id
stringlengths 30
36
| source
stringclasses 1
value | format
stringclasses 1
value | text
stringlengths 5
878k
|
|---|---|---|---|
no-problem/9906/astro-ph9906276.html
|
ar5iv
|
text
|
# Do galaxy mergers form elliptical galaxies? A comparison of kinematic and photometric properties
## 1 INTRODUCTION
The fate of mergers of gas-rich galaxies is an enduring question, which has attracted significant observational and theoretical interest. Such mergers must happen, and are probably common in at least some galaxy environments; but opinion is sharply divided on the end-product of this process. This is a possible mechanism for producing elliptical galaxies, but it has also been argued that diffuse, dynamically cold disk-dominated systems cannot become highly centrally-concentrated elliptical galaxies, and that the end-product of a disk-disk merger is likely to be something much more flimsy and diffuse (Ostriker 1980). Given this uncertainty, it is clearly of interest to study the properties of systems that are generally acknowledged as merger remnants, to determine their likely fates. This is the purpose of the present paper, with the ultimate aim of this programme being to investigate a possible evolutionary sequence between mergers and elliptical galaxies.
There are problems with using optical imaging and spectroscopy to study the dynamics of mergers. Mergers are extremely dusty systems, with visual extinctions of $``$10 magnitudes towards their nuclei in the early stages of the merging process, and the dust tends to be in a very irregular distribution. This causes great uncertainties in the interpretation of optical measurements of these systems, since the optical emission must come from an irregular outer shell of the galaxy which represents the most disturbed and unrelaxed component. This is particularly unfortunate for studies of the progression of the relaxation process in mergers, since relaxation will first occur in the central region where the characteristic timescales are shortest. These regions can in principle be observed in HI 21 cm emission, which is unaffected by the extinction problem, but many galaxies are known to have central deficiencies of atomic gas, and in very disturbed systems it is far from certain that the dynamics of the gas component bear any relation to the stellar component we are interested in here.
The near-infrared thus provides the best techniques for measuring directly the kinematics of the stellar component in the presence of extinction. K-band (2.2$`\mu `$m) light is affected by extinction an order of magnitude less than blue light, and thus can probe deep into the centres of these galaxies. This light is dominated by the old stellar population, at least in normal galaxies. Rix and Rieke (1993) find that in the disks of star-forming galaxies, there is a significant contribution from red supergiants, on the order of 15-25% of the K-band light, but this is very much less than the contribution of young stars to the optical emission in such systems. We consider the effects of supergiant emission in the K band for an extreme starburst system later in the paper. It is fortunate also that the K band contains the strong CO absorption feature which is well suited to the measurement of velocity dispersions of galaxies (Doyon et al. 1994b; Lester and Gaffney 1994; Gaffney, Lester & Doppmann 1995). Since this is a photospheric absorption, it traces directly the kinematics of the stellar population. Thus, by combining K band photometry with CO spectroscopy, it is possible to study the structure and kinematics of the stellar population in the central regions of merging galaxies. This is the aim of the present work.
## 2 PREVIOUS WORK
Many authors have contributed to the literature on this question. Toomre and Toomre (1972) made the cautious suggestion that dynamical friction and consequent merging of pairs of disk galaxies “seems like a scenario for nothing less than the delayed formation of some elliptical galaxies- or at least of major stellar halos from otherwise gas-rich disks.” Many numerical simulations of the merging process have been performed (e.g. Toomre and Toomre 1972; White 1979; Farouki and Shapiro 1982; Barnes 1988), and have almost unanimously concluded that the R<sup>1/4</sup> profile typical of luminous elliptical galaxies is the natural outcome of the merging process. This has received some observational support, with both optical (Schweizer 1982) and near-infrared (Wright et al. 1990; Stanford and Bushouse 1991) imaging showing the apparent emergence of such profiles in ongoing mergers. There are some problems with the merger/elliptical scenario, however. The phase-space argument of Ostriker (1980) remains unresolved, although the discovery of strong central concentrations of molecular gas in mergers (Sanders et al. 1986) and subsequent nuclear starbursts (Joseph and Wright 1985) may contribute the extra stellar density needed to resolve this discrepancy. The apparent deficit of globular clusters in spirals relative to ellipticals also remains problematic, with the suggestion that globulars may be made during the merging process (Schweizer 1986) now being seriously challenged (Whitmore and Schweizer 1995). It is also possible that the R<sup>1/4</sup> profiles in some or all of the mergers are due to pre-existing bulge components, and may not be indicative of the light profile of the entire system when relaxation has completed.
While this paper was in preparation, a very similar study was published by Shier and Fischer (1998). They also used the CO absorption feature to determine central velocity dispersions, in their case for a sample of 10 galaxy mergers. In general, their conclusions are very similar to ours, and the results will be compared throughout the present paper.
## 3 GALAXY SAMPLE
Galaxies were chosen on the basis of a variety of visual morphological criteria, which are generally accepted as being indicative of disk galaxy mergers. The most important of these is the presence of extended, narrow tails which are found in N-body simulations to be a general feature of galaxy mergers, and are caused by the strong tidal field generated during the merging process. The central regions of the mergers are visually chaotic, due in large part to the disrupted dust lanes of the pre-merger spirals. Redshifts were restricted to $`<`$15000 kms<sup>-1</sup> to ensure that the CO absorption edge, at a rest wavelength of 2.293$`\mu `$m, remained well within the K atmospheric window.
## 4 OBSERVATIONS
All observations presented here were made with the 3.8 metre United Kingdom Infrared Telescope (UKIRT) on Mauna Kea, Hawaii. K-band imaging was obtained on the 3 nights 1990 April 30 - May 2, using the near- IR camera IRCAM, which employed a 62$`\times `$58 pixel InSb array. The 60 mm focal length camera was used, giving a pixel scale of 1.24 arcsec. All the observations were made in the standard K filter. Exposure times were typically 10-15 seconds on chip before reading, to give sky-background-dominated noise, co-added to give 30-40 minutes total exposure per galaxy. Equal time was spent on nearby sky positions to build up sky flats. Telescope pointings were jittered around the mean position to increase the area observed around galaxies, and to enable the sky flats to be median filtered, thus minimising the effect of foreground stars. Seeing was typically $``$1.5 arcsec, but was difficult to estimate exactly due to the large pixel scale used.
High resolution spectroscopy at 2.3$`\mu `$m was obtained using the spectrometer CGS4 on UKIRT on the 3 nights 1994 June 6-8. At that time, CGS4 used a 62$`\times `$58 pixel InSb array. We used the high resolution echelle grating and the long focal length camera with a 2-pixel wide slit, giving a FWHM spectral resolution equivalent to 16 km s<sup>-1</sup> and a spectral range of 0.01 $`\mu `$m. Given the small spectral range, two grating positions were used for each galaxy, giving a total spectral coverage of 2.285-2.305 $`\mu `$m in the galaxy’s restframe. For each galaxy, between 15 and 20 integrations each with exposure times of $``$180 seconds were taken. The telescope was nodded so as to slide the target galaxy up and down the slit between integrations to facilitate sky subtraction. Observations were also made of A stars to enable atmospheric absorption features to be divided out, and BS 4737, a K1 III star, was observed to provide a template of the CO absorption profile.
## 5 DATA REDUCTION
### 5.1 Imaging
Imaging data reduction was standard, using the KAPPA and FIGARO packages. All frames were bias- and dark-subtracted, using dark frames with the same on-chip exposures as the science frames and scaled to the same number of coadds. Flat-fielding was performed using median sky flats taken in blank sky areas adjacent to the galaxy concerned, and with the same exposure time as the galaxy. Attempts to improve the signal-to-noise of the flats by including frames taken over a longer period of time yielded lower-quality final images, probably due to changes in the effective flat field pattern as a function of changing sky colour. Each galaxy was observed in at least 4 ‘jittered’ pointings, and these were combined into mosaics after flat-fielding and sky subtraction of the individual sub-arrays. Photometric calibration was taken from images of the standard stars GL 406, HD 77281, HD 84800, HD 105601, HD 106965, HD 136754, HD 161903 and HD 162208, which showed an overall photometric scatter of $`\pm `$0.04 mag, when the zero points for all three nights are combined. The galaxy images were calibrated using the zero point from the standard nearest in time and airmass, and differential airmass corrections were very small, typically 0.01 or 0.02 mag. The derived galaxy light profiles for NGC 2623 and Arp 220 were in excellent agreement with those in Wright et al. (1990), which had been obtained in a previous run with IRCAM on UKIRT. All K-band magnitudes and surface brightnesses were corrected for redshifting of the passband and for surface brightness dimming according to the prescription of Glazebrook et al. (1995) who derive a single correction formula applicable to all Hubble types. This correction has a maximum value of –0.11 mag. for the present sample, and a mean of –0.06 mag. For the present paper, the final stage of data reduction was to obtain multi-aperture photometry with circular software apertures, regardless of the ellipticity of the galaxies, for comparison with the similar measures taken by Mobasher et al. (1999) to derive log(D<sub>K</sub>) values (described in the section 6.3 below) for their sample of ellipticals. We did this for the mergers by using the Starlink routine APERADD, centring the apertures using the CENTROID package with a sampling window of 9$`\times `$9 pixels. Aperture sizes were iterated to find the value within which the mean galaxy surface brightness was 16.5 mag/square arcsec at K, as used by Mobasher et al. (1999). Column 4 of table 1 gives the derived log(D<sub>K</sub>) values, where D<sub>K</sub> is measured in arcsec, corrected to the distance of the Coma Cluster for direct comparability with the Coma ellipticals. This correction was done assuming an unperturbed Hubble Flow and a Coma Cluster redshift of 6942 km s<sup>-1</sup> Mpc<sup>-1</sup>. A Hubble constant of 75 km s<sup>-1</sup> Mpc<sup>-1</sup> was assumed throughout. Column 6 gives the log of the effective diameter A<sub>e</sub>, in arcsec and again corrected to the Coma distance. A<sub>e</sub> (the diameter encompassing half the total light for a pure R<sup>1/4</sup> profile) was approximated by a Petrosian diameter (Petrosian 1976) with index $`\eta =`$ 1.39 (Kjaergaard 1993). $`\eta `$ is the difference between the surface brightness at a given radius and the mean surface brightness within that radius. This definition of A<sub>e</sub> was used as it is insensitive to departures from pure R<sup>1/4</sup> profiles, and should be relatively unaffected by fading in the surface brightness of galaxies. Column 7 gives the mean K-band surface brightness within the diameter A<sub>e</sub>.
The other data listed in Table 1 are catalogued names (column 1), number in the Arp (1966) list of peculiar and disturbed galaxies (column 2), total K-band absolute magnitude (column 3), and best-fitting power-law index to the light profile, as described in section 6.1 below (column 5).
### 5.2 Spectroscopy
Initial data reduction was done at the telescope, using an automatic reduction process to dark- and bias- subtract all frames, to flat-field using internally-generated lamp flats, and to ratio by an A-star spectrum to remove telluric absorptions. The galaxies were slid up and down the slit to facilitate sky-subtraction, and the processed ‘Reduced Group’ frames contained two spectra, one positive and one negative, following the automatic subtraction of spectra taken at the two positions. Known bad pixels in the array were automatically set to magic values, so that they would be ignored in subsequent reduction, but it was also necessary to identify additional bad pixels and cosmic ray hits in the Reduced Groups. Wavelength calibration was done using an argon arc lamp in CGS4. No spectrophotometric calibration was attempted, as it was not required for this project.
The positive and negative spectra were then extracted from each Reduced Group frame and subtracted to give the final galaxy spectrum. For the template star BS 4737, a long baseline spectrum was obtained by observing at 12 grating positions. These spectra were combined using the ADJOIN routine in FIGARO, giving the final stellar template spectrum.
The final stage of data reduction is to determine the velocity dispersions in the galaxy mergers from the Doppler-smoothing of the CO absorption. This absorption has an intrinsically steep edge at 2.293 $`\mu `$m which makes it sensitive to such effects. Following Doyon et al. (1994b) and Gaffney et al. (1995) we assume that the velocity smoothing function can be approximated by a Gaussian of the form $`f(x)=exp((x^2/2\sigma ^2))`$, where $`x`$ is displacement in velocity from the systemic velocity of the galaxy and $`\sigma `$ is proportional to the width of the Gaussian and hence the velocity dispersion. To determine the velocity dispersion, we convolved the template spectrum of BS 4737 with Gaussians of different widths, and determined the best fit of the resulting spectrum to the measured galaxy spectrum. In each of these fits, the free parameters were a multiplicative scaling of the spectral fluxes, an additive constant to the spectral fluxes, and a wavelength shift due to redshift differences between the galaxy and template star. In each case, well-defined minima were found in the reduced chi-squared of the fits with respect to each of the four free parameters. Having identified the best fit, errors were calculated for the width of the velocity Gaussian by freezing the other parameters and changing the Gaussian width until the reduced $`\chi ^2`$ increased by the factors corresponding to 1- and 2-$`\sigma `$ uncertainties.
Table 2 lists parameters derived from spectroscopic measurements. Column 1 gives the galaxy names, column 2 the catalogued redshifts, and column 3 the redshifts derived from the spectroscopic fits described above, quite independently of the catalogued redshifts. Good agreement is found with the catalogued redshifts, giving confidence in the fitting process. Column 4 of Table 2 gives the derived velocity dispersions with 1-$`\sigma `$ errors, and column 5 the spectroscopic CO index, which is discussed further in section 6.3 below.
## 6 ANALYSIS
### 6.1 Light profiles
We first address the question of whether the elliptical-like light light profiles found for some mergers could be due to pre-existing bulges which have survived the merging process. Figure 1 shows a comparison of the best-fit indices of the K-band light profiles of the mergers (plotted as circles) with the bulges of spirals (crosses) of Hubble types Sa–Sd (Seigar and James 1998), and two ellipticals (squares), NGC 5845 and NGC 6023. The fitted profiles are generalised de Vaucouleurs profiles, and the index plotted is $`(1/n)`$ in the $`R^{(1/n)}`$ relation, such that a true de Vaucouleurs profile would correspond to 0.25. The errors were calculated by omitting 0, 1, 2 and 3 central points from the profiles and redoing the fits. This excluded a circular area of radius up to 3.7 arcsec, significantly larger than the seeing size, and generally had little impact on the measured indices. The spiral galaxy indices generally lie between 0.4 and 1.0, with some evidence for a bimodal distribution with $``$0.5 and $``$1.0 as preferred values. Similar results were found by de Jong (1996). However, the mergers show significantly brighter total luminosities and cuspier light profiles, with indices lying close to the value of 0.25, as do the two ellipticals for which we have K-band imaging. However, not all of the mergers plotted are actually well-fitted by a de Vaucouleurs profile, with some having significant ripples and irregularities in their profiles. The important result from Figure 1 is that even these systems show at least the degree of central light concentration expected of an elliptical, and their light distributions and luminosities are not those expected of pre-existing spiral bulges, or of galaxy disks. Indeed, there is even some evidence that the mergers have lower indices and cuspier profiles even than ellipticals, which may be evidence of an additional, centrally-concentrated component in the K-band light, resulting from a nuclear starburst. This possibility is discussed further in section 6.3.
The referee commented that there appears to be a trend in profile index with absolute magnitude, and suggests that the lower values found for the mergers may simply reflect a selection bias, since the mergers are significantly more luminous than the bulges, on average. However, if there is such an effect, it does not appear to be strong enough to cause the result we find. Regressing index on absolute magnitude, there is a weak trend both for the bulges and mergers, in the sense that more luminous systems have somewhat lower indices. However, the correlations have significances of less than 5% in both cases. Extrapolating the trend found for bulges to the mean absolute magnitude of the mergers predicts that the average profile index for the mergers should be 0.51, whereas the average of the measured indices is 0.23. Thus the mergers genuinely seem to be offset to lower profile indices than bulges, even if we were to compare with bulges of similar luminosity (which do not seem to exist in any case).
The merger with the large error bar in Figure 1 is NGC 6052, which has a highly irregular structure, even in K-band light. Since the errors are estimated from fitting the profile over different radial ranges, it would appear that the unusual and highly irregular profile of NGC 6052 is particularly sensitive to such changes. It is clear that this galaxy bears no resemblance to an elliptical galaxy, and appears quite different from the other mergers.
### 6.2 Velocity dispersions
Figure 2 shows the range of velocity dispersions found for ellipticals and bulges by Whitmore, McElroy & Tonry (1985). These are split into ellipticals (dashed line), lenticulars (solid line) and spirals (dotted line). The measured velocity dispersions for the mergers are also marked at the top of Figure 2. It can be seen that the majority of mergers have velocity dispersions between 100 and 200 kms<sup>-1</sup>, with a mean of 164$`\pm `$35 kms<sup>-1</sup>, where the standard error on the distribution is quoted. This compares with mean velocity dispersions of 222$`\pm `$5 kms<sup>-1</sup> for ellipticals, 190$`\pm `$4kms<sup>-1</sup> for lenticular bulges, and 142$`\pm `$4 kms<sup>-1</sup> for spiral bulges, from values in the Whitmore et al. (1985) compilation. NGC 6090 has the lowest velocity dispersion of any of the mergers, and this may well be related to the fact that it has two widely-separated nuclei. Thus for this galaxy we are almost certainly seeing the velocity dispersion of one of the pre-existing nuclei, and this value is likely to be significantly lower than the disperson of the final, relaxed merger remnant. At the other extreme, NGC 6240 has a velocity dispersion larger than any of the spiral galaxy bulges in the Whitmore et al. (1985) compilation, and the value is indeed in the high-velocity tail of the distribution even for elliptical galaxies. Again, this may not represent the final virialised velocity dispersion of this system. For example, there may be superposed components with different systemic velocities along the line of sight, which would be impossible to distinguish using the present data. Such effects cannot be gross, since then the spectroscopy would resolve two separate CO absorption features. However, given the width of the CO feature and limited signal-to-noise of the present data, no strong constraints can be placed.
The distribution of velocity dispersions of the merger sample is shown by a Kolmogorov-Smirnov test to be formally consistent with those of both spiral bulges and ellipticals, and so our velocity data alone do not permit any significant test of the dynamical evolution of mergers. In this, we differ somewhat from Shier and Fischer (1998), who conclude that their velocity dispersion measurements are not consistent with the values expected for L elliptical galaxies, since apart from NGC 6240, the highest velocity dispersion they measure is 151 km s<sup>-1</sup>.
### 6.3 Parameter correlations
A strong test of the merger hypothesis is made possible by the parameter correlations resulting from the ‘Fundamental Plane’ of elliptical galaxies (Dressler et al. 1987, Djorgovski & Davis 1987), since we know that any forming elliptical will have to arrive on this plane. This describes the tight 2-dimensional locus occupied by elliptical galaxies in the 3-dimensional space defined by velocity dispersion, luminosity and scale-size. These latter two parameters can be conveniently grouped together, and Dressler et al. (1987) found that a particularly useful quantity was D<sub>n</sub>, the diameter of a photometric aperture within which the mean surface brightness of the galaxy has some value $`n`$. Dressler et al. took $`n`$ to be 20.75 mag. per square arcsec in the optical B passband, and found a very tight correlation between the resulting D<sub>n</sub> values and velocity dispersion. Mobasher et al. (1999) have examined the equivalent relation (which they term the D<sub>K</sub>$`\sigma `$ relation) using K-band photometry of elliptical galaxies in the Coma cluster, and find a small scatter of 0.076 dex in the relation, similar to the optical D<sub>n</sub>$`\sigma `$ relation. They also present a plot which explicitly shows an edge-on projection of the K-band Fundamental Plane, thus reducing the scatter even below that of the D<sub>K</sub>$`\sigma `$ relation. Here the plotted quantities are 1.38log$`\sigma +`$0.3$`<`$SB<sub>K</sub>$`>_e+c_1`$ vs logA<sub>e</sub>, where A<sub>e</sub> is the effective diameter and $`<`$SB<sub>K</sub>$`>_e`$ the mean surface brightness within this diameter. The dataset of Mobasher et al. thus provides an excellent way to test whether our mergers obey the same parameter correlations as do elliptical galaxies, since all the parameters required for these relations are provided by our K-band photometry and spectroscopy.
D<sub>K</sub> values were calculated using circular aperture photometry on the K-band images, and finding the size in arcseconds of a circular aperture which enclosed a mean surface brightness of 16.5 mag. per square arcsec. This surface brightness, and the use of circular apertures, were identical to the procedure followed by Mobasher et al. (1999), and so the results should be directly comparable. Figure 3 shows the D<sub>K</sub>$`\sigma `$ relations for Coma ellipticals (crosses) and for the mergers (circles). In both axes, the quantities actually plotted are logarithms to base 10. The ellipticals show the expected correlation, whereas the mergers scatter well above this locus. This implies either that the mergers have anomalously low velocity dispersions, or anomalously large D<sub>K</sub> diameters, or a combination of both, when compared with the Coma ellipticals.
A similar result is shown in Figure 4, which compares the near-IR Fundamental Plane of Coma ellipticals with the present merger sample. Again the elliptical galaxies show a very tight correlation, whereas the mergers scatter away from the Fundamental Plane in the same sense as was found from Figure 3. We will now discuss whether the offset is likely to be due to anomalous velocity dispersions or surface brightnesses in the mergers.
As noted above, the distribution of velocity dispersions of the merger sample is consistent with that of elliptical galaxies (formally the means of the two distributions differ by 1.6 $`\sigma `$ but the Kolmogorov-Smirnov test shows the distributions to be indistinguishable), and is unlikely to evolve further in these relaxed central regions. This consistency is confirmed by Figure 3, with the possible exception of NGC 6090, which may evolve to a higher velocity dispersion as the two nuclei relax together. For the remainder of the mergers, any plausible evolution onto the elliptical locus in Figure 3 seems likely to be predominantly in the D<sub>K</sub> direction, and in every case would have to be in the sense of making the galaxies become smaller, at the same mean surface brightness level. This is quite plausible, given that mergers are widely acknowledged to be linked to highly luminous bursts of massive star formation (e.g. Joseph & Wright 1985; Barnes & Hernquist 1991; Solomon, Downes & Radford 1992; but see also the counter arguments of Thronson et al. 1990). Such a starburst will result in a temporarily enhanced luminosity and surface brightness of the system, and in the K-band this is likely to result principally from emission from supergiant stars. Thus we would argue that the most likely future evolution of the mergers will shift them vertically in Figure 3, in the direction of smaller D<sub>K</sub> values. Assuming the starburst population to be well mixed with the older stars, this fading will not affect the A<sub>e</sub> values, and will shift points vertically upwards in Figure 4, again consistent with moving them onto the Fundamental Plane defined by Coma ellipticals. We will now estimate the likely size of this effect.
The photometric contribution of supergiants in mergers can be quantified from the depth of the 2.3$`\mu `$m CO absorption feature, which is characteristic of such stars (Doyon, Joseph & Wright 1994a; Ridgway, Wynn-Williams & Becklin 1994; Goldader et al. 1995). Using a method described by Doyon et al. (1994a) it is possible to use low-resolution K-band spectroscopy to make an estimate of the fraction of K-band light contributed by the starburst population in a galaxy from the depth of the CO feature. This method assumes that the K-band light has contributions from an old stellar population with a small CO index, and a starburst population with a much larger CO index. From the observed index, which lies between these extremes, one can then solve for the fractional mix of the two populations. This simple form of population synthesis was applied to the 5 mergers for which CO indices were available (Mkn 273, Arp 220, NGC 6090 and NGC 6240 from Ridgway et al. 1994; IC 883 from Goldader et al. 1995), assuming a spectroscopic CO index (see Doyon et al. 1994a for a definition) of 0.2 for the old population and 0.34 for a pure starburst.
It was found that the starburst fraction varied from providing only $``$20% of the K-band luminosity in Mkn 273, to $``$85% in Arp 220. We then made the simplifying assumptions that all of this starburst population would ultimately disappear, and that this would uniformly depress the surface brightness of the galaxy in the K-band. It was then straightforward to predict the effect of such fading on the D<sub>K</sub> values for these 5 mergers, and the results are shown as arrows in Figure 3. Clearly, this correction moves the points towards the elliptical locus, and the movement is of the right order to explain the present discrepancies. The equivalent fading vectors in Figure 4 would shift the points vertically upwards by between 0.1 and 0.6 in the quantity plotted on the y axis, again of the correct order to shift points onto the Fundamental Plane. The corresponding arrows are not shown in this figure to avoid confusing the plot.
A further test of this fading hypothesis is shown in Figure 5, where log(A<sub>e</sub>) is plotted against log($`\sigma `$). Given the definition of A<sub>e</sub> as a Petrosian diameter, it should be relatively unaffected by fading, and certainly far less so than D<sub>K</sub> and $`<`$SB<sub>K</sub>$`>_e`$. Given this, it is reassuring to see that the distribution of mergers is much more consistent with that of ellipticals in this figure than in figures 3 and 4, and that the one very discrepant point, the double nucleus system NGC 6090, is offset to lower log($`\sigma `$), as suggested earlier.
If there are red supergiants in the numbers necessary to explain the observed CO indices, then for any plausible star formation history and IMF it is to be expected that younger OB stars should also be present. This is confirmed for the mergers plotted with arrows in Figure 3; Goldader et al. (1997) find strong Brackett $`\gamma `$ emission from all five, consistent with the presence of OB stars.
## 7 DISCUSSION
A study of the dynamics of mergers was undertaken by Lake & Dressler (1986), who compared the luminosity–velocity dispersion (L–$`\sigma `$) relation with that of ellipticals (Faber & Jackson 1976). They combined B and V photometry from the literature with Reticon spectroscopy of the Ca triplet, and found a remarkably good agreement between the distribution of velocity dispersions and optical luminosities for mergers and ellipticals. This result was argued to be at variance with a simple dynamical model, which predicted that merging should not lead to an increase in velocity dispersion. Thus they concluded that the merging process must involve significant dissipation for the merger products to follow the L–$`\sigma `$ relation as found.
The L–$`\sigma `$ relation is very closely related to the D<sub>n</sub>$`\sigma `$ relation we have investigated. The apparent difference in results is probably due to the merger sample selection. The only merger in common between the two studies is NGC 7252, and this is possibly the most evolved merger in our sample, but the youngest merger observed by Lake & Dressler. The majority of their objects are recognisable as elliptical galaxies, but with peculiarities such as strong shells and ripples, which make it likely that these are mergers in the late stage of the relaxation process, which is only beginning for most of our galaxies. Thus it is not surprising that different results are found by the two studies.
Shier and Fischer (1998) perform a K-band Fundamental Plane analysis of their merger sample, like that of the present paper, and reach very similar conclusions. They also find that most of the mergers are displaced from the plane defined by ellipticals, in the sense of being bright for a given velocity dispersion. From a population synthesis argument, they estimate that the mergers are likely to fade by 1.5–2.0 mag in the K band during the next 3 Gyr, and that this fading will place them on or near the elliptical locus, assuming no other scale size or dynamical changes.
## 8 CONCLUSIONS
The principal conclusions of this study are as follows. We confirm that in terms of their near-IR photometric properties, ongoing galaxy mergers more nearly resemble elliptical galaxies than they do spiral bulges. Even those poorly fitted by a de Vaucouleurs profile show a ‘cuspy’ light distribution which is likely to smooth into an elliptical-like profile through relaxation processes. Using velocity dispersions alone we are unable to distinguish whether mergers more closely resemble bulges or ellipticals, but this is unsurprising given the small number of mergers in our sample and the extensive overlap between bulge and elliptical $`\sigma `$ distributions. A clear discrepancy is found between the Fundamental Plane distributions and D<sub>K</sub>$`\sigma `$ relations of mergers and ellipticals, however. We suggest that this may be due to a short-lived population of supergiant stars temporarily increasing the surface brightness of the mergers, and find evidence for such a population at approximately the predicted level from literature measurements of the CO absorption strength.
It is clear that significant improvements on these results are attainable with existing and planned instrumentation. An intriguing possibility is that it may be possible, by combining samples of young and old mergers, to map out an evolutionary sequence from the highly distorted and dusty systems to the relaxed, old-star-dominated elliptical galaxies, from their positions in the Fundamental Plane parameter space. Key issues here would be the relative importance of star formation, stellar evolution effects, dynamical relaxation and dissipation in the formation of the Fundamental Plane.
## 9 ACKNOWLEDGMENTS
PJ thanks Sue Wild for many useful comments on a draft of this paper. The referee is thanked for a careful reading of the paper and several helpful suggestions which improved the content and clarity of presentation. The United Kingdom Infrared Telescope is operated by the Joint Astronomy Centre on behalf of the U.K. Particle Physics and Astronomy Research Council. This research has made use of the NASA/IPAC Extragalactic Database (NED) which is operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration.
## 10 REFERENCES
Arp H., 1966, ApJS, 14, 1
Barnes J.E., 1988, ApJ, 331, 699
Barnes J.E., Hernquist L.E., 1991, ApJ, 370, L65
de Jong R. S., 1996, A&A, 313, 45
Djorgovski S., Davis M., 1987, ApJ, 313, 59
Doyon R., Joseph R.D., Wright G.S., 1994a, ApJ, 421, 101
Doyon R., Wells M., Wright G.S., Joseph R.D., Nadeau D., James P.A., 1994b, ApJ, 437, L23
Dressler A., Lynden-Bell D., Burstein D., Davies R.L., Faber S.M., Terlevich R., Wegner G., 1987, ApJ, 313, 42
Faber S.M., Jackson R.E., 1976, ApJ, 204, 668
Farouki R.T., Shapiro S.L., 1982, ApJ, 259, 103
Gaffney N.I., Lester D.F., Doppmann G., 1995, PASP, 107, 68
Glazebrook K., Peacock J.A., Miller L.A., Collins C.A., 1995, MNRAS, 275, 169
Goldader J.D., Joseph R.D., Doyon R., Sanders D.B., 1995, ApJ, 444, 97
Goldader J.D., Joseph R.D., Doyon R., Sanders D.B., 1997, ApJS, 108, 449
Joseph R.D., Wright G.S., 1985, MNRAS, 214, 87
Kjaergaard P., Jorgensen I., Moles M., 1993, ApJ, 418, 617
Lake G., Dressler A., 1986, ApJ, 310, 605
Lester D.F., Gaffney N.I., 1994, ApJ, 431, L13
Mobasher B., Guzmán R., Aragón-Salamanca A., Zepf S., 1999, MNRAS, 304, 225
Ostriker J.P., 1980, ComAp, 8, 177
Petrosian V., 1976, ApJ, 209, L1
Ridgway S.E., Wynn-Williams C.G., Becklin E.E., 1994, ApJ, 428, 609
Rix H.-W., Rieke M.J., 1993, ApJ, 418, 123
Sanders D.B., Scoville N.Z., Young J.S., Soifer B.T., Schloerb F.P., Rice W.L., Danielson G.E, 1986, ApJ, 305, 45
Schweizer F., 1982, ApJ, 252, 455
Schweizer F., 1986, in Faber, S.M., ed., Nearly Normal Galaxies. Springer-Verlag, Berlin, p.18
Seigar M.S., James P.A., 1998, MNRAS, 299, 672
Shier L. M., Fischer J., 1998, ApJ, 497, 163
Solomon P.M., Downes D., Radford S.J.E., 1992, 387, L55
Stanford S.A., Bushouse H.A., 1991, ApJ, 371, 92
Thronson H.A., Majewski S., Descartes L., Hereld M., 1990, ApJ, 364, 456
Toomre A., Toomre J., 1972, ApJ, 179, 623
White S.D.M., 1979, MNRAS, 189, 831
Whitmore B.C., McElroy D.B., Tonry J.L., 1985, ApJS, 59, 1
Whitmore B.C., Schweizer F., 1995, AJ, 109, 960
Wright G.S., James P.A., Joseph R.D., McLean I.S., 1990, Nature, 344, 417
Figure 1. Best-fitting profile index vs K-band absolute magnitude for spiral bulges (crosses), ellipticals (squares) and mergers (circles).
Figure 2. Distribution of velocity dispersions taken from Whitmore et al. (1985), showing ellipticals (dashed line), lenticular bulges (solid line), and spiral bulges (dotted line). The measured velocity dispersions for the mergers are indicated at the top of the figure.
Figure 3. The D<sub>K</sub>$`\sigma `$ relation for Coma ellipticals (crosses), taken from Mobasher et al. (1999) and for the present sample of mergers (circles). The arrows show the effect of K-band fading inferred from CO indices as explained in section 6.3.
Figure 4. The Fundamental Plane relation for Coma ellipticals (crosses), taken from Mobasher et al. (1999) and for the present sample of mergers (circles).
Figure 5. Log(A<sub>e</sub>) vs Log($`\sigma `$) for Coma ellipticals (crosses), taken from Mobasher et al. (1999) and for the present sample of mergers (circles).
|
no-problem/9906/astro-ph9906234.html
|
ar5iv
|
text
|
# Iron line afterglows: general constraints
## 1 Introduction
At this meeting it has been reported the possible detection of an iron line in the X–ray afterglow spectra of two bursts: GRB 970508 detected by BeppoSAX (Piro et al. 1999, see also these proceedings) and GRB 970828 by ASCA (Yoshida et al. these proceedings). The significance of these features is admittedly not extremely compelling ($`99\%`$ and $`98\%`$ for the two events, respectively), but the implications that they bear are so important to justify a study on the radiation mechanisms that would produce it (Lazzati et al. 1999, see also these proceedings). We can also put strong constraints on the size and mass of the line emitting region in a model–independent way. These limits, which are very robust, point towards the presence, within a distance of $`10^{16}`$ cm from the burst, of at least $`10^3`$ $`M_{}`$ of iron. We also stress that these emission line features, observed in bursts which also have an optical afterglow, strongly suggest that the line emitting region is not spherically symmetric, but must have some degree of anisotropy. Finally, this constrains the possible anisotropy of the burst radiation, since the line emitting region samples different line of sights than our own.
The cosmological parameters will be set throughout this paper to $`H_0=65`$ km s<sup>-1</sup> Mpc<sup>-1</sup>, $`q_0=0.5`$ and $`\mathrm{\Lambda }=0`$.
## 2 General Constraints
We here prefer to discuss the implications of detecting an iron line during the X–ray afterglow of a generic burst, without referring in particular to the two cases mentioned in the introduction. Let us therefore assume that the flux of the X–ray afterglow is of the order of 10<sup>-13</sup> erg cm<sup>-2</sup> s<sup>-1</sup>, and assume a redshift of $`z=1`$. To be visible during the X–ray afterglow emission, the emission line should have a comparable flux<sup>1</sup><sup>1</sup>1Here and in the following we parametrize a quantity $`Q`$ as $`Q=10^xQ_x`$ and adopt cgs units., $`F_{Fe}=10^{13}F_{Fe,13}`$ erg cm<sup>-2</sup> s<sup>-1</sup>. This in itself constrains both the amount of line–emitting matter and the size of the emitting region.
### 2.1 Limit to the size
Assume for simplicity that the emitting region is a homogeneous spherical shell at a distance $`R`$ from the $`\gamma `$ray burst, with a width $`\mathrm{\Delta }RR`$. The fluence of the emission line cannot exceed the absorbed ionizing fluence $`q`$ (where $``$ is the total GRB fluence and $`q`$ is the fraction of it which is absorbed and reprocessed into the line). The observed duration of the emission line cannot be shorter than the light crossing time of the region $`R/c`$. From this we obtain the limit
$$R<3\times 10^{18}\frac{q_5}{F_{Fe,13}}\mathrm{cm}$$
(1)
Since $`q`$ is at most $`0.03`$, the emitting region is very compact, ruling out emission from interstellar matter, even assuming the large densities appropriate for star forming regions. We would like to stress that the above limit is independent of the variability of the line flux.
### 2.2 Limit to the mass
The total line photons produced at 6.4–6.9 keV in $`10^5t_5`$ seconds, for a GRB located at $`z=1`$, are $`3\times 10^{57}F_{Fe,13}t_5`$. Assuming that each iron atom produces $`k`$ line photons, this corresponds to $`150F_{Fe,13}t_5/kM_{}`$ of iron. The parameter $`k`$ depends on the details of the assumed scenario, but we can set some general limits. Assume in fact that each iron atom can emit photons only when illuminated by an ionizing flux, which is provided by the burst itself or by the high energy tail of the afterglow emission. Since the burst radiation has enough power to photoionize all the matter in the vicinity of the progenitor (see e.g. Boettcher et al. (1999)), line photons will be emitted only through the recombination process. But even if we assume that the recombination is instantaneous, the value of the parameter $`k`$ will not be larger than the total number of photoionizations an ion can undergo during the burst and/or the afterglow. For iron $`K`$–shell electrons, with cross section $`\sigma _K=1.2\times 10^{20}`$ cm<sup>2</sup> we have:
$$k\frac{qE}{4\pi ϵ_{ion}R^2}\sigma _K=6.5\times 10^6\frac{qE_{52}}{R_{16}^2}\left(\frac{9.1\mathrm{keV}}{ϵ_{ion}}\right)$$
(2)
where $`E`$ is the total energy emitted by the burst and/or afterglow and $`ϵ_{ion}`$ the energy of a single ionizing photon. This upper limit on $`k`$ translates in a lower limit on the iron mass:
$$M_{\mathrm{Fe}}\mathrm{\hspace{0.17em}2.3}\times 10^5F_{Fe,13}t_5\frac{R_{16}^2}{qE_{52}}M_{}$$
(3)
which, for a solar iron abundance, yields a total mass $`M0.013F_{Fe,13}t_5R_{16}^2/(qE_{52})M_{}`$, i.e. a third of a solar mass for $`q0.03`$.
Even if these numbers apparently do not rule out reverberation from a molecular cloud (Ghisellini et al. (1999); Mészáros & Rees (1998)), we remark that the recombination time has been assumed negligible in the above discussion. At densities typical of molecular clouds the recombination time is larger than the burst duration and the value of $`k`$ cannot exceed 12, set by the photoelectric yield of the iron atom (e.g. Boettcher et al. (1999)).
### 2.3 Limit on geometry and isotropy
Optical afterglow emission has been observed in about half the $`\gamma `$–ray burst events for which the X–ray afterglow has been detected. In particular, the optical afterglow of GRB 970508 lasted for hundreds of days (Galama et al. 1998). If the iron line emitting region were spherically symmetric, it would inevitably stop the fireball and the usual relatively slow transformation of bulk kinetic energy into radiation could not take place. For this reason it is necessary to assume some special geometry of the iron line emitting material, which has not to interfere with the observed optical afterglow. In other words, this region cannot be located along our line of sight, but, on the other hand, it has to be illuminated by the burst emission, in order to produce the iron line feature (e.g. a torus surrounding the central region, or a bicone). Therefore we conclude that the iron line feature is a powerful tool to know how isotropic the burst emission is.
## 3 Discussion
We have derived some model–independent limits on the size and mass of the material responsible for iron line features, assuming that they can be detected during the X–ray afterglow. This work has been stimulated by the recent claim of detection of such lines, but the validity of our conclusions are quite general. The most important implication of these lines, if confirmed, regard the progenitor of the burst. In fact it is inescapable to assume that about a solar mass of (probably iron rich) material is located in the close vicinity of the event. Among the proposed models, the Supranova of Vietri and Stella (1998) is the only one which naturally accounts for this. This in turn implies that at least a class of $`\gamma `$–ray burst have to be associated with supernovae exploded about a month earlier (assuming that the remnants have a velocity of about $`10^4`$ km s<sup>-1</sup>).
The possible association of GRB with supernovae has been investigated recently in detail by Bloom et al. (1998); Kippen et al. (1998) and Wang & Wheeler (1998), following the explosion of GRB 980425, likely associated with the type Ic SN 1998bw. Among these works, only Wang & Wheeler (1998) find evidence for a connection while the other two limit to a few percent the bursts possibly associated with supernovae. In the Supranova scenario, however, the association of supernovae with bursts suffer for a time delay which would smear the time correlation between the two phenomena.
Should the iron lines possibly detected in GRB 970508 and GRB 970828 be real and confirmed by other cases, then we have a strong case for the connection between supernovae and $`\gamma `$–ray bursts. The next generation of experiments and satellites, such as XMM, AXAF and ASTRO–E, will provide us with the necessary information to draw more accurate conclusion on the puzzling problem of the $`\gamma `$–ray burst progenitor.
|
no-problem/9906/gr-qc9906070.html
|
ar5iv
|
text
|
# Remark on formation of colored black holes via fine tuning
## Abstract
In a recent paper (gr-qc/9903081) Choptuik, Hirschmann, and Marsa have discovered the scaling law for the lifetime of an intermediate attractor in the formation of n=1 colored black holes via fine tuning. We show that their result is in agreement with the prediction of linear perturbation analysis. We also briefly comment on the dependence of the mass gap across the threshold on the radius of the event horizon.
Recently, Choptuik, Hirschmann, and Marsa have discovered a new type of critical behavior within the black hole regime of the spherically symmetric Einstein-Yang-Mills model . Using the horizon excision technique, they were able to follow the evolution of supercritical data beyond the formation of an event horizon. This allowed them to determine the ultimate behavior of the Yang-Mills hair which remains outside the black hole after the formation of an event horizon. In accordance with the no-hair property it is expected that this hair is lost either via radiation to infinity or by collapse to the black hole. Choptuik, Hirschmann, and Marsa found that the mechanism of loosing hair is different for black holes formed via two generalized (in the sense explained in ) types of collapse established in . For a generalized supercritical Type II collapse most of the hair is radiated off to infinity leaving the original black hole virtually unchanged. In contrast, for a generalized supercritical Type I collapse most of the hair collapses and consequently the original black hole gets bigger. These two kinds of behavior are separated by the $`n=1`$ colored black hole (the static spherically symmetric black hole solution of Einstein-Yang-Mills equations ) which plays the role of an intermediate attractor for near-critical solutions. Similarly to the Type I collapse, the lifetime $`T`$ of a near-critical solution staying in the vicinity of the intermediate attractor was found to satisfy the scaling law
$$T\lambda \mathrm{ln}|pp^{}|,$$
(1)
where the coefficient $`\lambda `$ is the characteristic time scale for the decay of the $`n=1`$ colored black hole, and $`|pp^{}|`$ is the distance from the threshold ($`p^{}`$) along a one-parameter family of interpolating initial data.
As described in , the coefficient $`\lambda `$ can be computed in two independent ways: either directly from the nonlinear evolution, or perturbatively via linear stability analysis. Since Choptuik, Hirschmann, and Marsa calculated $`\lambda `$ only in the first manner, here we would like to compare their result with the predictions of linear stability analysis. We think that this comparison not only validates the results of Choptuik, Hirschmann, and Marsa but, in addition, provides a useful test of the precision of their numerical estimates.
We recall that according to the picture of critical solution as a codimension-one attractor, the coefficient $`\lambda `$ is a reciprocal of the eigenvalue $`\sigma `$ corresponding to the single unstable mode, i. e., $`\lambda =1/\sigma `$. The single mode instability (within the magnetic ansatz) of the n=1 colored black holes was established almost a decade ago , however there are no (as far as we know) published results on how the eigenvalue of the unstable mode depends on the radius of the horizon (in one of us determined this relationship in reference with the issue of regularity of the unstable mode at the horizon but no numerical values were given there). We fill in this gap in the literature with Fig. 1, where the eigenvalue $`\sigma `$ is plotted as a function of the radius of the horizon $`r_h`$ of $`n=1`$ colored black hole. This plot was generated by numerically solving, after Straumann and Zhou, their eigenvalue equation for small perturbations .
Choptuik, Hirschmann, and Marsa estimated that $`\lambda 4.88`$ for $`r_h0.55`$ which compares well with our calculation of $`1/\sigma (0.55)=5.0736`$. Unfortunately, this is the only value of $`r_h`$ for which we can make the comparison because Choptuik, Hirschmann, and Marsa did not estimate $`\lambda `$ for other values of $`r_h`$. It is worth noting that since the eigenvalue $`\sigma `$ (and eo ipso the coefficient $`\lambda `$) varies with $`r_h`$, it is universal only among those interpolating families of initial data which share the same critical solution (i. e., have the same $`r_h`$).
Notice that, as follows from Eq.(1) and Fig. 1, for a given distance from the threshold, the lifetime $`T`$ increases with $`r_h`$, so colored black holes with large $`r_h`$ should be more clearly pronounced as intermediate attractors. However, this effect is partly offset by the fact that, in the limit of large $`r_h`$, colored black holes become almost indistiguishable from the Schwarzschild black hole because the energy of the Yang-Mills hair rapidly decreases with $`r_h`$. In other words, the mass gap across the threshold is expected to decrease with $`r_h`$. The upper bound for the mass gap is given by
$$\mathrm{\Delta }M(r_h)=M_{cbh}(r_h)\frac{r_h}{2},$$
(2)
which is the difference between the mass $`M_{cbh}(r_h)`$ of the $`n=1`$ colored black hole and the mass of the Schwarzchild black hole with the same radius of event horizon $`r_h`$. This maximal mass gap would be achieved if all hair remaining outside the horizon of near-critical solutions would either completely disperse, or completely collapse. Since in the real collapse there is always a fraction of hair which does otherwise, the actual mass gap is smaller than $`\mathrm{\Delta }M`$. The maximal mass gap given by Eq.(2) is plotted in Fig. 2.
It would be interesting to compare Fig. 2 with the actual mass gap. In particular, one might wonder whether the mass gap goes to zero for a large but finite $`r_h`$. If this really happened, then the line of colored black holes on the phase diagram shown in Fig. 4 in , beginning at the triple point for $`r_h=0`$, would terminate at a finite distance, in an amusing similarity to the gas-liquid boundary on phase diagrams for typical substances. We realize that the numerical verification of this speculation would be rather difficult.
Acknowledgments. This research was supported by the KBN grant 2 P03B 010 16.
|
no-problem/9906/astro-ph9906448.html
|
ar5iv
|
text
|
# Brown Dwarfs and the Cataclysmic Variable Period Minimum
## 1 Introduction
Cataclysmic variables (CVs) are semi–detached binaries with a white dwarf (WD) primary and a low–mass main–sequence companion that transfers mass to the WD through Roche–lobe overflow (e.g. Warner 1995). Orbital periods are known for about 400 systems. Two marked features stand out in the orbital period distribution: a dearth of systems in the range 2-3 hr, usually referred to as the “period gap”, and a sharp short–period cut–off at $`80`$ min, the “period minimum” $`P_{\mathrm{min}}`$ (e.g. Ritter & Kolb 1998). J0132–6554 at $`P=77.8`$ min currently marks the short end of the bulk of the distribution, while the single system V485 Cen at $`P=59`$ min is the notable exception. Six AM CVn–type CVs with yet shorter periods are interpreted as CVs with helium donors.
In a semi–detached binary the Roche lobe filling star’s mean density $`\rho `$ determines the orbital period $`P`$ almost uniquely (e.g. King 1988), $`P_h=k/\rho _{}^{1/2}`$ (where $`k8.85`$ is only a weak function of the mass ratio, $`P_h`$ is the period in hr, $`\rho _{}`$ the mean density in solar units). The density increases for donors evolving along the hydrogen–burning main sequence towards smaller mass. Approaching the hydrogen–burning minimum mass the increasing electron degeneracy induces structural changes such that further mass loss reduces $`\rho `$. Hence during the donor’s transition from a main–sequence star to a brown dwarf (BD) the secular mean orbital period derivative changes from negative to positive. This “period bounce” has long been identified with the minimum period of CVs (Paczyński 1981, Paczyński & Sienkiewicz 1981; Rappaport, Joss & Webbink 1982; D’Antona & Mazzitelli 1982).
The actual period $`P_{\mathrm{turn}}`$ at which a CV bounces depends on the ratio $`\tau =t_{\mathrm{KH}}/t_M`$ of the secondary’s thermal time $`t_{\mathrm{KH}}GM^2/RL`$ and mass transfer time $`t_M=M/(\dot{M}_2)`$, the characteristic timescales restoring and perturbing equilibrium. If $`\tau `$ is small $`P_{\mathrm{min}}`$ is short. Hence $`P_{\mathrm{turn}}`$ is sensitive both to the orbital angular momentum loss rate (which determines the transfer rate $`\dot{M}_2`$) and to the interior structure of the donor. Paczyński (1981) was the first to point out that $`P_{\mathrm{turn}}`$ is close to 80 min if gravitational wave radiation drives the mass transfer. Since then, models with different input physics have been employed to verify a quantitative agreement between the observed $`P_{\mathrm{min}}`$ and the calculated $`P_{\mathrm{turn}}`$. Recent calculations (Kolb & Ritter 1992; Howell, Rappaport & Politano 1997) notoriously give $`P_{\mathrm{turn}}`$ too short by typically 10%.
Most of these evolutionary calculations are based on input physics which is rather approximate for very low mass (VLM) objects, i.e. donors with mass $`\mathrm{\hspace{0.17em}1}\mathrm{M}_{}`$. Because of their relatively high central densities and low central temperatures, correlation effects between particles dominate and must be taken into account in the equation of state (cf. Chabrier & Baraffe 1997, and references therein). Below about $`T_{\mathrm{eff}}4000`$ K ($`M0.6\mathrm{M}_{}`$), molecules become stable and dominate the atmospheric opacity, being responsible for strong non-grey effects and significant departure of the spectral energy distribution from blackbody emission (cf. Allard et al. 1997, and references therein).
Because of significant efforts devoted to the complex physics of low–mass stellar/substellar objects, the theory describing them has improved considerably in the past few years. As a result, the latest generation of stellar models for low–mass stars and brown dwarfs is now able to reproduce observed properties of field M–dwarfs with unprecedented accuracy (cf. Allard et al. 1996; Marley et al. 1996; Chabrier & Baraffe 1997; Burrows et al. 1997; Baraffe et al. 1998).
In this paper we use the Baraffe et al. models (1995, 1997, 1998, henceforth summarized as BCAH) — briefly reviewed in Sect. 2 — to calculate the secular evolution of CVs in the vicinity of $`P_{\mathrm{turn}}`$ (Sect. 3.1). We test if tidally distorted stellar models lead to a significant increase of $`P_{\mathrm{turn}}`$ over the value for spherical stars, as claimed by Nelson et al. 1985 (Sect. 3.2), and address the problem of the missing “period spike” at $`P_{\mathrm{min}}`$ (Sect. 3.3). This predicted accumulation of systems at $`P_{\mathrm{turn}}`$ is caused by the slow velocity in period space close to period bounce which increases the detection probability. The spike has been a dominant feature in synthesized period distributions obtained from theoretical CV population models (Kolb 1993; Kolb et al. 1998), yet is absent in the observed CV period distribution. In these earlier population models the donor star was approximated as a (bi–)polytrope, and CVs forming with a donor mass smaller than $`0.10`$ $`\mathrm{M}_{}`$ were not considered. Here we investigate if the period spike persists in period distributions synthesized from full BCAH models. For the first time, we include CVs which form already with BD donors, i.e. systems which did not evolve through period bounce. In Sect. 4 alternative explanations for the missing period spike are discussed. In particular, we focus on the difference between magnetic and non–magnetic CVs.
## 2 The secondary’s internal structure
Baraffe et al. (1998, and references therein) give a brief account of the input physics used in the most recent low–mass star models we apply here. The main strengths of these models are in two areas: the microphysics determining the equation of state (EOS) in the stellar interior, and the outer boundary condition based on non-grey atmosphere models.
The models employ the Saumon, Chabrier and Van Horn (1995) EOS which is specifically calculated for VLM stars, BDs and giant planets. The EOS is an important ingredient for our analysis, since it determines the mechanical structure of the donor stars, and thus their mass–radius relation. Models based on this EOS have been successfully tested against stars in detached eclipsing binary systems (Chabrier & Baraffe 1995) and field M-dwarfs (cf. Beuermann et al. 1998). The Saumon et al. (1995) EOS has been successfully compared with recent laser-driven shock wave experiments performed at Livermore. These probe the complex regime of pressure dissociation and ionization which is so characteristic for the objects we are studying here (cf. Saumon et al. 1998).
The second important ingredient of our VLM star and BD models concerns the outer boundary condition, which determines the thermal properties of an object and thus the mass–effective temperature ( $`T_{\mathrm{eff}}`$) relation. As demonstrated by Chabrier & Baraffe (1997), evolutionary models with a grey atmosphere outer boundary, e.g. the standard Eddington approximation, overestimate $`T_{\mathrm{eff}}`$ for a given mass, and yield too large a minimum hydrogen burning mass (MHBM). Because of tremendous recent progress in the field of cool atmosphere models (see e.g. the review of Allard et al. 1997) these now provide realistic atmosphere profiles and synthetic spectra which we use as a more realistic outer boundary condition.
Several observational tests confirm the success of evolutionary models based on these improvements, e.g. mass–magnitude relationships, colour–magnitude diagrams (Baraffe et al. 1997, 1998), mass–spectral type relationships (Baraffe & Chabrier 1996), and the first cool BD GL 229B (Allard et al. 1996). Although some discrepancies between models and observations remain (see Baraffe et al. 1998), uncertainties attributed to the input physics now seem significantly reduced.
We employ these one–dimensional BCAH models in the context of CVs within the usual Roche approximation. The mass transfer rate is calculated as a function of the difference between donor radius and Roche radius, following Ritter (1988). In a separate set of calculations we apply tidal and rotational corrections to the one–dimension stellar structure equations (Chan & Chau 1979) to account for the non–spherical shape of Roche equipotential surfaces. This allows us to test the results by Nelson et al. (1985), who found that their distorted stellar models give a value for $`P_{\mathrm{turn}}`$ which is about $`10\%`$ longer than for their spherical stellar models.
## 3 Model calculations
### 3.1 Evolutionary sequences
Using the BCAH stellar evolution code we recalculated the secular evolution of short–period CVs for various initial component masses. We assume that orbital angular momentum is lost by gravitational wave emission (e.g. Landau & Lifshitz 1958) and via an isotropic stellar wind from the WD. The wind removes the accreted mass with the WD’s specific orbital angular momentum from the binary. Hence the WD mass is constant throughout the evolution.
As we focus on the period minimum we do not consider the evolution at periods longer than 2-3 hr. In a second paper (Baraffe & Kolb 1999; see also Kolb & Baraffe 1999) we discuss properties of such sequences with more massive donors. They depend on the rather uncertain strength of the magnetic braking thought to dominate the evolution of CVs above the period gap (see e.g. King 1988 for a review).
We calculated three sets of evolutionary sequences. In Set A the WD mass is 0.6 $`\mathrm{M}_{}`$, while the initial donor masses range from 0.27 $`\mathrm{M}_{}`$ to 0.04 $`\mathrm{M}_{}`$ (see Fig. 1). Set B is similar but for a WD mass $`1.0`$ $`\mathrm{M}_{}`$. In Set C the initial donor mass is fixed at $`0.21`$ $`\mathrm{M}_{}`$, while the WD masses range from 1.2 $`\mathrm{M}_{}`$ to 0.3 $`\mathrm{M}_{}`$; see Tab. 1. At turn–on of mass transfer the secondary was either on the ZAMS ($`M_20.085`$ $`\mathrm{M}_{}`$), or had an age of 2 Gyrs ($`M_2<0.085`$ $`\mathrm{M}_{}`$). All donors are hydrogen–rich and have solar metallicity ($`X=0.70,Z=0.02`$). The sequences were terminated when the donor’s effective temperature $`T_{\mathrm{eff}}`$ was smaller than 900 K. This is the lower limit of the temperature range covered by the present non-grey atmosphere models (Hauschildt et al. 1999) available for this study.
Some quantities of interest for a typical sequence starting immediately below the period gap are shown in Fig. 2 and given in Tab. 2 (see also Tab. 3 for a sequence initiating mass transfer from a BD donor).
Fig. 1 confirms the well known effect that systems with different initial donor masses rather quickly join a uniform evolutionary track (Stehle et al. 1996). Most systems of Set A undergo period bounce at $`P_{\mathrm{turn}}67`$ min, which is only slightly longer than the corresponding $`P_{\mathrm{turn}}=65`$ min found with Mazzitelli’s models (Kolb & Ritter 1992; see also the dotted curve in Fig 3, lower panel). Note that a grey outer boundary condition like the standard Eddington approximation yields a smaller radius for given donor mass $`M_20.2\mathrm{M}_{}`$, and a higher MHBM. A test calculation with fixed mass loss rate and the Eddington approximation gives $`P_{\mathrm{turn}}`$ shorter by 5 min than for the corresponding sequence based on non-grey atmosphere models. Note also that CVs forming with fairly old and massive brown dwarf donors (age $`2`$ Gyr, mass $`0.050.07\mathrm{M}_{}`$) would populate the period regime shortwards of $`P_{\mathrm{turn}}`$. V485 Cen might be such a system.
For Set B we have $`P_{\mathrm{turn}}=69`$ min., slightly longer than for Set A since gravitational radiation losses are higher for larger WD mass, hence the mass transfer rate is larger. The dependence of $`P_{\mathrm{turn}}`$ on WD mass is most easily seen in Fig. 3, upper panel.
The effective temperature of the donor as a function of period is shown in Fig. 3. The asymptotic convergence of sequences with different initial donor mass and the insensitivity of $`\dot{M}`$ to the WD mass result in an essentially unique relation $`T_{\mathrm{eff}}(P)`$ on the non–degenerate branch of the track (before period bounce occurrs). At $`P_{\mathrm{turn}}`$, and on the degenerate branch (after period bounce) there is a significant spread in $`T_{\mathrm{eff}}`$ for given $`P`$. In the same diagram dots indicate the claimed location of AL Com (Howell et al. 1998), WZ Sge, TY Psc, V592 Cas and HU Aqr (Ciardi et al. 1998). For these systems the effective temperature of the donor has been derived via spectral fittings in the near-infrared using the same atmosphere models (Hauschildt et al. 1999) that serve as outer boundary condition of our stellar models. Since the near–IR emission of CVs is produced by both the secondary and the accretion disc, the estimate of the secondary’s contribution may be rather uncertain because of the lack of reliable accretion disc models. We thus do not regard the discrepancy between model calculations and observational data points (cf. Fig.3) as significant.
### 3.2 Sequences with non–spherical donors
We implemented the rotational correction scheme of Chan & Chau (1979) to investigate the effect of rotation and tidal distortion on the donor star. Assuming solid–body rotation, we recalculated a sequence with $`0.21\mathrm{M}_{}`$ initial donor mass, $`1.0\mathrm{M}_{}`$ WD mass and $`\dot{J}=\dot{J}_{\mathrm{GR}}`$. We find that $`P_{\mathrm{turn}}`$ is hardly affected; it is longer by only 1 min compared to the corresponding sequence with spherical models (from Set B). We have performed several other experiments, with different initial secondary masses and mass transfer rates (constant, or as given by gravitational radiation), including the case considered by Nelson et al. (1985), i.e. $`M_1=1.0\mathrm{M}_{}`$, $`M_2=0.4\mathrm{M}_{}`$, $`\dot{J}=\dot{J}_{\mathrm{GR}}`$. In none of these sequences we could reproduce the significant effect ($`10\%`$) rotational and tidal corrections had on $`P_{\mathrm{turn}}`$ in the calculations by Nelson et al. (1985) based on the same scheme of Chan and Chau (1979).
We do not think that the discrepancy between Nelson et al. (1985) and our work results from differences in the EOS and opacities. Indeed, although our input physics differs from those adopted by Chan & Chau (1979), we find the same quantitative differences quoted by these authors between the properties ($`L`$, $`R`$ and $`T_{\mathrm{eff}}`$) of spherical and rotating ZAMS stars (cf. their Table 1). We also reproduce the effects quoted in the very recent paper by Sills & Pinsonneault (1999). They considered a 0.7 $`\mathrm{M}_{}`$ ZAMS star with equatorial rotational velocity $`145`$ km s<sup>-1</sup> and find a slightly increased surface luminosity and effective temperature ($`\mathrm{\Delta }\mathrm{log}L=3.2\%`$, $`\mathrm{\Delta }T_{\mathrm{eff}}=100`$ K) compared to non–rotating models. For the same case we find $`\mathrm{\Delta }\mathrm{log}L=4.5\%`$, $`\mathrm{\Delta }T_{\mathrm{eff}}=90`$ K.
We thus are confident in our results and conclude that rotational and tidal corrections as described by Chan and Chau (1979) can be neglected for CVs, even when the systems are close to period bounce. Note that, although the secondary spins up in evolving toward $`P_{\mathrm{turn}}`$, the key quantity defined by Chan and Chau (1979),
$$\alpha =\frac{2}{3}\frac{\omega ^2R^3}{GM_R},$$
which appears in the momentum equation and represents the ratio of rotational to gravitational potential gradient, remains essentially constant at the donor star surface during the evolution (recall that the angular velocity $`\omega ^2P^2M_2/R_2^3`$). The maximum value of $`\alpha `$, reached at the surface, never exceeds 0.07.
Given the small differences between spherical and tidally distorted models we use the less cpu–time intensive calculations with spherical stars to derive a period distribution and analyse its properties close to $`P_{\mathrm{min}}`$.
### 3.3 Generating a period histogram
When the evolutionary sequences presented in Sect. 3.1 are convolved with an appropriate CV formation rate they give a theoretically predicted period histogram in the vicinity of $`P_{\mathrm{min}}`$ (see Kolb 1993 for a detailed description of the population synthesis technique). De Kool (1992) and Politano (1996) calculated the formation of CVs with standard assumptions about common envelope evolution and magnetic braking (see e.g. Kolb 1996 for a review). We computed population models for all these formation rate models; in addition, to show the differential effect of the main parameters more clearly, we considered subpopulations for a given WD mass with a time–independent formation rate $`b(\mathrm{log}M_2)=`$ const. in certain $`\mathrm{log}M_2`$ intervals.
In Figs. 4, 5 we show period distributions for two different CV subpopulations — P1 and P2 — which allow one to study the main effects that determine the structure of the histogram at $`P_{\mathrm{min}}`$. The histograms are for either a volume–limited sample, or a sample where individual systems have been weighted by $`\dot{M}^\alpha `$, for various values of $`\alpha `$. This mimics selection effects which affect the observed sample (cf. Kolb 1996). The subpopulations formally correspond to a Galactic disc age of 6 Gyr. In a somewhat older population the edge of the volume–limited distributions at $`1.4`$ hr would appear at slightly longer $`P`$, while the $`\dot{M}^\alpha `$–weighted cases would hardly change.
#### P1. Period histograms obtained from Set A
(Figure 4). These correspond to the period distribution of a subset of CVs with WD mass $`0.6\mathrm{M}_{}`$ (assuming that this mass does not change through the evolution). The same analysis has been performed for sequences of Set B, with very similar results, which we therefore do not show.
As noted above we do not consider the evolution of CVs with donors $`>0.21\mathrm{M}_{}`$, i.e. of CVs that form above the period gap. In the disrupted magnetic braking model for the period gap (e.g. King 1988) these systems would all reappear $`10^8`$ yr after formation at the lower edge of the gap, with a donor in thermal equilibrium. A BCAH stellar model in thermal equilibrium with mass $`0.21\mathrm{M}_{}`$ would fill its Roche lobe at $`P=2.1`$ hr, the observed lower edge of the period gap. Hence we simulate the contribution of systems that form above the gap to the period distribution below the gap by increasing the formation rate in a narrow bin in $`M_2`$–space at 0.21 $`\mathrm{M}_{}`$ by a large factor.
For the first time, we explicitly include systems which form with a brown dwarf donor star. The formation rate of such BD CVs is not known. Survival of the common envelope phase is crucial and could be a problem as the maximum orbital energy available to eject the envelope, roughly $`M_2`$, is small. Simulations by Politano (1998, priv. comm.) show that the formation of BD CVs is possible when the same formalism is applied as for more massive donor stars. For our purposes we just extrapolate the birthrate function $`b_2(\mathrm{log}M_2)`$ down to the smallest initial donor mass ($`0.04\mathrm{M}_{}`$) considered here.
In particular, we use
$$b_2(\mathrm{log}M_2)=\{\begin{array}{cc}\hfill 0.368k_1:& 0.207M_2<0.210\hfill \\ \hfill 1:& 0.090<M_2/\mathrm{M}_{}0.207\hfill \\ \hfill k_2:& 0.040M_2/\mathrm{M}_{}0.090\hfill \end{array}$$
(1)
with various combinations of the free parameters $`k_1`$ and $`k_2`$, describing the relative weight of systems forming above the period gap and with a brown dwarf donor, respectively. The numerical factor in front of $`k_1`$ is chosen such that for $`k_1=1`$ as many CVs form with donor masses $`0.21<M_2/\mathrm{M}_{}1.0`$ (“above the gap”) as with masses $`0.09M_2/\mathrm{M}_{}0.21`$ (“below the gap”). The results are not sensitive to the choice of the boundaries at $`0.207\mathrm{M}_{}`$ and $`0.09\mathrm{M}_{}`$. So, chiefly, $`k_1`$ is the ratio of systems forming above the gap to systems forming below the gap with a non–degenerate donor, while $`k_2`$ is the ratio of the formation rate of CVs with degenerate and non–degenerate donors. Figure 4 shows period histograms for various combinations of $`k_1`$, $`k_2`$ and $`\alpha `$.
#### P2. A period histogram obtained from Set C,
with the full time–dependent formation rate calculated by de Kool (1992; his model 3). This distribution, shown in the top panel of Fig. 5, illustrates the fine structure of the period spike at $`P_{\mathrm{turn}}`$ as a result of the different WD masses in the sample. To obtain this histogram, we approximated sequences with an arbitrary initial donor mass $`M_2<0.21`$ $`\mathrm{M}_{}`$ by the $`0.21`$ $`\mathrm{M}_{}`$ sequence from $`M_2`$ onwards. This introduces a slight error at the onset of mass transfer and for sequences which would form close to $`P_{\mathrm{turn}}`$. An explicit comparison of results from this simplified method for $`0.6\mathrm{M}_{}`$ WD mass with the distribution obtained from the full set of sequences (set A) shows that these deviations are negligible as long as $`k_22`$. To estimate the contribution from systems that form above the period gap the actually calculated evolutionary tracks have been extended to donor masses $`>0.21\mathrm{M}_{}`$ by assuming a simple main–sequence mass–radius relation (such that $`PM_2`$) and a constant mass transfer rate ($`2\times 10^9`$ $`\mathrm{M}_{}`$yr<sup>-1</sup>).
####
The theoretical period distributions in Figs. 4, and 5 show that the collective period minimum $`P_{\mathrm{min}}`$ coincides with the bounce period $`P_{\mathrm{turn}}`$ of individual evolutionary sequences, and is about $`10`$ min. shorter than the observed period minimum (e.g. Ritter & Kolb 1998). This confirms tentative conclusions reached earlier by Kolb (1993) who considered populations constructed from simplified polytropic stellar models with artificial outer boundary conditions, and restricted to initial donor masses $`0.09\mathrm{M}_{}`$.
The subpopulations for WD masses 0.6 $`\mathrm{M}_{}`$ and 1.0 $`\mathrm{M}_{}`$ show that the period spike does not disappear or broaden significantly when CVs with brown dwarf donor stars are included in the population. This is true unless the majority of newly forming CVs actually are BD CVs ($`k_25`$). But if BD CVs were dominant the bulk of systems would populate the period regime shortwards of $`P_{\mathrm{turn}}`$, making this a fairly unlikely possibility in view of the observed value of $`P_{\mathrm{min}}`$.
In fact, in population models which do not explicitly overemphasize BD CVs ($`k_22`$) the effect of these sequences on the overall shape of the period spike is generally small. A fair representation of the period spike is obtained by the simple procedure used to generate the period distribution P2. The double spike at $`P_{\mathrm{min}}`$ in the P2 histogram (Fig. 5, upper panel) is a result of the double–humped WD mass distribution (see e.g. de Kool 1992): high–mass C/O WD systems cluster in the longer–period spike, low–mass He–WD systems in the shorter–period spike. This fine structure of course disappears altogether in small samples.
The overall shape of the weighted distribution is not sensitive to moderate $`\alpha `$; there is hardly any difference between histograms calculated with $`\alpha `$ between $`0.5`$ and $`1.5`$. Increasing $`\alpha `$ further tends to decrease the amplitude of the spike. We obtain a truly flat distribution, similar to the observed one (Fig. 6), only for $`\alpha 6`$.
## 4 Discussion
The results presented in the previous section show that despite the significant improvement of low–mass star and brown–dwarf models, the period histogram for short–period CVs synthesized from these models still appears to be in conflict with observations. If the evolution of short–period CVs is driven by the emission of gravitational waves, period bounce occurs at $`70`$ min for most systems, resulting in a collective minimum period $`P_{\mathrm{min}}`$ which is 10 min shorter than the observed one. The contribution of CVs forming with BD donors is either negligible or would cause $`P_{\mathrm{min}}`$ to be yet shorter. As a result of period bounce, the simulated period histograms show an accumulation of systems near $`P_{\mathrm{min}}`$, the period spike. The only way to remove the period spike effectively from the distribution is to impose a very steep dependence of the system’s detectability $`d`$ on one of the evolutionary parameters, e.g. $`\dot{M}^6`$ or steeper.
The catalogue of Ritter & Kolb (1998) lists 129 CVs with periods $`<2.10`$ hr. Among these are 33 polars, 78 dwarf novae (excluding intermediate polars), and 6 AM CVn systems. The remaining 12 systems are either intermediate polars, novae, novalikes, or of unknown type. Hence essentially all known short–period CVs are either dwarf novae, i.e. systems with an unstable accretion disc, or discless polars.
### 4.1 The missing period spike
Almost all dwarf novae are detected in outburst. It is tempting to identify the accretion disc and its outburst properties as the cause for a strong $`\dot{M}`$–dependent selection effect. A few short–period dwarf novae, sometimes referred to as WZ Sge stars, have very long outburst recurrence times $`t_{\mathrm{rec}}`$, the most extreme example being WZ Sge itself with $`t_{\mathrm{rec}}30`$ yr. It has long been noted that low–$`\dot{M}`$ CVs might have escaped detection if their outburst interval is significantly longer than the period since beginning of systematic monitoring and surveying of the sky with modern means – a few decades. For long $`t_{\mathrm{rec}}`$ the relative detection probability of dwarf novae scales as $`d1/t_{\mathrm{rec}}`$, suggesting that $`t_{\mathrm{rec}}\dot{M}^6`$ or steeper. In practice this very steep dependence would require that $`t_{\mathrm{rec}}\mathrm{}`$ for $`\dot{M}\mathrm{few}\times 10^{11}\mathrm{M}_{}`$yr<sup>-1</sup>, i.e. low–$`\dot{M}`$ CVs would not undergo outbursts at all.
Various plausible physical models which naturally account for this property have been suggested. Evaporation of accreted material into a hot corona could prevent the disc from accumulating the critical surface mass density required to launch a heating wave (Meyer–Hofmeister et al. 1998; see also Liu et al. 1997). Systems in such a permanent quiescence would be optically very faint but should still emit about $`10\%`$ of their accretion luminosity $`L_{\mathrm{acc}}=GM_1\dot{M}/R_1`$ ($`M_1,R_1`$ is the WD’s mass and radius) in X–rays. Alternatively, King (1999) argues that systems approaching the period minimum become magnetic ejectors even if the WD’s magnetic field is weak. In this case an accretion disc can no longer form, and the systems become unobservable. WZ Sge, where the WD is suspected to be weakly magnetic (see e.g. Lasota et al. 1999, and references therein), would be a marginal disc accretor (Wynn & Leach 1999).
Although this can explain the absence of a period spike at $`P_{\mathrm{min}}`$ for dwarf novae, it certainly does not apply to the discless polars. An independent observational selection effect operating on polars in the same period/mass transfer rate range and with the same net result as a steep increase of $`t_{\mathrm{rec}}`$ for dwarf novae seems a highly unlikely coincidence (but see Meyer & Meyer–Hofmeister 1999). In other words: polars should show a period spike, even though dwarf novae do not. Could small–number statistics hide the period spike for polars? To investigate this we performed a number of Monte Carlo experiments, where we drew a sample of either 33 or 78 systems from an underlying period distribution like the one shown in Fig. 5 (upper panel). As the middle and lower panels of this figure show, the smaller samples give a surprisingly wide variety of distributions, with typically $`20\%25\%`$ of them showing no sign of a period spike at all. In contrast, in the larger sample the spike is prominent in more than $`90\%`$ of all cases. A KS test confirms this impression: We wish to quantify the difference between model and observed distribution that is due to a different morphological shape. Therefore we have to exclude effects arising from different values for the lower edge $`P_{\mathrm{gap}}`$ of the period gap and for $`P_{\mathrm{min}}`$ — the latter effect is considered in Sect. 4.2. To achieve this we rescale the calculated distribution such that it matches the range of the observed distribution before performing the KS test. Specifically, the period axis $`C=\mathrm{log}P`$ of the theoretical distribution is rescaled according to
$$C(CO_l)\times \frac{O_uO_l}{C_uC_l}+O_l,$$
(2)
where $`C_l=\mathrm{log}P_{\mathrm{min}}`$, $`C_u=\mathrm{log}P_{\mathrm{gap}}`$ denote the calculated period minimum and lower edge of the gap, and $`O_l`$, $`O_u`$ the corresponding observed values. We applied (2) to the model distribution shown in Fig. 5 ($`P_{\mathrm{min}}=64`$ min, $`P_{\mathrm{gap}}=124`$ min), assuming the observed values $`P_{\mathrm{min}}=78`$ min, $`P_{\mathrm{gap}}=113`$ min. The maximum significance level for rejecting the null hypothesis that the observed sample is drawn from this rescaled model distribution is 0.34 for polars, 0.98 for dwarf novae. If we use $`P_{\mathrm{gap}}=130`$ min for the observed lower egde of the gap, the rejection significance for polars rises to 0.91. This is solely due to the so–called “114 min spike”, see e.g. Ritter & Kolb 1992, still a significant feature in the observed distribution.
### 4.2 The mismatch between observed and calculated $`P_{\mathrm{min}}`$
Neither the small number statistics for polars nor the suggested detectability function for dwarf nova can make the minimum period of the observed distribution significantly longer than in the underlying intrinsic distribution. To achieve this by the latter effect, dwarf novae have to become unobservable long before they reach $`P_{\mathrm{turn}}`$. With $`\dot{M}`$ as the most likely control parameter determining the detectability this is difficult to achieve, as $`\dot{M}`$ is almost constant on the non–degenerate branch above $`P_{\mathrm{turn}}`$. (This problem would be less severe if the main control parameter were the donor mass, the mass ratio, or the orbital separation. A particularly steep dependence of the discovery probability on any of these could be achieved if mass transfer cycles, similar to the ones discussed by King et al. 1996, 1997, existed in CVs below the period gap, but changed their character discontinuously before $`P_{\mathrm{turn}}`$.)
To test the effect of small number statistics we drew 1000 samples of either 33 or 78 systems from the theoretical distribution shown in Fig. 5 and registered the shortest period $`P_s`$ of each sample. The parent distribution has a short–period cut–off at $`64.4`$ min. We found that $`99.9\%`$ of the larger samples have $`P_s<67.0`$ min ($`99.0\%`$ have $`P_s<66.6`$ min, $`90.0\%`$ have $`P_s<65.6`$ min). The median is $`P_s=65`$ min, the longest $`P_s`$ we found is $`67.03`$ min. Similarly, $`99.9\%`$ of the smaller samples have $`P_s<71.0`$ min ($`99.0\%`$ have $`P_s<68.2`$ min, $`90.0\%`$ have $`P_s<66.8`$ min), with a median $`P_s=65.4`$ min, and $`71.2`$ min as the longest $`P_s`$. The rare cases where $`P_s`$ was as long as 70 min still fall well short of the observed value $`P_{\mathrm{min}}=78`$ min.
Given this, it might be that the $`P_{\mathrm{min}}`$ mismatch is caused by evolutionary effects after all — effects not accounted for in our models.
The chemical composition of the secondary affects the value of $`P_{\mathrm{turn}}`$, by changing the parameter of the donor at a given mass, hence changing both $`t_M`$ (via the angular momentum loss time) and $`t_{\mathrm{KH}}`$. Stehle et al. (1997) pointed out that $`P_{\mathrm{turn}}`$ is slightly shorter for CVs with low–metallicity donors. They found $`\mathrm{d}P_{\mathrm{turn}}/\mathrm{d}\mathrm{log}Z=0.084`$ hr. Our calculations confirm this ($`P_{\mathrm{turn}}=67.41`$ mins. for $`Z=0.02`$, while $`P_{\mathrm{turn}}=64.70`$ mins for $`Z=0.006`$; both with $`M_1=0.70\mathrm{M}_{}`$). Although $`P_{\mathrm{turn}}`$ increases for higher metallicity donors the effect is much too small to account for a mismatch of $`10`$ mins. It has been noted that a larger than expected fraction of CVs above the gap have a nuclear–evolved secondary (Beuermann et al. 1998, Baraffe & Kolb 1999, Kolb & Baraffe 1999). When these secondaries become fully convective their helium content is $`0.5`$, giving a $`P_{\mathrm{turn}}`$ which is in fact $`7`$ mins. shorter than for hydrogen–rich donors.
Residual shortcomings in the EOS and the atmosphere profiles cannot yet be ruled out as the cause for the $`P_{\mathrm{min}}`$ discrepancy. A quantitative estimate of uncertainties in the mass–radius relation from the treatment of the EOS (see Saumon et al. 1995) is difficult. The most profound observational test is against stellar parameters determined for components in (detached) eclipsing binaries; yet to date there are no such systems with $`M_2<0.2\mathrm{M}_{}`$. The region near $`P_{\mathrm{turn}}`$ involves effective temperatures lower than 2600 K (cf. Fig. 3), below which grains form. These affect both the atmosphere spectrum and profile. Although the atmosphere models used for this study are grainless, calculations based on preliminary atmosphere models including the formation and absorption of grains (Allard 1999; Baraffe & Chabrier 1999) do not yield a longer $`P_{\mathrm{turn}}`$. We thus do not expect an improvement of the situation with the forthcoming generation of dusty atmosphere models.
The $`P_{\mathrm{min}}`$ mismatch could also be due to uncertainties in the calculated value of $`P_{\mathrm{turn}}`$ inherent to the very concept of the Roche model. Strictly valid only for point masses, its applicability to extended donors relies on the fact that the stars are usually sufficiently centrally condensed. This is not necessarily a good approximation for fully convective stars which are essentially polytropes of index $`n=3/2`$. We note that Uryu & Eriguchi (1999) considered stationary configurations of fluids describing a synchronously rotating polytropic star in a binary with a point mass companion. For $`n=3/2`$ polytropes they find a Roche radius that is typically $`12\%`$ smaller than estimated from Eggleton’s (1983) approximation for a given orbital separation, but at the same time $`4\%`$ larger than the radius of a non–rotating polytrope with the same mass.
Alternatively, an obvious way to increase $`P_{\mathrm{min}}`$ is to increase the orbital angular momentum loss rate $`\dot{J}`$ over the value $`\dot{J}_{\mathrm{GR}}`$ set by gravitational radiation. We find $`P_{\mathrm{min}}83`$ min (up from $`69`$ min) for $`\dot{J}=4\times \dot{J}_{\mathrm{GR}}`$ and $`M_1=1\mathrm{M}_{}`$, a much smaller increase than quoted by Paczyński (1981). Patterson (1998) favoured a modest increase of $`\dot{J}`$ on grounds of space density considerations and the ratio of systems below and above the period gap. Standard population models typically give a local CV space density of up to $`10^4`$ pc<sup>-3</sup> (de Kool 1992, Kolb 1993, Politano 1996), with $`99\%`$ of all CVs below the period gap, $`70\%`$ of these past period bounce. Patterson argues that the observed space density is at least a factor 20 smaller, and that there is little evidence for the predicted large population of post–period minimum CVs. A higher transfer rate after period bounce would remove systems from the population as the donor can lose all its mass in the age of the Galaxy, thus resolving both problems and the $`P_{\mathrm{min}}`$ mismatch. However, postulating an as yet unknown $`\dot{J}`$ mechanism which conspires to produce almost the same value as $`\dot{J}_{\mathrm{GR}}`$ at the transition from non–degenerate to degenerate stars does not seem very attractive. Rather, it seems more likely that the bulk of short–period systems are indeed unobservable – at least in the optical. Watson (1999) shows that the presence of a population of CVs with a space density of $`10^4`$ pc<sup>-3</sup> that emits $`10\%`$ of its accretion luminosity in X–rays cannot be excluded from ROSAT and ASCA data. It should be possible to place tight limits on such a population and its potential contribution to the Galactic Ridge emission from deep XMM and AXAF surveys.
As a final note, we point out that the evolutionary effect of an irradiation–induced stellar wind from the donor (considered semi–analytically by King & van Teeseling 1998) could help to resolve the $`P_{\mathrm{min}}`$ mismatch and the period spike problem; detailed investigations are under way (Kolb et al. in preparation).
## 5 Summary
We have investigated the problem that the standard explanation for the CV period minimum as a result of period bounce of systems driven by gravitational radiation predicts a period minimum that is too short, and an accumulation of systems close to $`P_{\mathrm{turn}}`$, the “period spike”, which is not observed.
Using up–to–date stellar models by Baraffe et al. (1998) which successfully reproduce observed properties of single low–mass stars and brown dwarfs we confirm that $`P_{\mathrm{turn}}`$ is about 10 mins. shorter than the observed $`P_{\mathrm{min}}`$.
We have presented synthesised period histograms for CVs below the period gap which, for the first time, are based on evolutionary calculations obtained with full stellar models, and include CVs which form with brown dwarf donors. The period spike is always present.
Although there are ways to explain why the spike is not observed — small number statistics for polars, undetectability for dwarf novae — we find no satisfactory reason why $`P_{\mathrm{min}}`$ is longer than $`P_{\mathrm{turn}}`$ for both magnetic and non–magnetic CVs.
## Acknowledgments
We are grateful to H. Ritter, H. Spruit, F. Meyer, E. Meyer–Hofmeister and K. Beuermann for valuable discussions, and A. King for comments and for improving the language of the manuscript. We thank the referee, T. Marsh, for useful comments. IB thanks the Max–Planck–Institut für Astrophysik and the University of Leicester for hospitality during the realization of part of this work. The calculations were performed on the T3E at Centre d’Etudes Nucléaires de Grenoble.
|
no-problem/9906/astro-ph9906168.html
|
ar5iv
|
text
|
# Direct Analysis of Spectra of Type Ic Supernovae
## 1 Introduction
The photospheric–phase spectrum of a Type Ic supernova lacks the strong hydrogen lines of a Type II, the strong optical He I lines of a Type Ib, and the deep red Si II absorption of a Type Ia. A Type Ic (SN Ic) is thought to be the result of the core collapse of a massive star that either has lost its helium layer or ejects helium that remains insufficiently excited to produce conspicuous optical He I lines. For a review of observations of supernova spectra, see Filippenko (1997).
Since April, 1998, interest in SNe Ic has been very high because of the extraordinary SN 1998bw, which appears to have been associated with the gamma ray burst GRB 980425 (Galama et al. 1998; Kulkarni et al. 1998). Recently we (Millard et al. 1999) have used the parameterized supernova synthetic–spectrum code SYNOW to make a “direct analysis” of spectra of SN 1994I, a well observed, normal SN Ic. In this contribution, after summarizing our work on SN 1994I, I report some preliminary results of a similar analysis of the peculiar Type Ic SN 1997ef (Deaton et al. 1998; J. Millard et al., in preparation) and then consider the related but even more peculiar SN 1998bw. The emphasis here is on establishing line identifications and making spectroscopic estimates of the mass and kinetic energy of the line–forming layers of the ejected matter.
## 2 SYNOW
In its simplest form the SYNOW code assumes spherical symmetry and that line formation takes place by resonant scattering outside a sharp photosphere that radiates a blackbody continuum. To a good approximation the simple explosion velocity law, $`v=r/t`$, holds. Consequently the velocity gradient is isotropic and homogeneous (unlike the case of a stellar wind, where even in the constant–velocity case the velocity gradient is neither isotropic nor homogeneous). From the point of view of an observer, the (nonrelativistic) surfaces of constant radial velocity are planes perpendicular to the line of sight, and an unblended line formed by resonant scattering has an emission component that peaks at the line rest wavelength in the supernova frame and an absorption component whose minimum is blueshifted by an amount that corresponds to the velocity at the photosphere (unless the line is extremely weak or rather strong).
SYNOW treats line formation in the Sobolev approximation, which is a good one for this purpose. The profile of an unblended line is determined by the adopted radial dependences of the line optical depth and the line source function. The line optical depth determines the strength of the line and is given by
$$\tau _l=\frac{\pi e^2}{m_ec}f\lambda tn_l(v)=0.026f\lambda _\mu t_dn_l(v),$$
where $`f`$ is the oscillator strength, $`\lambda _\mu `$ is the line wavelength in microns, $`t_d`$ is the time since explosion in days, and $`n_l(v)`$ is the population of the lower level of the transition in cm<sup>-3</sup>. The correction for stimulated emission, although not written out here, is taken into account.
The line source function determines the extent to which the line is in emission or absorption. In the resonant–scattering approximation line photons are conserved, except for occultation effects, and the source function of an isolated line is just the product of the intensity of the photospheric continuum and the geometrical dilution factor. The source function of a line that interacts with one or more lines of shorter wavelength is altered by photons that are scattered by those lines. The essential role of SYNOW is to treat the multiple scattering, which in observers’ language is line blending.
Various fitting parameters are available for a SYNOW synthetic–spectrum calculation. $`T_{bb}`$ is the temperature of the blackbody continuum radiated by the photosphere. For each ion whose lines are introduced, the optical depth at the photosphere of a reference line is a parameter, and the optical depths of the other lines of the ion are calculated for Boltzmann excitation at excitation temperature $`T_{exc}`$ (which ordinarily is taken to be the same as $`T_{bb}`$). For the spectra shown here, the radial dependence of the line optical depths is a power law, $`\tau v^n`$. At each epoch, the velocity at the photosphere, $`v_{phot}`$, is a parameter, and maximum and minimum velocities also can be imposed on an ion; when the minimum velocity exceeds $`v_{phot}`$ the ion is said to be detached from the photosphere. The most interesting parameters are the velocity parameters and the “density” power–law index, $`n`$.
When deciding which ions to introduce, we are guided by experience and by the LTE calculations of line optical depths by Hatano et al. (1999), who considered six compositions that might be encountered in supernovae. The composition of interest here is the one in which hydrogen and helium have been burned to a mixture of carbon and oxygen, with the heavier elements present in their solar mass fractions. In this case, ions that are predicted to have lines of significant optical depth include Ca II, Fe II, O I, Si II, C II, Mg II, O II, and Ti II.
## 3 The Normal Type Ic SN 1994I
In Millard et al. (1999) we compare SYNOW synthetic spectra with observed spectra of SN 1994I obtained by Filippenko et al. (1995) from 5 to 35 days after the assumed explosion date of March 30, 1994. A density power–law index of $`n=8`$ is used for all epochs. Figure 1 compares a spectrum of SN 1994I obtained 16 days after explosion with a synthetic spectrum that has $`v_{phot}=10,000`$ km s<sup>-1</sup> and $`T_{bb}=8000`$ K. Most of the observed features are well matched. Ions that certainly must be introduced to account for observed features are Ca II, O I, Na I, Fe II, and Ti II. (The Na I D–line feature is not predicted to be significant by Hatano et al. (1999), but as usual in supernova spectra the Na I feature is observed to be stronger than predicted. We are confident of the identification.) In this particular synthetic spectrum, lines of C II also are used, but detached at 16,000 km s<sup>-1</sup> so that $`\lambda 6580`$ can account for most of the observed absorption near 6200 Å. Often it is difficult to decide between detached C II $`\lambda 6580`$ and undetached Si II $`\lambda 6355`$, and the C II identification is not considered to be definite. In order to account for the observed absorption around 7000 Å we would have to introduce lines of O II (see below). The excessive height of the synthetic peaks in the blue part of the spectrum is not of great concern; the number of lines of singly ionized iron–peak elements rises rapidly toward short wavelengths, so SYNOW spectra often are underblanketed in the blue due to missing lines of iron–peak ions that are not introduced.
In Millard et al. (1999) we consider the identification of the observed absorption near 10,250 Å, which has been identified as He I $`\lambda 10830`$ and taken as strong evidence that SN 1994I ejected helium (Filippenko et al. 1995). We find that it is difficult to account for even just the core of the observed 10,250 Å absorption with He I $`\lambda `$10830 without compromising the fit in the optical (see also Baron et al. 1999). We suggest that the observed feature may be a blend of He I $`\lambda `$10830 and C I $`\lambda `$10695, or perhaps a blend of Si I lines. This is an important issue but it will not be discussed further here, since we have no evidence for or against the presence of the feature in SNe 1997ef and 1998bw.
Another comparison of observed and synthetic spectra, but for just five days after the assumed explosion date, is shown in Figure 2. (We often consider photospheric–phase spectra in reverse chronological order, because at the earliest times line formation takes place in the highest–velocity layers and the blending is most severe.) The synthetic spectrum has $`v_{phot}=17,500`$ km s<sup>-1</sup> and $`T_{bb}=17,000`$ K. Ions that certainly are needed are Ca II, O I, and Fe II. In this synthetic spectrum, lines of C II, Na I, Mg II, Si II, and O II (with only \[O II\] $`\lambda `$7320,7330 having a significant effect on the spectrum) also are introduced; they are considered probable but not definite.
The adopted values of $`v_{phot}`$ can be used to estimate the mass and kinetic energy in the line–forming layers. For an $`r^n`$ density distribution, the mass (in $`M_{}`$) and the kinetic energy (in foe, where 1 foe $`10^{51}`$ erg) above the layer at which the electron–scattering optical depth is $`\tau _{es}`$ can be expressed as
$$M=1.2\times 10^4v_4^2t_d^2\mu _e\tau _{es}f_M(n,v_{max}),$$
$$E=1.2\times 10^4v_4^4t_d^2\mu _e\tau _{es}f_E(n,v_{max}),$$
where $`v_4`$ is $`v_{phot}`$ in units of $`10^4`$ km s<sup>-1</sup>, $`t_d`$ is the time since explosion in days, $`\mu _e`$ ($`Y_e^1`$) is the mean molecular weight per free electron, and the integration is carried out to velocity $`v_{max}`$. The functions $`f_M`$ and $`f_E`$ always exceed unity.
For SN 1994I, we use $`\mu _e=14`$ (a mixture of singly ionized carbon and oxygen), $`\tau _{es}=2/3`$ at the bottom of the line–forming layer, and integrate the steep density power–law to infinity ($`f_M(8,\mathrm{})=1.4`$, $`f_E(8,\mathrm{})=2.3`$). Figure 3 shows $`v_4`$, $`M`$, and $`E`$ plotted against time. ($`E`$ should increase monotonically with time. Its non–monotonic behavior just reflects the imprecision of our determinations of $`v_{phot}`$; recall that $`Ev_{phot}^4`$.) At 35 days after explosion, the mass moving faster than 7000 km s<sup>-1</sup> is estimated to be about 1.4 $`M_{}`$ and it carries a kinetic energy of about 1.2 foe. These numbers are reasonable, and similar to those that have been estimated for SN 1994I on the basis of light–curve studies (Nomoto et al. 1994; Iwamoto et al. 1994; Young, Baron, & Branch 1995; Woosley, Langer, & Weaver 1995).
Such spectroscopic estimates of mass and kinetic energy also come out to be reasonable for Type Ia supernovae (Branch 1980) and for SN 1987A (Jeffery & Branch 1990).
## 4 From SN 1994I to SNe 1997ef and 1998bw
Figure 4 compares spectra of SN 1994I at 16 days after explosion, SN 1997ef at 20 days after its assumed explosion date of November 15, 1997, and SN 1998bw at 16 days after its explosion date of April 25, 1998. It is clear that SNe 1997ef and 1998bw are spectroscopically related to each other and also, but less closely, to SN 1994I. Therefore it seems appropriate to refer to SNe 1997ef and 1998bw as “Type Ic peculiar”. The observed absorption features are much broader and bluer in SN 1997ef than in SN 1994I, and even moreso in SN 1998bw. This means that SNe 1997ef and 1998bw ejected more mass at high velocity. Figure 5 shows the effects, on the SYNOW synthetic spectrum of Figure 1 (for SN 1994I at 16 days), of raising $`v_{phot}`$ from 10,000 to 30,000 km s<sup>-1</sup>. Figure 6 shows the effects of dropping the density power–law index from $`n=8`$ to $`n=2`$. Raising $`v_{phot}`$ and dropping $`n`$ both cause the absorption features to become broader and bluer, and both appear to be necessary to obtain satisfactory SYNOW fits to spectra of SNe 1997ef and 1998bw. A value of $`n=2`$ is used for all of the synthetic spectra shown below.
### 4.1 Fitting spectra of SN 1997ef
Figure 7 compares a spectrum of SN 1997ef obtained 34 days after explosion with a synthetic spectrum that has $`v_{phot}=7000`$ km s<sup>-1</sup>, $`T_{bb}=7000`$ K, and uses only lines of Ca II, O I, Si II, and Fe II. This fit (and those to follow) could be improved by tuning the parameters, but as it stands it is good enough to indicate that we are on the right track.
Figure 8 compares a spectrum of SN 1997ef obtained 20 days after explosion with a synthetic spectrum that has $`v_{phot}=12,000`$ km s<sup>-1</sup>, $`T_{bb}=11,000`$ K, and uses lines of Ca II, O I, Si II, Fe II, and Mg II. In the red part of the spectrum, the good fit indicates that at this epoch just a few lines of Ca II, O I, and Si II are responsible for the features. In the blue, Fe II blends dominate. The synthetic spectrum is severely underblanketed in the blue due to missing lines of other singly ionized iron–peak elements.
Figure 9 compares a spectrum of SN 1997ef obtained 10 days after explosion with a synthetic spectrum that has $`v_{phot}=22,000`$ km s<sup>-1</sup>, $`T_{bb}=11,000`$ K, and uses the same ions as in Figure 8. This is a good example of why it can be instructive to work backward in time; this interpretation of the spectrum might seem arbitrary if the later–epoch spectra had not already been discussed.
Figure 10 is exactly like Figure 9 except that the Fe II lines have been turned off. Comparison of Figures 9 and 10 shows how strongly the Fe II lines affect the blue, while having practically no effect in the red. The same is true for the 20 and 34 day spectra discussed above.
### 4.2 Fitting spectra of SN 1998bw
Figure 11 compares a spectrum of SN 1998bw obtained 28 days after explosion with a synthetic spectrum that has $`v_{phot}=7000`$ km s<sup>-1</sup>, $`T_{bb}=6000`$ K, and uses lines of Ca II, O I, Si II, Na I, Ca II, and Fe II. The situation is much like that of SN 1997ef at 34 days.
Figure 12 compares a spectrum of SN 1998bw obtained 16 days after explosion with a synthetic spectrum that has $`v_{phot}=17,000`$ km s<sup>-1</sup>, $`T_{bb}=8000`$ K, and uses only lines of Ca II, O I, Si II, and Fe II. The situation is like that of SN 1997ef at 20 days, but here the blending is more severe due to the higher $`v_{phot}`$ of SN 1998bw.
Figure 13 compares a spectrum of SN 1998bw obtained 8 days after explosion with a synthetic spectrum that has $`v_{phot}=30,000`$ km s<sup>-1</sup>, $`T_{bb}=8000`$ K, and uses only lines of O I, Si II, Ca II, and Fe II. Again the situation is like that of SN 1997ef at 10 days but with more blending due to higher $`v_{phot}`$.
Figure 14 is like Figure 13 except that the Fe II lines have been turned off.
### 4.3 Masses and kinetic energies of SNe 1997ef and 1998bw
Figure 15 compares the adopted values of $`v_{phot}`$ versus time. Around 30 days after explosion the $`v_{phot}`$ values converge to about 7000 km s<sup>-1</sup>, but at earlier times the values for SN 1997ef are higher than those for SN 1994I, and the values for SN 1998bw are higher still.
To estimate the masses and kinetic energies of SNe 1997ef and 1998bw, the integration cannot be extended to infinity because $`n=2`$ has been used for the synthetic spectra. Instead the integration is carried out to $`v_{max}=2v_{phot}`$ (with $`f_M(2,2)=2,f_E(2,2)=4.7`$) for the earliest epoch considered, i.e., to 44,000 km s<sup>-1</sup> for SN 1997ef and to 60,000 km s<sup>-1</sup> for SN 1998bw. The results are shown in Figures 16 and 17. For SNe 1997ef and 1998bw, the masses above 7000 km s<sup>-1</sup> are estimated to be around 6 $`M_{}`$. For SN 1997ef the kinetic energy above 7000 km s<sup>-1</sup> comes out to be around 30 foe while that of SN 1998bw is around 60 foe. These kinetic energies are more likely to be too low than too high because most of the estimated kinetic energy comes from the earliest epochs considered, and the integrations are carried out to only $`2v_{phot}`$ while the synthetic spectra actually go to higher velocities. Of course, there also is more mass at velocities lower than 7000 km s<sup>-1</sup>, but not much more kinetic energy.
In a preprint, Iwamoto et al. (1998) compare observed spectra of SN 1997ef with synthetic spectra calculated for a hydrodynamical model that has an ejected mass of about 4.6 $`M_{}`$ and a kinetic energy of 1 foe. The prominent lines in their synthetic spectra are much the same as the ones that have been identified here but as they discuss, the lines in their synthetic spectra are much too narrow and not sufficiently shifted to the blue. Synthetic spectra calculated for models having more mass and kinetic energy give much better fits to the SN 1997ef spectra (P. Mazzali and K. Nomoto, personal communication). Iwamoto et al. (1999) compare observed spectra of SN 1998bw with synthetic spectra calculated for a hydrodynamical model that has an ejected mass of about 11 $`M_{}`$ and a kinetic energy of 30 foe. Their synthetic spectra match the SN 1998bw spectra fairly well, and it appears that more mass at high velocity would lead to even better fits.
## 5 Conclusion
The spectroscopic mass and kinetic–energy estimates presented here for SNe 1997ef and 1998bw are preliminary and approximate. Nevertherless, it seems clear that at least in the spherical approximation the kinetic energy of both events was much higher than the canonical one foe, as was reported by Deaton et al. (1998) for SN 1997ef and as in the models of Iwamoto et al. (1999) and Woosley, Eastman, & Schmidt (1999) for SN 1998bw.
Polarization spectra are much more sensitive than flux spectra to asymmetry. Core–collapse supernovae generally show detectable polarization, which indicates that they are significantly asymmetric (Wang et al. 1996). Höflich, Wheeler, & Wang (1999) calculate light curves of moderately asymmetric explosions and suggest that SN 1998bw was distinguished principally by having been viewed close to the symmetry axis, rather than by having a very high kinetic energy. It is true that to the extent that the ejecta of SNe 1997ef and 1998bw are “beamed”, the kinetic energy estimates presented here might be too high. However, because the spectra of SNe 1997ef and 1998bw can be matched fairly well in a straightforward way with the spherical symmetry assumption, and the corresponding kinetic–energy estimates are so very high, and the Lorentz factor of the ejecta is not high enough to be a factor in the energy estimates, it is likely that even if SNe 1997ef and 1998bw were non–spherical, they also were hyper–energetic.
I am grateful to Peter Garnavich and Yulei Qiu for providing spectra of SN 1997ef, to Ferdinando Patat for providing spectra of SN 1998bw, and to Eddie Baron, Kazuhito Hatano, David Jeffery, and Jennifer Millard for discussions and assistance.
|
no-problem/9906/cond-mat9906026.html
|
ar5iv
|
text
|
# REFERENCES
Comment on “Spin Dependent Hopping and Colossal Negative Magnetoresistance in Epitaxial $`\mathrm{𝐍𝐝}_{\mathbf{0.52}}\mathrm{𝐒𝐫}_{\mathbf{0.48}}\mathrm{𝐌𝐧𝐎}_\mathrm{𝟑}`$ Films in Fields up to 50 T”
Recently Wagner et al. proposed a modification of Mott’s original model to explain the magnetoresistance scaling in the ferromagnetic and paramagnetic regimes of the perovskite $`\mathrm{Nd}_{0.52}\mathrm{Sr}_{0.48}\mathrm{MnO}_3`$ films (in fields up to 50 T). These authors claimed that there is a hopping barrier which depends on the misorientation between the spins of electrons at the initial and the final states in an elementary process. They further claimed that using the model they can explain the observed scaling behavior– negative-magnetoresistivity scaling proportional to the Brillouin function $``$ in the ferromagnetic state and to $`^2`$ in the paramagnetic state. In this comment we argue that the modification needed for Mott’s original model is different from that proposed by Wagner et al. and further show that our picture will successfully explain the observed scaling in the two regimes.
Firstly, within a polaronic picture where hopping takes place from a polarized cloud, the wave function of the carrier in two dimensions corresponds to a super localized carrier and not to just a localized carrier. This claim of super localization follows from the correspondence of the energy equation to that of a simple harmonic oscillator. The polaronic free energy expression (up to a constant) that needs to be minimized in two dimensions is given by
$$\frac{\pi ^2\mathrm{}^2}{2mR_\xi ^2}+\frac{\pi R_\xi ^2}{a^2}[\mathrm{\Delta }IT\mathrm{\Delta }S],$$
(1)
where $`R_\xi `$ is the radius of the polaron, $`a`$ is the lattice constant, $`\mathrm{\Delta }I`$ is the change in interaction energy due to polaron formation, $`\mathrm{\Delta }S`$ is the change in entropy, and $`m=\mathrm{}^2/(2ta^2)`$ with $`t`$ being the hopping integral. At $`T=T_C`$ we get a ferromagnetic transition because the polarons try to align in the same direction and coalesce so as to minimize the kinetic energy and because the area spanned by the polarons is of the order of the area of the system. Also from Eq. (1) it follows that the wavefunction is of the form $`\mathrm{exp}[(\alpha R^2/R_\xi ^2)]`$. Thus the exponential part of the hopping conductivity is of the form $`\mathrm{exp}[2\alpha R^2/R_\xi ^2W_{ij}/(k_BT)]`$ which when minimized in the usual fashion yields $`\mathrm{exp}[(T_0/T)^{1/2}]`$.
Next, we observe that since the Hund’s coupling constant is much larger than the hopping term $`t`$, in an elementary hopping process the hopping probability gets modified by a multiplicative term $`(1+M^2/M_{S}^{}{}_{}{}^{2})/2`$ with $`M/M_S`$ being magnetization fraction. Furthermore, the potential energy difference $`W_{ij}`$ between the two (initial and final) hopping sites does not change with magnetization. Because of the large value of the Hund’s coupling, as now known , the mobile electron will always align its spin parallel to the localized spin. The above multiplicative factor results from the sum of the probabilities for the following four processes: (i)hopping from a randomly or paramagnetically oriented (P) site to a ferromagnetically aligned (F) site \[$`0.5\times (1M/M_S)\times (M/M_S)`$\]; (ii) hopping from a F site to a P site \[$`0.5\times (M/M_S)\times (1M/M_S)`$\]; (iii) hopping from a F site to a F site \[$`(M/M_S)\times (M/M_S)`$\]; and (iv) hopping from a P site to a P site \[$`0.5\times (1M/M_S)\times (1M/M_S)`$\]. The above mentioned multiplicative term has also been deduced by Appel . Thus the hopping conductivity finally takes the form
$`\sigma =e^2R^2\nu _{\mathrm{ph}}{\displaystyle \frac{(1+M^2/M_{S}^{}{}_{}{}^{2})}{2}}N(E_F)\mathrm{exp}[(T_0/T)^{1/2}].`$
Now assuming that in the paramagnetic regime the value of $`M/M_S`$ is small we get the magnetoresistance to be proportional to $`M^2/M_{S}^{}{}_{}{}^{2}`$. Next we assume that the contribution to the magnetization from the polarons is much larger than that from individual spins (which would be reasonable because of the size of the polarons and the fact that at $`T=T_C`$ the area occupied by the polarons is of the order of the area of the system). Thus we approximate $`M/M_S`$ by the Brillouin function $`[g\mu _BJB/(k_BT)]`$.
Lastly, in the ferromagnetic regime the transport is band like and not of hopping type as assumed in Ref. . In the ferromagnetic regime when the magnetization is sizeable, a decrease in resistivity due to an increase in magnetic field would be linear in the increase in magnetization. As the magnetic field in the ferromagnetic regime increases, although the increase in the size of the polaron is small (because Zeeman energy is smaller than interaction energy), more number of polarons get aligned in the direction of the magnetic field thereby increasing the size of the conducting domain. Since the increase in magnetization is mainly due to the magnetization of non-aligned polarons it can be taken to be proportional to $``$.
At higher temperatures larger polarons get aligned due to the weight of the Boltzmann factor. Thus the total spin $`J`$ of the polaron, as obtained from the Brillouin function, decreases as the temperature is lowered below $`T_C`$.
Sudhakar Yarlagadda
Saha Institute of Nuclear Physics, Calcutta, India
Received
PACS numbers: 73.50.Jt, 71.30.+h, 75.70.Pa, 75.70.Ak
|
no-problem/9906/patt-sol9906003.html
|
ar5iv
|
text
|
# Stability of multi-parameter solitons: Asymptotic approach
## I Introduction
Solitary waves (’solitons’) can appear when an initial excitation applied to a medium is strong enough to cause nonlinear response. Formally solitons are solutions of some nonlinear partial differential equations and their dynamics generally is a complex phenomenon, which can be described exactly only in the very special integrable situations . The problems of soliton stability and instability induced dynamics in nonintegrable Hamiltonian models have paramount importance for understanding of a wide range of physical phenomena covering such fields as propagation of electromagnetic, water and plasma waves, condensed matter physics and classical field theory . Several analytical approaches to the stability problem are known. For instance, in the nearly integrable situations the perturbation theory based on the inverse scattering transform can be used . Far from the integrable limit variety of methods can be applied. Among them asymptotic stability theory , method of adiabatically varying soliton parameters , Lyapunov and Evans methods.
Generally, stability of a solitary wave in a Hamiltonian model can be lost due to bifurcations involving appearance of a positive eigenvalue (stationary instability) in the soliton spectrum or a pair of complex conjugate eigenvalues with positive real parts (oscillatory instability) . Both types of these instabilities have been extensively studied in the different solitonic contexts proving their ubiquitousness and fundamental importance, see Refs. and Refs. , respectively, for the stationary and oscillatory instabilities. In the most of the known cases the loss of stability is associated with the collisions of the purely imaginary eigenvalues corresponding to the so called internal modes of the soliton spectrum (see for interesting exceptions).
Applying the above mentioned methods it was shown that in many cases a threshold of stationary instability of multi-parameter solitons is given by the zero of the determinant of the Jacobi matrix $`J_{ij}=_{\kappa _j}Q_i`$, where $`\kappa _j`$ are the soliton parameters and $`Q_i`$ are the associated motion integrals . The condition $`det(J_{ij})=0`$ is, in fact, the compatibility condition of the problem arising in the leading (zero) order of the asymptotic solution of the eigenvalue problem governing stability of the soliton . To find expressions for the eigenvalues it is necessary to proceed further and solve problems arising in higher (at least first) orders. Up to now this was done only for the specific class of model equations having single parameter soliton families . For stationary bifurcations of two-parameter solitons adiabatic method has been applied in Refs. . Linear approximation of this method actually gives an expression for eigenvalues, for more details see Section IV. However, all known developments of this method fail to give criterion indicating transition to the oscillatory instability, i.e. instability with complex eigenvalues. It is also difficult to extend this method beyond its first order because of the rather involved calculations.
The purpose of this work is to formulate a general asymptotic approach to stability of multi-parameter solitons in Hamiltonian models, to show how it can be used to find expressions for the instability growth rates with arbitrary accuracy and to formulate criterion for the oscillatory instability of solitons.
## II Model equations and symmetries
We will consider Hamiltonian equations in the form
$$i\frac{E_n}{z}=\frac{\delta H}{\delta E_n^{}},n=1,2\mathrm{}N,$$
(1)
which describes a wide range of physical phenomena related with self-action and interaction of slowly varying wave envelopes in a variety of nonlinear media , for general review of the Hamiltonian formalism see . Here $`E_n`$ are complex fields, $`z`$ the propagation direction of the interacting waves, $`x`$ the coordinate characterizing dispersion or diffraction, $`H=H(_xE_n,E_n,_xE_n^{},E_n^{})`$ is the Hamiltonian and means complex conjugation. We will assume that $`H`$ is invariant with respect to a set of $`(L1)`$ phase transformations:
$$E_nE_n\mathrm{exp}(i\gamma _{nl}\varphi _l),l=1,2,\mathrm{}(L1),$$
(2)
$`\varphi _l`$ are arbitrary real phases and $`\gamma _{nl}`$ are some constants. Because $`H`$ does not depend on $`x`$ explicitly, Eqs. (1) are also invariant with respect to arbitrary translations along $`x`$:
$$E_n(x,z)E_n(xx_0,z).$$
(3)
Symmetry properties (2), (3) together with Hamiltonian nature of our problem imply presence of $`L`$ conserved quantities, see, e.g., , which are the $`(L1)`$ energy invariants
$$Q_l=𝑑x\underset{n=1}{\overset{N}{}}\gamma _{nl}|E_n|^2,l=1,2,\mathrm{}(L1),$$
(4)
and the momentum
$$Q_L=\frac{1}{2i}𝑑x\underset{n=1}{\overset{N}{}}(E_n^{}_xE_nE_n_xE_n^{}).$$
(5)
Another important consequence of the invariances (2), (3) is that a certain class of solutions of Eqs. (1) can be sought in a form when $`x_0`$ and $`\varphi _l`$ are linear functions of $`z`$, i.e. $`x_0=\kappa _Lz`$ and $`\varphi _l=\kappa _lz`$, then
$$E_n(x,z)=a_n(x\kappa _Lz)\mathrm{exp}(i\underset{l=1}{\overset{L1}{}}\gamma _{nl}\kappa _lz),$$
(6)
where $`\{\kappa _l\}_{l=1}^{L1}`$ and $`\kappa _L`$ are real parameters characterizing, respectively, phase velocities of the interacting waves and the soliton group velocity. Functions $`a_n(\tau )`$ obey a system of the ordinary differential equations
$$(i\kappa _L_\tau +\alpha _n)a_n=\frac{\delta H_a}{\delta a_n^{}},$$
(7)
where $`H_aH(_\tau a_n,a_n,_\tau a_n^{},a_n^{})`$, $`\tau =x\kappa _Lz`$ and $`\alpha _n=_{l=1}^{L1}\gamma _{nl}\kappa _l`$. We assume now that in a certain domain of the parameter space $`(\kappa _1,\kappa _2,\mathrm{}\kappa _L)`$ Eqs. (7) have a family of the solitary solutions such that $`|a_n|0`$ for $`\tau \pm \mathrm{}`$.
## III Asymptotic stability analysis
To study stability of the solitons we seek solutions of Eqs. (1) in the form
$$E_n=(a_n(\tau )+\epsilon _n(\tau ,z))\mathrm{exp}(i\underset{l=1}{\overset{L1}{}}\gamma _{nl}\kappa _lz),$$
(8)
where $`\epsilon _n(\tau ,z)`$ are small complex perturbations. Linearizing Eqs. (1) and assuming that $`\epsilon _n(\tau ,z)=\xi _n(\tau )e^{\lambda z}`$, $`\epsilon _n^{}(\tau ,z)=\xi _{n+N}(\tau )e^{\lambda z}`$ we get the following nonselfadjoint eigenvalue problem (EVP)
$$i\lambda \stackrel{}{\xi }=\widehat{}\stackrel{}{\xi }\left(\begin{array}{cc}\widehat{S}& \widehat{R}\\ \widehat{R}^{}& \widehat{S}^{}\end{array}\right)\stackrel{}{\xi },$$
(9)
where $`\stackrel{}{\xi }=(\xi _1,\mathrm{}\xi _N,\xi _{N+1}\mathrm{}\xi _{2N})^T`$, and $`\widehat{R}`$, $`\widehat{S}`$ are $`N\times N`$ matrix operators with elements given by
$$\widehat{s}_{nl}=\delta _{nl}(\alpha _n+i\kappa _L_\tau )+\frac{\delta ^2H_a}{\delta a_n^{}\delta a_l},\widehat{r}_{nl}=\frac{\delta ^2H_a}{\delta a_n^{}\delta a_l^{}},$$
here $`\delta _{nl}`$ is the Kroneker symbol. Note, that the operator $`\widehat{S}`$ is a selfadjoint one, i.e. $`\widehat{S}=\widehat{S}^{}`$, and $`\widehat{R}`$ is a symmetric operator, i.e. $`\widehat{R}=\widehat{R}^T`$.
To solve EVP (9) we apply the asymptotic approach, which relies on expansion of the unknown eigenvector $`\stackrel{}{\xi }`$ into an asymptotic series near either neutral eigenmodes , i.e. zero-eigenvalue modes, of the operator $`\widehat{}`$, or modes of continuum , or both of them . The neutral modes can be generated by infinitesimal variations of the free parameters of the soliton and thus always be presented as explicit functions of the soliton solution. At the same time continuum eigenmodes are explicitly known in the very rare, normally in integrable, situations . This is an important fact which makes the asymptotic expansion near the neutral modes by the very practical tool of the stability theory. However, as any approximate method, it has a certain limitation. Namely, it describes only eigenvalues $`\lambda `$ corresponding to a specific class of the perturbations which in the zero approximation can be expressed as a linear superposition of the neutral eigenmodes. Thus, generally speaking, on the basis of this approach one can get only sufficient conditions for soliton instability or, in other words, necessary conditions for soliton stability. Therefore presence of other instabilities which can be captured only numerically can always be expected .
By infinitesimal variation of $`\varphi _l`$ and $`x_0`$ it can be shown that
$`\stackrel{}{u}_l=(\gamma _{1l}a_1,\mathrm{}\gamma _{Nl}a_N,\gamma _{1l}a_1^{},\mathrm{}\gamma _{Nl}a_N^{})^T,\stackrel{}{u}_L={\displaystyle \frac{\stackrel{}{a}}{\tau }},`$ (10)
$`\stackrel{}{a}(a_1,\mathrm{}a_N,a_1^{},\mathrm{}a_N^{})^T,l=1,\mathrm{}(L1)`$ (11)
are neutral modes of $`\widehat{}`$, i.e. $`\widehat{}\stackrel{}{u}_l=0`$ $`(l=1,\mathrm{}L)`$. $`\widehat{}`$ also has $`L`$ associated vectors $`\stackrel{}{U}_l=\stackrel{}{a}/\kappa _l`$ such that $`\widehat{}\stackrel{}{U}_l=\stackrel{}{u}_l,l=1,\mathrm{}L`$.
It is straightforward to see that any solution of EVP (9) must obey $`L`$ solvability conditions
$$\stackrel{}{w}_l|\lambda \stackrel{}{\xi }=0,l=1,\mathrm{}L,$$
(12)
where $`\stackrel{}{y}|\stackrel{}{z}=_{i=1}^{2N}𝑑xy_i^{}z_i`$ and $`\stackrel{}{w}_l`$ are the neutral modes of the operator $`\widehat{}^{}`$, $`\widehat{}^{}\stackrel{}{w}_l=0`$,
$`\stackrel{}{w}_l=(\gamma _{1l}a_1,\mathrm{}\gamma _{Nl}a_N,\gamma _{1l}a_1^{},\mathrm{}\gamma _{Nl}a_N^{})^T,\stackrel{}{w}_L=i{\displaystyle \frac{\stackrel{}{b}}{\tau }},`$ (13)
$`\stackrel{}{b}=(a_1,\mathrm{}a_N,a_1^{},\mathrm{}a_N^{})^T,l=1,\mathrm{}(L1).`$ (14)
Associated vectors of $`\widehat{}^{}`$ are $`\stackrel{}{W}_l=\stackrel{}{b}/\kappa _l`$ and they obey $`\widehat{}^{}\stackrel{}{W}_l=\stackrel{}{w}_l,l=1,\mathrm{}L`$.
Close to instability threshold it is naturally to assume that $`|\lambda |ϵ1`$. As it was already discussed above we will consider a special class of the perturbations which in the leading approximation can be presented as a linear combination of the neutral modes. Therefore we seek an asymptotic solution of EVP (9) in the following form
$$\stackrel{}{\xi }=\underset{m=0}{\overset{\mathrm{}}{}}ϵ^m\stackrel{}{\xi }_m(x),\stackrel{}{\xi }_0=\underset{l=1}{\overset{L}{}}C_l\stackrel{}{u}_l$$
(15)
where constants $`C_l`$ and vector-functions $`\stackrel{}{\xi }_{m>0}`$ have to be defined. Here and below $`l=1,2,\mathrm{}L`$. Substitution (15) into EVP (9) gives a recurrent system of equations for $`\stackrel{}{\xi }_m`$
$$\stackrel{}{\xi }_{m>0}=\left[\frac{i\lambda }{ϵ}\widehat{}^1\right]^m\stackrel{}{\xi }_0.$$
(16)
Substituting (15), (16) into conditions (12) one will find the homogeneous system of the $`L`$ linear algebraic equations
$$\lambda ^2\stackrel{}{w}_l|\underset{m=0}{\overset{\mathrm{}}{}}(\lambda ^2)^m\widehat{}^{2m}\underset{l=1}{\overset{L}{}}C_l\stackrel{}{U}_l=0$$
(17)
for $`L`$ unknown constants $`C_l`$. System (17) has a nontrivial solution providing that the corresponding determinant is equal to zero. This determinant is an infinite-order polynomial with respect to $`\lambda ^2`$, which, in fact, is the asymptotic expansion of Evans function . Zeros of this polynomial define the spectrum of the solitary wave linked with the chosen class of the perturbations. Thus the equation specifying eigenvalues $`\lambda `$ is
$$\lambda ^{2L}\underset{j=0}{\overset{\mathrm{}}{}}(\lambda ^2)^jD_j=0,$$
(18)
where $`D_j`$ are the real constants. Eq. (18) always has zero root of the $`2L`$-order. It indicates that each of the zero eigenvalues corresponding to the neutral modes $`\stackrel{}{u}_l`$ is doubly degenerate one. This degeneracy originates from the presence of the associated vectors $`\stackrel{}{U}_l`$.
To write the explicit expressions for $`D_j`$ it will be convenient to introduce vectors $`\stackrel{}{}_l^{(m)}=(_{l1}^{(m)}\mathrm{}_{lL}^{(m)})`$, where,
$$_{ll^{^{}}}^{(m)}=\stackrel{}{w}_l|\widehat{}^{2m}\stackrel{}{U}_l^{^{}},m=0,1,\mathrm{}\mathrm{}.$$
Now each $`D_j`$ can be presented as
$$D_j=\underset{m_1+\mathrm{}m_L=j}{}𝒟(\stackrel{}{}_1^{(m_1)},\mathrm{}\stackrel{}{}_L^{(m_L)}),$$
(19)
where $`𝒟(\stackrel{}{}_1^{(m_1)},\mathrm{}\stackrel{}{}_L^{(m_L)})`$ is the determinant of the $`L\times L`$ matrix consisting of the rows $`\stackrel{}{}_l^{(m_l)}`$ and the sum is taken over all such combinations of $`(m_1,\mathrm{}m_L)`$ that $`_{l=1}^Lm_l=j`$. $`_{ll^{^{}}}^{(0)}`$ can be readily expressed via derivatives of the conserved quantities with respect to the soliton parameters:
$$_{ll^{^{}}}^{(0)}=\frac{Q_l}{\kappa _l^{^{}}},$$
(20)
and practical calculation of $`_{ll^{^{}}}^{(m)}`$ for $`m>0`$ can be simplified: $`_{ll^{^{}}}^{(m)}=\stackrel{}{W}_l|\widehat{}^{(12m)}\stackrel{}{U}_l^{^{}}.`$ Note, that in most of the cases solitary solution itself can be found only numerically using any of the well established methods for solving the nonlinear ode’s. Recurrent calculations of $`\widehat{}^{(12m)}\stackrel{}{U}_l`$ can be readily reduced to the numerically even simpler problem of solving of the linear inhomogeneous ode’s.
Because $`|\lambda |`$ was assumed to be small, Eq. (18) has an asymptotic character. Therefore to make it work some additional assumptions must be made about orders of $`D_j`$. If these assumptions are satisfied then Eq. (18) describes correctly the soliton spectrum and predicts bifurcations of the soliton. The corresponding eigenvalues can be found using Eq. (18) with any degree of accuracy. For example, let us assume that $`D_0ϵ^2`$ and $`D_{j>0}O(1)`$. Then, presenting $`\lambda ^2`$ as
$$\lambda ^2=ϵ^2\underset{j=0}{\overset{\mathrm{}}{}}\zeta _j,\zeta _jϵ^{2j},$$
(21)
in the first order Eq. (18) gives a linear equation for $`\zeta _0`$,
$$D_0ϵ^2\zeta _0D_1=0,$$
(22)
which indicates a threshold of the stationary bifurcation at $`D_0=0`$. This is precisely the condition $`det(J_{ij})=0`$ discussed in the introduction. Continuing to the next order one obtains
$$\lambda ^2=\frac{D_0}{D_1}\left(1\frac{D_0D_2}{D_1^2}+O(ϵ^4)\right).$$
(23)
If $`D_1ϵ^2`$ then the asymptotic expression (23) fails. To have a balanced equation for $`\zeta _0`$, we must now assume that $`D_0ϵ^4`$. However, in such a case the Eq. 22 for $`\zeta _0`$ changes from linear to quadratic:
$$D_0ϵ^2\zeta _0D_1+ϵ^4\zeta _0^2D_2=0.$$
(24)
Eq. (24) gives two threshold conditions $`D_0=0`$ and $`D_1^2=4D_0D_2`$, see Fig. 1. The latter condition indicates onset of the oscillatory instability for
$$D_1^2<4D_0D_2.$$
(25)
Thus we have formulated analytic criterion for the oscillatory instability. It is also clear that the point $`D_{0,1}=0`$ is a source for the novel stationary instability, see rightmost region $`D_1^2>4D_0D_2`$, $`D_1>0`$ in Fig. 1, where an eigenvalue which is positive throughout this region can not be predicted by Eq. (23).
It follows by recurrence that if $`D_{j^{^{}}>0}ϵ^2`$ then to have a balanced equation for $`\zeta _0`$ we must assume that $`D_{j<j^{^{}}}ϵ^{2(1+j^{^{}})}`$. In other words asymptotic expansion near the neutral modes can only describe the soliton spectrum in regions of the parameter space which are close to codimension-$`(j^{^{}}+1)`$ bifurcation. If $`j^{^{}}=0`$ then only one condition must be satisfied and our asymptotic approach predicts presence of either two purely imaginary or two purely real eigenvalues, which can collide at zero. If $`j^{^{}}=1`$ then two conditions must be satisfied and the asymptotic approach predicts presence of two pairs of eigenvalues which can be real, imaginary or complex. In this situation the soliton becomes oscillation unstable providing that two pairs of imaginary eigenvalues collided. For each further $`j^{^{}}`$ two new eigenvalues come into play.
## IV Discussion
General formulae (18),(19) giving soliton eigenvalues with any degree of accuracy and criterion for the oscillatory instability (25) are main novel results of this work. At the same time expressions for the eigenvalues near the stationary instability threshold, analogs of the formula $`\lambda ^2=D_0/D_1+\mathrm{}`$, have been earlier obtained in a number of papers. It is instructive now to give explicit expressions for $`D_j`$ in the two simplest situations of one- and two-parameter solitons and to compare them with previously reported results. For the one parameter solitons: $`D_0=_{\kappa _1}Q_1`$, $`D_1=\stackrel{}{W}_1|\widehat{}^1\stackrel{}{U}_1`$, $`D_2=\stackrel{}{W}_1|\widehat{}^3\stackrel{}{U}_1`$. Using these formulae one can show that in the case when $`D_1O(1)`$ the first term in Eq. (23) gives the same expression for $`\lambda ^2`$ which was obtained in Refs. , where generalised Nonlinear Shrödinger equation and equations describing propagation in quadratically nonlinear media have been investigated. If $`D_1D_2>0`$ then it can be concluded that the second term in Eq. (23) indicates saturation of the growth rate when the distance from the instability threshold, $`D_0=0`$, growthes, which agrees with numerical results .
For the two-parameter solitons:
$`D_0=\left|\begin{array}{cc}\frac{Q_1}{\kappa _1}& \frac{Q_1}{\kappa _2}\\ \frac{Q_2}{\kappa _1}& \frac{Q_2}{\kappa _2}\end{array}\right|,`$ (28)
$`D_1=\left|\begin{array}{cc}\frac{Q_1}{\kappa _1}& \frac{Q_1}{\kappa _2}\\ _{21}^{(1)}& _{22}^{(1)}\end{array}\right|+\left|\begin{array}{cc}_{11}^{(1)}& _{12}^{(1)}\\ \frac{Q_2}{\kappa _1}& \frac{Q_2}{\kappa _2}\end{array}\right|,`$ (33)
$`D_2=\left|\begin{array}{cc}_{11}^{(1)}& _{12}^{(1)}\\ _{21}^{(1)}& _{22}^{(1)}\end{array}\right|+\left|\begin{array}{cc}\frac{Q_1}{\kappa _1}& \frac{Q_1}{\kappa _2}\\ _{21}^{(2)}& _{22}^{(2)}\end{array}\right|+\left|\begin{array}{cc}_{11}^{(2)}& _{12}^{(2)}\\ \frac{Q_2}{\kappa _1}& \frac{Q_2}{\kappa _2}\end{array}\right|.`$ (40)
The threshold condition $`D_0=0`$ has been previously found for two-parameter solitons in different physical contexts . However derivation of an accurate expression for the soliton eigenvalues near this threshold has remained a controvertial problem. Indeed, comparing eigenvalues given by Eqs. (23), (28), (33),(40) and eigenvalues which can be calculated from the ordinary differential equations for soliton parameters presented in one will discover that results are slightly different . It has also been argued that the sign of $`(_{11}^{(1)}_{22}^{(1)}_{12}^{(1)}_{21}^{(1)})`$, which is the first term in Eq. (40), plays an important role in stability of two-parameter solitons. However Eqs. (23),(28),(33),(40) apparently conflict with this finding.
Among open problems I would like to mention derivation of finite-dimensional normal forms describing dynamical evolution of the soliton parameters near the oscillatory instability threshold. A guideline for this work can be theory of G. Iooss for the normal forms of the reversible ordinary differential equations in vicinity of the codimension-2 bifurcation, wich is an equivalent of the our point $`D_0=D_1=0`$. The simplest case of the codimension-1 stationary instability, $`D_0=0`$, has only one homoclinic orbit separating regions of the periodic oscillations from the spreading or collapse . The vicinity of the codimension-2 point can contain the very reach dynamics, including multiple homoclinic orbits and stochastic regimes.
## V Summary
General form of the asymptotic approach to stability problem of multi-parameter solitons in Hamiltonian systems has been developed. It has been shown that the asymptotic study of the soliton stability reduces to the calculation of a certain sequence of determinants, where the famous determinant of the matrix consisting from the derivatives of the system invariants with respect to the soliton parameters is just the first in the series. Knowledge of these determinants allows to calculate eigenvalues governing soliton instability with arbitrary accuracy. The most important consequence is that the presented approach gives first analytic criterion for the oscillatory instability of solitons in Hamiltonian systems.
## VI Acknowledgment
Author acknowledges useful discussions with W.J. Firth and D.E. Pelinovsky and financial support from the Royal Society of Edinburgh and British Petroleum.
|
no-problem/9906/cond-mat9906010.html
|
ar5iv
|
text
|
# Interplay between orbital ordering and lattice distortions in LaMnO3, YVO3, and YTiO3
## I Introduction
Orbital ordering and concomitant Jahn-Teller (JT) distortions are observed in some perovskite-type 3$`d`$ transition-metal compounds such as KCuF<sub>3</sub> and LaMnO<sub>3</sub>. In the perovskite-type lattice, there are two possible JT distortions depending on the stacking of the elongated octahedra along the $`c`$-axis as shown in Fig. 1(a). In the $`d`$-type JT distortion, the elongated axes of the octahedra are parallel along the $`c`$-axis. On the other hand, the elongated axes are rotated by 90 along the $`c`$-axis in the $`a`$-type JT distortion. While LaMnO<sub>3</sub> ($`d^4`$), YVO<sub>3</sub> ($`d^2`$) and YTiO<sub>3</sub> ($`d^1`$) have the $`d`$-type JT distortion, LaVO<sub>3</sub> ($`d^2`$) has the $`a`$-type JT distortion. In KCuF<sub>3</sub>, both the $`d`$-type and the $`a`$-type JT distortions are observed. Hartree-Fock calculations which consider the hybridization between the transition-metal 3$`d`$ and oxygen 2$`p`$ orbitals predict that the orbital ordered state compatible with the $`a`$-type JT distortion is lower in energy than that with the $`d`$-type JT distortion for $`d^1`$ and $`d^2`$ systems and that the two states are degenerate for $`d^4`$ and $`d^9`$ systems. Therefore, one cannot explain why the $`d`$-type JT distortion is realized in LaMnO<sub>3</sub>, YVO<sub>3</sub> and YTiO<sub>3</sub> by considering the energy gain due to orbital ordering alone.
Perovskite-type $`AB`$O<sub>3</sub> compounds with relatively small $`A`$-site ions undergo the GdFeO<sub>3</sub>-type distortion which is caused by tilting of $`B`$O<sub>6</sub> octahedra as shown in Fig. 1(b). While LaMnO<sub>3</sub>, YVO<sub>3</sub>, LaVO<sub>3</sub> and YTiO<sub>3</sub> are accompanied by the GdFeO<sub>3</sub>-type distortion, KCuF<sub>3</sub> has no GdFeO<sub>3</sub>-type distortion. In addition, the magnitude of the GdFeO<sub>3</sub>-type distortion becomes larger in going from LaVO<sub>3</sub> and LaMnO<sub>3</sub> to YVO<sub>3</sub> and YTiO<sub>3</sub>. Here, one can notice that the compounds with the larger GdFeO<sub>3</sub>-type distortion tend to have the $`d`$-type JT distortion. It has been pointed out by Goodenough that the covalency between the $`A`$-site and oxygen ions ($`A`$-O covalency) is important in the GdFeO<sub>3</sub>-type distortion. In this paper, we have studied the relationship between the GdFeO<sub>3</sub>-type and JT distortions considering the $`A`$-O covalency and explored the reason why orbital ordering compatible with the $`d`$-type JT distortion are favored in LaMnO<sub>3</sub>, YVO<sub>3</sub> and YTiO<sub>3</sub> in terms of the interaction between the two distortions.
## II Method of calculation
We have employed lattice models for the perovskite-type structure in which the transition-metal 3$`d`$, the oxygen 2$`p`$ and the $`A`$-site cation $`d`$ orbitals are included. The on-site Coulomb interaction between the transition-metal 3$`d`$ orbitals, which is essential to make the system insulating and to cause orbital ordering, are expressed using Kanamori parameters $`u`$, $`u^{}`$, $`j`$ and $`j^{}`$. The charge-transfer energy $`\mathrm{\Delta }`$ is defined by $`ϵ_d^0ϵ_p+nU`$, where $`ϵ_d^0`$ and $`ϵ_p`$ are the energies of the bare transition-metal 3$`d`$ and oxygen 2$`p`$ orbitals and $`U`$ (=$`u`$-20/9$`j`$) is the multiplet averaged $`d`$-$`d`$ Coulomb interaction energy. The hybridization between the transition-metal 3$`d`$ and oxygen 2$`p`$ orbitals is expressed by Slater-Koster parameters ($`pd\sigma `$) and ($`pd\pi `$). The ratio ($`pd\sigma `$)/($`pd\pi `$) is fixed at -2.16. $`\mathrm{\Delta }`$, $`U`$, and ($`pd\sigma `$) can be deduced from cluster-model analysis of photoemission spectra. Although the error bars of these parameters estimated from photoemission spectra are not so small \[$`\pm `$1 eV for $`\mathrm{\Delta }`$ and $`U`$ and $`\pm `$0.2 eV for $`(pd\sigma )`$\], the conclusions obtained in the present calculations are not changed if the parameters are varied within the error bars.
In the present model, unoccupied $`d`$ orbitals of the $`A`$-site cation such as Y 4$`d`$ and La 5$`d`$ are taken into account. The hybridization term between the oxygen 2$`p`$ orbitals and $`A`$-site cation $`d`$ orbitals is expressed by $`(pd\sigma )_A`$ and $`(pd\pi )_A`$. The ratio $`(pd\sigma )_A`$/$`(pd\pi )_A`$ is also fixed at -2.16. The hybridization term between the oxygen 2$`p`$ orbitals is given by ($`pp\sigma `$) and ($`pp\pi `$) and the ratio ($`pp\sigma `$)/($`pp\pi `$) is fixed at -4. It is assumed that the transfer integrals $`(pd\sigma )`$ and $`(pd\sigma )_A`$ are proportional to $`d^{3.5}`$ and $`(pp\sigma )`$ is to $`d^2`$, where $`d`$ is the bond length. Without the JT and GdFeO<sub>3</sub>-type distortions, the bond length between two neighboring oxygens and that between the oxygen and the $`A`$-site cation are $`\sqrt{2}a`$, where $`a`$ is the bond length between the transition-metal ion and the oxygen. ($`pp\sigma `$) and $`(pd\sigma )_A`$ for bond length of $`\sqrt{2}a`$ are assumed to be -0.60 and -1.0 eV, respectively.
In the GdFeO<sub>3</sub>-type distortion, the four octahedra in the unit cell are rotated by angle of $`\omega `$ around the axes in the (0,1,1) plane in terms of the orthorombic unit cell. Here, we model the GdFeO<sub>3</sub>-type distortion by rotating the octahedra around the (0,1,0)-axis or the $`b`$-axis \[see Fig. 1(b)\]. The subsequent small rotation around the $`a`$-axis is required to retain the corner-sharing network of the octahedra. The magnitude of the GdFeO<sub>3</sub>-type distortion is expressed by the tilting angle $`\omega `$ around the $`b`$-axis. It is important that the $`A`$-site ions are shifted approximately along the $`b`$-axis to decrease the distance from the $`A`$-site ion to the three closest oxygen ions and increase the distance to the three next closest oxygen ions as shown in Fig. 1(b). Here, it is assumed that the shift is along the ($`\pm `$1/8,7/8,0)-direction. The magnitude of the shift is proportional to the tilting angle and is assumed to be $``$ 0.05$`a`$, 0.1$`a`$ and 0.15$`a`$ for the tilting angles of 5, 10 and 15, respectively, which are realistic values for the compounds studied in the present work. As for the Jahn-Teller distortion, it is assumed that the longest bond is by 0.1$`a`$ longer than the shortest bond which is reasonable for LaMnO<sub>3</sub> and is relatively large for LaVO<sub>3</sub> and YTiO<sub>3</sub>.
## III Results and Discussion
### A LaMnO<sub>3</sub>
In the high-spin $`d^4`$ system, in which one of the $`e_g`$ orbitals is occupied at each site, the $`A`$-type antiferromagnetic (AFM) states with $`3x^2r^2`$/$`3y^2r^2`$-type orbital ordering with considerable mixture of $`3z^2r^2`$ are predicted to be stable by theoretical calculations and are studied by x-ray and neutron diffraction measurements. Here, the $`z`$-direction is along the $`c`$-axis. The amount of the $`3z^2r^2`$ component decreases with the JT distortion. Different ways of stacking the orbitals along the $`c`$-axis give two types of orbital ordering: the one compatible with the $`d`$-type JT distortion and the other with the $`a`$-type JT distortion. These two types of orbital ordering are illustrated in Fig. 2. While, in the orbital ordering of the $`a`$-type, the sites 1, 2, 3, and 4 are occupied by $`3y^2r^2`$, $`3x^2r^2`$, $`3x^2r^2`$, and $`3y^2r^2`$ orbitals, the sites 1, 2, 3, and 4 are occupied by $`3y^2r^2`$, $`3x^2r^2`$, $`3y^2r^2`$, and $`3x^2r^2`$ orbitals in the orbital ordering compatible with the $`d`$-type JT distortion.
In Fig. 3, the energy difference between the orbital ordered states compatible with the $`d`$-type and $`a`$-type JT distortions is plotted as a function of the tilting angle of the octahedra, i.e., the magnitude of the GdFeO<sub>3</sub>-type distortion. $`\mathrm{\Delta }`$, $`U`$, and $`(pd\sigma )`$ are 4.0, 5.5, and -1.8 eV, respectively, for LaMnO<sub>3</sub>. Without the JT distortion and the shift of the $`A`$-site ion, the two states are degenerate within the accuracy of the present calculation ($`\pm `$ 1 meV/formula unit cell). This degeneracy is lifted when the JT distortion is included. With the JT distortion and the shift of the $`A`$-site ion, the orbital ordered state with the $`d`$-type JT distortion is lower in energy than that with the $`a`$-type JT distortion. If we tentatively switch off the shift of the $`A`$-site ion and include only the JT distortion, the orbital ordered state with the $`a`$-type JT distortion becomes slightly lower than that with the $`d`$-type JT distortion as shown in Fig. 3. Therefore, one can conclude that the shift of the $`A`$-site ion driven by the GdFeO<sub>3</sub>-type distortion is essential to stabilize the orbital ordered state with the $`d`$-type Jahn-Teller distortion.
The qualitative explanation of this behavior is as follows. In the $`d`$-type JT distortion, the four oxygen ions nearest to the $`A`$-site ion \[shaded in Fig. 1 (a) and (b)\] shift approximately in the same direction and, consequently, the system can gain the hybridization energy between the $`A`$-site and oxygen ions effectively. On the other hand, in the $`a`$-type JT distortion, the two of the four oxygen ions move in the other direction and the energy gain due to the hybridization is small compared to the $`d`$-type JT distortion. Another possible picture is that, in the $`d`$-type JT distortion, these four oxygen ions can push the $`A`$-site ion in the same direction since the JT distortion along the $`c`$-axis is in phase. On the other hand, in the case of the $`a`$-type JT distortion, the two oxygen ions in the upper plane push the $`A`$-site ion in the other direction than the two in the lower plane as shown in Fig. 4. Therefore, the stronger is the GdFeO<sub>3</sub>-type distortion, the more does it stabilize the $`d`$-type Jahn-Teller distortion and corresponding orbital ordering.
In the charge-ordered state of Pr<sub>0.5</sub>Ca<sub>0.5</sub>MnO<sub>3</sub>, the Mn<sup>3+</sup> and Mn<sup>4+</sup> sites are arranged like a checkerboard within the $`c`$-plane and the same arrangement is stacked along the $`c`$-axis. The Mn<sup>3+</sup> sites are accompanied by the JT distortion and the elongated axes are parallel along the $`c`$-axis just like the $`d`$-type JT distortion. Since the $`A`$-sites are occupied by Pr and Ca ions in Pr<sub>0.5</sub>Ca<sub>0.5</sub>MnO<sub>3</sub>, we cannot simply apply the present model calculation to it. However, it is reasonable to speculate that the stacking along the $`c`$-axis in Pr<sub>0.5</sub>Ca<sub>0.5</sub>MnO<sub>3</sub> is also determined by the interaction between the JT distortion and the shift of the $`A`$-site ion in the same way as in LaMnO<sub>3</sub>.
### B YVO<sub>3</sub>
In the $`d^2`$ system, the $`C`$-type AFM state in which the sites 1, 2, 3, and 4 are occupied by $`xy`$ and $`yz`$, $`xy`$ and $`zx`$, $`xy`$ and $`zx`$, and $`xy`$ and $`yz`$ orbitals and the $`G`$-type AFM state in which the sites 1, 2, 3, and 4 are occupied by $`xy`$ and $`yz`$, $`xy`$ and $`zx`$, $`xy`$ and $`yz`$, and $`xy`$ and $`zx`$ orbitals are competing. While the $`C`$-type AFM state is favored by the orbital ordering which is compatible with the $`a`$-type JT distortion, the $`G`$-type AFM state is favored by the orbital ordering of the $`d`$-type. The relative energy of the $`G`$-type AFM state with the $`d`$-type JT distortion to the $`C`$-type AFM state with the $`a`$-type JT distortion, $`E_dE_a`$, is plotted as a function of the tilting angle in Fig. 5. $`\mathrm{\Delta }`$, $`U`$, and $`(pd\sigma )`$ are 6.0, 4.5, and -2.2 eV, respectively, for LaVO<sub>3</sub> and YVO<sub>3</sub>. Without the GdFeO<sub>3</sub>-type distortion, the $`C`$-type AFM state with the $`a`$-type JT distortion is lower in energy than the $`G`$-type AFM state, indicating that the energy gain due to the orbital ordering is larger in the $`C`$-type AFM state than in the $`G`$-type AFM state. The energy difference becomes smaller with the tilting or the GdFeO<sub>3</sub>-type distortion. Finally, with the tilting of 15, the $`G`$-type AFM state with the $`d`$-type JT distortion becomes lower in energy than the $`C`$-type AFM state. The present calculation is in good agreement with the experimental result that the less distorted LaVO<sub>3</sub> is $`C`$-type AFM below 140 K and the more distorted YVO<sub>3</sub> is $`G`$-type AFM below 77 K.
This situation is illustrated in Fig. 6. When the GdFeO<sub>3</sub>-type distortion is large, the interaction between the $`d`$-type JT distortion and the shift of the $`A`$-site ion, namely, the energy gain due to $`A`$-O covalency becomes dominant just like in LaMnO<sub>3</sub> and, consequently, the $`G`$-type AFM with the $`d`$-type JT distortion is favored. On the other hand, when the GdFeO<sub>3</sub>-type distortion is small, the energy gain due to orbital ordering becomes dominant and the $`a`$-type JT distortion and the orbital ordering compatible with it are realized. The JT distortion may be suppressed if the system is located near the crossing point where the $`a`$-type and $`d`$-type JT distortions are almost degenerate. An interesting experimental result related to this point is that YVO<sub>3</sub> becomes $`C`$-type AFM between 77 K and 118 K. The present model calculation suggests that, if the JT distortion is switched off, the $`C`$-type AFM state is favored because of the orbital ordering. Therefore, YVO<sub>3</sub> is expected to be close to the crossing point and become $`C`$-type AFM when the $`d`$-type JT distortion is suppressed at elevated temperature. In this sense, between 77 K and 118 K, YVO<sub>3</sub> may be an ideal orbitally ordered system without JT distortion.
### C YTiO<sub>3</sub>
For the $`d^1`$ system, the ferromagnetic (FM) states with orbital ordering are favored in the model HF calculations. There are two possible orbital orderings compatible with the $`a`$-type and $`d`$-type JT distortions. The model HF calculation without the covalency between the $`A`$-site cation and the oxygen ion predicted that the orbital ordering of the $`a`$-type is lower in energy. However, the recent neutron diffraction measurement by Akimitsu et al. have shown that the orbital ordering compatible with the $`d`$-type JT distortion is realized in the FM insulator YTiO<sub>3</sub>. YTiO<sub>3</sub> has the considerable GdFeO<sub>3</sub>-type distortion and the tilting angle is expected to be larger than 15. In Fig. 7, the relative energy of the FM and orbital ordered state of the $`d`$-type to that of the $`a`$-type is plotted as a function of the tilting. $`\mathrm{\Delta }`$, $`U`$, and $`(pd\sigma )`$ are 7.0, 4.0, and -2.2 eV, respectively, for YTiO<sub>3</sub>. With the tilting of 15, the orbital ordered state of the $`d`$-type is lower in energy, in agreement with the experimental result. In this state, the sites 1, 2, 3, and 4 are occupied by $`c_1yz+c_2xy`$, $`c_1zx+c_2xy`$, $`c_1yzc_2xy`$, and $`c_1zxc_2xy`$ orbitals ($`c_10.8`$ and $`c_20.6`$). However, experimentally, less distorted LaTiO<sub>3</sub> has no or very small JT distortion and has a $`G`$-type AFM state. The present calculation cannot explain why the $`G`$-type AFM state can be stable compared to the FM state in LaTiO<sub>3</sub>.
## IV Conclusion
In conclusion, we have studied the relationship between orbital ordering and the JT and GdFeO<sub>3</sub>-type lattice distortions. It has been found that the covalency between the $`A`$-site cations and oxygen makes the $`d`$-type JT distortion (same orbitals along the $`c`$-direction) lower in energy than the $`a`$-type JT distortion (alternating orbitals along the $`c`$-direction) in the presence of the large GdFeO<sub>3</sub>-type distortion. As a result, the orbital ordered states compatible with the $`d`$-type JT distortion are favored in LaMnO<sub>3</sub>, YVO<sub>3</sub>, and YTiO<sub>3</sub> which have the relatively large GdFeO<sub>3</sub>-type distortion. On the other hand, in less distorted LaVO<sub>3</sub>, the orbital ordering compatible with the $`a`$-type JT distortion is favored because of the pure superexchange effect.
## Acknowledgment
The authors would like to thank useful discussions with K. Tomimoto, J. Akimitsu, Y. Ren and P. M. Woodward. This work was supported by the Nederlands foundation for Fundamental Research of Matter (FOM).
|
no-problem/9906/physics9906016.html
|
ar5iv
|
text
|
# Electromagnetic modes in cold magnetized strongly coupled plasmas
## I Introduction
The aim of this paper is to find the spectrum of electromagnetic waves propagating in a strongly coupled magnetized fully ionized hydrogen plasma without taking into account the ion motion. We make use of the dielectric tensor of cold magnetized plasmas constructed in Ref. by means of the classical theory of moments.
In neglect of thermal motion this dielectric tensor reads:
$$\epsilon _{\mu \nu }=\left(\begin{array}{ccc}\epsilon _{}& ig& 0\\ ig& \epsilon _{}& 0\\ 0& 0& \epsilon _{}\end{array}\right),$$
(1)
where $`\epsilon _{}`$ and $`\epsilon _{}`$ are the transverse and longitudinal (with respect to the external magnetic field, $`\stackrel{}{B}`$) components of the tensor.
We will consider the damping of the modes in question as negligibly small. This assumption obviously can be verified only experimentally. The damping can be essential and must be taken into account near the cyclotron resonances. Here the thermal motion of the particles leading to the spatial dispersion must be accounted for also. Thus our results are valid only far from the cyclotron resonances and in coupled plasma systems with the plasma parameter $`\mathrm{\Gamma }=e^2/aT1`$ ( -e is the electron charge, $`a`$ is the Wigner-Seitz radius and $`T`$ is the plasma temperature). For laboratory plasmas this condition implies the temperature $`T23eV`$ and the number density of electrons $`n10^{21}cm^3`$. The electrical conductivity $`\sigma `$ of such systems with strong Coulomb coupling is of the order of $`\omega _p`$, so that their effective collisional frequency $`\nu =\omega _p^2/4\pi \sigma `$ is at least an order of magnitude smaller than $`\omega _p`$. Similar conditions can also be realized in astrophysical systems (crust of neutron stars, the interior of white dwarfs and large planets with very strong magnetic fields, etc.). Further we will regard only long wavelengths modes for which the condition of a cold plasma holds. In addition the frequencies of these modes will be presumed to be much higher than the ion cyclotron frequency, so that the ion motion contribution could be neglected.
The components of the dielectric tensor of a system under consideration were found in, and within the first approximation in the ratio $`\sqrt{m/M}`$ ($`m`$ and $`M`$ being the electron and the ion masses):
$$\epsilon _{}=1\omega _p^2\frac{\omega ^2\mathrm{\Omega }_{}^2}{(\omega ^2\mathrm{\Omega }_{}^2)^2\omega ^2\omega _B^2},\epsilon _{}=1\frac{\omega _p^2}{\omega ^2\mathrm{\Omega }_{}},g=\omega _p^2\frac{\omega \omega _B}{(\omega ^2\mathrm{\Omega }_{}^2)^2\omega ^2\omega _B^2},$$
(2)
where $`\omega _p=(4\pi ne^2/m)^{1/2}`$ is the plasma frequency, and $`\omega _B=eB/mc`$ is the electron cyclotron frequency. The positive magnitudes $`\mathrm{\Omega }_{}`$ and $`\mathrm{\Omega }_{}`$ take into account the Coulomb correlations between the particles and are expressible via the second frequency moment of the magnetized plasma conductivity tensor Hermitian part, so that
$$\mathrm{\Omega }_{}^2=\frac{\omega _p^2}{2}\underset{\stackrel{}{q}0}{}S_{ei}(\stackrel{}{q})\frac{q_{}^2}{q^2},\mathrm{\Omega }_{}^2=\omega _p^2\underset{\stackrel{}{q}0}{}S_{ei}(\stackrel{}{q})\frac{q_{}^2}{q^2},$$
(3)
$`S_{ei}(\stackrel{}{q})`$ being the partial electron-ion static structure factor, and $`q_{}`$ (and $`q_{}`$) is the projection of the vector $`\stackrel{}{q}`$ on the direction perpendicular (parallel) to the external magnetic field. An analysis of these magnitudes is given in the third section, we mention here only that in the ideal plasma limit both $`\mathrm{\Omega }_{}`$ and $`\mathrm{\Omega }_{}0`$. We also wish to emphasize that the electron-ion correlations is the factor which guarantees the existence of nonvanishing parameters $`\mathrm{\Omega }_{}`$ and $`\mathrm{\Omega }_{}`$. Notice that the above expressions, Eqs. (2) for the dielectric tensor components coincides (within the first order in the ratio $`\sqrt{m/M}`$) with that of the quasilocalized charges model developed by Kalman and Golden .
## II Waves in strongly coupled magnetized plasmas
If we choose the Cartesian system of coordinates with the z-axis parallel to the external magnetic field $`\stackrel{}{B}`$, then the dispersion equation of electromagnetic waves propagating in a magnetized plasma takes the form:
$$AN^4+BN^2+C=\mathrm{\hspace{0.17em}\hspace{0.17em}0},$$
(4)
where $`N=\omega _0/\omega `$ is the scalar refraction index, $`\omega _0=|\stackrel{}{k}|c`$, and
$$A=\epsilon _{}\mathrm{sin}^2\theta +\epsilon _{}\mathrm{cos}^2\theta ,B=\epsilon _{}\epsilon _{}(1+\mathrm{cos}^2\theta )(\epsilon _{}^2g^2)\mathrm{sin}^2\theta ,C=\epsilon _{}(\epsilon _{}^2g^2),$$
(5)
and $`\theta `$ is the angle between the wavevector $`\stackrel{}{k}`$ and the magnetic field $`\stackrel{}{B}`$.
Eq.(4) has two different solutions:
$$N_\pm ^2=\left[B\pm (B^24AC)^{1/2}\right]/2A,$$
(6)
which are usually associated with ordinary and extraordinary waves: two different kinds of waves of a given frequency and with different refraction indices, which can propagate in magnetized plasmas. These waves are generally elliptically polarized, a wave which propagates along the external magnetic field is transverse polarized; the ordinary wave is characterized by the right-handed circular polarization, the extraordinary wave is left-handed polarized.
The frequencies that satisfy the relation $`A(\omega ,\stackrel{}{k})=0`$ are traditionally called the plasma resonances. Notice that one of the refraction indices tends to infinity as the frequency approaches the resonance value $`N_+^2=B/A`$, while the second one remains finite, $`N_{}^2=C/B`$. A cubic equation with respect to $`\omega ^2`$ can be obtained from Eq. (5). It determines three resonance frequencies. This is in contrast to ideal magnetized plasmas ($`\mathrm{\Omega }_{}=\mathrm{\Omega }_{}=0`$), where only two resonances exist (we neglect the ion motion!). For the case of near longitudinal propagation $`\theta 1`$ these resonances are
$$\omega _\pm ^{(p)}=\frac{1}{2}\left\{\omega _B+\left[\omega _B^2+4\mathrm{\Omega }_{}^2\right]^{1/2}\right\},\omega _3^p=\sqrt{\omega _p^2+\mathrm{\Omega }_{}^2}\left(1+\frac{\theta ^2}{2}\frac{\omega _p^2\omega _B^2}{\omega _p^4\omega _p^2\omega _B^2\mathrm{\Omega }_{}^2\omega _B^2}\right).$$
(7)
For the case of transverse propagation we found for the refraction indices poles:
$$\left(\omega _\pm ^{(p)}\right)^2=\mathrm{\Omega }_{}^2+\frac{1}{2}\left\{\omega _p^2+\omega _B^2\left[\left(\omega _p^2+\omega _B^2\right)^2+4\omega _B^2\mathrm{\Omega }_{}^2\right]^{1/2}\right\},\omega _3^{(p)}=\mathrm{\Omega }_{}.$$
(8)
The zeros of the $`N_\pm ^2`$ determine the boundaries between the domains of propagation for different waves. From Eq. (4) it follows that $`N_\pm =0`$, if the coefficient $`C`$ is equal to zero. We found three zeros,
$$\omega _\pm ^{(0)}=\frac{1}{2}\left\{\omega _B+\left[\omega _B^2+4\left(\omega _p^2+\mathrm{\Omega }_{}^2\right)\right]^{1/2}\right\},\omega _3^{(0)}=\left(\omega _p^2+\mathrm{\Omega }_{}^2\right)^{1/2}.$$
(9)
With the poles and zeros determined, and taking into account that $`N_\pm ^2(\omega =0)=1+\omega _p^2/\mathrm{\Omega }_{}N_0^2`$, and the relation $`N_\pm ^2(\omega \mathrm{})\mathrm{\hspace{0.17em}1}`$, the refraction indices can be plotted. In Fig.1 the frequency dependence of the refractive indices for an angle $`0<\theta <\pi `$ is shown.
The branches of propagation ($`N^2(\omega )>0`$) are associated with the eigenfrequencies $`\omega _k`$. The latter are given in Fig.2 vs. wavevector. The modes $`\omega _k`$ are determined by Eq. (4). Since in the present approximation Eq. (4) is the fifth order equation (with respect to $`\omega ^2`$), we find five eigenmodes. In ideal plasmas in neglect of the Alfvén wave only four eigenfrequencies can be found. From Fig. 2 we observe that in the case of a strongly coupled plasma the ideal plasma helicon wave splits into two branches, which we call the strongly coupled plasma whistling sound waves.
Thus in strongly coupled magnetized plasma there can exist five eigenmodes: ordinary and extraordinary whistling sound waves, the slow extraordinary, the ordinary and the fast extraordinary waves.
Consider now in more details the dispersion of the whistling sound waves at small wavenumbers, i.e., when $`\omega \omega _B`$. In this spectral region the dispersion equation reduces to a quadratic equation with respect to $`\omega ^2`$. For the case of parallel with respect to the external magnetic field propagation the corresponding solution reads:
$$\omega _k^{(4,5)}=\frac{1}{2}\left\{\pm \frac{\omega _B\omega _0^2}{\omega _p^2+\mathrm{\Omega }_{}^2+\omega _0^2}+\left[\frac{\omega _B^2\omega _0^4}{(\omega _p^2+\mathrm{\Omega }_{}^2+\omega _0^2)^2}+4\frac{\omega _0^2\mathrm{\Omega }_{}^2}{\omega _p^2+\mathrm{\Omega }_{}^2+\omega _0^2}\right]^{1/2}\right\}.$$
(10)
In ideal magnetized plasmas (i.e., when $`\mathrm{\Omega }_{}=0`$) the solution of Eq.(10) represents then the spiral wave - the helicon, or the whistler, the frequency of which equals
$$\omega _k^{(h)}=\frac{\omega _0^2\omega _B}{(\omega _p^2+\omega _0^2)},$$
(11)
and tends to zero as $`|\stackrel{}{B}|0`$.
For the case of strong interaction between the particles and at small wavenumbers, i.e. when $`\mathrm{\Omega }_{}>(\omega _B\omega _0)/\omega _p`$, the solutions of Eq.(10) describe the ordinary and extraordinary whistling sound waves propagating in strongly coupled plasmas,
$$\omega _k^{(4,5)}=v_sk\pm \frac{1}{2}\frac{\omega _B\omega _0^2\omega _p^2}{(\omega _p^2+\mathrm{\Omega }_{}^2)^2},$$
(12)
with the whistling sound velocity $`v_s=c\mathrm{\Omega }_{}/\sqrt{\omega _p^2+\mathrm{\Omega }_{}^2}.`$
The parameter $`\mathrm{\Omega }_{}`$ will be estimated in the next section.
## III Estimate of the frequencies $`\mathrm{\Omega }_{}`$ and $`\mathrm{\Omega }_{}`$.
For our purposes it is sufficient to make a simple estimate of the frequencies $`\mathrm{\Omega }_{}`$ and $`\mathrm{\Omega }_{}`$ without including their magnetic field dependence, within the random-phase approximation (RPA), and in the hydrogen plasma model. In this approximation they coincide
$$\mathrm{\Omega }_{}^2=\mathrm{\Omega }_{}^2=h_{ei}\left(0\right)\omega _p^2/3=\omega _p^2_{\stackrel{}{q}0}S_{ei}(q)/3,$$
(13)
and are directly related to the zero separation value of the electron-ion correlation function, $`h_{ei}\left(0\right).`$ The electon-ion structure factor can be estimated in a Coulomb system as :
$$S_{ei}(q)=\frac{4\pi e^2}{n\beta q^2}\frac{\mathrm{\Pi }_e(q)\mathrm{\Pi }_i(q)}{\epsilon (q)}.$$
(14)
Here $`\mathrm{\Pi }_e(q),\mathrm{\Pi }_i(q)`$, and $`\epsilon (q)`$ are the static electronic and ionic polarization operators (real parts), and the dielectric function, respectively, $`\beta ^1`$ is the system temperature in energy units. The ions can be considered as classical particles, and we put $`\mathrm{\Pi }_i(q)=n\beta .`$ For $`\mathrm{\Pi }_e(q)`$ we employ a rational interpolation
$$\mathrm{\Pi }_e(q)=\frac{\gamma ^4/4\pi e^2}{q^2+\gamma ^4\lambda _D^2},$$
(15)
constructed to satisfy both long- and short-wavelength limiting conditions of the RPA: $`\lambda _D^2=4\pi e^2n\beta `$ and $`\gamma ^4=16\pi ne^2m/\mathrm{}^2.`$
After a straightforward calculation we obtained:
$$h_{ei}\left(0\right)=2\alpha r_s\{\genfrac{}{}{0pt}{}{\frac{\sqrt{2}}{\sqrt{B+\sqrt{QS}}+\sqrt{B\sqrt{QS}}},\text{ if }Q>0}{\frac{1}{\sqrt{S}},\text{ if }Q0},$$
(16)
where $`S=B+\sqrt{2A/3}`$, and $`Q=B\sqrt{2A/3}`$, $`A=4\alpha r_s/\pi `$, $`\alpha =\left(4/9\pi \right)^{1/3}=\mathrm{0.\hspace{0.17em}521}`$, and $`B=\left(\pi /3\right)^{1/3}\left(A/4\mathrm{\Gamma }\right)+A/6\mathrm{\Theta }.`$ Usual notations are introduced here: $`k_F`$ is the Fermi wavenumber, $`r_s=a/a_B`$ is the Brueckner parameter, i.e., the Wigner-Seitz distance $`a`$ in the units of the Bohr radius, $`\mathrm{\Gamma }=\beta e^2/a`$, and $`\mathrm{\Theta }=\left(\beta E_F\right)^1`$, $`E_F`$ being the Fermi energy. Notice that $`r_s=\mathrm{\Gamma }\mathrm{\Theta }/0.543.`$
In the case of weakly coupled plasmas with $`\mathrm{\Gamma }0`$,
$$h_{ei}\left(0\right)\mathrm{3.\hspace{0.17em}4}\mathrm{\Gamma }\sqrt{\mathrm{\Theta }}\mathrm{13.\hspace{0.17em}1}\sqrt{\frac{eV}{T}},$$
(17)
with the temperature T measured in units of eV.
Eq.(16) (or Eq.(17) in the limit of weak coupling) together with Eq.(13) determine the magnitudes $`\mathrm{\Omega }_{}`$ and $`\mathrm{\Omega }_{}`$.
## IV Conclusions
In this note the dispersion laws for electromagnetic waves in cold magnetized plasmas are analyzed. Our analysis is based on the expression for the plasma dielectric tensor obtained from the classical theory of moments without using perturbation parameters. Thus both the cases of weak and strong Coulomb coupling are regarded. A qualitative distinction between systems with weak and strong Coulomb coupling is established in their low-frequency electromagnetic wave propagation spectra. It is shown that the weakly coupled plasma helicon branch splits in strongly coupled plasmas into two whistling sound branches. The coupling parameters thermodynamic dependence is estimated.
Acknowledgments. This work was partly financed by the Deutsche Forschungsgemeinschaft and the Polytechnic University of Valencia, Spain.
FIGURE CAPTIONS
Squares of refraction indices of strongly coupled magnetized plasma vs. frequency (in arbitrary units) 1-fast extraordinary wave; 2- ordinary wave; 3 - slow extraordinary wave; 4,5 - strongly coupled plasma whistling sound waves; ($`0<\theta <\pi `$).
Frequencies of various eigenmodes of strongly coupled and ideal magnetized plasma vs. wavevector (in arbitrary units) 1-5 see Fig.1; 3’ - slow extraordinary wave of ideal plasma; 4’ - helicon wave of ideal plasmas ($`0<\theta <\pi `$).
Figure 1. (Electromagnetic modes in cold magnetized strongly coupled plasmas authors: Tkachenko, Ortner, Rylyuk)
Figure 2. (Electromagnetic modes in cold magnetized strongly coupled plasmas authors: Tkachenko, Ortner, Rylyuk)
|
no-problem/9906/astro-ph9906092.html
|
ar5iv
|
text
|
# 1 Introduction
## 1 Introduction
The spectrum of diffuse photons is expected to have a deep of more than two orders of magnitude at energies $`10^{15}10^{17}`$ eV . This deep is similar in nature to the well-known Greisen-Zatsepin-Kuzmin (GZK) cutoff and is caused by electron pair production on the cosmic microwave background, $`\gamma \gamma _be^+e^{}`$. The cross section of the latter process reaches its maximum of $`0.3\sigma _T0.2\text{barn}`$ near the threshold at $`3\times 10^{14}`$ eV and decreases at higher energies (see, e.g., ref. ). The attenuation length of photons in the region of the deep is of order 100 kpc, so the deep in the spectrum is a universal feature of models in which high energy photons have extragalactic origin. Indeed, the existence of the deep is confirmed by simulations in various models of UHE CR (for a review see, e.g., refs.). For instance, in the region of the deep the top-down models typically give the photon flux<sup>3</sup><sup>3</sup>3Throughout this paper the “flux” denotes the quantity which is expressed in terms of th edifferential spectrum $`j(E)`$ by means of the relation $`F=E^2j(E)`$. of order $`10^3`$ eV cm<sup>-2</sup>s<sup>-1</sup>sr<sup>-1</sup>, while at ultra-high energies, $`E>10^{20}`$ eV, the predicted flux is more than two orders of magnitude larger and reaches $`(\text{a few})\times 10^1`$ eV cm<sup>-2</sup>s<sup>-1</sup>sr<sup>-1</sup>.
The spectrum of diffuse photons at energies above $`10^{11}`$ eV is known rather poorly. In the region of the deep the bounds have been obtained by EAS-TOP and CASA-MIA . At $`E10^{16}`$ eV the bound is of order $`0.5`$ eV cm<sup>-2</sup>s<sup>-1</sup>sr<sup>-1</sup>, about two orders of magnitude higher than has been predicted by top-down models. At ultra-high energies the photon flux can be as large as $`1`$ eV cm<sup>-2</sup>s<sup>-1</sup>sr<sup>-1</sup> if the observed UHE CR events are interpreted as photons (recent results from AGASA suggest this possibility ).
While experimentally the existence of the deep in the photon spectrum is an open question, theoretically it is not solid either. Calculations of the spectrum cited above do not take into account the possibility that high energy photons can be generated in our Galaxy by UHE electrons via synchrotron radiation in the Galactic magnetic field. As we argue below, the account for synchrotron emission may substantially change the photon spectrum at $`E10^{15}10^{17}`$ eV filling the deep and bringing the expected photon flux close to the existing experimental limit.
The synchrotron mechanism requires large flux of UHE electrons to hit the Galactic magnetic field. It has been recently pointed out in ref. that this condition is naturally satisfied in the halo models of UHE CR (these models explain UHE CR by decays of heavy relic particles clustered in the Galactic halo ). Due to the fragmentation process, the decay products of the superheavy particles contain a large fraction of UHE electrons.
In this paper we show that UHE electrons which are necessary for the synchrotron mechanism to work can be of extragalactic origin, provided extragalactic magnetic fields are small. We will see that, in fact, the large flux of UHE electrons is inherent in top-down models of UHE CR, so that the generation of high energy photons by the synchrotron mechanism is a generic prediction of top-down scenarios and is not specific to halo models of UHE CR .
The key observation is that UHE photons propagate in the extragalactic space via cascade process being converted to electrons and back with a small energy loss. As a result, the flux of UHE photons is necessarily accompanied by the flux of UHE electrons. At energies of order $`10^{22}10^{23}`$ eV and in the absence of extragalactic magnetic fields the electron flux is at least as large, or even much larger, than the photon one<sup>4</sup><sup>4</sup>4In the presence of extragalactic magnetic field the electrons rapidly loose energy via synchrotron radiation and the argument may not work (see Sect.4 for details).. While UHE photons reach the Earth and contribute to the observable flux of UHE CR, UHE electrons emit synchrotron radiation in the Galactic magnetic field and transfer their energy to high energy photons. As we will see below, the energy of produced photons lies in the region of the deep, while their flux is similar to that of UHE electrons (and, thus, of UHE photons). Hence, in the absence of extragalactic magnetic fields the flux of UHE photons at the level of $`1`$ eV cm<sup>-2</sup>s<sup>-1</sup>sr<sup>-1</sup> implies the flux of synchrotron photons at the same level which fills the deep in the photon spectrum. Since the large flux of UHE photons is one of the signatures of the top-down mechanisms of UHE CR, in the absence of extragalactic magnetic fields these models generically predict no deep in the spectrum of diffuse photons.
The mechanism we propose is sensitive to magnetic fields on distances up to $`50`$ Mpc from our Galaxy. If no deep in the photon spectrum is observed and halo models are ruled out, the extragalactic magnetic field on this distances must be smaller than $`2\times 10^{12}`$ G. This is two orders of magnitude better than the existing bounds . Inversely, if the deep is found and, at the same time, UHE CR have a large fraction of photons, the extragalactic magnetic field must be larger than $`2\times 10^{12}`$ G.
The paper is organized as follows. In Sect.2 we estimate the flux of UHE electrons given the flux of UHE photons and zero extragalactic magnetic field. In Sect.3 we calculate the spectrum of synchrotron radiation in the galactic magnetic field for injected electrons of given energy. In Sect.4 we estimate the effect of extragalactic magnetic fields. Sect.5 contains our conclusions.
## 2 The flux of UHE electrons.
Our aim in this section is to show that in the energy range $`10^{22}10^{23}`$ eV the flux of UHE photons is necessarily accompanied by a comparable or larger flux of UHE electrons, provided the extragalactic magnetic fields are absent. The argument is based on the observation that at these energies the photon propagation is a cascade process (see, e.g., ), i.e., propagating photon is converted to an electron and back with small energy loss. As this process is random, one should expect certain ratio of photons and electrons far from the source.
The main reactions driving the cascade are $`e^+e^{}`$ pair production (PP) on the radio background, $`\gamma \gamma _be^+e^{}`$, double pair production (DPP), $`\gamma \gamma _be^+e^{}e^+e^{}`$, and the inverse Compton scattering (ICS), $`e\gamma _be\gamma `$ . In the energy range of interest double pair production dominates, so one should expect to find more electrons than photons.
A simple estimate can be obtained if one neglects secondary particles and energy losses. In this (rather crude) approximation PP and DPP lead to the conversion of photon to electron with the rates $`a_{PP}`$ and $`a_{DPP}`$, respectively, while ICS converts electron back to photon with the rate $`b`$. The set of equations which describes propagation of photons and electrons reads
$`{\displaystyle \frac{dn_\gamma }{dR}}`$ $`=`$ $`an_\gamma +bn_e,`$
$`{\displaystyle \frac{dn_e}{dR}}`$ $`=`$ $`an_\gamma bn_e,`$
where $`R`$ is the distance from the source, $`n_\gamma (R)`$ and $`n_e(R)`$ are fractions of photons and electrons at the distance $`R`$, respectively, and $`aa_{PP}+a_{DPP}`$. The solution to this system is
$$\frac{n_e}{n_\gamma }=\frac{a\text{e}^{R(a+b)}C}{b\text{e}^{R(a+b)}+C},$$
(1)
where $`C`$ is an integration constant whose value is determined by the ratio $`n_e/n_\gamma `$ at $`R=0`$. Far from the source the value of this constant is irrelevant.
The observed fluxes $`F_{e,\gamma }`$ are given by integrals over the space of $`n_{e,\gamma }`$ multiplied by the particle injection rate. To estimate $`F_e/F_\gamma `$ we note that the integrals are dominated by large distances where the ratio $`n_e/n_\gamma `$ is constant,
$$\frac{n_e}{n_\gamma }\frac{a}{b}.$$
Therefore,
$$\frac{F_e}{F_\gamma }\frac{a}{b}.$$
(2)
Both $`a`$ and $`b`$ depend on energy. At $`E10^{22}`$ eV one has $`a2\times 10^2`$ Mpc<sup>-1</sup>, $`b8\times 10^3`$ Mpc<sup>-1</sup> , and thus
$$\frac{F_e}{F_\gamma }2\text{at}E=10^{22}\text{eV}.$$
(3)
At higher energies the rate $`a`$ is dominated by DPP process and tends to a constant, $`a8\times 10^3`$ Mpc<sup>-1</sup>, while the rate $`b`$ rapidly falls off . At $`E10^{23}`$ eV eq.(2) gives
$$\frac{F_e}{F_\gamma }10\text{at}E=10^{23}\text{eV}.$$
(4)
In a more accurate estimate one should take into account energy losses by leading particles and possible energy sharing in the leading $`e^+e^{}`$-pair in DPP. In this approximation the result depends on the energy distribution of initial particles, as well as on the energy dependence of the rates $`a_{PP}`$, $`a_{DPP}`$ and $`b`$. We have performed such estimate by dividing energy interval $`10^{21}10^{24}`$ eV into 10 energy bands and solving numerically the system of 20 coupled equations analogous to eqs.(1). We have found that the corrections to eqs.(3) and (4) are small and do not change our conclusions, unless the extragalactic magnetic filed is non-zero (the latter case is considered in Sect.4).
## 3 Synchrotron radiation in the galactic magnetic field.
Consider now the synchrotron radiation of UHE electrons in the Galactic magnetic field. An ultra-relativistic particle of energy $`E`$ moving in the magnetic field $`B`$ emits radiation at the characteristic frequency
$$\omega _c=\frac{3\sqrt{\alpha }B}{2m_e^3}E^2=6.7\times 10^{14}\left(\frac{E}{10^{20}\text{eV}}\right)^2\left(\frac{B}{10^6\text{G}}\right)\text{eV}.$$
(5)
The width of the frequency band is roughly $`\delta \omega \omega _c`$. As a result of this process, the particle looses energy at the rate
$$\frac{dE}{dx}=\frac{2\alpha ^2B^2}{3m^4}E^2.$$
(6)
Both equations are written for the case of particle momentum normal to the direction of the magnetic field. Generalization to other cases is straightforward.
Eq.(5) implies that in the Galactic magnetic field, $`B10^6`$ G, electrons with energy $`E10^{20}`$ eV radiate at the characteristic frequency $`\omega _c10^{15}`$ eV. This process is the source of high energy photons in the Galactic halo models of UHE CR . Since the Galactic magnetic field far from the Galactic center is smaller, the extragalactic electrons which we consider in this paper should have higher energy in order to produce synchrotron radiation in the same frequency range.
For the quantitative analysis of the synchrotron emission by the extragalactic electrons consider an UHE electron moving in a varying magnetic field $`B(x)`$ perpendicular to its velocity. Integration of eq.(6) gives
$$\frac{1}{E(x)}\frac{1}{E_0}=\frac{2\alpha ^2}{3m_e^4}_{\mathrm{}}^xB^2(x)𝑑x,$$
(7)
where $`E_0`$ is the initial energy of the electron.
