(x^r\right)}{s^2\left(x^c\right)+s^2\left(x^r\right ) } n\left[as^2\left(x^c\right)+s\left(x^c , \lambda\right)\right],\end{aligned}\ ] ] where @xmath225 denotes the sample covariance of the random variables @xmath31 and @xmath226 , which converges to @xmath132 almost surely by the independence assumption .
therefore , @xmath227\nonumber\\ & & \to c_1<0\label{eqn : consist-2}\end{aligned}\ ] ] almost surely , as @xmath149 . + by the similar calculation , we have that the second term @xmath228\nonumber\\ & & \to c_2<0\label{eqn : consist-3}\end{aligned}\ ] ] almost surely , as @xmath149 .
( [ eqn : consist-2 ] ) and ( [ eqn : consist-3 ] ) together imply that @xmath229 this completes the proof . | it has been some time since interval - valued linear regression was investigated . in this paper , we focus on linear regression for interval - valued data within the framework of random sets .
the model we propose generalizes a series of existing models .
we establish important properties of the model in the space of compact convex subsets of @xmath0 , analogous to those for the classical linear regression .
furthermore , we carry out theoretical investigations into the least squares estimation that is widely used in the literature . a simulation study is presented that supports our theorems .
finally , an application to a climate data set is provided to demonstrate the applicability of our model . |
some strongly coupled lattice fermion - gauge models with a charged scalar field , which break chiral symmetry dynamically , might be considered to be a possible alternative to the higgs mechanism for mass generation , as discussed in @xcite .
let us concentrate on a prototype with @xmath0 gauge group , a scalar of fixed modulus and one staggered fermion ( corresponding to 4 flavors ) , where both the scalar and fermion have charge one .
the action has been described in @xcite with three bare parameters @xmath1 .
the dynamical mass generation is meaningful only in the chiral limit @xmath2 .
we consider here the phase transition line * net * between two phases @xcite : \(1 ) dynamical mass generation ( nambu ) phase , below the * net * line , where chiral symmetry is spontaneously broken ( @xmath3 ) due to the strong gauge fluctuations so that the fermion mass @xmath4 is dynamically generated ; \(2 ) higgs phase , above the * nets * line , where the higgs mechanism is operative , but @xmath5 .
the scalar field induces a second order chiral phase transition * ne * line which opens the possibility for approaching the continuum . whether such a model can replace the higgs mechanism depends crucially on the existence and renormalizability of the continuum limit . to search for such a continuum theory and grasp its nature
, we need to make precise determination of the second order phase transition point with divergent correlation lengths .
for such a purpose , we have done extensive simulations using hybrid monte carlo ( hmc ) algorithm and developed some new methods for locating the * ne * line .
the hmc simulations have been done on @xmath6 and @xmath7 , where on @xmath6 , we have better statistics ( 1024 - 6500 trajectories ) for different @xmath1 .
the detailed results for the spectrum are reported in @xcite .
we have measured the following local observables : plaquette energy @xmath8 , link energy @xmath9 and chiral condensate @xmath10 , where for @xmath10 we use the stochastic estimator method .
however , it is very difficult to use the local quantities at finite @xmath11 to detect a critical behavior on the * ne * line , since they show smooth behavior as a function of @xmath12 or @xmath13 .
( one could expect the critical behavior only in the infinite volume and chiral limit . ) for @xmath14 near the point * e * , the peaks of susceptibility for different quantities develop and coincide , while the boson mass @xmath15 gets smaller .
concerning the location of the * et * line , on the @xmath6 and @xmath7 for @xmath16 or @xmath17 and @xmath18 , we find explicit two state signals from the thermo - cycle , time history and histogram analysis of the local quantities . on the * ne * line
, the @xmath19 meson shows more obviously the phase transition than other quantities . in the nambu phase
, the @xmath19 meson should obey the pcac relation . in the symmetric phase ,
the @xmath19 meson is no longer a goldstone boson , and one should observe a deviation from pcac . at @xmath20 , these properties are nicely seen in fig .
[ fig1 ] , from which one sees that for @xmath21 where the system is in the broken phase , we have goldstone bosons .
however , on @xmath6 , even at @xmath22 ( possibly in the chiral symmetric phase ) , a linear extrapolation leads to @xmath23 . for larger @xmath12
, the extrapolated result gets smaller ( e.g. at @xmath24 , @xmath25 ) and is expected to vanish in the @xmath26 limit . of course
, one should not expect the linear extrapolation to be valid at the critical point .
@xmath27 is not a convenient order parameter for the chiral transition of a finite system due to the sensitivity of chiral extrapolation .
we employ a different method for determining the chiral transition , namely we calculate the chiral susceptibility in the chiral limit , defined by @xmath28 if there is a second order chiral phase transition , @xmath29 should be divergent ( in other words , @xmath30 should be zero ) at the critical point and in the thermodynamical limit . in the chiral limit ,
the chiral susceptibility in the nambu phase is difficult to obtain , but it is calculable in the chiral symmetric phase @xcite .
it can be shown that in the symmetric phase @xmath30 , defined in eq .
( [ def ] ) , is the same as @xmath31 where @xmath32 are the positive eigenvalues of the massless fermionic matrix . approaching the * ne * line from the symmetric phase by fixing @xmath13 , @xmath33 should behave as @xmath34 corresponding to the divergent correlation length at the second order phase transition point in the thermodynamical limit @xmath26 . in the nambu phase
, it can also be shown that eq .
( [ order ] ) is equivalent to @xmath35 in the @xmath26 limit .
then in such a limit , @xmath33 should be zero since @xmath36 in the nambu phase .
therefore , @xmath33 defined in eq .
( [ order ] ) is a suitable order parameter for the chiral phase transition : it is zero in the broken phase , and it is nonzero in the symmetric phase .
let us again focus on the results at @xmath20 . to perform the calculation
, we generalize mfa @xcite , in which the chiral limit @xmath2 is accessible , to the fermion - gauge - scalar models . from fig .
[ fig2 ] , we observe that on @xmath37 the chiral transition appears at @xmath38 , being consistent with the observation of fig .
[ fig1 ] .
= 7.5 cm = 7.5 cm
the location of the * ne * line on the available lattices obtained by the above methods is summarized in fig .
[ fig3 ] , where the point * n * is plotted by interpolation .
we have determined the phase transition line * ne * with high precision and demonstrated that this second order chiral transition line joins the higgs phase transition line at the end point * e * being around @xmath39 , separating the higgs and nambu phases .
no finite size scaling analysis has been done , and larger lattices are required for such a purpose . from the spectroscopy @xcite , we know that @xmath40 scales to zero when crossing the chiral transition line * ne*. nevertheless , the susceptibility for @xmath9 and correlation length for the composite scalar @xmath41 remain finite on the whole * ne * line except approaching the end point * e*. therefore , the end point , hopefully being a second order point with divergent correlation lengths , is the most suitable candidate for the continuum limit .
further work to be done is to study the finite size effects , analyze the dependence of the end point * e * on the bare fermion mass , investigate the scaling properties and understand the nature of the end point , which is underway @xcite .
9 c. frick and j. jersk , these proceedings . c. frick and j. jersk , hlrz-94 - 52 . w. franzki and x.q .
luo , these proceedings .
v. azcoiti , private communications .
v. azcoiti , g. di carlo , a.f .
grillo , phys .
65 ( 1990 ) 2239 ; v. azcoiti , v. laliena , x.q .
luo , c.e .
piedrafita , g. di carlo , a. galante , a.f .
grillo , l.a .
fernandez and a. vladikas , phys .
d48 ( 1993 ) 402 .
w. franzki , c. frick , j. jersk and x.q .
luo , in preparation . | we report the recent results from the computer simulations of a fermion - gauge - scalar model with dynamical chiral - symmetry breaking and chiral transition induced by the scalar field .
this model might be considered to be a possible alternative to the higgs mechanism of mass generation .
a new scheme is developed for detecting the chiral transition .
our results show with higher precision than the earlier works that the chiral transition line joins the higgs phase transition line , separating the higgs and nambu ( chiral - symmetry breaking ) phases .
the end point of the higgs transition with divergent correlation lengths is therefore suitable for an investigation of the continuum limit . |
the kondo model and its periodic extension , the kondo lattice model ( klm ) , which describe spin - exchange interaction between a localized spin or a system of localized spins , respectively , and a band of itinerant electrons , has been subject of intense theoretical studies in the past @xcite .
this model has been applied to a variety of different problems in solid - state physics using both a ferromagnetic and antiferromagnetic coupling constant @xmath2 .
the model with @xmath0 is the one originally known as _ kondo lattice model _ or simply _ kondo model _ in its non - periodic form with a single impurity spin in the system .
it was used by kondo to explain the unusual temperature behavior of the resistivity of magnetic impurities in non - magnetic hosts @xcite .
the negative spin - exchange interaction can be derived from the hybridization of a correlated `` atomic '' level with a conduction band , the situation described by the anderson model @xcite . in the limit of a low - lying half - filled atomic level and strong correlations ,
the anderson model can be mapped onto the kondo model with a negative exchange constant @xcite .
the kondo lattice model is still subject to much theoretical work , the main objective is the understanding of the unusual physical behavior found in _ heavy - fermion _ materials @xcite . a model with identical operator structure in the hamiltonian , but with positive
exchange constant has been known in the literature for a long time by many different names ( double exchange model , @xmath3-@xmath4 model , @xmath3-@xmath5 model , ) @xcite . for clarity
, we will refer to this model in the following as _
ferromagnetic kondo lattice model_. the model with ferromagnetic exchange has to be understood as an effective one .
the origins of the exchange with @xmath1 are found in the interband coulomb correlations @xcite .
this situation is believed to dominate the physical properties of important systems such as the magnetic semiconductors @xcite ( eux ; x = o , s , se , te ) , the diluted magnetic semiconductors @xcite ( cd@xmath6mn@xmath7te , hg@xmath6fe@xmath7se ) , and the `` local moment '' metals @xcite ( gd , dy , tb ) .
to these problems , the ferromagnetic klm was successfully applied @xcite .
recently , this variant of the klm has gained a lot of interest with the discovery of the colossal magnetoresistance ( cmr ) materials @xcite . in these materials , typically manganese oxides with perovskite structure ( la@xmath6(ca , sr)@xmath7mno@xmath8 ) ,
the double - exchange model @xcite has been successfully applied to explain the origin of ferromagnetic order and is expected to be a good starting point to investigate the resistivity anomalies @xcite .
this double - exchange model , however , is nothing else than the kondo lattice model with ferromagnetic ( positive ) exchange constant in the strong coupling limit . in the cmr materials , the localized @xmath9-spin of the model represents the more or less localized manganese @xmath10-@xmath11 electrons , whereas the conduction band is formed by the @xmath12 electrons .
the interband - exchange interaction is nothing else but the intra - shell hund s rule coupling .
since the @xmath10-@xmath12 electrons of the manganese form a relatively narrow band ( theoretical results from band - structure calculations : @xmath13 @xcite and experimental estimates : @xmath14 @xcite ) and hund s coupling is assumed to be large , the model has to be taken in the intermediate to strong coupling regime .
there are few estimates about the value of the interaction constant in the literature , e.g. @xmath15 @xcite , but these are challenged as to be too small @xcite .
most theoretical papers of the last years concerned with colossal magnetoresistance assume classical spins @xmath16 @xcite .
this has been justified by the assumption of @xmath17 @xcite .
although it is true that the important energy scale is @xmath18 , there are much more implications of @xmath16 that are not justified in the strong - coupling limit for a @xmath9 system . in several papers
, it was stated that ``
the @xmath12 electrons are oriented parallel to the @xmath11 spins . ''
@xcite or equivalently `` so one only need consider configurations with @xmath12 electrons parallel to core spins . '' @xcite
. we will show below using exact results as well as several well - defined approximation methods , that for @xmath9 there is a considerable amount of spin-@xmath19 spectral weight located in the main region of the spin-@xmath20 states even for large interaction strengths .
the assumption of a half - metallic state @xcite , made in the two citations above can therefore never be met in the klm with quantum spins and is merely an effect of the ( unphysical ) limit of `` classical '' spins .
the recently discussed half - metallic behaviour of the manganites @xcite must have a different origin . however , for the opposite sign of @xmath2 , exactly the assumed effect happens in the strong - coupling limit : the lowest - lying excitations in the conduction band density of states will be purely spin-@xmath19 .
this already implies that results for the kondo lattice model with @xmath1 and @xmath0 can not simply be reverted into the respective other case .
the change of sign changes the whole physics of the system . for @xmath21 an antiparallel ( `` antiferromagnetic '' ) alignment of the conduction band spin and the localized spin lowers the internal energy . for a sufficient band filling
, this tends to a screening of the local moments by conduction electrons , well - known from the kondo effect that refers to a single magnetic impurity in a conduction electron sea . from this , the name `` kondo lattice model '' was originally derived for the @xmath0 case .
we will further show that already for comparatively low interaction strengths the spin - exchange interaction alone leads to an opening of a gap in the density of states .
this extraordinary correlation effect could give hints to the explanation of a recently discovered pseudogap in the managese oxides @xcite . to prove our claims already laid out so far
, we will first review two important non - trivial exactly solvable limiting cases of the kondo lattice model in sec .
[ sec : theo ] .
the first is the _ zero - bandwidth limit _ ( `` atomic limit '' ) where the bandwidth of the conduction band is set to @xmath22 @xcite .
the second exactly solvable limiting case is the so - called _ ferromagnetically saturated semiconductor _ @xcite .
this is essentially the zero - temperature limit of the model with vanishing electron density and fully aligned spin system . in this limit
, striking correlation effects can be observed and discussed .
these limiting cases will already give clear evidence to our propositions made above . in sec .
[ sec : dmft ] we will present a new _ dynamical mean - field theory _ (
dmft)-based approach for the klm with @xmath23 spins .
to circumstantiate our theory , we will introduce three more approximation schemes which also keep the spin as a quantum variable , not relying on the classical spin limit .
the first will be a _
second - order perturbation theory _ ( sopt ) for the klm based on the projector operator formalism @xcite .
the _ self - consistent cpa _
( coherent potential approximation ) is a straightforward extension to the well - known cpa for the klm @xcite .
it starts from the zero - bandwidth limit discussed before .
the third approximation method is a moment - conserving decoupling procedure for the equation of motion of the single - electron green function .
this approximation scheme continuously evolves from the exactly solvable limit of the ferromagnetically saturated semiconductor .
it will be called _ moment - conserving decoupling approximation _ ( mcda ) . the comparison of the results obtained by these three methods and our dmft scheme , each of which starts from a different limit allows to evaluate the range of applicability of the new approach , and to select the most trustworthy common features of all methods to gain a reliable picture of the physics of the ferromagnetic klm .
the kondo - lattice model ( or @xmath3-@xmath5 model ) traces back the characteristic features of the underlying physical system to an interband exchange coupling of itinerant conduction electrons to ( quasi- ) localized magnetic moments described by the following model hamiltonian @xcite @xmath24 @xmath25 ( @xmath26 ) and @xmath27 are , respectively , creation and annihilation operators of a band electron being specified by the lower indices .
the hopping integrals @xmath28 are connected by fourier transformation to the single - electron bloch energies @xmath29 .
the interband ( @xmath30 ) exchange with coupling strength @xmath2 is taken as an intra - atomic interaction between the conduction electron spin @xmath31 and the localized magnetic moment represented by the spin operator @xmath32 . for practical reasons
it is sometimes convenient to use the second quantization representation of the band electron spin @xmath31 which leads to the following form of the interband - interaction term : @xmath33 here we have used the abbreviations @xmath34 , @xmath35 and @xmath36 .
the first term in ( [ eq : hamil - sf_2 ] ) describes an ising - like interaction between the z - components of the spin - operators , while the second term incorporates spin exchange processes between the localized and the itinerant system .
the last two terms are an extension to the original model : @xmath37 leads to the correlated kondo lattice model , it introduces correlations between the conduction electrons in form of a hubbard - type interaction .
we will include this term in some parts of the discussion below .
the second term , @xmath38 represents a direct spin - exchange between localized moments on different lattice sites . although the @xmath30 exchange can lead to an effective rkky interaction between the spins for non - vanishing band occupation , a
`` superexchange''could also be included . in case of an empty conduction band ,
this term becomes essential as source of magnetic order .
@xmath39 are the superexchange integrals .
we state once more that the original kondo - lattice model ( or @xmath3-@xmath5 model ) is defined by @xmath40 , only .
the additional terms in eq .