For definiteness, let us take the exponentially decaying magnetic field,
$$B(x)=B_0\mathrm{exp}(x/x_0)$$
(8)
(note that we consider a particle propagating from $`x=\mathrm{}`$). This behavior is expected in some recent Galactic magnetic field models for the field in the direction normal to the Galactic disk. The scale $`x_0`$ is of order 4 kpc. Making use of eqs.(7) and (8) one finds the relation between $`E`$ and $`B`$ at a given point of particle trajectory,
$$\frac{1}{E}\frac{1}{E_0}=\frac{\alpha ^2x_0}{3m_e^4}B^2.$$
This equation, together with eq.(5), determines the dominant radiation frequency as a function of particle energy,
$$\omega _c(E)=\frac{9E_0^{3/2}}{2m_e\sqrt{3\alpha x_0}}f(E/E_0),$$
where
$$f(y)=y^{3/2}\sqrt{1y}.$$
This function is shown in Fig.1. As can be seen from the picture, most part of the electron energy is emitted at frequencies close to
$$\omega _{max}=\omega _c(3E_0/4)=\frac{27E_0^{3/2}}{32m_e\sqrt{\alpha x_0}}=0.8\times 10^{15}\left(\frac{E_0}{10^{22}\text{eV}}\right)^{3/2}\left(\frac{x_0}{4\text{kpc}}\right)^{1/2}\text{eV}$$
(9)
According to eqs.(5) and (9), at $`E=10^{22}`$ eV the electron looses most part of its energy in the region where the magnetic field is $`10^{10}`$ G, i.e., at the distance $`36`$ kpc from the galactic disk for $`B_010^6`$ G.
In the case of magnetic field not perpendicular to particle velocity, the spectrum of synchrotron photons is softer. The same is true for the magnetic field which falls off slower than in eq.(8), as usually assumed for the Galactic magnetic filed in the direction parallel to the Galactic plane. Thus, one should expect angular dependence of photon spectrum with more energetic photons coming from the direction normal to the Galactic plane, and softer spectrum from the directions in the Galactic plane.
Finally, let us estimate the flux of synchrotron photons assuming the flux of UHE photons which is typical for top-down scenarios, $`(\text{a few})\times 10^1`$ eV cm<sup>-2</sup>s<sup>-1</sup>sr<sup>-1</sup> at energies $`10^{22}10^{23}`$ eV. Eq.(3) implies that, outside the Galactic magnetic field, there is at least as large flux of UHE electrons which transfer their energy to high energy photons in the Galactic magnetic field. Since the synchrotron spectrum has $`\delta \omega \omega `$, the energy conservation implies that the flux of synchrotron photons is approximately the same as the flux of UHE electrons, which is larger by the factor $`F_e/F_\gamma `$ than the flux of UHE photons.
## 4 Effect of extragalactic magnetic field
As it was shown above, in the absence of extragalactic magnetic fields the observed flux of $`10^{15}10^{17}`$ eV photons is proportional to the flux of UHE photons and the ratio $`F_e/F_\gamma `$ at energies above $`E>10^{22}`$ eV. The presence of large enough extragalactic magnetic field can significantly decrease this ratio. Indeed, if $`\gamma e`$ conversion length, $`a^1=(a_{PP}+a_{DPP})^1`$, is large compared to the energy loss length of the electron due to the synchrotron radiation in the extragalactic magnetic field, the flux of UHE electrons should be much smaller than the flux of UHE photons.
Let us estimate the value of the extragalactic magnetic field at which the ratio $`F_e/F_\gamma 1`$ at $`E>10^{22}`$ eV. For this purpose note that the solution to eq.(6) in the constant magnetic field $`B`$ can be written in the form
$$l=\frac{3m^4}{2\alpha ^2B^2E}\left(1\frac{E}{E_0}\right),$$
(10)
where $`l`$ is the distance passed by the electron while its energy decreases from $`E_0`$ to $`E`$. Eq.(10) implies that electrons with energy $`E`$ can only come from distances smaller than
$$l_E=\frac{3m^4}{2\alpha ^2B^2E}50\left(\frac{2\times 10^{12}\text{G}}{B}\right)^2\left(\frac{10^{22}\text{eV}}{E}\right)\text{Mpc}.$$
(11)
Taking into account that the length $`a^1`$ of $`e\gamma `$ conversion is of order 50 Mpc at $`E=10^{22}`$ eV one finds that the extragalactic magnetic field should be smaller than $`2\times 10^{12}`$ G at the distances $`50`$ Mpc from our Galaxy in order that the ratio $`F_e/F_\gamma `$ to be comparable or larger than one. If the magnetic field at distances of order 50 Mpc is noticeably larger than $`2\times 10^{12}`$ G, the ratio $`F_e/F_\gamma `$ is much smaller than one and the flux of UHE electrons is not sufficient to fill the deep in photon spectrum by synchrotron mechanism (for the discussion of present limits on the extragalactic magnetic field see, e.g., ref. ).
It is worth noting that the mechanism we propose is not sensitive to magnetic fields at distances larger than $`50`$ Mpc because this distance is sufficient for generation of a large fraction of electrons, $`F_e/F_\gamma 1`$. The effect of distant magnetic fields is mere decreasing of the UHE photon flux, which is not important for our argument since we normalize UHE photon flux to the observed flux of UHE CR.
## 5 Conclusions
To summarize, we have shown that the expected deep in the photon spectrum at energies $`10^{15}10^{17}`$ eV may be absent due to the synchrotron radiation of UHE CR in the Galactic magnetic field. The mechanism we propose involves extragalactic UHE electrons produced during cascade propagation of UHE photons and thus requires small extragalactic magnetic fields and large fraction of photons in UHE CR at energies $`10^{22}10^{23}`$ eV. The latter is the characteristic feature of all top-down models of UHE CR. Note, however, that our mechanism does not rely on any peculiarities of these models.
When synchrotron radiation in the Galactic magnetic field is taken into account, the top-down models predict the flux of $`10^{15}10^{17}`$ eV photons which is close to the current experimental limits, provided that extragalactic magnetic fields are small. Thus, it is important to improve the sensitivity of the experiments in the energy range $`10^{15}10^{17}`$ eV. The detection of the diffuse photon flux at the level of $`10^1`$ eV cm<sup>-2</sup>s<sup>-1</sup>sr<sup>-1</sup> would strongly suggest that UHE CR are produced by a top-down mechanism. Moreover, this would imply that either the mechanism based on the halo model or the one proposed here works. The two possibilities can be distinguished by measuring the angular anisotropy of UHE CR and of produced high energy photons . Additional signature of our mechanism is the dependence of the photon spectrum on the arrival direction as discussed in Sect.3. On the contrary, if the photon flux in the region of the deep is smaller than $`10^3`$ eV cm<sup>-2</sup>s<sup>-1</sup>sr<sup>-1</sup> and, at the same time, UHE CR have a large fraction of photons, the extragalactic magnetic field must be larger than $`2\times 10^{12}`$ G.
## Acknowledgments
The authors are indebted to D.S. Gorbunov, O.E. Kalashev, M.V. Libanov, V.A. Rubakov and D.V. Semikoz for useful discussions and comments. The work of S.D. is supported in part by Russian Foundation for Basic Research under grant 99-02-18410, INTAS grant 96-0457 within the research program of the International Center for Fundamental Physics in Moscow, and ISSEP fellowship.
|
no-problem/9906/quant-ph9906116.html
|
ar5iv
|
text
|
# How to realize quantum superluminal communication?
## I Introduction
After having shown quantum superluminal communication must exist in our world, we will further study the possibility to realize such superluminal communication, this is undoubtedly a formidable task within the scope of our present knowledge about Nature, since the two foundation stones of modern physics—special relativity and quantum mechanics all reject superluminal communication, thus in order to achieve quantum superluminal communication, we must first revise them.
## II Revising special relativity
In order to admit quantum superluminal communication, special relativity needs to be changed only a little, and this does not limit its applicability in its previous territory at all, in fact, we only need to limit the scope of ”natural phenomena” in the principle of relativity, which is the first assumption of special relativity, namely the natural phenomena satisfying the principle of relativity will no longer involve all natural phenomena, and the quantum nonlocal influence is just such an exception, we call the new principle quantum relativity principle.
In fact, there exists nothing compelling in both theoretical considerations and experimental confirmations to require the validity of the principle of relativity for all natural phenomena, just as Einstein, the founder of special relativity, demonstrated himself, ”in view of the more recent development of electrodynamics and optics, it became more and more evident that classical mechanics affords an insufficient foundation for the physical description of all natural phenomena. At this juncture the question of the validity of the principle of relativity became ripe for discussion, and it did not appear impossible that the answer to this question might be in the negative.” Indeed, his worries become true when considering the quantum nonlocal influence in quantum theory.
On the other hand, although we have demonstrated that the quantum nonlocal influence rejects the relativity principle owing to the resulting causal loop, the deeper reasons need to be given, as Einstein denoted, the relativity principle originates from classical mechanics, the precondition of its validity will be ”all natural phenomena were capable of representation with the help of classical mechanics”, while quantum phenomena are evidently not such natural phenomena, and classical mechanics can no longer affords an sufficient foundation for the physical description of such phenomena either, concretely speaking, the relativity principle will hold true for the continuous motion of the real objects including particles and fields, while for the quantum nonlocal influence in quantum space-like measurement, no real objects are transmitted in the process, and the process is also essentially discontinuous, thus there does not exist any real objects in continuous motion for the principle to apply in such process, and the earth beneath the feet indeed disappears, then it is by no means a surprising fact that the quantum nonlocal influence does reject the relativity principle.
Indeed, if one principle is valid for all natural phenomena, it will be too absolute to be true, the original relativity principle is just such a principle; on the other hand, its founder Einstein also ignored one subtle possibility, namely the invalidity of the relativity principle for some natural phenomena, say the quantum nonlocal influence, will not influence its validity for other natural phenomena, say classical phenomena.
At last, even though special relativity is revised so as to permit the existence of quantum superluminal communication, it provides nothing helpful for realizing such superluminal communication, since the origin lies in the quantum nonlocal influence itself, thus we must turn to the quantum theory.
## III Revising quantum mechanics
According to the demonstrations about the existence of quantum superluminal communication, we know that in order to find the preferred Lorentz frame, which existence is predicted in theory, the distinguishability of nonorthogonal single states is required, but present quantum theory uncompromisingly rejects this requirement, thus we need to revise present quantum theory.
First, as to the normal linear evolution equation of the wave function in present quantum theory, even though it has been confirmed very precisely, we can not exclude the possibility of its inaccuracy yet, and there may exist some kind of deterministic nonlinear correction, which will result in the different predictions from present quantum mechanics, then quantum superluminal communication can be achieved in such revised quantum theory.
Secondly, as to the evolution of the wave function during quantum measurement, present quantum theory is by no means a complete theory, and the projection postulate, from which the quantum nonlocal influence appears, is just a makeshift, while the concrete dynamical process of the projection, where the availability of quantum superluminal communication may hide, is undoubtedly one of the most important unsettled problems in quantum theory, and the resulting revised theories are deeply studied recently, in which the linear evolution equation of the wave function is replaced by stochastic linear or nonlinear equation; on the other hand, there exists no essential reason to prevent the revised quantum theory including dynamical collapse process from permitting the distinguishability of nonorthogonal single states, on the contrary, this kind of distinguishability will disclose the mysterious veiling of the wave function, say endow with reality to it, and eventually settle the notorious interpretation problem of quantum mechanics.
## IV How to realize quantum superluminal communication?
Since no essential reason and experimental evidence can be found to revise the normal linear evolution equation of the wave function by some deterministic nonlinear correction, we will mainly analyze the possibility to achieve quantum superluminal communication by use of the revised quantum theory including dynamical collapse process.
First, even though present quantum theory, especially its projection postulate, is surely incomplete, but in case of its correctness it still imposes strong limitations for the availability of quantum superluminal communication, concretely speaking, once the quantum measurement process is completed, projection postulate will take effect, then little room is left for the possibility of quantum superluminal communication, thus the opportunity only exists in the quantum measurement process.
Secondly, since for any quantum measurement the measurement results undoubtedly need to be obtained by the observer, say our human being, or other conscious object, who wants to carry out such measurement, then the above opportunity requires that the state of the conscious object must be entangled with the measured state before the completion of the quantum measurement process, two essential advantages to resort to conscious are that, on the one hand, the conscious object will be the last identifier in the measurement process and its influence is inescapable before the dynamical collapse process is completed, on the other hand, only the conscious object may identify the intermediate process and state before the dynamical collapse process is completed owing to his self-conscious, and takes a different action from the action corresponding to one of the states in the superposition state.
Since the distinguishability of nonorthogonal single states is required to achieve quantum superluminal communication, here we assume what need to be measured and differentiated are the following nonorthogonal single states $`\psi _1+\psi _2`$ and $`\psi _1`$, and the entangled state of the whole system including measured system, measuring device and conscious object is respectively $`\psi _1\phi _1\chi _1+\psi _2\phi _2\chi _2`$and $`\psi _1\phi _1\chi _1`$, where $`\phi _1`$ and $`\phi _2`$ are the states of the measuring device, $`\chi _1`$ and $`\chi _2`$ are the conscious state of the conscious object, then it is evident that, in order to distinguish the above nonorthogonal single states, the only reasonable condition is the conscious time for the definite state $`\psi _1\phi _1\chi _1`$ is shorter than the collapse time for the superposition state $`\psi _1\phi _1\chi _1+\psi _2\phi _2\chi _2`$, and the time difference is long enough for the conscious object to identify<sup>*</sup><sup>*</sup>*This condition may result in the conclusion that conscious can not be explained by present ( quantum ) physical theory..
In the following, we will mainly demonstrate that this condition is not irrational, and can be satisfied in essence, first, we can assume the conscious time of the conscious object in Nature is generally independent of the collapse process of the observed state, since during the formation and evolution of conscious the input states will be always classical definite states coming from the outside classical world, the conscious object can only be trained to adapt to these classical definite states, and there exists no collapse process for the observed classical state at all, furthermore, with the natural selection the conscious time will turn shorter and shorter, while the universal collapse time formula is not changed, then it is reasonable that for some kind of conscious object the conscious time for the definite state $`\psi _1\phi _1\chi _1`$ is shorter than the collapse time of the superposition state $`\psi _1\phi _1\chi _1+\psi _2\phi _2\chi _2`$, and the time difference is long enough for the conscious object to identify, thus even if our human being can not satisfy this condition, other conscious objects may satisfy this condition.
On the other hand, if the above condition can not be satisfied in essence, then we must accept the following bizarre conclusion, namely the concrete collapse theory will limit both the mass, size of the conscious part and conscious time of any conscious object, this means that in case of a certain conscious time the mass and size of the conscious part of any conscious object can not be smaller than the minimal finite values, and in case of a certain mass and size the conscious time of any conscious object can not be shorter than the corresponding collapse time, these requirements are evidently irrational, since the above properties of any conscious object all originate from the natural selection in the environment of classical world, and even if the natural selection relating to the collapse process does take effect, then it is also more reasonable for the conscious object to be able to use quantum superluminal communication, since undoubtedly it will be useful for his existence and evolution.
## V A feasible experiment
In this section, we will present a feasible experiment to confirm the above possibility of quantum superluminal communication.
The experiment is based on the fact that the visual perceptual apparatus of many creatures including our human being possesses an extreme sensitivity, and as we know, for hoptoad only one photon can trigger a definite visual perception, for our human being the number is about seven.
We first consider a simple case in which a bunch of 10 photons coming from a region A propagates towards the eye of a human observer, which has been analyzed by Ghirardi from a different point of view, the photons hit the retina of the observer and trigger the definite perception “a luminous spot at A”; in a similar way, the photons coming from a region B, spatially separated and perceptively distinct from A, will trigger the definite perception of the observer “a luminous spot at B”.
Now, we consider a different situation, in which a superposition state of the above two states of the photons is prepared and kept long enough for the experimental aim using present quantum optics technology, and again the photons propagate towards the eye of the same human observer, hit his retina and trigger his visual perception, then according to the above analysis the observer may have a different perception from any one of the above perceptions.
Certainly, due to the brevity of the perceptual process, it may be very difficult for the observe to perceive the difference of the above two situations, but as we think, it is by no means impossible some conscious object in nature.
On the other hand, the human observer may be replaced by a trained hoptoad, and the above experiment may contain only one photon, which can be easily realized presently, then we can observe the different reactions of the hoptoad for the above two situations, according to our analysis, for the superposition state situation the hoptoad may take a different reaction from any one of the reactions corresponding to the definite states in the superposition state.
## VI Conclusions
We show that special relativity and present quantum theory can be revised to permit quantum superluminal communication, and when using the conscious object as one part of the measuring device, quantum superluminal communication may be a natural thing.
Acknowledgments
Thanks for helpful discussions with Dr S.X.Yu ( Institute Of Theoretical Physics, Academia Sinica ).
|
no-problem/9906/math-ph9906017.html
|
ar5iv
|
text
|
# Comment on ‘Dimensional expansion for the delta-function potential’
## References
|
no-problem/9906/astro-ph9906059.html
|
ar5iv
|
text
|
# THE EXTENDED NARROW LINE REGION OF 3C 299Based on observations made with the NASA/ESA Hubble Space Telescope, which is operated by the Association of Universities for Research in Astronomy, Inc, under NASA contract NAS 5-26555
## 1 Introduction
Over the last few years we have undertaken a systematic survey of extragalactic radio sources (de Koff et al. 1996; Martel et al. 1998a,b ; McCarthy et al. 1997) in the 3CR Catalogue (Bennett 1962a,b; Spinrad et al. 1985) using WFPC2 on HST. These data allow us to investigate the relationships between the radio and optical morphologies in a complete sample of powerful radio galaxies. Imaging of nearby Seyfert galaxies with HST, such as Mrk 3, Mrk 6, Mrk 573, NGC 1068, NGC 2992 and NGC 4151, (Capetti et al., 1995a, 1995b, 1996, 1997; Winge et al. 1997; Allen et al. 1998,1999; Axon et al, 1998) has shown an intimate connection between the radio structure and the extended NLR (ENLR). These studies show that the interaction of the radio jet with the ISM is the main source of UV photons that ionize the ENLR in Seyfert galaxies. We wish to investigate whether this scenario is also applicable to the high-power radio sources and in particular the radio galaxy 3C 299.
This galaxy is a FR II and one of the most asymmetric double-lobed radio AGNs (Liu et al. 1992) not only because of the significantly different distance between the nucleus and each of the lobes but also in terms of lobe flux and lobe length. The radio morphology consists of a compact radio-core and two lobes separated by 11<sup>′′</sup> at PA $``$ 63<sup>o</sup>. The NE lobe is larger ($``$ 1.5<sup>′′</sup>) and brighter (441.6 mJy at 8.4 Ghz, Akujor et al. 1995) than the SW lobe ($``$ 0.7<sup>′′</sup> and 20.1 mJy). The core does not appear to be located at the geometrical center of the lobes; it is 3<sup>′′</sup> nearer the NE lobe (Liu et al 1991). For some time, due to this large asymmetry, the NE lobe itself was believed to be the central component of a compact steep-spectrum source (CSS). But when the SW component was found by Laing (1981), and the central component by Liu & Pooley (1991) and van Breugel et al (1992), the true picture of 3C 299 as a double-lobed radio AGN fully emerged.
At optical wavelengths this galaxy appears to be a low-redshift analogue of the high redshift (z$`>`$1) radio galaxies studied by McCarthy et al. (1995) since its optical continuum emission is elongated and aligned with the radio source axis. The emission lines have a complex morphology and kinematics, with the \[OIII\]$`\lambda `$5007 line offset from the continuum peak (McCarthy et al. 1996).
In this paper we discuss the complex morphology of 3C 299 shown by the high spatial HST resolution images and we compare it with the radio data. We investigate the physical conditions of the ENLR through ground-based long-slit spectroscopy data and show that the ionization parameter is roughly constant throughout the ENLR.
We conclude by showing that a simple mechanism to provide the required ionizing photons are shocks triggered by the interaction of the radio jet with the local ISM (Capetti et al. 1995a, 1995b), as in the case of the lower redshift Seyfert galaxies.
## 2 Observations and data analysis
The HST/WFPC2 observations of 3C 299 were taken as part of the 3CR Snapshot Survey (PI: W. B. Sparks) which was conducted in cycles 4 to 7, in the F555W and F702W broad bands as well as narrow emission line bands. (de Koff et al. 1996; Martel et al. 1998a,b; McCarthy et al. 1997). The 3C 299 images were obtained with the WFPC2/PC and the F702W filter in May 1995 and the F555W filter in September 1996. Narrow-band images with the WFPC2/WF2 and the FR680N ramp filter were taken in August 1995. Table 1 shows the observation log.
At a redshift of z=0.367 , the galaxy is at a distance of 1.3 Gpc with a projected linear scale of 4.7 kpcs/arcsec (assuming $`H_0`$=65 km sec<sup>-1</sup> Mpc<sup>-1</sup> and $`q_0=0.5`$). The WFPC2/PC scale is 0<sup>′′</sup>.0455/pixel and for the WFPC2/WF2 the scale is 0<sup>′′</sup>.0996/pixel, and therefore, the physical scales for the images are 213 parsecs/pixel for the PC mode and 471 parsecs/pixel for the WF2 mode.
The reduction procedure for the F702W filter data, which is close to Cousins R filter, is fully discussed in Martel et al. (1998b). This reduction includes the standard WFPC2 pipeline processing and cosmic-ray removal. The data reduction was carried out using IRAF and the STSDAS package. At the redshift of 3C 299, the F702W data includes the H$`\beta `$ and \[OIII\] emission lines. The F555W filter data, which is closer to Johnson V, was reduced in the same way and then registered and added to produce one final image with higher S/N. Due to the redshift the F555W image includes the flux from the emission lines of \[OII\], \[NeIII\], H$`\delta `$ and H$`\gamma `$. In what follows we will define F702W as R, and F555W as V. The FR680N is a narrow-band ramp filter, which given the redshift of the galaxy and its position on the CCD, produces an \[OIII\] $`\lambda `$5007 image. These images were reduced as described above and the final image was re-scaled to the resolution of the WFPC2/PC.
The F702W and F555W filter data were flux-calibrated using values for the inverse sensitivities (PHOTFLAM) of 1.866 x 10<sup>-18</sup> ergs cm<sup>-2</sup> Å<sup>-1</sup> DN<sup>-1</sup>, 3.491410 x 10<sup>-18</sup> cm<sup>-2</sup> Å<sup>-1</sup> DN<sup>-1</sup>, and zero-points for the Vega system of of M<sub>Photzpt</sub>=-21.85, M<sub>Photzpt</sub>=-21.07 respectively. Non-rotated images were used for these flux calibrations. The FR680N filter was calibrated using the ramp filter calculator (Biretta et al, 1996). An approximate coordinate frame for the WFPC2 data is provided by the image header information, based on the HST guide stars. The accuracy at the coordinates is approximately 1<sup>′′</sup> (Biretta et al. 1996).
Images obtained with different filters were registered to a common frame using cross-correlation of the brighter elliptical structures. To derive the spatial shift between the broad-band filters images and the FR680N image, we used the filamentary NE structure, since this structure is well defined and bright in both filters. The sky background in each image was determined through the statistical analysis of a 3-sigma clipping of the average background value in several regions in each of the images.
To compare with radio data, we used the observations from Leahy (1997), obtained with MERLIN at 1534 MHz and posted on the DRAGN’s Web site maintained by Leahy, Bridle, & Strom R., at the Jodrell Bank Observatory.
## 3 Results
The optical broad-band images possess a complex morphology (Fig. 1). The V and R images show a central elliptical structure, with two bright cusps separated by an absorption lane, $``$ 0<sup>′′</sup>.3 x 1<sup>′′</sup>.4 (i.e. 1.2 x 5.7 kpcs). To the NE, separated by 1.4 arcsec, there is an extended filamentary structure. This region shows a shell-like structure with bright, sharp but resolved filaments enclosing what appears to be a bubble of lower intensity emission.
We found that the elliptical is well fitted by an $`r^{1/4}`$ surface brightness profile. From the isophotal fitting, the position of the nucleus is found to be hidden behind the central absorption lane. From the V-R map, we derive values for the color index. The elliptical galaxy has a color index of 1.0 $`<`$ (V-R) $`<`$ 1.33, which is compatible with the values from Rönnbach et al. (1996) for the stellar population of a elliptical galaxy at the redshift of 3C 299. The faint region between the brighter cusps has 1.53 $`<`$ (V-R) $`<`$ 2.0 and these values are also compatible with a stellar population reddened by a dust lane.
The FR680N filter shows a rather different picture, since the data is dominated by emission in the \[OIII\]$`\lambda `$5007 line. The image (Fig. 2) shows a bi-conical structure of ionized oxygen, bright and large to the NE and faint and small to the SW. The NE cone has a full opening angle of 45 degrees, from PA 30<sup>o</sup> to PA 75<sup>o</sup>. The geometry and the brightness intensity (in comparison with the broad-band filters data) of this structure in the \[OIII\]$`\lambda `$5007 emission-line data indicates that this flux came from the extended NLR (ENLR). This is consistent with the previous studies of McCarthy et al. (1995, 1996). The shape of the NE bright lobe in this data is the same as that in the broad-band filters. Since there is almost no continuum emission (as the elliptical galaxy profile vanished) in the NE bright lobe, the broad-band filter morphology is dominated by the line-emission, therefore we used it to register the broad-band data to the narrow-band data.
Using a bi-dimensional cross-correlation technique (CROSSCOR task in IRAF/STSDAS) we find that the center of the bi-conical structure coincides with the geometric center of the elliptical galaxy. Fig 3 shows the superposition of the \[OIII\] data over the R data; note that the NE structure coincides with the line emission. If we make the obvious assumption that the center (nucleus) of the elliptical galaxy is coincident with the radio core, it follows that the radio axis (PA 63<sup>o</sup>) is reasonably well aligned with the emission-line region (PA 52<sup>o</sup>). Furthermore the NE radio lobe is clearly related to the ENLR in the \[OIII\] image and with the large filamentary shell-like structure that is seen in the broad-band filters (Fig. 4a,b).
We derived values for the physical conditions of the emitting gas by analyzing and measuring line-emission ratios of the optical spectrum of McCarthy et. al. (1996), taken with the Lick 3 m telescope. The 2<sup>′′</sup> slit was positioned along the radio source axis (P.A. $``$ 70<sup>o</sup>), thus crossing the ENLR and covering a wavelength range from 4600 to 7400 Å. The McCarthy et al. data shows that the peak of the \[OIII\] and H$`\beta `$ lines are offset from the continuum (see Fig 5 for a plot of this spectrum in the range from 6300 to 7400 Å). Therefore the maximum of the line emission lies outside the nucleus and towards the ENLR. The high excitation line HeII $`\lambda `$ 4686 is present and shows a similar behavior with extensions towards the ENLR.
One interesting point is to understand what dominates the flux in the broad-band images at the ENLR. For the F555W data the \[OII\]$`\lambda `$3727 and the \[NeIII\]$`\lambda `$3869 are present, and there is no observable continuum in the ENLR spectrum. Since the \[NeIII\]$`\lambda `$3869 emission is low ($`<`$30%) compared to the \[OII\]$`\lambda `$3727 and also the \[OII\]$`\lambda `$3727/\[NeIII\]$`\lambda `$3869 ratio is quite constant over the entire ENLR, it is possible to use the broad-band F555W image as a good estimator of the \[OII\]$`\lambda `$3727 emission. We used this image, to derive surface brightness profiles and to determine the behavior of the \[OIII\]/\[OII\] ratio as an estimate of the spatial behavior of the ionizing parameter ($`U`$). In Fig. 6 we show the \[OIII\] and \[OII\] intensities as a function of distance from the nucleus along the NE radio jet, as well as a plot of the \[OIII\]/\[OII\] ratio. The \[OII\] flux profile includes the contribution from the stellar population of the elliptical galaxy to a distance of about 1 arcsec from the nucleus, where the $`r^{1/4}`$ profile vanishes. Therefore, from 1<sup>′′</sup> to 3<sup>′′</sup>.2, the measured flux arises only from ENLR emission lines.
The low amount of contamination of \[NeIII\] emission in the \[OII\] data does not affect the validity of the \[OIII\]/\[OII\] ratio. Models such of Dopita el al. 1995, show that the \[NeIII\] emission follows the trend of the \[OIII\] emission as the effect of the precursor become more important. Sophisticated photoionization models such as those at Binnette et al. (1996) show a relation between the \[NeIII\] and \[OIII\] emission. Therefore, the properties of the \[OIII\]/\[OII\] as the tracer of the ionization parameter $`U`$ are not affected by the low amount of \[NeIII\] emission.
The typical error for the \[OIII\] plot is $`\sigma =\pm `$ 0.7 for a reference flux of 4 $`DN`$ calculated with the Extend Exposure Time Calculator (Biretta et al. 1996) and taking into account two pixel binning. For the \[OII\] plot, $`\sigma =\pm `$ 0.3 for a 2 $`DN`$ flux and 8 pixel binning to match the spatial resolution of the upper plot (WF to PC resolution). The errors of the \[OIII\]/\[OII\] plot are a combination of the errors of the two upper plots and for example, a reference flux of 5 $`DN`$ has a maximum error of $`\sigma =\pm `$ 0.9, calculated as the large departure from the combination of fluxes and errors of the \[OIII\] and \[OII\] profiles.
## 4 Discussion
Our optical data shows that there is a close morphological relationship between the \[OIII\] ENLR and the radio emission of the NE lobe. Is is interesting to investigate the possible physical scenario for this close relationship. Taylor et al. (1992) proposed that fast bowshocks resulting from the interaction of the radio jet and the ISM were the source of ionizing photons of the emission-line gas. Capetti et al. (1995a, 1995b, 1996, 1997) and Winge et al. (1997) were the first to show that this mechanism best explains the optical emission in the NLR in nearby Seyfert galaxies (Mrk 3, Mrk 6, Mrk 573, NGC 1068, NGC 4151 and NGC 7319). Recently, this work has been confirmed by Aoki et al. (1999) and Kukula et al. (1999).
Dopita & Sutherland (1995a,b) have modeled in detail the ionization of the ENLR by shocks. In one scenario which has been shown to work for Seyfert galaxies, the radio jet interacts with the local interstellar medium and shocks the gas. In this scenario, the hot postshock plasma gas produces photons that can diffuse upstream and downstream of the jet. Photons diffusing upstream can encounter the preshocked gas and produce an extensive precursor HII region, while those traveling downstream will influence the ionization and temperature structure of the recombination of the shock.
This scenario appears very attractive in explaining the morphology of the radio and optical data in 3C 299. To check the validity of this interpretation, we derived values for the physical conditions of the emitting gas by measuring and analyzing the line-emission ratios from the optical spectrum of McCarthy et. al. (1996). In Table 2 we list the spatial behavior of the He II $`\lambda `$ 4686/H$`\beta `$ and \[OIII\] $`\lambda `$5007/H$`\beta `$ ratios derived from that spectrum. Figure 7 is the plot of the measure line-ratio at 3.6<sup>′′</sup> (from the nucleus) over the models of Dopita et al. (1995a), it is clear that the model which includes a high-speed shock ($``$ 400 km/sec) plus a precursor HII region photoionization (produced by photons traveling upstream from the shock) fits the observations rather well over the entire ENLR. The values of Table 2 are also fairly constant (within the errors) showing that physical conditions are similar over the jet path.
Another way to check the shock scenario is to explore the spatial behavior of the ionization condition of the gas by estimating the ionization parameter $`U`$. This parameter is defined as $`U=Q/4\pi r^2cN_e`$, where Q is the rate at which ionizing photons are emitted from a point source towards a gas cloud with electron density $`N_e`$ and distance $`r`$ from the ionizing source. If we consider the galactic nucleus to be the ionizing source, as the distance from the nucleus increases for a constant $`N_e`$, the parameter U will decrease and change the ionization condition of the gas. This change will be reflected in the values of the emission-line ratios (eg. \[OIII\]/\[OII\]). However, this parameter will have a very different behavior if the density changes or if there is another source of ionizing photons (such as shocks produced by the radio jet) or a combination of both (from the nucleus and local).
The two brighter peaks in the flux profiles of both \[OIII\] and \[OII\], at 1.7<sup>′′</sup> and 2.4<sup>′′</sup> from the nucleus arise from the bright filaments of the shell-like structure. Note that both profiles have a similar shape. The \[OIII\]/\[OII\] line ratio is rather high from 1.6<sup>′′</sup> to 2.6<sup>′′</sup> at the position of these filaments. As noted above this line ratio is strongly dependent on the ionization parameter $`U`$. If there is no local source of ionizing photons, due to geometrical dilution of the ionizing radiation field we expect $`UN_e^1r^2`$. To produce a constant value for $`U`$ the electron density should also follow an $`r^2`$ law.
Wellman et al. (1996) using radio data for 14 radio-galaxies and 8 radio-loud quasars obtained an estimate of the lobe magnetic field and the lobe speed propagation. From these parameters, they derived an estimate for the ambient gas density in the vicinity of the radio lobe, for the galaxies in their sample using the equation for ram pressure confinement. They found that the ambient gas density can be fitted with a modified King density profile. In particular in our region of interest, within the inner 11 $`h^1`$ kpc, the density still falls in the flat part of the density distribution, as the core radius of this King profile is 50 (or 100 depending on the model) $`h^1`$ kpc. From these, it seems extremely unlikely that the ambient density can vary as $`r^2`$, as would be required to keep the ionization parameter constant in the case that the only source of ionizing photons is the nuclear source. We therefore exclude that ionization by the nuclear source alone can be responsible for the observed ionization of the ENLR lobe.
Similar physical properties of the ENLR, namely a rather constant value for $`U`$, has been reported for some nearby Seyfert 2 galaxies. Capetti et al (1995a), Kukula et al. (1999) show a similar result for Mrk 3, where the radio jets and the optical emissions are very closely associated. In Mrk 6, Capetti et al. (1995b) show that there is also evidence for a transverse ionization structure to the south where the radio jet and the emission lines are co-spatial. In the case of Mrk 573 (Capetti et al. 1996), the ratio \[OIII\]/\[OII\] indicates that the ionization parameter actually increases with radius. These examples provide strong evidence for the scenario where line-emitting gas in the ENLR is compressed by shocks and causes the line emission to be highly enhanced in the region in which this interaction occurs. We can also compare 3C 299 to NGC 1068, one of the closest Seyfert 2 galaxies with a bright radio source and a prominent radio jet which terminates in a extended radio-lobe. For NGC 1068, Capetti et al. (1997) found that the geometrical dilution model of the nuclear radiation field cannot explain the high excitation core formed by the brightest knots of the NLR. An increase (3-4 higher) of the ionization parameter is also observed at a distance of $``$4<sup>′′</sup> from the nucleus, where the radio outflow changes its morphology from jet-like to lobe-like. This could be easily explained if there is a density increase where the jet enters the lobe, presumably due to the compression of the backflowing radio cocoon at the jet working surface.
In the case of 3C 299, morphology arguments i.e. the extent and filamentary nature of the ENLR, the structure of the ENLR and its association to the radio-lobe, as well as physical arguments, i.e. the value of the emission-line ratios HeII$`\lambda `$4686/H$`\beta `$ and \[OIII\]$`\lambda `$5007/H$`\beta `$ compared to the models of Dopita et al. (1995a) and the constant value for the ionization parameter $`U`$, which does not follow the dilution of the ionizing radiation field for a central source of UV photons, strongly support the same scenario, namely that the ENLR is a product of the interaction of the radio-lobe with the ISM gas.
In contrast to the nearby galaxies, 3C 299 has an intermediate redshift (z=0.367), and its ENLR is considerably larger ($``$ 12 Kpcs). Best et al. (1996,1998,1999), have carried out deep spectroscopic observations of a number of powerful radio galaxies with redshift $`z1`$ and show that the passage of the radio jet induces shocks through the host galaxy and plays a key role in producing the emission-line gas properties of these objects. Thus the physical connection between the radio jet and the ENLR appears to be at work not only for nearby galaxies, but also for those at higher redshifts. This could provide an explanation for the “alignment effect” found by McCarthy et al. (1987), for high-redshift radio galaxies.
## 5 Summary
We have shown that the elliptical galaxy 3C 299, has a prominent dust lane and bi-conical extended emission line morphology that is similar to those of some lower redshift Seyfert galaxies.
We found that the NE radio radio-lobe lies within the shell-like structure of the ENLR, suggesting a physical connection between the jet and the ENLR. We tested the scenario where the radio jet compresses and shocks the interstellar gas thus producing the observed morphology. From the spectra of McCarthy et al. (1996) we compared the line ratio behavior of the ENLR with the models of Dopita et al. (1995a). From these models we found that the observed spectrum is compatible with gas shocked by the radio jet plus ionization from a precursor HII region, produced by the ionizing photons on the preshocked gas. We also investigated the spatial behavior of the ionizing parameter $`U`$, by determining the \[OIII\]/\[OII\] line-ratio which is sensitive to the change of the ionization parameter, and we traced its behavior over the ENLR along the radio jet direction. We found that the \[OIII\]/\[OII\] ratio does not follow a simple dilution model, but rather that it is approximately constant over a large area (3 <sup>′′</sup> from the nucleus, $``$ 12 kpcs ) thus requiring a local source of ionization, and we found that shock models driven by the radio jet, provide the necessary ionizing flux.
Therefore, because of morphology arguments (the NE lobe is locate at the ENLR), physical arguments (the emission-line ratios are consistent with a shock plus an HII precursor region) and the constant value of the ionization parameter (which does not follow the dilution of the ionizing radiation field for a central source of photons), we conclude that the ENLR of 3C 299 is the result of the interaction of the radio jet with the ISM gas.
C.F. acknowledges the support from the STScI visitor program. The Fundacion Antorchas (Argentina) provided partial financial support of this work. We are very grateful to Mark G. Allen for providing us the emission-line diagnostic models for a shock plus an HII region precursor.
### Figure Captions
### Fig 1a
F702W (R) image of 3C 299. N is at the top, E is to the left.
### Fig 1b
F555W (V) image of 3C 299.
### Fig 2
\[OIII\]$`\lambda `$5007 image. The extended-narrow line region shows the bi-conical structure. It is centered on the galaxy nucleus and is bright and extended to the NE.
### Fig 3
Overlap between the \[OIII\] and the R filter images. Note that the NE structure consists mostly of line emission.
### Fig 4a
Overlap of the radio map at 1450 MHz with the R image.
### Fig 4b
Overlap of the radio map at 1450 MHz with the \[OIII\] image.
### Fig 5
Long-slit spectrum taken by McCarthy et al. (1996). Note that the peaks of the emission lines are offset from the continuum. Wavelength is redshift corrected.
### Fig 6
Flux profiles ($`DN`$ units) of \[OIII\], \[OII\] and \[OIII\]/\[OII\], as a function of distance from the nucleus along the radio jet. See text for discussions of errors bars.
### Fig 7
Plot of the emission-line ratios over the models of shock + precursor of Dopita et al. (1995). Models are for shock velocities of 200,300 and 500 km/sec and magnetic parameter $`B/n^{1/2}`$ of 0,1,2 and 4 $`\mu Gcm^{3/2}`$ respectively.
|
no-problem/9906/hep-ph9906410.html
|
ar5iv
|
text
|
# 1 Introduction
## 1 Introduction
The experience with data analysis in inclusive measurements of deep inelastic scattering (DIS) shows that radiative effects are very important and should be taken into account with maximally possible accuracy. Numerical results obtained for the case of the HERA collider have shown that the ratio of the observed and the Born cross sections varies from 0.5 to 2.0, and even larger values are reached at the borders of the kinematical region. The calculation of radiative corrections (RC) for semi-inclusive and exclusive reactions can not be easily reduced to the case of inclusive scattering, but rather requires separate considerations. There are two important approaches which can be applied for this task. The first one is the so-called semi-analytical approach, where covariant formulae for RC are obtained after applying the procedure of covariant cancellation of infrared divergences and integration over the photon phase space. No other approximation than the ultrarelativistic one is used within this approach. Explicit analytical formulae for RC in diffractive vector meson electroproduction were obtained in Ref. . The radiative correction factor has the following form
$$\eta =\frac{\sigma _{obs}}{\sigma _0}=\mathrm{exp}(\delta _{inf})(1+\delta _{vr}+\delta _{vac})+\frac{\sigma _{hard}}{\sigma _0}.$$
(1)
Here $`\delta _{vr}`$ and $`\delta _{vac}`$ come from so-called virtual corrections: the vertex function and vacuum polarization by lepton pairs and hadrons, resp. $`\delta _{inf}`$ appears after cancellation of infrared divergences. The exponential results from the summation of soft photon contributions over all orders of perturbation theory. $`\sigma _{hard}`$ is the contribution of hard photon radiation. In the general case it has the form of a three-dimensional integral over the photon phase space. In Ref. the numerical analysis was done for fixed target experiments. In this report we provide numerical results also for deep inelastic scattering at HERA. Moreover, we discuss the case of photoproduction within the semi-analytical approach.
Another possible approach for the calculation of RC in semi-inclusive and exclusive processes is based on using a Monte-Carlo generator for inclusive DIS. In this report we discuss how the RC can be calculated using information obtained during event generation. Also we show that, under certain assumptions, both methods give the same results. This fact is illustrated for the case of measurements at HERMES.
It should be noted that all numerical results are shown for the electron method of the reconstruction, when the kinematical variables are calculated using the measurement of the final lepton. There are other reconstruction methods (for example, the double angle method using for the case of diffractive vector meson production), which are, in the most of the phase space, less sensitive to these radiative corrections.
## 2 The code DIFFRAD
The FORTRAN code DIFFRAD was developed to calculate radiative corrections for diffractive vector meson electroproduction and is based on the semi-analytical approach. This code has two versions which are referred to as IDIFFRAD.F and MDIFFRAD.F. The first version allows to calculate the RC to the cross section ($`d^4\sigma /dxdQ^2dtd\varphi _h`$ and $`d^3\sigma /dxdQ^2dt`$) of diffractive vector meson electroproduction using usual (non Monte-Carlo) methods of numerical integration. The second one exploits the same formulae but using Monte Carlo methods for integration. It allows to integrate not only over the photon kinematic variables, but also over those of the final lepton. Integration over $`Q^2`$, $`W^2`$ and $`t`$ in arbitrary bins can be performed as well as over the full kinematically allowed phase space of the final lepton. Therefore MDIFFRAD.F has more possibilities in comparison to IDIFFRAD.F, but it requires more CPU time.
The code requires input for the model of the virtual photoproduction cross sections $`\sigma _{L,T}`$ and allows to apply a cut on inelasticity. As input for $`\sigma _{L,T}`$ we use the model presented originally in Ref. and refined in Ref. (DIPSI). The implementation of other models is possible and straightforward. A cut on the inelasticity $`v`$ ($`v=p_h^2M^2`$, where $`p_h`$ is the total momentum of unobserved particles and $`M`$ is the proton mass) is often suitable to reduce the RC. The value of the cut is related to the experimental resolution on $`v`$ and depends on details of the experimental procedure of non-exclusive background subtraction. A detailed discussion of this point can be found in Ref. . Both versions of DIFFRAD allow one to calculate RC with and without this cut. Note that two additional options also exist for the calculation of the inelasticity distribution with and without RC and for the fast evaluation of an approximate calculation of $`\sigma _{hard}`$.
The FORTRAN code DIFFRAD can be received upon request via email from aku@hep.by
## 3 Electroproduction
Results obtained with the help of DIFFRAD for the three-fold differential cross section are presented in Fig. 1 ($`\sqrt{s}`$=300 GeV). Different plots in this figure show that there is no strong $`W`$-dependence of the RC factor $`\eta `$. This is due to the fact that $`\sigma _L`$ and $`\sigma _T`$ are almost flat in the considered kinematical region. Moreover there is no large explicit dependence on $`W`$ in the RC factor itself. The quantities $`\delta _{vr}`$ and $`\delta _{vac}`$ in Eq. (1) are practically independent of $`W`$ and only a small dependence comes from $`\delta _{inf}`$ and $`\sigma _{hard}`$. The $`Q^2`$-dependence shown in this figure is typical for the case when the inelasticity cut is not applied. If this cut is used then the rise of $`\eta `$ when $`Q^2`$ goes down would be suppressed. From this figure one can also see that the $`t`$-dependence is rather important. Figures 3 and 3 illustrate this last property. $`\eta `$ crosses unity for $`t`$ 0.25–0.3 GeV<sup>2</sup> and rises with increasing $`|t|`$. The large positive correction in this case is a result of the large phase space for photon radiation. The cut applied to the inelasticity (or $`Ep_z`$) can again reduce the value of the RC factor for large $`|t|`$. As a consequence of the $`t`$-dependence of $`\eta `$, the observed slope parameter also receives large RC. This is illustrated in Fig. 3. The Born cross section has a steeper slope with respect to $`|t|`$ than the observed cross section. The radiative correction to the slope parameter is negative and $`10\%`$.
The model used for exclusive virtual photoproduction of vector mesons is based on the hypothesis of s-channel helicity conservation (SCHC)<sup>1</sup><sup>1</sup>1We should note that now there are experimental indications for the violation of this hypothesis as well as theoretical models beyond SCHC . Unfortunately the development of DIFFRAD for using such models is not straightforward, because the calculation of additional terms is required. The calculation for the general case of scalar meson production is considered in .. In the case of SCHC the Born cross section is independent of the azimuthal angle $`\varphi _h`$ between the scattering and the production plane. However, the observed cross section does depend on $`\varphi _h`$, because the formulae for RC includes this angle. Figure 5 shows an example. Since only $`\sigma _{hard}`$ has an essential dependence on the azimuth, $`\eta `$ has a visible $`\varphi _h`$ dependence at large $`|t|`$, where $`\sigma _{hard}`$ gives a relatively large contribution to the total RC.
We note finally that the RC to the cross section of diffractive vector meson electroproduction is very sensitive to the inelasticity cut. This cut suppresses the contribution of hard photon radiation, which is always positive. Thus the use of harder cuts leads to smaller values of the radiative correction factor. However, in the region of small $`|t|`$ where radiative corrections are negative ($`\eta <1`$) the influence of the cut is not essential or leads to larger absolute values of RC ($`\eta `$-1). We do not apply the cut for this numerical analysis.
All numerical results shown here were obtained for the case of exclusive $`\rho (770)`$ production. However there is no essential dependence on the type of the observed vector meson, and all discussed features are very similar for the production of heavier vector mesons.
## 4 Photoproduction
In photoproduction interactions, i.e. very small $`Q^2`$ (quasi-real photon exchanged), the final positron is scattered at very small angle and escapes detection in the main detector through the beam pipe. In this case QED radiative corrections can be calculated by integration of the analytical formulae over the phase space of the final positron. A numerical analysis shows that radiative corrections in this case are smaller than in the case of electroproduction. The results are shown in Fig. 5. Note that small values of RC in this case reflect the statement of the Kinoshita-Lee-Nauenberg theorem . For our case it says that if we integrate over the phase space of a particle which radiates, then all leading logarithms are mutually cancelled (see Ref. , for example).
## 5 Monte Carlo approach
The Monte Carlo approach considered here is based on a generator for inclusive processes with radiative effects included. Here we focus on the case of HERMES and use RADGEN as such a generator. This generator was developed for the fixed target experiments while for HERA experiments HERACLES which includes electroweak effects should be used.
Two Monte Carlo samples with and without radiative effects have to be prepared. The procedure is sketched in Fig. 7. Firstly, the kinematical variables $`\nu `$ and $`Q^2`$ defining the momentum of the final lepton are generated in accordance with the Born cross section. In the case of generation of the sample without RC these variables are used as an input to generate the momentum of the vector meson. For the generation of the sample with RC, a Monte Carlo generator including real photon emission has to be used. Three main tasks are performed by this generator:
1. simulation of the appropriate scattering channel (non-radiative; elastic, quasielastic or inelastic radiative tails) in accordance with their contribution to the total observed cross section;
2. calculation of the inclusive RC factor $`\eta ^{inc}`$ for the given values of $`\nu `$ and $`Q^2`$. Since for both samples we generate the same number of events this factor has to be used as a weight to obtain the observed cross section as a weighted sum of events<sup>2</sup><sup>2</sup>2Another algorithm without calculation of weights is possible. In this case both samples with and without RC are generated in accordance with the observed and the Born cross section, resp. Both approaches are equivalent, and everything discussed below is valid for this case as well.;
3. generation of the kinematical variables $`\nu _{true}`$ and $`Q_{true}^2`$ which are then used as input for generating the momentum of the vector meson. Note that $`\nu _{true}`$ and $`Q_{true}^2`$ differ from $`\nu `$, $`Q^2`$ for events which contain a radiated photon.
The events are collected in predefined bins. If the numbers of events in a given bin for the samples with and without radiative corrections are $`n_1`$ and $`n_2`$, resp., the RC factor for vector meson electroproduction $`\eta _\rho `$ can be estimated as
$$\eta _\rho =\frac{n_1<\eta ^{inc}>}{n_2}$$
(2)
where $`\eta ^{inc}`$ is the RC factor for the inclusive process. We note that due to photon radiation $`\nu _{true}\nu `$. Thus the exclusive production of a vector meson with a given value for its energy (which is related to $`|t|`$) is less probable for a radiative event than for a non-radiative event with the same meson energy. As a result $`n_1<n_2`$ and the typical value of the RC factor for vector meson production (for semi-inclusive processes in general) is smaller than for the inclusive case.
Figure 7 presents results for the HERMES experiment obtained within the two approaches. From this picture we can conclude that the Monte Carlo approach correctly reproduces the RC factor for vector meson electroproduction. The inelasticity cut was not applied for these calculations. We note that the parameter $`\mathrm{\Delta }`$ which separates soft from hard photon radiation and which is always necessary in the Monte Carlo approach (see discussion in Ref. ) should be chosen smaller than its default value $`\mathrm{\Delta }=100`$MeV. This value is close to the threshold energy for observing photons in the HERMES detector. In the case of vector meson electroproduction the maximal photon energy is smaller (especially for small $`|t|`$) in comparison with the inclusive case. So the parameter $`\mathrm{\Delta }`$ has to be reduced also. For the present calculation $`\mathrm{\Delta }=5`$MeV was chosen. For smaller values of $`\mathrm{\Delta }`$ we do not see an essential dependence on $`\mathrm{\Delta }`$.
## Acknowledgments
I am grateful to A.Brull and N.Shumeiko for help and support. Also I would like to thank N.Akopov, A.Borissov, S.Chekanov, L.Favart, E.Kuraev, P.Kuzhir, A.Nagaitsev, N.Nikolaev, A.Proskurykov, M.Ryskin, and H.Spiesberger for fruitful discussions and comments.
|
no-problem/9906/astro-ph9906341.html
|
ar5iv
|
text
|
# 1 Introduction
## 1 Introduction
Both binary eclipsing stars and star clusters are known as a very important source of information concerning fundamental problems of stellar astrophysics. Binary stars provide opportunity to derive basic stellar parameters such as dimensions, masses, distances and stellar composition. Observations of such stars allow for empirical test of stellar evolution models. Comprehensive discussion of these topics can be found in Andersen (1991).
Binary stars located in star clusters are especially well suited for many detailed studies of stellar evolution and dynamics. They may provide independent information on cluster distance, age, chemical composition, interstellar absorption and formation environment. Observations of binary systems in clusters help in studies of differential star formation, anomalies in the initial composition and the role of these stars in the dynamical evolution of star clusters. Moreover, analysis of distribution of binary stars in clusters may help in better understanding their formation and evolution.
An extensive list of binary stars in optical coincidence with galactic open clusters was provided by Gimenez and Clausen (1996). Unfortunately, no such catalogs exist for clusters from other galaxies, in particular the SMC. Only sporadic information about binary stars located in the SMC star clusters can be found in the literature (Hodge and Wright 1975, Sebo and Wood 1994).
The main goal of this paper is to provide for further detailed studies a list of eclipsing systems detected during the OGLE-II microlensing search and located in vicinity of star clusters. In the next section the observational material and used catalogs are described. The resulting list of 127 eclipsing systems from clusters of the SMC is presented in Section 3.
## 2 Observations
All photometric data were obtained with the 1.3 m Warsaw telescope located at the Las Campanas Observatory, Chile, which is operated by the Carnegie Institution of Washington. Detailed description of the instrumental system and reduction techniques were presented by Udalski, Kubiak and Szymański (1997). Udalski et al. (1998a) published the BVI-maps of the Small Magellanic Cloud. These maps contain precise photometric data for about 2.2 million stars from the central region of the SMC bar. Using these data the catalog of 238 star clusters was constructed by Pietrzyński et al. (1998). Udalski et al. (1998b) presented the sample of 1459 eclipsing binary stars from the observed region of the SMC. This catalog contains objects brighter than 20 mag in the I-band with periods ranging from about 0.3 to 250 days. Its average completeness is about 80%.
## 3 Eclipsing Stars in the SMC Clusters
The catalogs of star clusters and eclipsing systems described in the previous Section make it possible to select eclipsing binary stars that are located close to the star clusters. We searched for eclipsing systems located at the distance smaller than 1.5 radius of a given cluster (Pietrzyński et al. 1998). Altogether 124 objects passed this criterium. This group was extended by 3 eclipsing stars from the region of NGC 346. This cluster is situated at the edge of the observed field and it was not included in the catalog of clusters from the SMC (Pietrzyński et al. 1998). However, all three eclipsing stars were found to be in the region of the cluster and we decided to include them to our list. Resulting list comprises 127 objects. Their description and most important data are given in Table 1.
First column of Table 1 contains cluster name. Star ID number and distance from the cluster center, $`D`$, measured in units of cluster radius are presented in columns 2 and 3. Period in days, zero epoch corresponding to the deeper eclipse, $`I`$-band magnitude and $`BV`$ and $`VI`$ colors extracted from the catalog of eclipsing systems (Udalski et al. 1998b) are listed in columns 4, 5, 6, 7 and 8.
Based on location of each eclipsing star in the color-magnitude diagram (CMD) of the appropriate cluster preliminary classification was made. 85 eclipsing variable stars are located in the vicinity of relatively populous clusters. Position of 12 of them in the CMD allows to exclude their membership. They were classified as non-members (nm). Location of 73 systems in the CMDs of these clusters is consistent with their membership. These stars were designed as probable members (pm). The remaining 42 eclipsing systems are found around poorly populated clusters and no conclusive statements about their membership can be made. Information about cluster membership of eclipsing systems is given in column 9 of Table 1.
Figs. 1 and 2 present distribution of periods of eclipsing binary stars from clusters (those classified as probable members) and those from the field of the SMC. Similar shape of these distributions suggests common processes of formation and disruption of eclipsing stars located in the field and clusters of the SMC.
## 4 Summary
OGLE-II catalogs of eclipsing systems and star clusters from the regions covering about 2.4 square degree of the central parts of the SMC were used for selection of eclipsing systems located in star clusters. 127 eclipsing stars were found within 1.5 cluster radii. Based on their position in the cluster CMDs the preliminary classification was made. 73 and 12 eclipsing systems were labeled as probable members and non-members, respectively. Our photometric data do not allow to obtain information on membership of the remaining 42 objects. Distribution of periods of eclipsing stars from star clusters and field objects of the SMC are similar which may suggest similar processes of formation and disruption of eclipsing systems in the field and star cluster regions.
Table 1 and all photometric data can be obtained from the OGLE archive: http://www.astrouw.edu.pl/~ftp/ogle.
Acknowledgements. The paper was partly supported by the Polish KBN grant 2P03D00814 to A. Udalski. Partial support for the OGLE project was provided with the NSF grant AST-9530478 to B. Paczyński.
## REFERENCES
* Andersen, J. 1991, Astron. Astrophys. Rev., 3, 91.
* Gimenez, A., and Clausen, J.V. 1996, ASP Conference Ser., 90, 44.
* Hodge, P.W., and Wright, F.W. 1975, Astron. J., 80, 510.
* Pietrzyński, G., Udalski, A., Szymański, M., Kubiak, M., Woźniak, P., and Żebruń, K. 1998, Acta Astron., 48, 175.
* Sebo, K.M., and Wood, P.R. 1994, Astron. J., 108, 932.
* Udalski, A., Kubiak, M., and Szymański, M. 1997, Acta Astron., 47, 319.
* Udalski, A., Szymański, M., Kubiak, M., Pietrzyński, G., Woźniak, P., and Żebruń, K. 1998a, Acta Astron., 48, 147.
* Udalski, A., Soszyński, I., Szymański, M., Kubiak, M., Pietrzyński, G., Woźniak, P., and Żebruń, K. 1998b, Acta Astron., 48, 563.
|
no-problem/9906/astro-ph9906303.html
|
ar5iv
|
text
|
# Horizontal-Branch Stars as an Age Indicator
## 1. Introduction
For sufficiently old (t $`>`$ 8 Gyr) stellar populations, evolutionary models (e.g., Lee et al. 1994, LDZ) predict that the surface T$`_{eff}`$ distribution of horizontal-branch (HB) stars is very sensitive to age, which makes it an attractive age indicator. Because the HB stars are much brighter than main-sequence (MS) stars, the interpretation of HB morphology in terms of relative age differences would be of great value in the study of distant stellar populations where the MS turnoff is fainter than the detection limit. In the more distant stellar populations where even HB is fainter than the detection limit, HB stars still should have some crucial effect in spectrophotometric dating of old stellar populations based on integrated colors & spectra. In this paper, we report our progress in the use of HB as an age indicator, both from the interpretation of HB morphology in the color-magnitude (CM) diagrams, and from the integrated spectra.
## 2. Age as the Major Second Parameter
Some five years ago, LDZ have concluded that age is the most natural candidate for the global second parameter, because other candidates can be ruled out from the observational evidence, while supporting evidence do exist for the age hypothesis. There is still much debate about this issue, and Table 1 summarizes the current situation. Although it may not be a complete list, it is clear from Table 1 that many pieces of supporting evidence still suggest that age is the
major global second parameter. Recent addition to this list include red HB clusters in the outer halo, such as Palomar 3 & 4, and Eridanus from the HST photometry (Stetson et al. 1999), and Palomar 14 from the WIYN photometry (Sarajedini 1997). Before these observations, they used to say that these clusters will eventually solve the second parameter problem because they represent the most extreme cases of the second parameter effect. Now the observed age differences for these clusters, as estimated from the color difference method, are consistent with the age scenario of the 2nd parameter effect. Furthermore, J.-W. Lee & Carney (1999) and Yim et al. (1999) recently report that their high quality observations of blue HB clusters M2 and M13 are consistent with the age hypothesis when compared to the redder HB clusters of similar metallicity, such as M3.
On the other hand, we see that there are now some papers in the literature arguing that age may not be the major second parameter. Their argument, mostly based on M3-M13 pair, is that they did find some age difference, about 1$``$2 Gyr, but they think it is not enough to explain the difference in HB morphology. However, as described below, to within the observational errors, both in the age dating technique and in the HB type and metallicity measurements, our new models with the effects of recent developments are completely consistent with the observations. Other arguments include the HST photometry of Draco dwarf spheroidal galaxy (Grillmair et al. 1998), which has a red HB, in which they argue that Draco is not apparently younger than the blue HB clusters of similar metallicity. However, as they admit, the quality of their data is still very poor to reach any conclusion.
We are not arguing that age can explain every aspect of the HB. In fact, LDZ also admitted that something else is also required to explain some peculiar features on the HB, such as long “blue tail” and bimodal HB color distribution (see Table 1). However, these features are widely considerd to be a result of local effect, such as enhanced mass-loss in high density environment. Given the overwhelming supporting evidence for the age hypothesis, these local effects should be considered to be a third parameter effect, which is less important than the second parameter effect (see also Binney & Merrifield 1998). Certainly, there is noise, and it is important to understand the origin of this third parameter effect in order to use the HB as a more reliable age indicator. In this respect, it is encouraging to see that recent observations with the Keck HIRES spectrograph (Behr et al. 1999) provide some compelling evidence that the “blue tail” phenomenon is a natural result of abrupt diffusion mechanisms for stars hotter than 11,500K (see also Moehler et al. 1999). If, as suggested by these observations, the “blue tail phenomenon” is a universal characteristic of clusters with stars hotter than 11,500K, then this phenomenon should not be considered to be a third parameter effect, since it is then rather a general feature of the very blue HB clusters. Furthermore, several lines of evidence now suggest that some globular clusters with peculiar CM diagram morphology, such as M54 (Sarajedini & Layden 1995; Larson 1996) and $`\omega `$ Centauri (Lee et al. 1999), are nuclei of disrupted dwarf galaxies with internal age-metallicity relations. We suspect, therefore, a globular cluster with peculiar bimodal HB distribution, NGC 2808, may also represent such case. Note that it is among the most massive globular clusters in our Galaxy (Harris 1996). In any case, as demonstrated by Fusi Pecci et al. (1996), a procedure is available through which the measured HB morphology can be depurated from the effects of parameters (i.e., third parameter) other than age. So even in the clusters with peculiar HB morphology from unknown origin, a rough estimate of relative age is still possible from their HBs.
## 3. Recent Developments & New HB Population Models
There are several recent developments that can potentially affect the relative age dating technique from HB morphology. The following effects are included in our most recent update of the LDZ HB models (Yoon & Lee 1999):
1. There is now a reason to believe that absolute age of the oldest Galactic globular clusters is reduced to about 12 Gyr, as suggested by new Hipparcos distance calibration and other improvements in stellar models (Gratton et al. 1997; Reid 1997; Chaboyer et al. 1998). As LDZ already demonstrated in their paper, this has a strong impact in the relative age estimation from HB morphology, because the variation of the red giant branch (RGB) tip mass (after the mass-loss, HB mass) is more sensitive to age at younger ages.
2. Now, we have new HB tracks with improved input physics (Yi et al. 1997) and corresponding new Yale isochrones (Demarque et al. 1996).
3. It is now well established that $`\alpha `$-elements are enhanced in halo populations. Specifically, we adopt \[$`\alpha `$/Fe\] = 0.4 for clusters with \[Fe/H\] $`<`$ -1.0, and thereafter we assume that it steadily declines to 0.0 at solar metallicity (e.g., Wheeler et al. 1989). In practice, the treatment suggested by Salaris et al. (1993) was used to simulate the effect of $`\alpha `$-elements enhancement.
4. Finally, Reimers’(1975) empirical mass-loss law suggests more mass-loss at larger ages. The result of this effect was also presented in LDZ, but unfortunately most widely used diagram (their Fig. 7) is the one based on fixed mass-loss.
We found that all of the above effects make the HB morphology to be more sensitive to age (see Yoon & Lee 1999 for more details). Therefore, as illustrated in Figure 1, now the required age difference is much reduced compared to Figure 7 of LDZ. Now, only 1.2 Gyr of age difference, rather than 2 Gyr, is enough to explain the systematic shift of the HB morphology between the inner and outer halo clusters. To within the observational uncertainties, age differences of about 1.5$``$2 Gyr are now enough to explain the observed differences in HB morphology between the remote halo clusters (Pal 3, Pal 4, Pal 14, & Eridanus) and M3, and also between M3 and M13 (or M2). These values are consistent with the recent relative age datings both from the HST and high-quality ground-based data.
## 4. Effect of HB on the Integrated Spectra
If age is indeed the major parameter that controls HB morphology in addition to metallicity, then we expect some effect of it on the integrated spectra. For example, Figure 2 presents our models (Lee, Yoon, & Lee 1999) that illustrate the effect of age sensitivity of the HB on the H$`\beta `$ index. First, we have compared in panel (a) our models without the HB with those of Worthey (1994) in order to make sure that our calculations are consistent with previous investigations.
Thus, panel (a) illustrates only the effect of MS turnoff $`T_{eff}`$ variation with metallicity and age. Confirmed that there is no systematic difference with previous investigations, we then added HB stars in our models, along the HB isochrones similar to those in Figure 1b, but including some very old ages ($`\mathrm{\Delta }`$t = +2 & +4 Gyr). The result is very striking as shown in panel (b). Of course, the wave-like features are due to the variation of the HB with metallicity and age. For given age, HB gets bluer with decreasing metallicity, and when the mean $`T_{eff}`$ of the HB reaches around 9500K, the population models produce peak in the H$`\beta `$ index. But as metallicity decreases further, HB is becoming too hot to contribute significantly to the integrated H$`\beta `$ index.
The advent of large ground-based telescopes making it possible to obtain some absorption features of globular clusters in nearby giant ellipticals so that our models can be compared directly with the observations. Data plotted in panel (c) of Figure 2 were obtained at Keck telescope by Kissler-Patig et al. (1998) and Cohen et al. (1998), using the same instrument, both for the globular clusters in the giant elliptical galaxy NGC 1399 and those in the Milky Way Galaxy. At first glance, they may appear to be more or less the same, especially when considering still large errors in NGC 1399. However, some careful examination of the data indicates that there is some systematic difference in that the metal-rich clusters in NGC 1399 have higher H$`\beta `$ compared to the Galactic counterparts, while the opposite seems to be the case for more metal-poor clusters. Comparing this with our models, we can say that NGC 1399 system is probably several billion years older, in the mean, than the Galactic globular cluster system. Of course, given the large observational uncertainty, more observations are badly needed to confirm this possible scenario. If confirmed by future observations, this would indicate that the star formation in proto giant ellipticals started at an earlier epoch than in the less massive galaxies, such as our Milky Way Galaxy.
What is interesting to us is that a similar age difference is also inferred from the “metal-poor HB solution” of the UV upturn phenomenon of giant ellipticals (Park & Lee 1997; Yi et al. 1999). Our composite models (see also Lee et al., this volume) indicate that age difference of about 3 billion years can also explain the systematic difference in UV upturn between the giant ellipticals and the spiral bulges of the Local Group. Whether the UV upturn phenomenon is indeed a natural extension of the global second parameter effect observed in Galactic globular clusters (“metal-poor HB solution”), or it is rather due to the high $`\mathrm{\Delta }`$Y/$`\mathrm{\Delta }`$Z and enhanced mass-loss in super metal-rich population (“metal-rich HB solution”) is still under debate. Fortunately, there are several observational tests for this problem that will eventually provide us more concrete calibration of the far-UV dating for old stellar populations. First, as pointed out by Lee (1994), the “metal-rich HB solution” predicts many super metal-rich RR Lyrae stars in the Galactic bulge because super metal-rich HB stars must cross the instability strip as they move back to blue HB with increasing metallicity. In fact, detailed models by Lee & Lee (1999) indicate that the expected number of super metal-rich RR Lyraes is compatible with more metal-poor ones actually observed in the bulge. Although extensive $`\mathrm{\Delta }S`$ observations by Walker & Terndrup (1991) found not a single super metal-rich RR Lyrae stars in this field, more efficient observations with the Caby photometric system (Rey et al. 1999) would be useful to confirm this result. Second, UIT observations found strong radial gradients in UV colors within elliptical galaxies (O’Connell et al. 1992), and thus it is important to see whether they are correlated with internal metallicity gradients. Indeed, Ohl et al. (1998) report that they found no correlation between the far-UV gradients and internal metallicity gradients, from which they conclude that UV upturn phenomenon may not be simply related to overall metal abundances in galaxies. They even suggested age as an alternative, which is consistent with our “metal-poor HB solution”. Ongoing STIS/HST observations will provide more clear result on this problem. Finally, far-UV observations with the planned GALEX UV space facility will also provide useful database that is crucial in calibration of our UV dating technique for old stellar systems. Although the importance of new observations can not be overstated, it is already clear now that the detailed modeling of the HB is crucial in spectrophotometric dating of old stellar populations.
## References
Behr, B. B., Cohen, J. G., McCarthy, J. K. & Djorgovski, S. G. 1999, ApJ, 517, L135
Binney, J. & Merrifield M. 1998, in Galactic Astronomy (Princeton University Press), Ch.6
Bolte, M. 1989, AJ, 97, 1688
Buonanno, R., Buscema, G. Fusi Pecci, F. Richer, H. B. & Fahlman, G. G. 1990, AJ, 100, 1811
Buonanno, R., Caloi, V., Castellani, V., Corsi, C. E., Fusi Pecci, F. & Gratton, R. 1986, A&AS, 66, 79
Buonanno, R., Corsi, C. E., Fusi Pecci, F., Fahlman, G. G. & Richer, H. B. 1994, ApJ, 430, L121
Buonanno, R., Corsi, C. E., Fusi Pecci, F., Richer, H. B. & Fahlman, G. G. 1993, AJ, 105, 184
Buonanno, R., Corsi, C. E., Fusi Pecci, F., Richer, H. B. & Fahlman, G. C. 1995a, AJ, 109, 650
Buonanno, R., Corsi, C. E., Pulone, L., Fusi Pecci, F., Richer, H. B. & Fahlman, G. C. 1995b, AJ, 109, 663
Catelan, M. & de Freitas Pacheco, J. A. 1995, A&A, 297, 345
Chaboyer, B., Demarque, P., Kernan, P. J. & Krauss, L. M. 1998, ApJ, 494, 96
Chaboyer, B., Demarque, P. & Sarajedini, A. 1996, ApJ, 459, 558
Chaboyer, B., Sarajedini, A. & Demarque, P. 1992, ApJ, 394, 515
Cohen, J. G., Blakeslee, J. P. & Ryzhov, A. 1998, ApJ, 496, 808
Da Costa, G. S., Armandroff, T. E. & Norris, J. E. 1992, AJ, 104, 154
Davidge, T. J. & Courteau, S. 1999, AJ, 117, 1297
Demarque, P., Chaboyer, B., Guenther, D., Pinsonneault, M., Pinsonneault, L. & Yi, S. 1996, Yale Isochrone 1996
Ferraro, F. R., et al. 1997, ApJ, 484, L145
Ferraro, F. R., Paltrinieri, B., Fusi Pecci, F., Rood, R. & Dorman 1998, ApJ, 500, 311
Ferraro, I., Ferraro, F. R., Fusi Pecci, F., Corsi, C. E. & Buonanno, R. 1995, MNRAS, 275, 1057
Fusi Pecci, F., et al. 1996, in Formation of the Galactic Halo, Inside and Out, ASP Conf. Proc. 92, ed. H. Morrison & A. Sarajedini (ASP:San Francisco), 221
Gratton, R. G., Fusi Pecci, F., Carretta, E., Clementini, G., Corsi, C. E. & Lattanzi, M. 1997, ApJ, 491, 749
Gratton, R. G. & Ortolani, S. 1988, A&AS, 73, 137
Green, E. M. & Norris, J. E. 1990, ApJ, 353, L17
Grillmair, C. L., et al. 1998, AJ, 115,144
Harris, W. E. 1996, AJ, 112, 1487
Hesser, J. E., et al. 1997, BAAS, 191, 7901
Johnson, J. A. & Bolte, M. 1998, AJ, 115, 693
Kissler-Patig, M., Brodie, J. P., Schroder, L. L., Forbes, D. A., Grillmair, C. J. & Huchra, J. P. 1998, AJ, 115, 105
Kubiak, M. 1991, Acta Astron., 41, 231
Larson, R. B. 1996, in Formation of the Galactic Halo, Inside and Out, ASP Conf. Proc. 92, ed. H. Morrison & A. Sarajedini (ASP:San Francisco), 241
Lee, C.-H. & Lee, Y.-W. 1999, in preparation
Lee, H.-c., Yoon, S.-J. & Lee, Y.-W. 1999, in preparation
Lee, J.-W. & Carney, B. W. 1999, AJ, in press
Lee, Y.-W. 1994, ApJ, 430, L113
Lee, Y.-W., Demarque, P. & Zinn, R. 1988, in Calibration of Stellar Ages, ed. A. G. D. Philip (Schenectady: L. Davis), 141
Lee, Y.-W., Demarque, P. & Zinn, R. 1994, ApJ, 423, 248 (LDZ)
Lee, Y.-W., Joo, J.-M., Sohn, Y.-J., Rey, S.-C. & Walker, A. R. 1999, in preparation
Moehler, S, Sweigart, A. V., Landsman, W. B., Heber, U. & Catelan, M. 1999 A&A, 346, L1
O’Connell, R. W., et al. 1992, ApJ, 395, L45
Ohl, R. G., et al. 1998, ApJ, 505, L11
Paltrinieri, B., Ferraro, F. R., Fusi Pecci, F. & Carretta, E. 1998, MNRAS, 293, 434
Park, J.-H. & Lee, Y.-W. 1997, ApJ, 476, 28
Reid, I. N. 1998, AJ, 115, 204
Reimers, D. 1975, Mem. Soc. Roy. Sci. Leige, 6th Ser., 8, 369
Rey, S.-C., Lee, Y.-W., Joo, J.-M., Walker, A. R., & Baird, S. 1999, AJ, submitted
Rich, R. M., et al. 1997, ApJ, 484, L25
Rosenberg, A., Saviane, I., Piotto, G., Aparicio, A. & Zaggia, S. R. 1998, AJ, 115, 648
Salaris, M., Chieffi, A. & Straniero, O. 1993, ApJ, 414, 580
Sarajedini, A. 1997, AJ, 113, 682
Sarajedini, A. & Demarque, P. 1990, ApJ, 365, 219
Sarajedini, A. & King, C. R. 1989, AJ, 98, 1624
Sarajedini, A. & Layden, A. C. 1995, AJ, 109, 1086
Sarajedini, A. & Layden, A. C. 1997, AJ, 113, 264
Sosin, C., et al. 1997, ApJ, 480, L35
Stetson, P. B., et al. 1999, AJ, 117, 247
Stetson, P. B., VandenBerg, D. A. & Bolte, M. 1996, PASP, 108, 560
Stetson, P. B., VandenBerg, D. A., Bolte, M., Hesser, J. E. & Smith, G. H. 1989, AJ, 97, 1360
VandenBerg, D. A., Bolte, M. & Stetson, P. B. 1990, AJ, 100, 445
Walker, A. R. & Terndrup, D. M. 1991, ApJ, 378, 119
Wheeler, J. C., Sneden, C. & Truran, J. W. 1989, ARAA, 27, 279
Worthey, G. 1994, ApJS, 95, 107
Yi, S., Demarque, P. & Kim, Y.-C. 1997, ApJ, 482, 677
Yi, S., Lee, Y.-W., Woo, J.-H., Park, J.-H., Demarque, P. & Oemler, A., Jr. 1999, ApJ, 513, 128
Yim, H.-S., Chun, M.-S., Byun, Y.-I. & Lee, Y.-W. 1999, in preparation
Yoon, S.-J. & Lee, Y.-W. 1999, in preparation
|
no-problem/9906/astro-ph9906170.html
|
ar5iv
|
text
|
# A Method of Mass Measurement in Black Hole Binaries Using Timing and High Resolution X-ray Spectroscopy
## 1. Introduction
The existence of stellar-mass black holes in our Galaxy is almost exclusively established by measuring the mass of the compact object in several bright X-ray binaries. A compact object more massive than approximately $`3M_{}`$ is unstable and should collapse to form a black hole (Rhoades & Ruffini 1974, Chitre & Hartle 1976). Therefore, if the dynamically measured mass of a luminous X-ray source exceeds $`3M_{}`$, it is considered a firm black hole candidate.
In X-ray binaries, the orbital velocity of the compact object cannot be measured by means of optical spectroscopy. This means that the only directly measured quantity is the mass function $`f(M)`$,
$$f(M)=\frac{M_x^3\mathrm{sin}^3i}{(M_x+M_{\mathrm{opt}})^2},$$
(1)
where $`M_x`$ and $`M_{\mathrm{opt}}`$ are masses of the compact object and normal companion, respectively, and $`i`$ is the inclination angle of the orbit. The mass function is the lower limit of the compact object mass. In many high mass X-ray binaries, the mass function is uninterestingly low; for example, $`f(M)=0.25M_{}`$ for Cyg X-1 (Gies & Bolton 1986). A determination of $`M_x`$ is possible if $`M_{\mathrm{opt}}`$ and $`\mathrm{sin}i`$ are independently constrained. Extensive work has been done to estimate these parameters for several black hole candidates (Bolton 1972, Liutyj, Sunyaev & Cherepashchuk 1975, McClintock & Remillard 1986, Remillard, McClintock & Bailyn 1992, Herrero et al. 1995, Filippenko, Matheson & Barth 1995, Bailyn et al. 1995, Remillard et al. 1996). However, the derived compact object mass can be rather uncertain due to its strong dependence on $`M_{\mathrm{opt}}`$ and especially $`\mathrm{sin}i`$. For Cyg X-1, mass values in the range 6–16$`M_{}`$ are found in the literature (Dolan & Tapia 1989, Gies & Bolton 1986).
The mass measurement will be much more accurate if some information about the orbit of the compact object is available. Unfortunately, very fast motions of the material in the vicinity of the black hole make it impossible to measure the orbital velocity from the Doppler shifts of spectral lines. Instead, we propose to measure the orbital radius using the following approach. In high mass X-ray binaries, some fraction of the compact object X-ray luminosity is intercepted by the normal companion surface (the intercepted fraction in low mass binaries is too low for an application of our method). A significant fraction of this energy is re-radiated in the X-ray band either through Compton reflection or line fluorescence. This emission reaches the observer with a time delay roughly equal to $`a/c`$, where $`a`$ is the separation between the normal companion and the compact object. Since the compact object emission is usually variable, it may be possible to measure the time delay and thus estimate the orbital radius. We show below that this measurement can be carried out using temporally resolved high resolution spectroscopy of the fluorescent K line of neutral iron. Let us demonstrate how this new information refines the measurement of the compact object mass. Suppose for simplicity that the orbit is circular and the radius of the normal companion is negligible compared to the orbital radius. The time delay at orbital phase $`\varphi `$ is
$$\mathrm{\Delta }t=a/c\left[1-\mathrm{sin}(i\mathrm{cos}2\pi \varphi )\right].$$
(2)
Equations (1)–(2) together with the Kepler’s third law,
$$a^3=P^2(M_x+M_{\mathrm{opt}})G/4\pi ^2,$$
(3)
can be used to express $`M_x`$ as a function of just one parameter, $`M_{\mathrm{opt}}`$, $`i`$, or $`a`$. In particular, if the time delay measurement is performed at inferior conjunction ($`\varphi =0.5`$),
$$M_x=\frac{f^{1/3}g^{2/3}}{\mathrm{sin}i(1+\mathrm{sin}i)^2}\mathrm{or}M_x=\frac{(M_x+M_{\mathrm{opt}})f^{1/3}}{g^{1/3}-(M_x+M_{\mathrm{opt}})^{1/3}},$$
(4)
where $`g=4\pi ^2(c\mathrm{\Delta }t)^3/P^2G`$. If the time delay is measured at several orbital phases, all system parameters can be determined independently. Moreover, useful mass constraints are possible in this case even if no optical data are available (§ 4).
## 2. Measurement strategy
The X-ray emission reflected from the surface of the normal companion consists of the Compton-reflected continuum and the fluorescent emission. While the reflected continuum is weak compared to the direct emission from the black hole, the fluorescent line emission can be relatively strong. The strongest fluorescent line is the iron K doublet (Basko, Sunyaev & Titarchuk 1974). Basko (1978) estimates the equivalent width of $`7`$eV for this line in Cyg X-1. With $`5`$eV spectral resolution of X-ray calorimeters, it should be possible to measure the source light curve in a narrow spectral interval around K<sub>1</sub> and K<sub>2</sub>, where the fluorescent line contributes $`50\%`$ of the total flux. The time delay can be measured by cross-correlating the light curve in the K region with the continuum light curve which is dominated by the direct black hole emission.
K lines are primarily excited through photoelectric absorption in the K-edge ($`E_K=7.1`$ keV for neutral iron). Therefore, the energy band just above 7.1 keV is best for the continuum light curve measurement. The exact choice of the energy boundaries is unimportant as long as the source variability patterns do not change with energy, as is the case in Cyg X-1 (Nowak et al. 1999). Below, we assume that the continuum is measured in the 7.1–9.1 keV energy band.
The most challenging part of the time delay measurement is determining the light curve in the K line region. This line consists of two components, K<sub>1</sub> and K<sub>2</sub>, with energies $`E__1=6.404`$ and $`E__2=6.391`$ keV (Bambynek et al. 1972). Both components have approximately equal natural widths of 3.5 eV (FWHM). The Doppler broadening of the lines due to thermal motions of the emitting atoms $`\mathrm{\Delta }E_{\mathrm{th}}=0.1(T/10^5\mathrm{K})^{1/2}`$ eV is much smaller than the natural width of the lines. Therefore, more than 90% of the total K line flux is within a narrow, 20 eV, energy interval around $`E=6.398`$ keV. Selecting photons in this energy interval is easy with the 5 eV (FWHM) energy resolution of future X-ray calorimeters. An approximately 20 eV energy interval seems optimal for the time delay measurement. A wider band would include a larger contribution of the direct continuum without increasing the line flux significantly. Narrower intervals would miss flux in the line wings.
The cross-correlation function of the continuum and line light curves should contain two prominent peaks. A strong narrow peak around the delay $`\tau =0`$ corresponds to the auto-correlation of the emission coming directly from the black hole. Reflection from the normal companion results in a secondary broad peak in the correlation function. This peak should have a centroid at $`\tau a/c`$ and width $`\mathrm{\Delta }\tau R/c`$, where $`a`$ and $`R`$ are the radii of the orbit and normal companion, respectively. The amplitude of the secondary peak should be proportional to the equivalent width of the reflected K line and the $`\mathrm{𝑟𝑚𝑠}^2`$ of the source flux variability. Therefore, the time delay measurement is best performed when the reflected line is strong (i.e. near inferior conjunction, Basko 1978), and when the source is highly variable.
In the next section, we simulate an observation of the time delay in Cyg X-1 which could be performed by a telescope with an effective area of 1000 cm<sup>2</sup> in the 6–9 keV energy range. Such an instrument approximately corresponds to one spacecraft of the proposed *Constellation-X* mission.<sup>1</sup><sup>1</sup>1See http://constellation.gsfc.nasa.gov/ for information
## 3. Simulated Observation
We adapt the binary system parameters from Herrero et al. (1995) — a radius of the normal companion $`R=17R_{}`$, a separation between the components $`a=40R_{}`$, and an inclination angle $`i=35^{}`$. For simplicity, we assume that the orbit is circular and the normal companion is spherical. For a similar geometry, Basko (1978) predicted the equivalent width of the reflected iron line at inferior conjunction $`W_{\varphi =0.5}=7`$ eV. Basko assumed Solar chemical composition of the normal companion. However, some X-ray observations indicate that the iron abundance may be twice Solar (Kitamoto et al. 1984). Such an overabundance of iron would increase the fluorescent line flux to $`W14`$ eV. A narrow iron K line of similar amplitude was indeed observed by the *ASCA* SIS (Ebisawa et al. 1996). We adopt $`W_{\varphi =0.5}=10`$eV, which agrees with both theoretical predictions and current observations.
As was discussed in § 2, we assume that the K light curve is measured in the 20 eV energy interval centered at $`6.398`$ keV, and the continuum light curve is observed in the 7.1–9.1 keV band. The continuum spectrum is assumed to be a power law with a photon index $`\mathrm{\Gamma }=1.65`$ (Liang & Nolan 1984). We also allow for the possible presence of a broad iron line originating in the vicinity of the black hole, for example, in the outer part of the accretion disk. We assume that the equivalent width of the broad line is 0.1 keV (see, e.g., Dove et al. 1998) and that the line width is 0.08 keV (FWHM) which corresponds to the velocity dispersion of 2000 km s<sup>-1</sup>.
For the source variability, we adopt the shot noise model of Lochner et al. (1991). This model describes the variability of Cyg X-1 as a constant component plus a sequence of exponential flares with durations distributed in the 0.01–3.22 s interval. The variability level during the 1997 *HEAO 1* A-2 observation modeled by Lochner et al. was quite typical of the source (35% *rms*). The shot noise model is adequate for simulation of the direct emission from the black hole. The light curve of the reflected radiation is the convolution of the direct emission light curve with the reflection response to the instantaneous flare. This response function is calculated below.
Consider the fluorescent emission coming to the observer from the normal companion surface element $`dS`$ (Fig. 3). The distance between the surface element and the black hole is
$$l=\left[R^2+a^2-2Ra(\mathrm{cos}\psi \mathrm{sin}\theta \mathrm{cos}\phi +\mathrm{sin}\psi \mathrm{cos}\theta )\right]^{1/2}.$$
(5)
The reflected emission reaches the observer with a delay $`\tau `$
$$c\tau =l+a\mathrm{sin}\psi -R\mathrm{cos}\theta $$
(6)
relative to the direct emission from the black hole. The fraction of the X-ray luminosity intercepted by the surface element is
$$dL=LdS\mathrm{cos}\mathrm{\Theta }_i/4\pi l^2,$$
(7)
where $`\mathrm{\Theta }_i`$ is the incident angle of the X-rays,
$$\mathrm{cos}\mathrm{\Theta }_i=(\mathrm{cos}\psi \mathrm{sin}\theta \mathrm{cos}\phi +\mathrm{sin}\psi \mathrm{cos}\theta )a/l-R/d.$$
(8)
The ratio of intensities of the observed K line and the incident continuum above the K-edge is a strong function of $`\mathrm{\Theta }_i`$ and the fluorescence angle, $`\theta `$. This ratio, $`Y(\mathrm{\Theta }_i,\theta )`$, can be calculated analytically following the work of Basko. It was also derived by George & Fabian (1991) using Monte-Carlo simulations. George & Fabian present $`Y(\mathrm{\Theta }_i,\theta )`$ for an incident power-law spectrum with $`\mathrm{\Gamma }=1.7`$ on a grid of $`\mathrm{\Theta }_i`$ and $`\theta `$ values. We use numerical interpolation of the George & Fabian results in our calculations below.
The light curve of the fluorescent emission from the surface element is
$$f(t)=Y(\mathrm{\Theta }_i,\theta )\delta (t-\tau )dL$$
(9)
Integration of equation (9) over the normal companion’s surface provides the time response of the reflected emission. Figure 3 shows the results for the Cyg X-1 system at three orbital phases. A different set of responses is expected for a non-spherical normal companion, for example, the one filling its Roche lobe. However, the calculations in this case should be analogous to the procedure outlined above.
Our simulations proceed as follows. The continuum light curve in the 7.1–9.1 keV band is simulated using the shot noise model of Lochner et al. The reflected K line light curve is the convolution of the 7.1–9.1 keV light curve with the time response from Fig 3. The K line flux is rescaled so that the line equivalent width (relative to the $`\mathrm{\Gamma }=1.65`$ power law continuum) is 10 eV at $`\varphi =0.5`$. The 6.388–6.408 keV flux as a sum of the K line flux and the contribution of the direct emission calculated as the 7.1–9.1 keV light curve scaled according to the spectral model which consists of the power law and a broad 6.4 keV line (see above). Finally, we scale the simulated light curves to reproduce the observed Cyg X-1 flux at 7 keV (0.05 ph s<sup>-1</sup> cm<sup>-2</sup> keV<sup>-1</sup>, Ebisawa et al. 1996), multiply them by the effective area of 1000 cm<sup>2</sup>, and add Poisson noise.
The cross-correlation functions derived from the simulated $`1000`$ksec observations are presented in Fig 3. The secondary peak due to reflected emission is clearly visible at orbital phases $`\varphi =0.5`$ and 0.25. For $`\varphi =0`$, the secondary cross-correlation peak is weak and not separated from the autocorrelation of the direct emission (the average delay is only 14 s for the adopted system parameters).
## 4. Discussion
We have shown that the time delay between the direct and reflected emission can be measured in a Cyg X-1 like system. Arguably, this measurement requires long observations. For a convincing detection, a 250 ksec observation on a telescope with 1000 cm<sup>2</sup> effective area around 6 keV is required. A longer, 1000 ksec, exposure is needed for a quality measurement of the cross-correlation function (Fig 3). These exposures are longer than the 5.6 day orbital period of Cyg X-1, and therefore the observation needs to be split into several intervals, each performed at the same orbital phase. The required exposure is shorter for larger area telescopes. Unfortunately, the gain is smaller than a factor of $`(\mathrm{area})^{1/2}`$ because the internal source variability becomes the dominant source of noise. To achieve a detection significance similar to that in Fig. 3, $`300`$ ksec exposure is a minimum even for very large area telescopes.
In addition to the fluorescence from the surface of the normal companion, iron K lines can form in the accretion disk or stellar wind. These emissions are contaminants for the time delay measurement. Fortunately, it may be possible to separate them in the energy or time domains from the normal component fluorescence. First, lines originating in the accretion disk or stellar wind are likely to be significantly broadened because the reflecting material moves at high speed (a possible detection of the broadened iron line in Cyg X-1 is reported by Done & Zycki 1999). Thus, we can exclude a large fraction of the disk and possibly wind fluorescent emission by selecting a narrow range around the K line rest frame energy. Second, iron in both wind and disk is likely to be partially ionized because of the low matter density and high ionizing flux from the black hole. K lines for ionization states higher than $`\mathrm{Fe}\mathrm{XV}`$ are separated from the $`\mathrm{Fe}\mathrm{I}`$ lines by more than 20 eV (House 1969). Third, the accretion disk or stellar wind fluorescence should produce cross-correlation functions very different from those in Fig. 3. Since the outer disk radius is only a small fraction of the black hole Roche lobe size (Shapiro & Lightman 1976), the time delay of the fluorescent disk emission is much shorter than that in Fig. 3. In the case of stellar wind fluorescence, time delays should be distributed in the broad range $`\mathrm{\Delta }\tau r/c`$ around the average value $`\overline{\tau }=r/c`$, where $`r`$ is the characteristic radius of the fluorescence region. Therefore, a very broad feature in the cross-correlation function is expected due to stellar wind fluorescence. Such a feature should be easily distinguishable from the relatively narrow peak arising from the normal companion reflection. To conclude, the only consequence of the presence of the iron K-fluorescence from the stellar wind or accretion disk is to add noise to the cross-correlation function in the time delay range of interest.
The quality of the data in our simulated observation is sufficient to determine both the average delay and *shape* of the cross-correlation peak. The average delay provides the orbital separation between the components which can be used to refine the black hole mass estimates (§ 1). The width of the correlation peak is proportional to the radius of the normal companion. Its shape depends on the shape of the companion surface and the inclination angle (e.g., the $`\varphi =0.25`$ curve in Fig. 3 corresponds to all orbital phases in an $`i=0`$ system). The shift of the average delay as a function of the orbital phase also depends on the inclination angle and the companion radius (eq. 2, Fig. 3). Therefore, a much more detailed modeling than simple considerations outlined in eqs. (2)–(4) might be possible with high-quality time delay data. This should result in the compact object mass determination without the $`\mathrm{sin}i`$ or $`M_{\mathrm{opt}}`$ degeneracies of eq (4).
Basko (1978) pointed out that the orbital modulation of the reflected K line flux can be used to estimate the orbital parameters, in particular, the inclination angle. Our method is independent of Basko’s, because we do not use the line flux. The interpretation of the orbital modulation of the K line flux can be uncertain. For example, the fluorescence in the stellar wind reduces the orbital modulation of the line flux; a similar behavior is expected for low inclination angles (see Basko’s Fig. 4). On the contrary, our method is not directly affected by either stellar wind or accretion disk fluorescence.
The time delay technique opens the possibility of mass determination in X-ray systems without known optical counterparts, such as bright black hole candidates in the Galactic center region or LMC. In these systems, the binary period and the epoch of zero orbital phase can be determined by periodic variations of either the iron line intensity (Basko 1978) or location of the cross-correlation peak. Near the orbital phase $`\varphi =0.25`$, the cross-correlation function is independent of the inclination angle and can be used to measure the orbital radius, $`a`$, and the radius of the normal companion, $`R`$, (Fig. 3). The total system mass $`M_{\mathrm{opt}}+M_x`$ is then determined from the Kepler’s third law. Under the assumption that the normal companion almost fills its Roche lobe, which should be the case in X-ray luminous systems, the ratio $`R/a`$ defines $`M_x/M_{\mathrm{opt}}`$ and thus provides an estimate of $`M_x`$.
This work was supported by the Harvard-Smithsonian Center for Astrophysics postdoctoral fellowship.
|
no-problem/9906/cond-mat9906449.html
|
ar5iv
|
text
|
# Melting and Dimensionality of the Vortex Lattice in Underdoped YBa2Cu3O6.60
## Abstract
Muon spin rotation ($`\mu `$SR) measurements of the magnetic field distribution in the vortex state of the oxygen deficient high-$`T_c`$ superconductor YBa<sub>2</sub>Cu<sub>3</sub>O<sub>6.60</sub> reveal a vortex-lattice melting transition at much lower temperature than that in the fully oxygenated material. The transition is best described by a model in which adjacent layers of “pancake” vortices decouple in the liquid phase. Evidence is also found for a pinning-induced crossover from a solid 3D to quasi-2D vortex lattice, similar to that observed in the highly anisotropic superconductor Bi<sub>2+x</sub>Sr<sub>2-x</sub>CaCu<sub>2</sub>O<sub>8+y</sub>.
74.60.Ge, 74.25.Dw, 76.75.+i, 74.72.Bk
In the high-temperature superconductors there exist several exotic vortex lattice (VL) phases owing to the weak coupling between the superconducting CuO<sub>2</sub> layers (which gives rise to highly flexible vortices), a relatively short coherence length (which enhances their susceptibility to pinning) and high values of $`T_c`$ (which permit large thermal energies to be reached in the vortex state). To date much attention has been paid to VL melting and pinning effects in the highly anisotropic superconductor Bi<sub>2+x</sub>Sr<sub>2-x</sub>CaCu<sub>2</sub>O<sub>8+y</sub> (BSCCO) and in YBa<sub>2</sub>Cu<sub>3</sub>O<sub>7-δ</sub> (YBCO) for near optimal oxygen concentrations (i.e. $`\delta 0.05`$, giving the highest value of $`T_c`$). The vortices in BSCCO are often described as stacks of two-dimensional (2D) “pancakes”, which can become misaligned at elevated temperatures and/or in the presence of strong random pinning sites. On the other hand, in “optimally” doped YBCO the vortices behave in a three-dimensional (3D) manner. This difference is attributed to the mass anisotropy $`\gamma =(m_c/m_{ab})^{1/2}=\lambda _c/\lambda _{ab}5`$-$`7`$ in YBCO and $`\gamma 50`$-$`250`$ in BSCCO , where $`\lambda _{ab}`$ and $`\lambda _c`$ are the magnetic penetration depths describing the screening of flux by supercurrents flowing in and out of the CuO<sub>2</sub> layers, respectively. An added complication in many studies of YBCO is the presence of twin-plane boundaries which may act as extended pinning sites for vortices.
The muon spin rotation ($`\mu `$SR) technique measures the muon precession frequency distribution and thus the internal magnetic field distribution $`n(B)`$, also known as the $`\mu `$SR line shape, which has proven to be a powerful tool for investigating VL phases . In optimally doped YBCO, strong local pinning broadens the $`\mu `$SR line shape as predicted for small random displacements of 3D vortex lines . By contrast, random pinning in BSCCO leads to a field-induced reduction of the $`\mu `$SR line width associated with a dimensional crossover from a 3D-VL to a quasi-2D system consisting of independent weakly interacting VLs in different CuO<sub>2</sub> layers. In the quasi-2D region, the repulsive interaction between pancake vortices in the layers exceeds the strength of their interlayer coupling, so that random pinning results in a misalignment of the pancake vortices in the field direction.
The effect on the $`\mu `$SR line shape of thermal fluctuations of the vortex positions has been the focus of numerous studies in BSCCO . As explained in Ref. , the typical time scale for vortex fluctuations is short enough that the muon detects a fluctuation-averaged field. This results in a reduction of both the width and the skewness of the $`\mu `$SR line shape. In clean samples the VL melts in a first-order phase transition (see, for example, Ref. ). Recent $`\mu `$SR measurements in BSCCO have been interpreted as evidence for a two-stage VL transition: first the intralayer coupling of the pancake vortices is overcome by thermal fluctuations, then their interlayer coupling is lost .
In the less anisotropic compound YBCO, a melting transition has been observed by small-angle neutron scattering, magnetization, transport, specific heat and Hall probe ac susceptibility measurements (see, for example, Ref ). However, unlike in BSCCO, the vortex-liquid phase is found only in a very narrow region below the second-order phase transition at $`B_{c2}(T)`$. A melting transition near $`B_{c2}(T)`$ is difficult to study with the $`\mu `$SR technique, because even for a 3D VL, the $`\mu `$SR line shape narrows and becomes more symmetric due to the overlap of the vortices.
It is well established that removal of oxygen from the CuO chain layers in YBCO weakens the coupling between the superconducting CuO<sub>2</sub> layers. Here we report $`\mu `$SR measurements of the VL in the oxygen-deficient compound YBa<sub>2</sub>Cu<sub>3</sub>O<sub>6.60</sub>, which has an increased anisotropy ratio $`\gamma 22`$-$`36`$ (see, Ref. ). We observe an expanded vortex-liquid region and a pinning-induced crossover from a 3D to a quasi-2D system, qualitatively resembling the vortex phase diagram of BSCCO.
The $`\mu `$SR experiments were performed on the M15 and M20 surface muon beamlines at TRIUMF, Canada, using the experimental setup described in Ref. . Measurements of the internal magnetic field distribution were made in both twinned and detwinned crystals of underdoped YBa<sub>2</sub>Cu<sub>3</sub>O<sub>6.60</sub>, with superconducting transition temperatures $`T_c=59(0.1)`$ K. These samples were previously studied with $`\mu `$SR at low temperatures . The crystals were mounted with their $`\widehat{c}`$-axis parallel to both the applied magnetic field $`𝑯`$and the muon beam direction. The positive muons were injected into the sample with their initial spin polarization perpendicular to $`𝑯`$. As described fully elsewhere , $`𝒏\mathbf{(}𝑩\mathbf{)}`$ or the muon precession frequency distribution $`𝒏\mathbf{(}𝝎\mathbf{=}𝜸_𝝁𝑩\mathbf{)}`$, where $`𝜸_𝝁`$ is the muon gyromagnetic ratio, is obtained by monitoring the $`𝒆^\mathbf{+}`$ count rate as the muon decay pattern sweeps by the positron detectors.
Figure 1 shows the evolution of the fast Fourier transform (FFT) of the muon precession signal upon warming the sample, following field cooling to $`𝑻\mathbf{=}\mathbf{2.4}`$ K in a magnetic field $`𝑯\mathbf{=}\mathbf{1.49}`$ T. At $`𝑻\mathbf{=}\mathbf{20.5}`$ K the FFT shows the basic features expected for a well-ordered 3D solid VL — in particular a high-field “tail” associated with the region close to the vortex cores. However, at $`𝑻\mathbf{=}\mathrm{𝟒𝟎}`$ K the FFT is completely symmetric and the line width is drastically reduced — both of which characteristize a vortex-liquid phase. Despite these obvious changes in the $`𝝁`$SR line shape, the melting transition cannot usually be determined accurately by visual inspection, because the output frequency spectrum is artificially broadened by the finite time range and the “apodization” needed to eliminate “ringing” in the FFT . To quantify these changes in the field distribution we calculate the skewness parameter , $`𝜶\mathbf{=}\mathbf{}\mathbf{(}𝜹𝑩\mathbf{)}^\mathrm{𝟑}\mathbf{}^{\mathrm{𝟏}\mathbf{/}\mathrm{𝟑}}\mathbf{/}\mathbf{}\mathbf{(}𝜹𝑩\mathbf{)}^\mathrm{𝟐}\mathbf{}^{\mathrm{𝟏}\mathbf{/}\mathrm{𝟐}}`$, where $`\mathbf{}\mathbf{(}𝜹𝑩\mathbf{)}^𝒏\mathbf{}\mathbf{=}\mathbf{}\mathbf{(}𝑩\mathbf{}\mathbf{}𝑩\mathbf{}\mathbf{)}^𝒏\mathbf{}`$. To obtain reliable values for the moments $`\mathbf{}\mathbf{(}𝜹𝑩\mathbf{)}^𝒏\mathbf{}`$ we fit the $`𝝁`$SR spectra in the time domain to a polarization function calculated assuming a Ginzburg-Landau (GL) model for the VL field profile $`𝑩\mathbf{(}𝐫\mathbf{)}`$, as described in Ref. . The fitted function $`𝑩\mathbf{(}𝐫\mathbf{)}`$ properly accounts for the long high-field “tail” in $`𝒏\mathbf{(}𝑩\mathbf{)}`$ when the VL is 3D. This method of analysis also provides a means of monitoring the $`𝝁`$SR line width through the fitted value of $`\mathrm{𝟏}\mathbf{/}𝝀_{𝒂𝒃}^\mathrm{𝟐}`$. A hexagonal arrangement of vortices was assumed, consistent with recent neutron scattering measurements on untwinned YBCO .
Figure 2 shows the temperature dependence of $`𝜶`$ and $`\mathrm{𝟏}\mathbf{/}𝝀_{𝒂𝒃}^\mathrm{𝟐}`$ at three of the fields considered. The observed values of $`𝜶\mathbf{}\mathbf{1.2}`$ are in agreement with values previously obtained in the 3D-VL phase of BSCCO and that predicted for $`𝒔`$\- and $`𝒅`$-wave VLs at low reduced fields $`𝑩\mathbf{/}𝑩_{𝒄\mathrm{𝟐}}`$ . For temperatures above $`𝑻_𝐦`$ the fitted values of $`\mathrm{𝟏}\mathbf{/}𝝀_{𝒂𝒃}^\mathrm{𝟐}`$ and hence $`𝜶`$ drop to zero. This does not mean that the superfluid density has decreased to zero, but rather that the $`𝝁`$SR spectrum no longer contains a component corresponding to an ordered 3D vortex structure. For $`𝑻\mathbf{>}𝑻_𝐦`$ the $`𝝁`$SR spectra fit well to a single Gaussian function, with a line width much greater than that due to nuclear dipoles.
In Ref. a sharp transition was observed in ac susceptibility measurements on twinned YBa<sub>2</sub>Cu<sub>3</sub>O<sub>6.60</sub> crystals below a characteristic field $`𝑯^{\mathbf{}}\mathbf{}\mathbf{0.07}`$ T, possibly associated with a melting transition. Above $`𝑯^{\mathbf{}}`$, where the transition was found to be continuous, it was suggested that a quasi-2D system exists which is highly sensitive to pinning-induced disorder. In the present study, where $`𝑯\mathbf{>}𝑯^{\mathbf{}}`$, the field distributions for $`𝑻\mathbf{<}𝑻_𝐦`$ and $`\mathbf{0.1}\mathbf{<}𝑯\mathbf{<}\mathbf{1.49}`$ T fit well to the GL model for an ordered 3D VL. The rms deviation of the vortices from their ideal positions in the hexagonal VL was found to be $`\mathbf{<}\mathrm{𝟖}`$ %. These results suggest that the VL for our samples in this region of the $`𝑩`$-$`𝑻`$ phase diagram exhibits a 3D behavior, most likely consisting of stacks of strongly coupled pancake vortices. Above $`𝑻_𝐦`$, thermal fluctuations of the vortices lead to a loss of long-range spatial order.
The top panel of Figure 3 shows the $`𝝁`$SR line shape in the twinned sample of YBa<sub>2</sub>Cu<sub>3</sub>O<sub>6.60</sub> after cooling in a field $`𝑯\mathbf{}\mathbf{2.89}`$ T to $`𝑻\mathbf{=}\mathbf{2.5}`$ K, followed by an increase of $`𝚫𝑯\mathbf{}\mathbf{0.01}`$ T. Notice that the small background signal is positioned at the external field $`𝑯\mathbf{=}\mathbf{2.90}`$ T, whereas the signal originating from the sample looks as though the external field is still $`𝑯\mathbf{=}\mathbf{2.89}`$ T. This implies that the VL is strongly pinned. However, unlike in the line shapes observed at the lower fields, the “tail” appears on the low-field side of the cusp. Numerical calculations performed in Ref. , which account for the sample geometry effect, show that such a line shape can originate from a system of 2D pancake vortices that are ordered in the planes but uncorrelated between adjacent layers. The low-field “tail” is specifically associated with a lower flux density at the sample edges due to a nonuniform demagnetization. Thus clear evidence is observed for a pinning-induced dimensional crossover above a critical field $`𝑩_{\mathrm{𝐜𝐫}}\mathbf{(}𝑻\mathbf{)}`$.
By contrast, at the same field $`𝑯\mathbf{=}\mathbf{2.9}`$ T in the detwinned sample the “tail” extends to the high-field side of the distribution, as expected for an ordered 3D vortex structure. It is tempting to attribute the opposite skewness of the $`𝝁`$SR line shapes in the two samples to the presence of twin planes. However, because the twin planes extend the full depth of the sample, they would displace the 2D VLs by the same amount in all layers. Since the two samples were not from the same growth batch, the difference is more likely related to the concentration and randomness of pointlike defects. This result stresses the sensitivity of the VL structure to disorder at this field.
The middle panel of Fig. 3 shows that the vortices begin to depin at $`𝑻_𝐩\mathbf{}\mathbf{19.8}`$ K due to thermal fluctuations, and reposition themselves with an average field $`𝑩\mathbf{}\mathbf{2.90}`$ T at $`𝑻\mathbf{=}\mathrm{𝟐𝟑}`$ K (bottom panel of Fig. 3). Additional $`𝝁`$SR spectra taken without shifting the field show that the VL melts at $`𝑻_𝐦\mathbf{}\mathrm{𝟐𝟎}`$ K.
Figure 4 summarizes our results in a phase diagram. The upper critical field line represents an approximation assuming $`𝑩_{𝒄\mathrm{𝟐}}\mathbf{(}𝑻\mathbf{)}\mathbf{=}𝑩_𝒄\mathbf{(}\mathrm{𝟎}\mathbf{)}\mathbf{[}\mathrm{𝟏}\mathbf{}\mathbf{(}𝑻\mathbf{/}𝑻_𝒄\mathbf{)}^\mathrm{𝟐}\mathbf{]}`$ and $`𝑩_{𝒄\mathrm{𝟐}}\mathbf{(}\mathrm{𝟎}\mathbf{)}\mathbf{=}\mathrm{𝟕𝟎}`$ T. The data for the twinned and detwinned samples appear to fall on the same melting line. A fit of the data below $`𝑯\mathbf{=}\mathbf{2.9}`$ T to the phenomenological melting curve
$$𝑩_𝐦\mathbf{(}𝑻\mathbf{)}\mathbf{=}\frac{𝑲}{𝝀_{𝒂𝒃}^𝒎\mathbf{(}𝑻\mathbf{)}𝑻}\mathbf{,}$$
(1)
yields $`𝒎\mathbf{=}\mathbf{1.6}\mathbf{(}\mathrm{𝟏}\mathbf{)}`$, where $`𝝀_{𝒂𝒃}\mathbf{=}𝝀_{𝒂𝒃}\mathbf{(}\mathrm{𝟎}\mathbf{)}\mathbf{[}\mathrm{𝟏}\mathbf{}\mathbf{(}𝑻\mathbf{/}𝑻_𝒄\mathbf{)}^\mathrm{𝟒}\mathbf{]}^{\mathbf{}\mathrm{𝟏}\mathbf{/}\mathrm{𝟐}}`$, and $`𝑲`$ and $`𝒎`$ are constants. This is well below either the predicted value $`𝒎\mathbf{=}\mathrm{𝟒}`$ when interlayer coupling of the pancake vortices is dominated by electromagnetic interactions, or $`𝒎\mathbf{=}\mathrm{𝟑}`$ when Josephson coupling becomes important . As discussed in Refs. , for finite Josephson coupling an exponent $`𝒎\mathbf{=}\mathrm{𝟐}`$ is expected for an additional thermodynamic transition at $`𝑩\mathbf{<}𝑩_{\mathrm{𝐜𝐫}}`$ which decouples the layers in the liquid phase. The decoupling line is predicted to be
$$𝑩_{\mathrm{𝐝𝐜}}^𝐉\mathbf{(}𝑻\mathbf{)}\mathbf{=}\frac{𝚽_\mathrm{𝟎}^\mathrm{𝟑}𝒄_𝑫}{\mathrm{𝟒}𝝅𝝁_\mathrm{𝟎}𝒌_𝑩𝒔𝜸^\mathrm{𝟐}}\frac{\mathrm{𝟏}}{𝝀_{𝒂𝒃}^\mathrm{𝟐}}𝑻\mathbf{,}$$
(2)
where $`𝒄_𝑫\mathbf{}\mathbf{0.1}`$ is a decoupling constant and $`𝒔`$ is the interlayer spacing. The dashed curve in Fig. 4 is a fit to Eq. (2). Taking $`𝒔\mathbf{=}\mathbf{11.7}`$ Å, $`𝝀_{𝒂𝒃}\mathbf{(}\mathrm{𝟎}\mathbf{)}\mathbf{=}\mathrm{𝟏𝟔𝟒𝟐}`$ Å and $`𝜸\mathbf{=}\mathrm{𝟐𝟐}`$-$`\mathrm{𝟑𝟔}`$ , we calculate $`𝒄_𝑫\mathbf{=}\mathbf{0.025}`$-$`\mathbf{0.067}`$ from the fit. A value $`𝒄_𝑫\mathbf{=}\mathbf{0.076}`$ was obtained for BSCCO in Ref. . To determine whether the melting transition (i.e. to a liquid of vortex lines) coincides with the low-field interlayer decoupling transition, more measurements in the vicinity of the phase transition are needed. We note that, contrary to the conclusion in Ref. , recent detailed $`𝝁`$SR measurements in BSCCO suggest that $`𝑩_𝐦\mathbf{(}𝑻\mathbf{)}\mathbf{<}𝑩_{\mathrm{𝐝𝐜}}^𝐉\mathbf{(}𝑻\mathbf{)}`$.
In conclusion, we have observed changes in the $`𝝁`$SR line shape of underdoped YBa<sub>2</sub>Cu<sub>3</sub>O<sub>6.60</sub> which identify a VL melting transition far below that of optimally doped YBCO. More precisely, fits to the data suggest that there is a decoupling transition from a liquid of vortex lines to a liquid of 2D vortices. Our measurements also establish the existence of a pinning-induced dimensionl crossover to a quasi-2D vortex system, similar to that observed in the highly anisotropic material BSCCO.
The work at TRIUMF was supported by NSERC Canada and US AFOSR Grant No. F49620-97-1-0297. The work at Los Alamos was performed under the auspicies of the US Department of Energy.
FIGURE CAPTIONS
Figure 1. Fourier transform of the muon precession signal in twinned YBa<sub>2</sub>Cu<sub>3</sub>O<sub>6.60</sub> at $`𝑻\mathbf{=}\mathbf{20.5}`$ K and $`𝑻\mathbf{=}\mathrm{𝟒𝟎}`$ K, in a field $`𝑯\mathbf{=}\mathbf{1.49}`$ T.
Figure 2. Temperature dependence of the skewness $`𝜶`$ \[top\] and $`\mathrm{𝟏}\mathbf{/}𝝀_{𝒂𝒃}^\mathrm{𝟐}`$ \[bottom\] in detwinned YBa<sub>2</sub>Cu<sub>3</sub>O<sub>6.60</sub> at $`𝑯\mathbf{=}\mathbf{1.25}`$ T \[crosses\] and in twinned YBa<sub>2</sub>Cu<sub>3</sub>O<sub>6.60</sub> at applied magnetic fields $`𝑯\mathbf{=}\mathbf{0.74}`$ \[open circles\] and $`\mathbf{1.49}`$ T \[solid circles\].
Figure 3. Fourier transforms of muon precession signals in twinned YBa<sub>2</sub>Cu<sub>3</sub>O<sub>6.60</sub>. Top panel: after field cooling at $`𝑯\mathbf{}\mathbf{2.89}`$ T to $`𝑻\mathbf{=}\mathbf{2.5}`$ K followed by an increase in the field to 2.90 T. Middle and bottom panels: after subsequently warming the sample to $`𝑻\mathbf{=}\mathbf{19.8}`$ K and $`\mathbf{23.0}`$ K, respectively.
Figure 4. The vortex $`𝑩`$-$`𝑻`$ phase diagram for YBa<sub>2</sub>Cu<sub>3</sub>O<sub>6.60</sub>. Open and solid circles correspond to the twinnned and detwinned samples, respectively. The dashed curve $`𝑩_{\mathrm{𝐝𝐜}}^𝐉\mathbf{(}𝑻\mathbf{)}`$ is a fit to Eq. (2), the solid curve $`𝑩_{𝒄\mathrm{𝟐}}\mathbf{(}𝑻\mathbf{)}`$ is a theoretical line for the upper critical field (see text) and the solid curve $`𝑩_𝐦\mathbf{(}𝑻\mathbf{)}`$ is a fit of all of the data to Eq. (1). The transition line $`𝑩_{\mathrm{𝐜𝐫}}\mathbf{(}𝑻\mathbf{)}`$ is approximate.
|
no-problem/9906/cond-mat9906434.html
|
ar5iv
|
text
|
# 1 The Percolation model of stock market prices
## 1 The Percolation model of stock market prices
A wealth of models (to our knowledge, the first stock market simulation was performed by the economist Stigler in 1964 ), partially listed in , have been introduced in the financial and more recently in the physical community which attempt to capture the complex behavior of stock market prices and of market participants. Based on the competition between supply and demand, the effort is to model the main observed stylized facts: absence of two-point correlation of the returns , fat tail distribution of returns (probabilities higher than Gaussian ) and long-range volatility (standard deviation) correlations . The goal is to have, on the one hand, the simplest and most parsimonious description of the market and, on the other hand, the most faithful representation of the observed market characteristics.
In this spirit, Cont and Bouchaud introduced a percolation model which assumes that investors can be classified into groups (clusters) of the same opinion occurring with many different sizes. The simplest recipe to aggregate interacting or inter-influencing traders into groups is to assume that the connectivity between traders defining the groups can be seen as a pure geometrical percolation problem with fixed occupancy on a given network topology. Clusters are groups of neighboring occupied sites or investors. Then, random percolation clusters make a decision to buy or sell on the stock market, for all sites (corresponding to the individual investors and units of wealth) in that cluster together. Thus, the individual investors are thought to cluster together to form companies or groups of influence, which under the guidance of a single manager buy (probability $`a`$), sell (probability $`a`$), or refrain from trading (probability $`12a`$) within one time interval. The traded amount is proportional to the number $`s`$ of sites in the cluster, and
$$\mathrm{the}\mathrm{logarithm}\mathrm{of}\mathrm{the}\mathrm{price}\mathrm{changes}\mathrm{proportionally}\mathrm{to}$$
$$\mathrm{the}\mathrm{difference}\mathrm{\Delta }\mathrm{between}\mathrm{demand}\mathrm{and}\mathrm{supply}.$$
(1)
When the activity $`a`$ is small, at most one cluster trades at a time. As a consequence, the distribution $`P(R)`$ of relative price changes or “returns” $`R`$ scales as the well-known cluster size distribution $`n_s(p)`$ of percolation theory. In contrast, for large activity $`a`$ and without an infinite cluster, the relative price variation is the contribution (sum) of many clusters and the central limit theorem implies that the distribution $`P(R)`$ converges to the Gaussian law for large systems (except exactly at the critical point $`p_c`$).
For low activity $`a`$ and right at the site percolation threshold $`p=p_c`$, when the fraction $`p`$ of lattice sites occupied by an investor in a $`d`$-dimensional lattice of linear extent $`L`$ barely suffices to form an “infinite” cluster stretching from top to bottom, we observe power laws:
$$n_ss^\tau ,P(R)R^\tau \mathrm{for}1sL^D\mathrm{where}D=d/(\tau 1)$$
(2)
is the fractal dimension of the percolating cluster.
For concentrations $`p`$ different from $`p_c`$, the cluster numbers decay as an exponential (resp. stretched exponential) for $`p<p_c`$ (resp. for $`p>p_c`$). Their typical size $`s^{}`$ (not counting the “infinite” percolating cluster) is much smaller than at $`p=p_c`$ where it is solely controlled by the system size (i.e. total number of traders). Since a cluster of size roughly comparable to the total system size appears at and above the percolation threshold $`p_c`$, this value corresponds to a big crash in the market, and the region of $`p`$ below $`p_c`$ gives less volatile behavior.
Correlations of volatility in time are produced by letting the occupied sites (traders) diffuse slowly to empty neighbor sites on a lattice to reform or destroy new alliances. The volatility, i.e. the typical absolute value of the return, thus behaves similar to a mean cluster size and is correlated in time due to the slow diffusion with a typical decay slower than exponential.
This model at the percolation threshold thus agrees qualitatively (but not quantitatively especially on the exponents as discussed below) with the three stylized facts of real markets : the average return $`R`$ is zero (if inflation and other regular trends are subtracted); the return distribution $`P(R)`$ decays as a power law $`R^\mu `$ for intermediate $`R`$ with $`\mu 4`$ (the probability $`P_>(R)`$ of finding a change larger than $`R`$ then varies as $`R^{1\mu }`$), and the volatility
$$V(t)=<R(t)^2>^{1/2}$$
(3)
clusters in time in the sense that its autocorrelation function
$$C(\tau )=<\mathrm{\Delta }V(t)\mathrm{\Delta }V(t+\tau )>,\mathrm{with}\mathrm{\Delta }V=V<V>,$$
(4)
is positive and decays slowly to zero.
Two disadvantages of the model are:
* why should markets know the percolation threshold and work at $`p=p_c`$?
* How can a value of $`\mu =\tau `$ be nearly $`4`$ when $`\tau `$ varies only from 2 to 2.5 if the dimensionality $`d`$ increases from $`2`$ to infinity ?
A mechanism for self-organized criticality like invasion percolation would only solve the first and not the second problem and would be difficult to justify form an economic view point.
## 2 A Simple Self-Organizing Market
### 2.1 Percolation connectivity evolving with time
We thus return to an alternative mechanism which gives power laws without the need to tune $`p`$ to $`p_c`$ and which is very robust and simple. The new idea we propose in this context is that there is no reason a priori to expect that the parameter $`p`$ controlling the connectivity/influence between traders is fixed. The circle of professionals and colleagues to whom a trader is typically connected evolves as a function of time, not only in its structure at fixed average number of connections (corresponding to the diffusion effect discussed above) but also, in the average strength and number of interactions: at some times, traders are following strong herding behavior and the effective connectivity parameter $`p`$ is high; at other times, investors are more individualistic and smaller values of $`p`$ seem more reasonable. In order to take into account the complex dynamics of the network of interactions between traders, it thus seems reasonable to relax the hypothesis that $`p`$ is fixed at a given value but rather evolves with its own dynamics. The simplest version is to assume that $`p`$ is taken purely random at each time step. As a consequence, the distribution of relative price changes will be an average over those obtained for each sampled $`p`$’s. Averaging over an interval in $`p`$ containing the percolation threshold $`p_c`$ will give its main contribution to the number of large clusters from a narrow region (width $`1/s^\sigma `$) about $`p_c`$, and thus lead to an integrated cluster numbers $`s^{\tau \sigma }`$, where $`\sigma `$ varies from $`0.4`$ to $`0.5`$ if $`d`$ varies from $`2`$ to infinity. Now $`\mu =\tau +\sigma `$ varies from $`2.45`$ to $`3`$, closer to reality ($`\mu 4`$).
We stress that, as soon as $`p`$ samples an interval containing or close to $`p_c`$, the distribution of returns is a power law with no other truncation than given by the finite size of the total system. We thus obtain a robust critical behavior without any artificial adjustement of the connectivity parameter $`p`$. Notice that this mechanism in terms of a “sweeping of an instability” is different from what is usually called self-organized criticality which involves a dynamical feedback attracting the system dynamics to a dynamical critical point.
We thus distribute randomly our investors on the $`L\times L`$ square lattice, with concentration $`p`$. We sum up all the results obtained by varying $`p`$ in steps of one percent, from 1 to 59 percent where the percolation threshold $`p_c=0.593`$ is reached. For each concentration, we make 1000 iterations where at each iteration one percent of the investors try to move to a randomly selected neighbor site. For each cluster configuration obtained in this way, we sum over 1000 different realizations of buying and selling decisions of the cluster, which thus allows much better averaging than in real markets where history cannot be repeated so easily. Many such simulations are averaged over to give smooth results.
Fama has argued that the crash of Oct. 1987 on the US and other stock markets worldwide could be seen as the signature of an efficient reassessement of and convergence to the correct “fundamental” price after the long speculative bubble preceeding it. In this spirit, we assume that, as $`p`$ reaches the crash concentration $`p=p_c`$, “everything” changes and the network connectivity is afterwards reinitialized at a value $`p<p_c`$; therefore no data for $`p>p_c`$ are given here.
Fig.1 shows for the square lattice ($`d=2`$) the distribution of returns $`P(R)`$. The simulations confirm, for a range of about five orders of magnitude in $`P`$ similar to the range observed by Gopikrishnan et al , the predicted power law $`P(R)1/R^\mu `$ at intermediate $`R`$ with an exponent $`\mu 2.5`$. For the largest $`R`$, finite size effects reduce $`P(R)`$ and, for small $`R`$, the probability is roughly constant. Increasing the linear lattice size $`L`$ from 31 via 101 to 301 shifts the power-law region to larger $`R`$ without changing the effective exponent.
### 2.2 Size-dependent activity
The numerical deviation from the empirically observed $`\mu =4`$ is thus large, and a different approach is needed. Instead of taking the activity $`a`$ as a free parameter between zero and $`1/2`$ ($`0.005`$ in Fig.1) and the same independently of the cluster size, we assume the following size dependence
$$a=0.5/\sqrt{s},$$
(5)
thus getting rid of one free parameter. A priori, it is reasonable to consider that the big investors, such as the mutual and/or retirement funds with their prudent approach, their emphasis on low risk, and their enormous inertia due to the fact that large positions move the market unfavorably, have to and do trade less often than small professional investors who have to generate their income from active trading rather than from sheer mass. In this spirit, recent works have documented that the growth dynamics of business firms , the economies of countries and the university research activities depend on size, the smaller entities being the most active proportionally. Another not necessarily exclusive mechanism is that, within a large cluster, the $`s`$ investors have to agree by some majority to buy and sell, and do not trade if no such majority is reached. A random decision process then could lead to the square-root behavior given by (5).
With this modification, the exponent $`\mu `$ is predicted to be
$$\mu =\tau +\sigma +1/23.$$
(6)
In contrast, we measure in Fig.2 an effective exponent in the intermediate $`R`$ range equal to $`3.5`$, larger than the asymptotically expected value given by eq.(6) and close to the empirical value near $`4`$. The volatility clustering is not destroyed by our change, and Fig.3 shows the volatility auto-correlations to decay slowly towards zero, also in agreement with empirical facts.
Since eq.(4) implies a zero probability of the “infinite” cluster to act for an “infinite” system above $`p_c`$, we now can also integrate over the whole interval of $`p`$ by summing from 1 to 99 percent. Fig.4 documents a new phenomenon, namely wings with rare price changes determined mainly by large clusters containing nearly the whole lattice and choosing randomly to buy, sell, or sleep. The larger the lattice is, the smaller is the overall absolute weight of these wings. However, relative to the power law represented by the dashed straight line, the larger the lattice, the larger is the relative weight of the wings. This means that these large price changes are more and more “outliers” of the power law statistics holding for intermediate price changes.
To understand this observation, recall first that the connectivity parameter $`p`$ goes from a small number ($`1\%`$ in the simulations) to a larger number ($`99\%`$ in the simulations) above $`p_c`$ and the observed distribution of log-price changes is mapped one-to-one onto the distribution $`P_{\mathrm{sum}}`$ of cluster sizes obtained by averaging over all the cluster size distributions obtained for $`p`$ taken with equal weight between $`1\%`$ and $`99\%`$. If the $`n_s`$ of eq.(2) are the cluster numbers per lattice site, then $`L^2n_s`$ are those in the whole lattice and vary at $`p=p_c`$ as $`L^2s^\tau `$. The contribution $`P_{sum}(s)`$ of $`s`$-clusters averaged over all $`p`$, to the price changes reads
$$P_{\mathrm{sum}}L^2\frac{1}{s^{\tau +\sigma +1/2}}.$$
(7)
The two additional terms $`\sigma +1/2`$ in the exponent of (7) stem, as discussed above, from the two effects of “sweeping” of $`p`$ over $`p_c`$ and of the inverse square root dependence of the activity on the cluster size.
Now, extrapolating (7), we can estimate the probability $`cL^2P_{sum}(cL^2)`$ that this power law would predict for getting a cluster of size of order $`L^2`$ and compare it to the true probability of getting a cluster of this size. The factor $`cL^2`$ comes from the fact that one must count the large clusters around a typical size with proportional fluctuations, thus giving a true probability while $`P_{sum}(cL^2)`$ is the probability density to observe a cluster of size $`cL^2`$. From (7), we get
$$cL^2P_{sum}(cL^2)(cL^2)L^2\frac{1}{(cL^2)^{\tau +\sigma +1/2}},$$
(8)
where $`c`$ is a number of order unity. Counting powers of $`L`$ in (8) and expressing the result for two dimensions gives
$$cL^2P_{sum}L^{42\tau 2(1/2+\sigma )}=\frac{1}{L^{1.9}}.$$
(9)
In contrast, we know that there is exactly one infinite cluster in a quadratic lattice above $`p_c`$. Since the $`p`$’s are uniformely taken between $`1\%`$ and $`99\%`$, this shows directly that the probability to get a cluster close to the maximum size $`L^2`$ is a constant fraction, independently of the lattice size $`L`$, and its contribution to a price change is multiplied by its activity $`1/L`$. (Indeed our data of Fig.4 give a value near $`0.5/L`$ for the fraction of returns larger than $`L^2/2`$.) This argument thus shows that the ratio of the true frequency to observe the large “outlier” (stemming from the infinite cluster truncated to the size of the lattice above $`p_c`$) is larger than the extrapolation of the power distribution of intermediate cluster size by a factor $`L^{0.9}`$ which increases with the system size $`L`$, in qualitative agreement with Fig.4.
We suggest that the large wings might correspond to the “outliers” in the stock market like the Wall Street crashes of 1929 and 1987 . The normal autocorrelation functions of the volatility exhibiting long range dependence then apply to normal times on the stock market when no such outliers are relevant, and they are destroyed by the outliers. The wings vanish and the autocorrelations are restored if we follow and omit the largest cluster from the market.
### 2.3 Nonlinear price change dependence
All our previous results derive from the assumption (1) that the change of (the logarithm of the) price is proportional to the difference between supply and demand. This assumption is often made and can be in fact derived rigorously from the two assumptions that it is not possible to make profits by repeatedly trading through a circuit and that the ratio of prices before and after a transaction is a function of the difference $`\mathrm{\Delta }`$ between demand and supply alone.
However, many recent empirical studies suggest that the relationship between the change of the logarithm of price and $`\mathrm{\Delta }`$ is highly nonlinear, especially for large orders . Assuming that the time needed to complete a trade of size $`s`$ is proportional to $`s`$ and that the unobservable price fluctuations obey a diffusion process during that time, Zhang derives the relationship that the change of the logarithm of the price is proportional to the square root of the difference $`\mathrm{\Delta }`$ between demand and supply , i.e. to the square root of $`s`$ in our present formulation. This modifies all previous results as follows.
The result (2) for the “pure” percolation model becomes
$$n_ss^\tau ,P(R)R^\mu \mathrm{for}1sL^D\mathrm{where}R\sqrt{s},$$
(10)
giving with $`ds/dRR`$ (with numerical estimates in two dimensions) the exponent $`\mu =2\tau 1`$ around $`3.1`$, still smaller than the empirical value close to $`4`$.
The result obtained by the “sweeping” of the connectivity parameter $`p`$ transforms $`\mu `$ from $`\mu =\tau +\sigma `$ into $`\mu =2\tau 1+\sigma `$, giving a value $`3.5`$. Next, incorporating the size dependence (5) of the activity leads to the prediction $`\mu =2\tau +\sigma =4.5`$, in rough agreement with the empirical value 4.
We may even omit the size-dependent activity and use only this nonlinear price change dependence and $`0<p<p_c`$. Then the data of Fig.1 are transformed, without any additional simulations, into those of Fig.5 which give an effective $`\mu 3.9`$ in better agreement with the real $`\mu 4`$ than the theoretical prediction $`\mu =2\tau 1+\sigma 3.5.`$
## 3 Concluding Remarks
We have presented what we believe is probably the simplest and most robust model of stock market dynamics without tunable parameters that self-organizes into a regime where the most important empirical characteristics of stock market price dynamics are captured.
In this simplest version, we have chosen the most straightforward dynamics of the interaction/connectivity parameter $`p`$, i.e. a continuous increase up to the critical value $`p_c`$ followed by a reset to a low value and so on. Incorporating a size-dependence of the cluster activities has allowed us to let $`p`$ larger than $`p_c`$ for which we have documented the appearance of outliers corresponding to the infinite cluster truncated to the size of the lattice. This outlier might correspond to the large crashes observed in this century. A good agreement with empirical data is obtained alternatively by allowing for a nonlinear dependence of the change of (the logarithm of the) price as a function of the difference between supply and demand.
A random evolution of $`p`$, either pure white noise, or a random walk or with more correlation are interesting to investigate in the future, but will not change the most fundamental finding presented here of a power law distribution and long-range correlations of the volatility. More interesting presumably would be a dynamic of $`p`$ coupled to that of the price change, simulating the tendency to join a bullish market . Note also that the present model is by construction up-down symmetric, which means that rallies appear as often statistically and in the same shape as crashes. There is not sharp peak versus flat trough asymmetry . Such asymmetry can be easily incorporated by letting the trading activity be dependent on the function price(time), i.e. increasing prices causes more people to act than a decreasing price, but we have not persued this as this would imply adding novel ingredients in a model we have on purpose kept bare to its skeleton.
Ackowledgements: This idea originated at the meeting “Facets of Universality in Complex Systems: Climate, Biodynamics and Stock Markets”, organized at Giessen University by Armin Bunde and John Schellnhuber (June 1999). We thank NIC Jülich for time on their Cray-T3E, T.Lux for ref.3, and A. Johansen for comments. One of us (DS) wishes to point out that all errors are due to the other author (DS).
Captions:
Fig.1: Return distribution at constant activity $`a=0.005`$. The axis of relative price variations is scaled such that a buying cluster of $`s`$ investors produces an increase of the price by $`s`$.
Fig.2: Return distribution with activity decaying as $`1/\sqrt{s}`$.
Fig.3: Volatility autocorrelation function $`C(T)`$ versus time lag $`T`$ for the simulations of Fig.2.
Fig.4: As Fig.2, but with the cluster connectivity parameter $`p`$ varying from $`1`$ to $`99`$ percent.
Fig.5: Data from Fig.1 replotted by assuming a price change proportional to the square root of the (absolute value of the) difference $`\mathrm{\Delta }`$ between demand and supply (and with sign opposite to that of $`\mathrm{\Delta }`$).
|
no-problem/9906/gr-qc9906099.html
|
ar5iv
|
text
|
# Gravitating monopole and its black hole solution in Brans-Dicke Theory
## I Introduction
After the discovery of a particlelike solution in the Einstein-Yang-Mills (EYM) system by Bartnik and Mckinnon , a variety of particlelike solutions and black hole solutions with non-Abelian fields were found. Some of them are the solutions in which the self-gravity is taken into account for the solution in flat space-time. The others are new types of solutions which can exist in the presence of gravity. They show interesting properties, which can not be seen in the flat spacetime solutions or well known Kerr-Newman black hole, such as stability, thermodynamical properties, mass inflation phenomena and so on. In the context of a counterexample for the black hole no-hair conjecture, a monopole black hole may be the most important among them, since both the Reissner-Nortström (RN) black hole and the monopole black hole with the same mass and the gauge charge are stable within a certain parameter region and it is constructed in the EYM-Higgs (EYMH) system which is not a phenomenological model such as the Skyrme and the Proca model. The monopole black hole solution describes a space-time structure with an event horizon in the gravitating ’t Hooft-Polyakov magnetic monopole. The existence of such a black hole solution was first expected by examining a simple distribution of the energy density, where it is almost constant inside a core and decays as $`r^4`$ outside. Later Breitenlohner et al. , Aichelburg and we found the “exact” solutions numerically.
In the big bang scenario, the ’t Hooft-Polyakov monopole is overproduced during phase transitions in the early universe. This is one of the reasons why the inflationary scenario was proposed. Although such an inflationary scenario was discussed originally in general relativity (GR), it turns out that introduction of a scalar field (such as the Brans-Dicke (BD) scalar field) can make a great change in scenarios of the very early universe. Moreover, Vilenkin and Linde proposed that inflation occurs inside the ’t Hooft-Polyakov monopole, where no fine tuning of the initial value is necessary. Such a monopole inflation is also changed qualitatively by introducing a scalar field. So properties of the gravitating monopole and its black hole solution may be greatly modified in scalar-tensor theory compared with in GR.
As for the dilatonic black holes in an effective theory of superstring, it is known that such a scalar field also affects many features. In BD theory, it is also known that the Kerr or Kerr-Newman black hole is a unique solution in the vacuum case or in the case with the Maxwell field, because of the conformal invariance of the Maxwell field and the black hole uniqueness theorem. Recently, however, we found non-trivial black hole solutions with the non-Abelian field in BD theory. They are generalizations of neutral non-Abelian black holes in GR (Proca and Skyrme black holes). Then, it is important to study a self-gravitating monopole and its black hole solution in BD theory as the charged solution. In addition, we have the following reasons to consider the monopole black hole in BD theory. (i) For the neutral non-Abelian black holes, we can not find a more stable solution than a trivial solution, i.e., Schwarzschild solution even in BD theory. We expect that the monopole black hole can be more stable than a trivial solution, i.e., Reissner-Nortström (RN) solution in the BD-YMH system. So it would be more important than the neutral type. Particularly, when we discuss the Hawking radiation of the RN black hole in the EYMH system, it may evolve into the monopole black hole and eventually into the gravitating monopole, which would be a candidate for its remnant. The scalar field may cause a serious effect on this evolution. (ii) To investigate the stability of the neutral non-Abelian black holes, we can apply catastrophe theory even in BD theory. Using the variables in the Einstein frame, we find that the BD scalar field is less important from the perspective of catastrophe theory. However the solution is not evident in the charged monopole black hole case.
This paper is organized as follows. We introduce basic $`Ans\ddot{a}tze`$ and the field equations in the BD-YMH system in Sec. II. In Sec. III, we present numerical solutions for the regular self-gravitating monopole. In order to see their structure, we show their field configurations and compare them with the solution in GR. We also discuss the upper limit of the vacuum expectation value of the Higgs field for a solution to exist. In Sec. IV, we present the monopole black hole solution. In some parameter regions, there exist two types of non-trivial solutions, one of which is stable and the other is not. We show that the dependence of field configurations on the BD parameter is different in each type. This shows the difference of the black hole structures which can not be seen only by considering them in GR. We also discuss upper limits for the vacuum expectation value of the Higgs field for a static solution to exist in black hole cases. The concluding remarks will follow in Sec. V. Throughout this paper we use units of c=$`\mathrm{}`$=1. Notations and definitions follow Misner-Thorne-Wheeler .
## II Basic equations
The action of BD theory is written as
$`S={\displaystyle d^4x\sqrt{g}\left[\frac{1}{2\kappa ^2}\left(\varphi R\frac{\omega }{\varphi }_\alpha \varphi ^\alpha \varphi \right)+L_m\right]},`$ (1)
where $`\kappa ^28\pi G`$ with $`G`$ being Newton’s gravitational constant. $`\omega `$ is the BD parameter, and $`L_m`$ is Lagrangian of the matter fields. The dimensionless BD scalar field $`\varphi `$ is normalized by $`G`$. In our previous paper, we investigated two types of non-Abelian black hole solutions in BD theory (Proca black hole and Skyrme black hole) and found that there are several advantages to discussion in the Einstein conformal frame rather than in the BD frame. For example, the definition of thermodynamical variables of the solutions and the stability analysis using catastrophe theory. Although, of course, we can reach the same results in the BD frame, we transform the Einstein frame as a matter of convenience. The conformal transformation in the present model is written by
$`\widehat{g}_{ab}={\displaystyle \frac{\varphi }{\varphi _0}}g_{ab}.`$ (2)
Then, we find the equivalent action $`\widehat{S}=S/\varphi _0`$ given as
$$\widehat{S}=d^4x\sqrt{\widehat{g}}\left[\frac{1}{2\kappa ^2}\widehat{R}\frac{1}{2}\widehat{}_\alpha \phi \widehat{}^\alpha \phi +\frac{1}{\varphi _0}\widehat{L}_m\right],$$
(3)
where
$$\phi \frac{1}{\kappa \beta }\mathrm{ln}\left(\frac{\varphi }{\varphi _0}\right),$$
(4)
$$\beta \left(\frac{2\omega +3}{2}\right)^{1/2}.$$
(5)
Variables with a caret $`\widehat{}`$ denote those in the Einstein frame. $`\widehat{L}_m`$ is different from $`L_m`$ unless the matter field is conformally invariant, as with the Yang-Mills field. The constant $`\varphi _02(2+\omega )/(2\omega +3)`$ is the value of the BD scalar field $`\varphi `$ at spatial infinity, which guarantees that we can regard $`G`$ as Newton’s gravitational constant. In this paper we investigate the YMH system, where Lagrangian $`\widehat{L}_m`$ is written as
$$\widehat{L}_m=\frac{1}{4}F_{\mu \nu }^aF^{a\mu \nu }\frac{1}{2}e^{\kappa \beta \phi }(D_\mu \mathrm{\Phi }^a)(D^\mu \mathrm{\Phi }^a)\frac{\lambda }{4}e^{2\kappa \beta \phi }(\mathrm{\Phi }^a\mathrm{\Phi }^av^2)^2$$
(6)
$`F_{\mu \nu }^a`$ is the field strength of the SU(2) YM field and expressed by its potential $`A_\mu ^a`$ as
$$F_{\mu \nu }^a=_\mu A_\mu ^a_\nu A_\nu ^a+eϵ^{abc}A_\mu ^bA_\nu ^c,$$
(7)
with the gauge coupling constant $`e`$. $`\mathrm{\Phi }^a`$ is the real triplet Higgs field and $`D_\mu `$ is the covariant derivative:
$$D_\mu \mathrm{\Phi }^a=_\mu \mathrm{\Phi }^a+eϵ^{abc}A_\mu ^b\mathrm{\Phi }^c.$$
(8)
The theoretical parameters $`v`$ and $`\lambda `$ are the vacuum expectation value and the self-coupling constant of the Higgs field, respectively. Below we consider them only in the Einstein conformal frame, and drop the caret $`\widehat{}`$.
We assume that a space-time is static and spherically symmetric, in which case the metric is written as
$$ds^2=\left[1\frac{2Gm(r)}{r}\right]e^{2\delta (r)}dt^2+\left[1\frac{2Gm(r)}{r}\right]^1dr^2+r^2d\mathrm{\Omega }^2.$$
(9)
For the matter fields, we adopt the hedgehog ansatz given by
$$\mathrm{\Phi }^a=v𝒓^ah(r),$$
(10)
$$A_{\mathrm{\hspace{0.33em}0}}^a=0,$$
(11)
$$A_\mu ^a=\omega _\mu ^cϵ^{acb}𝒓^b\frac{1w(r)}{er},(\mu =1,2,3),$$
(12)
where $`𝒓^a`$ and $`\omega _\mu ^c`$ are a unit radial vector in the internal space and a triad, respectively.
Variation of the action (3) with the matter Lagrangian (6) leads to the field equations
$$\frac{d\delta }{d\stackrel{~}{r}}=\frac{8\pi \stackrel{~}{r}\stackrel{~}{v}^2\stackrel{~}{K}}{\varphi _0},$$
(13)
$$\frac{d\stackrel{~}{m}}{d\stackrel{~}{r}}=\frac{4\pi \stackrel{~}{r}^2\stackrel{~}{v}^2}{\varphi _0}\left[f\stackrel{~}{K}+\stackrel{~}{U}\right],$$
(14)
$$\frac{d^2w}{d\stackrel{~}{r}^2}=\frac{1}{f}\left[\frac{1}{2}\frac{\stackrel{~}{U}}{w}+\frac{8\pi \stackrel{~}{r}\stackrel{~}{v}^2\stackrel{~}{U}}{\varphi _0}\frac{dw}{d\stackrel{~}{r}}\frac{2\stackrel{~}{m}}{\stackrel{~}{r}^2}\frac{dw}{d\stackrel{~}{r}}\right],$$
(15)
$$\frac{d^2h}{d\stackrel{~}{r}^2}=\frac{dh}{d\stackrel{~}{r}}\left(\beta \frac{d\stackrel{~}{\phi }}{d\stackrel{~}{r}}\frac{1}{\stackrel{~}{r}}\right)+\frac{1}{f}\left[e^{\beta \stackrel{~}{\phi }}\frac{\stackrel{~}{U}}{h}+\frac{8\pi \stackrel{~}{r}\stackrel{~}{v}^2\stackrel{~}{U}}{\varphi _0}\frac{dh}{d\stackrel{~}{r}}\frac{1}{\stackrel{~}{r}}\frac{dh}{d\stackrel{~}{r}}\right],$$
(16)
$$\frac{d^2\stackrel{~}{\phi }}{d\stackrel{~}{r}^2}=\frac{2}{\stackrel{~}{r}}\frac{d\stackrel{~}{\phi }}{d\stackrel{~}{r}}\frac{4\pi \beta \stackrel{~}{v}^2e^{\beta \stackrel{~}{\phi }}}{\varphi _0}\left(\frac{dh}{d\stackrel{~}{r}}\right)^2+\frac{1}{f}\left[\frac{8\pi \stackrel{~}{v}^2}{\varphi _0}\left(\frac{\stackrel{~}{U}}{\stackrel{~}{\phi }}+\stackrel{~}{r}\stackrel{~}{U}\frac{d\stackrel{~}{\phi }}{d\stackrel{~}{r}}\right)\frac{2\stackrel{~}{m}}{\stackrel{~}{r}^2}\frac{d\stackrel{~}{\phi }}{d\stackrel{~}{r}}\right],$$
(17)
where
$$f1\frac{2\stackrel{~}{m}}{\stackrel{~}{r}},$$
(18)
$$\stackrel{~}{U}\frac{(1w^2)^2}{2\stackrel{~}{r}^4}+e^{\beta \stackrel{~}{\phi }}\left(\frac{wh}{\stackrel{~}{r}}\right)^2+\frac{\stackrel{~}{\lambda }}{4}e^{2\beta \stackrel{~}{\phi }}(h^21)^2,$$
(19)
$$\stackrel{~}{K}\frac{1}{\stackrel{~}{r}^2}\left(\frac{dw}{d\stackrel{~}{r}}\right)^2+\frac{e^{\beta \stackrel{~}{\phi }}}{2}\left(\frac{dh}{d\stackrel{~}{r}}\right)^2+\frac{\varphi _0}{16\pi \stackrel{~}{v}^2}\left(\frac{d\stackrel{~}{\phi }}{d\stackrel{~}{r}}\right)^2.$$
(20)
We have introduced the following dimensionless variables:
$$\stackrel{~}{r}=evr,\stackrel{~}{m}=Gevm,\stackrel{~}{\phi }=\kappa \phi ,$$
(21)
and dimensionless parameters:
$$\stackrel{~}{v}=v\sqrt{G},\stackrel{~}{\lambda }=\lambda /e^2.$$
(22)
Although the solution exists when $`vM_{Pl}`$, where $`M_{Pl}`$ is Planck mass, it can be described by a classical field configuration in the limit of a weak gauge coupling constant $`e`$, because its Compton wavelength $`e/v`$ is much smaller than the radius of the classical monopole solution $`1/ev`$ in this case. Moreover, since the energy density $`e^2v^4M_{Pl}^4`$, we can treat the system classically.
The boundary conditions at spatial infinity are
$$\underset{r\mathrm{}}{lim}m=M<\mathrm{},$$
(23)
$$\underset{r\mathrm{}}{lim}\delta =0,$$
(24)
$$\underset{r\mathrm{}}{lim}\kappa \beta \phi =\underset{r\mathrm{}}{lim}\mathrm{ln}\left(\frac{\varphi }{\varphi _0}\right)=0,$$
(25)
$$\underset{r\mathrm{}}{lim}h=1,$$
(26)
$$\underset{r\mathrm{}}{lim}w=0.$$
(27)
These conditions imply that the space-time approaches a flat one and that the solution is a charged object. The field equations are singular at the origin and at the event horizon for a self-gravitating monopole and its black hole solution, respectively.
For a self-gravitating monopole solution, we require regularity at the origin. Expanding the Eqs. (13)-(17) around the origin, we find that regularity is guaranteed when the fields behaves as
$$w(ϵ)1+\frac{1}{2}C_wϵ^2+O(ϵ^3),$$
(28)
$$h(ϵ)C_hϵ+O(ϵ^3),$$
(29)
$$\stackrel{~}{\phi }(ϵ)C_\phi \frac{2\pi \beta }{\varphi _0}e^{\beta C_\phi }\stackrel{~}{v}^2\left(C_h^2+\frac{\stackrel{~}{\lambda }}{3}e^{\beta C_\phi }\right)ϵ^2+O(ϵ^3),$$
(30)
$$\delta (ϵ)C_\delta \frac{4\pi \stackrel{~}{v}^2}{\varphi _0}\left(C_w^2+\frac{e^{\beta C_\phi }}{2}C_h^2\right)ϵ^2+O(ϵ^3),$$
(31)
$$\stackrel{~}{m}(ϵ)\frac{2\pi \stackrel{~}{v}^2}{\varphi _0}\left(C_w^2+e^{\beta C_\phi }C_h^2+\frac{\stackrel{~}{\lambda }}{6}e^{2\beta C_\phi }\right)ϵ^3+O(ϵ^4).$$
(32)
$`C_w`$, $`C_h`$ and $`C_\phi `$ are constant, which should be determined iteratively so that the boundary conditions (25)-(27) are satisfied. We also require that no singularity exists, i.e.,
$`m(r)<{\displaystyle \frac{r}{2G}}.`$ (33)
In the case of the monopole black hole, we assume the existence of a regular event horizon at $`r=r_H`$. For the metric
$$m_Hm(r_H)=\frac{r_H}{2G},$$
(34)
$$\delta _H\delta (r_H)<\mathrm{}.$$
(35)
For the matter fields, the square brackets in Eqs. (15)-(17) must vanish at the horizon. Hence we find that
$$\frac{dw}{d\stackrel{~}{r}}|_{\stackrel{~}{r}=\stackrel{~}{r}_H}=\frac{w_H\varphi _0}{F}\left(1w_H^2h_H^2\stackrel{~}{r}_H^2e^{\beta \stackrel{~}{\phi }_H}\right)$$
(36)
$$\frac{dh}{d\stackrel{~}{r}}|_{\stackrel{~}{r}=\stackrel{~}{r}_H}=\frac{h_H\varphi _0}{F}\left[2w_H^2+\stackrel{~}{\lambda }\stackrel{~}{r}_H^2(h_H^21)e^{\beta \stackrel{~}{\phi }_H}\right]$$
(37)
$$\frac{d\phi }{d\stackrel{~}{r}}|_{\stackrel{~}{r}=\stackrel{~}{r}_H}=\frac{2\pi \beta \stackrel{~}{v}^2e^{\beta \stackrel{~}{\phi }_H}}{F}\left[\stackrel{~}{\lambda }\stackrel{~}{r}_H^2e^{\beta \stackrel{~}{\phi }_H}(h_H^21)^2+2w_H^2h_H^2\right],$$
(38)
where
$`F`$ $`=`$ $`2\pi \stackrel{~}{v}^2\stackrel{~}{r}_H\left[2\stackrel{~}{r}_H^2(1w_H^2)^2+4e^{\beta \stackrel{~}{\phi }_h}w_H^2h_H^2+\stackrel{~}{\lambda }\stackrel{~}{r}_H^2(h_H^21)^2e^{2\beta \stackrel{~}{\phi }_H}\right]\varphi _0\stackrel{~}{r}_H.`$ (39)
We also require that no singularity exists outside the horizon, i.e.,
$`m(r)<{\displaystyle \frac{r}{2G}}\mathrm{for}r>r_H.`$ (40)
Hence, we should determine the values of $`C_w`$, $`C_h`$ and $`C_\phi `$ (self-gravitating monopole case) or $`\stackrel{~}{\phi }_H`$, $`w_H`$, and $`h_H`$ (monopole black hole case) iteratively so that the boundary conditions at infinity are fulfilled. In this sense these are shooting parameters. It is a remarkable property of the nontrivial structure of many systems discussed before that we can not choose arbitrarily the values both of the fields and of their derivatives at the horizon. When we investigate the internal structure of non-Abelian black holes, we may expect the existence of the inner (Cauchy) horizon. There, the fields must satisfy the same type of constraint such as Eqs. (36)-(38). However, the shooting parameters on the black hole horizon which satisfy the boundary condition at infinity do not necessarily fulfill the constraint at the inner horizon. By this behavior the mass function diverges and a mass inflation phenomenon is realized inside of the black hole event horizon for non-Abelian black holes .
A non-trivial solution does not necessarily exist for the given theoretical parameters. However, for arbitrary values of $`\stackrel{~}{v}`$ and $`\stackrel{~}{\lambda }`$, there exists a RN black hole solution such as
$`w=0,h=1,\delta =0,\varphi =\varphi _0,\stackrel{~}{m}(\stackrel{~}{r})=\stackrel{~}{M}{\displaystyle \frac{2\pi \stackrel{~}{v}^2}{\varphi _0\stackrel{~}{r}}}.`$ (41)
$`\stackrel{~}{M}`$ is the gravitational mass at spatial infinity and $`\stackrel{~}{Q}2\stackrel{~}{v}\sqrt{\pi /\varphi _0}`$ is the magnetic charge of the black hole. The radius of the event horizon of the RN black hole is constrained to be $`\stackrel{~}{r}_H\stackrel{~}{Q}`$. The equal sign case implies an extreme solution. If we equate $`\beta =0`$, (i.e., $`\omega \mathrm{}`$) the model recovers the EYMH system and we find a self-gravitating monopole and its black hole solution investigated in Ref. . In addition, if we put $`G=0`$, i.e., neglect the gravitational effect, the model recovers the YMH system, which has a famous ’t Hooft-Polyakov solution for a regular magnetic monopole. Further, assuming $`\lambda =0`$, we obtain the BPS monopole solution,
$`w(\stackrel{~}{r})={\displaystyle \frac{\stackrel{~}{r}}{\mathrm{sinh}\stackrel{~}{r}}},h(\stackrel{~}{r})=\mathrm{coth}\stackrel{~}{r}{\displaystyle \frac{1}{\stackrel{~}{r}}}.`$ (42)
In the next section, we show the solutions of a self-gravitating monopole both in GR and BD theory.
## III Self-gravitating monopole in Brans-Dicke theory
The self-gravitating monopole in GR were investigated in detail by several authors and found to have many interesting properties, some of which can not be seen for the monopole solution in flat space-time. We list their properties. (1) The distribution of the non-Abelian structure decays exponentially with respect to the distance from the origin and the monopole has its core radius $`1/ev`$. (2) The solutions are characterized by the node number of the YM potential, which is different the from topological winding number of the monopole. (3) There exists the maximum parameter $`v_{max}`$ beyond which there is no monopole solution. $`v_{max}`$ weakly depends on $`\lambda `$. (4) There exists the critical parameter $`v_{extreme}`$ for the small $`\stackrel{~}{\lambda }`$. At $`v=v_{extreme}`$, the monopole solution changes to extreme RN solution. Between $`v_{extreme}`$ and $`v_{max}`$, two monopole solutions with a different mass appear for each value of $`v`$. The higher mass solution is unstable while the lower one is stable. $`v_{extreme}`$ also weakly depends on $`\lambda `$.
Now we turn to the monopole solutions in BD theory and compare their properties with those in GR from the above points of view. Solving the system (13)-(20) with the boundary conditions (23)-(33), we obtain gravitating monopole solutions. Although they are found only numerically, there exists a rigorous proof of the existence of self-gravitating monopole and its black hole solution in the $`\stackrel{~}{\lambda }\mathrm{}`$ case in GR . We expect that the same discussion holds in the more general case including BD theory.
We show the field configurations of the gravitating monopole in Figs. 1. We chose the parameters as $`\stackrel{~}{v}=0.1`$, $`\stackrel{~}{\lambda }=0.1`$ and several values of $`\omega =1.4`$, $`1.0`$, $`0`$, $`\mathrm{}`$. The $`\omega =\mathrm{}`$ case corresponds to GR. (a) is $`\stackrel{~}{r}`$-$`h,w`$ diagram. We can see that the configurations of the YM potential and the Higgs field hardly depend on the BD parameter $`\omega `$. This implies that the solution can have almost the same structure as the ’t Hooft-Polyakov monopole in flat space-time, where the repulsive force of the gauge field balances with the attractive force of the Higgs field. (b) and (c) are $`\stackrel{~}{r}`$-$`\stackrel{~}{m}_g`$ and $`\stackrel{~}{r}`$-$`\delta `$ diagram, respectively. Since $`\stackrel{~}{m}_g`$ is defined as
$$\stackrel{~}{m}_g\stackrel{~}{m}+\frac{2\pi \stackrel{~}{v}^2}{\varphi _0\stackrel{~}{r}},$$
(43)
it diverges at the origin. The mass function $`\stackrel{~}{m}_g`$ strongly depends on $`\omega `$. For a different BD parameter each solution has a different effective global gauge charge $`\stackrel{~}{Q}`$. For the small $`\omega `$, the charge is small, i.e., the monopole in BD theory has a smaller charge than that in GR. Since there is a prefactor $`\stackrel{~}{Q}`$ in the $`\delta `$ and $`\stackrel{~}{m}`$ equations, it contributes directly to the behavior of the metric functions. There is another factor which may change the behavior of metric functions. It is the BD scalar field $`\phi `$. However from Fig. 1 (d), $`\stackrel{~}{\phi }\stackrel{<}{}0.07`$ even for $`\omega =1.4`$, which implies the $`e^{\beta \stackrel{~}{\phi }}1`$. $`d\stackrel{~}{\phi }/d\stackrel{~}{r}`$ term is also smaller than other terms in the field equations. Hence we can neglect the $`\stackrel{~}{\phi }`$ contribution to other fields in the zeroth approximation. However, we may not neglect the $`\stackrel{~}{\phi }`$ contribution for the parameter range $`vv_{extreme}`$ discussed later, where the gravitational effect plays an important roll.
In Fig. 2 we show the distribution of the energy density $`\rho T_{\mathrm{\hspace{0.33em}0}}^0`$ for the same parameters as Fig. 1 (It is normalized as $`\stackrel{~}{\rho }=\rho G/(ev)^2`$). An arrow shows the core radius of the monopole. The prefactor $`1/\varphi _0`$ for the matter Lagrangian can be absorbed by defining the new variables and the effective parameters as $`\overline{A}_\mu ^a=A_\mu ^a/\sqrt{\varphi _0}`$, $`\overline{\mathrm{\Phi }}=\mathrm{\Phi }/\sqrt{\varphi _0}`$, $`\overline{e}=\sqrt{\varphi _0}e`$, $`\overline{v}=v/\sqrt{\varphi _0}`$ and $`\overline{\lambda }=\varphi _0\lambda `$. Hence the core radius $`1/\overline{e}\overline{v}=1/ev`$ hardly depends on $`\varphi _0`$, i.e. BD parameter $`\omega `$. The energy density is almost constant inside the monopole core, but it decays as $`1/\stackrel{~}{r}^4`$ outside. Note that although $`w`$ decays exponentially, the YM potential decays $`1/r`$ for a large radius.
As for the monopole in flat space-time, the uniqueness of the solution was proven , which means there is no excited mode of this solution. As for gravitating counterpart, however, there exist radial excited solutions characterized by the node number $`n`$ of the YM potential for small $`\stackrel{~}{v}`$ in GR. We also find such excited solutions in BD theory. In the $`v0`$ limit, the Higgs field becomes constant and the YM field recovers its apparent gauge symmetry. Then the YM field is conformally invariant and the BD scalar field becomes trivial by the regularity. As a result, the $`n=1`$ excited solution approaches a regular Bartnik-McKinnon solution in the EYM system in this limit as discussed in Ref. . Hereafter, we focus on the solution with $`n=0`$, i.e., the ground state solution.
In GR, there is the maximum value of the vacuum expectation value $`\stackrel{~}{v}_{max}`$ for a fixed $`\stackrel{~}{\lambda }`$ where a regular monopole solution can exist . We can understand this intuitively as follows. We compare the mass of the monopole $`4\pi v/e`$ with its core radius $`1/ev`$. When $`v`$ becomes large, the mass of the monopole increases, as does the gravitational radius $`8\pi Gv/e`$, while its core radius shrinks. Hence, its core radius eventually gets into gravitational radius and the solution may develop into the RN black hole being swallowed in its non-trivial structure by the horizon. The existence of the boundary of the parameter is interesting also in the context of the inflation. Since there is no non-trivial static configuration beyond the maximum parameter, the solution must take a configuration of the static RN black hole or a nonstatic configuration. One of the possibilities in the latter case is the inflating solution. Actually it is confirmed that the boundary of the static solution almost coincides with the parameter region where the monopole inflation occurs .
In BD theory, it was shown that when we consider the global monopole, any inflating monopole eventually shrinks contrary to the case in GR. Although we consider gauge monopole here, the BD scalar field may give a serious effect. Moreover it is found that the gravitating monopole solution in $`N=4`$ supersymmetric low-energy superstring theory exists for arbitrary values of vacuum expectation value of the Higgs field, where the dilatonic field plays an important roll . Since the BD scalar field can be regarded as the dilaton field in a special coupling constant case as can be seen in Eq. (3), there is a possibility that $`\stackrel{~}{v}_{max}`$ disappears. Actually, taking $`\stackrel{~}{v}`$ larger, we find the BD scalar field gives a larger contribution. Hence we are interested in examining the maximum value $`\stackrel{~}{v}_{max}`$ in BD theory.
First, we discuss analytically the parameter region where the nontrivial solutions exist following the analysis in Ref. . For the monopole solution, the mass function behaves as Eq. (32) around the origin by regularity. At spatial infinity, the solution approaches RN solution,
$`\stackrel{~}{m}(\stackrel{~}{r})=\stackrel{~}{M}{\displaystyle \frac{2\pi \stackrel{~}{v}^2}{\varphi _0\stackrel{~}{r}}}+\mathrm{}.`$ (44)
From this boundary condition, $`g^{rr}`$ should have a minimum which corresponds to the maximum of the $`\stackrel{~}{m}/\stackrel{~}{r}`$. This occurs around $`\stackrel{~}{r}_m4\pi \stackrel{~}{v}^2/\varphi _0\stackrel{~}{M}`$. For the monopole in flat space-time, we have
$`\stackrel{~}{M}_{flat}={\displaystyle \frac{4\pi \stackrel{~}{v}^2}{\varphi _0}}g(\stackrel{~}{\lambda }),`$ (45)
$`g(\stackrel{~}{\lambda })`$ takes values from $`1`$ to $`1.787`$ when $`\stackrel{~}{\lambda }`$ changes from $`0`$ to $`\mathrm{}`$. We assume $`\stackrel{~}{M}\stackrel{~}{v}^2`$ even for the gravitating monopole in both theories. Under this assumption, $`\stackrel{~}{r}_m1/g1`$. So $`g^{rr}`$ takes a minimum value of about $`g^{rr}14\pi \stackrel{~}{v}^2/\varphi _0`$. When $`\stackrel{~}{v}=\stackrel{~}{v}_{extreme}=\sqrt{\varphi _0/4\pi }`$, $`g^{rr}`$ takes $`0`$ and the degenerate horizon appears at $`r=r_{extreme}`$. Because the region $`\stackrel{~}{r}<\stackrel{~}{r}_{extreme}`$ is separated infinitely in geodesic distance from the outer region, the monopole structure could be thought to be confined in the region $`\stackrel{~}{r}<\stackrel{~}{r}_{extreme}`$ and the solution looks like a RN black hole with the horizon radius $`r_{extreme}`$. Naively $`\stackrel{~}{v}_{extreme}`$ gives the maximum value of $`\stackrel{~}{v}`$. As a result, we can expect that the critical value $`\stackrel{~}{v}_{extreme}`$ becomes large for the large $`\varphi _0`$, i.e., the small value of $`\omega `$.
Breitenlohner et al. confirmed the above discussion numerically in GR ($`\varphi _0=1`$) and found the maximum value $`\stackrel{~}{v}_{max}`$ for the fixed $`\stackrel{~}{\lambda }`$ by examining the minimum of $`g^{rr}`$ . Moreover they found the parameter region of $`\stackrel{~}{\lambda }`$ where $`\stackrel{~}{v}_{extreme}`$ is not the maximum value of $`\stackrel{~}{v}`$. In the parameter region $`\stackrel{~}{v}_{extreme}<\stackrel{~}{v}<\stackrel{~}{v}_{max}`$, there are two different nontrivial solutions with the same $`\stackrel{~}{v}`$ which is not obvious in the above simple analysis. On $`\stackrel{~}{v}`$-$`\stackrel{~}{M}`$ diagram, these solutions construct two solution branches divided by a cusp structure at $`\stackrel{~}{v}=\stackrel{~}{v}_{max}`$. The lower mass branch ends up with $`\stackrel{~}{v}_{max}`$ and the higher mass branch is connected to the RN black hole branch at $`\stackrel{~}{v}=\stackrel{~}{v}_{extreme}`$. The cusp structure is important for the stability analysis by using catastrophe theory. For the larger $`\stackrel{~}{\lambda }`$, $`\stackrel{~}{v}_{max}`$ and $`\stackrel{~}{v}_{crit}`$ coincide. Since $`\stackrel{~}{v}_{max}`$ takes a value very close to $`\stackrel{~}{v}_{extreme}`$, we search $`\stackrel{~}{v}_{extreme}`$ by the behavior of $`g^{rr}`$ first and estimate $`\stackrel{~}{v}_{max}`$.
In the same manner, we examine $`\stackrel{~}{v}_{max}`$ in BD theory with $`\omega =0`$. Figs. 3 shows the field configurations of the monopole with $`\stackrel{~}{\lambda }=0.2`$ and around the maximum parameter $`\stackrel{~}{v}_{max}`$ ((a) $`\stackrel{~}{r}`$-$`g^{rr}`$ (b) $`\stackrel{~}{r}`$-$`h`$, $`w`$). We can find that $`g^{rr}`$ has the minimum around $`\stackrel{~}{r}=\stackrel{~}{r}_{extreme}1.53`$ and the minimum value decreases as $`\stackrel{~}{v}`$ becomes larger. When $`\stackrel{~}{v}`$ approaches $`\stackrel{~}{v}_{max}`$, the behavior of the YM potential and the Higgs field changes more rapidly near the $`\stackrel{~}{r}_{extreme}`$ and takes $`w0`$ and $`h1`$, i.e., the solution can be approximated by a RN black hole solution outside of $`\stackrel{~}{r}_{extreme}`$. When we have $`\stackrel{~}{v}_{extreme}`$, the horizon degenerates and the outer region becomes the extreme RN black hole. Since the gravitational effect becomes strong around this parameter region as seen from Fig. 3 (a), the amplitude of $`\stackrel{~}{\phi }`$ and the dependence of $`w`$ and $`h`$ on the BD parameter can not be negligible. We show the $`\stackrel{~}{v}_{max}`$-$`\stackrel{~}{\lambda }`$ diagram with $`\omega =0`$ and $`\mathrm{}`$ in Fig. 4 (the solid lines). Though $`\stackrel{~}{v}_{max}`$ is larger compared to the estimation $`\stackrel{~}{v}_{extreme}=\sqrt{\varphi _0/4\pi }0.282`$ (for GR), $`0.325`$ (for $`\omega =0`$) due to the naive estimation, we can obtain the relevant dependence of $`\stackrel{~}{v}_{max}`$ on $`\varphi _0`$ (or $`\omega `$). If we take a more accurate value of $`r_m`$ calculated numerically, we have the better estimation of $`\stackrel{~}{v}_{extreme}0.369`$ (for GR) and $`\stackrel{~}{v}_{extreme}0.497`$ (for $`\omega =0`$) As we describe later, the similar phenomena can be seen in monopole black hole. These are also shown in this diagram by the dotted lines (In this diagram, $`\stackrel{~}{r}_H=0.4`$.).
## IV Monopole black hole in Brans-Dicke theory
The monopole black hole solution in GR has the following properties in addition to (1)-(2) of Sec.III. The number of the solutions depends on the parameters complicatedly, since another parameter, the horizon radius, is added in the black hole case. We show the schematic diagram of $`\stackrel{~}{v}`$-$`\stackrel{~}{r}_H`$ for small $`\stackrel{~}{\lambda }`$ in Fig. 5 . In the shadowed region, the nontrivial solution exists. The $`\stackrel{~}{r}_H=0`$ axis corresponds to the regular monopole solution. (5) For a fixed value of $`\stackrel{~}{r}_H<\stackrel{~}{r}_H^{}`$ (See Fig. 5), we find two characteristic values $`\stackrel{~}{v}_{max}`$ and $`\stackrel{~}{v}_{extreme}`$. They are extensions of properties (3) and (4) in the regular monopole case discussed in the previous section. In the region (i) in Fig. 5 ($`\stackrel{~}{v}_{extreme}<\stackrel{~}{v}<\stackrel{~}{v}_{max}`$), there are two nontrivial solutions with different mass. The solution branch with higher mass is connected to the extreme RN black hole solution in the $`\stackrel{~}{v}`$-$`\stackrel{~}{M}`$ diagram although the horizon radius changes discontinuously. (6) The solution in the lower mass branch is stable while the higher one is unstable. (7) For a fixed value of $`\stackrel{~}{v}<\stackrel{~}{v}^{}`$ (See Fig. 5), there are two characteristic values $`\stackrel{~}{r}_{H,max}`$ and $`\stackrel{~}{r}_{H,RN}`$. In the region (ii) in Fig. 5 ($`\stackrel{~}{r}_{H,RN}<\stackrel{~}{r}<\stackrel{~}{r}_{H,max}`$), there are two nontrivial solutions. The solution branch with lower entropy (smaller $`\stackrel{~}{r}_H`$) is connected to the RN black hole at $`\stackrel{~}{r}_H=\stackrel{~}{r}_{H,RN}`$. (8) The solution in the high entropy branch is stable while the lower one is unstable. (9) The regions (i) and (ii) disappear for $`\stackrel{~}{\lambda }>\stackrel{~}{\lambda }_{crit}^{(i)}0.287`$ and $`\stackrel{~}{\lambda }>\stackrel{~}{\lambda }_{crit}^{(ii)}0.696`$, respectively. (10) When we consider the thermodynamics of the monopole black hole, its temperature diverges in the minimum mass limit unlike the RN black hole. Through a evaporation process from a large RN black hole, the solution experiences the first and/or second order phase transitions.
We turn to the monopole black hole in BD theory. Solving the system (13)-(20) numerically with the boundary conditions (23)-(27) and (34)-(40), we obtain a black hole solution. Figs. 6 are the field configurations with the parameters $`\stackrel{~}{\lambda }=0.1`$, $`\stackrel{~}{v}=0.1`$, $`\omega =0`$ and the horizon radius $`\stackrel{~}{r}_H=0.2,0.4`$ and $`0.6`$ ((a) $`\stackrel{~}{r}`$-$`h`$,$`w`$ (b) $`\stackrel{~}{r}`$-$`\stackrel{~}{m}_g`$ (c) $`\stackrel{~}{r}`$-$`\delta `$ (d) $`\stackrel{~}{r}`$-$`\stackrel{~}{\phi }`$) We also plot the gravitating monopole solution, i.e., $`\stackrel{~}{r}_H=0`$, for comparison. As the horizon radius becomes large, the curvature of the space-time outside the horizon would become small resulting in the variation of the BD scalar field being small. Hence the difference between the solutions in BD theory and in GR appears most conspicuously in the monopole solution case. The boundary value of $`w`$ and $`h`$ approaches $`w(r_H)0`$ and $`h(r_H)1`$ for a large horizon radius. Finally the solution becomes a RN black hole in the $`\stackrel{~}{r}_H=\stackrel{~}{r}_{H,RN}`$ limit. The behaviors of the solutions around this limit depend on the parameters complicatedly as discussed later.
In Fig.7(a), we show the $`\stackrel{~}{M}`$-$`\stackrel{~}{r}_H`$ diagram for $`\stackrel{~}{\lambda }=0.1`$, $`\stackrel{~}{v}=0.1`$ and $`\omega =1.4`$, $`1`$, $`0`$ and $`\mathrm{}`$. Dotted lines show RN solution branches. Note that the RN solution in BD theory is different from that in GR because it has a different effective gauge charge by the factor $`\varphi _0`$. Unlike the RN black hole, the monopole black hole does not have a extreme limit but has the $`\stackrel{~}{r}_H=0`$ limit, which corresponds to the gravitating monopole solution. Fig.7(b) is a magnification around the maximum horizon radius of each theory. As is the case of GR, there are two solutions in some ranges of horizon radius. This parameter region corresponds to the region (ii) in Fig. 5. We call the upper (lower) branch a high (low) entropy branch. In both theories, these branches form a cusp structure at the point $`A`$, where the solution has the maximum horizon radius $`\stackrel{~}{r}_{H,max}`$. At the point $`B`$, the solution coincides with the RN black hole solution with the horizon radius $`\stackrel{~}{r}_{H,RN}`$. The solutions have a qualitatively similar structure in both theories. The maximum mass and the maximum horizon radius decrease when $`\omega `$ becomes small. Furthermore we find that the range of horizon radius where the low entropy branch exists becomes larger when we take the $`\omega `$ as small. These behaviors are mainly due to the effect of the boundary value of $`\varphi _0`$. As we mentioned, $`\overline{v}=\stackrel{~}{v}/\sqrt{\varphi _0}`$ can be considered as an effective vacuum expectation value in BD theory. Since it becomes small when $`\omega `$ becomes small, we find that the width of region (ii) in Fig. 5 becomes large.
Next we consider a difference of solution in each branch. We consider the solution with $`\stackrel{~}{v}=0.1`$, $`\stackrel{~}{\lambda }=0.1`$ and $`\stackrel{~}{r}_h=0.59`$, and change the BD parameter $`\omega `$. Figs. 8 and 9 show configurations of the fields ((a) $`\stackrel{~}{r}`$-$`h`$, $`w`$ (b) $`\stackrel{~}{r}`$-$`\stackrel{~}{m}_g`$ (c) $`\stackrel{~}{r}`$-$`\stackrel{~}{\phi }`$ ). In the high entropy branch (Fig. 9), the YM potential and the Higgs field hardly depend on $`\omega `$. This is the same as the self-gravitating monopole case and the structure is supported by the balance between these fields. On the other hand, they have small amplitude in the low entropy branch (Fig. 10). The Higgs field can be approximated as $`h1`$, which means that it takes almost its vacuum value, while $`w0`$, where the YM potential decays as $`\stackrel{~}{r}^1`$. Hence, what attracts the YM field is not the Higgs field but the self-gravitational attractive force. Actually, the counterpart of the low entropy branch does not exist in flat space-time. The dependence on $`\omega `$ in the low entropy branch is due to the difference of the effective gauge charge again. The horizon radius $`\stackrel{~}{r}_h=0.59`$ line crosses the low entropy branch relatively near the point A (See Fig.8(b)). In the large $`\omega `$ (GR), however, it crosses around the point B, where the solution coincides with the RN black hole. Hence the field configurations approach a trivial solution.
In the present model, the YM field obtains the mass through a spontaneous symmetry breaking mechanism and the trace part of the energy-momentum tenser does not vanish. In this case the BD scalar field takes a nontrivial configuration and we can define the scalar mass $`M_s`$ by the asymptotic behavior of the BD scalar field as
$`\varphi \varphi _0\left(1+{\displaystyle \frac{2GM_s}{r}}\right).`$ (46)
Fig. 10 shows $`\omega `$ dependence of the gravitational mass $`M`$ and scalar mass $`M_s`$ for fixed horizon radius $`\stackrel{~}{r}_h=0.59`$. As the solution can be approximated by the RN black hole in the zeroth approximation, $`M`$ behaves as
$`\stackrel{~}{M}{\displaystyle \frac{1}{2\stackrel{~}{r}_h}}\left[\stackrel{~}{r}_h^2+4\pi \stackrel{~}{v}^2{\displaystyle \frac{2\pi \stackrel{~}{v}^2}{\omega +2}}\right].`$ (47)
This gives good agreement with the $`\omega `$ dependence in Fig. 10(a). $`\stackrel{~}{M}_s`$ can be expressed as
$`\stackrel{~}{M}_s={\displaystyle \frac{12s}{2\omega +3}}\stackrel{~}{M},`$ (48)
where $`s`$ is the sensitivity of a self-gravitating object. Though $`s`$ varies as $`\omega `$ changes, its configuration is negligible for large $`\omega `$ and $`\stackrel{~}{M}_s`$ behaves as $`1/\omega `$. Since $`s1/2`$ in the $`\omega 3/2`$ limit, the fraction in Eq. (48) is indeterminate. However, Fig. 11(b) shows that $`\stackrel{~}{M}_s`$ takes a non-zero finite value. As $`\stackrel{~}{M}`$ takes almost the same value in each branch, the solution in the high entropy branch has smaller sensitivity than the lower branch.
In GR, it is considered that when $`\lambda `$ becomes large, the parameter point where the $`\stackrel{~}{r}_{H,RN}=\stackrel{~}{r}_{H,max}`$ in Fig. 5s shifts to the left side, i.e., smaller $`\stackrel{~}{v}`$ and the region (ii) becomes small. When $`\stackrel{~}{\lambda }`$ takes the critical value $`\stackrel{~}{\lambda }_{crit}^{(ii)}`$, the region (ii) finally disappears . Although we do not examine the definite value of $`\stackrel{~}{\lambda }_{crit}^{(ii)}`$ in BD theory, the dependence on $`\omega `$ would be small because $`\stackrel{~}{\lambda }=\lambda /e^2=\overline{\lambda }/\overline{e}^2`$ does not depend on $`\varphi _0`$.