( [ eq : hamiltonian ] ) are used as soon as physical requirements do not allow to neglect them .
if we are mainly interested in the conduction electron properties then the single - electron green function @xmath41 is of primary interest .
its equation of motion reads for the correlated klm @xmath42 the two types of interaction terms in ( [ eq : hamil - sf_2 ] ) let appear the `` spinflip function '' @xmath43 and the `` ising function '' @xmath44 , while the `` hubbard - function '' @xmath45 comes into play only when the `` hubbard - interaction '' ( last term in eq . ( [ eq : hamiltonian ] ) ) is switched on . the three `` higher '' green functions on the right - hand side of eq . (
[ eq : eom ] ) prevent a direct solution of the equation of motion .
a formal solution for the fourier - transformed single - electron green function , @xmath46 defines the electronic selfenergy @xmath47 by the ansatz @xmath48_-;c_{{{\bi{k}}}\sigma}^{\dagger } \rangle \rangle_e = \sigma_{{{\bi{k}}}\sigma}(e ) g_{{{\bi{k}}}\sigma}(e)$ ] for the general case @xmath47 can not be determined rigorously . before introducing some approaches to the not exactly solvable many - body problem of the klm
let us discuss in the next sections two rather illustrative limiting cases which can help to test the unavoidable approximations .
let us assume that the arbitrarily filled conduction band is shrinked to an @xmath49-fold degenerate level @xmath50 : @xmath51 .
nevertheless , we consider the @xmath5-spin system as collectively ordered for @xmath52 by any direct or indirect exchange interaction . in this case
, the hierarchy of equations of motion decouples exactly and can rigorously be solved @xcite .
the resulting energies and respective spectral weights are @xmath53 the `` hubbard - u '' in @xmath54 and @xmath55 indicates that these excitations are bound to a double occupancy of the respective lattice site . @xmath56 and @xmath57
appear when our `` test electron '' enters an empty site .
if it orients its spin parallel to the local @xmath5-spin then the energy @xmath56 is needed . in case of an antiparallel spin orientation a triplet or a singlet state
is formed .
the first requires the energy @xmath56 , the second @xmath57 .
the latter is therefore two - fold degenerate .
the spectral weights are , contrary to the energy levels , strongly dependent on the magnetization state of the @xmath5 system and the band filling . for a complete solution one needs the average occupation number @xmath58 and the mixed correlation functions @xmath59 and @xmath60 .
the evaluation can selfconsistently be done by use of the spectral theorem for the green functions @xmath61 , @xmath62 and @xmath63 ( cf .
it is interesting to observe that in any case from the four poles only three do appear . for less than half - filled bands and @xmath1 ( @xmath0 ) @xmath64 ( @xmath65 ) vanishes , and for more than half - filled @xmath66 ( @xmath67 ) does .
it should be mentioned that the spectral weights @xmath68 do not explicitly depend on the coupling constants @xmath2 and @xmath69 .
that means , on the one side , that even for @xmath70 the @xmath3-@xmath5 interaction produces a splitting into four not coinciding quasiparticle levels . on the other hand
, there is a striking dependence of @xmath71 and @xmath58 on the sign of the @xmath3-@xmath5 coupling .
that transfers to the spectral weights giving them some indirect dependence on @xmath72 . for @xmath73 and @xmath21
the order of the energy levels is different resulting via the spectral theorem in different correlation functions .
note that the mentioned dependence on @xmath2 concerns only the sign of @xmath2 .
the spectral weights are not influenced by the absolute value @xmath74 .
there is another very instructive limiting case that can be treated rigorously .
it concerns the situation of a single electron in an otherwise empty conduction band at @xmath75 , when the local moment system is ferromagnetically saturated . in this case
the coulomb interaction is meaningless , the `` hubbard - function '' @xmath76 is identical to zero . in the zero - bandwidth limit , discussed in the last section , for the @xmath20-spectrum all spectral weights disappear except for @xmath77 , while for the @xmath19-spectrum the levels @xmath56 and @xmath57 survive with @xmath78 and @xmath79 . for finite bandwidth
the mentioned special case is that of a ferromagnetically saturated semiconductor @xcite . in this case , the spin-@xmath20 quasiparticle density of states @xmath80 is only rigidly shifted compared to the `` free '' bloch density of states .
@xmath81 consequently , the quasiparticle dispersion is undeformed with respect to the bloch energies , @xmath82 .
spectral densities ( quasiparticle bandstructures ) of the ferromagnetically saturated semiconductor for @xmath23 and @xmath2 as indicated along several symmetry directions . the left ( right ) column displays the @xmath0 ( @xmath1 ) case , the top - most picture shows the interaction - free spectral density.,scaledwidth=60.0% ] the @xmath19-spectrum is more complicated because a @xmath19-electron has several possibilities to exchange its spin with the antiparallel @xmath5 spins . therefore , the @xmath83-spinflip function is not at all trivial .
however , its equation of motion decouples exactly , producing a closed system of equations which can be solved after fourier transformation for the single - electron green function .
the corresponding selfenergy @xmath84 reads : @xmath85 @xmath86 are the spin wave energies following from the heisenberg exchange @xmath38 ( cf .
( [ eq : hamiltonian ] ) ) , @xmath87 .
@xmath88 is the fourier transform of the exchange integral @xmath39 . usually the spin wave energies will be smaller by about two orders of magnitude than other typical energies of the system as the exchange constant @xmath2 or the bloch bandwidth @xmath89 .
as a general result the spectral density @xmath90 consists of two structures corresponding to special elementary excitation processes of the @xmath19 electron . there is a rather broad structure built up by `` scattering states '' which result from magnon emission by the original @xmath19 electron .
thereby the excited electron reverses its spin becoming a @xmath20 electron .
such a process is possible only if there are @xmath20 band states within reach for the original @xmath19 electron to land after the spinflip .
the scattering states therefore occupy the same energy region as the @xmath20-dos ( [ eq : ro_up_magpol ] ) .
there is another possibility for the @xmath19 electron to flip its spin .
it can also be done by a repeated emission and reabsorption of a magnon by the conduction electron resulting in a `` dressed '' particle propagating through the lattice accompanied by a virtual cloud of magnons . for not too small positive ( negative ) @xmath2 the energy of this `` dressed '' particle lies above ( below ) the scattering spectrum giving even rise to a bound state , i.e. to a quasiparticle with infinite lifetime which we call the `` magnetic polaron '' . outside the scattering region
the polaron peak manifests itself as a @xmath91-function .
as soon as the peak dips into the scattering part the polaron gets a finite lifetime after which it decays into a @xmath20 electron plus magnon .
figure [ fig : mag_pol_spec ] shows the down - spin quasiparticle bandstructure as derived from the respective spectral density as a density plot .
the degree of blackening is a measure of the spectral density magnitude .
sharp dark lines refer to pronounced peaks in the spectral density representing quasiparticles with long life - time . for weak coupling @xmath92 scattering processes smear out a little bit the `` free '' dispersion but do not lead to strong deformations .
however , already for moderate couplings @xmath93 one recognizes for some @xmath94 vectors the appearance of a sharp polaron dispersion . for @xmath1 ( right column )
the magnetic polaron is stable on the high - energy side of the @xmath20 spectrum , for @xmath0 on the low - energy side .
in addition the scattering spectrum is clearly visible taking away a great part of the total spectral weight . in the antiferromagnetic
klm the magnetic polaron represents the ground state configuration @xcite . for still rather moderate couplings of @xmath95 the polaron dispersion has split off over the full brillouin zone .
the magnetic polaron has an infinite lifetime .
it is surprising that even the broad scattering structure is obviously bunched together as if it were a rather stable quasiparticle .
it is noteworthy to repeat that the results of fig .
[ fig : mag_pol_spec ] are exact and free of any uncontrollable approximation .
so we have to expect these quasiparticle effects in real systems , too .
this holds also for the quasiparticle density of states ( qdos ) plotted in fig .
[ fig : mag_pol_dos ] for several exchange couplings @xmath74 . according to eq .
( [ eq : ro_up_magpol ] ) the @xmath20-qdos is only rigidly shifted to higher ( lower ) energies for @xmath0 ( @xmath1 ) .
correlation effects appear exclusively in the @xmath19 spectrum . for @xmath96 a band splitting sets in .
one of the subbands occupies the same energy region as @xmath80 being therefore built up by the mentioned scattering states . in the ferromagnetic ( @xmath1 ) klm it is the low energy part of the spectrum containing a considerable amount of @xmath19-spectral weight .
this is not a specialty of the @xmath23 case or of weak and moderate couplings exhibited in fig .
[ fig : mag_pol_dos ] but holds equivalently , e.g. , for @xmath9 and in the strong coupling region ( @xmath97 , see last row in fig .
[ fig : mag_pol_dos ] ) .
after the band splitting has set in the weights of the two spin-@xmath19 subbands are close to the zero - bandwidth values @xmath98 for the lower , and @xmath99 for the upper part ( @xmath1 ) , i.e. independent of @xmath2 .
the very often used assumption , when the klm is applied to the manganites , that the @xmath12 electron orients its spin parallel to the @xmath11-@xmath9 spin @xcite , appears with respect to the exact results in fig .
[ fig : mag_pol_dos ] rather questionable . in the antiferromagnetic ( @xmath0 )
klm the low energy quasiparticle subband consists of stable polaron states which build the high energy part in the ferromagnetic ( @xmath1 ) klm .
although the @xmath1-qdos and the @xmath0-qdos appear as mirror - images the implied physics turns out to be rather different .
it is not possible to directly apply the methods of the dynamical mean - field theory @xcite to the klm .
one exception is the case of the classical - spin limit ( @xmath16 ) @xcite , thus removing the quantum nature of the spins .
the effect of quantum mechanics is strongest for @xmath23 , but also for @xmath100 , they dramatically influence the spectrum @xcite . the assumption of classical spins for @xmath9 , as done in refs .
@xcite , needs therefore careful analysis of the neglected effects .
another possibility to derive a dynamical mean - field theory for the kondo lattice model is the fermionization of the localized spin operators as suggested in @xcite .
this approach is , however limited to @xmath23 systems ( see also @xcite ) .
the first part of our approach will closely follow ref .
@xcite , but as discussed below , will then differ from the cited reference .
the localized spins in the hamiltonian ( [ eq : hamiltonian ] ) can be expressed in terms of auxiliary fermion operators @xmath101 ( @xmath102 ) @xcite : @xmath103 here , @xmath104 represents the pauli matrices .
this is the same transformation that led to ( [ eq : hamil - sf_2 ] ) where it was applied to the conduction electron spin @xmath31 .
the `` fermionization '' according to eq .
( [ eq : fermionization ] ) implies the introduction of the constraint : @xmath105 the inclusion of this constraint via a lagrange formalism corresponds to the addition of a hubbard - like interaction term for the @xmath5-fermions to the hamiltonian with the interaction constant @xmath106 @xcite .
the one - particle energy of the @xmath5-fermions is located at @xmath107 .
the `` fermionized klm '' takes the form : @xmath108 this hamiltonian describes a system of two different kinds of @xmath23 fermions coupled by a local spin - exchange interaction .
the @xmath5-fermions are additionally correlated via the hubbard - type of interaction to prevent double - occupancy .
this hamiltonian resembles the periodic anderson model ( pam ) @xcite in some way .
in fact the only difference in the operator structure is the `` coupling '' between conduction band and @xmath5 state . in the pam ,
this is simply a kinetic - energy term ( `` hybridization '' ) , whereas here , the two subsystems are coupled by the spin exchange . essentially , the `` fermionized klm '' is a rudimentary version of the general multi - band hubbard model with ( local ) inter - band interaction . for this model ,
the dmft is discussed by @xcite .
it is straightforward to apply the standard methods of the dmft to map model ( [ eq : klm+u ] ) onto an appropriate impurity model with the corresponding self - consistency condition to determine the parameters of the impurity model .
this leads to the following hamiltonian for the impurity model ( single - site kondo model , sskm ) with fermionized @xmath5-spins : @xmath109 with only one `` impurity '' site for the @xmath5-fermions .
the spin - exchange interaction acts only at this single site denoted by the site - index @xmath110 .
it is advisable to express the conduction band part of hamiltonian ( [ eq : sskm-1 ] ) in local space and single out the operators referring to the impurity site @xmath110 : @xmath111 for better readability , we will denote the construction operators at the site @xmath110 with the symbol @xmath112 ( @xmath113 ) , the on - site energy will be denoted @xmath114 .
finally we introduce a unitary transformation which diagonalizes the last term of the second line in eq .
( [ eq : transformek ] ) .
the transformed fermion operators will be denoted @xmath115 ( @xmath116 ) , the hopping @xmath117 and finally @xmath118 for @xmath119 . in the context of the dmft
, one does not need to know the explicit structure of this unitary transformation since @xmath120 and @xmath121 need not be known explicitly either .
the parameters of the single - site kondo model will be determined by the self - consistency condition of the dmft ( see below ) .
a direct solution of model ( [ eq : sskm-1 ] ) is , to our knowledge not possible .
however , it can be further simplified : the hubbard term originating from the constraint ( [ eq : constraint ] ) can be eliminated by simply reversing the `` fermionization '' procedure which led from hamiltonian ( [ eq : hamiltonian ] ) to ( [ eq : klm+u ] ) .
the auxiliary fermion operators get replaced by a local spin-@xmath122 operator at the impurity site , and the constraint ( [ eq : constraint ] ) can be dropped .
this leads to the final version of the single - site kondo model : @xmath123 this last step , which could be called `` de - fermionization '' , distinguishes our approach from the one used in ref .
@xcite and others .
this `` de - fermionization '' ensures the exact fulfillment of the constraint ( [ eq : constraint ] ) , which could for example only be kept on average ( @xmath124 instead of @xmath125 ) in @xcite . the parameters determining the conduction band in the hamiltonian ( [ eq : sskm ] ) , namely @xmath120 and @xmath121 have to be specified according to the dmft self - consistency condition : the local conduction band green function ( cf .
( [ eq : gf ] ) ) should be equivalent to the @xmath4-operator green function of the single - site model , @xmath126 : @xmath127 where the right equation follows from the formal solution of the equation of motion of @xmath128 .
so instead of the usual definition for the hybridization function , @xmath129 , it has to be determined so that eq .
( [ eq : seco ] ) holds .
one will see below that the knowledge of @xmath130 , which can become spin - dependent through this procedure , is sufficient to solve the single - site kondo model ( [ eq : sskm ] ) .
its dispersion @xmath120 and the hybridization parameter @xmath121 need not to be determined explicitly .
the dmft , i.e. the mapping of the klm onto the single - site model is , except for the limit of infinite spatial dimensions , an approximation , equivalent to the local approximation . in the exactly solvable case of the ferromagnetic semiconductor ( cf .
[ sec : magpol ] ) this is equivalent to neglecting the magnon energies which are assumed to be at least one order of magnitude smaller than the energy scales under consideration here , e.g. bandwidth or @xmath2 .
next , we will introduce an approximative method to solve the single - site kondo model defined by hamiltonian ( [ eq : sskm ] ) for an arbitrary hybridization function @xmath130 . in the following section , we derive an equation of motion - based method to solve the single - site kondo model ( [ eq : sskm ] ) .
since in this section we deal only with this model , we suppress all subscripts distinguishing between quantities in the lattice and the single - site model .
starting point is the equation of motion for the @xmath4-green function : @xmath131 where the higher green functions @xmath132 and @xmath133 , corresponding to the green functions @xmath134 and @xmath135 for the lattice case , are introduced .
the `` mixed '' green function @xmath136 can be eliminated by investigating its equation of motion : @xmath137 thereby defining the hybridization function @xmath138 .
this yields the final equation of motion : @xmath139 eq .
( [ eq : eomgd ] ) looks , except for the @xmath140 in the prefactor of @xmath128 , like the equation of motion of the zero - bandwidth limit .
the hybridization function @xmath140 is due to the @xmath121-term in the hamiltonian ( [ eq : sskm ] ) .
this term prohibits an exact solution .
in fact , the `` hybridization '' via @xmath121 is nothing else than the inter - site hopping which was neglected in the zero - bandwidth limit .
this term will force us to make certain approximations in the determination of the higher green functions on the right - hand side of eq .
( [ eq : eomgd ] ) .
we will exemplify this using the @xmath141-function .
its equation of motion reads @xmath142 where on the one hand , higher `` impurity - site '' green functions are introduced , but on the other hand also a higher `` mixed '' green function , @xmath143 . in analogy to the one - particle mixed green function , eq .
( [ eq : mixedgd ] ) , we use the following substitution : @xmath144 the justification of this procedure can be found in analogy to the `` self - energy substitution '' known from other approximation methods for the klm ( see e.g. @xcite ) by inspecting the spectral representations of the relevant green functions .
this reveals that all of them have the same single - particle poles , they differ only in the respective weights given by matrix elements of the type @xmath145 . here
@xmath146 are the energy eigenstates and a represents one of the relevant operators : @xmath112 , @xmath115 , @xmath147 or @xmath148 . from eq .
( [ eq : mixedgd ] ) follows that the hybridization function accounts for the differences of the matrix elements of the first two operators ( @xmath112 and @xmath115 ) . assuming that these differences are almost equal to those of the matrix elements built up by the latter two operators leads to the _ hybridization approximation _ ( [ eq : hybridapprox ] ) .
this is the only approximation necessary to decouple the hierarchy of equations of motion . in the case of @xmath23 spins
, there are only 6 different `` impurity - site '' green functions , whose equations of motion form a closed set of equations . besides the already introduced green functions @xmath128 , @xmath141 and @xmath149 , these are @xmath150 several expectation values , introduced into the theory via the inhomogeneities of the equations of motion can be expressed in terms of the above - mentioned green functions , a self - consistent solution has to be found . at this point ,
let us comment on the reliability of this approximation .
although only one approximation enters our decoupling procedure ( cf .
( [ eq : hybridapprox ] ) ) , it still has to be seen as an uncontrollable approximation meaning that there is no true small parameter .
there are two non - trivial limiting cases where the replacement ( [ eq : hybridapprox ] ) becomes exact : the first is the limit @xmath151 , representing the zero - bandwidth klm of sec .
[ sec : atomic ] .
but already a small , but finite bandwidth in the klm could lead to any ( unknown ) expression for @xmath140 implying that we can not necessarily assume @xmath121 to be small any more .
the second limit is the `` classical spin '' limit where the `` spin variable '' @xmath152 has no operator properties any more . here , the replacement of eq .
( [ eq : hybridapprox ] ) reduces to ( [ eq : mixedgd ] ) .
however since we are interested in the general case with finite @xmath152 and bandwidth , we have to confirm the trustworthiness of this approximation by a comparison with other well - tested methods .
we are now going to introduce three further approximation methods for the klm .
these are known from literature , their strengths and weaknesses have been identified .
the three methods differ substantially with respect to the theoretical assumptions made for an approximate solution of the klm .
common features following from these procedures and the above - introduced dmft scheme should then give some credit of reliability , in particular when they additionally fit the exact limiting cases discussed in sec .
[ sec : theo ] .
an application of the usual diagrammatic perturbation theory to the kondo model can not be performed due to the absence of wick s theorem . only for low temperatures ( spin - wave approximation )
@xcite or in the classical - spin limit @xcite , this method is applicable .
the projection operator formalism of mori @xcite is better suited for the kondo model .
it has been used successfully to describe correlation effects in the hubbard model in the weak coupling regime @xcite . the general formula for the second - order contribution @xmath153
can be found there ( eq . ( 3.12 ) in ref .
@xcite ) . to allow for a better comparison with the other approximations in this paper , we further approximate the self - energy taking only @xmath94 averaged occupation numbers into account ( local approximation ) .
next , we want to introduce a modification of the well - known `` coherent potential approximation '' ( cpa ) @xcite for the klm .
the cpa is a standard many - body approach that starts from a fictitious alloy in analogy to the interacting particle system .
starting point may be the zero bandwidth limit of sec .
[ sec : atomic ] .
we think of a four - component alloy each constituent of which is characterized by one of the energy levels @xmath154 in eq .
( [ eq : atomic_energies ] ) .
the spectral weights @xmath68 ( [ eq : atomic_energies ] ) are then to be interpreted as the `` concentrations '' of the alloy components as seen by a propagating @xmath155-electron .
the fictitious alloy for a @xmath155-electron is built up by the local moments and by the frozen ( @xmath156 ) electrons .
the cpa - selfenergy of the @xmath155-electron is found by the well - known formula @xcite .
@xmath157 as a consequence of the single - site aspect of the cpa the resulting selfenergy is wave - vector independent . according to eq .
( [ eq : atomic_energies ] ) the `` concentrations '' @xmath68 depend on a sum of the `` higher '' correlation functions @xmath62 and @xmath63 , which can rigorously be expressed by the single - electron green function .
@xmath158 the shortcomings of the cpa - procedure lie on hand .
the one is the same as that in the conventional alloy analogy of the hubbard model , namely the assumption of frozen ( @xmath156 ) electrons .
this is partially removed by our proposed modification of the standard cpa procedure for the klm , namely the selfconsistent calculation of the higher correlation functions @xmath159 and @xmath160 via eq .
( [ eq : spectheo_delta ] ) as well as the band - occupation @xmath161 via the spectral theorem for the one - electron green function .
maybe even more serious in the case of the klm is the blocking of repeated spin exchange with the local moment system .
magnon emission or absorption is not involved .
so we can not expect that the cpa - treatment correctly reproduces the exact limiting case of sec .
[ sec : magpol ]
. however , some general information about the quasiparticle bandstructure might be possible , in particular in the strong coupling ( `` split band '' ) regime . by construction
the method yields the correct zero - bandwidth limit .
a green function method which takes the spin dynamics correctly into account has been proposed in ref .
@xcite . for details about this approach , we refer the read to the cited paper , here we summarize the result shortly this decoupling approach yields finally a selfenergy of the following structure : @xmath162 the first term is linear in the coupling @xmath2 and proportional to the @xmath163 magnetization @xmath164 .
it just represents the result of a mean - field approximation being correct in the weak - coupling limit .
the second term in ( [ eq : sestruc ] ) contains all the spin exchange processes which may happen .
it is a complicated functional of the selfenergy itself .
so ( [ eq : sestruc ] ) is not at all an analytical solution but an implicit equation for the selfenergy .
we do not present here the lengthy expression for @xmath165 referring the reader for further details to ref .
it should be mentioned , however , that @xmath165 contains several expectation values which must be fixed to get a self - consistent solution .
no problems arise with the mixed correlation functions @xmath59 and @xmath166 .
they can rigorously be expressed by the spin flip function @xmath63 and the ising function @xmath62 defined previously .
both functions are already involved in the procedure , so that no further approximations are necessary to fix @xmath167 and @xmath168 . for pure local - moment correlations such as @xmath169 , @xmath170 , , however , a special treatment is necessary , e.g. as described in ref . @xcite .
in the following section , we discuss the results obtained for the dmft and the three approximation schemes of sec .
[ sec : approx ] for the ferromagnetic kondo lattice model with @xmath23 .
since we are interested in the reaction of the conduction band due to the magnetic order of the spin system , we have not calculated the latter self - consistently .
instead , we have simulated the magnetic order by determining @xmath171 using a brillouin function .
temperatures are given in units of @xmath172 .
within the cpa , our choice of @xmath171 can lead to unphysical results .
namely in the case of @xmath173 , some of the weights ( [ eq : atomic_energies ] ) can become negative .
there is an upper bound for @xmath164 @xcite .
we therefore had to limit @xmath164 to @xmath174 for some of the cpa calculations . within the dmft calculations , we experienced severe numerical problems which forced us to introduce a further approximation : `` mean - field''-decoupling the @xmath175 and @xmath176 functions simplifies the system of equations of motion : @xmath177 all dmft - results presented below were obtained using the hybridization approximation in combination with this unrestricted - mean - field decoupling of @xmath175 and @xmath176 .
as function of energy .
the temperature is set to @xmath178 , therefore @xmath179 ( cpa : @xmath180 , see text ) and the electron density @xmath181 , the chemical potential is at @xmath182 ; solid line : spin-@xmath20 , dotted line spin-@xmath19 dos , scaledwidth=70.0% ] in all calculations , the conduction band is described by a tight - binding dos for a simple - cubic lattice structure @xcite of unit width ( @xmath183 ) .
the curie temperature is taken as @xmath184 .
the quasiparticle densities of states ( dos ) for quarter - filling and different values of @xmath2 are plotted in figs .
[ fig : sf_t0 ] and [ fig : sf_tc ] for @xmath178 and @xmath185 , respectively .
as indicated , the columns correspond to dmft , mcda , cpa and sopt , respectively ( from left to right )
. we will begin the discussion with the dmft results shown in the left column of fig .
[ fig : sf_t0 ] .
a small value of @xmath2 leads to a spin - dependent shift of the spin-@xmath20 and @xmath19 dos , as one would obtain by a simple mean - field decoupling , i.e. by replacing @xmath186 by its mean value @xmath164 in hamiltonian ( [ eq : hamiltonian ] ) . with increasing @xmath2 ,
the dos show some striking correlation effects : first a broadening , later the onset of a splitting of the band can be observed .
whereas in general , the correlation effects are stronger for spin @xmath19 ( indicated by a stronger quasiparticle damping ) , the splitting is , in contrast to the other two methods , more pronounced in the spin-@xmath20 dos .
however , here the split - of ( upper ) peak has much less spectral weight than the original peak which , except for a band - narrowing still resembles strongly the free conduction band dos .
the picture is very similar in the mcda . again , a mean - field like spin - dependent bandshift is observed for small @xmath2 . on increasing @xmath2 , the spin-@xmath19 dos broadens , and
a two - peak structure emerges .
the spin-@xmath20 dos remains , except for a small tail at its upper edge , unchanged .
this behavior can easily be understood by comparing with the special case of the ferromagnetically saturated semiconductor as discussion in sec .
[ sec : magpol ] since the mcda develops continously into this special case for @xmath187 and @xmath188 .
the spin-@xmath19 dos splits into a scattering part ( low energies ) and the polaron - like part at higher energies above the conduction band . unlike the @xmath189 case ,
however , there are some , but weak modification in the @xmath20 dos , namely the above mentioned tail at its upper edge .
this can be interpreted as a scattering contribution for spin-@xmath20 electrons .
the origin of this is the finite number of spin-@xmath19 electrons . , but for @xmath185 , therefore @xmath190.,scaledwidth=70.0% ] also the cpa dos show a remarkably similar picture as the previously discussed theories .
again , there is the mean - field shift for small @xmath2 . for larger values of @xmath2 , the splitting of the spin-@xmath19 dos sets in similarly to the other two methods . here , this can be contributed to the two single - occupancy quasiparticle energies @xmath56 and @xmath57 known from the zero - bandwidth limit ( cf . sec .
[ sec : atomic ] ) . for @xmath191
a third quasiparticle energy carries non - vanishing weight , namely the @xmath55 peak .
this is located between the two other peaks since we switched off the conduction band coulomb interaction @xmath69 ( cf .
( [ eq : hamiltonian ] ) ) . for @xmath192 , the appearance of this third peak in between the two main peaks is vaguely visible .
a remarkable difference to the dmft and mcda results is the fact , that the spin-@xmath20 and scattering part of the spin-@xmath19 dos do not cover the same energy range .
this shortcoming of the cpa is due to the neglection of spin - dynamics ( magnons ) as mentioned in sec .
[ sec : self - consistent - cpa ] ( cf.ref .
@xcite ) .
the sopt results are , for the plotted values of @xmath2 in the weak- and intermediate coupling regime , not too far away from the other results . for small @xmath2 , where the sopt becomes by definition reliable , the mean - field shift as in the other methods is clearly observable .
similar to the other methods , the deformation of the spin-@xmath19 dos is much stronger than that of the spin-@xmath20 dos .
however , for larger @xmath2 , the sopt never shows a true band splitting , its range of validity is certainly restricted . for @xmath185 ,
where spin - symmetry is re - established , the dmft , mcda and cpa give a similar overall picture .
the resulting dos shows , already for @xmath193 , a two - peak structure ( dmft and mcda ) , in the cpa a third peak is dimly noticeable .
the most remarkable observation is the near - coincidence of the dmft and mcda results for all @xmath2 values .
the transition from the ( clearly distinct ) @xmath178 dos to the ( resembling ) @xmath185 dos is continuous for both theories .
the sopt , however , has to be seen as a complete failure for a paramagnetic system already for relatively small @xmath194 .
this can already be read of the formula for calculating the self - energy , which is more or less trivial .
the corresponding results can not be connected to any of the exactly solvable limiting cases ( cf .
[ sec : sopt ] ) .
a sound - standing interpretation of the observations is not difficult since the mcda can be traced back to the exactly solved limiting case of the ferromagnetically saturated semiconductor ( cf.sec .
[ sec : magpol ] ) and the cpa to that of the zero - bandwidth limit discussed in sec .
[ sec : atomic ] .
the sopt , of course , becomes reliable for small @xmath2 .
although all four presented methods are of approximate nature and their results do , at least for intermediate - to - large values of @xmath2 , differ , some common properties emerge : for small @xmath2 , the behavior is genuine mean - field like , a bandshift proportional to @xmath195 is observed . for larger @xmath2 ,
the onset of a band splitting occurs . at @xmath178
this primarily affects the spin-@xmath19 dos .
this fact is understandable by examination of the ferromagnetically saturated semiconductor where the band splitting was discussed in terms of a scattering band and the magnetic polaron . in that case
the spin-@xmath20 dos remains , except for a simple shift , unaffected by the interaction simply because spin - flip of spin-@xmath20 electrons is suppressed in this case .
now for finite @xmath196 , this is only approximately true .
the spin-@xmath19 dos still shows strong correlation effects , but also the spin-@xmath20 dos is affected due to the finite number of spin-@xmath19 electrons in the system .
this manifests itself differently in the various methods : in the mcda , a tail is seen at the upper edge of the spin-@xmath20 dos , in the cpa , a shoulder develops , and in the dmft approach , a band splitting is observable . from this
we conclude that correlation effects are much more pronounced in the dmft than in the other two theories . at @xmath185 , the dip indicating the onset of the band splitting is still existing for all but the sopt method . in cpa , mcda and
the dmft results this splitting is of similar size , which can easily be read off the cpa where it is simply given by the difference of the respective energies from the zero - bandwidth limit ( see eq . ( [ eq : atomic_energies ] ) ) .
the two most pronounced peaks correspond to the single - occupancy quasiparticle energies @xmath56 and @xmath57 , the band splitting is therefore approximately @xmath197 .
this is in contrast to the results obtained by dynamical mean - field theory for the @xmath16 klm ( klm with classical spins ) .
the emerging picture for classical spins is the following @xcite : at @xmath178 , the dos is characterized by a mean - field like `` zeeman '' splitting between the bands of both spin directions . with increasing temperature , at each of the respective spin-@xmath20 or @xmath19 band , spectral weight of the opposite spin direction appears , until finally at @xmath185 , the system becomes paramagnetic .
the splitting into two subbands separated by @xmath198 stays constant .
comparing our results with the @xmath16 ones , phenomenologically the dos s show completely different characteristics at @xmath178 , especially in the spin-@xmath19 band . for @xmath185 , the dos
s look quite similar .
however , the physics behind the scenes turn out to be completely different as can be seen , e.g. , by the inconsistent size of the band splitting @xmath199 .
whereas in the classical - spin limit , the splitting is always due to a mean - field like `` on - site zeeman splitting '' , in the case of quantum spins , the various elementary excitations as discussed in the case of the ferromagnetically saturated semiconductor , are responsible for the sub - band structure . for the ferromagnetically saturated system ,
the neglection of spin - flip processes and magnons in the @xmath16 calculation lead to the picture of a half - metal @xcite .
this does not apply for any finite value of @xmath152 and is therefore clearly an artifact of the @xmath16 limit .
let us shortly remark on the situation with anti - ferromagnetic coupling ( @xmath0 ) @xcite .
all four approximation methods presented here do not show any signs of kondo screening .
an investigation of the special low - temperature physics of the model with @xmath0 is therefore not possible .
however , some remarks about the behavior of the model for @xmath200 can be made : in general the excitation spectra are broader than in the @xmath1 case .
the resulting exchange splitting of @xmath201 is also larger than for @xmath1 which can already be seen in the zero - bandwidth limit .
this is again in sharp contrast to the @xmath202 results , where the size of the splitting is independent of the sign of @xmath2 .
in this paper , we investigated the kondo lattice model ( klm ) focusing on the model with positive ( ferromagnetic ) exchange constant @xmath2 .
we discussed two exactly solvable , but nevertheless non - trivial limiting cases as well as four different approximation methods .
the results obtained by these methods , maybe except for the perturbation theory , compare generally reasonably well .
sometimes even nearly perfect matches occur ( cf . fig .
[ fig : sf_tc ] ) . in general , the differences between the methods are smaller for the paramagnetic than the ferromagnetic system . from the comparison of common features of these approximation methods in combination with the exact results , the following picture emerges for the quasiparticle structure of the ferromagnetic kondo lattice model : for small @xmath74 , a mean - field like shift of @xmath20 and @xmath19 dos is visible .
already for intermediate coupling strengths , a band splitting of size @xmath203 occurs .
the relevant energy scale for the splitting is @xmath204 . in the case of ferromagnetic saturation ( @xmath178 ) , the splitting is more pronounced in the @xmath19 than in the @xmath20 dos .
this band - splitting should not be confused with the splitting found in the limit of `` classical spins '' ( @xmath16 ) . there , the splitting is simply due to a mean - field like zeeman - splitting and therefore of the size @xmath198 .
contrary to that , our results clearly show that the two emerging subbands can be traced back to the two elementary excitations known from the limit of the ferromagnetically saturated semiconductor ( see above ) .
the magnetic saturation of the spin - system suppresses these processes for spin-@xmath20 electrons which explains the stronger footprint of the correlations in the spin-@xmath19 dos . for finite temperatures ,
the magnetic polaron generally remains a well defined quasiparticle , represented by a rather sharp lorentzian peak in the spectral density .
now the spin-@xmath20 electrons can also participate in spin - exchange processes since the localized spins are not fully aligned any more .
the spin - symmetric dos at @xmath205 also show the characteristic splitting .
our results also confirm the fundamental differences between the @xmath1 and the @xmath0 case of the kondo lattice model : it can be read of both exactly solvable limiting cases that the ground state will be different depending on the sign of @xmath2 .
the approximative approaches of secs .