The stability of these exotic objects is one of the main interests. Aichelburg analyzed the monopole black hole and its stability via the linear perturbation method in GR and showed that most of the solutions are stable within the limit of $`\stackrel{~}{\lambda }\mathrm{}`$. They also obtained the solutions in the finite $`\stackrel{~}{\lambda }`$ case and suggested that if we have two monopole black hole solutions for a fixed value of $`\stackrel{~}{v}`$, the low energy one is stable while the other is not. Similar works are found in , where the authors considered the weak coupling limit. We examined stabilities via catastrophe theory and obtained a unified picture of the stabilities of the gravitating monopole and the monopole black hole. In catastrophe theory, we discuss the shape of a potential function as the variation of the control parameters. In that paper, we adopted $`\stackrel{~}{\lambda }`$, $`\stackrel{~}{v}`$ and $`\stackrel{~}{M}`$ as the control parameters, entropy $`S`$ as the potential function and $`\delta (r_h)`$ as the state variable, and found that the system is classified into a swallow tail catastrophe. In general, a swallow tail catastrophe has three control parameters $`a`$, $`b`$, $`c`$ and one state variable $`x`$ and its potential function is described as
$`V={\displaystyle \frac{1}{5}}x^5+{\displaystyle \frac{1}{3}}ax^3+{\displaystyle \frac{1}{2}}bx^2+cx.`$ (49)
We regard the parameters as
$`a=a(\stackrel{~}{\lambda },\stackrel{~}{v}),b=b(\stackrel{~}{M},\stackrel{~}{v}),c=0.`$ (50)
$`\stackrel{~}{\lambda }_{crit}^{(ii)}`$ is obtained by $`a(\stackrel{~}{\lambda }_{crit}^{(ii)},0)=0`$. Although we have a new control parameter $`\omega `$ in BD theory, the situation does not change. By analyzing the dependence of the solution on the parameters, Eq. (50) is changed to
$`a=a(\stackrel{~}{\lambda },\stackrel{~}{v},\omega ),b=b(\stackrel{~}{M},\stackrel{~}{v},\omega ),c=0,`$ (51)
i.e., the BD parameter is not an intrinsic parameter but a dummy parameter in catastrophe theory, and does not result in qualitative difference from GR. This is the same as in the neutral type non-Abelian black hole case. As a result, the monopole black hole in BD theory is also classified as a swallow tail catastrophe. From this we find that the solution in the high entropy branch and RN black hole with larger horizon radius than $`\stackrel{~}{r}_{H,RN}`$ are stable, while the solutions in the low entropy branch and RN black hole with smaller horizon radius than $`\stackrel{~}{r}_{H,RN}`$ are unstable.
These properties are interesting in the context of black hole no-hair conjecture and the uniqueness of the solution. For $`\stackrel{~}{\lambda }>\stackrel{~}{\lambda }_{crit}^{(ii)}`$, there are two solutions (monopole black hole and RN black hole) when we fix the global charge $`M`$ and $`Q`$. This is the counterexample of the strong no-hair conjecture, i.e., if the global charges are fixed, there exists only one solution. Moreover when $`\stackrel{~}{\lambda }`$ is smaller than $`\stackrel{~}{\lambda }_{crit}^{(ii)}`$, there is a mass range where two stable solutions (monopole black hole in the high entropy branch and RN black hole) exist for a fixed mass. This violates even the weak no-hair conjecture which states that if the global charges are fixed, there exists only one stable solution. These behaviors are shared by both theories. In GR, however, we have no idea how to identify whether a black hole is a monopole black hole or RN black hole from infinity, because the NA structure and the Higgs field decrease exponentially. On the other hand, in BD theory we can obtain the information about the scalar mass, which can be used to tell the type of black holes. Hence we can identify each black hole uniquely from infinity.
When we take into account the quantum effect, a black hole loses its energy by the Hawking evaporation process. The evolution of the monopole black hole by this process is rather complicated. First, we show an inverse temperature in the $`\stackrel{~}{\lambda }=1.0`$, $`\stackrel{~}{v}=0.1`$ and $`\omega =0`$, $`\mathrm{}`$ (GR) in Fig. 11(a). The inverse temperature of the RN black hole has the minimum $`1/T=6\sqrt{3}\pi Q`$ at $`\stackrel{~}{M}=2\stackrel{~}{Q}/\sqrt{3}`$ and diverges at the extreme limit $`\stackrel{~}{M}=\stackrel{~}{Q}`$, where the evaporation ends. On the other hand, the inverse temperature of the monopole black hole becomes zero in the $`\stackrel{~}{M}=\stackrel{~}{M}_{monopole}`$ limit as with the Schwarzschild black hole. Consider the evolution of a RN black hole with $`\stackrel{~}{r}_H>\stackrel{~}{r}_{H,max}`$ by evaporation process. In GR in this figure, it loses mass, and when the solution approaches the minimum of the inverse temperature, its heat capacity changes discontinuously. \[Note that its specific heat changes its sign continuously.\] This is the second order phase transition. The black hole loses its mass further and reaches the point $`B`$. Below this mass, a RN black hole becomes unstable and the solution traces the monopole black hole branch. The nontrivial structure comes out of the event horizon gradually. This is also the second phase transition. At this point, however, both the heat capacity and the specific heat change discontinuously. After that, the evaporation continues until the event horizon disappears. Finally, the regular monopole may remain as a remnant. For $`\omega =0`$, however, it experiences a different evolution, because before it reaches the minimum of the inverse temperature, it reaches the point $`B`$ and traces the monopole black hole branch continuously.
Furthermore, the $`\stackrel{~}{\lambda }<\stackrel{~}{\lambda }_{crit}^{(ii)}`$ case is different. Fig. 11(b) shows the inverse temperature of the solutions with the $`\stackrel{~}{\lambda }=0.1`$, $`\stackrel{~}{v}=0.1`$. In the $`\omega =1.4`$, $`1.0`$ and $`0`$ case, the RN black hole evolves to the point B as in the above case. However, since the RN black hole becomes unstable, the solution jumps to the stable monopole branch, increasing its entropy and temperature discontinuously (arrows in Fig. 11(b)). This is the first order phase transition. The nontrivial structure comes out suddenly (or catastrophecally) from the horizon. In the $`\omega =\mathrm{}`$ case, the solution experiences the second order phase transition at $`\stackrel{~}{M}=2\stackrel{~}{Q}/\sqrt{3}`$ before the first order phase transition. As shown above, the type and number of the phase transitions depend on the physical parameters. This property is more important for the stability of the solution in the system surrounded by a heat bath. In that system, stability change occurs at the point where the heat capacity diverges. Another thing that we should note is that the monopole black hole objects are forced to be produced by this process if the accretion to the black hole is very little.
Until now, we have studied the solution in the rather small value of $`\stackrel{~}{v}`$ corresponding to the left side of Fig. 5. and found the region (ii) in BD theory. Next we consider the $`\stackrel{~}{v}_{max}`$ in BD theory. By the same method in the gravitating monopole case, we found $`\stackrel{~}{v}_{max}`$. Fig. 4 shows the existence region of the black hole solution ($`\stackrel{~}{r}_H=0.4`$) in $`\stackrel{~}{v}`$-$`\stackrel{~}{\lambda }`$ plane for $`\omega =0`$ and $`\mathrm{}`$. In the left region of each dotted line, there are black hole solutions. When we take $`\stackrel{~}{\lambda }`$ as larger, the dependence of $`\stackrel{~}{v}_{max}`$ becomes smaller. We can not see the intrinsic difference also in this diagram. It is different to clarify the detailed structure of region (i) in BD theory because of the fine tuning of the asymptotic value of the BD field. In GR the region (i) disappears at $`\stackrel{~}{\lambda }=\stackrel{~}{\lambda }_{crit}^{(i)}0.287`$. As discussed before, since $`\stackrel{~}{\lambda }`$ does not depend on $`\varphi _0`$, $`\stackrel{~}{\lambda }^{(i)}`$ would not be changed much even in BD theory. However, since the gravitational effect becomes strong around $`\stackrel{~}{v}=\stackrel{~}{v}_{max}`$, the BD scalar field may affect the value of $`\stackrel{~}{\lambda }_{crit}^{(i)}`$. It needs further analysis to say definite thing.
## V Conclusion
We found the self-gravitating monopole and its black hole solution of monopole in BD theory and compared them with those in GR. We clarified the following important aspects. The configuration of the YM potential and the Higgs fields of a self-gravitating monopole and a monopole black hole hardly depend on the BD parameter in most of the mass range, which shows the structure resembles the ’t Hooft-Polyakov monopole. We found the parameter region (ii) in Fig. 5, where two monopole black hole solutions exist for fixed horizon radius. Each solution branch has a different dependence for the BD parameter. For the higher entropy branch, the configurations of the fields hardly depend on the BD parameter, while the configuration of the fields changes to those of RN black hole in the $`\stackrel{~}{v}\stackrel{~}{v}_{max}`$ limit for the lower entropy branch. However, we can not find the region (i) because fine tuning is needed to satisfy the boundary condition of the BD field at infinity. One of the characteristic properties in BD theory is lessening the effective global gauge charge. By this effect, the $`\stackrel{~}{M}`$-$`\stackrel{~}{r}_h`$ diagram (Fig. 7) shifts to the left side, i.e., the mass of the solution becomes small when $`\omega `$ gets small. From this diagram, we can derive information about stability by applying catastrophe theory. As in the GR case, the high (low) entropy branch is stable (unstable). And we found that the BD parameter is not an intrinsic parameter but a dummy parameter in catastrophe theory. The effective global charge also affects the configuration for the high entropy branch, and does affect the existence region of the solutions in parameter regions. We investigated a boundary value of $`\stackrel{~}{v}`$ above which there exist no static solution for $`\stackrel{~}{r}_h=0.4`$ and a self-gravitating solution. It is interesting that the allowed parameter region is extended in BD theory. We also discuss the thermodynamical properties of the black hole solutions. The number and the order of the phase transitions depend complicatedly on the theoretical parameters.
These aspects may give us some important information about monopole black holes in other theories of gravity. In particular, we can anticipate the results in dilaton gravity because of its similarity to BD theory. Though we can not see the intrinsic qualitative difference form GR, some effective theory of unified theory may cause important effects for monopole or its black hole solution, which may predict some information to the remnant of the Hawking evaporation.
There are some issues which we should comment on. In this paper, we transformed the system from the original BD frame to the Einstein frame and discussed only in the latter frame. However, the solutions in the former frame are easily obtained from our solution by the inverse transformation of Eq. (2). For example, $`M`$-$`r_H`$ diagram in the BD frame is obtained as follows. In the BD frame, a gravitational mass $`M_{BD}`$ is defined by use of the scalar mass as
$`M_{BD}=M+M_s.`$ (52)
The $`\stackrel{~}{M}_{BD}`$-$`\stackrel{~}{r}_H`$ diagram is shown in Fig. 12, where we normalize as $`\stackrel{~}{M}_{BD}=GevM_{BD}`$. Because of the existence of a scalar field, a cusp structure disappears at the points $`A`$. This may show that $`\stackrel{~}{M}_{BD}`$ is not appropriate to a control parameter of catastrophe theory. Since the solution in a high entropy branch has a larger scalar mass than that in a low entropy branch as shown in Fig. 10, the high entropy branch slides to the right of the lower branch in the BD frame.
We note the relation between our results and other topological defects such as a global monopole or a cosmic string in BD theory or in more general scalar-tensor theories. For global monopole, it is investigated that a massless dilaton field or the BD scalar field affects its structure, and is discussed that there is no solution except the special value of the coupling constant. Similar arguments are seen for a cosmic string in where they discussed that “wire approximation” which can be valid in GR will not be used. For a massive dilaton field, though both global monopole and cosmic strings would not be affected seriously compared to the massless case, it is discussed that it enables us to decide the dilaton mass from observation. As for the texture, though exact results are almost unknown but Dando using self-similar ansatz showed that regular solution exists even in a massless dilaton case for a suitable choice of coupling constant. Contrary to these cases, both the self-gravitating monopole and its black hole solution exist for any value of the BD parameter in our model. This would be due to the asymptotic mild behavior of the BD scalar field.
We finally note that much work on the monopole solution has been developing from the context of superstring theory. Recently monopole solutions in flat space-time, self-gravitating monopole solutions and its black hole solutions were obtained in Born-Infeld type action. However, little has been clarified about them. It should be interesting to investigate their structure, especially the internal structure around the central singularity, because of the remarkable form of its action.
## ACKOWLEDGEMENTS
We would like to thank J. Koga and T. Tachizawa for useful discussions. (T. T.)<sup>2</sup> are thankful for financial support from the JSPS. This work was supported partially by a Grant-in-Aid for Scientific Research Fund of the Ministry of Education, Science and Culture (Specially Promoted Research No. 08102010 and No. 09410217), by a JSPS Grant-in-Aid (No. 094162), and by the Waseda University Grant for Special Research Projects.
|
no-problem/9906/astro-ph9906402.html
|
ar5iv
|
text
|
# Ballerina – Pirouettes in Search of Gamma Bursts
## 1 Introduction
The crucial observation of an X-ray transient following a cosmic gamma ray burst (GRB) was made by the Italian-Dutch satellite Beppo–SAX (Costa et al. Costa (1997)). Several GRBs with X-ray and optical afterglows have subsequently been found. In several cases, it has been possible to identify a remote galaxy as a likely host to the burst source, and to derive its distance. The huge luminosities, and the very short time scales involved in GRBs, may possibly be explained by merging, or collapsing, compact objects such as neutron stars or black holes (e.g. Paczyński Pacz (1986)). However, the observed energy spectra of the bursts, and the details of the energy transport and conversion, are still to be explained. Models involving relativistic fireballs and jets have been proposed (e.g. Dar Dar (1998)). To date no theory has been able to explain all of the observed GRB phenomena, and it has become a challenge for modern astrophysics.
## 2 The Ballerina Mission
It remains an objective to ensure the earliest detection of the X-ray afterglow. Ballerina will for the first time allow systematic studies of the soft X-ray emission in the time interval from only a few minutes after the onset of the burst to a few hours later. Ballerina will, on its own provide observations in an uncharted region of parameter space. Positions of GRB sources with accuracy better than $`1\mathrm{}`$ will be distributed within a few minutes of the burst.
Secondary objectives of the Ballerina mission includes observations of the earliest phases of the outbursts of X-ray novae and other X-ray transients.
In addition to the autonomous observations of events detected on-board, Ballerina may on short notice be commanded from the ground to execute observations on objects identified by other observatories.
Ballerina is a spacecraft in the 100 kg class. The payload consists of an all-sky monitor and a grazing incidence X-ray telescope. A compact accommodation of these key elements is shown in Fig. 1. Ballerina is one of four missions currently under study for the Danish Small Satellite Programme. If selected, Ballerina will be launched in 2002.
### 2.1 The All-Sky Monitor
The all-sky monitor consists of four Rotation Modulation Collimator instruments, arranged in a tetrahedron configuration to cover the full sky. Each unit is similar to the WATCH instruments flown successfully on the EURECA and GRANAT missions (Lund Lund85 (1985); Castro-Tirado AJCT (1994); Sazonov et al. Sazonov (1998); Brandt Brandt (1994)). The detector used is a mosaic of NaI and CsI scintillators, with an energy range of 6–120 keV. The established threshold sensitivity of WATCH for a GRB of 5 second duration is $`10^5`$ $`\text{erg/cm}^2`$. Bursts of this fluence can be located to better than $`1\mathrm{°}`$ diameter ($`2\sigma `$).
Placed in a high (Molniya-type) orbit with an efficiency of 55%, Ballerina will observe $`80`$ bursts per year distributed over the full sky. About 10 of these will be too close to the Sun to be observed with the X-ray telescope, but the remaining 70 bursts will be located to better than 0.5 $`\mathrm{}`$.
The capacity of the telemetry link for Ballerina will be significantly better than available on GRANAT and on EURECA. We do not expect the telemetry to be a limiting factor for transmitting the data from the observed GRBs.
Efficient rejection of false triggers will be a high priority objective for the on-board software. The decision to slew the satellite to a new pointing will only be taken when the existence of a GRB (or X-ray transient) source has been confirmed by localizing the source consistently using two independent datasets. This is very efficient in rejecting false triggers. However, the real problem will be to manage the computational effort to search for two consistent localizations in the many data set combinations possible from a marginal trigger.
Satellite attitude slews will be executed solely by controlling the speed of each of the four WATCH modulators. Magneto-torquers will be used for momentum dumping.
### 2.2 The Grazing Incidence X-Ray Telescope
This instrument will be a smaller version of the instrument used on ROSAT with a focal length of 60 cm equipped with a CCD focal plane detector. Our design is based on a telescope with nested mirror shells, the required effective area being 50 $`\text{cm}^2`$. The field of view is $`1.5\mathrm{°}`$, and the mirror shape will be optimized for achieving a rms resolution $`<30\mathrm{}`$ over the entire field. The mirror fabrication technique is similar to the one, which has been used with great success for the mirrors for Beppo–SAX and for ESA’s XMM mission. The energy response will be in the range 0.5–2 keV.
The sensitivity of the Ballerina telescope will not be much different from the sensitivity of the Beppo–SAX telescopes. Consequently, it should be possible to follow the afterglows for 24–48 hours. In fact, the bursts detected by the Ballerina all-sky monitor will on average be brighter, than the bursts seen by the SAX Wide Field Cameras. Therefore, we expect that also the afterglows will be brighter, and may be followed for a longer time. The orbit of Ballerina will permit up to 7 hours of uninterrupted observations of an afterglow source. This will allow to detect possible deviations from the simple power law decay of the afterglow.
### 2.3 Burst Operations Sequence
The satellite normally operates in a three-axis stabilized mode, performing survey observations and follow-up on previously detected bursts, waiting for a new burst to occur. The GRB is detected by one of the four wide-field cameras. An initial position with accuracy better than $`1\mathrm{°}`$ is derived, and transmitted to the ground. A slew is initiated to acquire the afterglow with the pointed X-ray telescope. The satellite will be able to slew to the new target in 20–70 s, depending on the distance of the slew. A fine burst position (better than $`30\mathrm{}`$) is then determined.
The observations of the decaying afterglow are automatically scheduled to take spacecraft constraints into consideration. The full sky is accessible, except a cone around the Sun of $`45\mathrm{°}`$ half angle. Real time communication (at a low data rate) is continuously available. Source position information will be downlinked and distributed to the community as soon as it is available.
Modifications to the observation plan may also be uploaded from the ground.
## 3 Conclusion
Cosmic gamma ray bursts is a continuing enigma for astrophysics. We are proposing Ballerina, a small satellite to provide accurate GRB positions at a rate an order of magnitude larger than Beppo–SAX. The Ballerina spectral data will place powerful constraints on the theoretical work being done to understand GRBs. Establishing the cosmological distances to GRBs, and understanding their nature will provide us with a new and independent probe of the structure of the Universe – complementary to supernovae, quasars and clusters of galaxies.
###### Acknowledgements.
Part of this work was supported by the Danish Space Advisory Board
|
no-problem/9906/hep-ph9906354.html
|
ar5iv
|
text
|
# Are electroweak corrections at 1 TeV under control at the 1 % level?
## 1 Introduction
High precision experiments at LEP have been able to prove the quantum structure of the electroweak theory at the per mille level . These experiments have tested the Standard Model (SM) electroweak sector at energies close to the Z mass $`M_Z91`$ GeV. The typical magnitude of SM electroweak corrections is dictated at LEP by the perturbative series expansion parameter $`\frac{\alpha (M_Z)}{4\mathrm{sin}^2\theta _w\pi }2.7\times 10^3`$, where $`\alpha (M_Z)`$ is the QED effective coupling constant at the energy $`M_Z`$ and $`\mathrm{sin}\theta _w`$ is the Weinberg angle. Since the experimental accuracy is of comparable magnitude, the well known one-loop electroweak corrections are sufficient in general to allow for a comparison between theory and experiment. As an exception, some leading two loop electroweak corrections growing with the top mass $`m_t`$ turn out to be also relevant at LEP.
While experiments at LEP have tested the electroweak theory at its characteristic energy of about 100 GeV, future experiments will go much beyond this mass scale. This holds in particular for the generation of linear colliders . These colliders will feature high luminosities, allowing for precision experiments at energies ranging from 500 GeV to 2 TeV. The calculation of electroweak corrections at such high energies with a precision comparable to the experimental accuracy is then an important issue. Recent results seem to indicate that yet uncalculated higher order electroweak effects and/or their possible resummation are indeed important for future linear colliders.
## 2 IR divergences and double logs
What is the order of magnitude of electroweak corrections that one expects at a typical energy of, say, 1 TeV? Let us assume that we determine the SM parameters with high precision through a series of LEP experiments at the Z mass. Then we expect that perturbative corrections for an observable measured at a different c.m. energy $`\sqrt{s}`$ are enhanced by large logarithms of ultraviolet (UV) origin of the form $`\frac{\alpha (M_Z)}{4\mathrm{sin}^2\theta _w\pi }\mathrm{log}\frac{s}{M_Z^2}1.3\times 10^2`$ for $`\sqrt{s}=1`$ TeV. Since the one loop effects are of the order of 1 %, we expect higher order effects to be of the order of 0.1 %. Moreover, the large logarithms can be resummed at all orders through renormalization group equations (RGEs). Then, if we fix the expected experimental accuracy to be at the 1 % level at NLCs (which is probably a conservative assumption since the accuracy is expected to be better than this ), there is no need to worry at all: electroweak corrections are under control, i.e. they are theoretically known through one-loop results with an accuracy which is better than the experimental one.
However, this is not the end of the story. As has already been noticed, electroweak corrections also contain terms growing with the energy $`\sqrt{s}`$ like the square of a log, i.e. proportional to $`\mathrm{log}^2\frac{s}{M_{Z,W}^2}`$. This can be understood as follows: when the energy is much bigger than the mass of all the particles running in the loops, which means $`\sqrt{s}>>M_W,M_Z`$ if we don’t consider processes in which the top quark plays a role, the W and Z mass act as an effective cutoff for infrared (IR) divergences. Infrared divergences arise in perturbative calculations from regions of integration over the loop momentum $`k`$ where $`k`$ is small compared to the typical scales of the process. This is a well known fact in QED for instance where the problem of an unphysical divergence is solved by giving the photon a fictitious mass which acts a a cutoff for the IR divergent integral. When real (bremsstrahlung) and virtual contributions are summed, the dependence on this mass cancels and the final result is finite . The (double) logarithms coming from these contributions are large and, growing with the scale, can spoil perturbation theory and need to be resumed. They are usually called Sudakov double logarithms . In the case of electroweak corrections, similar logarithms arise when the typical scale of the process considered is much larger than the mass of the particles running in the loops, typically the $`W(Z)`$ mass . The expansion parameter results then $`\frac{\alpha }{4\mathrm{sin}^2\theta _w\pi }\mathrm{log}^2\frac{s}{M_W^2}`$, which is already 7 % for for energies $`\sqrt{s}`$ of the order of 1 TeV. In the case of corrections coming from loops with $`W(Z)`$s, there is no equivalent of “bremsstrahlung” like in QED or QCD: the $`W(Z)`$, unlike the photon, has a definite nonzero mass and is experimentally detected like a separate particle. In this way the full dependence on the $`W(Z)`$ mass is retained in the corrections. Let us consider IR divergences coming from vertex corrections for instance (also box diagrams are present in two fermion production, but I do not discuss them here). For simplicity, I consider a “SM-like case” in which a “W boson” having mass $`M`$ and coupling with fermions like the photon is exchanged. In the limit of massless fermions considered here, there is no coupling to the Higgs sector. Moreover, by power counting arguments, it is easy to see that the vertex correction where the trilinear gauge boson coupling appears is not IR divergent. The only potentially IR divergent diagram is then the one of fig. 1, where a gauge boson is exchanged in the t-channel. It is convenient to choose the momentum of integration $`k`$ to be the one of the exchanged particle, the boson in this case. Then, by simple power counting arguments it is easy to see that the IR divergence can only be produced by regions of integration where $`k0`$. The only potentially IR divergent integral is then the scalar integral, usually called $`C_0`$ in the literature . Any other integral with $`k_\mu ,k_\mu k_\nu `$ in the numerator cannot, again by power counting, be IR divergent. To understand the origin of the divergences, let us consider the diagram of fig.1 with all the masses set to zero. For $`k0`$ the leading term of the vertex amplitude is given by:
$$𝒱\frac{\alpha }{4\pi }𝒱_0\frac{d^4k}{i\pi ^2}\frac{(p_1p_2)}{k^2(kp_1)(kp_2)}\frac{\alpha }{2\pi }𝒱_0_0^1\frac{dx}{x}_0^{1x}\frac{dy}{y}$$
(1)
where $`𝒱_0`$ is the tree level vertex. We can see here the two logarithmic divergences that arise from the integration over the $`x,y`$ Feynmann parameters. As is well known , one of them is of collinear origin and the other one is a proper IR divergence. When we take some of the external squared momenta and/or masses different from zero, they serve as cutoffs for the divergences. The bottom line is that the Feynman diagram of fig. 1 produces a term proportional to $`\alpha \mathrm{log}^2\frac{s}{M^2}`$, where M is the exchanged boson mass. It is easy to see that the dependence on the IR logs simply factorizes for the cross section ($`\sigma _B`$ is the tree level cross section):
$$\sigma \frac{1}{s}_0^s\frac{dt}{s}|_0|^2[12\frac{\alpha }{4\pi }\mathrm{log}^2\frac{s}{M^2}]=\sigma _B[12\frac{\alpha }{4\pi }\mathrm{log}^2\frac{s}{M^2}]$$
## 3 Asymptotic behavior of two fermion processes
In and , the production of two massless fermions in an high energy lepton collider has been considered. In the coefficients of the leading terms, growing with the energy like the square of a log as we have seen, have been calculated. In also the coefficients of the subleading terms, growing like a single log and of collinear origin, have been calculated for various observables. In the following, I call “Sudakov-type” logs the single and double logs of IR and collinear origin, to distinguish them from the logs of UV origin. As an example, let me consider the total crossection for the process $`e^+e^{}\mu ^+\mu ^{}`$. From , we get:
$$\sigma _\mu \stackrel{\sqrt{s}>>M}{}\sigma _B\{1+\frac{\alpha }{4\pi \mathrm{sin}^2\theta _w}[0.6L_{UV}+9.4L_{IR}1.4L_{IR}^2]\}$$
(2)
Here, $`L_{UV}`$ and $`L_{IR}`$ are numerically the same: $`L_{UV}=L_{IR}=\mathrm{log}\frac{s}{M^2}`$ and M is the weak scale $`MM_WM_Z91`$ GeV. $`\sigma _B`$ is the Born cross section, precisely defined in . This formula is expected to describe the full one loop calculation better and better as the energy grows, since the subleading terms that have not been extracted become less and less important with respect to the leading logarithmic ones. Indeed, formula 2 well approximates the exact result coming from numerical programs for energies around 1 TeV(see ). However, for energies well above 1 TeV where the agreement is supposed to be even better, no numerical computation is available at the moment. The graph corresponding to eq. 2 is drawn in fig. 2, where the relative deviation for the total cross section $`\frac{\mathrm{\Delta }\sigma }{\sigma }\frac{\sigma \sigma _B}{\sigma _B}`$ is drawn as a function of the c.m. energy, and the various contributions are also separately plotted.
One evident feature of eq. 2 is that, while the coefficients of the single UV log and the double Sudakov log are of order 1, as one could expect a priori, the coefficent of the single Sudakov-type single log is of order 10. This has two immediate consequences:
* The contribution of the single log of UV origin is almost negligible with respect to the Sudakov logs ones. Thus, a naïve expectation of an asymptotic behavior dictated by the UV structure of the theory turns out to be wrong.
* Since the sign of the double and single Sudakov logs are opposite, there are big cancellations and the correction to $`\sigma _\mu `$ crosses a zero at an energy of about 2 TeV
These features are most easily seen by looking at fig. 2, where the net result is seen to result from cancellations of big contributions of Sudakov single and double logs of opposite signs, while the RGE driven logs are almost negligible.
Where does all this leave us with the question posed with the title? In the end, one could think that when doing perturbative calculations, big cancellations can always be present (between different graphs contributing to the same amplitude for instance). Then, since here the net effect is only a few percent in the considered energy range as one can see from fig. 2, there is no need to worry about higher order effects. However, the situation here is different in my opininon. Here we have terms that are separately gauge invariant and have different energy behavior. The fact that their contribution almost exactly cancels at an energy of about 2 TeV which is close to the energy of interest, is to be taken as accidental. Let us take another point of view: the relative effect of double logs is, from eq. 2, $`1.4\frac{\alpha }{4\pi \mathrm{sin}^2\theta _w}\mathrm{log}^2\frac{s}{M^2}0.1`$ at 1 TeV. A two loop calculation will produce a term growing like the fourth power of a log, of the order $`(0.1)^2=1`$ %. Until an higher order calculation will be done, one cannot say that electroweak corrections at 1 TeV are under control at the 1 % level.
## References
|
no-problem/9906/cond-mat9906254.html
|
ar5iv
|
text
|
# Percolation of Superconductivity.
## I introduction
In many attempts to construct a viable model for High Temperature Superconductors the notion of negative – $`U`$ centers is invoked . In this connection there is a simple, natural question that arises: How many such centers are needed to make a superconductor. In this contribution we shall argue that under certain circumstances there is a critical concentration $`c_0`$ below which there is no superconducting order. Moreover, we developed a strategy for investigating the factors which determine $`c_0`$.
In order to deal with a well posed problem we shall study a Random – $`U`$ Hubbard Model defined by the Hamiltonian
$$H=\underset{ij\sigma }{}t_{ij}c_{i\sigma }^+c_{j\sigma }+\underset{i\sigma }{}U_ic_{i\sigma }^+c_{i\sigma }c_{i\sigma }^+c_{i\sigma }\mu \underset{i\sigma }{}c_{i\sigma }^+c_{i\sigma },$$
(1)
where the coupling constant
$$U_i=\{\begin{array}{cc}\hfill |U|& \mathrm{with}\mathrm{probality}c\hfill \\ \hfill 0& \mathrm{with}\mathrm{probability}1c\hfill \end{array}.$$
The question we shall ask is: Is there a finite concentration $`c_0`$ such that for $`c<c_0`$ the cofigurationally averaged, superconducting long range order parameter $`\overline{\chi }`$ vanishes even at zero temperature?
As is natural we define $`\overline{\chi }`$ by the relation
$$\overline{\chi }=\frac{1}{N}\underset{i}{}\chi _i,$$
(2)
where the local pairing amplitude is given by
$$\chi _i=<c_ic_i>.$$
(3)
Following the conventional notation the bracket $`<\mathrm{}>`$ denotes a thermodynamic average and $`\overline{\mathrm{\Theta }}`$, for an arbitrary operator $`\widehat{\mathrm{\Theta }}`$, implies the average of $`\widehat{\mathrm{\Theta }}`$ over all configurations $`U_i`$, such that the fraction of negative $`U`$ sites is c, with equal weight. A sample will be said to be superconducting if $`\overline{\chi }0`$ This implies that $`\chi _i0`$ on a finite fraction of all sites. Namely, if $`\chi _i0`$ only on a finite number of sites $`\overline{\chi }`$ will go to zero as $`N\mathrm{}`$ and the system will be regarded as not superconducting.
To make progress we calculate $`\chi _i`$ within the Hartree – Fock – Gorkov (HFG) Decoupling scheme for the Greens functions and the averaging over the $`U`$–configurations is accomplished with the help of the Coherent Potential Approximation (CPA) . In short, at the risk of missing some important feature of the problem, like localization of electrons, we develop a mean field theory for the phenomena described by $`H`$ in Eq. (1) This approach may be justified by noting that little is known about the problem at hand systematically and hence as a preliminary study a mean field theory is called for.
Note that the simplest approximations to the problem would be to set $`U_i`$ at each site equal to its average value $`\overline{U}=cU`$. In some contexts this is called the Virtual Crystal Approximation . Since, as is well known , any amount of attraction leads to superconductivity, $`\overline{U}=cU`$ implies superconductivity for all non zero concentrations with the transition temperature $`T_C`$ decreasing albeit non-analytically, with $`c`$. Thus before setting out the details of the above theory it is worthwhile to pause, briefly, to consider a number of fairly general arguments which suggest that the above conclusion is premature and that there is a critical concentration $`c_0`$ of negative $`U`$ – centers for superconductivity.
i. Classical Percolation Theory for a mixture of two metals with resistivities $`\rho _1`$ and $`\rho _2`$ have been studied in the Effective Medium Approximation . For $`\rho _1=\rho _0`$ and $`\rho _2=0`$ (Fig. 1), namely in the case where metal 2 is a superconductor, it yields an effective resistivity given by
$$\rho _{eff}=\{\begin{array}{cc}\hfill \rho _0(1dc)& \mathrm{for}c<c_0=\frac{1}{d}\hfill \\ & \\ \hfill 0& \mathrm{for}c>c_0=\frac{1}{d}\hfill \end{array},$$
(4)
where $`d`$ is the number of spatial dimension in which percolation is allowed. Thus, this model predicts critical concentrations in $`d=1,2`$ and 3 dimensions, More over, $`c_0`$ depends on the dimensionality $`d`$. More generally $`\rho _{eff}(cc_0)^s`$ near $`c_0`$ but a mean field theory cannot be expected to deal with the critical exponent $`s`$ adequately.
ii. The propagation of Cooper pairs between negative $`U`$ centers by hopping from site to site, where $`U=0`$ on the intermediate sites, is depicted in Fig. 2. Assuming that the distance between two negative–$`U`$ centers is $`c^{1/d}`$, in units of the lattice constant $`a`$ on a $`d`$ dimensional lattice, we estimate the number of individual hops $`l`$ necessary to reach one such center from its nearest neighbors. Assuming random walk $`c^{1/d}=l^{1/(d1)}`$ (for $`d>1`$). If each hop takes $`\mathrm{}/W`$ seconds where $`W`$ is the bandwidth for the Cooper pairs, the time to travel between two negative $`U`$ centers is given by $`\tau =(\mathrm{}/W)c^{(d1)/d}`$. Now we note that in between two centers the Cooper pair is without its binding energy $`U`$. Consequently, such travel is allowed only for such times $`\delta t`$ that the energy uncertainty $`\delta E=\mathrm{}/\delta t>U`$. Taking $`\delta t=\tau `$ we conclude that for $`\delta E=Wc^{(d1)/d}>U`$ the pair will propagate for $`\delta E=Wc^{(d1)/d}<U`$ the pair will not propagate. Thus for $`c<c_0`$, where
$$c_0=\left(\frac{U}{W}\right)^{{\displaystyle \frac{d}{d1}}},$$
(5)
a system of negative $`U`$ centers will not be superconducting. Presumably, the Cooper pairs will be localized. On the other hand for $`c>c_0`$ it will be a superconductor.
iii. Localization of Cooper pairs by local charge and order – parameter – phase fluctuation is the third argument which we wish to recall briefly. It was explored in the present context by Doniach and Inui . In a Ginzburg Landau theory on a lattice the relative phases of the local order parameters $`\psi _i=|\psi _i|\mathrm{e}^{\mathrm{i}\mathrm{\Theta }_\mathrm{i}}`$ are determined by the quadratic term in the free energy function $`F(\{\psi _i\})`$. This may be written in the form of Josephson coupling energies $`F(\{\mathrm{\Theta }_i\})=\frac{1}{2}_{ij}E_{ij}^J\mathrm{cos}(\mathrm{\Theta }_i\mathrm{\Theta }_j)`$, where the precise relationship of the coefficients to various parameters of the theory need not concern us here. To describe charge fluctuations associated with Cooper pairs arriving and leaving a site a charging energy needs to be added to $`F`$. Because, the local potential is related to the phase by Josepson voltage relation $`V_i=\frac{\mathrm{}}{2\mathrm{e}}\dot{\mathrm{\Theta }}_i`$ this has the form of a kinetic energy term. Finally, to recover a microscopic description the local phases are treated as quantum mechanical variables by the Hamiltonian
$$H=\frac{1}{2}E_C\underset{i}{}\left(i\frac{}{\mathrm{\Theta }_i}\right)^2\frac{1}{2}E_J\underset{ij}{}\mathrm{cos}(\mathrm{\Theta }_i\mathrm{\Theta }_j).$$
(6)
This is a much studied Hamiltonian in connection with granular superconductivity. In particular it was investigated by Gosset and Györffy in the Hartree – approximation. In short they factorized the wave – function as shown below
$$\mathrm{\Psi }(\{\mathrm{\Theta }_i\})=\underset{i}{}\varphi _i(\mathrm{\Theta }_i)$$
(7)
and found the following self–consistent equation for the individual site wave function $`\varphi _0(\mathrm{\Theta })`$
$$\left[\frac{1}{2}E_C\left(\frac{}{\mathrm{\Theta }}\right)^2E_J\mathrm{cos}(\mathrm{\Theta }\overline{\mathrm{\Theta }})\right]\varphi _0(\mathrm{\Theta })=E_0\varphi _0(\mathrm{\Theta }),$$
(8)
where
$$e^{i\mathrm{\Theta }}=d\mathrm{\Theta }\varphi _0^{}\mathrm{e}^{\mathrm{i}\mathrm{\Theta }}\varphi _0(\mathrm{\Theta })=\rho \mathrm{e}^{\mathrm{i}\overline{\mathrm{\Theta }}}$$
The amplitude $`\rho `$, determined by solving the above equation numerically is shown in Fig. 1 (in Ref. ),as a function of the ratio $`E_J/E_C`$ ($``$ Josephson energy/charging energy) For $`E_J/E_C<0.125`$ we find $`\rho =0`$ and hence we conclude that the system of point superconductors we have been considering do not have long range superconducting order. Clearly, it is tempting to associate $`E_J`$ with the coupling between the negative – $`U`$ centers in our Hubbard model and assume that it goes to zero as $`c0`$. Evidently, this would imply a critical concentration determined by $`E_C=E_J(c_0)`$ . In short, charge fluctuations can destroy the phase coherence of superconducting order parameter if the coupling between the negative – $`U`$ centers drops below certain critical value. Indeed this was one of the main point of the paper by Doniach and Inui . In what follows we shall develop a strategy for investigating the link between the microscopic model defined by Eq. (1) and the above semi-phenomenological arguments.
In concluding this introduction we note that the specific task we shall undertake is a contribution to the general problem of treating disorder and electron–electron interactions simultaneously For a comprehensive discussion of the relevant issues in this field the reader is referred to the relatively recent review article by Belitz and Kirkpatrick .
## II The Coherent Potential Approximation for the Random–$`U`$ Hubbard Model
The physics described by this simple model appears to be exceedingly rich. For instance, one might expect that, under some circumstances, the Cooper pairs are subject to Anderson localization and hence they form a random set of Andreev scatterers for the quasi–particles . Such system of scattering centers may then Anderson localize the quasi–particles themselves and turn the system into an insulator below the critical concentration $`c_0`$ for superconductivity. However very little systematic fully microscopic work has been done on the problem and hence, as a preliminary exercise, a mean field theoretic treatment is called for even at the risk of failing to capture some of its important features. In any case, as we shall show, even such limited description turns out to be of physical interest.
Formally, the task is to find the Greens function
$$𝑮(i,j;\tau ;\{U_i\})=\left[\begin{array}{cc}<T\{c_i(\tau )c_i^+(0)\}>& <T\{c_i(\tau )c_i(0)\}>\\ <T\{c_i^+(\tau )c_i^+(0)\}>& <T\{c_i(\tau )c_i(0)\}>\end{array}\right],$$
(9)
where the creation and annihilation operators $`c_{i\sigma }(\tau )`$ and $`c_{i\sigma }^+(\tau )`$ evolve in complex time $`\tau `$ according the random–$`U`$ Hamiltonian $`H`$ in Eq. (1), $`T`$ is the $`\tau `$–ordering operator, brackets $`<\mathrm{}>`$ denote here the usual equilibrium thermal averages corresponding to $`H`$, and average the result with respect to all arrangement of the $`U`$-centers each denoted by $`\{U_i\}`$. In short we wish to find
$$\overline{𝑮}(i,j;\tau )=\underset{\{U_i\}}{}P(\{U_i\})𝑮(i,j;\tau ;\{U_i\}),$$
(10)
where the probability distribution is assumed to be of form
$`P(\{U_i\})`$ $`=`$ $`{\displaystyle \underset{i}{}}P(U_i)`$ (11)
$`\mathrm{with}P(U_i)`$ $`=`$ $`\{\begin{array}{c}\hfill c\mathrm{for}U_i=U\\ \hfill 1c\mathrm{for}U_i=0\end{array}.`$ (14)
Note that the local order parameter defined by Eq. (3) is given by
$$\overline{\chi _i}=\overline{G}_{12}(i,i;\tau =0^+),$$
(15)
and hence the knowledge of the averaged one particle Greens functions matrix is sufficient to address the question whether or not there is superconducting long range order at a given concentration $`c`$.
As we have indicated above we shall now proceed to a mean field approximation to the above problem. This consists of two steps. Firstly, we make use of the Hartree–Fock–Gorkov decoupling scheme to find the following ’mean-field’ equation of motion
$$\underset{l}{}\left[\begin{array}{c}(\omega _n+\mu \frac{1}{2}U_in_i)\delta _{il}+t_{il}\mathrm{\Delta }_i\delta _{il}\\ \mathrm{\Delta }_i^{}\delta _{il}(\omega _n\mu +\frac{1}{2}U_in_i)\delta _{il}t_{il}\end{array}\right]𝑮(l,j;\omega _n)=\mathrm{𝟏}\delta _{ij},$$
(16)
where
$`n_i`$ $`=`$ $`{\displaystyle \frac{2}{\beta }}{\displaystyle \underset{n}{}}\mathrm{e}^{\mathrm{i}\omega _n\delta }G_{11}(i,i;\omega _n),`$ (17)
$`\chi _i`$ $`=`$ $`{\displaystyle \frac{1}{\beta }}{\displaystyle \underset{n}{}}\mathrm{e}^{\mathrm{i}\omega _n\delta }G_{12}(i,i;\omega _n),`$ (18)
$`\mathrm{\Delta }_i`$ $`=`$ $`U_i\chi _i.`$ (19)
Secondly, we find average of the solution to Eq. (15), namely $`𝑮(i,j;\omega _n;\{u_i\})`$, over all $`U`$–configurations using the Coherent Potential Approximation (CPA). The justification for this second step is that the CPA is well known to be a reliable mean–field theory of disorder for wave propagation in a medium described by independent random variables .
To implement the CPA we rewrite Eq. (15) in the Dyson form
$$𝑮(i,j;\omega _n)=𝑮^0(i,j;\omega _n)+\underset{l}{}𝑮^0(i,l;\omega _n)𝑽_l𝑮(l,j;\omega _n),$$
(20)
where
$$V_l=\left(\begin{array}{cc}\frac{1}{2}U_ln_l& \mathrm{\Delta }_l\\ \mathrm{\Delta }_l^{}& \frac{1}{2}U_ln_l\end{array}\right).$$
(21)
The CPA recipe for $`\overline{𝑮}(i,j;\omega _n)`$ is to set it equal to the coherent Greens function $`𝑮^C(i,j;\omega _n)`$ which is the solution of Eq. (20) for the case where the random potential $`𝑽_l`$ is replaced by the energy dependent, complex coherent potential $`𝚺(\omega _n)`$, the same on every site. To determine the coherent potential (self– energy) we study, in turn a $`U_i=|U|`$ impurity in the coherent lattice. On the impurity site at $`i`$ we find
$$𝑮^\alpha (i,i;\omega _n)=\left[1𝑮^C(i,i;\omega _n)𝑽_i^\alpha \mathrm{\Sigma }(\omega _n)\right]^1𝑮^C(i,i;\omega _n),\mathrm{for}\alpha =0\mathrm{and}U,$$
(22)
where
$$𝑽_i^{\alpha =0}\mathrm{and}𝑽_i^{\alpha =U}=\left(\begin{array}{c}\frac{1}{2}Un_i\mathrm{\Delta }_i\\ \mathrm{\Delta }_i^{}\frac{1}{2}Un_i\end{array}\right).$$
(23)
Then, the usual CPA condition which determines the self–energy $`𝚺(\omega _n)`$ is given by
$$c𝑮^{(0)}(i,i;\omega _n)+(1c)𝑮^{(U)}(i,i;\omega _n)=𝑮^C(i,i;\omega _n).$$
(24)
Similar equations have been used to describe random superconductors by Lustfeld and more recently by Litak et al . The principle difference between our present concerns and that of these earlier authors is that we are focusing on the randomness of the interaction parameter $`U_i`$ and not on the random site energies $`ϵ_i`$ as was their aim. To put it on other way we are studying a problem analogues to that of ’spin–glass’ rather than that of dirty superconductors.
Equations (22,23) and Eq. (24) together with
$`n^\alpha `$ $`=`$ $`{\displaystyle \frac{2}{\beta }}{\displaystyle \underset{n}{}}\mathrm{e}^{\mathrm{i}\omega _\mathrm{n}\delta }\mathrm{G}_{11}^\alpha `$ (25)
$`\chi ^\alpha `$ $`=`$ $`{\displaystyle \frac{2}{\beta }}{\displaystyle \underset{n}{}}\mathrm{e}^{\mathrm{i}\omega _\mathrm{n}\delta }\mathrm{G}_{12}^\alpha `$ (26)
$`\mathrm{\Delta }^U`$ $`=`$ $`U^\alpha \chi ^\alpha `$ (27)
$`\overline{n}`$ $`=`$ $`cn^{(U)}+(1c)n^{(0)}`$ (28)
$`\overline{\chi }`$ $`=`$ $`c\chi ^{(U)}+(1c)\chi ^{(0)},`$ (29)
where $`\alpha =0`$ and $`U`$ as before (Eq. 22), are the fundamental equations of our theory.
Manipulating the $`CPA`$ equations yield the following gap equation:
$`\overline{\chi }`$ $`=`$ $`{\displaystyle \frac{U}{\beta }}{\displaystyle \underset{n}{}}\mathrm{e}^{\mathrm{i}\omega _n\delta }\left[{\displaystyle \frac{c}{2\omega _n}}\mathrm{Tr}\left\{𝑮^{(U)}\right\}{\displaystyle \frac{𝑮^{(U)}}{𝑮^C}}+𝑮^{(U)}\left(c{\displaystyle \frac{2\omega _n\mathrm{Tr}𝚺}{2\omega _n}}{\displaystyle \frac{𝑮^{(U)}}{𝑮^C}}1\right)\right]\overline{\chi }.`$ (30)
(31)
In what follows we present results of solving the above equations numerically for various interesting regimes. Of particular interest is the large $`U`$ limit. As $`U_i`$ change its values form 0 to $`|U|`$ there exists Mott–Hubbard metal–insulator transition for large enough interaction $`|U|`$. An other interesting feature of the problem at hand is that fluctuations of pairing potentials $`\mathrm{\Delta }_i`$, which changes randomly from 0 to $`\mathrm{\Delta }^U`$, invalidate the Anderson theorem and hence states appear in the gap .
## III ORDER PARAMETER FLUCTUATIONS
At first we have calculated $`T_C`$ and $`\overline{\chi }`$ for zero temperature $`(T=0)`$ by means of VCA where, as we mentioned in the introduction effective interaction between electrons $`U_{eff}=cU`$. Figs. 3$`a`$ and 3$`b`$ show the critical temperature $`T_C^{(c)}`$ normalized to the corresponding quantities of the clean system with $`U`$ on every site, namely $`T_C(c=1)`$ and averaged pairing parameter $`\overline{\mathrm{\Delta }}(T=0,c=1)`$, calculated for effective interaction $`U_{eff}`$ respectively as functions of concentration $`c`$. Calculations were done for various values of the interaction parameter $`U/W`$: -0.3, -0.5, -1.0, -2.0 and for a half filled band: $`n=1`$. One can see that for these approximations there is no evidence of critical concentration $`c_0>0`$ below which the systems is normal at $`T=0`$ i.e. no percolation.
As is clear from Eq. (16), $`U_i`$ fluctuating between 0 and $`|U|`$ has two distinct direct consequences. On the one hand it causes the Hartree potential $`\frac{1}{2}U_in_i`$ to fluctuate. On the other it gives rise to a fluctuating pairing parameter – $`\mathrm{\Delta }_i`$. As it turns out these two effects have very different influence on the solutions to Eqs. (23,24,26,31). Therefore, we examine them separately. As disorder was treated by CPA, at first we made calculations after neglecting Hartee potential $`\frac{1}{2}U_in_i`$, in Eq. (23) and studied the case of order parameter fluctuation on their own. This means that we took the impurity potential in Eq. (23) to be
$$𝑽_l=\left(\begin{array}{cc}0& \mathrm{\Delta }_l\\ \mathrm{\Delta }_l^{}& 0\end{array}\right).$$
(32)
In Figures 4$`a`$, $`b`$, $`c`$ we show the critical temperature $`T_C/T_C(c=1)`$ ($`a`$), the order parameter (for $`T=0`$) $`\overline{\chi }/\chi (c=1)`$ ($`b`$) and the local pairing potential on $`U`$ site $`\mathrm{\Delta }_U/\mathrm{\Delta }_U(c=1)`$ ($`c`$), versus concentration of negative centers $`c`$ for $`n=1`$. Calculations were done by means of CPA neglecting the Hartee term and using the same values of interaction parameters as in Fig. 3.
Surprisingly, our simplified $`CPA`$ results agree with the $`VCA`$ argument in the introduction in as much as we found non–zero local order parameter on both the $`U=0`$ and $`U<0`$ sites at all concentrations $`c0`$. That is to say we obtained finite $`\overline{\chi }0`$ and $`T_C`$ for any value of concentration $`c`$ and interaction $`U_i<0`$ and no evidence of percolation. The order parameter $`\overline{\chi }`$ increases gradually from 0 to its maximal value with changing the concentration $`c`$. Interestingly, in the large $`U`$ limit, $`|U|/W>0.5`$, $`T_C`$ and $`\mathrm{\Delta }_U`$ are nearly constant for various $`c`$ and they reach large finite values for arbitrary small concentrations of negative centers $`c`$.
To return to the full CPA solution we have used the full impurity potential $`𝑽_i^{\alpha =U}`$ as in Eq. (23). Figures 5$`a`$, $`b`$, $`c`$, $`d`$ show the critical temperature $`T_C^{(c)}/T_C(c=1)`$ ($`a`$), the order parameter $`\overline{\chi }/\chi (c=1)`$ ($`b`$) and the local pairing potential on $`U`$ site $`\mathrm{\Delta }_U/\mathrm{\Delta }_U(c=1)`$ ($`c`$), versus concentration of negative centers $`c`$ for $`n=1`$ and the same interactions as in Figs 3 and 4. Here one can clearly see that all that quantities $`T_C`$, $`\overline{\chi }(T=0)`$ and $`\mathrm{\Delta }_U`$ are tending to zero for some small enough concentration of negative $`U`$ centers $`c_0`$. Below this critical concentration the system is normal. For larger interaction ($`U/W=`$ -1.0, -2.0) $`c_0=0.5`$ and the order parameter scales as $`\chi (cc_0)^{1/2}`$. Decreasing $`|U|`$ we observe systematic decrease of $`c_0`$.
To investigate the case of critical concentration $`c_0`$ we have studied the density of quasiparticle states both in superconducting and in the normal states. In the latter case, for large enough interaction $`|U|>0.5W`$, there exists a band splitting in the system. With changing the concentration $`c`$ we observe Mott metal–insulator transition. It is caused by large fluctuations of Hartree term $`\frac{1}{2}U_in_i`$ (Eq. 16) as in the original paper of Hubbard . In Figs. 6 $`a`$, $`b`$, $`c`$ we plotted the densities of states (full line) and the local density of states on $`U`$ site (dashed line) for $`U/W=2.0`$, $`n=1`$ and $`T=0`$ for different concentrations $`c`$: $`c=0.4`$ ($`a`$ \- normal metal), $`c=0.5`$ ($`b`$ \- insulator) and $`c=0.6`$ ($`c`$ \- superconductor). The Fermi energy in these plots: $`ϵ_F=\mu =0`$. Thus changing $`c`$ from 0 to 1 system changes from normal metal (Fig. 6$`a`$) to a superconductor (Fig. 6$`c`$) through an insulator (Fig. 6$`b`$). Remarkably, for a low concentration of negative centers (Fig. 6$`a`$) $`c=0.4<c_0`$ (here $`c_0=0.5`$) in spite of finite and relatively large value of averaged density of states at the Fermi energy: $`\overline{D}(0)=\frac{1}{\pi }\mathrm{Im}G_{11}^C(0+\mathrm{i}\delta )`$, the local density of states on $`U`$ sites $`D^U(0)=\frac{1}{\pi }\mathrm{Im}G_{11}^U(0+\mathrm{i}\delta )`$ appears to be extremely small. Evidently the doubly occupied states form a lower ’Hubbard’ band split off from the upper band which is associated with the singly occupied sites.
This effect has been further investigated for other band fillings. The transition from normal to superconducting phase occurred for each of band fillings $`n`$ at some, specific, critical concentration $`c_0(n)`$. Figures 7$`a`$, $`b`$, $`c`$, $`d`$ show simultaneously the order parameter $`\overline{\chi }`$ ($`a`$), the local pairing potential on $`U`$ site $`\mathrm{\Delta }_U`$ ($`b`$), the local charge on $`U`$ site $`n_U`$ ($`c`$) and the chemical potential $`\mu `$ ($`d`$) plotted versus concentration $`c`$ for $`U/W=2.0`$, at $`T=0`$ and several values of $`n`$ ($`n=`$ $`1.8`$, $`1.6`$, $`1.4`$, $`1.2`$, $`1.0`$, $`0.8`$, $`0.6`$, $`0.4`$, $`0.2`$) Interestingly, that transition from superconducting to normal phase is accompanied by a large value of local charge occupation $`n_U`$ (Fig. 7$`c`$) and large jump of a chemical potential $`\mu `$ (from one subband to another) near $`c_0`$ (Fig. 7$`d`$). It appears that for $`c`$ below $`c_0`$ $`n_U2`$. Namely, every $`U`$ site is doubly occupied with a pair of electrons (Fig. 7$`c`$). Because there are no empty spare $`U`$ sites in the system these pairs cannot move. That is to say, they are localized on the $`U`$ sites.
Similar calculations have been performed for smaller interaction $`|U|`$ ($`U/W=0.5`$) The corresponding results are presented in Figs. 8 $`a`$$`d`$ respectively. Here the interaction $`|U|`$ is not large enough to create a band splitting effect but the tendency with $`\overline{\chi }0`$ is still observable as concentration $`c`$ is tending to some finite $`c_0>0`$ $`(cc_0)`$. Here $`c_0`$ is less that in former case of larger interaction $`|U|=2W`$ (Fig. 7). The occupation of negative centers is larger than $`n`$ but clearly less than 2 electrons per site. For small enough band filling $`n=0.2`$, the order parameter $`\overline{\chi }`$ was finite for all $`c`$ and we have not observed a percolation phenomenon. For larger band fillings we have obtained $`\overline{\chi }=0`$ below $`c<c_0`$ but instead of a square root behaviour $`\overline{\chi }(cc_0)^{1/2}`$ for $`c`$ close to $`c_0`$ for larger $`U`$ (U/W=-2.0 fig. 7$`a`$) here $`\chi `$ goes to zero rather in the asymptotic way (Fig 8$`a`$).
To investigate the demise of the superconducting state near $`c_0`$ we have studied the density of states in the appropriate region of parameter space. Figures 9$`a`$ and $`b`$ shows the quasiparticle densities of states ($`a`$) and the local densities of states on $`U`$ site ($`b`$) for $`U/W=0.5`$, $`n=0.4`$ and three values of $`c`$ specified in the figures. It is clearly visible how the superconducting gap is filled, due to pair breaking, in with $`c`$. Beginning from the clean system with interaction $`U`$ in every site $`c=1`$ we start from the sharp edges in the quasiparticle density of states (Fig. 9$`a`$) then for smaller values of $`c`$ ($`c=0.6`$) the gap parameter $`\mathrm{\Delta }`$ is of the same order (Fig 7$`b`$) but the real gap in the quasiparticle density of states $`\overline{D}(E)`$ changes significantly. The gap becomes smaller with smaller $`c`$ and looses its clear edges. For small enough $`c`$ ($`c=0.14`$) it nearly disappears. Clearly, the Anderson theorem for a superconuctor with nonmagnetic disorder is not satisfied in this case . As is well known, according to Anderson theorem the gap remains absolute in presence of disorder due to potential scattering provided the spatial fluctuations of $`\mathrm{\Delta }_i`$ about $`\overline{\mathrm{\Delta }}`$ are negligible . Clearly, in the random interaction case this is not true and this kind of disorder leads to pair breaking.
Thus on account for the large fluctuations of pairing potential $`\mathrm{\Delta }_i`$ in our system,due to disorder, we observe a qualitative change in quasiparticle density of states shown in Fig 9$`a`$. These fluctuations lead also to complicated gap equations where $`T_C`$ is determined not only by $`𝑮^C`$ but also by $`𝚺`$, $`𝑮^U`$ (Eq. 31).
Finally, we investigated the factors which determine the critical concentration $`c_0`$. In Fig. 10 we show $`c_0`$ as a function of band filling for two interaction parameters $`U/W=2.0`$ and $`0.5`$. In both cases function can be approximated by a straight line $`c_0=a+bn`$. In case of $`U/W=2.0`$, $`a=0`$ and $`b=0.5`$ but for $`U/W=0.5`$, $`a0.32`$ and $`b0.6`$.
## IV Conclusions
We have examined the question of percolating superconductivity in the context of a random $`U`$ Hubbard model. We have studied the case where $`U_i`$ is $`|U|`$ and 0 with probability $`c`$ and $`1c`$ respectively on a lattice whose sites are labelled $`i`$ using the Gorkov decoupling. Changing concentration $`c`$ we checked that simple averaging procedures like Virtual Crystal Approximations (VCA) do not lead to any zero temperature phase transition. Furthermore, we found that if charge fluctuations are neglected even a full mean field theory of disorder, like the CPA, does not predict a percolation transition. However, when the fluctuations in Hartree potential are included on equal footing with the fluctuations in the pairing potential $`\mathrm{\Delta }`$ and the problem is treated in the Coherent Potential Approximation a percolation phenomena, with a critical concentration $`c_0`$ of the negative $`U`$ centres, is discovered in our fully microscopic theory. For $`c<c_0`$ the lack of superconductivity is due to Mott localization of Cooper pairs and its high lights the qualitative difference between disorder in the crystal potential and the disorder in the interaction between the carriers. Having found the critical concentration $`c_0`$ we investigated its dependence on various parameters which defined the problem. In short we studied $`c_0(n,U)`$. For strong attractive interaction $`c_0=n/2`$ and $`\overline{\chi }(cc_0)^{1/2}`$ near $`c_0`$ but for smaller interaction $`\overline{\chi }0`$ (as $`cc_0`$) rather in a non polynomial manner. Calculations have been performed by a real space recursion algorithm which we developed for disordered superconductors in earlier publication.
Acknowledgements
This work has been partially supported by the Royal Society. Authors would like to thank prof. S. Alexandrov and prof. R. Micnas for helpful discussions.
|
no-problem/9906/cond-mat9906380.html
|
ar5iv
|
text
|
# Theoretical study of the (3×2) reconstruction of 𝛽-SiC(001)
## Abstract
By means of *ab initio* molecular dynamics and band structure calculations, as well as using calculated STM images, we have singled out one structural model for the $`(3\times 2)`$ reconstruction of the Si-terminated (001) surface of cubic SiC, amongst several proposed in the literature. This is an alternate dimer-row model, with an excess Si coverage of $`\frac{1}{3}`$, yielding STM images in good accord with recent measurements \[F.Semond et al. Phys. Rev. Lett. 77, 2013 (1996)\].
The reconstructions of SiC(001) surfaces have been widely studied in the last ten years, the characterization and understanding of growth mechanisms on the (001) substrate being prerequisites for technological applications. In the case of Si-terminated surfaces, several reconstructions have been found to occur; $`(2\times 1)`$, $`c(4\times 2)`$ and $`(n\times 2)_{n=3,5,7\mathrm{}}`$ periodicities have been observed in LEED, RHEED and STM measurements. Both (2$`\times `$1) and c(4$`\times `$2) reconstructions pertain to a complete Si monolayer at the top $`(\theta _{Si}=1)`$, as clearly indicated by all available experimental data. Unlike those of Si(001), the reconstructions of SiC(001) are characterized by weakly bonded, flat dimers ((2$`\times `$1) ) or by alternating symmetric dimers with different heights (c(4$`\times `$2)). Adsorption of additional Si produces successive $`(n\times 2)`$ reconstructions as a function of Si coverage, including $`(7\times 2)`$, $`(5\times 2)`$, and a combination of $`(5\times 2)`$ and $`(3\times 2)`$ periodicities.
The $`(3\times 2)`$ reconstruction seems to be the last stage before self-limitation of growth. Its atomic configuration and electronic structure are not clearly established, though they have been intensively investigated. Three different atomic configurations, depicted on Fig. 1, have been suggested in the literature. In the Double Dimer-Row (DDR) model, proposed by Dayan and possibly supported by other experimental studies, there are two Si ad-dimers on top of the full Si layer (Fig. 1.a). The resulting coverage $`\theta _{Si}=\frac{2}{3}`$ is in contradiction with the measured $`\theta _{Si}`$ value of $`\frac{1}{3}`$ reported by several groups. The straightforward extension of this model to the $`(5\times 2)`$ reconstruction is also inconsistent with the measured coverage. Moreover, this model is not supported by some STM studies. The DDR is favored by empirical molecular dynamics (MD) simulations, but the Tersoff potential used in these calculations is known to give a poor description of $`\beta `$-SiC surface reconstructions. Another model, the ADDed dimer-row (ADD), was first suggested in an early study by Hara et al. This configuration, with one Si ad-dimer per unit cell (Fig. 1.b), corresponds to the measured coverage for the $`(3\times 2)`$ and $`(5\times 2)`$ reconstructions. However, though it appears consistent with several experimental data, both empirical and *ab initio* calculations have shown that it is not energetically favored. Furthermore, STM investigations do not support this model. Another $`\frac{1}{3}`$ coverage model, the ALTernate dimer-row (ALT) (Fig. 1.c), was proposed by Yan et al. This configuration is supported both by calculations and by STM studies. However, it cannot account for the observed relation between single domain LEED patterns with $`(2\times 1)`$ and $`(3\times 2)`$ periodicities. It also fails to explain the $`(3\times 1)`$ reconstruction observed after O or H adsorption. Note that all three models involve Si ad-dimers which are perpendicular to the dimers on the underlying Si surface. Indeed, previous calculations have shown that a single parallel ad-dimer is energetically much less favored than a perpendicular one.
All three models show some discrepancies either with existing experiments or calculations, but none of them can be safely ruled out, owing to the lack of consistency between all available data. In this contribution, we report the results of self-consistent *ab initio* total energy calculations for all of the three structural models, including full geometrical optimizations. The surface energies are compared using Grand Canonical Potentials. The computed dispersion of electronic states, as well as STM images, are compared to experiments. Considering all the evidence, we conclude that the $`(3\times 2)`$ reconstruction of the Si-terminated surface of SiC(001) is best described by the ALT model.
Our calculations were performed at $`\text{T}=0`$ within the Local Density Approximation, using *ab initio* molecular dynamics codes employed in previous studies of SiC surfaces. Fully nonlocal norm-conserving pseudopotentials were used for Si (s and p nonlocality) and C (s nonlocality). The system was simulated by a slab in a periodically repeated supercell. The bottom layers were frozen in the p$`(2\times 1)`$ configuration determined earlier. Two different sets of calculations have been performed. To determine the relaxed surface atomic structures, we used a $`(6\times 4)`$ supercell with 8 atomic layers and a 10 Å vacuum region (the total number of atoms is 176 (184) for ALT and ADD (DDR) models). The plane-wave energy cutoffs for the wave functions and the charge density were 36 Ry and 130 Ry, respectively. Sums over occupied states were performed at the $`\mathrm{\Gamma }`$ point, which corresponds to 4 inequivalent k-points in the Brillouin Zone (BZ) for a $`(3\times 2)`$ cell. Next, the electronic band structure was computed in a $`(3\times 2)`$ supercell with 12 atomic layers and a 6 Å vacuum region, using an extension of ab-initio MD codes to finite wave functions vectors, and keeping all atoms fixed. Atomic positions in the six top layers and the two bottom layers were taken from the preceding ab-initio MD calculations, whereas those in the four central layers were assumed bulk-like. Wave functions and charge densities were expanded in plane waves with cutoffs of 40 Ry and 160 Ry, respectively. In these calculations, the electronic charge density was computed using 8 special k-points in the BZ, generated according to the Monkhorst-Pack scheme.
The relaxed atomic structures for the three models are shown on Fig. 1. In the DDR geometry, one ad-dimer is strongly tilted ($`\delta z=0.62`$ Å) and has a short bond length ($`d=2.26`$ Å) while the other, weakly bound ($`d=2.66`$ Å), is almost flat ($`\delta z=0.03`$ Å). The inequivalence of the two ad-dimers disagrees with simple expectations and with previous calculations by Kitabatake et al, who found two flat ad-dimers for the DDR model. Their use of the Tersoff potential could explain this disagreement, owing to the neglect of charge transfer between Si and C atoms. A single flat and weakly bonded ad-dimer ($`d=2.62`$ Å) is obtained in the ADD model, the geometry being close to that previously obtained by Yan et al in a calculation similar to ours. Note that two slightly different configurations are possible within the same model, since the weakly bonded Si dimers in the underlying layer can be arranged either all on one side, as originally proposed by Yan et al, or in a staggered pattern (see Fig. 1). Starting from different configurations, our calculations always converged to the staggered pattern. Finally, in the ALT model, the ad-dimer is strongly tilted ($`\delta z=0.5`$ Å) and strongly bound ($`d=2.24`$ Å), in good agreement with previous calculations. The length of the weak Si dimers in the underlying surface layer is close to the value computed for the $`(2\times 1)`$ reconstruction using the same method and a $`4\times 4`$ supercell.
In order to determine the most stable configuration, we have compared total energies. The ALT model is lower in energy than the ADD model by about 0.5 eV per $`(3\times 2)`$ cell; this energy difference is clearly in favor of the ALT, our error bar being 0.3 eV, as estimated from cutoff and BZ sampling tests. The drastic reduction from the 3.6 eV energy difference quoted by Yan et al is likely due to their poorer BZ sampling and, to a lesser extent, to the additional relaxation leading to the staggered pattern. A direct comparison with the DDR model is not possible since the corresponding Si coverage is different. This difficulty can be overcome by using the grand canonical scheme. The computed surface energy differences as a function of the Si chemical potential $`\mu _{\text{Si}}`$ are shown in Fig. 2. The value of $`\mu _{\text{Si}}`$ for bulk silicon was calculated with an energy cutoff of 40 Ry and 32 special k-points in the BZ, whereas we used the heat of formation of SiC $`\mathrm{\Delta }\text{H}=0.75`$ eV from a recent calculation, to obtain the Si chemical potential under C-rich conditions. Since $`(3\times 2)`$ growth is likely to occur under Si-rich conditions, a precise determination of $`\mu _{\text{Si}}`$ is required only in that limit. Our results show that the ALT model is the most stable configuration over the entire allowed range of Si chemical potential. However, the energy difference between ALT and DDR obtained under Si-rich conditions is only 77 meV, i.e. within our error bar. Consequently, the DDR model can not be definitely ruled out solely on the basis of total energy comparisons.
Several experimental STM studies of the $`(3\times 2)`$ reconstruction are currently available. In order to compare the three different models, we have calculated filled states constant-current STM images within the Tersoff-Hamann approximation. Representative images are shown in Fig. 3. In both the DDR and ADD models we find strings of peanut-shaped spots, originating from a slight overlap between maxima on adjacent flat ad-dimers. For the DDR model, additional maxima are located on the up adatoms of the tilted ad-dimers. The resulting images are incompatible with the experimental observations of a single oval spot stretched in the $`[110]`$ direction per $`3\times 2`$ cell. On the other hand, in the ALT model the spread out stretched spots located above up adatoms of the tilted ad-dimers are in accord with experimental STM images of filled states.
Additional insight can be obtained from analysis of the electronic states within a few eV of the Fermi level. All photoemission measurements agree about the presence of two occupied surface states in the band gap, 1 eV apart from each other. However, uncertainties exist about the location of these states with respect to the Valence Band Maximum (VBM). Recent angle-resolved photoemission spectroscopy (ARPES) measurements have shown that the dispersion of all identified surface states is very small ($`0.2`$ eV) along the $`[\overline{1}10]`$ and $`[110]`$ directions. Only the surface states of the DDR and ALT models have been considered here, the ADD model being higher in energy than the ALT model and exhibiting STM images which do not agree with experiment.
We find that in the DDR model the surface is metallic, within the Local Density Approximation. The highest occupied state, about 1 eV above the VBM at $`\mathrm{\Gamma }`$, is mainly localized on the flat ad-dimer and has a $`\pi ^{}`$ character with respect to the dimer axis. Its dispersion is very small along the $`[110]`$ direction ($`0.1`$ eV), but rather strong along the $`[\overline{1}10]`$ direction ($`1`$ eV). In the DDR model, we also find three additional surface states with energies between the highest occupied state and the VBM. Only one of them is localized above the up adatom of the tilted ad-dimer, and is essentially dispersionless. The other two states originate from backbond and dimer states of the underlying surface, and show strong dispersions. The presence of dispersive states in the band gap is in disagreement with ARPES evidence and points against the DDR model.
In the ALT model, the surface is semiconducting, with a direct gap at $`\mathrm{\Gamma }`$ of about 0.5 eV. The highest occupied state, 0.8 eV above the VBM at $`\mathrm{\Gamma }`$, is localized on the up adatom of the tilted ad-dimer and has a strong ‘s’ character. This surface state has a small dispersion along both $`[110]`$ and $`[\overline{1}10]`$ directions ($`0.1`$ eV). Close to the VBM we find another state, lying 0.7 eV below the highest occupied orbital. It is a $`\pi ^{}`$ state localized on the Si-Si dimer of the underlying surface which are not bonded to ad-dimers, and is nearly dispersionless ($`0.2`$ eV). This state is only present in the $`[110]`$ direction. Except for the energy difference between the two highest surface states (0.7 eV vs. 1 eV), agreement with ARPES experiments is definitely better for the ALT than the DDR model.
A discrepancy exists about the number and location of occupied surface states or resonances. Yeom et al argued that there should be a total of four states with pronounced surface character in their proposed DDR geometry. However, they considered the ideal non-relaxed geometry, with flat ad-dimers only, and, more importantly, they ignored backbond states and dimer-like states on the underlying surface which are also expected to exhibit surface character. These additional resonant states could be difficult to resolve because some are close in energy, and might have weak photoemission intensities. In our calculation we could identify resonant states, however we did not attempt to systematically analyse their character and dispersion.
Turning to unoccupied states, we find only one surface state in the band gap for the DDR model. It lies 2 eV above the VBM at $`\mathrm{\Gamma }`$, is localized around the down adatom of the tilted ad-dimer and has a predominant ‘p<sub>z</sub>’ character. Its dispersion is about 0.1 eV (0.4 eV) along the $`[110]`$ ($`[\overline{1}10]`$) direction. For the ALT model, two empty surface states have been identified. The lowest one, with energy 1.2 eV above the VBM at $`\mathrm{\Gamma }`$, is a $`\sigma `$-like state on the lone Si dimer of the underlying surface. It disperses about 0.1 eV along $`[110]`$ and 0.7 eV along $`[\overline{1}10]`$. The other one, 1.8 eV above the VBM, is a ‘p<sub>z</sub>’-like state on the down adatom of the tilted ad-dimer and has a very weak dispersion along both directions ($`0.2`$ eV).
The combination of all our results indicates that the ALT model is the most suitable candidate, since it explains the large majority of available measurements. The ADD and DDR models produce incorrect STM images. The ADD is energetically unfavorable, while the dispersion of the surface states calculated for the DDR model is not compatible with ARPES measurements.
To summarize, we have performed plane-wave pseudopotential calculations for three structural models of the $`(3\times 2)`$ reconstructed $`\beta `$-SiC(001) surface. In particular, relaxed atomic structures, surface energies, STM images, surface states and their dispersion have been calculated and compared with experiments. Our results strongly favor the ALT model and exclude the DDR and ADD models, although some ambiguities remain. More definitive conclusions could come from additional experimental studies, in particular investigations of unoccupied electronic states, and a convincing confirmation of the Si coverage corresponding to the $`(3\times 2)`$ reconstruction.
We are thankful to V. M. Bermudez, P. Soukiassian, G. Dujardin and H.-W. Yeom for fruitful discussions and/or preprints. One of us (L.P.) gratefully acknowledges Prof. H.-J. Güntherodt for the facilitities provided in his group, and the Swiss National Foundation for financial support under the NFP 36 program “Nanosciences”. This work has also been partially supported by the ”Consiglio Nazionale delle Ricerche” (Italy) and the Swiss Center for Scientific Computing (Manno, Switzerland). Part of this work was performed by the Lawrence Livermore National Laboratory under the auspices of the U. S. Department of Energy, Office of Basic Energy Sciences, Division of Materials Science, Contract No. W–7405–ENG–48.
|
no-problem/9906/cond-mat9906411.html
|
ar5iv
|
text
|
# The validity of the Landau-Zener model for output coupling of Bose condensates
## Abstract
We investigate the validity of the Landau-Zener model in describing the output coupling of Bose condensates from magnetic traps by a chirped radiofrequency field. The predictions of the model are compared with the numerical solutions of the Gross-Pitaevskii equation. We find a dependence on the chirp direction, and also quantify the role of gravitation.
Bose-Einstein condensation of alkali atoms in magnetic traps was first observed in 1995 . The first step towards coherent matter beams was taken when a controlled release of condensed atoms from a magnetic trap was demonstrated in MIT , and recently at Garching . The basic tool for output coupling has been spin-flipping induced by a radiofrequency (rf) magnetic field , but other approaches have also been used .
The MIT group demonstrated output couplings based on chirping the rf field, and on resonant rf pulses . In both cases the experimental results are understood in terms of simple models. The theory of the pulsed and cw output couplings has been studied extensively in the literature . In this Brief Report we examine the chirped output coupler. We find the validity conditions for the multistate Landau-Zener (MLZ) description and compare them with the numerical solutions of the Gross-Pitaevskii (GP) equation . We chart the combinations of the chirp speed $`\lambda =d\omega _{\mathrm{rf}}/dt`$ and the rf field amplitude $`\mathrm{\Omega }`$ for which the MLZ description fails. We show that this failure depends on the chirp direction. Also, the repulsion between particles in the condensate and gravitation are found to contribute to the validity conditions.
Magnetic trapping is achieved by spin-polarising the atoms. For simplicity we consider the case of $`F=1`$, where $`F`$ is the hyperfine quantum number. In the inhomogeneous magnetic field the three substates $`M=1,0,1`$ experience the potentials shown in Fig. 1(a). The atoms on the $`M=1`$ state are trapped in the harmonic potential. In the MIT trap one had $`\omega _x=\omega _z=(2\pi )320`$ Hz and $`\omega _y=(2\pi )18`$ Hz . When the spin of an atom is flipped from this state to the $`M=0`$ state, the atom moves away due to quantum mechanical dispersion and the repulsion between the condensed atoms. If the internal state of the atom is changed into the $`M=1`$ state, it feels the inverted parabolic potential as well.
We can eliminate the rf oscillations by making the rotating wave approximation, and then shifting each $`M`$ state in energy by an appropriate number of photons . Now the field-induced resonances appear as potential crossings. At the rf field frequency $`\omega _0`$ the atoms at the center of the trap are in resonance with the field. We define the field detuning as $`\mathrm{\Delta }=\omega _{\mathrm{rf}}\omega _0`$. For $`\mathrm{\Delta }<0`$ none of the atoms are in resonance, and for $`\mathrm{\Delta }>0`$ atoms at locations $`x=\pm x^{}`$ are in resonance \[Fig. 1(b) and (c)\].
In the chirped output coupler one sweeps $`\omega _{\mathrm{rf}}`$ so that all atoms in the trap feel a resonant field for a brief moment. The idea is to make this moment long enough for achieving a total or partial spin flip, and brief enough that the atoms do not have time to move due to changes in their internal state. Then the atoms remain stationary while they experience a time-dependent change of the energy difference between adjacent spin states.
In the MIT experiment a linear chirp was used: $`\mathrm{\Delta }=\lambda t`$. For two internal states this corresponds directly to the Landau-Zener (LZ) model. In moderate magnetic fields the Zeeman shifts are linear and thus all adjacent states are resonant simultaneously, so one has a genuine multistate problem . In this particular case, however, the spin dynamics of a stationary atom can be described analytically, by a simple multistate generalization of the Landau-Zener result (MLZ) . In the experiment the agreement with the MLZ prediction was good.
For simplicity we consider one spatial direction only. The three-state Hamiltonian for a stationary atom located at $`x`$ is
$$H(x,t)=\left(\begin{array}{ccc}\frac{1}{2}m\omega ^2x^2\mathrm{}\mathrm{\Delta }& \frac{\mathrm{}\mathrm{\Omega }}{\sqrt{2}}& 0\\ \frac{\mathrm{}\mathrm{\Omega }}{\sqrt{2}}& 0& \frac{\mathrm{}\mathrm{\Omega }}{\sqrt{2}}\\ 0& \frac{\mathrm{}\mathrm{\Omega }}{\sqrt{2}}& \mathrm{}\mathrm{\Delta }\frac{1}{2}m\omega ^2x^2\end{array}\right).$$
(1)
Here $`m`$ is the atomic mass, $`\omega `$ is the trap frequency and $`\mathrm{}\mathrm{\Omega }`$ is the rf field coupling. For a stationary atom we can solve the time-dependent Schrödinger equation with $`H`$. The MLZ theory predicts the final populations of the spin states:
$`P_1`$ $`=`$ $`\mathrm{exp}(2\pi \mathrm{\Gamma }),`$ (2)
$`P_0`$ $`=`$ $`2\mathrm{exp}(2\pi \mathrm{\Gamma })[1\mathrm{exp}(2\pi \mathrm{\Gamma })],`$ (3)
$`P_{+1}`$ $`=`$ $`[1\mathrm{exp}(2\pi \mathrm{\Gamma })]^2,`$ (4)
where $`\mathrm{\Gamma }=\mathrm{\Omega }^2/(2\lambda )`$ . In the MIT experiment $`\lambda =(2\pi )500`$ MHz/s, and $`\mathrm{\Omega }`$ changed from 0 to about $`(2\pi )11.3`$ kHz. The $`P_i`$’s depend only on $`\mathrm{\Gamma }`$ and not on the trap geometry; thus in asymmetric traps the output coupling takes place in all directions with the same efficiency.
In the Gross-Pitaevskii theory the single atom amplitudes $`\mathrm{\Psi }_i(x,t)`$ describe effectively the whole condensate (i.e., $`|\mathrm{\Psi }_i(x,t)|^2`$ gives the density distribution of the atoms on spin state $`i`$. Their time evolution is obtained from the time-dependent Schrödinger equation with the Hamiltonian
$$_{i,j}=\frac{\mathrm{}^2}{2m}\frac{^2}{x^2}\delta _{i,j}+H_{i,j}+\underset{l=1}{\overset{3}{}}C_{i,l}|\mathrm{\Psi }_l|^2.$$
(5)
The last term describes interactions between the particles. The parameters $`C_{i,l}`$ are proportional to atom numbers and the scattering lengths of the corresponding internal states $`i`$ and $`l`$. In our one-dimensional study this parameter does not match properly with the realistic three dimensional situation. But a trap with a very low frequency at one direction can be regarded quasi one-dimensional. In this case a reasonable estimate for our 1D $`C`$ parameter would be $`C=C_{3D}/A`$, where $`A`$ is the cross sectional condensate area. As an order of magnitude estimate it should be valid for other traps as well. For simplicity we take all $`C`$’s to be equal.
We solve the GP equation numerically for the output coupler. In the limit of large $`C`$ we can ignore the kinetic energy term and obtain the Thomas-Fermi solution, $`\mathrm{\Psi }(x,0)=\sqrt{[\mu U(x)]/C}`$, where $`U(x)`$ is the trapping potential. The condition $`\mu U(x)0`$ defines the edge of the condensate. The chemical potential $`\mu `$ is obtained from the normalisation of the wave function.
A breakdown of the MLZ model is expected if the atoms move during the transition process. A similar problem arises for diatomic molecules interacting with short laser pulses . In order to quantify this breakdown we consider the characteristic time scale $`\delta t`$ of the LZ process, $`\delta t\mathrm{\Omega }/\lambda `$ . The atoms need to remain stationary during this time. The term ”stationary” can be defined by transforming $`\delta t`$ into a region $`\delta x`$ around the location $`x_0`$ of the atom. For simplicity we assume that $`x_0>0`$. For parabolic potentials in the $`F=1`$ case we set $`\mathrm{}\mathrm{\Omega }=\frac{1}{2}m\omega ^2[(x_0+\delta x)^2x_0^2]m\omega ^2x_0\delta x`$, which defines $`\delta x`$. If the atom moves a distance $`\mathrm{\Delta }x`$ in time $`\delta t`$, it can be regarded stationary if $`\mathrm{\Delta }x\delta x`$.
The atomic motion can arise from quantum mechanical diffusion, repulsion between atoms, or from acceleration along the inverted parabolic potential. We consider the acceleration $`a`$ first. With Newtonian dynamics we get $`a(x_0)=(1/m)(U/x)|_{x=x_0}=\omega ^2x_0`$. In the small region around $`x_0`$ we have $`\mathrm{\Delta }x=a(\delta t)^2/2=\omega ^2x_0\mathrm{\Omega }^2/(2\lambda ^2)`$. Thus we get the condition $`\lambda ^2/\mathrm{\Omega }m\omega ^4x_0^2/\mathrm{}`$. This needs to be true for all $`x_0`$; the right-hand side is maximised at the edge of the condensate.
For small $`C`$ the edge is near the width of the ground state of the harmonic potential, max$`(x_0)\sqrt{\mathrm{}/m\omega }`$, which gives
$$\frac{\lambda ^2}{\mathrm{\Omega }\omega ^3}1.$$
(6)
For large $`C`$ we take the Thomas-Fermi approximation, and then max$`(x_0)\sqrt{2\mu /m\omega ^2}`$, which gives
$$\frac{\lambda ^2}{\mathrm{\Omega }\omega ^2}\omega _\mu ,$$
(7)
where $`\mu =\mathrm{}\omega _\mu `$. The MIT trap parameters satisfy condition (6) well in all directions, and to break down condition (7) would require an unrealistically large $`\mu `$. Since $`lim_{C0}\mu =\mathrm{}\omega /2`$, Eq. (6) is a special case of Eq. (7).
Diffusion and repulsion can give atoms a velocity which initially overcomes the acceleration. For the $`M=0`$ state these processes are naturally covered by the various studies of the ballistic expansion of condensates; see Ref. and references therein. As we are only looking for constraints it is sufficient to characterise the maximum speed $`v`$ of the atoms with the energy stored in the trapped condensate. We set $`\mu =mv^2/2`$. For small $`C`$ the speed $`v`$ reduces to the free-space momentum width of the Gaussian harmonic oscillator wave function, $`v\sqrt{\mathrm{}\omega /m}`$. Now $`\mathrm{\Delta }xv\delta t`$. On the other hand, at $`x_0=0`$ we have $`\delta x=\sqrt{2\mathrm{}\mathrm{\Omega }/(m\omega ^2)}`$. Eventually we get for diffusion/repulsion the conditions (6) and (7).
These conditions do not depend on the direction of the chirp. However, there exists another breakdown mechanism for the MLZ theory. Atoms that have interacted resonantly with the field can re-enter the resonance region (a reunion) due to their motion. Let us consider a positive chirp: a resonance emerges at $`x=0`$ and then separates into two points that move towards large $`|x|`$. This is demonstrated in Fig. 1 if we consider it as a sequency of snapshots. Strong acceleration on the $`M=1`$ state leads to the above problem. We need a condition on $`\lambda `$ and $`\mathrm{\Omega }`$ for avoiding the reunion until the transition probability is negligible.
Here the direction of the chirp is crucial. For negative chirps the resonances emerge at large $`|x|`$ and move towards $`x=0`$ where they disappear. Acceleration, however, moves the atoms in the opposite direction. Thus for negative chirps the reunion problem is absent. The role of chirp direction has been studied e.g. in the context of laser-molecule interactions .
We consider the case where the reunion happens at $`x=x_0`$. Energy conservation gives us roughly the local speed $`v_0`$ of the moving atoms: $`v_0\omega x_0`$. The slope of the local energy difference between adjacent states is $`\alpha _0=m\omega ^2x_0`$. The product of these quantities gives us the local motion-induced change in the energy levels for the atom: $`\mathrm{}\lambda _0=|dU/dt|=|(U/x)(x/t)|_{x=x_0}=m\omega ^3x_0^2`$. The motion of the resonance, $`\lambda `$, is small compared to $`\lambda _0`$ for realistic parameters, and we can ignore it. Basically, we want that the motion-induced LZ transition probability at $`x_0`$ is smaller than some fixed value $`\stackrel{~}{\mathrm{\Gamma }}`$:
$$\mathrm{\Gamma }=\frac{\mathrm{\Omega }^2}{\lambda _0}\stackrel{~}{\mathrm{\Gamma }}.$$
(8)
The time it takes for a resonance to reach $`x_0`$ is $`t_0=m\omega ^2x_0^2/(2\mathrm{}\lambda )`$. For a accelerating atoms we need to solve the Newtonian equation of motion: $`(^2x/t^2)\omega ^2x=0`$. For $`x^{}=\sqrt{2\mathrm{}\lambda t^{}/(m\omega ^2)}`$, $`v(t^{})=v^{}`$ we get
$$x=x^{}\mathrm{cosh}[\omega (tt^{})]+\frac{v^{}}{\omega }\mathrm{sinh}[\omega (tt^{})].$$
(9)
There are several possible values for $`x^{}`$ and $`v^{}`$, and the quantum mechanical diffusion/repulsion complicates this simple Newtonian picture. For large $`C`$ we can assume that the particles which reach the reunion first come from the edge of the initial condensate. This fixes $`x^{}`$ (and $`t^{}`$). We have simulated the problem numerically with the GP equation, and obtained the reunion time for the fastest atoms. As shown in Fig. 2(a) an effective trajectory $`v^{}C^{1/3}`$ locates the reunion well. The main dependence on $`C`$, however, arises from the location of the condensate edge, $`x^{}C^{1/3}`$. Some examples of the breakdown are shown in Fig. 2(b).
One should not make detailed conclusions from such trajectories. Apart from the crudeness of the Newtonian mechanics, the fastest atoms are only a fraction of the condensate. In Fig. 3 we show in the ($`\lambda ,\mathrm{\Omega }`$) plane where the MLZ prediction for $`P_{+1}`$ fails for more than 10 % for the large $`C`$ case. The constraint obtained by using the effective trajectory and condition (8) is clearly too demanding. Also, in a real experiment one can switch the field off before the reunion and thus avoid the problem. This is the case in the MIT experiment: after reaching the resonance $`\omega _0`$ the rf field is on for about $`\tau 0.5`$ ms. As $`\tau \omega _x=1`$ and $`\tau \omega _y=0.06`$, the field is off by the time of reunion, as Fig. 2(a) shows. For a negative $`\lambda `$ we saw no large deviation from the MLZ prediction for the values used in Fig. 3.
The effects of gravitation should be considered in output couplers . In the direction of gravitation ($`z`$) the trapping potentials are
$`V_1`$ $`=`$ $`{\displaystyle \frac{1}{2}}m\omega _z^2\left(z+{\displaystyle \frac{g}{\omega _z^2}}\right)^2{\displaystyle \frac{mg^2}{2\omega _z^2}},`$ (10)
$`V_0`$ $`=`$ $`mgz`$ (11)
$`V_{+1}`$ $`=`$ $`{\displaystyle \frac{1}{2}}m\omega _z^2\left(z{\displaystyle \frac{g}{\omega _z^2}}\right)^2+{\displaystyle \frac{mg^2}{2\omega _z^2}},`$ (12)
where $`g`$ is the gravitational acceleration. The $`M=\pm 1`$ potentials remain harmonic but have spatially shifted centers, as shown in Fig. 4. As gravitation affects only the external degrees of freedom the rf field resonance conditions still follow the purely magnetic potentials. In other words, the resonance points are located symmetrically in respect to the magnetic field minimum, but not in respect to the trap center. For a stationary atom the LZ parameter $`\mathrm{\Gamma }`$ remains unaffected by gravitation, but many other conditions change.
For the MIT trap one gets $`(\mathrm{\Delta }x)_g=g/\omega _z^2=2.5\mu `$m which is about the size of the trap ground state (1.7 $`\mu `$m) but clearly smaller than the condensate (17 $`\mu `$m). For gravitation to dominate acceleration on the $`M=1`$ state at the condensate edge we obtain the condition $`2\mu \omega _z^2<mg^2`$ (for large $`C`$), which reduces to $`\mathrm{}\omega _z^2<mg^2`$ as $`C0`$. If gravitation dominates, then the basic validity condition for the MLZ approach becomes ($`\mathrm{\Delta }z=g\mathrm{\Omega }^2/\lambda ^2`$, $`\delta z=\mathrm{}\mathrm{\Omega }/(m\omega _z^2z_0)`$, $`z_0g/\omega _z^2`$)
$$\frac{\lambda ^2}{\mathrm{\Omega }}\frac{mg^2}{\mathrm{}}.$$
(13)
This applies also for the motion on the $`M=0`$ state.
Figure 4 shows also that the shifts in the potential centers can be used for directed output coupling. With a very slow negative chirp one can leak the condensate from the earthside edge of the trap, as the chirped rf pulse becomes resonant there first. Due to the slowness the condensate edge will follow the resonance, and the other resonance point will not reach the diminishing condensate. This approach is used in the recent experiment . It is complementary to the situation considered by us, as there the chirp timescale must be clearly longer than the motional time scales.
In this Brief Report we have derived the validity conditions of the multistate Landau-Zener approach for chirped output couplers, with and without the presence of gravitation. Comparison with numerical results shows that these conditions allow one to achive ”safe” parameter regions easily in the experiments. Our study is for one dimension only, but our results should apply directly to three dimensions, especially in the case of strongly asymmetric traps, where the tightest trapping direction will dominate the expansion of the untrapped atom cloud.
We acknowledge the support by the Academy of Finland.
|
no-problem/9906/nucl-th9906067.html
|
ar5iv
|
text
|
# Directed flow of neutral strange particles at AGS
## Abstract
Directed flow of neutral strange particles in heavy ion collisions at AGS is studied in the ART transport model. Using a lambda mean-field potential which is 2/3 of that for a nucleon as predicted by the constituent quark model, lambdas are found to flow with protons but with a smaller flow parameter as observed in experiments. For kaons, their repulsive potential, which is calculated from the impulse approximation using the measured kaon-nucleon scattering length, leads to a smaller anti-flow than that shown in the preliminary E895 data. Implications of this discrepancy are discussed.
PACS number(s): 25.75.Ld, 13.75.Jz, 21.65.+f
Studies of kaon collective flow in heavy ion collisions have been very useful in extracting the kaon properties in hot dense matter . Such knowledge is important in understanding the possibility of kaon condensation in the core of neutron stars and the existence of mini black holes . Available experimental data from both the FOPI and the KaoS collaboration at SIS/GSI show a negligible kaon flow in collisions of Ni+Ni at $`E_{\mathrm{beam}}/\mathrm{nucleon}=1.93`$ GeV and Au+Au at 1 GeV, respectively. Based on transport models, it has been shown that a vanishing kaon flow is consistent with a repulsive kaon potential in medium . For heavy ion collisions at higher energies available from the AGS, the directed flow of neutral strange particles ($`K^0`$ and $`\mathrm{\Lambda }`$) is being studied experimentally by the E895 collaboration . Preliminary data have shown that lambdas flow in the same direction as protons but with a smaller flow parameter, and neutral kaons have a strong negative flow relative to that of nucleons. To understand these experimental observations, we have studied using A Relativistic Transport (ART) model the directed flow of these neutral strange particles in Au on Au collisions at a beam momentum of 6 GeV/c and an impact parameter of 4 fm. We note that particle directed flow is characterized by the average in-plane transverse momentum as a function of rapidity. In the central rapidity region, one can also introduce the flow parameter which measures the change of average in-plane transverse momentum with respect to the rapidity.
For the lambda potential, we take it to be 2/3 of that of protons. This follows from the light quark contents of proton and lambda, and is also consistent with that determined from the structure of hypernuclei . The nucleon potential used in the ART model is the standard Skyrme-type derived from either a soft nuclear equation of state with a compressibility of 200 MeV or a stiff nuclear equation of state with a compressibility of 380 MeV. In Fig. 1, we show the results for the proton and lambda directed flow from the ART model together with the preliminary E895 data. The latter are based on the analysis of 8500 semi-central Au(6 AGeV)+Au events. To improve the statistics, the ART result for a given rapidity has been obtained by averaging the values at both positive and negative rapidities. Typical statistical errors in the ART results for the proton flow are around $`0.1\%`$ and for lambda flow are around $`0.3\%`$. In our calculations, we have neglected the effect of rescatterings among lambdas. This should be a good approximation as the lambda to nucleon ratio is only about $`1\%`$. We find that at central rapidities, the soft nuclear equation of state gives a better description of the measured proton directed flow, while at high rapidities ($`y/y_{c.m.}>0.7`$), the stiff nuclear equation of state describes the data better. For the lambda flow, the ART model gives a smaller value at central rapidities than that for protons. This result is similar to that obtained in heavy ion collisions at lower energies of about 2 GeV/nucleon . As shown in Fig. 1, using the soft equation of state can indeed account for the experimentally observed weaker directed flow of lambdas.
In the ART model, the kaon mean-field potential is obtained from the impulse approximation. In this approximation, the kaon dispersion relation is related to the kaon-nucleon scattering length via
$$\omega (\stackrel{}{k},\rho )=\left[m_K^2+\stackrel{}{k}^24\pi \left(1+\frac{m_K}{m_N}\right)a_{KN}\rho \right]^{1/2},$$
(1)
where $`a_{KN}0.255`$ fm is the isospin-averaged kaon-nucleon scattering length, $`m_K`$ and $`m_N`$ are the kaon and nucleon masses, and $`\rho `$ is the local baryon density. The kaon mean-field potential is then defined as
$$U(\stackrel{}{k},\rho )=\omega (\stackrel{}{k},\rho )(m_K^2+\stackrel{}{k}^2)^{1/2}.$$
(2)
Because of the repulsive nature of the kaon potential, the predicted directed flow for $`K^0`$’s is opposite from that of protons and has almost zero value at the central rapidity. However, the charge-exchange reactions between $`K^0`$ and $`K^+`$ have not been included in the ART model. To obtain an upper limit on the effect of charge-exchange reactions, we have included the Coulomb interaction also for $`K^0`$’s. Results from such a calculation are compared with the preliminary data from E895, which are for $`K_S^0`$’s reconstructed from their charged pion decays in semi-central collisions. Since the number of $`\overline{K^0}`$’s is much smaller than that of $`K^0`$’s at AGS energies, the experimentally observed $`K_S^0`$’s thus come mostly from $`K^0`$’s.
In Fig. 2, we compare the kaon flow results from the ART model with the E895 data. Typical statistical errors are around $`1\%`$ for the kaon flow from the ART model. It is clearly seen that without kaon mean-field potential, $`K^0`$’s flow with nucleons. The repulsive kaon mean-field potential makes kaons flow away from nucleons. In addition, charge-exchange reactions enhance the anti-flow of kaons with respect to protons due to additional Coulomb repulsion. We see that even if all $`K^0`$’s are $`K^+`$’s in the reaction, $`K^0`$’s can get at most an average $`p_x/m`$ of about 0.01c which is almost five times smaller than that observed by the E895 collaboration. Our results on $`K^+`$ flow are, however, consistent with the preliminary data from the E917 collaboration at AGS . They are also similar to the theoretical results for lower energies , i.e., kaons have essentially vanishing flow. On the other hand, results from the RQMD model show a positive $`K_S^0`$ flow due to neglect of the kaon mean-field potential .
Since kaon directed flow is affected by the kaon potential in nuclear medium, using different potentials can give different values for the flow parameter. Various models have suggested that the kaon energy in the medium can be written as
$$\omega (\stackrel{}{k},\rho )=\left[m_K^2+\stackrel{}{k}^2a_K\rho _S+(b_K\rho )^2\right]^{1/2}+b_K\rho .$$
(3)
In the above, $`\rho _S`$ is the baryon scalar density, which is smaller than the baryon density $`\rho `$. The parameter $`b_K=3/(8f_\pi ^2)=0.333`$ GeV/fm<sup>3</sup> determines the repulsive kaon-nucleon vector interaction, while the parameter $`a_K`$ is used to adjust the strength of the attractive scalar interaction in order to effectively take into account higher-order corrections to the mean-field prediction. At SIS energies, $`a_K`$ was found to be $`0.22`$ GeV<sup>2</sup>fm<sup>3</sup> in order to reproduce the $`K^+`$ spectra . At normal nuclear matter density, Eq. (3) gives a kaon potential of about 30 MeV and is consistent with that from the impulse approximation using the empirical kaon-nucleon scattering length. As shown in Ref. , a large negative kaon directed flow can be obtained when higher-order contributions cancel a large fraction of the scalar interaction at high densities. Since preliminary E895 data show a very large increase in kaon anti-flow compared to the SIS data at lower beam energies, this abrupt change may give us new information about the kaon in-medium properties.
Another possible reason for the large kaon anti-flow is related to the kinematics of three-body phase space in kaon production from baryon-baryon interactions. A recent publication showed that the kinematics due to an isotropic three-body phase space might lead to a more isotropic kaon flow compared to that from the traditional cluster phase-space kinematics as used in the ART model. At SIS energies, this effect is not important as kaon production is dominated by meson-baryon interactions. Since kaon production from baryon-baryon interactions becomes more important at AGS energies, using the isotropic three-body phase space, which is not included in the ART model, is expected to make initial kaons stay away from the nucleons and thus lead to a larger kaon anti-flow. However, this may also affect the lambda directed flow. A detailed study of the directed flow of strange particles in conjunction with other variables, e.g., the differential and elliptic flow, and their dependence on the system size, beam energy, and impact parameter will help to understand the relative importance of these different mechanisms.
In summary, we have studied the directed flow of neutral strange particles in the ART model. Compared with preliminary E895 data, the ART results show that the smaller lambda directed flow relative to the proton flow can be accounted for by a weaker mean-field potential as in the constituent quark model and from the hypernuclei phenomenology. However, the $`K^0`$ flow cannot be explained by the kaon mean-field calculated from the impulse approximation using the empirical kaon-nucleon scattering length. This may need the introduction of different density dependence of the kaon potential at high baryon densities than the naive linear one.
We thank D. Best and R. Lacey for helpful discussions. This work was supported in part by NSF Grant PHY-9870038, the Robert A. Welch Foundation Grant A-1358, and the Texas Advanced Research Project FY97-10366-068.
|
no-problem/9906/nucl-th9906019.html
|
ar5iv
|
text
|
# Medium Effects in Coherent Pion Photo- and Electroproduction on ⁴𝐻𝑒 and ¹²𝐶
## I INTRODUCTION
Recently, much attention has been paid to experimental investigations of coherent $`\pi ^0`$ photo- and electroproduction off nuclei, i.e.
$$\gamma ^{}+A(gs)A(gs)+\pi ^0,$$
(1)
where $`\gamma ^{}`$ is a real or virtual photon and $`A(gs)`$ is a nucleus in its ground state. This reaction is of special interest for nuclei with zero spin and isospin. In this case the theoretical treatment simplifies and contains only a minimum number of ingredients. Since the description of the nuclear ground state is well under control, the reaction is then suitable for analyzing medium effects in the production and propagation of $`\mathrm{\Delta }`$ resonance.
The earliest theoretical and experimental studies of this reaction had the aim to extract additional information about the elementary amplitude, in particular about the values of the $`M_{1+}`$ and $`M_1`$ multipoles, which are related to the excitation of the resonances $`\mathrm{\Delta }(1232)`$ and $`N^{}(1440)`$, respectively. The general interest in this field was revived in the 80’s in the context of studying the photoproduction of neutral pions on the proton in order to check the threshold predictions of low energy theorems.
On the other hand, coherent $`\pi ^0`$ photoproduction off nuclei was considered as unique tool to examine the properties of the pion-nucleus interaction and to obtain information about the behavior of the pion in the nuclear medium. The theoretical method used for this purpose was the distorted wave impulse approximation (DWIA). While this method was originally formulated in coordinate space with a purely phenomenological optical potential, the modern approach to DWIA uses a representation in momentum space. The first application of the DWIA in the momentum space to pion photoproduction on nuclei was due to Eramzhyan and collaborators . Within this approach the pion-nucleus interaction (or final state interaction – FSI) is treated dynamically starting from the elementary pion-nucleon amplitude, and consistently taken into account of nonlocalities of the photoproduction operator and off-shell effects. In particular, this method was applied to the case of coherent $`\pi ^0`$ production by Ref. . In the same spirit, the authors of Refs. developed their dynamical model for pion photoproduction on the nucleon.
However, FSI is only one aspect of the medium effects due to the modification of the pion propagator in the nuclear medium. Another important issue is the modification of the resonance characteristics (width and position) inside a nucleus. This problem was the subject of numerous investigations, especially in the framework of the $`\mathrm{\Delta }`$-hole model. For a review and further literature, the reader is referred to the monographs . This issue will be the main subject of our study. For this purpose we apply, for the first time, the Unitary Isobar Model (UIM) of Ref. to pion photoproduction on nuclei. This recently developed model is particularly adopted to describe the elementary process on the nucleon in the nuclear environment. An important advantage of this model is that the contributions of the background (Born terms and vector meson exchange) and ”dressed” $`\mathrm{\Delta }`$ resonance are separated in a unitary way. As we will show below this allows us to incorporate medium effects in a self-consistent way with regard to both the ”bare” $`\gamma N\mathrm{\Delta }`$ vertices and $`\mathrm{\Delta }`$ excitations due to pion rescattering. Another advantage of the UIM is that it covers both pion photoproduction and pion electroproduction up to 4-momentum transfers of $`Q^2=k^24(GeV/c)^2`$. Therefore, this model provides quite a unique opportunity to investigate the $`Q^2`$ dependence of medium effects. Note that the first theoretical study of the medium effects in the coherent $`\pi ^0`$ electroproduction on the heavy nuclei was done in Ref..
The structure of this paper is as follows. In sections 2 and 3 we consider the main theoretical ingredients to describe pion photo- and electroproduction on nuclei. Our results and predictions are presented in section 4. We shall demonstrate that all the available data for coherent pion photoproduction on $`{}_{}{}^{4}He`$ and $`{}_{}{}^{12}C`$ can be described, in a self-consistent way, with the same parameter set for the $`\mathrm{\Delta }`$ self-energy. Finally, our conclusions are summarized in section 5.
## II General expressions
### A Differential cross sections
In the case of an unpolarized beam and unpolarized target, the 5-fold differential cross section in the $`lab`$ frame can be written as (see Ref. for details)
$$\frac{d\sigma }{d\mathrm{\Omega }_e^{}dE_e^{}d\mathrm{\Omega }_\pi }=\mathrm{\Gamma }\frac{d\sigma _V}{d\mathrm{\Omega }_\pi },$$
(2)
which defines the virtual photon cross section
$$\frac{d\sigma _V}{d\mathrm{\Omega }_\pi }=\frac{d\sigma _T}{d\mathrm{\Omega }_\pi }+ϵ\frac{d\sigma _L}{d\mathrm{\Omega }_\pi }+\sqrt{2ϵ(1+ϵ)}\frac{d\sigma _{TL}}{d\mathrm{\Omega }_\pi }\mathrm{cos}\mathrm{\Phi }_\pi +ϵ\frac{d\sigma _{TT}}{d\mathrm{\Omega }_\pi }\mathrm{cos}2\mathrm{\Phi }_\pi ,$$
(3)
with $`ϵ`$ and $`\mathrm{\Gamma }`$ the degree of transverse polarization and the flux of the virtual photon, respectively. Denoting the photon’s $`lab`$ energy by $`\omega _L=E_eE_e^{}`$ and its three-momentum by $`𝐤_L`$, we have
$$ϵ=\left[1+2\frac{𝐤_L^2}{Q^2}\mathrm{tan}^2\frac{\theta _e}{2}\right]^1,\mathrm{\Gamma }=\frac{\alpha }{2\pi ^2}\frac{E_e^{}}{E_e}\frac{K}{Q^2}\frac{1}{1ϵ}.$$
(4)
In accordance with our previous work, we define the flux by the photon ”equivalent energy” $`K=(s_AM_A^2)/2M_A`$, with the Mandelstam variable $`s_A=E^2=Q^2+M_A^2+2\omega _LM_A`$, $`M_A`$ the mass of the nuclear target, and the square of the four-momentum transfer $`Q^2=k^2=𝐤_L^2\omega _L^2>0`$.
The first two terms in Eq. (3) are the transverse (T) and longitudinal (L) cross sections, and the third and fourth terms are the transverse-longitudinal (TL) and transverse-transverse (TT) interference cross sections. The latter ones can be separated by use of their typical dependence on the azimuthal angle $`\varphi _\pi `$ of pion emission. As in the case of pion electroproduction on the free nucleon, these four cross sections may be expressed in terms of the nuclear tensor $`W_{\mu \nu }`$, with $`\mu `$ and $`\nu `$ corresponding to the Cartesian coordinates $`x,y`$ or $`z`$, which leads to the following relations in the $`cm`$ frame:
$$\frac{d\sigma _T}{d\mathrm{\Omega }_\pi }=\frac{k_\pi }{2k_\gamma ^{cm}}(W_{xx}+W_{yy}),\frac{d\sigma _{TT}}{d\mathrm{\Omega }_\pi }=\frac{k_\pi }{2k_\gamma ^{cm}}(W_{xx}W_{yy}),$$
(6)
$$\frac{d\sigma _L}{d\mathrm{\Omega }_\pi }=\frac{k_\pi }{k_\gamma ^{cm}}\frac{Q^2}{\omega _\gamma ^2}W_{zz},\frac{d\sigma _{TL}}{d\mathrm{\Omega }_\pi }=\frac{k_\pi }{k_\gamma ^{cm}}\frac{Q}{\omega _\gamma }ReW_{xz},$$
(7)
with $`k_\gamma ^{cm}=(E^2M_A^2)/2E`$ the ”photon equivalent energy” in the $`cm`$ frame and $`k_\pi =𝐤_\pi `$ the pion momentum. Note that we use a right-handed coordinate system with the positive $`z`$-axis along the virtual photon momentum $`𝐤_\gamma `$ and the $`y`$-axis along the vector $`[𝐤_\gamma \times 𝐤_\pi ]`$. The nuclear tensor $`W_{\mu \nu }`$ is defined in terms of the nuclear transition amplitudes $`F_{fi}^{(\mu )}`$ describing the transitions from the initial states $`i>=J_iM_i>`$ to the final states $`f>=J_fM_f>`$ with spin $`J_{i(f)}`$ and projection $`M_{i(f)}`$. After summing and averaging over the nuclear spin degrees of freedom we obtain
$$W_{\mu \nu }=\frac{1}{2J_i+1}Re\underset{M_i,M_f}{}F_{fi}^{(\mu )}F_{fi}^{(\nu )}.$$
(8)
For the purpose of the numerical calculations it is convenient to express the nuclear amplitudes in the covariant spherical basis $`𝐞=\{𝐞_{+1},𝐞_1,𝐞_0\}`$ by the relations
$$F_{fi}^{(\lambda )}=\frac{\lambda }{\sqrt{2}}(F_{fi}^{(x)}+i\lambda F_{fi}^{(y)}),F_{fi}^{(0)}=F_{fi}^{(z)},$$
(9)
where $`\lambda =\pm 1`$ is the helicity of the transverse photon.
In the case of targets with zero spin, the general symmetry properties and the pseudoscalar nature of the pion have the consequence that the transverse nuclear amplitude $`F_{fi}^{(\lambda )}`$ contains only a term proportional to $`[𝐤_\pi \times 𝐤_\gamma ]`$ in the pion-nucleus $`cm`$ frame. Hence
$$F_{fi}^{(\lambda )}(𝐤_\pi ,𝐤_\gamma ;Q^2)=F_0(𝐤_\pi ,𝐤_\gamma ;Q^2)[\widehat{𝐤}_\pi \times \widehat{𝐤}_\gamma ]𝐞_\lambda ,F_{fi}^{(0)}=0,$$
(10)
where $`\widehat{𝐤}_\pi `$ and $`\widehat{𝐤}_\gamma `$ are unit vectors in the directions of the pion and the virtual photon momentum, respectively.
In the right-handed coordinate system with the $`z`$-axis along $`\widehat{𝐤}_\gamma `$ and the $`y`$-axis along $`[\widehat{𝐤}_\gamma \times \widehat{𝐤}_\pi ]`$, only the $`W_{yy}`$ tensor component remains. Therefore, $`d\sigma _T/d\mathrm{\Omega }_\pi =d\sigma _{TT}/d\mathrm{\Omega }_\pi `$ and all other cross sections in Eq. (3) vanish. In this case the expression for the virtual photon cross section simplifies to
$$\frac{d\sigma _V}{d\mathrm{\Omega }_\pi }=\frac{d\sigma _T}{d\mathrm{\Omega }_\pi }(1ϵ\mathrm{cos}2\mathrm{\Phi }_\pi ).$$
(11)
It should be noted that Eqs. (8-9) are general. As a consequence the differential cross section has a $`\mathrm{sin}^2\theta _\pi `$ and $`(1ϵ\mathrm{cos}2\mathrm{\Phi }_\pi )`$ dependence in all cases, independently of the reaction mechanism. In fact, this peculiarity of coherent pion photo- and electroproduction on spin zero nuclei is often used to separate coherent and incoherent contributions, or to measure the degree of photon polarization, because the photon asymmetry $`\mathrm{\Sigma }=d\sigma _{TT}/d\sigma _T`$ equals unity.
### B PWIA and DWIA nuclear amplitudes
Let us first consider the nuclear amplitude in a simple plane wave impulse approximation (PWIA). The corresponding amplitude is denoted by $`V_{fi}^{(\lambda )}`$. It can be expressed in terms of the elementary pion electroproduction amplitudes $`f_{\gamma \pi }^{(\lambda )}`$ and reduced to the form
$$V_{fi}^{(\lambda )}(𝐤_\pi ,𝐤_\gamma ;Q^2)=𝒲_A𝑑𝐫\mathrm{\Psi }_f^{}(𝐫)\underset{j=1}{\overset{A}{}}e^{i𝐪𝐫_j}f_{\gamma \pi }^{(\lambda )}(𝐤_\pi ,𝐤_\gamma ,𝐩_j;Q^2)\mathrm{\Psi }_i(𝐫)$$
(12)
with $`𝐫=\{𝐫_1,𝐫_2,\mathrm{},𝐫_A\}`$ the set of nucleon coordinates, a phase space factor $`𝒲_A`$ given by
$$𝒲_A=\frac{W}{E}\sqrt{\frac{E_A(k_\pi )E_A(k_\gamma )}{E_N(p^{})E_N(p)}},$$
(13)
$`𝐩`$ and $`𝐩^{}`$ the initial and final nucleon momenta, $`E_{A(N)}`$ the nuclear (nucleon) energies, $`𝐪=𝐤_\gamma 𝐤_\pi `$ the momentum transfer to the nucleus, and $`Q^2=𝐤_\gamma ^2\omega _\gamma ^2`$ expressed by the $`cm`$ momentum and energy of the virtual photon.
The invariant amplitude $`f_{\gamma \pi }^{(\lambda )}`$ describes the elementary process on the free nucleon. It is evaluated within the framework of the UIM in terms of the standard CGLN amplitudes $`F_i(W,\mathrm{cos}\stackrel{~}{\theta }_\pi ,Q^2)`$, with the pion angle $`\stackrel{~}{\theta }_\pi `$ and the total energy $`W`$ in the $`\gamma N`$ $`cm`$ system,
$$W=\sqrt{(\omega _\gamma +E_N(𝐩))^2(𝐤_\gamma +𝐩)^2}.$$
(14)
Note that in the literature there are several prescriptions for the relations between total pion-nuclear energy $`E`$ and $`W`$ with various physical or mathematical motivations (see for example Refs.). Using this freedom we in principle can effectively take into account medium effects and improve by this way (or ”optimize”) impulse approximation. However, our main aim is the study the medium effects explicitly. Therefore, in the present work we will use physically transparent relation (12) which does not contain information about residual nucleus and about final state.
The next problem is related to the treatment of Fermi motion, i.e. the dependence of the amplitudes on the nucleon’s momentum $`𝐩`$. Previous studies of pion-nucleus scattering and pion photoproduction reactions showed that a good approximation to the proper averaging over Fermi motion is given by the ”factorization approximation”, which supposes that the main part of Fermi motion can be accounted for by evaluating the elementary amplitude at ”effective” nucleon momenta $`𝐩`$ and $`𝐩^{}`$ in the initial and final states,
$$𝐩=\frac{𝐤_\gamma }{A}\frac{A1}{2A}𝐪,𝐩^{}=\frac{𝐤_\pi }{A}+\frac{A1}{2A}𝐪,$$
(15)
where $`A`$ is the number of nucleons. The same approximation, but in the nuclear rest frame is taken in Refs. . Note that this approximation is based on the fact that for Gaussian nuclear wave functions (which reproduce the ground state of light nuclei sufficiently well) the replacement leads to an exact treatment of the terms linear in $`𝐩/2M`$ in the elementary amplitude. Moreover, it allows a consistent description of nuclear and nucleon kinematics, by means of a simultaneous conservation of energies and momenta for both the pion-nucleus and the pion-nucleon systems. As an example this factorization approximation was tested numerically for pion photoproduction on p-shell nuclei . Recently this approximation was also examined within a more elaborate relativistic model for coherent pion photoproduction on $`{}_{}{}^{12}C`$ .
In the case of coherent photo- and electroproduction of neutral pions on spin and isospin zero nuclei, the factorization approximation allows us to obtain the following simple expression for the PWIA amplitude $`V_{fi}^{(\lambda )}`$
$$V_{fi}^{(\lambda )}(𝐤_\pi ,𝐤_\gamma ;Q^2)=A𝒲_Af_2(𝐤_\pi ,𝐤_\gamma ;Q^2)F_A(q)[\widehat{𝐤}_\pi \times \widehat{𝐤}_\gamma ]𝐞_\lambda ,$$
(16)
where $`f_2=[f_2(p\pi ^0)+f_2(n\pi ^0)]/2`$ is the isoscalar non spin-flip part of the elementary amplitude boosted to the pion-nucleus $`cm`$ frame. The nuclear form factor is normalized to $`F_A(0)=1`$. It can be extracted from the nuclear charge form factor by the relation $`F_A^{ch}(q)=F_A(q)F_p^{ch}(q)`$, where $`F_p^{ch}(q)`$ is the proton charge form factor. The nuclear form factor $`F_A(q)`$ used in our present work is taken from Ref..
Let us now consider the full DWIA amplitude including pion distortion or FSI effects. In momentum space it takes the form
$$F_{fi}^{(\lambda )}(𝐤_\pi ,𝐤_\gamma ;Q^2)=V_{fi}^{(\lambda )}(𝐤_\pi ,𝐤_\gamma ;Q^2)+D_{fi}^{(\lambda )}(𝐤_\pi ,𝐤_\gamma ;Q^2),$$
(17)
where $`D_{fi}^{(\lambda )}`$ is the contribution from the pion-nucleus interaction expressed in terms of the pion-nucleus elastic scattering amplitude $`F_{\pi A}`$,
$$D_{fi}^{(\lambda )}(𝐤_\pi ,𝐤_\gamma ;Q^2)=\frac{a}{(2\pi )^2}\underset{M_f^{}}{}\frac{d𝐤_\pi ^{}}{(k_\pi ^{})}\frac{F_{\pi A}(𝐤_\pi ,𝐤_\pi ^{})V_{fi}^{(\lambda )}(𝐤_\pi ^{},𝐤_\gamma ;Q^2)}{E(k_\pi )E(k_\pi ^{})+iϵ}.$$
(18)
In this equation, the relativistic reduced mass of the pion-nucleus system is given by $`(k_\pi )=\omega _\pi (k_\pi )E_A(k_\pi )/E(k_\pi )`$. Note that the main difference compared to standard DWIA is the factor $`a=(A1)/A`$, which eliminates double counting of pion rescattering on one and the same nucleon. Such effects are in fact already included in the elementary amplitude. Finally, the pion scattering amplitude $`F_{\pi A}`$ is constructed as solution of a Lippmann-Schwinger integral equation (for details see Refs.).
## III Elementary amplitude and $`\mathrm{\Delta }`$ self-energy
### A Elementary amplitudes on- and off-shell
The detailed information about the elementary reaction amplitude for the free nucleon is given in Ref.. In the coherent $`\pi ^0`$ photo- and electroproduction on spin zero nuclei only the spin independent part contributes . Therefore, in the pion-nucleon $`cm`$ frame we have
$$\stackrel{~}{f}_{\gamma \pi }^{(\lambda )}=F_2(\stackrel{~}{𝐤}_\pi ,\stackrel{~}{𝐤}_\gamma ;W,Q^2)[\widehat{\stackrel{~}{𝐤}}_\pi \times \widehat{\stackrel{~}{𝐤}}_\gamma ]𝐞_\lambda ,$$
(19)
where $`F_2`$ is the standard CGLN amplitude and $`\stackrel{~}{𝐤}_{\gamma (\pi )}`$ is the photon (pion) momentum in the $`\pi N`$ $`cm`$ frame. These momenta can be boosted to an arbitrary frame by the Lorentz transformation
$$\stackrel{~}{𝐤}_{\gamma (\pi )}=𝐤_{\gamma (\pi )}+\alpha _{\gamma (\pi )}𝐏,\alpha _{\gamma (\pi )}=\frac{1}{W_{i(f)}}\left(\frac{𝐏𝐤_{\gamma (\pi )}}{E_{i(f)}+W_{i(f)}}\omega _{\gamma (\pi )}\right),$$
(20)
where $`E_{i(f)}=\omega _{\gamma (\pi )}+E_N^{i(f)}`$ and $`W_{i(f)}`$ are the total energies for the photon-nucleon (pion-nucleon) systems in the arbitrary and $`cm`$ frames, respectively. The total momentum $`𝐏=𝐤_\gamma +𝐩=𝐤_\pi +𝐩^{}`$ is given in the arbitrary frame. Using the factorization approximation (13) and the transformation (18), we can get a simple connection between the $`f_2`$ amplitude of Eq. (14) with the CGLN amplitude $`F_2`$,
$$f_2=\frac{k_\gamma k_\pi }{\stackrel{~}{k}_\gamma \stackrel{~}{k}_\pi }\left[1+\frac{A1}{2A}(\alpha _\gamma +\alpha _\pi )\right]F_2.$$
(21)
Within the UIM the amplitude $`F_2`$ contains contributions from Born and vector meson exchange terms, $`F_2^{(B)}`$, and from ”dressed” $`N^{}`$ resonances, $`F_2^{(R)}`$. In our present work we consider the photon’s $`lab`$ energy range $`E_\gamma <400`$ MeV, for which the s-channel $`\mathrm{\Delta }(1232)`$ resonance gives the most important contribution. The multipole $`M_{1+}^{(3/2)}`$ related to the resonance is parametrized in a unitary way by a standard Breit-Wigner form, i.e. $`F_2^{(R)}=2M_{1+}^{(\mathrm{\Delta })}+M_1^{(Roper)}`$, where
$$M_{1+}^{(\mathrm{\Delta })}(W,Q^2)=F_{\gamma N\mathrm{\Delta }}(W,Q^2)\frac{\mathrm{\Gamma }_\mathrm{\Delta }(W)M_\mathrm{\Delta }e^{i\varphi }}{M_\mathrm{\Delta }^2W^2iM_\mathrm{\Delta }\mathrm{\Gamma }_\mathrm{\Delta }}F_{\pi N\mathrm{\Delta }}(W),$$
(22)
and $`M_1^{(Roper)}`$ accounts for the small contribution of the Roper resonance $`N^{}(1440)`$. In Eq. (20), $`F_{\gamma N\mathrm{\Delta }}`$ and $`F_{\pi N\mathrm{\Delta }}`$ are the $`\gamma N\mathrm{\Delta }`$ and $`\pi N\mathrm{\Delta }`$ vertex function, respectively, $`M_\mathrm{\Delta }`$ and $`\mathrm{\Gamma }_\mathrm{\Delta }(W)`$ the energy and total width of the $`\mathrm{\Delta }`$ resonance as defined in Ref.. The unitary phase $`\varphi (W,Q^2)`$ adjusts the phase of the total multipole (background+resonance) to the corresponding pion-nucleon scattering phase, in accordance with the Fermi-Watson theorem.
As has been mentioned in the Introduction, it is one of the advantages of the DWIA approach in momentum space that it allows us to account for the nonlocality of the elementary operator and off-shell effects. Such effects originate mainly from the principal value part of the Eq. (16), the nonlocality being related with the dependence of the amplitude on the pion momentum $`𝐤_\pi `$. For the background contribution $`F_2^{(B)}`$, this dependence is in principle fixed by the form of the Lagrangians. However, the UIM background contribution is derived from Lagrangians with a zero-range interaction. In this case the standard way (often used in coupled channels , dynamical or meson exchange models) is to introduce the finite range aspects of the interaction by an off-shell form factor. In this spirit we have multiplied the background contribution by a dipole-like form factor
$`f(\stackrel{~}{k}_\pi ,\stackrel{~}{k}_W)=\left({\displaystyle \frac{\mathrm{\Lambda }^2+\stackrel{~}{k}_W^2}{\mathrm{\Lambda }^2+\stackrel{~}{k}_\pi ^2}}\right)^2,`$ (23)
where $`\stackrel{~}{k}_W`$ is the on-shell value of the pion momentum in the $`\pi N`$ $`cm`$ frame corresponding to the total energy $`W`$ for the $`\gamma N`$ system. The standard value for the cut-off parameter $`\mathrm{\Lambda }`$, which provides the best fit in many calculations of pion-nucleon scattering and pion photoproduction, is in the range of 400-500 MeV. In our present work we shall fix this parameter at $`\mathrm{\Lambda }=450`$ MeV. For consistency with our off-shell extrapolation of the pion-nucleon scattering amplitude in the $`P_{33}`$ channel , we shall use the prescription
$`M_{1+}^{(\mathrm{\Delta })}(\stackrel{~}{k}_\pi ,W,Q^2)=M_{1+}^{(\mathrm{\Delta })}(W,Q^2){\displaystyle \frac{\stackrel{~}{k}_\pi }{\stackrel{~}{k}_W}}f(\stackrel{~}{k}_\pi ,\stackrel{~}{k}_W)`$ (24)
for the $`\mathrm{\Delta }`$ contribution, with the factor $`\stackrel{~}{k}_\pi /\stackrel{~}{k}_W`$ providing the correct threshold behavior in the off-shell case.
### B Self-energy of the $`\mathrm{\Delta }`$
It has been known from numerous $`\mathrm{\Delta }`$-hole model calculations of pion scattering, pion photoproduction and photoabsorption reactions that the properties of the bound $`\mathrm{\Delta }`$ isobar in the nuclear medium differ substantially from those of the free $`\mathrm{\Delta }`$ (see Ref. and references therein).
In the case of pion photo- and electroproduction, two main mechanisms excite the $`\mathrm{\Delta }`$ isobar excitation, i) a direct excitation with a ”bare” $`\gamma N\mathrm{\Delta }`$ vertex, which corresponds to diagram (b) in Fig. 1, and ii) a vertex renormalization mechanism, for which the $`\mathrm{\Delta }`$ is excited with pions produced by background terms (diagram (c) in Fig. 1). As has been shown recently contributions of these two diagrams are equally important and, therefore, medium effects have to account for both mechanisms. The natural way to do so is to introduce medium effects in the form of dressed resonance contributions (diagram (a)) as defined by the UIM and given by Eq. (20). However, as a first step, the relativistic $`\mathrm{\Delta }`$ propagator with the unitary phase $`\varphi `$ has to be reduced to the nonrelativistic Breit-Wigner form. This can be achieved by the transformation
$`{\displaystyle \frac{e^{i\varphi }}{M_\mathrm{\Delta }^2W^2iM_\mathrm{\Delta }\mathrm{\Gamma }_\mathrm{\Delta }(W)}}={\displaystyle \frac{1}{W+M_\mathrm{\Delta }}}{\displaystyle \frac{1}{W\overline{M}_\mathrm{\Delta }(W)+i\overline{\mathrm{\Gamma }}_\mathrm{\Delta }(W)/2}},`$ (25)
where
$$\overline{M}_\mathrm{\Delta }(W)=W(WM_\mathrm{\Delta })\mathrm{cos}\varphi \frac{M_\mathrm{\Delta }\mathrm{\Gamma }_\mathrm{\Delta }(W)}{W+M_\mathrm{\Delta }}\mathrm{sin}\varphi ,$$
(27)
$$\overline{\mathrm{\Gamma }}_\mathrm{\Delta }=\frac{2M_\mathrm{\Delta }\mathrm{\Gamma }_\mathrm{\Delta }(W)}{W+M_\mathrm{\Delta }}\mathrm{cos}\varphi 2(WM_\mathrm{\Delta })\mathrm{sin}\varphi .$$
(28)
The next step is to modify DWIA amplitude by introducing the so-called $`\mathrm{\Delta }`$ self-energy $`\mathrm{\Sigma }_\mathrm{\Delta }`$ on the $`rhs`$ of Eq. (23), i.e.
$`{\displaystyle \frac{1}{W\overline{M}_\mathrm{\Delta }+i\overline{\mathrm{\Gamma }}_\mathrm{\Delta }(W)/2}}{\displaystyle \frac{1}{W\overline{M}_\mathrm{\Delta }+i\overline{\mathrm{\Gamma }}_\mathrm{\Delta }(W)/2\mathrm{\Sigma }_\mathrm{\Delta }}}.`$ (29)
In the literature we find many approaches to calculate $`\mathrm{\Sigma }_\mathrm{\Delta }`$ within simple models based on the local density approximation or more refined microscopical calculations based on the $`\mathrm{\Delta }`$-hole model. Such models were quite successful in describing pion scattering and pion photoproduction on heavy nuclei. However, one of the peculiarities of these approaches is that they finally always incorporate elements of phenomenology. In our present work we will do so from the very beginning by looking for a phenomenological parametrization of $`\mathrm{\Sigma }_\mathrm{\Delta }`$ which should be simple, common for all nuclei and able to describe the available data. For this purpose we shall test two types of parametrization
$$\mathrm{\Sigma }_\mathrm{\Delta }(E_\gamma ,q^2)=V_1(E_\gamma )F(q^2),F(q^2)=e^{\beta q^2},$$
(31)
and
$$\mathrm{\Sigma }_\mathrm{\Delta }(E_\gamma ,r)=(A1)V_2(E_\gamma )\rho (r)/\rho _0,\rho _0=0.17fm^3,$$
(32)
where $`V_{1,2}`$ is a (complex and energy-dependent ) free parameter. The first parametrization, Eq. (26a), is convenient to use in momentum space. Here we assume that $`F(q^2)`$ is the $`s`$-shell harmonic oscillator form factor with $`\beta =0.54`$ fm<sup>2</sup> and that $`\mathrm{\Sigma }_\mathrm{\Delta }`$ is already saturated for $`{}_{}{}^{4}He`$. This has the consequence that $`V_1`$ and $`F(q^2)`$ have the same values for both $`{}_{}{}^{4}He`$ and $`{}_{}{}^{12}C`$. The second parametrization, Eq. (26b), results from the local density approximation, the nuclear density $`\rho (r)`$ being normalized to $`\rho (r)d^3r=1`$. Therefore, the $`\mathrm{\Delta }`$ self-energy will differ for $`{}_{}{}^{4}He`$ and $`{}_{}{}^{12}C`$.
Finally we note that in the present context we neglect medium effects due to the Roper resonance, because this contribution is very small in our energy range, i.e. of the order of 2-3% at a photon’s $`lab`$ energy $`E_\gamma <400`$ MeV.
## IV Results and Discussion
One of the attractive features of coherent pion photo- and electroproduction on nuclei is that we can obtain clear signals from modifications of the $`\mathrm{\Delta }`$ isobar in the nuclear medium. For this purpose, Fig. 2 presents our results for the reaction $`{}_{}{}^{4}He(\gamma ,\pi ^0)^4He`$ and $`{}_{}{}^{12}C(\gamma ,\pi ^0)^{12}C`$ at a photon’s lab energy $`E_\gamma =290`$ MeV. In this energy region the standard DWIA calculations substantially overestimate the measured differential cross sections. Our DWIA results (dashed curves) in the maximum are larger than the experimental data by approximately a factor of 2. As was shown in previous studies, one of the reasons for this large discrepancy is the interaction of the $`\mathrm{\Delta }`$ isobar with the surrounding nucleons.
The $`\mathrm{\Delta }`$-nucleus interaction leads to a renormalization of the $`\mathrm{\Delta }`$ propagator and can be described in terms of the $`\mathrm{\Delta }`$ self-energy $`\mathrm{\Sigma }_\mathrm{\Delta }`$. The recently measured differential cross sections for the reaction $`{}_{}{}^{4}He(\gamma ,\pi ^0)^4He`$ and forthcoming new data for $`{}_{}{}^{12}C(\gamma ,\pi ^0)^{12}C`$ to be measured over a wide energy region ($`200<E_\gamma <400`$ MeV) provide us with a unique opportunity to determine $`\mathrm{\Sigma }_\mathrm{\Delta }`$ with good accuracy. As has been pointed out in the last section, we shall now test two types of parametrizations given by Eq. (26).
First, let us consider $`\rho type`$ parametrization (26b) conventionally used in the analysis of elastic pion-nucleus scattering. Note that a Taylor expansion of Eq. (25) in $`\rho (r)`$ leads to medium corrections in $`\rho ^2(r)`$, which were often used in analyses of pion-nucleus scattering. One of the consequences of this parametrization is that for finite nuclei it gives a good description of the differential cross sections in forward direction but fail at large angles and in the resonance region. On the other hand pion scattering on light nuclei at backward angles is usually better described without the phenomenological $`\rho ^2`$ term (or without $`\mathrm{\Delta }`$ medium effects).
¿From Fig. 2 we can see that the coherent $`\pi ^0`$ photoproduction situation is similar to the case described above (see the dash-dotted curves). This result clearly indicates that for light nuclei the treatment of the $`\mathrm{\Delta }`$-nucleus interaction in terms of $`\rho (r)`$ becomes less satisfactory. Therefore, nuclei like $`{}_{}{}^{4}He`$ require a more sophisticated microscopical calculation of the $`\mathrm{\Delta }`$ self-energy, e.g. the $`\mathrm{\Delta }`$-hole or meson exchange models . We would like to stress that the first microscopical $`\mathrm{\Delta }`$-hole calculation performed some 15 years ago by J. Koch and E. Moniz gives excellent agreement with the results of the recent measurements (see the dotted curve in Fig. 2).
However, a consistent extension of the microscopical calculations to heavier nuclei meets serious difficulties, and at some level it always requires elements of phenomenology. Therefore, we decide to use another approach in our present work. First, with the use of the $`Ftype`$ parametrization (26a) in momentum space, we extracted $`\mathrm{\Sigma }_\mathrm{\Delta }`$ from the data for the reaction $`{}_{}{}^{4}He(\gamma ,\pi ^0)^4He`$. Then, we assume that the $`\mathrm{\Delta }`$-nucleus interaction is the same also for the heavier nuclei, i.e., that it saturates already for $`{}_{}{}^{4}He`$. Our calculations at $`E_\gamma =290`$ MeV indicate that this is indeed a realistic assumption, and allows us to describe the coherent $`\pi ^0`$ photoproduction on $`{}_{}{}^{4}He`$ and $`{}_{}{}^{12}C`$ over a wide range of pion angles (see the solid curves in Fig. 2). The obtained values for the strength of the $`\mathrm{\Delta }`$-nucleus interaction, $`ReV_1`$=19 MeV and $`ImV_1`$= -33 MeV, are in reasonable agreement with the values obtained in pion-nucleus scattering.
This result inspires us to test our model at other energies assuming that $`V_1`$ is an energy dependent function to be determined by fitting the experimental data for $`{}_{}{}^{4}He`$. In Fig. 3 we present results of our fit in the energy range $`200MeV<E_\gamma <400`$ MeV. In Fig. 4 the prediction of our model is compared with the new experimental data from Ref.. The energy dependence for the total cross section and the parameters $`V_1`$ are shown in Fig. 5. These results for the potential parameter $`V_1`$ agree qualitatively well with the predictions of Ref..
Our most interesting result is shown in Fig. 6, where the prediction is compared to the data of the A2 collaboration at MAMI for the reaction $`{}_{}{}^{12}C(\gamma ,\pi ^0)^{12}C`$ obtained at a pion angle $`\theta _\pi =60^0`$. Even at such a large angle where the differential cross section has dropped by more than an order of magnitude (see Fig. 2), our assumption about the saturation of the $`\mathrm{\Delta }`$-nucleus interaction is still quite acceptable. Of course, a better understanding of this phenomenon will require additional data at smaller angles where the differential cross section reaches a maximum, as well as data for other nuclei.
A new feature of the $`\mathrm{\Delta }`$-nucleus interaction can be investigated and new information about the $`Q^2`$ dependence of $`\mathrm{\Sigma }_\mathrm{\Delta }`$ can be obtained by using virtual photons (or coherent $`\pi ^0`$ electroproduction). Clearly, as the $`\mathrm{\Delta }`$ isobar should not remember how it was excited, the $`V_1`$ does not depend on $`Q^2`$. Therefore, the strength of its interaction with the surrounding nucleons is universal for all processes (electroproduction, pion scattering, Compton scattering, etc.) and the $`Q^2`$ dependence enters the parametrization of Eq. (26a) only in the form factor $`F(q)`$, with $`q=|𝐤_\gamma 𝐤_\pi |`$ the momentum transferred to the nucleus and $`𝐤_\gamma `$ the virtual photon momentum as function of $`Q^2`$,
$$𝐤_\gamma ^2=Q^2+\frac{(s_AM_A^2Q^2)^2}{4s_A}=Q^2+\omega _\gamma ^2.$$
(33)
In Fig. 7 we depict the predictions of our model for: i) an equivalent photon energy $`E_\gamma `$=228 MeV and $`Q^2`$=0, 0.054, 1.81 $`(GeV/c)^2`$, and ii) $`E_\gamma `$=289 MeV and $`Q^2`$=0, 0.074, 1.61 $`(GeV/c)^2`$. This corresponds to the kinematics of the recent and forthcoming data from NIKHEF
## V Conclusion
In this paper we have presented a first nuclear application of our recently developed unitary isobar model to pion photo- and electroproduction. This model describes all the existing data for $`(\gamma ,\pi )`$ and $`(e,e^{}\pi )`$ on the nucleon reasonably well and furthermore compares very well with the partial wave analysis of the VPI group up to $`W_{cm}=1700MeV`$.
We have performed a DWIA calculation for coherent $`\pi ^0`$ photo- and electroproduction from nuclei and applied it to experimentally investigated reactions on $`{}_{}{}^{4}He`$ and $`{}_{}{}^{12}C`$. Due to the very precise data recently measured at Mainz over a large angular range of photon $`lab`$ energies from 207 MeV up to 397 MeV, we are able to determine, for the first time, the energy dependence of the $`\mathrm{\Delta }`$-nucleus potential from pion photoproduction reactions. Comparing $`{}_{}{}^{4}He`$ and $`{}_{}{}^{12}C`$ we find no A-dependence of the potential and conclude that the $`\mathrm{\Delta }`$-nucleus interaction saturates already for $`{}_{}{}^{4}He`$ .
In our study of the dynamical mechanism of the $`\mathrm{\Delta }`$ self-energy in the nuclear medium we find that a form factor type $`\mathrm{\Delta }`$-nucleus interaction is preferred as compared to a density type of medium modification often used in the local density approximation of coordinate space calculations.
First experimental results on electroproduction obtained at NIKHEF are in good agreement with our calculations and no readjustment of the $`\mathrm{\Delta }`$-nucleus interaction is needed.
For future experiments it will be very interesting to see how the $`\mathrm{\Delta }`$-nucleus interaction saturates by measurements on the deuteron and on $`{}_{}{}^{3}He`$ and, secondly, if the saturation observed for $`{}_{}{}^{12}C`$ is valid for the heavier nuclei.
###### Acknowledgements.
We would like to thank M. Kirchbach, F. Wissman, F. Rambo, H. Blok, T. Botto and D. H. Lu for helpful discussions. S.S.K. and L.T. are grateful to the Physics Department of the NTU for the hospitality extended to them during their visit. This work was supported in part by the Deutsche Forschungsgemeinschaft (SFB 443) and by the National Science Council/ROC under grant NSC 88-2112-M002-015.
|
no-problem/9906/physics9906041.html
|
ar5iv
|
text
|
# Untitled Document
ELECTRON COLLISIONAL BROADENING OF ISOLATED LINES FROM MULTIPLY-IONIZED ATOMS<sup>*</sup><sup>*</sup>*Partially supported by the US-National Science Foundation and the Israeli Academy of Sciences and INTAS.
H. R. GRIEM† and Yu. V. RALCHENKO‡
†Institute for Plasma Research, University of Maryland, College Park, MD 20742, USA
‡Department of Particle Physics, Weizmann Institute of Science, Rehovot 76100, Israel
> Abstract—Recent experimental and theoretical (both improved semiclassical and fully quantum-mechanical) line broadening calculations for B III and Ne VII $`\mathrm{\Delta }n=0`$ transitions with $`n=2`$ and 3, respectively, are discussed. The disagreements by about a factor of 2 between the fully quantum-mechanically calculated and both measured and semiclassically calculated widths can be explained in terms of violations of validity criteria for the semiclassical calculations and nonthermal Doppler effects. Only the quantum calculations allow a clear separation of elastic and inelastic scattering contributions to the width. For B III, elastic scattering contributes about 30%, for Ne VII inelastic scattering dominates. This allows rather direct comparisons with benchmark electron-ion scattering experiments. Additional independent determinations of line widths for multiply-ionized, nonhydrogenic ions are called for, but meanwhile caution should be exercised in the use of corresponding semiclassically calculated widths, e.g., in opacity calculations or for plasma density diagnostics.
1. INTRODUCTION
Most spectral lines observed in laboratory experiments or lines emanating from stellar atmospheres are, as far as their Stark broadening is concerned, in the category of isolated lines. Such lines, per definition, have widths, not to mention shifts, which are smaller than separations between the perturbed and the relevant perturbing levels. Relevant in this context are especially levels connected to upper or lower levels of the transition by electric dipole matrix elements, i.e., levels which would contribute to the Stark shift of the line in an external electric field. Furthermore, if these widths are small enough, such that the required timescale for the Fourier transform of the line profile (in angular frequency units) is large compared with the duration of important electron-ion collisions, then the impact approximation to the general theory can be used to calculate widths (and shifts) of the resulting Lorentzian profiles.
Most important is generally the (full) width, $`\gamma `$, which according to Baranger<sup>1</sup> can be written as
$`\gamma `$ $`=`$ $`N_e{\displaystyle _0^{\mathrm{}}}vF(v)`$ (1)
$`\times \left({\displaystyle \underset{u^{}u}{}}\sigma _{uu^{}}(v)+{\displaystyle \underset{\mathrm{}^{}\mathrm{}}{}}\sigma _{\mathrm{}\mathrm{}^{}}(v)+{\displaystyle |f_u(\theta ,v)f_{\mathrm{}}(\theta ,v)|^2𝑑\mathrm{\Omega }}\right)dv,`$
in terms of cross sections $`\sigma _{uu^{}}`$ and $`\sigma _{\mathrm{}\mathrm{}^{}}`$ for electron-collisional inelastic transition rates from upper ($`u`$) and lower ($`\mathrm{}`$) levels into the perturbing levels $`u^{}`$ and $`\mathrm{}^{}`$, respectively. These cross sections are functions of the initial electron velocity $`v`$, and the $`\sigma v`$-product is averaged using the electron velocity distribution function $`F(v)`$ as weight function. The ensuing inelastic contribution to the line broadening, often called lifetime broadening, is as expected, e.g., from the original work of Lorentz.<sup>2</sup> However, the last (elastic) scattering term is less intuitive. It involves the difference of elastic scattering amplitudes $`f`$ on upper and lower levels. These quantities depend on velocity $`v`$ and scattering angle $`\mathrm{\Theta }`$, and were it not for the cross term $`f_u^{}f_{\mathrm{}}^{}f_u^{}f_{\mathrm{}}^{}`$, the integrals over $`v`$ and $`\mathrm{\Theta }`$ would simply give the sum of total elastic scattering rates on upper and lower levels. However, this cross term can be very important, especially for $`\mathrm{\Delta }n=0`$ transitions, and cause strong cancellations. This could already have been inferred from the semiclassical, adiabatic phase shift limit,<sup>3,4</sup> which had been applied also for the case of monopole-dipole interactions<sup>6,7</sup> (quadratic Stark effect) at the time of the first realistic Stark broadening calculations of isolated lines from neutral atoms<sup>6</sup> and ions<sup>8,9</sup> in low charge states. These calculations all used the semiclassical approximation and were only done to the lowest nonvanishing order in the interaction, the phase-shift calculations being a notable exception to this.<sup>6,7,10</sup> Results of such calculations, and a dispersion relation<sup>11</sup> between widths and shifts, were actually used to optimize the choice of minimum impact parameters and strong-collision terms in perturbative semiclassical calculations of widths and shifts.<sup>12</sup> For neutral and singly-ionized atoms this procedure is surprisingly accurate, especially for the widths.<sup>13</sup> In other words, remaining errors are much smaller than the strong-collision contributions given in Tables IVa and V of Ref. 12, which are rather large.
For lines from multiply-ionized atoms, both measurements and calculations have been much more difficult and are still relatively rare. The experimental problems lie in the development of suitable, i.e., optically thin and homogeneous, plasma sources capable of obtaining the high densities and temperatures required, and in the application of density and temperature diagnostics. Most successful in this regard has been the gas-liner pinch work at Bochum;<sup>14</sup> comparisons<sup>15</sup> of some of these results with linear-pinch-discharge experiments<sup>16</sup> are also possible and interesting. As to calculations, various versions of a semiempirical method<sup>17</sup> had been rather successful<sup>13</sup> in predicting widths of many lines from low charge states. However, along isoelectronic sequences, and for $`\mathrm{\Delta }n=0`$, $`n=3`$ transitions in multiply-ionized Li-like,<sup>18</sup> Be-like<sup>19</sup> and B-like<sup>20</sup> ions, significant disagreements were found, measured widths being larger than most calculated widths for the higher members of the isoelectronic sequences. For $`\mathrm{\Delta }n=0`$, $`n=2`$ transitions, measured widths for (Li-like) B III<sup>21</sup> and, perhaps, also for C IV<sup>22</sup> disagree with most semiclassical and semiempirical estimates in a similar way.
The semiempirical estimates were based on Eq. (1) in the sense that the effective Gaunt factor approximation<sup>23,24</sup> was used for the cross sections of dipole-allowed collision-induced transitions, e.g., from the upper level of the line, with an extrapolation to below-threshold energies (velocities) to account for elastic scattering from ions in the upper state of the line. At least in Ref. 17, the lower level width calculated in the same way was simply added, i.e., the interference term, $`f_u^{}f_{\mathrm{}}^{}f_u^{}f_{\mathrm{}}^{}`$, mentioned above was neglected, as were collision-induced transitions caused by other than dipole $`(\mathrm{\Lambda }=1)`$ interactions, i.e., monopole $`(\mathrm{\Lambda }=0)`$, quadrupole ($`\mathrm{\Lambda }=2`$), etc., interactions. The major difference between these semiempirical and previous versions of semiclassical calculations is that in the latter the equivalents of the Gaunt factors are actually calculated using classical path integrals of the dipole interaction, supplemented by estimates of strong-collision contributions. Recently proposed improvements to this procedure will be discussed in Sec. 3 after a more detailed discussion of experiments in Sec. 2. Fully quantum-mechanical calculations are reviewed in Sec. 4, followed by a summary and some recommendations.
2. EXPERIMENTS
The stringent requirements for accurate measurements of Stark broadening parameters have been summarized before, in connection with the first critical data evaluation for isolated ion lines,<sup>25</sup> but are difficult to reconcile with the high densities and temperatures necessary for multiply-ionized atoms. Moreover, at that time, electron densities remained well under $`10^{18}`$ cm<sup>-3</sup> and temperatures below 5 eV. Additional critically reviewed data<sup>26-27</sup> (see also Ref. 28 for an updated bibliography) range to higher densities, by about an order of magnitude, and to temperatures as high as 20 eV. The high-aspect ratio, cylindrical, gas-liner (gas-puff) z-pinch<sup>14</sup> was designed to facilitate measurements of multiply-ionized ions under optimal conditions, which can be well and independently (of Stark broadening) measured.
In this gas-liner pinch, a driver gas, hydrogen or helium, is injected axially as a concentric hollow gas cylinder near the wall of the discharge vessel, while a test gas is injected in controlled (small) amounts near the axis. In this manner one ideally obtains optically thin line emission from a nearly homogeneous, albeit transient, plasma near the discharge axis. Great care is taken to have sufficient pre-ionization for the driver gas to facilitate uniform breakdown of the main discharge and to avoid, hopefully, instabilities during the implosion of this low-aspect ratio gas-puff pinch. Typical implosion velocities are about 10<sup>7</sup> cm/sec, corresponding to Reynolds numbers over $`10^4`$, suggesting the development of hydrodynamic turbulence.<sup>29</sup> (Magnetic field effects are not likely to inhibit such turbulence because of the relatively large ion-ion collision frequency, namely $`\nu 10^9`$ sec<sup>-1</sup> for $`100`$ eV proton-proton collisions, compared to proton cyclotron frequencies of $`\omega _{ci}10^8`$ sec<sup>-1</sup> at typical fields of 1 T.) The timescale for development and decay of the turbulence<sup>30</sup> is $`\tau \mathrm{}/\mathrm{\Delta }v`$, if $`\mathrm{}`$ is a characteristic length, say 1 cm, and $`\mathrm{\Delta }v10^6`$ cm/sec a typical spatial difference of velocities. This suggests $`\tau 1\mu `$sec, somewhat less than the time to peak compression, but longer than the duration of significant line emission.
While the driver gas may therefore be highly turbulent during the time of interest, the heavier test-gas ions have collisional mean-free-paths of $`\stackrel{>}{}`$ 0.1 cm against collisions with protons; they may therefore average over several eddies, especially during the implosion. In any case, the test gas ions can have significant nonthermal velocities, which could be of the same order as thermal velocities of the driver gas and therefore much larger than their thermal velocities.
It is not clear to what extent the otherwise very powerful (collective) Thomson-scattering diagnostic<sup>31,32</sup> is capable of distinguishing between thermal and nonthermal (test gas) ion motions. (This would be different in case of plasma turbulence associated with nonlinear plasma waves.) Moreover, it usually measures temperatures and electron density only in a rather small fraction of the emitting volume, which may not be entirely representative of the average conditions along the line of sight for the spectroscopic measurements. Only very recently,<sup>22,33</sup> with the help of a 2-dimensional detector array, have scattering spectra also been taken radially-imaged to check the homogeneity in regard to density and temperature and to verify that, e.g., argon as test-gas was restricted to a small, near-axial region. Axial imaging of emission spectra with krypton as test gas<sup>34</sup> as function of test-gas concentration suggests a stable plasma column for concentrations below 1%, but inhomogeneities at higher levels, indicating macroscopic instabilities. However, these observations cannot be interpreted as evidence against fully developed, fine-scale turbulence in case of small test-gas concentrations as used for line profile measurements. In principle, the effective temperature for the Doppler width of lines of test-gas ion could be determined from the width of the impurity peak in the scattering spectrum,<sup>31,32</sup> but the corresponding error is fairly large at the small concentrations needed to avoid self-absorption of strong lines. Finally, to conclude<sup>19</sup> from the absence of bulkshifts of scattering spectra (i.e., from the absence of net velocities of the emitting plasma volume) that there is no additional (to thermal Doppler broadening) Doppler broadening of the lines from test-gas ions, is premature. However, these observations do indicate that any turbulent eddies are smaller than 1 mm.
For narrow lines, the possibility of systematic errors in Stark widths due to underestimates of the Doppler broadening<sup>29</sup> evidently remains. It is made even more plausible, in the case of C IV $`\mathrm{\Delta }n=0`$, $`n=2`$ transitions,<sup>22</sup> by the observation of a Doppler splitting of the lines corresponding to less than or about $`4\times 10^6`$ cm/sec, i.e., about half the implosion velocity. So far it is not understood why these streaming velocities are not completely randomized by collisions or turbulence.
3. SEMICLASSICAL CALCULATIONS
Stimulated by the various disagreements between measured widths and semiclassical calculations, which were already discussed in Sec. 1, two improvements to these semiclassical calculations have been proposed. The first of these was an attempt by Alexiou<sup>35</sup> to improve the estimates for strong-collision impact parameter and strong-collision term by insisting that unitarity of the path integrals of the electron-ion interaction energy be preserved also during the interaction, not only over the entire collision as in previous semiclassical calculations. This distinction is important for hyperbolic perturber paths, because of the compensation between effects during approach and separation occurring especially for dipole interactions. (These cancellations had been noticed already in the context of electron-collisional broadening of ionized-helium lines.<sup>36</sup>) Strong-collision impact parameters in the usual calculations are of the order<sup>12</sup>
$$\rho _{\mathrm{min}}=\frac{\mathrm{}}{mv}\frac{n^2}{Z}=\frac{\lambda }{2\pi }\frac{n^2}{Z},$$
(2)
$`\lambda `$ being the de Broglie wavelength of the perturbing electron and $`Z`$ the (spectroscopic) charge, e.g., $`Z=3`$ for B III, etc. This first improvement is therefore probably not realistic unless $`n^2/2\pi Z\stackrel{>}{}1`$, which is not fulfilled for the lines showing large disagreements with previous semiclassical calculations. The $`2\pi `$ in Eq. (2) was apparently omitted in Ref. 35, as was the effect of curvature on the distance of closest approach in estimating the validity of the long-range dipole approximation.
The second improvement,<sup>37</sup> by the same author, was to replace the lowest order, dipole interactions, perturbation theory by a numerical solution of the time-dependent equations to all orders, and to include long-range quadrupole interactions. This nonperturbative approach no longer requires the unitarity check, but of course still a minimum impact parameter, etc., to avoid errors from breakdown of the semiclassical approximation and of the long-range approximation to the exact electron-ion Coulomb interactions. Both of these errors were again severely underestimated<sup>15</sup> in the same manner as discussed in the preceding paragraph.
In spite of these potentially serious problems, the nonperturbative semiclassicalcalculations<sup>37,38</sup> give widths, e.g., for the Ne VII 2s3s-2s3p singlet and triplet lines which agree within reasonable errors with gas-liner experiments,<sup>38</sup> whereas previous calculations yielded smaller widths by as much as a factor of 2. A similar discrepancy is found if comparison is made with a fully quantum-mechanical close-coupling (CC) calculation,<sup>39</sup> albeit only for the rather similar 3s-3p transition in Ne VIII.
As will be discussed in the following sections, such large disagreements may be the rule rather than the exception for strong isolated lines from multiply-ionized atoms, i.e., for lines for which $`n^2/Z`$ is near or below 1.
4. QUANTUM-MECHANICAL CALCULATIONS
Fully quantum-mechanical, CC, calculations of electron impact broadening were published almost 30 years ago for resonance lines of Mg II<sup>40</sup> and Ca II,<sup>41</sup> and a few years later also of Be II.<sup>42</sup> It was already noticed at that time that the close-coupling (CC) calculations gave widths which were less than half the measured width for Be II, and less than a third of the semiclassically calculated width. Except for the impact approximation,<sup>1</sup> the colliding electron is treated in these calculations exactly and on an equal footing with the (active) bound electrons, with which it interacts via the complete Coulomb interaction rather than only its long-range multipole expansion. In other words, except for numerical errors from using, e.g., an insufficiently complete system of basis functions, insufficient number of active bound electrons, or insufficient energy resolution, the CC calculations should be almost exact. They have therefore also been continued as part of the Opacity Project with results published for lines from some transitions in Li- and Be-like ions<sup>39</sup> and for $`\mathrm{\Delta }n=0`$, $`n=2`$ transitions in C III.<sup>43</sup> For the latter lines, elastic scattering contributions to the width are particularly important, providing 20–55% of the line widths.
Meanwhile, completely quantum-mechanical calculations were also made for the electron broadening of H-like<sup>44,45</sup> and He-like<sup>46,47</sup> multiply-charged ions (overlapping lines). Here the emphasis was on line spectra from $`Z\stackrel{>}{}6`$ ions, encouraging the use of the distorted-wave<sup>44,47</sup> approximation, thus significantly reducing the numerical effort in comparison with CC-calculations. As a matter of fact, it turned out that the additional phase shifts even for $`L=0`$ partial waves remained small so that no unitarization or higher order calculation was required. ($`L`$ is the angular momentum, in units of $`\mathrm{}`$, of the scattered electron.) The $`L`$ value corresponding to the minimum impact parameter in semiclassical calculations, see Eq. (2), is
$$L_{\mathrm{min}}=\frac{n^2}{Z},$$
(3)
and a corresponding effective strong collision term was determined in Ref. 44 by a fit to the sum over partial waves of the expression for the line width. (Remember, however, that $`n^2/Z\stackrel{>}{}2\pi `$ and therefore $`L_{\mathrm{min}}\stackrel{>}{}2\pi `$ is required for the semiclassical approximation.) All of these calculations used the full Coulomb interaction, and (red) shifts were calculated as well in Refs. 45-47. They are mostly due to the $`\mathrm{\Lambda }=0`$, penetrating monopole interactions, which for the $`1s`$ and $`1s^2`$ lower states are also important for the interference between elastic scattering terms.
Returning to isolated lines, the discrepancies between measured widths and nonperturbatively, semiclassically, calculated widths on one side, and the fully quantum-mechanical calculations just mentioned, have encouraged some new calculations<sup>29,48</sup> using the converging close-coupling (CCC) method.<sup>49-51</sup> This method is a standard close-coupling approach, except that discrete and continuum target states are obtained by diagonalizing the Hamiltonian in a large orthogonal Laguerre basis. The coupled equations are formulated in momentum space; the convergence can therefore be easily tested by increasing the basis size. This method has been shown to give very similar results<sup>52</sup> for inelastic and elastic electron-ion cross sections to the RMPS (R-matrix with pseudo states) method,<sup>53</sup> but avoids the difficulties associated with the oscillatory behavior of wave functions in coordinate space.
The first use of the CCC method for line broadening calculations was for the case of the B III 2s-2p (Li-like) resonance doublet measured<sup>21</sup> on the gas-liner pinch. Special attention was paid to the elastic scattering term in Eq. (1), which decreases much faster with electron energy $`E`$ than the $`1/E`$ decay of the 2s and 2p (non-Coulomb) elastic cross sections. The cancellation between upper and lower level elastic scattering is substantial; simply using the sum of elastic cross sections would lead to an overestimate of the elastic scattering contribution to the line width by a factor of about 6 at $`T_e=10`$ eV. However, because of this cancellation and the somewhat erratic energy dependence, there is more uncertainty in the elastic than in the inelastic contributions. Collision-induced 2s-2p excitation and deexcitation transitions give the major contribution to the line width, and CCC, RMPS and Coulomb-Born with exchange<sup>54</sup> (CBE) calculations all give very similar cross sections, the latter being typically 20% larger than the two strong-coupling results. Most importantly, over 90% of the total cross section comes from $`L\stackrel{<}{}6`$, i.e., from the nonclassical region. Finally, inelastic scattering associated with $`\mathrm{\Delta }n1`$ transitions, estimated using the CBE approximation, was found to contribute only 5% to the calculated width, vs. typically 35 and 60% from elastic and $`\mathrm{\Delta }n=0`$ inelastic scattering, respectively. Moreover, the CCC width was even smaller, by less or about 10%, than the first $`R`$-matrix method results<sup>39</sup> in the temperature range of interest (5–10 eV).
Further evidence for the severe discrepancies mentioned above is obtained by a comparison of new CCC calculations<sup>48</sup> for the 2s3s-2s3p singlet and triplet lines of Ne VII (Be-like) with experiments<sup>19,38</sup> and nonperturbative semiclassical calculations.<sup>37</sup> In this case, two bound electrons are actively involved in the scattering process, while the 1s<sup>2</sup> innershell electrons can be considered as part of a frozen core. (This was verified<sup>48</sup> by comparisons between the full Hartree-Fock and frozen-core Hartree-Fock calculations using the Cowan code.<sup>55</sup>) In analogy to the B III case, 3s-3p inelastic and super-elastic transition rates are again a major contribution to the line width, followed by 3p-3d transition rates. For these dipole $`(\mathrm{\Lambda }=1)`$ cross sections about 50% of the total cross section arises from $`L9`$ for the singlet s-p and triplet p-d transitions, and from $`L7`$ for the triplet s-p and singlet p-d transitions. The $`L\stackrel{>}{}2\pi `$ criterion is thus fulfilled only marginally. Also, the distance of closest approach for, e.g., an $`L=5`$ classical electron is about equal to the boundstate radius,<sup>15,48</sup> which causes the long-range interaction used in the semiclassical calculations to fail. The semiclassical results are therefore again rather questionable. Note also that CCC and CBE cross sections for these large dipole cross sections are within 10% of each other, and that dipole cross sections for 2s3$`\mathrm{}`$-2p3$`\mathrm{}`$ transitions contribute about 10%,<sup>48</sup> although they were ignored in semiclassical calculations. The $`\mathrm{\Delta }n1`$ collisional cross sections for excitation or deexcitation of the $`3\mathrm{}`$ electron are at the percent level, but were also included in Ref. 48.
The 2s3s-2s3d quadrupole ($`\mathrm{\Lambda }=2`$) transition rates contribute only about 3% to the total quantum-mechanical inelastic transition rate,<sup>48</sup> also at $`T_e=20`$ eV, in part because of the relatively large threshold energies ($`9`$ eV). This is in contrast to the semiclassical result,<sup>37</sup> to which the quadrupole ($`\mathrm{\Lambda }=2`$) channel contributes 15% of the total width. Because this total width is about twice the quantum-mechanical result, the $`\mathrm{\Lambda }=2`$ cross sections actually differ by an order of magnitude. This is not very surprising, because in this case $`L\stackrel{<}{}5`$ partial waves are most important; semiclassical and long-range interaction approximations are both clearly inappropriate. This might have been inferred by analyzing Fig. 11 of Ref. 56, in which nonperturbative, semiclassical contributions to the line width are shown as a function of impact parameter. The excluded region here corresponds to $`L3`$ for the cases of only $`\mathrm{\Lambda }=1`$ and for both $`\mathrm{\Lambda }=1`$ and 2 interactions.
This leaves the elastic scattering contribution, which is essentially ignored in the semiclassical calculations. For this contribution, only $`L\stackrel{<}{}2`$ partial waves are important, which are entirely in the quantum-mechanical and short-range interaction ($`\mathrm{\Lambda }=0`$, etc.) regimes. As for B III, there is again strong cancellation of upper- and lower-level scattering amplitudes, by about an order of magnitude, for the electron energies of interest. The contribution to the electron collisional line widths is only about 8 and 10%, respectively, for singlet and triplet lines. The estimated errors in the quantum calculation,<sup>48</sup> using a comparison of results obtained from Eq. (1) and from a formulation in terms of T-matrix elements,<sup>12,57</sup> from CCC calculations, are actually $`\pm `$15%, mostly related to resonances in the electron-ion scattering.
5. SUMMARY AND RECOMMENDATIONS
It should be clear to the reader that in contrast to a previous assertion,<sup>58</sup> there is still no convincing convergence between measurements and calculations of electron-collisional widths of strong isolated lines from multiply-ionized atoms, and for that matter, not even for the 2s-2p resonance line of Be II.<sup>42</sup> As far as the measurements are concerned, the observed trends along isoelectronic sequences are particularly striking. This observed slower decay with spectroscopic charge $`Z`$ than predicted by most calculations is especially pronounced for lines which are relatively insensitive to electron broadening in comparison with higher $`n`$-lines. This slower decay occurs for $`n=3`$ lines and probably also for $`n=2`$ lines. It is, therefore, suggested that some of the other line broadening mechanisms compete with the broadening by electrons. Of these, Stark broadening by ions comes to mind, but has been shown theoretically<sup>12</sup> not to be sufficient to explain discrepancies by a factor of about 2. (Note that there is no question concerning the validity of semiclassical calculations for perturbing ions.) The proposed additional Doppler broadening due to turbulent or other nonthermal motions of the emitting ions provides a more likely explanation. It had been invoked before to explain the unexpectedly large widths of singlet and triplet 2s-2p lines of CV in laser-produced plasmas.<sup>59</sup>
Hopefully the uncertainties in the interpretation of measurements of relatively small Stark widths will be removed, e.g., by measurements on other plasma light sources, by precision measurements of the impurity peak in collective Thompson scattering spectra, or by the measurement of Doppler widths of lines with negligible Stark widths. If then the discrepancy with, e.g., CCC calculations should still persist, one would have to accept the possibility of some broadening mechanism not yet considered. Finding any substantial error in the quantum calculations for conditions well inside the regime of validity of the impact approximation is not a realistic option, since they agree very well with, e.g., benchmark experiments of the most important inelastic cross section for the B III 2s-2p lines, namely, the 2s-2p excitation cross section.<sup>60</sup> Although the elastic scattering contribution remains less accurate, even a 50% error in this contribution would increase the calculated width only by 15%. (For the Ne VII 3s-3p lines, such correction would require an increase of the elastic scattering contribution by a factor 2.5.)
Although numerical improvements in the calculation of elastic scattering amplitudes would be very desirable, the pattern appears to be that the corresponding contribution to the width decreases with $`Z`$, while inelastic, dipole-allowed inelastic contributions dominate. They are quite accurately represented by the CBE approximation, which in turn explains the surprising accuracy of some semiempirical estimates. However, the perturbative or nonperturbative semiclassical calculations cannot be trusted, if partial-wave contributions $`L\stackrel{<}{}2\pi `$ are important and/or if distances of closest approach approaching boundstate radii are responsible for much of the broadening. A practical criterion for the validity of these calculations is $`n^2/Z\stackrel{>}{}2\pi `$. This appears to be consistent with our recent quantum-mechanical calculation for the N IV $`\mathrm{\Delta }n=0`$, $`n=3`$ singlet line, which accounts for 70% of the measured width. For a more detailed analysis of the transition from the quantum-mechanical to the semiclassical regime, a good example is provided by an analogous discussion of electron scattering on bare nuclei.<sup>61</sup>
REFERENCES
1. Baranger, M., Phys. Rev., 1958, 112, 855.
2. Lorentz, H. A., Proc. Roy. Acad. Sci. (Amsterdam), 1906, 8, 591.
3. Lindholm, E. Ark. Mat. Astron. Fys., 1941, 28B, No. 3; 1945, No. 17.
4. Foley, H. M., Phys. Rev., 1946, 69, 616.
5. Griem, H. R., Baranger, M., Kolb, A. C. and Oertel, G. K., Phys. Rev., 1962, 125, 177.
6. Roberts, D. E. and Davis, J., Phys. Lett., 1967, 25A, 175.
7. Bréchot, S. and Van Regemorter, H., Ann. Astrophys., 1964, 27, 432.
8. Sahal-Bréchot, S., Astron. Astrophys., 1969, 1, 91; 2, 322.
9. Roberts, D. E., Astron. Astrophys., 1970, 6, 1.
10. Griem, H. R. and Shen, C. S., Phys. Rev., 1962, 125, 196.
11. Griem, H. R., Spectral Line Broadening by Plasmas, Academic, New York (1974).
12. H. R. Griem, Principles of Plasma Spectroscopy, Cambridge University Press, Cambridge (1997).
13. Kunze, H.-J., in Spectral Line Shapes, Vol. 4, ed. R. J. Exton, Deepak Publish, Hampton (1987).
14. Griem, H. R., in Spectral Line Shapes, Vol. 10, ed. R. Herman, AIP Conf. Proc. (in press).
15. Purić, J, Djeniẑe, S., Srecković, A., Platiša, M. and Labat, J., Phys. Rev. A, 1988, 37, 498.
16. Griem, H. R., Phys. Rev. A, 1968, 165, 258.
17. Glenzer, S., Uzelac, N. I. and Kunze, H.-J., Phys. Rev. A, 1992, 45, 8795.
18. Wrubel, Th., Ahmad, I., Büscher, S., Kunze, H.-J. and Glenzer, S., Phys. Rev. E, 1998, 57, 5972.
19. Glenzer, S., Hey, J. D. and Kunze, H.-J., J. Phys. B, 1994, 27, 413.
20. Glenzer, S. and Kunze, H.-J., Phys. Rev. A, 1996, 53, 2225.
21. Büscher, S., Wrubel, Th., Ahmad, I. and Kunze, H.-J., in Spectral Line Shapes, Vol. 10, ed. R. Herman, AIP Conf. Proc. (in press).
22. Van Regemorter, H., Astrophys. J., 1962, 136, 906.
23. Seaton, M. J., in Atomic and Molecular Processes, ed. D. R. Bates, Academic Press, New York (1962).
24. Konjević, N. and Wiese, W. L., J. Phys. Chem. Ref. Data, 1976, 5, 259.
25. Konjević, N. and Dimitrijević, M. S., J. Phys. Chem. Ref. Data, 1984, 13, 649.
26. Konjević, N. and Wiese, W. L., J. Phys. Chem. Ref. Data, 1990, 19, 1307.
27. Fuhr, J. R. and Felrice, H. R., http://physics.nist.gov.linebrbib.
28. Griem, H. R., Ralchenko, Yu. V. and Bray, I., Phys. Rev. E, 1997, 56, 7186.
29. Landau, L. D. and Lifshitz, E. M., Hydrodynamics, Nauka, Moscow (1986).
30. DeSilva, A. W., Baig, T. J., Olivares, I. and Kunze, H.-J., Phys. Fluids B, 1992, 4, 458.
31. Wrubel, Th., Glenzer, S., Büscher, S. and Kunze, H.-J., J. Atmos. Terr. Phys., 1996, 58, 1077.
32. Wrubel, Th., Büscher, S., Ahmad, I. and Kunze, H.-J., in Proceedings of the 11th APS Topical Conference on Atomic Processes, ed. M. S. Pindzola and E. Oks, AIP Conf. Proc., New York (1998).
33. Ahmad, I., Büscher, S., Wrubel, Th. and Kunze, H.-J., Phys. Rev. E, 1998, (in press).
34. Alexiou, S., Phys. Rev. A, 1994, 49, 106.
35. Griem, H. R. and Shen, K. Y., Phys. Rev., 1961, 122, 1490.
36. Alexiou, S., Phys. Rev. Lett., 1995, 75, 3406.
37. Wrubel, Th., Glenzer, S., Büscher, S., Kunze, H.-J. and Alexiou, S., Astron. Astrophys., 1996, 306, 1028.
38. Seaton, M. J., J. Phys. B, 1988, 21, 3033.
39. Bely, O. and Griem, H. R., Phys. Rev. A, 1970, 1, 97.
40. Barnes, K. S. and Peach, G., J. Phys. B, 1970, 3, 350; Barnes, K. S., J. Phys. B, 1971, 4, 1377.
41. Sanchez, A., Blaha, M. and Jones, W. W., Phys. Rev. A, 1973, 8, 774; see also Hadziomerspahic, D., Platisa, M., Konjević, N. and Popovic, M., Z. Physik, 1973, 262, 169.
42. Seaton, M. J., J. Phys. B, 1987, 20, 6431.
43. Griem, H. R., Blaha, M. and Kepple, P. C., Phys. Rev. A, 1979, 19, 2421.
44. Nguyen, H., Koenig, M., Benredjem, D., Caby, M. and Coulaud, G., Phys. Rev. A, 1986, 33, 1279.
45. Koenig, M., Malnoult, P. and Nguyen, H., Phys. Rev., 1988, 38, 2089.
46. Griem, H. R., Blaha, M. and Kepple, P. C., Phys. Rev. A, 1990, 41, 5600.
47. Ralchenko, Yu. V., Griem, H. R., Bray, I. and Fursa, D. V., Phys. Rev. A (submitted).
48. Bray, I., Phys. Rev. A, 1994, 49, 1066.
49. Bray, I. and Stelbovics, A. T., Adv. At. Mol. Phys., 1995, 35, 209.
50. Fursa, D. V. and Bray, I., J. Phys. B, 1997, 30, 757.
51. Bartschat, K., Burke, P. G. and Scott, M. P., J. Phys. B, 1997, 30, 5915.
52. Bartschat, K., Hudson, E. T., Scott, M. P., Burke, P. G. and Burke, V. M., J. Phys. B, 1996, 29, 115.
53. Shevelko, V. P. and Vainshtein, L. A., Atomic Physics for Hot Plasmas, IOP, Bristol (1993).
54. Cowan, R. D. The Theory of Atomic Structure and Spectra, Univ. of California Press, Berkeley (1981).
55. Alexiou, S., Calisti, A., Gauthier, P., Klein, L., LeBoucher-Dalimier, E., Lee, R. W., Stamm, R. and Talin, B., J. Quant. Spectrosc. Radiat. Transfer, 1997, 58, 399.
56. Peach, G. in Atomic, Molecular and Optical Physics Handbook, ed. G. W. F. Drake, AIP Press, New York (1996).
57. Alexiou, S., in Spectral Line Shapes, Vol. 9, ed. M. Zoppi and L. Ulivi, AIP Conf. Proc. 386, AIP, New York (1997).
58. Iglesias, E. J. and Griem, H. R., Phys. Rev. A, 1988, 38, 301.
59. Woitke, O., Djurić, N., Dunn, G. H., Bannister, M. E., Smith, A.C.H., Wallbank, B., Badnell, N. R. and Pindzola, M. S., Phys. Rev. A, 1998 (in press).
60. Williams, E. J., Rev. Mod. Phys., 1945, 17, 217.
|
no-problem/9906/hep-ph9906405.html
|
ar5iv
|
text
|
# Gluon topology and the spin structure of the constituent quark
## 1 INTRODUCTION
The small value of the flavour-singlet axial charge $`g_A^{(0)}`$ which is extracted from the first moment of $`g_1`$ (the nucleon’s first spin dependent structure function)
$$g_A^{(0)}|_{\mathrm{pDIS}}=0.20.35.$$
(1)
has inspired much theoretical and experimental effort to understand the internal spin structure of the nucleon . A key issue is the role of the axial anomaly in the transition from parton to constituent quark degrees of freedom in low energy QCD. In this paper I explain why some fraction of proton’s spin may be carried by gluon topology. The topological contribution has support only at Bjorken $`x`$ equal to zero.
In deep inelastic processes the internal structure of the nucleon is described by the QCD parton model . The deep inelastic structure functions may be written as the sum over the convolution of “soft” quark and gluon parton distributions with “hard” photon-parton scattering coefficients. The (target dependent) parton distributions describe a flux of quark and gluon partons carrying some fraction $`x=p_{+\mathrm{parton}}/p_{+\mathrm{proton}}`$ of the proton’s momentum into the hard (target independent) photon-parton interaction which is described by the hard scattering coefficients.
In low energy processes the nucleon behaves like a colour neutral system of three massive constituent quark quasi-particles interacting self consistently with a cloud of virtual pions which is induced by spontaneous chiral symmetry breaking .
One of the most challenging problems in particle physics is to understand the transition between the fundamental QCD “current” quarks and gluons and the constituent quarks of low-energy QCD. The fundamental building blocks are the local QCD quark and gluon fields together with the non-local structure associated with gluon topology .
Relativistic constituent-quark pion coupling models predict $`g_A^{(0)}0.6`$ — two standard deviations greater than the value of $`g_A^{(0)}|_{\mathrm{pDIS}}`$ in Eq.(1). Can we reconcile these two values of $`g_A^{(0)}`$ without abandoning the constituent quark picture of the nucleon ?