[ sec : dmft ] and [ sec : approx ] do not allow a deeper investigation of the model with @xmath0 due to the inability to reproduce the special low - temperature properties of that model ( `` kondo physics '' ) .
another important conclusion from our calculations can be drawn : there is always finite spin-@xmath19 spectral weight in the region of the spin-@xmath20 dos .
this spectral weight does not disappear in the limit @xmath206 but only in the limit @xmath16 ( `` classical spins '' ) .
for the ferromagnetic klm ( @xmath1 ) , this implies that the klm for @xmath9 and large @xmath2 , corresponding to the situation found in manganites , will not be a half - metal for @xmath178 , contrary to the predictions of @xmath16 calculations . financial support by the _ volkswagen - foundation _ within the project _
`` phasendiagramm des kondo - gitter - modells '' _ is gratefully acknowledged .
this work also benefitted from the financial support of the _ sonderforschungsbereich sfb 290 _ ( `` metallische dnne schichten : struktur , magnetismus und elektronische eigenschaften '' ) of the deutsche forschungsgemeinschaft .
one of us ( d. m. ) wants to thank the _ friedrich - naumann - foundation _ for supporting his work . | a new `` dynamical mean - field theory '' based approach for the _ kondo lattice model _ with quantum spins is introduced .
the inspection of exactly solvable limiting cases and several known approximation methods , namely the _ second - order perturbation theory _ , the _ self - consistent cpa _ and finally a _ moment - conserving decoupling _ of the equations of motion help in evaluating the new approach .
this comprehensive investigation gives some certainty to our results : whereas our method is somewhat limited in the investigation of the @xmath0-model , the results for @xmath1 reveal important aspects of the physics of the model : the energetically lowest states are not completely spin - polarized .
a band splitting , which occurs already for relatively low interaction strengths , can be related to distinct elementary excitations , namely magnon emission ( absorption ) and the formation of magnetic polarons .
we demonstrate the properties of the ferromagnetic kondo lattice model in terms of spectral densities and quasiparticle densities of states . |
the standard picture for galaxies such as our own milky way is one where the visible matter is embedded in a halo of _ dark matter _ of as of yet unknown particle nature .
many theoretically well - motivated particle theories beyond the standard model predict the existence of particles that could be the dark matter . in many of these theories , such particles can pair - annihilate , producing , as a result , standard model particles that can be detected .
these detectable final states include antimatter particles , high - energy neutrinos , gamma rays , electrons and positrons ( @xmath0 ) , and secondary products of @xmath0 energy losses , such as radiation at x - ray frequencies from inverse compton and at radio frequencies from synchrotron .
collectively , the search for a signature from dark matter looking at its annihilation products is known as _
indirect detection_. the key question for a particle physicist is : can we do fundamental physics with indirect dark matter detection ? specifically
, i will address the sub - question of whether this is possible employing cosmic ray @xmath0 and gamma rays .
it is relevant to note that , historically , big leaps in our understanding of the underlying structure of particle physics models have been made possible by cosmic ray studies .
these leaps include the discovery of antimatter , of a second generation of matter fermions , of mesons , and , more recently , of neutrino masses and mixing .
indirect dark matter detection is a multi - wavelength and multi - messenger endeavor .
key to doing fundamental physics is to control astrophysical backgrounds to signatures possibly stemming from dark matter .
not surprisingly , the most exciting places to look for a dark matter signature correspond to some of the places with the `` worse '' astrophysical background . in this contribution
, i address the high - energy positron excess , dark matter annihilation in the galactic center , and the 130 gev line .
the pamela telescope has observed a significant excess of positrons in the 10 - 100 gev range , well above the standard galactic cosmic - ray expectations @xcite .
a notable recent result is the confirmation , by the fermi large area telescope ( lat ) , of this excess @xcite , a result which extends the rise in the positron fraction ( the fraction of positrons to electrons plus positrons as a function of energy ) up to @xmath1200 gev .
the notable experimental technique employed by the fermi - lat team hinges upon the use of the earth shadow , as a function of the cosmic ray @xmath0 arrival direction and energy : for particular arrival directions on the detector , electrons or positrons of given energy are completely forbidden , their trajectories , bent by the geomagnetic field , being shadowed by the earth .
the result of ref .
@xcite is notable from at least three standpoints : 1 . it confirms the pamela result , that was potentially plagued by positron mis - identification ( for every 50 gev cosmic - ray positron , one has about 10 electrons and @xmath2 protons , the latter being in principle easily mis - identified with positrons ) ; 2 .
it extends the pamela result to higher energy , with no evidence for a turnover or suppression in the source of high - energy positrons ; 3 .
it rules out certain models for the positron fraction excess , for example supernova remnant inhomogeneities in the vicinity of the sun @xcite , predicting a turnover of the positron fraction between 100 and 200 gev .
the question of the nature of the additional source of big - energy @xmath0 remains to be addressed : numerous studies have advocated dark matter annihilation , despite serious model building issues , including : 1 .
the need for very large annihilation cross sections , @xmath3 times larger than the natural scale needed for a standard thermal freeze - out scenario to produce dark matter with the correct relic density ; 2 . the need for rather large dark matter masses ( even larger after the fermi @xmath0 result ) ; 3 .
the need to suppress excess antiprotons and therefore to select very special annihilation final states , such as @xmath4 . despite these shortcomings ,
several possible scenarios have been proposed .
any such scenario is constrained by the necessary occurrence of final state radiation and of inverse compton from the @xmath0 needed in the annihilation final state to explain the positron fraction .
one of the best handles to constrain dark matter explanations of the high - energy positron excess is observing nearby galaxy clusters at gamma - ray frequencies , including the case of dark matter decay @xcite .
the size of clusters of galaxies is such that the energy loss time scale is much shorter than the time - scale associated to the diffusion of @xmath0 out of the system : clusters are thus perfect `` calorimeters '' for cosmic - ray @xmath0 in the 10 - 100 gev range @xcite . it should be noted that the resulting limits depend on assumptions on substructure , but provide rather stringent constraints on the range of masses and cross sections that could explain the @xmath0 excess @xcite .
additional constraints come from observations of the cosmic microwave background and of the diffuse isotropic extragalactic gamma - ray background ( see e.g. @xcite ) .
an alternative scenario , proposed among other studies in ref .
@xcite , postulates that the excess @xmath0 originate from existing nearby radio pulsars , in certain ranges of age and distance , and for plausible @xmath0 injection efficiencies .
the fermi - lat discovery of multiple radio - quiet pulsars has further strengthen the pulsar scenario , providing a number of previously unknown nearby pulsars that likely contribute to the local cosmic - ray @xmath0 population @xcite .
the pulsar interpretation of the @xmath0 thus remains a viable and probable explanation to the cosmic - ray positron excess , which is , in particular , compatible with the new fermi determination of the positron fraction @xcite .
can we discriminate between dark matter and pulsars as the origin of the @xmath0 excess ?
a smoking gun for the pulsar interpretation is the detection of an anisotropy in the arrival direction of the cosmic ray @xmath0 , especially if such anisotropy pointed towards a known pulsar @xcite . for dark matter , the general expectation is a high - latitude inverse compton emission from annihilation of dark matter in the halo , with a peculiar spectrum .
none of these smoking guns has been observed so far , although limits on the anisotropy of the @xmath0 arrival directions are close to the level expected if the excess were due to one single nearby pulsar @xcite . at present , both interpretations are compatible with data , and the origin of the positron excess remains still unsettled ( albeit with increasingly tight constraints on the dark matter explanation ) .
looking at the nearby future , the alpha magnetic spectrometer ams-02 @xcite , successfully launched and deployed on the international space station , is undergoing calibration and alignment .
the expectation from ams-02 is to deliver preliminary data on the positron fraction in early 2013 , with the next priorities antiproton data and high - z antimatter searches @xcite .
multiple studies have identified a possible signal from dark matter annihilation at the center of the galaxy using fermi - lat data ( see e.g. @xcite ) .
the results of these studies illustrate how crucial a proper modeling of the diffuse emission and of the emission from the very center of the galaxy is to reliably pinpoint an `` anomalous '' signal from dark matter .
an _ under - fitting _ of the astrophysical background leads to the possibly misleading identification of excesses to be attributed to dark matter .
the opposite issue is the _ over - fitting _ of the diffuse gamma - ray model , that would lead to fine - tuning the astrophysical background to incorporate at all costs any possible excess .
a key element in searching for dark matter in the galactic center is understanding the gamma - ray source associated with sgr a*. this source , if fueled by protons , can be responsible for a significant diffuse emission from the protons accelerated by the central supermassive black hole in the innermost galactic center region , which is of crucial importance to look for a signal from dark matter annihilation .
a careful treatment of the target density of gas in the region shows that a key diagnostic to unveil whether sgr a * accelerates protons is the gamma - ray source morphology @xcite .
only future cherenkov telescope arrays will have the needed angular resolution to pinpoint a hadronic origin for sgr a * , leading to possibly understanding some of the diffuse gamma - ray emission in the region @xcite .
if the dark matter interpretation of the diffuse excess in the galactic center is correct , a likely mass range is about 10 gev .
this mass range resonates with a series of recent direct and indirect dark matter detection results ( notably , recently , at radio frequencies @xcite ) .
albeit there is no conclusive evidence for new physics for any of these signals , this is an interesting possibility @xcite worth investigating on a case - by - case basis and definitely worth model - building efforts ( for recent examples see e.g. @xcite ) .
evidence for a monochromatic gamma - ray line at 130 gev has emerged from the analysis , in ref .
@xcite , of optimized signal - to - noise regions in the gamma - ray sky , at about the 3-@xmath5 level , once the look - elsewhere effect is taken into account @xcite . in ref .
@xcite , myself and linden pointed out that the relevant region of interest overlaps with the so - called fermi bubbles @xcite , which are thus an important source of background to the line .
it was also noticed that a broken power - law spectrum ( as what observed in the fermi bubbles @xcite ) could be mis - identified , or could create confusion , with a gamma - ray line @xcite .
it is important to note that the outline of the relevant regions for the fermi bubbles only partly overlaps with the region where most of the 130 gev photons originate from .
however , the fermi bubble regions were specifically outlined outside the galactic plane ( to suppress the bright galactic diffuse background ) , while the observed x - ray counterpart to the fermi bubbles @xcite indicates precisely that most of the emission associated with the mechanism fueling the bubbles should originate squarely from the galactic center region where the 130 gev line is most prominent .
it is currently unclear whether the 130 gev line can be associated with an instrumental effect ( for a recent study see e.g. @xcite ) perhaps issues with the energy - reconstruction algorithm employed by the current version of the fermi tools .
likely , the forthcoming `` pass 8 '' fermi tools software suite will address relevant aspects associated with instrumental systematics
. future observations with ground based cherenkov telescopes with a low energy threshold ( e.g. h.e.s.s .- ii and , further in the future , cta ) could also be decisive in deciding of the validity and nature of the 130 gev line observation .
the search for dark matter is in an extraordinarily exciting phase : * direct dark matter experiments are tapping into the exciting low - mass region , and carving into interesting regions of particle dark matter parameter space ; * the large hadron collider at cern is producing results that are sharpening the predictions of certain dark matter models ( e.g. universal extra dimensions ) and could soon provide clues on the correct beyond - the standard model framework ; * results from ams-02 will help understanding how to properly treat cosmic ray propagation , and will provide more information on cosmic ray antimatter , especially positrons , antiprotons and antideuterons ; * post - fermi - lat high - energy telescopes will cover the high - mass end of the particle dark matter parameter space ( cta ) , with the low - mass end possibly under scrutiny with the new hard x - ray facilities @xcite .
the indirect search for dark matter resonates with a famous quote by painter ren magritte : `` everything we see hides another thin , we always want to see what is hidden by what we see '' .
sp is partly supported by an outstanding junior investigator award from the us department of energy and by contract de - fg02 - 04er41268 , and by nsf grant phy-0757911 .
9 o. adriani _ et al . _
[ pamela collaboration ] , nature * 458 * , 607 ( 2009 ) [ arxiv:0810.4995 [ astro - ph ] ] .
m. ackermann _ et al .
_ [ fermi lat collaboration ] , phys . rev .
lett . * 108 * , 011103 ( 2012 ) [ arxiv:1109.0521 [ astro-ph.he ] ] .
n. j. shaviv , e. nakar and t. piran , phys .
lett . * 103 * , 111302 ( 2009 ) [ arxiv:0902.0376 [ astro-ph.he ] ] .
m. ackermann , m. ajello , a. allafort , l. baldini , j. ballet , g. barbiellini , d. bastieri and k. bechtol _ et al .
_ , jcap * 1005 * , 025 ( 2010 ) [ arxiv:1002.2239 [ astro-ph.co ] ] ; l. dugger , t. e. jeltema and s. profumo , jcap * 1012 * , 015 ( 2010 ) [ arxiv:1009.5988 [ astro-ph.he ] ] .
s. colafrancesco , s. profumo and p. ullio , astron .
astrophys . * 455 * , 21 ( 2006 ) [ astro - ph/0507575 ] .
s. profumo , phys .
d * 77 * , 103510 ( 2008 ) [ arxiv:0801.0740 [ astro - ph ] ]
. t. e. jeltema , j. kehayias and s. profumo , phys .
d * 80 * , 023005 ( 2009 ) [ arxiv:0812.0597 [ astro - ph ] ] . s. profumo and t. e. jeltema , jcap * 0907 * , 020 ( 2009 ) [ arxiv:0906.0001 [ astro-ph.co ] ]
. s. profumo , central eur .
j. phys .
* 10 * , 1 ( 2011 ) [ arxiv:0812.4457 [ astro - ph ] ] .
d. grasso _ et al .
_ [ fermi - lat collaboration ] , astropart . phys . *
32 * , 140 ( 2009 ) [ arxiv:0905.0636 [ astro-ph.he ] ] .
l. gendelev , s. profumo and m. dormody , jcap * 1002 * , 016 ( 2010 ) [ arxiv:1001.4540 [ astro-ph.he ] ] . m. ackermann _ et al .
_ [ fermi - lat collaboration ] , phys . rev .
d * 82 * , 092003 ( 2010 ) [ arxiv:1008.5119 [ astro-ph.he ] ] .
see http://www.ams02.org and http://ams.cern.ch .
l. goodenough and d. hooper , arxiv:0910.2998 [ hep - ph ] .
d. hooper and l. goodenough , phys .
b * 697 * , 412 ( 2011 ) [ arxiv:1010.2752 [ hep - ph ] ] .
d. hooper and t. linden , phys .
d * 83 * , 083517 ( 2011 ) [ arxiv:1011.4520 [ astro-ph.he ] ] .
t. linden , e. lovegrove and s. profumo , astrophys .
j. * 753 * , 41 ( 2012 ) [ arxiv:1203.3539 [ astro-ph.he ] ] . t. linden and s. profumo , arxiv:1206.4308 [ astro-ph.he ] . t. linden , d. hooper and f. yusef - zadeh , astrophys .
j. * 741 * , 95 ( 2011 ) [ arxiv:1106.5493 [ astro-ph.he ] ] .
d. hooper , a. v. belikov , t. e. jeltema , t. linden , s. profumo and t. r. slatyer , arxiv:1203.3547 [ astro-ph.co ] .
d. hooper , arxiv:1201.1303 [ astro-ph.co ] . m. s. boucenna and s. profumo , phys .
d * 84 * , 055011 ( 2011 ) [ arxiv:1106.3368 [ hep - ph ] ] . c. weniger , jcap
* 1208 * , 007 ( 2012 ) [ arxiv:1204.2797 [ hep - ph ] ] . s. profumo and t. linden , jcap * 1207 * , 011 ( 2012 ) [ arxiv:1204.6047 [ astro-ph.he ] ] .
m. su , t. r. slatyer and d. p. finkbeiner , astrophys .
j. * 724 * , 1044 ( 2010 ) [ arxiv:1005.5480 [ astro-ph.he ] ] .
d. p. finkbeiner , m. su and c. weniger , arxiv:1209.4562 [ astro-ph.he ] . t. e. jeltema and s. profumo , arxiv:1108.1407 [ astro-ph.he ] . | can we learn about new physics with astronomical and astro - particle data ? understanding how this is possible is key to unraveling one of the most pressing mysteries at the interface of cosmology and particle physics : the fundamental nature of dark matter .
i will discuss some of the recent puzzling findings in cosmic - ray electron - positron data and in gamma - ray observations that might be related to dark matter .
i will argue that recent cosmic - ray data , most notably from the pamela and fermi satellites , indicate that previously unaccounted - for powerful sources in the galaxy inject high - energy electrons and positrons .
interestingly , this new source class might be related to new fundamental particle physics , and specifically to pair - annihilation or decay of galactic dark matter .
this exciting scenario is directly constrained by fermi gamma - ray observations , which also inform us on astrophysical source counterparts that could also be responsible for the high - energy electron - positron excess .
observations of gamma - ray emission from the central regions of the galaxy as well as claims on a gamma - ray line at around 130 gev also recently triggered a wide - spread interest : i will address the question of whether we are really observing signals from dark matter annihilation , how to test this hypothesis , and which astrophysical mechanisms constitute the relevant background . |
let @xmath0 be a lie algebra over a field of characteristic zero .
for any functional @xmath1 there is an associated skew bilinear form @xmath2 on @xmath0 defined by @xmath3)$ ] for @xmath4 .
the index of @xmath0 is defined to be @xmath5 the lie algebra @xmath0 is _ frobenius _ if @xmath6 ; equivalently , if there is a functional @xmath1 such that @xmath7 is non - degenerate .
frobenius lie algebras were first studied by ooms in @xcite where he proved that the universal enveloping algebra @xmath8 is primitive ( i.e. admits a faithful simple module ) provided that @xmath0 is frobenius and that the converse holds when @xmath9 is algebraic .
the relevance of frobenius lie algebras to deformation and quantum group theory stems from their relation to the classical yang - baxter equation ( cybe ) .
suppose @xmath7 is non - degenerate and let @xmath10 be the matrix of @xmath11 relative to some basis @xmath12 of @xmath0 .
belavin and drinfeld showed that @xmath13 is a ( constant ) solution of the cybe , see @xcite .
thus , each pair consisting of a lie algebra @xmath0 together with functional @xmath1 such that @xmath2 is non - degenerate provides a solution to the cybe , see @xcite and @xcite for examples .
the index of a semisimple lie algebra @xmath14 is equal to its rank and thus such algebras can never be frobenius
. however , there always exist subalgebras of @xmath14 which are frobenius .
in particular , many amongst the class of _ biparabolic _ subalgebras of @xmath14 are frobenius .
a biparabolic subalgebra is the intersection of two parabolic subalgebras whose sum is @xmath14 .
they were first introduced in the case @xmath15 by dergachev and kirillov in @xcite where they were called lie algebras of _ seaweed _ type . associated to each seaweed algebra
is a certain graph called the _ meander_. one of the main results of @xcite is that the algebra s index is determined by graph - theoretical properties of its meander , see section [ sec : meander ] for details . using different methods
, panyushev developed an inductive procedure for computing the index of seaweed subalgebras , see @xcite .
in the same paper , he exhibits a closed form for the index of a biparabolic subalgebra of @xmath16 .
one may also see @xcite .
tauvel and yu found in @xcite an upper bound for the index of a biparabolic subalgebra of an arbitrary semisimple lie algebra , and they conjectured that this was an equality .
joseph proved the tauvel - yu conjecture in @xcite .
the methods of @xcite , @xcite , @xcite , and @xcite are all combinatorial in nature .
yet even with the this theory available , it is difficult in practice to implement this theory to find families of frobenius biparabolic lie algebras .
in contrast , for many cases it is known explicitly which biparabolic algebras have the maximum possible index .
for example , the only biparabolics in @xmath17 and @xmath16 which have maximal index are the levi subalgebras . in contrast
, the problem of determining the biparabolics of minimal index is an open question in all cases .
our focus in this note is on the seaweed lie algebras
these are the biparabolic subalgebras of @xmath17 .
the only known families of frobenius seaweed lie algebras that seem to be in the literature will be outlined in section [ sec : families ] , although the unpublished preprint @xcite may offer more examples .
we shall examine these families using the meander graphs of dergachev and kirillov .
our methodology provides new proofs that these algebras are indeed frobenius .
we also exhibit a new infinite family of frobenius seaweed lie algebras in section [ sec : submaximal ] .
in this section we introduce the seaweed lie algebras of @xcite . recall that a composition of a positive integer @xmath18 is an unordered partition @xmath19 .
that is , each @xmath20 and @xmath21 .
let @xmath22 be an @xmath18-dimensional vector space with a basis @xmath23 .
let @xmath19 and @xmath24 be two compositions of @xmath18 and consider the flags @xmath25 where @xmath26 and @xmath27 .
the subalgebra of @xmath17 preserving these flags is called a seaweed lie algebra and is denoted @xmath28 .
a basis - free definition is available but is not necessary for the present discussion .
the name seaweed lie algebra was chosen due to their suggestive shape when exhibited in matrix form .
for example , the algebra @xmath29 consists of traceless matrices of the form @xmath30 where the entries marked by the dots are zero .
many important subalgebras of @xmath17 are of seaweed type , as illustrated in the following example .
[ seaweed - examples ] * the entire algebra @xmath31 has index @xmath32 . *
the cartan subalgebra of traceless diagonal matrices is @xmath33 , where @xmath34 and has index @xmath32 .
* the borel subalgebra is @xmath35 and has index @xmath36 . *
a maximal parabolic subalgebra is of the form @xmath37 .
elashvili proved in @xcite that its index is @xmath38 .
the only explicitly known frobenius examples in the above list are the maximal parabolic algebras @xmath37 with @xmath39 and @xmath18 relatively prime .
of course , another infinite family of frobenius seaweed algebras occurs when @xmath40 , @xmath41 , and @xmath18 is odd .
a similar case is @xmath42 , @xmath43 , and @xmath18 is even .
these two families are detailed in @xcite .
a tantalizing question is how to classify which seaweed algebras are frobenius , especially given their importance in the general theory of lie algebras and applications to deformations and quantum groups .
as stated earlier , dergachev and kirillov have developed a combinatorial algorithm to compute the index of an arbitrary @xmath44 from its associated _ meander _ graph @xmath45 determined by the compositions @xmath46 and @xmath47 .
the vertices of @xmath48 consist of @xmath18 ordered points on a horizontal line , which can be called @xmath49 .
the edges are arcs above and below the line connecting pairs of different vertices .
more specifically , the composition @xmath50 determines arcs above the line which we will call the top edges .
the component @xmath51 of @xmath46 determines @xmath52 arcs above vertices @xmath53 .
the arcs are obtained by connecting vertex 1 to vertex @xmath51 , vertex 2 to vertex @xmath54 , and so on .
if @xmath51 is odd then vertex @xmath55 has no arc above it . for the component @xmath56 of @xmath57
, we do the same procedure over vertices @xmath58 , and continue with the higher @xmath59 . the arcs corresponding to @xmath60 are drawn with the same rule but are under the line containing the vertices .
these are called the bottom edges .
it is easy to see that every meander consists of a disjoint union of cycles , paths , and isolated points , but not all of these are necessarily present in any given meander .
the index of the lie algebra of seaweed type @xmath61 is equal to the number of connected components in the meander plus the number of closed cycles minus 1 .
the presence of the minus one in the theorem is due to our use of seaweed subalgebras of @xmath17 rather than of @xmath62 as used by dergachev and kirillov @xcite
. the index drops by one by the restriction to @xmath17 from @xmath62 .
[ modified ] figure [ meanderfig1 ] shows the meander @xmath45 corresponding to the compositions @xmath63 and @xmath64 .
we see that there is a single path and a single cycle . using the theorem above
, the index is @xmath65 .
hence , @xmath66 is not a frobenius algebra .
it is easy to see that to obtain a frobenius algebra , the only possibility for the meander is that it consist of a single path with no cycles and no isolated points .
the following illustrates this point .
[ examplea ] consider the algebra @xmath67 .
its meander is given in figure [ meanderfig2 ] .
labeling the vertices with @xmath68 from left to right , notice that @xmath69 is the single path @xmath70 ( if we start with @xmath71 ) or its reversal @xmath72 if we start with @xmath73 .
in particular , the index is @xmath74 and so this is a frobenius algebra .
what are the conditions on the compositions @xmath46 and @xmath47 so that the meander @xmath45 consists of a single path with no cycles or isolated points ? as stated , this seems to be an elementary question involving nothing more that the basics of graph theory .
however , the apparent simplicity of the question is misleading since an answer would provide a complete classification of frobenius seaweed algebras - a difficult problem .
even so , it is easy to give some necessary conditions on @xmath75 and @xmath76 for @xmath77 to be a single path .
for example , exactly two elements of the set @xmath78 must be odd .
this is because a path must have a starting and ending point , and these corresponds to vertices of degree one .
a vertex of degree one is either missing a top edge or bottom edge connecting to it , and this happens only if some @xmath59 or @xmath79 is odd .
another necessary condition for @xmath80 to be a single path is that @xmath81 . in this case ,
@xmath82 and thus @xmath83 is not frobenius since the index is additive for direct sums of lie algebras . more generally , if @xmath84 for some @xmath85 then the meander is not a single path .
other necessary conditions can be given , but none seems to shed light on what is sufficient .
in this section we revisit some known families of frobenius seaweed algebras in terms of meanders . at the end
we also provide two new families . first consider panyushev s example with @xmath86 , @xmath87 , and @xmath18 is odd . again , numbering as in example [ examplea ] , the top edges connect 2 to 4 , 4 to 6 , etc . while the bottom edges connect 1 to 3 , 3 to 5 , etc .
hence , the meander consists of the single path @xmath88 . a similar argument verifies that the meander for @xmath89 and @xmath90 with @xmath18 even is also the path @xmath91 .
suppose @xmath93 and @xmath94 are compositions of @xmath18 .
the modified meander @xmath95 is the graph @xmath96 appended with a loop corresponding to each odd @xmath59 and @xmath79 . specifically , for all odd @xmath59 , add a loop connecting @xmath97 to itself .
similarly , for all odd @xmath79 , add a bottom loop connecting @xmath98 to itself .
each modified meander determines two bijections of @xmath102 to itself .
define a `` top '' bijection @xmath103 of @xmath104 by @xmath105 , where @xmath106 is the unique vertex incident with the same top edge as @xmath107 . if @xmath107 is joined to itself by a top loop , then @xmath108 . in a similar way ,
define a `` bottom '' bijection @xmath109 of @xmath104 by @xmath110 , where @xmath106 is the unique vertex incident with the same bottom edge as @xmath106 . if @xmath107 is joined to itself by a bottom loop , then @xmath111 .
clearly the maps @xmath103 and @xmath109 are well - defined . for instance , in example [ modified2 ] , we have @xmath112 and @xmath113 .
consider the meander permutation @xmath117 with @xmath93 and @xmath94 as in example [ modified2 ] .
we can write @xmath117 as a product of disjoint cycles in @xmath118 : @xmath119 ( note the different use of the term `` cycle '' ) .
suppose the meander @xmath77 is the single path @xmath121 . by switching @xmath93 and @xmath94 if necessary
, we can assume that @xmath122 .
then the meander permutation is the @xmath18-cycle @xmath123 if @xmath18 is even and if @xmath18 is odd it is the @xmath18-cycle if @xmath124 conversely suppose @xmath117 is an @xmath18-cycle but @xmath96 is not a single path .
then @xmath96 contains either an isolated point , a path of length less than @xmath18 , or a cycle .
we shall show that each of these possibilities leads to a contradiction .
if @xmath126 is a path in @xmath96 with @xmath127 then , depending on whether @xmath128 is even or odd , either the @xmath129 or @xmath130 appears in the cycle decomposition of @xmath117 .
since @xmath127 we conclude that @xmath117 is not an @xmath18-cycle . now if @xmath96 contains a cycle @xmath131 then the meander permutation contains either the @xmath132 cycle @xmath133 if @xmath18 is even or the @xmath128-cycle @xmath134 if @xmath18 is odd . if @xmath127 then @xmath117 is not an @xmath18-cycle .
if @xmath135 is even , then the same argument shows that @xmath117 is not an @xmath18-cycle .
the remaining case is that @xmath135 is odd . if this happens though , we must have @xmath136 , and consequently all components @xmath59 and @xmath79 are even .
since @xmath137 we have a contradiction .
thus , in all cases when @xmath96 is not a single path , the meander permutation @xmath117 is not an @xmath18-cycle , which is a contradiction .
the proof is complete .
recall elashvili s result asserting that the maximal parabolic algebra @xmath144 is frobenius if and only if @xmath145 . an immediate corollary of the previous lemma gives a new simple proof of elashvili s result . by theorem [ cycle ]
it suffices to show that the meander permutation is an @xmath18-cycle . according to lemma [ modn ] , @xmath146 for all @xmath107
thus , the meander permutation is an @xmath18-cycle if and only if the sequence @xmath147 forms a complete residue system modulo @xmath18 .
this occurs precisely when @xmath145 .
the proof is complete .
we now use the same ideas to present another family of frobenius seaweed algebras each of which is an intersection of a positive and negative maximal parabolic algebra .
such algebras are of the form @xmath148 and are called _ opposite maximal parabolic subalgebras_. the bottom and top maps are given by @xmath152 and @xmath153 there are four possible compositions @xmath154 , depending on and whether @xmath155 or @xmath156 and whether @xmath157 or @xmath158 .
it is an easy calculation to see that in each case @xmath159 .
the argument is exactly as that used in corollary [ relprime ] .
namely , that the meander permutation is an @xmath18-cycle if and only if the sequence @xmath162 is a complete residue system modulo @xmath18 , and this is the case if and only if @xmath161 . at this time , the above line of reasoning does not easily extend to compositions @xmath93 and @xmath94 with more than two components . however , some calculations offer hope of producing more families of frobenius lie algebras using methods similar to those above .
we conclude with another new family of frobenius algebras . these are of the form @xmath166 , so they are parabolic algebras omitting exactly two simple roots .
we use a different technique than for maximal or opposite maximal algebras to analyze this family .
our result is the following classification theorem .
we first establish some conditions on the degrees of the vertices @xmath168 of the meander @xmath169 .
since the vertices of @xmath10 are viewed as the numbers @xmath170 on a line the _ interval _ between vertices @xmath171 and @xmath172 makes sense .
suppose for a moment that there exists a vertex @xmath174 of degree @xmath175 .
this vertex must have no bottom edge , meaning that @xmath18 is odd and @xmath176 .
we also know @xmath174 has no top edge so @xmath109 is odd and @xmath174 is halfway between @xmath177 and @xmath178 .
this implies that @xmath179 so @xmath180 , a contradiction .
hence , we get exactly one vertex of degree @xmath173 for each integer in @xmath181 which is odd .
if @xmath18 is odd , the vertex @xmath182 has degree @xmath173 .
if all three of @xmath39 , @xmath109 and @xmath183 are odd , then @xmath184 and @xmath185 are both even , meaning they have a common factor of @xmath71 , a contradiction .
this implies that exactly one of @xmath39 , @xmath109 or @xmath183 must be odd .
then there is exactly one other vertex of degree one as desired .
if @xmath18 is even , the bottom edges form a perfect matching .
if all three of @xmath39 , @xmath109 and @xmath183 are even , then @xmath184 and @xmath185 are again even , a contradiction .
this implies that exactly two of @xmath39 , @xmath109 or @xmath183 are odd , meaning there are two vertices of degree @xmath173 as desired . by lemma [ twoones ] , one component of @xmath10
must be a path and there are possibly more components which are all cycles .
let @xmath186 be this path and suppose @xmath186 has @xmath187 vertices in the first part of the partition , and @xmath188 and @xmath189 vertices in the other parts respectively .
note that one of @xmath190 or @xmath191 may be zero .
label the vertices of @xmath186 with @xmath192 where latexmath:[$n ' = @xmath171 ) of the vertices .
notice that the path @xmath186 forms a meander graph on its own .
this means that , by the proof of lemma [ twoones ] , we know that exactly two of the integers in @xmath194 are even and two are odd .
now suppose there exists at least one component of @xmath10 that is a cycle .
let @xmath195 be the set of all vertices in cycles of @xmath10 .
suppose @xmath195 has @xmath196 vertices in the interval between @xmath197 and @xmath198 .
for the moment , let us suppose that @xmath199 .
following the bottom edges , this means that @xmath195 must also have @xmath196 vertices in the interval between @xmath200 and @xmath201 . using this argument
, we will show that @xmath195 has @xmath196 vertices in almost every interval .
define a _ dead end _ in @xmath10 to be an interval @xmath197 to @xmath198 such that @xmath10 contains an edge joining @xmath202 and @xmath203 .
in particular , if @xmath204 is even , then the interval between @xmath205 and @xmath206 is a dead end .
a dead end is formed by two consecutive vertices of @xmath186 which are adjacent .
each occurrence of a dead end coincides with one of @xmath207 or @xmath204 being even , and we know that exactly two of these are even
. thus , there are exactly two dead ends and the proof is complete .
call an interval a _ partition interval _ if it is the meeting point of two parts of our partition .
namely , the partition intervals are from @xmath208 to @xmath209 and from @xmath210 to @xmath211 . now suppose @xmath195 has @xmath196 vertices in the interval from @xmath197 to @xmath198 . for the moment
we suppose this interval is not a dead end . as mentioned before
, this means that , by following bottom edges , @xmath195 must also have @xmath196 vertices in the interval from @xmath200 to @xmath201 .
also , by following top edges , @xmath195 must have @xmath196 vertices in another interval ( depending where the top edges go ) .
if the interval from @xmath197 to @xmath198 happens to be one of the two partition intervals ( for example suppose @xmath212 ) then this means @xmath195 must have @xmath213 vertices in the interval outside @xmath214 and at least @xmath215 vertices in the interval from @xmath210 to @xmath211 where @xmath216 .
this then implies that @xmath195 has @xmath213 vertices in the interval beyond @xmath217 ( following bottom edges ) and another @xmath213 vertices in the interval from @xmath210 to @xmath218 ( following top edges ) for a total of @xmath196 vertices in the interval @xmath210 to @xmath218 .
see figure [ meanderfig ] for an example . in this figure ,
the dark lines represent the edges of @xmath186 while light lines represent edges of @xmath195 .
the unlabeled light lines represent @xmath196 edges each . here
@xmath219 , @xmath220 , @xmath221 and @xmath222 .
alternating following top and bottom edges , we see that the cycle @xmath195 has exactly @xmath196 vertices in every interval between vertices and possibly @xmath223 vertices on each end beyond @xmath214 and beyond @xmath224 . carefully counting
, we see that the first part of our partition has @xmath225 vertices . similarly , the second part has @xmath226 and the third part has @xmath227 .
this means that @xmath228 and @xmath229 and these have a common factor of @xmath230 , a contradiction .
this shows that @xmath195 must be empty so @xmath231 is simply the path @xmath186 .
here we have broken the top into @xmath233 pieces while leaving the bottom in one piece . notice that we can add a cycle to this structure which does not pass through all the intervals .
this happens because , as the number of pieces we have increases , the number of dead ends also increases , allowing more flexibility in the placement of the cycles .
the above illustrates the complexity of the meander graphs @xmath96 as the number of parts of @xmath93 and @xmath94 grow . at the moment , the problem of classifying all frobenius seaweed lie algebras seems to be out of reach .
of late , there has been a great deal of interest in frobenius lie algebras .
perhaps these recent developments will be instrumental in the development of a classification theory . | the index of a seaweed lie algebra can be computed from its associated meander graph .
we examine this graph in several ways with a goal of determining families of frobenius ( index zero ) seaweed algebras .
our analysis gives two new families of frobenius seaweed algebras as well as elementary proofs of known families of such lie algebras .