## 2 GLUON TOPOLOGY AND $`g_A^{(0)}`$
The flavour-singlet axial charge $`g_A^{(0)}`$ is measured by the proton forward matrix element of the gauge invariantly renormalised axial-vector current
$$J_{\mu 5}^{GI}=\left(\overline{u}\gamma _\mu \gamma _5u+\overline{d}\gamma _\mu \gamma _5d+\overline{s}\gamma _\mu \gamma _5s\right)_{GI}$$
(2)
viz.
$$2ms_\mu g_A^{(0)}=p,s|J_{\mu 5}^{GI}|p,s_c$$
(3)
In QCD the axial anomaly induces various gluonic contributions to $`g_A^{(0)}`$. The flavour-singlet axial-vector current satisfies the anomalous divergence equation
$$^\mu J_{\mu 5}^{GI}=2f^\mu K_\mu +\underset{i=1}{\overset{f}{}}2im_i\overline{q}_i\gamma _5q_i$$
(4)
where
$$K_\mu =\frac{g^2}{16\pi ^2}ϵ_{\mu \nu \rho \sigma }\left[A_a^\nu \left(^\rho A_a^\sigma \frac{1}{3}gf_{abc}A_b^\rho A_c^\sigma \right)\right]$$
(5)
is a renormalised version of the Chern-Simons current and $`f=3`$ is the number of light-flavours. Eq.(4) allows us to write
$$J_{\mu 5}^{GI}=J_{\mu 5}^{\mathrm{con}}+2fK_\mu $$
(6)
where
$$^\mu K_\mu =\frac{g^2}{32\pi ^2}G_{\mu \nu }\stackrel{~}{G}^{\mu \nu }.$$
(7)
and
$$^\mu J_{\mu 5}^{\mathrm{con}}=\underset{l=1}{\overset{f}{}}2im_i\overline{q}_l\gamma _5q_l$$
(8)
The partially conserved axial-vector current $`J_{\mu 5}^{\mathrm{con}}`$ and the Chern-Simons current $`K_\mu `$ are separately gauge dependent. Gauge transformations shuffle a scale invariant operator quantity between the two operators $`J_{\mu 5}^{\mathrm{con}}`$ and $`K_\mu `$ whilst keeping $`J_{\mu 5}^{GI}`$ invariant.
One would like to isolate the gluonic contribution to $`g_A^{(0)}`$ associated with $`K_\mu `$ and thus write $`g_A^{(0)}`$ as the sum of “quark” and “gluonic” contributions. Here we have to be careful because of the gauge dependence of $`K_\mu `$.
Whilst $`K_\mu `$ is a gauge dependent operator, its forward matrix elements are invariant under the “small” gauge transformations of perturbative QCD. In the QCD parton model one finds
$$g_A^{(0)}|_{\mathrm{partons}}=\left(\underset{q}{}\mathrm{\Delta }qf\frac{\alpha _s}{2\pi }\mathrm{\Delta }g\right)_{\mathrm{partons}}$$
(9)
Here $`\frac{1}{2}\mathrm{\Delta }q`$ and $`\mathrm{\Delta }g`$ are the amount of spin carried by quark and gluon partons in the polarised proton.
The polarised gluon contribution to Eq.(9) is characterised by the contribution to the first moment of $`g_1`$ from two-quark-jet events carrying large transverse momentum squared $`k_T^2Q^2`$ which are generated by photon-gluon fusion . The polarised quark contribution $`\mathrm{\Delta }q_{\mathrm{parton}}`$ is associated with the first moment of the measured $`g_1`$ after these two-quark-jet events are subtracted from the total data set.
The QCD parton model formula (9) is not the whole story. Choose a covariant gauge. When we go beyond perturbation theory, the forward matrix elements of $`K_\mu `$ are not invariant under “large” gauge transformations which change the topological winding number . The topological winding number is a non-local property of QCD. It is determined by the gluonic boundary conditions at “infinity” — a large surface with boundary which is spacelike with respect to the positions $`z_k`$ of any operators or fields in the physical problem — and is insensitive to any local deformations of the gluon field $`A_\mu (z)`$ or of the gauge transformation $`U(z)`$ — that is, perturbative QCD degrees of freedom. When we take the Fourier transform to momentum space the topological structure induces a light-cone zero-mode which has support only at $`x=0`$ . Hence, we are led to consider the possibility that there may be a term in $`g_1`$ which is proportional to $`\delta (x)`$.
It remains an open question whether the net non-perturbative quantity which is shuffled between the $`J_{\mu 5}^{\mathrm{con}}`$ and $`K_\mu `$ contributions to $`g_A^{(0)}`$ under “large” gauge transformations is finite or not. If it is finite and, therefore, physical then we find a net topological contribution $`𝒞`$ to $`g_A^{(0)}`$
$$g_A^{(0)}=\left(\underset{q}{}\mathrm{\Delta }qf\frac{\alpha _s}{2\pi }\mathrm{\Delta }g\right)_{\mathrm{partons}}+𝒞$$
(10)
The topological term $`𝒞`$ has support only at $`x=0`$. It is missed by polarised deep inelastic scattering experiments which measure $`g_1(x,Q^2)`$ between some small but finite value $`x_{\mathrm{min}}`$ and an upper value $`x_{\mathrm{max}}`$ which is close to one. As we decrease $`x_{\mathrm{min}}0`$ we measure the first moment
$$\mathrm{\Gamma }\underset{x_{\mathrm{min}}0}{lim}_{x_{\mathrm{min}}}^1𝑑xg_1(x,Q^2).$$
(11)
This means that the singlet axial charge which is extracted from polarised deep inelastic scattering is the combination $`g_A^{(0)}|_{\mathrm{pDIS}}=(g_A^{(0)}𝒞)`$. In contrast, elastic $`\mathrm{Z}^0`$ exchange processes such as $`\nu p`$ elastic scattering and parity violation in light atoms measure the full $`g_A^{(0)}`$. One can, in principle, measure the topology term $`𝒞`$ by comparing the flavour-singlet axial charges which are extracted from polarised deep inelastic and $`\nu p`$ elastic scattering experiments.
If some fraction of the spin of the constituent quark is carried by gluon topology in QCD, then the constituent quark model predictions for $`g_A^{(0)}`$ are not necessarily in contradiction with the small value of $`g_A^{(0)}|_{\mathrm{pDIS}}`$ extracted from deep inelastic scattering experiments.
The presence or absence of topological $`x=0`$ polarisation is intimately related to the dynamics of $`U_A(1)`$ symmetry breaking in QCD. A simple dynamical mechanism for producing topological $`x=0`$ polarisation is provided by Crewther’s theory of quark-instanton interactions . There, any instanton induced suppression of $`g_A^{(0)}|_{\mathrm{pDIS}}`$ is compensated by a net transfer of axial charge or “spin”from partons carrying finite momentum fraction $`x>0`$ to the flavour-singlet topological term at $`x=0`$ .
A large positive $`\mathrm{\Delta }g`$ ($`+1.5`$ at $`Q^2=1`$GeV<sup>2</sup>) and topological $`x=0`$ polarisation are two possible explanations for the small value of $`g_A^{(0)}|_{\mathrm{pDIS}}`$. Measurements of both quantities are urgently needed!
|
no-problem/9906/hep-ph9906556.html
|
ar5iv
|
text
|
# References
Recently it was suggested that the fundamental scale of quantum gravity $`M_{Pf}`$ may be as low as a few TeV, which would provide an alternative way of understanding the hierarchy problem . In this proposal the observed weakness of gravity at large distances is due to the presence of $`n`$ large new dimensions (of size $`R>>M_{Pf}^1`$) in which gravity can propagate. The relation between the observed Planck scale $`M_P`$ and the fundamental Planck scale $`M_{Pf}`$ is given by
$$M_P^2M_{Pf}^{n+2}V_n,$$
(1)
where $`V_nR^n`$ ($`n`$ must be 2 or larger) is the volume of the $`n`$ extra spatial dimensions (i.e., of the bulk). In this picture all the Standard Model fields must reside inside a brane (or a set of branes) with $`3`$ extended spatial dimensions<sup>1</sup><sup>1</sup>1In a somewhat different context an attempt to lower the string scale down to a TeV (without lowering the fundamental Planck scale) was considered in based on an earlier observation in .. A general discussion of possible embeddings of such a scenario within the string theory context was given in .
In this framework the role of supersymmetry might be somewhat modified. A priori it is no longer needed as a solution to the hierarchy problem. However, it might still play an important role in, say, formulating the fundamental theory that incorporates quantum gravity above a TeV. At present the only known candidate theory for such a unified picture is superstring theory. Thus, we may expect string theory to take over at energies above a TeV with supersymmetry broken not much below the fundamental string scale $`M_s`$. If so, what is the content of the supersymmetric theory just below $`M_s`$ where stringy modes can be integrated out?
The purpose of this note is to point out that in theories with quantum gravity (or strings) at the TeV scale, there is a logical possibility to identify the electroweak Higgs with the scalar superpartner of one of the lepton doublets (or a linear combination thereof). However, we cannot identify the $`SU(2)`$-breaking slepton with the superpartner of a known lepton. Rather, we propose the existence of a fourth generation, whose slepton plays the role of the Higgs boson. In this way, the Higgs and flavor structures of the theory are more closely related than in the usual supersymmetric standard model. Other proposals have been made wherein a sneutrino belonging to one of the first three generations has a small $`SU(2)`$-breaking vacuum expectation value . Our proposal differs in that the fourth generation sneutrino is the sole source of $`SU(2)`$ breaking.
Already in the early days of supersymmetric model building, the intriguing similarity of the quantum numbers of the weak lepton doublets ($`L^a`$) and the electroweak Higgs $`H`$ made it tempting to speculate that the Higgs might actually be the superpartner of a lepton (that is, a slepton). Unfortunately, within the usual paradigm with the fundamental Planck scale $`10^{19}`$ GeV this is impossible because of the well known rules for writing supersymmetric Lagrangians. Thus, the couplings
$$L^{}QU_c,$$
(2)
which in these scenarios are required to give masses to the up-type quarks, would be suppressed by the factor $`M_{susy}^2/M_P^2`$, which can at best be $`10^{32}`$ or so.
However, the situation is dramatically different if the fundamental Planck scale is around a TeV. Now the desired couplings of the slepton Higgs to the fermions can be generated through the couplings in the Kähler potential
$$d^4\theta g_{ab}^u\frac{X^{}L^{}}{M_{Pf}^2}Q^aU_c^b,$$
(3)
where $`X`$ is the “spurion” superfield which breaks supersymmetry through the F-term $`F_X\theta ^2M_{Pf}^2`$. Then the above terms generate effective Yukawa couplings of the form $`g_{ab}^uL_{\mathrm{scalar}}^{}Q_{\mathrm{fermion}}^aU_{\mathrm{c},\mathrm{fermion}}^b`$, where $`L`$ is the lepton superfield whose scalar component is identified with the Higgs, and $`U_c`$ is the charge-conjugated up-type quark. Note that in this context no additional suppression appears. The couplings that generate masses of the down quarks and charged leptons can come (depending on the details of a given model) from the superpotential (as usual)
$$\mathrm{\Delta }W=g_{ab}^dLQ^aD^b+g_{ab}^eLL^ae_c^b,$$
(4)
and/or from the Kähler potential
$$\mathrm{\Delta }K=\frac{X^{}}{M_{Pf}^2}(g_{ab}^dLQ^aD^b+g_{ab}^eLL^ae_c^b).$$
(5)
Small neutrino masses (if they are non-zero) are more subtle and may involve intrinsically higher dimensional mechanisms which will not be discussed here. Here we should point out that the mass of the fermionic component of $`L`$ (that is, of the lepton superfield whose scalar component is identified with the Higgs) cannot be generated from the above couplings. Indeed, couplings $`LLe_c^a`$ and $`X^{}LLe_c^a`$ vanish due to the antisymmetry of the weak $`SU(2)`$ contractions. However, there are other terms in the Kähler potential that give rise to the desired Yukawa couplings $`L_{\mathrm{scalar}}L_{\mathrm{fermion}}e_{\mathrm{c},\mathrm{fermion}}^a`$. For instance, consider the term
$$\frac{(𝒟X)X^{}}{M_{Pf}^4}L(𝒟L)e_c^a,$$
(6)
where $`𝒟`$ denotes the covariant superderivative. In fact, if there is some hierarchy between the supersymmetry breaking scale and $`M_{Pf}`$, then this might be (partially) responsible for the fermion mass hierarchy between the quark and lepton sectors.
A priori within this framework there might arise problems associated with lepton flavor violation. However, this problem might be solved by introducing appropriate flavor symmetries. We specialize to the case where the electroweak Higgs is identified with a fourth generation slepton $`L^4`$. (It will become clear momentarily why we cannot identify the Higgs boson with the superpartner of a known neutrino.) Then the couplings (4) that generate charged lepton masses will also generate the couplings of the form
$$g_{ab}^eL_{\mathrm{fermion}}^4L_{\mathrm{scalar}}^ae_{\mathrm{c},\mathrm{fermion}}^b,$$
(7)
which will generically induce various flavor violating transitions. The same will be true if masses are generated from (5). Thus, we have to impose flavor symmetries. For instance, we can attempt to impose individual $`e`$-, $`\mu `$\- and $`\tau `$-lepton number symmetries. This would guarantee that the individual lepton numbers are conserved. Alternatively, we may attempt to solve the flavor violation problem along the lines of <sup>2</sup><sup>2</sup>2Fermion Yukawa hierarchy from flavor symmetry breaking on distant branes can also be combined with this scenario.. Here we should point out that the baryon number violating terms must be adequately suppressed to avoid too rapid proton decay. Generically this requires imposing additional symmetries (see, e.g., ). However, it appears that this can be done consistently with generating acceptable fermion masses in the above framework .
The spectrum of new particles in this class of models differs from that found in the minimal supersymmetric standard model. In particular, the model has only a single Higgs boson and only three neutralinos. The charginos of the model are a linear combination of the charged gauginos and the fourth generation lepton. Their mass matrix has the form
$$\left(\begin{array}{cc}M_2& 0\\ \sqrt{2}m_W& m_4\end{array}\right)$$
(8)
in the $`(\stackrel{~}{W},L_4)`$ basis, where $`M_2`$ is the $`SU(2)`$ gaugino mass, and $`m_4`$ is the fourth generation lepton mass. The neutralinos are mixtures of the neutral gauginos and the fourth generation neutrino. In the $`(\stackrel{~}{B},\stackrel{~}{W}^3,\nu _4)`$ basis, their mass matrix has the form
$$\left(\begin{array}{ccc}M_1& 0& m_Z\mathrm{sin}\theta _W\\ 0& M_2& m_Z\mathrm{cos}\theta _W\\ m_Z\mathrm{sin}\theta _W& m_Z\mathrm{cos}\theta _W& 0\end{array}\right).$$
(9)
We can now explain why it is impossible to identify the Higgs boson with the superpartner of a known neutrino. Indeed, had we identified the Higgs doublet with, say, the tau lepton doublet, then we would find that tau neutrino typically acquires a see-saw type mass of order $`m_Z^2/M_{1,2}`$. This mass, of order 10 GeV, is far too large. Hence we choose the alternate route of introducing a fourth generation, and identifying its sneutrino with the Higgs boson.
Limits from collider searches for new particles constrain the parameter space of this class of models. The charginos have masses that can exceed LEP-II collider limits , provided that $`M_2`$ and $`m_4`$ are sufficiently large. However, the neutralinos cannot be made arbitrarily heavy, and indeed this class of models always has at least one relatively light neutralino. As an example, take the values
$$M_1=45\mathrm{GeV},M_2=110\mathrm{GeV},m_4=185\mathrm{GeV}.$$
(10)
We then have $`m_{\chi _{1,2}^+}=90,226`$ GeV, while the neutralino masses are
$$m_{\chi _1^0}=56\mathrm{GeV},m_{\chi _2^0}=57\mathrm{GeV},m_{\chi _3^0}=156\mathrm{GeV}.$$
(11)
The lightest neutralino is predominantly $`U(1)`$ gaugino, the next to lightest is mostly neutrino, and the heaviest is mostly $`SU(2)`$ gaugino. In $`e^+e^{}`$ collisions at $`\sqrt{s}=183`$ GeV, the production cross sections for the lightest two neutralinos are 0.002 pb and 0.31 pb respectively, provided that the sfermions are heavy so that $`Z`$ exchange is the dominant production mechanism. The relative size of the cross sections is easy to understand, since the lightest neutralino is mostly bino and therefore couples very weakly to the $`Z`$. In this particular case, neutralino production will be dominated by $`\chi _2^0`$ pairs. The $`\chi _2^0`$ may either decay promptly to three-body final states, or may decay first to $`\chi _1^0Z^{}`$ or $`\chi _1^0\stackrel{~}{\nu }_4^{}`$, leading to five (or more)-body final states. The decays of the $`\chi _{1,2}^0`$ are highly model dependent, but we can delineate the main possibilities.
1. R-parity is strictly preserved for the first three generations, and broken only by the Yukawa couplings of the fourth generation lepton. Here we have two possibilities, depending on the mass of the gravitino $`\stackrel{~}{G}`$. If decays to the gravitino are allowed, we expect decays like $`\chi _{1,2}^0\overline{f}f\stackrel{~}{G}`$, where $`f`$ can be a lepton or quark. If decays to the gravitino are kinematically forbidden, the lightest neutralino is absolutely stable. In this case, $`\chi _2^0`$ pair production will result in events with rather soft jets and leptons coming from the $`\chi _2^0\chi _1^0Z^{},\chi _1^0\stackrel{~}{\nu }_4^{}`$ decay. Also note that the lightest neutralino in this case is the SUSY dark matter candidate.
2. If R-parity is broken for all four generations, then $`\chi _{1,2}^0`$ will decay to the standard model particles. The decays are mediated by couplings of the form $`LLe_c`$ or $`LQD`$, where the superfields can belong to any of the four generations. The cross section for $`\chi _2^0`$ pair production for the sample point given above is 0.31 pb. This exceeds the experimental limit of roughly 0.1 pb if the neutralino decays primarily to leptonic final states via $`LLe_c`$ couplings. Hence the neutralinos must decay primarily to hadronic final states via couplings like $`LQD`$. In this case, the cross section limits are about 0.6 pb , and the sample point is not excluded. Hence we expect decays like $`\chi ^0q\stackrel{~}{q}^{}qq\mathrm{}`$, so that the final states for $`\chi _2^0`$ pair production will contain 4 jets together with either charged leptons or missing energy if the $`\chi _2^0`$ decays promptly. If it decays first to $`\chi _1^0Z^{}`$ or $`\chi _1^0\stackrel{~}{\nu }_4^{}`$ there will be additional particles in the final state from $`Z,\stackrel{~}{\nu }f\overline{f}`$. Conservative limits on the dimensionless coefficients of $`LQD`$ couplings among the first three generations are at the level of 0.4-0.001 . Under less conservative assumptions, tighter constraints can be derived . For instance, bounds from neutrinoless double beta decay imply $`\lambda _{113}^{}\lambda _{131}^{}<8\times 10^8`$. Hence it seems not unlikely that the $`b`$-quark Yukawa coupling $`L_4Q_3D_3`$ is among the largest of the R-parity violating terms that can contribute to neutralino decay. If this is the case, the final states will frequently include $`b`$-flavored hadrons.
Although we have illustrated the qualitative features of this scenario for a particular point in parameter space, certain predictions are fairly model independent. In particular, we expect that models of this type will always possess at least one neutralino with mass below 65 GeV. Indeed, varying $`M_1`$ and $`M_2`$ between 0 and 1 TeV, we find that the neutralino mass matrix always has at least one eigenvalue below 65 GeV. Also, we note that the neutralino production cross section is not likely to be too small. The cross section from $`Z`$ exchange, summed over neutralino species, is larger than 0.23 pb for all points in parameter space where $`M_{1,2}1`$ TeV, where charginos cannot be pair produced at LEP-II, and where $`Z`$ decays to pairs of neutralinos are kinematically forbidden. The cross section could be smaller if there are large cancellations between $`Z`$ exchange and $`t`$-channel sfermion exchange. If we permit lighter charginos, then of course we expect still more light particles and a richer phenomenology. We have emphasized the case of heavy charginos and lighter neutralinos since constraints on R-parity violating charginos are quite stringent .
We would like to end this note by summarizing some of the model independent predictions of the above scenario.
1. R-parity is necessarily broken by the Yukawa couplings. As a result, the production and decay of superpartners is modified compared to what one finds in R-parity conserving models. We have outlined the most novel aspects of this scenario above in connection with neutralino pair production. Apart from this, one expects the usual signatures of R-parity violating supersymmetry, such as the absence of missing energy in SUSY particle decays.
2. A light Higgs is no longer a robust prediction of supersymmetry. This is much in the spirit of where it was pointed out that $`M_{Pf}`$ suppressed operators can ameliorate an upper bound on the Higgs mass in the non-supersymmetric context. In our case the Higgs potential can come from the Kähler potential
$$\frac{X^{}X}{M_{Pf}^2}(LL^{}+\frac{a}{M_{Pf}^2}(LL^{})^2+\mathrm{}),$$
(12)
which can in principle tolerate any mass the Higgs could have in the context of the non-supersymmetric Standard Model.
3. Although we no longer necessarily expect a light Higgs, we do in general expect one or more light neutralinos, with mass less than 65 GeV. If the charginos are too heavy to be pair produced at LEP-II, then at least one of the neutralinos is likely to have a production cross section of 0.25 pb or larger, which is not too far from present experimental limits.
Finally, let us point out that the entire scheme discussed in this note is possible because supersymmetry in the context of TeV scale gravity is broken softly but (almost) completely. Thus, for instance, the Yukawa couplings not allowed by supersymmetry are generated via higher dimensional operators in the Kähler potential once supersymmetry is broken. On the other hand, we have shown that the underlying supersymmetry leads to non-trivial predictions in this scenario, such as the presence of light neutralinos. It is conceivable that this scenario can be confirmed or excluded by LEP-II or the Tevatron.
We would like to thank Gia Dvali for collaboration at initial stages of this work and valuable discussions. We would also like to thank Howard Georgi, Henry Tye, and Cumrun Vafa for useful discussions. The work of A.K.G. was supported by NSF grant PHY-9802709. The work of Z.K. was supported in part by NSF grant PHY-96-02074, and the DOE 1994 OJI award. This work was completed during Z.K.’s visit at the Enrico Fermi Institute. Z.K. would also like to thank Albert and Ribena Yu for financial support.
|
no-problem/9906/astro-ph9906013.html
|
ar5iv
|
text
|
# References
The degree scale feature in the CMB spectrum in the fractal universe
D.L. Khokhlov
Sumy State University, R.-Korsakov St. 2,
Sumy 244007, Ukraine
E-mail: khokhlov@cafe.sumy.ua
## Abstract
The position of the degree scale feature in the CMB spectrum is determined within the framework of the fractal universe with a power index of 2.
The presence of any feature in the cosmic microwave background (CMB) anisotropy spectrum whose physical scale is known provides us with the ability to perform the angular diameter distance test . Any feature progects as an anisotropy onto an angular scale associated with multipole
$$\mathrm{}_{eff}=kd$$
(1)
where $`k`$ is the size of the feature in k-space, $`d`$ is the angular diameter distance to the feature. Anisotropy measurements on degree scales pin down the feature in the CMB spectrum. The position of the feature is $`\mathrm{}_{eff}=263`$ , $`\mathrm{}_{eff}=260`$ . In the Friedmann universe, the degree scale feature is considered to arise due to acoustic oscillations in the photon-baryon fluid at last scattering. The feature represents the sound horizon when the universe recombines.
The potential fluctuation $`\mathrm{\Delta }M/M`$ is a relation of the scales. The mass $`M`$ defines the scale of homogeneity. The mass $`\mathrm{\Delta }M`$ defines the scale of fluctuations. The scale of homogeneity can be specified with the distance $`r`$ covered by the CMB photon. The distance $`r`$ gives the mass of radiation $`M`$ restricted within $`r`$. Then the size of the potential fluctuation $`\mathrm{\Delta }r`$ is defined via the value $`\mathrm{\Delta }M/M`$. The size of the potential fluctuation $`\mathrm{\Delta }r`$ represents the feature in the CMB spectrum, with the value $`r`$ being the distance to the feature.
Determine the above feature in the CMB spectrum within the framework of the universe with the linear evolution law . The model is based on the premise that the evolution of the universe is not defined by the matter but is a priori specified. The scale factor of the universe is a linear function of time
$$a=ct.$$
(2)
The scale of mass is a linear function of the scale factor
$$Ma.$$
(3)
This leads to that the universe has the fractal structure with a power index of 2 . In the fractal universe, the relation of the masses within radii $`R_1`$ and $`R_2`$ is given by
$$\frac{M_1(<R_1)}{M_2(<R_2)}=\left(\frac{R_1}{R_2}\right)^2.$$
(4)
Then the size of the potential fluctuation is given by
$$\frac{\mathrm{\Delta }r}{r}=\left(\frac{\mathrm{\Delta }M}{M}\right)^{1/2}.$$
(5)
In view of this, the multipole in the CMB spectrum is given by
$$\mathrm{}_{eff}=\left(\frac{\mathrm{\Delta }r}{r}\right)^1=\left(\frac{\mathrm{\Delta }M}{M}\right)^{1/2}.$$
(6)
According to , the value of potential fluctuations of radiation is $`\mathrm{\Delta }M/M=1.49\times 10^5`$. From this calculations yield $`\mathrm{}_{eff}=259`$.
|
no-problem/9906/cond-mat9906298.html
|
ar5iv
|
text
|
# REFERENCES
Self-Organization of Value and Demand
R. Donangelo<sup>a</sup> and K. Sneppen<sup>b</sup>
<sup>a</sup> Instituto de Física, Universidade Federal do Rio de Janeiro,
C.P. 68528, 21945-970 Rio de Janeiro, Brazil
<sup>b</sup> Nordita, Blegdamsvej 17, 2100 Copenhagen, Denmark
Abstract: We study the dynamics of exchange value in a system composed of many interacting agents. The simple model we propose exhibits cooperative emergence and collapse of global value for individual goods. We demonstrate that the demand that drives the value exhibits non-Gaussian “fat tails” and typical fluctuations which grow with time interval as $`\mathrm{\Delta }t^H`$, with $`H0.7`$.
PACS numbers: 02.50, 05.10.-a, 05.40.-a, 05.45.-a, 05.65.+b, 05.70.Ln, 87.23.Ge, 89.90.+n
The self organizational patterns of large non-equilibrium systems are of large current interest. One aspect of such systems is their tendency to display cooperative behavior, evidenced, for example, in the form of the occasional coordinated activity throughout the system . Economic systems provide examples of such cooperative behavior, that indeed show some of the characteristics of self organizing systems. More specifically, agents reach a collective agreement about what should be considered as valuable, and then use it as a common trade object (money) . Furthermore, the fluctuation of relative value with time $`t`$ displays anomalous Hurst exponents (typical $`\mathrm{\Delta }D\mathrm{\Delta }t^H`$ with $`H>1/2`$) and non-Gaussian statistics. The latter fact was first noticed by Mandelbrot and later quantified in the observation that the succession of daily, weekly and monthly distributions of exchange values possibly converges towards a Gaussian .
Recently a number of theoretical approaches have been developed to deal either with the emergence of money as a cooperative phenomenon, or with the non-Gaussian fluctuations associated to it. In particular the work of Yasutomi deals with the stochastic nature of emergence of certain goods as means of common exchange between all the agents, i.e. how they become money. Yasutomi makes the observation that if agents tends to accept what others accept, then trade is facilitated. Therefore, emergence of money becomes related to history dependent processes, which, as discussed by B. Arthur tend to lock the market into certain trade patterns. This, however does not explain how one popular product may be replaced by another, and thus result in an open and ever fluctuating system. Yasutomi suggestion to solve this problem is to include both trade costs and an evolving threshold for transactions. In this way he obtains a bi-stable system where money emerges and collapses quasiperiodically. Dealing with the nature of fluctuations alone, Bak, Paczuski and Shubik suggested a model with only one product, and concluded that non-Gaussian fluctuations could be associated with the crowding obtained when agents imitate each other’s prize assessments . They further obtained the anomalous Hurst exponents $`H=0.65`$ in a scenario where local fluctuations are amplified by the global activity level. Levy, Levy and Solomon emphasize the effect of heterogeneous expectations for the traders in a stock-bond market. However, their specific model develops an unrealistic periodicity for large systems . As we will see, our model self organizes heterogenity in a market where each agent has its own limited history. Furthermore, it predicts the emergence of non-periodic but persistent fluctuations in a market where many products compete for attention.
In the present work we suggest a simple model that suggests that emergence of “money” and its anomalous fluctuation in value are two sides of the same cooperative phenomena. We suggest that value emerges through agents which make simple decisions based on their individual memory of earlier encounters with other agents. The agents are not assumed to be smart, the only trade that occurs is a one-good-for-one-good trade, and agents basically act in order to keep stock of all products. The model leads to emergence and collapse of money, it exhibits non-Gaussian statistics and also displays long time fluctuations quantified by anomalous Hurst exponents.
The model we propose is as follows. We consider $`N_{ag}`$ agents and $`N_{pr}`$ different products. Initially we give $`N_{unit}`$ units of the products to each agent. The number $`N_{unit}`$ is fixed, but the products are chosen at random, so the individuals are not in exactly the same situation.
At each timestep we select two random agents and let them attempt to perform a trade between them. The trade starts by comparing the list of goods that each agent lacks and therefore would like to get from the other agent in exchange for goods it has in stock.
We first consider the simple barter exchange procedure: when each of the agents has products that the other needs, then one of these products, chosen at random, is exchanged.
In case such a barter exchange is not possible we consider the “money” exchange procedure: one or both of the agents accept goods which they do not lack, but consider useful for future exchanges.
In order to determine the usefulness of a product, each agent $`i`$ keeps a record of the last requests for goods it received in encounters with other agents. This memory is finite, having a length of $`N_{mem}`$ positions, each of which registers a product that was requested. As the memory gets filled, the record of old transactions is lost. Agents accept products they already have in stock with a probability based on its memory record. The chance of accepting such a good $`j`$ is taken to be proportional to the number of times $`T_{ij}`$ that good $`j`$ appears on the memory list of agent $`i`$:
$$p_{ij}=\frac{T_{ij}}{N_{mem}},$$
(1)
where we have used the fact that $`_jT_{ij}=N_{mem}`$. These two exchange mechanisms define our model.
In order to understand the dynamics of this model we first show, in Fig. 1, the time evolution of the number of different kinds of trades. The case depicted corresponds to $`N_{ag}=200`$ agents trading $`N_{pr}=200`$ types of products between them. The memory list of each agent was chosen to be $`N_{mem}=400`$ items long and the total “richness” of the economy was fixed by setting the numbers of products possessed by each agent to $`N_{unit}=400`$. This may be considered a typical set of parameters, and the behavior is similar for a wide range of number of goods, memory sizes and richness values, as long as memories are not too short and richness not too high or too small. If the economy is too rich, barter exchange disappears completely as all agents will always have all types of goods. On the other hand, if the economy is very poor, barter exchange dominates completely and “money” exchange does not emerge. Besides, in order for money to appear at all, the memory of each agent must not be much shorter than the number of different products, $`N_{pr}`$, (however, it can be much smaller than $`N_{ag}`$).
Fig. 1 shows that after a relatively short equilibration time the ratio of barter to money transactions saturates. The fact that money transactions dominate means that the agents have already distributed their holdings in an efficient way, thus allowing money exchange to become the dominating mode of transaction. We also notice that the sum of the two types of exchanges is less than 100% , meaning that approximately 40% of encounters between agents did not lead to trades, for this set of parameters.
The interesting feature of our model is however not the relative number of these two types of transactions, but the the fact that some products become valuable as means of exchange. In order to quantify this we define the monetary value of a good $`j`$ as the number of agents $`M=M(j)`$ that consider this good to be the easiest to trade. In other words, each agent $`i`$ indicates the good $`j`$ with highest $`p_{ij}`$.
In Fig. 2a,b we show the value of two particular goods as function of time, which here we measure as number of encounters between agents. Notice that the time scale for the evolution is a factor 100 larger than that of Fig. 1. This shows that, although the number of money and barter exchanges has equilibrated almost instantaneously in the timescale of Fig. 2, the evolution of the value of the goods displays an interesting dynamics on this larger timescale. In Fig. 2c we have plotted the number of agents that accept the most desirable good. From comparison with Figs. 2a and 2b, we notice that the good in Fig. 2a becomes the most accepted in terms of exchange early in the evolution of the system, and remains in that condition until more than 900000 encounters have taken place, in this particular history. At this point, the good from Fig. 2b takes over, for a briefer period, until time 1200000. There is a couple of additional crossovers between these two products, but at the end another product arises and takes over as being the most popular.
We stress that an important feature of the model is that often, and over long periods, one particular good is considered valuable by a majority of the agents. This appears without any a priori property of this product; the value of the good develops through a cooperative agreement between the agents about what is valuable.
How is this agreement reached? In our model the agents only transfer information about what they need to replenish their stocks. Thus the common concept of value emerges when many agents need a good, e.g. because it is concentrated by some of the agents. This accumulation is possible in our simple model because an agent accepts goods even when having them in sufficient stock, if it feels they will be useful for future transactions. This property in itself makes a desired product to circulate, and possibly concentrate, more than other goods, thereby making it more needed across the system.
However goods in high demand also collapse, as it was seen at time $`800000`$ in Fig. 2a. This is because, due to the large number of money transactions, the most valuable good may, through random fluctuations, become better distributed than some other good, which then replaces it as the system’s money.
We now try to quantify these fluctuations. Fig. 3a displays the changes in the value of $`M`$, which time evolution was seen in Fig. 2c, for various time intervals $`\mathrm{\Delta }t`$. The figure shows that, for sufficiently large values of $`\mathrm{\Delta }t`$, these fluctuations have exponential tails. This is in contrast to the normal Gaussian behavior expected for value assigned by independent agents. The fluctuations obtained for the different values of $`\mathrm{\Delta }t`$ could not be made to collapse into one curve. This signals that the short and long time statistics of $`M`$ cannot be described by the same Hurst exponent.
In order to examine the statistics of the underlying demand of goods we show, in Fig.3b, the fluctuations in the number of times $`D`$ that a given product $`j`$ appears in the memory of the system,
$$D(j)=\underset{i}{}T_{ij},$$
(2)
where the sum runs over all agents $`i`$ in the system. In Fig. 3b we show a data collapse of fluctuations in $`D`$, i.e. $`P(\mathrm{\Delta }D)\mathrm{\Delta }t^HP(\frac{\mathrm{\Delta }D}{\mathrm{\Delta }t^H})`$ for 3 different values of $`\mathrm{\Delta }t`$. We notice that the curves collapse with Hurst exponent $`H=0.70`$, an observation that we confirmed by finding numerically that $`\left((D(t+\mathrm{\Delta }t)D(t))^2\right)^{\frac{1}{2}}\mathrm{\Delta }t^{0.68\pm 0.02}`$, over more than three orders of magnitude. The fact that $`H>0.5`$ indicates that fluctuations in demand exhibit persistency, i.e. that trends are amplified by a self organized cooperative feedback in our model. This relatively large anomalous Hurst exponent requires that there is a sizeable number of different products. We have found that persistence exists for nearly all system sizes. The value of the Hurst exponent was found to 0.6 for a size 10 system whereas it was found to be 0.7 for system size 100 and 1000. For size dependence we mean here that we scale equally the parameters of the model: $`N_{ag},N_{pr},N_{mem},N_{unit}`$. We have also checked that a change in $`N_{ag}`$ alone from 100 to 1000 also leads to a Hurst exponent of 0.7. Only in the case of $`N_{pr}=2`$ we found $`H=0.5`$, and this for any value of the other parameters.
We further notice, from Fig. 3b, that the fluctuations exhibit fat tails. In particular, the short time scale statistics tend to have exponential tails whereas the long timescale fluctuations are more Gaussian-like. The difference of Hurst scaling between value $`M`$ and demand $`D`$ reflects the fact that changes in the most wanted product are faster, but less persistent than structural changes in the composition of demand.
We have examined variations of the above model, as, for example, including reluctance for the agents to trade away goods they consider valuable, i.e. with high $`p_{ij}`$. Also we have considered the possibility that agents include in $`T_i`$ only trades that were effectively performed, and not just requests. Finally, we have tested the case where the $`T_i`$ lists also include requests for “money trades”, arising from what other agents consider valuable. All these cases change the information exchanged between agents about what is valuable. However, the qualitative behavior of our model was in all cases similar, indicating its robustness.
It is tempting to compare the emergence of value in the above model with the emergence of monetary systems in the real world. A beautiful historical example of the collapse of one currency and the appearance of a new one in a XVII century chinese town is given in Yasutomi’s work . More familiar to the reader may be similar events in the market for tulips in the Netherlands, or the market for gold in more recent times. In all these cases the acceptability of the good or currency may drop dramatically, without any deeper reason than the fact that nobody considers them valuable any more. More quantifiable are the values of currencies, where now, due to events that have taken place in this century, the U.S. Dollar has become globally accepted. In that regards one should mention that fluctuations in monetary value seem to exhibit fat tails , and long time fluctuations that can be characterized by a Hurst exponent $`H0.55`$ . All these features are consistent with our simple scenario. We stress that our model is schematic and does not include any development of strategy by the agents, strategies which would naturally influence the emergence of cooperativity .
Furthermore, the present version of the model does not include effects related to production or consumption of goods. We have verified that the main result of including these processes, is that products that are easily produced never become valuable. Thus, if all goods are produced and consumed at a high rate, no common “money” will emerge. On the other hand, if products are produced at a low rate the results presented above remain valid, indicating again the robustness of the present model. In fact, the addition of a slow production and consumption of goods allows the system to settle into a statistical stationary state. This avoids products from falling, one by one, into the absorbing state consisting of having the product present in the stock of all agents, and, consequently, being gradually removed from the memories. As soon as such a product is removed from all memories, it is effectively not anymore present in the economy. The effect of the decay mentioned above is negligible in the time scale of the calculations presented in this work.
Finally we would like to emphasize that our work purports to model behavior in a social setting where information exchange is a key feature. The fact that some forms of information exchange lead to self-organization has already been proposed by e.g. Bonabeau et al. . There sociological hierarchies emerge through comparisons of a dominance index assigned to each individual. Our model is more adapted for market behavior, in particular because it also considers the exchange of goods, and not only of information. By emphasizing the interplay between exchange of products and information, we have constructed a minimalistic market model. However, since we explicitly require that our agents do not develop different strategies, and that all the products can be exchanged only in a one-to-one manner, our model may not give answers to some questions that may be adressed in more detailed simulation models e.g. employing minority games , or the more elaborate models of strategy and investments programs also existing in the economic literature, e.g. the one by Kim and Markowitz .
In summary, we have constructed a simple model for the cooperative concept of money. The model suggests that money emerges and collapses as a simple consequence of trade of goods between interacting agents with memory. The concept of value arises from the probability that a local dynamical pattern (the need for certain products by individual agents) results in a global one (the general acceptability of products for exchange). An important consequence, possibly valid for the dynamics of other large systems, is that a tendency to draw information from previously encountered patterns may result in a dynamic behavior that exhibits non-Gaussian fluctuations and persistency, evidenced by anomalous Hurst exponents.
We thank Maya Paczuski and Dietrich Stauffer for constructive comments to the manuscript. R.D. thanks Nordita for its hospitality and financial support during the time that this work was performed.
Figure Captions
* Fig. 1. Probability, measured as a percentage of the encounters between agents that result in barter (full line) or money exchanges (dashed line), as defined in the text.
* Fig. 2. Number of agents $`M`$ that consider a particular product as the most valuable (a and b), and the maximum number of agents that, at a given time, consider the same product as the most valuable (c).
* Fig. 3. (a) Probability of having changes in the value $`M`$ as a function of their size, for the three different time steps $`\mathrm{\Delta }t=200000`$ (full line), $`\mathrm{\Delta }t=20000`$ (long dashed line), and $`\mathrm{\Delta }t=2000`$ (short dashed line). (b) Probability of having changes in the demand $`D`$. The demand is here measured in units of $`\mathrm{\Delta }D/\mathrm{\Delta }t^H`$, where $`\mathrm{\Delta }t`$ takes the values $`200`$, $`2000`$, and $`20000`$ (three thin lines), and we took for the Hurst exponent the value $`H=0.7`$. The thick line depicts the normal probability distribution having the same mean and standard deviation as that for $`\mathrm{\Delta }t=200`$. The peaks near 0 observed in the other distributions are attributed to border effects, arising from transitions near $`D=0`$.
|
no-problem/9906/hep-th9906116.html
|
ar5iv
|
text
|
# Untitled Document
DUALITY IN LIOUVILLE THEORY AS A REDUCED SYMMETRY
L. O’Raifeartaigh and V. V. Sreedhar
School of Theoretical Physics
Dublin Institute for Advanced Studies
10, Burlington Road, Dublin 4
Ireland
Abstract
The origin of the rather mysterious duality symmetry found in quantum Liouville theory is investigated by considering the Liouville theory as the reduction of a WZW-like theory in which the form of the potential for the Cartan field is not fixed a priori. It is shown that in the classical theory conformal invariance places no condition on the form of the potential, but the conformal invariance of the classical reduction requires that it be an exponential. In contrast, the quantum theory requires that, even before reduction, the potential be a sum of two exponentials. The duality of these two exponentials is the fore-runner of the Liouville duality. An interpretation for the reflection symmetry found in quantum Liouville theory is also obtained along similar lines.
PACS: 11.10Kk; 11.15-q; 11.25Hf; 11.10Lm; 04.60Kz DIAS-STP-99-05
A single real scalar field governed by the so-called Liouville Action plays a ubiquitous role in the study of two-dimensional conformal and integrable field theories, and is of considerable importance in the context of quantum gravity and string theory . Amongst its many unusual properties is a particularly intriguing one, namely, the invariance of the quantum theory under a duality symmetry which is absent in the classical theory. This duality symmetry is responsible for the doubling of a one-dimensional lattice of poles (in the parameter space) of the coefficient of the three-point function of vertex operators, as explained in detail in . It was also shown there that the three-point function, with the correct pole structure, may be derived by incorporating the duality into the definition of the path integral for the Liouville theory on a sphere.
As is well-known, the classical Liouville theory may be obtained by imposing a set of linear first class constraints on the currents of an SL(2, R) Wess-Zumino-Witten (WZW) model . In the present paper, we investigate the origin of the Liouville duality by considering the Liouville theory as the reduction of a WZW-like theory in which the form of the potential for the Cartan field is not fixed a priori. It will be shown that the conformal invariance of the classical reduction forces this theory to be the SL(2, R) WZW model, but an essential difference between the classical and quantum reductions lies at the heart of the duality symmetry of the quantum Liouville theory.
The WZW-like theory we start with is defined by the following Action:
$$S=kd^2x\sqrt{g}[\frac{1}{2}g^{\mu \nu }(_\mu \varphi )(_\nu \varphi )+g^{\mu \nu }(_\mu a)(_\nu c)V(\varphi )]$$
$`(1)`$
In the above equation, $`\varphi ,a`$ and $`c`$ are real scalar fields and $`V(\varphi )`$ is an arbitrary function of $`\varphi `$. Obviously, (1) describes a conformally invariant model at the classical level. Upon the addition of a topological term $`ikϵ^{\mu \nu }(_\mu a)(_\nu c)V(\varphi )`$, the Action becomes
$$S=kd^2z[\frac{1}{2}(\overline{}\varphi )(\varphi )+(\overline{}a)(c)V(\varphi )]$$
$`(2)`$
where we have used complex coordinates $`z=\frac{x_0+ix_1}{2},\overline{z}=\frac{x_0ix_1}{2}`$ in terms of which $`=_0i_1,\overline{}=_0+i_1`$ and $`d^2z=2idzd\overline{z}`$. Note that choosing the topological term with the opposite sign would have resulted in $`\overline{}`$ in (2). The classical energy-momentum tensor, $`T^{\alpha \beta }\frac{1}{\sqrt{g}}\frac{\delta S}{\delta g_{\alpha \beta }}`$, is given by
$$\begin{array}{cc}\hfill T^{\alpha \beta }=\frac{k}{2}& g^{\alpha \beta }g^{\mu \nu }[\frac{1}{2}(_\mu \varphi )(_\nu \varphi )+(_\mu a)(_\nu c)V(\varphi )]\hfill \\ & kg^{\mu \alpha }g^{\beta \nu }[\frac{1}{2}(_\mu \varphi )(_\nu \varphi )+(_\mu a)(_\nu c)V(\varphi )]\hfill \end{array}$$
$`(3)`$
As usual, it is traceless, symmetric, and conserved, and does not receive any contribution from the topological term as the latter is independent of the metric.
Note that for $`V(\varphi )=e^\varphi `$, Eq. (2) simplifies to the SL(2, R) WZW Action, with $`a`$ and $`c`$ locally parametrising the nilpotent subgroups, and $`\varphi `$ the abelian subgroup, respectively, in a Gauss decomposition of the group-valued WZW field. In the following, however, we shall allow $`V(\varphi )`$ to be arbitrary to begin with, and examine the restrictions imposed on its form, by requiring the Action (2) to be amenable to a conformally invariant reduction. It will be shown that for this requirement to be satisfied, $`V(\varphi )`$ is governed by a second order functional differential equation whose two solutions are dual to each other. In the classical limit only one of the two solutions of this equation has a non-trivial ($`\varphi `$-dependent) value which corresponds to the standard SL(2, R) WZW case.
We shall first present the analysis of conformal invariance in the classical case. The expressions for the momenta $`\pi `$ conjugate to the fields $`\varphi ,a,c`$ follow from the Action (2) and are given by
$$\pi _\varphi =k\dot{\varphi },\pi _a=kV(\varphi )c,\text{and}\pi _c=kV(\varphi )\overline{}a$$
$`(4)`$
The non-vanishing Poisson brackets, evaluated at equal time, are given by
$$\{\varphi (x,t),\pi _\varphi (y,t)\}=\{a(x,t),\pi _a(y,t)\}=\{c(x,t),\pi _c(y,t)\}=\delta (xy)$$
$`(5)`$
The Virasoro generator $`T`$ is given by
$$T=𝑑zϵ(z)[\frac{k}{2}(\varphi )^2k\pi _a\pi _cV^1+2i\pi _a(_1a)](z)$$
$`(6)`$
The other Virasoro generator $`\overline{T}`$, may be obtained by making the exchanges $`ac,\overline{},ϵ\overline{ϵ}`$. As usual, the Poisson bracket of the two Virasoros vanishes, and it suffices to focus our attention on one Virasoro.
The conformal transformation of a field $`𝒪`$, generated by $`T_ϵ`$, is given by $`\delta _ϵ𝒪\{T_ϵ,𝒪\}`$. Hence the fields have the following conformal variations:
$$\delta _ϵa=ϵa,\delta _ϵc=ϵc,\delta _ϵ\varphi =ϵ\varphi ,\delta _ϵV(\varphi )=ϵV(\varphi )$$
$`(7)`$
$$\delta _ϵ\pi _a=ϵ\pi _a+(ϵ)\pi _a,\delta _ϵ\pi _c=0$$
$`(8)`$
The momentum $`\pi _\varphi `$ does not transform like a conformal primary field:
$$\delta _ϵ\pi _\varphi =ϵ\pi _\varphi +(ϵ)\frac{k}{2}\varphi $$
This is expected because $`\pi _\varphi =(k\dot{\varphi })`$ is a mixture of $``$ and $`\overline{}`$ derivatives of a scalar field. As can be seen from the following equation, however, $`\varphi `$ transforms like a conformal primary field:
$$\delta _ϵ\varphi =ϵ\varphi +(ϵ)\varphi $$
$`(9)`$
It is also easy to check that
$$\delta _ϵT=ϵT+2(ϵ)T$$
$`(10)`$
It follows from the above equations that classically $`a,c,\varphi ,\pi _c`$ have conformal weights zero; $`\pi _a,\varphi `$ have weights one, and $`T`$ has a weight two – as expected. The weights with respect to the other Virasoro are given by the same numbers if the exchanges $`ac`$, $`ϵ\overline{ϵ}`$, and $`\overline{}`$ are done.
At this juncture it is worth noting that the Virasoro (6) is unique if we insist on keeping $`V(\varphi )`$ arbitrary. However, if we are interested in a special class of theories for which $`V`$ is restricted to be of the form
$$V=e^{\lambda \varphi }$$
$`(11)`$
the Virasoro is no longer unique.<sup>*</sup><sup>*</sup>Note that by scaling the fields $`\varphi ,a,\text{and}c`$, appropriately, (2) can be seen to represent the SL(2, R) WZW model for this choice of $`V`$. This is because it is not necessary that $`\pi _a`$ and $`\pi _c`$ should have conformal weights (1,0) and (0,1) respectively, for the Action (2) to be conformally invariant; it is only necessary that the combination $`\pi _a\pi _cV^1(\varphi )`$ has a weight (1,1). This freedom is expressible in terms of improvement terms which, when added to the standard Virasoro (6), redistribute the weights of the various fields in a way which preserves the weight of the above combination of the fields. It is straightforward to see that the improvement term $`t`$, for the Virasoro $`T`$, is given by
$$t_ϵ=\alpha 𝑑z(ϵ)[k\varphi \lambda a\pi _a](z)$$
$`(12)`$
where $`\alpha `$ is an arbitrary parameter. Clearly, with respect to the full Virasoro $`T+t`$, $`\pi _a`$ behaves like a primary field of weight $`(1+\lambda \alpha ,0)`$. It is also easy to see that the field $`e^{\lambda \varphi }`$ transforms as follows under the action of the full Virasoro $`T+t`$:
$$\delta _ϵe^{\lambda \varphi }=\lambda \alpha (ϵ)e^{\lambda \varphi }+ϵe^{\lambda \varphi }$$
$`(13)`$
Thus the combination $`\pi _a\pi _cV^1`$ has weight one with respect to the full Virasoro $`T+t`$. Note that the ratio of the coefficients of the two terms in (12) is fixed by the parameter $`\lambda `$ in the potential. It is straightforward to see that this ratio is also equal to the corresponding ratio in the improvement terms $`\overline{t}`$ for the other Virasoro $`\overline{T}`$. Further, the requirement that the Poisson bracket of the two Virasoros is zero implies that the improvement terms $`t`$ and $`\overline{t}`$ have the same overall coefficient $`\alpha `$ relative to $`T`$ and $`\overline{T}`$. It may also be noted that for the theory to remain conformally invariant, it is not possible to have only $`\lambda `$ to be zero. If $`\lambda `$ is zero, then $`\alpha `$ has to be necessarily zero.
It is worth mentioning that there is a good reason for considering the improved Virasoro $`T+t`$, and the concomitant freedom in redistributing the weights of the various fields in the theory, namely, that it is necessary for the conformal reduction of the SL(2, R) WZW theory to the Liouville theory. The constraints that accomplish this reduction are
$$\pi _a=m_a\text{and}\pi _c=m_c$$
$`(14)`$
where $`m_a`$ and $`m_c`$ are constants. Note that unless the improvement terms are present with $`\lambda \alpha =1`$, it is not possible to impose the above constraints in a conformally invariant manner since $`\pi _a`$ and $`\pi _c`$ have non-zero conformal dimensions. Thus one is forced to use the full Virasoro $`T+t`$ with $`\lambda \alpha =1`$. Indeed, the improvement term with $`\lambda \alpha =1`$ automatically provides the usual improvement terms of the Liouville theory upon reduction. Having thus motivated the need for the improvement terms, it may be noted that the argument can be reversed: Given the full Virasoro $`T+t`$ and the Action (2), conformal invariance requires $`V`$ to be of the form (11).
The situation in the quantum theory is remarkably different from that in the classical theory. The entire analysis leading up to Eq. (10) can be carried over to the quantum theory in a relatively straightforward way by replacing Poisson brackets with commutators and the appropriate factors of $`i\mathrm{}`$, as long as care is exercised in taking the ordering of the operators. The commutator of $`\pi _a`$ with the Virasoro can be calculated without ado as the Virasoro is linear in $`a`$, the conjugate field. As expected, $`\pi _a`$ remains a conformal primary field of weight (1, 0). The commutator of $`V^1`$ with the Virasoro is trickier because $`T`$ is quadratic in $`\pi _\varphi `$. A short calculation shows that
$$\delta _ϵV^1[T_ϵ,V^1]_{}=(ϵ)\left(\frac{i\mathrm{}^2}{2k}\frac{\delta ^2V^1}{\delta \varphi ^2}\right)+i\mathrm{}ϵ(V^1)$$
$`(15)`$
where we have used $`\delta (zz^{})=i[\delta (zz^{})]^2`$ . Thus, at the quantum mechanical level, $`V^1`$ ceases to behave like a conformal scalar for general $`V`$, with respect to $`T_ϵ`$. However, as in the classical case, if one adds improvement terms, it is sufficient to require that the combination $`\pi _a\pi _cV^1(\varphi )`$ has a weight (1,1). It is easy to check then that with respect to the full Virasoro $`T+t`$, $`\pi _a`$ behaves like a primary field of weight $`(1+\lambda \alpha ,0)`$ even at the quantum level. The conformal variation of $`V^1`$ with respect to the full quantum Virasoro is
$$\delta _ϵV^1[T_ϵ+t_ϵ,V^1]_{}=(ϵ)\left(\frac{i\mathrm{}^2}{2k}\frac{\delta ^2V^1}{\delta \varphi ^2}+i\mathrm{}\alpha \frac{\delta V^1}{\delta \varphi }\right)+i\mathrm{}ϵ(V^1)$$
$`(16)`$
The necessary condition for $`V^1`$ to have the correct conformal properties, can be read off from the above equation to be
$$\frac{\mathrm{}}{2k\alpha }\frac{\delta ^2V^1}{\delta \varphi ^2}+\frac{\delta V^1}{\delta \varphi }=\lambda V^1$$
$`(17)`$
Note that the above equation is a quadratic equation and has two solutions while its classical counterpart obtained by setting $`\mathrm{}=0`$ is a first order equation which leads to $`V=e^{\lambda \varphi }`$. Thus the quantum theory is slightly less stringent about the form of $`V`$, than the classical theory, with respect to the demands imposed by conformal invariance. The general solution for $`V^1`$ takes the form
$$V^1=e^{\omega _+\varphi }+\mu e^{\omega _{}\varphi }$$
$`(18)`$
where
$$\omega _\pm =\frac{\alpha k}{\mathrm{}}\frac{\alpha k}{\mathrm{}}\sqrt{1+\frac{2\mathrm{}\lambda }{\alpha k}}$$
$`(19)`$
and $`\mu `$ is a constant. Note that $`\omega _+\omega _{}=\frac{2k\lambda \alpha }{\mathrm{}}`$ — showing that the two solutions are dual to each other. Also note that in the classical limit, $`\mathrm{}0,\omega _+\lambda \text{and}\omega _{}\mathrm{}`$. In this limit the duality disappears and only (11) survives.
In passing we mention that the above results can be readily applied to work out the transformation properties of the vertex operators $`U(\varphi )=e^{2\gamma \varphi }`$ of the field $`\varphi `$. It follows from Eq. (16) that
$$\delta _ϵ(e^{2\gamma })[T_ϵ+t_ϵ,e^{2\gamma }]_{}=(ϵ)(i\mathrm{})\left(\frac{2\gamma }{k}(k\alpha \mathrm{}\gamma )\right)e^{2\alpha }+i\mathrm{}ϵ(e^{2\gamma })$$
$`(20)`$
Thus the conformal weight of the vertex operator is given by
$$\mathrm{\Delta }(e^{2\gamma })=\frac{2\gamma }{k}(k\alpha \mathrm{}\gamma )$$
$`(21)`$
which is manifestly symmetric under the exchange $`\gamma k\alpha \gamma `$ in units of $`\mathrm{}=1`$. This is the reflection symmetry of the conformal weight of a vertex operator in the quantum theory.
Substituting the solution for $`V`$ from Eqs. (18) and (19) in Eq. (2), we have
$$S=d^2z[\frac{1}{2}(\overline{}\varphi )(\varphi )+(\overline{}a)(c)(e^{\omega _+\varphi }+\mu e^{\omega _{}\varphi })^1]$$
$`(22)`$
What we have shown is that the above highly non-linear Action is conformally invariant with respect to the full Virasoro $`T+t`$ in the quantum theory, and reduces to the standard SL(2, R) WZW case in the classical limit. It easily follows that the constraints (14) reduce (22) to
$$S=d^2z[\frac{1}{2}(\overline{}\varphi )(\varphi )+m_am_c(e^{\omega _+\varphi }+\mu e^{\omega _{}\varphi })]$$
$`(23)`$
This is precisely the theory proposed in to explain the duality symmetry of quantum Liouville theory and the pole structure of its three-point function.
To summarise the results of this paper, we have shown how the duality symmetry that appears rather mysteriously in quantum Liouville theory may be interpreted as a reduced symmetry. Thus the origin of this symmetry can be traced to the fact that the conditions imposed by conformal invariance on the structure of the potential in the Liouville theory are different in the classical and quantum reductions, the latter being less stringent than the former. This is because the $`\mathrm{}`$-dependent operator ordering effects in quantum theory produce a second order functional differential equation for the potential, whose solutions are dual to each other, while the absence of such effects in the classical theory produces a first order functional differential equation whose unique solution does not allow for the possibility of a duality symmetry. A natural interpretation for the reflection symmetry in quantum Liouville theory was also obtained along these lines. It would be interesting to see if the results of this paper can be generalised to discuss the duality and reflection invariance of Toda theories as reduced symmetries.
REFERENCES
1. H. Poincare, J. Math. Pure App. 5 se 4 (1898) 157; N. Seiberg, Notes on Quantum Liouville Theory and Quantum Gravity, in ”Random Surfaces and Quantum Gravity”, ed. O. Alvarez, E. Marinari, and P. Windey, Plenum Press, 1990.
2. A. Polyakov, Phys. Lett. B103 (1981) 207; T. Curtright and C. Thorn, Phys. Rev. Lett. 48 (1982) 1309; E. Braaten, T. Curtright and C. Thorn, Phys. Lett 118 (1982) 115; Ann. Phys. 147 (1983) 365; J.-L. Gervais and A. Neveu, Nuc. Phys. B238 (1984) 125; B238 (1984) 396; 257\[FS14\] (1985) 59; E. D’Hoker and R. Jackiw, Phys. Rev. D26 (1982) 3517.
3. H. Dorn and H.-J. Otto, Phys. Lett B291 (1992) 39; Nuc. Phys. B429 (1994) 375; A. and Al. Zamolodchikov, Nuc. Phys. B477(1996) 577; L. O’Raifeartaigh, J. M. Pawlowski, and V. V. Sreedhar, Duality in Quantum Liouville Theory, hep-th/9811090; to appear in Annals of Physics.
4. F. A. Bais, T. Tjin and P. Van Driel, Nuc. Phys. B 357 (1991) 632; V. A. Fateev and S. L. Lukyanov, Int. J. Mod. Phys. A3 (1988) 507; S. L. Lukyanov, Funct. Anal. Appl. 22 (1989) 255; P. Forgács, A. Wipf, J. Balog, L. Fehér, and L. O’Raifeartaigh, Phys. Lett. B227 (1989)214; J. Balog, L. Fehér, L. O’Raifeartaigh, P. Forgács and A. Wipf, Ann. Phys. 203 (1990) 76; Phys. Lett B244(1990)435; L. Fehér, L. O’Raifaertaigh, P. Ruelle, I. Tsutsui and A. Wipf, Phys. Rep. 222 No. 1, (1992)1; P. Bouwknegt and K. Schoutens Eds. W Symmetry (Advanced series in mathematical physics, 22) World Scientific, Singapore (1995); L. O’Raifeartaigh and V. V. Sreedhar, Nuc. Phys. B520 (1998) 513; Phys. Lett. B425 (1998) 291.
5. C. Ford and L. O’Raifeartaigh, Nuc. Phys. B460 (1996) 203.
|
no-problem/9906/astro-ph9906236.html
|
ar5iv
|
text
|
# On the CCD Calibration of Zwicky galaxy magnitudes & The Properties of Nearby Field Galaxies
## 1 Introduction
Some important aspects of galaxy evolution can only be understood by studying the statistical properties of nearby field galaxies, in particular its luminosity function (LF). As well as providing vital information for galaxy evolution studies, an accurate knowledge of the present day LF is needed to normalize the number counts of galaxies at fainter magnitudes, and to understand the clustering and large scale structure of the galaxy distribution. Moreover, the colour distribution of the nearby galaxies provides a basic reference to determine star formation rates in galaxies.
Much recent work has been directed towards studying the evolution of the LF using samples of faint galaxies at high-redshift, but a consideration of the ensemble of available estimates of the LF at low redshifts suggests that there are still large inconsistencies which must be reconciled before reliable conclusions can be drawn about the implications of deep surveys.
It is common practice to fit the LF to the so called Schechter (1976) form:
$$\varphi (L)=\varphi _{}\left(\frac{L}{L^{}}\right)^\alpha \mathrm{exp}\left(\frac{L}{L^{}}\right),$$
(1)
were luminosity is related with magnitude in the usual way, $`\frac{L}{L^{}}=10^{0.4(M_{}M)}`$. Determinations of the low redshift $`B`$-band LF are available from the Stromlo-APM Survey (SAPM, Loveday et al., 1992), the Southern Sky Redshift Survey (Da Costa et al., 1994), the CfA2 redshift survey (Marzke et al., 1994), and more recently from the ESO Slice Project (Zucca et al., 1997) and the 2dF Galaxy Redshift Survey (Folkes et al., 1999). Determinations in the $`R`$-band are available from the Las Campanas Redshift Survey (LCRS, Lin et al., 1994) and from the Century Survey (CS, Geller et al., 1997). The best determinations of the Schechter function parameters from each of these surveys are listed in Table 1. The SSRS2 is based on the ESO-Uppsala photometry of Lauberts & Valentijn (1989), but transformed to the $`B(0)`$ system of Huchra (1976). Da Costa et al. (1994) ascribe the difference in the LFs of the two surveys to the large colour-term used by Lauberts & Valentijn to relate the two systems, and so we therefore use $`m_Z`$ for Zwicky magnitudes and $`B(0)`$ for SSRS2 magnitudes. Da Costa et al. (1994) quote the $`rms`$ difference between $`B(0)`$ and $`b_\mathrm{J}`$ as $`\mathrm{}<0.2\mathrm{mag}`$. We quote the results for the red LFs of the Las Campanas and Century surveys here for completeness, and do not concern ourselves with the details of the transformations between red and blue passbands. We note however that Geller et al., 1997 find the LF obtained from the CS to be in excellent agreement with a prediction based on colour transfmorations applied to the SSRS2 LF.
Allowing for this adjustment between blue and red magnitudes, the final six rows of Table 1 are all in reasonable agreement, with only the CfA2 result showing a large deviation, particularly in the value of $`\varphi _{}`$. The difference in the LFs deduced from the CfA2 and SAPM surveys is illustrated in Figure 1.
At face value, the CfA2 LF seems to have less intrinsically bright galaxies (at least 10 times less M=-21 galaxies). Of course, this depends on the transformation between Zwicky $`m_Z`$, APM $`b_j`$ or LCRS $`R`$ magnitudes. Efstathiou et al. (1988) adopted the relation $`m_Z=B+0.29`$ for the CfA1 survey (Huchra et al. 1983). This transformation would move the CfA2 LF in the correct direction to reconcile it with other measurements, although Lin et al. (1994) point out that a shift of $`0.7\mathrm{mag}`$ in $`M`$ is required to reconcile $`M_{}`$ with the other surveys. However, a simple shift in the magnitude zero-point would not correct for the apparent discrepancy in the normalisation, which would suggest that a more subtle effect is present in the CfA2 data.
In this paper we present the results of a photometric survey of bright galaxies in the overlap region of CfA2 and the equatorial extension of the APM Galaxy Survey (Maddox et al. 1990a,b; Maddox et al. 1991), which we will use to investigate this apparent discrepancy. In Section 2 we begin by comparing the APM photometry with the Zwicky measurements. We describe our CCD observations in Section 3, and compare our observations with Zwicky’s photometry in Section 4. In Section 5 we give some other properties for the galaxies in our fields and dicuss the implications for the LF. In Section 6 we compare our results with the findings of other authors and discuss the possible implications for other results. No direct CCD calibration has yet been published for this part of the APM Galaxy Survey. We will present a detailed comparison of our CCD photometry with the APM data down to the survey completeness limit in a separate paper.
## 2 A Comparison of Zwicky and APM magnitudes
From the information used to calibrate the APM survey we would expect $`Bb_\mathrm{J}^{\mathrm{APM}}+0.2`$, given the mean $`(BV)0.7`$ and the colour equation $`b_\mathrm{J}=B0.28(BV)`$ (Blair & Gilmore 1982; Maddox et al. 1990b; although this shift could increase by as much as $`0.07\mathrm{mag}`$ according to the findings of Metcalfe, Fong & Shanks, 1995). Adopting the relation between $`m_Z`$ and $`B`$ used by Efstathiou, Ellis & Peterson (1988), this transformation would imply
$$m_Zb_\mathrm{J}+0.5,$$
(2)
which is close to the shift in $`M_{}`$ adopted by Lin et al. (1996), and, if taken as they stand, would suggest that the results of the two surveys might be consistent, at least in $`M_{}`$.
We investigated the usefulness of this relation by considering the subset of $`100`$ galaxies found in the part of the APM Galaxy Survey equatorial extension which overlaps with the Southernmost region of CfA2 (The S+3 sample of Marzke et al., 1994). This sample is drawn entirely from volume V of the Zwicky catalogue. We have used the publically available CfA1 catalogue as our source for Zwicky galaxies, with magnitudes corrected as described in CfA1. The CfA1 includes redshifts only for galaxies with $`m_Z<14.5`$, but has magnitudes and positions for $`m_Z<15.5`$. Marzke et al find this sample to be representative of the whole of the Southern part of CfA2 (their Figure 3). We inspected this sample of galaxies on the film copies of the UKST IIIa-J Sky Survey plates, and examined the image maps as reconstructed from the APM survey data. We removed from our sample all those galaxies which had either been broken up into multiple image components or which were composed of multiple components which had been merged into a single object by the APM software. The distribution of these objects in the $`m_Z,b_\mathrm{J}`$ plane is shown in Figure 2. A simple least squares fit to these data with the slope constrained to be unity gives a zero-point shift of
$$m_Z=b_\mathrm{J}0.6\pm 0.5,$$
(3)
which is more than a magnitude in the opposite sense to that implied by equation 2. As will be noted below a detailed calibration requires a correction for Malmquist bias.
The APM Survey is internally calibrated by matching images in plate overlaps, with the overall zero-point fixed by a number of CCD frames distributed over the survey region (Maddox et al., 1990b). At bright magnitudes ($`b_\mathrm{J}\mathrm{}<17`$) the calibrations are less well determined due to a combination of variations in galaxy surface brightness profiles and the smaller number of bright galaxies found in the plate overlap regions. The equatorial survey extension has been calibrated by matching to the original survey using plate overlaps and the adopted zero-point for the whole survey taken from the original CCD photometry. Maddox et al. (1990b) also used their CCD frames to perform an internal consistency check on the quality of the plate-plate matching technique, and found the residual zero point errors on individual plates to be $`\mathrm{}<0.04\mathrm{mag}`$. The individual uncertainties in galaxy magnitudes in the APM system are therefore expected to be much smaller than effect shown in Figure 2. It is possible that an improvement in the quality of the plate material used for the more recent plates of the equatorial extension coupled with changes in the plate copying process could result in a small change in the saturation correction required for bright galaxies, and that this could produce an effect in this direction (Maddox, private communication). However, account was taken of such effects in the construction of the equatorial extension, and any residual effect is likely to be much smaller than the discrepancy shown here.
The transformation of Zwicky magnitudes to $`B`$ deduced by Efstathiou, Ellis & Peterson (1988) are based on a comparison of the CfA1 survey data with the Durham–AAT redshift survey using 139 galaxies (Peterson et al. 1986). These authors find that the different volumes of the Zwicky catalogue have large variations in the zero-points, although volume V is found to be representative of the calibration as estimated from all volumes. These data are limited to $`m_Z<14.5`$, and the luminosity function parameters inferred for the CfA1 are consistent with the parameters listed for the other surveys in Table 1. It is not possible to draw a direct comparison of the APM data with Zwicky data brighter than $`m_Z14`$, due to heavy saturation of galaxies this bright in the APM data. The most plausible inference from the comparison described above would therefore be a difference in the magnitude scales of the two surveys for $`b_\mathrm{J}\mathrm{}>14`$, with the calibration of Efstathiou, Ellis & Peterson (1988) being appropriate for the Zwicky system at brighter magnitudes.
## 3 CCD Data
### 3.1 Observations
We used the galaxies from the sample discussed in the previous section as the basis for a CCD survey to investigate the above discrepancy by providing an independent calibration for both surveys. This overlap region is essentially defined as $`21^\mathrm{h}50<\alpha <3^\mathrm{h}40`$ and $`0.25^{}<\delta <0.25^{}`$. We obtained images with the 2.5m Isaac Newton Telescope(INT) and 1.0m Jacobus Kapteyn Telescope(JKT) in October 1994. The decision to use two telescopes was motivated by the number of close groups of Zwicky galaxies in the region, so that we use the $`10^{}`$ field of view available at the INT to image cluster fields containing groups of bright galaxies and the $`6^{}`$ field of view of the JKT to image individual galaxies. We used identical Tektronix $`1024\times 1024`$ detectors (TEK3 and TEK4) and Harris $`B`$ and $`R`$ filters on the two telescopes. We obtained observations of 58 fields in two nights at the INT and 73 fields in three nights at the JKT with exposure times of 360s and 600s, respectively. We observed a number of photometric standards from the list of Landolt (1992), at hourly intervals throughout each night, as well as the field of the Active Galaxy AKN120 which contains a number of photometric standards (Hamuy & Maza, 1989).
The data were reduced using standard techniques as implemented in the IRAF ccdred packages, with the exception that we used a modified version of the overscan correction routine to compensate for a saturation of the preamplifier used with TEK4 on the JKT. This effect manifested itself as a sudden drop in the overscan level following readout of particularly bright stellar objects in the field, and subsequent exponential recovery to the normal level as the next $`100`$ rows were read out. We found that this problem could be corrected by fitting the overscan regions on either side of the drop for those fields where the effect occurred.
Extinction coefficients were derived each night, giving values in the range $`0.10k_B0.12`$. We determined zero-points for the two combinations of telescope and detector to be
$$B_{0,INT}=24.16\pm 0.02,$$
(4)
and
$$B_{0,JKT}=21.77\pm 0.02.$$
(5)
We also obtained $`R`$-band images of our standards and target fields, with zero points determined to be $`R_{0,INT}=24.33\pm 0.1`$ and $`R_{0,JKT}=22.11\pm 0.02`$, but we were unable to determine any significant colour term from our standard stars, consistent with previous experience with this combination of CCD and filters which are designed to be very close to the Johnson–Cousins system.
### 3.2 Image Detection and Photometry
We used the STARLINK PISA image detection and photometry software for our photometric analysis. This is essentially the same as the image detection software used in the construction of the APM Galaxy Survey (Irwin, 1986). The basic input parameters for the detection are the threshold intensity per pixel or surface brightness, $`I_s`$, and the minimum size of the object to be detected, $`A_s`$. Amongst other parameters PISA returns total area $`A_T`$, the ellipticities and the isophotal magnitude $`B_I`$ or the corresponding total magnitude $`B`$ resulting from a curve of growth analysis for each detected object after removal of any overlapping objects.
Given that our CCD survey was designed to provide a calibration of both the APM and Zwicky data, we were necessarily interested in a wide range of magnitudes ($`14B20`$) and image sizes. For this range of objects there was no unique combination of $`I_s`$ and $`A_s`$ that could deblend the faint objects without breaking the bright ones. In order to automate this process as much as possible for the whole range of magnitudes, we ran PISA several times with different combinations of $`I_s/A_s`$. These are listed in Table 2.
We chose a larger (smaller) area for the fainter (brighter) isophotes so as to select similar objects in all runs. Objects in the final catalogue were selected from the INT4 and JKT4 runs. The information from different isophotes was then used to perform an automatic rejection of broken or contaminated images. After rejection, the total magnitude and size were the largest remaining estimates of $`B`$ and $`A_T`$, which typically correspond to the faintest isophote left. The error in $`B`$ was taken to be the variance in the different estimates for total magnitudes. Objects with rejected isophotes were automatically labeled and checked visually. We refer to total magnitudes estimated in this way as $`B_{INT}`$ and $`B_{JKT}`$ for the INT and JKT sets, respectively, or $`B_{CCD}`$ in general. These effectively correspond to total magnitudes detected at isophotes $`26.1\mathrm{mag}\mathrm{arcsec}^2`$ and $`25.0\mathrm{mag}\mathrm{arcsec}^2`$, respectively, unless stated otherwise.
To check if our isophotes are low enough we have also computed total magnitudes determined by PISA using higher isophotes. We find a small zero-point shift of the total $`B_{INT}`$ scale as a function of the isophote $`I`$, but for a given $`I`$ this shift is not a function of $`B`$ over the wide magnitude range considered here. This is illustrated in Figure 3 which shows a change of $`0.2\mathrm{mag}`$ in the mean $`B_{INT}(I)`$ when we change the isophote from $`I=26.1\mathrm{mag}\mathrm{arcsec}^2`$ to $`I=24.0\mathrm{mag}\mathrm{arcsec}^2`$. Changing the isophote from $`I=26.1\mathrm{mag}\mathrm{arcsec}^2`$ to $`I=25\mathrm{mag}\mathrm{arcsec}^2`$ introduces a change of only $`0.1\mathrm{mag}`$. Changing the detection isophote from from $`I=26.1\mathrm{mag}\mathrm{arcsec}^2`$ to $`I=24.4\mathrm{mag}\mathrm{arcsec}^2`$ also introduces a change in $`B_{INT}`$ of about $`0.2\mathrm{mag}`$. We therefore conclude that the residuals associated with our final choice of detection isophote are small compared to the shifts illustrated in Figure 2.
This analysis also implies that there should be a mean zero-point shift in the JKT data relative to the INT data of $`0.1\mathrm{mag}`$ due to difference in the isophotes used. We therefore apply this shift to all objects observed with the JKT to transform these data to our $`B_{26.1}`$ system.
Finally, we investigated the possibility that there could be a small effect due to the change in mean redshift of the samples at fainter magnitudes by searching for a change in the mean galaxy colour. Figure 4 shows ($`B_{INT}`$-$`R_{INT}`$) as a function of $`B_{INT}`$. There is only a small colour evolution within the errors. A linear fit to the binned data in Figure 4 yields $`BR=0.016B+1.23`$ with a mean $`BR=1.50`$ (the mean weighted by each galaxies is $`BR=1.61`$ as is dominated by the faint objects).
## 4 Comparison with the Zwicky Catalogue
In Figure 5 we show the number counts histograms for the 204 Zwicky galaxies in our sample to the different magnitude systems: Zwicky (top), APM (middle), and CCD (bottom). Comparison of the APM and CCD data shown here suggests that the effect shown in Figure 2 is unlikely to be an artefact of saturation effects in the equatorial APM Survey data at bright magnitudes (see Section 2).
We searched for possible correlations between the magnitude differences, $`B_{CCD}m_Z`$, and other properties of our sample. The top panel of Figure 6 shows the $`B_{CCD}m_Z`$ difference as a function of CCD colour $`BR`$. The colour of Zwicky galaxies is similar to the mean colour in Figure 4, $`BR1.5`$ and there is no apparent correlation with $`B_{CCD}m_Z`$ within the scatter. This is a useful check, as any large differences that were due to processing errors might be expected to show up as objects of extreme colour. The bottom panel of Figure 6 shows the scatter in $`B_{CCD}m_Z`$ as a function of right ascension. Objects were observed in order of increasing RA on each night of our observing run, and so we would expect temporal drifts to show up as a correlation with RA. Again, there is no evidence of any trend. We also investigated possible variations with Galactic latitude (not shown here), and again found no evidence for trends in our data.
The top panel of Figure 7 shows the $`Bm_Z`$ error as a function of the mean surface brightness for all Zwicky galaxies. Surface brightness is defined here as the ratio of isophotal magnitudes to the area above a threshold of $`25.0\mathrm{mag}\mathrm{arcsec}^2`$ in the INT (triangles) and with a threshold of $`24.1\mathrm{mag}\mathrm{arcsec}^2`$ in the JKT (circles). The difference in the threshold explains the systematic shift in the bulk value of mean high surface brightness of the JKT images (as we are closer to the galaxy core). As can be seen in the Figure, there is no apparent variation of magnitude error with surface brightness.
The bottom panel of Figure 7 shows the $`Bm_Z`$ error as a function of the sequential run number. This is close to observational time for the INT (triangles) or JKT (circles) when considered separately (in practice, some of the INT and JKT observations were made on the same night). There is no apparent variation of magnitude error with run number.
Figure 8 is the main result of this paper. It shows the Zwicky magnitude error $`B_{CCD}m_Z`$ as a function of $`B_{CCD}`$, for all Zwicky galaxies with $`B_{CCD}<17.5`$. For the Zwicky magnitudes we have used a constant error: $`\mathrm{\Delta }m_Z=0.05\mathrm{mag}`$ (as Zwicky quoted magnitudes with a precision of $`0.1\mathrm{mag}`$). For the CCD data we used the error described in Section 3.2. A very similar trend is found for the separate INT (closed circles) and JKT images (open squares).
The dashed line in Figure 8 shows a direct least square fit to the data:
$$B_{CCD}m_Z0.6(B_{CCD}15.1),$$
(6)
which has a scatter of $`0.55\mathrm{mag}`$.
This fit is the result of the interplay between the scatter in the two magnitude systems and the Zwicky survey limit (shown as dotted line in Figure 8). As a result of this scatter, objects with faint $`B_{CCD}`$ and bright $`m_Z`$ can be included in the survey, but there is a deficit of objects with faint $`m_Z`$ and bright $`B_{CCD}`$. i.e. the fit suffers from a Malmquist type of bias. However, for a linear relation, it is apparent that Malmquist bias is insufficient to account for all of the effect shown in Figure 8. We next develop a scheme to correct this fit for the effects of Malmquist bias.
### 4.1 Malmquist bias correction
Different magnitude system are subject to different systematic errors, and even in the best of the situations intrinsic differences in the morphology, environment and spectrum of the galaxies tend to introduce stochastic fluctuations in any magnitude system. We want to find a best fit linear relation between the Zwicky, $`m_Z`$, and CCD, $`B_{CCD}`$, system:
$$m_Z=\lambda B_{CCD}+Z,$$
(7)
where $`\lambda `$ will account for any scale difference and $`Z`$ is a zero point shift. In general, both $`m_Z`$ and $`B_{CCD}`$ are stochastic variables and equation 7 is just a mean relation. As is common practice we will assume that there is Gaussian scatter around the above mean relation (due to the accumulation of multiple uncorrelated factors). That is, given a measured magnitude $`m_Z`$, the error is given by:
$$P(\mathrm{\Delta })=\frac{1}{N}\mathrm{exp}\left[\frac{\mathrm{\Delta }^2}{2\sigma ^2}\right],$$
(8)
where $`\mathrm{\Delta }m_Z\overline{m_Z}`$ is a stochastic variate, $`\overline{m_Z}`$ is some mean best fit value in the linear relation of equation 7, $`\sigma `$ is the $`rms`$ error and $`N`$ is a normalization factor. For a sample that is not magnitude limited we have $`N=\sqrt{2\pi }\sigma `$. For a magnitude limited sample, where $`m_Z<m_Z^{lim}`$, the probability is the same but has a different normalization:
$$N=_{\mathrm{}}^{m_Z^{lim}\overline{m_Z}}𝑑m_Z\mathrm{exp}\left[\frac{\mathrm{\Delta }^2}{2\sigma ^2}\right],$$
(9)
because not all magnitudes are possible, so that $`\mathrm{\Delta }m_Z\overline{m_Z}<m_Z^{lim}\overline{m_Z}`$. Thus, we can write $`P(m_Z)`$ in terms of the complementary error function:
$$P(\mathrm{\Delta })=\frac{2\mathrm{exp}\left[\frac{\mathrm{\Delta }^2}{2\sigma ^2}\right]}{\sqrt{2\pi }\sigma \mathrm{erfc}\left[\frac{\overline{m_Z}m_Z^{lim}}{\sqrt{2}\sigma }\right]}$$
(10)
so that for $`m_Z^{lim}=\mathrm{}`$ we recover the standard Gaussian result.
We are now able to perform a likelihood analysis to find the best fit values of $`\lambda `$ and $`Z`$ in the linear relation of equation 7. We define a likelihood as:
$$=\underset{i}{}P(\mathrm{\Delta }_i)$$
(11)
where $`i`$ runs over all galaxies in the survey, and:
$$\mathrm{\Delta }_i=m_Z^i(\lambda B_{CCD}^i+Z)$$
(12)
with $`m_Z^i`$ and $`B_{CCD}^i`$ the measured Zwicky and CCD magnitudes for galaxy $`i`$. In analogy with the standard $`\chi ^2`$ test we define a Malmquist bias “corrected chi-square”:
$$\chi _{Malm}^2\underset{i}{}\left[\frac{\mathrm{\Delta }_i^2}{2\sigma ^2}+2\mathrm{log}(\mathrm{erfc}\frac{\lambda B_{CCD}^i+Zm_Z^{lim}}{\sqrt{2}\sigma })\right]$$
(13)
Note that the measured magnitude errors, $`\sigma _i`$, are added in quadrature to the stochastic error in the linear relation, $`\sigma `$, which is one of the parameters we want to determine with this fit. The best fit values that we find on maximising the likelihood, using the data in Figure 8, are:
$`\lambda `$ $`=`$ $`0.62\pm 0.05`$ (14)
$`Z`$ $`=`$ $`5.9\pm 0.7`$
$`\sigma `$ $`=`$ $`0.40\pm 0.05,`$
the errors correspond to approximate 99% confidence levels. The values in each uncertainty range are strongly correlated: Lower values of $`\lambda `$ (eg larger scale errors) are (linearly) correlated with larger values of $`Z`$, and both are obtained for the smaller $`\sigma `$. The inverse relation gives:
$$B_{CCD}1.61m_Z9.5.$$
(15)
Figure 8 shows as a continuous line the best fit model for the corresponding best fit magnitude difference:
$$B_{CCD}m_Z=(0.38\pm 0.02)(B_{CCD}15.53)\pm 0.4$$
(16)
Figure 9 shows the residuals of this fit. For comparison we display a Gaussian with same dispersion, $`\sigma =0.4`$. As mentioned above the Gaussian does not provide a good fit because of the Malmquist bias, which produces a deficit of faint objects. It is not possible to show in this Figure a comparison with the Malmquist bias corrected version for the error probability of equation 10, because this depends not only on the differences, $`\mathrm{\Delta }`$, but also on the measured value, $`B_{CCD}^i`$.
Thus, after correcting for the effects of Malmquist bias, the above analysis indicates both a zero-point shift and a change of magnitude scale. The zero point different $`Z_0`$ could be obtained by taking the mean of the magnitude differences over $`m_Z`$ magnitudes (which are the ones that define the survey limit):
$$Z_0<B_{CCD}m_Z>=0.35\pm 0.15.$$
(17)
which is in good agreement with Efstathiou et al. (1988). The scaling relation between the two magnitude systems is then:
$$\lambda \frac{\mathrm{\Delta }m_Z}{\mathrm{\Delta }B_{CCD}}0.62\pm 0.05.$$
(18)
We can also test the above model for the scale error by estimating the magnitude correlations as a function of projected separation. Figure 10 shows the mean correlation $`<\mathrm{\Delta }(\theta _i)\mathrm{\Delta }(\theta _j)>`$ as a function of the pair separation $`|\theta _j\theta _i|`$, in arcmin. The uppermost panel shows (as continuous lines) the autocorrelation for magnitude differences in the Zwicky system, $`\mathrm{\Delta }_i^Z=m_Z^i\overline{m_Z}`$, where $`m_Z^i`$ is the Zwicky magnitude for the $`i`$th galaxy and $`\overline{m_Z}=<m_Z>`$ is the mean Zwicky magnitude in the Survey. The middle panel shows the corresponding autocorrelation for the CCD system: $`\mathrm{\Delta }_i=B_{CCD}^i\overline{B}_{CCD}`$. The lower panel shows the autocorrelation of the magnitude errors: $`ϵ_i=B_{CCD}^im_Z^i`$. The long-dashed in each case show the zero values for reference. The short-dashed line in the lower panel shows the prediction based on applying equation 18 to the data, which is extremely close to the observed result.
We can see in Figure 10 that there is a significant angular correlation between nearby magnitudes and positions with $`\theta \begin{array}{c}<\hfill \\ \hfill \end{array}5^{}`$. This correlation is followed by the errors, indicating that the above model is valid. The Zwicky system shows smaller correlations and smaller variance than the CCD system. This can also be understood in the context of the model, as in the Zwicky system the “effective” magnitude scale is about a factor of two smaller (equation 18).
Note that at the typical depth of the survey, $`𝒟80h^1\mathrm{Mpc}`$, the above magnitude correlations are only significant at physical scales smaller than $`\begin{array}{c}<\hfill \\ \hfill \end{array}100h^1\mathrm{Kpc}`$. This correlation has little effect on typical galaxy clustering scales, $`r_05h^1\mathrm{Mpc}`$, but might be relevant for the inversion of angular clustering on smaller scales. For example, Szapudi & Gaztañaga (1998) find that on scales $`\theta \begin{array}{c}<\hfill \\ \hfill \end{array}0.1^{}`$ there is a significant disagreement between the APM and the EDSGC that is attributed to differences in the construction of the surveys, most likely the dissimilar deblending of crowded fields. At the depth of these Surveys, $`𝒟400h^1\mathrm{Mpc}`$, the above magnitude correlations correspond to $`\theta \begin{array}{c}<\hfill \\ \hfill \end{array}1^{}`$ and could also have a significant effect on the angular clustering and its interpretation on these small scales.
### 4.2 An illustration
An illustration of the scale error in the Zwicky system can be seem in the galaxies shown in Figures 12 and 11. Table 3 gives the Zwicky and CCD magnitudes for each of the Zwicky galaxies as labelled in the Figures. As can be seen from the table, the range of Zwicky galaxies in Figure 11 is $`\mathrm{\Delta }m_Z=0.3`$ which is almost 6 times smaller than the CCD range $`\mathrm{\Delta }B_{CCD}=1.7`$. The range for the whole cluster is $`\mathrm{\Delta }m_Z=1.3`$, almost a factor of 3 smaller that the CCD range: $`\mathrm{\Delta }B_{CCD}=3.5`$. These illustrations are strongly suggestive of an observer-bias effect, whereby some fainter galaxies are included in the same due to their proximity to brighter objects, e.g. object #9 in Figure 12.
## 5 The properties of nearby Field Galaxies
We will now present some further properties of the galaxies in our sample: colours, sizes and ellipticities, and discuss the implications for the Luminosity function. This will help us to understand the systematic effects present in the Zwicky magnitude system. As mentioned in § 1, these local properties are interesting in the context of galaxy evolution and star formation rates.
There are still only few samples that are both homogeneous and large enough to provide estimates of the statistical properties of nearby samples, most of which have been selected from photographic plates. Besides the redshift catalogues listed in Table 1, one of the more extensive catalogues of bright galaxies is contained in the Second Reference Catalogue of Bright Galaxies (de Vaucouleurs, de Vaucouleurs & Corwin 1976, known as RC2), which gives magnitudes, colours, ellipticity and morphology for well over a thousand galaxies to a limiting isophote of around $`25.0\mathrm{mag}\mathrm{arcsec}^2`$. The problem with this catalogue is that it is a mere compilation of data and was not intended to be complete to any specified limiting magnitude, diameter, or redshift. Moreover it is based on photographic plates.
Our sample is homogeneous enough and extends over a large enough area ( $`400`$ square degrees) to provide a fair sample. Thus the new results based on the CCD magnitudes should be a good local reference for the fainter studies of galaxy evolution.
We will compare the properties of galaxies in the Zwicky ($`m_Z<15.5`$) sample with the corresponding properties of galaxies in the INT fields selected using the CCD blue magnitude $`B=B_{CCD}`$. We choose two magnitude cuts: $`B<16`$ which includes most Zwicky galaxies and $`19<B<20`$, which includes faint galaxies in the same fields.
### 5.1 Sizes and Ellipticities
Figure 13 shows an histogram of the frequency distribution of galaxies as a function of its area (number of galaxy pixels above detection threshold in the CCD image). Top panel shows Zwicky selected galaxies only. Middle and bottom panels include all the galaxies in the INT fields selected with CCD magnitudes of $`B<16`$ and $`19<B<20`$, respectively.
As can be seem in the Figure, Zwicky selected galaxies have a long tail of small objects which is not present in the bright subsample of $`B<16`$ CCD selected objects. This again illustrates the scale error (and selection biases) in the Zwicky system mentioned above (see also § 6).
The distribution of sizes in logarithmic scale for both CCD sub-samples are well approximated by Gaussians (shown as continuous lines in the figure). The mean size is about $`60`$ $`\mathrm{arcsec}^2`$ for $`19<B<20`$ and $`1600`$ $`\mathrm{arcsec}^2`$ for $`B>16`$, with a rms dispersion of about $`20\%`$ and $`27\%`$ respectively (recall that these areas correspond to a nominal threshold of $`26.1\mathrm{mag}\mathrm{arcsec}^2`$, fainter thresholds could be needed to sample the size of the lower surphace halos).
Figure 14 shows the corresponding frequency distribution for the ellipticities as measure in the galaxy shapes (with a threshold of $`26.1\mathrm{mag}\mathrm{arcsec}^2`$). These Figures are in rough agreement with the results by Binney & Vaucouleurs (1982) over the Second Reference Catalogue (RC2).
The local distribution seems remarkably similar to the faint one, given the large differences in sizes shown in Fig.13. This is a good indication that our isophote ($`26.1\mathrm{mag}\mathrm{arcsec}^2`$) is low enough, as otherwise we would expect a large excess of round faint objects. There seems to be nevertheless a slight ($`5\%`$) excess of round faint objects. This is probably not due to the seeing (or pixel resolution) as most of the faint objects here have more than 30 $`\mathrm{arcsec}^2`$ (or 100 pixels) of area (see Fig.13). The lack of correlation between ellipticities $`ϵ`$ and areas $`𝒜`$ is illustrated in the bottom panel of Figure 13. A least-square-fit to the points give $`ϵ0.277+0.018\mathrm{log}_{10}(𝒜)`$.
### 5.2 Colours
Both the morphology and a more detailed study of colours will be presented elsewhere. Here we just show the colour frequencies and discuss some of its implications.
Figure 15 shows the frequency distribution of CCD colour ($`BR`$) for all Zwicky galaxies (top) in our sample. We then show the corresponding distributions for the same Zwicky galaxies separated into two sets, corresponding to the JKT (middle) and INT (bottom) CCD frames. As mentioned above, the INT field of view (10’) was larger than the JKT (6’) and so we used the INT to target groups and clusters of galaxies (which could take up several overlaping CCD frames), while the JKT was used for more isolated galaxies. As can be seem in the Figure the INT galaxies are significantly redder, $`BR1.57`$, than the JKT, $`BR1.36`$. There are about $`100`$ objects in each subset which roughly corresponds to a random sampling of the total 600 Zwicky galaxies in our survey area (the 600 targets were split in half between the two fields and then more or less randomly selected during each night). The mean colour of our 200 Zwicky galaxies is $`BR1.47`$. Neither the mean colour or the frequency distribution change significantly when we exclude all the galaxies that are in the same CCD frame (eg to avoid nearby ellipticals). This indicate that selection effects are not important for this distribution.
Given the large area covered (several hundreds of square degree), these results should be a good estimate of the overall mean colour of bright local galaxies. How do these values compare with previous studies? This is a difficult question because it requieres both accurate colours and the accurate fraction of galaxies in different enviroments (eg different morphological types). Previous studies were limited to inhomogeneous samples (eg RC2) or to small numbers of galaxies. For example Kennicutt (1992) presented results for 8 early-type galaxies and 17 spirals to irregulars. Our mean $`BR1.47`$ can be compared againts the synthetic $`BR`$ colours presented by Fukugita et al. (1995). We have used Harris filters which were designed to be very close to the Johnson–Cousins (with $`R`$ Johnson) system. In these broadbands, Table 3 of Fukugita et al. show $`B_R=1.67`$ for Ellipticals, $`1.48`$ for S0, $`1.21.1`$ for Spirals and $`0.6`$ for Irregulars. The range seems in rough agreement with Figure 15, although we find a significant number of objects with $`BR>2`$. Notice that these objects are mostly in the INT sample, that is in clusters or groups. While most of the bluer objects with $`BR<1`$ are in the JKT fields, that is, they are more isolated (by angular separations larger than $`6^{}`$, which corresponds to a mean separation larger than $`140h^1\mathrm{Kpc}`$. Nevertheless, this is seems to be a small effect.
Note that Fukugita et al synthetic colours seem to be slightly less red than the Kennicutt (1992) observations (according to Table 2 in Fukugita et al., they are about 0.1 redder in B-V and this could be larger in B-R). Also notice that we are using total magnitudes, while the synthetic colours are based in small aperture spectra. Galaxies could have quite a different colour distribution (or spectra) in their low-surface halos.
Figure 16 shows the frequency distribution of CCD colour ($`BR`$) for the Zwicky galaxies (top panel) in comparison with the same galaxies selected with CCD magnitudes: with $`B<16`$ (middle panel) and a sub-set of INT galaxies with $`19<B<20`$ (bottom panel), which corresponds to the points in Figure 4. The continuous line in all cases shows for reference a Gaussian distribution with mean $`BR=1.45`$ and rms deviation of $`\sigma =0.4`$, which roughly matches the Zwicky frequencies.
It is interesting to notice the peak of local galaxies with $`BR=1.5`$ and the spread and shift to the red of the faint objects in the bottom panel. These magnitudes are not k-corrected, which could well explain the relative redening of the faint objects.
### 5.3 The local CfA2 Luminosity Function (LF)
Zwicky magnitudes have been used to estimate the LF in the CfA2 redshift catalogue (Marzke et al. 1994). Marzke et al. used a Monte Carlo method to estimate how the LF parameters would change if the dispersion $`\sigma _M`$ were $`0.65\mathrm{mag}`$ (closer to what we find here than the nominal $`0.3\mathrm{mag}`$ they used). They found that the true $`M_{}`$ should be about $`0.6\mathrm{mag}`$ brighter, and that a similar conclusion would be reached for a small scale error. Marzke et al. used this estimate to conclude that a combination of incompleteness and a small ($`0.2\mathrm{mag}`$) scale error would be sufficient to move the CfA2-North values to those found from CfA2-South.
A detailed analysis of the implications of our new Zwicky magnitude calibration on the LF would require the redshift information and this is left for future work. It is nevertheless possible to use the mean relation that we found to show how we expect the LF to change.
In practice, when fitting the Schechter parameters in equation 1 to observational data, the value of $`\varphi _{}`$ is correlated with the values of $`M_{}`$ and $`\alpha `$. In our case we do not use a fit to data, but just model how the LF changes with an homogeneous linear shift in the magnitude scale. A zero point shift $`Z_0`$ in the magnitude scale, as in equation 17, will just change the value of $`M_{}`$:
$$M_{}(corrected)=M_{}+Z_0.$$
(19)
As the LF measures the number density of galaxies per magnitude inverval, a linear change in the magnitude scale will shift the amplitude of the LF proportionally to the shift in the magnitude interval, i.e. by $`\lambda `$ in equation (18). Thus we have:
$$\varphi _{}(corrected)=\lambda \varphi _{}.$$
(20)
Thus the scale and zero point error in the Zwicky system give the following corrections for the CfA2 South results of Marzke et al. (1994)
$$\varphi _{}(corrected)=\lambda \varphi _{}0.0124\pm 0.006h^3\mathrm{Mpc}^3$$
(21)
$$M_{}(corrected)=M_{}+Z_019.3\pm 0.3.$$
(22)
If applied to the LF for the whole CfA2 survey:
$$\varphi _{}(corrected)=\lambda \varphi _{}0.025\pm 0.009h^3\mathrm{Mpc}^3$$
(23)
$$M_{}(corrected)=M_{}+Z_019.1\pm 0.2,$$
(24)
where we have added the errors in quadrature. Thus the corrected values are now closer the SAPM and LCRS (see Table 1). This can also be seem in Figure 17, where we compare the corrected CfA2 luminosity function given above with the one corresponding to the SAPM. Note that the later is in the APM $`b_J`$ band, so that there could be some additional zero-point shifts between them (see §2).
## 6 Conclusion
In this paper we have presented CCD magnitudes for galaxies around $`204`$ Zwicky galaxies which sample over 400 square degrees, extending 6 hours in right ascension. This subsample is drawn entirely from volume V of the Zwicky catalogue. We find evidence for a significant scale error as pointed out by Bothun & Schommer (1982) and Giovannelli & Haynes (1984). This is found by direct likelihood analysis, corrected for Malmquist bias, and also by noticing the angular correlations between the errors (Fig.10) or the long tail of small objects in the frequency distributions of sizes (Fig.13). The mean scale magnitude error is quite large: $`\mathrm{\Delta }m_Z0.62\mathrm{\Delta }B_{CCD}`$, i.e. an error of 0.38 mag per magnitude, while the mean zero point is about $`0.35`$ (equation 17), in agreement with Efstathiou et al. (1988).
An illustration of the scale error in the Zwicky system can be seem in the galaxies shown in Figures 12 and 11. These illustrations are strongly suggestive of an observer-bias effect, whereby some fainter galaxies are included in the sample due to their proximity to brighter objects, e.g. object #9 in Figure 12. This bias can also be seem in statistical terms as a long tail of small objects in the frequency distribution of Fig.13.
Huchra (1976) found only a 0.08 mag per magnitude scale error in a callibration of Zwicky galaxies with a photoelectric photometry of 181 sample of Markarian galaxies, which are preferentially spirals. As spiral galaxies are bad tracers of clusters or groups, it is unlikely that these Markarian galaxies include any of the fainter galaxies that contribute to the proximity observer-bias mentioned above (which are mostly ellipticals). This effect would be hard to notice with photoelectric photometry which samples only one object at a time. Note that our analysis is restricted to a narrow range of magnitudes $`13.5<m_Z<15.5`$, as compared to the wider range in Huchra (1976), but this narrow range contains the majority of the galaxies with $`m_Z<15.5`$ and therefore dominates all the relevant statistical properties (such as the luminosity function).
Bothun & Cornell (1990) have studied the calibration of the Zwicky magnitude scale using a sample of 107 spiral galaxies. They suggest that the errors in $`m_Z`$ are minimized if $`m_Z`$ is an isophotal magnitude at $`B=26.0\mathrm{mag}\mathrm{arcsec}^2`$, although it is clear from their Figure 2 that even within such a small sample of objects there is a $`5\mathrm{mag}`$ range of isophotes which give isophotal magnitudes corresponding to $`m_Z`$. This suggests that our isophotal detection limit should be optimal for this comparison.
Takamiya et al. (1995), also with photoelectric photometry, find evidence for a large scale shift between Volume I and Volumes II and V of the Zwicky catalogue. This shift appears to be of order $`0.5\mathrm{mag}\mathrm{mag}^1`$ over the range $`14<m_Z<15.7`$, which is comparable to our findings for Volume V, although they find very little effect for Volumes II and V. However, we note that Figure 4 of Takamiya et al. shows $`Bm_Z`$ vs. $`B`$, rather than $`m_Z`$. A similar representation of Figure 8 of this paper looks very similar to Figure 4a of Takamiya et al., which is reasonable, given that Figure 8 implies a strong compression of the $`m_Z`$ axis, which effectively hides the scale error. Unfortunately, the overlap between the 204 objects in our sample and the 155 objects in the Takamiya et al. data is too small to draw any detailed comparison, given that there are $`600`$ Zwicky galaxies within the region of our survey.
We have also estimated how this scale error could change the CfA2 luminosity function in §5.3. Figure 17 shows how the corrected estimation is now closer to other local estimates, such as the SAPM. Finally, in §5 we give properties of the galaxies in our sample: colours, sizes and ellipticities, providing one of the largest samples of this kind. The local colour frequency distribution can be well approximated by a Gaussian distribution with mean $`BR=1.45`$ and rms deviation of $`\sigma =0.40`$. These colours compare well with synthetic $`BR`$ colours presented Fukugita et al (1995). But there is a significant number of galaxies with $`BR>2`$ which are preferably found in clusters and groups, while most of the bluer objects with $`BR<1`$ (spirals to irregulars) seem more isolated. These local properties are interesting in the context of galaxy evolution and star formation rates. This will be studied in more detail in future work.
Acknowledgements
The Isaac Newton Telescope (INT) and Jacobus Kapteyn Telescope(JKT) are operated on the island of La Palma by the Isaac Newton Group in the Spanish Observatorio del Roque de los Muchachos of the Instituto de Astrofisica de Canarias. We thank Steve Maddox, Will Sutherland, John Loveday, Cedric Lacey, Michael Vogeley, and Gary Wegner for useful discussions. We also thank the referee for helpful and constructive comments. This work has been supported in part by CSIC, DGICYT (Spain), projects PB93-0035 and PB96-0925, in part by PPARC (UK), and by a bilateral collaboration (Accion Integrada HB1996-0091) between CSIC (Spain) and the British Council (UK). Part of the data reduction and analysis was carried out at the University of Oxford, using facilities provided by the UK Starlink project, funded by PPARC.
## 7 References
Blair, M., & Gilmore, G., 1982, PASP, 94, 7423
Binney, J., & de Vaucouleurs, G., 1981, MNRAS, 194, 679
Bothun, G.D., & Cornell, M.E., 1990, AJ, 99, 1004
Bothun, G.D., & Schommer, R., 1982, ApJ, 255, L23
Da Costa, L.N., Geller, M.J., Pellegrini, P.S., Latham, D.W., Fairall, A.P., Marzke, R.O., Willmer, C.N.A., Huchra, J.P., Calderon, J.H., Ramella, M., Kurtz, M.J., 1994, ApJ, 424, L1
de Vaucouleurs, G., de Vaucouleurs, A., Corwin Jr., H.G., 1976, Second Reference Catalogue of Bright Galaxies, University of Texas Press, Austin.
Efstathiou, G., Ellis, R.S., & Peterson, B.J., 1988, MNRAS,232, 431
Folkes, S., Ronen, S., Price, I., Lahav, O., Colless, M.M., Maddox, S.J., Deeley, K., Glazebrook, K., Bland-Hawthorn, J., Cannon, R.D., Cole, S.M., Collins, C.A., Couch, W.J., Driver, S.P., Dalton, G.B., Efstathiou, G., Ellis, R.S., Frenk, C.S., Kaiser, N., Lewis, I.J., Lumsden, S.L., Peacock, J.A., Peterson, B.A., Sutherland, W.J., Taylor, K., 1999, MNRAS, 208, 459
Fukugita, M., Shimasaku, K., Ichikawa, 1995, T., PASP, 107, 945
Geller, M.J., Kurtz, M.J., Wegner, G., Thorstensen, J.R., Fabricant, D.G., Marzke, R.O., Huchra, J.P., Schild, R.E., Falco, E.E., 1997, AJ, 114, 2205
Giovanelli, R., & Haynes, M.P., 1984, AJ, 89, 1
Hamuy, M., & Maza, J., 1989, AJ, 97, 720
Huchra, J.P. 1976, A.J,81, 952
Huchra, J.P., Davis, M., Latham, D. & Tonry, J. 1983, ApJS, 52, 89
Kennicutt, R.C., 1992, ApJS, 79, 255
Landolt, A.U., 1992, AJ, 104, 340
Lin, H., Kirshner, R.P., Shectman. S.A., Landy, S.D., Oemler, A., Tucker, D.L., Schechter, P.L., ApJ,464, 60
Loveday, J., Peterson, B.A., Efstathiou, G., & Maddox, S.J., 1992, ApJ,390, 338
Maddox, S.J., Sutherland, W.J., Efstathiou, G., & Loveday, J., 1990a, MNRAS,243, 692
Maddox, S.J., Sutherland, W.J., Efstathiou, G., & Loveday, J., 1990b, MNRAS,246, 433
Maddox, S.J., Sutherland, W.J., Efstathiou, G., & Loveday, J., 1991, in “The Early Observable Universe from Diffuse Backgrounds”, eds. Rocca-Volmerange, B., Deharveng, J.M., & Trân Thanh Vân, J.
Marzke, R.O., Huchra, J.P., & Geller, M.J., 1994, ApJ,428, 43
Metcalfe, N., Fong, R., & Shanks, T., 1995, MNRAS, 274, 769
Peterson, B.A., Ellis, R.S., Efstathiou, G., Shanks, T., Bean, A.J., Fong, R., & Zen-Long, Z., 1986, MNRAS, 221, 233
Schechter, P., 1976, ApJ,203, 297
Szapudi, I., Gaztañaga, E., 1998, MNRAS 300, 493
Takamiya, M., Kron, R.G., & Kron, G.E., 1995, AJ, 110, 1083 A&A,342, 15.
Zucca, E., Zamorani, G., Vettolani, G., Cappi, A., Merighi, R., Mignoli, M., Stirpe, G.M., MacGillivray, H., Collins, C., Balkowski, C., Cayatte, V., Maurogordato, S., Proust, D., Chincarini, G., Guzzo, L., Maccagni, D., Scaramella, R., Blanchard, A., Ramella, M., 1997, A&A, 326, 477
Zwicky, F., Herzog, E., Wild, P., Karpowicz, M., & Kowal, C.T., 1968, Catalogue of Galaxies and Clusters of Galaxies.
|
no-problem/9906/cond-mat9906079.html
|
ar5iv
|
text
|
# Evidence of structural instability near 𝑇^∗∼250 K in the resistivity of Bi2Sr2CaCu2O8 whiskers
## Abstract
We report anomalous features in the normal-state resistivity of single crystalline Bi<sub>2</sub>Sr<sub>2</sub>CaCu<sub>2</sub>O<sub>8</sub> whiskers near $`T^{}250`$ K. From these varied oxygen-doping samples, the resistance ($`R`$) was found to deviate upward from the linear-$`T`$ relation for $`T<T^{}`$, while $`R`$ keeps decreasing with $`T`$. The deviation, $`\mathrm{\Delta }RR(a+bT)`$, reaches a maximum in a temperature range of $`25`$ K, just before $`RT`$ follows the well-known descent associated with the pseudogap opening in underdoped samples. A second kink was also found in some samples immediately following the first one, resulting in an S-shaped feature in the $`RT`$ curve. The resistance then resumes nearly linear-$`T`$ decrease below the anomalies in contrast to the first case. We interpret these features as being related to crystal structure transformation (or lattice distortion), as was evidenced in thermal and elastic property measurements. These structural instabilities seem to be connected to the subsequent dynamics as $`T`$ is further lowered.
74.25.Fy, 74.72.Hs, 75.30.Kz
The normal state of high-$`T_c`$ cuprates provides rich information for an understanding of superconducting mechanism. Two of the many interesting normal-state properties are the so-called pseudogap opening and crystal structural transformation. In resistivity ($`\rho `$) measurement, the former was believed to manifest itself as a downturn deviation from the linear $`\rho T`$ relation below some characteristic temperature $`T^{}>T_c`$ in underdoped samples. Evidence for crystal structure transformation (or lattice distortion) is more difficult to detect because of such factors as crystal quality and the weakness of its physical effects. Structural transformations are generally studied by elastic and acoustic methods . In most of the cuprates, anomalous features corresponding to structural instabilities in the normal state have been observed. In La<sub>2-x</sub>Sr<sub>x</sub>CuO<sub>4</sub>, a tetragonal to orthorhombic structure transition upon cooling occurred at 220 K . Around the same temperature for YBa<sub>2</sub>Cu<sub>3</sub>O<sub>7-δ</sub>, the sound velocity shows a step-like change and was associated with the instability corresponding to a structural order-disorder transition . A step-like change in Young’s modulus (hysteresis) also appeared in elastic measurements of single crystals at 200 - 240 K . In Bi cuprates, similar evidence was also accumulated. Dominec et al. reported sound velocity deviations around 240 K in Bi<sub>2</sub>Sr<sub>2</sub>CuO<sub>6</sub> (Bi-2201). Acoustic studies revealed sound velocity softening along the $`a`$\- axis around 250 K in Bi<sub>2</sub>Sr<sub>2</sub>CaCu<sub>2</sub>O<sub>8</sub> (Bi-2212) samples . In elastic property measurements of similar Bi-2212 whiskers as studied in the present work, Tritt et al. found anomalous peaking of Young’s modulus and stress-strain hysteresis near 270 K, and interpreted it as evidence for a phase transition. Similar results were also obtained for Bi<sub>2</sub>Sr<sub>2</sub>Ca<sub>2</sub>Cu<sub>3</sub>O<sub>10</sub> (Bi-2223) . More directly, by using acoustic and x-ray diffraction methods, Titova et al. consistently found a small increase of lattice dimensions at $`T190270`$K for the cuprates: YBCO (123 and 124), Bi-2201, 2212, and 2223, and Ti-series compounds. The structural instabilities of the cuprates observed at comparable temperature ranges (about 200 to 300 K) demonstrate common features consistent with their similar crystalline structures. Obviously, structural transformation affects transport properties of the material and should be detectable by resistivity measurement. Indeed, resistivity jumps were observed above $`T_c`$ in La<sub>2-x</sub>Sr<sub>x</sub>CuO<sub>4</sub>, which are caused by a structural phase transition . However, few results were reported for other cuprates in the temperature regions where lattice instabilities were found by mechanical methods described above. This situation can be understood in two ways. First, the influence of structural instabilities on sample resistivity is small, because at such structural transformation, the crystal symmetry usually remains the same, only the lattice parameters change (distortion), leaving weak traces on electron scattering. Secondly, the small change in resistivity can be easily covered up by effects such as percolation, and is undetectable unless the crystal is of very high quality. Therefore, in the present work on pure single crystalline Bi-2212, it is interesting that we observed unusual features in the normal-state resistivity in temperature regions where x-ray and mechanical studies had revealed anomalous structural properties. This is a clear indication for a correlation between these features probed by different methods. Moreover, drastically different temperature dependence of resistivity appeared below these anomalies, showing crucial connections between the phenomena in these temperature regimes.
In our extensive resistivity measurements of both Bi-2212 and Bi-2223 whiskers, we often noticed the data were fluctuating in a temperature region between 180 and 270 K. Outside this window, the signal is much more stable. Since this temperature range corresponds to the important regime where the resistivity shows dramatic changes and is central in the normal-state theory, we were motivated to check the region more closely. In this work we report five samples selected from many others which exhibit anomalous behavior in this temperature window.
Samples of Bi-2212 whiskers were grown by a sintering method . They are rectangular needle-like tapes with typical dimensions: 3 mm $`\times `$ 10 $`\mu `$m $`\times `$ 0.5 $`\mu `$m. The three crystalline axes ($`a`$, $`b`$ and $`c`$) are along these directions, respectively. Various characterization by scanning and transmission electron microscopy, and energy dispersive x-ray analysis indicate that the whiskers are good single crystals with perfect surface morphology throughout the whole length ($``$3 mm). The as-grown whiskers are usually oxygen overdoped. Before making the first measurements, some samples were annealed in flowing nitrogen, making them slightly underdoped. The resistivity was measured by the standard four-wire method. The electrical leads were made by magnetron sputtering of silver through a mask. A dc current (0.5 $`\mu `$ A) was driven along the $`a`$-axis and voltage was also measured in this direction.
Figure 1a shows the resistance ($`R`$)data of one whisker (sample A, as-grown), together with a segment of $`R`$ above 200 K. At high temperature, $`RT`$ is well fitted to a linear line (fitting range: 260 - 308 K) with $`R(\mathrm{\Omega })=129.62+2.475T`$. As the temperature is further lowered, $`R`$ shows a clear upward curving just before it follows the well-known downward deviation. To examine the curve in more detail, we plotted the temperature dependence of the resistance deviation $`\mathrm{\Delta }RR(a+bT)`$, where $`(a+bT)`$ is the high temperature fit as described. The sharp turning point and the slope change are more evident in the plot of $`\mathrm{\Delta }R`$ as shown in Fig. 1b. We can see that, while $`R`$ keeps decreasing with $`T`$, $`\mathrm{\Delta }R`$ starts to deviate upward at $`T_0260`$ K and reaches a maximum at $`T^{}=236`$ K, it then falls and becomes negative, i.e. $`R<a+bT`$. The peak $`\mathrm{\Delta }R/R`$(300 K) is about 0.5%. Although the choice of fitting range of temperature may slightly affect the kink, the two major features that $`R`$ curves upward from linear-$`T`$ dependence and reaches a sharp turning point are clearly identifiable. Such small bumps could be easily smeared out in samples with defects such as grain boundaries, where percolation is not negligible, to yield a fully linear-$`T`$ resistance. In our twined samples, which have a two-step superconducting transition, linear-$`T`$ dependence was observed without any feature in the normal state. This agrees with the studies by Gorlova and Timofeev who made combined electron diffraction and resistivity measurement on the same sample, and showed that a wider portion of linear-$`T`$ resistivity appears in twinned whiskers or samples covered with a polycrystalline film than in purer samples. This explains the often-observed discrepancy in bulk materials between transport and other measurements in which the sample’s polycrystalline nature is not important. The smallness in the change of resistivity puts high requirement on sample quality to observe these anomalies. The features were identified only after more significant kinks around this temperature region were found in our Bi-2223 single crystal samples . The kink in $`\mathrm{\Delta }R`$ seems to follow a power-law relation $`(1T/T_0)^2`$ (dotted line in Fig. 1b, where $`T_0=260`$ K). However, it could also fit to a weak exponential increase of the form: $`R_0=\mathrm{exp}[c(T_0T)]`$, with $`c=0.1`$ to 0.2.
In Fig.2 we plotted $`\mathrm{\Delta }R/R`$(300) vs. $`T`$ for four more samples plus sample A from Fig. 1b. The annealing status of these samples are as follows: A and B: as-grown in ambient pressure; C: annealed in flowing N<sub>2</sub> ($`450^{}`$C, 8 h); D: flowing N<sub>2</sub> annealing ($`450^{}`$C, 4 h); and E: flowing O<sub>2</sub> annealing ($`450^{}`$C, 10 h). The doping level may not solely depend upon the annealing conditions because the sample size varies. Nevertheless, data from long-time annealed samples are reliable. Two cases in the temperature dependence of $`\mathrm{\Delta }R`$ can be classified, as shown in the separate panels of Fig. 2. In the lower panel, $`\mathrm{\Delta }R`$ reveals one kink as discussed. $`\mathrm{\Delta }R`$ falls steeply right after the sharp peak, which gives rise to the well-known downward turning $`RT`$ in underdoped cuprates. Our identification of this transition temperature is much clearer than the data available in the literature . The curves in the top panel show another kink following the first one (samples B and C), which resulted in an S-shaped feature in the $`RT`$ curve. At the end of this curving, $`R`$ nearly resumes the previous linear-$`T`$ descent. The origin of this anomaly is unclear. It does not show up in the underdoped cases. For the sample annealed in oxygen (E), the kink around 250 K is absent, but the kink at lower temperature ($``$ 220 K) is consistent with the second kinks in the other two samples. A very similar single kink was also found in the as-grown data of sample C (refer to the inset of Fig. 3), which is also oxygen overdoped. These features in the normal state thus seem to be intrinsic to the Bi-2212 samples. This result joins our previous results from Bi-2223 samples to give a consistent picture about the anomaly around this characteristic temperature region in Bi-series cuprates. The peaking temperature $`T^{}`$ shows a trend to increase with the extent of underdoping. $`T^{}=254`$ K is the highest for sample C which experienced the longest N<sub>2</sub> annealing, whereas the oxygen-annealed sample E showed the lowest $`T^{}`$ (220 K). The temperature $`T_0`$, at which $`\mathrm{\Delta }R`$ starts to climb up, also shows the similar dependence. The detailed characteristic temperatures for the samples are listed in Table I. In light of the thermal and elastic measurement results in the literature, which showed evidence of structural phase transformation around the same temperature region as is reported here, we interpret such upward deviation as due to structural transformation or lattice distortion. Similar kinks in resistivity had been observed in La-214 due to structural transformation . The signal is often fluctuating immediately following the kink, in the region around 200 K as shown by the oscillating $`\mathrm{\Delta }R`$. This is obviously correlated to the $`\mathrm{\Delta }R`$ jump, and may well mean an influence on electron scattering due to structural instabilities.
The peak temperature $`T^{}`$ clearly separates two regions in which different mechanisms are switched on. Especially for the noisy region below $`T^{}`$, if it is caused by the opening of a pseudogap in the spin excitation, the initial fluctuating $`R`$ data may be a clue for this mechanism, when the antiferromagnetic correlation is gradually frozen out. In the scenario of stripe phase formation, on the other hand, the stripe phase may be triggered by the lattice distortion. Charge or spin excitation following a structure transformation (or lattice distortion) was established by neutron scattering studies . Bianconi et al. have shown for Bi-2212 that 1D modulation of the CuO<sub>2</sub> plane is formed by alternating stripes of low temperature orthorhombic and low temperature tetragonal lattices. However, our $`T^{}(3T_c)`$ is higher than their data ($`1.4T_c`$). Whatever the model will be, the present result shows that the lattice structure undergoes some change in all the samples, which initiates new resistive mechanisms. However, the subsequent dynamics depends on the oxygen doping level, not on this structural change.
The upward deviation is not due to thermal stress in the sample, which is anchored onto the glass substrate by silver leads. We estimate the thermal stress to be a small value $``$ 0.018 GPa at 250 K, by taking the Young’s modulus $`Y=20`$ GPa , the thermal expansion coefficient along the $`a`$-axis $`\alpha =14.4\times 10^6`$ K<sup>-1</sup> , and a length between voltage leads of about 0.3 mm at room temperature. Chen et al. had measured the effect of uniaxial stress (strain) on the resistivity of similar whiskers, and found that the influence is very small ($`1/12000`$ at room temperature). Moreover, the pressure generally causes a decrease in resistivity.
To further study the influence of doping on $`T^{}`$, we annealed one sample (C) successively in flowing nitrogen. The resistance data from this sample are as shown in Fig. 3. The result from the as-grown sample shows a slope change near 200 K, $`RT`$ is linear above this temperature. After one annealing, a new feature appeared around $`T^{}220`$ K, while the slope change around 200 K remains. After the second annealing, $`T^{}`$ shifted to 245 K, as indicated by the arrows, showing the tendency that $`T^{}`$ increases with further underdoping, which agrees with the widely observed behavior of resistivity anomaly and the pseudogap.
All the above discussion is based upon crystal structure instability in the normal state. In light of the magnetic structure of the CuO<sub>2</sub> planes, an alternative interpretation may come from the competition between quasiparticle scattering and the increasing correlation with antiferromagnetic (AF) ordering as the temperature is decreasing. It is interesting to compare the present results with the anomalous resistivity of magnetic materials . In fact, resistivity kinks were commonly found near magnetic phase transitions in the latter. For example, in dysprosium single crystals , a kink in the linearly decreasing $`\rho T`$ appears near the Néel point when the sample transforms from the paramagnetic into the AF state. Similar anomalies were also reported in europium chalcogenides . It is surprising that the Néel temperature of perovskite KNiF<sub>3</sub> is about 253 K , very close to the anomalous temperatures in the present work. This may point to a common origin of the resistivity kinks for the materials with similar crystal structures.
In the scenario of free electron scattering with spin excitations, the total resistivity consists of three additive terms: the $`T`$-independent residual resistivity $`\rho _0`$ due to elastic impurity scattering, the lattice contribution (phonon) $`\rho _LT`$, and the spin-disorder resistivity $`\rho _M`$. The resistivity kink (exponentially increasing with decreasing $`T`$ ) is associated with the spin-disorder resistivity ($`\rho _M`$). In the critical region of magnetic transition, the behavior of $`\rho _M`$ reflects the strongly temperature dependent short-range spin-spin correlation. This is in the regime of spin-disorder due to the large long-range magnetic fluctuations near the transition temperature . The resistivity thus takes the form:
$$\rho (T)=\rho _0b|t1|\mathrm{ln}|t1|^1,$$
(1)
where $`t=T/T_c`$. For $`b>0`$, $`\rho (T)`$ shows a concave cusp at the magnetic transition temperature $`T_c`$. If this idea is valid, then the observed anomaly near $`T^{}250`$ K must come from the antiferromagnetic ordering in the CuO<sub>2</sub> planes. Especially, the sharp turning point in all the curves may well signify a similar magnetic phase transition. Currently only the spin fluctuation theory is a close approach to this picture.
In conclusion, an upward deviation from the linear-$`T`$ dependence were observed around $`T^{}250`$ K in the normal-state resistivity of Bi-2212 single crystals. This is usually followed by a fluctuating region near 200 K. We interpret such behavior as due to structural instabilities in accordance with many results from other measurement techniques (thermodynamic, elastic, acoustic, and x-ray). A peak in the resistance deviation ($`\mathrm{\Delta }R`$) is clearly identifiable. Below this sharp turning point, the resistivity drops more steeply, signifying a new mechanism in underdoped samples. This turning in resistivity had been widely identified as pseudogap opening. However, the scenario of stripe formation preceded with lattice instabilities was also proposed . The result shows that the lattice instability exists in all the samples regardless of their doping level. The same result was also found in our Bi-2223 samples. This gives a consistent picture intrinsic to similar crystal structures of Bi-cuprates. Although the structural transformation seems to trigger new mechanism, it is the oxygen doping that controls the subsequent behavior of $`RT`$. It is noticeable that our $`T^{}`$ data are consistently higher than the literature values (typically 120 -165 K, with the highest for Bi-2212 being 220 K ). There is no identifiable feature (smooth $`dR/dT`$) in our $`RT`$ curves below 180 K. An alternative interpretation of the resistance kinks is the correlation between electrons and AF spin fluctuations. Similar mechanisms in magnetic materials are known to give rise to the kinks in the resistivity curves at magnetic phase transition. Our work should stimulate theoretical interest in this direction.
|
no-problem/9906/astro-ph9906222.html
|
ar5iv
|
text
|
# Dark Bulge, Exponential Disk, and Massive Halo in the Large Magellanic Cloud
## 1 Introduction
Rotation curves of dwarf galaxies are known to increase monotonically toward their outer edge, and the mass-to-luminosity ratio, and therefore, the dark matter fraction is larger than that in usual spiral galaxies (Persic et al 1996). However, little is known of their internal distribution of dynamical mass, because of the insufficient spatial resolution due to their small angular extents as well as for the low surface brightness in HI and H$`\alpha `$ lines.
HI observations of the Large Magellanic Cloud (LMC), the nearest dwarf galaxy, have provided us with a unique opportunity to study detailed kinematics (e.g., Luks and Rohlfs 1992; Putman et al 1998). Recently, Kim et al (1998) have obtained high-resolution HI kinematics from interferometer observations. They have shown that the rotation characteristics is nearly axisymmetric, and this fact allows us to derive a reliable rotation curve by using a position-velocity diagram along the major axis.
In the present paper, we derive a new rotation curve for the LMC, and discuss the distribution of dynamical mass and dark matter.
## 2 Position-Velocity Diagrams and Rotation Curve
Fig. 1 shows the HI velocity field as reproduced from Kim et al (1998), superposed on a smoothed DSS (Digitized Sky Survey) image in B band. The cross indicates the kinematical center defined by Kim et al.. We have, here, determined a position, RA=5h 13.8m, Dec=$`68^{}38^{}`$ (J2000), as indicated by the asterisk, at which the velocity gradient attains the maximum, and around which the velocity field becomes most symmetric. From the symmetric velocity field, we may assume that the gas is circularly rotating around this kinematical center in the central 2 kpc radius region. The interferometer observations may miss the intensities from extended gas disk. However, since the extended gas and the gaseous features detected with the interferometer are thought to be rotating at the same speed, we assume that the observed velocity field represents that of the total gas.
In Fig. 2 we reproduce the PV diagram across a position close to the adopted kinematical center from Kim et al (1998). The thin line represents a rotation curve fitted by Kim et al. by using an AIPS utility, which adopts a simple function. However, the innermost part of the PV diagram shows a much steeper variation from positive to negative velocities, which appears to be not well traced. In order to obtain a more precise rotation velocities in the innermost region, we apply the envelope-tracing method of maximum-velocities (Sofue 1996), which we have done by eye-estimates on the PV diagram. The newly determined rotation velocities are indicated by the thick curve, by which the steep velocity variation is now well traced. Besides the bright features traced by this curve, a higher-velocity feature is found at -60 to -70 km s<sup>-1</sup> and $`+20^{}`$. If we trace this feature, the velocity variation around the center becomes steeper, and the rotation curve will have much steeper rise at the center.
Using the thick curve in Fig. 2 and correcting for the inclination angle of 33, we obtain a rotation curve as shown in Fig. 3. The outer rotation curves beyond the HI disk up to 8 kpc has been taken from Kunkel et al (1997). In the following discussion, we adopt the thick curve in Fig. 3, which is drawn by smoothing the observed curve (thin line). The thus-obtained rotation curve is found to be similar to those of usual disk galaxies (Rubin et al 1982; Sofue 1996, 1997; Sofue et al 1998, 1999) except for the absolute values. The rotation curve of the LMC is characterized by
(1) a steep central rise,
(2) a flat part in the disk, and
(3) a gradual rise toward the edge.
## 3 Surface Mass Distribution: Bulge, Disk, and Massive Halo
Given a rotation curve from the center to the outer edge with a sufficient resolution, we can directly calculate the surface mass density (SMD)(Takamiya and Sofue 1999). This method is not intervened by any models such as the Plummer potentials and exponential disks. An extreme case is to assume a spherical symmetry: the rotation velocity is used to calculate the total mass involved within a radius, which is then used to calculate the SMD. Another extreme case is to assume a thin rotating disk: the SMD can be directly calculated by using the Poisson equation (Binney and Tremaine 1987). We assume that the true mass distribution will lie in between these two cases, which are indeed found to coincide within a factor of 1.5. Fig. 4 shows the calculated SMD for the disk (full line) and spherical cases (dashed line). The full line will better represent a disk component as the HI appearance indicates, while the spherical case would be better for an inner bulge. The obtained SMD distribution can be summarized as the following.
(1) Dark bulge
The distribution of SMD shows a compact, high-density component tightly concentrated around the kinematical center. This component is not associated with an optical counterpart, and we call it a ”dark bulge”. The $`e`$-folding scale radius is about 120 pc, when fitted with an exponential function. The mass involved within 240 pc radius, at which the SMD becomes equal to that of the disk component, is $`1.7\times 10^8M_{}`$. Except for its small radius, the dark bulge is as dense as a central bulge of normal disk galaxies: the SMD is estimated to be $`1.0\times 10^3M_{}\mathrm{pc}^2`$ at its scale radius, which is comparable to or only by a factor of ten smaller than the values of usual galactic bulges ($`10^{34}M_{}\mathrm{pc}^2`$) at their scale radii (a few hundred pc) (Takamiya and Sofue 1999).
(2) Exponential disk
The SMD, then, is followed by an exponentially decreasing part with an $`e`$-folding scale radius of 1.0 kpc. This ”exponential disk” in mass is dominant up to a radius of 2 kpc, and extends to 3 to 4 kpc radius, being gradually replaced by an extended halo component. The total mass involved within 2 kpc radius is $`2\times 10^9M_{}`$, consistent with the value obtained by Kim et al (1998). The SMD of $`200M_{}\mathrm{pc}^2`$ at its scale radius is comparable to the values of usual disk galaxies ($`300500M_{}\mathrm{pc}^2`$) at their scale radii ($`35`$ kpc).
The total $`B`$ magnitude of LMC is 0.6 mag. within an angular extent of $`650^{}\times 550^{}`$, or within $`4.3`$ kpc radius (de Vaucouleurs et al 1991). For a distance modulus of 18.50 (50 kpc) and $`BV`$ color of 0.55, we obtain a $`V`$ magnitude $`18.45`$. This leads to a $`V`$ band luminosity of $`2\times 10^9L_{}`$. Since the mass involved within 4.3 kpc is $`4.5\times 10^9M_{}`$, we obtain a mean M/L ratio of about 2.2 for the disk component.
(3) Massive halo
The SMD in the outer halo obeys an exponentially decreasing function with an $`e`$-folding scale radius of 5.2 kpc. The total mass involved within 8 kpc radius is $`1\times 10^{10}M_{}`$. Since the outer envelope is not well visible in optical photographs (DSS), this may suggest again a high M/L ratio in the halo. This is consistent with the high M/L ratio in many other dwarf galaxies, for which gradually-rising rotation curves are generally observed (Persic et al 1996). Kunkel et al (1997) have suggested that the high velocities in the outer region could be due to a tidal effect with the Small Magellanic Cloud. Alternatively, tidal warping could also cause apparently higher (or lower) velocities, because the galaxy is nearly face on. In these cases, the presently estimated mass would be significantly overestimated.
## 4 Discussion
The dark bulge is significantly displaced from the stellar bar. No optical counterpart is visible, despite that the extinction is supposed to be not large for the nearly face-on orientation. Since the surface luminosity appears to be not significantly enhanced from the average in the disk part, where M/L ratio is $`2`$, the bulge’s M/L would be as large as $`2050M_{}/L_{}`$. Such a high M/L indicates a significant excess of the dark matter fraction in the bulge. As discussed in Section 2, if the faint and higher-velocity features in the PV diagrams are taken into account, the rotation curve will have a much steeper rise in the center than that traced by the thick line in Fig. 2. If this is the case, the bulge’s mass density will be still higher, and so is the dark mass fraction.
The center of the stellar bar is about 1.2 kpc away from the center of the dark bulge (Fig. 1). No significant mass enhancement, which might be related to the bar, is detected at radii 1 to 2 kpc in Fig. 4. In fact, neither a stream motion in the velocity field, nor an anomaly in the PV diagrams are observed along the stellar bar (Kim et al 1998). We may argue that the M/L ratio in the stellar bar is significantly lower than that in the surrounding regions.
We finally comment on an alternative possibility to explain the displacement of the optical and kinematical centers. Ram-pressure stripping of the gas disk either due to an intergalactic wind (Sofue 1994) and/or a gaseous inflow from the Small Magellanic Cloud (Gardiner et al 1994; Putman et al 1998) could cause such a displacement. In order for the HI velocity field being kept unperturbed, it must occur as quickly as in a crossing time of the innermost disk, or within $`tr/v_{\mathrm{rot}}2\times 10^6`$ years, where $`r120`$ pc is the radius of the innermost HI disk and $`v_{\mathrm{rot}}55`$ km s<sup>-1</sup>is the rotating velocity. This requires a wind velocity greater than $`v_{\mathrm{ram}}d/t500`$ km s<sup>-1</sup>, where $`d1`$ kpc is the displacement. Such a high velocity would be possible for an intergalactic wind. However, the ram-stripping condition, $`\rho _{\mathrm{ram}}v_{\mathrm{ram}}^2>\rho _{\mathrm{HI}}v_{\mathrm{rot}}^2`$, seems to be not satisfied: for $`\rho _{\mathrm{ram}}10^4m_\mathrm{H}`$ cm<sup>-3</sup>, $`v_{\mathrm{ram}}500`$ km s<sup>-1</sup>, $`\rho _{\mathrm{HI}}0.1m_\mathrm{H}`$ cm<sup>-3</sup>, and $`v_{\mathrm{rot}}55`$ km s<sup>-1</sup>, we have $`0.5\times 10^{12}`$ and $`5\times 10^{12}`$ dyne cm<sup>-2</sup> for the first and second terms, respectively. Therefore, the ram-stripping appears not a likely origin. In any other circumstances, where no gravitational attraction is present, the inner rotation feature will disappear within about one million years. So, in so far as the velocity field is assumed to represent rotation, a high-density mass concentration at the kinematical center, and therefore, the presence of a dark bulge, will be an inevitable consequence.
Another possible idea is to attribute the velocity gradients in the kinematical center to local velocity anomalies such as due to wiggles by HI shells and/or turbulence. However, Kim et al’s modulus map of velocity residuals indicates that density waves and warping are more dominant to cause the residuals. Neither the largest HI shells nor 30 Dor are associated with the largest velocity residuals and wiggles. Moreover, there appear no particular HI shells or associated starforming regions around the kinematical center where the velocity variation is steepest.
Hence, we here rely more on our assumption that the HI kinematics of the LMC directly manifests the dynamics of the galaxy. We also comment that, if we adopt the smooth rotation curve of Kim et al. (1998) as an alternative case, the mass distribution would be similar to the one represented by the disk component in Fig. 4. Even in this case the mass center is still significantly displaced from the optical bar, and no particular optical enhancement corresponding to the exponential disk is recognized around the kinematical center.
The author thanks T. Takamiya for calculating the SMD from the rotation curve data. He is also indebted to Prof. M. Fujimoto of Nagoya University for invaluable discussion.
## References
Binney, T., Tremaine, S. 1987, in Galactic Astronomy (Princeton Univ. Press).
de Vaucouleurs G., de Vaucouleurs A., Corwin H. G. Jr., et al., 1991, in Third Reference Catalogue of Bright Galaxies (New York: Springer Verlag)
Gardiner, L. T., Sawa, T. Fujimoto, M. 1994 NMRAS, 266, 567.
Kim, S., Stavely-Smith, L., Dopita, M. A., Freeman, K. C., Sault, R. J., Kesteven, M. J., and McConnell, D. 1998, ApJ 503, 674.
Kunkel, W. E., Demers, S., Irwin, M. J., and Albert, L. 1997, ApJ 488, L129.
Luks, Th. and Rohlfs, K. 1982 AA, 263, 41L. @
Persic, M., Salucci, P., Stel, F. 1996, MNRAS, 281, 27.
Putman, M. E., Gibson, B. K., Staveley-Smith, L., Banks, C., Barnes, D. G. et al. 1998 Nature 394, 752.
Rubin, V. C., Ford, W. K., Thonnard, N. 1982, ApJ, 261, 439
Sofue, Y. 1994, PASJ, 46, 431
Sofue, Y. 1996, ApJ, 458, 120
Sofue, Y. 1997, PASJ, 49, 17
Sofue, Y., Tomita, A., Honma, M.,Tutui, Y. and Takeda, Y. 1998, PASJ 50, 427.
Sofue, Y., Tutui, Honma, M., Tomita, A., Takamiya, T., Koda, J. and and Takeda, Y. 1999, ApJ. Sept. issue.
Takamiya, T., and Sofue, Y. 1999, submitted to ApJ.
Figure Captions
Fig. 1: HI velocity field of LMC (Kim et al 1998) superposed on a smoothed DSS image in B band. The cross indicates the kinematical center of Kim et al (1998), and the asterisk is the adopted center position in this paper. Note the significant displacement of the bar from the kinematical center.
Fig.2: Position-velocity diagram of the LMC along the major axis (Kim et al 1998). We trace the maximum-velocity envelope by the thick line to derive the rotation curve.
Fig.3: Rotation curve of the LMC. Note the steep central rise, and flat disk rotation. The thin line denotes the observation, and the thick line is a smoothed rotation curve, which we adopt for calculating the surface-mass distribution.
Fig.4: Radial distributions of the surface mass density for a thin disk assumption (full line) and for a spherical assumption (dashed line). There appear three components: (1) a dense and compact bulge at radii $`<240`$ pc, (2) an exponential disk until 2 to 3 kpc, and (3) a massive halo.
|
no-problem/9906/hep-ph9906519.html
|
ar5iv
|
text
|
# References
SAGA-HE-151-99
Nonlinear effects in Schwinger-Dyson Equation
Hiroaki Kouno, Akira Hasegawa, Masahiro Nakano<sup>∗∗</sup> and Kunito Tuchitani
Department of Physics, Saga University Saga 840, Japan
\**University of Occupational and Environmental Health, Kitakyushu 807, Japan
\*e-mail address:kounoh@cc.saga-u.ac.jp
Abstract
We study nonlinear effects in the QED ladder Schwinger-Dyson(SD) equation. Without further approximations, we show that all nonlinear effects in the ladder SD equation can be included in the effective couplings and how a linear approximation works well. The analyses is generalized in the case of the improved ladder calculation with the Higashijima-Miransky approximation.
The dynamical symmetry breaking (DSB) is widely studied by the Schwinger-Dyson (SD) equation. (See references and , and the references therein. ) In QED, the lowest-order approximation is a ladder approximation. It was pointed out that there is DSB of the chiral symmetry in the QED ladder SD equation with a finite cut off. (See, e.g., . ) In QCD, the calculation was modified by introducing a running coupling. The modified calculation is called an improved ladder approximation and gives a good results even quantitatively.
Although the original SD equation is an integral equation, it is sometimes rewritten in the form of a nonlinear differential equation. (See, e.g., . ) Even in the most simple ladder approximation, the nonlinear equation can not be solved exactly and a linear approximation is widely used. (See, e.g., . ) The aim of this paper is to re-examine the nonlinear effects in the QED ladder SD equation and show how the linear approximation works well. After that, we generalize our analyses in the case of the improved ladder calculation with the Higashijima-Miransky approximation.
In Landau gauge, the QED ladder Schwinger-Dyson equation is given by
$$B(p^2)=3e^2\frac{d^4q}{(2\pi )^4}\frac{B(q^2)}{q^2+B^2(q^2)}\frac{1}{(pq)^2}+m_0,$$
(1)
where $`B(p^2)`$ is a fermion self-energy with four momentum $`p`$ and $`m_0`$ is a bare mass of the fermion. We restrict our discussions on Euclidean region only. Putting $`p=me^{\frac{t}{2}}`$($`m`$ is a some energy scale) and rescaling $`x(t)=B(p^2)/p`$, after some algebra, we get a nonlinear second differential equation
$$\ddot{x}+2\dot{x}+\frac{3}{4}x+\lambda \frac{x}{1+x^2}=0$$
(2)
with two boundary conditions
$$v_1e^{\frac{3t}{2}}(\dot{x}+\frac{1}{2}x)0(t\mathrm{})$$
(3)
and
$$v_3e^{\frac{t}{2}}(\dot{x}+\frac{3}{2}x)\frac{m_0}{m}(t\mathrm{}).$$
(4)
In these equations, $`\lambda =\frac{3e^2}{16\pi ^2}`$, $`\dot{x}=\frac{dx}{dt}`$ and $`\ddot{x}=\frac{d^2x}{dt^2}`$. Since we are interested in DSB of chiral symmetry, we put $`m_0=0`$. The ultra-violet boundary condition (UVBC) (4) becomes
$$v_3e^{\frac{t}{2}}(\dot{x}+\frac{3}{2}x)0(t\mathrm{}).$$
(5)
The equation (2) can be rewritten as
$$\frac{dv_1}{dt}=\lambda \frac{xe^{\frac{3t}{2}}}{1+x^2}\mathrm{or}\frac{dv_3}{dt}=\lambda \frac{xe^{\frac{t}{2}}}{1+x^2}.$$
(6)
We can construct a potential $`V(x)`$ as
$$V(x)=\frac{3}{8}x^2+\frac{\lambda }{2}\mathrm{log}(1+x^2);\frac{dV(x)}{dx}=\frac{3}{4}x+\lambda \frac{x}{1+x^2}$$
(7)
and a hamiltonian $`H(y,x)`$
$$H(y,x)=\frac{1}{2}y^2+V(x),$$
(8)
where $`y=\dot{x}`$ is a conjugate momentum of $`x`$. This system is a dissipative system and it is easy to derive
$$\frac{dH}{dt}=2y^20$$
(9)
The equality in (9) is hold only $`y=0`$. Because $`y=0`$ and $`x0`$ is not a stable point, $`H`$ decreases. Since the potential $`V(x)`$ has a minimum only at $`x=0`$, Eq. (9) gives
$$H(y,x),V(x)0$$
(10)
in the limit of $`t\mathrm{}`$. Next, we define
$$r\sqrt{x^2+y^2}$$
(11)
In the large $`t`$ limit, $`r^2`$ also approaches zero, since
$$\frac{3}{8}r^2H(y,x)\mathrm{max}(\frac{1}{2},\frac{3}{8}+\frac{\lambda }{2})r^2.$$
(12)
Therefore, $`x`$ and $`y`$ approach zero in this limit. Similarly, we can show
$$r^2\mathrm{},$$
(13)
when $`t`$ approaches $`\mathrm{}`$. Since $`x=0`$ and $`y=\dot{x}\pm \mathrm{}`$ is not a stable point, we consider the case $`x^2\mathrm{}`$ in the limit $`t\mathrm{}`$. In this case, the nonlinear equion (2) is reduced to a linear equation
$$\ddot{x}+2\dot{x}+\frac{3}{4}x=0.$$
(14)
This equation can be solved analytically and we get
$$x(t)=C_1e^{\frac{t}{2}}+C_3e^{\frac{3t}{2}},$$
(15)
where $`C_1`$ and $`C_3`$ are some constants. To satisfy the infra-red boundary condition (IRBC) (3), $`C_3=0`$. We remark that, due to Eq. (6), $`v_1`$ and $`v_3`$ are conserved in this limit. On the other hand, in the high-energy region in which $`x0`$, the nonlinear equation (2) reduces to another linear equation as
$$\ddot{x}+2\dot{x}+\frac{3}{4}x+\lambda x=0$$
(16)
The general solution in this limit is given by
$`x(t)`$ $`=`$ $`c_1e^{(1a)t}+c_2e^{(1+a)t}(\lambda <1/4)`$
$`x(t)`$ $`=`$ $`c_3e^t+c_4te^t(\lambda =1/4)`$
$`x(t)`$ $`=`$ $`c_5e^t\mathrm{cos}(bt+c_6)(\lambda >1/4),`$ (17)
where $`a=\sqrt{1/4\lambda }`$, $`b=\sqrt{\lambda 1/4}`$ and $`c_1c_6`$ are some constants. Therefore, in the high-energy region, the solution has oscillating behavior in the case of $`\lambda >\lambda _c1/4`$ and it can connect the line $`v_1=0`$ ($`y=x/2`$) of IRBC and the line $`v_3=0`$ ($`y=3x/2`$) of UVBC with $`m_0=0`$. In that case, the equation (1) has a non-trivial solution $`B(p^2)0`$ and DSB of chiral symmetry takes place. Therefore, $`\alpha _c4\pi \lambda _c/3=\pi /3`$ is called a critical coupling. Below, for simplicity, we call $`\lambda _c`$ itself a ”critical coupling”.
The condition $`\lambda >\lambda _c`$ is necessary but is not sufficient for DSB. If the solution continues to oscillate in the ultraviolet region, it may go beyond the line $`y=3x/2`$ of UVBC. On the other hand, the integral (1) may diverge. To converge the integral, a form factor (or a cutoff) which switches off the interaction at very large $`t`$ may be introduced. Due to the form factor, the equation (2) becomes (14) again and the solution ceases to oscillate. If the solution satisfies UVBC when the interaction is switched off by the form factor, it evolutes on the line $`y=3x/2`$ after that time. This is a sufficient condition for DSB. The aim of this paper is to re-examine the nonlinear effects in this scenario and show how the linear approximation works well. To achieve this, we define
$$x=r\mathrm{cos}\theta \mathrm{and}y=r\mathrm{sin}\theta .$$
(18)
We get the differential equations in this polar coordinate.
$$\frac{dr}{dt}=\mathrm{sin}\theta \left(2r\mathrm{sin}\theta +\frac{1}{4}r\mathrm{cos}\theta \lambda \frac{r\mathrm{cos}\theta }{1+r^2\mathrm{cos}^2\theta }\right).$$
(19)
$$\frac{d\theta }{dt}=12\mathrm{cos}\theta \mathrm{sin}\theta +\frac{\mathrm{cos}^2\theta }{4}\lambda \frac{\mathrm{cos}^2\theta }{1+r^2\mathrm{cos}^2\theta }.$$
(20)
The equation (20) can be rewritten as
$$\frac{d\theta }{dt}=\frac{y^2+2xy+(3/4+\lambda _{eff}(x))x^2}{r^2}=\frac{y^2+2xy+\omega ^2(x)x^2}{r^2},$$
(21)
where $`\lambda _{eff}(x)=\lambda /(1+x^2)(\lambda )`$ is an effective coupling and $`\omega ^2(x)=\frac{dV(x)}{dx}/x=3/4+\lambda _{eff}(x)`$. The $`\frac{d\theta }{dt}`$ becomes negative unless a second order equation for $`y`$
$$y^2+2xy+(3/4+\lambda _{eff}(x))x^2=0$$
(22)
has real solutions. This condition gives
$$\lambda _{eff}(x)1/4=\lambda _c.$$
(23)
In this case, the real solutions are given by
$$y=x\pm \sqrt{x^2(1\omega ^2(x))}=x\pm \sqrt{x^2\left(\frac{1}{4}\lambda _{eff}(x)\right)}.$$
(24)
These are boundaries of a region where $`\frac{d\theta }{dt}>0`$. We call this region ”typhoon-like” because the solution has a $`left`$-handed rotation in this region. On the boundaries $`\frac{d\theta }{dt}=0`$. For convenience, we include the boundaries in the typhoon-like region. In the remaining part of $`x`$-$`y`$ plane, $`\frac{d\theta }{dt}<0`$. We call this region ”anti-typhoon-like” because the solution has a $`right`$-handed rotation in this region. It is interesting that two boundaries lie between the line $`y=x/2`$ of IRBC and the line $`y=3x/2`$ of UVBC on $`x`$-$`y`$ plane. In particular, two boundaries coincide with the line $`y=x/2`$ and the line $`y=3x/2`$ when $`\lambda =0`$. In fig. 1, we show the boundaries in the cases of $`\lambda <\lambda _c`$ and $`\lambda >\lambda _c`$. The typhoon-like region is continuous at the origin of $`x`$-$`y`$ plane, if $`\lambda \lambda _c`$. It is discontinuous at the origin, if $`\lambda >\lambda _c`$.
Our picture of DSB is as following. In the infra-red limit, the solution is given by $`x=C_1e^{t/2}`$. As $`t`$ becomes large, the interaction is switched on and the conservation of $`v_1`$ is violated, i.e., the solution deviates from $`x=C_1e^{t/2}`$. As $`x`$ becomes smaller, the effective coupling $`\lambda _{eff}(x)`$ becomes larger. However, if $`\lambda \lambda _c`$, the effective coupling $`\lambda _{eff}(x)`$ can not be larger than $`\lambda _c`$. Therefore, there is a typhoon-like region just between the line $`y=x/2`$ of IRBC and the line $`y=3x/2`$ of UVBC. Since the solution has a left-handed rotation within the typhoon-like region, the solution is prevented from connecting the line $`y=x/2`$ with the line $`y=3x/2`$ in the same quadrant. On the other hand, the anti-typhoon-like region prevents the solution connecting the line $`y=x/2`$ with the line $`y=3x/2`$ in the different quadrant. If $`\lambda >\lambda _c`$, at some $`t`$, $`\lambda _{eff}(x)`$ becomes greater than $`\lambda _c`$. The typhoon-like region disappears. The solution begins to oscillate and can connect IRBC and UVBC. The DSB may happen. Therefore, $`\lambda >\lambda _c`$ is a necessary condition for DSB.
As is seen above, the nonlinear effects can be included in the effective coupling $`\lambda _{eff}(x)`$. The typhoon-like region also prevents the solution deviating much from the line $`y=x/2`$ of IRBC. Therefore, $`x=C_1e^{t/2}`$ may be a good approximation for the exact solution, until $`\lambda _{eff}(x)`$ becomes $`\lambda _c`$. Furthermore, if $`\lambda `$ is not much larger than $`\lambda _c`$, $`x`$ becomes sufficiently small before $`\lambda _{eff}(x)`$ becomes $`\lambda _c`$, and, therefore, the linear approximation (16) works well in the region where $`\lambda _{eff}(x)>\lambda _c`$. In this case, the effective coupling $`\lambda _{eff}(x)`$ can be approximated as $`\lambda /(1+(C_1)^2e^t)`$. This is nothing but the linear approximation used by Fomin, Gusynin and Miransky.
Next, we consider a sufficient condition for DSB. Suppose that $`\lambda `$ is slightly larger than $`\lambda _c`$ and $`x`$ does not deviate much from $`y=\dot{x}=x/2`$ until some $`t_0`$ where $`r`$ becomes sufficiently small. We can re-define $`t_0`$ as an origin of $`t`$. Neglecting the $`r`$-dependence, we integrate the equation (20) with an initial conditions $`\dot{x}(0)=x(0)/2`$ and impose a final condition $`\dot{x}(t_{cut})=3x(t_{cut})/2`$ where $`t_{cut}`$ is a cutoff parameter. The result gives the condition
$$bt_{cut}=n\pi \theta ^{}(n=1,2\mathrm{}),$$
(25)
where $`b`$ is the same as in Eq. (17) and $`\theta ^{}=\mathrm{arctan}\frac{4b}{14b^2}`$. This is an approximate version of the condition for DSB in .
Our analyses can be generalized in more general equation as
$$\ddot{x}+2\gamma (t,x)\dot{x}+\omega ^2(t,x)x=\ddot{x}+2\gamma (t,x)\dot{x}+\frac{V(t,x)}{x}=0.$$
(26)
In this case, we get
$$\frac{dH}{dt}=2\gamma (t,x)y^2+\frac{V(t,x)}{t}$$
(27)
$$\frac{d\theta }{dt}=\frac{y^2+2\gamma (t,x)xy+\omega ^2(t,x)x^2}{r^2}$$
(28)
As is already seen, if we put $`\gamma (t,x)=1`$ and $`\omega (x,t)=3/4+\lambda /(1+(C_1)^2e^t)`$ in Eq. (26), we get the linear approximation used by Fomin, Gusynin and Miransky. If we put $`\gamma (t,x)=1`$ and $`\omega (x,t)=3/4+\lambda \mathrm{\Theta }(t)`$ where $`\mathrm{\Theta }(t)`$ is a step function, it reproduces the results in Eq. (25).
To apply our analyses to the improved ladder calculation, consider the integral
$$B(z)=\frac{1}{z}_0^z𝑑u\frac{\lambda (z,u)B(u)u}{u+B^2(u)}+_z^{\mathrm{}}𝑑u\frac{\lambda (z,u)B(u)}{u+B^2(u)}+m_0,$$
(29)
where $`z=p^2=m^2e^t`$ and $`u=q^2`$. For $`\lambda (x,y)`$, we use the Higashijima-Miransky approximation
$$\lambda (z,u)=\lambda (z)\mathrm{\Theta }(zu)+\lambda (u)\mathrm{\Theta }(uz),$$
(30)
Differentiating $`B(z)`$ twice with respect to $`z`$, we get the differential equation as Eq. (26) with
$$\gamma (t,x)=\gamma (t)=1\frac{1}{2}U(t);U(t)\frac{\dot{\lambda }^{}(t)}{\lambda ^{}(t)}$$
(31)
and
$$\omega ^2(t,x)=\frac{1}{x}\frac{V(t,x)}{x};V(t,x)\frac{1}{2}\left(\frac{3}{4}\frac{1}{2}U(t)\right)x^2+\frac{\lambda ^{}(t)}{2}\mathrm{log}(1+x^2),$$
(32)
where $`\lambda ^{}(t)=\lambda (t)\dot{\lambda }(t)`$. The IRBC and UVBC are modified as
$$v_1^{}e^{\frac{3t}{2}}\frac{(\dot{x}+\frac{1}{2}x)}{\lambda ^{}(t)}0(t\mathrm{})$$
(33)
and
$$v_3^{}e^{\frac{t}{2}}\left(\frac{\lambda (t)}{\lambda ^{}(t)}\dot{x}+\left(1+\frac{1}{2}\frac{\lambda (t)}{\lambda ^{}(t)}\right)x\right)\frac{m_0}{m}(t\mathrm{}).$$
(34)
We remark that IRBC (33) is reduced to the original one $`v_1=0`$ if $`\lambda ^{}(t)`$ is finite in the limit $`t\mathrm{}`$. Due to (28), if
$$D=x^2(\gamma ^2(t)\omega ^2(t,x))=x^2\left(\frac{1}{4}(1U(t))^2\lambda _{eff}^{}(t,x)\right);\lambda _{eff}^{}(t,x)\frac{\lambda ^{}(t)}{1+x^2}$$
(35)
becomes negative, a typhoon-like region disappears. The condition (35) gives $`\lambda _{eff}^{}(t,x)>1/4(1U(t))^2`$. In the case of $`D0`$, the boundaries of the typhoon-like region is given by
$$y=x\gamma (t)\pm \sqrt{D}.$$
(36)
If $`\dot{\lambda }(t)`$ and $`\ddot{\lambda }(t)`$ is small compared with $`\lambda (t)`$, we get $`\gamma (t)1`$, $`\omega ^2(x,t)3/4+\lambda (t)/(1+x^2)`$ and $`v_3^{}v_3`$. In this case, approximately, $`\mathrm{max}(\lambda (t))>\lambda _c`$ is a necessary condition for DSB.
In summary, we have re-examined the solution of DSB of chiral symmetry in the ladder SD equation. Without further approximation, we have showed that the nonlinear effects are included in the effective couplings $`\lambda _{eff}(x)`$. If $`\lambda _{eff}(x)>\lambda _c`$, the typhoon-like region disappears and DSB takes place. The typhoon-like region prevents the solution deviating much from IRBC and makes the linear approximation works well. The analyses is generalized in the case of the improved ladder calculation with the Higashijima-Miransky approximation.
Acknowledgements
The authors thank K-I. Aoki and M. Maruyama for useful discussions. One of the author (H.K.) also thank O. Kiriyama and F. Takagi for useful discussions.
|
no-problem/9906/astro-ph9906307.html
|
ar5iv
|
text
|
# Average Emissivity Curve of BATSE Gamma-Ray Bursts with Different Intensities
## 1 Introduction
Gamma-ray bursts are known to have very different intensities. The peak flux of bursts varies more than 2 orders of magnitude from the BATSE trigger threshold of about 0.3 ph cm<sup>-2</sup> s<sup>-1</sup> up to the highest measured values of about 50 ph cm<sup>-2</sup> s<sup>-1</sup> (Fishman and Meegan (1995)). The rich statistics of the BATSE 4B Catalog (Paciesas et al. (1998)) enables one to divide all detected bursts into several intensity groups with reasonably large numbers of events ($`150`$) in each of them. Although individual bursts have very different time profiles, a characteristic average temporal signature can be produced for each group. Using these average signatures, a comparison can be made between the duration of different brightness groups of gamma-ray bursts. The main goal of this comparison is to test the brightness-dependent stretching of gamma-ray bursts.
It is well-known that individual pulses from radio pulsars have quite variable emission time profiles, so that the characteristic periodic pulsar signal is hardly recognizable by examining a short interval of real-time data. The randomized time profile is transformed into the stable signature of a pulsar light curve by using the epoch folding technique, averaging the data over many periods. A similar signature has been proposed for cosmic gamma-ray bursts: averaging the time profiles of individual events by the normalized peak-alignment technique, where each time profile is normalized by the peak number of counts $`C_{\mathrm{max}}`$, aligned at peak time bins $`t_{\mathrm{max}}`$ and then averaged for all bins along the time scale (Mitrofanov et al. 1996, hereafter Paper I ). This technique produces an Average Curve of Emissivity, or ACE, and the corresponding signature represents the averaging of observed time histories of bursts.
ACE curves have already been studied for different energy ranges and for different intensity groups of bursts, and they have been found to be rather convenient signatures to describe the basic properties of the slow temporal variation of bursts (Mitrofanov et al. (1997)). This signature averages out the fast variations of bursts in the sample and describes the general envelope of rising emission before the main peaks of bursts and the subsequent decaying tail. An analytic approximation to the ACE has been found, which results in a very acceptable fit to the observations. The ACE equivalent width parameter describes the mean time scale of the slow variation of GRBs.
The comparison of the generic time-dependent properties of gamma-ray bursts with different brightnesses has recently become an issue of common interest. Indeed, if the difference in intensities represents the difference in distances to the emitters, then the difference between the average time signatures of different brightness groups should manifest effects of cosmological time-dilation, where dimmer bursts are observed to be broader than brighter events because they were emitted at larger cosmological distances. The time-dilation test based on the comparison of the averaged temporal structures for dim and for bright sets has already been done by two groups, which have drawn two opposite conclusions: in the first case a large time-dilation effect $`2`$ was seen (Norris et al. (1994, 1997)), in the other case none was observed (Mitrofanov et al. (1994) & Paper I ).
The effect of cosmological time-dilation should also be accompanied by the effect of photon energy red-shift, which could influence the time-stretching value (Fenimore & Bloom 1995a ; Mitrofanov et al. (1997)). On the other hand, sources of bright and dim gamma-ray bursts could be non-standard candles with intrinsic luminosity-based correlations (Brainerd (1997)). In this case, the observed time-stretching of dim bursts could result from the fact that shorter outbursts correlate with smaller intrinsic luminosity of their sources. Also, one should take into account possible distance-related evolution of bursts emitters: in the co-moving reference frames, more distant sources could have shorter time profiles than nearer sources. These intrinsic effects could interfere with the cosmological effects.
Therefore, in order to make conclusive statements about generic time-dependent properties of gamma-ray bursts, one should distinguish between the physical effect of time-dilation, which results from the expansion of the Universe, and the phenomenological effect of time-stretching, which is found by the comparison of bursts with different intensities. We think that the first logical step should be the investigation of the stretching/intensity phenomenon by comparing time profiles for different intensity groups of bursts. If a significant time-stretching effect is found between different intensity groups, then the next logical step can be made. This step should be the physical interpretation of the observed stretching, according to predictions of a cosmological model, taking into account the geometrical effects of time-dilation and red-shift, the physical effects of time-energy dependence of emission, the effects of a broad luminosity distribution of emitters, and finally, the effects of distance-related evolution. In this paper, we restrict ourself to purely phenomenological studies of the stretching/intensity correlation of bursts, using the average emissivity curves for different brightness groups of bursts as the tool for measuring the stretching.
A total set of 887 bursts with $`t_{90}>2`$ s from the BATSE 4B Catalog (Meegan et al. (1994); Paciesas et al. (1998)) have data available on the 1024 ms timescale. We have divided these into 6 intensity groups (Table 1), using the peak flux parameter $`F_{\mathrm{max}}^{(1024)}`$ as the selection criterion. For each of these groups, an ACE profile is produced using counts observed in the broad energy range 50-300 keV. The technique used to produce ACE profiles is described in Paper I . Below, the stretching/intensity effect is investigated and stretching factors are evaluated between each ACE<sub>i</sub> for the dimmer groups ($`i=2`$ – 6) and the reference group (ACE<sub>1</sub>). We have done this analysis for the total profile as well as for the rise fronts and the back slopes separately.
## 2 Analytical approximation of ACE profiles
We have previously determined that there is a simple analytical form that fits the ACE profiles of all burst intensity groups in several different spectral ranges quite well (Mitrofanov et al. (1997)). For each selected group ($`i`$) of bursts, the analytical approximation to the ACE is
$$f^{(i)}[t]=\left(\frac{\mathrm{\Delta }t^{(i)}}{\mathrm{\Delta }t^{(i)}+|tt_{\mathrm{max}}|}\right)^{a_{\mathrm{RF}}^{(i)},a_{\mathrm{BS}}^{(i)}},$$
(1)
with different power indices ($`a_{\mathrm{RF}}^{(i)}`$, $`a_{\mathrm{BS}}^{(i)}`$) at the rise front (RF, with $`t<t_{\mathrm{max}}`$) and at the back slope (BS, with $`t>t_{\mathrm{max}}`$), respectively.
Another approximation for ACE-like profiles with exponential wings has been suggested by Stern et al. (1996). A direct comparison between these two approximations has shown that a power law (eq. 1) fits the 1024 ms time scale ACE profiles over the range of 20 time bins before and after the peak much better than the exponential model. For example, with intensity group 1 (Table 1) we found a reduced $`\chi ^2=1.51`$ for an exponential-type law, while Eq. 1 provided a better fit, with reduced $`\chi ^2=0.88`$. For this reason, we use Eq. 1 to model ACE profiles in this paper.
## 3 Two procedures to estimate stretch factors between ACE curves of different intensity groups
When cosmological models were first suggested for GRBs, the observational tests for cosmological signatures were expected be quite obvious; for the case of ACE stretching coefficients, a factor of at least 2 should be observed between the brightest and dimmest burst intensity groups. Therefore, a simple procedure was devised to look for evidence for this stretching (see Paper I ). However, a stretching factor of the expected magnitude was not found; indeed, recently-developed cosmological models for bursts predict much smaller average stretching, despite the fact that some emitters should exist at large cosmological red-shifts (see e.g. Brainerd (1997)). In order to make further progress, a more sensitive comparison method between different ACE profiles has been developed, allowing the measurement of weak-signature stretching factors (less than 1.5). To be successful, this method must take into account several statistical biases associated with the limited number of bursts in different intensity group samples and with the poor counts statistics of weak events.
We have developed two such procedures to make a robust comparison of ACE profiles between different brightness groups of bursts. Procedure (a), which is relatively time-consuming, requires the following steps to compare intensity groups $`i=2`$ – 6 with the reference group $`i=1`$:
(a1) An arbitrary stretching coefficient $`Y`$ is selected, and each burst of the reference group 1 is stretched by the factor $`Y`$.
(a2) An artificially-dimmed reference group, which has the same signal-to-noise ratio as group $`i`$, is created from bursts of the stretched reference group (see Section 5).
(a3) ACE profiles are produced for the stretched and dimmed version of the reference group and for the original group $`i`$. The value of $`\chi ^2`$ is used to measure the difference between them. To find the best fit stretching coefficient $`Y_i`$, the value of the stretching coefficient $`Y`$ is changed, and the cycle (a1)–(a3) is performed again to minimize $`\chi ^2`$.
The time-efficient procedure (b) is more simple and straight-forward than the first one. It consists of the following steps:
(b1) An artificially-dimmed reference group that has equal signal-to-noise ratio to the the group $`i`$ is created from the reference group (see Section 5).
(b2) An ACE profile is produced for the artificially-dimmed reference group, and the parameters of the best-fit analytical model (Eq. 1) are computed.
(b3) The stretching parameter Y for the time constant $`\mathrm{\Delta }t^{(1)}`$ is introduced into the analytical model (Eq. 1) for the artificially dimmed reference group 1:
$$f_{dim}^{(i)}[t]=\left(\frac{Y_i\mathrm{\Delta }t^{(1)}}{Y_i\mathrm{\Delta }t^{(1)}+|tt_{\mathrm{max}}|}\right)^{a_{\mathrm{RF}}^{(1)},a_{\mathrm{BS}}^{(1)}}.$$
(2)
Fixing all other parameters, the best fit value for $`Y`$, as a free parameter, is found for the ACE of each group $`i=26`$. This value is taken to be the stretch coefficient $`Y_i`$ between the group $`i`$ and the reference group.
A direct comparison of these two procedures has been performed for the back slopes (time bins after the peak in the ACE) of bursts using the 3B Catalog data base (Litvak et al. 1997a ). Excellent agreement between the best-fit stretch factors ($`Y`$) estimated from both procedures was found. The difference between them is much less than 1$`\sigma `$ for choice statistics (see Section 4). Therefore, the computationally time-efficient procedure, (b), is used below in studies of ACE stretching.
## 4 Choice statistics and the related errors of the stretching coefficients
The proposed method suggests that we may use the ACE as the signature representing the slow variability of a selected sample of gamma-ray bursts. Individual bursts have very different time profiles, and one must use a very large number of individual events to build up a representative ACE that will be the same for different samples of bursts with similar brightnesses.
The errors associated with ACE curves may be estimated from the sample variance for the selected groups of bursts. ACE profiles for two independent groups are statistically indistinguishable, provided the difference between them is within these errors. The direct comparison of ACE profiles for different groups has shown that they are much more different than might expected from the variance within each sample only. This means that these groups are not representative samples, and a random selection of bursts within these groups does not ensure well-weighted contributions from all kinds of possible profiles.
To study the random choice statistics of bursts, a Monte Carlo test has been performed using the total set of 603 bursts from the 3B Catalog (Meegan et al. (1997)) that have catalog values of peak flux and durations $`T_{90}>2`$ s (Litvak et al. 1997b ). Different numbers of bursts, $`N=100`$, 150, 200, 250, 300, were assumed. We randomly selected $`10^4`$ times, two testing groups of $`N`$ events each from the total set of bursts, produced ACEs for these groups, and then determined the best-fit stretch factor $`Y_{\mathrm{choice}}`$ between them. The distributions of $`Y_{\mathrm{choice}}`$ for two different values for $`N`$ are presented in Fig. 1.
The spread of stretching coefficients $`Y_{\mathrm{choice}}`$ due to random choice statistics is found to be much broader than would be expected from the sample variance for each individual group. Of course, the distribution of $`Y_{\mathrm{choice}}`$ becomes narrower for larger $`N`$, but even for the largest value, $`N=300`$, it is still quite broad.
Therefore, we conclude that for any two intensity groups, with $`N`$ bursts each, the estimated stretching coefficient must be compared with standard deviations from the random choice distribution, evaluated for $`N`$ events. For $`N=150`$, the level of 1$`\sigma `$ significance corresponds to the stretching coefficient $``$1.10. Therefore, for $`Y1`$, one should use the errors $`\delta Y0.10`$ for the 1$`\sigma `$ errors of the stretching coefficients. In the arbitrary case where $`N100`$, one can use following for estimations of errors:
$$\frac{\delta Y}{Y}=0.10\sqrt{\frac{150}{N}}.$$
(3)
Below, these errors were used to estimate the significance of stretching between selected brightness groups (Tables 2 and 3).
## 5 Count-noise produced effects on the ACE
The procedure to create each ACE includes the selection of the highest peak of each burst, with counting rate $`C_{\mathrm{max}}`$, and the normalization of the time profile by the $`C_{\mathrm{max}}`$ value. Therefore, the ACE is sensitive to a bias where positive fluctuations of counts ($`C+\delta C`$) dominate the selection of $`C_{\mathrm{max}}`$. When the normalization is performed, a profile is lowered $`(C_{\mathrm{max}}+\delta C)/C_{\mathrm{max}}`$ times. The effect should be larger for dimmer bursts, where the ratio of $`(C_{\mathrm{max}}+\delta C)/C_{\mathrm{max}}`$ is larger.
To evaluate the effect, the reference group 1 (Table 1) has been transformed by the procedure of Monte Carlo noisification into 5 artificial reference groups, which have peak flux distributions similar to the corresponding distributions for the 5 observed dim intensity groups, $`i=`$2–6 (Table 1). The Monte Carlo transformation procedure includes the following steps:
a) For each original burst of the reference group (1), some counterpart event is randomly selected inside the testing group $`i`$.
b) The ratio of peak fluxes $`f=F_{\mathrm{max}}^{(1)}/F_{\mathrm{max}}^{(i)}`$ is estimated between the peak fluxes of the original burst and its weaker counterpart.
c) The time profile of the original bright burst is divided by the factor $`f`$, and counts for an artificially-dimmed version of the original burst are simulated using Poisson statistics:
$$D_j=\frac{C_{\mathrm{S},j}}{f}+C_{\mathrm{B},j},$$
(4)
where $`C_{\mathrm{S},j}`$, $`C_{\mathrm{B},j}`$ are the signal and background counts accumulated during the $`j`$th time bin for a burst from the reference group.
Using equation (2), stretching coefficients $`Y_{\mathrm{noise}}`$ have been estimated between the ACE<sub>1</sub> for the original reference group and ACE$`{}_{}{}^{\mathrm{art}}{}_{i}{}^{}`$ for the artificial reference groups ($`i=2`$–6).
One can see the largest effect is that the ACE$`{}_{}{}^{\mathrm{art}}{}_{6}{}^{}`$ for the dimmest group ($`i=6`$) must be broadened by a factor of 1.22 to equal the width of the ACE<sub>1</sub> for the original reference group. Therefore, if the observed stretching factor between ACEs for the actual observed dim group and the reference group is equal to 1, it should be interpreted as evidence for real stretching on the order of $`Y_{\mathrm{noise}}`$ between the dim and reference groups, because the ACE for the dim group is known to be narrowed systematically due to larger relative Poisson noise. This effect could be ignored when testing for much larger stretching coefficients above $`2.0`$ (Paper I ), but it should certainly be taken into account when the expected stretching might be as small as the noise-produced effect (see below). Both procedures (a) and (b) (Section 3) take this effect into account.
## 6 Stretching factors between the ACE for different intensity groups
### 6.1 Stretching with respect to the reference group
Using procedure (b) to compare between ACEs, stretching coefficients were estimated for each intensity group (Figure 2 and Table 2). The group of brightest events (1) is used as a $`\mathrm{𝑟𝑒𝑓𝑒𝑟𝑒𝑛𝑐𝑒}`$ and the corresponding stretching coefficient for this group was defined to be 1. For each group, the stretching factors $`Y_{\mathrm{RF}}^{(1)}`$, $`Y_{\mathrm{BS}}^{(1)}`$ and $`Y_{\mathrm{TOT}}^{(1)}`$ were estimated with respect to group 1 for the RF and BS wings and for the total ACE profiles, respectively (Figure 2 and Table 2) .
The rise front portions of bursts do not manifest any significant increase of stretching with decreasing intensity of bursts. The stretch factor for the rise front of group 5 with respect to the reference group is the only one that is above the 3$`\sigma `$ level (in this case, 3$`\sigma =1.34`$). On the other hand, at the back slopes the stretching factors $`Y_{\mathrm{BS}}^{(1)}`$ are larger than $`3\sigma `$ already between groups 2 and 1, and increase up to $``$2 with decreasing brightness (Table 2). The noise-produced systematic narrowing of the ACE is taken into account for these values, because the test groups ($`i=2`$ – 6) are compared with the correspondingly noisified reference groups. In the case of the comparison with the original reference group (1), the stretching would be less. The lack of stretching during the rising portion of the ACE of the dimmest group (6) may be an instrumental effect, resulting from the deficit of slow-rising events near the trigger threshold (e. g. Higdon & Lingenfelter, (1996)). We know from the non-triggered burst sample of Kommers et al. (1997) that the threshold effect may change the estimate for the rise-front stretching factor only for the dimmest group 6 in our sample.
When the total ACE profiles are compared, the stretching factors have intermediate values between the corresponding factors for rise fronts and back slopes. One can see that the back slope stretching factors for dimmer groups ($`i=3`$–5) are $`2\sigma `$ larger than those associated with the rise front (Table 2). That is why one should study the stretching phenomena of bursts separately for the rise front and back slopes.
### 6.2 Stretching with respect to the second brightest group
With decreasing burst intensities, the largest increase of stretching between successive intensity groups happens between the brightest and the second brightest groups (Table 2 and Figure 1). Of course, this could result from a random choice fluctuation: the number of bursts in each group is still not large enough to exclude this possibility. Larger burst samples from subsequent BATSE catalogs should be used to check the jump of stretching for the two intensity groups bounded by the flux of $`4.1`$ ph cm<sup>-2</sup> s<sup>-1</sup>. The jump may completely disappear for larger statistics; or, more interestingly, it could be confirmed as a real phenomenological effect.
However, one has to be sure that this jump-like effect between the test group ($`i=2`$) and the reference ($`i=1`$) does not result from some systematic effect in the comparison procedure. Indeed, the reference group has been transformed by adding noise in order to be to be compared with dimmer groups, and some unknown systematics in this procedure could result in a jump of stretching between the reference group and dimmer groups.
To check this possibility, the same procedure used for the estimations of stretching factors has been reproduced; however, this time assuming the second bright group (2) to be the reference (Table 3). The stretching factors $`Y_{\mathrm{RF}}^{(2)}`$ and $`Y_{\mathrm{BS}}^{(2)}`$ for group 2 have been defined as 1.0 in this case. To facilitate comparison between stretching coefficients based on group 1 and group 2, the re-normalized factors $`1.19Y_{\mathrm{RF}}^{(2)}`$ and $`1.35Y_{\mathrm{BS}}^{(2)}`$ are also presented in Table 3. There is very good agreement between stretching factors $`Y^{(1)}`$ (Table 2) and $`Y^{(2)}`$ (Table 3), based on the reference groups 1 and 2, respectively.
Therefore, one can exclude the possibility that some systematic effect takes place when the reference group is compared with dimmer ones. One may conclude that in the 4B catalog data set a large stretching factor of about 1.35 really exists between the back slopes of bursts of intensity groups 1 and 2, separated by a flux $`4.1`$ ph cm<sup>-2</sup> s<sup>-1</sup>.