* 2000 msc : * 17b05 , 17b08 |
low - dimensional quantum - spin systems have attracted continuous attention due to the wealth of unusual magnetic properties that result from the interplay of low dimensionalty , competing interactions and strong quantum fluctuations . among these systems ,
the diamond chain has been of particular interest , where triangular arrangements of spin @xmath12 = 1/2 entities with exchange coupling constants @xmath13 , @xmath6 and @xmath14 , are connected to form chains .
@xcite in recent years , great interest has surrounded the discovery of azurite , cu@xmath0(co@xmath0)@xmath1(oh)@xmath1 , @xcite as a model system of a cu@xmath15(@xmath12 = 1/2)-based distorted diamond chain with _ @xmath13 @xmath16 @xmath6 @xmath16 @xmath14_.@xcite the observation of a plateau at 1/3 of the saturation magnetization @xcite is consistent with a description of azurite in terms of an alternating dimer - monomer model @xcite . two characteristic temperatures ( energies ) have been derived from peaks in the magnetic susceptibility @xmath17(@xmath18 ) .
@xcite whereas the peak at @xmath19 25k , has been assigned to the dominant intra - dimer coupling @xmath20 , the one at @xmath21 5k has been linked to a monomer - monomer coupling along the chain @xmath8 axis.@xcite there have been conflicting results , however , as for the appropriate microscopic description of the relevant magnetic couplings of azurite.@xcite very recently , jeschke _ et al . _
@xcite succeeded in deriving an effective microscopic model capable of providing a consistent picture of most available experimental data for not too low temperatures , i.e. , distinctly above the transition into long - range antiferromagnetic order at @xmath22 = 1.86k @xcite . according to this work ,
the exchange couplings @xmath13 , @xmath6 and @xmath14 are all antiferromagnetic , thus placing azurite in the highly frustrated parameter regime of the diamond chain . within the `` refined model '' proposed there , @xmath20/@xmath23
= 33k and an effective monomer - monomer coupling @xmath24/@xmath23 = 4.6k were found .
another intriguing property of azurite , not accounted for so far from theory , refers to the strong magneto - elastic couplings in this compound .
these couplings manifest themselves , e.g. , in a pronounced structural distortion accompanying the magnetic transition at @xmath25 , as revealed by thermal expansion @xcite and neutron scattering experiments @xcite . here
we present a study of these magneto - elastic couplings of azurite by means of temperature - dependent measurements of the elastic constant and uniaxial thermal expansion coefficients .
these data are supplemented by thermal expansion and susceptibility measurements under hydrostatic pressure conditions .
the salient results of our study is the observation of an extraordinarily large magneto - elastic coupling constant of the intra - dimer coupling @xmath6 with respect to intra - chain deformations .
this coupling manifests itself in pronounced anomalies in the elastic constant and uniaxial thermal expansion coefficients , the latter are characterized by a negative poisson effect .
we propose that the anomalous magneto - elastic behavior of azurite is a consequence of the material s structural peculiarities , in particular , the presence of unusually stretched cu@xmath1o@xmath11 dimer units .
the single crystals ( samples # 1 - # 4 ) used for the measurements described in this paper were cut from a large high - quality single crystal which was also studied by neutron scattering and muon spin resonance ( @xmath26sr ) @xcite . for the ultrasonic experiments two parallel surfaces normal to the [ 010 ] direction were prepared and two piezoelectric polymer - foil transducers were glued to these surfaces .
longitudinal sound waves for frequencies around 75mhz were propagated along the [ 010 ] direction to access the acoustic @xmath2 mode . by using a phase - sensitive detection technique @xcite the relative change of the sound velocity and the sound attenuation were simultaneously measured as the function of temperature for 0.08k @xmath27 310k .
a top - loading dilution refrigerator was used for measurements at @xmath28 1.5k , whereas a @xmath29he bath cryostat , equipped with a variable temperature insert , was employed for accessing temperatures @xmath30 1.5k
. the elastic constant @xmath31 is obtained from the sound velocity @xmath32 by @xmath31 = @xmath33 where @xmath34 is the mass density . for measurements of the uniaxial thermal expansion coefficients , @xmath35 = @xmath36(@xmath37/@xmath38 ) , where @xmath39 is the sample length along the @xmath40 axis , two different dilatometers were used .
experiments under ambient pressure along the @xmath41 , @xmath8 and @xmath42 axes , where @xmath41 and @xmath42 are perpendicular to the ( @xmath4302 ) and ( 102 ) crystallographic planes , respectively , were carried out by means of an ultrahigh - resolution capacitive dilatometer , built after ref .
, with a resolution of @xmath44 10@xmath45 .
in addition , measurements along the @xmath8 axis were performed by using a different dilatometer @xcite , with a slightly reduced sensitivity of @xmath44 5@xmath4610@xmath45 , enabling measurements to be performed under helium - gas pressure .
the magnetic susceptibility at ambient pressure and at various finite pressure values was measured with a squid magnetometer ( quantum design mpms ) . for the measurements under pressure , a cube piston cylinder clamped cell
was used with daphne oil 7373 as a pressure - transmitting medium . at low temperature ,
the pressure inside the pressure cell was determined by measuring the superconducting transition temperature of a small piece of indium .
( color online ) . ( a ) temperature dependence of the longitudinal elastic @xmath47 mode ( open circles ) of single crystalline azurite ( sample # 1 ) .
the red solid line represents the non - magnetic elastic background @xmath48(@xmath18 ) derived from fitting eq.[eq1 ] to the experimental data for @xmath49 250k , see text for details .
( b ) normalized magnetic contribution to the elastic @xmath2 mode @xmath50@xmath47 = [ @xmath51(@xmath18)]/@xmath48(@xmath52 ) as a function of temperature for @xmath28 125k .
the solid blue line represents a fit to the experimental data ( open symbols ) based on eq.[eq2 ] .
inset to ( b ) : the longitudinal mode @xmath47 on a logarithmic temperature scale .
the arrows mark the positions of the elastic anomalies in @xmath2 of azurite which are connected to the various energy scales ( @xmath53 , @xmath54 ) , the phase transition into long - range antiferromagnetic order ( @xmath22 ) , and a transition of unknown , most likely magnetic origin ( @xmath55 ) . ]
figure [ hinh1](a ) shows the experimental results ( open symbols ) of the longitudinal elastic constant @xmath47 of azurite ( sample # 1 ) over the whole temperature range investigated . upon cooling ,
the @xmath2 mode initially increases ( hardening ) as expected for materials where anharmonic phonon interactions dominate . upon further cooling
, however , a pronounced softening becomes visible below about 250k which is accompanied by various anomalies at lower temperatures @xmath28 30k .
these anomalies can be discerned particularly clearly in the inset of fig.[hinh1](b ) , where the low - temperature data are shown on a logarithmic temperature scale .
most prominent is a distinct minimum at a temperature around 27k .
moreover , the data disclose a small dip slightly below 10k .
the position of this feature is difficult to estimate due to its smallness and the strong variation of @xmath2 with temperature caused by the nearby anomalies .
we assign these features ( labeled @xmath53 and @xmath54 in the inset to fig.[hinh1](b ) ) to the characteristic temperatures @xmath56 and @xmath57 of azurite , as revealed by susceptibility measurements@xcite .
in addition , the elastic data highlight a sharp minimum around 1.9k , reflecting the transition into long - range antiferromagnetic ordering at @xmath25 = 1.88k , and a step - like softening of comparable size to that at @xmath25 around @xmath58 = 0.37k .
the latter feature is likely to be of magnetic origin as well @xcite . for a quantitative analysis of the @xmath2 data , we determined the magnetic contribution by subtracting the non - magnetic ( normal ) elastic background @xmath48 . for @xmath48 a phenomenological expression @xmath59
was used @xcite which is fitted to the experimental data at temperatures high enough so that magnetic interactions can be neglected . here
@xmath60 is the value of the elastic constant at @xmath18 = 0 , and @xmath61 and @xmath62 are constants .
the quantity @xmath62 is usually set to @xmath63/2 , where @xmath63 is the debye temperature ( see ref . for details ) . by choosing @xmath62 = 175k , corresponding to @xmath63 = 350k as derived from specific heat @xcite , and by using @xmath64 and @xmath61 as free parameters , eq.[eq1 ] was fitted to the data for temperatures 250k @xmath65 310k .
the elastic background obtained by this fitting procedure is displayed as the solid red line in fig.[hinh1](a ) .
( color online ) .
( a ) uniaxial coefficients of thermal expansion of single crystalline azurite ( sample # 3 ) along the chain @xmath8 axis and two orthogonal axes perpendicular to @xmath8 labeled @xmath41 , @xmath42 .
inset : temperature dependence of the @xmath8-axis expansivity measured on crystal # 4 for 20k @xmath27 100k at ambient pressure ( 10@xmath66gpa ) ( yellow circles ) and at a helium - gas pressure of 0.104gpa ( blue diamonds ) in a double - logarithmic representation , cf .
the arrows mark the position of the local maximum in @xmath67 at @xmath56 derived from the fourth - order polynomial ( black solid line ) fitted to the experimental data .
( b ) volume expansivity , @xmath68 = @xmath69 + @xmath67 + @xmath70 determined from the data in ( a ) .
arrows labeled @xmath71 and @xmath72 mark the position of anomalies ( minima ) in @xmath68 .
inset : details of the phase transition anomaly in @xmath68 at @xmath22 . ]
fig.[hinh1](b ) shows the temperature dependence of the magnetic contribution to the elastic constant @xmath47 for @xmath73 125k obtained by subtracting the elastic background from the experimental data .
the so - derived magnetic contribution @xmath74 reveals a large softening of more than four percent on cooling down to @xmath53 .
this effect indicates an extraordinarily strong magneto - elastic coupling which is likely of exchange - striction type for the following reason . for longitudinal modes , such as
the @xmath2 mode investigated here , a two - ion magneto - elastic coupling arises from the modulation of the distance or bond angles between the magnetic ions which changes the interaction .
furthermore the single - ion magneto - elastic coupling for cu@xmath15 is small because of the vanishing quadrupole matrix elements @xcite . in order to quantitatively evaluate the corresponding magneto - elastic coupling constant ,
we introduce in eq.[eq2 ] a generalized strain susceptibility @xmath75 . this model accounts , within a random - phase approximation ( rpa ) @xcite , for the temperature dependence of the elastic constants of coupled dimers characterized by an intra - dimer coupling constant corresponding to the dimers singlet - triplet excitation gap @xmath50 .
this results in a temperature dependence for the elastic constant @xcite of : @xmath76 where @xmath77 is the free energy , @xmath78 is the strain along the [ 010 ] direction , and @xmath3 = @xmath79/@xmath80 is the variation of the singlet - triplet energy gap @xmath50 of the dimers upon applying a @xmath8-axis strain , i.e. , @xmath3 = @xmath4/@xmath80 for the case of azurite .
@xmath81 is the density of dimers and @xmath82 is the generalized strain susceptibility . here
@xmath83(@xmath18)= 3e@xmath84/(@xmath85@xmath18@xmath86 ) denotes the strain susceptibility of a single dimer with @xmath87 = 1 + 3e@xmath84 the partition function and @xmath88 the strength of the effective magneto - elastic dimer - dimer interaction .
the solid blue line in fig.[hinh1](b ) shows a fit to the experimental data using eq.[eq2 ] with a dimer density @xmath81 = 0.9918@xmath4610@xmath89 @xmath90 , corresponding to 2 dimers per unit cell , and @xmath91 = 7.94@xmath4610@xmath92 erg/@xmath93 , the latter is taken from the fitting procedure of the elastic background . from a nonlinear least - squares
fit the parameters @xmath50 = @xmath6 , @xmath88 and @xmath94 can be derived .
the fit was constrained to the temperature range 20k @xmath27 125k where the intra - dimer coupling represents the dominant magnetic interaction in the system .
the so - derived curve provides a good overall description of the data in the selected temperature window , especially it reproduces well the observed minimum at @xmath95 27k .
the deviations of the fit from the data become significant at temperatures @xmath18 @xmath96 22 k ( not shown in fig.[hinh1](b ) ) .
this is attributed to the simultaneous action of additional magnetic couplings at lower temperatures , especially those related to @xmath57 . from the fit
we obtain @xmath6 = ( 57 @xmath10 5)k and a coupling constant @xmath97 ( 3650 @xmath10 150)k .
the so - derived coupling constant @xmath6 is somewhat larger than the one found in the dft calculations @xcite .
most remarkable , however , is the exceptionally large value obtained for the coupling constant @xmath94 , reflecting an extraordinarily large strain dependence of the dominant energy scale in azurite .
this value , corresponding to a softening of approximately 4@xmath98 , exceeds the coupling constants typically revealed for other low - dimensional quantum spin systems @xcite by one to two orders of magnitude .
it is even about four times bigger than the very large coupling constant found for the longitudinal mode in the coupled - dimer system srcu@xmath1(bo@xmath0)@xmath1 @xcite . for the parameter @xmath88 in eq.[eq3 ]
the fit yields @xmath99(200 @xmath10 10)k , indicating that the effective magneto - elastic dimer - dimer interaction is antiferrodistortive in azurite . + in order to obtain supplementary information about the magneto - elastic couplings corresponding to the various energy scales , we performed measurements of the coefficient of thermal expansion and the magnetic susceptibility both at ambient- and under hydrostatic - pressure conditions .
figure [ hinh2](a ) shows the results of the temperature dependence of thermal expansion coefficients @xmath100(@xmath18 ) ( @xmath40 = @xmath41 , @xmath8 , @xmath42 ) for temperatures 1.6k @xmath27 75k measured at ambient pressure .
similar to the elastic constant , three distinct anomalies are observed in this temperature range along all three crystallographic axes . upon cooling ,
the @xmath8-axis data show a broad positive maximum around 20k whereas along the two other axes ( @xmath41 and @xmath42 ) a negative minimum appears . upon further cooling to lower temperatures ,
a minimum shows up in the @xmath8-axis data , which contrasts with maxima around 3.5k in the data along the @xmath41- and @xmath42-axis . despite having opposite signs , the anomalies in @xmath101 and those in @xmath102 and @xmath103
do not cancel each other out in the volume expansion coefficient @xmath68(@xmath18 ) = @xmath102 + @xmath101 + @xmath103 , shown in the main panel of fig.[hinh2](b ) .
the volume expansivity exhibits a pronounced negative contribution , giving rise to a change of sign around 50k , and a broad minimum around 25k followed by a second minimum at @xmath104 5k .
the strong upturn in @xmath68(@xmath18 ) and @xmath100 at lower temperatures is due to the antiferromagnetic phase transition at @xmath22 .
this is shown more clearly in the inset of fig.[hinh2](b ) where @xmath68(@xmath18 ) is displayed on expanded scales around @xmath22 .
an extraordinarily large @xmath105-type anomaly , lacking any hysteresis upon cooling and warming , is observed which demonstrates the second - order character of the phase transition at @xmath22 . from the coincidence of the anomalies in @xmath68(@xmath18 ) ( and @xmath100(@xmath18 ) ) with those revealed in the magnetic susceptibility @xcite and elastic constant ( fig.[hinh1 ] ) , we conclude that these anomalies reflect the characteristic temperatures @xmath56 and @xmath57 which are related to the energy scales @xmath6 and @xmath24 , respectively .
thus from the evolution of these anomalies under pressure we may determine the pressure dependencies of these energy scales .
in the inset of fig.[hinh2](a ) we compare , on a logarithmic temperature scale , the temperature dependence of @xmath67 for a small pressure of at 10@xmath66gpa with that of 0.104gpa . these data , which have been taken on sample # 4 by employing a different dilatometer , especially designed for measurements under helium - gas pressure @xcite ,
disclose two remarkable features .
first , we find a strong change in the temperature dependence of @xmath67(@xmath18 ) , accompanied by a considerable suppression in its absolute value over the whole temperature range investigated , i.e. , for @xmath28 100k .
note that due to the solidification of the pressure medium , the measurements at 0.104gpa were limited to @xmath106 14.5k . since changes of the lattice expansivity under a pressure of 0.104gpa for a material with a normal bulk modulus are expected to be less than 1% ,
see ref . and the discussion therein
, we attribute these effects to the influence of pressure on the material s magnetic properties and their coupling to the lattice degrees of freedom .
the presence of significant magnetic contributions at elevated temperatures @xmath107 which strongly couple to the lattice is consistent with the acoustic behavior revealed for the @xmath2(@xmath18 ) elastic mode , yielding an onset temperature for the pronounced softening as high as 250k , cf . , fig.[hinh1](a ) .
second , a thorough inspection of the data around the maximum reveals a shift of the position of the maximum to lower temperatures on increasing the pressure from @xmath108 = 0.1 to 0.104gpa .
this can be quantified more precisely by fitting both data sets in the immediate surrounding of the maximum by fourth - order polynomials @xcite , depicted as solid lines in fig.[hinh2](a ) . by identifying the position of the maximum with @xmath56
, we find a pressure dependence @xmath109/@xmath110 = @xmath99(0.08 @xmath10 0.03)k / gpa .