### 6.3 Consistency between stretching factors for intensity groups of the 2B and 4B database
A comparison between ACE profiles was done previously (Paper I ) using the database of 338 events from the BATSE 2B catalog (Meegan et al. (1994)), with durations $`T_{90}>1`$ s. These were divided at the peak flux value $`1`$ ph cm<sup>-2</sup> s<sup>-1</sup> into two intensity groups with 143 bright bursts and 179 dim bursts. The estimated equivalent time widths $`t_{ETW}`$ (see Paper I ) of ACE rise fronts, back slopes and total profiles for these old samples are presented in Table 4. The values labeled “Old samples” are taken from Paper I but the errors are estimated according to random choice statistics as described above (Section 4). No significant stretching is seen between the ACE for these groups.
The 4B catalog allows us to do a similar analysis for new samples with as many as 480 bright bursts and 464 dim events, and to check the consistency between the estimations of equivalent time widths for the old and new samples. One concludes from Table 4 that the “old” and “new” results are consistent for intensity groups divided by the same peak flux value $`1`$ ph cm<sup>-2</sup> s<sup>-1</sup>. There is no evidence for stretching for the “old” groups with 143 bright and 179 dim events either during the rise front or the back slope. Using the much better statistics available with the “new” intensity groups, there still is no significant stretching during the rise fronts, but there is some effect when $`t_{ETW}^{(\mathrm{BS})}`$ are compared.
The separation of all bursts into bright and dim groups by the peak flux $`1`$ ph cm<sup>-2</sup> s<sup>-1</sup> was appropriate when the goal was to test for an obvious signal of stretching by factors large as $`2`$. On the other hand, the 2B catalog sample was too small for more accurate sampling. Now, with much better statistics, more subtle stretching effects can be seen for several intensity groups (Table 2), in particular at the back slope of ACE profiles. In the half-to-half separation, a stretching effect of $`2\sigma `$ is also seen at the back slope (Table 4).
## 7 Conclusions
Six burst intensity groups with $`150`$ events each have been compared to determine a stretching effect between the ACE profiles for dim and bright groups. To study the stretching, a separate comparison is preferable for the average rise fronts and back slopes. The Pearson $`\chi ^2`$ statistic between the reference and the stretched dimmer groups allows us to find the most probable stretching factors for the ACE of different intensity groups (Table 2), and the resulting reduced $`\chi ^2`$ values are significantly smaller when the rise fronts and back slopes are each compared separately. There is a significant difference between the corresponding stretching factors for the rise fronts and back slopes in different intensity groups. During the rise front a stretch factor of $`1.39\pm 0.14`$ is found between intensity groups 5 and 1, which seems to be marginally significant. The other factors determined for the rise fronts, taken together, also indicate some stretching effect, but not significantly.
This study is based on BATSE 1 s resolution discriminator data. Higher temporal resolution datatypes begin either at the trigger time or 2 s before. Our choice of the continuously available 1 s data avoids possible systematic effects in the rise front due to mixing datatypes of differing temporal resolutions.
There may be an instrumental triggering bias that selects against those dim bursts that rise more slowly on the average. This has been corroborated in the study of non-triggered bursts by Kommers et al. (1997), who found that only some of the dimmest bursts failed to trigger due to slow rise times. The intensities of these events correspond to our dimmest intensity group 6. Since only group 6 is incomplete due to missing slow risers, and since this incompleteness affects only the rise front results, we base our conclusions on rise front stretching on groups 1 to 5. Therefore, omitting the dimmest group, there is the indication of a trend for the rise-front stretching factors (Figure 3).
At the back slopes the estimated stretch factors are quite significant for all dimmer groups ($`i=2`$ – 6) with respect to the brightest one ($`i=1`$), and the largest factor between the ACE profiles of dimmest 150 bursts and the brightest 147 bursts is about 2.1$`\pm `$ 0.2. The non-stretching hypothesis may be significantly rejected for the back slopes of bursts. Our results are in qualitative agreement with recent estimations of time-stretching by Norris et al. (1994), where a different energy range was used for burst averaging and a different time scale was applied for burst selection.
We conclude that models of gamma-ray bursts should explain two phenomenological results:
1) Back slopes for the 150 dimmest bursts, with peak fluxes $`<0.45`$ ph cm<sup>-2</sup> s<sup>-1</sup>, are on the average $``$2 times longer in duration than the 150 brightest bursts, with peak fluxes $`>4.1`$ ph cm<sup>-2</sup> s<sup>-1</sup>.
2) While a trend of increasing stretching factor, relative to group 1, may exist for rise fronts for groups 2 to 5, the magnitude of the stretching factor is less than $`1.4`$ and is therefore inconsistent with the stretching factor of the back slope.
3)Finally, there is definitely no full-profile stretching between dim and bright bursts as large as $`2`$ or more. Whatever stretching exists is weak and so the correct determination of the stretching factor will require careful treatment of statistical and systematic effects that are comparable with the physical effect of stretching that we are looking for. Using the 4B data, we conclude that a significant stretching by a factor of $``$2 between the different brightness groups may only be resolved for the back slopes of the average light curves of GRB time profiles.
Our primary concern has been assessing the observational evidence for time stretching in the average profiles of GRBs. The observation of stretching in the back slopes but not in the rise fronts of GRBs cannot be solely caused by cosmological effects – at least one of these phenomena seems to be intrinsic to GRBs, perhaps indicating a slope correlation with absolute intensity or the existence of source evolution. Whatever the explanation, determining the distance scale of GRBs using cosmological tests is proving to be more difficult than had been hoped.
## 8 Acknowledgments
The work in USA was supported by NASA grant CRO-96-173. The work in Russia was supported by RFBR grant 96-02-18825.
|
no-problem/9906/astro-ph9906264.html
|
ar5iv
|
text
|
# Water vapor absorption in early M-type stars
## 1 Introduction
Water is one of the most abundant molecules in the atmosphere of late M-giants, and it is a dominant absorber in the near-infrared (near-IR) region. Water vapor in stellar atmospheres has been studied by theoretical and observational methods. The strength of the H<sub>2</sub>O absorption possibly correlates with the spectral type, the effective temperature of the star, and the near-IR color (Kleinmann & Hall kleinmann (1986); Lançon & Rocca-Volmerange lancon (1992)), even though such correlations were not clearly found by Hyland (hyland (1974)). According to Spinrad & Wing (spinrad (1969)) and Hyland (hyland (1974)), H<sub>2</sub>O could only be detected in giants with spectral type M6 or later. This was consistent with hydrostatic models of the atmosphere of red-giants (Tsuji tsuji (1978); Scargle & Strecker scargle (1979)). As a further complication, most late M-giants are long period variables, and in general, the band strength of the H<sub>2</sub>O absorption features depends on stellar variability. For example, Mira variables show very deep H<sub>2</sub>O absorption, and the depths of H<sub>2</sub>O features in Miras vary from phase to phase (Hyland hyland (1974)). High-resolution spectroscopic observations of the Mira variable R Leo by Hinkle & Barnes (hinkle (1979)) revealed that a significant fraction of the H<sub>2</sub>O molecules were in a component with a distinct velocity, and a cooler excitation temperature than molecules near the photosphere. They interpreted this ‘cool component’ as an overlying layer above the photosphere.
These previous studies were mostly based on ground or airborne observations, where the terrestrial H<sub>2</sub>O interferes with a detailed study of the center of the stellar water bands. In contrast, observations from space are ideal for investigations of stellar H<sub>2</sub>O features. Using the Infrared Space Observatory (ISO), Tsuji et al. (tsuji97 (1997)) discovered a weak H<sub>2</sub>O absorption in the early M-type star, $β$~Peg (M2.5II-III). They argued that the observed H<sub>2</sub>O is in a ‘warm molecular layer’ above the photosphere.
In this paper, we present the results of a study of H<sub>2</sub>O absorption features, using data from the Infrared Telescope in Space (IRTS, Murakami et al. murakami96 (1996) and references therein).
## 2 Sample Data
This study is based on data from the two grating spectrometers onboard the IRTS: the Near-Infrared Spectrometer (NIRS), and the Mid-Infrared Spectrometer (MIRS). The IRTS was launched in March 1995, and it surveyed about 7% of the sky with four instruments during its 26 day mission. The NIRS covers the wavelength region from 1.43 to 2.54 and from 2.88 to 3.98 $`\mu `$m in 24 channels with a spectral resolution for point sources of $`\mathrm{\Delta }\lambda =0.100.12`$ $`\mu `$m. The MIRS covers the range from 4.5 to 11.7 $`\mu `$m in 32 channels with a resolution of $`\mathrm{\Delta }\lambda =0.230.36\mu `$m. Both spectrometers have a rectangular entrance aperture of $`8\mathrm{}\times 8\mathrm{}`$. The total number of detected point sources is about 50,000 for the NIRS (Freund et al. freund (1997)) and about 1,000 for the MIRS (Yamamura et al. yamamura96 (1996)). The estimated absolute calibration errors are 5% for the NIRS, and 10% for the MIRS. This may cause systematic errors in the colors and the H<sub>2</sub>O index discussed in this paper. The spectra are not color corrected.
All stars used in this study were observed both by the NIRS and the MIRS between April 9 and 24. We only include stars at high galactic latitudes ($`|b|>10\mathrm{°}`$) in this sample, to minimize source confusion and interstellar extinction. Each selected star has a unique association with the IRAS Point Source Catalog (PSC, iras-psc (1988)), within 8$`\mathrm{}`$ from the nominal NIRS position, and each associated PSC entry has reliable 12 and 25 $`\mu `$m band fluxes (FQUAL$`=`$3; IRAS Explanatory Supplement iras-es (1988)). Known carbon and S-type stars (Stephenson stephenson89 (1989), stephenson84 (1984)), as well as non-stellar objects were discarded from the sample. The current sample contains 108 stars. The signal-to-noise ratio for all stars is larger than 5 in all NIRS bands.
Of the 108 stars used in this study, 43 have a unique association in the Bright Star Catalogue (BSC, Hoffleit & Jaschek bsc (1991)). The distribution of spectral types for these 43 stars are: 1 B, 2 F, 4 G, 22 K, and 14 M-types. A total of 31 M-giants are found in the General Catalogue of Variable Stars (GCVS, Kholopov et al. gcvs (1988)), and the distribution of variable types in the GCVS is as follows: 10 Miras, 13 semi-regulars (SR, SRa or SRb, hereafter SR), and 8 irregulars (L or Lb, hereafter L). No M-dwarfs or M-supergiants were found in the BSC or the GCVS associations.
## 3 Results
In Fig. 1, we show the composite spectra of six representative M-giants observed by the NIRS and the MIRS. We indicate the position of the molecular absorption features due to CO (1.4, 2.3, 4.6 $`\mu `$m), H<sub>2</sub>O (1.5, 1.9, 2.7, 6.2 $`\mu `$m), and SiO (8.2 $`\mu `$m). The broad-band emission at 9.7 $`\mu `$m is due to silicate dust. The H<sub>2</sub>O bands at 1.9 and 2.7 $`\mu `$m are visible in two early M-giants, AK~Cap (M2, Lb) and V~Hor (M5, SRb), where no H<sub>2</sub>O in the photosphere was expected to be detectable.
One could argue that the early M-type stars with H<sub>2</sub>O absorption had spectral types later than M6 at the time of observation, because of their variability. However, this is not the case. We can estimate the spectral types, using the relation between the spectral type and color (Bessell & Brett bessell88 (1988)). We use the color $`C_{2.2/1.7}`$ instead of the photometric color $`HK`$, where $`C_{2.2/1.7}`$ is defined as:
$$C_{2.2/1.7}=\mathrm{log}(F_{2.2}/F_{1.7}).$$
(1)
$`F_{2.2}`$ and $`F_{1.7}`$ are the IRTS/NIRS fluxes at 2.2 and 1.7 $`\mu `$m in units of Jy, respectively. For the NIRS wavelength region, the 2.2 and 1.7 $`\mu `$m bands are least affected by stellar H<sub>2</sub>O absorption bands. Fig. 2 shows $`C_{2.2/1.7}`$ versus the spectral types from BSC and GCVS of all the known K- and M-giants (59 giants) in the sample. There is a clear increase in $`C_{2.2/1.7}`$ toward later spectral type except for Miras. One M9 star (KP~Lyr $`=`$ ADS 11423) deviates from this relation, but we regard this star as M5, according to Abt (abt (1988)).
All stars of spectral type M6 and later are above $`C_{2.2/1.7}=0.085`$. Thus, stars bluer than $`0.085`$ are expected to have spectral types earlier than M6. There are 67 stars in the $`<`$ M6 region, defined by $`C_{2.2/1.7}<0.085`$. $`C_{2.2/1.7}`$ for AK~Cap and V~Hor is $`0.130`$ and $`0.102`$, respectively. These numbers confirm that these 2 stars have spectral types earlier than M6 at the time of the IRTS observation, even though both stars show clear H<sub>2</sub>O absorption.
We now discuss the relationship between the H<sub>2</sub>O absorption strength and the spectral type. For this, we define the H<sub>2</sub>O index $`I_{\mathrm{H}_2\mathrm{O}}`$ as follows:
$$I_{\mathrm{H}_2\mathrm{O}}=\mathrm{log}(F_{\mathrm{cont}}/F_{1.9}),$$
(2)
where $`F_{\mathrm{cont}}`$ is the continuum flux level at 1.9 $`\mu `$m in units of Jy, which is evaluated by linear interpolation between $`F_{1.7}`$ and $`F_{2.2}`$. $`F_{1.9}`$ is the observed flux (Jy) at 1.9 $`\mu `$m. In Fig. 3, we plot $`I_{\mathrm{H}_2\mathrm{O}}`$ as a function of $`C_{2.2/1.7}`$. The dominant measurement errors for $`C_{2.2/1.7}`$ and $`I_{\mathrm{H}_2\mathrm{O}}`$ are due to by the slitless spectroscopy, and they are roughly 0.01 and 0.002 for stars $`<`$ M6, respectively.
Fig. 3 can be used to find candidate stars ($`<`$ M6) with H<sub>2</sub>O absorption. Since we estimated $`F_{\mathrm{cont}}`$ by linear interpolation, $`I_{\mathrm{H}_2\mathrm{O}}`$ is not zero even in the absence of H<sub>2</sub>O. We evaluate the relation between $`I_{\mathrm{H}_2\mathrm{O}}`$ and $`C_{2.2/1.7}`$ for stars without H<sub>2</sub>O, using a linear fit for the 67 stars in the $`<`$ M6 region, by minimizing a merit function $`_i^n|y_iabx_i|`$ for $`n`$ data points $`(x_i,y_i)`$ (Press et al. press (1986)). This fit is robust against outliers, i.e. stars with H<sub>2</sub>O. The result is indicated as a thick line. The thin lines indicate $`\pm 2\sigma `$ level from the fit, where $`\sigma `$ is standard deviation of $`I_{\mathrm{H}_2\mathrm{O}}`$. Here, we use the $`+2\sigma `$ level as a threshold to find candidate stars with H<sub>2</sub>O, because no star appears below the $`2\sigma `$ line. The stars above the $`+2\sigma `$ are supposed to be stars with H<sub>2</sub>O absorptions. There are 6 such stars in the region of $`<`$ M6. The value of $`2\sigma `$ is almost equal to that of 3 standard deviations after the outliers are excluded. The spectra of the stars are shown in Fig. 4. These 6 stars above the $`+2\sigma `$ line show clear evidence of H<sub>2</sub>O absorption bands at 1.9 and 2.7 $`\mu `$m. Only AK~Cap and V~Hor have identifications of spectral type in BSC or GCVS. On the basis of their $`C_{2.2/1.7}`$ (ranging from $`0.113`$ to $`0.086`$), the other four stars are probably M-type stars, and not K-type stars (see Fig. 2).
## 4 Discussion
In Fig. 5, we plot $`I_{\mathrm{H}_2\mathrm{O}}`$ versus the color defined by the NIRS 2.2 $`\mu `$m band flux and IRAS 12 $`\mu `$m flux ($`F_{12}`$) as:
$$C_{12/2.2}=\mathrm{log}(F_{12}/F_{2.2}).$$
(3)
$`C_{12/2.2}`$ is a measure of the IR excess due to circumstellar dust, and is roughly equivalent to $`K[12]`$, which is an indicator of mass-loss rate in Miras (Whitelock et al. whitelock (1994)). Fig. 5 shows boundary at $`C_{12/2.2}1.0`$ between early M-type stars with H<sub>2</sub>O and those without H<sub>2</sub>O (we regard all stars below the $`+2\sigma `$ line in Fig. 3 as early M-type stars without H<sub>2</sub>O). Furthermore, there is a clear correlation between $`I_{\mathrm{H}_2\mathrm{O}}`$ and $`C_{12/2.2}`$, which implies that the H<sub>2</sub>O absorption is related to the circumstellar dust emission. However, as we show below, the H<sub>2</sub>O molecules are not necessarily located in the circumstellar envelope.
We estimate the excitation temperature ($`T_{\mathrm{ex}}`$), and the column density ($`N`$) of H<sub>2</sub>O molecules in early M-type stars. The spectrum of a representative star AK~Cap is normalized with respect to the spectra of two early M-type stars, which have the same spectral types and similar near-IR color ($`C_{2.2/1.7}`$), but do not show H<sub>2</sub>O absorption ($`I_{\mathrm{H}_2\mathrm{O}}`$), or dust excess ($`C_{12/2.2}`$). The resulting normalized spectrum of AK~Cap is fitted by a simple plane parallel model with a uniform molecular layer assuming local thermodynamic equilibrium (LTE) (Fig. 6). The H<sub>2</sub>O line list is taken from Partridge & Schwenke (partridge (1997)). The turbulent velocity is assumed to be 3 $`\mathrm{km}\mathrm{s}^1`$. For AK~Cap, we obtain a reasonable fit for $`T_{\mathrm{ex}}1000`$$`1500`$K and $`N5\times 10^{19}`$ cm<sup>-2</sup>. A similar analysis for V~Hor (M5III) results in $`T_{\mathrm{ex}}1000`$$`1500`$K and $`N1\times 10^{20}`$ cm<sup>-2</sup>. If we assume that the H<sub>2</sub>O molecules were in the circumstellar envelope, the mass-loss rate obtained from these column densities would exceed $`10^6\mathrm{M}_{\mathrm{}}\mathrm{yr}^1`$ (assuming an abundance ratio of H<sub>2</sub>O/H<sub>2</sub> $`=8\times 10^4`$; Barlow et al. barlow (1996)), which is about a factor of 10–100 larger than those expected for Ls and SRs ($`10^7`$$`10^8\mathrm{M}_{\mathrm{}}\mathrm{yr}^1`$, Jura & Kleinmann jurab (1992)). Thus, the H<sub>2</sub>O molecules responsible for observed feature cannot be in the circumstellar shell.
In contrast, our measured values for $`T_{\mathrm{ex}}`$ and $`N`$ are in good agreement with results by Tsuji et al. (tsuji97 (1997)), who suggested that the H<sub>2</sub>O molecules in M-type stars are located in the layer above the photosphere. Our numbers $`T_{\mathrm{ex}}`$ are also consistent with $`T_{\mathrm{ex}}1150\pm 200`$ by Hinkle & Barnes (hinkle (1979)) for the ‘cool component’ of the Mira variable R~Leo, which is an overlaying layer of photosphere. Because the H<sub>2</sub>O molecules responsible for the near-IR absorption cannot be in the circumstellar shell, and because our results are consistent with Tsuji et al. (tsuji97 (1997)) and Hinkle & Barnes (hinkle (1979)), we conclude that they are in an outer atmosphere, i.e. the layer above the photosphere, but below the circumstellar envelope.
Numerical calculations of the atmospheres of Miras (e.g. Bowen bowen (1988); Bessell et al. bessell96 (1996)) show that the pulsation of the star extends the stellar atmosphere. Our observations show that such an extended region could also be present in some SRs and Ls. The dependence of H<sub>2</sub>O intensity on variable types as seen in Fig. 3 and 5 may result from differences in the physics of pulsation. Not all Ls and SRs show H<sub>2</sub>O absorption, which is not surprising, because of their complexity (e.g. Jura & Kleinmann jurab (1992)). Using the light curve of V~Hor (Mattei mattei (1998)), we confirm that V~Hor is probably an SR, although it shows a sudden increase of its visual magnitude before the IRTS observation. Unfortunately, no light curve is available for AK~Cap.
Hinkle & Barnes (hinkle (1979)) found that in Miras the H<sub>2</sub>O in an outer layer is responsible for the near-IR absorption, and we find the same situation in some early M-type stars. If we assume a constant abundance ratio of H<sub>2</sub>O$`/`$H<sub>2</sub>, then $`I_{\mathrm{H}_2\mathrm{O}}`$ is a measure of the total column density in the outer layer. Furthermore, $`C_{12/2.2}`$ is a measure of the amount of the hot circumstellar dust, if the circumstellar shells of the stars in Fig. 5 have a similar dust composition. Therefore, the total column density of the outer layer correlates with the thickness of the circumstellar shell. This may suggest that the outer layer influences the mass loss of the star.
In conclusion, we demonstrate that H<sub>2</sub>O absorption can be seen in early M-type stars, and that the H<sub>2</sub>O molecules are located in the outer atmosphere. The observed correlation between the intensity of the H<sub>2</sub>O absorption and the mid-infrared excess implies that the extended atmosphere is connected to the mass loss of the stars.
###### Acknowledgements.
The authors acknowledge Drs. M. Cohen and M. Noda for their efforts on the NIRS calibration. M.M. thanks the Research Fellowships of the Japan Society for the Promotion of Science for the Young Scientists. I.Y. acknowledges financial support from a NWO PIONIER grant. M.M.F. thanks Dr. H.A. Thronson at NASA Headquarters for discretionary funding, as well as the Center of Excellency of the Japanese Ministry of Education.
|
no-problem/9906/astro-ph9906502.html
|
ar5iv
|
text
|
# Young stars and non-stellar emission in the aligned radio galaxy 3C 2561footnote 11footnote 1Based in part on observations made with the NASA/ESA Hubble Space Telescope, obtained at the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy, Inc. under NASA contract NAS 5-26555, and on observations made with the W. M. Keck Observatory
## 1 Introduction
The radio galaxy 3C 256 ($`z=1.824`$; Dey et al. 1996, hereafter D96) was first identified by Spinrad & Djorgovski (1984), at which time it was the highest redshift galaxy known. It is one of the most luminous radio galaxies so far discovered ($`L_{178MHz}=3.5\times 10^{29}`$ W Hz<sup>-1</sup>; we adopt $`H_0=50`$ km s<sup>-1</sup> Mpc<sup>-1</sup>, $`q_0=0.1`$, and $`\mathrm{\Lambda }=0`$<sup>2</sup><sup>2</sup>2For this cosmology, the angular size scale at the distance of 3C 256 is 11.3 kpc arcsec<sup>-1</sup>, and the look-back time is 11.9 Gyr to a Universe that is 4.6 Gyr old.), and also one of the best examples of the alignment effect, i.e., the tendency for the optical and radio emission from high-redshift radio galaxies to share a common position angle (Chambers, Miley, & van Breugel 1987; McCarthy et al. 1987). However, unlike most luminous radio galaxies, where the alignment is much weaker in the near-infrared (Rigler et al. 1992; Dunlop & Peacock 1993), in 3C 256 the continuum shows a clear elongation along the radio position angle of $`140\mathrm{°}`$ even in the $`K`$-band (Eisenhardt & Dickinson 1992). The radio and optical sizes of 3C 256 are also very similar, with a 4″ (45 kpc) separation between the two radio lobes.
The nature of the alignment effect is not yet fully understood. Various mechanisms have been proposed, including star formation induced by the expanding radio source (e.g., Rees 1989; De Young 1989; Begelman & Cioffi 1989), scattering of an anisotropic continuum from the active nucleus (e.g., di Serego Alighieri et al. 1989; Tadhunter et al. 1989; Fabian 1989), inverse Compton scattering of microwave background photons (Daly 1992a,b) and nebular continuum emission (e.g., Dickson et al. 1995). All of these models explain why the alignment effect is normally observed to be stronger in the rest-frame ultraviolet, since the aligned continuum is much bluer than the evolved stellar population which is believed to emit most of the near-infrared light in these radio galaxies. Since 3C 256 displays such a strong alignment at $`K`$ ($`\lambda _{\mathrm{rest}}7800`$ Å), is unusually blue ($`RK=2.4`$; Eisenhardt & Dickinson 1992), and is underluminous in the $`K`$-band Hubble diagram ($`K19`$; Chokshi & Eisenhardt 1991), it may be an object which has yet to form a substantial population of evolved stars. Such arguments led Elston (1988) and Eisenhardt & Dickinson (1992) to suggest that 3C 256 is undergoing its first major burst of star formation, i.e., is a protogalaxy.
Imaging polarimetry observations by Jannuzi et al. (1995) have shown the spatially extended emission to be strongly polarized ($`P=11.7\pm 1.5`$% in a 3$`\stackrel{}{\mathrm{.}}`$6 aperture; $`17.6\pm 2.2`$% in a 7$`\stackrel{}{\mathrm{.}}`$8 aperture) with the electric vector perpendicular to the radio and optical axes of 3C 256. This result is naturally explained if the extended polarized radiation is produced by the scattering of emission from a source located near the center of 3C 256. Similar polarimetric observations of other radio galaxies have been used to support the scattering hypothesis for the alignment effect. Spectropolarimetry by D96 further supports this scenario in the case of 3C 256 by demonstrating that the spatially extended polarized radiation has a power law form consistent with the emission having been originally produced by an active nucleus ($`S_\nu \nu ^{1.1\pm 0.1}`$). These authors also suggested a protogalactic nature for 3C 256, proposing that the galaxy has yet to undergo the major burst(s) of star formation that will convert most of its mass into stars.
In this paper, we combine new and existing optical and near-infrared observations to identify and model the components of the spectral energy distribution of 3C 256. In §2, we present the new data obtained for this paper. In §3, we construct the rest-frame ultraviolet/optical spectral energy distribution of 3C 256 and discuss its properties and the salient morphological points. In §4, we identify the components necessary to explain the observed SED and morphology and estimate their strengths. In §5, we discuss the implications of our decomposition, and in §6, we provide a summary of our results.
## 2 Observations and reduction
### 2.1 Optical spectroscopy
3C 256 was observed with the double spectrograph (Oke & Gunn 1982) on the Hale 5-m telescope at Palomar Observatory over the course of 5 nights in January 1990 and April 1991. A 2″-wide slit was used, oriented along $`\mathrm{PA}=148\mathrm{°}`$ (i.e., along the optical major axis). The blue spectra were obtained with a 300 lines mm<sup>-1</sup> grating blazed at 4000 Å and the red channel employed a 316 lines mm<sup>-1</sup> grating blazed at 7500 ÅĖach exposure lasted 3000–4000 s and the object was recentered in the slit by offsetting from a nearby star every 2 hours. The spectra were reduced and calibrated in a standard manner. The flux calibration was based on observations of Ross 374, Feige 67 and BD +26°2606. All of the 1-D spectra (extracted along a 5$`\stackrel{}{\mathrm{.}}`$5 aperture) were combined, after scaling those observed through cirrus (present on 2 of the 5 nights) to match the spectra taken under photometric conditions. We present the final, reduced spectrum in Fig. 1.
### 2.2 Infrared spectroscopy
A low-resolution near-infrared spectrum was obtained on the night of 1997 Jun 15 using the Near-Infrared Camera (NIRC; Matthews & Soifer 1994) on the Keck I telescope. The 150 lines mm<sup>-1</sup> JH grism was used with a 0$`\stackrel{}{\mathrm{.}}`$7 wide slit, giving a resolution $`R80`$. The slit was oriented at $`\mathrm{PA}=112\mathrm{°}`$, which was the closest to the major axis that could be achieved while still being able to acquire a guide star. Individual 200 s exposures were obtained at five different positions along the slit, separated from each other by 5″. The process was then repeated, to obtain a total exposure time of 2000 s. Each set of five exposures was median-filtered and this median image subtracted from each individual exposure to provide first-order sky subtraction. More accurate subtraction of sky lines was performed on each exposure by fitting a quadratic function of position to each column of the image, which was then subtracted. The individual frames were registered and a spectrum extracted through a 3″ aperture.
Five exposures were also taken of the G5V star BD +32°2290 and reduced in a similar manner, and used to correct the 3C 256 spectrum for atmospheric absorption and provide flux calibration. Short exposure images in the $`J`$ and $`K`$ filters image were taken immediately before and after the spectroscopy to enable an accurate determination of the slit location. The reduced spectrum is presented in Fig. 2.
### 2.3 Optical imaging
A list of the ground-based imaging observations used in this paper is presented in Table 1; a number of these have been presented elsewhere (Spinrad & Djorgovski 1984; Le Fèvre et al. 1988). In addition, an image was taken with the Mayall 4-m telescope at Kitt Peak National Observatory though an interference filter with central wavelength 8800 Å and FWHM 550 Å, and a $`V`$-band image was also obtained at the Hale Telescope. Both sets of data were reduced in a similar manner, using standard techniques. Flux calibration was performed using observations of the field of M 92 (Christian et al. 1985; $`V`$-band image) or spectrophotometric standards from the list of Massey & Gronwall (1990; 8800 Å image). The best images taken through the $`B`$, $`V`$, $`R`$, and $`I`$ filters are shown in Fig. 3.
### 2.4 Infrared imaging
A number of new near-infrared images were also obtained for this paper, as listed in Table 1. These include $`J`$ and $`K`$-band images taken with NIRC on the Keck I telescope, $`J`$ and $`K`$ images taken with the infrared array camera IRIM (a $`62\times 58`$ InSb array with a plate scale of 0$`\stackrel{}{\mathrm{.}}`$39 pixel<sup>-1</sup>) on the Mayall Telescope, and an $`H`$-band image taken using the Prime Focus Infrared Camera (PFIRCAM; Jarrett et al. 1994) on the Hale Telescope. Observations and subsequent reduction were performed following standard ‘jittering’ methods. Flux calibration was determined from observations of photometric standards from the lists of Elias et al. (1982; Mayall data), Persson et al. (1998; Keck data), or Casali & Hawarden (1992; Hale data). The $`K`$-band image from the Mayall telescope is the coaddition of several mosaics taken on the nights of 1989 Apr 17, 1991 May 24, and 1992 Mar 22. The $`J`$ and $`K`$ images from Keck are presented in Fig. 3.
### 2.5 HST imaging
Initial HST imaging observations of 3C 256 were made with WF/PC in Cycle 1, but no useful and reliable morphological information could be extracted from them, due to a poor signal-to-noise ratio, lack of good flatfields, and the complexity of the pre-refurbishment point spread function. These data are therefore not discussed in this paper.
The source was re-observed with HST/WFPC2 in the F336W filter on 1996 May 22. Ly$`\alpha `$, at a wavelength of 3433 Å, lies close to the peak transmission of this filter and, since the observed equivalent width of the emission line is $`1500`$ Å and the width of the filter is only 381 Å, it dominates the observed flux. The source was placed on the WF3 chip, providing a scale of 0$`\stackrel{}{\mathrm{.}}`$100 pixel<sup>-1</sup>. Eight orbits were spent on the source, for a total of 21,400 s of integration. The data were processed through the STScI pipeline (Voit et al. 1997) and cosmic rays were rejected and the data combined in the normal manner using the iraf task crrej. In Fig. 4 we show an overlay of the F336W and $`K`$-band images and the 5 GHz radio map of D96.
## 3 Results
### 3.1 Spectroscopy
We have used the specfit software (Kriss 1994; available as part of iraf/stsdas) to measure line fluxes and equivalent widths from our spectra. We fitted a Gaussian line profile and linear continuum level to a 200 Å-wide region of the optical spectrum (1000 Å-wide for the infrared spectrum), centered on the emission line. In instances where two emission lines were separated by less than 100 Å (500 Å), we fitted the continuum and two Gaussians simultaneously to a region covering 100 Å (500 Å) on both sides of each line. It was confirmed in all cases that the flux of the best-fitting Gaussian was very close to that produced by a direct measurement of the flux above the best-fitting continuum level. The emission-line fluxes and rest-frame equivalent widths are presented in Table 2.
The power law which provides the best fit to our spectrum has a spectral index $`\alpha =1.66\pm 0.04`$ ($`S_\nu \nu ^\alpha `$), very much redder that the $`\alpha =1.1\pm 0.1`$ found by D96. However, if we restrict our wavelength range to that between the C IV and Mg II lines, common to both spectra, a much bluer spectral index ($`\alpha =1.29\pm 0.13`$) is found, consistent with D96’s value. This indicates a change in spectral shape near 1500 Å (rest-frame). We return to this point in the next section.
### 3.2 Spectral energy distribution
We co-registered the various images by computing a geometric transformation based on the locations of stars in the images, and constructed the spectral energy distribution (SED) of 3C 256 in the optical–near-infrared (rest-frame ultraviolet–optical) by making photometric measurements through a 4″ aperture, centered on the peak of the $`K`$-band emission. This aperture is large enough to encompass all the high surface brightness emission from the galaxy, and therefore errors introduced by poor registration of the images will be smaller than the photometric uncertainties.
We correct the observed broad-band fluxes for the presence of strong emission lines. We smooth the F336W image, which is dominated by Ly$`\alpha `$, to the resolution of our ground-based images and compare the fluxes in the spectroscopic and photometric apertures<sup>3</sup><sup>3</sup>3Recently-obtained HST F555W images confirm that the ultraviolet continuum shares the Ly$`\alpha `$ morphology (Jannuzi et al. 1999).. By assuming the other emission lines share the Ly$`\alpha `$ morphology, we can thus use their spectroscopic fluxes to estimate their contributions to the broad-band photometry. We correct the optical photometry and $`J`$-band photometry using the spectra of Figs 1 and 2, respectively. The $`H`$-band flux is corrected for the presence of the \[O I\] $`\lambda \lambda `$6300,6364 doublet assuming \[O I\] $`\lambda `$6300/\[O III\] $`\lambda 5007=0.03`$. No major emission lines lie within the $`K`$-band. We estimate the uncertainty in our aperture correction to be $`10`$%, based on the variations which result from different smoothing scales. However, this usually translates to a $`1`$–2% error in the line-corrected flux, much less than the random photometric uncertainty. The exception is the line-corrected flux from the F336W image ($`1.8\pm 0.7\mu `$Jy), whose large uncertainty is due to the high equivalent width of the Ly$`\alpha `$ line.
When constructing the SED, we usually take the weighted mean of fluxes measured through similar filters. However, there is a significant difference between the $`J`$-band filters used for the Mayall and Keck observations, in that the Keck filter extends to longer wavelengths, and includes the strong \[O III\] $`\lambda \lambda `$4959,5007 doublet. Even though these lines lie in a region of poor and uncertain atmospheric transmission, their equivalent width is so large that they contribute significantly to the measured flux. We therefore use the Mayall measurement of the $`J`$-band flux alone, whose bluer cutoff excludes these strong lines. We present the complete SED in Fig. 5.
The SED exhibits three major features which must be reproduced in any viable model. First, the optical–infrared color is very blue ($`VK=2.9`$); in fact 3C 256 is bluer even than a star forming dwarf Irr at this redshift. Secondly, longward of 1500 Å (rest-frame), the SED can be approximated by a power law with $`\alpha 1`$. Finally, there is a pronounced decrease in flux shortward of 1500 Å (seen most obviously in the spectrum rather than the $`\mathrm{F336W}B`$ continuum color), which we shall refer to as the “UV rollover”.
The importance of the UV rollover will become apparent later, and we spend some time here confirming its reality, since it is not seen in D96’s spectrum. In Fig. 6, we plot our spectrum and that of D96 (scaled to match ours redward of C IV since their spectrum was taken in non-photometric conditions through a slightly smaller aperture) in the wavelength region of interest. Both spectra agree very well around the C IV line and show a decrease in flux at $`\lambda _{\mathrm{rest}}1480`$ Å; however, D96’s rises again below $`1450`$ Å, and is $`50`$% higher than ours shortward of this wavelength. This rise produces a broad absorption feature which D96 interpret to be a BAL cloud seen in the scattered light spectrum. We reject the possibility that our spectrum is incorrectly flux calibrated, since the spectroscopic Ly$`\alpha `$ flux agrees well with that measured from images. The difference between the spectra may be due to aperture effects (D96 used a $`4\stackrel{}{\mathrm{.}}1\times 1\mathrm{}`$ aperture, compared to our $`5\stackrel{}{\mathrm{.}}5\times 2\mathrm{}`$), but this does not concern us since we are attempting to determine the overall SED, as measured in a 4″ aperture, rather than the detailed spatial distribution of individual components. Importantly, the spectrum of Spinrad et al. (1985) also shows the UV rollover, and rules out the possibility of differential atmospheric refraction as the cause of the rollover, since their observing technique positioned the major axis of the entrance aperture along the parallactic angle.
As with D96, we note the absence of any strong absorption lines in the spectrum, with the possible exception of a feature at 1720 Å (a blend of N IV, Si IV and Al II; e.g., Fanelli et al. 1992), although this is a marginal detection at best. Our typical signal-to-noise ratio of 4 per 2 Å pixel implies a rest-frame equivalent width detection limit of $`2`$ Å if the absorption lines are a few hundred km s<sup>-1</sup> in width.
### 3.3 Morphology
The HST F336W image reveals two strong emission peaks, separated by 1$`\stackrel{}{\mathrm{.}}`$3, which can be identified with components $`a`$ and $`b`$ from Le Fèvre et al. (1988). There is also a third peak, further to the NW, which is not seen in Le Fèvre et al.’s image but might plausibly be the cause of the extension of $`b`$, and some diffuse emission which contributes about 20% of the total flux. If the Ly$`\alpha `$ emission is produced by photoionization from a central source, it should be located within two oppositely-directed cones, whose common apex marks the location of the hidden active nucleus. From Fig. 4, we measure the half-opening angles of such cones to be 28°, and the projected axis of the cones to be at $`\mathrm{PA}=146\mathrm{°}`$, slightly offset from the radio axis position angle of 132° (D96).
Despite having rather poorer spatial resolution, our $`B`$ and $`V`$ images show a similar structure to Le Fèvre et al.’s data. The structure at $`K`$ is very different, however, with a central peak appearing between the Ly$`\alpha `$-emitting regions. The color of the central region of 3C 256 is $`RK=3.4`$, while the more extended emission has $`RK2.7`$. This difference should be considered a lower limit to the true color variation, as seeing effects will smear the more strongly-peaked $`K`$-band image, causing some redistribution of flux from the center to the outer regions.
In addition, there is an object located 4″ NW of 3C 256 along the radio axis. There appears to be low surface brightness emission connecting the companion to the radio galaxy, visible in both the $`J`$ and $`K`$ images, and so we speculate that the two are physically connected. The companion is only convincingly seen in the near-infrared images and, while there may be a marginal detection in the deep CFHT $`I`$-band image, the fringing precludes any photometry.
## 4 Analysis
Our goal in this paper is to disentangle the various components which contribute to the overall spectral energy distribution of 3C 256 by using our new imaging and spectroscopy data, together with published polarimetry. Although it would clearly be helpful to model distinct regions of the source separately, the F336W image indicates that significant structure occurs on a scale of about 0$`\stackrel{}{\mathrm{.}}`$5. Our seeing-limited ground-based observations therefore preclude such an analysis, and instead we investigate the integrated SED of 3C 256 through a 4″ aperture. All the fluxes used in this section are either measured directly through such an aperture, or have been scaled from the optical spectrum by a factor of 1.2, to account for the difference in aperture sizes.
### 4.1 Scattered light
The polarimetric properties of 3C 256 all suggest that the polarized radiation is produced by the scattering of emission from a central source (Jannuzi et al. 1995; D96), presumably non-stellar emission from the active nucleus. Jannuzi et al. had no clear preference for electrons or dust as the scattering particles, while D96 favored electrons. In addition to the arguments advanced by D96, the spectral index of the polarized flux ($`\alpha =1.1\pm 0.1`$; D96) is redder than the mean for radio-loud quasars (e.g., Baker & Hunstead 1995; Willott et al. 1998), which argues against the substantial bluening that dust scattering often produces. Also, theoretical models of dust scattering predict a fairly sharp change in polarization at $`\lambda 2000`$ Å (Kartje 1995; Manzini & di Serego Alighieri 1996), which is not seen in 3C 256 (D96). We therefore assume electron scattering throughout the remainder of this paper. Since this mechanism is achromatic, the scattered light spectrum will be the same as the spectrum of the central AGN seen by the scatterers (i.e., at angles close to the radio axis). We model this with the core-dominated quasar spectrum of Baker & Hunstead (1995), artificially “reddened” by 0.6 powers of $`\nu `$ to match the observed spectral index of the polarized flux (D96), and extrapolated as a power law longward of H$`\alpha `$.
We can estimate the fraction of scattered light by considering the lack of prominent broad lines in our spectrum. The broad permitted lines from the quasar nucleus should be scattered into our line of sight together with the non-stellar continuum. We fit the spectral region around C IV $`\lambda `$1549 with narrow and broad Gaussians and determine a $`3\sigma `$ upper limit of 25 Å to the rest-frame equivalent width of any broad line. Like the limits D96 found for Mg II $`\lambda `$2800, this is substantially lower than the mean value for the equivalent width of C IV in radio-loud quasars (e.g., Steidel & Sargent 1991; Baker & Hunstead 1995), but within the observed range. The absence of a broad emission line in our spectrum is therefore suggestive of the presence of non-scattered flux, but is certainly not conclusive. The lower signal-to-noise ratio in our spectrum at Ly$`\alpha `$ means that even a broad line with an equivalent width of 100 Å (cf., a mean of 70 Å; Baker & Hunstead 1995) is consistent with the data.
### 4.2 Nebular continuum
Broadly speaking, the overall SED is consistent with an $`\alpha 1`$ power law. Since this is also the spectral shape of the polarized flux, it appears possible to ascribe the total flux spectrum entirely to scattered light with a constant fractional polarization of $`11`$%. However, Dickson et al. (1995) have shown that nebular continuum emission can be very important in powerful radio galaxies, and it is essential that we consider its contribution to the ultraviolet flux of 3C 256.
After applying an aperture correction, we determine the H$`\beta `$ flux in our 4″ aperture to be $`(7.6\pm 2.3)\times 10^{19}`$ W m<sup>-2</sup>. This implies Ly$`\alpha `$/H$`\beta 9`$, a value typical of high-redshift radio galaxies (e.g., McCarthy, Elston, & Eisenhardt 1992). We assume $`N_{\mathrm{He}}/N_\mathrm{H}=0.1`$ and that 5% of the helium is in the form of He II, based on the inferred ratio He II $`\lambda `$1640/H$`\beta 1`$ and the recombination coefficients of Osterbrock (1989), although the result is rather insensitive to this value. Using the tabulated emission coefficients of Aller (1987) for $`T_\mathrm{e}=10^4`$ K, the nebular continuum emission is shown by the solid curve in Fig. 5. In order to see more clearly the significance of the nebular emission to the overall SED, in Fig. 7 we plot the SED with the nebular emission (assumed to be unreddened) removed. The shape is dramatically different, with a clear break appearing at $`\lambda _{\mathrm{rest}}4000`$ Å, which was previously obscured by the hydrogen recombination continuum.
Since this break is the most important evidence in favor of a substantial stellar population in 3C 256, we investigate how its strength is affected by different assumptions about the nebular continuum emission. Given the uncertainty in the H$`\beta `$ flux, we construct alternative SEDs where the nebular continuum has been reduced or increased by one-third (although note that the high observed \[O III\]/H$`\beta =22`$ ratio argues against our H$`\beta `$ flux being too high by a large amount). We also produce an SED where we have assumed that the observed ratio of Ly$`\alpha `$/H$`\beta 9`$ deviates from the low-density Case B recombination value of 23 (e.g., Ferland & Osterbrock 1985) due to a foreground dust screen, which also affects the observed nebular continuum. We use Pei’s (1992) Small Magellanic Cloud extinction law (the reason for this choice will become apparent later), although nearly identical results are obtained if we use Pei’s Galactic law or Calzetti’s (1997) empirical “recipe” which is based on observations of starburst galaxies. The implied reddening is $`E(BV)=0.13`$. All three alternative SEDs are also plotted in Fig. 7, where it can be seen that the variation caused by these different assumptions is comparable to the photometric uncertainties. Furthermore, the Balmer jump persists throughout all the different assumptions, as shown in Fig. 7. We therefore adopt our original assumptions throughout the remainder of this analysis.
### 4.3 A young stellar population
The pronounced jump in the continuum level near 4000 Å implies starlight is a significant contributor to the total SED, both above and below this jump. Any stellar population must be fairly young (identifying this jump as the Balmer jump, not the 4000 Å break), since neither the 2600 nor 2900 Å breaks displayed by more evolved stellar populations are observed in our spectrum (see also D96), and the continua on both sides of the break are quite blue. However, three features of the revised SED are inconsistent with a population of arbitrarily young (hot) stars, namely the strong Balmer jump, relatively flat UV continuum and the UV rollover. All three of these are typical of main sequence stars of late B or early A type (e.g., Fanelli et al. 1992). Such stars only dominate the UV flux after the $`100`$ Myr needed for hotter stars to complete their evolution.
While no stellar absorption features are detected in our spectrum, the measured upper limit of 2 Å is typical of the narrow blends seen in A-type main sequence stars (Fanelli et al. 1992). This is before considering the effects of dilution from the scattered and nebular components, and the masking of absorption features by the strong emission lines. The observations are therefore not at odds with the notion that stars contribute significantly to the ultraviolet flux.
We compare our observations with synthetic stellar spectra produced with the GISSEL96 code (Bruzual & Charlot 1993, 1999), adopting a Salpeter (1955) IMF which extends from 0.1–125 $`M_{}`$. The upper mass cutoff is a fairly critical parameter in determining the age of the stellar population, since it is possible to reach the same main sequence turnoff much more rapidly than 100 Myr if more massive stars are never formed. On the other hand, the lower mass cutoff is important only in determining the total mass of stars.
We estimate the age of the young stellar population using the observed $`8800J`$ and $`JH`$ colors, after correcting for the nebular continuum and scattered emission. In Fig. 8, these colors are compared with those of a stellar population formed in an instantaneous burst and seen at $`z=1.824`$. The age of the stellar population is estimated to be between 70 and 400 Myr for an instantaneous burst (effectively independent of the assumed metallicity), with a best-fitting value of 220 Myr. It is clear from Fig. 8 that the age of the stellar population is constrained almost entirely by the $`8800J`$ color (i.e., the strength of the Balmer jump) and, since this wavelength baseline is short, the results are fairly insensitive to the spectral shape of the scattered continuum. Even if the scattered radiation does deviate significantly from our adopted spectrum, a young stellar population is still required to fit the SED of 3C 256. This analysis also provides an estimate of the strength of the scattered component, and our conclusion that $`20`$% of the observed optical continuum flux is provided by this component is in line with the estimate of D96. Approximately half the rest-frame ultraviolet continuum therefore comes from starlight.
Although we assume an instantaneous burst, the uncertainty in this age gives some indication of the possible duration over which star formation could have occurred. Models where the star formation rate (SFR) follows an exponential decline cannot provide a good fit to the overall SED unless the characteristic time scale is small ($`20`$ Myr), since the UV rollover demands that most of the 1500 Å flux arises from stars older than about 100 Myr, and hence the present star formation rate must be low. Models where the SFR is constant for a finite period of time and then drops to zero are more successful, with there being little difference in the SEDs of a 200 Myr-old population whose stars formed instantaneously, and one in which they formed over a period of 50 Myr.
### 4.4 The red core
All three components we have so far discussed are too blue to explain the red color at the center of 3C 256, which is also associated with a change in the galaxy’s morphology. We therefore need to address the possibility of a fourth component.
The red central color and strong $`K`$-band peak can be simply explained by a compact red component superimposed on a spatially flat blue continuum. If we assume that the spatial color variations in this blue continuum are small (supported by the similarities between the F336W and F555W HST images), we can estimate the flux of the putative core component by scaling the $`R`$ image to match the extended flux in the $`K`$ image and subtracting it. The residual source has a flux of $`3\mu `$Jy, and is slightly resolved compared to the point spread function of a star in the image. The core cannot therefore be due to a lightly reddened quasar nucleus, as has been seen in other 3C radio galaxies (Simpson, Rawlings, & Lacy 1999, and references therein; see §5.3 for the implications of this result), and is presumably stellar in origin.
The core could either be intrinsically red, and dominated by old stars, or be reddened by dust. The presence of dust is suggested by the Ly$`\alpha `$ morphology, which has a brightness minimum at the location of the core. Rest-frame ultraviolet continuum images of other distant radio galaxies show similar morphologies, and Dickinson, Dey, & Spinrad (1995) have argued that the $`z=1.206`$ host galaxy of the radio source 3C 324 must suffer substantial reddening ($`A_V1`$) for it not to be seen in their WFPC2 image. On the other hand, if the companion object is associated with 3C 256, its red colors support an intrinsically red stellar population. This source is redder than 3C 256 by $`0.38\pm 0.11`$ mag in $`JK`$ and more than 1.2 mag in $`RK`$ (after correcting for the presence of emission lines in the radio galaxy).
We somewhat arbitrarily prefer the dust-lane model to explain the red color of the central regions of 3C 256. In attempting to model the SED, we assume that the underlying population has the same age as the blue population of the previous section. Since the observed morphology of 3C 256 indicates that this red component does not contribute significant flux shortward of the $`K`$-band, its nature and spectral shape have little impact on our determination of the strengths of the other components.
### 4.5 The four component model
Since we have been able to infer a substantial amount of information about the various components already, it is not difficult to find the four-component model which provides the best fit to the observations. The difficulty in objectively weighting the spectroscopic and photometric data forces us to determine the best fit by eye. However, we have also performed a grid search to minimize the $`\chi ^2`$ quality-of-fit statistic between the broad-band photometry and the model, and confirmed that this was not significantly better than our adopted fit. The characteristics of this fit are summarized in Table 3, and the fit is shown graphically in Fig. 9.
We determine a mass of $`6.6\times 10^{10}M_{}`$ for the UV-bright stellar component. This either requires a peak SFR of $`3000M_{}`$ yr<sup>-1</sup> for exponentially declining star formation, or a constant SFR of $`1300M_{}`$ yr<sup>-1</sup> for a period of $`50`$ Myr. Although high, we note for comparison that Dey et al. (1997) estimate that the SFR could be as high as $`1100M_{}`$ yr<sup>-1</sup> in the $`z=3.80`$ radio galaxy 4C 41.17.
### 4.6 The three component model
Although we argued above for a fourth component to explain the red core, we investigate whether a reddened, centrally-peaked, young stellar population can explain the structure of 3C 256. The dust responsible for this reddening would need to have an extinction law which is steep in the far UV (1200–1600 Å) and devoid of a 2200 Å bump. This rules out a Galactic extinction law, since the strength of the bump correlates with the steepness of the law in the far UV (Cardelli et al. 1989). We therefore consider the extinction law of the Small Magellanic Cloud (SMC) and adopt the parametrized SMC extinction law of Pei (1992).
The UV rollover and $`\alpha 1`$ power law can be reproduced with an $`\alpha 0`$ power law seen through $`A_V0.5`$ mag of extinction. The intrinsically bluer color obviously requires a younger stellar population than determined in the previous section, and Fig. 8 indicates that foreground reddening also lowers the age of the stellar population needed to fit the $`8800J`$ color to $`120`$ Myr. This scenario obviates the need for a fourth component to explain the red core, since the color of the reddened stars is $`RK=3.2`$, consistent with that observed in the central regions of the galaxy.
We find the best-fitting three-component model, again adopting a by-eye fit and confirming that it is not significantly worse than a $`\chi ^2`$-minimization fit to the broad-band photometry. The results of this fit are presented in Table 3 and Fig. 10. Our present data do not favor either model over the other, although it is hoped that HST imaging and polarimetry will enable more detailed modeling and a conclusion to be drawn. However, since both models require a substantial young stellar population, our conclusion that one is present in 3C 256 is not dependent on which is correct, but rather on the accuracy of our H$`\beta `$ flux measurement and the observed UV rollover.
## 5 Discussion
### 5.1 The alignment effect in 3C 256
In both of our models, half of the observed $`V`$-band light arises from either scattered light or nebular continuum. In schemes which attempt to unify extragalactic radio sources, both of these should both be aligned with the radio source, since the rest-frame ultraviolet radiation escapes in two oppositely-directed cones along the radio axis. We have observational evidence to support this, since both the polarized continuum (which traces the scattered quasar light) and the line emission (which traces the nebular continuum) are extended along the radio axis. The optical-radio alignment seen in 3C 256 is therefore naturally explained. Although the fraction of the total light at $`K`$ produced by the scattered and nebular emission is lower, at 35%, it is still high enough to produce a pronounced near-infrared alignment effect, due to the absence of a strong, unaligned red component such as an evolved stellar population.
While the aligned optical light is provided in approximately equal amounts by the nebular continuum and scattered light, the latter component produces the majority of the aligned $`K`$-band light. In fact, the scattered quasar light provides a larger fraction of the total light at $`K`$ than at $`V`$, and therefore the fractional $`K`$-band polarization should be larger, if the intrinsic polarization of the scattered light remains constant. Simply scaling by the fractional contributions from the scattered component, we predict that the $`K`$-band polarization should be about 16% (in a 4″ aperture), although the exact value is somewhat dependent on the specific geometry.
Even without the scattered emission, the nebular emission could produce a strong optical alignment by itself: it would contribute over 40% of the total $`V`$-band flux in the absence of the scattered light. This alignment should be most pronounced at an observed wavelength of $`1`$ µm, where the Balmer continuum is strongest. Unfortunately, our 8800 Å image has insufficient signal-to-noise to make a quantitative statement about the alignment strength, but it should be possible to confirm spectroscopically the importance of nebular emission in 3C 256. Blanketing from high-order Balmer lines causes the effective wavelength of the Balmer jump from the young stellar population to be longer than the Balmer limit at 3646 Å. As both Figs 9 and 10 show, this results in a deep (the observed equivalent width is $`140`$ Å), broad absorption feature around 3700 Å, which should be detectable with moderate-resolution spectroscopy ($`R300`$) even at a fairly low signal-to-noise ratio.
What then of the jet-induced star formation scenario? On a cursory inspection, this appeared quite favorable for 3C 256, due to the similar extents of the radio and optical emission, and the near-infrared alignment. However, the age of the stellar population, at 100–200 Myr is older than any plausible estimate for the age of the radio source, as we now demonstrate.
The separation of the radio lobes is 4$`\stackrel{}{\mathrm{.}}`$3, which means that the hotspots have propagated an average distance of $`24(\mathrm{sin}\theta )^1`$ kpc, where $`\theta `$ is the angle between the radio axis and the line of sight. This angle is likely to be large because no radio core is seen in D96’s radio map with a limit $`\mathrm{log}R<3.36`$ ($`R`$ is the ratio of core to extended radio luminosity at $`\nu _{\mathrm{rest}}=5`$ GHz; we use D96’s spectral indices for the lobes and assume a flat core spectrum to determine the $`K`$-corrections). This is much lower than typical values of $`\mathrm{log}R2.5`$ (e.g., Morganti et al. 1997; Simpson 1998), ruling out the Doppler boosting which occurs when the radio axis is close to the line of sight (e.g., Orr & Browne 1982). Estimates of the hotspot advance speed in powerful radio galaxies are typically $`v_{\mathrm{hs}}0.03`$–0.15$`c`$ (Liu, Pooley, & Riley 1992; Scheuer 1995), although the speed may be $`0.01c`$ in young sources still within their host galaxies (e.g., Fanti & Fanti 1994). A realistic age for the radio source is therefore
$$t1.6\times 10^6(v_{\mathrm{hs}}/0.05c)^1(\mathrm{sin}\theta )^1\mathrm{yr},$$
which is about two orders of magnitude lower than the age we have determined for the stellar population.
Of course, the ‘age’ of the stellar population is, in fact, merely a measure of the location of the main sequence turnoff, which the UV rollover and Balmer jump strength constrain to be late B/early A-type. The true age could therefore be much shorter if the IMF is biased against high-mass stars. Such a suggestion is, however, counter to claims made on theoretical grounds that the IMF should be biased against low-mass stars (Larson 1977). In addition, the spectrum of 4C 41.17, the most convincing case for jet-induced star formation, clearly shows features associated with hot, young stars (Dey et al. 1997) which are strong enough to have been detectable in our spectrum. The discrepancy in the ages of the stellar population and the radio source therefore demands that the trigger for the most recent episode of star formation must have come from elsewhere. It can be shown, following Scheuer & Williams (1968), that 100 Myr is sufficient time for relic lobes to have dissipated through synchrotron losses, thereby allowing the possibility of multiple episodes of radio source activity (e.g., Roettiger et al. 1994). On the other hand, Daly (1990,1992b) has shown that galactic rotation will disrupt any radio–optical alignment over such a timescale, and so this scenario appears unlikely.
One question we have not answered is whether the young stellar population in 3C 256 is aligned with the radio source. This is impossible to determine with images of different depths and resolutions, especially when the contribution from the aligned components is $`25`$%, even in the least-affected (J) band. Also, the presence of a dust lane perpendicular to the radio source, as we have postulated might be the cause of the red core, would produce an alignment even when the underlying population is not aligned. Multicolor images at $`0\stackrel{}{\mathrm{.}}1`$ resolution might help us to disentangle the structures of the various components and address this question.
### 5.2 The stellar mass of 3C 256
The fact that 3C 256 is underluminous in the $`K`$$`z`$ diagram (it is more than a magnitude fainter than the locus of other 3C radio galaxies; e.g., Eales et al. 1997) has been used to infer a protogalactic nature for the source, where it has yet to form the bulk of its eventual stellar mass. Our analysis lends further weight to this picture.
Best, Longair, & Röttgering (1998) find that the optical–infrared SEDs of 3CR radio galaxies can be fairly well-modeled by a flat-spectrum aligned component and an old stellar population which provides nearly all of the $`K`$-band light. On the other hand, about 30% of the $`K`$-band light in 3C 256 is non-stellar, and at least half of the remainder originates from young stars whose mass-to-light ratio is much lower than that of more evolved populations. Although we cannot rule out the presence of an additional stellar component which is so heavily reddened as to contribute very little flux, there is no evidence to support its existence in 3C 256, or its absence in other radio galaxies. It therefore appears that the stellar mass of 3C 256 is less than that of similar radio galaxies by a factor of a few.
We attempt to quantify this by calculating the stellar mass of our model SED, and comparing it with the masses derived by Best et al. (1998). Since these authors also used the Bruzual & Charlot spectral synthesis code, there will be no systematic differences introduced. For the four-component model, the mass of the young, blue stellar component is fairly well-determined ($`6.6\times 10^{10}M_{}`$), but the same is not true for the red component which is responsible for the near-infrared core. We modeled it earlier as a reddened version of the same young stellar population, and derived a similar mass as for the unreddened component, but this mass depends on the extinction, which is poorly constrained. In addition, the lack of spectral information means that it could instead be a population of older stars, with a higher mass-to-light ratio. The best-fitting unreddened “old” population is 1–1.5 Gyr old and has a mass of $`5`$$`6\times 10^{10}M_{}`$, slightly dependent on the duration over which the star formation took place. The total stellar mass is therefore very similar for both models of the red component. For the three-component model, the mass of the single stellar component is $`8.2\times 10^{10}M_{}`$.
These are lower limits to the total stellar mass for each model, since there may be additional starlight, either heavily reddened or from an old stellar population, which contributes very little visible light. For a short burst of star formation at $`z=\mathrm{}`$, a mass of $`7\times 10^{10}M_{}`$ is required to produce 1 $`\mu `$Jy of (observed-frame) $`K`$-band light. This mass increases by 70% for every magnitude of foreground visual extinction. On balance we doubt that the total stellar mass is likely to be greater than $`2\times 10^{11}M_{}`$, much less than the $`7\times 10^{11}M_{}`$ which is the median stellar mass for the lower-redshift 3C sample of Best et al. (1998; we modify their results to account for the difference in our adopted cosmologies), and more than a factor of two lower than any of their estimates.
Although our uncertainty as to the nature of the red core means we cannot confirm the putative protogalactic nature of 3C 256, the low inferred stellar mass compared to other radio galaxies does suggest that it is not yet mature. Since the observed polarization requires a substantial mass of gas (Jannuzi et al. 1995; D96), and even the gas mass needed to produce the observed line emission is large ($`6\times 10^{10}M_{}`$ for $`n_\mathrm{e}=5`$ cm<sup>-3</sup>), it is quite conceivable that 3C 256 could undergo further bursts of star formation. Alternatively, it could acquire more mass through mergers. The companion object may be a galaxy which is in the early stages of a merger with 3C 256. Although spectroscopic confirmation that the companion is at the same redshift as 3C 256 may not be feasible (it has $`R25`$), the red optical–infrared colors ($`RJ>2.0`$) suggest the presence of a spectral break around 1 µm, supporting a similar redshift. It is worth noting that the near-infrared colors of the companion are very similar to those of the red core (see Fig. 9), and may therefore be a coeval stellar population, if the red core is old, rather than reddened.
### 5.3 Implications for unification scenarios
We briefly discuss what implications the absence of an unresolved near-infrared core has for unification schemes. The two quasars in the 3CR-based catalogue of Laing, Riley & Longair (1983) with redshifts closest to 3C 256 both have $`V18.4`$. Assuming that 3C 256 would be similarly bright if its quasar nucleus were seen directly, a power law spectrum with $`\alpha 1`$ extended into the near-infrared implies a $`K`$-band flux of 0.6 mJy for the nucleus alone. We therefore require $`A_V12`$ mag of rest-frame extinction to produce a core with a flux below 1 $`\mu `$Jy. Most $`z1`$ radio galaxies have nuclear obscurations in excess of 15 mag (Simpson et al. 1999), and therefore the extinction required to hide the quasar nucleus from direct view at $`K`$ is not unreasonably large.
In both of our models, approximately one-fifth of the rest-frame UV light is non-stellar radiation scattered into our line of sight. We should therefore expect to see broad lines whose equivalent widths are one-fifth of their intrinsic values. The limits we determined in §4.1 are consistent with this fraction, even if the intrinsic equivalent widths are significantly higher than the average values for radio-loud quasars. A normal quasar-like central engine is therefore not ruled out by their absence.
## 6 Summary
We have modeled the spectral energy distribution of the $`z=1.824`$ radio galaxy 3C 256. Although the overall SED is consistent with a single power law, a clear break appears at $`\lambda _{\mathrm{rest}}4000`$ Å after subtraction of nebular continuum emission which is constrained by our infrared spectroscopy. Although no stellar absorption lines are seen in our spectrum, this break indicates that starlight is a major contributor to the overall SED. Our model includes starlight, nebular continuum emission, and scattered AGN light, and provides a good fit to the data. The dominant stellar population has an age of 100–200 Myr, and was formed in a fairly short burst with a peak star formation rate of $`1000M_{}`$ yr<sup>-1</sup>.
In our model, the alignment effect seen in 3C 256 is due to a combination of nebular emission and scattered light, with the latter dominating at longer wavelengths. The pronounced near-infrared alignment is due to the absence of a bright evolved stellar population. Simple arguments based on size show that the radio source must be younger than 100 Myr and therefore is unlikely to have influenced the most recent burst of star formation. We predict that 3C 256 should have substantial near-infrared polarization, estimating a value of about 16% in the $`K`$-band. We also expect it to show a strong absorption feature (the observed-frame equivalent width is 140 Å in our model) comprised of the Balmer edge at 3646 Å in emission, and blanketing from high order Balmer lines in the young stellar population.
We have estimated the total stellar mass of 3C 256 to be no more than $`2\times 10^{11}M_{}`$, rather lower than estimates for other 3C radio galaxies. Arguments based on the polarization properties suggest that there is a large mass of gas associated with 3C 256, and therefore further major bursts of star formation are possible.
The specific case of 3C 256 presented here indicates the complex nature of powerful radio galaxies and the importance of obtaining data spanning as large a wavelength baseline as possible. In a future paper, we shall present the results of HST imaging polarimetry in the F555W filter, which should help us to understand more about this interesting source.
This work has been supported by the National Aeronautics and Space Administration, through HST grants GO-2698 and GO-5925 from the Space Telescope Science Institute. B. T. J. acknowledges support for his research from the National Science Foundation through their cooperative agreement with AURA, Inc. for the operation of the National Optical Astronomy Observatories. The W. M. Keck Observatory is operated as a scientific partnership among the California Institute of Technology, the University of California and the National Aeronautics and Space Administration. The Observatory was made possible by the generous financial support of the W.M. Keck Foundation. Parts of this work were performed at the Jet Propulsion Laboratory, California Institute of Technology under a contract with NASA. We thank Arjun Dey and Olivier Le Fèvre for supplying us with some of the data used in this paper. C. S. wishes to thank Clive Tadhunter for an invaluable discussion which helped to direct this work.
|
no-problem/9906/hep-ph9906555.html
|
ar5iv
|
text
|
# I Introduction
## I Introduction
Gauge field theories exhibit many different patterns of infrared behavior. During the past few years, there has been much progress in understanding the possibilities in supersymmetric gauge theories . For non-supersymmetric gauge theories, less is known, but it is expected that the infrared behavior will vary according to the number of massless fermions ($`N_f`$) coupled to the gauge fields. For a vector-like theory such as QCD, it is known that for low values of $`N_f`$, the theory confines and chiral symmetry breaking occurs. On the other hand, for large $`N_f`$ the theory loses asymptotic freedom. In between, there is a conformal window where the theory does not confine, chiral symmetry is restored, and the theory acquires a long range conformal symmetry. It has been proposed that for an $`SU(N)`$ gauge theory, there is a transition from the confining, chirally broken theory to the chirally symmetric theory at $`N_f4N`$ . Recent lattice simulations, however, seem to indicate that the amount of chiral symmetry breaking decreases substantially (for $`N=3`$) when $`N_f`$ is only about $`4`$.
Assuming that a single transition takes place at some critical value of $`N_f`$, we can ask questions about the spectrum of the theory near the transition. In Reference , it was argued by studying Weinberg spectral function sum rules that for near-critical theories parity partners become more degenerate than in QCD-like theories. This leads naturally to the idea that parity doublets might form as chiral symmetry is being restored. Lattice studies also indicate such a possibility .
In this paper we observe using an effective-Lagrangian as a guide, that the formation of degenerate parity partners is associated with the appearance of an enhanced global symmetry in the spectrum of states. We also note that this new symmetry could play a key role in describing a possible strong electroweak Higgs sector. Whether the new symmetry can be shown to emerge dynamically from an underlying gauge theory with $`N_f`$ near a critical value remains an open question.
It is worth noting that there exist examples of extra symmetries, not manifestly present in the underlying theory, but dynamically generated at low energies. For instance, by using duality arguments, it has been argued that a supersymmetric $`SU(2)`$ gauge theory with $`N_f`$ matter fields and global symmetry $`SU(2N_f)`$ is dual to a $`SU(N_f2)`$ gauge theory with $`N_f`$ matter fields. For $`N_f5`$, the ultraviolet flavor symmetry of the latter theory is $`SU_L(N_f)\times SU_R(N_f)\times U_B(1)`$. Since its infrared global symmetry must be $`SU(2N_f)`$ (that of the dual), its infrared symmetry is enhanced.
In Section II we discuss the appearance of enhanced global symmetry. Confinement is assumed and the symmetry of the underlying gauge theory, $`SU_L(N_f)\times SU_R(N_f)`$, is built into an effective Lagrangian describing the physical states of the theory. Parity invariance is imposed and the usual pattern of chiral symmetry breaking ($`SU_L(N_f)\times SU_R(N_f)SU_V(N_f)`$) is assumed. The $`N_f^21`$ Goldstone bosons appear together with scalar chiral partners. We augment the spectrum with a set of vector fields for both the $`SU_L(N_f)`$ and $`SU_R(N_f)`$ symmetry groups. The Lagrangian thus takes the form of a linear sigma model coupled to vectors. It could be expanded to include fields corresponding to other states as well. The natural mass scale of this strongly interacting system is expected to be of order $`2\pi v`$, where $`v`$ is the vacuum expectation value.
We examine the spectrum and recognize that there is a particular choice of the parameters that allows for a degenerate vector and axial-vector, while enlarging the global symmetry to include an additional (unbroken) $`SU_L(N_f)\times SU_R(N_f)`$. This happens as the spectrum of the theory splits into two sectors with one displaying the additional symmetry. We then briefly review the arguments (see Ref. ) that an underlying near-critical $`SU(N)`$ gauge theory might naturally lead to a more degenerate vector-axial spectrum than in QCD, and to an enhanced symmetry. Finally we note that even a discrete additional symmetry, $`Z_{2L}\times Z_{2R}`$, of the effective theory is adequate to insure the mass degeneracy of the vector and axial vector<sup>§</sup><sup>§</sup>§We thank Noriaki Kitazawa for suggesting the possibility of a discrete symmetry..
The possible appearance of an additional, continuous symmetry was considered by Casalbuoni et al in Refs.. These papers were restricted to the case $`N_f=2`$ and did not include discussion of the possible connection to a near-critical underlying theory. The treatment in Ref. made use of a nonlinear realization for the Goldstone degrees of freedom, using hidden gauge symmetry methods . We could generate the effective Lagrangian of Ref. by integrating out the massive scalar degrees of freedom, but that would keep some massive degrees of freedom (the vector fields) and neglect others. When we focus on low energy consequences (in Section III), we will integrate out all the massive degrees of freedom leading to the electroweak chiral Lagrangian. The treatment of Ref. utilized a linear realization for the scalars and focused on the decoupling of the vectors as they are made heavy relative to the weak scale. We do not take this limit here since we assume the vector and scalar masses to be of the same order.
In Section III we embed the electroweak gauge group within the global symmetry group. We observe that the enhanced symmetry of the strongly interacting sector, which now provides electroweak symmetry breaking, plays an important role. The additional symmetry is a partial custodial symmetry for the electroweak $`S`$ parameter, in the sense that the parity doubled part of the strong sector, by itself, makes no contribution to $`S`$. This is shown by integrating out the massive physics to construct the terms in the low energy electroweak chiral Lagrangian. The $`S`$ parameter corresponds to one such term.
We extend the study to fermions in a pseudoreal representation of the underlying gauge group in Section IV. In this case parity is automatically enforced. The pseudoreal representations allow for the lowest number of colors (i.e. $`N=2`$) and consequently for the lowest possible number of flavors for which the theory might show a dynamically enhanced symmetry. The enhanced global symmetry is $`\left[SU(2N_f)\right]^2`$ spontaneously broken to $`Sp(2N_f)\times SU(2N_f)`$.
In Section V we conclude and suggest some directions for future work. In Appendix A we provide an explicit representation for the $`Sp(4)`$ generators.
## II Effective Lagrangian for $`SU_L(N_f)\times SU_R(N_f)`$ global symmetry
To discuss the possible appearance of enhanced symmetry in a strongly interacting spectrum, some description of the spectrum is needed. We will find it helpful to use an effective Lagrangian possessing $`SU_L(N_f)\times SU_R(N_f)`$ symmetry, the global invariance of the underlying gauge theory. We assume that chiral symmetry is broken according to the standard pattern $`SU_L(N_f)\times SU_R(N_f)SU_V(N_f)`$. The $`N_f^21`$ Goldstone bosons are encoded in the $`N_f\times N_f`$ real traceless matrix $`\mathrm{\Phi }_j^i`$ with $`i,j=1,\mathrm{},N_f`$ . The complex matrix $`M=S+i\mathrm{\Phi }`$ describes both the Goldstone bosons as well as associated scalar partners $`S`$. It transforms linearly under a chiral rotation:
$$Mu_LMu_R^{},$$
(1)
with $`u_{L/R}`$ in $`SU_{L/R}(N_f)`$.
To augment the massive spectrum, we introduce vector and axial vector fields following a method outlined in Ref. . We first formally gauge the global chiral group introducing the covariant derivative
$$D^\mu M=^\mu Mi\stackrel{~}{g}A_L^\mu M+i\stackrel{~}{g}MA_R^\mu ,$$
(2)
where $`A_{L/R}^\mu =A_{L/R}^{\mu ,a}T^a`$ and $`T^a`$ are the generators of $`SU(N_f)`$, with $`a=1,\mathrm{},N_f^21`$ and $`\mathrm{Tr}\left[T^aT^b\right]={\displaystyle \frac{1}{2}}\delta ^{ab}`$. The left and right couplings are the same since we assume parity invariance. Under a chiral transformation
$$A_{L/R}^\mu =u_{L/R}A_L^\mu u_{L/R}^{}\frac{i}{\stackrel{~}{g}}^\mu u_{L/R}u_{L/R}^{}.$$
(3)
The effective Lagrangian needs only to be invariant under global chiral transformations. Including terms only up to mass dimension four, it may be written in the form
$`L=`$ $`{\displaystyle \frac{1}{2}}\mathrm{Tr}\left[D_\mu MD^\mu M^{}\right]+m^2\mathrm{Tr}\left[A_{L\mu }A_L^\mu +A_{R\mu }A_R^\mu \right]`$ (4)
$`+`$ $`h\mathrm{Tr}\left[A_{L\mu }MA_R^\mu M^{}\right]+r\mathrm{Tr}\left[A_{L\mu }A_L^\mu MM^{}+A_{R\mu }A_R^\mu M^{}M\right]`$ (5)
$`+`$ $`i{\displaystyle \frac{s}{2}}\mathrm{Tr}\left[A_{L\mu }\left(MD^\mu M^{}D^\mu MM^{}\right)+A_{R\mu }\left(M^{}D^\mu MD^\mu M^{}M\right)\right].`$ (6)
The parameters $`h,r`$ and $`s`$ are dimensionless real parameters, while $`m^2`$ is a common mass term. To this, we may add a kinetic term for the vector fields
$$L_{\mathrm{Kin}}=\frac{1}{2}\mathrm{Tr}\left[F_{L\mu \nu }F_L^{\mu \nu }+F_{R\mu \nu }F_R^{\mu \nu }\right],$$
(7)
where
$$F_{L/R}^{\mu \nu }=^\mu A_{L/R}^\nu ^\nu A_{L/R}^\mu i\stackrel{~}{g}[A_{L/R}^\mu ,A_{L/R}^\nu ],$$
(8)
along with vector-interaction terms respecting only the global symmetry. Finally, we may add the double trace term,
$$\mathrm{Tr}\left[MM^{}\right]\mathrm{Tr}\left[A_L^2+A_R^2\right],$$
(9)
at the dimension-four level. To arrange for symmetry breaking, a potential $`V(M,M^{})`$ must be added. When the effective Lagrangian is extended to the dimension-six level and higher, many new operators enter.
Parity is also a symmetry and it acts on the fields according to
$`PM(𝒙)(P)^1`$ $`=`$ $`M^{}(𝒙),`$ (10)
$`PA_{L/R}^\mu (𝒙)(P)^1`$ $`=`$ $`ϵ(\mu )A_{R/L}^\mu (𝒙),`$ (11)
where $`ϵ(\mu )=1`$ for $`\mu =0`$ and $`1`$ for $`\mu =1,2,3`$.
The spectrum described by this effective Lagrangian consists of Goldstone bosons, a set of scalars, and massive vector and axial vectors. With its massive vectors and axial vectors, it is of course not renormalizable, but it can nevertheless provide a reasonable description of low-lying states. (It is worth noting that a Lagrangian of this type does this for the low-lying QCD resonances ). While it cannot be a complete description of the hadronic spectrum, it has sufficient content to guide a general discussion of enhanced symmetries.
Keeping only terms quadratic in the fields and temporarily neglecting the massive scalars, the Lagrangian Eq. (6) takes the form
$`L=`$ $`{\displaystyle \frac{1}{2}}\mathrm{Tr}\left[_\mu \mathrm{\Phi }^\mu \mathrm{\Phi }\right]+\sqrt{2}(s\stackrel{~}{g})v\mathrm{Tr}\left[_\mu \mathrm{\Phi }A^\mu \right]+M_A^2\mathrm{Tr}\left[A_\mu A^\mu \right]+M_V^2\mathrm{Tr}\left[V_\mu V^\mu \right],`$ (12)
where $`M=v+i\mathrm{\Phi }`$, $`v`$ is the vacuum expectation value and we have defined the new vector fields
$$V=\frac{A_L+A_R}{\sqrt{2}},A=\frac{A_LA_R}{\sqrt{2}}.$$
(13)
The vector and axial masses are related to the effective Lagrangian parameters via
$`M_A^2`$ $`=`$ $`m^2+v^2\left[r+\stackrel{~}{g}^22s\stackrel{~}{g}{\displaystyle \frac{h}{2}}\right],`$ (14)
$`M_V^2`$ $`=`$ $`m^2+v^2\left[r+{\displaystyle \frac{h}{2}}\right]`$ (15)
where the contribution from Eq. (9) has been absorbed into $`m^2`$. The terms proportional to $`v^2`$ are Higgs-like contributions, arising from the spontaneous breaking.
The second term in Eq. (12) mixes the axial vector with the Goldstone bosons. This kinetic mixing may be diagonalized away by the field redefinition
$$AA+v\frac{\stackrel{~}{g}s}{\sqrt{2}M_A^2}\mathrm{\Phi },$$
(16)
leaving the mass spectrum unchanged . The vector-axial vector mass difference is given by
$$M_A^2M_V^2=v^2\left[\stackrel{~}{g}^22\stackrel{~}{g}sh\right].$$
(17)
In QCD this difference is known experimentally to be positive, a fact that can be understood by examining the Weinberg spectral function sum rules (see Ref. and references therein). The effective Lagrangian description is of course unrestrictive. Depending on the values of the $`\stackrel{~}{g}`$, $`s`$ and $`h`$ parameters, one can have a degenerate or even inverted mass spectrum.
What kind of underlying gauge theory might provide a degenerate or inverted spectrum? Clearly, it has to be different from QCD, allowing for a modification of the spectral function sum rules. In Reference , an $`SU(N)`$ gauge theory (with $`N>2`$) and $`N_f`$ flavors was considered. If $`N_f`$ is large enough but below $`11N/2`$, an infrared fixed point of the gauge coupling $`\alpha _{}`$ exists, determined by the first two terms in the $`\beta `$ function. For $`N_f`$ near $`11N/2`$, $`\alpha _{}`$ is small and the global symmetry group remains unbroken. For small $`N_f`$, on the other hand, the chiral symmetry group $`SU_L(N_f)SU_R(N_f)`$ breaks to its diagonal subgroup. One possibility is that the transition out of the broken phase takes place at a relatively large value of $`N_f/N(4)`$, corresponding to a relatively weak infrared fixed point . An alternate possibility is that the transition takes place in the strong coupling regime, corresponding to a small value of $`N_f/N`$ . The larger value emerges from the renormalization group improved gap equation, as well as from instanton effects , and saturates a recently conjectured upper limit . It corresponds to the perturbative infrared fixed point $`\alpha _{}`$ reaching a certain critical value $`\alpha _c`$. A similar result has also been obtained by using a suitable effective Lagrangian .
These studies also suggest that the order parameter, for example the Goldstone boson decay constant $`F_\pi v`$, vanishes continuously at the transition relative to the intrinsic renormalization scale $`\mathrm{\Lambda }`$ of the gauge theory. In the broken phase near the transition, the fact that one is approaching a phase with long range conformal symmetry suggests that all massive states scale to zero with the order parameter relative to $`\mathrm{\Lambda }`$ .
In Reference the spectrum of states in the broken phase near a large-$`N_f/N`$ transition was investigated using the spectral function sum rules. It was shown that the ordering pattern for vector-axial hadronic states need not be the same as in QCD-like theories (small $`N_f/N`$). The crucial ingredient is that these theories contain an extended ”conformal region” extending from roughly $`2\pi F_\pi `$ to the scale $`\mathrm{\Lambda }`$ where asymptotic freedom sets in. In this region, the coupling remains close to an approximate infrared fixed point and the theory has an approximate long range conformal symmetry. It was argued that this leads to a reduced vector-axial mass splitting, compared to QCD-like theories. This suggests the interesting possibility that parity doublets begin to form as chiral symmetry is being restored. That is, the vector-axial mass ratio approaches unity as the masses decrease relative to $`\mathrm{\Lambda }`$. Lattice results seem to provide supporting evidence for such a possibility , although at smaller values of $`N_f/N`$.
If a parity doubled spectrum does appear, it is natural to expect it to be associated with some new global symmetry. While we have not demonstrated the appearance of a new global symmetry using the underlying degrees of freedom, we can explore aspects of parity doubling at the effective Lagrangian level. Returning to this description, we note that vector-axial parity doubling corresponds to the parameter choice (see Eq. (17)),
$$\stackrel{~}{g}^2=2\stackrel{~}{g}s+h.$$
(18)
This condition does not yet reveal an additional symmetry and therefore there is no reason to expect parity degeneracy to be stable in the presence of quantum corrections and the many higher dimensional operators that can be added to the effective Lagrangian in Eq. (6).
However, for the special choice $`s=\stackrel{~}{g}`$, $`r=\stackrel{~}{g}^2/2`$ and $`h=\stackrel{~}{g}^2`$, the effective Lagrangian acquires a new continuous global symmetry that protects the vector-axial mass difference. The effective Lagrangian at the dimension-four level takes the simple form
$$L=\frac{1}{2}\mathrm{Tr}\left[_\mu M^\mu M^{}\right]+m^2\mathrm{Tr}\left[A_{L\mu }A_L^\mu +A_{R\mu }A_R^\mu \right],$$
(19)
along with vector kinetic and interaction terms, the interaction term Eq. (9), and the symmetry breaking potential $`V(M,M^{})`$. The theory now has two sectors, with the vector and axial vector having their own unbroken global $`SU_L(N_f)\times SU_R(N_f)`$. The two sectors interact only through the product of singlet operators. The full global symmetry is $`\left[SU_L(N_f)\times SU_R(N_f)\right]^2\times U_V(1)`$ spontaneously broken to $`SU_V(N_f)\times U_V(1)\times \left[SU_L(N_f)\times SU_R(N_f)\right]`$. The vector and axial vector become stable due to the emergence of a new conservation law. This enhanced symmetry would become exact only in the chiral limit. For finite but small (relative to $`\mathrm{\Lambda }`$) values of the mass scales in Eq. (19), there are additional, smaller terms giving smaller mass splittings and small width-to-mass ratios.
It is of course a simple observation that a new symmetry and conservation law emerge if a theory is split into two sectors by setting certain combinations of parameters to zero. But here we were led to this possibility by looking for a symmetry basis for the parity doubling that has been hinted at by analyses of the underlying gauge theory. Although we have used a relatively simple effective Lagrangian, we anticipate that the conclusion is true in general, that is, that parity doublets form in the spectrum of a strongly interacting theory with chiral symmetry breaking only if the spectrum splits into two sectors, one exhibiting the spontaneous breaking and the other, parity doubled, sector exhibiting an unbroken additional symmetry.
We next observe that along with the additional global symmetry $`SU_L(N_f)\times SU_R(N_f)`$, the effective Lagrangian Eq. (19) possesses a discrete $`Z_{2L}\times Z_{2R}`$ symmetry. Under $`Z_{2L}\times Z_{2R}`$ the vector fields transform according to
$$A_Lz_LA_L,A_Rz_RA_R,$$
(20)
with $`z_{L/R}=1,1`$ and $`z_{L/R}Z_{2L/R}`$. Actually, the discrete symmetry alone is enough to insure vector-axial mass degeneracy and stability against decay. In that case, additional interaction terms, such as the single trace term
$$r\mathrm{Tr}\left[A_{\mu L}A_L^\mu MM^{}+A_{\mu R}A_R^\mu M^{}M\right],$$
(21)
are allowed, but degeneracy and stability are still insured. Of course, trilinear vector interactions will not respect this discrete symmetry. Nevertheless, one can not rule out the possibility that it is only this smaller, discrete symmetry that appears as an effective infrared symmetry of an underlying gauge theory near the chiral/conformal transition.
From the point of view of the underlying theory, the appearance of any additional symmetry in the spectrum, at criticality, would seem mysterious. The composite degrees of freedom in both sectors are made of the same fundamental fermions with a single underlying $`SU_L(N_f)\times SU_R(N_f)`$ symmetry. If the symmetry of the parity doubled sector is an unbroken $`SU_L(N_f)\times SU_R(N_f)`$, it would look as though the chiral symmetry is being realized there in the Wigner-Weyl mode. If that is the case, chiral dynamics would have to be influenced by confinement and bound state formation in an interesting new way. Whether a near-critical gauge theory can lead to this behavior is an unresolved question.
## III Strongly Interacting Electroweak Sector
We next discuss the consequences of enhanced symmetry for a strong symmetry breaking sector of the standard electroweak theory, embedding the $`SU_L(2)\times U_Y(1)`$ gauge symmetry in the global $`SU_L(N_f)\times SU_R(N_f)`$ chiral group. In this Section, for simplicity, we will restrict attention to the $`SU_L(2)\times SU_R(2)`$ subgroup of the full global group . The electroweak gauge transformation then takes the form
$$Mu_WMu_Y^{},$$
(22)
where $`M`$ is now a $`2\times 2`$ matrix which can be written as $`M={\displaystyle \frac{1}{\sqrt{2}}}\left[\sigma +i\stackrel{}{\tau }\stackrel{}{\pi }\right]`$ , where $`u_W=u_L=\mathrm{exp}\left({\displaystyle \frac{i}{2}}ϵ^a\tau ^a\right)`$ with $`\tau ^a`$ the Pauli matrices, and where $`u_Y=\mathrm{exp}\left({\displaystyle \frac{i}{2}}ϵ_0\tau ^3\right)`$. The weak vector boson fields transform as
$`W^\mu `$ $``$ $`u_LW^\mu u_L^{}{\displaystyle \frac{i}{g}}^\mu u_Lu_L^{},`$ (23)
$`B^\mu `$ $``$ $`u_YB^\mu u_Y^{}{\displaystyle \frac{i}{g^{}}}^\mu u_Yu_Y^{},`$ (24)
where $`g`$ and $`g^{}`$ are the standard electroweak coupling constants, $`W_\mu =W_\mu ^a{\displaystyle \frac{\tau ^a}{2}}`$ and $`B_\mu =B_\mu {\displaystyle \frac{\tau ^3}{2}}`$.
A convenient method of coupling the electroweak gauge fields to the globally invariant effective Lagrangian of Section II is to introduce a covariant derivative, which includes the $`W`$ and $`B`$ fields as well as the strong vector and axial-vector fields,
$$\mathrm{D}^\mu M=^\mu MigW^\mu M+ig^{}MB^\mu i\stackrel{~}{g}cC_L^\mu M+i\stackrel{~}{g}c^{}MC_R^\mu ,$$
(25)
where we have defined the new vector fields
$$C_L^\mu =A_L^\mu \frac{g}{\stackrel{~}{g}}W^\mu ,C_R^\mu =A_R^\mu \frac{g^{}}{\stackrel{~}{g}}B^\mu ,$$
(26)
and where $`c`$ and $`c^{}`$ are arbitrary real constants. Since the $`A_{L/R}^\mu `$ transform as Eq. (3), the $`C_{L/R}^\mu `$ transform under the electroweak transformations as
$$C_L^\mu u_LC_L^\mu u_L^{},C_R^\mu u_YC_R^\mu u_Y^{}.$$
(27)
By requiring invariance under the parity operation exchanging the labels $`LR`$ we have the extra condition $`c=c^{}`$.
The effective Lagrangian is constructed to be invariant under a local $`SU_L(2)\times U_Y(1)`$ as well as $`CP`$. The $`CP`$ transformation properties of the fieldsHere we summarize the $`CP`$ field transformations: $`CPM(𝒙)(CP)^1`$ $`=`$ $`\eta M^{}(\text{ }𝒙),`$ (28) $`CPA_{L/R\mu }(𝒙)(CP)^1`$ $`=`$ $`A_{L/R}^\mu (𝒙),`$ (29) $`CPW_\mu (𝒙)(CP)^1`$ $`=`$ $`W^\mu (𝒙),`$ (30) $`CPB_\mu (𝒙)(CP)^1`$ $`=`$ $`B^\mu (𝒙),`$ (31) where $`\eta `$ is an arbitrary $`C`$ phase. insure that the covariant derivative transforms as $`M`$, i.e.
$$CP\mathrm{D}_\mu M(𝒙)(CP)^1=\eta \left(\mathrm{D}^\mu M(𝒙)\right)^{}.$$
(32)
The effective Lagrangian is then obtained by replacing in Eq. (6) the covariant derivative with the new one in Eq. (25). To make the theory electroweak gauge invariant, one substitutes the $`A_{L/R}`$ with the $`C_{L/R}`$, giving, through dimension four,
$`L=`$ $`{\displaystyle \frac{1}{2}}\mathrm{Tr}\left[\mathrm{D}_\mu M\mathrm{D}^\mu M^{}\right]+m^2\mathrm{Tr}\left[C_{L\mu }C_L^\mu +C_{R\mu }C_R^\mu \right]`$ (33)
$`+`$ $`h\mathrm{Tr}\left[C_{L\mu }MC_R^\mu M^{}\right]+r\mathrm{Tr}\left[C_{L\mu }C_L^\mu MM^{}+C_{R\mu }C_R^\mu M^{}M\right]`$ (34)
$`+`$ $`i{\displaystyle \frac{s}{2}}\mathrm{Tr}\left[C_{L\mu }\left(M\mathrm{D}^\mu M^{}\mathrm{D}^\mu MM^{}\right)+C_{R\mu }\left(M^{}\mathrm{D}^\mu M\mathrm{D}^\mu M^{}M\right)\right].`$ (35)
To this we add a kinetic term
$$L_{\mathrm{Kin}}=\frac{1}{2}\mathrm{Tr}\left[F_{L\mu \nu }F_L^{\mu \nu }+F_{R\mu \nu }F_R^{\mu \nu }\right]\frac{1}{2}\mathrm{Tr}\left[W_{\mu \nu }W^{\mu \nu }\right]\frac{1}{2}\mathrm{Tr}\left[B_{\mu \nu }B^{\mu \nu }\right],$$
(36)
where
$`W_{\mu \nu }`$ $`=`$ $`_\mu W_\nu _\nu W_\mu ig[W_\mu ,W_\nu ],`$ (37)
$`B_{\mu \nu }`$ $`=`$ $`_\mu B_\nu _\nu B_\mu ,`$ (38)
with the $`F_{L/R}`$ for the fields $`A_{L/R}`$ defined in Eq. (8), along with other interaction terms involving the $`C_{L/R}`$ fields, the interaction term
$$\mathrm{Tr}\left[MM^{}\right]\mathrm{Tr}\left[C_L^2+C_R^2\right],$$
(39)
and a symmetry breaking potential.
One can show that this is the most general dimension-four, CP-invariant Lagrangian describing a strongly interacting set of scalars, vectors, and axial vectors with a spontaneously broken $`SU_L(2)\times SU_R(2)`$ symmetry, and possessing electroweak gauge invariance. It describes weak mixing between the $`A_{L/R}`$ fields and the $`W`$ and $`Z`$, and, through the mixing, conventional electroweak charges for the $`A_{L/R}`$. The extension of this effective Lagrangian to the relevant case of the larger symmetry group $`SU_L(N_f)\times SU_R(N_f)`$ with $`N_f>2`$, is straightforward.
Replacing $`M`$ by its vacuum value $`v/\sqrt{2}`$, and keeping only terms quadratic in the fields, the Lagrangian Eq. (35) takes the form
$`L=`$ $`M_A^2\mathrm{Tr}\left[A^2\right]+M_V^2\mathrm{Tr}\left[V^2\right]{\displaystyle \frac{\sqrt{2}}{\stackrel{~}{g}}}(1\chi )M_A^2\mathrm{Tr}\left[\left(gWg^{}B\right)A\right]`$ (40)
$``$ $`{\displaystyle \frac{\sqrt{2}}{\stackrel{~}{g}}}M_V^2\mathrm{Tr}\left[\left(gW+g^{}B\right)V\right]+{\displaystyle \frac{M_V^2}{2\stackrel{~}{g}^2}}\mathrm{Tr}\left[\left(gW+g^{}B\right)^2\right]`$ (41)
$`+`$ $`{\displaystyle \frac{M_A^2}{2\stackrel{~}{g}^2}}\left(1+\delta \right)\mathrm{Tr}\left[\left(gWg^{}B\right)^2\right]+\mathrm{},`$ (42)
where we have defined:
$`M_V^2`$ $`=`$ $`m^2+v^2\left[r+{\displaystyle \frac{h}{2}}\right],`$ (43)
$`M_A^2`$ $`=`$ $`m^2+v^2\left[r+\stackrel{~}{g}^2c^22s\stackrel{~}{g}c{\displaystyle \frac{h}{2}}\right],`$ (44)
$`\chi `$ $`=`$ $`{\displaystyle \frac{v^2}{2M_A^2}}\stackrel{~}{g}\left[\stackrel{~}{g}cs\right],`$ (45)
$`\delta `$ $`=`$ $`{\displaystyle \frac{v^2}{2M_A^2}}\left[\stackrel{~}{g}^2(12c)+2s\stackrel{~}{g}\right].`$ (46)
The vector $`V`$ and axial $`A`$ fields are defined in Eq. (13). This quadratic Lagrangian describes masses for the V and A, weak mass mixing with the $`W`$ and $`B`$, and a mass matrix for the $`W`$ and $`B`$. There is no further, kinetic energy mixing among these fields. The vector and axial vector masses, $`M_V^2`$ and $`M_A^2`$, are arbitrary, depending on the choice of parameters, although generically we expect them and the scalar masses to be of order $`4\pi ^2v^2`$.
The weak mixing terms in Eq. (42) provide a contribution from physics beyond the standard model to the oblique electroweak corrections. These may be described by the $`S`$, $`T`$, and $`U`$ parameters, but the last two vanish in the present model because there is no breaking of weak isospin in the strong sector. While this is not apparent in Eq. (42), it is insured by the Ward identities and easily revealed through the mixing effects. The $`S`$ parameter receives contributions from all the physics beyond the standard model, including, in the model being used here, loops of pseudo-Goldstone bosons (PGB’s), the strongly interacting massive scalars, and the vector and axial vector. The direct, vector-dominance contribution of the vector and axial vector may be read off from Eq. (42) together with the kinetic term for the $`V`$ and $`A`$. One finds
$$S_{vectdom}=\frac{8\pi }{\stackrel{~}{g}^2}\left[\frac{M_A^2\left(1\chi \right)^2}{M_Z^2M_A^2}\frac{M_V^2}{M_Z^2M_V^2}\right]\frac{8\pi }{\stackrel{~}{g}^2}\left[1\left(1\chi \right)^2\right].$$
(47)
Clearly, this contribution to the $`S`$ parameter can take on any value depending on the choice of parameters. Its typical order of magnitude, with the strong coupling estimate $`\stackrel{~}{g}^24\pi ^2`$, is expected to be $`O(1)`$. This expression can be seen to be equivalent to the familiar vector-dominance formula $`S_{vectdom}4\pi \left[{\displaystyle \frac{F_V^2}{M_V^2}}{\displaystyle \frac{F_A^2}{M_A^2}}\right]`$ , with the identifications $`F_V^2={\displaystyle \frac{2}{\stackrel{~}{g}^2}}M_V^2`$ and $`F_A^2={\displaystyle \frac{2}{\stackrel{~}{g}^2}}M_A^2\left(1\chi \right)^2`$.
We next observe that the choice
$$s=\stackrel{~}{g}c,h=\stackrel{~}{g}^2c^2$$
(48)
gives $`\chi =0`$, leading immediately to the degeneracy of the vector and axial vector (see Eq. (46)), the relation $`F_A=F_V`$, and the vanishing of $`S_{vectdom}`$. The further choice $`r={\displaystyle \frac{\stackrel{~}{g}^2c^2}{2}}`$ leads to the collapse of the general effective Lagrangian into the simple form
$$L=\frac{1}{2}\mathrm{Tr}\left[D_\mu MD^\mu M^{}\right]+m^2\mathrm{Tr}\left[C_{L\mu }C_L^\mu +C_{R\mu }C_R^\mu \right],$$
(49)
along with the kinetic terms of Eq.(36), interactions among the $`C_{L/R}^\mu `$ fields, the interaction term Eq.(39), and a symmetry breaking potential. Here, $`DM=MigWM+ig^{}MB`$ is the standard electroweak covariant derivative, and $`C_{L/R}^\mu `$ are given by Eq. (26).
The strongly interacting sector has split into two subsectors, communicating only through the electroweak interactions. One subsector consists of the Goldstone bosons together with their massive scalar partners. The other consists of the degenerate vector and axial vector described by the $`A_{L/R}^\mu `$ fields. The mass mixing in Eq. (35) insures that they have conventional electroweak couplings. In the absence of electroweak interactions, there is an enhanced symmetry $`\left[SU_L(2)\times SU_R(2)\right]\times \left[SU_L(2)\times SU_R(2)\right]`$, breaking spontaneously to $`SU_V(2)\times \left[SU_L(2)\times SU_R(2)\right]`$. The electroweak interactions explicitly break the enhanced symmetry to $`SU_L(2)\times U_Y(1)`$. All of this may be generalized to $`N_f>2`$, necessary to yield a near-critical theory.
The additional symmetry has an important effect on the $`S`$ parameter, suppressing contributions that are typically large in QCD-like theories. It doesn’t suppress all contributions, of course, since the symmetry breaking subsector gives contributions that are expected to be of order unity. The parity-doubled subsector, however, cannot by itself contribute to $`S`$, because $`S`$ relies on electroweak symmetry breaking for its existence. It is the coefficient of an operator in the low-energy electroweak chiral Lagrangian ($`L_1`$ in Ref. ), which may be written in the form $`\mathrm{Tr}W^{\mu \nu }UB_{\mu \nu }U^{}`$, where $`W^{\mu \nu }`$ and $`B_{\mu \nu }`$ are defined in Eq. (38) and $`U`$ is the Goldstone matrix field satisfying the nonlinear constraint $`UU^{}=U^{}U=1`$. Clearly the $`U`$ operator, with its vacuum value $`U=1`$, is necessary to couple $`W^{\mu \nu }`$ to $`B_{\mu \nu }`$.
Among the contributions to $`S`$ remaining in the limit of enhanced symmetry, are loops of pseudo-Goldstone bosons, present when $`N_f>2`$. They may be estimated using chiral perturbation theory, with the standard-model corrections removed by convention. While they typically give contributions to $`S`$ of order unity, their specific value depends on details such as mass estimates for the PGB’s that arise from electroweak, QCD, and others interactions . An interesting new feature in the limit of enhanced symmetry is that the PGB contribution is not related to a direct, vector-dominance effect (which is now zero). There will also be contributions from the strongly interacting TeV physics, represented in our effective Lagrangian by the massive scalars. Our purpose here is not to make these estimates, but only to point out that an enhanced symmetry, leading to vector-axial vector degeneracy, will suppress contributions to $`S`$ purely from the parity doubled sector. These include the typically large vector dominance contribution discussed above.
Finally we note that, as we discussed at the end of Section III, it could be that only a lesser, discrete symmetry emerges in the physical spectrum. Even this would be sufficient to insure vector-axial degeneracy and the vanishing of the vector dominance contribution to the $`S`$ parameter. The discrete symmetry of Section III would only be possible if trilinear vector interactions are somehow suppressed. It will be interesting to explore the phenomenology of this possibility, in particular the effect on the self interactions of the $`W`$ and $`Z`$.
## IV $`SU(2N_f)`$ global symmetry
In this section we adapt the above discussion to the interesting case of fermions in pseudoreal representations of the gauge group. The simplest example is provided by an underlying $`SU(2)`$ gauge theory, a choice that will also offer the smallest value for the critical $`N_f`$ . Such theories are currently being investigated on the lattice (see Ref. ). The quantum global symmetry for $`N_f`$ matter fields in the pseudoreal representation of the gauge group is $`SU(2N_f)`$. We expect the gauge dynamics to create a non vanishing fermion-antifermion condensate which breaks the global symmetry to $`Sp(2N_f)`$. Since $`SU(2N_f)SU_L(N_f)\times SU_R(N_f)`$, the left-right independent groups are unified and parity invariance is automatic.
This breaking pattern gives $`2N_f^2N_f1`$ Goldstone bosons which are contained in the antisymmetric matrix $`M^{ij}`$ and $`i,j=1,\mathrm{},2N_f`$. With $`uSU(2N_f)`$ we have
$$MuMu^T.$$
(50)
We associate a vector field $`A_\mu =A_\mu ^aT^a`$ with $`T^a`$, a generic generator of $`SU(2N_f)`$, ($`a=1,\mathrm{},4N_f^21`$) and $`\mathrm{Tr}\left[T^aT^b\right]={\displaystyle \frac{1}{2}}\delta ^{ab}`$. Following the procedure outlined in the previous sections, we define a formal covariant derivative as
$$D_\mu M=_\mu Mi\stackrel{~}{g}A_\mu Mi\stackrel{~}{g}MA_\mu ^T,$$
(51)
where $`A`$ transforms as
$$A_\mu uA_\mu u^{}\frac{i}{\stackrel{~}{g}}_\mu uu^{}.$$
(52)
With electroweak interactions turned off, the effective Lagrangian reads
$`L=`$ $`{\displaystyle \frac{1}{2}}\mathrm{Tr}\left[DMDM^{}\right]+m^2\mathrm{Tr}\left[A^2\right]+r\mathrm{Tr}\left[A^2MM^{}\right]`$ (53)
$`+`$ $`h\mathrm{Tr}\left[AMA^TM^{}\right]+is\mathrm{Tr}\left[A\left(MDM^{}DMM^{}\right)\right],`$ (54)
together with the kinetic term
$$L_{\mathrm{Kin}}=\frac{1}{2}\mathrm{Tr}\left[F_{\mu \nu }F^{\mu \nu }\right],$$
(55)
where $`F_{\mu \nu }=_\mu A_\nu _\nu A_\mu i\stackrel{~}{g}[A_\mu ,A_\nu ]`$, along with globally invariant vector interaction terms, an interaction term proportional to $`\mathrm{Tr}A^2\mathrm{Tr}MM^{}`$, and a symmetry breaking potential.
The global symmetry is enhanced to $`SU(2N_f)\times SU(2N_f)`$ for the parameter choice $`s=\stackrel{~}{g},h=\stackrel{~}{g}^2\mathrm{and}r=\stackrel{~}{g}^2`$. The effective Lagrangian then takes the form
$$L=\frac{1}{2}\mathrm{Tr}\left[MM^{}\right]+m^2\mathrm{Tr}\left[A^2\right]$$
(56)
together with the same terms as above. The spontaneous breaking leads to the vacuum symmetry $`Sp(2N_f)\times SU(2N_f)`$.
To proceed further, we simplify the notation by choosing $`N_f=2`$. We divide the generators $`\{T\}`$ of $`SU(4)`$ into two classes, calling the generators of $`Sp(4)`$ $`\{S^a\}`$ with $`a=1,\mathrm{},10`$ and the broken generators $`\{X^i\}`$ with $`i=1,\mathrm{},5`$. We have
$$S^TE+ES=0,$$
(57)
with
$$E=\frac{1}{2\sqrt{2}}\left(\begin{array}{cc}\mathrm{𝟎}& \mathrm{𝟏}\\ \mathrm{𝟏}& \mathrm{𝟎}\end{array}\right).$$
(58)
In Appendix A we provide a convenient representation for the $`\{S\}`$ and $`\{X\}`$ generators. We define the antisymmetric meson matrix $`M=(M^T)`$ as
$$M=\sqrt{2}\left[\sigma +i\mathrm{\hspace{0.17em}2}\sqrt{2}X^i\mathrm{\Pi }^i\right]E,$$
(59)
where the five $`\mathrm{\Pi }^i`$ fields are the Goldstone bosons associated with the breaking of $`SU(4)Sp(4)`$.
It is convenient to divide the vector field $`A`$ in the following way:
$$A=A_X+A_S,$$
(60)
where $`A_X=A_X^iX^i,\mathrm{and}A_S=A_S^aS^a`$. The $`A_X`$ are the axial vector fields while the $`A_S`$ are the vectors. Then expanding $`M`$ around its vacuum value $`\sqrt{2}vE`$ and keeping only terms quadratic in the fields, the Lagrangian Eq. (54) takes the form
$`L=`$ $`{\displaystyle \frac{1}{2}}\left[\sigma \sigma +\mathrm{\Pi }^i\mathrm{\Pi }^i\right]v{\displaystyle \frac{\left(\stackrel{~}{g}s\right)}{\sqrt{2}}}\mathrm{\Pi }^iA_X^i+M_X^2\mathrm{Tr}\left[A_X^2\right]+M_S^2\mathrm{Tr}\left[A_S^2\right],`$ (61)
with
$`M_S^2`$ $`=`$ $`m^2+{\displaystyle \frac{v^2}{4}}\left[rh\right],`$ (62)
$`M_X^2`$ $`=`$ $`m^2+{\displaystyle \frac{v^2}{4}}\left[2\stackrel{~}{g}^2+r4\stackrel{~}{g}s+h\right].`$ (63)
For the choice of parameters associated with an additional $`SU(4)`$ global symmetry (i.e. $`s=\stackrel{~}{g},h=\stackrel{~}{g}^2`$ and $`r=\stackrel{~}{g}^2`$) the vector-axial vector mass difference vanishes, as do the width to mass ratios.
We next treat the above theory as an electroweak symmetry breaking sector by gauging the $`SU_L(2)\times U_Y(1)`$ subgroup. It is convenient to introduce a vector field $`G_\mu `$. If we were to gauge the entire $`SU(4)`$ flavor symmetry then $`G_\mu `$ would transform under chiral rotations in the standard way
$$G_\mu uG_\mu u^{}\frac{i}{g}_\mu uu^{}.$$
(64)
We identify the electroweak gauge transformations in the following way:
$$u=\left(\begin{array}{cc}u_L& \mathrm{𝟎}\\ \mathrm{𝟎}& u_R^{}\end{array}\right),$$
(65)
with $`u_{L/R}SU_{L/R}(2)`$. Then
$$G_\mu =\left(\begin{array}{cc}W_\mu & \mathrm{𝟎}\\ \mathrm{𝟎}& \frac{g^{}}{g}B_\mu ^T\end{array}\right),$$
(66)
where $`W_\mu =W_\mu ^a{\displaystyle \frac{\tau ^a}{2}}`$ and $`B_\mu =B_\mu {\displaystyle \frac{\tau ^3}{2}}`$, and $`g`$ and $`g^{}`$ are the electroweak couplings. It is easy to verify that the electroweak transformation properties of the gauge bosons are respected (see Eq. (24)).
Using the left-right generators defined in Eq. (A10) we have
$$G=W^aL^a\frac{g^{}}{g}B^3R^{3T},$$
(67)
with $`a=1,2,3`$. In terms of the axial and vector type generators we have
$$G=G_X+G_S,$$
(68)
with
$$G_X=\frac{1}{\sqrt{2}}\left(W^a\frac{g^{}}{g}B^a\right)X^a,G_S=\frac{1}{\sqrt{2}}\left(W^a+\frac{g^{}}{g}B^a\right)S^a.$$
(69)
The covariant derivative including the weak vector bosons and the composite vector fields is
$$\mathrm{D}_\mu M=_\mu Mig\left(G_\mu M+MG_\mu ^T\right)i\stackrel{~}{g}c\left(C_\mu M+MC_\mu ^T\right),$$
(70)
where $`c`$ is a real coefficient and we have introduced the vector $`C_\mu `$
$$C=A\frac{g}{\stackrel{~}{g}}G,$$
(71)
transforming covariantly under electroweak rotations.
We extend the effective Lagrangian of Eq. (54) to include electroweak interactions by replacing the old covariant derivative with the one in Eq. (70). To render the full theory invariant under electroweak transformations we also substitute $`A`$ with $`C`$ giving
$`L=`$ $`{\displaystyle \frac{1}{2}}\mathrm{Tr}\left[\mathrm{D}M\mathrm{D}M^{}\right]+m^2\mathrm{Tr}\left[C^2\right]+r\mathrm{Tr}\left[C^2MM^{}\right]`$ (72)
$`+`$ $`h\mathrm{Tr}\left[CMC^TM^{}\right]+is\mathrm{Tr}\left[C\left(M\mathrm{D}M^{}\mathrm{D}MM^{}\right)\right].`$ (73)
To this we add kinetic terms, interaction terms involving the $`C`$ fields, the interaction term $`\mathrm{Tr}C^2\mathrm{Tr}MM^{}`$, and a symmetry breaking potential.
Replacing $`M`$ by its vacuum value and retaining only quadratic mass terms for the vectors we have
$`L=`$ $`M_X^2(1+\delta ){\displaystyle \frac{g^2}{\stackrel{~}{g}^2}}\mathrm{Tr}\left[G_X^2\right]+M_X^2\mathrm{Tr}\left[A_X^2\right]2{\displaystyle \frac{g}{\stackrel{~}{g}}}M_X^2(1\chi )\mathrm{Tr}\left[G_XA_X\right]`$ (74)
$`+`$ $`M_S^2\mathrm{Tr}\left[A_S^2+{\displaystyle \frac{g^2}{\stackrel{~}{g}^2}}G_S^22{\displaystyle \frac{g}{\stackrel{~}{g}}}G_SA_S\right]+\mathrm{},`$ (75)
where we have identified
$`M_S^2`$ $`=`$ $`m^2+{\displaystyle \frac{v^2}{4}}\left[rh\right],`$ (76)
$`M_X^2`$ $`=`$ $`m^2+{\displaystyle \frac{v^2}{4}}\left[r+h+2\stackrel{~}{g}^2c^24s\stackrel{~}{g}c\right],`$ (77)
$`\delta `$ $`=`$ $`{\displaystyle \frac{v^2}{2M_X^2}}\left[\stackrel{~}{g}^22\stackrel{~}{g}\left(\stackrel{~}{g}cs\right)\right],`$ (78)
$`\chi `$ $`=`$ $`{\displaystyle \frac{v^2}{2M_X^2}}\stackrel{~}{g}\left(\stackrel{~}{g}cs\right).`$ (79)
The generalization of this discussion to the case $`N_f>2`$, necessary for near-criticality of the underlying gauge theory, is straightforward.
From this point on, the discussion of enhanced symmetry, parity degeneracy, and the estimate of the $`S`$ parameter proceeds as in the previous section. The choice of parameters $`s=\stackrel{~}{g}c`$ and $`h=r=\stackrel{~}{g}^2c^2`$ leads to the enhanced symmetry of the strongly interacting sector and to parity degeneracy. The contribution to the $`S`$ parameter from the parity doubled sector by itself is zero. The contribution from the symmetry breaking sector is modified by the presence of the larger number of pseudo-Goldstone bosons associated with the $`SU(2N_f)`$ global symmetry.
## V Conclusions
In this paper we used an effective Lagrangian to explore some features that might arise in a strongly coupled gauge theory when the number of fermions $`N_f`$ is near a critical value for the transition to chiral symmetry. It has been argued that this transition is second order or higher and that a long range conformal symmetry also sets in at the transition. It has also been suggested that near the transition, parity doublets may begin to form .
We explored this possibility using as a guide an effective Lagrangian with a linear realization of the global chiral symmetry. The spectrum was taken to consist of a set of Goldstone particles, associated massive scalars, and a set of massive vectors and axial vectors. It was observed that parity doubling is associated with the appearance of an enhanced global symmetry, consisting of the spontaneously broken chiral symmetry of the underlying theory ($`SU_L(N_f)\times SU_R(N_f)`$) together with an additional, unbroken symmetry, either continuous or discrete. The additional symmetry leads to the degeneracy of the vector and axial vector, and to their stability with respect to decay into the Goldstone bosons.
It is worth noting that the effective Lagrangian employed here, while describing the global symmetries, does not accurately describe the dynamics of a chiral/conformal transition. That is, it cannot be used directly as the basis for a Landau-Ginzburg theory of this transition with its expected nonanalytic behavior . In Ref. , an approach to such a Landau-Ginzburg theory was developed, restricted to only the scalar degrees of freedom, and it described the usual global symmetries. This approach could perhaps be extended to include the vectors of the present effective Lagrangian. We expect that it would describe the same symmetries we have considered here, both the spontaneously broken symmetry and the additional, unbroken symmetry.
Despite the hints in Refs. , it has not been established that an underlying gauge theory leads to these enhanced symmetries as $`N_f`$ approaches a critical value for the chiral transition. If it is to happen, an unusual and interesting interplay between confinement and chiral symmetry breaking would have to develop at the transition.
We also noted, by electroweak gauging of a subgroup of the chiral flavor group, that the enhanced symmetry provides a partial custodial symmetry for the $`S`$ parameter, in that there is no contribution from the parity-doubled sector by itself. It could be interesting to explore further the consequences of an enhanced symmetry for electroweak precision measurements.
###### Acknowledgements.
We thank Sekhar Chivukula for a careful reading of this manuscript. We also thank Joseph Schechter, Adriano Natale, Noriaki Kitazawa, Zhiyong Duan, Erich Poppitz, and Alan Chodos for discussions. We thank Roberto Casalbuoni for helpful comments and for bringing to our attention some relevant literature. One of us (F.S.) thanks the theoretical group of Tokyo Metropolitan University for their warm hospitality. The work of T.A. and F.S. has been partially supported by the US DOE under contract DE-FG-02-92ER-40704. The work of F.S. was supported in part by the Grant-in-Aid for Scientific Research \# 0945036 under the International Scientific Research Program, Inter- University Cooperative Research. The work of P.S.R.S. has been supported by the Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP) under contract # 98/05643-5.
## A Explicit Realization of the $`Sp(4)`$ Generators
We conveniently represent the generators of $`SU(4)`$ in the following way
$$S^a=\left(\begin{array}{cc}𝐀& 𝐁\\ 𝐁^{}& 𝐀^T\end{array}\right),X^i=\left(\begin{array}{cc}𝐂& 𝐃\\ 𝐃^{}& 𝐂^T\end{array}\right),$$
(A1)
where $`A`$ is hermitian, $`C`$ is hermitian and traceless, $`B=B^T`$ and $`D=D^T`$. The $`\{S\}`$ are also a representation of the $`Sp(4)`$ generators since they obey the relation $`S^TE+ES=0`$. We define
$$S^a=\frac{1}{2\sqrt{2}}\left(\begin{array}{cc}\tau ^a& \mathrm{𝟎}\\ \mathrm{𝟎}& \tau ^{aT}\end{array}\right),a=1,2,3,4.$$
(A2)
For $`a=1,2,3`$ we have the standard Pauli matrices, while for $`a=4`$ we define $`\tau ^4=\mathrm{𝟏}`$. These are the generators for $`SU_V(2)\times U_V(1)`$. For $`a=5,\mathrm{},10`$
$$S^a=\frac{1}{2\sqrt{2}}\left(\begin{array}{cc}\mathrm{𝟎}& 𝐁^a\\ 𝐁^a& \mathrm{𝟎}\end{array}\right),a=5,\mathrm{},10$$
(A3)
and
$$\begin{array}{ccc}B^5=1& B^7=\tau ^3& B^9=\tau ^1\\ B^6=i\mathrm{\hspace{0.17em}1}& B^8=i\tau ^3& B^{10}=i\tau ^1\end{array}$$
(A4)
The five axial type generators $`\{X^i\}`$ are
$$X^i=\frac{1}{2\sqrt{2}}\left(\begin{array}{cc}\tau ^i& \mathrm{𝟎}\\ \mathrm{𝟎}& \tau ^{iT}\end{array}\right),i=1,2,3.$$
(A5)
$`\tau ^i`$ are the standard Pauli matrices. For $`i=4,5`$
$$X^i=\frac{1}{2\sqrt{2}}\left(\begin{array}{cc}\mathrm{𝟎}& 𝐃^i\\ 𝐃^i& \mathrm{𝟎}\end{array}\right),i=4,5,$$
(A6)
and
$$D^4=\tau ^2,D^5=i\tau ^2.$$
(A7)
The generators are normalized as follows
$$\mathrm{Tr}\left[S^aS^b\right]=\mathrm{Tr}\left[X^aX^b\right]=\frac{1}{2}\delta ^{ab},\mathrm{Tr}\left[X^iS^a\right]=0.$$
(A8)
The $`SU_{L/R}(2)`$ generators are readily identified as
$`L^a`$ $``$ $`{\displaystyle \frac{S^a+X^a}{\sqrt{2}}},`$ (A9)
$`R^a`$ $``$ $`{\displaystyle \frac{X^{aT}S^{aT}}{\sqrt{2}}},`$ (A10)
and $`a=1,2,3`$.
|
no-problem/9906/cond-mat9906367.html
|
ar5iv
|
text
|
# Fractional-differential equations of motion
## Abstract
An elementary system leading to the notions of fractional integrals and derivatives is considered. Various physical situations whose description is associated with fractional differential equations of motion are discussed.
PACS: 05.70.Ln, 47.53.+n, 64.60.Ak
Key words: Nonideal memory, Fractal, Fractional integral/derivative
Consider a medium exhibiting memory. Such a situation arises in describing the structure of a solid far from thermodynamic equilibrium: in amorphous materials, under structural relaxation of high-$`T_c`$ oxide superconductors, under plastic deformation and destruction of solids, in solid solutions and in martensite macrostructures, and so on . The existence of memory implies that if a force $`f(t^{})`$ acts on a system at an instant $`t^{}>0`$, then a flux $`J`$ arises whose magnitude at a subsequent instant $`t>t^{}`$ is given by
$$J(t)=_0^tM(tt^{})f(t^{})\text{d}t^{}.$$
(1)
For a system without memory, the time dependence of the memory function $`M(tt^{})`$ has the form
$$M(tt^{})=\gamma \delta (tt^{}),$$
(2)
where $`\gamma `$ is a positive constant and $`\delta (tt^{})`$ is the Dirac $`\delta `$-function. On substituting (2) into (1), we obtain the relation
$$J(t)=\gamma f(t),$$
(3)
according to which in the absence of memory only a force value at the same instant $`t`$ has an effect on the flux $`J(t)`$. With memory inclusion, the $`\delta `$-function in (2) spreads into a bell-shaped function with a width determining an interval $`\tau `$ during which the force $`f`$ has an effect on the flux $`J`$. For systems with ideal memory, we have $`\tau \mathrm{}`$, i.e., the flux $`J(t)`$ is formed over all the course of the action of the force $`f(t^{})`$ up to the instant $`t`$. Formally, this is expressed by specifying the kernel of the integral relation (1) in the form
$$M(tt^{})=\gamma /t.$$
(4)
Here, the dependence on the instant $`t^{}`$ of the force $`f`$ action is absent, and the dependence of the memory on the flux measurement time $`t`$ is given so as to satisfy the normalization condition
$$_0^tM(tt^{})\text{d}t^{}=\gamma .$$
(5)
Relation (1) as written in the time domain, is inconvenient because of the convolution (integral over $`t^{}`$). This can be eliminated by using the Laplace transformation
$$f(t)=_i\mathrm{}^{+i\mathrm{}}f(\lambda )\text{e}^{\lambda t}\frac{\text{d}\lambda }{2\pi i},f(\lambda )=_0^{\mathrm{}}f(t)\text{e}^{\lambda t}\text{d}t,$$
(6)
which enables one to move from the time $`t`$ to the complex frequency $`\lambda `$. After transforming both sides of the definition (1), this converts to the algebraic form
$$J(\lambda )=M(\lambda )f(\lambda ).$$
(7)
Obviously the Laplace transform of the kernel (2), which corresponds to the absence of memory, reduces to a constant:
$$M(\lambda )=\gamma .$$
(8)
For ideal memory, we obtain from (4), in the limit $`|\lambda |t1`$
$$M(\lambda )=\gamma /\lambda t.$$
(9)
Thus, the inclusion of memory leads to the transformation of the constant kernel (8) into the hyperbolic form (9).
In this connection, one could ask: How does the function $`M(\lambda )`$ stand up if memory is complete but nonideal? This means that the memory manifests itself within the interval $`(0,t)`$ preceding $`t`$ but not at all instants $`t^{}`$. Assume, for example, that memory holds only at the points of a Cantor set. One can then expect that its fractal dimension $`D`$ will be associated with a measure of memory preservation.
Although elementary for simple time dependences (2) and (4), the problem of finding the Laplace transform for the memory acting at the points of the Cantor set is much more difficult. The appropriate calculations arrive at the result
$$M(\lambda )=A_Dz^D,z=(1\xi )\lambda t.$$
(10)
Here, the fractal dimension $`D`$ is defined by usual equality $`D=\mathrm{ln}j/\mathrm{ln}\xi ^1`$, $`j`$ is the branching index of the corresponding hierarchical tree, $`\xi `$ is the similarity parameter, and the constant $`A_D=2^{D/2}`$ for the Cantor discontinuum with the number of elementary blocks $`j=2`$.
It can easily be seen that the memory function (10) satisfies the similarity condition $`M(\xi \lambda )=\xi ^DM(\lambda )`$ with the exponent $`D`$. Being the fractal dimension of the Cantor set at whose points memory is switched on, this exponent thus represents a quantitative measure of the manifestation of memory effects. For an empty Cantor set ($`\xi =0`$), one has $`D=0`$, and the dependence (10) reduces to the constant (8), corresponding to the entire absence of memory. The exponent $`D`$ increases with the similarity parameter $`\xi >0`$, and the Laplace transform (10) becomes an increasingly more rapidly varying function. The limiting value $`\xi =j^1`$ of the similarity parameter yields the dimension $`D=1`$, which corresponds to the ideal memory governed by the function (9).
Thus, systems with residual memory are described by the Laplace transform (10), where the exponent value $`0D1`$ determines the extent of memory preservation. Using the inverse Laplace transformation (6), we find for the time-dependent memory function in the simplest case of the Cantor set with $`j=2`$ (see , )
$$M(tt^{})=(\gamma /\pi )\left[\sqrt{2}(1\xi )t\right]^D\mathrm{\Gamma }(1D)(tt^{})^{D1},$$
(11)
where $`\mathrm{\Gamma }(1D)`$ is the gamma-function. Substituting this into equality (1), we bring this equality in the form
$$J(t)=(\gamma /\pi )\left[\sqrt{2}(1\xi )\right]^D\widehat{D}^Df(t),$$
(12)
where the integral is introduced
$`\widehat{D}^Df(t)`$ $`=`$ $`\mathrm{\Gamma }(1D){\displaystyle _0^1}(1u)^{D1}f(ut)\text{d}u`$ (13)
$`=`$ $`D^1\mathrm{\Gamma }(1D){\displaystyle _0^1}f\left((1u)t\right)\text{d}u^D,`$ (14)
whose fractional nature is reflected by the presence of the exponent $`D`$ in the argument $`u`$ of differential .
In the foregoing discussion we have used an integral representation for memory effects. It is easy to show, however, that an equivalent differential equation can be written that corresponds to this integral representation. For example, if we have some conserved quantity $`n`$ (like the number density of atoms), then it space–time dependence $`n(𝐫,t)`$ is governed by the continuity equation
$$\frac{n}{t}=𝐉,$$
(15)
where the flux J is given by equality (1). In the absence of memory effects, equality (3) is satisfied, where the force f is given by
$$𝐟=\mu ,$$
(16)
where $`\mu `$ is the chemical potential, and (14) leads to a conventional equation of diffusion type
$$\frac{n}{t}=(\gamma \mu ).$$
(17)
According to the earlier discussion, when memory is switched on, a fractional integral appears on the right-hand side of (16). The treatment of the resulting integro-differential equation, with partial derivatives and furthermore, of fractional order, presents a very difficult problem.
However, this problem can be simplified when it is considered that the memory function $`M(tt^{})`$ vanishes within some intervals, thus leading to a fixing of the flux J in (14) and correspondingly to the vanishing of the velocity $`n/t`$. Therefore, it would appear natural that loss of memory can be taken into account by a fractional property, not only of the flux integral (1) but also of the time derivative in (14). Because equation (16) corresponds to the entire absence of memory ($`D=0`$), when it is switched on, as is reflected by the growth of the exponent $`D`$, the rate of $`n`$ variation must decrease, and it is reasonable to postulate a fractional-differential equation of the form (see further)
$$\frac{^qn}{t^q}=(\gamma \mu ),q=1D.$$
(18)
As would be expected, the growth of the memory parameter $`D`$ leads to lowering the order $`q`$ of this equation. In systems with ideal memory ($`D=1`$), the derivative on the left-hand side of (17) disappears so that one should treat the behavior of the flux (1) itself which is what we have done above.
Our considerations in favor of the specific form $`q=1D`$ for the fractional degree in equation (17) act as guidelines. To confirm these, we will show that equation (12) for the flux in the form of the fractional integral (13) is in agreement with the solution of the fractional-order differential equation (17). Indeed, replacing the variable $`u`$ by $`(1u)t`$ in the integrand of the second integral of (13), one can easily show the time dependence of the flux to have the form $`Jt^D`$. Substituting this dependence into (14), we find the solution
$$nt^{1D},$$
(19)
satisfying the tested equation (17). This confirms conclusively the equivalence of the mutually complementary concepts of the fractional integral (13) and the fractional-differential equation (17). Notably is displayed in the same form the geometric relation $`Ll^{1D}`$ for coastline length $`L(l)`$ and in the physical dependence $`n(t)`$ obtained. In particular, it follows that the time $`t`$ plays the some role as the length of the segments $`l`$, covering a $`D`$-dimensional fractal set in the space of states, the points of this set determining the memory of the system.
When interpreting equation (17) in fractional derivatives, we start from the fact that in the usual case $`D=0`$ this equation governs irreversible processes like diffusion, heat conductivity, and so forth . In the latter processes, the microscopic memory with respect to time inversion is completely lost: the mechanical equations of motion for an isolated object exhibit ideal memory, manifesting itself in the invariance with respect to the replacement $`tt`$, whereas such invariance is completely violated in the thermodynamic equation (16). It follows that a decrease in the derivative order $`q=1D`$ in (17) corresponds to the switching on of memory channels whose part is determined by the fractal dimension $`D`$. The remaining $`q=1D`$ channels provide the irreversibility of the system. In the initial form, the residual irreversibility can be taken into account by using the fractional form of the Liouville equation
$$\frac{i\mathrm{}}{t}\frac{^D\rho }{u^D}=[H,\rho ],$$
(20)
where $`\mathrm{}`$ is the Planck constant, $`H`$ is the Hamiltonian, $`\rho `$ is the nonequilibrium statistical operator, $`t`$ is the evolution time of a macroscopic system, and $`u=t^{}/t`$ is the dimensionless microscopic time bounded by the condition $`u<1`$. It is apparent that the condition $`D<1`$ enables one, in the framework of (19), to take into account the irreversible effects in the interval $`(0,t)`$ which are due to the loss of $`q=1D`$ deterministic channels. Such a means of allowing for the irreversibility is obviously preferable to the phenomenological addition to the right-hand side of (19) of the relaxation term $`\rho /\tau `$, where $`\tau `$ is the relaxation time, or to the inclusion of an infinitesimal source ).
Apart from equations of the diffusion type, one might expect the appearance of fractional derivatives in a case of the description of the motion of particles under inelastic collisions. Performing calculations similar to those that led to the flux of the form (12), one can show that, if a force $`F`$ acts on a particle of mass $`m`$ in each collision, then the change in its velocity
$$\mathrm{\Delta }v=\left[\sqrt{2}(1\xi )\right]^D(t/m)\widehat{D}^DF$$
(21)
is determined by a fractional integral of the form (13). From the above example of the diffusion equation, one can see that, in order to switch to the corresponding fractional-differential equation, it is sufficient to operate on equation (20) with the operator $`\widehat{D}^D=^D/u^D`$ which is inverse with respect to the fractional integral $`\widehat{D}^D`$. In the case of an elastic force $`𝐅=\lambda \mathrm{\Omega }^2𝐫`$, where $`\lambda `$ is the elastic constant and $`\mathrm{\Omega }`$ is the atomic volume, we then obtain the generalized transport equation for the particle coordinate $`𝐫`$
$$\frac{\mathrm{d}^{1+D}𝐫}{\mathrm{d}u^{1+D}}+(ct)^2^2𝐫=0,$$
(22)
where the dimensionless time $`u=t^{}/t<1`$ is used and the characteristic velocity $`c`$, defined by
$$c^2=(\lambda \mathrm{\Omega }/m)\left[\sqrt{2}(1\xi )\right]^D.$$
(23)
is introduced. Equation (21) governs a new type of wave motion that is intermediate between conventional diffusion ($`D=0`$) and classical waves ($`D=1`$). Accordingly, formula (22) determines the diffusion coefficient $`D=c^2t`$ in the first case and the wave velocity $`c`$ in the second case.
The fractional-differential equations (17) and (21) are written as applied to the study of the space–time behavior of conserved quantities. As we known , this is reflected in the presence of a second derivative with respect to the coordinate in the right-hand sides of these equations. For a nonconserved quantity $`\eta `$, this derivative vanishes, and the diffusion equation (17) converts to the generalized Landau–Khalatnikov relaxation equation
$$\frac{\mathrm{d}^{1D}\eta }{\mathrm{d}u^{1D}}=(\gamma _Dt)\eta ,$$
(24)
and the wave equation (21) converts to the generalized equation for an harmonic oscillator
$$\frac{\mathrm{d}^{1+D}\eta }{\mathrm{d}u^{1+D}}+(\omega _Dt)^2\eta =0;$$
(25)
here, as above, $`u=t^{}/t`$ is the dimensionless time, and the parameters $`\gamma _D`$, $`\omega _D\left[\sqrt{2}(1\xi )\right]^D`$ determine the relaxation time $`\tau _D=\gamma _D^1`$ and the oscillator frequency $`\omega _D`$. The fractal dimension $`D`$ specifies a measure of the residual memory of the system. In its absence ($`D=0`$), both equations (23) and (24) reduce to the equation of Debye’s relaxation which governs the behavior of the simplest thermodynamic systems ($`\gamma _0=\omega _0^2t`$ in this case). For ideal memory ($`D=1`$), equation (23) degenerates into the condition of phase equilibrium $`\eta =\eta _0`$ and (24) degenerates into the ordinary oscillator relation.
The above consideration shows that, in limiting cases, one can recognize the following equations of motion, which correspond to the different types of behavior of a continuous medium. First, there is the continuity equation, being the condition for the conservation of particles in the medium. In the context of a mechanistic approach, the oscillatory behavior is governed either by an equation of oscillation, which involves a second derivative with respect to time, or by a wave equation involving second derivatives with respect to both time and space coordinates. Under the thermodynamic description of a system determined by a hydrodynamic mode, its amplitude (order parameter) may be characterized both by reactive and by dissipative regimes. In the first regime, the equation of motion reduces to an oscillatory equation for a nonconserved parameter and to a wave equation for a conserved one. Upon entering the dissipative regime, the time-derivative order is lowered to the first order.
Although these equations govern quite different physical situations, our discussion shows them to transform smoothly from one to another , . Thus, the question arises: Is it possible to give a model of the medium whose equation of motion includes all the above types as special cases? Following , we will show how such an equation can be obtained.
We start from the continuity equation $`\dot{\eta }+𝐣=0`$ for the order parameter field $`\eta (𝐫,t)`$ characterized by the flux $`𝐣(𝐫,t)`$. In the general case, the latter is nonlocally related to the distribution of the chemical potential $`\mu (𝐫,t)`$. In the framework of the Fourier representation with respect to the space coordinate and of the Laplace representation with respect to time (we take into account the causality condition), this relation reduces to the equality $`𝐣(𝐤,\lambda )=𝐌(𝐤,\lambda )\mu (𝐤,\lambda )`$, where $`𝐤`$ is the wave vector and $`\lambda `$ is the complex frequency.
To find the dependence of the memory function $`𝐌(𝐤,\lambda )`$ on $`𝐤`$, we will consider the simplest one-dimensional model of the medium in the form of a segment $`0xl`$. Then the flux through the right boundary is given by (cf.Eq.(1))
$$j(l)=_0^lM(lx)\mu (x)dx.$$
(26)
We assume that the function $`M(x)`$ takes on nonzero values only at the points of the Cantor set which is obtained from the original segment $`[0,l]`$ by the convolution method. Then, as in Ref., it can be shown that the desired dependence is given by the equality similar to (10):
$$M(k)=\mathrm{e}^{\alpha D}\kappa ^D,\kappa =ikl(1\xi ).$$
(27)
Here, the constant $`\alpha 1`$ is introduced (according to , $`\alpha =2^1\mathrm{ln}2=0.347`$ for the simplest Cantor set), $`D`$, $`\xi `$, and $`l`$ are the fractal dimension, the similarity parameter, and the scale, respectively, which describe the set of points on the $`x`$-axis where the order parameters conserved. Using (26), we find the dispersion law from the continuity equation:
$$\lambda =B_Dk^{1D},B_D=i^{1D}\mathrm{e}^{\alpha D}(1\xi )^Dl^{1D}\mathrm{d}\mu /\mathrm{d}\eta .$$
$`(27a)`$
When we reduce the fractal set to a continuum ($`\xi =1/2`$), its dimension is then $`D=1`$, and relation (27a) corresponds to the Landau–Khalatnikov equation for the nonconserved order parameter. In the general case of $`\xi <1/2`$, upon reduction of the fractal dimension of the set with ordering points, the vanishing of long-wave harmonics occurs, which is inherent in the order parameter conserved. However, even the maximum value of the exponent $`1D`$, corresponding to the empty set ($`\xi =0`$), proves to be half as large as the exponent 2 inherent in the order parameter conserved.
To raise the exponent in the dispersion law (27a), we assume that the order parameter becomes conserved after going to the empty set as a result of the reducing the similarity parameter $`\xi `$ from $`1/2`$ to 0. Next, using a procedure inverse to convolution, we increase the fractal dimension to a finite value $`D>0`$. Obviously, this can be done by decreasing the similarity parameter $`\xi `$ from $`\mathrm{}`$ to a finite value $`\xi 2`$. The condition $`\xi >1`$ implies that we stretch the segments rather than convolute them. Then, the functional equation $`M(\kappa /\xi )=2^1M(\kappa )`$ takes the form $`M(\xi \kappa )=2M(\kappa )`$. Its solution $`M(\kappa )=\mathrm{e}^{\alpha D}\kappa ^D`$ has a positive exponent whose value is now determined by the relation
$$D=\mathrm{ln}2/\mathrm{ln}\xi $$
(28)
instead of the usual equation $`D=\mathrm{ln}2/\mathrm{ln}\xi `$. Accordingly, the dispersion law for the partially conserved order parameter has the form
$$\lambda =B_Dk^{1+D},B_D=i^{1+D}\mathrm{e}^{\alpha D}|1\xi |^Dl^{1+D}\mathrm{d}\mu /\mathrm{d}\eta .$$
$`(27)`$
As seen from (27a), the distinction from the case of the conserved order parameter reduces to the change of signs of $`D`$ and $`\xi 1`$. According to (28), this corresponds to the transition from contraction ($`\xi <1`$) to stretching ($`\xi >1`$) when constructing a fractal.
Thus, if it is assumed that the similarity parameter $`\xi `$ is defined in the regions $`(0,1/2]`$ and $`[2,\mathrm{})`$, the dispersion law will encompass both the cases of the nonconserved ($`D<0`$) and conserved ($`D>0`$) order parameter. The transition between the limiting cases occurs due to a smooth variation in the similarity parameter $`\xi `$. At $`\xi =1/2`$, there is $`D=1`$, and we have complete nonconservation of the order parameter. A decrease in $`\xi `$ down to values $`0<\xi <1/2`$ partially violates the nonconservation conditions, thus leading to an increase in the fractal dimension in the range $`1<D<0`$ and to the rise in the exponent of the dispersion law (27). With the empty Cantor set ($`D=0`$), we go from the nonconserved order parameter ($`\xi 0`$) to the conserved one ($`\xi \mathrm{}`$). A decrease in the similarity parameter within the segment $`2<\xi <\mathrm{}`$ ($`0<D<1`$) provides the rise in the number of conservation channels, thus increasing the exponent of the dispersion law (27). Finally, with the complete conservation of the order parameter ($`\xi =2`$, $`D=1`$), we obtain, as expected, $`\lambda k^2`$.
It is easy to understand that this variation in the order of the derivative with respect to the space coordinate can be extended to the derivative with respect to time. Indeed, as shown in , the memory switching on relative to the time inversion $`tt`$ in the equation of particle motion, leads to the appearance in (27) of the exponent $`1D_0`$ at frequency $`\lambda `$, where $`D_0`$ is the dimension of a fractal set on the time axis, at the points $`t`$ of which the mechanical memory is switched on. The set itself is represented by the similarity transformation $`F(z/\xi )=2^1F(z)`$, where $`z=\lambda t(1\xi _0)`$ is the argument and $`0\xi _01/2`$. The transition from the contraction region $`\xi _01/2`$ into the stretching region $`\xi _02`$, corresponding to the functional equation $`M(\xi _0z)=2M(z)`$, occurs on the empty set ($`D_0=0`$) and implies the replacement of the corpuscular memory by the wave memory. The latter is enhanced with a decrease in the similarity parameter $`\xi _0`$ and becomes ideal at $`\xi _0=2`$. Then, the exponent at the frequency $`\lambda `$ in (27) takes on the value 2 inherent in the wave equation.
Thus, in the general case of partial nonconservation/conservation of the order parameter and of the non-ideal memory of particle/wave, the dispersion law for a one-dimensional fractal medium takes the form
$`\lambda `$ $`=`$ $`c^{1/(1+D_0)}(ik)^q;`$ (29)
$`q`$ $`=`$ $`(1+D)/(1+D_0),`$ (30)
$`c`$ $`=`$ $`{\displaystyle \frac{\mathrm{e}^{\alpha D}|\xi 1|^Dl^{1+D}}{\mathrm{e}^{\alpha _0D_0}|\xi _01|^{D_0}\tau ^{1+D_0}}}{\displaystyle \frac{\mathrm{d}\mu }{\mathrm{d}\eta }},\alpha _0,\alpha 1,`$ (31)
where $`l`$ and $`\tau `$ are the characteristic space-time scales, and the dependence of the dimension $`D_0`$ upon the parameter $`\xi _0`$ is determined by the same equality (28) as for the space component $`D(\xi )`$. Accordingly, the fractional-differential equations of motion are written in the form<sup>*</sup><sup>*</sup>*In the $`d`$-dimensional case, the second term of equation (30) is replaced by a sum of terms characterized by scale $`l_i`$, fractional dimension $`D_i`$, similarity parameter $`\xi _i`$ and constant $`\alpha _i`$, for $`i=1,\mathrm{},d`$.
$$\frac{\mathrm{d}^{1+D_0}\eta }{\mathrm{d}t^{1+D_0}}+i^{1+D}\frac{\mathrm{d}^{1+D}(c\eta )}{\mathrm{d}x^{1+D}}=0.$$
(32)
In the linear approximation $`\mu \eta `$, the constant $`c`$ may be factored out of the differentiation sign.
It can be easily seen that, in the limiting cases of the integer-valued space-time dimensions $`D_0`$ and $`D`$, the linearized equation (30) encompasses the main regimes of medium evolution. For example, in the case $`D_0=1`$, this governs the wave motion at $`D=1`$, uniform oscillations at $`D=1`$, and localized oscillations at $`D=0`$. The constant $`c`$ equals the square of wave velocity, the square of oscillation frequency, and the product of localization region size and oscillation frequency, respectively. In the case $`D_0=0`$, we have the Landau–Khalatnikov equation for the conserved ($`D=1`$) and nonconserved ($`D=1`$) order parameters and the traveling-wave equation ($`D=0`$). The constant $`c`$ reduces to the diffusion coefficient in the first case, to the inverse relaxation time in the second case, and to the wave velocity in the third case. And finally, in the static case $`D_0=1`$, we obtain the Poisson equation at $`D=1`$, the equilibrium condition at $`D=1`$, and the Debye screening equation at $`D=0`$. Here, $`c`$ reduces, respectively, to the square of screening radius, to the inverse susceptibility $`\mathrm{d}\mu /\mathrm{d}\eta `$, and to the screening radius.
It is evident that the approach outlined enables one to represent not only linear equations of motion but also to generalize these to the nonlinear case and to take into account the higher dispersion terms. For example, if, by analogy with the theory of solitons , we set that, in addition to the term $`k^D`$, the flux involves the term $`k^{2+\mathrm{\Delta }}`$ with the exponent $`1\mathrm{\Delta }1`$, then a dispersion component of a higher order $`3+\mathrm{\Delta }`$ appears. The nonlinearity is reflected in the dependence $`\mu (\eta )`$ of the chemical potential on the order parameter. One should take into account that the dimension $`\delta `$ of a nonlinear channel does not reduce to the corresponding quantity $`D`$ for a linear one. As a result, equation (30) takes the form
$`{\displaystyle \frac{\mathrm{d}^{1+D_0}\eta }{\mathrm{d}t^{1+D_0}}}+i^{1+D}c_0{\displaystyle \frac{\mathrm{d}^{1+D}\eta }{\mathrm{d}x^{1+D}}}+\beta {\displaystyle \frac{\mathrm{d}^{3+\mathrm{\Delta }}\eta }{\mathrm{d}x^{3+\mathrm{\Delta }}}}+{\displaystyle \frac{\mathrm{d}^{1+\delta }f(\eta )}{\mathrm{d}x^{1+\delta }}}=0,`$ (33)
$`f(\eta )={\displaystyle \frac{\mathrm{e}^{\alpha _\delta \delta }\left|\xi _\delta 1\right|^\delta l_\delta ^{1+\delta }}{\mathrm{e}^{\alpha _0D_0}\left|\xi _01\right|^{D_0}\tau ^{1+D_0}}}{\displaystyle \frac{\mathrm{d}\stackrel{~}{\mu }}{\mathrm{d}\eta }},`$ (34)
where $`c_0`$ corresponds to $`\eta =0`$, $`\beta >0`$ is the dispersion parameter, and $`\stackrel{~}{\mu }(\eta )`$ is the nonlinear contribution into the chemical potential. This equation reduces to the Korteweg–de Vries equation at $`D_0=D=\mathrm{\Delta }=\delta =0`$ and $`f\eta ^2`$, to the nonlinear Schrödinger one at $`D_0=D=0`$, $`\mathrm{\Delta }=\delta =1`$, $`f=\eta ^3`$, and finally to the sine-Gordon equation at $`D_0=D=1`$, $`\beta =0`$, $`\delta =1`$, $`f=\mathrm{sin}\eta `$. As is known, these equations are notable in that their symmetry within the framework of the inverse problem of the scattering theory provides the existence of soliton solutions . Here the question arises: Will this symmetry also manifest itself at other values of the differentiation orders $`D`$, $`D_0`$, $`\mathrm{\Delta }`$, and $`\delta `$ and/or with different choices of $`\beta `$ and $`f(\eta )`$? To answer this question demands investigation of the symmetry of multidimensional fractal sets and, to our knowledge, this has not been done.
Now, one can note only that conformal symmetry (see ) states the follow connection between the above exponents:
$$D=D_0=2(1+\delta n)^1.$$
(35)
Here exponent $`n`$ determines the nonliniar contribution $`f(\eta )\eta ^{n1}`$. Within framework of the Landau $`\eta ^4`$-theory, one has $`n=4`$ for a thermodynamic system characterised by the additive noise which intensity is the temperature. But with passing to the multiplicative noise with the amplitude proportional to $`\eta ^a`$, $`0<a<1`$ the powers of $`\eta `$ for all terms in Eq.(31), excluding the first, are lowered by the value $`2a`$ .
|
no-problem/9906/hep-ph9906386.html
|
ar5iv
|
text
|
# 1 Outline and summary of sum rule approach to 𝜇(𝐵)
## 1 Outline and summary of sum rule approach to $`\mu (B)`$
The rapidly growing precision of measurements of the electroweak coupling constants, like magnetic moments or the $`G_A/G_V`$– ratios in baryon semi-leptonic decays , may be the basis of new, more subtle and detailed information on the internal structure of baryons. In this report we consider some consequences of sum rules for the static, electroweak characteristics of baryons following from the theory of broken internal symmetries and common features of the quark models including relativistic effects and corrections due to nonvalence degrees of freedom – the sea partons and/or the meson clouds at the periphery of baryons.
In Ref. , the following parametrization was introduced for magnetic moments $`\mu (B)`$ of baryons :
$`\mu (B)=\mu (q_e)g_2+\mu (q_o)g_1+C(B)+\mathrm{\Delta },`$ (1)
$`\mu (\mathrm{\Lambda })=\mu (s)({\displaystyle \frac{2}{3}}g_2{\displaystyle \frac{1}{3}}g_1)+(\mu (u)+\mu (d))({\displaystyle \frac{1}{6}}g_2+{\displaystyle \frac{2}{3}}g_1)+\mathrm{\Delta },`$ (2)
$`\mu (\mathrm{\Lambda }\mathrm{\Sigma })={\displaystyle \frac{1}{\sqrt{3}}}(\mu (u)\mu (d))({\displaystyle \frac{1}{2}}g_2g_1)+C(\mathrm{\Lambda }\mathrm{\Sigma }),`$ (3)
$`\mathrm{\Delta }={\displaystyle \underset{q=u,d,s}{}}\mu (q)\delta (N),`$ (4)
where $`\mu (q)`$ are the effective quark magnetic moments defined without any nonrelativistic approximations, $`q_e=q_{even}=u;d;s`$ for $`P`$ and $`\mathrm{\Sigma }^+`$; $`N`$ and $`\mathrm{\Sigma }^{}`$; $`\mathrm{\Xi }^o`$ and $`\mathrm{\Xi }^{}`$, $`q_o=q_{odd}=u;d;s`$ for $`N`$ and $`\mathrm{\Xi }^o`$; $`P`$ and $`\mathrm{\Xi }^{}`$; $`\mathrm{\Sigma }^+`$ and $`\mathrm{\Sigma }^{}`$, respectively, so that $`P=u^2d`$, $`N=d^2u`$, etc. if only valence quarks are retained, $`g_{2(1)}`$ are ”reduced” dimensionless coupling constants obeying exact $`SU(3)`$–symmetry and related with the SU(3) $`F`$ and $`D`$ type constants via $`g_2=2F`$ and $`g_1=FD`$, $`\delta (B)`$ is a matrix element of the OZI–suppressed $`\overline{q}q`$–configuration for a given hadron: $`\delta (B)=<B|\overline{q}q|B>`$, where $`q\{q_e^2,q_o\}`$, e.g. $`\delta (N)=<N|\overline{s}s|N>`$, etc. For the models, explicitly including the meson degrees of freedom, we introduce the exchange current contribution to $`\mu (B)`$ due only to the dominant pion currents (these contributions are represented by the diagrams with the photon touching the charged pion line that connects different nonstrange quark lines). By inspection, this contribution is absent for all octet, three-quark hyperons except for the nondiagonal $`\mu (\mathrm{\Lambda }\mathrm{\Sigma })`$–magnetic moment. So that $`C(P)=C(N)`$ and $`C(\mathrm{\Lambda }\mathrm{\Sigma })`$ are the isovector contributions of the charged–pion exchange current to $`\mu (P),\mu (N)`$ and the $`\mathrm{\Lambda }\mathrm{\Sigma }`$–transition moment $`\mu (\mathrm{\Lambda }\mathrm{\Sigma })`$. In the following we make use, consecutively, two pictures of the baryon internal composition.In the first one, all baryons are considered as consisting of three massive, ”dressed” constituent quarks, locally coupled with lightest goldstonions – the pion fields.In the second picture only fundamental QCD quanta – the quarks and gluons – are there, the meson component of the baryon state vectors being represented by the properly correlated ”current” quarks and gluons.The use of one picture or another will be reflected in a particular parametrization of contributions due to corresponding nonvalence degrees of freedom. Below, we shall use the particle and quark symbols for corresponding magnetic moments. Equations (1) – (4) lead to the following sum rules :
$`P+N+\mathrm{\Xi }^0+\mathrm{\Xi }^{}3\mathrm{\Lambda }+{\displaystyle \frac{1}{2}}(\mathrm{\Sigma }^++\mathrm{\Sigma }^{})=0,`$ (5)
$`(\mathrm{\Sigma }^+\mathrm{\Sigma }^{})(\mathrm{\Sigma }^++\mathrm{\Sigma }^{}PN)(\mathrm{\Xi }^0\mathrm{\Xi }^{})(\mathrm{\Xi }^0+\mathrm{\Xi }^{}PN)=0,`$ (6)
$`\alpha ={\displaystyle \frac{D}{F+D}}={\displaystyle \frac{g_22g_1}{2(g_2g_1)}}={\displaystyle \frac{1}{2}}\left(1{\displaystyle \frac{\mathrm{\Xi }^0\mathrm{\Xi }^{}}{\mathrm{\Sigma }^+\mathrm{\Sigma }^{}\mathrm{\Xi }^0+\mathrm{\Xi }^{}}}\right),`$ (7)
$`C(P)={\displaystyle \frac{1}{2}}(C(P)C(N))={\displaystyle \frac{1}{2}}(PN+\mathrm{\Xi }^0\mathrm{\Xi }^{}\mathrm{\Sigma }^++\mathrm{\Sigma }^{}),`$ (8)
$`C(\mathrm{\Lambda }\mathrm{\Sigma })=\mu (\mathrm{\Lambda }\mathrm{\Sigma })+{\displaystyle \frac{1}{\sqrt{3}}}(\mathrm{\Xi }^0\mathrm{\Xi }^{}{\displaystyle \frac{1}{2}}(\mathrm{\Sigma }^+\mathrm{\Sigma }^{})),`$ (9)
$`{\displaystyle \frac{ud}{us}}={\displaystyle \frac{\mathrm{\Sigma }^+\mathrm{\Sigma }^{}\mathrm{\Xi }^0+\mathrm{\Xi }^{}}{\mathrm{\Sigma }^+\mathrm{\Xi }^0}},`$ (10)
$`{\displaystyle \frac{u}{d}}={\displaystyle \frac{\mathrm{\Sigma }^+(\mathrm{\Sigma }^+\mathrm{\Sigma }^{})\mathrm{\Xi }^0(\mathrm{\Xi }^0\mathrm{\Xi }^{})}{\mathrm{\Sigma }^{}(\mathrm{\Sigma }^+\mathrm{\Sigma }^{})\mathrm{\Xi }^{}(\mathrm{\Xi }^0\mathrm{\Xi }^{})}},`$ (11)
$`{\displaystyle \frac{s}{d}}={\displaystyle \frac{\mathrm{\Sigma }^+\mathrm{\Xi }^{}\mathrm{\Sigma }^{}\mathrm{\Xi }^0}{\mathrm{\Sigma }^{}(\mathrm{\Sigma }^+\mathrm{\Sigma }^{})\mathrm{\Xi }^{}(\mathrm{\Xi }^0\mathrm{\Xi }^{})}},`$ (12)
$`\mathrm{\Delta }^{++}={\displaystyle \frac{u}{s}}\mathrm{\Omega }^{}={\displaystyle \frac{u}{d}}\mathrm{\Delta }^{}.`$ (13)
Reserving the possibility of $`g_i(N)g_i(Y),Y=\mathrm{\Lambda },\mathrm{\Sigma },\mathrm{\Xi }`$, due to a more prominent role of the pion degrees of freedom in the nucleon and combining Eqs. (5)–(6), we propose also the following, probably, most general sum rule of this approach
$`(\mathrm{\Sigma }^+\mathrm{\Sigma }^{})(\mathrm{\Sigma }^++\mathrm{\Sigma }^{}6\mathrm{\Lambda }+2\mathrm{\Xi }^0+2\mathrm{\Xi }^{})`$
$`(\mathrm{\Xi }^0\mathrm{\Xi }^{})(\mathrm{\Sigma }^++\mathrm{\Sigma }^{}+6\mathrm{\Lambda }4\mathrm{\Xi }^04\mathrm{\Xi }^{})=0.`$ (14)
Eqs.(11) – (13) are obtained provided $`\delta (N)=0`$. Hence, we relate them to the chiral constituent quark models, where we assume that a given baryon consists of three ”dressed” constituent quarks, and also to the validity of the OZI (or the quark–line) rule. Now, we list some consequences of the obtained sum rules. By definition, the $`\mathrm{\Lambda }`$–value entering into Eqs. (5) and (14) should be ”refined” from the electromagnetic $`\mathrm{\Lambda }\mathrm{\Sigma }^0`$–mixing affecting $`\mu (\mathrm{\Lambda })_{exp}`$. Hence, the numerical value of $`\mathrm{\Lambda }`$, extracted from Eq. (14), can be used to determine the $`\mathrm{\Lambda }\mathrm{\Sigma }^0`$–mixing angle through the relation
$`\mathrm{sin}\theta _{\mathrm{\Lambda }\mathrm{\Sigma }}\theta _{\mathrm{\Lambda }\mathrm{\Sigma }}={\displaystyle \frac{\mathrm{\Lambda }\mathrm{\Lambda }_{exp}}{2\mu (\mathrm{\Lambda }\mathrm{\Sigma })}}=(1.43\pm 0.31)10^2`$ (15)
in accord with the independent estimate of $`\theta _{\mathrm{\Lambda }\mathrm{\Sigma }}`$ from the electromagnetic mass-splitting sum rule . Equation (11) shows that owing to interaction of the $`u`$– and $`d`$– quarks with charged pions the ”magnetic anomaly” is developing, i.e. $`u/d=1.80\pm 0.02Q_u/Q_d=2.`$ Evaluation of the lowest order quark–pion loop diagrams gives : $`u/d=(Q_u+\kappa _u)/(Q_d+\kappa _d)=1.77`$, where $`\kappa _q`$ is the quark anomalous magnetic moment in natural units. The meson–baryon universality of the quark characteristics, Eqs. (10) – (12), suggested long ago , is confirmed by the calculation of the ratio of $`K^{}`$ radiative widths
$`{\displaystyle \frac{\mathrm{\Gamma }(K^+K^+\gamma )}{\mathrm{\Gamma }(K^+K^+\gamma )}}=\left({\displaystyle \frac{u/d+s/d}{1+s/d}}\right)^2=0.42\pm 0.03(\text{vs}0.44\pm 0.06\text{[1]}).`$ (16)
The experimentally interesting quantities $`\mu (\mathrm{\Delta }^+P)=\mu (\mathrm{\Delta }^0N)`$ and $`\mu (\mathrm{\Sigma }^0\mathrm{\Lambda })`$ are affected by the exchange current contributions and for their estimation we need additional assumptions. We use the analogy with the one–pion–exchange current, well–known in nuclear physics, to assume for the exchange magnetic moment operator
$`\widehat{\mu }_{exch}={\displaystyle \underset{i<j}{}}[\stackrel{}{\sigma }_i\times \stackrel{}{\sigma }_j]_3[\stackrel{}{\tau }_i\times \stackrel{}{\tau }_j]_3f(r_{ij}),`$ (17)
where $`f(r_{ij})`$ is a unspecified function of the interquark distances, $`\stackrel{}{\sigma }_i(\stackrel{}{\tau }_i)`$ are spin (isospin) operators of quarks. Calculating the matrix elements of $`\widehat{\mu }_{exch}`$ between the baryon wave functions, belonging to the 56–plet of $`SU(6)`$, one can find
$`C(P)={\displaystyle \frac{1}{\sqrt{2}}}C(\mathrm{\Delta }^+P)=\sqrt{3}C(\mathrm{\Lambda }\mathrm{\Sigma }),`$ (18)
$`\mu (\mathrm{\Delta }^+P;\{56\})={\displaystyle \frac{1}{\sqrt{2}}}\left(PN+{\displaystyle \frac{1}{3}}(P+N){\displaystyle \frac{1u/d}{1+u/d}}\right).`$ (19)
where Eq.(19) may serve as a generalization of the well–known $`SU(6)`$–relation . We close this section presenting the set of sum rules following from Eqs.(2)–(4) with C(B’s) =$`\delta `$(B’s) = 0, u/d=-2:
$`\mathrm{\Sigma }^+[\mathrm{\Sigma }^{}]=P[PN]+(\mathrm{\Lambda }{\displaystyle \frac{N}{2}})(1+{\displaystyle \frac{2N}{P}}),`$ (20)
$`\mathrm{\Xi }^0[\mathrm{\Xi }^{}]=N[PN]+2(\mathrm{\Lambda }{\displaystyle \frac{N}{2}})(1+{\displaystyle \frac{N}{2P}}),`$ (21)
$`\mu (\mathrm{\Lambda }\mathrm{\Sigma })={\displaystyle \frac{\sqrt{3}}{2}}N.`$ (22)
The numerical values of Eqs.(20)–(22) coincide almost identically with the results of the $`SU(6)`$–based NRQM taking account of the $`SU(3)`$ breaking due to the quark–mass differences . We stress, however, that no NR assumption or explicit $`SU(6)`$-wave function are used this time. The ratio $`\alpha `$ ( cf. Eq.(7) ) equals .61 in this case, demonstrating a substantial influence of the nonvalence degrees of freedom on this important parameter.
In calculations we used baryon magnetic moments from the PDG–tabulation and new values of the $`\mu (\mathrm{\Xi }^{})`$ and $`\mu (\mathrm{\Sigma }^+)`$, from and , respectively.
## 2 The OZI-rule violation in magnetic and axial couplings of baryons
Here, we follow a complementary view of the nucleon structure, absorbing C(N’s) into products of the corresponding $`\mu (q)`$ and g(N’s), keeping the constraint u/d=–2,and $`\delta (B^{}s)`$ non-zero. We shall refer to this approach as a correlated current quark picture of nucleons. Then, instead of Eq.(11) we have (in n.m.)
$`\mathrm{\Delta }(N)={\displaystyle \frac{1}{6}}(3(P+N)\mathrm{\Sigma }^++\mathrm{\Sigma }^{}\mathrm{\Xi }^0+\mathrm{\Xi }^{})=.07\pm .01,`$ (23)
$`\mu _N(\overline{s}s)=\mu (s)N|\overline{s}s|N=(1{\displaystyle \frac{d}{s}})^1\mathrm{\Delta }=.13,`$ (24)
where independence of the sum P+N of the C(N’s) and the ratio d/s=1.55 from the correspondingly modified Eq.(12) were used. By definition, $`\mu _N(\overline{s}s)`$ represents the contribution of strange (”current”) quarks to nucleon magnetic moments.Numerically, Eq.(24) agrees fairly well with other more specific models ( see,e.g. ). The calculated quantity indicates violation of the OZI rule and the strange current quark contribution $`\mu _N(\overline{s}s)`$ is seen to constitute a sizable part of the isoscalar magnetic moment of nucleons (or, which is approximately the same, of the nonstrange constituent quarks)
$`{\displaystyle \frac{1}{2}}(P+N)=\mu (\overline{u}u+\overline{d}d)+\mu (\overline{s}s)=.44,`$ (25)
This observation helps to understand the unexpectedly large ratio
$`BR\left({\displaystyle \frac{\overline{P}N\varphi +\pi }{\overline{P}N\omega +\pi }}\right)(10\pm 2)\%,`$ (26)
reported for the s–wave $`\overline{N}N`$ – annihilation reaction.
Indeed, the transition $`(\overline{P}N)_{swave}V+\pi `$, where $`V=\gamma ,\omega ,\varphi `$, is of the magnetic dipole type. Therefore, the transition operator should be proportional to the isoscalar magnetic moment contributions from the light u– and d–quarks and the strange s–quark, Eq.(25). The transition operators for the $`\omega `$– and $`\varphi `$–mesons are obtained from $`\mu (\overline{q}q)`$ and $`\mu (\overline{s}s)`$ through the well–known vector meson dominance model ( VDM ).Using the ”ideal” mixing ratio $`g_\omega :g_\varphi =1:\sqrt{2}`$ for the photon–vector–meson junction couplings and $`\mu _\omega :\mu _\varphi =\mu (\overline{q}q):\mu (\overline{s}s)`$ according to Eq.(24) and Eq.(25), we get
$`BR\left({\displaystyle \frac{\overline{P}N\varphi +\pi }{\overline{P}N\omega +\pi }}\right)\left({\displaystyle \frac{\mu (\overline{s}s)}{\sqrt{2}\mu (\overline{u}u+\overline{d}d)}}\right)^2\left({\displaystyle \frac{p_\varphi ^{c.m.}}{p_\omega ^{c.m.}}}\right)^36\%,`$ (27)
which is reasonably compared with data.
The structure constants connected with the vector part of the weak neutral current ( NC ) are obtained from the electromagnetic ones by the substitution
$`Q(q)V^{NC}(q)={\displaystyle \frac{1}{2sin\theta _Wcos\theta _W}}(t_L(q)2Q(q)\mathrm{sin}^2\theta _W),`$ (28)
where Q(q) is the quark electric charge, $`t_L`$–the 3rd component of the weak isospin, $`sin^2\theta _W=.23`$, $`\theta _W`$–the weak angle. In this way, for the NC analogue of magnetic moments of the proton, neutron and deuteron we get ( in the units of n.m.)
$`\mu ^{NC}(P)1.49(1.25),`$ (29)
$`\mu ^{NC}(N)1.52(1.73),`$ (30)
$`\mu ^{NC}(d)0.04(.47),`$ (31)
where the values in the parentheses refer to the neglect of the strange quark contributions. Therefore the planned detailed investigation of the $`\gamma Z^0`$ interference effects and measuring $`\mu _N(\overline{s}s)`$ via the P- odd effects in polarized electron–nucleon scattering will give an important information on the strangeness content of the nucleon.
The value of $`N|\overline{s}s|N`$ in Eq.(25) can be considered as the difference of the averaged values $`N|l_z(s)+\sigma _z(s)|N`$ of strange quarks and antiquarks, respectively, $`l_z(s)`$ and $`\sigma _z(s)`$ being the orbital and spin operators of corresponding quarks in the polarized nucleon. The product of this combination of the quark angular momenta and the effective magnetic moments of quarks, in which quark energies are used instead of quark masses, was proposed to define the magnetic moment operator in a relativistic model of quark composites. Taking, for the sake of qualitative estimates, $`\mu (s)`$ =$`e/(6\epsilon _s)`$ $`.7`$ n.m., where we put $`\epsilon _s(m_{s}^{}{}_{}{}^{2}+p_{s}^{}{}_{}{}^{2})^{\frac{1}{2}}`$, $`m_s(150MeV)`$ – the ”current” quark mass, $`p_s(400MeV)`$–the mean momentum of sea quarks, we obtain
$`{\displaystyle \underset{s,\overline{s}}{}}N,J_z=+{\displaystyle \frac{1}{2}}|\sigma _z(s)+l_z(s)\sigma _z(\overline{s})l_z(\overline{s})|N,J_z=+{\displaystyle \frac{1}{2}}.17,`$ (32)
which can be compared with the other strange quark spin characteristics $`\mathrm{\Delta }s.1`$, as given by the polarized DIS data, e.g. . We wish here to note the following alternative. As is known , to obtain the contributions of the u-,d-, and s-flavoured quarks to the proton spin, denoted by $`\mathrm{\Delta }u(p),\mathrm{\Delta }d(p)`$ and $`\mathrm{\Delta }s(p)`$, the use is usually made of baryon semileptonic weak decays treated with the help of the exact $`SU(3)`$-symmetry. It has been shown earlier that when both the strangeness-changing $`(\mathrm{\Delta }S=1)`$ and strangeness-conserving $`(\mathrm{\Delta }S=0)`$ transitions are taken for the analysis, then $`(D/F+D)_{ax}^{\mathrm{\Delta }S=0,1}=.635\pm .005`$ while $`(D/F+D)_{ax}^{\mathrm{\Delta }S=0}=.584\pm .035`$, which is close to $`(D/D+F)_{mag}.58`$, according to Eq.(7). We list below two sets of the $`\mathrm{\Delta }q`$-values, we have obtained from the data with inclusion of the QCD radiative corrections (e.g. and references therein) : $`\mathrm{\Delta }u(p).82(.83),\mathrm{\Delta }d(p).44(.37),\mathrm{\Delta }s=.10\pm .04(.19\pm .05)`$, where the values in the parentheses correspond to $`\alpha _D=(D/D+F)=.58`$. At the same time, the problem of difference of the following two expressions
$`FD=\mathrm{\Delta }d(p)\mathrm{\Delta }s(p)=G_A^{exp}(\mathrm{\Sigma }^{}n)=.34\pm .02,`$ (33)
$`FD=\mathrm{\Delta }d(p)\mathrm{\Delta }s(p)=G_A^{exp}(np)\sqrt{6}G_A^{exp}(\mathrm{\Sigma }\mathrm{\Lambda })=.19\pm .04,`$ (34)
of which we prefer the second one when we postulate $`\alpha _D^{ax}=\alpha _D^{mag}`$, remains largely open.The intriguing possibility can, however, be mentioned that the numerical value of the $`G_A^{exp}(\mathrm{\Sigma }^{}n)`$, coinciding with Eq.(34), was in fact found in , if the weak–electric dipole form factor, referred as one of the second class current effects, is included in the joint analysis of all experimental data.
We note in passing, that the production of the axial – vector meson composed of strange quark-antiquark pairs in antinucleon annihilation may also be useful to probe the strange content of the nucleon and the dynamics of the OZI rule violation.The $`q+(\overline{s}s)`$\- configuration with $`J^P=\frac{1}{2}^+`$ composing in part the nonstrange constituent quark can evolve via a chain of transitions
$$q+(\overline{s}s)_{vac}(\overline{s}q)_0^{}+sq+(\overline{s}s)_{J^{PC}}$$
(35)
with $`J^{PC}=0^{++},1^{++},1^+`$.A simple recoupling of the angular momenta enable, through the pertinent Clebsch–Gordan coefficients, to obtain then the ratio
$$w(1^{++}):w(1^+):w(0^{++})=2:1:1$$
(36)
which may serve as a qualitative measure of relative yields of corresponding meson states, composed mainly of the ”strange matter”.The relevant mesons with $`J^P=1^+`$ might be the $`f_1(1420;1^{++})`$– or $`f_1(1510;1^{++})`$– and $`h_1(1380;1^{+(\mathrm{?})})`$–resonances .
## 3 Remarks
1. The deviation of the ratio $`F/D=.75`$, Eq.7, from the $`SU(6)`$ –value $`2/3`$ shows, that despite the validity of the celebrated $`SU(6)`$–ratio $`\mu (P)/\mu (N)=3/2`$, the $`SU(6)`$–symmetry is strongly broken. The importance of taking into account the nonvalence degrees of freedom in relevant parametrization of the observables within the (broken) internal symmetries is demonstrated
2. The real meaning of all failures ( and relative successes, of course ) of the ”naive” NRQM results ( e.g.), coinciding with values from Eqs.(20)–(22), is the neglect of contributions due to the nonvalence degrees of freedom ( the sea partons and/or meson clouds at the periphery of hadrons).
3. The strange current quarks considered as a part of the constituent quark internal structure should be explored by the probes with the resolution capability comparable with the constituent quark size, i.e. in the processes with high enough momentum transfers or the energy release.The OZI–rule violating $`\varphi `$ \- meson production in the proton–antiproton annihilation gives further evidence of the polarized strange quark sea inside the polarized nucleon.
The author is grateful to M.G.Sapozhnikov for a useful discussion of the data on the OZI – rule violation in the antinucleon annihilation reactions. This work was supported in part by the Russian Foundation for Basic Researches (grant No.96-02-18137).
|
no-problem/9906/patt-sol9906005.html
|
ar5iv
|
text
|
# Modulational instability of solitary waves in non-degenerate three-wave mixing: The role of phase symmetries.
## Abstract
We show how the analytical approach of Zakharov and Rubenchik \[Sov. Phys. JETP 38, 494 (1974)\] to modulational instability (MI) of solitary waves in the nonlinear Schröedinger equation (NLS) can be generalised for models with two phase symmetries. MI of three-wave parametric spatial solitons due to group velocity dispersion (GVD) is investigated as a typical example of such models. We reveal a new branch of neck instability, which dominates the usual snake type MI found for normal GVD. The resultant nonlinear evolution is thereby qualitatively different from cases with only a single phase symmetry.
One of the central issues of solitary wave theory is the question of stability . Stability of solitary waves (’solitons’) in nonintegrable Hamiltonian models is often governed by the derivative of a certain invariant with respect to an associated parameter of the solution . For instance, positivity of the derivative of the total energy with respect to the nonlinearity-induced wave-number shift is sufficient for stability of bright solitary waves in such fundamental processes as self-action of radiation in media with intensity dependent refractive index and in degenerate three-wave mixing in quadratic nonlinear media .
A wide range of parametric processes conserve not only the total energy of the interacting waves but also other energy invariants. These can be, for instance, energies of individual waves or energy unbalance of two waves. Typical examples of such processes are non-degenerate three-wave and four-wave mixings . By Noether’s theorem every integral of the motion in Hamiltonian models originates from a corresponding symmetry property. Energy invariants are often associated with phase (gauge) symmetries. Although many issues of the dynamics of the multi-wave solitons in models with several symmetries still remain to be understood, their stability threshold in many situations is given by a zero of the determinant of the Jacobi matrix $`_{\kappa _j}Q_i`$ . Here $`Q_i`$ and $`\kappa _j`$ are respectively the integrals of motion and their associated parameters.
Soliton dynamics in a space where the soliton is localised in some dimensions but extended in one or more others, e.g. consideration of a soliton stripe, raises the problem of modulational instability (MI) along the extra dimensions . MI of solitary waves has been studied in many fields, including plasma physics , fluid dynamics and optics . For instance in optics the above-mentioned extra dimensions might be associated with diffraction and/or group velocity dispersion (GVD). An analytical approach to the low-frequency limit of MI for bright solitary waves was originally developed by Zakharov and Rubenchik for the generalised NLS equation, which has a single phase symmetry. This approach is based on asymptotic expansion near neutrally stable eigenmodes (Goldstone modes) of the solitary wave excitation spectrum. The presence of such neutral modes is directly linked to the symmetries of the model. It follows that systems with a single phase symmetry should be qualitatively similar to the NLS case , where the soliton always shows MI in the extra dimension. The unstable mode is of neck type (the soliton stripe breaks into a chain of spots) or snake type (the stripe distorts in zig-zag fashion) depending on the relative signs of the dispersive terms in the localised and extended dimensions. All previous studies of solitary wave MI have been restricted to situations with a single phase symmetry, and all do indeed exhibit NLS-like MI. For example, in degenerate three-wave mixing (3WM), which has a single phase symmetry, the first analytical results on MI of solitary waves reported by Kanashov and Rubenchik and recently extended and supported by numerical results , typically show the neck/snake scenario.
The influence of extra phase symmetries on MI of solitary waves is still an open issue and is the main subject of the present Letter. We will concentrate on non-degenerate 3WM as a typical and practically important example of a solitonic model with two phase symmetries, motivated by recent theoretical and experimental advances in the study of quadratic optical solitons.
In this Letter we show that MI of solitary waves in non-degenerate 3WM reveals a new branch of neck-type instability which can be ascribed to the extra phase symmetry and its corresponding neutral mode. We show that in media with normal GVD this new instability dominates the snake-mode responsible for MI in the corresponding degenerate 3WM model. Our theoretical approach and the phenomena predicted by it should extend to other solitonic systems of broad interest with similar symmetry properties, e.g. to incoherently coupled NLS equations and non-degenerate 4WM .
In non-degenerate 3WM the evolution of suitably normalised slowly varying field envelopes $`E_m`$ ($`m=1,2,3`$) of three waves with carrier frequencies $`\omega _m`$ ($`\omega _1+\omega _2=\omega _3`$) propagating in a dispersive and diffractive quadratic nonlinear medium can be modeled by the following system of dimensionless equations:
$`i_zE_1+\alpha _1_x^2E_1+\gamma _1_t^2E_1+E_2^{}E_3=0,`$ (1)
$`i_zE_2+\alpha _2_x^2E_2+\gamma _2_t^2E_2+E_1^{}E_3=0,`$ (2)
$`i_zE_3+\alpha _3_x^2E_3+\gamma _3_t^2E_3+E_1E_2=\beta E_3.`$ (3)
Transverse $`x`$, longitudinal $`z`$ and retarded time $`t`$ coordinates are respectively measured in units of a suitable beam width, diffraction length and GVD parameter. The $`\alpha _m`$ and $`\gamma _m`$ are diffraction and dispersion coefficients referred to these scales, and the wave-vector mismatch is characterised by $`\beta `$. We neglect spatial and temporal walk-off effects, implicitly assuming that either their spatial and temporal scales are much longer than those associated with diffraction and GVD, or that walk-off is compensated by special techniques, as in recent experiments on temporal solitons in degenerate 3WM . Henceforth we make the experimentally appropriate choice $`\alpha _{1,2}=2\alpha _3=0.5`$. We also assume all $`\gamma _m`$ either negative (normal GVD) or positive (anomalous GVD), leaving the case of mixed dispersion for future work. For the sake of simplicity we have restricted our model to one transverse dimension, as in a planar waveguide.
Suppressing the time derivatives for the moment, Eqs. (1) have a family of non-diffracting solitonic solutions $`E_m=A_m(x)e^{i(\kappa _mz+\varphi _m)}`$, where the $`A_m`$ are real, $`\kappa _{1,2}=\kappa \pm \delta `$ are positive parameters, $`\kappa _3=2\kappa >\beta `$ and $`\varphi _{1,2}=\phi \pm \psi `$, $`\varphi _3=2\phi `$, where $`\phi `$ and $`\psi `$ are arbitrary real constants. In general functions $`A_m(x)`$ must be found numerically or approximated variationally . The free choice of $`\phi `$ and $`\psi `$ implies two phase (gauge) symmetries, which by Noether’s theorem leads to two conserved quantities. These are the total energy $`Q=Q_1+Q_2+2Q_3`$, and energy unbalance $`Q_u=Q_1Q_2`$, or equivalent combinations of the $`Q_m=𝑑x|E_m|^2`$. Note that the degenerate case forces $`E_1=E_2`$, and thus $`\psi =0`$, which suppresses one phase symmetry.
Our primary aim here is to study temporal MI due to GVD of these spatial solitons. This is most interesting and important when they are spatially stable. Their stability against purely spatial perturbations has been studied , yielding a stability boundary which in our notation is given by $`_\delta Q_\kappa Q_u=_\kappa Q_\delta Q_u`$. Spatially stable domain is in fact almost the entire domain of soliton existence, excluding only a small range of $`\kappa ,\delta `$ values with $`\beta <0`$ . Close to this region there is also a small domain of bistability . Therefore the existence and nature of temporal MI is the important question for almost all parameter values, and in particular for the entire half-space with $`\beta 0`$.
To study MI due to GVD we seek solutions of Eqs. (1) in the form of spatial solitons weakly modulated in time at frequency $`\mathrm{\Omega }0`$: $`E_m=(A_m(x)+(V_m(x,z)+iW_m(x,z))\mathrm{cos}\mathrm{\Omega }t)e^{i(\kappa _mz+\varphi _m)}`$. Setting $`V_m=v_me^{\lambda z}`$, $`W_m=w_me^{\lambda z}`$, we obtain two eigenvalue problems $`\widehat{L}_{}\widehat{L}_+\stackrel{}{v}=\lambda ^2\stackrel{}{v}`$ and $`\widehat{L}_+\widehat{L}_{}\stackrel{}{w}=\lambda ^2\stackrel{}{w}`$, where $`\stackrel{}{v}=(v_1,v_2,v_3)^T`$, $`\stackrel{}{w}=(w_1,w_2,w_3)^T`$ and
$`\widehat{L}_\pm =\left[\begin{array}{ccc}\pm \widehat{L}_1& A_3& \pm A_2\\ A_3& \pm \widehat{L}_2& \pm A_1\\ \pm A_2& \pm A_1& \pm \widehat{L}_3\end{array}\right],`$ (7)
$`\widehat{L}_m=\alpha _m_x^2\gamma _m\mathrm{\Omega }^2\xi _m`$, where $`\xi _{1,2}=\kappa \pm \delta `$ and $`\xi _3=2\kappa +\beta `$. Note that $`\widehat{L}_\pm `$ are self-adjoint, so $`\widehat{L}_{}\widehat{L}_+`$ and $`\widehat{L}_+\widehat{L}_{}`$ are adjoint operators with identical spectra. It is thus enough to consider the spectrum of one of these operators, e.g. $`\widehat{L}_+\widehat{L}_{}`$. We are particularly interested in the discrete spectrum, with eigenfunctions exponentially decaying at $`x\pm \mathrm{}`$.
In general the stability problem can only be solved numerically, but for small absolute values of $`\lambda `$ we can obtain some analytical results. Our two phase symmetries, plus the Galilean one, generate three neutrally stable $`(\lambda =0)`$ eigenmodes of $`\widehat{L}_+\widehat{L}_{}`$ at $`\mathrm{\Omega }=0`$. These are: $`\stackrel{}{w}_\phi =(A_1,A_2,2A_3)^T`$, $`\stackrel{}{w}_\psi =(A_1,A_2,0)^T`$, $`\stackrel{}{w}_x=(xA_1,xA_2,2xA_3)^T`$. Assuming that $`\mathrm{\Omega }1`$ we can express $`\stackrel{}{w}`$ as a linear combination of $`\stackrel{}{w}_\phi `$, $`\stackrel{}{w}_\psi `$, $`\stackrel{}{w}_x`$ and then use an asymptotic approach to find eigenvectors and corresponding eigenvalues. This approach can be applied only if all other eigenmodes have eigenvalues obeying the condition $`|\lambda |>\mathrm{\Omega }`$. In practice this only excludes a small neighborhood of the spatial stability boundary discussed above.
Three eigenvalue pairs $`\pm \lambda `$ are obtained from solvability conditions of the first order problems. One, associated with the asymmetric eigenvector $`\stackrel{}{w}_x`$, obeys
$$\lambda _x^2\frac{2\mathrm{\Omega }^2}{Q}𝑑x\underset{m=1}{\overset{3}{}}\gamma _m(_xA_m)^2.$$
(8)
Clearly the asymmetric mode is unstable for normal GVD, which corresponds to the snake instabilities found in NLS and degenerate 3WM models.
The other two eigenvalue pairs are associated with linear combinations of the spatially symmetric vectors $`C_\phi \stackrel{}{w}_\phi +C_\psi \stackrel{}{w}_\psi `$, and thus with neck-type instabilities. They are the roots of
$$a\lambda ^4+b\mathrm{\Omega }^2\lambda ^2+c\mathrm{\Omega }^4=0,$$
(9)
where $`a=(_\delta Q_\kappa Q_u_\kappa Q_\delta Q_u)/2`$, $`b=_\kappa Q(\gamma _1Q_1+\gamma _2Q_2)+_\delta Q_u(\gamma _1Q_1+\gamma _2Q_2+4\gamma _3Q_3)+(_\kappa Q_u+_\delta Q)(\gamma _2Q_2\gamma _1Q_1)`$, and $`c=8(\gamma _1\gamma _2Q_1Q_2+\gamma _2\gamma _3Q_2Q_3+\gamma _1\gamma _3Q_1Q_3)`$. These expressions are quite complicated, but yield some important general results. Clearly $`c`$ is negative when all $`\gamma _m`$ of the same sign. Since $`a>0`$ throughout the spatially monostable domain, it follows that the two roots $`\lambda ^2`$ are always real and of opposite sign, so that there is always an unstable neck-type mode. Thus we establish coexistence and competition of neck and snake instabilities for normal GVD. This is a novel feature of the present model, quite different from previous analytical results for NLS and degenerate 3WM , where the snake instability is the only one for normal GVD. This is because only one symmetric neutral mode exists in models with a single phase symmetry, and it generates instability only for anomalous GVD. In non-degenerate 3WM there are two neck modes, one stable for normal GVD and unstable for anomalous GVD, and vice versa, since $`b`$ is odd in the $`\gamma _m`$.
Simple analytic expressions for growth rates of these neck modes can be obtained in several special cases, e.g. in the case of second harmonic generation ($`\omega _1=\omega _2`$). Setting $`\gamma _1=\gamma _2`$ and $`\delta =0`$ (readily achieved in experiment ) the two eigenmodes have either $`C_\phi =0`$ or $`C_\psi =0`$, with eigenvalues
$$\lambda _\psi ^22\gamma _1\mathrm{\Omega }^2\frac{Q_1}{_\delta Q_1},\lambda _\phi ^2\frac{4\mathrm{\Omega }^2}{_\kappa Q}(\gamma _1Q_1+2\gamma _3Q_3).$$
(10)
For $`\delta =0`$, $`_\kappa Q`$ is positive and $`_\delta Q_1`$ is negative in the spatially monostable domain. Thus the novel neck instability for $`\gamma _1<0`$ can be directly attributed to the gauge symmetry in the differential phase $`\psi `$ and its associated neutral mode $`\stackrel{}{w}_\psi `$. On the other hand $`\lambda _\phi `$ is associated with the usual neck MI for anomalous GVD in models with a single gauge symmetry . The expression for $`\lambda _\psi ^2`$ holds also (when $`Q_u=0`$) for other solitonic models with a differential phase symmetry. Note, however, that $`_\delta Q_1`$ can generally have either sign, leading to instability with either normal or anomalous GVD.
Solving the eigenvalue problem numerically, we find that in low-frequency limit the instability growth rates precisely match those predicted by our perturbation theory, see Fig. 1(a), (b). As $`\mathrm{\Omega }`$ is increased each MI gain curve reaches a maximum and then decreases. A typical example of the maximal MI growth rate vs $`\beta `$ is presented in Fig. 1(c). Similar plots for $`Q_u0`$ and across wide range of $`\gamma _m`$ values show the same behaviour . Thus we conclude that for normal GVD the new neck instability strongly dominates the snake one. Note that its growth rate is maximised, as Fig. 1(a), (b) illustrate, for $`Q_u=0`$. For normal GVD the unstable eigenfunctions become weakly confined and develop oscillating tails as $`\mathrm{\Omega }`$ increases. Because this increases computer demand, we have plotted growth rates in Fig. 1(a) only for $`\mathrm{\Omega }`$ values corresponding to well-localised eigenmodes. Physically, broader eigenmodes have weaker overlap with the soliton, and hence lesser gain.
Spatial profiles of the symmetric eigenfunctions at maximum gain ($`\mathrm{\Omega }=\mathrm{\Omega }_{max}`$) are presented in Fig. 2. Despite this being well beyond the perturbative limit in which expressions (4) apply, the novel neck MI eigenmode still has qualitatively the same form as $`\stackrel{}{w}_\psi `$, i.e. $`w_1=w_2`$, $`w_3=0`$, indicating that the $`\psi `$ phase symmetry underlies the instability through the whole range of $`\mathrm{\Omega }`$. Similarly, the ustable neck mode for anomalous GVD is evidently associated with the $`\phi `$ symmetry.
To test our linear stability analysis and study the nonlinear evolution we performed an extensive series of computer simulations of the system (1) with initial conditions in form of a soliton stripe perturbed by spatio-temporal white noise of order $`1\%`$. Typical simulation results are presented in Fig. 3 and they fully support our predictions. We chose the size of the computational window in the time domain to be $`18\pi /\mathrm{\Omega }_{max}`$, and the initial soliton stripe rapidly develops nine humps, in accord with the stability analysis. During further evolution the modulated stripe forms into a train of pulses which either spread (normal GVD) or form a persistent chain of three-wave optical bullets (anomalous GVD). Due to the initial noise, modes from a band of frequencies close to $`\mathrm{\Omega }_{max}`$ are able to grow and compete, and hence the modulations in Fig. 3 are somewhat irregular.
A striking difference between Figs. 3(a,c) is that the initially imposed translational symmetry of the solitary stripe along the time dimension is broken in different ways. For normal GVD interleaved intensity peaks of $`E_1`$ and $`E_2`$ are formed, while for anomalous GVD the intensity peaks coincide. (Each amplitude is modulated with period $`2\pi /\mathrm{\Omega }_{max}`$.)
This difference is directly related with the spatial form of the most unstable eigenvectors. In the case of normal GVD $`w_1`$ and $`w_2`$ are out of phase and $`w_3=0`$, see Fig. 2(a), leading to the interleaving. Since $`E_1E_2`$ drives $`E_3`$, the intensity profile of the second harmonic becomes modulated with period $`\pi /\mathrm{\Omega }_{max}`$, see Fig. 3(a<sub>3</sub>). Because the overlap of the three fields is diminished by this evolution, mutual trapping becomes impossible and the whole structure eventually spreads through diffraction and dispersion, see Fig. 3(b). For anomalous GVD, all three components of most unstable eigenvector are in phase, see Fig. 2(b), and thus all three intensities become modulated with the same temporal period, see Fig. 3 (c<sub>1</sub>), (c<sub>2</sub>), (c<sub>3</sub>). This provides conditions for mutual self-trapping of the filaments, see Fig. 3(d).
The predicted instabilities can be experimentally observed in a nonlinear material longer than the MI gain length $`l_g\lambda ^1`$ and with pulse width order of several $`2\pi /\mathrm{\Omega }_{max}`$ or more. Following Ref. , typical values for KTP $`l_g1cm`$ and $`2\pi /\mathrm{\Omega }_{max}10^{12}s`$. Fig. 1(c) shows that the new neck instability has the highest gain and is thus most easily observable. Artificially birefrigent semiconductor materials, which are highly nonlinear, seem quite promising, based on a recent experiments on 3WM . Waveguides containing Bragg structures might be very suitable for observation of the MI phenomena predicted here because of their large and controllable dispersion .
In summary, we have analysed and described dispersive MI of spatial solitons due to non-degenerate 3WM. Using this model as a typical example we generalised a previous analytical approach to MI to models with two phase symmetries. We found that the extra neutral mode associated with the additional phase symmetry gives rise to a new branch of MI. This is symmetric (of neck type), and is found to dominate the asymmetric (snake) instability which is the only MI for normal GVD in systems possessing just one phase symmetry. This result enables a new understanding of the dynamics of multi-component solitary waves in terms of their phase symmetry properties. The MI phenomena which we have described for 3WM are likely to be generic in other solitonic and nonlinear wave models with two phase symmetries.
This work was partially supported by EPSRC grant GR/L 27916.
|
no-problem/9906/hep-ex9906042.html
|
ar5iv
|
text
|
# A Measurement of 𝑥𝐹₃^𝜈-𝑥𝐹₃^𝜈̄ and 𝑅 with the CCFR Detector To be published in proceedings of the 7th International Workshop on Deep Inelastic Scattering and QCD, Zeuthen, Germany, Apr. 1999.
## 1 Introduction
Wide band neutrino beams provide an unique opportunity to measure $`R`$, the ratio of the longitudinal and transverse structure functions, and the difference between $`xF_3^\nu `$ and $`xF_3^{\overline{\nu }}`$. A measurement of $`R`$ provides a test of perturbative QCD at large $`x`$, and probes the gluon density at small $`x`$. A non-zero $`\mathrm{\Delta }xF_3=xF_3^\nu xF_3^{\overline{\nu }}`$ originates from the difference between the contributions of the strange and charm seas in the nucleon to massive charm quark production. Thus, a measurement of $`\mathrm{\Delta }xF_3`$ can be used to extract the strange sea, and to understand massive charm production in neutrino-nucleon scattering. Previously, information on the strange sea came from the exclusive dimuon events channel ($`\nu N\mu ^{}cX`$, $`c`$ $``$ $`\mu ^+X^{}`$. The dimuon analysis requires both an understanding of charm fragmentation and an acceptance correction for the detection of muons from charm decays. In previous extractions of structure functions , a leading order slow rescaling correction for heavy charm production was applied in order to extract a theoretically corrected $`F_2^\nu `$ which could be directly compared with $`F_2^\mu `$. However, because of theoretical uncertainties in the treatment of massive charm production , this correction is no longer applied in this analysis. Furthermore, the value of $`\mathrm{\Delta }xF_3`$ plays a crucial role in the extraction of $`F_2`$ at low $`x`$, where there is a long-standing discrepancy between CCFR and NMC $`F_2`$ data. Therefore, it is important to measure the structure functions without any model-dependent slow rescaling corrections.
## 2 The CCFR experiment
The CCFR experiment collected data using the Fermilab Tevatron Quad-Triplet wide-band $`\nu _\mu `$ and $`\overline{\nu }_\mu `$ beam. The CCFR detector consists of an unmagnetized steel-scintillator target calorimeter instrumented with drift chambers, followed by a toroidally magnetized muon spectrometer. The hadron energy resolution is $`\mathrm{\Delta }E/E=0.85/\sqrt{E}`$(GeV), and the muon momentum resolution is $`\mathrm{\Delta }p/p=0.11`$. By measuring the hadronic energy ($`E_h`$), muon momentum ($`p_\mu `$), and muon angle ($`\theta _\mu `$), we construct three independent kinematic variables $`x`$, $`Q^2`$, and $`y`$. The relative flux at different energies obtained from the events with low hadron energy ($`E_h<20`$ GeV) is normalized so that the neutrino total cross section equals the world average $`\sigma ^{\nu N}/E=(0.677\pm 0.014)\times 10^{38}`$ cm<sup>2</sup>/GeV and $`\sigma ^{\overline{\nu }N}/\sigma ^{\nu N}=0.499\pm 0.005`$ . The total data sample used for the extraction of structure functions consists of 1,030,000 $`\nu _\mu `$ and 179,000 $`\overline{\nu }_\mu `$ events after fiducial and kinematic cuts ($`E_\mu >15`$ GeV, $`\theta _\mu <0.150`$, $`E_h>10`$ GeV, and 30 GeV $`<E_\nu <`$ 360 GeV). Dimuon events are removed because of the ambiguous identification of the leading muon for high-$`y`$ events.
## 3 Measurement of differential cross sections
The differential cross sections are determined in bins of $`x`$, $`y`$, and $`E_\nu `$ ($`0.01<x<0.65`$, $`0.05<y<0.95`$, and $`30<E_\nu <360`$ GeV). Over the entire $`x`$ region, differential cross sections are in good agreement with NLO Mixed Flavour Scheme (MFS) QCD calculation using the MRST PDFs. This calculation includes an improved treatment of massive charm production. Figure 1 shows typical differential cross sections at $`E_\nu =150`$ GeV. Also shown are the prediction from a LO QCD model, which is only used in the calculation of acceptance and resolution smearing correction. The CCFR data exhibit a quadratic $`y`$ dependence at small $`x`$ for neutrino and anti-neutrino, and a flat $`y`$ distribution at high $`x`$ for the neutrino cross section. Note that the $`y`$ distributions of the CDHSW differential cross sections data do not agree with those of CCFR data, or with the MRST predictions (this difference is crucial in any QCD analysis).
## 4 Measurement of $`\mathrm{\Delta }xF_3`$ and $`R`$
Values of $`\mathrm{\Delta }xF_3`$ and $`R`$ are extracted from the sums of neutrino and anti-neutrino differential cross sections. The sum of the two differential cross sections can be written as:
| $`F(ϵ)`$ | $`\left[\frac{d^2\sigma ^\nu }{dxdy}+\frac{d^2\sigma ^{\overline{\nu }}}{dxdy}\right]\frac{(1ϵ)\pi }{y^2G_F^2ME}`$ |
| --- | --- |
| | $`=2xF_1[1+ϵR]\frac{y(1y/2)}{1+(1y)^2}\mathrm{\Delta }xF_3`$ |
where $`ϵ2(1y)/(1+(1y)^2)`$ is the polarization of virtual $`W`$ boson. To fit $`\mathrm{\Delta }xF_3`$, $`R`$, and $`2xF_1`$ at a given $`x`$ and $`Q^2`$ is very challenging because of the strong correlation between the $`\mathrm{\Delta }xF_3`$ and $`R`$ terms, unless the full range of $`ϵ`$ is covered by the data. Covering this range (especially the low $`ϵ`$ region correspoing to high $`y`$) is difficult because of the low acceptance at high $`y`$.
Since the contribution of $`\mathrm{\Delta }xF_3`$ to $`F(ϵ)`$ increases with $`Q^2`$ while that from $`R`$ decreases, our strategy is to fit $`\mathrm{\Delta }xF_3`$ and $`2xF_1`$ for $`Q^2>5`$ where the $`\mathrm{\Delta }xF_3`$ contribution is relatively larger, constraining $`R`$ with $`R_{world}`$ . ($`\mathrm{\Delta }xF_3`$ fits are done only for $`x<0.1`$, because its contribution is small for $`x>0.1`$.) It is reasonable to use $`R_{world}`$ (an empirical fit for all available $`R`$ data), because above $`Q^2=5`$, $`R`$ for neutrino and muon scattering are expected to be the same, and $`R_{world}`$ is in good agreement with the NMC muon data .
In the $`Q^2<5`$ region, where the contribution from $`R`$ is larger, we fit $`R`$ and $`2xF_1`$ with $`\mathrm{\Delta }xF_3`$ constrained to the model predictions. This allows us to study the $`Q^2`$ dependence of $`R`$ for neutrino scattering at low $`Q^2`$. In the $`Q^2>5`$ region, we test which scheme is favored for the massive charm treatment, based on the extracted $`\mathrm{\Delta }xF_3`$ results. Then the favored scheme is used in constraining $`\mathrm{\Delta }xF_3`$ for the $`Q^2<5`$ analysis.
Before the structure function extraction, we apply an electroweak radiative correction (Bardin) and corrections for the $`W`$ boson propagator, and a non-isoscalar target (the 6.85% excess of neutrons over protons in iron; this effect is valid only at high $`x`$). The values of $`\mathrm{\Delta }xF_3`$ and $`R`$ are sensitive to the energy dependence of the neutrino flux, but are insensitive to the absolute normalization. The uncertainty on the flux shape is estimated by using the constraint that $`F_2`$ and $`xF_3`$ should be flat over $`y`$ (or $`E_\nu `$) for each $`x`$ and $`Q^2`$ bin.
In the $`\mathrm{\Delta }xF_3`$ fit, we use the NLO MFS calculation with MRST PDFs for the $`Q^2`$ dependence of $`\mathrm{\Delta }xF_3`$ (for $`Q^2>5`$), and fit for the level of $`\mathrm{\Delta }xF_3`$ for each $`x`$ bin below $`x=0.1`$. Figure 2 shows the $`\mathrm{\Delta }xF_3/2`$ data as a function of $`x`$. The CCFR data favor the Fixed Flavour Scheme (FFS) calculation with the GRV94 PDFs, and also the Mixed Flavour Scheme (MFS) using MRST PDFs. The data do not favor the VFS calculation with CTEQ4HQ as implemented by Kramer, Lampe, and Spiesberger (KLS) . The difference among the various schemes does not come from the different parton distributions, but rather from the different treatments of massive charm. In fact, all strange sea distributions in these parton distributions agree with the CCFR dimuon data . Recently Olness and Kretzer pointed out that KLS did not include the LO charm sea diagram in the VFS implementation, as required in the original ACOT’s VFS calculation . The ACOT’s VFS calculations agree with data after the inclusion of the charm sea contribution, as shown in Fig. 2.
Since all three schemes except KLS’ VFS show a good agreement with data, we use a MFS calculation with MRST PDFs (arguably the most reasonable theoretical description of massive charm production currently available) to constrain $`\mathrm{\Delta }xF_3`$ for the extraction of $`R`$ for $`Q^2<5`$ and $`x<0.1`$. Even for $`Q^2>5`$, we extract $`R`$ in order to investigate its $`Q^2`$ dependence. The extracted values of $`R`$ at fixed $`x`$ vs $`Q^2`$ are shown in Fig. 3. The new data reveal the $`Q^2`$ dependence of $`R`$ at $`x<0.1`$ for the first time ($`R`$ decreases as $`Q^2`$ increases for fixed $`x`$). The NMC data shown in Fig. 3 are integrated over $`Q^2`$, and the two nearest NMC $`x`$ bins are shown together. At higher $`x`$, our measurements agree with the other world data for $`R`$ . Figure 3 also shows a comparison with $`R_{world}`$ with $`m_c=0`$ (muon scattering or with slow rescaling correction in neutrino scattering) and $`m_c=1.3`$ (without slow rescaling correction in neutrino scattering). The CCFR $`R`$ data at $`x=0.015`$ do not show $`R`$ approaching zero as $`Q^2`$ goes zero (as expected in neutrino scattering, but not in electron scattering).
## 5 Conclusions
New measurements of $`\mathrm{\Delta }xF_3`$ and $`R`$ have been extracted from the CCFR differential cross section data. The $`R`$ data are extended to lower $`x`$ and higher $`Q^2`$ than previous measurements. The $`Q^2`$ dependence of $`R`$ at lower $`x`$ region has been measured for the first time. The $`\mathrm{\Delta }xF_3`$ data from $`Q^2>5`$ region agree with various schemes for the treatment of massive charm production. Further reduction of the errors in $`\mathrm{\Delta }xF_3`$ is expected by including lower $`Q^2`$ data. The effect of $`\mathrm{\Delta }xF_3`$ on the extraction of $`F_2`$ is currently under study. In addition, new data from the recent NuTeV run (1996-97), taken with sign selected neutrino beams, are expected to yield more precise determinations of $`\mathrm{\Delta }xF_3`$, $`F_2`$ and $`R`$ at low $`x`$.
|
no-problem/9906/astro-ph9906437.html
|
ar5iv
|
text
|
# Optical Spectroscopy of GX 339-4 during the High-Soft and Low-Hard States
## 1 Introduction
Almost all known black hole candidates (BHCs) in our Galaxy are X-ray transients. Although the triggering for the X-ray outbursts is not fully understood, models invoking accretion disk instability seem to provide an explanation. Other important issues concerning the BHCs are the transition between the X-ray spectral states and its relation to the physical conditions at the accretion disk.
GX 339$``$4 is an X-ray transient (Markert et al. 1973; Harmon et al. 1994), and is also a radio source (e.g. Hannikainen et al. 1998) which shows jet-like features (Fender et al. 1997). It is classified as a BHC, because of its short-term X-ray and optical variability (Makishima et al. 1986), the transition between high-soft and low-hard X-ray spectral states (Markert et al. 1973), and the extended high energy power-law tail in its X-ray spectrum (Rubin et al. 1998). In fact it is also one of the few BHCs that have shown four X-ray spectral states: off, low-hard, high-soft and ultra-high. Its off state is generally characterised by very weak, hard X-ray emission. The optical counterpart is faint with $`V19`$–21. In the low-hard state, the source has a very hard X-ray spectrum, with an extended power-law component with a photon index $`1.5`$. The 2–10 keV X-ray flux is about $`0.4\times 10^9\mathrm{erg}\mathrm{cm}^2\mathrm{s}^1`$. The optical brightness is $`V16`$. In the high-soft state, the X-ray spectrum is dominated by a soft thermal component. The power-law tail is weaker than that in the low-hard state and has a photon index $`2`$. The 2$``$10 keV X-ray flux is about 20 times higher than that of the low-hard state. The optical brightness is $`V16`$, similar to that in the low-hard state. In the ultra-high state, both the thermal and the power-law components are very strong in the X-ray spectrum. The 2$``$10 keV X-ray flux is about 50 times the X-ray flux in the low state, and the photon index of the power law is $`2.5`$. On one occasion Mendez & van der Klis (1997) reported that the source was in a state intermediate between the low-hard and the high-soft state in which the 2$``$10 keV X-ray flux is about a factor of 5 below the X-ray flux in the high state. (For reviews of the optical and X-ray properties of the system, see e.g. Motch et al. 1985; Ilovaisky et al. 1986; Makishima et al. 1986; Corbet et al. 1987; Tanaka & Lewin 1995; Mendez & van der Klis 1997).
Despite the fact that GX 339$``$4 is optically bright ($`V16`$) when it is X-ray active, it is not well studied in the optical bands. Its mass function has not yet been determined, and so the black hole candidacy is not verified in terms of the orbital dynamics. From a photometric observation during the off state, Callanan et al. (1992) detected a periodicity of 14.8 hr, and attributed it to the orbital period. If this is true, the companion star would be a late-type star with a mass $`<1.6`$ M.
Optical spectra taken by Smith, Filippenko & Leonard (1999) in 1996 May, when the system was in a low-hard state, show a strong H$`\alpha `$ emission line with a broad flat-topped profile. Within the resolution and the signal-to-noise ratio limit of the spectra, the line resembles the double-peaked lines that characterise the accretion disk in binary systems. By fitting two gaussians to the H$`\alpha `$ line profile, a velocity separation of $`370\pm 40`$ km $`\mathrm{s}^1`$ is deduced. As the separation of the peak in this data set is not as clearly seen as in the other BHCs, e.g., GRO J1655$``$40 (Soria, Wu & Hunstead 1999) and A0620$``$00 (Johnston, Kulkarni & Oke 1989), there is a possibility that the flat-topped profile seen in 1996 May is not intrinsically double-peaked, just as in the case of Cyg X-1 (Smith, Filippenko & Leonard 1999).
Here we report spectroscopic observations of GX 339$``$4 which we carried out in 1997 May and in 1998 April and August. The system was in a low-hard state in 1997 May and was in a high-soft state in 1998 April and August (see Fig. 1). During the 1998 August observations, we also obtained simultaneous photometric data. Our data show that GX 339$``$4 has distinct optical spectral features in the low-hard and in the high-soft states, an indication of different physical conditions in the line emission regions or perhaps different line emission regions in the two X-ray spectral states.
## 2 Observations and Data Reduction
The 1997 May 6 and 8 observations of GX 339$``$4 were carried out with the RGO spectrograph and Tektronix 1k$`\times `$1k thinned CCD on the 3.9 m Anglo-Australian Telescope (AAT). The seeing condition was 2 arcsec. Series of 600 sec spectra were taken in the $`53556950`$ Å region. A grating with 300 grooves/mm was used with the 25 cm camera, giving a resolution of 3 Å FWHM.
On 1998 April 28 $``$ 30 we observed the system again, with the Double Beam Spectrograph (DBS) on the ANU 2.3 m Telescope at Siding Spring Observatory. The detectors on the two arms of the spectrograph were SITe 1752$`\times `$532 CCDs. Gratings with 1200 grooves/mm were used for both the blue ($`41505115`$ Å) and the red ($`62007150`$ Å) bands, giving a resolution of $`1.3`$ Å FWHM. The average seeing on each of the three nights was about 2 arcsec. Further observations were carried out on 1998 August 20 and 23, with the same instrumental setup as in the April observations. Simultaneous photometric observation were conducted with the ANU 40 in telescope on 1998 August 20 and 23. The seeing was 2 and 3 arcsec on the two nights respectively. A log of our spectroscopic observations is shown in Table 1.
Standard data reduction procedures were followed, using the IRAF tasks. After removing the bias and pixel-to-pixel gain variations from each exposure, we subtracted the sky background and extracted the spectra with the APALL routine. CuAr and FeAr lamp spectra were taken to carry out wavelength calibration. We removed atmospheric spectral features using the standard star LTT 7379.
## 3 Results
### 3.1 Spectrum in the low-hard state
Figure 2 shows the summed spectrum from the observations we conducted on 1997 May 6. The total exposure time was 19,800 sec. The H$`\alpha `$, He I $`\lambda `$ 6678 and He I $`\lambda `$ 5876 emission lines are easily identified. The emission feature at about 6500 Å is probably the N II $`\lambda `$ 6505 emission line. (The main features identified in the spectra and their equivalent widths are listed in Table 2.) Our data do not reveal two clearly resolved peaks for H$`\alpha `$, He I $`\lambda `$ 5876 and He I $`\lambda `$ 6678. Although the suspected N II $`\lambda `$ 6505 line appears to have two maxima, it is uncertain whether they are purely statistical or intrinsic. We do not detect the Li I $`\lambda `$ 6708 absorption line, which was present in the spectra of the BHCs GRO J1655$``$40, A 0620$``$00, V404 Cyg and Cen X-4 (Martin et al. 1992; Shahbaz et al. 1999; Smith et al. 1999), nor any other obvious stellar absorption lines. All absorption features in our spectra appear to be interstellar.
The H$`\alpha `$ line is asymmetric, with its peak slightly skewed towards the red. The line profile is more appropriately described as asymmetric and round-topped rather than as unevenly double-peaked. The equivalent width (EW) of the line is $`7.2\pm 0.3`$ Å (negative values are taken to indicate emission). The EW and the overall shape are similar to those found in the spectra obtained by Smith et al. (1999) on 1996 May 12, when the system was also in a low-hard state, except for the fact that in those observations the line peak was skewed towards the blue instead. We considered a two-gaussian fit using the QDP routine (Tennant 1991) and obtained the gaussian centres at 6560.15 and 6567.57 Å. If we assume that the line truly has a double-peaked profile, the deconvolved peak separation is then roughly 7.4 Å. This value is consistent with the $`8.0\pm 0.8`$ Å separation obtained by Smith et al. (1999) with a two-gaussian fit to their 1996 data.
### 3.2 Spectrum in the high-soft state
In Figure 3 we show the summed spectrum in the H$`\alpha `$ region for the data obtained on 1998 April 28 $``$ 30; in Figure 4, we show the summed spectrum in the the H$`\beta `$/He II region for the data obtained on the same nights. The total exposure times in both case were 25,500 sec on the first night, 19,500 sec on the second and 4,000 on the third. Strong H I Balmer emission lines are seen in the spectra. Other prominent emission lines are He I $`\lambda `$ 6678, He I $`\lambda `$ 7065, He II $`\lambda `$ 4686 and N III $`\lambda \lambda `$ 4641,4642.
The H$`\alpha `$, H$`\beta `$, H$`\gamma `$, He I $`\lambda `$ 6678, He II $`\lambda `$ 4686 and N III $`\lambda \lambda `$ 4641,4642 lines all show double-peaked profiles, and the peaks are resolved visually. The peaks are still clearly separated even when we bin the data to mimic a spectrum with a resolution of 4 Å FWHM, lower than the 3 Å resolution of our 1997 May spectra (see Fig. 2, cf. also Smith et al. 1999). It is therefore the first time that double-peaked lines are unambiguously detected in the optical spectra of GX 339$``$4. The peak separations of the emission lines are listed in Table 3.
We note that while H$`\alpha `$ and H$`\beta `$ has similar velocity separation, that of H$`\gamma `$ is significantly larger. As the signal-to-noise ratio of the H$`\gamma `$ line is relatively low, the unexpectedly large velocity separation is probably caused by contamination.
The general features of the 1998 August spectra (not shown) are similar to those in the 1998 April spectra. The H$`\alpha `$, H$`\beta `$, H$`\gamma `$, He I $`\lambda `$ 6678, He II $`\lambda `$ 4686 and N III $`\lambda \lambda `$ 4641,4642 lines are prominently seen in emission and generally have double-peaked profiles. Because of poorer observing conditions – cloudy nights – the peaks are not as well resolved as those appearing in the 1998 April spectra.
### 3.3 Line profile morphology
In Figure 5, we show a sequence of H$`\alpha `$ line profiles from the 1997 May 6 and 8 observations. The H$`\alpha `$ line profiles are asymmetric, with a slightly red-shifted peak and a round top. There is no strong indication of double-peaked profiles in any set of data obtained in 1997 May (cf. the H$`\alpha `$ and the He II 4686 lines in Fig. 6, where the double-peaked structure is instead very clear). The line shows detectable variations from night to night. If the line profiles are fitted with a single gaussian, one may obtain a series of velocity shifts. Judging from the central location of the line bases, we argue that such variations are statistical, i.e. our data do not show systematic velocity modulation due to orbital motion. The base-width of the line does not show obvious variations over the two nights of observations, either. The variations in the line top are therefore likely to be due to opacity rather than kinematic effects.
The signal-to-noise ratios of the He I $`\lambda `$ 6678 line profiles in the 1997 May observations (not shown) are not as good as those measured at H$`\alpha `$. In spite of this, the properties of the He I $`\lambda `$ 6678 line profile sequence is still consistent with our “argument” of no systematic velocity modulations and no line-width variations.
The sequence of H$`\alpha `$ and He II $`\lambda `$ 4686 line profiles from the 1998 April observations are shown in Figure 6. The presence of a double-peak in each profile is not as obvious as it appears in the summed spectra. However, the trace of two peaks can still be easily seen. The relative prominence of the two peaks varies from one spectrum to another. As changes of the relative contribution of the two peaks have also been observed in other BHCs, e.g. GRO J1665$``$40, when they were in a high-soft state (see Soria, Wu & Hunstead 1999), this may imply that similar activities are occurring in the accretion discs of these BHCs during that state.
The profile of the H$`\alpha `$ line is slightly skewed, with an apparently wider blue wing. We suspect that the skewness is caused by the presence of a weak broad component whose red wing is partially obscured/absorbed along the line of sight, an indication that opacity effects play a significant role in the high-soft state as well as in the low-hard state. A hint of a third peak superimposed on the blue wing is detected in some spectra, particularly in the last series of observations from April 28. The profile sequence in Figure 5 reveals that the strength of the emission above the continuum increased from the first to the second night, and then decreased, to the previous levels, from the second to the third night. The average EW of the H$`\alpha `$ line was $`6.6\pm 0.3`$ Å on April 28, $`9.0\pm 0.2`$ Å on April 29, and $`6.0\pm 0.3`$ Å on April 30. We do not have photometric data from 1998 April, and we are therefore unable to determine whether the variations in the EW are a consequence of changes in the intensity of the continuum or of the emission line itself.
The presence of two peaks in the He II $`\lambda `$ 4686 line is more obvious than in the case of H$`\alpha `$. They are clearly separated most of the time, and their separation is wider than that of the two peaks in the H$`\alpha `$ line profile. The red and blue wings are more symmetric with respect to each other, in comparison with the wings of the H$`\alpha `$ line. The average EW of He II $`\lambda `$ 4686 was $`5.5\pm 0.2`$ Å on April 28, $`6.0\pm 0.4`$ Å on April 29, and $`4.5\pm 0.3`$ Å on April 30.
In Figure 7 we show the profiles of the H$`\alpha `$ and the He II $`\lambda `$ 4686 lines from the 1998 August observations. The corresponding peak locations are consistent with those found in April. The red peak is generally weaker than that observed in 1998 April, while the blue peak seems to be stronger.
## 4 Discussion
### 4.1 Accretion disk and line emission
The non-detection of Li I $`\lambda `$ 6708 and of other stellar absorption lines above the noise level (S/N $`30`$) suggests that the emission from the companion star gives a negligible contribution to the optical continuum. This is not surprising: in 1981 the system was observed at an optical brightness as low as $`V21`$ mag (Hutchings, Cowley & Crampton 1981; Ilovaisky & Chevalier 1981), which is approximately the brightness of a $`1M_{}`$ main sequence star at a distance of $`4`$ kpc (Zdziarski et al. 1998); if that is the intrinsic brightness of the companion star, it would have contributed only $`1/50`$ of the optical flux detected in our observations ($`V16.5`$).
The spectra are dominated by emission lines in both the low-hard and the high-soft states. We argue that these lines are emitted from the accretion disc and/or its corona and wind. Thus, the accretion disc is present and active in both X-ray spectral states.
The emission lines in the 1997 May spectra appear to have a round-topped profile, skewed towards the red. Although we cannot rule out the interpretation that the round-topped lines actually consist of two components, we do not see clearly resolved peaks in our data. An alternative explanation is that the round-topped lines are the consequences of an outflow, a dense wind from an inflated non-Keplerian disk or an evaporating corona. If the outflowing matter is sufficiently dense to produce a large opacity, lines emitted from the accretion disc beneath can be masked. The lines will then be dominated by emission from the outflowing material instead of by emission from the accretion disc. As the kinematic velocity of the outflowing matter should be of the same order of magnitude as the local Keplerian velocities, the widths of the lines from the outflowing material may not differ much from the widths of lines from the accretion disc. However, the lines will not show double-peaked profiles, and will instead appear to be round-topped or flat topped, depending on the outflow velocity and density profiles (see Chapter 14 in Mihalas 1978), similar to those of lines from windy massive stars.
The 1998 spectra show a strong optical continuum originated from an optically thick accretion disc. For a Keplerian disc, the rotational velocity and the temperature increase radially inward. If the disc is only viscously heated at the middle layer of the disc, a temperature gradient will be set up, such that the temperature decreases from the central plane to the disc surface. This will result in a spectrum dominated by absorption lines (see e.g. la Dous 1989). However, if there is a temperature inversion near the disc surface — which may be caused by external irradiation — the spectrum will then be dominated by emission lines. The 1998 April spectra clearly show prominent emission lines with resolved double-peaks superimposed on a strong continuum. This implies that the accretion disc is irradiatively heated. The increase in the velocity separation with the ionisation state of the accreting matter is strong evidence that the disc is irradiated by a central source, as the higher excitation lines originate from the inner regions, where the Keplerian velocities are larger and the temperature higher due to the smaller distance from the irradiation source. For a discussion on irradiation heating of accretion disks, see e.g. Dubus et al. (1999).
The velocity separation of the H$`\alpha `$ peaks in the high-soft state is smaller than the velocity separation obtained by a two-gaussian fit to the line profiles obtained in the low-hard state (see also Smith et al. 1999). It is, however, roughly the same as the velocity separation of the H$`\beta `$ line observed in the same epoch. As the velocity separation of the peaks in the low-hard state is not well-defined, we cannot draw any firm conclusions from the comparison. However, we have found that the velocity separation of the peaks in the H$`\alpha `$ line in the high-soft state is a factor of two smaller than that observed in GRO J1655$``$40 in the same X-ray spectral state (Soria, Wu & Hunstead 1999). The latter is a high orbital inclination system ($`i70^o`$, e.g. van der Hooft et al. 1998). If the masses of the black holes in GX 339$``$4 and GRO J1655$``$40 are similar, say 7–10 M, the smaller velocity separation for GX 339$``$4 implies that either the accretion disc in GX 339$``$4 is larger or the orbital inclination of GX 339$``$4 is lower. As the companion star of GX 339$``$4 is not visible during its off state (Callanan et al. 1992), it is likely to be a low mass star, less massive than the companion star in GRO J1655$``$40. This leads us to propose that GX 339$``$4 is a system with a low orbital inclination. A low orbital inclination for GX 339$``$4 is in fact consistent with the null detection of the orbital modulation in our 1997/1998 spectroscopic data. It is also supported by the fact that the radio spectrum of GX 339$``$4 is relatively flat (Fender et al. 1997), unlike the steep spectra expected from a superluminal synchrotron jet source.
Although a 14.8-h periodicity was detected when GX 339$``$4 was observed in an off state (Callanan et al. 1992), both our 1997/1998 spectroscopic data and our 1998/1999 photometric data (Soria, Wu & Johnston 1999a,b) do not show any obvious periodicities. Our photometric observations in 1998 August show variations in the brightness of the source similar to the flaring activities seen in a previous observation by Corbet et al. (1987). On August 20 its brightness varied between $`V=16.45\pm 0.01`$ and $`V=16.52\pm 0.01`$, while on August 23 the source brightened up from $`V=16.61\pm 0.01`$ at the beginning of the night to a maximum of $`V=16.22\pm 0.01`$ 5 hours later, then started to decline again. Further photometric observations during the off state are required to verify if GX 339$``$4 has a 14.8-h orbital period. If we accept the argument that GX 339$``$4 has a low orbital inclination, the 14.8-h periodicity detected by Callanan et al. (1992) may perhaps also be due to the precession of a weak accretion disc and/or its associated jet.
### 4.2 Summary
We carried out spectroscopic observations of the BHC binary GX 339$``$4 during its low-hard and high-soft states. Our data show that the optical spectra in the low-hard state are characterised by emission lines with slightly asymmetric, round-topped profiles; on the other hand, the spectra in the high-soft state show emission lines with unambiguously resolved doubled-peaked profiles. We do not see obvious stellar absorption lines: this implies that the optical spectrum in both X-ray spectral states is dominated by emission from accreting matter around the black hole. The round-topped lines seen in the low-hard state are probably formed in opaque matter outflowing from the central black hole and/or the accretion disc. The double-peaked lines seen in the high-soft state, however, indicate the presence of a bright, active accretion disc. The trend of increasing velocity separation with line ionisation supports the model of an accretion disc irradiatively heated by soft X-rays from a central source. We do not see 14.8-h modulations in our spectroscopic data and we are therefore unable to verify the 14.8-h orbital period. The null detection of the 14.8-h periodicity and of stellar absorption lines, together with the relatively small velocity separation of the line peaks, lead us to believe that GX 339$``$4 is low-mass system with a low orbital inclination.
## 5 Acknowledgements
We thank Richard Hunstead for discussions and Michelle Buxton for taking the 1998 August 23 spectra. KW acknowledges the support from the Australian Research Council through an Australian Research Fellowship and an ARC grant.
|
no-problem/9906/hep-th9906020.html
|
ar5iv
|
text
|
# Mirror symmetry for 𝒩=1 QED in three dimensions
## I Introduction
Mirror symmetry provides a relation between the infrared descriptions of different supersymmetric gauge theories in three dimensions. The first example of this kind was discovered in Ref. . $`𝒩=4`$ supersymmetric QED with $`N_f`$ flavors has a dual description in terms of another three dimensional gauge theory. The Coulomb branch of one theory is mapped into the Higgs branch of the other and vice versa, which implies that the IR descriptions of these theories agree after suitable identifications. The exact change of variables relating these theories in the infrared is not known, but recently some progress in that direction has been made .
While mirror symmetry was first discovered in a purely field theoretic context, it proved very useful to rederive and generalize the results of Ref. using brane configurations in type IIB string theory (see also for different approaches and for a list of such brane configurations). The three-dimensional field theory is realized on D3-branes that terminate on NS5-branes and intersect D5-branes. The S-dual of this configuration gives rise to the mirror theory.
Mirror symmetry can be generalized to include theories with $`𝒩=2`$ supersymmetry in three dimensions . These theories can also be realized on D3-branes in the presence of D5- and NS5-branes, and S-duality relates brane constructions that give rise to theories that are mirror pairs. A different kind of IR duality between such $`𝒩=2`$ theories was discussed in .
$`𝒩=4,2`$ theories in three dimensions can be obtained by dimensional reduction of four-dimensional $`𝒩=2,1`$ theories. Just like in the four-dimensional case, there are strong constraints from R-symmetries, holomorphy, and non-renormalization theorems that simplify the analysis of the vacuum structure. For theories with eight supercharges ($`𝒩=4`$ in three dimensions) it is possible to show that the exact metric on the Coulomb branch of the original theory agrees with the metric on the Higgs branch of the mirror. In the $`𝒩=2`$ case (four supercharges) one can show that the complex structure of corresponding branches agrees.
In three dimensions one can also study $`𝒩=1`$ theories with two real supercharges. These theories cannot be obtained by dimensional reduction of higher dimensional supersymmetric field theories. The $`𝒩=1`$ superconformal algebra does not contain any R-symmetries, there are no constraints from holomorphy, and there are no know non-renormalization theorems. Nonetheless one can learn something about the infrared behavior of some of these theories (see e.g. ).
In this note we analyze three-dimensional $`𝒩=1`$ QED with $`N_f`$ flavors and its mirror using 3-branes and 5-branes of type IIB string theory. This theory does not have a Coulomb branch, since the $`𝒩=1`$ vector multiplet does not contain any scalars. We argue that parity invariance prevents the Higgs branch from being lifted by quantum corrections<sup>*</sup><sup>*</sup>*This argument is due to M. Berkooz, A. Kapustin, and M. Strassler . . The S-dual brane configuration gives rise to the mirror theory with a Coulomb branch that also cannot be lifted by quantum corrections. We suggest that these two theories are equivalent in the infrared. Unfortunately it is hard to assemble additional evidence for this duality, since the metric on Higgs branch receives perturbative corrections at all orders and it does not have a complex structure we could compare to that of the mirror branch. However, it seems highly plausible that a pair of S-dual brane configurations should give rise to the same infrared physics on the 3-branes.
## II $`𝒩=1`$ Mirror Symmetry
A brane configuration consisting of NS-branes in (012345), NS-branes in (012389), $`N_f`$ D5-branes in (012348), and D3-branes in (0126) preserves two supercharges and gives rise to a $`𝒩=1`$ theory in three dimensions. The brane configuration is shown in Fig. 1. It is straightforward to determine the matter content of this theory. Each D5-brane provides one $`𝒩=4`$ hypermultiplet consisting of two complex scalars, $`Q`$, $`\stackrel{~}{Q}`$, and fermions. We can combine these fields into a complex $`SU(2)`$ doublet $`𝒬=(Q,\stackrel{~}{Q}^{})`$. This $`SU(2)`$ is a remnant of the $`SU(2)_R`$ symmetry of the $`𝒩=4`$ theory, but here it acts as an ordinary global symmetry of the theory. Without the D5-brane, this configuration is $`𝒩=2`$ supersymmetric, so the theory contains an $`𝒩=2`$ vector multiplet which contains an $`𝒩=1`$ vector superfield and a real scalar superfield. The expectation value of the real scalar, $`\varphi `$, corresponds to the $`X_3`$ position of the three brane. In an $`𝒩=2`$ theory the scalar in the gauge multiplet couples to the matter fields via terms of the form $`|\varphi Q|^2+|\varphi \stackrel{~}{Q}|^2`$ in the potential. In our configuration moving the 3-branes in the $`X_3`$ direction does not give a mass to the fundamentals, so our field theory differs from the $`𝒩=2`$ case by turning off this coupling. Since we are considering a $`U(1)`$ theory, this scalar is uncharged under the gauge group and we expect it to decouple from the low energy dynamics of the gauge theory.
The brane configuration has no rotational symmetries, reflecting the fact that the $`𝒩=1`$ superconformal algebra in three dimensions does not include an R-symmetry. The Higgs branch of the theory can be seen geometrically by breaking the 3-branes on the D5-branes and separating the pieces. For example, if there is only one D5-brane, we can separate the two halfs of the 3-brane in the $`X_{3,4,8}`$ directions. The $`X_3`$ position of the center of mass corresponds to the expectation value of the decoupled scalar. We expect a three-dimensional Higgs branch in this case, because it can be parametrized by expectation value of the fundamentals modulo gauge transformations, which eliminate one of the four real parameters in $`𝒬`$.
The theory has three real mass parameters for each flavor. In the brane picture they correspond to motions of the D5-branes in $`X_{5,7,9}`$. These mass terms appear in the Lagrangian in the same way as in a $`𝒩=2`$ theory, i.e., they provide a complex and a real mass. Unlike in the $`𝒩=2`$ case, the real mass cannot be absorbed into a shift of the real scalar in the gauge multiplet, because there is no coupling of this field to the fundamentals in the theory we discuss here.
This exhausts the list of possible brane motions. In particular there is no brane motion that corresponds to turning on FI parameters in the field theory. This is consistent with our expectations, because the $`𝒩=1`$ vector multiplet does not contain any scalars.
To summarize, the classical 3-brane theory is a $`𝒩=1`$ supersymmetric $`U(1)`$ gauge theory with $`N_f`$ complex fundamentals, $`Q_i,\stackrel{~}{Q}^i`$, a decoupled real scalar, $`\varphi `$, and no potential. The classical Higgs branch has $`4N_f1`$ real dimensions. There are no non-renormalization theorems that protect the Higgs branch in theories with two supercharges, so the classical Higgs branch could be lifted by quantum corrections. However, discrete symmetries of the field theory provide some constraints .
Using the notation in Ref. , we can define a parity transformation, $`P`$, that acts on the coordinates as $`P(X_0,X_1,X_2)(X_0,X_1,X_2)`$ and on the spinors as $`P\chi \pm \sigma _3\chi `$. The Lagrangian of the theory we described above is invariant under this transformation if all fields are parity even. There is no parity anomaly, since the number of charged fields is even. This ensures that parity is preserved in the full quantum theory. The superpotential in a parity invariant theory must be parity odd, because the the measure $`d\theta ^2`$ is odd . Since the theory we consider here does not contain any parity odd fields, we conclude that no superpotential can be generated. This in turn implies that the Higgs branch cannot be lifted by quantum corrections. Of course this does not mean that there are no quantum corrections to the metric on the Higgs branch.
In order to find a candidate mirror theory of three-dimensional QED with $`N_f`$ flavors, we perform an S-duality on the brane configuration above. This turns the NS-branes into D5-branes and vice versa. The field theory on the 3-brane is now a $`U(1)^{N_f1}`$ gauge theory. Each $`U(1)`$ factor contains an $`𝒩=4`$ vector multiplet, which decomposes into three real scalar superfields and one vector superfield in $`𝒩=1`$ language. There is a $`𝒩=4`$ hypermultiplet in the bifundamental representation at every intersection of the 3-brane with an NS brane. These fields couple in the $`𝒩=4`$ supersymmetric way to the vector multiplets except at the first and last NS brane. We discuss the $`N_f=1,2`$ cases is some detail. The generalization to higher $`N_f`$ is straightforward.
For $`N_f=2`$ the mirror is a $`U(1)`$ gauge theory with two $`𝒩=4`$ hypermultiplets $`q_L`$, $`\stackrel{~}{q}_L`$ and $`q_R`$, $`\stackrel{~}{q}_R`$ from strings stretched across the left and right NS-branes respectively. The theory also contains two uncharged complex scalars, $`M_{L,R}`$. We identify their expectation values with the position of the left and right sections of the 3-brane in $`X_3+iX_4`$ and $`X_3+iX_8`$ relative to the middle section.
Expectation values for the three scalars in the $`𝒩=4`$ gauge multiplet correspond the the position of the middle section of the 3-brane in $`X_{3,4,8}`$. Giving an expectation value to either the neutral scalar fields or the scalars in the gauge multiplet makes the hypermultiplets massive. This implies a superpotential of the form
$`W`$ $`=`$ $`(M_L+\varphi _3+i\varphi _4)q_L\stackrel{~}{q}_L+\varphi _8(q_L^{}q_L\stackrel{~}{q}_L\stackrel{~}{q}_L^{})`$ (2)
$`+(M_R+\varphi _3+i\varphi _8)q_R\stackrel{~}{q}_R+\varphi _4(q_R^{}q_R\stackrel{~}{q}_R\stackrel{~}{q}_R^{}),`$
where the subscripts on the $`\varphi _i`$ denote the directions they correspond to in the brane construction. $`\varphi _3`$ can be absorbed into a redefinition of $`M_{L,R}`$, so it is again a free decoupled field. The Coulomb branch of this theory can be parametrized by $`M_{L,R}`$, $`\varphi _{4,8}`$ and the dual photon, which provides seven real parameters in agreement with the count in the original theory. To ensure parity invariance of this theory, we take the fundamentals to be parity even and the other fields to be parity odd.
The superpotential can be recast in a manifestly $`SU(2)`$ invariant way. The hypermultiplets combine into $`SU(2)`$ doublets $`𝒫_{L,R}=(q_{L,R},\stackrel{~}{q}_{L,R}^{})`$, and the components of $`M_{L,R}`$ and $`\varphi _{4,8}`$ can be rearranged into two triplets $`\mathrm{\Phi }_{L,R}^i`$. In terms of these fields the superpotential reads
$$W=\mathrm{\Phi }_L^i𝒫_L^{}\sigma _i𝒫_L+\mathrm{\Phi }_R^i𝒫_R^{}\sigma _i𝒫_R.$$
(3)
The Coulomb branch can be parametrized by the expectation values of $`\mathrm{\Phi }_{L,R}`$. Since the fundamentals are massive on this branch, any dynamically generated superpotential can be expressed in terms of $`\mathrm{\Phi }_{L,R}`$. However, there are no parity odd and $`SU(2)`$ invariant combinations of these fields. Thus we conclude that the Coulomb branch cannot be lifted by quantum corrections and corresponds to the Higgs branch of the original theory.
Moving the NS branes in $`X_{5,7,9}`$ corresponds to adding terms of the form
$$W_\xi =\mathrm{\Phi }_L\xi _L+\mathrm{\Phi }_R\xi _R$$
(4)
to the superpotential. The $`\xi _{L,R}`$ are similar to FI parameters in the $`𝒩=4`$ theory, because they force the matter fields to develope an expectation value. However, since the $`𝒩=1`$ vector multiplet does not contain any scalars, they are not FI parameters in the original sense. Nonetheless we will abuse notation and refer to the fields $`\xi _{L,R}`$ as FI parameters.
The FI parameters correspond to the three real mass terms in the original theory. As a consistency check on our superpotential we turn on $`\xi _L`$ while keeping $`\xi _R=0`$. This breaks the $`U(1)`$ gauge symmetry and the resulting theory is a WZ model with a real triplet $`\mathrm{\Phi }_R`$, a complex doublet $`𝒫_R`$, and a superpotential
$$W=\mathrm{\Phi }_R^i𝒫_R^{}\sigma _i𝒫_R.$$
(5)
This agrees with the superpotential we read off from the S-dual of the $`N_f=1`$ brane configuration,
$$W=M_Lq\stackrel{~}{q}+\mathrm{Im}[M_R](qq^{}\stackrel{~}{q}^{}\stackrel{~}{q}),$$
(6)
up to field redefinitions.
We can readily construct non-Abelian theories using our brane configuration. If we put $`N_c`$ 3-branes between the NS-branes, we get a $`SU(N_c)`$ gauge theory in three dimensions. However, the analysis of the field theory in this case is much more involved. The scalar in the $`𝒩=2`$ vector multiplet no longer decouples, since it is now charged under the gauge group. Since it does not couple to the fundamentals, they do not become massive if we give an expectation value to this field. This implies that we can move onto the Higgs branch (parametrized by the expectation value of the fundamentals) at any point on the Coulomb branch (parametrized by the expectation value of the adjoint scalar). Thus classically we find an interacting conformal field theory at every point of the moduli space, which is very difficult to analyze. We will not discuss the non-Abelian case any further in this paper.
## III Discussion
In this note we discussed a type IIB brane configuration that gives rise to $`𝒩=1`$ QED with $`N_f`$ flavors in three dimensions. We argued that the Higgs branch of this theory cannot be lifted by quantum corrections, using the parity invariance of the theory. Since there are no non-renormalization theorems for theories with two supercharges, the metric on the Higgs branch receives quantum corrections but our argument indicates that this does not change its dimension.
For $`𝒩=2,4`$ theories, S-dual pairs of brane configurations give rise to three dimensional theories that are mirror pairs. We suggest that this is true also for the $`𝒩=1`$ theory we consider here. To support this claim we showed that the S-dual brane configuration gives rise to another $`𝒩=1`$ theory that has a Coulomb branch of the same dimension as the Higgs branch of original theory. Parity invariance ensures that neither of these branches can be lifted by quantum corrections. The mass parameters of the original theory are mapped into FI parameters of the mirror. To the best of our knowledge, this is the first proposed duality between theories with two supercharges.
Unfortunately the low degree of supersymmetry makes it hard to check our proposal any further. The metric on the moduli spaces on both sides have quantum corrections and to establish their equivalence, these would have to be computed. Since these branches are real manifolds, there are no complex structures we can compare. On the other hand, some support for our claim comes from the fact that the version of three dimensional QED we discuss here is very similar to the theory one obtains by giving a mass to all scalars in the vector multiplet of $`𝒩=4`$ QED. The only difference between this theory and the theory we discussed here is the presence of the decoupled scalar in our case. Since it does not participate in the dynamics we expect these two theories to be closely related. In fact, adding a term $`W=m^i\varphi _i`$, where $`m^i`$ are uncharged scalars and $`\varphi _i`$ are the three scalars in the $`𝒩=4`$ vector multiplet to the superpotential of $`𝒩=4`$ QED and perturbing the mirror by the corresponding FI terms, generates the mirror pairs we discussed above . While this lends some support to our claims, it is not a very strong argument. Theories with two supercharges can undergo phase transitions as we vary the mass of the three scalars in the $`𝒩=4`$ vector multiplet, so it is not clear that perturbing a pair of mirror theories with $`𝒩=4`$ supersymmetry causes them to flow to mirror pairs of $`𝒩=1`$ theories. The analysis in this paper suggests that this does happen in the specific case we study here.
###### Acknowledgements.
It is a pleasure to thank Anton Kapustin for several very helpful discussions. This work was supported in part by DOE grants #DF-FC02-94ER40818 and #DE-FC-02-91ER40671.
|
no-problem/9906/cond-mat9906281.html
|
ar5iv
|
text
|
# Phonon and Elastic Instabilities in MoC and MoN
## I Introduction
The transition metal carbides and nitrides, such as NbC and MoN, represent a technologically important series of materials, often revealing interplay between their interesting properties and the incipient instabilities that seem to drive those properties. The important features of these materials include extreme hardness and high melting temperatures, as well as superconductivity in many of them. In some of these materials, the atomistic properties (e.g. bonding properties) that drive particular macroscopic behaviors can also lead to instabilities that inhibit the stoichiometric B1 (sodium chloride) structure from forming. MoC and MoN are good examples of this circumstance. In this paper, we report new theoretical results for the MoC and MoN systems indicating that their lack of stability as stoichiometric compounds in the B1 structure is directly correlated with phonon instabilities. We also present electronic structure results for the B1-structure carbides and nitrides with zirconium, niobium and molybdenum. We discuss the trends related to variables such as the number of valence electrons per formula unit, as well as the effect of the particular non-metal atom (nitrogen or carbon). We discuss these results on the basis of the known properties of these materials.
Using results from first principles electronic structure calculations, Pickett et al. argued that MoN in the B1 structure was a prime candidate for a “high-temperature” superconductor with a predicted transition temperature of approximately $`30^{}`$ K. There were two aspects of the electronic structure that motivated this prediction for MoN—the high density of electronic states at the Fermi level, N(E<sub>F</sub>), and the large electron-phonon matrix elements associated with the strong bonding in this material. Both of these features are indicative of a potentially high T<sub>c</sub>, but they are also hallmarks of either structural instabilities that can frustrate the formation of stoichiometric structures, or of magnetic instabilities that can destroy superconductivity. Experimentally, it was found that these features lead instead to structural instabilities in B1 MoN that are manifested by the fact that only highly N deficient or vacancy/defect-rich B1 structures have been made in the laboratory. Alternatively, the more thermodynamically stable hexagonal phase of MoN is often formed close to stoichiometry. Similar systems, such as NbC and NbN, which exhibit superconductivity are also prone to poor stoichiometry and must be prepared carefully to obtain a ratio of niobium to carbon or nitrogen that is close to one-to-one. The substoichiometry in these systems normally occurs as a deficiency of carbon or nitrogen atoms.
Following the initial prediction of a potentially high T<sub>c</sub> in MoN, there was a flurry of unsuccessful experimental activity aimed at fabricating this compound close to stoichiometry. Subsequent theoretical work by Chen and coworkers shed light on the cause of the experimental difficulties, showing that perfectly stoichiometric MoN was elastically unstable in the B1 structure. They calculated the total energies for small strains away from the B1 structure and showed that the cubic elastic constant C<sub>44</sub> exhibited an instability (negative value) for MoN. In contrast, their calculations of the three cubic elastic constants for NbC showed that there were no elastic instabilities in that material.
Earlier studies related the experimentally observed phonon anomalies (dips in the phonon dispersion curves) in NbC and TaC to their Fermi surfaces. The so-called Fermi surface nesting effects refer to the fact that parallel sheets of the Fermi surface with (relatively) large areas provide a mechanism for enhancing the phonon renormalization (decrease in square frequency) resulting from the screening effects of the electron-phonon interactions. Enhanced screening effects can lead to anomalous structure in the phonon dispersion curves or, in the extreme case, to instabilities manifested by negative square frequencies at wave vectors related to the nesting vectors of the Fermi surface. In the extreme case of Fermi-surface-induced instabilities, a given phase is unstable and will not form at stoichiometry. Presumably, aspects of the instabilities in MoC and MoN are Fermi surface related, but as we mention below the quantitative manifestation of Fermi surface nesting in these materials is complicated and, unlike NbC, not amenable to a simple interpretation (e.g. matrix element effects are apparently key to triggering the instability).
Various experimental investigators attempted to form the metastable B1 phase of MoN, but most samples still suffered from large amounts of vacancies or defects, or a high pressure hexagonal phase formed. Attempts at forming a cubic metastable bulk phase have been unsuccessful to date. There was hope, however, that MoN might be metastable at high pressure in the B1 structure, with the elastic instabilities mitigated by the increased bonding that would result. However, our calculations demonstrate that, up to pressures well above 400 GPa, the C<sub>44</sub> instability for MoN actually increases, indicating that a metastable structure of B1-MoN is not possible even at the highest pressures attainable with novel experimental methods.
In contrast to the situation for B1-MoN, we show below that MoC does not have an elastic instability in C<sub>44</sub> (or any of the cubic elastic moduli, for that matter) despite the fact that experimentally MoC is nearly as unstable as MoN with regard to the limits on obtainable stoichiometry or defect/vacancy-free crystals. We show instead that the instabilities in MoC are driven by unstable phonons near the X point in the Brillouin zone (BZ) and not by elastic instabilities (long wavelength acoustic phonon instabilities) as in MoN. Moreover, we show that MoN also has phonon instabilities in addition to its elastic instabilities that have already been demonstrated. We discuss these instabilities in MoC and MoN in terms of their electronic structure, including their densities of states and Fermi surfaces.
## II computational details
Our calculations were performed using the full-potential linear-augmented-plane-wave (LAPW) method within the local density approximation (LDA). The core states were treated fully relativistically and the valence states were treated semi-relativistically (without spin-orbit interaction). No shape approximations were made for the potential or the charge density. The exchange-correlation potential used was that of Ceperley and Alder as parameterized by Perdew and Zunger.
For the lattice constant and bulk modulus calculations, 47 k-points (after the method of Blöchl) were used in the irreducible wedge of the Brillouin Zone (BZ) (equivalent to 1000 points in the entire BZ). Approximately 270 LAPW basis functions were used per atom.
A set of monoclinic strains was used to calculate the cubic elastic constant C<sub>44</sub> using the method outlined by Mehl et al. Because the energy differences in these types of calculations are very small, great care was taken in selecting k-point meshes to ensure convergence. For the results shown here, 1088 k-points in the irreducible wedge were used.
Phonon frequencies were calculated using a supercell approach (frozen phonons). Calculating the energy as a function of several different well-chosen distortions allows the phonon frequencies at high symmetry points to be determined. Calculations for X point phonons were done by calculating the energy differences for three different sets of displacements which were determined using the program ISOTROPY. The energy versus distortion results were fitted to polynomial expansions and the phonon frequencies were then determined. When the square of the frequency of an unstable phonon is negative, this indicates that the B1 structure is unstable with respect to that particular set of atomic displacements.
## III Results and discussion
### A Bonding in the carbo-nitrides
As discussed below, the B1 carbides and nitrides of zirconium, niobium, and molybdenum have very similar density of states (DOS) profiles, the main difference being a systematic increase in the DOS at the Fermi level. The same systematic trends can be seen in the lattice constants and bulk moduli. Fig. 1 shows the bulk moduli of these materials as a function of the lattice constant. There is a general trend of decreasing lattice constant as the number of valence electrons increases, with the nitrides having the smaller lattice constants in the case of the isoelectronic systems (ZrN & NbC, NbN & MoC). Despite the smaller lattice constants of the nitride systems, the corresponding increases in the bulk moduli are smaller than expected compared to the isoelectronic carbide system (compare ZrN and NbC or NbN and MoC in Fig. 1). In a strict rigid-band view, the bulk moduli would be purely a function of volume, leading to one smooth curve in Fig. 1 rather than separate curves for the carbides and the nitrides. This circumstance demonstrates the subtle differences in the bonding characteristics between the carbide and nitride systems despite very similar densities of states. In the carbide systems, the covalent bonding charge between the carbon atoms is more localized and closer to the atoms than the corresponding bonding charge in the nitrides. Thus, the more itinerant, less tightly bound charge in the nitrides is more easily deformed than the covalent bonding charge in the carbides, leading to a smaller than expected increase in the bulk modulus of the nitrides relative to the carbides.
The above considerations illustrate that, although the major contributions to the bonding and resultant properties of the carbo-nitride systems can be accounted for by general considerations of atom size and overall valence of the formula unit, the chemical composition is also important for understanding differences in the physical properties of these materials. For example, although the B1 compounds NbN and MoC have the same number of valence electrons per formula unit and very similar DOS profiles, the former can be made close to stoichiometry while the latter cannot.
### B Densities of states
The DOS for isoelectronic pairs of these compounds are qualitatively very similar, the quantitative difference being the total DOS at the Fermi level. (A representative DOS plot is shown in Fig. 2.) By definition, in a purely rigid band picture, the isoelectronic pairs (ZrN & NbC, NbN & MoC) would be identical. While the DOS profiles look very similar and the Fermi level occurs at similar places on the DOS profile, the DOS at the Fermi level is higher in the nitride systems compared to the isoelectronic carbide systems. The DOS at the Fermi level increases systematically with the number of valence electrons (ZrC, which has only 8 valence electrons per unit cell, has the lowest DOS at the Fermi level and the highest is for MoN which has 11 valence electrons per unit cell.)
The increase in the DOS at the Fermi level as the valence is increased is a rigid-band-like effect resulting from the positive slope of the DOS as a function of energy in this energy range, while the increase between isoelectronic pairs is a chemical effect due to the bonding differences between a nitride and its isoelectronic carbide. This increase in N(E<sub>F</sub>) is reflected in the superconducting transition temperatures in the series up to NbN. (NbN has the highest T<sub>c</sub> \[17 K\] of the B1 superconductors.) Using simple BCS arguments, it is expected that T<sub>c</sub> will increase with electron count due to the larger values of N(E<sub>F</sub>), resulting in T<sub>c</sub> being higher for the nitrides. However, in B1-MoC and B1-MoN samples, the stoichiometry is poor and/or there are a large number of defects. Consequently, the superconducting transition temperatures in MoC and MoN are much lower than predicted from the perfect crystal DOS arguments using quantitative rigid-muffin-tin calculations.
### C Elastic instability of MoN
The systematic increase in the DOS at the Fermi level led to the proposal of MoN and MoC as good candidates for “high temperature” superconductors. However, efforts to make high quality, stoichiometric B1-MoN crystals were unsuccessful. In most samples, the defects are primarily nitrogen substitutional defects on the molybdenum sites, nitrogen atoms in the interstitial regions, and vacancies on the nitrogen sites. Some workers tried to improve the crystals by applying pressure in an attempt to drive nitrogen defects from interstitial sites into the nitrogen sites of the ideal crystal. Unfortunately, this results in a hexagonal structure less favorable for high T<sub>c</sub>. In an effort to explain the experimental “instability” of B1-MoN, Chen et al. performed theoretical studies of the cubic elastic constants for MoN. They found that the cubic elastic constant C<sub>44</sub> for MoN was negative. This result clearly shows that B1-MoN is unstable at zero pressure, and is in good agreement with the experiments in this regard.
It is perhaps not unreasonable that a metastable state might exist, however, given the successes in forming metastable phases of other carbo-nitride compounds, and others have suggested that the C<sub>44</sub> instability may be mitigated in MoN by applying high pressures. To test this, we calculated the change in energy as a function of a monoclinic strain at various pressures. The results are shown in Fig. 3. As is evident from the figure, the instability persists and even increases at pressures high enough (above 400 GPa) to contract the lattice constant by 3%. This indicates that any attempts at fabricating a metastable phase, even at the highest available laboratory pressures using novel techniques, will be unsuccessful.
### D Phonon instabilities of MoC and MoN
Because the stability of MoC samples is not much better than that of MoN, and the predicted superconductivity transition temperature is smaller than expected, it is natural to suspect that C<sub>44</sub> for MoC also reveals the same instability observed in MoN. However, our calculations show that for MoC the C<sub>44</sub> elastic constant is large and positive. The other two elastic constants, C<sub>11</sub> and C<sub>12</sub>, are also relatively large and positive, indicating an elastically stable structure.
Given that the elastic constants for MoC are stable, the question arises: why can stoichiometric B1-structure MoC not be made? In particular, is B1-MoC a possible metastable phase, or is it intrinsically unstable for some reason other than elastic behavior? As a first step toward answering this question, we examined the phonon frequencies of MoC at selected points in the Brillouin Zone. Using the frozen phonon approach, we first determined the phonon frequencies of MoC and MoN at the $`\mathrm{\Gamma }`$ point, and NbC at the X point to compare with experiment and previous calculations. The results within a few percent of the experimental results as well as Savrasov’s calculations for NbC. The frequency of the calculated zone center optical phonon in MoC shows no indication of a phonon anomaly and is very close to that of NbC (as one might expect). On the other hand, the frequency of the calculated zone center optical phonon in MoN is less than half that of MoC and NbC. (Of course, zone center acoustic modes in MoN are unstable as shown by the C<sub>44</sub> calculation.)
Even when a high density of states at the Fermi level does not indicate a structural or magnetic instability, it may indicate anomalous phonons. Though NbC is quite stable and samples can be fabricated with good stoichiometry, the phonon spectrum contains a very distinct anomalous region near $`k=.7(2\pi /a)`$ along $`\mathrm{\Gamma }\mathrm{X}`$ that is related to Fermi surface nesting as shown by Gupta and Freeman and Klein et al. In the rigid band picture, NbN is similar to NbC but with an extra valence electron per unit cell. Adding an extra electron to the system (NbC $``$ NbN) causes more hybridized niobium-d/non-metal p-states to be occupied which leads to an increase in the density of states at the Fermi level. The anomalous region in the phonon spectrum becomes more pronounced and shifts toward the X point. (See Fig. 4.) If the phonon spectra of these carbo-nitride systems also follow systematic trends as for the lattice constants, bulk moduli, and DOS at the Fermi level, then it is reasonable to expect that the anomalous region of the phonon spectrum may become even more pronounced in MoC—perhaps even to the point that some phonon modes become unstable causing the crystal to spontaneously distort at finite temperatures. A schematic of this idea is shown in Fig. 4. Given these ideas, a good candidate BZ region for phonon instabilities in MoC would be that near the X point.
To test this hypothesis, we calculated the frequencies of the optical and acoustic longitudinal phonons at the X point for NbC and MoC. The calculated frequencies for NbC agree within a few percent of the experimental values of Smith and Gläser and frequencies calculated using other methods. Our NbC calculation is simply a “proof of principle” check for the frozen phonon calculations. For MoC, our calculations show a frequency for optical longitudinal phonons that is nearly half that of NbC, and the calculated frequency for acoustic longitudinal phonons is imaginary (367 cm<sup>-1</sup>) indicating that these latter phonons are unstable at the X point. It can be reasonably surmised, then, that there exists a finite region of the spectrum near the X point for which the acoustic phonons are unstable. This is consistent with the experimental instability that exists despite the fact that LDA-based calculations of the elastic constants indicate that B1-MoC is elastically very stable. While the Fermi surfaces of MoC and MoN which we calculated (not shown here) show indications of nesting effects with a wave vector at or near the X-point, the nesting features are not as pronounced as those found in NbC and TaC. The phonon anomalies cannot be explained as simple nesting effects; They are a complicated combination of nesting (phase space considerations) and bonding (the strength of the electron-phonon matrix elements).
## IV Conclusions
We discussed systematic trends in some of the physical properties in the series of compounds TMC and TMN, TM = Zr, Nb, Mo. Lattice constants, bulk moduli, DOS at the Fermi level, etc. change systematically through the series. Corresponding to these systematic changes, the superconducting transition temperatures of the materials increase with increasing numbers of electrons per unit cell, but the compounds become increasingly unstable through the series.
While Chen et al. showed that perfectly stoichiometric B1-MoN was unstable, we have also shown that these instabilities are not mitigated by increased pressure—that is, the C<sub>44</sub> instability actually is enhanced by the application of hydrostatic pressure.
Unlike MoN, there is no indication of a similar elastic instability in B1-structure MoC despite the fact that fabrication of high quality MoC crystals is also very difficult. We have shown that this experimental instability is related to an extreme phonon softening near the X point. Consequently, as with MoN, a stable state of B1-MoC cannot be expected.
## ACKNOWLEDGMENTS
The authors would like to thank the Campus Laboratory Collaboration Program of the University of California for financial support. Generous computer resources were also provided by Lawrence Livermore National Laboratories. One of us, G. L. W. H., would like to thank H. T. Stokes of Brigham Young University for providing his program ISOTROPY, part of which was used in determining the symmetries of distortions for the frozen phonon calculations reported here.
|
no-problem/9906/astro-ph9906466.html
|
ar5iv
|
text
|
# The Lick Planet Search : Detectability and Mass Thresholds
## 1 Introduction
In the past few years, high precision radial velocity surveys have had remarkable success in the discovery of planetary-mass companions around nearby solar-type stars (for reviews, see Marcy & Butler 1998 and Marcy, Cochran & Mayor 1999). Searches for companions (Campbell et al. 1988; McMillan et al. 1994; Mayor & Queloz 1995; Walker et al. 1995; Cochran et al. 1997; Noyes et al. 1997; Marcy & Butler 1992, 1998) have been carried out with Doppler velocity precision $`10\mathrm{m}\mathrm{s}^1`$, although $`3\mathrm{m}\mathrm{s}^1`$ has been achieved at Lick observatory for chromospherically quiet stars (Butler et al. 1996). There are now 17 companions known with masses (the observable is $`M_p\mathrm{sin}i`$, where $`i`$ is the angle of inclination of the orbit with respect to the line of sight) below 10 Jupiter masses. In total, several hundred stars have been monitored for timescales of 3 years to more than 11 years. The detections so far suggest that a few percent of solar type stars harbor companions of a Jupiter mass or more within a few AU.
These objects have raised many questions regarding the distribution of the mass and orbital radius of planetary-mass companions, and the relation of these systems to our own solar system and its giant planets. For example, a surprise was the discovery of Jupiter-mass companions in close proximity to their host star. Of the 17 companions within $`2.5\mathrm{AU}`$, 13 have semi-major axis $`a<0.5\mathrm{AU}`$ and five have $`a<0.1\mathrm{AU}`$. The archetypal example is the companion orbiting 51 Pegasi (Mayor & Queloz 1995) which has a mass ($`M_p\mathrm{sin}i`$) of 0.44 Jupiter masses ($`M_J`$) and an orbital radius $`a=0.05\mathrm{AU}`$, eight times closer than Mercury’s orbit about the Sun. The orbital parameters of the 17 planetary-mass companions are listed in Table 1.
Unfortunately, gleaning the true distribution of companions is complicated by selection effects which favor the detection of massive, close companions. It is necessary to establish detection thresholds for searches for planetary-mass companions before the observations can be fully interpreted. Walker et al. (1995) monitored 21 bright solar type stars for 12 years. They carried out a detailed statistical analysis, and from their upper limits could exclude companions with $`M_p\mathrm{sin}i1`$$`3M_J`$ for periods less than the duration of their observations ($`12`$ years). Nelson & Angel (1998) used a simple analytic formalism, together with comparisons with real data, to investigate the dependence of detection thresholds on the number and duration of observations and the Doppler errors.
Our aim in this paper is to place the confirmed companions from the Lick radial velocity survey in context by an analysis of the null detections. The ongoing Lick survey consists of more than eleven years of precision Doppler velocity measurements of 107 stars (a few of which have been added or dropped as the survey progressed), and 200 new stars have recently been added (Fischer et al. 1999). The original program has so far identified six planetary-mass companions (70 Vir, 47 UMa, 55$`\rho `$ Cnc, $`\tau `$ Boo, $`v`$ And, GL876, they are marked in Table 1), codiscovered the companion to 16 Cyg B, and confirmed two discoveries made by other groups (51 Peg and $`\rho `$ CrB, see Table 1 for orbital parameters and references).
We search for periodicities and place upper limits using the “floating-mean periodogram”, an extension of the well-known Lomb-Scargle periodogram (Lomb 1976; Scargle 1982) in which we fit sinusoids to the data, but allow the zero-point of the sinusoid to “float” during the fit. This approach was adopted by Walker et al. (1995), who were interested in obtaining correct upper limits for periods greater than the duration of the observations. We show here that allowing the mean to float is crucial to account for statistical fluctuations in the mean of a sampled sinusoid. The traditional Lomb-Scargle periodogram fails in precisely the regime where we demand it be robust, namely when the number of observations is small, the sampling is uneven or the period of the sinusoid is comparable to or greater than the duration of the observations. We carefully consider the correct normalization of the periodogram, an issue which has been of some debate in the literature (Horne & Baliunas 1986; Gilliland & Baliunas 1987; Schwarzenberg-Czerny 1996).
The plan of the paper is as follows. We begin in §2 by describing the observations and estimating the velocity variability we expect to see due to measurement error and intrinsic stellar effects. We discuss in §3 our methods for searching for the signatures of companions in the radial velocity data. We look for variability in excess of our prediction, long term trends and periodicities. We list those stars whose data show interesting variability or periodicities which may indicate the presence of a yet unconfirmed companion. In §4, we place upper limits on the mass of a possible companion as a function of orbital radius for each star in the sample. We continue in §5 with a discussion of the implications of our results for the occurrence rate of planetary-mass companions to solar type stars, and their distribution in mass and orbital radius. We present our conclusions in §6. In Appendix A, we give a brief derivation of the periodogram, and in Appendix B, we discuss its normalization.
## 2 The Observations
The Lick radial velocity survey is now more than eleven years old (Marcy & Butler 1992, 1998). Precise radial velocity measurements (current precision $`5\mathrm{m}\mathrm{s}^1`$, Butler et al. 1996) are made with the Lick 3m telescope by using an echelle spectrograph and a comparison iodine reference spectrum. The exposure time is ten minutes for a star with $`V=5`$, allowing several observations per year for each star in the sample. In this paper, we present an analysis of observations of 76 F, G, and K type stars in the original survey<sup>1</sup><sup>1</sup>1There have been observations of 29 M dwarfs as part of the survey, but we do not include these in our analysis as they are faint ($`V>7`$) and suffer from large measurement uncertainties. These stars are part of a sample being monitored with the Keck telescope. An analysis of their radial velocity variability will be presented elsewhere.. A summary of the observations is given in Table 2. For each star, we give its HR and HD catalog number, spectral type and rotation period. We list the number and duration of the observations, the typical internal velocity error and the rms scatter of the data. Radial velocities are available upon request from G. Marcy.
### 2.1 The Distribution of Errors and Intrinsic Stellar Variability
Two sources of variability can mask velocity variations due to a companion: measurement uncertainties and intrinsic stellar variability. In this section, we attempt to quantify these effects.
The uncertainty in the radial velocity measurement $`v`$ is estimated for each observation from the dispersion of the velocities measured by different spectral segments of the spectrometer. An upgrade to the spectrograph optics and improvements in modeling in November 1994 led to an increase in the Doppler precision from $`\sigma _D10`$$`15\mathrm{m}\mathrm{s}^1`$ to $`\sigma _D5\mathrm{m}\mathrm{s}^1`$, with $`\sigma _D3\mathrm{m}\mathrm{s}^1`$ achievable in the best cases. This improved Doppler precision is dominated by photon statistics (for a detailed discussion, see Butler et al. 1996). In this paper, we shall refer to the pre-November 1994 data as “pre-fix” and post-November 1994 data as “post-fix”. Table 2 gives the average internal error $`<\sigma _D>`$ before and after November 1994 for each star.
Intrinsic stellar variability arises from magnetic activity or rotation of features across the stellar surface, such as sunspots or inhomogeneous convective patterns (Saar & Donahue 1997). Saar, Butler & Marcy (1998, hereafter SBM98) used the Lick radial velocity variations (post-fix data only) to characterize the relationship between the rotation period, $`P_{\mathrm{rot}}`$, of a star and its intrinsic velocity variability, $`\sigma _V`$. They found that the variability in excess of the internal errors could be explained by simple models of sunspot rotation and inhomogeneous convective flows. We use their results to estimate the typical intrinsic variability associated with such effects as a function of stellar rotation period<sup>2</sup><sup>2</sup>2 At first sight, it may seem circular to use the work of SBM98 which was based on the Lick data set (the post-fix data only). However, our approach is to use their results to characterize the average variability typical for a star in the survey with a particular rotation period.. After inspecting Figure 2 of SBM98, we find $`\sigma _V=10\mathrm{m}\mathrm{s}^1(12\mathrm{days}/P_{\mathrm{rot}})^{1.1}`$ for G and K type stars and $`\sigma _V=10\mathrm{m}\mathrm{s}^1(10\mathrm{days}/P_{\mathrm{rot}})^{1.3}`$ for F stars. The rotation period for each star (taken from Soderblom 1985, Baliunas, Sokoloff & Soon 1996, or Fischer 1999, private communication) is given in Table 2.
We obtain the total estimated variability for each data point by adding the intrinsic variability to the internal error in quadrature, $`\sigma _j^2=\sigma _V^2+\sigma _{D,j}^2`$ (here we label each data point with the index $`j`$). How well does this estimate reproduce the scatter in the data? After a preliminary analysis of the post-fix data (using the methods of §§3.1 and 3.2), we selected a subset of 26 stars that had no excess variability or long term trends. In Figure 1, we plot a histogram (solid line) of individual radial velocity measurements $`v_j`$ divided by the estimated variability $`\sigma _j`$ for this subset of stars. For each star, we have subtracted the weighted mean of the velocities. The upper panel shows the pre-fix data; the lower panel shows the post-fix data. The dotted histogram in each case shows a Gaussian distribution with unit variance. If the scatter in the velocities were Gaussian with variance $`\sigma _j^2`$ for each point, the dotted and solid histograms would match. They do not, indicating more scatter in the data than we expect. In addition, the pre-fix and post-fix $`v/\sigma `$ distributions are different.
The excess scatter may be due to a combination of underestimated internal errors and systematic errors in the velocities, particularly for the early observations. We have chosen to augment the internal errors by multiplying by a constant factor to force the observed $`v/\sigma `$ distribution to match a Gaussian with unit variance. In this way, we bring the pre-fix and post-fix $`v/\sigma `$ distributions into agreement, and we are confident that we have not underestimated the errors. For all 76 stars in our sample, we multiply the pre-fix internal errors by 1.7 and the post-fix internal errors by 1.4. The dashed histograms in Figure 1 show the $`v/\sigma `$ distributions using these rescaled internal errors. The distribution is unchanged if we remove the 10 chromospherically-active stars in this subset which have $`P_{\mathrm{rot}}12\mathrm{days}`$. Hereafter, we refer to the augmented internal errors as simply “internal errors”.
In Figure 2, we show the effect of including the intrinsic variability prediction $`\sigma _V`$ for stars of different rotation period. For each star in the subset of 26 stars of Figure 1, we plot $`\chi _\nu ^2=(v_j/\sigma _j)^2/(N1)`$ as a function of $`P_{\mathrm{rot}}`$. We show $`\chi _\nu ^2`$ evaluated using the internal errors only (crosses) and including the intrinsic variability added in quadrature (squares). The extra variability of stars with short rotation periods ($`P_{\mathrm{rot}}12`$ days) is shown by the large uncorrected $`\chi ^2`$ values for these stars. The mean value of $`\chi _\nu ^2`$ in Figure 2 is less than unity, indicating that our procedure may have overestimated the internal errors somewhat. However, we prefer to err on the side of overestimation.
We use the variability estimate $`\sigma _j`$ in two ways. The first is to identify those stars which show much more variability than we might expect given their rotation periods (§3.2). The second is to weight the data points when we look for periodicities (§3.4). The large difference between the pre-fix and post-fix data makes it important to give less weight to the early data points. Not only are the pre-fix errors larger, they are less well characterized than the post-fix errors. One might question the value of including the low quality early data points at all. However, they are important because they extend the time baseline, allowing us to search for longer period companions.
## 3 Search for Companions
The velocity amplitude $`K`$ of a star of mass $`M_{}`$ due to a companion with mass $`M_p\mathrm{sin}i`$ with orbital period $`P`$ and eccentricity $`e`$ is
$$K=\left(\frac{2\pi G}{P}\right)^{1/3}\frac{M_p\mathrm{sin}i}{(M_p+M_{})^{2/3}}\frac{1}{\sqrt{1e^2}}.$$
(1)
For a circular orbit with $`M_pM_{}`$, the velocity variations are sinusoidal with amplitude
$$K=28.4\mathrm{m}\mathrm{s}^1\left(\frac{1\mathrm{year}}{P}\right)^{1/3}\left(\frac{M_p\mathrm{sin}i}{M_J}\right)\left(\frac{M_{}}{M_{}}\right)^{2/3},$$
(2)
where $`M_J`$ is the mass of Jupiter. The orbital period is related to the orbital radius by Kepler’s law,
$$P=1\mathrm{year}(a/\mathrm{AU})^{3/2}(M_{}/M_{})^{1/2}.$$
(3)
For example, the companion to 51 Peg ($`a=0.05`$ AU, $`M_p\mathrm{sin}i=0.44`$) induces a velocity amplitude $`K=56\mathrm{m}\mathrm{s}^1`$, whereas Jupiter ($`a=`$5.2 AU, $`P=11.9`$ years) gives $`K=12.5\mathrm{m}\mathrm{s}^1`$ for the Sun.
In this section, we describe the methods we use to search for such a signal, and present our results. For each star in Table 2, we first ask if there is a significant long term trend in the data, and if so we subtract it (§3.1). We then ask if the observed scatter in the data is consistent with the expected variability (§3.2). To search for periodicities (§3.4), we fit sinusoids to the data, employing a generalization of the well-known Lomb-Scargle periodogram (§3.3). By using sinusoids as our basic model for the data, we are strictly assuming circular orbits, although we find that the periodogram gives a good estimate of the orbital period even for eccentric orbits. This is because, to lowest order in the eccentricity, the radial velocity signal from an eccentric orbit has its main component at the orbital frequency (with smaller components at multiples of the orbital frequency). We discuss a possible extension of the periodogram to Kepler orbits in the conclusions (§6). The normalization of the periodogram has been of some question in the literature, so we discuss our choice of normalization in Appendix B. We close this section by summarizing our results and indicating those stars that show interesting variability or evidence for companions (§3.5).
### 3.1 Long Term Trends
We first ask if there is a significant long term trend, on a timescale much greater than the duration of the observations (i.e. $``$ 10 years). For each star, we fit a straight line $`v_j=at_j+b`$ to the measured velocities. When calculating $`\chi ^2`$ for the fit, we weight each point by the inverse square of the estimated error, $`w_j=1/\sigma _j^2`$. We give the best-fit slope and its uncertainty (as derived from the least-squares fit) in Table 3.
To assess the significance of each slope, we ask if the coefficient of the linear term is significantly non-zero. We use the F-test to compare the weighted sum of squares of residuals from the straight line fit $`\chi _{N2}^2`$ (two free parameters) to the weighted sum of squares about the mean $`\chi _{N1}^2`$ (one free parameter). If there is no long term trend and the residuals are Gaussian distributed, the statistic
$$F=(N2)\frac{\chi _{N1}^2\chi _{N2}^2}{\chi _{N2}^2}$$
(4)
follows Fisher’s $`F`$ distribution (Bevington & Robinson 1992) with $`1`$ and $`N2`$ degrees of freedom. Given the distribution $`F_{1,N2}`$, we calculate the probability that $`F`$ would exceed the observed value $`F_{\mathrm{obs}}`$ purely by chance (the false alarm probability). We give the F-test false alarm probabilities for each star in Table 3.
The slopes listed in Table 3 contain much information about possible companions at long periods. For the purposes of this paper, however, we are interested in identifying slopes which would directly affect our search for companions with $`P30`$ years. Thus, we mark with $``$ in Table 3 those stars which have a slope $`5\mathrm{m}\mathrm{s}^1\mathrm{yr}^1`$ and an F-test false alarm probability $`<10^5`$. We adopt a higher threshold ($`10^5`$) than elsewhere in this paper because systematic effects in the pre-fix data or variations due to magnetic activity on long timescales can imitate a slope. The stars marked with $``$ in Table 3 are HR 219a, HR 753, HR 1325, HR 2047, HR 5544a, HR 5544b, HR 6623, HR 7672 and HR 8086<sup>3</sup><sup>3</sup>3HR 8085, the companion to HR 8086, would have exhibited a slope had we removed the effect of secular acceleration, see §3.5..
### 3.2 Excess Variability
We now ask if the scatter in the velocities is consistent with our predicted variability for each star (§2.1). The $`\chi ^2`$ about the mean, $`\chi _{N1}^2`$, or (for those stars marked $``$ in Table 3) about the best-fit straight line, $`\chi _{N2}^2`$, gives us a measure of variability. We use the $`\chi ^2`$ distribution with $`Nm`$ degrees of freedom to test if the velocities are consistent with being drawn from a Gaussian distribution with variance $`\sigma _j^2`$. Here $`m`$ is the number of parameters in the model of the data, $`m=2`$ for a straight line or $`m=1`$ for a mean. A small false alarm probability indicates there is more variability in the data than we expect. The results of this test are shown in Table 4. We indicate with a $``$ those stars which have false alarm probabilities less than 1%. We choose a 99% threshold so that there will be no more than one false signal in our sample of 76 stars. The stars which show significant variability are HR 88, HR 166, HR 2047, HR 4345, HR 5273, HR 5544b, HR 5553, HR 7061, GL 641, GL 688, and the stars with confirmed planetary-mass companions (listed in Table 1).
### 3.3 The Floating-Mean Periodogram
In this section, we test for periodicity in the data using what we refer to as the “floating-mean periodogram”, a generalization of the well-known Lomb-Scargle periodogram (Lomb 1976; Scargle 1982). We define the floating-mean periodogram as follows. For each trial frequency $`\omega =2\pi /P`$, we start with a simple model for the data, namely a sinusoid plus a constant term,
$$f_j=A\mathrm{cos}\omega t_j+B\mathrm{sin}\omega t_j+C,$$
(5)
where $`t_j`$ are the observation times. We perform a linear least squares fit of this model to the data, to determine the constants $`A`$,$`B`$ and $`C`$. The periodogram is an “inverted” plot of the $`\chi ^2`$ of this fit as a function of frequency. We define the floating-mean periodogram power $`z(\omega )`$ as
$$z(\omega )=\frac{(N3)}{2}\frac{\chi _{N1}^2\chi ^2(\omega )}{\chi ^2(\omega _0)},$$
(6)
where $`\chi ^2(\omega )=w_j[v_jf_j(\omega )]^2`$ is the $`\chi ^2`$ of the fit, $`\omega _0`$ is the best-fit frequency (i.e. the frequency that gives the maximum periodogram power) and $`\chi _{N1}^2`$ is the weighted sum of squares about the mean. When calculating $`\chi ^2`$, we weight each data point by $`w_j1/\sigma _j^2`$ as in §3.1. We use a linear least squares fitting algorithm from Press et al. (1992) to fit equation (5) to the data and find $`\chi ^2(\omega )`$. The choice of normalization of the periodogram has been a subject of some debate in the literature; we discuss this in detail in Appendix B. We normalize by the weighted sum of squares of the residuals to the best fit sinusoid, $`\chi ^2(\omega _0)`$.
The Lomb-Scargle periodogram is obtained by considering a fit of a sinusoid only, i.e. the case $`C0`$ in equation (5) (we sketch the derivation of the Lomb-Scargle formula from the least-squares approach in Appendix A). This means that the zero-point of the sinusoid is assumed to be known already. In practice, the mean of the data is taken as an estimate of the zero-point and is subtracted from the data before applying the Lomb-Scargle formula. In our approach, the zero-point of the sinusoid is an additional free parameter at each frequency, i.e. the mean of the data is allowed to “float” during the fit.
This approach has been adopted by other authors. Ferraz-Mello (1981) was the first to do so, defining the “date-compensated discrete Fourier transform”. Walker et al. (1995) generalized to the case where a straight line or quadratic function was subtracted from the data, defining a “correlated periodogram”. Most recently, Nelson & Angel (1998) included a constant term in their Monte Carlo experiments. These authors were concerned about the suppression of periodogram power for periods greater than the duration of the observations. We show here that allowing the mean to float is important under much more general circumstances.
We now provide some examples which show that allowing the mean to float is crucial if the number of observations is small, the sampling is uneven or there is a period comparable to the duration of the observations or longer. Figure 3 shows simulated data of a companion in a circular orbit with $`P=9.6`$ years and $`K=15\mathrm{m}\mathrm{s}^1`$ ($`a=4.5\mathrm{AU}`$, $`M_p\mathrm{sin}i=1.1M_J`$). We use the observation times and errors for one of the stars in our sample, HR 222. The upper panel shows the velocity measurements as a function of time. The dashed line shows the mean of the data, which is about $`10\mathrm{m}\mathrm{s}^1`$ greater than the correct zero-point of the sine wave. The solid line shows the best-fit sinusoid when the mean is allowed to float. The traditional and floating-mean periodograms are shown in the lower panel. The power at long periods is significantly less in the traditional periodogram than the floating mean periodogram. Black & Scargle (1982) first noted this effect in their analysis of astrometric data, where they showed that proper motions could significantly reduce the measured amplitude of long period signals (see Figure 3 of Black & Scargle 1982).
The upper panel in Figure 4 shows 20 velocities obtained at Lick for the star HR 5968, which has a known companion with $`K=67\mathrm{m}\mathrm{s}^1`$ and $`P=39.6`$ days (Noyes et al. 1997, Table 1). The duration of these observations is 1.2 years. By chance, most of the measurements lie above the zero-point of the orbit, giving a $`20\mathrm{m}\mathrm{s}^1`$ difference between the mean of the data and the correct zero-point. We plot the best-fit sinusoid with a fixed mean as a dashed line and with a floating mean as a solid line. The traditional periodogram does not detect a significant period; it identifies a period of $`43.9`$ days, but with false alarm probability 8%. The floating-mean periodogram gives a very significant detection at the correct period of $`40.0`$ days (see Table 5). Thus allowing the mean to float is not only important at long periods, but is crucial to account for statistical fluctuations when the number of observations is small. A similar situation could occur due to uneven sampling. If a sinusoid is sampled at nearly the same phase each cycle (e.g. for $`P1`$ year), the mean of the data could be significantly different from the zero-point of the sinusoid, and the periodicity thus go undetected.
If the number of data points is large and the periodicity is well-sampled, the mean of the data does give a good estimate of the correct zero-point. However, the traditional periodogram fails in precisely the regime where we require it to be robust; namely, when the number of observations is small, the duration of the observations is limited, or the sampling uneven. For this reason, we adopt the floating-mean periodogram, despite it being less computationally-efficient than the simple Lomb-Scargle formula.
Following Walker et al. (1995), we make a further generalization in the case where we subtract a straight line from the data (those stars marked $``$ in Table 3). In this case we must fit a sinusoid plus a straight line to the data at each frequency (i.e. add a term $`t_j`$ to eq. ). For these stars, we use $`\chi _{N2}^2`$ in place of $`\chi _{N1}^2`$ in our definition of the periodogram, and replace $`\chi ^2(\omega )`$ by the $`\chi ^2`$ from the straight line plus sinusoid fit. This gives a general formula for the floating-mean periodogram power (as Walker et al. 1995, eq. \[A2\])
$$z(\omega )=\frac{(Nm2)}{2}\frac{\chi _{Nm}^2\chi _{Nm2}^2(\omega )}{\chi _{Nm2}^2(\omega _0)},$$
(7)
where $`m=1`$ for a mean or $`m=2`$ for a straight-line. Comparing equation (7) with equation (4), we see that the periodogram is similar to the F-statistic, measuring how much the fit is improved by introducing the two extra sinusoid parameters.
### 3.4 Application of the Periodogram
#### 3.4.1 Search for Periodicities
For each star, we plot the periodogram and look for the frequency which gives the maximum power $`z_{\mathrm{max}}`$. The periodogram power $`z`$ is a continuous function of frequency $`f=\omega /2\pi `$. However, the finite duration of the observations $`T`$ gives each periodogram peak a finite width $`\mathrm{\Delta }f1/T`$. Thus in a frequency range $`\mathrm{\Delta }f`$, there are roughly $`T\mathrm{\Delta }f`$ peaks. To make sure we sample all the peaks, we evaluate $`4T\mathrm{\Delta }f`$ periods between 2 days and 30 years<sup>4</sup><sup>4</sup>4The “average” Nyquist period of our observations ($`P_{\mathrm{Nyq}}2T/N`$) is a few months. However, the uneven sampling gives information on much shorter periods (perhaps much shorter than the minimum spacing between observations, see Eyer & Bartholdi 1998). The minimum period we investigate is 2 days. The maximum period of 30 years is a few times greater than the typical duration of the observations.. Monte Carlo tests indicate that this gives adequate sampling of the periodogram. We evaluate $`z`$ at evenly-spaced frequencies using a linear least squares fitting algorithm from Press et al. (1992).
To assess the significance of a possible detection, we test the null hypothesis that the data are pure noise. We ask, what is the false alarm probability associated with $`z_{\mathrm{max}}`$, or how often would noise fluctuations conspire to give a maximum power larger than that observed? We use Monte Carlo tests to find the false alarm probability<sup>5</sup><sup>5</sup>5The false alarm probability increases as the frequency range searched $`\mathrm{\Delta }f`$ increases (Schwarzenberg-Czerny 1996, 1997a, 1998 refers to this as the bandwidth penalty). As we discuss in Appendix B, although the distribution of $`z`$ at one particular frequency is easy to write down analytically, the number of independent frequencies in a frequency range $`\mathrm{\Delta }f`$ is not. For this reason, we adopt a Monte Carlo approach. This also allows us to check for non-Gaussian effects.. For each star, we make fake data sets of either a mean or straight line plus noise. We then perform the same analysis as for the original data. We find the maximum periodogram power, and ask, in what fraction of trials does $`z_{\mathrm{max}}`$ exceeds the observed value? This fraction is the false alarm probability.
We add noise to the simulated data sets in two ways. One is to add Gaussian deviates with the same variance as the observed velocities, keeping the same relative weights and observation times. The second is to take the observed velocities and randomize them, keeping the sample times fixed (the so-called bootstrap method, Press et al. 1992). In this case, we randomize the pre-fix and post-fix velocities separately, to account for their different $`v/\sigma `$ distributions. We find that both approaches give false alarm probabilities which are similar for most stars. We have also applied the analytic distribution given in Appendix B to our results (see Table 9), and fit for the number of independent frequencies $`M`$. The resulting analytic false alarm probabilities agree well with our Monte Carlo calculations.
The results are given in Table 5. For each star, we give the maximum power $`z_{\mathrm{max}}`$, the corresponding best-fit period and velocity amplitude, and the false alarm probabilities determined from 400 Monte Carlo trials for each star. We mark with $``$ those stars which show false alarm probabilities less than 1%. Apart from those stars with confirmed companions (Table 1), these are HR 509, HR 937, HR 996, HR 1084, HR 1729, HR 2047, HR 3951, HR 4345, HR 4496, HR 4540, HR 4983, HR 5019, HR 5273, HR 5544a, HR 5568, HR 6623, HR 7061, HR 7602, HR 7672, and HR 8086. Again, our motivation for choosing a 99% detection threshold is that we then expect no more than one false detection in our sample of 76 stars.
We find that 20 stars have more than one peak in the periodogram with false alarm probability $`<1\%`$. The search for multiple companions is beyond the scope of this paper, however, we carried out a simple test of whether these secondary peaks were due to aliasing of the primary period by the finite sampling. We subtracted off the best fit sinusoid from the data and looked at a periodogram of the residuals to see if the secondary peaks remained. Only in two cases did they remain, HR 458 which has a second peak at 1210 days, and HR 509 which has a second peak at 60 days.
#### 3.4.2 Variability
The average power $`\overline{z}`$ (averaged over frequency) is an additional indicator of variability. If the data are drawn from a Gaussian distribution, we expect the mean power to be $`\overline{z}1`$ (the mean value of the $`F`$-distribution). Again using a Monte Carlo method, we calculate the false alarm probability associated with the observed $`\overline{z}`$. Namely, we ask in what fraction of simulated data sets is the mean power larger than the observed value? A low false alarm probability in this test indicates either a periodicity may be present (the uneven sampling results in “leakage” which contaminates the background level), some non-Gaussian behavior or some kind of “broad-band” variability (for example as might be expected from magnetic activity). The results are shown in Table 6. Several stars show significantly high $`\overline{z}`$ at the 99% level. Apart from those stars with confirmed companions (Table 1), they are HR 88, HR 166, HR 509, HR 937, HR 996, HR 1084, HR 1729, HR 2047, HR 4345, HR 4496, HR 4540, HR 4983, HR 5019, HR 5273, HR 5544a, HR 5568, HR 5868, HR 5914, HR 5933, HR 6623, HR 7061, and HR 7602.
### 3.5 Results of our Search for Companions
Our sample of 76 stars contains eight stars with companions of planetary-mass, HR 3522, HR 4277, HR 458, HR 5072, HR 5185, HR 5968, HR 7504, and HR 8729. All of these companions give highly significant periodogram peaks. The orbital period and velocity amplitude obtained from the periodogram agree well with a Keplerian fit, although not surprisingly, the velocity amplitude is less than that of the Kepler fit for eccentric orbits. The orbital parameters for these planetary-mass companions and references are given in Table 1.
Five stars have companions with $`M_p\mathrm{sin}i>15M_J`$ which have been reported by previous authors. They are HR 2047 ($`M\mathrm{sin}i=0.15M_{}`$, $`P=14.2\mathrm{yr}`$; Irwin, Yang, & Walker 1992), HR 5273 ($`M\mathrm{sin}i=0.4M_{}`$, $`a=5\mathrm{AU}`$; Kamper 1987), HR 5553 ($`K20\mathrm{km}\mathrm{s}^1`$, $`P=125`$ d; Beavers & Salzer 1983), HR 6623 (Cochran & Hatzes 1994 report a slope of $`40\pm 5\mathrm{m}\mathrm{s}^1\mathrm{yr}^1`$) and GL688 ($`K=5.7\mathrm{km}\mathrm{s}^1`$, $`P=83.7`$ d; Tokovinin 1991). These massive companions explain the variability seen in these stars, and the trend in those two, HR 2047 and HR 6623, for which we have not yet seen a complete orbital period. They provide interesting examples of the limitations of the periodogram. There are 20 observations of HR 5553. The orbital period is correctly identified by the periodogram as 125 days, but it is not deemed a significant detection. GL 688 has only 7 observations, and the periodogram is unable even to identify the correct period.
Seven stars have significant long-term trends, the slopes of which are given in Table 7. These are HR 219a, HR 753, HR 1325, HR 5544a, HR 5544b, HR 7672, and HR 8086. Long-term trends such as these most likely indicate a companion with $`a10\mathrm{AU}`$ and $`M_p\mathrm{sin}i>15M_J`$. Of course many of the other slopes listed in Table 3 may also be due to companions of lower mass (but still with periods $`P11\mathrm{yr}`$). The slopes of HR 5544a and HR 5544b are of almost the same magnitude but opposite sign, as expected for orbiting companions. HR 8086 is an interesting case, because much of its slope is due to secular acceleration, a geometrical effect that stems from the exchange of proper motion for radial velocity. This effect is negligible for the other stars in our sample, but we see it for HR 8086 because it is close and has a large velocity relative to us. Secular acceleration also explains why we don’t see a corresponding negative slope in HR 8085, the companion to HR 8086.
The 20 remaining stars which show significant variability or periodicities are listed in Table 7. These stars are candidates for having planetary-mass companions. We include those stars with a significant periodicity, and those without a significant periodicity but with variability in excess of our prediction of §2.1. We have divided these stars into two groups, chromospherically active ($`P_{\mathrm{rot}}14`$ days) and chromospherically quiet ($`P_{\mathrm{rot}}>14`$ days). The best fit velocity amplitudes for the chromospherically quiet stars are of order the predicted scatter in the velocities due to Doppler errors and intrinsic variability. This makes it difficult, even for very significant periodogram peaks, to confidently identify the observed velocity variations with the Keplerian orbit of a companion. For the chromospherically active stars ($`P_{\mathrm{rot}}14`$ days), there is an additional complication. Even though we account for the excess variability expected in chromospherically active stars, intrinsic velocity variations are likely to be periodic on many different timescales. This renders periodicities seen in these stars suspect, as they may be related to the rotation period, convective motions or magnetic activity. Thus, despite having extremely low false alarm probabilities in the periodogram analysis, none of the periodicities listed in Table 7 are convincing as companions. We are currently making more observations of these stars, to attempt to confirm or rule out these periodicities.
## 4 Upper Limits on Companion Mass
For those stars without a confirmed companion, we would like to know the upper limit on the signal amplitude $`K`$. In this section, we use the periodogram to place upper limits on the velocity amplitude as a function of period. Strictly, our upper limits are on the amplitude of circular orbits because we assume a sinusoidal periodicity. However, we expect our upper limits would not be significantly different for eccentric orbits. This is because for a given period, eccentric orbits have a larger velocity amplitude $`K`$, but an extended duration of roughly constant velocity near apastron. These two effects will cancel each other to some extent for a companion of given mass and orbital radius. Of course, this is not true for orbital periods longer than the duration of the observations, for which our upper limits are strictly for circular orbits. We first describe our method in detail and then present our upper limits on companion mass as a function of orbital radius for each star in the survey.
### 4.1 Method
Our method for placing upper limits uses the fact that, because of measurement errors, different observations of the same signal give different periodogram powers. Most of the time, a large signal amplitude will give a large power. Sometimes, because of noise fluctuations, a large signal will give a small power, but less and less often as the signal amplitude increases. This means that, given a particular observation, we can rule out very large signal amplitudes because they have a very small chance of giving a power as small as the observed value<sup>6</sup><sup>6</sup>6This approach to placing upper limits is a “frequentist” one, as discussed by de Jager (1994) (see also Caso et al. 1998), and applied to the Rayleigh test by Protheroe (1987) and Brazier (1994). It was used in searches for pulsations in the quiescent emission from low mass X-ray binaries by Leahy et al. (1983) and Vaughan et al. (1994). In their work, the noise is dominated by photon counting (Poisson) noise and the time series is evenly sampled. This allows the noise level to be “read off” the power spectrum, giving a natural normalization (Leahy et al. 1983). This is not true in the case of uneven sampling, hence the different approaches in the literature to normalizing the periodogram (see Appendix B).. We define the 99% upper limit to be the signal amplitude such that the periodogram power would be less than or equal to the observed value only 1% of the time.
We proceed as follows. For each star, we find the maximum periodogram power $`z_{\mathrm{max}}`$ from the data (as given in Table 5) for periods between 2 days and 30 years. Then for different trial frequencies, we make simulated data sets of a sinusoid (with frequency $`\omega `$, amplitude $`K`$ and randomly-selected phase $`\varphi `$) and noise. We find the periodogram power $`z(\omega )`$ for each simulated data set and ask how often is $`z`$ larger than the observed value $`z_{\mathrm{max}}`$? The 99% upper limit to the velocity amplitude at frequency $`\omega `$ is that $`K`$ which gives $`z>z_{\mathrm{max}}`$ in 99% of trials. In other words, the observed $`z_{\mathrm{max}}`$ lies at the lowest one percentile of the distribution of $`z`$ that stems from the upper limit to the velocity.
For each trial, we evaluate the periodogram at only one frequency, the trial frequency $`\omega `$. This assumes that the maximum periodogram power will occur at the trial frequency. We have tested this assumption by evaluating the upper limits using 100 frequencies centered on the trial frequency, with the same spacing as used in the periodogram. The upper limits from both methods agree well, and hence we adopt the former as it is computationally faster.
To add noise to the simulated data sets, we utilize the residuals to the sinusoid which best fits the data (i.e. the residuals after subtracting from the data the sinusoid corresponding to the maximum periodogram power $`z_{\mathrm{max}}`$). We motivate this in Appendix B, where we show that, if the data consist of a sinusoid plus noise, the best estimate of the noise variance is the variance of the residuals to the best fit sinusoid. This allows us to obtain the noise variance directly from the data, without having to rely on the estimated errors of §2.1. We find that the variance of the residuals is typically less than the estimated error by a factor of 1.5, consistent with our finding that we overestimated the internal errors (see §2.1). We are thus confident that the residuals give a good estimate of the noise variance for each star.
For each observation, we add noise by selecting at random from the $`v/\sigma `$ distribution of the residuals, scaling by the expected variability $`\sigma _j`$ for each data point. When choosing at random, we select from the pre-fix and post-fix data separately, as appropriate. We have also tried adding Gaussian noise (with appropriate relative weighting between points), scaling by the variance of the residuals to the best fit sinusoid. The upper limits from the two methods agree to 10% or better.
### 4.2 Results
The 99% upper limits are plotted in the $`M_p\mathrm{sin}i`$$`a`$ plane as a solid line in Figure 5 for each star in the sample. For clarity, we state again the meaning of our upper limit. For each star, we imagine repeating the observations many times, each with identical sampling, errors and duration as the real observations. The 99% upper limit is the mass ($`M_p\mathrm{sin}i`$) of a companion which, if present at orbital radius $`a`$, would give a periodogram power larger than that observed in the real data ($`z_{\mathrm{max}}`$) in 99% of our repeated observations. Thus a companion more massive than the upper limit is excluded by the data at better than 99% confidence.
For each star, we calculate the upper limit to the velocity amplitude at 500 logarithmically-spaced periods between 2 days and 30 years. To convert velocity and period $`(K,P)`$ to mass and orbital radius $`(M_p\mathrm{sin}i,a)`$, we use equations (1) and (3). We estimate the mass of each star from its spectral type (Table 2) using a simple empirical formula,
$$M/M_{}=1.3\left(\frac{s}{38}\right)+\frac{3}{10}\left(1\frac{s}{30}\right)^5,$$
(8)
where $`s`$ parameterizes the spectral type, ranging from $`s=0`$ for F0 to $`s=30`$ for M0 (for example, $`s=22`$ refers to spectral type K2). This formula reproduces Table 9.6 of Lang (1991) to within 10%. Metallicity renders our masses uncertain by a further 10% (see Carney et al. 1994). This is adequate for our purposes, however.
Inspection of the upper limits shows that, for many periods, the upper limit to the velocity amplitude $`K`$ is roughly independent of period. For each star, we list in Table 8 the average velocity upper limit $`\overline{K}`$ calculated for periods less than half the duration of the observations ($`P<T/2`$). We show the corresponding line of constant velocity in Figure 5 as a dotted line. The average velocity upper limit is a good estimate of the upper limit for most periods less than the duration of the observations ($``$ 11 years for most stars, see Table 2).
The upper limit deviates from this “constant velocity” behavior for two reasons. First, the upper limit is larger at periods where the sampling of the observations gives poor phase coverage. For example, many stars show an increase in the upper limit at $`a1\mathrm{AU}`$ because of the tendency for observations to take place at the same time each year. As expected, these aliasing effects are more important in stars with fewer observations. Second, the upper limit to the velocity amplitude increases for periods greater than the duration of the observations. In Figure 5, we show the orbital radius for which $`P=T`$ for each star by a vertical dashed line. The upper limit at long periods ($`PT`$) is sensitive to how we choose the phase in the simulated data sets. The data contain some information about the best fit phase at each period, but we (conservatively) ignore that and choose the phase at random for each trial.
In Figure 5 we do not show results for the 8 stars with confirmed planetary-mass companions (Table 1), or for the 5 stars with companions with $`M_p\mathrm{sin}i>15M_J`$ and $`P20\mathrm{yr}`$ (HR2047, HR5273, HR5553, HR6623 and GL 688). An interesting question is whether these stars have second companions. We are currently generalizing our approach to the case of two companions. For now, we note that the residuals to Keplerian fits to the velocities for the five stars with $`M_p\mathrm{sin}i>15`$ are less than $`30\mathrm{m}\mathrm{s}^1`$ in each case. Searches for second companions to the stars with planetary-mass have been made (see Table 1 for references), without success except HR 3522 ($`\rho ^1`$ 55 Cnc) and HR 458 (Ups And) show evidence for long period second companions ($`P3`$ years).
Walker et al. (1995) (hereafter W95) also used a floating-mean periodogram to place upper limits on companion mass, but took a different approach based on subtracting a sinusoid directly from the data<sup>7</sup><sup>7</sup>7This kind of approach to placing confidence limits is discussed by Lampton, Margon & Bowyer (1976), including the cases where the noise level must be estimated from the data and some of the parameters are “uninteresting” (see also Cline & Lesser 1970, Avni 1976, Cash 1976 and the discussion in Press et al. 1992).. We have applied our method to the data of W95 and find good agreement with their published upper limits. Comparison with their Figure 5 shows that our upper limits on companion mass are about 10–20% lower for $`P<T`$ and 20–30% higher for $`P>T`$. In addition, our upper limits, which show less variability from one period to the next, seem less affected by the sampling of the data. The reason for these differences is not clear. Our comparison shows that there is a “theoretical uncertainty” in the upper limits of about 20%. The good agreement of the two techniques, which are quite different, is encouraging.
## 5 Discussion
So far, we have searched for companions and determined upper limits on the mass of companions for each individual star. We now investigate the implications of our results for the population of planetary-mass companions as a whole.
The observed distribution of the mass and orbital radius of all known planetary mass companions is shown in Figure 9, in which we plot the 17 confirmed companions listed in Table 1 in the $`M_p\mathrm{sin}i`$$`a`$ plane. The dashed lines show constant velocity contours of $`10`$ and $`40\mathrm{m}\mathrm{s}^1`$ for a 1 $`M_{}`$ star. Four features of the observed distribution are interesting. First, all confirmed companions have $`K40\mathrm{m}\mathrm{s}^1`$. Second, no companions have been detected with orbital radius $`a2.5\mathrm{AU}`$. Third, there is a “piling-up” of companions at small orbital radii; for example, of the 17 companions within $`2.5\mathrm{AU}`$, 13 have semi-major axis $`a<0.5\mathrm{AU}`$ and 5 have $`a<0.1\mathrm{AU}`$. Fourth, there is a paucity of companions with orbital radii between $`0.3`$ and $`1\mathrm{AU}`$. In this section, we ask whether our results help to explain these features.
### 5.1 Confirmed and Candidate Companions
First, we summarize the results of our search for companions. We began with a sample of 76 stars from the original Lick Survey. Two stars, HR 5968 ($`\rho `$ CrB) and HR 8729 (51 Peg) were included in the sample because of the discovery of the companions to these stars by other groups (Table 1), and for this reason cannot be considered part of a statistically unbiased sample. Of the remaining 74 stars, six have confirmed planetary-mass companions, or about 8% of our sample. Our periodogram analysis of §3.4 reveals several candidate periodicities (marked “$``$” in Table 5 and listed in Table 7), which are yet to be confirmed or ruled out as being due to companions.
We plot the confirmed companions (circles) and candidate periodicities (squares) in the $`M_p\mathrm{sin}i`$$`a`$ plane in Figure 10. We do this for purely illustrative purposes: we stress that it may be that none of the candidate periodicities are actually due to companions. Open and filled squares indicate candidate periodicities for chromospherically active and quiet stars, respectively. The error bars show the 99% upper limit on $`M_p\mathrm{sin}i`$ obtained as described in §4.1. Broken circles show those confirmed companions not included in our sample of 74 stars. The dashed lines show constant velocity contours of 5, 10, 20 and 40$`\mathrm{m}\mathrm{s}^1`$ for a 1$`M_{}`$ star.
Inspection of Table 7 shows that there are several candidates for companions with velocity amplitude between the Doppler errors, $`K5`$$`10\mathrm{m}\mathrm{s}^1`$, and the lowest velocity amplitude detected, $`K40\mathrm{m}\mathrm{s}^1`$. Thus, one interpretation of our results is that there is an effective detection threshold of $`K40\mathrm{m}\mathrm{s}^1`$. This could be caused by the fact that confirmation of these low amplitude orbits is difficult, as we discussed in §3.5. However, there is an interesting difference in the velocity amplitudes of the candidate periodicities from chromospherically quiet versus chromospherically active stars. Figure 10 shows that there are no candidate periodicities with $`K15\mathrm{m}\mathrm{s}^1`$ in the quiet stars, whereas all of the candidate periodicities in the active stars have $`K15\mathrm{m}\mathrm{s}^1`$. Thus another interpretation is that there is a paucity of companions with velocity amplitudes $`K15`$$`40\mathrm{m}\mathrm{s}^1`$, as none are seen in the subset of quiet stars for which these velocity amplitudes could have been detected. However, we cannot draw conclusions from the present data, as the small number of detections is subject to $`\sqrt{N}`$ fluctuations.
Figure 10 also shows that there are fewer candidates with orbital radii between $`a0.2\mathrm{AU}`$ and $`a1\mathrm{AU}`$ than at other orbital radii. If these periodicities are due to intrinsic stellar variations, a possible explanation is that these variations naturally occur on two timescales, the stellar rotation period ($``$ 1 month) and the timescale of the magnetic cycle ($``$ years). However, the apparent paucity of confirmed companions and candidate periodicities between $`a0.2\mathrm{AU}`$ and $`a1\mathrm{AU}`$ is interesting, and may indicate a real paucity of companions at these orbital radii.
### 5.2 Upper Limits and Detectability
We now turn to the upper limits which we calculated in §4. In Figure 11, we show a histogram (left panel) and cumulative histogram (right panel) of the mean velocity upper limits listed in Table 8. For most stars, we can exclude companions with $`K10`$$`20\mathrm{m}\mathrm{s}^1`$ at the 99% level. Nine stars (15% of the sample) have a mean upper limit $`\overline{K}<10\mathrm{m}\mathrm{s}^1`$ and 38 stars (60% of the sample) have $`\overline{K}<20\mathrm{m}\mathrm{s}^1`$. About 10 stars have $`\overline{K}40\mathrm{m}\mathrm{s}^1`$: inspection of Tables 2 and 8 shows that these stars have either a small number of observations, poor internal errors (for example, because they are faint), short rotation period or a large rms $`>40\mathrm{m}\mathrm{s}^1`$ (for example, because of magnetic activity).
In Figure 12, we plot the cumulative histogram in the $`M_p\mathrm{sin}i`$$`a`$ plane. We take into account the different stellar masses and the effects of sampling (i.e. we use the upper limits as a function of period, Figure 5, rather than just $`\overline{K}`$, Table 8). The solid lines show contours of the number of stars from which a companion of given mass and orbital radius can be excluded at the 99% level (plotted on a 40 x 40 grid). Each contour is labeled by the number of stars from which we can exclude companions above and to the left of the solid line. The dashed lines show constant velocity contours of 5, 10, 20 and 40$`\mathrm{m}\mathrm{s}^1`$ for a 1$`M_{}`$ star. The filled squares show the six confirmed planetary-mass companions in our sample (Table 1). In Figure 13, we show sections of this contour map. We plot the number of stars from which a companion can be excluded (at the 99% level) as a function of orbital radius for different masses, and as a function of mass for different orbital radii.
Figure 13 demonstrates the effect of the finite duration of the observations. Even for massive companions (e.g. $`M_p\mathrm{sin}i>5M_J`$), the number of stars from which a companion can be excluded decreases rapidly for $`a6\mathrm{AU}`$. In Figure 12, we plot with an open circle the point $`M_p\mathrm{sin}i=1M_J`$, $`a=5.2\mathrm{AU}`$, or where our solar system would lie in this diagram if $`\mathrm{sin}i=1`$. This shows that analogs of our solar system are excluded from only a handful of stars once the distribution of $`\mathrm{sin}i`$ is taken into account. This is not true for more massive companions, with $`M_p\mathrm{sin}i>3`$$`4M_J`$. Our results show that these companions are not common at orbital radii $`a1\mathrm{AU}`$, as found by Walker et al. (1995) in their survey of 21 stars. For example, companions with $`M_p\mathrm{sin}i>3M_J`$ ($`>5M_J`$) can be excluded from 80% (95%) of our sample of 63 stars.
Our results give some sense of how the detectability of a companion of a particular mass falls off with radius. For example, for $`M_p\mathrm{sin}i=1M_J`$ ($`0.5M_J`$), the number of stars excluded falls off for $`a0.3`$$`0.5\mathrm{AU}`$ ($`0.1\mathrm{AU}`$). Thus we have detected all companions in the sample with $`a0.3\mathrm{AU}`$ and $`M_p\mathrm{sin}i1M_J`$. For the 40 stars with $`\overline{K}20\mathrm{m}\mathrm{s}^1`$, we can exclude $`1M_J`$ companions out to $`1\mathrm{AU}`$, and $`2M_J`$ companions out to $`4\mathrm{AU}`$. Unfortunately, we cannot assume that the detectability of companions is the same in other surveys. Thus we must restrict our attention to the six companions in our sample.
We have performed a simple test of whether the six companions in our sample could have been drawn from a parent population which is uniformly distributed in orbital radius. First, we assume that the mass distribution of companions is uniform in $`\mathrm{log}M_p\mathrm{sin}i`$ between 0.1 and $`4M_J`$. This seems consistent with the observed distribution of $`M_p\mathrm{sin}i`$ (at least for small orbital radii where detectability is good). We select companions at random from this mass distribution and a uniform distribution in orbital radius from 0 to $`2.5\mathrm{AU}`$. We assign the companion to a star in the sample at random, simulate observations of the star using the observation times and errors from the real data, and ask whether the periodogram power is larger than that observed. The “corrected” orbital radius distribution is then the distribution of orbital radii of those companions which give periodogram powers larger than the observed values. We then use the Kolmogorov-Smirnov (KS) test (Press et al. 1992) to find the probability that the six companions are drawn from this “corrected” distribution.
We find that the probability that the six companions are drawn from a parent population with a uniform distribution in orbital radius is $`10\%`$. If we do not include 70 Vir, which may have a low $`\mathrm{sin}i`$ and thus a mass $`>15M_J`$, the probability is 5%. Thus we cannot rule out that the distribution of companions is uniform in radius. Also, we emphasize again that we are dealing with a small number of detections which are subject to $`\sqrt{N}`$ fluctuations. Future detections will enable us to use the KS test in this way to rule out distributions in mass and orbital radius. For now, without knowledge of the detection thresholds of other surveys, we cannot rule out the possibility that the “pile-up” of companions at low orbital radius or the lack of companions between $`0.2\mathrm{AU}`$ and $`1\mathrm{AU}`$ are due to selection effects or small number statistics.
## 6 Summary and Conclusions
We have presented an analysis of eleven years of precision Doppler velocity measurements of 76 G, K and F type stars from the Lick radial velocity survey. We have performed tests for variability, long term trends, and periodicities. Our sample contains eight confirmed planetary-mass companions, six of which were discovered at Lick, and five companions of stellar or sub-stellar mass (§3.5). Seven stars have significant long term trends, likely indicating a companion with $`a10\mathrm{AU}`$ and $`M_p\mathrm{sin}i>15M_J`$, and 20 stars show variability or periodicities which may indicate a planetary-mass companion (Table 7). We are currently making more observations of these stars, to attempt to confirm or rule out these periodicities. For those stars without a confirmed companion, we have calculated upper limits to the mass of a companion as a function of orbital radius (Figure 5). For most stars, the mean limit on the velocity amplitude is between 10 and $`20\mathrm{m}\mathrm{s}^1`$ or $`0.35`$ and $`0.7M_J(a/\mathrm{AU})^{1/2}`$ (for stellar mass $`MM_{}`$).
We have searched for periodicities and placed upper limits using a “floating-mean” periodogram, in which we fit sinusoids to the data, allowing the zero-point of the sinusoid to “float” during the fit. Allowing the mean to float is crucial to account for statistical fluctuations in the mean of a sampled sinusoid. The traditional Lomb-Scargle periodogram fails when the number of observations is small, the sampling is uneven or the period of the sinusoid is comparable to or greater than the duration of the observations. This may lead to missed detections, or inaccurate upper limits. We have also expanded on the recent discussion by Schwarzenberg-Czerny (1997a, 1997b, 1998) of the correct way to normalize the periodogram and the resulting distribution of periodogram powers. The three different prescriptions in the literature for normalizing the periodogram are statistically equivalent, and all three give a distribution of periodogram powers for Gaussian noise which is significantly different from the usually assumed exponential distribution (see Figure 14). Unfortunately, it is not possible in general to write a simple analytic formula for the false alarm probability, making Monte Carlo methods essential.
Our results help to explain the observed distribution of mass and orbital radius of companions (see Figure 9). The confirmed companions so far have velocity amplitudes $`K>40\mathrm{m}\mathrm{s}^1`$, whereas the Doppler errors lie between $`5`$$`15\mathrm{m}\mathrm{s}^1`$. This is most likely because there is an effective detection threshold which comes about because of the ambiguity introduced by intrinsic velocity flutter (which may be periodic). Confirmation of an orbit is then difficult when the velocity amplitude is similar to the scatter predicted because of intrinsic variability and Doppler errors. We note, however, that in the chromospherically quiet stars ($`P_{\mathrm{rot}}14`$ days), there are no candidate periodicities in the range $`15`$$`40\mathrm{m}\mathrm{s}^1`$. This may reflect a real paucity of companions in this range.
The finite duration of the observations makes it difficult to detect Jupiter-mass companions with orbital radius $`a3\mathrm{AU}`$. Thus the four known companions with $`a>1\mathrm{AU}`$ may be only the first of a population of Jupiter-mass companions at large orbital radii. This is not true for more massive companions, however. It is striking that companions with $`M_p\mathrm{sin}i>3M_J`$ are rare at orbital radii $`4`$$`6\mathrm{AU}`$; we could have detected such objects in $`90\%`$ of stars, yet found none.
A few more years of observations will allow detection of Jupiter-mass companions at $`a5\mathrm{AU}`$, particularly as the poorer quality pre-fix observations become less important. Already, we are able to exclude velocity amplitudes of $`10\mathrm{m}\mathrm{s}^1`$ from 15% of the stars in the sample (for $`a3\mathrm{AU}`$). The velocity amplitudes which can be detected are ultimately limited by intrinsic stellar variability. Even for chromospherically inactive stars ($`P_{\mathrm{rot}}14`$ days), there is intrinsic flutter of a few m/s. Detectability of short period companions should improve in the future as there is more feedback between the observed variability and future observations (ie. stars which show variability are observed more often). Care must be taken to include this effect in assessment of detectabilities.
Our analysis has assumed that the orbits of companions are circular. Yet eccentric orbits appear to be the norm for many of the planetary-mass companions (see Table 1 and Marcy et al. 1999). We find, however, that the periodogram gives a good estimate of the period for eccentric orbits. In addition, we do not expect our upper limits to change substantially for eccentric orbits, as we argued in §4, except for long period orbits for which eccentricities are important. The possibility of a non-zero eccentricity makes identification of an orbital period impossible for periods more than two or three times the duration of the observations. A possible extension of the periodogram to non-circular orbits would be to fit a Kepler orbit at each frequency, and define the periodogram power in terms of the $`\chi ^2`$ of the Kepler fit (see eq. ). It is not clear if the gain in detectability would outweigh the computing power needed for this non-linear least squares fit. A better approach may be to fit higher harmonics as suggested by Schwarzenberg-Czerny (1996).
The observed distribution of companions in mass and orbital radius shows a “piling-up” towards small orbital radii and a paucity of companions between orbital radii $`a0.2\mathrm{AU}`$ and $`a1\mathrm{AU}`$. Because of the small number of companions in our sample, it is not possible for us to say whether these features are due to selection effects. Unfortunately, without knowledge of the detection thresholds of the other radial velocity surveys, we cannot include the other confirmed companions in our analysis. The candidate periodicities we find also show fewer candidates between $`a0.2\mathrm{AU}`$ and $`a1\mathrm{AU}`$, which is intriguing. Future detections will show whether these features reflect the parent population of planetary-mass companions. It may be that companions are to be found at all orbital radii, or it may be that there are two populations of Jupiter-mass companions, one at orbital radii $`a0.2\mathrm{AU}`$ and one at orbital radii $`a1\mathrm{AU}`$.
Either scenario presents an interesting challenge to theorists. Orbital migration models have been proposed to explain the presence of giant planets at small orbital radii (Lin et al. 1999; Murray et al. 1998; Trilling et al. 1998; Ward 1997). These models naturally predict a “piling-up” at small orbital radii (Trilling et al. 1998) because the orbital migration timescale grows progressively shorter as the planet spirals inwards. However, it is not clear exactly how such migration might be halted, in particular at orbital radii as large as $`0.2\mathrm{AU}`$. Indeed, the inevitability of migration may be responsible for the low percentage of solar type stars which have close planetary-mass companions (Ward 1997). If migration depends on gap formation, one would expect migration only to occur for companions above a certain mass. As yet, there is no observed dependence of the mass distribution on orbital radius, except for the lack of companions with $`M_p\mathrm{sin}i3`$$`4M_J`$ at large orbital radius, $`a3`$$`5\mathrm{AU}`$.
There have also been suggestions that gravitational scattering of planets by other planets, companion stars or neighbouring stars in a young star cluster may play a role in determining the final distribution of orbital radii (Rasio & Ford 1996; Weidenschilling & Marzari 1996; Lin & Ida 1997; Laughlin & Adams 1998; Levison, Lissauer & Duncan 1998). The large range of orbital eccentricities of the observed companions may be evidence for this type of scenario (Marcy et al. 1999). One way to lose enough energy to allow a planet to move from $`5\mathrm{AU}`$ to $`<1\mathrm{AU}`$ may be interaction with the protoplanetary disk during the last stages of dissipation, as suggested by Marcy et al. (1999). It is not known to what extent these different physical mechanisms play a role in determining the distribution of planet masses and orbital radii. Clear theoretical predictions are needed if the discovery of more planetary mass companions is to allow us to distinguish between these different pictures.
We would like to thank Lars Bildsten, Debra Fischer, Evan Scannapieco, Jonathan Tan and Andrew Youdin for useful discussions as this work progressed. Debra Fischer provided rotation periods for several stars, for which a program developed by Phil Shirts was used to measure the Ca IR triplet emission. Jeff Scargle and the referee, Andy Nelson, provided thoughtful comments on the manuscript. Alex Schwarzenberg-Czerny kindly provided a copy of his paper (Schwarzenberg-Czerny 1998). AC was partly supported by the Victor F. Lenzen Memorial Scholarship Fund at UC Berkeley. GWM was partly supported by NSF grant AST 95-20443 and NASA grant NAGW-3182.
## Appendix A Derivation of the Lomb-Scargle Periodogram
In this section, we sketch the derivation of the traditional Lomb-Scargle formula from a least squares fit of a sinusoid to the data, and show how it relates to the floating mean periodogram used in this paper. Following Lomb (1976), for each trial frequency $`\omega =2\pi /P`$, our model for the velocities is
$$f_j(t_j)=A\mathrm{cos}\omega (t_j\tau )+B\mathrm{sin}\omega (t_j\tau ).$$
(A1)
The constant $`\tau `$ is introduced to simplify the calculation of $`\chi ^2`$, as the cross terms cancel if we choose $`\tau `$ such that
$$w_j\mathrm{sin}\omega (t_j\tau )\mathrm{cos}\omega (t_j\tau )=0,$$
(A2)
or equivalently
$$\mathrm{tan}(2\omega \tau )=\frac{w_j\mathrm{sin}2\omega t_j}{w_j\mathrm{cos}2\omega t_j}.$$
(A3)
The best fit values of the parameters $`A`$ and $`B`$ are those which minimize $`\chi ^2=w_j(v_jf_j)^2`$, where $`w_j1/\sigma _j^2`$ is the weight for data point $`j`$. Setting $`\chi ^2/A=0`$ and $`\chi ^2/B=0`$ and using equation (A2) gives
$`A={\displaystyle \frac{w_jv_j\mathrm{cos}\omega (t_j\tau )}{w_j\mathrm{cos}^2\omega (t_j\tau )}},`$
$`B={\displaystyle \frac{w_jv_j\mathrm{sin}\omega (t_j\tau )}{w_j\mathrm{sin}^2\omega (t_j\tau )}}.`$ (A4)
The amplitude $`K`$ and phase $`\varphi `$ of the sinusoid are obtained from $`K^2=A^2+B^2`$ and $`\mathrm{tan}\varphi =A/B`$.
Using equation (A) to substitute the best fit $`A`$ and $`B`$ into $`\chi ^2=w_j(v_jf_j)^2`$, we find the minimum value of $`\chi ^2`$ at each frequency is
$$\chi _{\mathrm{min}}^2(\omega )=w_jv_j^2w_jf_j^2,$$
(A5)
where $`f_j`$ is now the best fit sinusoid. As in §3.3, we define the unnormalized periodogram power $`\widehat{z}`$ as the reduction in the sum of squares (Lomb 1976),
$$\widehat{z}(\omega )w_jf_j^2=w_jv_j^2\chi _{\mathrm{min}}^2(\omega ),$$
(A6)
or
$`\widehat{z}(\omega )={\displaystyle \frac{\left[w_jv_j\mathrm{cos}\omega (t_j\tau )\right]^2}{w_j\mathrm{cos}^2\omega (t_j\tau )}}+`$
$`{\displaystyle \frac{\left[w_jv_j\mathrm{sin}\omega (t_j\tau )\right]^2}{w_j\mathrm{sin}^2\omega (t_j\tau )}}.`$ (A7)
Equation (A) is the (unnormalized) Lomb-Scargle periodogram (Lomb 1976; Scargle 1982; Horne & Baliunas 1986) modified for unequally-weighted data (Gilliland & Baliunas 1987; Irwin et al. 1989; Scargle 1989). Equation (A6) shows that the floating-mean periodogram we adopt in this paper (§3.3) is a straightforward generalization of the traditional periodogram (see also Walker et al. 1995, eq. \[A2\]).
## Appendix B Normalization of the Periodogram and The Distribution of Noise Powers
There are three different prescriptions in the literature for normalizing the periodogram, namely dividing by (i) the sample variance (Horne & Baliunas 1986; Irwin et al. 1989; Walker et al. 1995), (ii) the variance of the residuals to the best fit sinusoid (Gilliland & Baliunas 1987, this paper), or (iii) the variance of the residuals to the best fit sinusoid at each frequency (Schwarzenberg-Czerny 1996). The motivation for these normalizations is to give the periodogram power $`z`$ a simple statistical distribution when the data are pure noise. The goal is to assess the false alarm probability associated with a given periodogram power. Recent work by Schwarzenberg-Czerny (1997a, 1997b, 1998) has shown that all three normalizations are statistically equivalent, and, at a given frequency, each leads to a simple analytical distribution for Gaussian noise. We now expand on this discussion and describe our choice of normalization.
We point out that because of non-orthogonality between different frequencies, it is difficult to estimate the number of independent frequencies for a given data set. The commonly used empirical formula of Horne & Baliunas (1986) for the number of independent frequencies is based on an inaccurate probability distribution, and is valid only for a particular frequency range. We stress that, unfortunately, it is not possible to write a simple analytic form for the false alarm probability, making Monte Carlo methods essential.
Why normalize the periodogram at all? Scargle (1982) and Horne & Baliunas (1986) showed that if the data points $`X_j`$ are independent Gaussian deviates with variance $`\sigma _0^2`$, the distribution of unnormalized periodogram powers is
$$f(\widehat{z})d\widehat{z}=\frac{1}{\sigma _0^2}\mathrm{exp}(\frac{\widehat{z}}{\sigma _0^2})d\widehat{z}$$
(B1)
(or simply $`\chi ^2`$ with $`2`$ degrees of freedom). This analysis extends to the weighted periodogram, where $`\sigma _0^2`$ is a measure of the overall normalization of the weights. We would like to stress that if the noise variance were somehow known in advance<sup>8</sup><sup>8</sup>8This is often the situation in X-ray astronomy, for example, where the noise is dominated by Poisson photon statistics, giving a well-defined background power level (Leahy et al. 1983). See also the discussion in Lampton, Margon & Bowyer (1976)., one could simply normalize by the known variance $`\sigma _0^2`$, thus obtaining
$$f(z)dz=\mathrm{exp}(z)dz,$$
(B2)
an exponential distribution of periodogram powers. However, in many cases $`\sigma _0^2`$ is not known accurately in advance and must be estimated from the data. The idea is to normalize the noise powers to a known level<sup>9</sup><sup>9</sup>9As we noted in §4.1, in principle the background noise level in the power spectrum gives a direct measure of $`\sigma _0^2`$. However, in practice the uneven sampling results in contamination of the noise powers by the signal because of spectral leakage and aliasing effects., aiding identification of localized features in the power spectrum (e.g. periodic signals).
To estimate the noise level from the data set, we use an “analysis of variance” approach (Schwarzenberg-Czerny 1989; Davies 1990). In §A, we showed that
$$w_jv_j^2=\widehat{z}(\omega )+\chi _{\mathrm{min}}^2(\omega ).$$
(B3)
We now define
$$s^2\frac{1}{Nm}w_jv_j^2,s_f^2\frac{\widehat{z}}{2},s_n^2\frac{1}{Nm2}w_j(v_jf_j)^2,$$
(B4)
and rewrite equation (B3) in terms of variances, giving
$$(Nm)s^2=2s_f^2+(Nm2)s_n^2.$$
(B5)
We have partitioned the variance into two pieces, together with their respective degrees of freedom; one piece from the signal $`s_f^2`$ and one piece from the noise $`s_n^2`$. Schwarzenberg-Czerny (1989, 1997a, 1998) showed that under the null hypothesis, $`s_n^2`$ and $`s_f^2`$ are statistically independent by Fisher’s Lemma. This is also true if $`v_j`$ consists of a sinusoidal signal plus Gaussian noise. In this case, $`s_n^2`$ is an unbiased estimate of the noise variance. The different definitions of the normalized periodogram are simply different ratios of the variances in equation (B5). The “traditional” normalization by the sample variance is $`zs_f^2/s^2`$ whereas normalizing by the residuals to the noise gives $`zs_f^2/s_n^2`$.
The distribution of powers when the $`X_j`$ are independent Gaussian deviates can be written down analytically (Schwarzenberg-Czerny 1997a, 1997b). The partitioning of the degrees of freedom means that $`s^2`$, $`s_f^2`$ and $`s_n^2`$ are $`\chi ^2`$ distributed with $`Nm`$, $`2`$ and $`Nm2`$ degrees of freedom respectively. Thus $`z=s_f^2/s_n^2`$ follows an $`F`$ distribution with $`2`$ and $`Nm2`$ degrees of freedom (for example, see Abromowitz & Stegun 1971, §26.6). The distribution of $`z=s_f^2/s^2`$ is more complex as $`s^2`$ is correlated<sup>10</sup><sup>10</sup>10This correlation was neglected by Koen (1990), who incorrectly presumed an $`F`$ distribution in this case. with $`s_f^2`$. The distribution of $`z`$ in this case is an incomplete beta function (see Abromowitz & Stegun 1971, §26.5). The probability that the periodogram power $`z`$ is larger than a given value $`z_0`$ is given in Table 9 for the different normalizations.
Given a probability distribution for the periodogram power, we can write down an expression for the false alarm probability. If the probability that a periodogram power $`z`$ is above some value $`z_0`$ is $`\mathrm{Prob}(z>z_0)`$ (as in Table 9), then the false alarm probability is
$$F=1(1\mathrm{Prob}(z>z_0))^M,$$
(B6)
where $`M`$ is the number of independent frequencies that were examined. In Figure 14, we show the distribution of maximum periodogram powers for sets of evenly spaced data with $`N=20`$. The crosses are the results of our Monte Carlo simulations. We used three different normalizations for the periodogram. The lines show the theoretical distributions. In the case where the noise variance $`\sigma _0^2`$ is known, we use equation (B6) with $`\mathrm{Prob}(z>z_0)=\mathrm{exp}(z_0)`$ and $`M=N`$ (dotted line). For the normalizations by $`s^2`$ and $`s_n^2`$ (dot-dashed and dashed lines), we use the distributions given in Table 9 and find the best fit value of $`M`$, fitting to the tail ($`\mathrm{Prob}>0.5`$) of the distribution. Both distributions give the same value, $`M=23.5`$. Notice that normalizing by $`s_n^2`$ broadens the distribution of maximum powers (because of the extra uncertainty in the value of $`s_n^2`$), whereas normalizing by $`s^2`$ narrows the distribution (because of the correlation with $`s_f^2`$).
Horne & Baliunas (1986) used Monte Carlo simulations to find the false alarm probability for evenly-spaced data sets<sup>11</sup><sup>11</sup>11Note that due to a typographical error, the values of $`M`$ given in the tables in HB86 are incorrect (Baliunas 1998, private communication).. Normalizing by the sample variance, they assumed $`\mathrm{Prob}(z>z_0)=\mathrm{exp}(z_0)`$ and fit for $`M`$ as a function of $`N`$. However, as Figure 14 shows, the distribution of $`z`$ is different from exponential, especially in the tails of the distribution. Thus the relations of Horne & Baliunas (1986) give inaccurate estimates of false alarm probabilities or detection thresholds. Because the distribution is squashed, the effect is to overestimate both detection thresholds and false alarm probabilities.
In agreement with Press et al. (1992) and Marcy & Benitz (1989), Horne & Baliunas found that $`MN`$ when the period range searched was from the average Nyquist period to the duration of the data set. However, in this paper, we evaluate frequencies several times larger than the average Nyquist frequency. A naive estimate of the number of independent frequencies is $`T\mathrm{\Delta }f`$. Our numerical results show that in general the number of independent frequencies is less than this estimate, hence our Monte Carlo approach for finding the false alarm probabilities in §3.4. It would be useful to have a method, for a given data set, of estimating the number of independent frequencies, allowing one to write down false alarm probabilities analytically. One approach may be to look at the correlations between residuals (Schwarzenberg-Czerny 1991).
Finally, we discuss the difference between normalizing by $`s_n^2`$ at each frequency, and normalizing by $`s_n^2`$ evaluated at the best fit frequency. This choice is really a matter of taste. The distribution of maximum periodogram powers is the same, by definition. However, it seems to us that it is fairer to make comparisons between frequencies using the same normalization for the noise in $`\chi ^2`$. Thus our choice of normalization is that of Gilliland & Baliunas (1987, eq. ): we normalize by the same factor for each frequency, namely $`s_n^2`$ evaluated at the best fit period.
|
no-problem/9906/hep-ph9906490.html
|
ar5iv
|
text
|
# Abelian Family Symmetries and Leptogenesis
## I Introduction
There is now strong evidence for atmospheric neutrino oscillations. The data suggests that $`\nu _\mu \nu _\tau `$ oscillations occur with near maximal mixing $`\mathrm{sin}^22\theta _{23}1`$ and a mass splitting of $`\mathrm{\Delta }m_{23}^22.2\times 10^3`$ eV<sup>2</sup>. The measured solar neutrino flux can be explained by oscillations of $`\nu _e`$ to the other two generations ($`x=2,3`$). In the case of matter oscillations (MSW) there are two solutions: (1) the small mixing angle (SMA) solution for which $`\mathrm{\Delta }m_{1x}^25\times 10^6`$ eV<sup>2</sup> and $`\mathrm{sin}^22\theta _{1x}6\times 10^3`$, and (2) the large mixing angle (LMA) solution for which $`\mathrm{\Delta }m_{1x}^22\times 10^5`$ eV<sup>2</sup> and $`\mathrm{sin}^22\theta _{1x}0.8`$. In the case of vacuum oscillations (VO) the mass-squared difference is much smaller $`\mathrm{\Delta }m_{1x}^28\times 10^{11}`$ eV<sup>2</sup> and the mixing angle is also large, $`\mathrm{sin}^22\theta _{1x}0.8`$. The largeness of the mixing $`\theta _{23}`$ and possibly in $`\theta _{1x}`$ and the apparent hierarchy in the associated masses presents something of a dilemma, since one would expect that large mixing of order one occur when the eigenvalues (neutrino masses) are roughly degenerate. Many models have been proposed to account for the neutrino oscillation data, and it is interesting to explore whether these models can account in a natural way for the baryon asymmetry of the universe through the process of leptogenesis. In this paper we explore the implications for Abelian family symmetries on lepton asymmetries generated in the early universe. In particular we argue that a discrete $`Z_2`$ component can not only resolve the dilemma of large mixing together with a large hierarchy mentioned above, but it can also lead to an enhanced baryon asymmetry.
## II The Baryon Asymmetry and Leptogenesis
The lightness of the three known neutrinos can be understood as arising from the see-saw mechanism where right-handed neutrinos, being Standard Model singlets, have a very large mass. The addition of right-handed singlet neutrinos to the Standard Model leads to lepton number violation. The existence of very heavy right-handed neutrinos are predicted by grand unified theories based on the gauge group $`SO(10)`$, and the lightness of the observed neutrinos can be explained via a see-saw mechanism. Since the heavy right-handed neutrinos offer a reasonable basis for the observed oscillations and neutrino masses, it motivates the consideration of their possible cosmological effects. Since these particles would naturally occur in the early universe, it is of interest to determine whether it is possible that the decays of these heavy particles could be the source of the baryon asymmetry of the universe.
The nonzero net baryon density $`n_Bn_{\overline{B}}`$ of the universe can be accounted for in theories that satisfy Sakharov’s conditions: 1) baryon number is violated, 2) charge conjugation symmetry (C) and CP are violated, and 3) there is a departure from thermal equilibrium. A nontrivial requirement on any particle theory satisfying these three conditions is that a sufficient asymmetry in $`n_B`$ and $`n_{\overline{B}}`$ be produced to explain the observed value of the ratio of net baryon density to the entropy density $`s`$ of the universe
$`Y_B={\displaystyle \frac{n_Bn_{\overline{B}}}{s}}=(0.61)\times 10^{10}.`$ (1)
The Standard Model in the early universe satisfies all three conditions, but it is generally agreed that the produced asymmetry is too small. Therefore one is motivated to look beyond the Standard Model at theories that contain new sources of baryon number violation and CP-violation and/or for theories that have a new mechanism for producing the asymmetry. If one instead considers the Minimal Supersymmetric Standard Model (MSSM) then the regions of parameter space where sufficient baryon asymmetry is produced is quite small. Consequently various proposals have been made for new physics capable of producing the baryon asymmetry of the universe. One of the most attractive of these is the possibility that CP violating decays of heavy neutrinos can produce an excess of leptons over antileptons (or vice versa). The lepton asymmetry produced in the early universe via out-of-equilibrium decays of the right-handed neutrinos is subsequently recycled into a baryon asymmetry by sphaleron transitions (which violated both baryon number and lepton number). A straightforward analysis of chemical potentials for equilibrating processes including the sphaleron transition relates the baryon asymmetry $`Y_B`$ to the original lepton asymmetry $`Y_L=(n_Ln_{\overline{L}})/s`$ via
$`Y_B=aY_{BL}={\displaystyle \frac{a}{a1}}Y_L,a={\displaystyle \frac{8N_F+4N_H}{22N_F+13N_H}},`$ (2)
where $`N_F`$ is the number of fermion families and $`N_H`$ is the number of Higgs doublets. So the final baryon asymmetry present in the universe today is related to the lepton asymmetry $`Y_L`$ by an order one parameter. If one accepts the presence of heavy Majorana neutrinos in nature, then CP-violation naturally occurs and the question becomes whether or not the lepton asymmetry that results is the right order of magnitude for producing the observed baryon asymmetry in Eq. (1). In the MSSM with heavy right-handed neutrinos, the resulting lepton asymmetry has been shown to be sufficient to explain the observed baryon asymmetry in a natural way in a number of models.
Most work in trying to understand the structure of the fermion masses and mixings has tried to fit the low energy data, e.g. the fermion masses and the CKM matrix as well as the neutrino data (especially the solar neutrino oscillation data and the atmospheric neutrino oscillation data). If one accepts the notion that leptogenesis is the source of the baryon asymmetry of the universe, then this mechanism imposes another rather strong constraint on the details of the family symmetry (this kind of symmetry is also called a horizontal symmetry). For example the lepton asymmetry produced by the decay of heavy Majorana neutrinos is sensitive to the texture pattern of the Yukawa matrices as well as the details of the mass and mixing hierarchies. In the next section we apply the strategy of employing an Abelian family symmetry to describe the hierarchies and discuss the implications for leptogenesis.
## III Horizontal Symmetries
One attempt at accounting for the fermion mass spectrum makes use of broken family symmetries. The most common approach is to take an Abelian $`U(1)`$ as the horizontal symmetry, but nonabelian groups and discrete groups (and combinations of these) have been tried with varying degrees of success. Since an Abelian symmetry alone cannot generate a nearly degenerate set of neutrinos, we assume here that the $`\mathrm{\Delta }m_{23}^2`$ and $`\mathrm{\Delta }m_{1x}^2`$ are indicating that the neutrino masses are arranged in a hierarchical pattern. This hierarchical structure of the fermion masses suggests that it might be produced by an expansion in a small parameter, and one widely adopted strategy is to have this parameter arise from a family symmetry spontaneously broken at a scale $`\mathrm{\Lambda }_L`$. In this paper we consider the possibility that the horizontal symmetry is an Abelian anomalous gauge symmetry, where the anomaly is cancelled by the Green-Schwarz mechanism. In this scenario there is field $`\mathrm{\Phi }`$ that is a singlet under the Standard Model gauge symmetries. The contribution of the Fayet-Iliopoulos term to the D-term cancels against the contribution from the vev, $`<\mathrm{\Phi }>`$. The ratio of this vev to the Planck scale naturally provides a small parameter $`\lambda =<S>/m_{\mathrm{Pl}}`$. The field $`\mathrm{\Phi }`$ is charged under the horizontal symmetry, and without loss of generality it charge can be taken to be $`1`$. In this approach the hierarchy is generated by nonrenormalizable terms that transform as singlets under the horizontal symmetry and therefore produce contributions to the mass matrices that contain integer powers of the small parameter $`\lambda `$. In this scenario it is often the case that only the $`(3,3)`$ entry of one or more mass matrices receives a contribution from a renormalizable coupling to the Higgs boson. So by assigning quantum numbers for the horizontal symmetry for each Standard Model field, one can generate a hierarchy in the Yukawa matrices as powers of the small parameter $`\lambda `$.
The heavy Majorana neutrino mass matrix $`M_N`$ is obtained by inverting the type-I see-saw formula
$$m_\nu =m_D(M_N)^1m_D^T,$$
(3)
where $`m_D`$ is the neutrino Dirac mass matrix. CP-asymmetries in neutrino decays arise from the interference between the tree level and one-loop level decay channels. In the mass basis where the right-handed Majorana mass matrix is diagonal the asymmetry in heavy neutrino $`N_i`$ decays
$`ϵ_i`$ $`=`$ $`{\displaystyle \frac{\mathrm{\Gamma }(N_i\mathrm{}H_2)\mathrm{\Gamma }(N_i\mathrm{}^cH_2^c)}{\mathrm{\Gamma }(N_i\mathrm{}H_2)+\mathrm{\Gamma }(N_i\mathrm{}^cH_2^c)}},`$ (4)
is given by
$`ϵ_i`$ $`=`$ $`{\displaystyle \frac{3}{16\pi v_2^2}}{\displaystyle \frac{1}{(m_D^{}m_D)_{ii}}}{\displaystyle \underset{ni}{}}\mathrm{Im}\left[(m_D^{}m_D)_{ni}^2\right]{\displaystyle \frac{M_i}{M_n}}.`$ (5)
The masses $`M_i`$ are the three eigenvalues of the heavy Majorana mass matrix and $`v_2`$ is the vev of the Higgs giving Dirac masses to the neutrinos and up-type quarks. $`M_1`$ is the mass of the lightest of the three heavy Majorana neutrinos, and Eq. (5) is an approximate formula valid for $`M_n>>M_i`$. The most common scenario that occurs is that the lightest Majorana neutrino $`N_1`$ has a mass such that $`M_1<<M_2,M_3`$, and the lepton asymmetry producedIn some cases inverted hierarchies in the Majorana mass matrix can occur where $`M_2<M_1`$, which can produce a larger asymmetry if $`ϵ_2>ϵ_1`$. We do not consider this possibility in this paper. comes almost entirely from the decays of $`N_1`$. So the CP-asymmetry of most interest to the discussion of lepton asymmetry generation is $`ϵ_1`$.
The other parameter of most interest is the mass parameter
$`\stackrel{~}{m}_1`$ $`=`$ $`{\displaystyle \frac{(m_D^{}m_D)_{11}}{M_1}},`$ (6)
which controls the decay width of the lightest right-handed neutrino $`N_1`$ since
$`\mathrm{\Gamma }_{N_i}=\mathrm{\Gamma }(N_i\mathrm{}H_2)+\mathrm{\Gamma }(N_i\mathrm{}^cH_2^c)={\displaystyle \frac{1}{8\pi }}(m_D^{}m_D)_{ii}{\displaystyle \frac{M_i}{v_2^2}},`$ (7)
and $`\stackrel{~}{m}_1`$ also largely controls the amount of dilution caused by the lepton number violating scattering. The parameter $`\stackrel{~}{m}_1`$ can therefore be called the dilution mass. These two constraints bound the possible values of $`\stackrel{~}{m}_1`$ such that a sufficient asymmetry is produced to agree with Eq. (1). The generated lepton asymmetry is given by
$`Y_L={\displaystyle \frac{n_Ln_{\overline{L}}}{s}}=\kappa {\displaystyle \frac{ϵ_1}{g^{}}},`$ (8)
where $`g^{}`$ is the number of light (effective) degrees of freedom in the theory ($`106\frac{3}{4}`$ in the Standard Model or $`228\frac{3}{4}`$ in the MSSM), and $`\kappa `$ is a dilution factor that can be reliably calculated by solving the full Boltzmann equations. The dilution depends critically on the parameter $`\stackrel{~}{m}_1`$ because it governs the size of the most important Yukawa coupling in the $`\mathrm{\Delta }L=2`$ scattering processes, as shown in Ref. .
### A Leptogenesis with a $`U(1)`$ family symmetry
Assume now that the lepton fields have charges under a $`U(1)`$ family symmetry
$`\begin{array}{ccccccccc}e_{R1}^c& e_{R2}^c& e_{R3}^c& \mathrm{}_{L1}& \mathrm{}_{L2}& \mathrm{}_{L3}& \nu _{R1}^c& \nu _{R2}^c& \nu _{R3}^c\\ E_1& E_2& E_3& L_1& L_2& L_3& 𝒩_1& 𝒩_2& 𝒩_3\end{array}.`$ (11)
We assume here that the quantum numbers satisfy the hierarchies $`E_1E_2E_30`$, $`L_1L_2L_30`$, and $`𝒩_1𝒩_2𝒩_30`$ (This last condition will guarantee that no light neutrino masses are enhanced because a right-handed neutrino mass is suppressed).
Given lepton doublet charges $`L_i`$ and right-handed neutrino charges $`𝒩_i`$ one has the following pattern for the neutrino Dirac mass matrix
$`m_D\left(\begin{array}{ccc}\lambda ^{L_1+𝒩_1}& \lambda ^{L_1+𝒩_2}& \lambda ^{L_1+𝒩_3}\\ \lambda ^{L_2+𝒩_1}& \lambda ^{L_2+𝒩_2}& \lambda ^{L_2+𝒩_3}\\ \lambda ^{L_3+𝒩_1}& \lambda ^{L_3+𝒩_2}& \lambda ^{L_3+𝒩_3}\end{array}\right)v_2,`$ (12)
and the following pattern for the Majorana mass matrix
$`M_N\left(\begin{array}{ccc}\lambda ^{2𝒩_1}& \lambda ^{𝒩_1+𝒩_2}& \lambda ^{𝒩_1+𝒩_3}\\ \lambda ^{𝒩_1+𝒩_2}& \lambda ^{2𝒩_2}& \lambda ^{𝒩_2+𝒩_3}\\ \lambda ^{𝒩_1+𝒩_3}& \lambda ^{𝒩_2+𝒩_3}& \lambda ^{2𝒩_3}\end{array}\right)\mathrm{\Lambda }_L.`$ (13)
Then one obtains the following form for the light neutrino mass matrix via the see-saw formula Eq. (3)
$`m_\nu \left(\begin{array}{ccc}\lambda ^{2L_1}& \lambda ^{L_1+L_2}& \lambda ^{L_1+L_3}\\ \lambda ^{L_1+L_2}& \lambda ^{2L_2}& \lambda ^{L_2+L_3}\\ \lambda ^{L_1+L_3}& \lambda ^{L_2+L_3}& \lambda ^{2L_3}\end{array}\right){\displaystyle \frac{v_2^2}{\mathrm{\Lambda }_L}},`$ (14)
Clearly if $`L_2=L_3`$ one can obtain $`𝒪(1)`$ mixing in the 2-3 sector, or if $`L_2=L_3`$ one has a pseudo-Dirac neutrino and maximal mixing in the 2-3 sector.It is also possible that one has only an approximate equality $`L_2\pm L_3`$ in which case the mixing is not truly order one, but could be sufficiently large to be phenomenologically relevant without assuming accidental cancellations.
The Super-Kamiokande collaboration measurements of the atmospheric neutrino flux indicates large mixing $`\mathrm{sin}^22\theta _{23}1`$ and a mass-squared difference $`\mathrm{\Delta }m_{23}^22\times 10^3\mathrm{eV}^2`$. The SMA solution to the solar neutrino oscillations requires $`\mathrm{\Delta }m_{1x}^25\times 10^6\mathrm{eV}^2`$. If one assumes that the light neutrino masses are hierarchical, then one can identify $`m_{\nu _\tau }^22\times 10^3\mathrm{eV}^2`$ and $`m_{\nu _\mu }^25\times 10^6\mathrm{eV}^2`$; it is then difficult to naturally explain the separation of masses simultaneously with the large mixing angle. The suppression of one of the neutrino masses can always result from a fine-tuning of the parameters.
The dilution parameter $`\stackrel{~}{m}_1`$ defined in Eq. (6) can be described in terms of the $`U(1)`$ quantum numbers by constructing the Yukawa coupling squared matrix
$`m_D^{}m_D\left(\begin{array}{ccc}\lambda ^{2𝒩_1}& \lambda ^{𝒩_1+𝒩_2}& \lambda ^{𝒩_1+𝒩_3}\\ \lambda ^{𝒩_1+𝒩_2}& \lambda ^{2𝒩_2}& \lambda ^{𝒩_2+𝒩_3}\\ \lambda ^{𝒩_1+𝒩_3}& \lambda ^{𝒩_2+𝒩_3}& \lambda ^{2𝒩_3}\end{array}\right)\lambda ^{2L_3}v_2^2,`$ (15)
so that
$`\stackrel{~}{m}_1{\displaystyle \frac{\lambda ^{2(L_3+𝒩_1)}v_2^2}{M_1}}{\displaystyle \frac{\lambda ^{2(L_3+𝒩_1)}}{\lambda ^{2𝒩_1}}}{\displaystyle \frac{v_2^2}{\mathrm{\Lambda }_L}}\lambda ^{2L_3}{\displaystyle \frac{v_2^2}{\mathrm{\Lambda }_L}},`$ (16)
When $`L_2=L_3`$ then this parameter is the same order of magnitude as the neutrino masses $`m_{\nu _\mu }`$ and $`m_{\nu _\tau }`$<sup>§</sup><sup>§</sup>§We use the notation $`\nu _\mu `$ and $`\nu _\tau `$ for the eigenstates even though they have large mixing., and it is consistent to take the parameter $`\stackrel{~}{m}_1(m_{\nu _\mu }m_{\nu _\tau })^{1/2}`$. Typically one needs a fine-tuning to produce the hierarchy $`m_{\nu _\mu }<<m_{\nu _\tau }`$. The CP-violating parameter is given by
$`ϵ_1{\displaystyle \frac{3}{16\pi }}\lambda ^{2(L_3+𝒩_1)}.`$ (17)
Comparing to Eq. (16), one sees that $`ϵ_1`$ can be simply expressed in terms of the dilution mass $`\stackrel{~}{m}_1`$, the mass $`M_1`$ of the lightest Majorana neutrino, and the electroweak scale vev $`v_2`$. Since $`\stackrel{~}{m}_1`$ is tied to the light neutrino masses, a connection between these quantities is established at the order-of-magnitude level.
The problem with the situation outlined is well-known: it seems to predict that $`m_{\nu _\mu }`$ is the naturally of the same order as $`m_{\nu _\tau }`$, and one would need to have an accidental cancellation to get the hierarchy $`m_{\nu _\mu }<<m_{\nu _\tau }`$. The charged lepton matrix is given by
$`m_\mathrm{}^\pm \left(\begin{array}{ccc}\lambda ^{L_1+E_1}& \lambda ^{L_1+E_2}& \lambda ^{L_1+E_3}\\ \lambda ^{L_2+E_1}& \lambda ^{L_2+E_2}& \lambda ^{L_2+E_3}\\ \lambda ^{L_3+E_1}& \lambda ^{L_3+E_2}& \lambda ^{L_3+E_3}\end{array}\right)v_1,`$ (18)
where $`v_1`$ is the vev of the other Higgs doublet. So the relevant rotation to get to the basis where the charged lepton mass is diagonal is also order one when $`L_2=L_3`$. Hence the large mixing in the 2-3 sector is connected in this approach to near degeneracy of two of the light neutrino masses.
### B Leptogenesis with a $`Z_2\times U(1)`$ family symmetry
Ref. proposed that a discrete Abelian family symmetry could be employed to enhance a mass or mixing angle above what would be otherwise obtained if the family symmetry was the usual continuous $`U(1)`$ symmetry, and this idea was pursued further in a specific model. If the family symmetry is $`Z_m`$ then entries in the mass matrices can be enhanced by factors of the small parameter $`\lambda `$ to the $`m`$th power. With this approach the discrete $`Z_m`$ symmetry can result in the enhancement of entries in the light neutrino mass matrix. A consequence for leptogenesis is that this will also change the relationship between the light neutrino masses and the dilution parameter $`\stackrel{~}{m}_1`$ by a factor of $`\lambda ^m`$. For example take the following $`Z_2\times U(1)`$ charges for the lepton fieldsThe second group factor does not need to be continuous, but could be replaced by a second $`Z_n`$ with $`n`$ sufficiently large.
$`\begin{array}{ccccccccc}e_{R1}^c& e_{R2}^c& e_{R3}^c& \mathrm{}_{L1}& \mathrm{}_{L2}& \mathrm{}_{L3}& \nu _{R1}^c& \nu _{R2}^c& \nu _{R3}^c\\ (0,E_1)& (0,E_2)& (0,E_3)& (0,L_1)& (0,L_2)& (1,L_31)& (0,𝒩_1)& (0,𝒩_2)& (1,𝒩_31)\end{array}`$ (21)
and later we will take $`L_2=L_3`$. Assume the symmetry breaking is characterized by the single expansion parameter $`\lambda `$. The formulas given above for the heavy neutrino mass matrix, $`M_N`$, the neutrino Dirac mass matrix, $`m_D`$, and the resulting light neutrino mass matrix, $`m_\nu `$ are modified. With the above assignments one finds that
$`M_N\left(\begin{array}{ccc}\lambda ^{2𝒩_1}& \lambda ^{𝒩_1+𝒩_2}& \lambda ^{𝒩_1+𝒩_3}\\ \lambda ^{𝒩_1+𝒩_2}& \lambda ^{2𝒩_2}& \lambda ^{𝒩_2+𝒩_3}\\ \lambda ^{𝒩_1+𝒩_3}& \lambda ^{𝒩_2+𝒩_3}& \lambda ^{2𝒩_32}\end{array}\right)\mathrm{\Lambda }_L,`$ (22)
so that
$`(M_N)^1\left(\begin{array}{ccc}\lambda ^{2𝒩_1}& \lambda ^{𝒩_1𝒩_2}& \lambda ^{𝒩_1𝒩_3+2}\\ \lambda ^{𝒩_1𝒩_2}& \lambda ^{2𝒩_2}& \lambda ^{𝒩_2𝒩_3+2}\\ \lambda ^{𝒩_1𝒩_3+2}& \lambda ^{𝒩_2𝒩_3+2}& \lambda ^{2𝒩_3+2}\end{array}\right)\mathrm{\Lambda }_L^1.`$ (23)
Furthermore one has
$`m_D\left(\begin{array}{ccc}\lambda ^{L_1+𝒩_1}& \lambda ^{L_1+𝒩_2}& \lambda ^{L_1+𝒩_3}\\ \lambda ^{L_2+𝒩_1}& \lambda ^{L_2+𝒩_2}& \lambda ^{L_2+𝒩_3}\\ \lambda ^{L_3+𝒩_1}& \lambda ^{L_3+𝒩_2}& \lambda ^{L_3+𝒩_32}\end{array}\right)v_2,`$ (24)
Then it is straightforward to show that the light neutrino mass matrix $`m_\nu `$ is modified so that only one component is enhanced,
$`m_\nu \left(\begin{array}{ccc}\lambda ^{2L_1}& \lambda ^{L_1+L_2}& \lambda ^{L_1+L_3}\\ \lambda ^{L_1+L_2}& \lambda ^{2L_2}& \lambda ^{L_2+L_3}\\ \lambda ^{L_1+L_3}& \lambda ^{L_2+L_3}& \lambda ^{2L_32}\end{array}\right){\displaystyle \frac{v_2^2}{\mathrm{\Lambda }_L}}.`$ (25)
Also one finds that
$`m_D^{}m_D\left(\begin{array}{ccc}\lambda ^{2𝒩_1}& \lambda ^{𝒩_1+𝒩_2}& \lambda ^{𝒩_1+𝒩_32}\\ \lambda ^{𝒩_1+𝒩_2}& \lambda ^{2𝒩_2}& \lambda ^{𝒩_2+𝒩_32}\\ \lambda ^{𝒩_1+𝒩_32}& \lambda ^{𝒩_2+𝒩_32}& \lambda ^{2𝒩_34}\end{array}\right)\lambda ^{2L_3}v_2^2.`$ (26)
So then using our previous definitions, one sees that $`\stackrel{~}{m}_1\lambda ^{2L_3}v_2^2/\mathrm{\Lambda }_Lm_{\nu _\mu }`$, whereas $`m_{\nu _\tau }\lambda ^{2L_32}v_2^2/\mathrm{\Lambda }_L`$. More specifically when the atmospheric neutrino constraint $`\mathrm{\Delta }m_{23}^22\times 10^3`$ eV<sup>2</sup> is interpreted as the mass-squared of the heaviest light neutrino $`m_{\nu _\tau }`$, then in the case of a horizontal $`U(1)`$ symmetry, one has that $`\stackrel{~}{m}_1^22\times 10^3`$ eV<sup>2</sup>. In the case of the discrete $`Z_2`$ symmetry, the Yukawa coupling related to $`\stackrel{~}{m}_1^2`$ via Eq. (6) can be reduced by a factor $`\lambda ^2`$ thereby substantially reducing the amount of dilution from the $`\mathrm{\Delta }L=2`$ processes and reducing the decay rate of $`N_1`$. More generally, a $`Z_m`$ symmetry can arrange for a suppression of $`\stackrel{~}{m}_1^2`$ by a factor $`\lambda ^m`$. The CP-violation asymmetry $`ϵ_1`$ is easily obtained from Eqs. (5) and (26),
$`ϵ_1{\displaystyle \frac{3}{16\pi }}\lambda ^{2(L_3+𝒩_1)2}.`$ (27)
So the ratio of $`m_{\nu _\tau }`$ to $`ϵ_1`$ is unaffected by the discrete symmetry. Since $`m_{\nu _\tau }`$ is being fixed by the experimental data for $`\mathrm{\Delta }m_{23}^2`$, the expected value of $`ϵ_1`$ is expected to be unchanged for the same quantum number $`𝒩_1`$ (and thus the same heavy neutrino mass $`M_1`$).
A phenomenologically viable solution has been presented in Ref. : taking $`L_1=3`$, $`L_2=L_3=1`$, $`E_1=5`$, $`E_2=4`$, and $`E_3=2`$ yields mass matrices of the form
$`m_\nu \left(\begin{array}{ccc}\lambda ^6& \lambda ^4& \lambda ^4\\ \lambda ^4& \lambda ^2& \lambda ^2\\ \lambda ^4& \lambda ^2& 1\end{array}\right){\displaystyle \frac{v_2^2}{\mathrm{\Lambda }_L}},m_\mathrm{}^\pm \left(\begin{array}{ccc}\lambda ^8& \lambda ^7& \lambda ^5\\ \lambda ^6& \lambda ^5& \lambda ^3\\ \lambda ^6& \lambda ^5& \lambda ^3\end{array}\right)v_1,`$ (28)
which give the correct order of magnitude for the SMA solution (after rotating to the charged lepton mass basis)
$`{\displaystyle \frac{\mathrm{\Delta }m_{1x}^2}{\mathrm{\Delta }m_{23}^2}}\lambda ^4,\mathrm{sin}\theta _{12}\lambda ^2,\mathrm{sin}\theta _{23}1,\mathrm{sin}\theta _{13}\lambda ^2,`$ (29)
when the small parameter is identified as the Cabibbo angle, i.e. $`\lambda 0.2`$.
For a sufficient amount of leptogenesis to occur two conditions must be satisfied: (1) $`|ϵ_1|10^6`$ and (2) $`10^5\stackrel{~}{m}_110^2`$. The first condition guarantees that there is sufficient CP-violation in the heavy $`N_1`$ neutrino decay (c.f. Eq. (8), while the second condition guarantees that the dilution is not too large ($`\kappa 10^2`$) and that a sufficient number of heavy neutrino are produced out-of-equilibrium. The condition on $`\stackrel{~}{m}_1`$ is equivalent to a condition on the relevant Yukawa coupling $`(h_\nu ^{}h_\nu )_{11}=m_D^{}m_D/v_2^2`$ that governs the rates of these two processes.
The mass $`\stackrel{~}{m}_1`$ arising in the case of the $`U(1)`$ symmetry is identified with $`m_{\nu _\tau }`$, and thus is near the top of the required range. The resulting lepton asymmetry is smaller than it would be if the dilution mass $`\stackrel{~}{m}_1`$ could be reduced. Lowering the mass parameter $`\stackrel{~}{m}_1`$ by using the horizontal $`Z_2\times U(1)`$ rather than the $`U(1)`$ symmetry has the following effects on the Boltzmann evolution: 1) The lightest Majorana neutrino $`N_1`$ decays more slowly, and stays out of thermal equilibrium for a longer period of time. 2) The dilution of the generated lepton asymmetry is reduced since the relevant Yukawa coupling controlling the strength of the interactions is reduced. These two factors can result in a remnant lepton asymmetry that is enhanced over that which is obtained in the case of the $`U(1)`$ symmetry.
## IV Numerical Simulation
The lepton asymmetry that results can be obtained by integrating the full set of Boltzmann equations. These differential equations, incorporating the Majorana neutrino decay rates as well as all lepton number violating scattering processes in the MSSM, has been given in Ref. . The above discussion gives an overall order of magnitude estimate for the CP-violation parameter $`ϵ_1`$ and the dilution parameter $`\stackrel{~}{m}_1`$. The parameter $`ϵ_1`$ depends on a CP-phase (see Eq. (5)); this phase is not specified by the family symmetry and we assume that it is order one.