( color online ) .
temperature dependence of magnetic susceptibility of single crystalline azurite ( sample # 2 ) measured at various hydrostatic pressure values ( from bottom to top ambient pressure , 0.36 , 0.62gpa ) in a magnetic field of @xmath111 = 2 t applied parallel to the @xmath8-axis .
red solid lines correspond to fits to the data based on a model for interacting dimers given in eq.[eq4 ] .
inset : blow - up of the data ( symbols ) and model curves ( red lines ) in the vicinity of the characteristic temperature @xmath112 . besides the data at ambient pressure ( black circles ) , @xmath108 = 0.360gpa ( green triangles ) and 0.62gpa ( purple triangles ) , also data at 0.26gpa ( orange squares ) are shown .
arrows labeled @xmath113 and @xmath114 mark the position of the maxima in the various data sets .
broken lines indicate the shift in the maximum position with pressure . ]
figure [ hinh3 ] displays the data of the magnetic susceptibility @xmath17(@xmath18 , @xmath108 = const ) as a function of temperature for 2k @xmath115 300k at varying pressure values from ambient pressure up to 0.62gpa .
the ambient - pressure data are consistent with those reported by kikuchi _
@xcite , yielding an increase in @xmath17 with decreasing temperature and two broadened maxima at @xmath113 and @xmath114 . upon increasing the pressure
we observe a progressive increase of the low - temperature susceptibility , cf.fig.[hinh3 ] .
in addition , a closer look at the low - temperature data in the inset of fig.[hinh3 ] discloses a shift of the position of both maxima to lower temperatures albeit at different rates . for a quantitative analysis of the susceptibility data for not too low temperatures around @xmath56 and up to 300k
, we again use an rpa - molecular field expression for coupled dimers , in analogy to the procedure applied for analyzing the elastic constant @xmath47 data : @xmath116 here @xmath117(@xmath18 ) = 2e@xmath118 is the magnetic susceptibility of an isolated dimer , @xmath87 the partition function and @xmath119 an average magnetic inter - dimer coupling . a fit to the experimental data at ambient pressure was performed for 15k @xmath27 300k by using eq.[eq4 ] and a curie contribution , according to the amount of cu - monomer spins , with @xmath50 = @xmath6 and @xmath119 as free parameters .
this fit provides a very good description of the data including the height and the position of the maximum at @xmath113 , cf . the solid red line running through the ambient- pressure data points in fig.[hinh3 ] . from the fit
we obtain @xmath6/@xmath85 = ( 40.5 @xmath10 0.7)k , consistent with the value suggested by dft calculations for the `` full model '' discussed there @xcite , and @xmath119 = ( 4.1 @xmath10 0.8)k .
note that this value of @xmath119 indicates that the average magnetic dimer - dimer interaction is small and ferromagnetic .
the rpa - molecular field description remains good also for the data taken at finite pressure .
for these fits the same curie susceptibility as used for the ambient - pressure data was used to account for the magnetic contributions of the monomers .
the evolution of the parameters @xmath6 and @xmath119 with pressure , derived from fitting eq.[eq4 ] to the finite - pressure data , is shown in fig.4 .
we find a suppression of @xmath6 with increasing pressure which can be approximated by @xmath4/@xmath110 @xmath120 @xmath99(0.077 @xmath10 0.007)k / gpa . for the average magnetic inter - dimer interaction
we find an approximate linear increase under pressure with a larger rate of @xmath121/@xmath122 ( 0.216 @xmath10 0.002)k / gpa .
this pronounced increase of @xmath123/@xmath110 is responsible for the growth of the low - temperature susceptibility with pressure .
as will be discussed in more detail below ( sec.c ) , the dominant ferromagnetic character of the dimer - dimer interaction can be assigned to a particular cu - o - cu exchange path connecting dimers of adjacent chains along the @xmath124-direction . + ( color online ) .
evolution of the fit parameters , the intra - dimer coupling constant @xmath6 ( open squares , left scale ) and the average magnetic dimer - dimer interaction energy @xmath119 ( full triangles , right scale ) , with pressure as derived from fits of eq.[eq2 ] to the data in fig.3 . ] ( color online ) .
poisson ratio ( black full circles , right scale ) together with the relative volume change around @xmath22 ( blue open squares , left scale ) . ]
as shown in the previous section , the phase transition into the long - range antiferromagnetic order in azurite at @xmath22 = 1.88k is accompanied by a pronounced anomaly in the elastic @xmath2 mode , corresponding to a softening of about 0.1@xmath98 ( cf.inset to fig.[hinh1](b ) ) which exceeds by far the features usually revealed at a magnetic transition in low - dimensional spin systems where often only a kink - like anomaly is observed , see refs .
, , for typical examples , and ref . for a review .
this goes along with an extraordinarily large anomaly in the uniaxial coefficients of thermal expansion @xmath100 , see fig .
[ hinh2](b ) .
as a result of the strong magneto - elastic coupling , there is a large @xmath105-type anomaly in the volume expansion coefficient @xmath125 , ( see the inset of fig.[hinh2](b ) ) , corresponding to a relative reduction of the volume upon cooling from 2 to 1.6k of @xmath126 = @xmath995.7 @xmath46 10@xmath127 as shown in fig.5 . + according to the ehrenfest relation , the discontinuities at this second - order phase transition in @xmath68 , @xmath128 , and that in the specific heat , @xmath129 , can be used to determine the pressure dependence of the nel temperature in the limit of vanishing pressure @xmath130 by using @xmath128 = ( 550 @xmath10 30 ) @xmath131 10@xmath132 k@xmath133 and @xmath129 = ( 6.18 @xmath10 0.4 ) @xmath131 j mol@xmath133k@xmath133 taken from ref . , we find @xmath134 = ( 0.15 @xmath10 0.02)k / gpa .
this extraordinarily large pressure dependence , exceeded only by some exceptional cases , such as the coupled antiferromagnetic / structural transition in co - substituted cafe@xmath1as@xmath1 @xcite , highlights the strong magneto - elastic coupling and the unusual elastic properties of azurite .
+ the large @xmath128 at @xmath22 , tantamount to a large pressure dependence of @xmath22 in eq.[eq6 ] , is partly due to the fact that in azurite at @xmath22 the discontinuities in the uniaxial expansion coefficients , @xmath135 , all have the same ( positive ) sign .
hence , this phase transition is characterized by an anomalous poisson effect .
+ in general for an isotropic material , the poisson ratio , which measures the material s cross section under tension , is defined as : @xmath136 where @xmath137 is the strain in stretching direction and @xmath138 perpendicular to it . in most materials
@xmath139 is positive , which reflects the fact that an expansion along one axis is usually accompanied by a compression in the perpendicular direction , to keep the overall volume change small . in many materials @xmath139 values
are found in the range 0.2 @xmath140 0.5 @xcite .
the latter value corresponds to the situation that the material keeps its volume under tension .
a positive @xmath141 means that the material becomes thinner when it is stretched , the behavior encountered for most materials .
in contrast , materials with a negative poisson ratio become thicker when they are stretched .
those compounds , called auxetic materials , are of interest due to potential technical applications @xcite . in fig.5
we show the poisson ratio @xmath139 of azurite for temperatures around @xmath22 . for the present anisotropic case ,
@xmath142 has been determined by using the relative length changes ( @xmath143/@xmath144)@xmath145 ( corresponding to the integral of @xmath67(@xmath18 ) with respect to temperature ) for the strain @xmath137 along the stretching @xmath8 direction and @xmath146 = ( @xmath147 + @xmath148)/2 = ( ( @xmath143/@xmath144)@xmath149 + ( @xmath143/@xmath144)@xmath150 perpendicular to it .
we stress that @xmath139 of azurite reaches a normal value of 0.23 around room temperature ( not shown ) .
however , as indicated in fig.5 , @xmath139 exhibits a large negative value of @xmath990.8 for temperatures slightly above @xmath22 .
this value increases to @xmath990.65 upon cooling to @xmath22 below which it further decreases , reaching almost @xmath991 at 1.6 k. note that this value is close to the stability limit of any elastic linear , isotropic material where the requirement of positive young s shear and bulk moduli dictates @xmath139 @xmath151 @xmath991 .
+ ( color online ) .
sections of the structure of azurite .
a ) the dominant inter - dimer interaction , connecting dimers ( marked by yellow planes ) in adjacent chains along the @xmath124-direction , is mediated via the cu2-o3-cu2 exchange path .
b ) arrangement of relevant structural units of azurite including two cu(ii ) monomers ( cuo@xmath152 containing cu1 ) and one dimer ( cu@xmath1o@xmath11 containing cu2 ) forming chains along the @xmath8-axis viewed perpendicular to the @xmath153 plane .
the labels correspond to the room temperature structure reported in ref .. ] [ cols="^,^,^,^ " , ] we start the discussion by considering the dimer - dimer interaction @xmath119 revealed from the analysis of the susceptibility measurements under variable pressure .
the dominant ferromagnetic character of this interaction is assigned to the cu2-o3-cu2 exchange path ( cf .
6a ) connecting dimers of adjacent chains along the @xmath124-axis .
the corresponding cu2-o3-cu2 bond angle amounts to 91.57@xmath154 at 5k , consistent with a weak ferromagnetic interaction as revealed for hydroxo - bridged cu(ii ) complexes @xcite .
note that also the dft calculations for the `` full model '' of azurite exhibit a small ferromagnetic exchange @xcite .
it is likely that under hydrostatic pressure the structure will deform in a way such that this angle decreases with increasing pressure .
this is consistent with an increase of the ferromagnetic inter - dimer coupling @xmath119 derived from susceptibility measurements under pressure , cf.fig.4 .
+ in the following we will argue that the auxetic behavior at @xmath22 and the huge magneto - elastic coupling of azurite is likely due to peculiarities of the molecular arrangement in this compound , in particular that of the cu@xmath1o@xmath11 dimer units , cf.fig.6 . as shown in fig.6b ,
cuo@xmath152 monomer units ( containing cu1 ) are connected via o2 ions to cu@xmath1o@xmath11 dimers ( containing cu2 and marked by yellow planes in fig.6 ) to form chains along the @xmath8 axis . according to structural data at room temperature @xcite and 5k @xcite , cf.table 1 , the structural parameters of the monomers , involving o1-cu1 and o2-cu1 bonds , change very little upon cooling and are very close to those values ( 1.90 - 1.93@xmath155 ) typically found in isolated cu(ii ) complexes @xcite .
this indicates a stable and rigid configuration of the cuo@xmath152 monomer units .
in contrast , the structural parameters of the cu@xmath1o@xmath11 dimers are rather unusual . in particular the cu2-o2 bonds , mediating the intra - dimer coupling @xmath6 , are significantly longer ( 1.9947 @xmath155 and 1.9675 @xmath155 at 300k ) ( see table 1 ) than those found in isolated dimer complexes , reflecting an unusually `` stretched '' arrangement .
it is thus obvious to suspect that this dimer unit represents the flexible part in the structure which may accommodate itself accordingly when the surrounding structure is exposed to external stimuli .
this is the case , e.g. , when a @xmath8-axis strain is applied or , alternatively , upon cooling through @xmath56 which is accompanied by a significant contraction of the @xmath8 axis ( large positive anomaly in @xmath67 , cf.fig.2 ) .
in fact , according to structural data ( table1 ) , it is the o2-o2 distance , as indicated by the blue arrows in fig.6b , across the dimer which yields the strongest reduction upon cooling from at 300k to 5k , reflecting the softness of these dimer units .
note that changes of the cu2-o2 geometry imply changes of the dominant magnetic interaction @xmath6 .
as fig.6b suggests , such a contraction along the @xmath8-direction will cause not only a reduction of the o2-o2 distance .
it will also lead to changes of the cu2-o2 bond length and the cu2-o2-cu2 bond angle ( 97.77@xmath154 at 300k ) within the dimers .
in fact , the 5k structure reveals an cu2-o2-cu2 angle of 98.41(7)@xmath154 . at the same time it is conceivable that upon the reorientation of the dimer in response to a @xmath8-axis strain , the tensile forces acting on the dimers get weaker , allowing the dimers to adopt a more natural , compact configuration , consistent with the observed reduction of the cu2-o2 bond lengths .
this less - strained dimer configuration is also visible in the intra - dimer o3-o3 distance which is reduced as well ( see table 1 and the yellow arrows in fig .
due to the orientation of the dimer unit in the @xmath156 plane , a reduction of the @xmath8 axis is connected with a reduction of the @xmath124- and @xmath157-axis which then explains the anomalous poisson effect @xcite at low temperatures .
+ as for the anomalously large magneto - elastic coupling @xmath94 , one might expect that the stretched arrangement of the dimers may alter the relationship between the exchange coupling constant and the dimers structural parameters @xmath158 , such as the inter - atomic distances or bonding angles , and with it the generalized derivatives @xmath159/@xmath160 . according to crawford _
@xcite , who investigated various stable hydroxo - bridged dimer complexes , there is a linear correlation between the intra - dimer coupling and both the cu - cu distance as well as the cu - o - cu bond angle .
all of these materials exhibit nearly the same typical cu - o distances .
it is likely that these relations may change significantly for a strongly stretched configuration , as realized for the dimer units in azurite .
this may result in strongly enhanced derivatives @xmath159/@xmath160 such as the large magneto - elastic coupling constant @xmath159/@xmath5 revealed here .
a microscopic theory , addressing the relationship between the coupling constants and the dimers structural parameters , is necessary to confirm this conjecture .
note that the presence of pre - strained structural units in azurite is consistent with the fact that a hydrothermal synthesis technique operating at a pressure of about 0.350gpa has to be applied for growing single crystals of this mineral @xcite .
measurements of the longitudinal elastic constant @xmath2 and the uniaxial thermal expansion coefficients @xmath100 on single crystalline azurite reveal pronounced anomalies associated with the intra - dimer coupling constant @xmath6 . from a quantitative analysis of the elastic constant data ,
an exceptionally large value of the magneto - elastic coupling @xmath3 = @xmath4/@xmath5 of @xmath161 ( 3650 @xmath10 150)k has been derived . by lacking a microscopic theory , we tentatively assign this large value , exceeding corresponding magneto - elastic couplings for other low - dimensional quantum spin systems by two to three orders of magnitude , to structural peculiarities of azurite .
we propose that it is the cu@xmath1o@xmath11 dimer unit , which is incorporated in the structure in an unnaturally stretched manner , which is responsible for the exceptionally large magneto - elastic coupling in this system .
we acknowledge financial support by the deutsche forschungsgemeinschaft via the sfb / tr49 and b. lthi , r. valent and h. jeschke for useful discussions .
h. jeschke , i. opahle , h. kandpal , r. valenti , h. das , t. saha - dasgupta , o. janson , h. rosner , a. brhl , b. wolf , m. lang , j. richter , s. hu , x. wang , r. peters , t. pruschke , and a. honecker , phys .
lett . * 106 * , 217201 ( 2011 ) .
m. c. r. gibson , k. c. rule , a. u. b. wolter , j .- u .
hoffmann , o. prokhnenko , d. a. tennant , s. gerischer , m. kraken , f. j. litterst , s. sllow , j. schreuer , h. luetkens , a. brhl , b. wolf , and m. lang , phys .
b * 81 * , 140406(r ) ( 2010 ) .
h. kikuchi , y. fujii , m. chiba , s. mitsudo , t. idehara , t. tonegawa , k. okamoto , t. sakai , t. kuwai , k. kindo , a. matsuo , w. higemoto , k. nishiyama , m. horvatic , and c. bertheir , prog .