A concrete example of how the discrete symmetry can change the produced lepton asymmetry is shown in Figs. 1 and 2. First consider the case where the family symmetry is $`U(1)`$: the values of the parameters are $`(m_D^{}m_D)_{11}=0.2`$ GeV<sup>2</sup> and $`ϵ_1=4.0\times 10^6`$ for the case of the continuous $`U(1)`$ symmetry. The values for the CP-violation parameter $`ϵ_1`$ in Eq. (17) and $`(m_D^{}m_D)_{11}\lambda ^{2(L_3+𝒩_1)}v_2^2`$ are consistentThese relationships involving $`ϵ_1`$ and $`(m_D^{}m_D)_{11}`$ only determine the leading order contribution in the small parameter $`\lambda `$ and there is an undetermined coefficient of order one. We choose the values of the parameters here as an example to illustrate that an enhancement occurs. The exact values the unknown order one coefficients take are not important; the enhancement of the lepton asymmetry is generic. with taking $`L_3=0`$ and $`𝒩_1=3`$. Taking the scale $`\mathrm{\Lambda }_L`$ to be near a supersymmetric grand unified scale $`3\times 10^{15}`$ GeV, one finds the mass of the lightest heavy neutrino $`N_1`$ that is decaying asymmetrically to be $`M_1=2\times 10^{11}`$ GeV. This then yields a dilution mass of $`\stackrel{~}{m}_1=5\times 10^3`$ eV, which is the same order of magnitude as the mass splitting $`\mathrm{\Delta }m_{23}^22.2\times 10^3`$ eV<sup>2</sup> as expected from Eqs. (14) and (16).
Fig. 1. The neutrino density $`Y_{N_1}`$ as a function of the temperature $`T`$ of the universe for the case of a horizontal $`U(1)`$ symmetry (solid), and for the $`Z_2\times U(1)`$ symmetry (dashed). The dotted curve is the equilibrium value $`Y_{N_1}^{\mathrm{eq}}`$ of the neutrino density. The discrete symmetry results in a smaller decay rate for $`N_1`$ and it requires a longer time before it comes into thermal equilibrium.
When the $`U(1)`$ symmetry is replaced with $`Z_2`$ the dilution mass is suppressed by an additional factor of $`\lambda ^2`$ so that $`\stackrel{~}{m}_1=2.2\times 10^4`$ eV (for the case of the $`Z_2\times U(1)`$ symmetry, we take $`L_3=1`$ and keep $`𝒩_1=3`$ so that $`m_{\nu _\tau }`$ and $`ϵ_1`$ remain the same, but $`\stackrel{~}{m}_1`$ is reduced by a factor $`\lambda ^2`$ relative to the $`U(1)`$ symmetry case.). Figure 1 shows the neutrino density $`Y_{N_1}`$ of the lightest Majorana neutrino that is decaying to produce to produce the lepton asymmetry shown in Fig. 2. The densities are plotted against the dimensionless ratio $`z=M_1/T`$ where $`T`$ is the temperature of the universe, so the universe evolves toward the present day as $`z`$ becomes larger. For the quantitative results shown in the figures, the unknown CP phase (see Eq. (5) is chosen so as to maximize the lepton asymmetry; another phase would just scale the curves in Fig. 2 by some overall factor. The $`Z_2`$ symmetry results in $`N_1`$ decaying more slowly, and thus $`N_1`$ can remain out-of-equilibrium for a greater period of time in the early universe. The lepton asymmetry produced in each case begins with one sign, then goes through zero, and finally asymptotes to a final value. For the particular example shown in Figs. 1 and 2, the lepton asymmetry is enhanced by about a factor between seven and eight when the continuous family symmetry is replaced by a discrete one. The enhancement (or suppression) that can result in general (from suppressing $`\stackrel{~}{m}_1`$) is a sensitive function of the values of dilution parameter $`\stackrel{~}{m}_1`$ and the mass $`M_1`$, as shown in Ref. .
Fig. 2. The lepton asymmetry in fermions $`Y_{L_f}`$ and in scalars $`Y_{L_s}`$ produced for a horizontal $`U(1)`$ symmetry (solid), and for the $`Z_2\times U(1)`$ symmetry (dashed). The generated asymmetry in the latter case is smaller at earlier times (larger temperatures) since the decay rate of the lightest Majorana neutrino $`N_1`$ is suppressed, but ultimately a larger asymmetry is produced as the neutrino density remains out of thermal equilibrium for a longer period. The equality $`Y_{L_f}=Y_{L_s}`$ is maintained by MSSM processes $`f+f\stackrel{~}{f}+\stackrel{~}{f}`$, e.g. neutralino exchange.
## V Conclusion
The predominance of matter over antimatter in the universe can be produced CP-violation in the decays of heavy neutrinos followed by sphaleron processes that recycle the resulting lepton asymmetry into a baryon asymmetry. We have shown that if the fermion mass matrices are determined by imposing an Abelian family symmetry then there are simple order-of-magnitude estimates of the CP-violation parameter $`ϵ_1`$ and the dilution mass $`\stackrel{~}{m}_1`$ that are critically important for determining the size of the lepton asymmetry produced in the early universe. In the most straightforward case these parameters are given by universal formulas in terms of the $`U(1)`$ quantum numbers ($`ϵ_1(3/16\pi )\lambda ^{2(𝒩_1+L_3)}`$ and $`\stackrel{~}{m}_1\lambda ^{2L_3}v_2^2/\mathrm{\Lambda }_L`$), and $`\stackrel{~}{m}_1`$ can be simply related to the experimentally determined light neutrino masses.
A $`Z_2`$ horizontal symmetry can be employed to reconcile (a) the large mixing that must be present to explain the atmospheric neutrino data with (b) a hierarchy in neutrino masses. We have shown here that employing this same $`Z_2`$ horizontal symmetry can enhance the lepton asymmetry that results from heavy right-handed neutrino decays. This results in an enhanced baryon asymmetry in the universe. The change in the generated lepton asymmetry comes about because when a Yukawa coupling $`(h_\nu ^{}h_\nu )_{11}`$ can be suppressed or enhanced compared to the usual expectation when the horizontal symmetry is $`U(1)`$. This affects the decay rate of the lightest Majorana neutrino $`N_1`$, as well as the amount of subsequent dilution of the asymmetry by lepton number violating scattering. A particular example where the generated asymmetry was explicitly calculated using the supersymmetric Boltzmann equations was given, and an enhancement of the lepton asymmetry (and hence ultimately the baryon asymmetry) by a factor seven was derived quantitatively.
## Acknowledgments
This work was supported in part by the U.S. Department of Energy under Grant No. No. DE-FG02-91ER40661.
|
no-problem/9906/cs9906017.html
|
ar5iv
|
text
|
# Generalization of automatic sequences for numeration systems on a regular language
## 1 Introduction
In , P. Lecomte and I have defined a numeration system as being a triple $`S=(L,\mathrm{\Sigma },<)`$ where $`L`$ is an infinite regular language over a totally ordered alphabet $`(\mathrm{\Sigma },<)`$. The lexicographic ordering of $`L`$ gives a one-to-one correspondence $`r_S`$ between the set of the natural numbers $`\mathrm{I}\mathrm{N}`$ and the language $`L`$.
For a given subset $`X`$ of $`\mathrm{I}\mathrm{N}`$, a question arise naturally. Is it possible to find a numeration system $`S`$ such that $`\mathrm{r}_S(X)`$ is recognizable by finite automata ? (In this case, $`X`$ is said to be $`S`$-recognizable.) For example, the set $`\{n^2:n\mathrm{I}\mathrm{N}\}`$ is $`S`$-recognizable for some $`S`$ and the arithmetic progressions $`p+q\mathrm{I}\mathrm{N}`$ are $`S`$-recognizable for any $`S`$. An interesting question is thus the following: is there a system $`S`$ such that the set of primes is $`S`$-recognizable ?
To answer this question I show that a subset of $`\mathrm{I}\mathrm{N}`$ is $`S`$-recognizable if and only if its characteristic sequence can be generated by an ‘automatic’ method. The term automatic refers, as we shall see further, to a generalization of the $`k`$-automatic sequences for numeration systems on a regular language.
The $`k`$-automatic sequences are well-known and have been extensively studied since the 70’s . The construction of this kind of sequences is based on the representation of the integers in the base $`k`$. For a given integer $`n`$, one represents this number in base $`k`$ using the greedy algorithm and obtains a word $`[n]_k`$ over the alphabet $`\{0,\mathrm{},k1\}`$. Next one gives $`[n]_k`$ to a deterministic finite automaton with output and obtains the $`n^{th}`$ term of a sequence which is said to be a $`k`$-automatic sequence.
These sequences have been already generalized in different ways . In particular, a method used by J. Shallit to generalize the $`k`$-automatic sequences is to consider some kind of linear numeration system instead of the standard numeration system with integer base $`k`$ . Two properties of the systems encountered in are precisely that the set of all the representations is regular and that the lexicographic ordering is respected.
Here, instead of giving $`[n]_k`$ to a deterministic finite automaton with output, we feed it with $`\mathrm{r}_S(n)`$ to obtain an output which is the $`n^{th}`$ term of an $`S`$-automatic sequence for a numeration system $`S`$. Having thus introduced the concept of $`S`$-automatic sequences, we can follow two paths. Learn their intrinsic properties but also use them as a tool to check if a subset of $`\mathrm{I}\mathrm{N}`$ is $`S`$-recognizable.
Our article has the following articulation. In the first section, we recall some definitions and we introduce a teaching example which could be very instructive for the reader not familiar with automatic sequences. In the second section, we adapt the classical results concerning the fiber and the kernel of an automatic sequence.
Initially, A. Cobham showed the equivalence between the $`k`$-automatic sequences and the sequences obtained by iterating a uniform morphism (also called uniform tag system ). In the third section, we show that an $`S`$-automatic sequence is always generated by a substitution (i.e., an iterated non-uniform morphism followed by one application of another morphism). From this, we deduce that the number of distinct factors of length $`l`$ in an $`S`$-automatic sequence is in $`O(l^2)`$. We also show how to construct $`S`$-automatic sequences with at least the same complexity that infinite words obtained by iterated morphisms.
In the last section, we will be able to show that for any numeration system $`S`$, the set of primes is never $`S`$-recognizable. We use the fact that to be $`S`$-recognizable, the characteristic sequence of the set must be generated by a substitution. Hence we use some results of C. Mauduit about the density of the infinite words obtained by substitution .
## 2 Basic definitions and notations
In this paper, capital greek letters represent finite alphabet. We denote by $`\mathrm{\Sigma }^{}`$ the set of the words over $`\mathrm{\Sigma }`$ ($`\epsilon `$ is the empty word) and by $`\mathrm{\Sigma }^\omega `$ the set of the infinite words over $`\mathrm{\Sigma }`$. If $`K`$ is a set then $`\mathrm{\#}K`$ denotes the cardinality of $`K`$ and if $`w`$ is a string then $`|w|`$ denotes the length of $`w`$. For $`1i|w|`$, $`w_i`$ is the $`i^{th}`$ letter of $`w`$. The same notation holds for infinite words, in this case $`i\mathrm{I}\mathrm{N}\{0\}`$.
First, recall some definitions about the numeration systems we are dealing with. For more about these systems see .
###### Definition 1
A numeration system $`S`$ is a triple $`(L,\mathrm{\Sigma },<)`$ where $`L`$ is an infinite regular language over the totally ordered alphabet $`(\mathrm{\Sigma },<)`$.
For each $`n\mathrm{I}\mathrm{N}`$, $`\mathrm{r}_S(n)`$ denotes the $`(n+1)^{th}`$ word of $`L`$ with respect to the lexicographic ordering and is called the $`S`$-representation of $`n`$.
Remark that the map $`\mathrm{r}_S:\mathrm{I}\mathrm{N}L`$ is an increasing bijection. For $`wL`$, we set $`\mathrm{val}_S(w)=\mathrm{r}_S^1(w)`$. We call $`\mathrm{val}_S(w)`$ the numerical value of $`w`$.
Examples of such systems are the numeration systems defined by a recurrence relation whose characteristic polynomial is the minimum polynomial of a Pisot number . (Indeed, with this hypothesis, the set of representations of the integers is a regular language.) The standard numeration systems with integer base and also the Fibonacci system belong to this class.
###### Definition 2
Let $`S`$ be a numeration system. A subset $`X`$ of $`\mathrm{I}\mathrm{N}`$ is $`S`$-recognizable if $`\mathrm{r}_S(X)`$ is recognizable by finite automata.
Let us introduce the concept of $`S`$-automatic sequence which naturally generalizes the $`k`$-automatic sequences based on the representation of the integers in base $`k`$. For more about $`k`$-automatic sequences see for instance .
###### Definition 3
A deterministic finite automaton with output (DFAO) $`M`$ is a $`6`$-uple $`(K,s,\mathrm{\Sigma },\delta ,\mathrm{\Delta },\tau )`$ where $`K`$ is the finite set of the states, $`s`$ is the start state, $`\mathrm{\Sigma }`$ is the input alphabet, $`\delta :K\times \mathrm{\Sigma }K`$ is the transition function, $`\mathrm{\Delta }`$ is the output alphabet and $`\tau :K\mathrm{\Delta }`$ is the output function.
###### Definition 4
Let $`S=(L,\mathrm{\Sigma },<)`$ be a numeration system. A sequence $`u\mathrm{\Delta }^\omega `$ is $`S`$-automatic if there exists a DFAO $`M=(K,s,\mathrm{\Sigma },\delta ,\mathrm{\Delta },\tau )`$ such that for all $`n\mathrm{I}\mathrm{N}`$,
$$u_{n+1}=\tau (\delta (s,\mathrm{r}_S(n))).$$
If the context is clear, we write $`\tau (w)`$ in place of $`\tau (\delta (s,w))`$.
###### Remark 1
A subset $`X\mathrm{I}\mathrm{N}`$ is $`S`$-recognizable if and only if its characteristic sequence $`\chi _X\{0,1\}^\omega `$ is $`S`$-automatic.
In the following we will often encounter two more ‘classical’ ways of obtaining infinite sequences.
###### Definition 5
Let $`\phi :\mathrm{\Sigma }\mathrm{\Sigma }^{}`$ be a morphism of monoid such that for some $`\sigma \mathrm{\Sigma }`$, $`\phi (\sigma )\sigma \mathrm{\Sigma }^{}`$. The word $`u_\phi =\phi ^\omega (\sigma )`$ is a fixed point of $`\phi `$ and we say that $`u_\phi `$ is generated by an iterated morphism.
A morphism is uniform if $`|\phi (\sigma _1)|=\mathrm{}=|\phi (\sigma _n)|`$, $`\mathrm{\Sigma }=\{\sigma _1,\mathrm{},\sigma _n\}`$.
###### Definition 6
A substitution $`T`$ is a triple $`(\phi ,h,c)`$ such that $`\phi :\mathrm{\Sigma }\mathrm{\Sigma }^{}`$ and $`h:\mathrm{\Sigma }\mathrm{\Delta }^{}`$ are morphisms of monoids. Moreover $`c\mathrm{\Sigma }`$, $`\phi (c)c\mathrm{\Sigma }^{}`$ and for any $`\sigma \mathrm{\Sigma }`$, $`h(\sigma )=\epsilon `$ or $`h(\sigma )\mathrm{\Delta }`$ ($`h`$ is said to be a weak coding). We said that the word $`u_T=h(\phi ^\omega (c))`$ over $`\mathrm{\Delta }`$ is generated by the substitution $`T`$.
If $`h(\sigma )=\epsilon `$ for some $`\sigma `$ then $`h`$ is said to be erasing otherwise $`h`$ is said to be non-erasing.
### 2.1 A teaching example
We consider the numeration system $`S=(a^{}b^{},\{a,b\},a<b)`$, the alphabets $`\mathrm{\Sigma }=\{a,b\}`$, $`\mathrm{\Delta }=\{0,1,2,3\}`$ and the following DFAO
As usual the start state is indicated by an unlabeled arrow. The first words of $`a^{}b^{}`$ are
$$\epsilon ,a,b,aa,ab,bb,aaa,aab,abb,bbb,\mathrm{}$$
and thus feeding the automaton with these words we obtain the first terms of the sequence $`u\mathrm{\Delta }^\omega `$,
$$u=01023031200231010123023031203120231002310123010123\mathrm{}.$$
###### Remark 2
The sequence $`u`$ is not ultimately periodic. One can observe that the distance between two occurrences of the block ‘$`00`$’ is not bounded. Indeed,
$$\tau (w)=0r,s\mathrm{I}\mathrm{N}:w=a^{4r}b^s$$
(1)
thus a block ‘$`00`$’ comes from two consecutive words $`b^{4r1}`$ and $`a^{4r}`$, $`r1`$ and the number of words of length $`n`$ in $`a^{}b^{}`$ is $`n+1`$, $`n\mathrm{I}\mathrm{N}`$.
###### Remark 3
The sequence $`u`$ is not generated by an iterated morphism $`\phi `$. First observe that
$$\tau (w)=\left\{\begin{array}{c}1\hfill \\ 2\hfill \\ 3\hfill \end{array}\right\}r,s\mathrm{I}\mathrm{N}:w=\{\begin{array}{c}a^{4r+1}b^{3s},a^{4r+2}b^{3s+1},a^{4r+3}b^{3s+2}\hfill \\ a^{4r+1}b^{3s+2},a^{4r+2}b^{3s},a^{4r+3}b^{3s+1}\hfill \\ a^{4r+1}b^{3s+1},a^{4r+2}b^{3s+2},a^{4r+3}b^{3s}.\hfill \end{array}$$
Suppose that there exists a morphism $`\phi `$ such that $`u=lim_{n+\mathrm{}}\phi ^n(0)`$.
1) If $`\phi (0)0102\mathrm{\Delta }^{}`$ then the block ‘$`0102`$’ must appear at least twice in $`u`$ since ‘$`0`$’ appears twice in $`u`$. If the first ‘$`0`$’ of the block is obtained from a word $`a^{4r}b^s`$ with $`r1`$ then the second ‘$`0`$’ is obtained from $`a^{4r2}b^{s+2}`$ which leads to a contradiction in view of (1). If the first ‘$`0`$’ is obtained from $`b^s`$ with $`s1`$ then the second ‘$`0`$’ come from $`a^sb`$ and we have $`s=4t`$. The ‘$`2`$’ is obtained from $`a^{4t1}b^2`$ which also leads to a contradiction.
2) If $`\phi (0)=01`$ then in view of the first terms of $`u`$, $`\phi (1)023031200231\mathrm{\Delta }^{}`$. We show that ‘$`023031200`$’ appears only once in $`u`$. Suppose that we can find another block of this kind. Thus the last two ‘$`0`$’ come from words $`b^{4r1}`$ and $`a^{4r}`$ with $`r2`$. Since we consider all the words of $`a^{}b^{}`$ lexicographically ordered, the first ‘$`0`$’ of the block come from $`a^7b^{4r8}`$ which is in contradiction with (1).
3) If $`\phi (0)=010`$ then $`\phi (1)23031200231\mathrm{\Delta }^{}`$ and $`\phi (010)01023031200\mathrm{\Delta }^{}`$. The block ‘$`010`$’ appears at least twice in $`u`$ but we know that ‘$`023031200`$’ appears only once.
We shall see further that $`u`$ is generated by a substitution.
## 3 First results about $`S`$-automatic sequences
Some classical results about $`k`$-automatic sequences can be easily restated .
###### Definition 7
Let $`a\mathrm{\Delta }`$ and $`S=(L,\mathrm{\Sigma },<)`$ , the $`S`$-fiber $`_S(u,a)`$ of a sequence $`u\mathrm{\Delta }^\omega `$ is defined as follows
$$_S(u,a)=\{\mathrm{r}_S(n):u_n=a\}.$$
###### Theorem 8
Let $`u`$ be an infinite sequence over $`\mathrm{\Delta }`$ and $`S=(L,\mathrm{\Sigma },<)`$. The sequence $`u`$ is $`S`$-automatic if and only if for all $`a\mathrm{\Delta }`$, $`_S(u,a)`$ is a regular subset of $`L`$.
Proof. If $`u`$ is $`S`$-automatic then we have a DFAO $`M=(K,s,\mathrm{\Sigma },\delta ,\mathrm{\Delta },\tau )`$ which is used to generate $`u`$. Let $`L(M^{})`$ be the language recognized by the DFA $`M^{}=(K,s,\mathrm{\Sigma },\delta ,F)`$ where the set of final states $`F`$ only contains the states $`k`$ such that $`\tau (k)=a`$. Therefore $`_S(u,a)`$ is regular since it is the intersection of the two regular sets $`L(M^{})`$ and $`L`$.
The condition is sufficient. Let $`\mathrm{\Delta }=\{a_1,\mathrm{},a_n\}`$. Remark that if $`ij`$, $`_S(u,a_i)_S(u,a_j)=\mathrm{}`$ and $`L=_{i=1}^n_S(u,a_i)`$. For all $`i=1,\mathrm{},n`$, $`_S(u,a_i)`$ is accepted by a DFA $`M_i=(K_i,s_i,\mathrm{\Sigma },\delta _i,F_i)`$. From these automata we construct a DFAO $`M=(K,s,\mathrm{\Sigma },\delta ,\mathrm{\Delta },\tau )`$ to generate $`u`$ using the numeration system $`S`$. The set $`K`$ is $`K_1\times \mathrm{}\times K_n`$, the initial state is $`(s_1,\mathrm{},s_n)`$. For all states $`(q_1,\mathrm{},q_n)K`$ and for all $`\sigma \mathrm{\Sigma }`$, $`\delta ((q_1,\mathrm{},q_n),\sigma )=(\delta _1(q_1,\sigma ),\mathrm{},\delta _n(q_n,\sigma ))`$. If there is a unique $`i`$ such that $`q_iF_i`$ then $`\tau ((q_1,\mathrm{},q_n))=a_i`$ otherwise the state cannot be reached by a word of $`L`$ and the output is not important. The sequence $`u`$ is obtained from $`S`$ and the DFAO $`M`$ thus $`u`$ is $`S`$-automatic. $`\mathrm{}`$
The notion of $`k`$-kernel of a $`k`$-automatic sequence can be transposed as follows.
###### Definition 9
Let $`S=(L,\mathrm{\Sigma },<)`$ and $`u`$ be an infinite sequence. For each $`w\mathrm{\Sigma }^{}`$, we set $`𝒦_w=\{vL|z\mathrm{\Sigma }^{}:v=wz\}.`$ One can enumerate $`𝒦_w`$ lexicographically with respect to $`<`$, $`𝒦_w=\{wz_0<wz_1<\mathrm{}\}`$. Thus for each $`w\mathrm{\Sigma }^{}`$, one can construct the subsequence $`nu_{\mathrm{val}_S(wz_n)}`$ (remark that the subsequence can be finite or even empty).
###### Theorem 10
Let $`S=(L,\mathrm{\Sigma },<)`$. A sequence $`u\mathrm{\Delta }^\omega `$ is $`S`$-automatic if and only if $`\{nu_{\mathrm{val}_S(wz_n)}:w\mathrm{\Sigma }^{}\}`$ is finite.
Proof. If $`u`$ is $`S`$-automatic, we have a DFAO $`M=(K,s,\mathrm{\Sigma },\delta ,\mathrm{\Delta },\tau )`$ used to generate $`u`$ and we define the equivalence relation $`_1`$ over $`\mathrm{\Sigma }^{}`$ by $`x_1y`$ if and only if $`\delta (s,x)=\delta (s,y)`$. In the same way, the minimal automaton of $`L`$ provides an equivalence relation $`_2`$. The two relations have a finite index thus the relation $`_{1,2}`$ given by $`x_{1,2}y`$ if and only if $`x_1y`$ and $`x_2y`$ has also a finite index. Remark that each class of $`_{1,2}`$ gives one of the sequences $`nu_{\mathrm{val}_S(wz_n)}`$. Indeed, $`x_2y`$ implies that $`\{z\mathrm{\Sigma }^{}:xzL\}=\{z\mathrm{\Sigma }^{}:yzL\}`$ thus $`𝒦_x=\{xz_0<xz_1<\mathrm{}\}`$ and $`𝒦_y=\{yz_0<yz_1<\mathrm{}\}`$ with the same $`z_0,z_1,\mathrm{}`$.
The condition is sufficient. We show how to construct a DFAO. The states are the subsequences $`q_w=(nu_{\mathrm{val}_S(wz_n)})`$. The initial state is $`q_\epsilon `$ (i.e., the subsequence obtained from the empty word). The transition function $`\delta `$ is given by $`\delta (q_w,\sigma )=q_{w\sigma }`$ and the output function $`\tau `$ is given by $`\tau (q_w)=u_{\mathrm{val}_S(w)}`$. $`\mathrm{}`$
## 4 Complexity of $`S`$-automatic sequences
The complexity function $`p_u`$ of an infinite sequence $`u`$ maps $`n\mathrm{I}\mathrm{N}`$ to the number $`p_u(n)`$ of distinct factors of length $`n`$ which occur at least once in $`u`$. In this section, we will show that the complexity of an $`S`$-automatic sequence is in $`O(n^2)`$ as a consequence that every $`S`$-automatic sequence is generated by a substitution.
Recall that an infinite word $`w`$ generated by iterated morphism has a complexity such that
$$c_1f(n)p_w(n)c_2f(n)$$
where $`f(n)`$ is one of the following functions $`1`$, $`n`$, $`n\mathrm{log}\mathrm{log}n`$, $`n\mathrm{log}n`$ or $`n^2`$ . For a survey on the complexity function, see for example .
The next remark shows that an $`S`$-automatic sequence can reach at least the same complexity as a word generated by morphism.
###### Remark 4
For every infinite word $`w`$ generated by an iterated morphism $`\phi `$ over an alphabet $`\mathrm{\Delta }`$ we can construct an $`S`$-automatic sequence $`u`$ such that $`n\mathrm{I}\mathrm{N},`$ $`p_w(n)p_u(n)`$.
We show how to proceed on the following example,
$$\mathrm{\Delta }=\{0,1\},\phi :\{\begin{array}{c}00101\hfill \\ 111.\hfill \end{array}$$
It is well-known that $`w=\phi ^\omega (0)`$ is such that $`p_w`$ is of complexity $`O(n\mathrm{log}\mathrm{log}n)`$ . To the morphism $`\phi `$, we associate a finite automaton $`M`$ (if the morphism is not uniform then $`M`$ is not deterministic). The set of states is $`\mathrm{\Delta }`$, all the states are final and the transition function $`\delta `$ is obtained by reading the productions of $`\phi `$ from left to right. For this purpose, we introduce a new ordered alphabet $`\mathrm{\Sigma }`$ such that $`\mathrm{\#}\mathrm{\Sigma }=sup_{x\mathrm{\Delta }}|\phi (x)|`$. Here, $`0`$ gives the initial state (for we consider the word $`\phi ^\omega (0)`$) and $`1`$ the other state. Thus with $`\mathrm{\Sigma }=\{a<b<c<d\}`$, we have $`\delta (0,a)=[\phi (0)]_1=0`$, $`\delta (0,b)=[\phi (0)]_2=1`$, $`\mathrm{}`$ and $`M`$ is then
As is customary, the final states are denoted by double circles. The language accepted by $`M`$ is $`L=\{a,c\}^{}\{b,d\}\{a,b\}^{}\{a,c\}^{}`$. The numeration system $`S`$ is thus $`(L,\mathrm{\Sigma },a<b<c<d)`$. This kind of construction can also be found in . Now from $`M`$ we simply construct a DFAO $`M^{}`$
The way we find the output can be easily understood. The third state can have any output for this state is never reached with a word belonging to $`L`$. One remarks that the $`S`$-automatic sequence obtained with $`M^{}`$ and $`S`$ is
$$u=\phi (0)\phi ^2(0)\phi ^3(0)\mathrm{}$$
and thus every factor of $`w=\phi ^\omega (0)`$ belongs to $`u`$.
We now show that every $`S`$-automatic sequence is generated by a substitution.
###### Lemma 11
Let $`\mathrm{\Sigma }=\{\sigma _1<\mathrm{}<\sigma _n\}`$, $`M=(K,s,\mathrm{\Sigma },\delta ,F)`$ be a DFA and $`\alpha K`$. The morphism $`\phi _M:K\{\alpha \}(K\{\alpha \})^{}`$ defined by
$$\{\begin{array}{c}\alpha \alpha s\hfill \\ k\delta (k,\sigma _1)\mathrm{}\delta (k,\sigma _n),kK\hfill \end{array}$$
produces the sequence $`u_\phi `$ of the states reached by the words of $`\mathrm{\Sigma }^{}`$ i.e., $`i\mathrm{I}\mathrm{N}\{0\}`$, $`u_{i+1}=\delta (s,w_i)`$ where $`w_i`$ is the $`i^{th}`$ element of $`(\mathrm{\Sigma }^{},<)`$.
Proof. One can check easily by constructing $`\phi (\alpha )`$, $`\phi ^2(\alpha )`$, $`\phi ^3(\alpha )`$ (which are prefixes of $`u_\phi `$) that $`u_\phi `$ satisfies the property. $`\mathrm{}`$
###### Proposition 12
Every $`S`$-automatic sequence is generated by a substitution.
Proof. Let $`S=(L,\mathrm{\Sigma },<)`$, $`M_L=(K,s,\mathrm{\Sigma },\delta ,F)`$ be a DFA accepting $`L`$ and $`u`$ be an $`S`$-automatic sequence obtained with the DFAO $`=(K^{},s^{},\mathrm{\Sigma },\delta ^{},\mathrm{\Delta },\tau )`$. From these two automata, we construct the product automaton $`M=(K\times K^{},(s,s^{}),\mathrm{\Sigma },\nu )`$ where $`\nu ((k,k^{}),\sigma )=(\delta (k,\sigma ),\delta ^{}(k^{},\sigma ))`$. We do not give explicitly the final states of $`M`$. By Lemma 11, we associate to this automaton a morphism $`\phi _M:(K\times K^{})\{\alpha \}((K\times K^{})\{\alpha \})^{}`$. To conclude the proof, we construct the erasing morphism $`h:(K\times K^{})\{\alpha \}\mathrm{\Delta }^{}`$ defined by
$$\{\begin{array}{ccc}\hfill h(\alpha )& =\epsilon \hfill & \\ \hfill h((k,k^{}))& =\epsilon \hfill & \mathrm{if}kF\hfill \\ & =\tau (k^{})\hfill & \mathrm{otherwise}.\hfill \end{array}$$
Indeed, $`\phi ^\omega (\alpha )`$ is the sequence of the states reached by the words of $`\mathrm{\Sigma }^{}`$ in $`M`$ but we are only interested in the words belonging to $`L`$ and in the corresponding output of $``$. Thus $`u`$ is generated by $`(\phi _M,h,\alpha )`$. $`\mathrm{}`$
Dealing with erasing morphisms whenever one wants to determine the complexity function of a sequence is painful. So the next lemma permits to get rid of erasing morphisms.
###### Lemma 13
If $`f`$ and $`g`$ are arbitrary morphisms with $`f(g^\omega (a))`$ an infinite word, then there exists a non-erasing morphism $`k`$ and a coding $`h`$ (i.e., a letter-to-letter morphism $`h`$) such that $`f(g^\omega (a))=h(k^\omega (a))`$. $`\mathrm{}`$
###### Theorem 14
The complexity of an automatic sequence is in $`O(n^2)`$. Moreover, there exists an automatic sequence $`v`$ and a positive constant $`d^{}`$ such that $`n>0`$, $`p_v(n)d^{}n^2`$.
Proof. Let $`u`$ be an $`S`$-automatic sequence. By Proposition 12, $`u`$ is generated by a substitution $`(\phi ,h,\alpha )`$ and by Lemma 13 we can suppose that $`h`$ is non-erasing. The word $`u_\phi =\phi ^\omega (\alpha )`$ is generated by an iterated morphism and thus $`p_{u_\phi }(n)dn^2`$. To conclude, since $`u=h(u_\phi )`$, recall that if $`v`$, $`w`$ are two infinite words and if $`h`$ is a non-erasing morphism such that $`h(v)=w`$ then there exist positive constants $`a`$, $`b`$ such that $`p_w(n)ap_v(n+b)`$ .
2) We show that there exist a language $`L`$ over an ordered alphabet and a DFAO such that the corresponding automatic sequence $`v`$ has a complexity function $`p_v(n)d^{}n^2`$.
The morphism
$$\phi :\{\begin{array}{c}001\hfill \\ 112\hfill \\ 22\hfill \end{array}$$
generates the word $`w=\phi ^\omega (0)`$. Since $`2`$ is a bounded letter (i.e., $`|\phi ^n(2)|`$ is bounded) and $`2^n`$ is a factor of $`w`$ for an arbitrary $`n`$, there exists a positive constant $`d^{}`$ such that $`p_w(n)d^{}n^2`$ (see ). Using the same technique as in Remark 4, we construct an $`S`$-automatic sequence $`v`$ such that $`p_v(n)p_w(n)`$. One find easily that the regular language used in the numeration system $`S`$ is $`L=a^{}a^{}ba^{}a^{}ba^{}ba^{}`$. $`\mathrm{}`$
To conclude this section, we refine in a very simple way Proposition 12 to give a characterization of the $`S`$-automatic sequences.
Let $`T=(\phi ,h,c)`$ and $`T^{}=(\phi ^{},h^{},c^{})`$ be two substitutions such that $`\phi :\mathrm{\Sigma }\mathrm{\Sigma }^{}`$, $`h:\mathrm{\Sigma }\mathrm{\Delta }^{}`$,$`\phi ^{}:\mathrm{\Sigma }^{}\mathrm{\Sigma }^{}`$ and $`h^{}:\mathrm{\Sigma }^{}\mathrm{\Delta }^{}`$. A morphism of substitutions $`m:TT^{}`$ is a surjective morphism $`m:\mathrm{\Sigma }\mathrm{\Delta }\mathrm{\Sigma }^{}\mathrm{\Delta }^{}`$ such that
1. $`m(c)=c^{}`$, $`m(\mathrm{\Sigma })=\mathrm{\Sigma }^{}`$, $`m(\mathrm{\Delta })=\mathrm{\Delta }^{}`$
2. $`m(\phi (\sigma ))=\phi ^{}(m(\sigma ))`$, $`\sigma \mathrm{\Sigma }`$
3. $`m(h(\sigma ))=h^{}(m(\sigma ))`$, $`\sigma \mathrm{\Sigma }`$.
For a regular language $`L`$ on the totally ordered alphabet $`(\mathrm{\Sigma },<)`$ and for a DFAO $`M=(K,s,\mathrm{\Sigma },\delta ,\mathrm{\Delta },\tau )`$, one can construct the canonical substitution $`T_{(L,<,M)}`$ by proceeding in the same way as in Proposition 12 with $`M_L`$ equals to the minimal automaton of $`L`$ and the DFAO $``$ equals to a reduced and accessible copy of $`M`$.
To reduce $`M`$, one have to merge the states $`p`$, $`q`$ such that for all $`w\mathrm{\Sigma }^{}`$, $`\tau (\delta (p,w))=\tau (\delta (q,w))`$.
###### Definition 15
A substitution $`T`$ is an $`(L,<,M)`$-substitution if there exists a morphism $`m:TT_{(L,<,M)}`$. This kind of construction has already been introduced in for linear numeration systems based on a Pisot number.
The next theorem is obvious and we state it without proof.
###### Theorem 16
Let $`S=(L,\mathrm{\Sigma },<)`$. The sequence $`u\mathrm{\Delta }^\omega `$ is $`S`$-automatic if and only if $`u`$ is generated by a $`(L,<,M)`$-substitution for some DFAO $`M`$. $`\mathrm{}`$
## 5 Application to $`S`$-recognizable sets of integers
Proposition 12 gives a necessary condition for a set $`X`$ of integers to be $`S`$-recognizable. The characteristic sequence $`\chi _X\{0,1\}^\omega `$ has to be generated by a substitution. Thus this proposition can be used as an interesting tool to show that a subset of $`\mathrm{I}\mathrm{N}`$ is not $`S`$-recognizable for any numeration system $`S`$.
In the following $`𝒫`$ is the set of primes and $`\chi _𝒫`$ is its characteristic sequence. We show that $`𝒫`$ is never $`S`$-recognizable but first we construct by hand a subset of $`\mathrm{I}\mathrm{N}`$ which cannot be $`S`$-recognizable for its characteristic sequence is too complex.
###### Example 1
For $`n3`$, consider the $`\left(\begin{array}{c}n\\ 3\end{array}\right)`$ words belonging to $`\{0,1\}^n`$ which contains exactly three ‘$`1`$’ and concatenate these words lexicographically ordered to obtain the word $`w_{n3}`$. To conclude consider the infinite word
$$w=w_0w_1w_2\mathrm{}=\underset{w_0}{\underset{}{111}}\underset{w_1}{\underset{}{\mathrm{0111\hspace{0.17em}1011\hspace{0.17em}1101\hspace{0.17em}1110}}}\underset{w_2}{\underset{}{\mathrm{00111\hspace{0.17em}01011}\mathrm{}}}\mathrm{}$$
By construction, it is obvious that for all positive constants $`C`$, there exists $`n_0`$ such that $`nn_0:p_w(n)>Cn^2`$. Thus $`w`$ cannot be generated by a substitution and the corresponding subset $`W`$ such that $`\chi _W=w`$,
$$W=\{0,1,2,4,5,6,7,9,10,11,12,14,15,16,17,21,22,23,25,27,28,\mathrm{}\},$$
is never $`S`$-recognizable.
###### Proposition 17
For any numeration system $`S`$, $`𝒫`$ is not $`S`$-recognizable.
Proof. In , C. Mauduit shows using some density arguments that $`\chi _𝒫\{0,1\}^\omega `$ is not generated by a substitution $`(\phi ,h,\alpha )`$ where $`h`$ sends all the letters on $`0`$ except one. A slight adaptation of the proof leads to the conclusion for any letter-to-letter morphism $`h`$. $`\mathrm{}`$
## 6 Acknowledgments
The author would like to warmly thank J.-P. Allouche for pointing out very useful references and for some fruitful comments and also P. Lecomte for his support and for fruitful conversations.
|
no-problem/9906/astro-ph9906421.html
|
ar5iv
|
text
|
# A Measurement of the Angular Power Spectrum of the CMB from ℓ=100to400
## 1. Introduction
It is widely recognized that the characterization of the cosmic microwave background (CMB) anisotropy is essential for understanding the process of cosmic structure formation (e.g. Hu et al. (1997), Bennett et al. (1997), Turner & Tyson (1999)). If some of the currently popular models prove correct, the anisotropy may be used to strongly constrain cosmological parameters (e.g. Jungman et al. (1995), Bond et al. (1998)). Summaries of the state of our knowledge of the CMB (e.g. Bond et al. 1998b (BJK), Page & Wilkinson (1999)) suggest the existence of a peak in the angular spectrum near $`l=200`$. In particular, BJK show $`150l_{\mathrm{peak}}350`$. Since their analysis there have been additional results at $`l>200`$ that lend support to their picture (Baker et al. (1999) (CAT), Glanz (1999) (VIPER), Wilson et al. (1999) (MSAM)). Here, we report the results from the TOCO98 campaign of the Mobile Anisotropy Telescope (MAT) which probes from $`l100`$ to $`l400`$.
## 2. Instrument
The MAT telescope, based on the design in Wollack et al. (1997), is described briefly in Torbet et al. (1999) and Devlin et al. (1998) and is documented on the web<sup>1</sup><sup>1</sup>1Details of the experiment, synthesis vectors, likelihoods, data, and analysis code may be found at http://www.hep.upenn.edu/CBR/ and http://physics.princeton.edu/~cmb/.. In this paper, we focus on results from the two $`D`$-band (144 GHz) channels. The receivers use SIS mixers designed and fabricated by A. R. Kerr and S.-K. Pan of NRAO and A. W. Lichtenberger of the University of Virginia (Kerr et al. (1993)). The six other detectors in the focal plane are 30 and 40 GHz high electron mobility transistor (HEMT) amplifiers designed by M. Pospieszalski (Pospieszalski (1992), Pospieszalski et al. (1994)).
The mixers, which operate in double sideband mode, are fed with a 25% bandwidth corrugated feed cooled to 4.5 K. The 144 GHz local oscillator (LO) is cavity stabilized and thermally controlled. The cryogenic IF HEMT amplifier, which operates between 4 and 6 GHz, is also of NRAO design. The resultant passband has been measured (Robertson (1996)) to be approximately 138-140 and 148-150 GHz. The total system sensitivity (including atmospheric loading) for each receiver is $`1.3\mathrm{mK}\mathrm{s}^{1/2}`$ (Rayleigh-Jeans) with the SIS body operating at $`4.4`$ K.
The $`D`$1 feed ($`az=207\stackrel{}{\mathrm{.}}47`$, $`el=40\stackrel{}{\mathrm{.}}63`$ at the center of the chop) is near the center of the focal plane, resulting in $`\theta _{\mathrm{FWHM}}0\stackrel{}{\mathrm{.}}2`$ ($`\mathrm{\Omega }_{\mathrm{D1}}=1.36\times 10^5`$ sr) while $`D`$2 is displaced from the center by 2.9 cm ($`az=205\stackrel{}{\mathrm{.}}73`$, $`el=40\stackrel{}{\mathrm{.}}13`$), resulting in $`\theta _{\mathrm{FWHM}}0\stackrel{}{\mathrm{.}}3`$ ($`\mathrm{\Omega }_{\mathrm{D2}}=2.93\times 10^5`$ sr). $`D`$1 is polarized with the $`E`$-field in the horizontal direction and $`D`$2 with the field in the vertical direction. No use is made of the polarization information in this analysis.
In the 1997 campaign (Torbet et al. (1999)), a microphonic coupling rendered the $`D`$-band data suspect. The problem was traced to a combination of the azimuth drive motor and the chopper. The coupling was effectively eliminated for the 1998 campaign. In addition, the chopper amplitude was reduced from 2$`\stackrel{}{\mathrm{.}}`$96 to 2$`\stackrel{}{\mathrm{.}}`$02 and the frequency reduced from 4.6 Hz to 3.7 Hz. In all other respects, the instrument was the same as for 1997.
The telescope pointing is established through observations of Jupiter and is monitored with two redundant encoders on both the azimuth bearing and on the chopper. The absolute errors in azimuth and elevation are 0$`\stackrel{}{\mathrm{.}}`$04, and the relative errors are $`<0`$$`\stackrel{}{\mathrm{.}}`$01. The chopper position, which is calibrated in the field, is sampled 80 times per chop. When its rms position over one cycle deviates more than 0$`\stackrel{}{\mathrm{.}}`$015 from the average position (due to wind loading), we reject the data.
The analysis uses data between 20 and 200 Hz. These frequencies are well removed from the refrigerator cycle frequency at 1.2 Hz, the chopper frequency, and the Nyquist frequency at 592 Hz. The amplitude of the electronic transfer function varies by $`<2`$% over this band.
## 3. Observations and Calibration
Data were taken at a 5200 m site<sup>2</sup><sup>2</sup>2 The Cerro Toco site of the Universidad Católica de Chile was made available through the generosity of Prof. Hernán Quintana, Dept. of Astronomy and Astrophysics. It is near the proposed MMA site. on the side of Cerro Toco (lat = -22$`\stackrel{}{\mathrm{.}}`$95, long = 67$`\stackrel{}{\mathrm{.}}`$775), near San Pedro de Atacama, Chile, from Aug. 26, 1998 to Dec. 7, 1998. For the anisotropy data, the primary optical axis is fixed at az = 207$`\stackrel{}{\mathrm{.}}`$41, el = 40$`\stackrel{}{\mathrm{.}}`$76, $`\delta `$ = -60$`\stackrel{}{\mathrm{.}}`$9 and the chopper scans 6$`\stackrel{}{\mathrm{.}}`$12 of sky. We present here the analysis of data from Sept. 3, 1998 to Oct. 28, 1998.
Jupiter is used to calibrate all channels and map the beams. Its brightness temperature is 170 K in $`D`$-band (Griffin et al. (1986), Ulich et al. (1981)) and the intrinsic calibration error is $`5\%`$. We account for the variation in angular diameter. To convert to thermodynamic units relative to the CMB, we multiply by $`1.67\pm 0.03`$. The error is due to incomplete knowledge of the passbands.
After determining the beam parameters from a global fit of the clear weather Jupiter calibrations, the standard deviation in the measured solid angle is 5.5% for $`D`$1 and 4% for $`D`$2. Jupiter is observed on average within 2 hours of the prime observing time (approximately 10 PM to 10 AM local). The responsivity varies $`20`$% over two months. In all, there are $`35`$ Jupiter calibrations in each channel.
To verify the calibration between observations of Jupiter, a 149 GHz tone is coupled to the detectors through the LO port for 40 msec every 200 seconds. Its effective temperature is $`1`$K. There is good long term agreement between the Jupiter and pulse calibrations. The short-term ($`<`$ 1 day) calibration is determined with a fit of the pulses to the Jupiter calibrations. The measurement uncertainty in the calibration is 7%.
The total 1$`\sigma `$ calibration error of 10% for $`D`$1 and 9% for $`D`$2 is obtained from the quadrature sum of the above sources. In the full analysis, $`D`$1 and $`D`$2 are combined; thus the uncorrelated component of the error adds in quadrature yielding an error for the combination of 8%.
## 4. Data Reduction
The data reduction is similar to that of the TOCO97 experiment (Torbet et al. (1999)). We use the terminology discussed there and in Netterfield et al. (1997), though we now use Knox filters (Knox (1999)) (in place of window functions) to determine the $`l`$-space coverage. For $`D`$1 we form the 2-pt through 16-pt synthesized beams and for $`D`$2, the 2-pt through 17-pt synthesized beams. In practice, atmospheric contamination precludes using the 2-pt through 4-pt data and the achieved sensitivity renders the 17-pt and higher uninteresting. If there is atmospheric contamination in the 5-pt data, its level is under the 1 sigma error. For the 7-pt and higher, atmospheric contamination is negligible.
The phase of the time ordered data relative to the beam position is determined with observations of Jupiter and the Galaxy. In the analysis, we use the phase for each harmonic obtained when the quadrature signal from the Galaxy is minimized.
A quantity useful in assessing sensitivity to the beam shape is $`l_{\mathrm{eff}}\theta _{\mathrm{FWHM}}`$. For SK at $`l_{\mathrm{eff}}=256`$, $`l_{\mathrm{eff}}\theta _{\mathrm{FWHM}}=2.5`$. For TOCO98 at $`l_{\mathrm{eff}}=409`$, $`l_{\mathrm{eff}}\theta _{\mathrm{FWHM}}=1.5`$ for $`D`$1 and 2.2 for $`D`$2. This corresponds to a separation of lobes in the synthesized beam of $`2\theta _{\mathrm{FWHM}}`$ for $`D`$1 and $`1.3\theta _{\mathrm{FWHM}}`$ for $`D`$2.
As with TOCO97, the harmonics are binned according to the right ascension at the center of the chopper sweep. The number of bins depends on the band and harmonic as shown in Table 1. For each night, we compute the variance and mean of the data corresponding to a bin. These numbers are averaged over the 25 good nights and used in the likelihood analysis.
After cuts based on pointing, the data are selected according to the weather. For each n-pt data set, we find the mean rms of 6.5 sec averages over 15 minute segments. When this value exceeds 1.2 of the minimum value for a given day, the data from that 15 min segment, along with the previous and subsequent 15 min segments are cut. The effect is to keep 5-10 hour blocks of continuous good data in any day, and to eliminate transitions into periods of poor atmospheric stability. Increasing the cut level adds data to the beginning and end of the prime observing time.
The stability of the instrument is assessed through internal consistency checks and we examine it with the distribution of the offset of each harmonic. The offset is the average of a night of data after the cuts have been applied (the duration ranges from 5-10 hours) and is typically of magnitude $`150\mu `$K with standard deviation $`75\mu `$K. The offsets for these data were stable over the campaign. The resulting $`\chi ^2/\nu `$ is typically 1-4. For the offsets of the quadrature signal, $`\chi ^2/\nu `$ is typically 1-2. The stability of the offset led to a relatively straightforward data reduction.
To eliminate the potential effect of slow variations in offset, we remove the slope and mean for each night for each harmonic. This is accounted for in the quoted result following Bond et al. (1991).
## 5. Analysis and Discussion
In the analysis, we include all known correlations inherent in the observing strategy. In computing the “theory covariance matrix” (BJK) which encodes the observing strategy, we use the measured two dimensional beam profiles. From the data, we determine the correlations between harmonics due to the atmosphere. Because the S/N is only 2-5 per synthesized beam, and the noise is correlated between beams, we work with groups of harmonics. This is similar to band averaging, though we use the full covariance matrix so as to include all correlations.
Table 1 gives the results of separate analyses of $`D`$1 and $`D`$2. Both channels show a fall in the angular spectrum above $`l=300`$. The fact that the results agree is an important check as the receivers (other than the optics) are independent. It is not possible to compute $`D`$1$``$$`D`$2 directly from the data because of the different beam sizes. The eventual production of a map will facilitate the comparison.
In the full analysis, $`D`$1 and $`D`$2 are combined. The resulting likelihoods are shown in Figure 1 along with the results of the null tests. Because $`D`$1 and $`D`$2 observe the same section of sky at different times, some care must be taken in computing the correlation matrices. The correlation coefficients between $`D`$1 and $`D`$2 due to the atmosphere are of order $`0.05`$. The largest off-diagonal terms of the theory covariance matrix are $`0.4`$. The quoted results are insensitive to the precise values of the off-diagonal terms of the covariance matrix. The combined analysis affirms what is seen in $`D`$1 and $`D`$2 individually and shows a peak in the angular spectrum near $`l=200`$.
The angular spectrum of the TOCO97 and TOCO98 data agree in the regions of common $`l`$. We compute the spectral index of the fluctuations by comparing band powers. We find $`\beta _{\mathrm{CMB}}=\mathrm{ln}(\delta T_{144}/\delta T_{36.5})/\mathrm{ln}(144/36.5)=0.04\pm 0.25`$, (including calibration error), where $`\delta T_{144}`$ is the weighted mean of the two highest points for TOCO98 and $`\delta T_{36.5}`$ is a similar quantity for TOCO97 (36.5 GHz is the average TOCO97 frequency). For the CMB, $`\beta _{\mathrm{CMB}}=0`$. For dust, $`\beta _{\mathrm{RJ}}=1.7`$ corresponds to $`\beta _{\mathrm{CMB}}=2.05`$; for free-free emission $`\beta _{\mathrm{RJ}}=2.1`$ corresponds to $`\beta _{\mathrm{CMB}}=1.75`$. Though it is possible for spinning dust grains (Drain & Lazarian (1999)) to mimic this spectrum for our frequencies, the amplitude of this component is small (de Oliveira-Costa et al. (1998)). In addition, the spatial spectrum of diffuse sources like interstellar dust falls as $`l^{3/2}`$ (Gautier et al. (1992)), so the observed peak is inconsistent with our observations at $`l100`$.
The frequency spectral index $`\beta _{\mathrm{RJ}}`$ of unresolved extra-Galactic sources is typically between 2 and 3, inconsistent with the measured index. In addition, the spatial spectrum of sources rises as $`\delta T_ll`$, inconsistent with our observations at $`l400`$. Moreover, recent analyses (e.g. Tegmark (1999)) estimate the level of point source contamination to be much lower than the fluctuations we observe. We therefore conclude that the source of the fluctuations is the CMB.
We assess the statistical significance of the decrease in $`\delta T_l`$ for $`l>300`$ by comparing just the likelihood distributions at $`l=226`$ ($`L_{226}`$, Fig. 1), for which $`\delta T_l=83\mu `$K, and $`l=409`$, for which $`\delta T_l<67\mu `$K (95%). These two distributions are effectively uncorrelated. The point at which $`L_{409}=L_{226}`$ is, coincidentally, the $`2\sigma `$ lower limit on $`\delta T_{226}`$ and the 95% upper limit on $`\delta T_{409}`$. Thus, there is a 0.97 probability that $`\delta T_{226}`$ is greater than the 95% upper limit on $`\delta T_{409}`$. In addition, the probability that $`\delta T_{409}83\mu `$K (the peak of $`L_{226}`$) is 0.996. When all the data in Figure 2 are considered, the significance of a decrease in $`\delta T_l`$ for $`l>300`$ will be even higher.
The weighted mean of data from TOCO97, TOCO98, and SK between $`l=150`$ and 250, is $`\overline{\delta T}_{\mathrm{peak}}=82\pm 3.3\pm 5.5\mu `$K (the second error is calibration uncertainty). This is consistent with, though slightly higher than, the value from the Wang et al. (1999) concordance model plotted in Figure 2, which gives $`\overline{\delta T}_{\mathrm{peak}}75\mu `$K. In the context of this model, the high $`\overline{\delta T}_{\mathrm{peak}}`$ favors a smaller $`\mathrm{\Omega }_mh^2`$ (e.g. larger “cosmological constant”) or more baryons.
Figure 2 shows results taken over six years and seven observing campaigns and three different experiments. Though a detailed confrontation with cosmological models will have to await a thorough analysis and comparison with other experiments, a straightforward read of the data indicates a rise to $`\delta T_{\mathrm{peak}}85\mu `$K at $`l200`$ and a fall at $`l>300`$. The data strongly disfavor models with a peak in the spectrum at $`l=400`$. Future work will include the analysis of the TOCO98 HEMT and remaining $`D`$-band data.
We gratefully acknowledge conversations with and help from Dave Wilkinson, Norm Jarosik, Suzanne Staggs, Steve Myers, David Spergel, Max Tegmark, Angel Otárola, Hernán Quintana, the Princeton Machine Shop, Bernard Jones, Harvey Chapman, Stuart Bradley, and Eugenio Ortiz. The experiment would not have been possible without NRAO’s site monitoring and detector development. We also thank Lucent Technologies for donating the radar trailer. This work was supported by an NSF NYI award, a Cottrell Award from the Research Corporation, a David and Lucile Packard Fellowship (to L. P.), an NSF Career award (AST-9732960, to M. D.), a NASA GSRP fellowship to A. M., an NSF graduate fellowship to M. N., a Robert H. Dicke fellowship to E. T., NSF grants PHY-9222952, PHY-9600015, and the University of Pennsylvania. The data will be made public upon publication of this Letter.
|
no-problem/9906/quant-ph9906076.html
|
ar5iv
|
text
|
# References
Several remarks on “Comments” by A. Moroz
P. Šťovíček
Department of Mathematics, Faculty of Nuclear Science, CTU,
Trojanova 13, 120 00 Prague, Czech Republic
Abstract
We make a couple of remarks on “Comments” due to A. Moroz which were addressed to our recent letter .
PACS. 03.65.Nk – Nonrelativistic scattering theory
In this note we wish to make a couple of remarks on “Comments on ’Differential cross section for Aharonov–Bohm effect with non-standard boundary conditions’ “ (referred to as Comments in what follows) due to A. Moroz and addressed to our recent letter . In fact, these remarks have been revised as a consequence of the revision of Comments. In the very beginning it should be emphasized that in our letter we didn’t pretend and attempt to do anything more than claimed in the introductory part, and this was to complete the results of the preceding paper by doing some elementary numerical analysis, plotting a couple of graphs, and providing them with some basic discussion. However all the involved formulae were derived in (of course, the rotational symmetry was mentioned there, too). The basic result of consists in finding a general form of boundary conditions, depending on four parameters, imposed on a wave function in the case of idealized Aharonov–Bohm effect in the plane, and furthermore in derivation of a formula for the differential cross section with the corresponding Hamilton operator. The value of the magnetic flux is then another parameter, the fifth one, occurring in the problem.
It should be noted that approximately at the same time another paper appeared, , treating the same problem as we did in , using practically the same mathematical tools, and naturally arriving at the same results. So Comments by Moroz might have been better addressed to the papers and rather than to .
The mathematical machinery applied in and is based on the theory of self-adjoint extensions of symmetric operators which is nowadays quite a common tool suited for this type of problems. For example, just for the purpose of illustration, let us mention that in another relatively recent paper, , in its nature a similar problem, concerning this time a magnetic monopole in three-dimensional space, has been solved by the same method.
As I understood from Comments, Moroz had applied in his analysis a completely different technique based on a limit procedure when sending the radius of the flux tube to zero. There is no doubt that this approach is highly interesting, too. For example one may hope to give this way the involved parameters a more concrete interpretation. Unfortunately I am not able to compare the results directly since I didn’t deduce from Comments what was the precise definition of Hamiltonian, particularly what kind of boundary conditions were imposed on wave functions. Nevertheless some aspects seem to be clear. First of all, Comments are apparently concerned with the case when the s and p-wave are decoupled. Then the corresponding family of Hamiltonians depends only on two parameters, called $`\mathrm{\Delta }_n`$ and $`\mathrm{\Delta }_{n1}`$ in formula (3) of Comments, while the complete solution admits coupling of the s and p-wave via boundary conditions, and consequently depends, as already mentioned, on four parameters.
As far as the rotational symmetry is concerned, this notion is interpreted somewhat differently in on one side and in Comments on the other side. This was not our aim to study symmetries of the differential cross section itself, for example with respect to the reflection $`\phi \phi `$. A basically more essential observation has been made in (and graphically illustrated in ). The point is that the angular momentum need not be conserved, or, in other words, the angular momentum need not commute with the Hamiltonian. This effect happens provided the s and p-wave are coupled, and this is what we called the violation of the rotational symmetry. For the differential cross section this means that it depends non-trivially on both the scattering and incident angle, $`\phi `$ and $`\phi _0`$, and not merely on their difference $`\phi \phi _0`$. Note that only the scattering angle $`\phi `$ occurs in formula (3) of Comments, with $`\phi _0`$ being set to 0.
Furthermore, let us make a short remark on the number of bounded states. The situation is slightly more complicated than mentioned in Comments. The possible number of bounded states is 0, 1 or 2. This means that even with non-standard boundary conditions it may happen that there are no bounded states. The dependence of the number of bounded states on the choice of boundary conditions has been analyzed quite explicitly both in and .
As for the longer history, the fact that the two critical sectors of angular momentum admit more general boundary conditions than the regular one has been known for a long time, see for example the now classical paper , though this possibility was not exploited systematically until recently. We have to admit that we missed the work . No doubt it should be included among the references of . On the other hand I’d like to point out that the case with the s and p-wave decoupled was studied in several works prior to the paper referred to in Comments (see ).
|
no-problem/9906/physics9906003.html
|
ar5iv
|
text
|
# Scaling properties of three-dimensional magnetohydrodynamic turbulence
## Abstract
The scaling properties of three-dimensional magnetohydrodynamic turbulence are obtained from direct numerical simulations of decaying turbulence using 512<sup>3</sup> modes. The results indicate that the turbulence does not follow the Iroshnikov-Kraichnan phenomenology. The spectrum is consistent with $`k^{5/3}`$. In the case of hyperresistivity the structure functions exhibit a clear scaling range yielding absolute values of the scaling exponents $`\zeta _p`$, in particular $`\zeta _31`$, consistent with a recent analytical result. The scaling exponents agree with a modified She-Leveque model $`\zeta _p=p/9+1(1/3)^{p/3}`$, corresponding to Kolmogorov scaling but sheet-like geometry of the dissipative structures.
Magnetic turbulence is the natural state of a plasma in motion, especially in astrophysical systems. The convenient framework to describe such turbulence is magnetohydrodynamics (MHD). For high magnetic Reynolds number $`Rm=vl_0/\eta `$, where $`v`$ is a typical turbulent velocity, $`l_0`$ the integral scale and $`\eta `$ the magnetic diffusivity, there is a broad range of scales $`l`$ between $`l_0`$, and the dissipative scale length $`l_d`$, $`l_0ll_d`$, called the inertial range which exhibits characteristic self-similarity or scaling properties.
The concept of inertial-range scaling was introduced by Kolmogorov for hydrodynamic turbulence, which is called the Kolmogorov (K41) phenomenology. Assuming homogeneity and isotropy of the turbulence as well as locality of the turbulent cascade process, he obtains $`ϵ(\delta v_l)^2/\tau _l=(\delta v_l)^3/l`$ yielding the scaling law $`\delta v_lϵ^{1/3}l^{1/3}`$. Here $`ϵ`$ is the energy dissipation rate and, to be specific, $`\delta v_l=[𝐯(𝐱+𝐥)𝐯(𝐱)]𝐥/l`$ is the longitudinal velocity increment. A direct consequence is the Kolmogorov energy spectrum $`E_kϵ^{2/3}k^{5/3}`$.
For MHD turbulence the Iroshnikov-Kraichnan (IK) phenomenology , takes into account the Alfvén effect, the coupling of small-scale velocity and magnetic fluctuations by the integral-scale field $`B_0`$. Hence the natural variables are the Elsässer fields $`𝐳^\pm =𝐯\pm 𝐁`$, which describe Alfvén waves. In the IK phenomenology the spectral transfer is reduced by the factor $`\tau _A/\tau _l`$, $`ϵ(\tau _A/\tau _l)(\delta z_l)^2/\tau _l`$, where $`\tau _A=l/v_A`$, $`v_A`$ = Alfvén velocity in the field $`B_0`$, $`\delta z_l\delta v_l\delta B_l(ϵv_A)^{1/4}l^{1/4}`$, and the IK energy spectrum becomes $`E_k(ϵv_A)^{1/2}k^{3/2}`$. This spectrum can also be written in the form $`E_kϵ^{2/3}k^{5/3}(kl_0)^{1/6}`$ with the integral scale $`l_0`$ defined by $`l_0=v_A^3/ϵ`$, which illustrates the nonlocal character of the energy cascade in MHD turbulence.
It is, however, well known that these qualitative scaling relations for $`\delta v_l`$ or $`\delta z_l`$ are not exactly valid in a statistical sense because of intermittency, which implies that the distribution of turbulent scales is not strictly self-similar. A quantitative measure is provided by the scaling exponents $`\zeta _p`$ of the structure functions, the moments of the field increments. For hydrodynamic turbulence She and Leveque proposed a model leading to the expression $`\zeta _p^{\mathrm{SL}}=p/9+2[1(2/3)^{p/3}]`$, which fits the experimental results surprisingly well, reproducing in particular the exact result $`\zeta _3=1`$. This model has been modified for MHD incorporating the IK effect , , which yields $`\zeta _p^{\mathrm{IK}}=p/8+1(1/2)^{p/4}`$, in particular $`\zeta _4^{\mathrm{IK}}=1`$.
The IK phenomenology has been supported by direct numerical simulations of 2D MHD turbulence at moderate Reynolds numbers . However, recent developments in MHD turbulence theory cast some doubt on the general validity of the IK scaling. 2D simulations at considerably higher Reynolds numbers reveal an anomalous scaling behavior , , indicating that the results of Ref. are not asymptotic. There have also been theoretical arguments in favor of a Kolmogorov scaling, e.g., , . Even more convincingly, exact relations have been derived for moments of certain triple products of $`\delta z_l`$ , which are shown to be proportional to $`l`$, i.e., $`\zeta _3=1`$, analogous to the well-known 4/5-relation in hydrodynamic turbulence, thus excluding the IK result $`\zeta _4=1`$. Scaling exponents for MHD turbulence have also been derived from observations in the solar wind . Here agreement with the IK exponents has been claimed , but in this comparison the observational results were normalized assuming $`\zeta _4=1`$. Actually the error bars seem to be too large to reach a definite conclusion.
To clarify the issue of scaling in 3D MHD turbulence direct numerical simulations are desirable with higher Reynolds numbers than studied previously, for instance in , , , , . In this Letter we present a numerical study of freely decaying turbulence with spatial resolution of $`512^3`$ modes. The scaling properties are analyzed by considering the time-averages of the normalized spectra and structure functions. We solve the incompressible MHD equations
$$_t𝐁\times (𝐯\times 𝐁)=\eta _\nu (1)^{\nu 1}^{2\nu }𝐁,$$
(1)
$$_t𝐰\times (𝐯\times 𝐰)\times (𝐣\times 𝐁)=\mu _\nu (1)^{\nu 1}^{2\nu }𝐰,$$
(2)
$`𝐰=\times 𝐯,𝐣=\times 𝐁,𝐯=𝐁=0,`$
by applying a pseudo-spectral method with spherical mode truncation as conveniently used in 3D turbulence simulations (instead of full dealiasing by the 2/3 rule used in most 2D simulations). The generalized magnetic Prandtl number $`\eta _\nu /\mu _\nu `$ has been set equal to unity. Initial conditions are
$$𝐁_𝐤=a\mathrm{e}^{k^2/k_0^2i\alpha _𝐤},𝐯_𝐤=b\mathrm{e}^{k^2/k_0^2i\beta _𝐤},$$
(3)
which are characterized by random phases $`\alpha _𝐤`$, $`\beta _𝐤`$ and satisfy the conditions $`𝐤𝐁_𝐤=𝐤𝐯_𝐤=0`$, $`E=E^V+E^M=1`$ and $`E^V/E^M=1`$. Further restrictions on $`𝐁_𝐤`$ arise by requiring a specific value of the magnetic helicity $`H=d^3x𝐀𝐁`$. We believe that finite magnetic helicity is more typical than $`H0`$, since MHD turbulence usually occurs in rotating systems. The wavenumber $`k_0`$, the location of the maximum of the initial energy spectrum, is chosen $`k_0=4`$, which allows the inverse cascade of $`H_𝐤`$ to develop freely during the simulation time of 10 time units (about 7 eddy turnover times, defining the eddy turnover time as the time required to reach the maximum dissipation from the smooth initial state). Though this choice implies a certain loss of inertial range, the sacrifice is unavoidable in the presence of inverse cascade dynamics, since assuming $`k_01`$ would lead to magnetic condensation in the lowest-$`k`$ state, which would also affect the dynamics of higher-$`k`$ modes. Both normal diffusion $`\nu =1`$ and hyperdiffusion $`\nu =2`$ have been used, $`\nu =1`$ to discuss the spectral properties and $`\nu =2`$ to determine the scaling of the structure functions. All runs presented in this Letter have finite $`H`$, $`H/H_{\mathrm{max}}0.6`$, and negligible alignment. Table I lists the important parameters of the simulation runs, where the magnetic Taylor Reynolds number is $`Rm_\lambda =Rm^{1/2}`$. Since $`Rm_\lambda `$ is not stricly constant during turbulence decay but increases slowly $`Rm_\lambda t^{1/8}`$, we give the values taken at a specific time $`t=4`$.
We first discuss the spectral properties considering the angle-averaged energy spectrum $`E_k`$. Figure 1 shows the scatter plot of the normalized spectrum (the normalization is discussed below), compensated by $`k^{3/2}`$, taken from run 3 over the period $`t=410`$ of fully developed turbulence. The spectrum exhibits a clear scaling range of almost one decade with a spectral law, which is definitely steeper than the IK spectrum $`k^{3/2}`$, close to (in fact slightly steeper than) $`k^{5/3}`$ indicated by the dashed line. In order to form the time average in a system of decaying turbulence, the spectrum must be normalized to eliminate the time variation of the macroscopic quantities. In hydrodynamic turbulence the only such quantity is $`ϵ`$, which leads to the universal form of the Kolmogorov spectrum $`\widehat{E}(\widehat{k})=E_k/(ϵ\eta ^5)^{1/4}`$, $`\widehat{k}=kl_K`$, where $`l_K=(\eta ^3/ϵ)^{1/4}`$ is the Kolmogorov length. However, when normalized in this way the MHD energy spectrum is found to change during turbulence decay and even more strongly so when comparing runs of different $`Rm_\lambda `$. For finite magnetic helicity the spectrum may also depend on $`H`$, which introduces a second macroscopic length scale $`l_1=H/v_A^2`$ in addition to $`l_0`$, i.e, the spectrum may contain some function of $`l_0/l_1`$. To determine this function we propose the following argument. Since the Alfvén effect is clearly present in the simulations, kinetic and magnetic energy spectrum being nearly equal at small scales, while on the other hand the scaling $`E_kk^{5/3}`$ is observed, we modify the nonlinear transfer in the IK ansatz by a factor $`(l/l_1)^\delta `$,
$$\frac{\tau _A}{\tau _l}\left(\frac{l}{l_1}\right)^\delta \frac{\delta z_l^2}{\tau _l}=ϵ,$$
(4)
and determine $`\delta `$ by requiring the observed scaling $`\delta z_ll^{1/3}`$, which gives $`\delta =1/3`$ and hence
$$E_kϵ^{2/3}(l_0/l_1)^{1/6}k^{5/3}.$$
(5)
Also the dissipation scale length is slightly changed. Balancing nonlinear transfer and dissipation gives
$$l_K^H=l_K(l_0/l_1)^{1/8}.$$
(6)
Using these relations we obtain a the normalized energy spectrum $`\widehat{E}(\widehat{k})=E_k/[(ϵ\eta ^5)^{1/4}(l_0/l_1)^{3/8}]=\widehat{k}^{5/3}F(\widehat{k})`$, $`\widehat{k}=kl_K^H`$. Normalized in this way the spectra at different times of run 3 coincide very well as seen in Fig. 1 and so do the time-averaged normalized energy spectra of runs 1-3 shown in Fig. 2, which vary only by the extent of the inertial range, apart from statistical oscillations. Relations (5), (6) are not valid for $`H0`$, where we expect the pure Kolmogorov normalization to be valid.
A more complete picture of the inertial-range distribution of turbulent structures is provided by the scaling exponents $`\zeta _p`$ of the structure functions, where the second order exponent $`\zeta _2`$ is related to the inertial-range spectral law $`k^{(1+\zeta _2)}`$. To be definite we consider the moments of the absolute value of the longitudinal increments $`\delta z_l^\pm `$ discussing only the runs 4 and 5 with the highest Reynolds numbers. The normalized structure functions $`\widehat{S}_p^\pm (\widehat{l})=|\delta z_l^\pm |^p/E^{p/2}`$, $`\widehat{l}=l/l_K^H`$, are averaged over time, exhibiting a similar weak scatter as for the spectrum in Fig. 1.
For normal diffusion $`\nu =1`$ no scaling range is visible. (Note that the structure function $`S_2`$ corresponds to the one-dimensional spectrum $`E_{k_x}`$, which has a shorter inertial range than the angle-averaged spectrum shown in Fig. 1.) For $`\nu =2`$, however, there is a scaling range $`30<\widehat{l}<200`$, as seen in Fig. 3, where the time-averaged curves $`\widehat{S}_p^+`$ are plotted for $`p=3,4`$. The inserts give the logarithmic derivatives, where the central quasi-constant parts determine the scaling coefficients, the dashed horizontal lines indicating the most probable values $`\zeta _3^+0.95`$, $`\zeta _4^+1.15`$. These results are consistent with the spectral law derived from Fig. 1 and are close to the analytical prediction $`\zeta _3=1`$. It is true that the analytical theory refers to third-order moments different from $`|\delta z|^p`$ discussed here, but the scaling coefficients should not depend thereof (the scaling range, however, does). One might object that the use of hyperdiffusion affects the inertial-range scaling, if the scaling range is not very broad. In fact, the energy spectrum law tends to be polluted by the bottleneck effect, which is particularly pronounced for hyperdiffusion (see e.g., ). Thus the energy spectrum in run 4 (not shown) is effectively flatter than $`k^{5/3}`$ expected from the value of $`\zeta _2`$. However, there is, to our knowledge, no argument for a similar effect in the structure functions.
Assuming the exact result $`\zeta _3=1`$ allows to obtain rather accurate values of $`\zeta _p`$ by using the property of ESS (extended self-similarity) plotting $`S_p`$ as function of $`S_3`$. (It should be noted that ESS usually results in almost perfect scaling behavior, but the scaling coefficients thus derived vary in time, hence time averaging is required.) The results are shown in Fig. 4, which gives the ESS results of $`\zeta _p^+`$ for run 4 (diamonds) and the ESS values $`\xi _{3p}^+=\zeta _p^+/\zeta _3^+`$ from for 2D MHD turbulence (triangles). (In 2D MHD the absolute values of $`\zeta _p`$ are found in to decrease with $`Rm`$, while the relative values $`\xi _{3p}`$ appear to be independent of $`Rm`$.) The results indicate that in 3D MHD turbulence is less intermittent than 2D, but it is more intermittent than hydrodynamic turbulence, the continuous curve, which gives the She-Leveque result $`\zeta _p^{\mathrm{SL}}`$. As shown by Politano and Pouquet , the She-Leveque concept contains effectively three parameters: $`g`$ related to the scaling $`\delta z_ll^{1/g}`$, $`x`$ related to the energy transfer time at the smallest, the dissipative scales $`t_ll^x`$, and $`C`$, the codimension of the dissipative structures,
$$\zeta _p=\frac{p}{g}(1x)+C\left(1(1x/C)^{p/g}\right).$$
(7)
Our results for the 3D MHD case suggest Kolmogorov scaling $`g=3`$, $`x=2/3`$, while different from hydrodynamic turbulence the dissipative structures are sheet-like, hence the codimension is $`C=1`$,
$$\zeta _p^{\mathrm{MHD}}=p/9+1(1/3)^{p/3}.$$
(8)
This is the dashed curve in Fig. 4, which fits the numerical values very well.
In conclusion we have studied the spatial scaling properties of 3D MHD turbulence using direct numerical simulations with resolution of $`512^3`$ modes. The results indicate that the turbulence does not follow the Iroshnikov-Kraichnan (IK) phenomenology. The energy spectrum is consistent with a $`k^{5/3}`$ law. For hyperresistivity the structure functions exhibit a clear scaling range yielding absolute values of the scaling exponents $`\zeta _p`$, in particular $`\zeta _31`$, consistent with recent analytical predictions. The scaling exponents agree well with a modified She-Leveque model $`\zeta _p^{\mathrm{MHD}}=p/9+1(1/3)^{p/3}`$, corresponding to Kolmogorov scaling, but sheet-like geometry of the dissipative structures. The results are also consistent with observations of turbulence in the solar wind, which typically show a $`k^{1.7}`$ spectrum.
The authors would like to thank Andreas Zeiler for providing the basic version of the code, Antonio Celani for developing some of the diagnostics, and Reinhard Tisma for optimizing the code for the CRAY T3E.
|
no-problem/9906/astro-ph9906030.html
|
ar5iv
|
text
|
# INTERMEDIATE MASS STARS: UPDATED MODELS
## 1 INTRODUCTION
The comprehension of the evolutionary status of the various stellar systems, from the simplest stellar clusters up to the more complex galaxies, are mainly based on the comparison between theoretical stellar models and observational data. It follows that any improvement and/or assumptions in the basic input physics (equation of state, opacity, and the like) included in the computations of the stellar models directly influences the interpretation of the observed data. Moreover, since the population synthesis requires the availability of stellar models in a large range of masses and chemical compositions, and since such an homogeneous database is still missing, people involved in such kind of studies have been forced to collect models computed by different authors.
In this paper we present new evolutionary sequences for stellar masses ranging between 1.2 and 9 $`M_{}`$; these models have been included in our database of stellar evolution which is available by anonymous ftp. This database also includes fully homogeneous models for low mass stars (Straniero, Chieffi & Limongi 1997) and those for massive stars (Chieffi, Limongi & Straniero, 1998, Limongi, Straniero & Chieffi, 1998). All these models have been obtained by means of the FRANEC, an acronym of Frascati Raphson Newton Evolutionary Code (Chieffi & Straniero, 1989, and Chieffi, Limongi & Straniero 1998). The input physics (equation of state, opacity, neutrino losses and the like) we are adopting at the moment are discussed in Straniero, Chieffi & Limongi (1997).
Several papers described sets of models of intermediate mass stars with various chemical compositions. However, owing to the continues improvement of the input physics, the (re)computation of the same stellar models becomes necessary. Major contributions are from: Kippenhahn, Thomas & Weigart 1965; Iben 1967a (and references therein); Paczynski 1970a, 1970b, 1971; Trimble, Paczynski & Zimmerman 1973; Alcock & Paczynski 1978; Becker & Iben 1979; Becker 1981; Vandenberg 1985; Maeder & Meynet, 1987, 1988, 1989, 1991; Castellani, Chieffi & Straniero 1990, 1992; Lattanzio, 1991; Stothers & Chin 1990, 1991a, 1991b, 1992; Vassiliadis & Wood 1992; Alongi et al. 1991, 1992; Bressan et al. 1993; Schaller et al. 1992. Thus, our new models will be compared with the most recent compilations of similar evolutionary sequences. Whenever possible, we analyze the origin of the resulting differencies. The main goal of this study is to constraint the present level of uncertainty of the stellar evolution in the range of intermediate mass stars. In such a way, we provide a basic tool to check the reliability of our understanding of the galactic history as emerging from the study of population synthesis.
This is the plan of the paper: in the next section we revise the possible sources of uncertainties for H and He burning intermediate mass stellar models; in section 3 we describe our latest evolutionary computations for intermediate mass stars, from the ZAMS up to the AGB; in section 4 we discuss the comparisons among different evolutionary models; selected tests of the evolutionary sequences are presented in section 5. Final remarks follow.
## 2 INPUT PHYSICS AND CONVECTION
If the theoretical investigation of low mass stars appears well anchored to the result of helioseismology (see Straniero, Chieffi & Limongi, 1997) this is not the case for intermediate mass stars. It is commonly believed that the many uncertainties in the theory of turbulent convection still affects our understanding of the internal structure of these kind of stars. Owing to the lack of a conclusive test for the adequacy of the current theory of convection, the astrophysical literature presents a variety of different approaches to the computation of stellar models. It has been early recognized that the mixing of material in the core of a given star significantly alters its lifetime and in turn it could modify the age estimations of the various galactic components. The instability against turbulent convection is classically handled by means of a thermodynamical criterion (namely the Schwarzschild criterion for a chemically homogeneous fluid). As it is well known, this criterion is based on the evaluation of the expected gradient of temperature produced by the radiative transport of energy: when the required gradient is too high, the radiative flux cannot account for the whole energy transport and hence convection is settled on.
First of all let us emphasize that the correct evaluation of the size of an unstable region is primarily dependent on the accuracy of the input physics. Any improvement of the stellar physics (eos, opacity, cross section and the like) could imply a variation of the estimated value of the temperature gradient and in turn it could modify the location of the borders of the convective regions, with sizable consequences on the computed stellar lifetime.
A second question concerns the possibility that the convective motion is not drastically inhibited in a stable region located just outside the convective core. In fact, although out of the Schwarzschild border a moving element of matter is subject to a strong deceleration, it might be possible that a non zero velocity is maintained for a certain path. In such a case this mechanical overshoot might induce a mixing of material in a radiatively stable region and might also contribute to the energy transport. A large number of papers have been devoted to the inclusion of such phenomenon in the computation of stellar models. Major contributions are from: Shaviv & Salpeter (1973); Maeder (1975); Cloutman & Whitaker (1980); Bressan et al (1981); Stothers & Chin (1981, 1990); Matraka, Wassermann & Weigert (1982); Xiong (1983, 1986); Doom (1982, 1985); Alongi et al. (1991); Maeder & Meynet (1987, 1988, 1989, 1991); Shaller et al. (1992).
Unfortunately the available convection theory is still inadequate for a reliable description of this phenomenon (see Renzini 1988 for a critical discussion on this argument), so that the evaluation of the degree of both matter and energy transport out of the unstable regions must be obtained by comparing the result of parameterized models with the measurements of some selected observable quantities (see e.g. Bressan et al. 1993). As for the mixing length (Chieffi, Straniero & Salaris 1995), the calibration of the free parameters used to describe the mixing in the overshooting region, is model dependent. In fact, since the assumed input physics affects the size of the unstable regions, the calibration of the overshooting depends on these assumptions. For example, the larger is the opacity the larger is the convective region and in turn the lower is the required amount of overshooting. But the opacity coefficients are likely underestimated rather then overestimated. For this reason, if at the beginning of the eighties Becker & Methews (1983) claimed a relatively strong overshoot in order to reconcile the theory with the observed distribution of stars in the young Globular Clusters of the Magellanic Clouds, the latest attempts to derive the size of the convective core overshoot for H-burning stars indicate that, if it’s present, it should be ”mild” (see e.g. Sthothers & Chin 1992; Castellani, Chieffi & Straniero 1992; Schaller et al. 1992; Bressan et al. 1993; Demarque et al. 1994, Mermilliod et al. 1994; Schröder et al. 1997). In particular these studies generally found that the best reproduction of the various indicators of the convective core size is obtained with models including an overshoot roughly confined in between 0 and 0.3 (in units of pressure scale height and measured from above the stability border defined by means of the Schwarzschild criterion).
The situation is still more controversial for the central He-burning phase. In such a case, when He is converted into C within the core, the opacity increases and, in turn, the convective core size must increase (Schwarzschild 1970; Paczynski 1970b; Castellani et al. 1971a). As the He burning proceeds, a minimum in the radiative gradient settles on, so that mixing occurs in two separated regions: an internal one, which is fully convective, and an external one, in which the resulting mixture of C and He is just that needed to allow the convective neutrality (Castellani et al, 1971b; see also Iben 1986 and references therein). Such phenomenon was called He burning semiconvection. Close to the He exhaustion (namely when the central He becomes lower then approximately 0.1), some instabilities at the border of the convective core appears in stellar model computations (Castellani et al. 1985 a,b; Iben 1986). These instabilities were called breathing pulses (BP). As a consequence of both semiconvection and BPs, a larger amount of fuel is available for the central He-burning. There is some debate concerning the actual occurrence of the BPs in real stars (Renzini & Fusi Pecci, 1988; Caputo et al., 1989). In any case, the inclusion of these phenomena might affect some important results of stellar evolution: the estimated central He-burning and AGB evolutionary times, the final amount of C and O in the core, the final WD mass. Note that the efficiency of both semiconvection and BPs also depend on the adopted input physics. For example, as firstly discussed by Iben (1972), the use of different prescriptions for the $`{}_{}{}^{12}C(\alpha ,\gamma )^{16}O`$ reaction rate alters the duration of the final part of the He-burning phase during which the BPs occur. Then, a larger value of this rate will enhance the effects of the BPs.
In summary, when comparing the evolutionary results obtained by different authors, the connection between the assumed mixing scheme and the adopted input physics must be taken into account. This will be done in section 4.
## 3 THE NEW MODELS
Models from 1.2 to 9 $`M_{}`$ and metallicity ranging between $`10^4`$ and $`2\times 10^2`$ have been computed from the ZAMS (Zero Age Main Sequence) up to the end of the E-AGB (Early Asymptotic Giant Branch) phase. The evolutionary tracks in the HR diagram are reported in figures 1, 2, 3 and 4. The runs of the central temperature versus the central density are shown in figures 5, 6, 7 and 8. Some examples of the evolution of the fully convective regions are illustrated in figures 9, 10, 11 and 12.
In tables 1 to 4 we have reported the fundamental properties of the evolutionary sequences, namely from column 1 to 10: the total mass, the central H-burning lifetime (in Myr), the maximum size (in solar mass unit) of the convective core during central H-burning, the surface He mass fraction after the first dredge-up, the tip luminosity of the first RGB (red giant branch), the He core mass at the beginning of the He-burning, the central He-burning lifetime (in Myr), the He core mass (in solar unit) at the end of the He burning, the surface He mass fraction after the second dredge-up and the He core mass (in solar unit) at the beginning of the TP-AGB (thermally pulsing asymptotic giant branch) phase.
In the following part of this section we briefly summarize the main features of the computed sequences of models, revising the dependence of the various evolutionary phases on the stellar mass and on the chemical composition. As already recalled in the introduction of this paper, the evolutionary history of an intermediate mass star, crossing the HR diagrams from the main sequence up to the AGB, is well known. For a more accurate description of the various evolutionary phases and an exhaustive list of references we remind the reader to the review paper by Iben (1991).
### 3.1 THE CENTRAL H-BURNING
All the sequences of models having mass larger than 1.2 $`M_{}`$ develop a convective core during the central H burning independently on the initial chemical composition. As a consequence, at variance with low mass stars for which the H burning occurs in a radiative environment, the evolutionary tracks evolve off the ZAMS towards lower temperature and larger luminosity. As the H is converted into He in the central region of the star, the opacity decreases and the convective core recedes (in mass). The convective instability in the core is retained until the H mass fraction is reduced down to about 0.1. Then an overall contraction occurs and the star rapidly move towards the radiative main sequence. A maximum in luminosity is reached at the time of the H exhaustion.
One interesting quantity characterizing the H-burning phase is the maximum extension of the convective core. As already recalled, this maximum is attained just after the ZAMS (see figures 9, 10, 11 and 12). In the last 30 years the computed values for this quantity have been systematically increased, mainly due to the increasing values of the adopted radiative opacity coefficients. Our present results are listed in column 3 of tables 1 to 4. Note that the size of the convective core decreases monotonically as the initial He increases while it initially increases as the metallicity decreases (up to Z=0.001) and then it decreases at smaller metallicities. The corresponding H-burning evolutionary times are reported in column 2.
### 3.2 THE H-BURNING SHELL
When the H-burning shell settles on, the convective envelope penetrates more than 80 % in mass of the star, bringing to the surface the products of the H burning. As firstly pointed out by Iben (1964, 1967b), the main consequences of this first dredge-up are: the increase of the surface abundances of <sup>4</sup>He, <sup>3</sup>He and <sup>14</sup>N and a decrease of those of <sup>12</sup>C and <sup>16</sup>O. The modification of the surface composition is stronger in low mass stars due to the minor size of the envelope. The surface amount of He resulting after the first dredge-up in our models is listed in column 4 of tables 1 to 4. The subsequent evolution up to the onset of the He burning phase is mainly characterized by the equation of state governing the He core. In the strong degenerate regime a quite large He core mass is necessary to ignite He (namely about 0.5 $`M_{}`$, but it depends on the chemical composition). This is the case of a low mass star (i.e. $`M1.5M_{}`$). For more massive stars the degree of degeneracy in the core is reduced and in turn the He ignition is more rapidly attained. In the asymptotic limit of non degenerate matter the minimum mass needed to ignite He is about 0.35 $`M_{}`$. For this reason the luminosity of the RGB tip and the He core mass attained at the He ignition (columns 5 and 6 of tables 1 to 4) decrease when the total mass increases from 1.5 up to 2.5 $`M_{}`$. When the stellar mass is larger than 2.5-3 $`M_{}`$, the off main sequence evolution is not further controlled by the growth of the He core. In such a case, owing to the internal mixing occurring during the main sequence, the H-burning shell at the beginning of the RGB settles well outside the minimum mass needed to ignite He. Hence, the RGB tip and the mass of the He core at the He ignition rise as the total mass increases.
The minimum resulting by the combination of these two behaviors marks the so called RGB phase transition (Iben 1967c). Such an occurrence is illustrated in figure 13. According to the classical results, we found that the minimum core mass is attained for a total mass of 2.3-2.5 $`M_{}`$, value slightly increasing with the metallicity. Note that the almost constant minimum core mass at very different metallicities is the consequence of the opposite influence of Z and Y on this quantity. In fact, as pointed out by Sweigart, Greggio & Renzini (1990), the transition mass increases as the metallicity increases and decreases as the He increases.
### 3.3 THE CENTRAL HE-BURNING
As it is well known, the larger the core mass at the He ignition the brighter is the star during the central Helium burning phase and the shorter is the central He burning lifetime. Hence a maximum in such a lifetime is expected, corresponding to the minimum He core mass occurring at the RGB phase transition. In figure 14 we have reported this lifetime for our models with Z=0.02 and Y=0.28. They are listed in column 7 of tables 1 to 4 for the full set of models.
During central He burning the evolutionary tracks move towards the blue part of the HR diagrams on a Kelvin-Helmotz time scale and then moves back to the red giant branch when the central He vanishes. The extension of this loop depends on both the stellar mass and the chemical composition (see Alcock & Paczynski, 1978, and references therein). The larger the mass the hotter the left border of the loop, but an interesting exception is worth to be noted (Castellani, Chieffi & Straniero, 1990). At the lower metallicities (Z=0.0001, 0.001 and 0.006), the more massive stars ignite He before they can reach the RGB, thus skipping the first dredge-up. It occurs for stellar masses $`M2.7M_{}`$, $`M4M_{}`$ and $`M5M_{}`$, at Z=0.0001, 0.001 and 0.006 respectively. Since the larger is the surface He amount the lower is the opacity, the fate of the first dredge-up affects the blue loop extension: those models in which the first dredge up do not occur have a narrower blue loop. Such an occurrence is clearly shown in figure 2, which reports the HR diagrams for Z=0.006: note that the He burning evolutionary track of the $`4M_{}`$ sequence have a significantly larger blue loop than the $`5M_{}`$ one.
### 3.4 THE DOUBLE SHELL PHASE
During the whole central He burning phase the H-burning shell moves outward so that the longer is the central He burning lifetime the greater will be the increment of the He core mass (see column 8 in tables 1 to 4). For this reason, the minimum in the $`M_{He}`$-initial mass relation, which is evident at the beginning of the central He burning, is smoothed away at the end of the E-AGB phase (see figures 15). As a consequence all the stars with $`M3`$ $`M_{}`$ starting the thermally pulsing AGB phase have a rather similar He core mass, namely $`0.55\pm 0.05M_{}`$ depending on the metallicity (column 10 of tables 1 to 4). Such a value provides us a lower limit to the expected mass of a CO WD.
For a sufficiently high initial stellar mass a second dredge-up occurs during the early AGB (Kippenhahn et al. 1965, Paczynski 1970, Becker & Iben 1979). In such a case, the convective envelope penetrates the H discontinuity located just below the H-burning shell so that the resulting He core mass is lowered with respect to the value attained at the end of the central He-burning (column 10 and 8 of tables 1 to 4, respectively). The second dredge-up takes place only if the H-burning shell extinguishes. This is not the case for a low mass stars, in which the expansion induced by the He-burning shell did not induce a sufficient cooling of the H-burning one. We found that the minimum mass for the occurrence of the second dredge-up is 4 $`M_{}`$ for the three lower metallicities and a bit larger (about 5 $`M_{}`$) in the case of $`Z=0.02`$. The surface He mass fraction after the second dredge-up is listed in column 9 of tables 1 to 4.
### 3.5 THE FINAL MASSES AND $`M_{UP}`$
The white dwarf (WD) masses resulting from the evolution of low and intermediate mass stars are very important quantities for the purpose of the study of population synthesis, Planetary Nebula, Novae, Super novae, and the like (see e.g. Iben, 1991).
In figure 16 we compare our theoretical final masses with the relation reported by Weidemann (1987) and the updated one by Herwig (1995). Squares represent the He core masses at the end of the E-AGB, while arrows show the growth of the He core masses during the TP-AGB phase, as derived by using our thermally pulsing models (Straniero et al. 1996). The number reported at the top of each arrow is the number of thermal pulses computed up to the end of the AGB phase. It was determined according to the mass loss rate prescriptions of Groenewegen & de Jong (1994).
Note that the Weidemann predictions (as well as the recent improvements incorporated by Herwig) are based on a semi empirical approach. They make use of various methods to evaluate the WD masses in nearby Open Clusters, whose initial mass is derived from the turnoff age. Our final masses are instead the result of a pure theoretical calculation and, hence, they are mainly dependent on the adopted input physics. Such a difference should be well kept in mind when comparing our results with those of Weidemann (or Herwig), as we do in figure 16. Despite the two different approaches, there is an acceptable agreement between our theoretical final masses and those of Weidemann (Herwig). In the next section we will discuss our final core masses in comparison with the ones obtained by other authors by means of alternative stellar evolutionary codes.
Stars with higher masses ignite the Carbon before the onset of the thermally pulsing phase (Paczynski 1971; Alcock & Paczynski 1978; Becker & Iben 1979,1980; Castellani, Chieffi & Straniero 1991; Bressan et al. 1993; García-Berro, Ritossa, Iben 1997). In tables 1 to 4 we have distinguished among the models which experience an off center Carbon ignition and those with a central Carbon ignition. We found that $`M_{up}`$, i.e. the maximum mass for which the concurrent action of the pressure of a strong degenerate electron component and the neutrino energy loss in the core prevent the onset of the C-burning, ranges between 6.5 and 8 $`M_{}`$, the lower and the larger values being obtained for Z=0.0001 and 0.02, respectively
## 4 THE PRESENT LEVEL OF UNCERTAINTY
Owing to the large amount of numerical algorithms and physical ingredients commonly used in the computation of stellar models, the evaluation of their reliability is not trivial. A first idea of the possible sources of uncertainties can be obtained by comparing the evolutionary sequences obtained by different authors by means of different evolutionary codes and/or input physics. This might be also useful to evaluate the correctness of merging different sets of stellar models. As already recalled, there exist a rather large number of papers which present set of models for intermediate mass stars. In the following we will compare our results with the most recent and widely adopted collections of these stellar models.
### 4.1 OLD AND NEW PHYSICS
Let us firstly compare the present computations to the ones of Castellani, Chieffi & Straniero (1990 and 1992), which were obtained by means of almost the same evolutionary code, but by adopting an ”old” physics. Such a comparison will provide us with an evaluation of the importance of the most recent (last decade) improvements of the input physics. In figure 17 we have compared the evolutionary tracks for Z=0.02. The two sets of models appear rather similar except for some (important) details. Concerning the main sequence, the new tracks are slightly brighter and the convective path (i.e. from the ZAMS up to the beginning of the overall contraction) is longer.
The new H-burning lifetimes are generally lower ($`5`$ %) than the old ones, but this is mainly due to the slightly lower amount of He used in our old computations (namely Y=0.27). Concerning the He burning, the most striking difference is the extension of the blue loop in the more massive sequences, the new ones being significantly wider. The He burning lifetime is substantially unchanged.
### 4.2 DIFFERENT EVOLUTIONARY CODES
The second step in the evaluation of the reliability of the evolutionary sequences will be the comparison with the most recent and widely adopted compilations of stellar models. Let us distinguish between models with and without mechanical convective core overshoot.
The most recent sets of intermediate mass stellar models without overshoot have been published by Lattanzio (1991, L91) and Vassiliadis and Wood (1993, VW93). Despite the differences in the chemical composition and in the input physics there is a good agreement between our H-burning models and the ones found in the two papers cited above. For example, by interpolating on the grid published by Lattanzio we derive for a $`2.5M_{}`$ (Z=0.02 and Y=0.28) an H-burning lifetime of 512 Myr to be compared with our result, namely 505 Myr. For the same stellar mass, but Y=0.25, Vassiliadis and Wood found 619 Myr, Since in this range of mass and metallicity we found that by increasing the original helium of $`\delta Y=0.1`$ the corresponding $`t_H`$ must be reduced of about 25 Myr, the quoted value correspond to about 545 Myr at Y=0.28. Similar differences are found for other masses and other chemical compositions. Concerning the He burning models the situation is more complicated. The He burning lifetime is strongly dependent on both the assumed scheme for convection and the $`{}_{}{}^{12}C(\alpha ,\gamma )^{16}O`$ reaction rate. As in VW93, we allow semiconvection and suppress breathing pulses, whereas they are both allowed in the Lattanzio’s computation. On the other hand, we use the $`{}_{}{}^{12}C(\alpha ,\gamma )^{16}O`$ reaction rate of Caughlan et al. (1985) which is about three time larger than the rates adopted by Lattanzio (1991) and Vassiliadis and Wood (1993). We recall that the larger this reaction rate the larger the He-burning lifetime. Bearing in mind these differences in the input physics, for the $`5M_{}`$ with Z=0.02 (0.016 in VW93) one find 20.8, 23.5 and 30.6 in the present paper, VW93 and L91, respectively. For lower masses and/or metallicity the differences are similar. For example, for $`M=1.5M_{}`$ and Z=0.001 we found $`t_{He}=97.4`$ Myr to be compared to 122.2 Myr (VW93) and 118.3 (L91). In summary, our H-burning lifetimes appear in good agreement with those obtained in other studies, being the differences in the evolutionary time scales always lower than $`10\%`$. Note that similar differences were found with respect to our old computations. On the contrary, the present uncertainty in the theoretical evaluation of the He-burning lifetime is definitely larger. Differences up to $`30\%`$ are found in $`t_{He}`$. In principle they should be primarily attributed to the uncertainties in the convective algorithm and/or in the major He burning reaction rates. In practice, due to the connection between input physics and mixing efficiency, it is rather complicated to disclose the origin of such differences.
Recent models including a moderate amount of convective core overshoot have been published in a series of papers by the Padua group (see Bressan et al., 1993, and references therein; in the following B93) and by the Geneve one (Schaller et al. 1992 and reference therein; in the following S92). We recall that the B93 models were obtained by extending, in practice, the mixed central region of an H-burning or He burning intermediate mass star by approximately 0.25 $`H_p`$ over the unstable zone, while Schaller et al. assume 0.2 $`H_p`$.
Concerning the H-burning, when the convective core overshoot is taken into account, a larger amount of fuel is available in the burning region of the star. However, since the larger is the mixed region the brighter is the star, this additional fuel is more rapidly burned, so that the effect of the core overshoot on the H- burning lifetime is partially counterbalanced. In figure 18 we compare our H-burning lifetimes to those resulting from the B93 and S92 models. As expected, overshoot models are generally older than the corresponding classical ones. Note, however, that despite the similar amount of overshoot assumed by Bressan et al. and Schaller et al., the differences between these two sets of models are comparable to the ones found with respect to our (no overshoot) models.
Another important consequence of the convective core overshoot during the central H-burning is the reduction of the mass at which the RGB transition occurs. Because the RGB evolution is faster if the star has a non degenerate He core, the number of stars lying on the RGB of galactic open clusters might be used, in principle, to derived the value of the transition mass and, in turn, to discriminate in between models with and without core overshoot (see e.g. Mermilliod et al., 1994 and references therein). The comparison between our models and those of Padua, shows that the difference is presently quite small. For example, at Z=0.02 we found a transition mass of 2.4 $`M_{}`$ while Bressan et al. (1993) found 2.2 $`M_{}`$. Thus, minor differences are aspected in the synthetic RGB populations. Note that a similar comparison cannot be made with the Schaller et al. models because their set is not spaced enough in mass.
At variance with the H-burning phase, the inclusion of a moderate overshoot in computing He burning models do not significantly alter the resulting He-burning lifetime. In fact, if a moderate overshoot is taken into account, the semiconvective layer is hidden by this extra mixing, but the total amount of fuel (He) available for the central nuclear burning should be practically the same than that found in models without core overshoot but including semiconvection. Our He-burning evolutionary times and those obtained by Bressan et al. and Schaller et al. are compared in figure 19. Note that B93 adopt the rather low Caughlan & Fowler (1988) rate for the $`{}_{}{}^{12}C(\alpha ,\gamma )^{16}O`$ reaction, whereas S92 use our preferred rate (i.e. Caughlan et al., 1985). The differences in the stellar lifetimes are larger than those found in the case of the H-burning phase (up to 60%). Also in this case, the origin of the disagreement is not easily recognized.
Let us conclude by noting that the typical differences which we found when comparing our models (no overshoot) with those by L91, VW93, B93 and S92 are of the same order of magnitude of those found in the comparisons between the two set of models with convective core overshoot. In other words, in the range of intermediate mass stellar models, the current uncertainty due to a possible not negligible occurrence of a convective core overshoot appears less severe, or of the same order of magnitude, than those induced by other input physics.
### 4.3 THE RELIABILITY OF THE THEORETICAL CORE MASSES
In the previous section we have compared the final masses obtained by evolving our models up to the end of the AGB to the semiempirical initial/final mass relation (Weidemann, 1987). These quantities depend on the core mass attained at the beginning of the thermally pulsing AGB phase and on the AGB mass loss rate (see e.g. Iben & Renzini 1983). For the more massive stars also the efficiency of the second dredge-up should be taken into account (Paczynski 1971, Becker & Iben 1979,1980) In the following we compare our evolutionary core masses at the first TP with the ones obtained by other authors. Let us recall that the larger the duration of the He burning phase the larger is the time available for the shell H-burning to advance in mass. Hence, the large uncertainty on the current estimation of the stellar lifetime (as illustrated in section 4.2) might affects the theoretical previsions of the final masses. Concerning the convective algorithm, the lowest lifetime, and in turn the smallest He core mass, is obtained when semiconvection, BPs and overshooting are neglected, while an approximate doubling of the He-burning lifetime is found when, as in our models, only the semiconvection is taken into account. However by comparing the RGB, HB and AGB theoretical lifetime ratios to the observed stellar number ratios of well studied galactic Globular Clusters, it is possible to discriminate among the various mixing hypothesis (see e.g. Renzini & Fusi Pecci 1988). In such a way, there is a support to the classical semiconvection scheme (no BPs), but a moderate overshooting, which could mimic the effect of semiconvection, cannot be ruled out. The current uncertainty in the resonant contribution to the $`{}_{}{}^{12}C(\alpha ,\gamma )^{16}O`$ reaction rate might also change the estimated He burning lifetime (Iben 1967). In such a case, by varying the astrophysical factor in the range of value compatible with the available measurements for this reaction rate (see Buchmann, 1996 and 1997), we have obtained a variation of the He burning lifetime of about 5-10 %.
In figure 20 we compare our core masses at the end of the E-AGB to those computed by Lattanzio (1991) and Bressan et al (1993). In spite of the rather large discrepancies found in the He-burning evolutionary time scales, there is a good agreement between the core masses obtained by the different authors. Only the core masses of the more massive models by Bressan et al. are rather larger than ours (i. e. for $`M5M_{}`$). From table 5 of B93 we see that the maximum core masses attained before the onset of the second dredge-up in the 5 and 7 $`M_{}`$ sequences are slightly lower than in our models. Thus their larger core masses at the first TP should be a consequence of a less efficient second dredge-up. This is also confirmed by the smaller changes induced by the dredge-up on the surface composition. Note that B93 even include 0.7 $`H_p`$ of undershooting in their computations.
## 5 TEST OF THE EVOLUTIONARY SEQUENCES
Although a detailed comparison of our stellar models with the observed properties of different stellar systems is well beyond the purpose of the present paper, let us discuss two interesting tests of the evolutionary sequences recently proposed by different authors, which allow us to check the reliability of the current theoretical scenario. To do that we have computed selected isochrones and synthetic diagrams based on the present set of stellar models.
### 5.1 THE PLEIADES AND THE BROWN DWARF TEST
The certain identification of brown dwarfs would provide important informations on the star formation rate of very low mass stars and contribute to shed some light on the dark matter problem. For this reason, in the last few years, the search of these objects in nearby stellar clusters has been intensified. Brown dwarfs are (quasi) stars for which H burning does not occur or, at least, it does not reach the full equilibrium. How to certificate such an occurrence? A low mass stars approaching the main sequence is fully convective, so that the products of the internal nuclear burning should appear at the surface. Thus, the best indicators of the occurrence of the internal nuclear burning are the secondary elements of the pp-chain. In this context, Li is a good tracer, since it is a very volatile element and it is easily observable in faint objects. Stars with $`M1M_{}`$ deplete Li even in pre-main sequence. But in very low mass stars, when the internal temperature never exceeds $`23\times 10^6`$ K, the Li remains unburnt. Therefore, the presence of observable Li lines in stellar spectra is a certain identification of a brown dwarf (Rebolo, Martin & Magazzu 1992). In turn the reappearance of Li in the lower main sequence of stellar clusters provides an interesting test for the age estimated by means of stellar models. In fact the upper luminosity for which Li is measured depends on the age of the cluster: the larger is the age the fainter is the Li cutoff. Basri et al. (1996) by means of accurate infrared photometry coupled to high resolution spectroscopy of brown dwarfs candidates in the Pleiades, have been able to identify the Li cutoff. By using this Li test they found an age of $``$ 115 Myr. They concludes that this value is definitely larger than the age estimated by comparing the turnoff luminosity with that predicted by canonical stellar models.
In figure 21 we report the isochrones fitting obtained by using our present stellar models (no overshoot) and that derived by Bertelli et al. (1994) by using the B93 models. Accordingly to the Hypparcos parallaxes, we adopt a true distance modulus of 5.33. A reddening of 0.04 has been assumed. In both cases the isochrones are computed for Z=0.02 and Y=0.28. Concerning the canonical models, by excluding the isolated bright star at V=2.87, or $`M_V=2.58`$, (a blue straggler?), one can get an age of at least 120 Myr and not exceeding 140 Myr, i.e. a value in very good agreement with the brown dwarf test. It is worth noting that owing to the few amount of stars in the turnoff region, a precise age cannot be derived by means of the isochrone fitting. However, even taking into account such an uncertainty, we can exclude ages lower than 100 Myr. Similarly with the Bertelli et al. (1994) isochrones one may get an age in between 150 and 200 Gyr, which is a bit larger then that implied by the brown dwarf Li cutoff. Obviously, also in this case the uncertainty due to the few statistical significance of the number of turnoff stars might be claimed, so that we can’t definitely rule out the presence of a moderate overshoot. Let us finally note that this canonical estimation of the Pleiades age, based on the new distance, removes the old controversy of the lack of an evident lower main sequence turnon formed by those stars approaching the ZAMS (Herbig, 1962; Stauffer, 1984). In fact the corresponding lower limit claimed by Stauffer (i.e. 100 Myr) is well in agreement with the present determination.
### 5.2 THE WHITE DWARFS LUMINOSITY FUNCTION
An intermediate mass star ($`M56M_{}`$) ends its life as a CO white dwarf. Thus, if this star is a member of an old stellar system (say 1 Gyr or older), it spent most of its life as a WD so that its cooling time might be used as an age indicator for the stellar system. A search for the cutoff of the WD luminosity function has been recently performed by Von Hippel et al. (1995) by means of the HST with the Wide Field Planetary Camera 2. They were able to identify this cutoff in two old open cluster, namely NGC 2420 and NGC 2477. Then, by means of the theoretical cooling sequences computed by Wood (1994), they estimated the ages of these two clusters and concluded that they are in contrast with all the available stellar models. Owing to the existence of an accurate CCD photometry only for one of these two cluster, namely NGC 2420 (Anthony-Twarog et al 1990), we will focus our attention on this one. From the paper by Von Hippel et al. we derive an age based on the WD luminosity function cutoff of 1.5-1.6 Gyr.
In figure 22 we show the isochrones fitting to the Color Magnitude diagram of NGC 2420. According to Anthony-Twarog et al. (1990), we have assumed a metallicity of Z=0.008, which correspond to about \[M/H\]=-0.4, and an E(B-V)=0.05. Then the distance modulus was derived by reproducing the clump of the He burning stars which is a feature almost independent on the age (see Castellani, Chieffi & Straniero, 1992). The resulting age is of the order of 1.6 ($`\pm `$ 0.2) Gyr, value in very good agreement with the age derived from the WD cooling sequence. A slightly larger value would be obtained by adopting isochrones including a moderate amount of convective core overshoot. For example, Carraro & Chiosi (1994) found 2.1 Gyr while Friel (1995) reported 2.8 Gyr. However, as already noted by Demarque et al. (1994), the canonical isochrones cannot account for the distribution of stars near the turnoff of NGC 2420. In particular the path of the isochrones just before the overall contraction appears shorter than the observed one. Demarque et al. showed that a moderate amount of convective core overshoot (namely $`\lambda =0.23H_p`$) make longer the isochrones path, but they are forced to use an age of 2.4 Gyr.
We argue that the isochrones provide us just the locus ”permitted” to the single stars in the CMD. In order to understand if and how this permitted locus is really populated or not (and at what extent) a comparison between observed and synthetic CMD is absolutely required. When a suitable mass function as well as binary stars are considered, a very good reproduction of the observed sequences of NGC 2420 is obtained (see figure 23). In particular the contribution of the binaries leads to a larger spread in the main sequence. Note the effect on the turnoff region: the convective path of single stars seems prolonged by the presence of the binary main sequence and the bluer region after the overall contraction gap is depopulated. Also in this case we obtain an age of 1.6 Gyr which is in very good agreement with the value derived from the WD luminosity function cutoff.
## 6 CONCLUSIONS
In this paper we have illustrated the main properties of our latest set of stellar models for intermediate mass stars as obtained by means of the FRANEC code (Chieffi & Straniero 1989). By comparing the most recent evolutionary sequences computed by using different evolutionary codes and/or input physics, we found a rather large disagreement, which is partially due to the influence of the theoretical assumptions on the estimated extension of the convective regions. We would again remark that not only the difference in the adopted convective algorithm (Schwarzschild criterion, overshooting, semiconvection and the like) is responsible of such a disagreement. The connection among the various ingredients of the model cooking must be understood in order to recognize the origin of the theoretical uncertainties. In some case we found that models obtained by using very different schemes for the treatment of the convective instabilities are more similar than models obtained by using the same algorithm, but different input physics (eos, opacity, nuclear reaction rate and the like).
In spite of this disagreement, since the brighter is a model the lower is the lifetime, many differencies are smoothed away when transposing the evolutionary tracks into the isochrones. This has been already shown in the previous section where we compare our theoretical isochrones and those by Bertelli et al (1994) with the Color Magnitude diagrams of some well studied Open Clusters (see figure 21). Although the evolutionary features of these two sets of models are rather different, the resulting isochrone fittings are quite similar in the two cases. This means that a quite similar isochrone path may be obtained simply by rescaling the mass (or the age). Such an occurrence is clearly illustrated in the example reported in figure 24. In this figure we compare our isochrones of 0.7 Gyr with the ones of Bertelli et all having 0.8 Gyr.
We recall that the Bertelli isochrones were obtained by assuming a moderate amount of overshooting, whereas our models do not include any extra mixing with respect to the instability boundary. As discussed in the previous section, the only evident difference is in the shape of the turnoff.
Another quantities which is well established in the framework of the current theory of the stellar evolution is the core mass attained at the beginning of the AGB phase. In fact, since the luminosity of an off main sequence star is mainly controlled by the size of its He-core mass, the H-burning shell have more time to advance in mass in models with lower core mass,
Let us finally comment that the use of the Color Manitude diagrams to check the reliability of a particular set of models cannot be barely made by means of the isochrone fitting. In fact the best photometric studies of open clusters include few thousand of stars. Then the ”permitted” locus do not necessarily coincide with the ”populated” locus. In addition many Open Clusters have a huge population of binary stars which contribute to determine the shape of the observed Color Magnitude diagrams. The case of NGC2420, as discussed in the previous section, is a template of such a situation.
FIGURE CAPTIONS
|
no-problem/9906/cond-mat9906268.html
|
ar5iv
|
text
|
# Systematic Evolution of the Magnetotransport Properties of Bi2Sr2-xLaxCuO6 with Carrier Concentration
Magnetotransport Properties of Bi<sub>2</sub>Sr<sub>2-x</sub>La<sub>x</sub>CuO<sub>6</sub>
## Abstract
We report that it is possible to obtain a series of high-quality crystals of Bi<sub>2</sub>Sr<sub>2-x</sub>La<sub>x</sub>CuO<sub>6</sub>, of which the transport properties have been believed to be “dirtier” than those of other cuprates. In our crystals, the normal-state transport properties display behaviors which are in good accord with other cuprates; for example, in the underdoped region the in-plane resistivity $`\rho _{ab}`$ shows the pseudogap feature and in the overdoped region the $`T`$ dependence of $`\rho _{ab}`$ changes to $`T^n`$ with $`n>`$1. The characteristic temperatures of the pseudogap deduced from the resistivity and the Hall coefficient data are presented.
PACS numbers: 74.25.Fy, 74.62.Dh, 74.72.Hs
A useful way to elucidate the origin of the peculiar normal-state properties of high-$`T_c`$ cuprates is to study their systematic evolution upon changing the carrier concentration. Bi<sub>2</sub>Sr<sub>2</sub>CuO<sub>6</sub> (Bi-2201) system is an attractive candidate for such studies, because the carrier concentration can be widely changed by partially replacing Sr with La (to underdope) or Bi with Pb (to overdope) . Moreover, this system allows us to study the normal-state in a wider temperature range, because the optimum $`T_c`$ (achieved in Bi<sub>2</sub>Sr<sub>2-x</sub>La<sub>x</sub>CuO<sub>6</sub> with $`x`$$``$0.4 ) is about 30 K, which is lower than the optimum $`T_c`$ of La<sub>2-x</sub>Sr<sub>x</sub>CuO<sub>4</sub>. However, a number of problems have been known so far for Bi-2201 crystals: (i) the transport properties of Bi-2201 are quite non-reproducible even among crystals of nominally the same composition ; (ii) the residual resistivity of $`\rho _{ab}`$ is usually large (the smallest value reported to date is 70 $`\mu \mathrm{\Omega }`$cm ), as opposed to other systems where the residual resistivity in high-quality crystals is negligibly small; and (iii) the temperature dependence of the Hall coefficient $`R_H`$ is weak and thus the cotangent of the Hall angle $`\theta _H`$ does not obey the $`T^2`$ law , while $`\mathrm{cot}\theta _H`$$``$$`T^2`$ has been almost universally observed in other cuprates.
Here we report the transport properties of a series of high-quality Bi-2201 crystals, in which the normal-state transport properties display behaviors that are in good accord with other cuprates. We show data on $`\rho _{ab}(T)`$ and $`R_H(T)`$ for a wide range of carrier concentrations, from which we extract the characteristic temperatures for the pseudogap.
The single crystals of Bi<sub>2</sub>Sr<sub>2-x</sub>La<sub>x</sub>CuO<sub>6</sub> (BSLCO) are grown using a floating-zone technique in 1 atm of flowing oxygen. Note that pure Bi-2201 is an overdoped system , and increasing La doping brings the system from overdoped region to underdoped region. The actual La concentrations in the crystals are determined with the ICP analysis. Here we report crystals with $`x`$=0.24, 0.30, 0.44, 0.57, 0.66, and 0.74, for which the zero-resistance $`T_c`$ is 24, 30, 33.3, 29.2, 21.4 K, and 17.3 K, respectively. The optimum doping is achieved with $`x`$$``$0.4, which is consistent with previous reports on BSLCO . The optimum zero-resistance $`T_c`$ of 33 K (which is in agreement with the Meissner-onset $`T_c`$) is, to our knowledge, the highest value ever reported for Bi-2201 or BSLCO system.
Figure 1(a) shows the $`T`$ dependence of $`\rho _{ab}`$ for the six $`x`$ values in zero field. Clearly, both the magnitude of $`\rho _{ab}`$ and its slope show a systematic decrease with increasing carrier concentration (decreasing $`x`$). One can see that it is only at the optimum doping that $`\rho _{ab}`$ shows a good $`T`$-linear behavior: In the underdoped region, $`\rho _{ab}(T)`$ shows a downward deviation from the $`T`$-linear behavior, which has been discussed to mark the pseudogap . In the overdoped region, $`\rho _{ab}(T)`$ shows an upward curvature in the whole temperature range; the $`T`$ dependence of $`\rho _{ab}`$ in the overdoped region can be well described by $`\rho _{ab}`$=$`\rho _0+AT^n`$ (with $`n`$=1.14 and 1.27 for $`x`$=0.30 and 0.24, respectively), which is a behavior known to be peculiar for the overdoped cuprates . Shown in Fig. 1(b) is the $`T`$ dependence of $`R_H`$ for the six samples. Here again, a clear evolution of $`R_H`$ with $`x`$ is observed; the change in the magnitude of $`R_H`$ at 300 K suggests that the carrier concentration is actually reduced roughly by a factor of 4 upon increasing $`x`$ from 0.24 to 0.74. Note that the $`T`$ dependence of $`R_H`$ is stronger than those previously reported . In our data, a pronounced peak in $`R_H(T)`$ is clearly observed for all carrier concentrations and the position of the peak shifts systematically to higher temperatures as the carrier concentration is reduced.
We observed that the cotangent of the Hall angle, $`\mathrm{cot}\theta _H`$, obeys a power-law temperature dependence, $`T^\alpha `$, where $`\alpha `$ is nearly 2 in underdoped samples ($`x`$=0.74 and 0.66) but shows a systematic decrease with increasing carrier concentration . Figure 2 shows the plots of $`\mathrm{cot}\theta _H`$ vs $`T^\alpha `$ for the two underdoped samples ($`x`$=0.74 and 0.66) and the optimally-doped sample ($`x`$=0.44). We note that the $`T^2`$ law of $`\mathrm{cot}\theta _H`$ is confirmed for the first time for Bi-2201 in our crystals. A particularly intriguing fact here is that $`\mathrm{cot}\theta _H`$ of the optimally-doped sample changes as $`T^{1.7}`$, not as $`T^2`$, while $`\rho _{ab}`$ shows a good $`T`$-linear behavior. This suggests that the Fermi-liquid-like behavior of the Hall scattering rate, $`\tau _H^1`$$``$$`T^2`$, may not be a generic feature of the optimally-doped cuprates. The upward deviation from the $`T^2`$ behavior evident in Fig. 2(a) for the most underdoped sample ($`x`$=0.74) is likely to be related to the opening of the pseudogap .
Due to the limited space, we will concentrate below on the implication of our data to the pseudogap in Bi-2201. As we noted above, the downward deviation from the $`T`$-linear behavior in $`\rho _{ab}(T)`$ has been associated with the pseudogap and the onset of the deviation at $`T^{}`$ gives a characteristic temperature for the pseudogap . Figure 3(a) shows the plot which emphasizes the deviation from the $`T`$-linear behavior to determine $`T^{}`$. We should mention that this type of plot is subject to some arbitrariness and thus the errors in $`T^{}`$ are inevitably large. (Interestingly, in Fig. 3(a), even the optimally-doped sample shows a $`T^{}`$ which is well above $`T_c`$.) It has recently been recognized that there are two different characteristic temperatures for the pseudogap , and $`T^{}`$ corresponds to the higher characteristic temperature for the pseudogap. Also, it was proposed very recently that the peak in the $`T`$ dependence of $`R_H`$ may mark the lower characteristic temperature $`T_0`$ for the pseudogap , as does the NMR relaxation rate, ARPES, or the tunnelling spectroscopy . In our samples, the peak in $`R_H(T)`$ moves to higher temperatures as the carrier concentration is reduced, which is consistent with the behavior of the pseudogap. We note, however, that it is not clear whether the peak in $`R_H(T)`$ observed in the overdoped samples really corresponds to the pseudogap. For a more detailed discussion on the relation between the peak in $`R_H(T)`$ and the pseudogap, please refer to Ref. .
Figure 3(b) shows the phase diagram, $`T`$ vs $`1x`$, for our BSLCO. Note that the horizontal axis is taken to be $`1x`$ for convenience; this way, the left hand side of the plot corresponds to underdoping. One can see in Fig. 3(b) that a significant portion of the phase diagram has been covered and all the plotted temperatures, $`T^{}`$, $`T_0`$, and $`T_c`$, show good systematics with the carrier concentration. It is intriguing that the magnitudes of the characteristic temperatures for the pseudogap is quite similar to other cuprates , while the $`T_c`$ of the present system is the lowest among the major cuprates.
In summary, we present the data of the in-plane resistivity, Hall coefficient, and the Hall angle of a series of high-quality La-doped Bi-2201 crystals in a wide range of carrier concentrations. The normal-state transport properties of our Bi-2201 crystals show systematics that can be considered to be “canonical” to cuprates. The characteristic temperatures for the pseudogap were deduced from the data to construct a phase diagram.
|
no-problem/9906/cond-mat9906212.html
|
ar5iv
|
text
|
# Phase diagram of the integer quantum Hall effect in p-type Germanium
\[
We experimentally study the phase diagram of the integer quantized Hall effect, as a function of density and magnetic field. We used a two dimensional hole system confined in a Ge/SiGe quantum well, where all energy levels are resolved, because the Zeeman splitting is comparable to the cyclotron energy. At low fields and close to the quantum Hall liquid-to-insulator transition, we observe the floating up of the lowest energy level, but no floating of any higher levels, rather a merging of these levels into the insulating state. For a given filling factor, only direct transitions between the insulating phase and higher quantum Hall liquids are observed as a function of density. Finally, we observe a peak in the critical resistivity around filling factor one.
PACS numbers: 73.40.Hm, 71.70.Di, 72.15.Rn, 71.30.+h, 71.55.Jv, 73.20.Dx
\]
Two dimensional electronic systems, subject to a perpendicular magnetic field exhibit a large variety of phenomena. The most interesting one, is the quantization of the Hall resistivity, $`\rho _{xy}`$, in terms of the quantum unit of resistance $`h/e^2`$. In a pioneering work, Kivelson, Lee and Zhang (KLZ) , proposed a theoretical global phase diagram (GPD) of the quantum Hall effect expressed in terms of “laws of corresponding states”. Their phase diagram is composed of two different types of stable zero temperature (T) phases, the quantum Hall liquid phases and the insulating phase. In a quantum Hall liquid phase, the diagonal resistivity, $`\rho _{xx}=0`$, and the Hall resistivity $`\rho _{xy}=h/\nu e^2`$. (We will only consider the case in which the filling factor $`\nu `$ is an integer.) In the insulating phase $`\rho _{xx}`$ diverges at zero T. The different phases in KLZ’s GPD are a function of disorder strength and magnetic field (B). At high enough disorder only the insulating phase is present at finite B. At lower disorder, when different quantum Hall liquids exist, only transitions between the $`\nu =1`$ quantum Hall liquid and the insulating phase are allowed. This means, for example, that the $`\nu =3`$ state has to first undergo a transition to $`\nu =2`$ and $`\nu =1`$ before reaching the insulating phase. This rule relies on the “floating up” of the extended states at vanishing B, first discussed by Khmel’nitzkii and Laughlin . The “floating” argument was later questioned by recent experiments, in which direct transitions between $`\nu 3`$ states and the insulating phase have been observed .
During the past years, there has been a number of theoretical and experimental papers focusing on the existence of the “floating up” of extended states. Some experiments in n-type GaAs and in Si-MOSFET are consistent with the “floating up” scenario and other experiments in SI-MOSFET and in p-type GaAs are inconsistent with it. Theoretically, this issue is also under debate . To analyze this aspect further, we present new results for a different system, i.e., a two dimensional hole system (2DHS) confined in a Ge/SiGe quantum well , in which the lowest energy level “floats up” but no higher energy level.
Addressing this issue experimentally, involves extracting the zero T physics. A phase is insulating, when $`\rho _{xx}`$ diverges as $`T0`$ as opposed to the quantum Hall liquid phases, where $`\rho _{xy}h/\nu e^2`$ and $`\rho _{xx}0`$ with vanishing $`T`$. Therefore, the transition can be defined as the B-field, where the T-dependence of $`\rho _{xx}`$ changes direction . This method was later successfully used by several groups to determine parts of the phase diagram . We will follow the same method and discuss the differences compared to other methods in connection with our data.
In order to illustrate our procedure, we have plotted in fig. 1a, $`\rho _{xx}`$ and $`\rho _{xy}`$ as a function of $`B`$ for different T ’s, ranging from 0.4 K to 4.2 K. The insulating-to-quantum Hall liquid transition at $`B_C^L`$ can easily be identified from the crossing point of the $`\rho _{xx}`$ curves at different T ’s. With increasing $`B`$, $`\rho _{xy}`$ first develops a plateau at $`\nu =3`$, then at $`\nu =2`$ and finally at $`\nu =1`$. This is an example of a direct transition between the $`\nu =3`$ state and the insulating phase at $`B<B_C^L`$. At higher fields, the quantum Hall liquid $`\nu =1`$ state-to-insulator is shown in fig. 1b. Here again the transition at $`B_C`$ is obtained from the crossing point of the $`\rho _{xx}`$ traces. In both cases (fig. 1a and fig. 1b) $`\rho _{xy}`$ shows no features across the transition. In the case of $`B_C`$, $`\rho _{xy}`$ even remains quantized as was noted previously . In the inset of fig. 1b, we observe that the transition does not occur at the maximum of the diagonal conductivity, $`\sigma _{xx}`$. On the other hand, at $`B_C`$ and at $`B_C^L`$, the Hall conductivity, $`\sigma _{xy}`$, is T-independent. One could, therefore, alternatively use the T-dependence of $`\sigma _{xy}`$ to define the transition point , but this method has the disadvantage, that $`\sigma _{xy}`$ is not measured directly, because the conductivities are obtained by inverting the resistivity tensor.
When extracting the transition from the crossing point we imply that only two types of states exist in our system: a quantum Hall and an insulating state. The crossing point, which defines a $`T`$-independent point, can be reliably taken as the transition point. However, in recent experiments a zero-field metallic-like state in a 2D system was observed . In these systems, for high enough densities and at $`B=0`$ the resistance decreases and saturates with decreasing $`T`$, which invalidates the use of the crossing point to extract the boundary. Therefore, alternatively the peak in the conductivity was used to determine the transition . The question of the existence of only an insulating state at $`B=0`$ is even more fundamental. In our Ge/SiGe system the insulating behavior is well defined in the low density limit with an increase of $`\rho _{xx}`$ by more than an order of magnitude when $`T`$ is lowered from 1K to 0.3 K. This is in contrast to our highest density, where the increase in $`\rho _{xx}`$ is only a few percent for the same $`T`$ span. Are these two behaviors characteristics of the same insulator? If yes, our procedure is well defined. On the contrary, if these two regions belong to different zero-$`T`$ phases one could expect signatures at our experimental $`T`$. However, we do not observe any qualitative change of the $`T`$-dependence in our entire density range, which indicates a transition or cross-over to a metallic-like state at $`B=0`$. This is correlated with the existence of a well defined crossing point as a function of $`B`$. On the quantum Hall side we have the reverse $`T`$-dependence with a similar magnitude close to the transition.
The next step is to obtain the entire phase diagram, by varying the density and $`B`$. In most previous experimental phase diagrams the degeneracy of the Landau levels could not be resolved down to the insulating phase. Typically the Zeeman splitting was much smaller than the cyclotron gap . In holes there is an additional degeneracy of the valence band, which leads to a two band system, which affects the phase diagram . We avoid these difficulties because our 2DHS is confined in a strained Ge layer, where the strain removes the degeneracy of the valence band. The 150 Å thick Ge layer is sandwiched in between Si<sub>0.4</sub>Ge<sub>0.6</sub> layers, where Boron modulation doping is placed. The crucial property of this system is that the Zeeman activation energy is only 30 % smaller than the cyclotron activation energy. As a result, the Zeeman energy is resolved down to our lowest density so that we observe all even and odd filling factors. To vary the density we use a top gate with an insulating silicon oxynitride layer grown between the cap layer and the gate. By applying a gate voltage between 0 V and 6 V we could vary the density between $`n=0.55\times 10^{11}\text{cm}^2`$ with no leakage. The highest mobility we obtained was $`20\times 10^3\text{ cm}^2/Vs`$ at zero gate voltage. The measurements were performed in a dilution refrigerator and a He<sub>3</sub> system at T ’s ranging between 40 mK and 4.2 K, using standard AC lock-in technique with an excitation current of 0.1 nA. DC measurements were performed in order to check for consistency.
We determine the transition between two quantum Hall liquids from the local Max{$`d\rho _{xy}/dB`$} at 40 mK. This point is T-independent at low $`T`$ and the size of the dots are representative of the differences obtained from the 200mK and 400mK curves. However, other choices for determining the transition are possible, such as Max{$`\rho _{xx}`$} or Max{$`d\sigma _{xy}/dB`$} . When using these other methods we found that they give similar results. The insulator-to-quantum Hall transitions (solid dots) were extracted from the crossing points of $`\rho _{xx}`$ at three different T ’s (40mK, 200mK and 400mK), which also gives an estimate of the experimental inaccuracy represented by the size of the dots. The results are summarized in fig. 2.
At a fixed $`\nu `$ and with decreasing density, there are direct transitions between the quantum Hall phases $`\nu =1,2,3,\mathrm{}`$, and the insulating phase. This is the main result of fig. 2a and is in strong contrast to KLZ’s GPD, where the insulating phase is only bordered by the $`\nu =1`$ quantum Hall liquid. These results are very similar to those obtained in Si MOSFET , where a merging of the transition states into the insulating phase and an absence of floating was observed. However, in Si MOSFET each Landau level is doubly degenerate due to the additional valley band degeneracy. Therefore, the $`\nu =5`$ state corresponds to the second Landau level. Other results with Si MOSFET (using an arbitrary cut off value of the conductivity to determine the transition) were interpreted as consistent with the floating up scenario like in n-type GaAs/AlGaAs . p-type GaAs/AlGaAs has a similar valley band degeneracy, but no floating was observed and mainly even filling factors were resolved . An interesting result was obtained by Lee et al., who used a multiple GaAs quantum well and obtained similar transitions for high even filling factors. As an alternative method to determine the transition, the disappearance of the activated behavior was used in ref. . Transitions up to $`\nu =3`$ were observed in a similar Ge/SiGe system by Song et al..
To analyze the termination of the Landau levels as a function of B, we plotted in fig. 2b the same data as in fig. 2a, but as a function of B. At high $`B`$, the experimental points (open and solid dots) follow very precisely the $`n=(i+\frac{1}{2})\frac{eB}{h}`$ lines, where $`i=0,1,2,\mathrm{}`$. The insulating phase below the lowest energy level ($`i=0`$) can be understood in terms of simple energy level physics. The energy levels are broadened by disorder, which leads to a smooth density of states around them and to the formation of mini-bands. The states at the center of the mini-bands are extended, but the states with energies away from the band center are localized . In this way, at $`T=0`$ and when the Fermi energy ($`F_F`$) is below the lowest energy level, $`E_F`$ is pinned to the localized states and the system becomes insulating. The deviation from the linear dependence ($`i=0`$), when $`B`$ is decreased, is due to the disorder broadening. Lowering $`B`$ decreases the gaps and leads to an overlap of the energy levels. Therefore, some states from higher energy levels increase the density of states below the lowest energy level ($`i=0`$), which leads to the apparent “floating up” of the lowest energy level, when $`E_F`$ is pinned to the $`i=0`$ level. When assuming a symmetric broadening, the transition occurs at $`\nu =1/2+\nu _h`$. In this case, $`\nu _h`$ is simply the density of states of the higher energy levels (normalized to one energy level) integrated between zero and the $`i=0`$ level. Hence, $`\nu _h`$ is a measure of the overlap between the density of states of higher energy levels and the lowest energy level. Applying the same argument to higher energy levels does not hold, because the contribution of states from even higher levels is compensated by the states lost from lower levels. Therefore, broadening cannot affect the higher energy levels and, indeed, we observe no floating of higher energy levels. These higher energy levels simply merge straight into the insulating phase.
Before discussing our results in relation to existing theoretical ones, we need to make an important point. Most theories, starting with KLZ’s GPD, are a function of the B-field strength and the disorder strength. At first sight our system is very different, as we replaced the disorder axis by the density. One could argue that, when the density is reduced, screening becomes smaller. However, in two dimensions and at $`B=0`$ the screening is largely density independent. The situation is slightly different with a quantizing B-field because the screening becomes non-linear . In addition, screening is also $`T`$-dependent. On the other hand, it is the ratio of the disorder potential fluctuation over the quantizing energy, which can be effectively tuned by the density. Therefore, our results suggest that at strong level mixing (which can be estimated by $`\nu _h`$), the extended states at the center of each energy level disappear and localization takes over. This observation is in agreement with recent theoretical and numerical calculations. Fogler, for instance, showed that the levitation of extended states remains very weak even at low B . Using tight binding models, Liu et al. and more recently Sheng and Weng showed evidence for the disappearance of extended states without floating. Inspired by recent experimental results, a numerical phase diagram based on the tight binding model has been obtained . The similarity between their GPD and ours (fig. 2) is very striking.
We now turn to the analysis of the resistivities at the transition. In fig. 3, we have plotted the critical resistivities, $`\rho _{xx}^C`$ and $`\rho _{xy}^C`$ defined as the value of the resistivity at the quantum Hall liquid-to-insulator transition. This work was inspired by a recent letter by Song et al. , who obtained a similar curve for $`\nu >1`$. Our striking new result is the peak of $`\rho _{xx}^C`$ close to $`\nu =1`$ before $`\rho _{xx}^C`$ saturates around the quantized value $`h/e^2`$. In addition, $`\rho _{xy}^C`$ follows the classical expression for $`\nu >1`$ indicative of a Hall insulator, which becomes closely quantized for $`\nu <1`$, bordering the quantized Hall insulator. A similar peak in $`\rho _{xx}^C`$ was obtained very recently by Hanein et al. in systems exhibiting a zero B metal-insulator transition. But in their case $`\rho _{xx}^C`$ tends to $`h/e^2`$, with vanishing $`B`$, as opposed to zero in our case. The physical origin of this peak is not understood.
Summarizing, we have experimentally mapped out the phase diagram of the integer quantum Hall effect, as a function of density and magnetic field. At low fields and close to the quantum Hall liquid-to-insulator transition, we have observed the floating up of the lowest energy level, but no floating of any higher levels, rather a merging of these levels into the insulating state. These results are consistent with the disappearance of extended states without levitation due to the broadening of the energy levels. Along the transition, we observe a peak in the critical resistivity around filling factor one.
We would like to acknowledge M.M. Fogler, H.W. Jiang, S. Kivelson and D.N. Sheng for helpful discussions. This work was supported in part by the National Science Foundation.
|
no-problem/9906/astro-ph9906124.html
|
ar5iv
|
text
|
# Spectrophotometric Dating of Elliptical Galaxies in the Ultraviolet
## 1. Introduction
Recent observations on the origin of the UV upturn phenomenon of elliptical galaxies showed that hot HB and post-HB stars are the dominant UV sources in these systems (Ferguson et al. 1991; O’Connell et al. 1992; King et al. 1992; Bertola et al. 1995; Brown et al 1997). As demonstrated in our recent investigation (Yi et al. 1999), it is still not clear, however, whether the dominant UV sources are metal poor (“metal-poor HB model”; Lee 1994; Park & Lee 1997) or metal rich (“metal-rich HB model”; Bressan et al. 1994; Dorman et al 1995; Yi et al 1998). The “metal-poor HB model” suggests that the dominant UV sources are very old, hot metal-poor HB stars and their post-HB progeny, although metal-rich PAGB stars also contribute some UV flux. In this picture, nearby giant ellipticals are about 3 Gyr older than the Milky Way Galaxy, which would imply non-zero cosmological constant from time scale test. On the other hand, the “metal-rich HB model” suggests that the dominant UV sources are super metal-rich hot HB stars that experienced enhanced mass-loss and helium enrichment. If so, nearby giant ellipticals are similar in age to the Milky Way Galaxy, and there is no conflict with zero $`\mathrm{\Lambda }`$ cosmology from time scale test. Clearly, it is of considerable importance to understand the origin of the UV upturn. Considering this situation, we have recently investigated the evolution of UV flux with look-back time to see if two models predict substantially different evolutionary patterns. We found that this would indeed provide some strong observational test on the origin of the UV upturn phenomenon (Yi et al. 1999). The major uncertainty in these model calculations is the mass of PAGB stars, since it is only poorly constrained from direct observations. In this paper, we report our progress in investigating the effect of PAGB mass in the modeling of UV evolution of elliptical galaxies.
## 2. Variation of the Dominant UV source with Look-Back Tim
Our model calculations show that hot HB and post-HB stars contribute most of the UV light in nearby giant ellipticals (Fig. 1). As lookback time increses, however, PAGB stars are getting more dominat UV sources since the mean masses of helium burning stars are too high to be hot enough HB stars at younger ages. (At very large look-back times, hot main sequence stars eventually become the major UV source.) Since the contribution from PAGB stars overwhelms the UV spectrum at higher redshifts, the model SED is strongly affected by the current uncertainty of the PAGB mass as look-back time increases.
## 3. 1500Å-V Color Evolution of Giant Elliptical Galaxies
We present, in Figure 2, the 1500Å-V color evolution of giant ellipticals predicted by both the “metal-poor” and “metal-rich” HB models under two assumptions regarding the PAGB mass. It is clear from Figure 2 that the 1500Å-V is not strongly affected by the uncertainty of the PAGB mass for the nearby giant ellipticals since the PAGB contribution to the total UV flux is less than 10%. PAGB treatment becomes significant, however, for the ellipticals at $`z0.15`$. Fortunately, both “metal-poor” and “metal-rich” models predict more or less the same 1500Å-V colors at $`z0.35`$, and thus far-UV observations at these redshifts (cf. Brown et al. 1998) would provide crucial constraint on the PAGB mass. Our models in Figure 2 suggest that the observations at lower redshifts ($`0.15z0.30`$) would still provide strong constraints on two models on the origin of the UV upturn. Future observations of ellipticals from the STIS/HST and planned GALEX (Martin et al 1998) space UV facility will provide crucial database required for more concrete calibration of our UV dating techniques for old stellar systems.
## References
Bertola, F., et al. 1995, ApJ, 438, 680
Bressan, A., Chiosi, C., & Fagotto, F. 1994, ApJS, 94, 63
Brown, T. M., Ferguson, H. C., Davidsen, A. F., & Dorman, B. 1997, ApJ, 482, 685
Brown, T. M., Ferguson, H. C., Deharveng, J.-M., & Jedrzejewski, R. 1998, ApJ, 508, L139
Ferguson, H. C., et al. 1991, ApJ, 382, L69
King, I. R., et al. 1992, ApJ, 397, L35
Lee, Y.-W. 1994, ApJ, 430, L113
Martin, C., et al. 1998, BAAS, 29, 1309
O’Connell, R. W., et al. 1992. ApJ, 395, L45
Park, J.-H., & Lee, Y.-W. 1997, ApJ, 476, 28
Yi, S., Demarque, P., & Oemler, A. 1998, ApJ, 492, 480
Yi, S., Lee, Y.-W., Woo, J.-H., Park, J.-H., Demarque, P., & Oemler, A. 1999, ApJ, 513, 128
|
no-problem/9906/astro-ph9906420.html
|
ar5iv
|
text
|
# Structure in the First Quadrant of the Galaxy: an Analysis of “TMGS” Star Counts using the “SKY” Model
## 1 Introduction
Although star counts have long been used for the determination of Galactic structure (see Paul 1993), there are still a large number of parameters which are not well determined. Reviews of star count surveys can be found in Bahcall (1986), Majewski (1993) and Reid (1993). Many visible studies have been directed towards the Galactic poles where extinction in the Galactic plane can largely be ignored. Star counts, velocity distributions and metallicity have been used to determine and separate the various Galactic populations, such as the old disc, young disc and halo. Detailed reviews are given, for example, by Gilmore, Wyse & Kuijken (1989) and Freeman (1987), and references therein. Hence, the vertical structure of the local Galaxy is well measured (although there remain contentious issues such as the thick disc). The horizontal structures are less well studied. Robin, Crézé & Mohan (1992) obtained deep visible star counts in the anticentre region and concluded that the disc has a scale length of 2.5 kpc and a radial cutoff at about 14 kpc. However, apart from individual clear windows, visible star counts in the plane are limited to a maximum heliocentric distance of about 3 kpc (Schmidt-Kaler 1977), because of the high extinction. A method of overcoming this, at least in part, is to observe in the infrared, where the extinction is significantly lower. Infrared source counts detect sources to a greater distance and are far less affected by local dust, so the measured distribution of sources is closer to the true distribution.
To date, infrared surveys of the Galactic plane capable of detecting point sources have been limited. The largest IR survey to date was that performed by the Infrared Astronomical Satellite, $`IRAS`$, which mapped nearly the entire sky at 12, 25, 60 and 100 $`\mu `$m. Much work on Galactic structure has been enabled by $`IRAS`$ particularly on the distribution of very late type stars (see Beichman, 1987, for a review). However, at large distances, $`IRAS`$ could detect only extremely luminous sources (e.g. OH/IR stars) and was severely confusion-limited over much of the Galactic plane. The Two Micron Sky Survey (Neugebauer & Leighton 1969: hereafter TMSS) also had a low limiting magnitude ($`m_K3`$) and so could not detect sources in the inner Galaxy. Other surveys with greater sensitivity have tended to be restricted to the Galactic Centre (hereafter GC) region. Those that have, at least in part, examined the disc (e.g. Eaton et al. 1984; Kawara et al. 1982) have covered only very small areas of sky, typically a few hundred arcmin<sup>2</sup>. These surveys ran the risk of sampling an area with unusually high extinction, as well as suffering from low source counts, and were unable to show the large-scale distribution of sources.
Garzón et al. (1993) have described the Two Micron Galactic Survey (hereafter TMGS). This is a $`K`$–band survey with an approximate limit of $`m_K`$10 which represents a valuable addition to our knowledge of the near–infrared Galaxy, deepening the broad sweep coverage of the plane by a factor of almost 1000 compared with the TMSS. It fulfils an important transitional role, providing a homogeneous data base over a number of regions in the Galactic plane at a time when both the hemispheric DENIS (Epchtein 1998) and all-sky 2MASS (Skrutskie 1998) three–band surveys are in routine operation. These new surveys will go about 4–5 magnitudes fainter than TMGS and offer simultaneous observations in $`IJK^{}`$ or $`JHK^{}`$. Nevertheless, a $`K`$-band survey with a magnitude limit of $`m_K`$10 such as the TMGS can usefully probe the structure of our Galaxy.
Hammersley et al. (1994) and Calbet et al. (1995) explored the morphology of star counts from the TMGS by heuristic methods, while Hammersley et al. (1995) deduced the displacement of the Sun from the plane from the slight north–south asymmetry of counts crossing the plane. However, these papers are based principally in looking at the very obvious features seen in the the TMGS. In order to understand the $`K`$ star counts more fully, a detailed model is required which allows the total counts to be broken down into counts from the various Galactic geometric components in each magnitude range. The model that we have chosen to work with is described by Cohen (1994), who has enhanced the “SKY” model for the point source sky originally described by Wainscoat et al. (1992: hereafter WCVWS), adding molecular clouds for mid–infrared realism, extending the Galactic structure to model the far–ultraviolet sky, and displacing the Sun 15 pc north of the plane (Cohen 1995).
The purpose of this paper is to describe our use of the SKY model to provide a formal basis for interpreting the TMGS source counts. Our eventual objective is to refine the model, guided by any discrepancies that we find with the TMGS data base. To do this we contrast the TMGS data from almost 600 patches of sky (some 296 deg<sup>2</sup>) in the vicinity of the plane with the predictions of the SKY model, to investigate the stellar content and probable geometry of the Galactic plane. The advantage of this approach is that we can emphasize otherwise slight deviations from expectation, and thereby define those regions in which the TMGS has uncovered substantive evidence of structure contrary to that in a formal model. We present analyses with SKY of TMGS star counts along entire survey strips, from 10 to 30 in length, cutting the plane at a variety of longitudes. We conclude that SKY provides a very good overall description of $`K`$ counts in the plane, which is chiefly dominated by the exponential disc, but that the TMGS offers important insights into the character of structure very close to the Galactic plane and the nature of the bulge. We find four basic styles of statistically significant deviation of SKY’s predictions from the observed counts. One appears to arise from a significant constituent of the ring, or arm, populations not currently in SKY, and strongly confined in latitude close to the plane in the inner Galaxy (longitudes 20–31). A second type of behaviour suggests that the luminosity function (hereafter LF) of the inner bulge is not that implicitly represented by SKY. The third set of departures from an ideal model is caused by spiral arms that do not lie in the plane (cf. WCVWS; Cohen 1994). The remaining type of pattern is due both to off–plane dark clouds not included in SKY’s smooth analytic expression of the extinction, and lines of sight through the Galaxy which have unusually high or low distributed (rather than localized) extinction.
## 2 The TMGS and the SKY model
### 2.1 The TMGS
The TMGS (Garzón et al. 1993) is a collaborative project between the Instituto de Astrofísica de Canarias, Tenerife, and Imperial College, London, to map large parts of the Galactic bulge and plane. Observations were made between 1988 and 1995 on the 1.5-m Carlos Sánchez Telescope of the Observatorio del Teide, Tenerife. Some 350 deg<sup>2</sup> of sky have been mapped down to about 10 mag in $`K`$, resulting in a data set of about 700,000 sources. The main areas are in strips about 30 in RA (centred on the Galactic plane) by about 1.5 in declination. Most TMGS sources lie within a few degrees of the Galactic plane, the majority having no optical counterparts to at least $`m_V`$=16, implying that they are sources deep within the Galaxy. The star counts presented in this paper were made using all of the data in each area. Although the data were taken in strips in RA, the TMGS catalogue has been converted to Galactic coordinates and the counts made using 1 bins in latitude.
### 2.2 SKY: the Model
The original SKY model described in WCVWS incorporates six fully detailed Galactic components: an exponential disc; a Bahcall–type bulge; a de Vaucouleurs halo; a 4–armed spiral with 2 local spurs; Gould’s Belt and a circular molecular ring. Each geometrical element may contain up to 87 source categories, derived from detailed analyses of the content of the $`IRAS`$ sky (Walker et al. 1989). These comprise 33 “normal” stellar types; 42 types of AGB star, both oxygen- and carbon-rich; 6 types of object that are distinct from others only in their MIR high luminosity; and 6 types of exotica including H II regions, planetaries, and reflection nebulae. Every category has its own set of absolute magnitudes and dispersion in ten hardwired passbands, and an individual scale height and volume density in the local solar neighbourhood. Some sources are absent from some components (the arms and ring are made rich in high–mass stars; the halo is deficient in these).
We have used version 4 of SKY as detailed by Cohen (1994,1995) and Cohen et al. (1994), the principal attributes are summarized in Table 1. This is different from WCVWS in that it has the solar offset from the plane and reduced halo:disc ratio, however, we have not included the additional molecular cloud populations essential to obtain a high-fidelity representation of the mid–infrared sky. This omission is justified because the clouds were inserted by Cohen (1994) for $`IRAS`$ wavelengths of 12 and 25 $`\mu `$m, which are essentially insensitive to extinction. Therefore only the extra sources were included and not the extra extinction through the clouds, yet the extra populations associated with these clouds are surely accompanied by extra extinction not accounted for within SKY. Cohen et al. (1994) demonstrated in the far–ultraviolet that one could assign a value to the additional extinction encountered through local molecular clouds using SKY. Within the TMGS archive one can clearly recognize obscuring clouds beyond which no 2-$`\mu `$m sources are detected, although at $`K`$ only the dense cores cause a significant effect on the counts and so the effect of the extinction is generally limited to single TMGS zones. Consequently, without a detailed prior knowledge of the additional extinction along the specific TMGS lines of sight, it is preferable to model without including these molecular complexes at wavelengths susceptible to appreciable attenuation through clouds of $`A_V`$$`>`$10 mag. In this way one can more readily identify those lines of sight that intercept dark clouds at $`K`$. Furthermore, the $`IRAS`$ sources detected in the clouds had warm dust which makes the sources very luminous at 12 $`\mu `$m. This warm dust does not have as large an effect on the absolute magnitudes at 2.2 $`\mu `$m, so the sources will be detected at $`K`$ but will not be that much brighter than the sources without warm dust. Hence they will be more difficult to see against the background of disc or bulge sources. This is the sole modification made to the published description of SKY4.
### 2.3 Why Confront SKY and the TMGS?
The SKY model was originally built to reproduce the $`IRAS`$ 12- and 25-$`\mu `$m source counts. However, it has been successfully compared with observations at many other wavelengths including the far-ultraviolet (Cohen 1994; Cohen et al. 1994). The model has already been compared with DENIS $`K`$ star counts in small regions (Ruphy et al. 1997) for the magnitude range $`m_K`$=10–13. However, the comparison with the TMGS is possibly the most challenging for the model because its magnitude range, $`m_K`$=5–9, probes the brightest sources in the LFs of the geometric components such as the arms and bulge. At $`m_K`$, these LFs are largely unknown, particularly for the arms, but the contrast with the disc will be maximal so any discrepancy with the model will readily be seen. Further, the areas to be studied are long cuts through the arms and disc, providing nearby reference regions out of the arms. This, too, allows differences to be explored in a manner that is impossible given only a single region. The final important advantage is the relatively large area of sky covered which, in many regions, affords meaningful star counts to be produced as bright as $`m_K`$=5.
The fact that SKY produces good matches to $`IRAS`$ (Cohen 1995) and visible stars counts towards the NGP (WCVWS) does not guarantee that it will successfully match the TMGS star counts. The $`K`$ LF is dominated by photospheres whereas, for IRAS, the principal populations detected at significant distances are all dust-enshrouded. Clearly, the $`IRAS`$ sources will be detected by the TMGS but are many magnitudes fainter, greatly diminishing their importance. Thus, whilst SKY may get the LF of the bulge correct for $`IRAS`$, this does not necessarily imply correctness for the TMGS. Garzón et al. (1993) already noted that the Galaxy at 2.2 $`\mu `$m looks very different from that in the visible because of extinction in the plane. Furthermore, the principal sources detected by the TMGS are K4–M0 giants or M-supergiants, which are far less important in visible star counts when coupled with extinction.
## 3 Confrontation of the Model by the Observations
In the Appendix (Figures A1–A7, D1–D9), we present TMGS counts along nine strips in the first quadrant. These illustrate our interpretations of the counts by direct comparisons with the predictions of SKY. Differential counts were scaled per square degree per magnitude, using entirely independent TMGS data in 0.2-mag wide bins.
Table 2 summarizes for each strip: the J2000 declination; the longitude of the Galactic plane crossing; the bounding coordinates; the number of useful (i.e. with tolerable implicit Poisson errors) zones we present in the relevant figure; the typical area sampled at each position along the strip (this varies slightly along a strip); and the figure which illustrates the star counts and SKY predictions. Table 3 amplifies this information by indicating the closest approach of each line of sight to the Galactic Centre and which (non-disc) geometric components contribute to the predicted counts.
Our analyses depend on three basic analytical tools: differential star counts (Figures A1–A7, D1–D9); profiles of cumulative counts in a specific $`K`$-magnitude bin as a function of latitude along each strip (Figures B1–B9); and similar profiles created after subtracting the disc component predicted by SKY from the observed counts (Figures C1–C9). In both Figures B and C, we represent the observations by solid lines and the model counts by dashed lines.
Figures 1(a)-(h) present eight representative fields, taken from the montages shown in the Appendix, for which we explicitly plot the Poisson error bars appropriate to the original counts (which were scaled by 5, to yield counts per mag from the 0.2-mag binned data, and again, according to the actual area sampled, to yield counts per deg<sup>2</sup>). Figure 1 distinguishes the counts arising from separate geometric components as follows: solid line – total count; dotted or less heavy line – disc; long dashes – spiral arms, local spurs and Gould’s Belt; short dashes – bulge; short dashes and dots – molecular ring; long dashes and dots – halo. The figures were chosen to be representative of the various regions covered as well as showing specific features in their form which are discussed later in the paper.
As with all surveys, Malmquist bias can arise in TMGS and we see its effects in Figure 1(h) where the counts in the final few faint bins rise abruptly before the limit of completeness is attained. Note also that this is a region in which the actual source density is relatively low. In general, the direct comparison of SKY and TMGS counts indicates that TMGS is complete to a magnitude between 9.2 and 9.8, depending on the local surface density of sources.
## 4 Extinction
### 4.1 Off-Plane Dark Clouds
As with all previous versions of the model, SKY4 represents interstellar extinction simply by the product of two exponentials; one radial, with scale length of 3.5 kpc (matching that of the stars); one vertical, with scale height of 100 pc (matching that of the youngest stars still associated with dark clouds). This version of SKY does not include off-plane dark clouds so, if a TMGS strip runs through the core of such a cloud, the counts will be significantly decreased when compared with the model (however, the greatly reduced extinction at $`K`$ compared with the visible implies it requires many magnitudes of visible extinction to cause a significant effect on the TMGS counts). Figures A5,B5,D5 illustrates this phenomenon for the area centred on $`l=29.3^{}`$, $`b=4^{}`$ from the $`l=31^{}`$ strip. One sees that the TMGS counts fall well below the predictions of SKY yet, at $`b=2.5^{}`$ or $`b=5^{}`$, the model and the TMGS are in good agreement. Figure 1(a) enlarges the plot for $`b=4^{}`$, the deepest discrepancy, and adds the Poisson error bars. This location corresponds to a known off-plane cloud which clearly shows up in emission in longer-wavelength $`COBE`$ maps, whereas the area is almost devoid of visible stars. In fact, the core of this cloud does not fill the TMGS bin at this location and sampling at higher resolution would decrease the TMGS counts even more at the centre of the cloud. The only potential H I counterpart (Weaver 1998, priv. com.) is a fairly uniformly dark patch (i.e. an absence of H I) well-centred in an H I emission ring, at an LSR velocity of $``$15 km s<sup>-1</sup>, and about 1–1.5 in diameter. Figure 1(b) shows another example, this time of a more localized off-plane cloud, whose influence is recognizable by an abrupt change in the observed source counts fainter than a specific magnitude. Here one sees an offset in source counts by 0.2 dex for $`m_K>`$8.5. Analysis of this line of sight by SKY indicates that the dominant contributors at this magnitude are disc K0-3 giants, at a probable distance of about 700 pc. Consequently, we suggest that one encounters a cloud with incremental $`A_V`$ 5 mag in this direction and at this distance. Again, H I maps (Weaver 1998, priv. comm.) show a sizeable void, bigger than that seen above (about 2 in diameter), well-centred on this TMGS field, in an emission ring at an LSR velocity of +38 km/s. In both cases, the presence of H I emission with central voids suggests that the hydrogen in these clouds is largely molecular rather than atomic.
### 4.2 On-Plane Dark Clouds and General Extinction in the Plane
The TMGS reveals a number of “on-plane” ($`|b|<2^{}`$) dark clouds, particularly towards the inner Galaxy. Hammersley et al. (1996) discussed the dark clouds which cause the major loss of stars at $`l=1^{}`$, $`b=0^{}`$, and showed that most of the extinction comes from individual dark clouds close to the GC. However, there are individual clouds on, or very near, the plane in many TMGS areas which locally cause a noticeable loss of counts ($`l=7^{}`$, $`b=1^{}`$ or $`l=31^{}`$, $`b=1^{}`$: Figures A2, B2, D2, A5, B5 and B5)
SKY includes only the general extinction along the line of sight. This is modeled as a double exponential which continues all the way into the GC but is normalized in the solar neighbourhood. Figures B1– B9 show the cumulative TMGS and SKY star counts (the latter separated into total and disc contributions) for various strips, as a function of Galactic latitude, and the ratios of these totals at different limiting magnitudes. The effect of extinction on SKY’s counts is clearly visible in these cuts as a dip in counts at $`b=0^{}`$. For lines of sight with $`l>30^{}`$, the dip on the plane is small, a few percent of the total counts, and there is very good agreement with the TMGS star counts. In the inner Galaxy, however, SKY predicts a major loss of counts on the plane, principally due to extinction in front of the bulge sources (the difference between SKY’s total and disc curves for $`m_K>7`$ and the $`l=7^{}`$ and $`1^{}`$ strips is almost totally due to the bulge). The model does reasonably well near $`l=1^{}`$, $`b=0^{}`$ although, as already stated, most of the extinction appears to come from identifiable dark clouds close to the GC, not represented in SKY. However, SKY substantially over-predicts the extinction on the plane at $`l=7^{}`$, $`b=0^{}`$. The ratios of model-to-observed star counts for this region show a narrow peak which the original plot indicates is due to extra extinction in SKY.
Hammersley et al. (1998) show that the extinction at the radius of the molecular ring ($`0.4R_{}<R<0.5R_{}`$) is about a factor of three above that predicted by a simple exponential model yet SKY’s analytical function for the extinction in the outer Galaxy (e.g. R$`>`$ 0.5R) gives a reasonable representation for the distributed extinction in the plane. Therefore, the only explanation for why SKY so significantly overestimates the extinction in the line of sight towards $`l=7^{}`$, $`b=0^{}`$, is that the extinction for $`R>0.4R_{}`$ drops to almost zero until a few hundred pc from the GC, where the features associated with the Centre begin (e.g. the molecular disc). The idea of there being a hole in the extinction in the inner Galaxy is not new: the CO maps clearly show a significant fall-off in density inside the molecular ring (Clemens, Sanders & Scoville 1988; Combes 1991). Freudenreich (1998) also needed a hole in the extinction in the inner Galaxy to match the DIRBE surface brightness maps.
## 5 The Exponential Disc for $`b>5^{}`$
SKY incorporates a single exponential disc with 87 categories of point source, each with its own scale height, colours, and local space density. The scale length of all sources is set to 3.5 kpc. SKY does not include a thick disc for several reasons. Firstly, it provides excellent fits to counts at a wide variety of wavelengths without using a thick disc. Secondly, the proportion of disc counts contributed by the thick disc, even at optical wavelengths where metallicities and kinematics may help separate these two populations, is minimal ($`<2\%`$; e.g. Gilmore, Wyse & Kuijken 1989). Thirdly, the stars of the thick disc contribute negligibly in the infrared, far below the levels of Poisson errors of the TMGS (and many other) data sets.
SKY separates young and old components by assigning the young population (OBA dwarfs and supergiants) a significantly smaller scale height, 90 pc, than the 325 pc of the old populations. Hence, the vast majority of the sources that SKY predicts that the TMGS should see in the exponential disc are the old K- and M-giants for regions well away from the bulge and more than $`5^{}`$ from the plane. The relevant figures are those for the strips at $`l=21^{},31^{}`$ and 65 (Figures A3, A5, A7, D3, D5 and D8). Figures B3, B5 and B8 show the associated cumulative counts for the TMGS and the model for these regions, to $`m_K=9`$.
Both the form of the curves and the differential star counts show that SKY gives a good fit to the TMGS at all magnitudes in these longitudes, for $`|b|>5^{}`$.
In most areas of the sky, the disc dominates the star counts, making the remaining geometric components more difficult to identify and analyse. However, the accuracy of the fit of SKY’S disc to the TMGS counts makes it entirely reasonable to subtract the model’s disc counts from the TMGS counts. This highlights the distribution of sources in the arms, ring, and bulge, if accomplished with care.
Alternatively, some argue that an exponential disc is no longer justifiable at the lowest latitudes in the Galaxy, and that the sech$`{}_{}{}^{2}z`$ form might be physically more defensible. There are a number of papers comparing the sech$`{}_{}{}^{2}z`$ and exponential in the discs of external galaxies (e.g. van de Kruit 1988; De Grijs, Peletier & van de Kruit 1997). However, the results, even in the infrared, are complicated by extinction when close to the plane of the galaxy. The sech$`{}_{}{}^{2}z`$ function asymptotically approaches an exponential as stellar densities fall far from the plane but, for latitudes within about 10 of the plane, it fails utterly to reproduce the observed counts. We illustrate this failure in Figure 2. The $`l`$=65 strip is one in which we see only disc and arms, the latter falling almost two orders of magnitude below the counts from the former. Thus, it is an ideal location to experiment with a sech$`{}_{}{}^{2}z`$ distribution function. Figure 2 shows the cut through the plane at $`m_K`$=9 for two model discs for $`l`$=65, $`|b|15^{}`$, one is exponential in height (this is a simplified model based on SKY but it is not the SKY model) and the second is identical except that it has a sech$`{}_{}{}^{2}z`$ height distribution. We have normalized the sech$`{}_{}{}^{2}z`$ curve to the wings of the cumulative latitude profile in the regions with $`|b|>10^{}`$ as both functions work well, away from the plane. Yet, for all inner latitudes, the sech$`{}_{}{}^{2}z`$ function delivers no more than half the requisite counts. This is equivalent to saying that, everywhere in the disc, even in those locations where spiral arms and other components do not make appreciable contributions to counts, more than 50% of the observed stars cannot come from the disc. Were we to have carried out this test at other TMGS positions in the plane, the excess of counts above the sech$`{}_{}{}^{2}z`$ law would be even larger than that shown in Figure 2.
We further note that the SKY model produces good fits to the DENIS $`K`$ star counts in the plane to $`m_K`$=13.5 (Ruphy et al. 1997), significantly deeper than the TMGS. In regions away from the Galactic bulge, the contribution from the disc will be at least an order of magnitude greater than from any other component, e.g. the arms. If the disc were not exponential, or at least very close to exponential, all the way to the plane, then SKY would significantly overestimate the counts on the plane.
The $`sech^2z`$ function is strictly valid only for an isolated, self-gravitating, isothermal system. Self-consistent dynamical models of the disc (e.g. Haywood et al. 1997) show the dependence of the resulting z-distribution on star formation rate (SFR). For all but declining SFRs, the vertical distributions can mimic exponentials. It is our suspicion that the $`sech^2z`$ function from a locally isothermal disc achieved its popularity in the context of overall light distributions across the planes of edge-on spirals at optical wavelengths (e.g. van der Kruit & Searle 1982). In such a situation, heavy and patchy extinction militates against any meaningful comparison between a $`sech^2z`$ law and observations and, far from the planes of such galaxies, one cannot distinguish between $`sech^2z`$ and exponential functions. Perhaps the only redeeming feature of a $`sech^2z`$ law is that it offers a null second spatial derivative at zero latitude, yet this is precisely the latitude range in which optical observations are invalid for the purposes of making such comparisons. Further, even in the field of edge-on external galaxies, there is evidence that the infrared light distributions follow exponential rather than $`sech^2z`$ functions (e.g. Wainscoat et al 1989).
## 6 The plane of the disc
For $`|b|<2^{}`$, at all longitudes, the TMGS has more counts on the plane than are predicted by SKY. The profiles of counts (Figure B) show three distinct distributions at these latitudes:
* A distribution extending from about $`|b|<4^{}`$ that is seen at all longitudes between $`l=70^{}`$ and $`1^{}`$.
* A spiked distribution about 1 wide centred on the plane at $`l=21^{}`$ and $`27^{}`$.
* The bulge at $`l=1^{}`$ and $`l=7^{}`$ (which we treat in section 9).
### 6.1 The Spiral Arms
SKY predicts that the 2-$`\mu `$m star counts should form a continuous distribution from $`b=15^{}`$ to $`b=1^{}`$ with the only serious deviations due to extinction in the plane. However, the scans $`l=21^{},31^{}`$ and 65 all show a distinct transition at about $`|b|=5^{}`$. (These can be seen both in the cuts \[Figures B3, B5 and B8\] and in the counts \[Figures A3, A5, A7, D3, D5, D8 \].) The star counts for $`|b|>5^{}`$ approximate an exponential very well but, at $`b=5^{}`$, it is clear that a second contribution is required with a far smaller scale height than typifies the stars of the disc (i.e. an element rising more rapidly than the disc as latitude decreases). Figure 1(e) represents this phenomenon from the $`l`$=65 strip, some 3.5 off the plane.
A feature so close to the plane has to be due to a young component. There are two possibilities to explain this distribution; a young disc or spiral arms. SKY already has a young component within its representation of the disc and these young stars (O, B and A) have a much smaller scale height than the older stars. Thus, if the extra sources in the plane were young disc stars, this would imply an enhancement in these populations. Indeed, examination of Figures B and C shows that, particularly at the brighter magnitudes, the numbers of young sources would need to be increased by orders of magnitude to reproduce the counts near the plane. This would create a major distortion of the LF, clearly recognizable in the local vicinity. Figure C shows the TMGS counts after subtracting the SKY disc. Note that, for both $`l=31^{}`$ and $`l=65^{}`$, the width of the non-disc source distribution is independent of magnitude, suggesting that these sources are grouped at a specific distance rather than spread along the line-of-sight. For the above reasons, we prefer to attribute the excess of stars on the plane to the arms, rather than to a boost in the young disc.
The prediction from SKY is that the arms are always substantially below the disc counts, although there are certainly locations in which the arms’ contribution is second only to that of the disc. However, particularly at the brighter $`m_K`$ magnitudes ($`m_K<7.5`$) the TMGS finds that the counts from the arms are at least as important as those from the disc within 2 of the plane. The two major TMGS regions (i.e. with sufficient area to give reliable statistics at the brightest magnitudes) that run only through spiral arms and through none of the other inner Galaxy components are at $`l=31^{}`$ and $`l=65^{}`$. At $`l=31^{}`$, $`b=0^{}`$ at $`m_K`$=7, the TMGS has nearly three times SKY’s predicted counts. At $`l=65^{}`$ at $`m_K`$=7, the excess on the plane is only 50% but at $`m_K`$=5 the TMGS is again a factor of 3 too high. This can be clearly seen in Figure B.
By $`m_K`$=9 in both the $`l=31^{}`$ and 65 strips the arm-to-disc ratio has dropped significantly. In the range $`8<K<9`$ at $`b=0^{}`$ there are some 331 predicted disc sources compared with only 40 arm sources. At $`l=31^{}`$ the loss of contrast is not as great (540 disc sources to 350 arm sources) probably because this line of sight is almost tangential to the Scutum arm, looking into the inner Galaxy, whereas the $`l=65^{}`$ line of sight roughly maintains a Galactocentric distance of R for over 8 kpc.
Another interesting conclusion can be reached by detailed comparison of the star counts at $`l=31^{}`$ and 65. In Figures C5 and C6, the dominant contribution to the disc-subtracted counts is from the arms. At $`m_K=5`$, we see roughly the same absolute number (within a factor of 1.5) of luminous non-disc sources in both directions yet, by the fainter magnitudes (e.g. $`m_K=9`$), the $`l=31^{}`$ line of sight shows about 6 times as many spiral arm sources as towards $`l=65^{}`$. Thus, the LF of the spiral arms must be a function of Galactocentric distance, rather than constant as in SKY4. If this is truly the case, one might suppose that metallicity was implicated. There is a precedent for this argument in the observed confinement of the latest WC-type Wolf-Rayet stars to the immediate environs of the GC, a phenomenon attributed by Smith & Maeder (1991) to the high metallicity near the Centre which biases evolution of massive stars to the formation of WC9 stars.
At $`l=65^{}`$ (Figure C6), only the Perseus arm is crossed and at a Galactocentric distance of about 8.5 kpc. Here the excess is around 80 sources deg<sup>-2</sup> to $`m_K`$=8. It is noticeable that the peak is not at $`b=0^{}`$; rather the centre of the distribution is shifted some 1.5 to positive latitude. Also, the observed counts drawn from the $`l=37^{}`$ strip are parallel to, but above, SKY’s predictions at all magnitudes. Both features are symptomatic of directions in which the spiral arms do not lie in the formal Galactic plane but arch above it. This same effect was detected in the IRAS counts by WCVWS, who were able to match it to specific regions in the H I sky (e.g. Weaver 1974) where neutral gas lies almost 1 kpc off the plane. They cited, in particular, the $`l=50^{}`$ to 90 region, which includes the off-plane peak of TMGS $`K`$ counts along the $`l=65^{}`$ strip.
Figure 3 shows the peak source count detected in each strip after the SKY disc has been subtracted. Vertical dashed lines are the points tangential to the Scutum and Sagittarius spiral arms. The aim is to show primarily the arm sources on the plane, therefore, at $`l=7^{}`$ and $`l=1^{}`$ only sources to $`m_K=6`$ have been shown to avoid bulge sources. Also because of the spike (Section 6.2) the values at $`l=21^{}`$ have been estimated as if the spike were not there and the region at $`l=27^{}`$ has been ignored. The $`l=31^{}`$ and $`l=37^{}`$ lines of sight are very close together, yet the $`l=31^{}`$ region has almost twice the number of arm sources as at $`l=37^{}`$. The only significant difference is that the $`l=31^{}`$ direction crosses the Sagittarius and Scutum arms whilst $`l=37^{}`$ intercepts only the Sagittarius arm. The extra sources to $`m_K`$=7 and 8 suggest that just under half the excess sources can be attributed to the Scutum arm (i.e. about 350 sources deg<sup>-2</sup> to $`m_K`$=8 on the plane). The effect of the arms on the counts can be clearly seen in all regions, in particular the influence of the number of arms which cut the line of sight. Directions crossing two arms have nearly twice the counts of those crossing only one. Furthermore, the inner spiral arms are far stronger than the arms at the distance of the solar circle (cf. the difference between $`l`$=65 and the other regions). Another curious feature of Figure 3 is that at $`m_K`$=5 and 6 the number of spiral arm sources does not alter significantly with longitude between $`l`$=$`1^{}`$ and $`l`$=31. Clearly there is a strong interplay between the distance to the arm and the LF but one would expect the $`l`$=$`1^{}`$ and $`l`$=$`7^{}`$ cuts to be significantly less than the $`l`$=31 which is almost tangential to the Scutum arm.
We attribute these excess sources observed near the plane by the TMGS to the presence of an additional population in the arms, not present in SKY4. From the estimated distances to these spiral arms, this population would have to be of supergiant luminosity. Consequently, the spiral arms clearly provide an important contribution to the $`K`$ star counts, particularly towards the inner Galaxy.
### 6.2 The “Spikes”
The most dramatic deviations from the SKY model anywhere in the area covered here are the “spikes” on the plane at $`l`$=21 and $`l`$=27. Figures A3 and A4 portray the differential star counts at $`l`$=21, $`b`$=0, and $`l`$=27, $`b`$=0, where the excess is substantial at all magnitudes. If the SKY disc is removed (Figures C3 and C4), then the TMGS has 6 times the counts of the remaining SKY components at $`l`$=21. In the cuts (Figures B3 and B4), it can be seen that the distribution of the excess counts is very narrow, about 1 wide, hence the term “spike”. It is significant that the spike only appears at $`l`$=21 and $`l`$=27, there is no evidence for it at $`l`$=31 (or larger longitudes). Neither is there evidence for the spike at $`l`$=7 or $``$1 although the arms are clearly visible at the brighter magnitudes in both strips so the absence of the spike cannot be just a line of sight effect. This almost rules out the feature being due to a continuous feature circling the inner Galaxy. The implications of the spikes were discussed by Hammersley et al. (1994), who argued that these must be features associated with the inner 3–4 kpc of the Galactic disc. The scale height must be around 40 pc, suggestive of a very young population, and the $`M_K`$ must attain at least $`11`$ to $`13`$. Mikami et al. (1982) suggested that there is a clustering of supergiants at $`l`$=27, $`b`$=0. Garzón et al. (1997) obtained visible spectra of some of the brightest TMGS sources with optical counterparts, proving that there are a very high number of supergiants in this region. There are two possible explanations for what causes these spikes: bar or ring.
In order to match the $`IRAS`$ 12- and 25-$`\mu `$m counts in this region WCVWS introduced another component, the molecular ring (Dame et al. 1987). However, it is clear that, even with this representation of the ring, SKY4 is well below the TMGS counts. The most significant deviations are confined to $``$b$``$$``$1.5 on the l=21 strip, but persist to higher latitudes along the l=27 strip ($``$b$``$$``$3.5). Note that the slope of the observed counts in these regions more closely parallels that of SKY’s ring component than the steeply sloped disc counts. Were we to posit that an additional stellar population attends the ring, close to the plane, then this population must contribute some 5 times SKY’s predicted ring counts near l=21 but almost an order of magnitude more than SKY’s ring near l=27. The excess of sources above SKY’s expectation is very similar on the l=27 and 21 strips, but shows a conspicuous additional bias toward very bright sources near l=27.
We note that Ruphy et al. (1997) describe an analysis of DENIS near-infrared counts using a new version of SKY, with a non-uniform elliptical ring that replaces the old circular molecular ring in SKY4. In a later paper we plan to exercise SKY5 (after implementing all the changes discussed in the present paper) to investigate what TMGS counts tell us about the geometry of this ring. In brief, DENIS counts probe the lower-density portions of this ring and help to constrain the overall geometry of the ellipse. But TMGS offers direct lines of sight across the first quadrant dense ansa included in SKY5. Given sufficient TMGS strips, and new deeper $`JHK`$ images, the density enhancements around this new ring should be definable. The inclusion of these star-rich ansae provides SKY with a measure of asymmetry (triaxiality) not present in earlier versions, and can be used to provide a more formal examination of the bar. However, we defer this discussion to a future paper using new data and further developments of SKY.
Hammersley et al. (1994) suggested that such spikes could be associated with a possible bar distribution on the grounds that, where the bar interacts with the disc, there is a high probability of strong star formation. Note this “bar” is not the triaxial bulge, often called a bar, but rather is a stick-like bar with a half-length of nearly 4 kpc and an angle of 75 to the GC-sun line. Neither can the star formation be related to the triaxial bulge somehow triggering star formation in the inner disc (as suggested by Freudenreich 1998) as the geometrical constraints alone make this nearly impossible. Hammersley it et al. (1994) also examined the $`COBE`$ 2.2-$`\mu `$m surface brightness maps and argued that such a bar far more simply and naturally predicts the observed features (both images and star counts), whilst the form of a ring would have to be contrived using ad hoc solutions to make it fit.
It is beyond the scope of the current paper to engage in a discussion of ring versus bar. To make any significant advance, new data are required. However, it should be noted that many external bars have strong ansae of star formation at either ends and so the discussions in the previous two paragraphs are not mutually exclusive.
## 7 The Bulge
Two TMGS strips run through the bulge: those at $`l`$=$`1^{}`$ and $`l`$=7. However, when examining these two regions, one must remember that the strips cut the plane at an appreciable angle so there is a significant variation in $`l`$ as well as $`b`$, which is important this close to the GC. These are the most complex regions to analyse because disc, arms, and bulge contribute to the counts. The first step is to subtract the SKY disc to highlight the bulge population. This is entirely reasonable because SKY gives a good fit to the exponential disc, and because of the high bulge-to-disc ratio fainter than $`m_K`$=7 (e.g. near the plane, there are twice as many bulge stars as disc stars).