. suppl . * 159 * , 1 ( 2005 ) .
note that the anomalous length changes around @xmath22 discussed above , giving rise to a negative poisson effect at low temperatures , are of the order of @xmath143/@xmath144 @xmath96 10@xmath127 and therefore beyond the resolution of standard structural investigations . | we present results of ultrasonic measurements on a single crystal of the distorted diamond - chain compound azurite cu@xmath0(co@xmath0)@xmath1(oh)@xmath1 .
pronounced elastic anomalies are observed in the temperature dependence of the longitudinal elastic mode @xmath2 which can be assigned to the relevant magnetic interactions in the system and their couplings to the lattice degrees of freedom . from a quantitative analysis of the magnetic contribution to @xmath2 the magneto - elastic coupling @xmath3 = @xmath4/@xmath5
can be determined , where @xmath6 is the intra - dimer coupling constant and @xmath7 the strain along the intra - chain @xmath8 axis .
we find an exceptionally large coupling constant of @xmath93650 @xmath10 150)k highlighting an extraordinarily strong sensitivity of @xmath6 against changes of the @xmath8-axis lattice parameter .
these results are complemented by measurements of the hydrostatic pressure dependence of @xmath6 by means of thermal expansion and magnetic susceptibility measurements performed both at ambient and finite hydrostatic pressure .
we propose that a structural peculiarity of this compound , in which cu@xmath1o@xmath11 dimer units are incorporated in an unusually stretched manner , is responsible for the anomalously large magneto - elastic coupling . |
resonance properties of nanoparticles have been observed for centuries thanks to beautiful colors of gold- and silver - patterned stained glasses . over the last decade
nanopatterned materials have attracted even increased attention due to their unique electronic and optical characteristics .
novadays , they are considered as promising candidates for wide variety of applications in subwavelength waveguiding @xcite , enhanced raman scattering spectroscopy 2005_prb_bachelier , non - linear optics @xcite , photovoltaics @xcite and biological / medical sensing 2004_natbio_alivisatos and many others . a characteristic size of metallic nanoparticles @xmath0 is about an order of magnitude smaller than the wavelength of incoming light @xmath1 , which can excite collective oscillations of electron density inside the particle , - plasmons . the plasmon excitation results in an enhanced extinction ( extinction = absorption + scattering ) as well as an increased intensity of the electromagnetic field near the particle @xcite .
the important issue that makes nanoparticles so attractive for sensing applications is the effect of the geometry and size of nanoparticles and surrounding environment on a position of the plasmonic resonance 1995_book_kreibig , 2005_os_khlebtsov , 2005_jpcb_lee , 2003_jpcb_kelly .
for example , the presence of antibodies in cells affected by cancer modifies the environment for the gold nanoparticles placed on a tissue and results in a shift of extinction peak that can be easily imaged by conventional microscopy @xcite .
recently it has also been demonstrated @xcite that embedding metallic nanoparticles into a polymeric matrix provides the larger contrast in the effective refractive index of the blend material , being much lower or higher than that of a pure polymer .
developing such the materials can facilitate creating high - contrast - index photonic polymer crystals .
nanoparticles assembled in nanochains can also be applied as subwavelength waveguides @xcite . in the case of closely placed particles the coupling ( and light propagation )
arises from the evanescent dipole field from each particle , which excites a plasmon on its neighbour .
this excitation travels along the chain , making the electron density within all the particles oscillate in resonance . in the present paper
we will focus on light propagation in large arrays of infinitely long nanorods .
prototypes of such the arrays have been recently fabricated experimentally @xcite .
these arrays represent randomly oriented or aligned long rods ( or spikes ) of a material ( dielectric or metal ) , several tens of nanometers in diameter . despite of significant progress in nanofabrication technologies , to our knowledge , however , the theoretical description of light propagation in nanorod arrays is still missing .
the paper is organized as follows . in section
ii we outline transmittance properties of nanorod arrays within the framework of the maxwell - garnett effective - medium theory . in section
iii we present numerical modeling of light propagation through periodic arrays of nanorods and compare the results with the predictions of the maxwell - garnett theory . in section
iv the effect of various types of disorder is studied .
we consider a gelatin matrix with an embedded two - dimensional array of silver nanorods .
the effective dielectric function @xmath2 of that composite can be estimated from developed for more than 100 years ago maxwell - garnett theory @xcite : @xmath3 where @xmath4 is the filling factor of the nanorods embedded into the matrix , @xmath5 is the active area of the matrix and @xmath6 is the total cross - section area of the nanorods .
the dielectric function of the gelatin matrix is @xmath7 .
the dielectric function @xmath8 of the nanorods is taken from the sopra database @xcite for the bulk material .
the maxwell - garnet theory is valid for relatively small nanoparticles ( nanorods ) ( up to several tens of nanometers ) at low concentrations ( less then 30% ) .
the dielectric function ( here and hereafter all the spectra are given in respect of light wavelength in vacuum @xmath9 ) of the ag(10%)-gelatin blend is presented in fig .
[ fig : eff_medium ] ( a ) .
thick layer of ag(10%)-gelatin blend .
inset in ( b ) outlines the system under study . ]
the dielectric function in fig .
[ fig : eff_medium ] ( a ) characterizes the blend as a higly - dispersive lossy material with an absorption peak centered around 414 nm . according to the mie s theory this peak corresponds the plasmon resonance of a single ag spherical nanoparticle in gelatin , the position of the peak obeys the well - known relation @xmath10 @xcite . in order to study light propagation through the layer of the blend we consider a 2d `` sandwich - like '' structure consisting of semi - infinite gelatin `` waveguides '' connected to the blend region
[ see inset to fig .
fig : eff_medium ( b ) ] .
the structure is assumed to be infinite in @xmath11-direction , thus the solution to the maxwell s equations decouples into te ( vector of a magnetic field is parallel to @xmath11 ) and tm ( vector of an electric field is parallel to @xmath11 ) . the transmission , reflection and absorption for both polarizations are given in fig .
[ fig : eff_medium ] ( b ) and ( c ) respectively .
it is easy to see that for both te and tm polarizations there exists a gap ( or a stop - band ) in the transmission caused by the enhanced absorption near the extinction resonance peak .
however , the reflectance and absorption within the stop - band possess distinct behavior for different polarizations .
when the real part of the dielectric constant of the blend becomes negative ( @xmath12 nm ) the reflectance of the te mode increases due to increased contrast against the dielectric function of the gelatin matrix ( which causes a dip in the absorption ) . at the same time , for tm - polarized light the reflectance sharply increases up to 1 because of the metallic character of the blend in this region and enhanced skin effect .
for both polarizations bragg s reflections from the boundaries of the blend region , manifested themselves as minima and maxima , are clearly seen for @xmath13 nm . despite its adequacy for small isolated circular nanoparticles , a simple maxwell - garnet theory ,
however , has certain limitations .
namely , it does not account for the shape and distribution of metal clusters in the dielectric medium , neglecting important polarization properties of both single non - circular particles and their arrangements @xcite . in order to incorporate these features and
study transmission characteristics of periodic and disordered nanorod arrays we apply the recursive green s function technique @xcite .
we now focus on 2d arrays of infinitely long silver nanorods arranged as a square lattice in a gelatin background . keeping the filling factor of ag , @xmath14 constant , we consider two cases , ( a ) a finite - size lattice with thickness @xmath15 of nanorods with the diameter @xmath16 , and ( b ) the lattice of the same width assembled from nanorods of 60 nm in diameter , see fig .
[ fig : nanorod_arrays ] . lattice constants are 29 and 175 nm for cases ( a ) and ( b ) respectively . and the filling factor @xmath17=10% . ] such the choice of nanorod sizes is motivated by the essential difference in polarization properties of small and large nanoparticles 2003_jpcb_kelly . if the nanoparticle is small enough ( @xmath18 ) , according to the mie s theory , only the dipole plasmonic oscillations contribute to the extinction spectra , whereas for larger particles higher - order resonances contribute to the spectra as well 2003_jpcb_kelly . using the recursive green s function technique we perform numerical simulations for both te and tm polarizations of light falling normally from the left to the boundary between gelatin and the blend .
when a nanosized metallic nanoparticle is illuminated by light , the electric components of an electromagnetic field excite collective oscillations of electronic plasma inside the particles plasmons .
if the particles are arranged into chains , these plasmonic oscillations possess a resonant character that facilitates the propagation of light along the chain .
such the chains have been intensively studied in literature @xcite as promising candidates for sub - wavelength wave - guiding .
let us irradiate the array of infinitely long nanorods with te - polarized light . in this case @xmath19 and @xmath20 components of the electromagnetic field excite coherent plasmonic oscillations on each nanorod .
figure fig : te_modes_ordered shows the calculated transmittance , reflectance and absorption of a te mode propagating through the arrays of nanorods . for details ) . ]
_ small nanorods .
_ let us first concentrate on an array of nanorods with the diameter 10 nm [ fig .
[ fig : te_modes_ordered ] ( a ) ] . in the spectra
one can clearly distinguish two regions , namely the region of high absorption ( @xmath21 nm ) , containing a wide main absorption peak at 414 nm , two minor peaks at 350 and 530 nm and the region of high transmittance ( @xmath22 nm ) .
now we will take a closer look at these regions separately .
the position of the main extinction resonance agrees well with that obtained from eq .
( [ eq : maxwell_garnet ] ) .
however , in contrast to the maxwell - garnett theory , the spectrum contains two minor peaks near 350 and 530 nm . in order to explain them one needs to account the effect of coupling between several nanorods . for this sake
we compare light propagation through ( a ) a single isolated nanorod ( diameter 10 nm ) , ( b ) two coupled nanorods aligned parallel to the light propagation direction , ( c ) those aligned perpendicularly , and ( d ) four coupled nanorods .
these four cases are presented in fig .
[ fig : single_nanoparticles ] . .
] for the case of a single isolated rod [ fig .
[ fig : single_nanoparticles ] ( a ) ] only one peak near 410 nm emerges which is in a good agreement with the analytical value of 414 nm . for the cases ( b ) and
( c ) of twin coupled nanorods the additional peaks , centered at 355 and 620 nm respectively , appear .
their origin has been thoroughly studied in @xcite and clarified in terms of enhanced [ case ( b ) ] and weaken [ case ( c ) ] restoring forces between the particles . however , for the system of four particles ( d ) these forces partially compensate each other , and the minor resonances move closer towards the main peak .
let us now focus on the wavelength region @xmath23 nm , where te - polarized light propagates at high transmittance . in order to understand this behavior ,
we complement the transmission coefficient with the band diagram of the nanorod array .
it should be mentioned that , in general , a band diagram represents propagating bloch states ( states with real eigenvalues ) . however , as the metallic rods ( or nanoparticles ) are absorbing , all the states in the blend will be decaying and eventually die off at the infinity . yet
all the bloch eigenvalues in such systems have imaginary components . in fig .
[ fig : disp_rel ] ( a ) we represent a band structure ( real parts of eigenvalues ) in @xmath24-direction for the states with the smallest imaginary parts .
the dispersion curve in fig . [ fig : disp_rel ] ( a ) has a small bump around 550 nm , which is caused by the minor extinction resonance .
the band is located very close to the light line that results in a rather strong coupling between the incoming light and the plasmonic bloch states of the blend region .
such the strong coupling explains high transmittance in the red wavelength region .
_ large nanorods _ for nanorods with the diameter 60 nm , the position of the main extinction peak agrees with that one of the small particles
. however , there is an essential difference in physics behind . when the diameter of a nanoparticle increases , higher - order dipole oscillations now contribute the resulting extinction spectrum @xcite
. it has been recently shown @xcite that the peak centered at @xmath25 nm is due to the _ quadruple _ resonance of a nanorod , whereas the dipole resonance is redshifted and overlaps with the region of the enhanced reflectance ( @xmath26 nm ) .
the indication in favor of this interpretation is a narrower width of the stop - band in the transmission ( 60 nm against 100 nm in case of small rods ) .
this is because the higher - order dipole interactions causing the stop - band behaviour for the case of large nanorods are generally weaker .
now let us clarify the origin of the high - reflectance region .
the lattice constant for this structure is 175 nm .
this is of the same order as the wavelength of light , such that the structure effectively represents a two - dimensional photonic crystal .
the plasmonic band in fig .
fig : disp_rel ( b ) extends from @xmath27 to 0.4 ( @xmath28 nm ) where it experiences a _ photonic bandgap _ that causes the high reflectance of the structure .
this bandgap overlaps with the tail of the extinction peak near 500 nm ( see fig . [
fig : te_modes_ordered ] ) .
let us now consider the tm - polarization of the incoming light .
figure fig : tm_modes_ordered ( a ) shows the transmittance , reflectance and absorption of the tm - polarized light for the small nanorods .
in contrast to the maxwell - garnett picture ( fig .
[ fig : eff_medium ] ) , almost for the whole wavelength range under study light does not penetrate the region occupied by nanorods and gets fully reflected back , resulting in zero transmittance . this discrepancy can be explained by the skin - effect on the silver rods . at the same time
, the maxwell - garnett theory ( eq : maxwell_garnet ) fully disregards the important screening properties of the rods , simply averaging the effective dielectric constant over the structure .
it is also worth mentioning , that as we consider infinitely long nanorods , incoming tm - mode _ does not excite _ any plasmons on the rods and thus there is no plasmonic contribution in overall transmission .
nm . due to the skin - effect light
does not penetrate the blend region . for @xmath29
nm the real part @xmath30 of the dielectric function of silver ( b ) becomes positive and the transmission coefficient abruptly increases . ] however , for very short wavelengths ( @xmath31 nm ) the real part of the dielectric function of silver becomes positive ( see fig .
fig : tm_modes_ordered ( b ) ) and the blend behaves like a lossy dielectric rather than a metal .
this results in non - zer transmission in this region .
the obtained results clearly show that resonant plasmonic oscillations in periodic nanorod arrays represent a dominating light propagation mechanism for the te - polarized light , whereas for the tm modes the nanorod structure represents practically a perfect screen .
this features can be utilized in a nearly 100% effective polarizer .
diameter randomly varies from of 5 to 20 nm .
( b , d ) nanorods with fixed @xmath32 nm are randomly distributed within the layer .
insets show the actual geometries of the structures . ] as we have demonstrated in the preceding section , te - polarized incident light in off - resonance wavelength region propagates through periodic arrays of small nanorods at very high transmission .
now we introduce some disorder in this array and consider two separate cases , namely when nanorods are arranged in a square lattice but have randomly varying diameter , and rods of equal diameter , randomly distributed within the layer . for both cases the filling factor @xmath3310% is kept constant , the distribution is taken uniform .
figures [ fig : disordered_str ] ( a , c ) and ( b , d ) demonstrate the transmittance , reflectance and absorbtion for both cases . the transmission characteristics for the structure with the random diameter of nanorods [ fig .
[ fig : disordered_str ] ( a , c ) ] closely resemble those for the array of fixed - sized nanoparticles [ fig .
[ fig : te_modes_ordered ] ( a ) ] .
the main difference is that weaker dipole interactions between adjacent particles of different diameter cause a minor narrowing of the stop - band ( 60 nm versus 100 nm in the ideal case ) and a slight degradation of the minor extinction peak .
it should be emphasized that the transmission properties of arrays with different distributions of the nanorod diameter [ fig .
[ fig : disordered_str ] ( a ) and ( c ) ] are virtually the same . for the case of the randomly distributed equal - sized nanorods [ fig .
fig : disordered_str ( b , d ) ] the situation changes . in contrast to the previous case of the ordered nanorod array with random diameter , the absorption spectra in the region @xmath23 nm are extremely sensitive to the geometry of the structure .
it is interesting to note how clustering of nanorods manifests itself .
the overall absorption in the region @xmath23 nm is much higher ( and the transmision is lower ) in comparison to the periodic lattice , as it consists of the averaged multiple absorbtion peaks of closely situated , touching or overlapping nanorods .
since the inter - rod distances are not constant any longer , each single rod is now affected by many dipole interactions of different strengths from neighboring rods .
reflectances , however , are practically identical and not significantly higher than for the periodic case .
it can be explained that due to its non - periodicity this structure absorbs better than reflects .
the clustering and more complex interactions of nanorods influence the region @xmath34 as well .
the main and minor absorption peaks for the structure fig . [ fig : disordered_str ] ( b ) almost overlap , whereas for fig .
[ fig : disordered_str ] ( d ) are still well - separated .
we should specially mention that in order to incorporate the effect of nanoparticle aggregates into maxwell - garnett theory , several approaches were suggested @xcite ( see also references therein ) . in that case
the effective medium dielectric function is derived by inserting the total aggregate polarizabilities instead of that of a single isolated particle into the maxwell - garnett theory .
we have studied propagation of te- and tm - polarized light in two - dimensional arrays of silver nanorods in a gelatin background . in order to calculate transmittance , reflectance and absorption in arrays of ordered and disordered nanorods we applied the recursive green s function technique and compared the obtained numerical results with predictions of the maxwell - garnett effective - medium theory .
we have demonstrated that this theory describes adequately only the case of the te - polarized light propagating in ordered arrays of small ( @xmath35 nm ) , well - separated nanorods and only in the frequency interval outside the main plasmonic resonance .
our numerical calculations outline the importance of geometrical factors such as the size of the rods and their distribution .
in particular , we have demonstrated , that interaction between adjacent nanorods brings the significant contribution to the transmission spectra , which is manifested as additional absorption peaks ( that are missing in the effective - medium approach ) .
the maxwell - garnett theory also disregards both the impacts of higher - order dipole contributions and formation of photonic band gaps in the case of arrays of large nanorods .
we have also studied the effect of disorder on the transmittance of the nanorod arrays .
we have introduced two types of disorder , ( a ) ordered array with randomly varying nanorod diameters , and ( b ) a random distribution of nanorods of the same size within the blend .
the disorder in rod placement leads to a strong suppression of the transmission ( and the enhanced absorption ) due to plasmonic resonances related to the clustering of the rods .
we have demonstrated that clustering effects are sensitive to the actual geometry of the structure .
in contrast , the impact of randomly varying diameters of the rods is much less profound . despite its partial adequacy for the te - polarized light ,
the maxwell - garnett effective - medium theory is shown to be invalid for the case of tm polarization .
it simply averages the effective dielectric function inside the blend , missing the important screening properties of the metallic nanorods and characterizing the blend as a ( partially ) transparent medium .
in contrast , the numerical modelling shows the strong skin effect that fully prohibits the propagation of the tm modes through the structure .
the region of high transmittance for the te modes and the strong skin effect for the tm modes makes the nanorod arrays promising candidates for high - quality polarizers .
we would like to thank olle ingans for stimulating and fruitful discussions .
we acknowledge a partial financial support from the center for organic electronics at linkping university .
useful conversations with nils - christer persson are also appreciated . | we study propagation of tm- and te - polarized light in two - dimensional arrays of silver nanorods of various diameters in a gelatin background .
we calculate the transmittance , reflectance and absorption of arranged and disordered nanorod arrays and compare the exact numerical results with the predictions of the maxwell - garnett effective - medium theory .
we show that interactions between nanorods , multipole contributions and formations of photonic gaps affect strongly the transmittance spectra that can not be accounted for in terms of the conventional effective - medium theory .
we also demonstrate and explain the degradation of the transmittance in arrays with randomly located rods as well as weak influence of their fluctuating diameter . for tm modes we outline the importance of skin - effect , which causes the full reflection of the incoming light .
we then illustrate the possibility of using periodic arrays of nanorods as high - quality polarizers . |
Subsets and Splits
No saved queries yet
Save your SQL queries to embed, download, and access them later. Queries will appear here once saved.