Figures B1 and B2 show the cuts for the two strips; Figures C1 and C2 show the residuals after subtracting SKY’s disc from the TMGS counts. At $`m_K`$=5 and 6 the residuals in the TMGS are clearly due to spiral arm sources because they are as narrow in spatial extent, and as large in size, as the residuals at $`l`$=31. However, at $`m_K`$=6, SKY’s predicted residuals clearly show a wider distribution than seen in the TMGS, due to the predicted bulge sources. Even at $`l`$=7 the SKY residual is far wider than seen in the TMGS, yet by $`m_K`$=9 the residuals are far closer together. The explanation for these discrepancies is that SKY’s bulge LF has too many of the most luminous sources. A quantitative analysis of the $`K`$ LF, based on the TMGS, is presented in López-Corredoira et al. (1997,1999) and shows that SKY’s bulge LF starts 1 to 2 magnitudes too bright at $`K`$. Consequently, it is planned to implement a new LF for the bulge and to re-examine this issue in the future.
The SKY bulge is axially symmetric, with a projected axial ratio of 0.6. Therefore, if one were to draw ellipses of constant predicted bulge star counts, then these would have a ratio of major-to-minor axis of 0.6; i.e. all positions on the locus $`\sqrt{0.6^2l^2+b^2}`$ would have the same star counts. The bulge is the only component which has this type of projected spatial distribution and so plotting this function is an unambiguous test of whether the TMGS is actually detecting bulge sources. It should be noted that as the function $`\sqrt{0.6^2l^2+b^2}`$ is always positive this means that the line for a TMGS strip will double back on itself as it crosses the plane. Figure 4 shows SKY’s bulge counts plotted against $`\sqrt{rat^2l^2+b^2}`$ (right column) together with the TMGS strips (left column) through $`l=7^{}`$ (dashed line) and $`l=1^{}`$ (solid line) for various ratios, $`rat`$, between 1.5 and 0.2. In both cases this is after subtracting the SKY disc and for $`m_K`$=9 so that the bulge will dominate all the remaining components. More than a few degrees from the GC, the best fit for the bulge (i.e. the lines at the same position most nearly overlie each other is for a projected axial ratio of 0.6). There is some evidence from the central few degrees (Figure 4, $`l`$=$`1^{}`$) that the fit is better for an axial ratio of 1.0, but this is consistent with a triaxial bulge (as will be discussed below), which would depress the portion from negative longitudes.
Figure 5 shows the same type of plot as Figure 4, this time varying the $`K`$-magnitude at a fixed ratio of 0.6. The left column shows the TMGS whilst the right column shows the ratio of (TMGS–SKY disc) to (SKY total–SKY disc). At $`m_K`$=5, where there is no bulge component, there are two separate peaks caused by the spiral arm component in the plane. Spiral arms clearly will not follow the bulge locus. At $`m_K`$=6, SKY (as discussed above), predicts that there should be a bulge-like component but the TMGS plots show that the only significant population is that of the arms. For this reason, apart from the plane crossings, the TMGS is below the counts predicted by the model. $`m_K`$=7 is a transitional magnitude where the bulge begins to become important, although in the plane there is still a significant arm component. On average, the TMGS residuals are below the predicted SKY residuals; i.e. the TMGS bulge counts are below those of SKY. On average at $`m_K`$=8 and 9 away from the plane, the ratio of SKY to TMGS is about 1 suggesting that by these magnitudes the SKY LF is close to that observed by the TMGS. Note that the peak in the ratio where the TMGS strips cut the plane (where the curves for each strip double back on themselves) at $`m_K`$=8 and 9 is caused by the model overestimating the extinction in the inner Galaxy (see Section 4.2).
SKY predicts that the bulge distribution in the two cuts should be asymmetric due to the change in $`l`$ as well as $`b`$. However, a comparison of the TMGS residuals, taking account of the discrepancy in the LF, shows that for $`b<0^{}`$ ($`l>0^{}`$ by comparison with the point at $`b=0^{}`$) the TMGS is invariably above SKY, whereas for $`b>0^{}`$ ($`l<0^{}`$) the TMGS is well below SKY (Figures B1, B2, C1 and C2). It should be noted that the effect is very obvious particularly at $`m_K`$=7 and 8 and so we are not discussing some subtle feature in the counts. Initially it was supposed that this was due to the extinction by major dark cloud systems above the Galactic plane in this direction. However, recent studies have shown that in general there is insufficient extinction in these clouds to account for the observed effect we see (Mahoney 1998). Typically, the additional $`A_V`$ above the plane, as opposed to below it, is 1 mag (i.e. an extra $`A_K`$ of about 0.1 mag). The asymmetry seen in the TMGS would require a difference of extinction in $`K`$ of at least 0.5 mag, which is impossible to attribute to extinction alone. Figure 5 shows that at $`l`$=$`1^{}`$ (the solid line) the difference between the TMGS/SKY ratio for $`b<0^{}`$ ($`l>0^{}`$, the upper part of the solid line) and $`b>0^{}`$ ($`l<0^{}`$) increases with distance from the Galactic plane (between 2 and 7 on the x-axis). However, the total extinction, as well as the difference in extinction above and below the plane, must decrease away from $`b=0^{}`$. Hence, if the effect were due to extinction both the upper and lower parts of the solid line would come together with increasing distance from $`b=0^{}`$, and not diverge as is seen. Furthermore, if the differential extinction above and below the Galactic plane were significant, it would clearly show in the $`COBE`$ 2.2-$`\mu `$m surface brightness maps, which is not the case (see Freudenreich 1998).
A closer analysis of the difference between SKY and the TMGS shows that, at $`m_K`$=8 and 9 (where the LFs are in closer agreement), the TMGS is above SKY for $`l>0^{}`$ and below SKY for $`l<0^{}`$. Furthermore, at $`m_K`$=7, where the asymmetry in the TMGS is the greatest, the distribution is more bulge-like for $`b<0^{}`$ (i.e. greater $`l`$) than $`b>0^{}`$, where the distribution is almost narrow enough to be due solely to spiral arms, particularly in the $`l=1^{}`$ strip. A number of papers have argued that the bulge is triaxial with the near end in the first quadrant at an angle of 12 to 30 to the GC-Sun line of sight (Binney et al. 1991: Dwek et al. 1995: Freudenreich 1997: López-Corredoira et al. 1997). This would on average make the bulge sources at $`l>0^{}`$ closer than at $`l<0^{}`$. If the bulge were triaxial, we would expect asymmetries in the TMGS star counts. The sources from the closer part of the bulge ($`l>0^{}`$) would appear at a significantly brighter magnitude than those with $`l<0^{}`$. A rough estimate suggests a magnitude difference as high as 0.5 mag. The steep rise in the bulge LF for the brightest sources means the counts cut on very sharply around $`m_K`$=6.5, but then rise less steeply from $`m_K>`$7.5. Thus we would expect to see a very asymmetric distribution in the $`m_K`$=7 cut but a far more symmetric one at $`m_K`$=9, precisely as seen in the TMGS counts. Similarly, the ratios shown in Figure 5 have the features that one would expect when ratioing star counts from a triaxial bulge with those from an axisymmetric model; i.e. the more positive the longitude, the more the counts rise above the predicted values and the more negative the longitude, the more the counts fall below the predictions. $`K`$ star counts give access to the LF so we suggest that an analysis of the bulge based on star counts is likely to be far more sensitive than one based solely on surface brightness maps because it avoids the ambiguities in the size and orientation of the triaxial bulge inherent in an analysis of its surface brightness (Zhao 1997). Recently López-Corredoira et al. (1997,1999) have analysed the TMGS star counts by inverting the fundamental star count equation - very different from the qualitative analysis presented here - and have found that the bulge is triaxial with an angle to the GC-Sun line of 12.
For the innermost strip ($`l=1^{}`$) almost sampling the GC, the latitude range in which there is statistically significant deviation from SKY that cannot arise from off–plane dark clouds, is confined to $``$b$``$$``$2.0 (Figures A1, B1 and D1). From these plots, it is apparent that, if the bulge component were simply enhanced by factor of 2, the total predicted curve would much more satisfactorily match the counts. This bears directly on the bulge LF implicit within SKY. One can see from Figure 1(g) how it is possible, quite directly, to probe this LF because the observed counts are well-matched by SKY except in the restricted range $`8.4>m_K>6.2`$. TMGS counts exceed SKY’s here by about 50% at most, but this is naturally attributable to the shape of the bulge LF which declines abruptly at bright $`K`$ magnitudes. We note, however, that there is some interplay between changes in the density function and changes in the bulge LF.
## 8 Conclusions
We conclude that, overall, SKY provides a very reliable simulator of the TMGS Galaxy. Only rarely are the predictions wrong by as much as 50% and, in all these cases, there is a fairly obvious change that can be implemented in SKY to match the observations by more realistically representing the astrophysical situation. Consequently, the wide-area coverage offered by TMGS provides substantive constraints on the model. This potential interaction between SKY and the real Galaxy is extremely valuable in the development of a better model and in better understanding the structures in the inner Galaxy and their relative importance to $`K`$ star counts.
Specific changes that we plan to investigate in more detail and to implement in a new version of SKY are: the inclusion of an additional population of supergiants in the cores of the spiral arms; a modified LF for the bulge; a triaxial bulge; more non-coplanar arms and plane than represented by WCVWS; and the non-uniformly dense elliptical ring.
Perhaps the single most important result is that the star counts at 2.2 $`\mu `$m can clearly be modelled using simple analytical functions to describe the 4 Galactic components of disc, arms, bulge and ring.
## 9 Acknowledgments
The Carlos Sánchez Telescope is operated on the island of Tenerife by the Instituto de Astrofísica de Canarias in the Spanish Observatorio del Teide of the Instituto de Astrofísica de Canarias.
## Appendix A The plots showing the comparison between TMGS and SKY
The appendix presents the majority of the the plots showing the detailed comparison between TMGS and SKY. From these plots it is possible to see how the relative importance of the various Galactic components vary with magnitude and position.
Figures A1 to A7 (9 plots) illustrate our interpretations of the counts by direct comparisons with the predictions of SKY. The plots are for a positions every few degrees along each scan (the full data is presented in figure D). Differential counts were scaled per square degree per magnitude, using entirely independent TMGS data in 0.2-mag wide bins. The counts arising from the separate geometric components are distinguished as follows: solid line – total count; dotted or less heavy line – disc; long dashes – spiral arms, local spurs and Gould’s Belt; short dashes – bulge; short dashes and dots – molecular ring; long dashes and dots – halo. The position for each area is given in each plot. See section 3 and Tables 2 and 3 for more details.
Figures B1 to B9 show in the left column the cumulative star counts from the TMGS for each magnitude between $`m_K`$=5 and 9 plotted as the solid line. The longitude at which the strip crosses the plane is given in the upper box. The long dashed line shows the total prediction from SKY and the dotted line the counts from the SKY disc alone. In the right column is shown the ratio of TMGS counts to SKY total counts for each magnitude.
Figures C1 to C6 show the cumulative counts after subtracting the SKY disc from each magnitude between $`m_K`$=5 and 9. The longitude at which the strip crosses the plane is given in the upper box. The solid line is TMGS – SKY disc, The long dashed line is SKY total – SKY disc and the dotted line is the SKY bulge.
Figures D1 to D9 (33 plots) illustrate our interpretations of the counts by direct comparisons with the predictions of SKY every half a degree in latiude along the full TMGS scans. Differential counts were scaled per square degree per magnitude, using entirely independent TMGS data in 0.2-mag wide bins. The counts arising from the separate geometric components are distinguished as follows: solid line – total count; dotted or less heavy line – disc; long dashes – spiral arms, local spurs and Gould’s Belt; short dashes – bulge; short dashes and dots – molecular ring; long dashes and dots – halo. The position for each area is given in each plot. See section 3 and Tables 2 and 3 for more details. Due to the number of plots these are only available electronically.
|
no-problem/9906/astro-ph9906062.html
|
ar5iv
|
text
|
# Intrinsic Parameters of GRB990123 from Its Prompt Optical Flash and Afterglow
## 1 Introduction
The current standard model for gamma-ray bursts (GRBs) and their afterglows is the fireball-plus-shock model (see Piran 1999 for a review). It involves that a large amount of energy, $`E_010^{5154}`$ ergs, is released within a few seconds in a small volume with negligible baryonic load, $`Mc^2E_0`$. This leads to a fireball that expands ultra-relativistically with a Lorentz factor $`\mathrm{\Gamma }_0E_0/Mc^2>100`$ required to avoid the attenuation of hard $`\gamma `$-rays due to pair production (e.g. Woods $`\&`$ Loeb 1995; Fenimore, Epstein $`\&`$ Ho 1993). A substantial fraction of the kinetic energy of the baryons is transferred to a non-thermal population of relativistic electrons through Fermi acceleration in the shock ($`\mathrm{M}\stackrel{´}{\mathrm{e}}\mathrm{sz}\stackrel{´}{\mathrm{a}}\mathrm{ros}`$ $`\&`$ Rees 1993). The accelerated electrons cool via synchrotron emission and inverse Compton scattering in the post-shock magnetic fields and produce the radiation observed in GRBs and their afterglows (e.g. Katz 1994; Sari et al. 1996; Vietri 1997; Waxman 1997a; Wijers et al. 1997). The shock could be either $`\mathrm{𝑖𝑛𝑡𝑒𝑟𝑛𝑎𝑙}`$ due to collisions between fireball shells caused by outflow variability (Paczy$`\stackrel{´}{\mathrm{n}}`$ski $`\&`$ Xu 1994; Rees $`\&`$ $`\mathrm{M}\stackrel{´}{\mathrm{e}}\mathrm{sz}\stackrel{´}{\mathrm{a}}\mathrm{ros}`$ 1994), or $`\mathrm{𝑒𝑥𝑡𝑒𝑟𝑛𝑎𝑙}`$ due to the interaction of the fireball with the surrounding interstellar medium (ISM; $`\mathrm{M}\stackrel{´}{\mathrm{e}}\mathrm{sz}\stackrel{´}{\mathrm{a}}\mathrm{ros}`$ $`\&`$ Rees 1993). The radiation from internal shocks can explain the spectra (Pilla $`\&`$ Loeb 1998) and the fast irregular variability of GRBs (Sari $`\&`$ Piran 1997), while the synchrotron emission from the external shocks provides a successful model for the broken power law spectra and the power law decay of afterglow light curves (e.g. Waxman 1997a,b; Wijers, Rees $`\&`$ $`\mathrm{M}\stackrel{´}{\mathrm{e}}\mathrm{sz}\stackrel{´}{\mathrm{a}}\mathrm{ros}`$ 1997; Vietri 1997; Dai $`\&`$ Lu 1998a,b,c; Dai et al. 1999; Wang et al. 1999a,b; Huang et al. 1998a,b; 1999a,b).
The properties of the synchrotron emission from GRB shocks are determined by the magnetic field strength, $`B`$, and the electron energy distribution behind the shock. Both of them are difficult to estimate from first principles, and so the following dimensionless parameters are often used to incorporate modeling uncertainties (Sari et al. 1996), $`ϵ_B\frac{U_B}{e_{th}}`$, $`ϵ_e\frac{U_e}{e_{th}}`$. Here $`U_B`$ and $`U_e`$ are the magnetic and electron energy densities and $`e_{th}=nm_pc^2(\gamma _p1)`$ is the total thermal energy density behind the shocks, where $`m_p`$ is the proton mass, $`n`$ is the proton number density, and $`\gamma _p`$ is the mean thermal Lorentz factor of the protons. In spite of these uncertainties, an important assumption that $`ϵ_B`$ and $`ϵ_e`$ do not change with time, has been made in the standard external shock model. Through the computation, Wijers $`\&`$ Galama (1998; hereafter WG99) even suggested that they may be universal parameters, i.e. the same for different bursts.
The BeppoSAX satellite ushered in 1999 with the discovery of GRB990123 (Heise et al. 1999), the brightest GRB seen by BeppoSAX to date. This is a very strong burst. An assumption of isotropic emission and the detection of the source’s redshift $`z=1.6004`$ , lead to a huge energy release about $`1.6\times 10^{54}`$ergs (Briggs et al. 1999; Kulkarni et al. 1999a) in $`\gamma `$-rays alone. GRB990123 would have been amongest the most exciting GRBs even just on the basis of these facts. Furthermore, ROTSE discovered a prompt optical flash of 9-th magnitude (Akerlof et al. 1999). It is the first time that a prompt emission in another wavelength apart from $`\gamma `$-rays has been detected from GRB. Such a strong optical flash was predicted to arise from a reverse external shock propagating into the relativistic ejecta ($`\mathrm{M}\stackrel{´}{\mathrm{e}}\mathrm{sz}\stackrel{´}{\mathrm{a}}\mathrm{ros}`$ $`\&`$ Rees 1997; Sari $`\&`$ Piran 1999a,b, hereafter SP99a,b). This is the so called “early afterglow”. The five last exposures of ROTSE show a power law decay with a slope of $``$2.0, which can also be explained by the reverse shock model (SP99b; $`\mathrm{M}\stackrel{´}{\mathrm{e}}\mathrm{sz}\stackrel{´}{\mathrm{a}}\mathrm{ros}`$ $`\&`$ Rees 1999). The usual afterglows in X-ray, optical, IR and radio bands were also detected after the burst. They have two distinguishing features: 1) the radio emission is unique both due to its very early appearance and its rapid decline; 2) the temporal decaying behaviour of the $`r`$-band light curve after two days steepens from about $`t^{1.1}`$ to $`t^{1.8}`$ (Kulkarni et al. 1999a; Fruchter et al. 1999; Casrto-Tirado et al. 1999), and this steepening might be due to a jet which has transited from a spherical-like phase to sideways expansion phase (Rhoads 1999; Sari et al. 1999; Wei $`\&`$ Lu 1999) or a dense medium which has slowed down the relativistic expansion of shock quickly to a non-relativistic one (Dai $`\&`$ Lu 1999a,b).
Galama et al. (1999) assumed that the radio emission is produced by the forward shock as usual and then reconstructed the radio-to-X-ray afterglow spectrum on January 24.65 UT. However, later work shows that the simplest interpretation of this “radio flare” is that it arises in the reverse shock and that such radio emission is an inevitable consequence of the prompt bright optical flash seen by ROTSE (Kulkarni et al. 1999b; Sari $`\&`$ Piran 1999b). Kulkarni et al. (1999b) also constrained two key parameters of the forward shock, the peak flux $`F_{\nu _m}`$ and the peak frequency $`\nu _m`$, to within a factor of two. For previous bursts, we have no other information apart from the afterglow to infer the intrinsic parameters of the external shocks. But now the optical flash of GRB990123 has been fortunately detected, which enables us to determine another two key parameters, the initial Lorentz factor $`\mathrm{\Gamma }_0`$ and the Lorentz factor of the reverse shock $`\mathrm{\Gamma }_{rs}`$. WG99 computed the intrinsic parameters of GRB970508 and GRB971214 in terms of their afterglow data, and found that $`ϵ_e`$ is nearly the same for these two bursts, suggesting it may be a universal parameter. Granot, Piran $`\&`$ Sari (1998a,b; hereafter GPS99a,b) modified the set of equations derived by WG99, and inferred the electron energy density fraction and the magnetic energy density fraction of GRB970508 to be: $`ϵ_e=0.57`$; $`ϵ_B=8.2\times 10^3`$. Here we apply the set of equations of GPS99a to GRB990123 and try to determine some intrinsic parameters based on the information of the two aspects of GRB990123—the optical flash and the afterglow.
The initial Lorentz factor $`\mathrm{\Gamma }_0`$ is also an important physical parameter of GRBs. It is a crucial ingredient for constraining models of the source itself, since it specifies how “clean” the fireball is as the baryonic load is $`ME_0/\mathrm{\Gamma }_0c^2`$. Unfortunately, the spectrum of GRBs can provide only a lower limit to this Lorentz factor ($`\mathrm{\Gamma }_0>100`$). Moreover, the current afterglow observations, which detect radiation from several hours after the burst, do not provide a verification of the initial extreme relativistic motion. A possible method to estimate $`\mathrm{\Gamma }_0`$ of GRBs has been suggested by Sari $`\&`$ Piran (1999a), based on identifying the “early afterglow” peak time. In this paper, the initial Lorentz factor has been inferred more precisely from the full set of equations describing the reverse shock region.
In section 2 and 3, we compute the intrinsic parameters of GRB990123 from its afterglow and optical flash information. In the final section, we give our conclusions.
## 2 Parameters from the afterglow
Intrinsic parameters like the magnetic energy density fraction $`ϵ_B`$, electron energy density fraction $`ϵ_e`$, energy in the forward external shock $`EE_{52}\times 10^{52}`$ ergs and ambient density $`n`$ can be determined from the afterglow spectrum (GPS99a,b; WG99), i. e. if we know all three break frequencies (not necessary at the same time) and the peak flux of the afterglow, we can infer all these parameters.
¿From the observations of the afterglow of GRB990123, Kulkarni et al. (1999b) have estimated two key parameters of the forward shock: $`\nu _m1.1\times 10^{13}\mathrm{Hz}`$ and $`F_{\nu _m}170\mu \mathrm{Jy}`$ at the time $`t_{}=1.25\mathrm{days}`$ after the burst. The cooling frequency $`\nu _c`$ cannot be seen from the radio-to-X-ray spectrum obtained by Galama et al (1999). This indicates that $`\nu _c`$ is at or above the X-ray frequencies. We need to determine it more precisely. The X-ray afterglow, observed 6 hours after the burst, decayed with $`\alpha _X=1.44\pm 0.07`$ ( Heise et al. 1999), while the optical afterglow with $`\alpha _r=1.10\pm 0.03`$ (Kulkarni et al., 1999). An X-ray afterglow decay slope steeper by $`\frac{1}{4}`$ than an optical decay, which seems to be the case in this burst, is predicted by Sari, Piran and Narayan (1998), if the cooling frequency is between the X-rays and the optical. So at the time 6 hours after the burst , $`4\times 10^{14}\nu _c\nu _X(4.444)\times 10^{17}`$Hz. Extrapolating it to the time $`t_{}`$ as $`\nu _ct^{\frac{1}{2}}`$, we get $`1.8\times 10^{14}`$Hz$`\nu _c(t_{})(220)\times 10^{17}`$Hz. Another speculative constraint on $`\nu _c`$ is obtained from the GRB spectrum itself by Sari $`\&`$ Piran (1999b), who constrained $`\nu _c2\times 10^{19}`$Hz at the time $`t50`$ sec. Extrapolating it to $`t_{}`$, we get $`\nu _c(t_{})0.4\times 10^{18}`$. Now $`\nu _c`$ is almost determined, and we take the approximate value $`\nu _c=0.5\times 10^{18}`$Hz, which is in agreement with the estimate of Galama et al (1999). In addition, Kulkarni et al. (1999a) have inferred the electron index $`p`$ (defined as $`N(\gamma _e)\gamma _{e}^{}{}_{}{}^{p}`$) to be $`p=2.44`$. Now, apart from the self-absorption frequency $`\nu _a`$, we have all other three quantities of the afterglow spectrum required to calculate the intrinsic parameters:
$$\nu _m1.1\times 10^{13}\mathrm{Hz},\nu _c0.5\times 10^{18}\mathrm{Hz},F_{\nu _m}170\mu \mathrm{Jy},p2.44$$
(1)
Following GPS99b, we adopt the formulas of $`\nu _m`$ and $`F_{\nu _m}`$ from GPS99a and $`\nu _c`$ from Sari, Piran $`\&`$ Narayan (1998):
$$\nu _m2.9\times 10^{15}(1+z)^{1/2}(\frac{p2}{p1})^2ϵ_{B}^{}{}_{}{}^{1/2}ϵ_{e}^{}{}_{}{}^{2}E_{52}^{}{}_{}{}^{1/2}t_{d}^{}{}_{}{}^{3/2}\mathrm{Hz};$$
(2)
$$F_{\nu _m}1.7\times 10^4(1+z)ϵ_{B}^{}{}_{}{}^{1/2}E_{52}n_{1}^{}{}_{}{}^{1/2}(\frac{d_L}{10^{28}\mathrm{cm}})^2\mu \mathrm{Jy};$$
(3)
$$\nu _c2.7\times 10^{12}ϵ_{B}^{}{}_{}{}^{3/2}E_{52}^{}{}_{}{}^{1/2}n^1t_{d}^{}{}_{}{}^{1/2}(1+z)^{1/2}\mathrm{Hz},$$
(4)
where $`z`$ is the redshift of the burst and $`d_L=2c(1+z\sqrt{1+z})/H_0`$ is the luminosity distance.
Now we have three equations (2), (3) and (4) with four unknowns: $`ϵ_e`$, $`ϵ_B`$, $`E_{52}`$ and $`n`$. To solve these equations, we here assume that the electron density fraction $`ϵ_e`$ is the same for different bursts, just an argument of WG99, though in which a different set of equations are used. Since we here use the formulas of GPS99a, we adopt the value of $`ϵ_e`$ from that of GRB970508 inferred according to the above formulas, i.e. $`ϵ_e0.57`$(GPS99b).
By combining Eqs.(1) and (2)-(4), we get the values of four intrinsic parameters of the forward shock region:
$$ϵ_e0.57ϵ_B3.1\times 10^3n0.01E_{52}5$$
(5)
We astonishingly find that the value $`ϵ_B`$ inferred is very close to that of GRB970508 ($`ϵ_B8.2\times 10^3`$; GPS99b), considering the uncertainties in the $`\nu _m`$ and $`F_{\nu _m}`$ of a factor of two. This result supports our above adoption of the value of $`ϵ_e`$ and implies that $`ϵ_e`$ and $`ϵ_B`$ may be universal parameters (i.e. constants for every GRBs), favouring the argument of WG99.
The ambient density $`n`$ inferred for GRB990123 is $`n0.01`$. This density is on the low side of normal for a disc of galaxy but definitely higher than expected for a halo, lending a further support to the notion that bursts occur in gas-rich environment. The inferred isotropic energy left in the adiabatic forward shock is $`E5\times 10^{52}`$ergs, about thirty times less than the isotropic energy in $`\gamma `$-rays. This case is very similar to GRB971214 (WG99). We think, for GRB990123, there are two processes causing $`E_{52}<E_{52,\gamma }`$. One is that there is a radiative evolution phase before the adiabatic phase, causing it to emit more of the initial explosion energy and leaving less for the adiabatic phase. According to Sari, Piran $`\&`$ Narayan (1998), we can estimate the reduced energy of the forward shock in the self-similar deceleration stage, i.e.
$$E_{f,52}0.02ϵ_{B}^{}{}_{}{}^{3/5}ϵ_{e}^{}{}_{}{}^{3/5}E_{i,52}^{}{}_{}{}^{4/5}(\mathrm{\Gamma }_A/100)^{4/5}n^{2/5},$$
(6)
where $`E_{i,52}`$ denotes the initial isotropic energy of the forward shock during the self-similar stage and $`E_{f,52}`$ denotes the final energy after the radiative phase. The value of $`E_{i,52}`$ is difficult to be determined precisely and it may be less than $`E_{52,\gamma }`$ for GRB990123, according to Freedman & Waxman (2000). If we use $`E_{i,52}E_{52,\gamma }`$ , $`ϵ_e0.57`$, $`ϵ_B8.2\times 10^3`$ (instead of $`ϵ_B3.1\times 10^3`$, since the latter is less reliable than the former, considering the uncertainties in the $`\nu _m`$ and $`F_{\nu _m}`$ of a factor of two for GRB990123.), $`\mathrm{\Gamma }_A300`$ (see the next section) and $`n0.01`$, we then get $`E_{f,52}56`$. The other process is the sideways expansion of the fireball jet (Kulkarni et al.1999a,b), which can also reduce the energy per solid angle, hence the isotropic energy $`E_{52}`$. Since the opening angle of the jet is $`\theta =\theta _0+c_st_{proper}/ct\theta _0+\gamma ^1`$ (Sari, Piran $`\&`$ Halpern 1999; Rhoads 1999), at the time $`t=1.25`$days (very near the break time of the jet evolution $`t_b2.1`$days), $`\gamma \frac{1}{\theta _0}`$ (note that $`\gamma t^{3/8}`$), then $`\theta 2\theta _0`$. Therefore, the real isotropic energy $`E_{52}`$ left in the late adiabatic forward shock should be $`56(\frac{2\theta _0}{\theta _0})^214`$ , in rough agreement with the above value inferred from the afterglow spectrum. An additional possible loss of energy may be the reverse shock itself if it is radiative.
## 3 Parameters from the optical flash
An optical flash is considered to be produced by the reverse external shock, which heats up the shell’s matter and accelerates its electrons (SP99b; M$`\stackrel{´}{e}`$sz$`\stackrel{´}{a}`$ros $`\&`$ Rees 1999). BATSE’s observations triggered ROTSE via BACODINE system (Akerlof et al. 1999). An 11.82 magnitude optical flash was detected on the first 5 seconds exposure, 22.18 seconds after the onset of the burst. Then the optical emission peaked in the following 5 seconds exposure, 25 seconds later, which revealed an 8.95 magnitude signal ($``$1Jy). The optical signal decayed to 10.08 magnitude 25 seconds later and continued to decay down to 14.53 magnitude in the subsequent three 75 seconds exposures that took place up to 10 minutes after the burst. The five last exposures depict a power law decay with a slope $`2.0`$ (Akerlof et al. 1999; SP99b). Sari $`\&`$ Piran (1999b) and M$`\stackrel{´}{e}`$sz$`\stackrel{´}{a}`$ros $`\&`$ Rees (1999) assumed that the ejecta shell follows the Blandford-McKee (1976) self-similar solution after the reverse shock has passed through it and explained the decay of $`t^{2.0}`$.
So we assume that at the optical emission peak time ($`t=50`$ sec) the reverse shock had just passed through the ejecta shell. At this time, the Lorentz factor of the reverse shock $`\mathrm{\Gamma }_{rs}`$ is approximately given by
$$\mathrm{\Gamma }_{rs}=\frac{\mathrm{\Gamma }_0}{\mathrm{\Gamma }_A},$$
(7)
where $`\mathrm{\Gamma }_0`$ is the initial Lorentz factor of the ejecta and $`\mathrm{\Gamma }_A`$ is the Lorentz factor of the ejecta at the optical flash peak time. Then the random minimum Lorentz factor $`\gamma _{min}`$ of the electrons in the reverse shock region is:
$$\gamma _{min}=\frac{m_p}{m_e}\frac{p2}{p1}ϵ_e\frac{\mathrm{\Gamma }_0}{\mathrm{\Gamma }_A},$$
(8)
The formulas of $`\nu _m`$ at the reverse shock were given by Sari $`\&`$ Piran (1999b). We here add the correction for redshift.
$$\nu _m=1.2\times 10^{13}(\frac{ϵ_e}{0.1})^2(\frac{ϵ_B}{10^3})^{1/2}(\frac{\mathrm{\Gamma }_0}{300})^2n^{1/2}(1+z)^1\mathrm{Hz}5\times 10^{14}\mathrm{Hz}.$$
(9)
The observed flux at $`\nu _m`$ can be obtained by assuming that all the electrons in the reverse shock region contribute the same average power per unit frequency $`P_{\nu _m}^{}`$ at $`\nu _m`$, which is given by $`P_{\nu _m}^{}=\frac{\sqrt{3}e^3B^{}}{m_ec^2}`$, where $`B^{}=\mathrm{\Gamma }_Ac\sqrt{32\pi nm_pϵ_B}`$. Adding one factor of $`\mathrm{\Gamma }_A`$ to transform to the observer frame and accounting for the redshift, we have:
$$F_{\nu _m}=\frac{N_e\mathrm{\Gamma }_AP_{\nu _m}^{}(1+z)}{4\pi d_{L}^{}{}_{}{}^{2}},$$
(10)
where $`N_e`$ is the total number of radiating electrons in the ejecta shell, and $`d_L=2c(1+z\sqrt{1+z})/H_0`$ is the luminosity distance. Please note that $`N_e`$ here is different from the $`N_e`$ adopted in the forward shock region, which is the total number of swept-up electrons by the forward external shock. We consider $`N_e`$ here to be the total number of electrons contained in the baryonic load:
$$N_e=\frac{M}{m_p}\frac{E_\gamma }{\mathrm{\Gamma }_0m_pc^2}=1.08\times 10^{54}(\frac{E_\gamma }{1.6\times 10^{54}})(\frac{\mathrm{\Gamma }_0}{1000})^1,$$
(11)
where $`E_\gamma `$ is the total energy in $`\gamma `$-rays. Substituting the expression of $`P_{\nu _m}^{}`$ and $`z=1.6`$ into Eq.(10), we get
$$F_{\nu _m}=0.74(\frac{E_\gamma }{1.6\times 10^{54}})(\frac{\mathrm{\Gamma }_0}{1000})^1(\frac{\mathrm{\Gamma }_A}{100})^2n^{1/2}(\frac{ϵ_B}{10^3})^{1/2}\mathrm{Jy}.$$
(12)
Please note that this formula always holds whether the ejecta is jet-like or spherical, because the beaming factor in Eq.(10) and (11) will cancel out each other in the jet-like case. According to the jump condition of the shock, the Lorentz factor of the shocked shell should be approximately equal to that of the shocked ISM (Piran 1999). The Lorentz factor of the forward shocked ISM can be obtained from the standard afterglow model (e. g. Sari, Piran $`\&`$ Narayan 1998):
$$\mathrm{\Gamma }_{A,fs}(t)6(\frac{E_{52}}{n})^{1/8}(\frac{t_d}{1+z})^{3/8}$$
(13)
For $`E_{52}5`$ and $`n0.01`$, we get
$$\mathrm{\Gamma }_{A,rs}(50s)=\mathrm{\Gamma }_{A,fs}(50s)=300$$
(14)
Taking the above inferred value $`ϵ_B3.1\times 10^3`$, from the Eqs.(9), (12), and (14) with the conditions: $`\nu _m5\times 10^{14}`$Hz and $`F_{\nu _m}1`$Jy, we finally get:
$$\mathrm{\Gamma }_A=300,\mathrm{\Gamma }_0=1200,ϵ_e0.6.$$
(15)
Please note that here the value $`ϵ_e0.6`$ inferred from the optical flash data is consistent with that inferred independently from the afterglow information. On the other hand, if we substitute the value $`ϵ_e=0.57`$ into Eq.(9), we find that the peak frequency of the reverse shock $`\nu _m`$ is almost located at the optical band. In addition, our inferred initial Lorentz factor $`\mathrm{\Gamma }_0`$ is six times larger than that obtained by Sari $`\&`$ Piran (1999b), who have used the ambient density $`n`$ of GRB970508. Consequently, at the time the reverse shock has just passed through the ejecta shell, its Lorentz factor was $`\mathrm{\Gamma }_{rs}=\frac{\mathrm{\Gamma }_0}{\mathrm{\Gamma }_A}4`$. This indicates that the reverse shock had become relativistic before it crossed the entire shell. This result is also different from that obtained by Sari $`\&`$ Piran (1999b), who found the Lorentz factor of the reverse shock of GRB990123 was only near one. However we argue that our result is reasonable according to the criterion presented by Sari $`\&`$ Piran (1995) (also see Kobayashi et al. 1998). They defined a dimensionless parameter $`\xi `$ constructed from $`l`$, $`\mathrm{\Delta }`$ and $`\mathrm{\Gamma }_0`$:
$$\xi (l/\mathrm{\Delta })^{1/2}\mathrm{\Gamma }_{0}^{}{}_{}{}^{4/3},$$
(16)
where $`l=(\frac{E}{nm_pc^2})^{1/3}`$ is the Sedov length, $`\mathrm{\Delta }=c\delta T`$ is the width of the shell ($`\delta T`$ is the duration of GRB) and $`\mathrm{\Gamma }_0`$ is the initial Lorentz factor of the ejecta. If $`\xi <1`$, the reverse shock becomes relativistic before it crosses the shell; otherwise ($`\xi >1`$), the reverse shock remains Newtonian or at best mildly relativistic during the whole energy extraction process. For GRB990123, we find $`\xi 0.3<1`$. So the reverse shock of GRB990123 had become relativistic before it crossed the shell, consistent with our calculated result.
## 4 Conclusions and Discussions
We have constrained some intrinsic parameters, such as the magnetic energy density fraction ($`ϵ_B`$), the electron energy density fraction ($`ϵ_e`$), the isotropic energy in the adiabatic forward shock $`E_{52}`$ and the ambient density $`n`$. Due to the lack of the value of the self-absorption frequency $`\nu _a`$, we made an assumption that $`ϵ_e`$ of GRB990123 is the same as that of GRB970508, then astonishingly find that the inferred value of $`ϵ_B`$ is also nearly equal to that of GRB970508. This result favours the argument proposed by WG99 that the magnetic energy fraction and the electron density fraction may be universal parameters.
Another two important intrinsic parameters of GRB990123 are also inferred from the optical flash information: the initial Lorentz factor $`\mathrm{\Gamma }_0`$ and the Lorentz factor $`\mathrm{\Gamma }_A`$ at the prompt optical emission peak time of the ejecta. They are: $`\mathrm{\Gamma }_0=1200`$, $`\mathrm{\Gamma }_A=300`$. Our inferred value of the $`\mathrm{\Gamma }_0`$ is six times larger than that obtained by Sari $`\&`$ Piran (1999b), who used the ambient density $`n`$ inferred for GRB970508. A larger initial Lorentz factor is reasonable in consideration of the huge energy of this burst. The Lorentz factor of the reverse shock at the optical flash peak time is $`\mathrm{\Gamma }_{rs}\frac{\mathrm{\Gamma }_0}{\mathrm{\Gamma }_A}4`$, which shows that the reverse shock had become relativistic rather than mildly relativistic before it crossed the entire ejecta shell. This result is in agreement with the criterion presented by Sari $`\&`$ Piran (1995) to judge the RRS case or NRS case.
Prompt optical flash has added another dimension to GRB astronomy. Prompt observations in the optical band during and immediately after GRB may provide more and more events of optical flash in the near future, and they will enable us to make more detailed analyses, make more precise determination of intrinsic parameters and test the reverse—forward external shock model.
## Acknowledgments
We would like to thank the referee for his valuable suggestions and improvement on our manuscript. X.Y. Wang also thank Dr. D. M. Wei for helpful discussion. This work was supported by the National Natural Science Foundation of China under grants 19773007 and 19825109 and the foundation of Ministry Education of China.
|
no-problem/9906/hep-ex9906045.html
|
ar5iv
|
text
|
# Search for Rare and Forbidden Dilepton Decays of the 𝐷⁺, 𝐷_𝑠⁺, and 𝐷⁰ Charmed Mesons
## Abstract
We report the results of a search for flavor-changing neutral current, lepton-flavor violating, and lepton-number violating decays of $`D^+`$, $`D_s^+`$, and $`D^0`$ mesons (and their antiparticles) into modes containing muons and electrons. Using data from Fermilab charm hadroproduction experiment E791, we examine the $`\pi \mathrm{}\mathrm{}`$ and $`K\mathrm{}\mathrm{}`$ decay modes of $`D^+`$ and $`D_s^+`$ and the $`\mathrm{}^+\mathrm{}^{}`$ decay modes of $`D^0`$. No evidence for any of these decays is found. Therefore, we present branching-fraction upper limits at 90$`\%`$ confidence level for the 24 decay modes examined. Eight of these modes have no previously reported limits, and fourteen are reported with significant improvements over previously published results.
FERMILAB-Pub-99/183-E
UMS/HEP/99–020
, , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , and
The SU(2)$`\times `$U(1) Standard Model of electroweak interactions qualitatively accounts for the known decays of heavy quarks and can often quantitatively predict the decay rates. However, this model is incomplete in that it does not account for the number of quark and lepton families observed, nor their hierarchy of mass scales. Also unknown is the mechanism responsible for breaking the underlying gauge symmetry. One way to search for physics beyond the Standard Model is to search for decays that are forbidden or else are predicted to occur at a negligible level. Observing such decays would constitute evidence for new physics, and measuring their branching fractions would provide insight into how to modify our theoretical understanding, e.g., by introducing new particles or new gauge couplings.
In this letter we present the results of a search for 24 decay modes of the neutral and charged $`D`$ mesons (which contain the heavy charm quark). These decay modes<sup>1</sup><sup>1</sup>1Charge-conjugate modes are included implicitly throughout this paper. fall into three categories:
1. FCNC – flavor-changing neutral current decays ($`D^0\mathrm{}^+\mathrm{}^{}`$ and $`D_{(d,s)}^+h^+\mathrm{}^+\mathrm{}^{}`$, in which $`h`$ is $`\pi `$ or $`K`$);
2. LFV – lepton-flavor violating decays ($`D^0\mu ^\pm e^{}`$, $`D_{(d,s)}^+h^+\mu ^\pm e^{}`$, and $`D_{(d,s)}^+h^{}\mu ^+e^+`$, in which the leptons belong to different generations);
3. LNV – lepton-number violating decays ($`D_{(d,s)}^+h^{}\mathrm{}^+\mathrm{}^+`$, in which the leptons belong to the same generation but have the same sign charge).
Decay modes belonging to (1) occur within the Standard Model via higher-order diagrams, but the estimated branching fractions are $`10^8`$ to $`10^6`$ . Such small rates are below the sensitivity of current experiments. However, if additional particles such as supersymmetric squarks or charginos exist, they could contribute additional amplitudes that would make these modes observable. Decay modes belonging to (2) and (3) do not conserve lepton number and thus are forbidden within the Standard Model. However, lepton number conservation is not required by Lorentz invariance or gauge invariance, and a number of theoretical extensions to the Standard Model predict lepton-number violation . Many experiments have searched for lepton-number violation in $`K`$ decays, and for lepton-number violation and flavor-changing neutral currents in $`D`$ and $`B`$ decays. The limits we present here for rare and forbidden dilepton decays of the $`D`$ mesons are typically more stringent than those obtained from previous searches , or else are the first reported.
The data are from Fermilab experiment E791 , which recorded $`2\times 10^{10}`$ events with a loose transverse energy trigger. These events were produced by a 500 GeV/$`c`$ $`\pi ^{}`$ beam interacting in a target consisting of five thin foils that had 15 mm center-to-center separation along the beamline. The most upstream foil was 0.5 mm thick platinum. It was followed by four foils consisting of 1.6 mm thick diamond. Momentum analysis was provided by two dipole magnets that bent particles in the horizontal ($`x`$-$`z`$) plane. Position information for track and vertex reconstruction was provided by 23 silicon microstrip detectors (6 upstream and 17 downstream of the target) along with 10 planes of proportional wire chambers (8 upstream and 2 downstream of the target), and 35 drift chamber planes. The experiment also included electromagnetic and hadronic calorimeters, a muon detector, and two multi-cell Čerenkov counters that provided $`\pi /K`$ separation in the momentum range $`660`$ GeV/$`c`$ . The kaon identification criteria varied by search decay mode. We typically required that the momentum-dependent light yield in the Čerenkov counters be consistent with that of a kaon track measured in the spectrometer.
Electrons were identified by an electromagnetic calorimeter that consisted of lead sheets and liquid scintillator located 19 m downstream of the target. Electron identification was based on energy deposition and transverse shower shape in the calorimeter. The electron identification efficiency varied from 62$`\%`$ for momenta below 9 GeV/$`c`$ to 45$`\%`$ for momenta above 20 GeV/$`c`$. The decrease in efficiency with increasing momentum reflects the fact that higher momentum electrons populate a more congested region of the spectrometer. The pion misidentification rate was approximately 0.8$`\%`$, independent of pion momentum.
Muon identification was obtained from two planes of scintillation counters. The plane that measured vertical coordinates ($`y`$) consisted of 16 scintillation counters, each 3 meters long and 14 cm wide. The plane that measured horizontal coordinates ($`x`$) consisted of 14 counters, each 3 meters long and covering a full width of 5.5 meters in the $`x`$-direction. The counters were located behind shielding with a thickness equivalent to 2.5 meters (15 interaction lengths) of iron. Candidate muon tracks projected into the muon system were required to pass a series of muon quality criteria that were optimized with $`D^+\overline{K}^0\mu ^+\nu __\mu `$ decays from our data . Timing information from the $`y`$-coordinate counters was used to improve the position resolution in the $`x`$-direction. The efficiencies of the muon counters were measured in special runs using muons originating from the primary beam dump, and were found to be $`(99\pm 1)\%`$ for the $`y`$-coordinate counters and $`(69\pm 3)\%`$ for the $`x`$-coordinate counters. The probability for misidentifying a pion as a muon decreased as momentum increased, from about 6$`\%`$ at 8 GeV/$`c`$ to $`(1.3\pm 0.1)\%`$ for momenta greater than 20 GeV/$`c`$.
After reconstruction, events with evidence of well-separated production (primary) and decay (secondary) vertices were retained for further analysis. To separate charm candidates from background, we required the following: that secondary vertices be well-separated from the primary vertex and located well outside the target foils and other solid material; that the momentum vector of the candidate charm meson point back to the primary vertex; and that the decay track candidates pass approximately 10 times closer to the secondary vertex than to the primary vertex. A secondary vertex had to be separated from the primary vertex by greater than $`20\sigma __L`$ for $`D^+`$ decays and greater than $`12\sigma __L`$ for $`D^0`$ and $`D_s^+`$ decays, where $`\sigma __L`$ is the calculated resolution of the measured longitudinal separation. In addition, the secondary vertex had to be separated from the closest material in the target foils by greater than $`5\sigma __L^{}`$, where $`\sigma __L^{}`$ is the uncertainty in this separation. The sum of the vector momenta of the tracks from the secondary vertex was required to pass within $`40\mu `$m of the primary vertex in the plane perpendicular to the beam. Finally, the net momentum of the charm candidate transverse to the line connecting the production and decay vertices had to be less than 300 MeV/$`c`$ for $`D^0`$ candidates, less than 250 MeV/$`c`$ for $`D_s^+`$ candidates, and less than 200 MeV/$`c`$ for $`D^+`$ candidates. These selection criteria and, where possible, the kaon identification requirements, were the same for the search mode and for its normalization signal.
For this study we used a “blind” analysis technique. Before our selection criteria were finalized, all events having masses within a mass window $`\mathrm{\Delta }M_S`$ around the mass of $`D^+`$, $`D_s^+`$, or $`D^0`$ were “masked” so that the presence or absence of any potential signal candidates would not bias our choice of selection criteria. All criteria were then chosen by studying signal events generated by a Monte Carlo simulation program (see below) and background events from real data. Events within the signal windows were unmasked only after this optimization. Background events were chosen from a mass window $`\mathrm{\Delta }M_B`$ above and below the signal window $`\mathrm{\Delta }M_S`$. The criteria were chosen to maximize the ratio $`N_S/\sqrt{N_B}`$, where $`N_S`$ and $`N_B`$ are the numbers of signal and background events, respectively. We used asymmetric windows for the decay modes containing electrons to allow for the bremsstrahlung low-energy tail. The signal windows are:
$$\begin{array}{c}1.84<M(D^+)<1.90\mathrm{GeV}/c^{\mathrm{\hspace{0.17em}2}}\mathrm{for}D^+h\mu \mu ,\hfill \\ 1.78<M(D^+)<1.90\mathrm{GeV}/c^{\mathrm{\hspace{0.17em}2}}\mathrm{for}D^+hee\mathrm{and}h\mu e,\hfill \\ 1.95<M(D_s^+)<1.99\mathrm{GeV}/c^{\mathrm{\hspace{0.17em}2}}\mathrm{for}D_s^+h\mu \mu ,\hfill \\ 1.91<M(D_s^+)<1.99\mathrm{GeV}/c^{\mathrm{\hspace{0.17em}2}}\mathrm{for}D_s^+hee\mathrm{and}h\mu e,\hfill \\ 1.83<M(D^0)<1.90\mathrm{GeV}/c^{\mathrm{\hspace{0.17em}2}}\mathrm{for}D^0\mu \mu ,\hfill \\ 1.76<M(D^0)<1.90\mathrm{GeV}/c^{\mathrm{\hspace{0.17em}2}}\mathrm{for}D^0ee\mathrm{and}\mu e.\hfill \end{array}$$
(1)
We normalize the sensitivity of our search to topologically similar Cabibbo-favored decays. For the $`D^+`$ decays we use $`D^+K^{}\pi ^+\pi ^+`$; for $`D_s^+`$ decays we use $`D_s^+\varphi \pi ^+`$; and for $`D^0`$ decays we use $`D^0K^{}\pi ^+`$. The widths of our normalization modes were 10.5 MeV/$`c^{\mathrm{\hspace{0.17em}2}}`$ for $`D^+`$, 9.5 MeV/$`c^{\mathrm{\hspace{0.17em}2}}`$ for $`D_s^+`$, and 12 MeV/$`c^{\mathrm{\hspace{0.17em}2}}`$ for $`D^0`$. The events within the $`5\sigma `$ window are shown in Figs. 1a–c. The upper limit for each branching fraction $`B_X`$ is calculated using the following formula:
$$B_X=\frac{N_X}{N_{\mathrm{Norm}}}\frac{\epsilon _{\mathrm{Norm}}}{\epsilon _X}B_{\mathrm{Norm}}$$
(2)
where $`N_X`$ is the 90$`\%`$ CL upper limit on the number of decays for the rare or forbidden decay mode $`X`$, and $`\epsilon _X`$ is that mode’s detection efficiency. $`N_{\mathrm{Norm}}`$ is the fitted number of normalization mode decays; $`\epsilon _{\mathrm{Norm}}`$ is the normalization mode detection efficiency; and $`B_{\mathrm{Norm}}`$ is the normalization mode branching fraction obtained from the Particle Data Group .
The ratio of detection efficiencies is
$$\frac{\epsilon _{\mathrm{Norm}}}{\epsilon _X}=\frac{N_{\mathrm{Norm}}^{\mathrm{MC}}}{N_X^{\mathrm{MC}}}$$
(3)
where $`N_{\mathrm{Norm}}^{\mathrm{MC}}`$ and $`N_X^{\mathrm{MC}}`$ are the fractions of Monte Carlo events that are reconstructed and pass the final selection criteria, for the normalization and decay modes respectively. The simulations use Pythia/Jetset as the physics generator and model the effects of resolution, geometry, magnetic fields, multiple scattering, interactions in the detector material, detector efficiencies, and the analysis selection criteria. The efficiencies for the normalization modes varied from approximately $`0.5\%`$ to $`2\%`$ depending on the mode, and the efficiencies for the search modes varied from approximately $`0.1\%`$ to $`2\%`$.
Monte Carlo studies show that the experiment’s acceptances are nearly uniform across the Dalitz plots, except that the dilepton identification efficiencies typically drop to near zero at the dilepton mass threshold. While the loss in efficiency varies channel by channel, the efficiency typically reaches its full value at masses only a few hundred MeV/$`c^{\mathrm{\hspace{0.17em}2}}`$ above the dilepton mass threshold. We use a constant weak-decay matrix element when calculating the overall detection efficiencies. Two exceptions to the use of the Monte Carlo simulations in determining relative efficiencies are made: those for Čerenkov identification when the number of kaons in the signal and normalization modes are different, and those for the muon identification. These efficiencies are determined from data.
The 90$`\%`$ CL upper limits $`N_X`$ are calculated using the method of Feldman and Cousins to account for background, and then corrected for systematic errors by the method of Cousins and Highland . In these methods, the numbers of signal events are determined by simple counting, not by a fit. All results are listed in Table 1 and shown in Figs. 2 and 3. The kinematic criteria and removal of reflections (see below) are different for the $`D^+`$, $`D_s^+`$, and $`D^0`$. Thus, the $`D^+`$ and $`D_s^+`$ rows in Fig. 2 with the same decay particles are different, and the seventh row of Fig. 2 is different from the bottom row of Fig. 1.
The upper limits are determined by both the number of candidate events and the expected number of background events within the signal region. Background sources that are not removed by the selection criteria discussed earlier include decays in which hadrons (from real, fully-hadronic decay vertices) are misidentified as leptons. In the case where kaons are misidentified as leptons, candidates have effective masses which lie outside the signal windows. Most of these originate from Cabibbo-favored modes $`D^+K^{}\pi ^+\pi ^+`$, $`D_s^+K^{}K^+\pi ^+`$, and $`D^0K^{}\pi ^+`$ (and charge conjugates). These Cabibbo-favored reflections are explicitly removed prior to the selection-criteria optimization. There remain two sources of background in our data: hadronic decays with pions misidentified as leptons ($`N_{\mathrm{MisID}}`$) and “combinatoric” background ($`N_{\mathrm{Cmb}}`$) arising primarily from false vertices and partially reconstructed charm decays. After selection criteria were applied and the signal windows opened, the number of events within the window is $`N_{\mathrm{Obs}}=N_{\mathrm{Sig}}+N_{\mathrm{MisID}}+N_{\mathrm{Cmb}}`$.
The background $`N_{\mathrm{MisID}}`$ arises mainly from singly-Cabibbo-suppressed (SCS) modes. These misidentified leptons can come from hadronic showers reaching the muon counter, decays-in-flight, and random overlaps of tracks from otherwise separate decays (“accidental” sources). We do not attempt to establish a limit for $`D^+K^{}\mathrm{}^+\mathrm{}^+`$ modes, as they have relatively large feedthrough signals from copious Cabibbo-favored $`K^{}\pi ^+\pi ^+`$ decays. Instead, we use the observed signals in $`K^{}\mathrm{}^+\mathrm{}^+`$ channels to measure three dilepton misidentification rates under the assumption that the observed signals (shown in Figs. 1d–f) arise entirely from lepton misidentification. The curve shapes were determined from Monte Carlo simulations. The following misidentification rates were obtained: $`r_{\mu \mu }=(7.3\pm 2.0)\times 10^4`$, $`r_{\mu e}=(2.9\pm 1.3)\times 10^4`$, and $`r_{ee}=(3.4\pm 1.4)\times 10^4`$. Using these rates we estimate the numbers of misidentified candidates, $`N_{\mathrm{MisID}}^h\mathrm{}\mathrm{}`$ (for $`D^+`$ and $`D_s^+`$) and $`N_{\mathrm{MisID}}^{\mathrm{}\mathrm{}}`$ (for $`D^0`$), in the signal windows as follows:
$$N_{\mathrm{MisID}}^h\mathrm{}\mathrm{}=r_{\mathrm{}\mathrm{}}N_{\mathrm{SCS}}^{h\pi \pi }\mathrm{and}N_{\mathrm{MisID}}^{\mathrm{}\mathrm{}}=r_{\mathrm{}\mathrm{}}N_{\mathrm{SCS}}^{\pi \pi },$$
(4)
where $`N_{\mathrm{SCS}}^{h\pi \pi }`$ and $`N_{\mathrm{SCS}}^{\pi \pi }`$ are the numbers of SCS hadronic decay candidates within the signal windows. For modes in which two possible pion combinations can contribute, e.g., $`D^+h^+\mu ^\pm \mu ^{}`$, we use twice the above rate. These misidentification backgrounds were typically small or negligible.
To estimate the combinatoric background $`N_{\mathrm{Cmb}}`$ within a signal window $`\mathrm{\Delta }M_S`$, we count events having masses within an adjacent background mass window $`\mathrm{\Delta }M_B`$, and scale this number ($`N_{\mathrm{\Delta }M_B}`$) by the relative sizes of these windows:
$$N_{\mathrm{Cmb}}=\frac{\mathrm{\Delta }M_S}{\mathrm{\Delta }M_B}N_{\mathrm{\Delta }M_B}.$$
(5)
To be conservative in calculating our 90$`\%`$ confidence level upper limits, we take combinatoric backgrounds to be zero when no events are located above the mass windows. In Table 1 we present the numbers of combinatoric background, misidentification background, and observed events for all 24 modes.
The sources of systematic errors in this analysis include: statistical errors from the fit to the normalization sample $`N_{\mathrm{Norm}}`$; statistical errors on the numbers of Monte Carlo generated events for both $`N_{\mathrm{Norm}}^{\mathrm{MC}}`$ and $`N_X^{\mathrm{MC}}`$; uncertainties in the calculation of misidentification background; and uncertainties in the relative efficiency for each mode, including lepton and kaon tagging efficiencies. These tagging efficiency uncertainties include: 1) the muon counter efficiencies from both Monte Carlo simulation and hardware performance; 2) kaon Čerenkov identification efficiency due to differences in kinematics and modeling between data and Monte Carlo simulated events; and 3) the fraction of signal events (based on simulations) that would remain outside the signal window due to bremsstrahlung tails. The larger systematic errors for the $`D_s^+`$ modes, compared to the $`D^+`$ and $`D^0`$ modes, are due to the uncertainty in the branching fraction for the $`D_s^+`$ normalization mode. The sums, taken in quadrature, of these systematic errors are listed in Table 1.
In summary, we use a “blind” analysis of data from Fermilab experiment E791 to obtain upper limits on the dilepton branching fractions for flavor-changing neutral current, lepton-number violating, and lepton-family violating decays of $`D^+`$, $`D_s^+`$, and $`D^0`$ mesons. No evidence for any of these decays is found. Therefore, we present upper limits on the branching fractions at the 90$`\%`$ confidence level. These limits represent significant improvements over previously published results. Eight new $`D_s^+`$ search modes are reported. A comparison of our 90$`\%`$ C.L. upper limits with previously published results is shown in Fig. 3.
We gratefully acknowledge the assistance of the staffs of Fermilab and of all the participating institutions. This research was supported by the Brazilian Conselho Nacional de Desenvolvimento Científico e Tecnológico, CONACyT (Mexico), the Israeli Academy of Sciences and Humanities, the U.S. Department of Energy, the U.S.-Israel Binational Science Foundation, and the U.S. National Science Foundation. Fermilab is operated by the Universities Research Association, Inc., under contract with the U.S. Department of Energy.
|
no-problem/9906/astro-ph9906494.html
|
ar5iv
|
text
|
# Thermonuclear Burning Regimes and the Use of SNe Ia in Cosmology
## Introduction
For decades, Type Ia supernovae (SNe Ia) are treated as very suitable objects for distance measurements in the Universe and for deriving cosmological parameters (Sandage, Tammann, 1997; Ruiz-Lapuente, 1997). There are several reasons for this. The first one is the intrinsic brightness of the SNe Ia. Due to their luminosity, one can measure an appreciable flux even if they are at high redshift, like $`z0.51`$. Besides that, the spectra and light curves of SNe Ia seem very uniform at first glance. However, more thorough investigation demonstrates the diversity within this class of objects (Pskovskii 1977, Bartunov & Tsvetkov 1986, Branch 1987, Barbon et al. 1989, Phillips et al. 1987; Filippenko, et al. 1992, Leibundgut, et al. 1993).
Quite a while ago, Pskovskii (1977) has shown that there is a correlation between the maximum luminosities of nearby SNe I and their post-maximum decline rates. This dependence was confirmed subsequently from the observations of many low-$`z`$ supernovae and studied by several workers (Bartunov & Tsvetkov 1986, Phillips 1993, Hamuy et al. 1995,1996a, Hamuy & Pinto 1999).
Modern observational technique allows discovering and studying distant, high-$`z`$, supernovae. The first results were obtained by Nørgaard-Nielsen et al. (1989): a few years of their observations have brought up only two supernovae at $`z0.3`$.
Presently, there are several search groups in the world which observe distant supernovae with the largest terrestrial telescopes (and sometimes with the help of the Hubble Space Telescope). The observational technique is advanced to the level allowing the discovery of dozens high-$`z`$ supernovae per one observational period of a couple of weeks. The observational material produced by those groups allowed to estimate the cosmological parameters such as the Hubble constant $`H_0`$, the matter density $`\mathrm{\Omega }__\mathrm{M}`$ and the vacuum energy $`\mathrm{\Omega }_\mathrm{\Lambda }`$ in units of the critical density (for definitions see, e.g., Carroll et al. 1992). The dependent quantities, like the deceleration parameter $`q`$, the ratio of the local value of $`H_0`$ to the global one, etc., can be derived based on this. E.g., Kim et al. (1997) have estimated $`H_0`$ using the first seven SNe Ia at $`z0.35`$ and refuted the suspicion that the local value of $`H_0`$ is appreciably larger than the average one. Perlmutter et al. (1997) have estimated a probable ratio of the $`\mathrm{\Omega }_\mathrm{M}`$ and $`\mathrm{\Omega }_\mathrm{\Lambda }`$.
Quite recently, several interesting papers appeared based on a richer statistics of distant supernovae (Schmidt et al. 1998; Garnavich et al. 1998a,b; Riess et al. 1998; Perlmutter et al. 1999). An extremely important result is claimed in that work: the analysis of the observations implies with high confidence that the expansion of the Universe is accelerating at present epoch.
One should note, though, that those results on distant supernovae were based on the maximum luminosity – decline rate relation derived from the analysis of nearby objects (in the work by Perlmutter et al. 1999, the high redshift SNe Ia are also used in obtaining such a relation, but this does not help in excluding possible effects of evolution, see Drell et al. 1999). Even for the nearby SNe Ia the deviations of individual objects from the relation cannot be explained solely by observational errors. The physical understanding of the peak luminosity – decline rate correlation is crucial for estimating the cosmological results obtained with SNe Ia.
In principle, the slower decline rate for the higher maximum luminosity can be explained from theoretical point of view: both are caused mainly by <sup>56</sup>Ni produced during the explosion. Radioactive decay of <sup>56</sup>Ni forms the SN Ia light curve and determines its maximum luminosity. On the other hand, large amount of nickel should enhance the opacity of matter. It takes longer time for radiation to diffuse through the stellar matter and the light curve becomes less steep. However, the decline rate is influenced not only by the amount of nickel (and of other heavy elements), but also by their distribution inside the expanding star and by the velocity distribution of ejected gas. These distributions are dependent, in turn, on the mode of burning propagation in the star.
After the pioneering work by Arnett (1969), Ivanova et al. (1974), Nomoto et al. (1976) the theory of thermonuclear burning in supernovae has been advanced appreciably, yet many problems are still not solved, see, e.g., Niemeyer and Woosley (1997), Reinecke et al. (1999), Niemeyer (1999). A plenty of models for SN Ia explosions were built, with various masses (Chandrasekhar and sub-Chandrasekhar), burning modes (detonation, deflagration, and a variety of their combinations) in a range of explosion energies and expansion velocities. As a result of dissimilar burning, chemical elements are produced in various proportions and distributed differently throughout the ejecta, even if conditions in a star prior to the explosion are identical. This leads to noticeable variations in theoretical light curves. From the comparison of calculated light curves with the observed ones one can judge which mode of explosion is preferred in nature. Different explosion scenarios imply a scatter in the dependencies between the observable parameters of the burst, and could explain the events deviating from the standard relations to a larger extent than a margin anticipated from the observational errors.
An extensive set of the models for light curves of SNe Ia was produced by Höflich et al. (1996,1997), giving some theoretical insight into the observed correlations of the peak luminosity and the post-maximum decline rate. We have undertaken an independent study of Chandrasekhar and sub-Chandrasekhar SN Ia light curve models. The results will be published elsewhere (Sorokina et al. 1999), some of our calculations for sub-Chandrasekhar models are discussed by Ruiz-Lapuente et al. (1999). In this report we present the results of the calculations of the light curves for two well-known SN Ia models: the deflagration model W7 (Nomoto et al. 1984) and the delayed detonation one, DD4 (Woosley & Weaver 1994). They both are Chandrasekhar mass models and contain similar amount of <sup>56</sup>Ni, but the latter is distributed differently due to different burning modes. Our result is that the two light curves are not similar in their decline rate, despite the close absolute luminosities at maximum light. The B-band flux of W7 goes down too slowly and agrees with the observations of typical SNe Ia only marginally. We explain why the results of other authors somewhat disagree with ours for this model.
Discussing the possibility of the use of SNe Ia in cosmology, we suggest that the available statistics of distant supernovae does not yet allow drawing firm conclusions on the geometry of the Universe. In terrestrial experiments, one cannot predict with certainty the outcome of an explosion. The situation for supernovae can be similar: it is quite likely that initial conditions influence only the probability of the development of a certain burning mode, but do not determine it exactly. Since the light curve shape depends strongly on the regime of burning, one cannot predict deterministically the decline rate knowing only the initial conditions. The probability of the specific decline rate, which is crucial for obtaining the cosmological parameters, would be found only with a sufficiently rich statistics of SNe Ia at various redshifts.
## Numerical method
The approach used for light curve modeling in our study is the multi-energy group radiation hydrodynamics. The code we have used is called stella (Blinnikov & Bartunov 1993; Blinnikov et al. 1998). It solves the time-dependent equations for the angular moments of intensity averaged over fixed frequency bands, using up to $`200`$ zones for the Lagrangean coordinate and up to 100 frequency bins (i.e., energy groups). This number of frequency groups allows one to have a reasonably accurate representation of non-equilibrium continuum radiation. There is no need to ascribe any temperature to the radiation: the photon energy distribution can be quite arbitrary.
The coupling of multi-group radiative transfer with hydrodynamics means that we can obtain UBVRI fluxes in a self-consistent calculation, and that no additional estimates of thermalization depth as in the one-energy group models are needed. Variable Eddington factors are computed, which fully take into account scattering and redshifts for each frequency group in each mass zone. The parameters of the decays of <sup>56</sup>Ni to <sup>56</sup>Co and of <sup>56</sup>Co to <sup>56</sup>Fe are taken from the work of Nadyozhin (1994). The positron energy is deposited locally. The gamma-ray transfer is calculated using a one-group approximation for the non-local deposition of the energy of radioactive nuclei. Here we follow Swartz et al. (1995) and we use purely absorptive opacity in the gamma-ray range. Swartz et al. addressed this question using in particular the W7 model. They found a very weak dependence on the optical depth and derived the error of only 10% in heating rate due to the gamma deposition in the SN Ia model up to 1200 days after the explosion for the effective opacity $`\kappa _\gamma =0.050Y_e`$ ($`Y_e`$ is the total number of electrons per baryon).
Local Thermodynamic Equilibrium (LTE) for ionization and atomic level populations is assumed in our modeling. In the equation of state, LTE ionization and recombination are taken into account. The effect of line opacity is treated as an expansion opacity according to the prescription of Eastman & Pinto (1993).
The main limitation of our current light curve code is the LTE approximation. To simulate some of the non-LTE effects we used the approximation of the absorptive opacity in spectral lines, following the results of non-LTE computations by Baron et al. (1996), see the discussion in Blinnikov et al. (1998). We do not pretend yet that the light curves in this approximation are reproduced absolutely correctly, especially later than $`2`$ months after the explosion. But our experience in the light curve modeling of other types of supernovae (Blinnikov et al. 1998, Blinnikov 1999, Blinnikov et al. 1999) shows that our results are very reliable for a month or so after the maximum light. One can see in Fig. 4 below, that the photosphere in B band for the day 35 in our models is still in the outer layers, so the bulk of the flux is born in the conditions close to LTE. What is also important, we are interested here mostly in relative changes of the light curves for two classical regimes of burning. We believe that the relative difference is reproduced quite reliably in our computations.
## Burning modes in thermonuclear supernovae
Currently, there is no doubt that the light of SNe Ia is produced in the decays of <sup>56</sup>Ni to <sup>56</sup>Co and then to <sup>56</sup>Fe. The sufficient amount of initial <sup>56</sup>Ni as well as the required energy of the explosion can be naturally produced during the thermal runaway in a degenerate carbon-oxygen mixture. Yet, the regime of the nuclear burning in supernovae is still a controversial issue.
Arnett (1969) was the first to model the supersonic combustion, i.e. detonation, in supernovae. Later, Ivanova et al. (1974) obtained a sub-sonic flame (deflagration) propagating in a spontaneous regime with pulsations and switching subsequently to the detonation. Nomoto et al. (1976) modeled deflagration propagating due to a convective heat transfer with a parameterized flame speed. Both detonation and deflagration have their merits and problems in explaining the supernova phenomenon (see, e.g., the review of Woosley & Weaver 1986).
A prompt detonation, born near the center of a star, would incinerate almost the whole star to ashes consisting of iron peak elements (Arnett 1969). This is in clear contradiction with observations which show that intermediate mass elements are abundant in the ejecta. Thus at some stage (if not always) the burning must be sub-sonic. Here another problem arises: it is clear that the combustion must be much faster than suggested by an analysis of propagation of a laminar one-dimensional flame (Timmes & Woosley 1992).
From the microscopic point of view the one-dimensional nuclear flame is a wave described essentially in the same way as that formulated by Zeldovich & Frank-Kamenetsky (1938) in spite of complications introduced by nuclear kinetics and a very high conductivity of dense presupernova matter. It is found that the conductive flame propagates in a presupernova far too slowly to explain the supernova outburst correctly: the flame Mach number is of the order of a few percent and less (Timmes & Woosley 1992).
A natural way to accelerate the fuel consumption is the development of instabilities inherent to the flame front. As explained in the classical paper by Landau (1944), the hydrodynamic instability leads to wrinkling or roughening of the front surface, i.e., to increasing of its area with respect to the smooth front, and consequently to acceleration of the flame propagation. In extreme cases, the instabilities can lead to a transition from the regime of slow flame propagation to the regime of detonation. The role of the Landau instability in supernovae was pointed out by Blinnikov et al. (1995) and Niemeyer & Hillebrandt (1995b). As the instability grows, the front becomes wrinkled and fractal. The detailed consideration of non-linear stage of Landau instability and the calculation of the fractal dimension of the flame front for that case is given by Blinnikov & Sasorov (1996).
Since the flame propagates in the gravity field and the burned ashes have lower density than the unburned fuel, the Rayleigh–Taylor (RT) instability is often considered to be the main process governing the corrugation of the front (see Müller & Arnett 1986, Woosley 1990a, Woosley 1990b, Livne & Arnett 1993, Khokhlov 1993). The RT instability gives birth to a turbulent cascade providing for an acceleration to the flame and additional difficulties in modeling the SN event (Niemeyer & Hillebrandt 1995a, Niemeyer 1995, Niemeyer & Woosley 1997). The turbulent mixing of hot products of burning with the cold fuel might trigger a detonation after a period of slow expansion of the partly burned degenerate star (Niemeyer & Woosley 1997). Here the transition to the detonation may happen via a stage of supersonic spontaneous flame propagation (Blinnikov & Khokhlov 1986). However, recent computations by Reinecke et al. (1999) show that the RT instability alone is not sufficient for the explosion: it drives a flame front too weakly and the needed subsonic speed is not achieved. Additional sources for the turbulence or alternative routes for accelerating flames seem to be needed (Niemeyer 1999). Even if detonation is established, it cannot be described by a classical Zeldovich–von Neumann wave, which is unstable in the conditions of SNe Ia explosions (see Imshennik et al. 1999 and references therein).
It is clear that a full direct modeling of the burning is not possible because the involved scales range from the thickness of a flame $`10^5`$ cm up to the dimension of a white dwarf, $`10^9`$ cm. Moreover, in all cases where turbulence is involved, the chaotic behavior sets in inevitably, and a subtle difference in initial conditions can lead to drastically unsimilar outcomes. This is well known from terrestrial experiments: after a series of tens or even hundreds cases of slow burning, sometime a detonation sets in when all the conditions of the experiment are quite the same (e.g., Mader 1979).
## Theoretical Models of Supernovae and Light Curves
Initial models corresponding to two different regimes of burning were used in our calculations, namely, the classical deflagration model W7 (Nomoto et al. 1984) and a delayed detonation model DD4 (Woosley & Weaver 1994). The main parameters of these two models are compared in Table 1, and their resulting chemical compositions, in Table 2. Figures 1, 2 display the distributions of density, velocity, and basic chemical elements in the ejecta for those two models. It is clear from the figures and tables that the hydrodynamical parameters are approximately the same for both models, while differences in chemistry are due to the relevant burning mode in a white dwarf (though, the amount of <sup>56</sup>Ni is almost the same in both models, so maxima in luminosity are close). Both W7 and DD4 are simple 1D models, and each individual SN Ia can be a more complicated 3D fractal event in reality. Yet our computed light curves have a good resemblance to the observed ones and reproduce the broad band fluxes quite well.
The curves in the Fig. 3 display the numerical results for the models W7 and DD4, while asterisks show the observational points for SN 1998bu (Suntzeff et al. 1999; see also Meikle & Hernandez 1999, and Jha et al. 1999) and crosses the template fluxes for SN 1992A (Hamuy et al. 1996b). The results demonstrate clearly that in spite of similar maximum luminosities, the models have very unequal decline rates in the B filter, though they are much more similar in the V band.
It is not easy to explain this scatter in the predicted decline rates by a single factor. The outgoing flux of the SN Ia explosion is influenced by many physical parameters. The differences in all parameters for the models are small for each instant. Yet the combination of the minute differences seems to be sufficient to change the light curve shapes. For example, the distributions of <sup>56</sup>Ni in outer layers, as well as lighter elements closer to the center, are not the same, as shown in the Fig. 2. This causes slight changes in gamma-ray deposition and in diffusion of soft photons, thus implying unsimilar temperature distributions. The latter are shown in Fig. 4 for the day 35 after the explosion. The solid dots mark the layers where the optical depth becomes equal to $`2/3`$, if measured from the outer boundary of the ejecta. The temperature for W7 is 10% higher, and the radius is also a bit larger than for DD4 in this point. This can explain quite well the differences in B fluxes for this epoch (see Fig. 3).
Let us compare our computed light curves with observations of real supernovae. There are several quantitative relations to describe their shapes. One of the most popular is the dependence suggested by Phillips (1993), $`M_{\mathrm{max}}(\mathrm{\Delta }m_{15}^B)`$, where $`\mathrm{\Delta }m_{15}^B`$ is a drop of the B flux (in stellar magnitudes) 15 days after the maximum. The linear regression fit
$$M_{\mathrm{max}}=a+b\mathrm{\Delta }m_{15}^B$$
is usually used for the interpretation of observations. The values of $`a`$ and $`b`$ vary strongly, when obtained with different sets of SNe Ia by different authors. Moreover, the dispersion of their values is large even within one set. We are more interested in the value of $`b`$, the slope of the light curve. Its variation is rather large, e.g., Phillips (1993) obtains $`b^B=2.698\pm 0.359`$ from the observations of 9 nearby supernovae. In contrast, analysis by Hamuy et al. (1995) of only six supernovae from the same set yields $`b^B=1.624\pm 0.582`$. This scatter cannot be explained only by the observational errors, they are not so big. See, however, Hamuy et al. (1996a) where they obtain much smaller scatter and even lower value of $`b^B=0.784\pm 0.182`$ using 26 Calán/Tololo distant supernovae. The value of $`a`$ is influenced by the Hubble parameter $`H_0`$: $`a=a_0+5\mathrm{lg}(H_0/85)`$ for Phillips (1993), and $`a=a_0+5\mathrm{lg}(H_0/65)`$ for Hamuy et al. (1996a). In reality, supernovae do not lie on a straight line, they occupy an extended region on the diagram $`M_{\mathrm{max}}(\mathrm{\Delta }m_{15}^B)`$. Most probably, this reflects real variations of the explosion physics. One can come to the same conclusion after analyzing the correlations between other key parameters of SN Ia, like <sup>56</sup>Ni mass, peak bolometric luminosity, rise time to the maximum light, explosion energy, etc (Nadyozhin, 1997).
The slope of our calculated light curves (especially for DD4 model) and the luminosity are comparable with the observable values. We can estimate the decline rate from the results of Phillips (1993) and Hamuy et al. (1995) with the formula $`\mathrm{\Delta }m_{15}^B=(M_{\mathrm{max}}^Ba)/b`$. Comparing this with the rates for W7 and DD4 in our computations we find for $`H_0=85`$ km/s/Mpc:
$$\mathrm{\Delta }m_{15}^B=\{\begin{array}{cccc}1.04(\text{Phillips})\hfill & 0.94(\text{Hamuy95})\hfill & 1.07(\text{model DD4})\hfill & \text{for }M_{\mathrm{max}}^B=18.9;\hfill \\ 1.12(\text{Phillips})\hfill & 1.06(\text{Hamuy95})\hfill & 0.66(\text{model W7})\hfill & \text{for }M_{\mathrm{max}}^B=18.7.\hfill \end{array}$$
So, we see that the model W7 behaves in the sense opposite to the Pskovskii-Phillips relation: being a bit weaker than DD4 it declines slower, instead of fading down faster.
If we take more recent data from Hamuy et al. (1996a) and $`H_0=65`$ km/s/Mpc, then we get higher values for $`\mathrm{\Delta }m_{15}^B`$, predicted by ‘inverse Phillips’ relations, $`\mathrm{\Delta }m_{15}^B=(M_{\mathrm{max}}^Ba)/b`$, for the $`M_{\mathrm{max}}^B`$ found in our modeling, but this reflects only the fact that both models are underluminous for the long cosmological distance scale. We are interested in relative differences between two models, so it is more instructive to compare the absolute magnitudes from the direct formula $`M_{\mathrm{max}}=a+b\mathrm{\Delta }m_{15}^B`$. We present the comparison for the two values of the Hubble parameter. For $`H_0=85`$ km/s/Mpc:
$$M_{\mathrm{max}}^B=\{\begin{array}{cccc}18.8(\text{Phillips})\hfill & 18.7(\text{Hamuy96})\hfill & 18.9(\text{model DD4})\hfill & \text{for }\mathrm{\Delta }m_{15}^B=1.07;\hfill \\ 19.9(\text{Phillips})\hfill & 19.0(\text{Hamuy96})\hfill & 18.7(\text{model W7})\hfill & \text{for }\mathrm{\Delta }m_{15}^B=0.66\hfill \end{array}$$
Thus, for the short distance scale DD4 is OK, while W7 will look underluminous. For $`H_0=65`$ km/s/Mpc:
$$M_{\mathrm{max}}^B=\{\begin{array}{cccc}19.4(\text{Phillips})\hfill & 19.3(\text{Hamuy96})\hfill & 18.9(\text{model DD4})\hfill & \text{for }\mathrm{\Delta }m_{15}^B=1.07;\hfill \\ 20.5(\text{Phillips})\hfill & 19.6(\text{Hamuy96})\hfill & 18.7(\text{model W7})\hfill & \text{for }\mathrm{\Delta }m_{15}^B=0.66\hfill \end{array}$$
Now, both models are too weak, and this means that for the long distance scale a standard model, like DD4, which is 0.4 magnitudes underluminous according to the peak luminosity–decline rate relation of Hamuy et al. (1996a), must have perhaps 30 to 40% more <sup>56</sup>Ni (Arnett 1982, Cappellaro et al. 1997, Mazzali et al. 1998). At the same time W7 is underluminous already by 0.9 magnitudes. If we assume the same relative difference between delayed detonation and deflagration models with the same enhancement of <sup>56</sup>Ni abundance, then the W7-like events will still look 0.5 magnitudes underluminous. This can have drastic implications for the use of SNe Ia in cosmology.
One can see in the Fig. 3 that the behavior of both calculated light curves is qualitatively correct when compared with a couple of well studied individual SNe Ia. In particular, the rise time corresponds to the observed values: the flux of observed SNe Ia grows up to the maximum during 15–20 days (Maza 1981, Pskovskii 1984, Bartunov & Tsvetkov 1997). As it is shown in Table 3, our calculations predict 16–17 days for this value in the $`B`$ band, and this is in much better agreement with the observations than yielded by previous computations of Höflich et al. (1997). (Their fluxes grow for only 10–15 days for the majority of models, including W7.) One can also note the realistic behavior of our computed light curves in different spectral bands. The decline rate is less steep for redder bands in both models (see a simple analytical explanation for this given by Arnett 1982) But it is also evident that the deflagration model fades down in the B-band too slowly, and does not fit the decline rates for typical type Ia supernovae, while the delayed detonation model agrees with the observations very well.
Until recently W7 has been in frequent use in many papers where light curves of SNe Ia were presented, as well as when supernova remnants were calculated subsequently. The reason for such a popularity is that this model seemed to reproduce the observed fluxes quite well (see, e.g., Höflich 1995 and Eastman 1997). Our results for W7 (especially in the B band) differ from the ones obtained in the previously published papers. This can be explained most probably by the differences in including the expansion opacity (Karp et al. 1977) into the energy equation. The correct treatment of the energy equation is described by Blinnikov (1996, 1997) who has shown that there is no effect of expansion opacity when line extinction is taken into account in the energy equation for photons. The expansion effect is important only in the flux equation. Extinction in lines is much stronger than other sources of opacity for the SNe Ia. So when the effect by Karp et al. is included into the energy equation, this leads to unjustifiable enhancement of the energy exchange between radiation and matter. The photon energy is in this case artificially redistributed towards longer wavelengths, and escapes the system much easier. As a result, the flux from a supernova grows up and falls down faster, and the slope of the light curve becomes steeper.
To check the importance of this effect, we have calculated the model W7 once more, but this time we treated the extinction (including the one in lines and the expansion effect) as a pure absorption. In such a way, we have made the energy exchange stronger, and so we have spoiled the energy equation. Because of this, the decline rate of the light curve in the B band has become “better”, although the rise time is shortened and its agreement with the observations is much worse. One can notice here that the results of the calculations of our “spoiled” W7 model resemble those by Höflich. Fig. 5 shows how the enhanced energy exchange influences the light curve of W7. It is easy to see from the figure that although the flux in shorter wavelengths could be called “good” (but with shortened rise time; see also Table 3), it would be absolutely untrue regarding the flux in I. The secondary maximum in I band, which is observed in almost all real SNe Ia, has disappeared in the new light curve. Physical processes responsible for the presence of this feature in the light curves should yet be investigated. Nevertheless, the lack of the secondary maximum clearly demonstrates that some important processes were not included (or were included incorrectly) in the treatment of radiative transport.
Thus, we are able to reproduce the results of other researchers when the wrong energy equation is used in our code. When the appropriate physics is included instead, the decline rate of the deflagration model W7 is too slow, and does not correspond to the behavior of a typical SN Ia. Two possible reasons for such a result could be proposed: either the real explosions are not due to the pure deflagration or, most probably, too many simplifications were supposed while constructing W7. More complicated deflagration model could, perhaps, describe the reality better (cf. Niemeyer & Woosley 1997, Reinecke et al. 1999).
## Using SNe Ia in Cosmology
Due to the maximum light – decline rate dependence, SNe Ia are now in active use for finding cosmological distances. Their observations can help to determine the geometry of the Universe, as well as to check the viability of some cosmological models. The idea is to compare the photometric distance to a supernova with its value in a given cosmological model when its redshift $`z`$ is measured. The simple formula $`d_{\mathrm{ph}}=\left(L/4\pi F\right)^{1/2}`$ defines the photometric distance to an observed object if its absolute luminosity $`L`$ is known. At the same time, the photometric distance can be calculated theoretically: if $`z`$ is fixed, then $`d_{\mathrm{th}}`$ is a function of the cosmological parameters $`H_0`$, $`\mathrm{\Omega }_\mathrm{M}`$ and $`\mathrm{\Omega }_\mathrm{\Lambda }`$ (Carroll et al. 1992):
$$d_{\mathrm{th}}=\frac{c}{H_0}(1+z)\frac{1}{\sqrt{\mathrm{\Omega }_k}}\mathrm{sinh}\left\{\sqrt{\mathrm{\Omega }_k}\underset{0}{\overset{z}{}}\left[\mathrm{\Omega }_\mathrm{M}(1+z)^3+\mathrm{\Omega }_\mathrm{\Lambda }+\mathrm{\Omega }_k(1+z)^2\right]^{1/2}dz\right\}.$$
Here $`\mathrm{\Omega }_k=1\mathrm{\Omega }_\mathrm{M}\mathrm{\Omega }_\mathrm{\Lambda }`$, and for $`\mathrm{\Omega }_k<0`$ the hyperbolic sine ($`\mathrm{sinh}`$) transforms, according to elementary formulae, to the trigonometric sine ($`\mathrm{sin}`$), while $`\sqrt{\mathrm{\Omega }_k}`$ should read $`\sqrt{|\mathrm{\Omega }_k|}`$. For $`\mathrm{\Omega }_k=0`$ the limit $`\mathrm{\Omega }_k0`$ is easily taken, so $`\mathrm{sinh}`$ disappears from the expression for $`d_{\mathrm{th}}`$, and only the integral is left.
At present, several search groups in the world do observe the high-$`z`$ supernovae. The peak luminosity – decline rate relation is used by them to estimate the cosmological parameters from these observations. The relation is calibrated with the help of nearby supernovae samples, since rich data are available in this case (see, e.g., Phillips 1993), but it is used to interpret the observations of the much more distant supernovae as well. We already discussed that even nearby examples are not fully homogeneous, though they seem to be equivalent at first glance. The investigations of the influence of the evolutionary effects on the light curve shape are only at the beginning so far.
The maximum light – decline rate relation could be different for younger galaxies. To understand how it could change, one should know how the shape of the light curve depends on the initial chemical composition of the exploded white dwarf. And what are the initial conditions in general, what are the progenitors of supernovae, do those conditions change with the age of Universe?
The influence of the initial conditions just before the explosion on the emission of the SNe Ia was studied by Domínguez et al. (1999). The role of evolution of SNe Ia is discussed by Riess et al. (1998), Schmidt et al. (1998) and Drell et al. (1999). (See also the discussion on different pre-supernova models in von Hippel et al. 1997, Livio 1999, and Umeda et al. 1999.)
Schmidt et al. (1998) compare the distances to 8 SNe Ia in early type galaxies with 19 in late type ones and find no significant difference. But here are two points that need be yet clarified. First, it is hard to select the progenitors of the same ages in the late type hosts. Second, the statistics of such progenitors is too small. One has to build distributions of SNe Ia at different $`z`$ with good statistics and to compare them. The statistical analysis by Drell et al. (1999) shows convincingly that it is virtually impossible to distinguish the effects of evolution from the effects of the geometry of the Universe expressed through $`\mathrm{\Omega }_\mathrm{M}`$ and $`\mathrm{\Omega }_\mathrm{\Lambda }`$. Drell et al. (1999) write that without physical understanding, the additional data, such as source spectra, cannot be used to argue against evolution, while observers pretend that there is no compelling evidence for evolution in the SNe Ia samples, based on these data. The physical insight in the problem is being developed, e.g. by Höflich, Wheeler, and Thielemann 1998, Domínguez et al. 1999, Umeda et al. 1999.
Domínguez et al. (1999) find that due to the changes in metallicity and C/O ratio, the light curve shape can also change. This leads to the drift of $`0.2^\mathrm{m}`$ in distance modulus determined with the averaged luminosity – light curve shape relation. This effect was considered for only one set of delayed detonation models, and the results were found to be influenced mostly via the change of <sup>56</sup>Ni mass.
All previous work on the evolution of the SN Ia progenitors is based on the assumption that chemistry and hydrodynamic structure of the pre-supernovae exactly fix the behavior of the light curve. This is a deterministic approach. With this approach the net effect is not necessarily large. As we find in this work, the supernova emission may also depend drastically on the mode of burning, and this is most probably the factor that cannot be predicted in a deterministic way in principle. The main difference in our approach is a conjecture that even for the same initial conditions (or minute differences in the initial conditions) the outcome of the explosion can be vastly different. For the same chemistry and for the same age of progenitors there can be a wide distribution of different SN Ia events, and this is confirmed by the observations (see, e.g, Hamuy et al. 1995, 1996a). If this distribution (there is no sense to speak about individual events!) changes with $`z`$ then one can obtain a wrong value of $`\mathrm{\Lambda }`$. Fig. 3 is a bright illustration of the possibility to erroneously determine the absolute flux from a SN Ia and, therefore, its distance. It shows the strong deviation in the decline rate among the SNe Ia with similar maximum light due to the differences in the burning regime, which is hardly predictable in principle, as was discussed above. One can only predict a trend for the preferred mode of burning, which can indeed depend on the age and the chemistry of the progenitor.
Let us assume that W7 and DD4 represent two limiting cases for real Chandrasekhar-mass explosions. There can be a continuous set of supernovae between those limits, or there can be a clustering to the two extreme cases. This does not matter for our purpose now. What matters is the indeterminacy principle governing the behavior of the flame inside a star. Thus, when the initial conditions are fixed, there can be a trend in the distribution of the real events to be more “DD4-like” or more “W7-like”. If, for instance, we have more W7-like events at high $`z`$ then we can obtain a wrong result when we try to derive the distance, and hence, the cosmological constant from SNe Ia, based on the local calibration samples, because of different luminosity – slope dependence in the past.
In the recent work on the cosmological supernovae (Riess et al. 1998, Perlmutter et al. 1999) the conclusion is reached, that the distance to the high-$`z`$ supernovae is about 10–15% larger than expected for the model with $`\mathrm{\Omega }_\mathrm{M}=0.2`$, $`\mathrm{\Omega }_\mathrm{\Lambda }=0`$. It is claimed that a model with $`\mathrm{\Omega }_\mathrm{\Lambda }>0`$ is needed to avoid such a deviation. In reality, $`\mathrm{\Lambda }`$ could be exactly zero if the fraction of slowly burning supernovae would be larger for higher $`z`$. The true luminosity would be then smaller than found from the peak luminosity – light curve shape calibrations for the local sample, and this could mimic the effect of positive $`\mathrm{\Lambda }`$. We are persuaded that so far there is no sufficient statistics of distant supernovae to make firm conclusions on the geometry of the Universe. Not only the non-zero $`\mathrm{\Lambda }`$ can reconcile the theoretical distances $`d_{\mathrm{th}}`$ with the observed photometric ones, but also the evolution of the probability of the burning regimes, making the W7-like explosions in the present epoch less likely than in the younger Universe.
## Conclusions
We have chosen as initial configurations for SNe Ia two models of carbon-oxygen white dwarfs with different regimes of burning during the explosion, and simulated the light curves till $`2`$ months after the burst. We find that the post-maximum decline rate is very sensitive to the burning mode. Even minute differences in the <sup>56</sup>Ni distribution seem to be sufficient to heat the material differently inside the ejecta and to produce different slopes of the light curve thereby, despite the similar maximum brightness.
The combustion theory on its modern level is not able to predict the mode of burning if the initial conditions are fixed. It is conceivable that this prediction is impossible in principle. The different burning modes, in turn, can result in unlike temperature distributions. If this is true, then the initial conditions do not determine the absolute maximum – decline rate dependence exactly. They fix only the probability to obtain one or another slope of the light curve for the fixed maximum brightness. This probability can be changed by variation in the initial conditions.
One can speculate that the future combustion theory will be able to predict only a proportion between “fast” and “slow” supernovae, and this proportion can be very sensitive to the initial conditions. Astronomers should not wait for such an advancement of the theory, but could already now check the distribution function of maximum brightness over the decline rate at different $`z`$.
What to do in this situation? First, the statistics must be improved to such an extent that, at a given high redshift, one could compare the distributions of SNe on the maximum luminosity – decline rate diagram. I.e., at each $`z`$ one should build the distributions of the decline rates in a broad range of the peak luminosity. The hardest thing is, of course, to observe the intrinsically weak SNe Ia for sufficiently long time. The distributions found for high $`z`$ could coincide with the one for nearby supernovae, but could also differ from it. The latter will mean that nearby statistical relations are not applicable to the distant events. But even in the case when the brightest supernovae at, for instance, $`z=1`$ have the same distribution of their light curve slopes as the brightest samples at $`z<0.1`$, this will tell no word about the coincidence of their absolute magnitudes if one does not use any additional data.
Second, the individual differences are larger in the I band. This spectrum interval is more sensitive to the explosion physics. Perhaps, the influence of the burning mode on the I band will become more transparent in future theoretical investigations. The theorists should simulate the light curves in I for extended range of models and compare them with the observations. It is hard to measure the rest frame I flux for large $`z`$, since then it goes to the far infrared on Earth. But this must be possible with new orbiting telescopes, like the Next Generation Space Telescope (Dahlén & Fransson 1998, Rees 1998). The observations in I should help to put useful constraints on the supernova models for each real explosion. But the thorough study is still needed both to understand the role of the non-LTE effects in the emission of SNe Ia in near infrared and to develop the realistic 3D models of thermonuclear burning. Ultimately, the radiative transfer computations for those 3D-models will show the level of accuracy of 1D models considered here.
We are grateful to P.Ruiz-Lapuente, S.Woosley, K.Nomoto, K.Iwamoto for providing us with the models used in the calculations, to P.Lundqvist for the possibility of doing calculations at Stockholm Observatory and for his warm hospitality during our stay in Sweden. We are also thankful to D.Tsvetkov for very useful comments. The work was supported in part by INTAS–RFBR project 95-0832 ‘Thermonuclear Supernovae’, by ISTC grant 370–97, by the Royal Swedish Academy, and by the University of Barcelona.
References
Arnett W.D. 1969, Astrophys. Space Sci., 5, 180
Arnett W.D. 1982, ApJ, 253, 785
Baron E., Hauschildt P.H., Nugent P., & Branch D. 1996, MNRAS, 283, 297
Bartunov O.S. & Tsvetkov D.Yu. 1986, Astrophys. Space Sci., 122, 343
Bartunov O.S. & Tsvetkov D.Yu. 1997, in Thermonuclear Supernovae (ed. Ruiz-Lapuente P. et al.), Dordrecht: Kluwer Acad. Pub., p.87
Blinnikov S.I. 1996, Astron. Letters 22, 79 (translated from Russian: Pisma Astron. Zh., 1996, 22, 92)
Blinnikov S.I. 1997, in Thermonuclear Supernovae (ed. Ruiz-Lapuente P. et al.), Dordrecht: Kluwer Acad. Pub., p.589
Blinnikov S.I. 1999, Astron. Letters, in press (translated from Russian: Pisma Astron. Zh., 1999)
Blinnikov S.I. & Bartunov O.S. 1993, A&A, 273, 106
Blinnikov S.I., Eastman R., Bartunov O.S., Popolitov V.A., & Woosley S.E. 1998, ApJ, 496, 454
Blinnikov S.I. & Khokhlov A.M. 1986, Sov. Astron. Lett., 12, 131 (translated from Russian: Pisma Astron. Zh., 1986, 12, 318)
Blinnikov S.I., Lundqvist P., Bartunov O.S., Nomoto K., & Iwamoto K. 1999, ApJ, in press
Blinnikov S.I. & Sasorov P.V. 1996, Phys. Rev. E, 53, 4827
Blinnikov S.I., Sasorov P.V., & Woosley S.E. 1995, Space Science Rev., 74, 299
Branch, D., 1987 ApJ, 316, 81L
Barbon, R., Rosino, L. and Iijima, T. 1989, A&A, 220, 83
Cappellaro E., Mazzali P.A., Benetti S., Danziger I.J., Turatto M., Della Valle M., & Patat F. 1997, A&A, 328, 203
Carroll S.M., Press W.H., & Turner E.L. 1992, Ann. Rev. Astron. Astrophys., 30, 499
Dahlén T. & Fransson C. 1998, in The Next Generation Space Telescope: Science Drivers and Technological Challenges, (astro-ph/9809379)
Domínguez I., Höflich P., Straniero O., & Wheeler C. 1999, in Future Directions of Supernovae Research: Progenitors to Remnants. Assergi (Italy) Sep-Oct 1998 (astro-ph/9905047)
Drell P.S., Loredo T.J., & Wasserman I. 1999, ApJ, in press (astro-ph/9905027)
Eastman R.G. 1997, in Thermonuclear Supernovae (ed. P.Ruiz-Lapuente et al.), Dordrecht: Kluwer Acad. Pub., p.571
Eastman R.G. & Pinto P.A. 1993, ApJ, 412, 731
Filippenko A.V., et al. 1992, ApJ, 384, L15
Garnavich P.M. et al. 1998a, ApJ, 493, L53
Garnavich P.M. et al. 1998b, ApJ, 509, 74
Hamuy M., Phillips M.M., Maza J., Suntzeff N.B., Schommer R.A., & Aviles R. 1995, AJ, 109, 1
Hamuy M., Phillips M.M., Suntzeff N.B., Schommer R.A., Maza J., & Aviles R. 1996a, AJ, 112, 2391
Hamuy M., Phillips M.M., Suntzeff N.B., Schommer R.A., Maza J., Smith R.C., Lira P., & Aviles R. 1996b, AJ, 112, 2438
Hamuy M. & Pinto P.A. 1999, AJ, 117, 1185
von Hippel T., Bothun G.D., & Schommer R.A. 1997, AJ, 114, 1154
Höflich P. 1995, ApJ, 443, 89
Höflich P., Khokhlov A., Wheeler J.C., Phillips M.M., Suntzeff N.B., & Hamuy M. 1996, ApJ, 472, L81
Höflich P., Khokhlov A., Wheeler J.C., Nomoto K., & Thielemann F.K. 1997, in Thermonuclear Supernovae (ed. Ruiz-Lapuente P. et al.), Dordrecht: Kluwer Academic Publishers, p.659
Höflich P., Wheeler J.C. & Thielemann F.K. 1998, ApJ, 495, 617
Ivanova L.N., Imshennik V.S., & Chechetkin V.M. 1974, Astrophys. Space Sci., 31, 497
Imshennik V.S., Kal’yanova N.L., Koldoba A.V., & Chechetkin V.M. 1999, Astronomy Letters, 25, 206
Jha S. et al. 1999, to appear in ApJS (astro-ph/9906220)
Karp A.H., Lasher G., Chan K.L., & Salpeter E.E. 1977, ApJ, 214, 161.
Khokhlov A.M. 1993, ApJ, 419, L77.
Kim A.G. et al. 1997, ApJ, 476, L63
Landau L.D. 1944, ZhETF, 14, 240; Acta Physicochim. USSR, 1944, 19, 77
Leibundgut B. et al. 1993, AJ, 105, 301
Livio M. 1999, astro-ph/9903264
Livne E. & Arnett W.D. 1993, ApJ, 415, L107
Mader C.L. 1979, Numerical Modeling of detonations, Bercley: Univ. of California Press
Maza J. 1981, IAU Circ. 3583
Mazzali P.A., Cappellaro E., Danziger I.J., Turatto M., & Benetti S. 1998, ApJ 499, L49
Meikle P. & Hernandez M. 1999, to appear in the Journal of the Italian Astronomical Society (astro-ph/9902056)
Müller E. & Arnett W.D. 1986, ApJ, 307, 619
Nadyozhin D.K. 1994, ApJS, 92, 527
Nadyozhin D.K. 1997, colloquium “Supernovae and Cosmology”, Basel
Niemeyer J.C. 1995, Ph.D. Thesis, Garching: MPA-911
Niemeyer J.C., 1999, ApJL in press (astro-ph/9906142)
Niemeyer J.C. & Hillebrandt W. 1995a, ApJ, 452, 769
Niemeyer J.C. & Hillebrandt W. 1995b, ApJ, 452, 779
Niemeyer J.C. & Woosley S.E. 1997, ApJ, 475, 740
Nomoto K., Sugimoto D., & Neo S. 1976, Astrophys. Space Sci., 39, L37
Nomoto K., Thielemann F.–K., & Yokoi K. 1984, ApJ, 286, 644
Nørgaard-Nielsen H.U. et al. 1989, Nature, 339, 523
Perlmutter S. et al. 1997, ApJ, 483, 565
Perlmutter S. et al. 1999, ApJ, 517, 565
Phillips M.M. et al. 1987, PASP, 99, 592
Phillips M.M. 1993, ApJ, 413, L105
Pskovskii Yu.P. 1977, Sov. Astronomy, 21, 675 (translated from Russian: Astron. Zh. 1977, 54, 1188)
Pskovskii Yu.P. 1984, Sov. Astronomy, 28, 658 (translated from Russian: Astron. Zh. 1984, 61, 1125)
Rees M.J. 1998, in The Next Generation Space Telescope: Science Drivers and Technological Challenges, (astro-ph/9809029)
Reinecke M., Hillebrandt W., & Niemeyer J.C. 1999, A& A, in press (astro–ph/9812120)
Riess A.G. et al. 1998, AJ, 116, 1009
Ruiz-Lapuente P. 1997, in proc. of Casablanca Winter School From Quantum Fluctuations to Cosmological Structures, December 1996, (ed. D. Valls-Gabaud, M. Hendry, P. Molaro and K. Chamcham), ASP Conference Series, vol. 126, p. 207
Ruiz-Lapuente P., Blinnikov S., Canal R., Mendez J., Sorokina E., Visco A., & Walton N. 1999, in Cosmology with type Ia Supernovae, proc. of the workshop held at the University of Chicago, October 29 - 31, 1998 (ed. J. Niemeyer & J. Truran)
Sandage G., Tammann G.A. 1997, in Critical Dialogues in Cosmology, (ed. N. Turok), Singapore: World Scientific, p. 130
Schmidt B.P. et al. 1998, ApJ, 1998, 507, 46
Sorokina E.I., Blinnikov, S.I., Ruiz-Lapuente P., & Nomoto, K. 1999, in preparation
Suntzeff N.B. et al. 1999, AJ, 117, 1175
Swartz D.A., Sutherland P.G., & Harkness R.P. 1995, ApJ, 446, 766
Timmes F.X. & Woosley S.E. 1992, ApJ, 396, 649
Umeda H., Nomoto K., Kobayashi C., Hachisu I., Kato M. 1999, ApJ Letters, accepted (astro-ph/9906192)
Woosley S.E. 1990a, in Supernovae (ed. Petschek A.G.), Astron. Astrophys. Library, p.182
Woosley S.E. 1990b, in Gamma-ray Line Astrophysics (ed. Durouchoux P., Prantzos N.), New York: Am. Inst. of Phys., p.270
Woosley S.E. 1997, in Thermonuclear Supernovae (ed. Ruis-Lapuente P. et al.), Dordrecht: Kluwer Academic Publishers, p.313
Woosley S.E. & Weaver T.A. 1986, Ann. Rev. Astron. Astrophys., 24, 205
Woosley S.E. & Weaver T.A. 1994, in Supernovae (ed. Audouze J. et al.), Amsterdam: Elsevier Science Publishers, p.63
Zeldovich Ya.B. & Frank-Kamenetsky D.A. 1938, Acta Physicochim. USSR, 9, 341
|
no-problem/9906/hep-th9906022.html
|
ar5iv
|
text
|
# Holography in General Space-times
## 1 Holographic Principle
### 1.1 The holographic principle for general spacetimes
In Ref. a covariant entropy bound was conjectured to hold in general space-times. The bound can be saturated, but not exceeded, in cosmology and in collapsing regions. Applied to finite systems of limited self-gravity, it reduces to Bekenstein’s bound . For a $`D`$-dimensional Lorentzian space-time, the covariant bound can be stated as follows:
##### Covariant Entropy Conjecture
Let $`A`$ be the area of a connected $`(D2)`$-dimensional spatial surface $`B`$. Let $`L`$ be a hypersurface bounded by $`B`$ and generated by one of the four null congruences orthogonal to $`B`$. Let $`S`$ be the total entropy contained on $`L`$. If the expansion of the congruence is non-positive (measured in the direction away from $`B`$) at every point on $`L`$, then $`SA/4`$.
The conjecture can be viewed as a generalization of an entropy bound proposed by Fischler and Susskind . It differs in that it considers all four light-like directions and selects some of them by the criterion of non-positive expansion. Several concepts crucial to a light-like formulation were recognized earlier by Corley and Jacobson , who distinguished between “past and future screen maps” and drew attention to the importance of caustics. We should also point out a number of recent proposals for entropy bounds in cosmology (see Sec. 3.4).
The covariant entropy bound is manifestly invariant under time reversal. This property cannot be understood if the bound applies only to thermodynamic entropy. One is thus forced to interpret the bound as a limit on the number of degrees of freedom ($`N_{\mathrm{dof}}`$) that constitute the statistical origin of any thermodynamic entropy that may be present on $`L`$. Since no assumptions about the microscopic properties of matter were made, the limit is fundamental . There simply cannot be more independent degrees of freedom on $`L`$ than $`A/4`$, in Planck units. This conclusion compels us to embrace the holographic conjecture of ’t Hooft and Susskind , and it motivates the following background-independent formulation of their hypothesis:
##### Holographic Principle
Let $`A`$ be the area of a connected $`(D2)`$-dimensional spatial surface $`B`$. Let $`L`$ be a hypersurface bounded by $`B`$ and generated by one of the four null congruences orthogonal to $`B`$. Let $`𝒩`$ be the number of elements of an orthonormal basis of the quantum Hilbert space that fully describes all physics on $`L`$. If the expansion of the congruence is non-positive (measured in the direction away from $`B`$) at every point on $`L`$, then $`𝒩e^{A/4}`$.
Simplifying slightly,<sup>1</sup><sup>1</sup>1We thank Gerard ’t Hooft for suggesting a formulation in terms of Hilbert space. one could state that $`N_{\mathrm{dof}}A/4`$, where $`N_{\mathrm{dof}}`$ is the total number of independent quantum degrees of freedom present on $`L`$.
The holographic principle thus assigns at least two light-like hypersurfaces to any given spatial surface $`B`$ and bounds $`N_{\mathrm{dof}}`$ on those hypersurfaces. The relevant hypersurfaces can be constructed as follows (see Ref. for a more detailed discussion). There will be four families of light-rays orthogonal to $`B`$: a past-directed and a future-directed family on each side of $`B`$. Consider only the families with non-positive expansion away from $`B`$. Generically there will be two such families, but if the expansion is zero in some directions, there may be as many as three or four. Pick one of the allowed families and follow each light-ray until the expansion becomes positive or a boundary of space-time is reached. The null hypersurface thus generated is called a light-sheet. $`N_{\mathrm{dof}}`$ on a light-sheet of $`B`$ will not exceed a quarter of the area of $`B`$.
In general, the holographic principle associates the area of a surface $`B`$ with $`N_{\mathrm{dof}}`$ on null hypersurfaces, not spatial regions, bounded by $`B`$. Under certain conditions, however, the bound does apply to space-like hypersurfaces as well. This was shown in Ref. for the covariant entropy bound. For the holographic principle, the derivation can be repeated, with “entropy” replaced by “$`N_{\mathrm{dof}}`$.” It yields the following theorem:
##### Spacelike Projection Theorem
Let $`A`$ be the area of a closed surface $`B`$ possessing a future-directed light-sheet $`L`$ with no boundary other than $`B`$. Let the spatial region $`V`$ be contained in the intersection of the causal past of $`L`$ with any spacelike hypersurface containing $`B`$. Let $`N_{\mathrm{dof}}`$ be the total number of independent quantum degrees of freedom present on $`V`$. Then $`N_{\mathrm{dof}}A/4`$.
The theorem can be widely applied and easily understood. Under the stated conditions, none of the degrees of freedom on $`V`$ can escape through holes in $`L`$ or be destroyed on a singularity. Then causality and the second law of thermodynamics require all degrees of freedom on $`V`$ to be present on $`L`$ as well. On $`L`$ their number is bounded by the holographic principle, so it must be bounded also on $`V`$. The spacelike projection theorem will be of use in a number of spacetimes, including de Sitter and AdS (Sec. 3).
### 1.2 Outline
The holographic principle is a relation between space-time geometry and the number of degrees of freedom. It is not equivalent to the statement that there exists a conventional theory without gravity, living on the boundary of a space-time region, with one degree of freedom per Planck area, by which all bulk phenomena including quantum gravity can be described. The holographic principle is clearly necessary for the existence of such a theory, but as we will argue below, it is not sufficient.
The holographic principle does imply, however, that all information contained on $`L`$ can be stored on the surface $`B`$, at a density of no more than one bit per Planck area. (We neglect factors of $`\mathrm{ln}2`$.) We shall work with this interpretation. For a given space-time, we ask whether the information in the interior can be completely projected (in accordance with our formulation of the holographic principle) onto suitable hypersurfaces which will be called screens. We are led to a construction (Sec. 2) under which the space-time is sliced into null hypersurfaces $`L`$. Each light-ray on $`L`$ is followed in the direction of non-negative expansion until the expansion becomes zero. This yields a preferred location for a screen encoding the information on $`L`$. By repeating this procedure for every slice $`L`$, one obtains one or more screen-hypersurfaces. They will either be located on the boundary or will be embedded in the interior of the spacetime. We establish conditions under which projection along spacelike directions is possible.
In Sec. 3 we apply the construction to examples of space-times, including Anti-de Sitter space, Minkowski space, de Sitter space, cosmological solutions, and black holes. We find that AdS has highly special properties under our construction, as it admits spacelike projection onto timelike screens of constant area. For all examples we find that the information in the entire spacetime can be projected onto preferred screens. For spaces with black holes, inequivalent slicings lead to different screen structures. We relate this to the question of information loss.
We discuss the structure of holographic screens in Sec. 4, and draw some conclusions. In a number of examples, the screen-hypersurfaces are spacelike. In other examples they are null or timelike, with the spatial area depending on time. One would not expect such screens to admit a conventional Lorentzian quantum field theory with one degree of freedom per Planck area, because the number of degrees of freedom would have to be time-dependent.
This suggests that a distinction should be made between a dual theory, and the holographic theory (Sec. 4.2). In both types, the holographic principle would be manifest. A dual theory would be characteristic of a certain class of space-times. It would be a conventional theory without gravity, living on the geometric background defined by a holographic screen of the space-time and containing one degree of freedom per Planck area. It would complement, or be equivalent to, a quantum gravity theory living in the bulk, and could thus be used to describe bulk physics. The conformal field theory on the boundary of AdS is an example of a dual theory.
The existence of a dual theory in a given class of space-times will depend on certain properties of the projection and the screen which we aim to expose. More generally, the structure of screens points to a fundamental theory in which quantum degrees of freedom are a derived concept, and their number can change. The theory must give rise to gravity by permitting the unique reconstruction of space-time geometry from the effective number of degrees of freedom in such a way that the holographic principle is manifestly satisfied (Sec. 4.3). We call this the holographic theory. Clearly it cannot be a conventional quantum field theory living on a pre-defined geometric background. Perhaps an indication of its character can be gained from certain proposals of ’t Hooft .
##### Notation and conventions
We work with $`D`$-dimensional Lorentzian manifolds $`M`$. The terms light-like and null are used interchangeably. Any $`(D1)`$-dimensional submanifold $`HM`$ is called a hypersurface of $`M`$ . If $`D2`$ of its dimensions are everywhere spacelike and the remaining dimension is everywhere timelike (null, spacelike), $`H`$ is called a timelike (null, spacelike) hypersurface. By a surface we always refer to a $`(D2)`$-dimensional spacelike submanifold $`BM`$; by area we mean the proper volume of a surface. By a light-ray we do not mean an actual electromagnetic wave or photon, but simply a null geodesic. We use the terms null congruence, and family of light-rays, to refer to a congruence of null geodesics . The term light-sheet is defined in Sec. 1.1; the terms projection, screen, and screen-hypersurface in Sec. 2.1; dual theory and holographic theory in Sec. 4.2. We set $`\mathrm{}=c=G=k=1`$.
## 2 Holographic projection
### 2.1 Screens
The construction described in the previous section answers the following question: Given a surface $`B`$ of area $`A`$, what is the hypersurface $`L`$ on which $`N_{\mathrm{dof}}`$ is bounded by $`A/4`$? We will now consider space-times globally, and ask a different question: which surfaces store the information contained in the entire space-time? To answer this question, the above prescription should be inverted. Given a null hypersurface $`L`$, one should follow the geodesic generators of $`L`$ in the direction of non-negative expansion. One can stop anytime, but one must stop when the expansion becomes negative. This procedure will be called projection. The $`(D2)`$-dimensional spatial surface $`B`$ spanned by the points where the projection is terminated will be called a screen of the projection. If the expansion vanishes on every point of $`B`$, it will be called a preferred screen.
Preferred screens are of particular interest for a simple reason. The expansion of the projection typically changes sign on a preferred screen $`B`$. Therefore $`B`$ will be a preferred screen for projections coming from two directions, e.g., the past-directed outgoing and future-directed ingoing directions. It will thus be particularly efficient in encoding global information. (Actually, there may be a deeper reason why preferred screens play a special role. We suspect that they are precisely the surfaces for which the holographic bound, $`N_{\mathrm{dof}}A/4`$, is saturated. This is suggested by considerations in Sec. 4.2 of Ref. . There it was found that the covariant entropy bound can be saturated on the future light-sheet of an apparent horizon, but not on the future light-sheets of smaller spheres inside the apparent horizon. Under the projection that generates those light-sheets, the apparent horizon is a preferred screen. This argument should be viewed with caution, however, because there might be independent, practical reasons why the thermodynamic entropy cannot be made as large as $`N_{\mathrm{dof}}`$ for the smaller spheres.)
By following all generators of the null hypersurface $`L`$ in a non-contracting direction to a screen, we obtain a projection of all information on the hypersurface onto one or more screens, which may be embedded in the hypersurface, or may lie on its boundary. The number of screens can be minimized by using preferred screens whenever possible.
### 2.2 Screen-hypersurfaces
In order to project the information in a space-time $`M`$, our strategy will be to slice $`M`$ into a one-parameter family of null hypersurfaces, $`\{L\}`$. This will be possible in all examples we consider. Usually the slicing is highly non-unique, but the symmetries of most space-times of interest reduce the number of inequivalent slicings considerably. To each slice $`L`$, we apply the projection rule. This procedure yields a number of one-parameter families of $`(D2)`$-dimensional screens. Each family forms a $`(D1)`$-dimensional screen-hypersurface embedded in $`M`$ or located on the boundary of $`M`$. (This sounds a lot more complicated than it is—see the “recipe” in Sec. 2.3 and the figures in Sec. 3 below.) The screen-hypersurfaces can be time-like, null, or space-like; in Sec. 3 examples of each type will be found. In general, the causal character can change from time-like to space-like within the screen-hypersurface.
Usually it will be clear whether we are talking about a screen (a spatial surface), or a $`(D1)`$-dimensional hypersurface formed by a one-parameter family of screens. Therefore we will often refer to a screen-hypersurface loosely as a “screen” of $`M`$. If the hypersurface consists of preferred screens, we call it a preferred screen-hypersurface, or loosely a preferred screen of $`M`$. If the expansions of both independent pairs of orthogonal families of light-rays vanish on a screen, it will be preferred under all four projections that end on it. We will call such a screen, and hypersurfaces formed by such screens, optimal.
So far we have discussed only null projection, i.e., projection of information along null hypersurfaces. It is sometimes possible to project information along spacelike hypersurfaces. Namely, spacelike projection of the information in a spatial region $`V`$ onto a screen $`B`$ is allowed if $`V`$ and $`B`$ satisfy the conditions set forth in the “spacelike projection theorem” (Sec. 1.1). This will be significant in a number of space-times, in particular in de Sitter and Anti-de Sitter space.
### 2.3 The recipe
To construct screens, one must slice a space-time into null hypersurfaces. In view of the spherical symmetry of all metrics considered below, it will be natural to slice them into a family $`\{L\}`$ of light-cones centered at $`r=0`$.<sup>2</sup><sup>2</sup>2In Minkowski-space, we will also consider a family of light-rays orthogonal to a flat $`(D2)`$-plane. The family can be parametrized by time. This will leave two inequivalent null projections, namely along past or future-directed light-cones. Often the light-cones will be truncated by boundaries of the space-time and will not include $`r=0`$, but this does not matter. In the case of spherical symmetry, one thus obtains the following recipe for the construction of screens:
1. Draw a Penrose diagram. Every point represents a $`(D2)`$-sphere. Each diagonal line represents a light-cone. The two inequivalent null slicings can be represented by the ascending and descending families of diagonal lines.
2. Pick one of the two families. Now the question is in which direction to project along the diagonal lines.
3. Identify the apparent horizons, i.e., hypersurfaces on which the expansion of the past or future light-cones vanishes. They will divide the space-time into normal, trapped, and anti-trapped regions. In each region, draw a wedge whose legs point in the direction of negative expansion of the cones.
4. On a given diagonal line (i.e., light-cone), project each point towards the tip of the local wedge, onto the nearest point (i.e., sphere) $`B_i`$ where the direction of the tip flips, or onto the boundary of space-time as the case may be.
5. Repeat for every line in the family. The surfaces $`B_i`$ will form (preferred) screen-hypersurfaces $`H_i`$.
Below we will strive to make these steps explicit by including two or three Penrose diagrams for most examples. In the first diagram, the apparent horizons will be identified and the wedges placed. For each inequivalent family of light-cones, we will then provide a diagram in which the projection directions are indicated by thick arrows. We invite the reader to verify that these directions are uniquely determined by the wedges. Screens will be denoted by thick points, preferred screen-hypersurfaces by thick lines.
## 3 Examples
### 3.1 Anti-de Sitter space
Type IIB string theory on the background $`\text{AdS}_5\times 𝐒^5`$, with $`N`$ units of flux on the $`𝐒^5`$, appears to be dual to $`(3+1)`$-dimensional $`U(N)`$ supersymmetric Yang-Mills theory with 16 real supercharges . One can consider this theory to live on the boundary of the AdS space. The correspondence between bulk and boundary relates infrared effects in the bulk to ultraviolet effects on the boundary . This feature was exploited by Susskind and Witten to show that the boundary theory has only one degree of freedom per Planck area, as required by the holographic principle in the traditional, “spacelike” form in which it has often been expressed.
We wish to understand some of these properties from the perspective of the general formulation of the holographic principle given in Sec. 1.1. From this point of view, the bulk information is projected along null directions in general, and along spacelike directions only if certain conditions are met. We will verify that these conditions are indeed satisfied in AdS. Moreover, we will find that the boundary at spatial infinity is a preferred (and optimal) screen under our construction. Finally, we will note that AdS admits screen-hypersurfaces of constant spatial area that encode their space-time interior. The concurrence of these properties is special to AdS (and to some unstable solutions identified in Sec. 3.5), and may be a necessary condition for the existence of the kind of duality that has been found in this space-time.
Anti-de Sitter space can be scaled into the direct product of an infinite time axis with a unit spatial ball . In this form it has the metric
$$ds^2=R^2\left[\frac{1+r^2}{1r^2}dt^2+\frac{4}{(1r^2)^2}\left(dr^2+r^2d\mathrm{\Omega }^2\right)\right].$$
(3.1)
The constant scale factor $`R`$ is the radius of curvature. The spacelike hypersurfaces are open balls given by $`t=\text{const},0r<1`$. The boundary of space is a two-sphere residing at $`r=1`$. The proper area of spheres diverges as $`r1`$.
Consider the past directed radial light-rays emanating from a caustic ($`\theta =+\mathrm{}`$) at $`r=0,t=t_0`$ (Fig. 1).
They form a past light-cone $`L`$ with a spherical boundary. The cone grows with affine time until the light-rays reach the boundary of space at $`r=1`$. It is straightforward to check that the expansion, $`\theta `$, is inversely proportional to the affine time. It thus decreases monotonically, but remains positive; one finds that
$$\theta 0\text{as}r1.$$
(3.2)
Consider a sphere $`B`$, of area $`A_B`$, on the lightcone $`L`$. The part of $`L`$ in the interior of $`B`$, $`L_B`$, has negative expansion in the direction away from $`B`$, and therefore constitutes a light-sheet of $`B`$. By the holographic principle, the number of degrees of freedom on $`L_B`$ does not exceed a quarter of the area of $`B`$:
$$N_{\mathrm{dof}}(L_B)\frac{A_B}{4}.$$
(3.3)
Because the cone closes off at $`r=0`$, it has no boundary other than $`B`$. The spatial interior of $`B`$ on any spacelike hypersurface through $`B`$, $`V_B`$, lies entirely in the causal past of $`L_B`$. Therefore the conditions for the spacelike projection theorem are met. It follows that area bounds $`N_{\mathrm{dof}}`$ on spatial regions of AdS:
$$N_{\mathrm{dof}}(V_B)\frac{A_B}{4}.$$
(3.4)
Since the light-cone expansion is positive for all values of $`r`$, these conclusions remain valid in the limit as the sphere $`B`$ moves to the boundary of space, $`BB_{\mathrm{}}`$. By Eq. (3.2), $`B_{\mathrm{}}`$ is a preferred screen. As expected, the preferred screen is precisely the one which encodes the entire space. By time reversal invariance of Eq. (3.1), the expansion of future-directed radial lightrays arriving at $`B_{\mathrm{}}`$ will also vanish; thus $`B_{\mathrm{}}`$ is an optimal screen.
So far we have considered screens bounding $`N_{\mathrm{dof}}`$ on a particular light-cone or spatial hypersurface. A screen-hypersurface encoding the entire space-time is obtained by repeating the construction for every single light-cone in the slicing of AdS. By the time-translation invariance of Eq. (3.1), this repetition is trivial. The family of finite screens $`B(t)`$ of constant area $`A_B`$ thus forms a timelike screen-hypersurface $`H`$ of topology $`𝐑\times 𝐒^{D2}`$. By the holographic principle, $`N_{\mathrm{dof}}`$ in the enclosed space-time region does not exceed $`A_B/4`$. From the spacelike projection theorem it follows that it does not matter whether one counts degrees of freedom on null or on spacelike hypersurfaces intersecting $`H`$. After taking the limit $`B(t)B_{\mathrm{}}(t)`$, one finds that the timelike boundary at $`r=1`$, $`H_{\mathrm{}}`$, is an optimal screen of Anti-de Sitter space. It encodes the entire information in the bulk, by spacelike or null projection.
Let us briefly discuss what happens when a black hole forms. This is shown in the upper part of Fig. 1. Let us assume that the constant screen area, $`A_B`$, is so large that the black hole never engulfs the screen-hypersurface $`H`$ formed by the screens $`B(t)`$. Then the past-directed ingoing light-sheet of any $`B(t)`$ lies outside the event horizon, and has no boundary other than $`B(t)`$. By arguments similar to those leading to the spacelike projection theorem , this implies that $`N_{\mathrm{dof}}`$ in the region between the black hole and $`H`$ never exceeds $`A_B/4`$. Generic space-like hypersurfaces passing through the interior of the black hole, however, are not contained in the causal past of any complete future-directed light-sheet of $`B`$ (see dotted line in Fig. 1). The spacelike projection theorem does not apply to those regions, and therefore the spacelike projection of the interior of the event horizon onto $`H`$ is not possible. Of course, the entire black hole interior can be encoded on $`H`$ by null projection along past light-cones. (Alternatively, it can be projected onto the apparent horizon; we discuss this in more detail in Sec. 3.2 for the case of Schwarzschild black holes.) This discussion remains valid in the limit as $`BB_{\mathrm{}}`$, and thus applies to the boundary of Schwarzschild-AdS at spatial infinity.
### 3.2 Minkowski space
We now turn to space-times which are asymptotically flat. The discussion of finite bound systems in Minkowski space does not differ much from the treatment in AdS. The space-time region occupied by them can be projected onto a screen-hypersurface of topology $`𝐑\times 𝐒^{D2}`$, formed by a spherical screen circumscribing the system. As long as no black holes form, the projection can be spacelike. A spherical screen of finite size is not preferred unless the interior is on the verge of gravitational collapse (see Sec. 3.5).
A bound system can also be projected along past-directed light-rays onto a remote flat plane. All families of null-geodesics orthogonal to the plane have zero expansion; therefore the screen is optimal. It can encode bulk information on both sides. This projection was originally proposed by Susskind and was further investigated in Ref. . If the system does not contain black holes, it can be projected onto the plane along future-directed light-rays as well.
For the discussion of scattering processes (Fig. 2) we shall follow the recipe given in Sec. 2.3.
By following past light-cones centered at $`r=0`$, all of Minkowski space is projected onto past null infinity, $`I^{}`$, where $`\theta 0`$. Similarly, by following future light-cones one can project the bulk onto future null infinity, $`I^+`$. Both infinities are preferred screens of Minkowski space. Each screen alone suffices to store all information in the interior of the space-time; one can interpret this as a statement of the unitarity of the S-matrix in the absence of black holes.
Let us now assume that a black hole forms during scattering (Fig. 3a).
The past light-cones still project all points in the spacetime onto $`I^{}`$, including the interior of the black hole (Fig. 3b). The screen $`I^+`$, however, encodes only the exterior of the black hole, via future-directed outgoing light-rays (Fig. 3c). This discrepancy can be interpreted as information loss in classical black holes. The black hole interior can be encoded onto the apparent horizon, which forms a preferred screen, by future-directed outgoing and past-directed ingoing lightrays.
The picture becomes more interesting when the quantum radiation of black holes, as well as its back-reaction, is included. This restores the possibility of unitarity. After the black hole has formed from classical matter, the apparent horizon shrinks due to the quantum pair creation of particles . In this process a positive energy particle escapes to infinity, while its negative energy partner crosses into the black hole. Unlike positive energy matter, this particle anti-focusses light: it violates the null convergence condition and causes the expansion of light-rays to increase. Outgoing light-rays immediately inside the horizon can thus change from negative to positive expansion without going through a caustic (Fig. 4).
(If the null convergence condition holds, the expansion becomes positive only at “caustics,” or focal points, of the light-rays. Caustics thus are the generic endpoints of light-sheets .) This leads to a situation which would not be possible in a classical space-time. There exists a hypersurface $`H`$, namely the black hole apparent horizon during evaporation, from which one has to project away in both directions. Thus, past-directed ingoing light-rays map $`H`$ onto a different part of the apparent horizon ($`h`$), and future-directed outgoing light-rays map it onto a part of $`I^+`$.
#### A digression on unitarity
Let us examine the evaporation process in more detail. When a negative mass particle enters the horizon, the expansion of the generators of the horizon changes from zero to a positive value. There will be a nearby null congruence, inside the black hole, whose expansion is changed from a negative value to zero by the same process. This congruence will now generate the apparent horizon. Since it has smaller cross-sectional area, the horizon has shrunk. The movement of the apparent horizon will leave behind a trace in the Hawking radiation, causing a deviation from a thermal spectrum over and above the deviation caused by greybody factors. This is similar to the distortion in the thermal spectrum of radiation enclosed in a cavity, while the wall of the cavity is being moved.
The amount by which the apparent horizon decreases during a given pair-creation event depends on the profile, $`\theta (𝒜)`$, of the expansion of the outgoing future-directed null geodesics near the horizon. The cross-sectional area $`𝒜`$ of null congruences becomes smaller, and the expansion $`\theta `$ more negative, the further inside the black hole they are located. If the black hole was formed by a system of low entropy, for example by the collapse of a homogeneous dust ball of zero temperature, the profile will be a featureless monotonic function, and the horizon will decrease very smoothly during evaporation. The back-reaction, in this case, will not imprint a significant signature onto the thermal spectrum. However, if the black hole was formed by a highly enthropic system, the profile will be more complex.
Consider a shell of matter falling into a black hole. For now, assume that the shell contains only radial modes, and is thus exactly spherically symmetric even microscopically. If a lot of entropy is stored in the shell, its density will be a complicated function of the radius. By Raychauduri’s equation , the density profile of the infalling shell will be imprinted on the expansion profile of the outgoing future-directed null geodesics that eventually pass through the apparent horizon during evaporation. Correspondingly, the same type of pair creation process will sometimes cause the horizon area to decrease by a larger step, sometimes by a smaller amount, depending on the expansion profile of the null geodesics passing through $`H`$ at the pair creation event. The back-reaction will be irregular, and the corresponding deviations of the Hawking radiation from the thermal spectrum will be complex. There is thus a signature in the radiation which encodes the irregularity of the back-reaction, which in turn encodes the complexity of the matter system that formed the black hole.
It is easy to extend this discussion to systems containing also angular modes. They will deflect outgoing lightrays into angular directions. The expansion will now be a local function of the cross-sectional area, $`\theta (\delta 𝒜;\vartheta ,\phi )`$. The back-reaction will not be spherically symmetric, and the apparent horizon will develop dents and bulges. This leaves a non-spherical signature in the Hawking radiation.
In this way information about the material falling into the black hole may be transferred onto the outgoing Hawking radiation. The information will be encoded in a subtle way and it will typically be necessary to measure the entire radiation emitted by the black hole before the ingoing state can be reconstructed. Of course, we have sketched only a qualitative picture, and we have taken the pair creation model of black hole evaporation rather literally. Moreover, no mechanism can copy ingoing information onto outgoing radiation unless one implicitly assumes that the fundamental theory evades the “quantum Xeroxing” no-go theorem , for example by non-locality .<sup>3</sup><sup>3</sup>3We thank Lenny Susskind for pointing this out, and for a number of related discussions. We have aimed to outline a specific mechanism by which information may be transferred in the semi-classical picture. In general terms, our discussion is strongly related to the approach of ’t Hooft .
### 3.3 de Sitter space
de Sitter space is the maximally symmetric solution of the vacuum Einstein equation with a positive cosmological constant $`\mathrm{\Lambda }`$. It may be visualized as a $`(D1,1)`$-hyperboloid embedded in $`(D+1)`$-dimensional Minkowski space. A metric covering the entire space-time is given by
$$ds^2=dt^2+H^2\mathrm{cosh}^2Htd\mathrm{\Omega }_{D1}^2,$$
(3.5)
where
$$H=\sqrt{\frac{\mathrm{\Lambda }}{3}}$$
(3.6)
is the Hubble parameter, or inverse curvature radius. In this metric, the spacelike hypersurfaces are spheres, $`𝐒^{D1}`$. They contract, and then expand, at an exponential rate. de Sitter space also admits metrics with maximally symmetric spatial sections of zero or negative curvature, as well as a static metric,
$$ds^2=(1H^2r^2)d\tau ^2+\frac{dr^2}{1H^2r^2}+r^2d\mathrm{\Omega }_{D2}^2.$$
(3.7)
Those metrics cover only certain portions of the spacetime.
The metric on the spatial $`(D1)`$-sphere is given by:
$$d\mathrm{\Omega }_{D1}^2=d\chi ^2+\mathrm{sin}^2\chi d\mathrm{\Omega }_{D2}^2,$$
(3.8)
whence
$$r=H^1\mathrm{cosh}Ht\mathrm{sin}\chi .$$
(3.9)
A geodesic observer is immersed in a bath of thermal radiation of temperature $`T=H/(2\pi )`$, which appears to come from the cosmological horizon surrounding the observer. The causal structure of de Sitter space is shown in Fig. 5.
The only boundaries are past and future infinity, $`I^{}`$ and $`I^+`$; they are both spacelike. The space-time is divided in half by the event horizon, $`E`$, of a geodesic observer, who can be taken to live at $`\chi =r=0`$.
Consider the past light-cones centered at $`\chi =0`$, i.e., at one of the two poles of the $`𝐒^{D1}`$. Just as for AdS and Minkowski, the expansion starts with the value $`+\mathrm{}`$ and decreases. Any surface on the light-cone bounds $`N_{\mathrm{dof}}`$ on the part of the cone it encloses. By the spacelike projection theorem, it also bounds $`N_{\mathrm{dof}}`$ on any spatial hypersurface in its interior. Perhaps surprisingly, this holds even for surfaces which are both near the event horizon and near the past singularity (see Fig. 5a, dashed line). Their area will be $`H^2`$, but they enclose an exponentially large spatial region.
When the light-cone reaches $`I^{}`$, the expansion approaches zero. Thus the boundary surface on past infinity is a preferred screen. (Actually it is optimal because a past light-cone arriving from the other pole, $`\chi =\pi `$, will also have $`\theta 0`$ near $`I^{}`$.) Repeating this projection for all times, one finds that half of de Sitter space, namely the region within the event horizon of an observer at $`\chi =0`$, can be projected onto past infinity. The projection of only half of the space-time is peculiar to de Sitter space. By contrast, past light-cones project all of Minkowski, or all of AdS, onto their respective infinities (Secs. 3.1 and 3.2). By using past light-cones emanating from the other pole of the spatial $`𝐒^{D1}`$, an additional portion of de Sitter can be projected onto past infinity. But this still leaves out the antitrapped region beyond the cosmological horizons. It can only be projected by future-directed lightrays onto future infinity. A global null projection of de Sitter space can thus be achieved by using two optimal screen-hypersurfaces: past and future infinity. This is shown in Fig. 5b. Indeed, the potential holographic role of these boundaries has been speculated upon for some time . Both global screens are spacelike hypersurfaces. Because surfaces near future infinity are anti-trapped, one cannot encode global de Sitter space on a finite number of timelike or null screens.
However, we can apply the spacelike projection theorem (Sec. 1.1) to the screens that form the event horizon $`E`$ of an observer at $`\chi =0`$. They are all of constant area, $`4\pi H^2`$. (Since $`E`$ never reaches $`\chi =r=0`$, this argument strictly requires a limiting procedure starting from spheres in the vicinity of $`E`$; see Fig. 5a.) Because $`E`$ is generated by light-rays of zero expansion, all screens on it are manifestly preferred. Thus, the null hypersurface $`E`$ is a preferred screen of constant area. Because of its degeneracy, $`E`$ encodes only itself under null projection along $`E`$. Under spacelike projection, however, it encodes half of the space-time (Fig. 5c), namely the region within the event horizon. One can reasonably argue that the region beyond the event horizon has no meaning because it cannot be observed, and that de Sitter space should not be treated globally (Fig. 5c); thus the screen $`E`$ should suffice for a holographic description of de Sitter space.
In inflationary models, the de Sitter phase is followed by a matter or radiation dominated phase, and the entire space-time during this era can be projected onto the screens available in the relevant FRW models (see Sec. 3.4). — Black holes in de Sitter space can be treated much like black holes in AdS or Minkowski space.
### 3.4 FRW cosmologies
Friedmann-Robertson-Walker (FRW) cosmologies are described by a metric of the form
$$ds^2=a^2(\eta )\left[d\eta ^2+d\chi ^2+f^2(\chi )d\mathrm{\Omega }^2\right].$$
(3.10)
Here $`f(\chi )=\mathrm{sinh}\chi `$, $`\chi `$, $`\mathrm{sin}\chi `$ corresponds to open, flat, and closed universes respectively. FRW universes contain homogeneous, isotropic spacelike slices of constant (negative, zero, or positive) curvature. We will not discuss open universes, since they display no significant features beyond those arising in the treatment of closed or flat universes.
The matter content will be described by $`T_{ab}=\text{diag}(\rho ,p,p,p)`$, with pressure $`p=\gamma \rho `$. We assume that $`\rho 0`$ and $`1/3<\gamma 1`$. The case $`\gamma =1`$ corresponds to de Sitter space, which was discussed in Sec. 3.3. The apparent horizon is defined geometrically as the spheres on which at least one pair of orthogonal null congruences have zero expansion. It is given by
$$\eta =q\chi ,$$
(3.11)
where
$$q=\frac{2}{1+3\gamma }.$$
(3.12)
The solution for a flat universe is given by
$$a(\eta )=\left(\frac{\eta }{q}\right)^q.$$
(3.13)
Its causal structure is shown in Fig. 6.
The interior of the apparent horizon, $`\eta q\chi `$, can be projected along past light-cones centered at $`\chi =0`$, or by space-like projection, onto the apparent horizon. The exterior, $`\eta q\chi `$, can be projected by the same light-cones, but in the opposite direction, onto the apparent horizon. The apparent horizon is thus a preferred screen encoding the entire space-time. Alternatively, one can use future light-cones to project the entire universe onto future null infinity, another preferred screen.
By Eq. (3.11), the apparent horizon screen is a timelike hypersurface for $`1/3<\gamma <1/3`$, null for $`\gamma =1/3`$, and spacelike for $`1/3<\gamma 1`$. In a universe dominated by different types of matter in different eras, the causal character of the apparent horizon hypersurface can change from timelike to spacelike or vice versa (see, e.g., Fig. 5 in Ref. ).
For a closed universe, the solution is given by
$$a(\eta )=a_{\mathrm{max}}\left(\mathrm{sin}\frac{\eta }{q}\right)^q.$$
(3.14)
In addition to Eq. (3.11), a second apparent horizon emanates from the opposite pole of the spatial $`𝐒^{D1}`$, at $`\chi =\pi `$; it is described by
$$\eta =q(\pi \chi ).$$
(3.15)
The two hypersurfaces formed by the apparent horizons divide the space-time into four regions, as shown in Fig. 7.
Let us choose the first apparent horizon, Eq. (3.11), as a (preferred) screen-hypersurface. On one side, $`\eta q\chi `$, lies a normal region and a trapped region. These regions can be projected onto the screen by past-directed radial light-rays moving away from the South pole ($`\chi =0`$). The other half of the universe, $`\eta q\chi `$, can be projected onto the same screen by future directed radial light-rays moving away from the North pole ($`\chi =\pi `$). Therefore the preferred screen given by Eq. (3.11) encodes the entire closed universe.
A number of cosmological entropy bounds have been proposed which are based on the idea of defining a horizon-size spatial region to which Bekenstein’s bound can be directly applied. We have emphasized the importance of these bounds in Ref. , where we also discuss their relation to the covariant entropy bound (Sec. 1.1). Those bounds can be given a holographic interpretation by considering them as limits on $`N_{\mathrm{dof}}`$ in the specified ken. Because they refer to limited regions, however, it is not clear how global screen-hypersurfaces could be constructed.
### 3.5 Einstein static universe
The Einstein static universe (ESU) is a closed FRW space-time containing ordinary matter as well as a positive cosmological constant of a certain critical value . Its metric can be written as a direct product of an infinite time axis with a $`(D1)`$-sphere of constant radius $`a`$:
$$ds^2=dt^2+a^2d\mathrm{\Omega }_{D1}^2.$$
(3.16)
The causal structure is shown in Fig. 8.
Each hemisphere can be projected along past- or future-directed light-rays, or by spacelike projection, onto the equator. This screen is optimal, because all four families of orthogonal light-rays have vanishing expansion. Moreover, the screen forms a timelike hypersurface, with spatial slices of constant finite size.
This is reminiscent of the properties of the screen at the boundary of Anti-de Sitter space: the screen is optimal, timelike, of constant size, and encodes the entire space-time by space-like projection. The difference is that the AdS screen has infinite proper area, while the equator in the ESU is a $`(D2)`$-sphere of finite area $`a^{D2}`$. It lies not on a boundary of space (there is none in the ESU), but is embedded in the interior.
The properties of the projection might give rise to the hope that a boundary theory, dual to the bulk description, could be formulated on the equator of the ESU. The example may be of limited use, however, because the ESU is not a stable solution . Another unstable solution with similar properties is given by a static spherical system just on the verge of gravitational collapse. Its radius will be equal to its gravitational radius, and the expansion of both past- and future-directed outgoing light-rays goes to zero at the surface of the system.
## 4 Holographic Theory
### 4.1 Summary
From a universal entropy bound found in Ref. , we obtained a background-independent formulation of the holographic principle . This led us to a construction of hypersurfaces (screens) on which all information contained in a space-time can be stored. The screens are embedded, or lie on the boundary of the space-time, and contain no more than one bit of information per Planck area. In this sense, the world is a hologram.
The construction was applied to a number of examples. For Anti-de Sitter space it yields the timelike boundary at spatial infinity as a preferred screen. In Minkowski space, past or future null infinity, or a flat plane, can encode all information. de Sitter space is mapped along light-rays onto the spacelike infinities in the past and future; alternatively, all information in the observable half of the de Sitter space can be stored on the event horizon (a null hypersurface of finite area) via spacelike projection. Cosmological spacetimes may not have a boundary, but embedded screens can be found; they may be spacelike, timelike, or null, depending on the matter content. The information in a black hole can be mapped onto the apparent horizon, or onto past null infinity.
From the examples one can draw the following observations:
* Holographic screens can be spacelike hypersurfaces.
* If they are timelike or null, the spatial area is not necessarily constant in the induced, or any other, time-slicing.
Before explaining why these features may be significant, let us briefly discuss a tempting but misguided conclusion. One might argue that holographic projection onto spacelike screens is a trivial accomplishment, because in any conventional theory one can specify initial conditions on a Cauchy surface and predict, or retrodict, the past and future development of the system. That is true, but it is a different kind of information storage. In that case, one stores not only the information at one moment of time, but also a machine (namely the theory) which is capable of recovering the state of the system at all other times. A holographic construction, on the other hand, feigns ignorance of any theory describing the matter evolution, and simply encodes all information, at all times, onto screens of dimension $`D2`$. The space-time is sliced into null hypersurfaces; slice by slice the information is encoded onto $`(D2)`$-dimensional spatial surfaces at a maximum density of one bit per Planck area. These surfaces form a $`(D1)`$ dimensional screen hypersurface which may be timelike, spacelike, or null, but from the point of view of holographic information storage its causal character is irrelevant.
### 4.2 Theories on the screen
Our interpretation, so far, has centered on the information needed to describe a state. This is measured by the number of degrees of freedom. Therefore we can use the holographic principle (which refers to $`N_{\mathrm{dof}}`$) to project all information in the space-time onto screen-hypersurfaces. In this sense, the holographic principle implies a drastic reduction of the complexity of nature compared to naive expectations of, perhaps, one degree of freedom per Planck volume.
We did not, however, use the holographic principle to describe nature. The holographic principle is far from manifest in the description of the world in terms of general relativity and quantum field theory; yet these theories are very successful. Working within their frame, one finds a number of non-trivial effects which appear to insure that the entropy bound implied by the holographic principle is always satisfied ; but these results could not have been immediately inferred from the basic axioms of GR and QFT.
As a kind of external restriction imposed on physical theories, holography is interesting but unsatisfactory. If the number of degrees of freedom is limited by the holographic principle, there ought to be a description of nature in which this restriction is manifest. Let us call this hypothetical description the holographic theory. One would expect the holographic theory to remain valid when semi-classical gravity breaks down ; in this regime it may be the only possible description. These are good reasons to search for a holographic theory.
The simplest idea would be to define a theory on the geometric background given by the screen-hypersurface(s). If the theory contained one degree of freedom per Planck area, and was related by a kind of dictionary (“duality”) to the space-time (“bulk”) physics, the holographic principle would be manifest. Let us call this type of theory a dual theory.
This idea works for certain asymptotically Anti-de Sitter space-times. The screen encoding the entire bulk information is the timelike hypersurface formed by the boundary of space (Sec. 3.1). According to Maldacena’s remarkable conjecture , a super-Yang-Mills theory living on this hypersurface describes the bulk physics completely. By considering a finite boundary and taking the limit as it moves to spatial infinity, one can show that the theory contains no more than one degree of freedom per Planck area . Therefore it is a dual theory in the sense of our definition.
Perhaps it will be possible to find dual theories for some other classes of space-times; certainly this would be be an important contribution to the understanding of holography and of quantum gravity. In general, however, the “dual theory” approach will not work. The theories we usually think of have a fixed number of degrees of freedom built into them; these degrees of freedom evolve in Lorentzian time. But consider the cosmological solutions studied in Sec. 3.4. The area of the screens is time-dependent. The screen theory would have to be capable of “creating” or “activating” degrees of freedom. Moreover, the area can decrease, as seen in the closed universe example. In the screen theory this would correspond to the destruction, or de-activation, of degrees of freedom. Eventually their number would approach zero, and the second law of thermodynamics would be violated in the screen theory.<sup>4</sup><sup>4</sup>4We are grateful to Andrei Linde for stressing this point to us. (Note that this does not, of course, imply a violation of the second law in the bulk. Rather, it is related to the creation, or destruction, of degrees of freedom at the initial and final singularities of the universe.)
This suggests that one should not in general think of the screen theory as a conventional theory with a fixed number of degrees of freedom. Rather, one might expect it to be a theory with a varying number of “active” degrees of freedom. Thus, its properties would be very different from those of ordinary physical theories. Moreover, since the screen hypersurfaces can be spacelike or null, one should not expect the theory to live in Lorentzian time.
### 4.3 Geometry from entropy
We would like to advocate a more radical approach. One should not be thinking about a “screen theory” (a theory defined on some hypersurface of space-time) at all. The screen theory approach cannot be fundamental, because it presumes the existence of a space-time background, or at least of an asymptotic structure of space and time. In order to use the holographic principle for a full description of nature, we suggest it should be turned around. Loosely speaking, one should not constrain entropy by geometry, but construct geometry from entropy. (Strictly, “number of degrees of freedom” should replace “entropy” here.) The construction must be such that the holographic principle, in the form given in Sec. 1.1, is automatically satisfied.
It thus appears that two problems must be overcome if a holographic theory is to be found. First, one must formulate a theory with a varying number of degrees of freedom. A possibility may be that the theory can activate or de-activate degrees of freedom from an infinite reservoir.<sup>5</sup><sup>5</sup>5We thank Lenny Susskind for this suggestion and related discussions. An extreme but perhaps more satisfying resolution would be to treat quantum degrees of freedom not as fundamental ingredients, but as a derived concept. ’t Hooft has long been advocating that models should be sought in which quantum degrees of freedom arise as a complex, effective structures (see Ref. and references therein). It would be natural for $`N_{\mathrm{dof}}`$ to vary in such models.
The second challenge is to find a prescription that allows the unique reconstruction of space-time geometry from the varying number of degrees of freedom (see, e.g., Refs. and references therein for a discussion of related questions). Part of this prescription will be to equate $`N_{\mathrm{dof}}`$ with the proper area of an embedded $`(D2)`$-dimensional preferred or optimal screen. A more difficult question is how the intrinsic geometry of screen-hypersurfaces can be recovered. It may be undesirable to identify the discrete steps of, say, a cellular automaton with Lorentzian time. But the number and character of the degrees of freedom provide information about the matter content. Therefore a complete reconstruction of space-time geometry is not inconceivable.
## Acknowledgments
I am indebted to Gerard ’t Hooft, Andrei Linde, and Lenny Susskind for very helpful discussions, as well as important criticism and suggestions. I thank Lenny Susskind for valuable comments on a draft of this paper.
|
no-problem/9906/astro-ph9906370.html
|
ar5iv
|
text
|
# 1 INTRODUCTION
## 1 INTRODUCTION
Since the BeppoSAX detection of GRB 970228, X-ray afterglows have been observed from about 15 gamma-ray bursts (GRBs), of which 10 were detected optically and 5 even also in radio wavelengths (Costa et al. 1997; Kulkarni et al. 1998; Bloom et al. 1998; Piran 1998; and references therein). The cosmological origin of at least some GRBs is thus firmly established. The so called fireball model (Goodman 1986; Paczyński 1986; Rees & Mészáros 1992, 1994; Mészáros & Rees 1992; Katz 1994; Sari, Narayan & Piran 1996) is strongly favoured, which is found successful at explaining the major features of the low energy light curves (Mészáros & Rees 1997; Vietri 1997; Tavani 1997; Waxman 1997; Wijers, Rees & Mészáros 1997; Sari 1997; Huang et al. 1998; Dai & Lu 1998a; Dai, Huang & Lu 1998). A variant of this model, where central engines (e.g., strongly magnetized millisecond pulsars) supply energy to postburst fireballs through magnetic dipole radiation, has been proposed to account for the special features of the optical afterglows from GRB 970228 and GRB 970508 (Dai & Lu 1998b, c).
Since the expansion of a fireball may be either adiabatic or highly radiative, extensive attempts have been made to find a common model applicable for both cases (Blandford & McKee 1976; Chiang & Dermer 1998; Piran 1998). As a result, a conventional model was suggested by various authors (see for example, Chiang & Dermer 1998; Piran 1998). A dynamical model should be correct not only in the initial ultra-relativistic phase, which is well described by those simple scaling laws (Mészáros & Rees 1997; Vietri 1997; Waxman 1997), but also in the consequent non-relativistic phase, which is correctly discussed by using the Sedov solution (Sedov 1969; Wijers, Rees & Mészáros 1997). Although the conventional model is correct for the ultra-relativistic phase, we find it could not match the Sedov solution in the non-relativistic limit. So in this paper, we will construct a dynamical model that is really capable of describing generic fireballs, no matter whether they are radiative or adiabatic, and no matter whether they are ultra-relativistic or non-relativistic.
## 2 CONVENTIONAL DYNAMIC MODEL
A differential equation has been proposed to depict the expansion of GRB remnants (Chiang & Dermer 1998; Piran 1998),
$$\frac{d\gamma }{dm}=\frac{\gamma ^21}{M},$$
(1)
where $`m`$ is the rest mass of the swept-up medium, $`\gamma `$ is the bulk Lorentz factor and $`M`$ is the total mass in the co-moving frame, including internal energy $`U`$. Since thermal energy produced during the collisions is $`dE=(\gamma 1)dmc^2`$, usually we assume: $`dM=(1ϵ)dE/c^2+dm=[(1ϵ)\gamma +ϵ]dm`$, where $`ϵ`$ is defined as the fraction of the shock generated thermal energy (in the co-moving frame) that is radiated (Piran 1998). It is putative that equation (1) is correct in both ultra-relativistic and non-relativistic phase, for both radiative and adiabatic fireballs. However, after careful inspection, we find that during the non-relativistic phase of an adiabatic expansion, equation (1) could not give out a solution consistent with the Sedov results (Sedov 1969).
### 2.1 Radiative case
In the highly radiative case, $`ϵ=1`$, $`dM=dm`$, equation (1) reduces to,
$$\frac{d\gamma }{dm}=\frac{\gamma ^21}{M_{\mathrm{ej}}+m},$$
(2)
where $`M_{\mathrm{ej}}`$ is the mass ejected from the GRB central engine. Then an analytic solution is available (Blandford & McKee 1976; Piran 1998):
$$\frac{(\gamma 1)(\gamma _0+1)}{(\gamma +1)(\gamma _01)}=\left(\frac{m_0+M_{\mathrm{ej}}}{m+M_{\mathrm{ej}}}\right)^2,$$
(3)
where $`\gamma _0`$ and $`m_0`$ are initial values of $`\gamma `$ and $`m`$ respectively. Usually we assume $`\gamma _0\eta /2`$, $`m_0M_{\mathrm{ej}}/\eta `$, where $`\eta E_0/(M_{\mathrm{ej}}c^2)`$ and $`E_0`$ is the total energy in the initial fireball (Waxman 1997; Piran 1998).
During the ultra-relativistic phase, $`\gamma 1`$, $`M_{\mathrm{ej}}m`$, equation (3) gives $`(\gamma +1)mM_{\mathrm{ej}}`$, or equivalently the familiar power-law $`\gamma R^3`$, where $`R`$ is the radius of the blast wave. In the later non-relativistic phase, $`\gamma 1`$, $`mM_{\mathrm{ej}}`$, we have: $`m^2\beta ^2=4M_{\mathrm{ej}}^2`$, or $`\beta R^3`$, where $`\beta =v/c`$ and $`v`$ is the bulk velocity of the material. This is consistent with the late isothermal phase of the expansion of supernova remnants (SNRs) (Spitzer 1968). From these approximations, we believe that equation (2) is really correct for highly radiative fireballs.
### 2.2 Adiabatic case
In the adiabatic case, $`ϵ=0`$, $`dM=\gamma dm`$, equation (1) also has an analytic solution (Chiang & Dermer 1998) :
$$M=[M_{\mathrm{ej}}^2+2\gamma _0M_{\mathrm{ej}}m+m^2]^{1/2},$$
(4)
$$\gamma =\frac{m+\gamma _0M_{\mathrm{ej}}}{M}.$$
(5)
During the ultra-relativistic phase, $`\gamma _0M_{\mathrm{ej}}mM_{\mathrm{ej}}/\gamma _0`$, $`\gamma 1`$, this solution can produce the familiar power-law $`\gamma R^{3/2}`$, which is often quoted for an adiabatic blastwave decelerating in a uniform medium. In the non-relativistic limit ($`\gamma 1`$, $`m\gamma _0M_{\mathrm{ej}}`$), Chiang & Dermer (1998) have derived $`\gamma 1+\gamma _0M_{\mathrm{ej}}/m`$, so that they believe it also agrees with the Sedov solution (Lozinskaya 1992). However we find that their approximation is not accurate enough, because they have omitted some first-order infinitesimals of $`\gamma _0M_{\mathrm{ej}}/m`$. The correct approximation could be obtained only by retaining all the first and second order infinitesimals, which in fact gives: $`\gamma 1+(\gamma _0M_{\mathrm{ej}}/m)^2/2`$, then we have $`\beta R^3`$. This is not consistent with the Sedov solution! We have also evaluated equation (1) numerically, the result is consistent with equations (4) and (5), all pointing to $`\beta R^3`$, not the relation of $`\beta R^{3/2}`$ as declared popularly in the literature (Chiang & Dermer 1998; Piran 1998).
This question is serious. First, it means that equation (1) is not a dependable model for non-radiative fireballs, although it can reproduce the major features in the ultra-relativistic phase. Second, the expansion of a realistic fireball is widely believed to be highly radiative at first, but after only several days, the expansion will become non-radiative (Sari, Piran & Narayan 1998; Dai, Huang & Lu 1998). So in the non-relativistic phase, the fireball is more likely to be adiabatic rather than highly radiative. However, it is just in this condition that the conventional model fails. So any calculation made according to equation (1) will lead to serious deviations in the light curves in the non-relativistic phase.
## 3 OUR GENERIC MODEL
Equation (1) is not consistent with the Sedov solution, we need to revise it. In the fixed frame, since the total kinetic energy of the fireball is $`E_\mathrm{K}=(\gamma 1)(M_{\mathrm{ej}}+m)c^2+(1ϵ)\gamma U`$ (Panaitescu, Mészáros & Rees 1998), and the radiated thermal energy is $`ϵ\gamma (\gamma 1)dmc^2`$ (Blandford & McKee 1976), we have:
$$d[(\gamma 1)(M_{\mathrm{ej}}+m)c^2+(1ϵ)\gamma U]=ϵ\gamma (\gamma 1)dmc^2.$$
(6)
For the item $`U`$, it is usually assumed: $`dU=(\gamma 1)dmc^2`$ (Panaitescu, Mészáros & Rees 1998). Equation (1) has been derived just in this way. However, the jump conditions (Blandford & McKee 1976) at the forward shock imply that $`U=(\gamma 1)mc^2`$, so we suggest that the correct expression for $`dU`$ should be: $`dU=d[(\gamma 1)mc^2]=(\gamma 1)dmc^2+mc^2d\gamma `$. Here we simply use $`U=(\gamma 1)mc^2`$ and substitute it into equation (6), then it is easy to get:
$$\frac{d\gamma }{dm}=\frac{\gamma ^21}{M_{\mathrm{ej}}+ϵm+2(1ϵ)\gamma m}.$$
(7)
We expect this equation should describe a generic fireball correctly.
Indeed, in the highly radiative case ($`ϵ=1`$), equation (7) reduces to equation (2) exactly. While in the adiabatic case ($`ϵ=0`$), equation (7) reduces to :
$$\frac{d\gamma }{dm}=\frac{\gamma ^21}{M_{\mathrm{ej}}+2\gamma m}.$$
(8)
This equation has an analytic solution:
$$(\gamma 1)M_{\mathrm{ej}}c^2+(\gamma ^21)mc^2E_{\mathrm{K0}},$$
(9)
where $`E_{\mathrm{K0}}`$ is the initial value of $`E_\mathrm{K}`$. In the ultra-relativistic phase ($`\gamma _0M_{\mathrm{ej}}mM_{\mathrm{ej}}/\gamma `$), we get the familiar relation of $`\gamma R^{3/2}`$. And in the non-relativistic phase ($`mM_{\mathrm{ej}}`$), we get $`\beta R^{3/2}`$ as required by the Sedov solution.
For any other $`ϵ`$ value between 0 and 1, equation (7) describes the evolution of a partially radiative fireball. Unfortunately, we now could not find an exact analytic solution for equation (7). But in the non-relativistic phase, by assuming $`mM_{\mathrm{ej}}`$, we still can get $`m(\gamma 1)^{(2ϵ)/2}\mathrm{const}`$, that is:
$$\beta R^{3/(2ϵ)}.$$
(10)
## 4 NUMERICAL RESULTS
We have evaluated equation (7) numerically, bearing in mind that (Huang et al. 1998):
$$dm=4\pi R^2nm_\mathrm{p}dR,$$
(11)
$$dR=\beta c\gamma (\gamma +\sqrt{\gamma ^21})dt,$$
(12)
where $`n`$ is the number density of interstellar medium, $`m_\mathrm{p}`$ is the mass of proton, and $`t`$ is the time measured by an observer. We take $`E_0=10^{52}`$ ergs, $`n=1`$ cm<sup>-3</sup>, $`M_{\mathrm{ej}}=2\times 10^5`$ M. Figures (1) – (4) illustrate the evolution of $`\gamma `$, $`v`$, $`R`$, and $`E_\mathrm{K}`$ respectively. In these figures, we have set $`ϵ=0`$ (full lines), 0.5 (dotted lines), and 1 (dashed lines). It is clearly shown that our generic model overcomes the shortcoming of equation (1).
For example, for highly radiative expansion, the dashed lines in these figures approximately satisfy $`\gamma t^{3/7}`$, $`Rt^{1/7}`$, $`\gamma R^3`$, $`E_\mathrm{K}t^{3/7}`$ when $`\gamma 1`$, and $`vt^{3/4}`$, $`Rt^{1/4}`$, $`vR^3`$, $`E_\mathrm{K}t^{3/4}`$ when $`\gamma 1`$. While for adiabatic expansion, the full lines satisfy $`\gamma t^{3/8}`$, $`Rt^{1/4}`$, $`\gamma R^{3/2}`$ when $`\gamma 1`$, and satisfy $`vt^{3/5}`$, $`Rt^{2/5}`$, $`vR^{3/2}`$ when $`\gamma 1`$.
## 5 CONCLUSION AND DISCUSSION
The conventional dynamic model is successful at describing highly radiative GRB remnants, however it has difficulty in reproducing the Sedov solution for adiabatic fireballs. This is completely unnoticed in the literature. We have constructed a new generic model to overcome this shortcoming. Numerical evaluation has proved that our model is highly credible. We hope this work would remind researchers of the importance of the transition from ultra-relativistic to non-relativistic phase, which might occur as early as $`10^6`$$`10^7`$ s since the initial burst (Huang, Dai & Lu 1998).
In the above analysis, for simplicity, we have assumed that $`ϵ`$ is a constant. But in realistic fireballs, $`ϵ`$ is expected to evolve from 1 to 0 due to the changes in the relative importance of synchrotron-radiation-induced and expansion-induced loss of energy (Dai, Huang & Lu 1998). Assuming electrons in the co-moving frame carry a fraction $`\xi _\mathrm{e}=1`$ of the total thermal energy and that the magnetic energy density is a fraction $`\xi _\mathrm{B}^2=0.01`$ of it, we re-evaluate equation (6) numerically. The results are plotted in Figs (1) – (4) with dash-dotted lines. We see from Fig. (4) that the evolution of $`ϵ`$ changes $`E_\mathrm{K}(t)`$ dramatically.
It is worth mentioning that SNRs evolve from non-radiative to radiative stage, but GRB remnants are just on the contrary. This is not surprising because GRB remnants radiate mainly through synchrotron radiation while SNRs lose energy due to excited ions. It is reasonable to deduce that at very late stages, when the cooling due to ions becomes important, GRB remnants may become highly radiative again, just in the same way that SNRs do. The transition may occur when the temperature drops to below $`10^6`$ K and the velocity is just several tens kilometer per second. This needs to be addressed in more detail.
Another interesting problem is the possibility that HI supershells might be highly evolved GRB remnants (Loeb & Perna 1998; Efremov, Elmegreen & Hodge 1998). Our Figs (3) and (4) have shown that typical adiabatic GRB fireballs can evolve to $`R1`$ kpc at $`t10^6`$$`10^7`$ yr, with $`v10`$ km/s, but highly radiative fireballs are obviously not powerful enough. To discuss this possibility in detail, we should pay attention to the possible adiabatic-to-radiative transition mentioned just above.
We are very grateful to R. Wijers for his valuable comments and suggestions. This work was partly supported by the National Natural Science Foundation of China, grants 19773007 and 19825109, and the National Climbing Project on Fundamental Researches.
REFERENCES
Blandford R.D., McKee C.F., 1976, Phys. Fluids, 19, 1130
Bloom J.S. et al., 1998, ApJ, submitted (astro-ph/9808319)
Chiang J., Dermer, C.D., 1998, ApJ, submitted (astro-ph/9803339)
Dai Z.G., Huang Y.F., Lu, T., 1998, ApJ, in press (astro-ph/9806334)
Dai Z.G., Lu T., 1998a, MNRAS, 298, 87
Dai Z.G., Lu T., 1998b, A&A, 333, L87
Dai Z.G., Lu T., 1998c, Phys Rev Lett, 81, 4301
Costa E. et al., 1997, Nat, 387, 783
Efremov Y.N., Elmegreen B.G., Hodge P.W., 1998, ApJ, in press (astro-ph/9805236)
Goodman J., 1986, ApJ, 308, L47
Huang Y.F., Dai Z.G., Lu T., 1998, A&A, 336, L69
Huang Y.F., Dai Z.G., Wei D.M., Lu T., 1998, MNRAS, 298, 459
Katz J., 1994, ApJ, 422, 248
Kulkarni S.R. et al., 1998, Nat, 393, 35
Loeb A., Perna R., 1998, ApJ, in press (astro-ph/9805139)
Lozinskaya T.A., 1992, Supernovae and Stellar Winds in the Interstellar Medium (New York, AIP), Chap. 9
Mészáros P., Rees M.J., 1992, MNRAS, 257, 29P
Mészáros P., Rees M.J., 1997, ApJ, 476, 232
Paczyński B., 1986, ApJ, 308, L43
Panaitescu A., Mészáros P., Rees M.J., 1998, ApJ, in press (astro-ph/9801258)
Piran T., 1998, Phys Report, in press (astro-ph/9810256)
Rees M.J., Mészáros P., 1992, MNRAS, 258, 41P
Rees M.J., Mészáros P., 1994, ApJ, 430, L93
Sari R., 1997, ApJ, 489, L37
Sari R., Narayan R., Piran T., 1996, ApJ, 473, 204
Sari R., Piran T., Narayan R., 1998, ApJ, 497, L17
Sedov L., 1969, Similarity and Dimensional Methods in Mechanics (Academic, New York), Chap. IV
Spitzer L., 1968, Diffuse Matter in Space (Wiley, New York), p200
Tavani M., 1997, ApJ, 483, L87
Vietri M., 1997, ApJ, 488, L105
Waxman E., 1997, ApJ, 485, L5
Wijers R.A.M.J., Rees M.J., Mészáros P., 1997, MNRAS, 288, L51
Figure Captions
Figure 1. Evolution of the bulk Lorentz factor $`\gamma `$. We take $`E_0=10^{52}`$ ergs, $`n=1`$ cm<sup>-3</sup>, $`M_{\mathrm{ej}}=2\times 10^5`$ M. The full, dotted, and dashed lines correspond to $`ϵ=0`$ (adiabatic), 0.5 (partially radiative), and 1 (highly radiative) respectively. The dash-dotted line is plotted by allowing $`ϵ`$ to evolve with time (see Sect. (5) in the text).
Figure 2. Evolution of the bulk velocity $`v`$. Parameters and line styles are the same as in Fig. 1.
Figure 3. Evolution of the shock radius $`R`$. Parameters and line styles are the same as in Fig. 1.
Figure 4. Evolution of the total kinetic energy $`E_\mathrm{K}`$. Parameters and line styles are the same as in Fig. 1.
|
no-problem/9906/astro-ph9906401.html
|
ar5iv
|
text
|
# A small universe after all?
\[
## Abstract
The cosmic microwave background radiation allows us to measure both the geometry and topology of the universe. It has been argued that the COBE-DMR data already rule out models that are multiply connected on scales smaller than the particle horizon. Here we show the opposite is true: compact (small) hyperbolic universes are favoured over their infinite counterparts. For a density parameter of $`\mathrm{\Omega }_o=0.3`$, the compact models are a better fit to COBE-DMR (relative likelihood $`20`$) and the large-scale structure data ($`\sigma _8`$ increases by $`25`$%).
\]
Measurements of the cosmic microwave background radiation (CMBR) provide a powerful probe of the geometry and topology of the universe. The geometry of the universe is reflected in the height, position and spacing of acoustic peaks in the CMBR angular power spectrum, while the topology of the universe is betrayed by matched circles in the microwave sky. The topology of the universe also influences the angular power spectrum by “quantising” the spectrum of density fluctuations, and in many cases, by imposing a long wavelength cut-off. An infrared cut-off in the spectrum of density fluctuations translates into a suppression of large angle CMBR fluctuations on the surface of last scatter. This effect has been used to rule out flat and hyperbolic models with toroidal topologies.
Earlier we conjectured that similar negative conclusions would not apply to generic compact hyperbolic models as the majority of the large angle CMBR power in a hyperbolic universe comes not from the surface of last scatter, but from the decay of curvature perturbations along the line of sight. To a lesser extent, the same will be true in the newly popular flat models with a cosmological constant, or in flat models with other forms of exotic dark matter such as tangled string networks. In what follows we shall not only verify our conjecture, but show that finite hyperbolic models actually provide a significantly better fit to the COBE-DMR data than their infinite counterparts.
Our main focus will be on universes with compact hyperbolic spatial sections as these are the most appealing from a theoretical standpoint. However, the growing body of observational evidence favouring a flat universe with a cosmological constant prompts us to reconsider models with three-torus topology. We refer to a model as being “small” if its comoving spatial volume is less than the comoving volume enclosed by the particle horizon (measured in the covering space). The ratio of the horizon volume to the volume of the space exceeds 500 for several of the models we looked at. Small universe models are obtained from the usual FRW models by making identifications between different points in space. These identifications break global isotropy and homogeneity, but do not alter the evolution history. There is one caveat to the last statement: by altering the mode spectrum, the topological identifications will alter the vacuum structure, leading to a Casimir-like vacuum energy that could alter the dynamics. If an effective cosmological constant could be linked to the universe having non-trivial topology, we would have a strong motivation for re-considering flat models.
On large angular scales, temperature fluctuations in the CMBR are related to the fluctuations in the gauge-invariant gravitational potential $`\mathrm{\Phi }`$ by the Sachs-Wolfe equation
$$\frac{\delta T(\theta ,\varphi )}{T}=\frac{1}{3}\mathrm{\Phi }(\eta _{sls},r_{sls},\theta ,\varphi )+2_{\eta _{sls}}^{\eta _o}\mathrm{\Phi }^{}(\eta ,r,\theta ,\varphi )𝑑\eta .$$
(1)
Here $`\eta `$ denotes the conformal time, $`\mathrm{\Phi }^{}=_\eta \mathrm{\Phi }`$, and $`\eta _{sls}`$ and $`\eta _o`$ denote recombination and the present day respectively. The evolution of the gauge-invariant potential from last scatter until today is described by
$$\mathrm{\Phi }^{\prime \prime }+3(1+c_s^2)\mathrm{\Phi }^{}c_s^2\mathrm{\Phi }+(2^{}+(1+c_s^2))\mathrm{\Phi }=0,$$
(2)
where $``$ is the conformal Hubble factor and $`c_s`$ is the sound speed in the cosmological fluid. In a flat matter dominated universe we have $`c_s=0`$, $`=2/\eta `$ and (2) tells us that $`\mathrm{\Phi }^{}=0`$. Consequently, the second term in (1) vanishes and the temperature fluctuations are all imprinted when matter and radiation decouple. The presence of either curvature or a cosmological constant alters the time evolution of the expansion rate $``$, leading to a decay of the potential $`\mathrm{\Phi }`$. In these models, the line-of-sight integral (Integrated Sachs-Wolfe, or ISW effect) in (1) can be the dominant source of large angle temperature fluctuations.
The gravitational potential can be expanded in terms of the eigenmodes, $`\mathrm{\Psi }_q`$, of the Laplacian:
$$\mathrm{\Phi }=\underset{q}{}\underset{n}{}\delta _q^n\mathrm{\Psi }_q.$$
(3)
In the above equation $`n`$ denotes the multiplicity of each eigenmode and $`\delta `$ denotes the amplitude. The eigenmodes are found as solutions of the equation $`\mathrm{\Psi }_q=q^2\mathrm{\Psi }_q`$. The geometry of the space determines the form of the Laplacian operator $``$, while the topology determines the boundary conditions. Though it is a simple exercise to write down the eigenmodes for any of the 10 flat topologies, hyperbolic manifolds have defied description. The first breakthrough came last year when Inoue found the 14 lowest eigenmodes of the Thurston space using a numerical method originally developed for 2-dimensional manifolds. By refining the method and using a more powerful computer, Inoue extended the count to include the first 36 eigenmodes. These modes have since been used to study the CMBR in a universe with Thurston topology. Using the same numerical method, Aurich has studied the CMBR in a small hyperbolic orbifold with tetrahedral topology. Recently a new, fully automated, algorithm for finding the eigenmodes was discovered and implemented. The list of solutions has grown from 36 to several thousand in the past two weeks. Our cosmological simulations use the first 100+ modes for each of 4 spaces selected from the SnapPea census of compact hyperbolic manifolds. Our selections are: the Weeks space, m003(-3,1), as it is smallest know; the Thurston space, m003(-2,3), in order to compare our results to Inoue’s; and two larger examples, s718(1,1) and v3509(4,3), to see how the size of the space influences the CMBR. In units of the curvature radius cubed, the volumes of our selections are 0.9427, 0.9814, 2.2726 and 6.2392 respectively. To get a feel for how the eigenmodes look, Figure 1 shows the lowest eigenmode of the Weeks space extended across the entire Poincaré ball. The Poincaré representation is a conformal mapping of $`H^3`$ into a unit ball in $`E^3`$.
The cosmological simulations are readily performed in spherical coordinates, where the eigenmodes can be expanded:
$$\mathrm{\Psi }_q=\underset{\mathrm{}=0}{\overset{\mathrm{}}{}}\underset{m=\mathrm{}}{\overset{\mathrm{}}{}}A_{q\mathrm{}m}R_q\mathrm{}(r)Y_\mathrm{}m(\theta ,\varphi ).$$
(4)
Here the $`Y_\mathrm{}m`$’s are spherical harmonics and the radial function $`R_q\mathrm{}`$ are either spherical (flat space) or hyperspherical (hyperbolic space) Bessel functions. It is worth emphasising how dramatically different the eigenmodes look in compact flat spaces and compact hyperbolic spaces. In flat space the expansion coefficients $`A_{q\mathrm{}m}`$ are peaked around certain values of $`\mathrm{}`$, and we can always rotate the coordinate system so that $`A_{q\mathrm{}m}=0`$ for $`m0`$. In other words, global isotropy is badly broken. In contrast, the $`A_{q\mathrm{}m}`$’s for hyperbolic eigenmodes are statistically independent of $`\mathrm{}`$ and $`m`$. The $`A_{q\mathrm{}m}`$’s are gaussian distributed pseudo-random numbers with variance proportional to $`1/k^2`$. While it is possible to assign a wavenumber $`k=\sqrt{q^21}`$ to hyperbolic eigenmodes, it is impossible to assign a wavevector. They are essentially omni-directional. You do not need inflation to explain why the microwave sky is nearly isotropic in a small hyperbolic universe.
Once the eigenmodes have been found, it is a simple though laborious task to evaluate equation (1) using equations (2) and (3). We took the mode amplitudes to be gaussian random numbers with variance $`\sqrt{k}/(k^2+4)`$ (hyperbolic space) or $`1/k^{3/2}`$ (flat space). These are the standard inflationary power spectra.
Figure 2 shows the angular power spectrum for a universe with density parameter $`\mathrm{\Omega }_o=0.3`$ and Weeks topology. Also shown is the angular power spectrum for the corresponding infinite model, along with the COBE-DMR data points. The effect of varying the density parameter is shown in Figure 3, while the effect of varying the volume of the space is shown in Figure 4. Notice that the large volume space v3509(4,3) produces an angular power spectrum very similar to the infinite model shown in Figure 1.
We quantified how well each model fit the COBE-DMR data by modeling the spread in the $`\mathrm{\Delta }T`$’s by smooth probability distributions (the statistics are not gaussian) and performing a standard likelihood analysis. Our results are based on 1000 realizations of each topology for five values of the density parameter. The likelihoods relative
to the corresponding infinite model are listed in Table 1, along with the likelihoods of the infinite hyperbolic models relative to a fiducial flat matter dominated universe.
The lower the density, the better the compact models fare relative to their infinite counterparts. The mechanism behind this result can be seen at work in Figure 1. As we decrease the density, the infrared cut-off in the mode spectrum reduces the contribution from the first term in equation (1), while the line of sight contribution from the second term becomes increasingly important. Rather miraculously, the two effects almost precisely cancel out for the low volume models, resulting in a flat or mildly tilted spectrum. On small scales there is no difference in the shape of the angular power spectra for compact and infinite models, but the compact models have a higher COBE-DMR normalization. This helps raise the predicted size of density fluctuations on 8 $`h^1`$ Mpc scales from the low value of $`\sigma _8=0.6`$ for an infinite model, to the larger value of $`\sigma _8=0.75`$ for the Weeks model (both with $`\mathrm{\Omega }_o=0.3`$). Measurements of the present day cluster abundance favour $`\sigma _8=0.9\pm 0.1`$ if $`\mathrm{\Omega }_o=0.3`$. At this stage we should stress that our results primarily affect the fit to the COBE-DMR data. On small scales there is a growing body of observational evidence for an acoustic peak at $`\mathrm{}220`$, which is consistent with the universe being flat, not hyperbolic. We should also mention that our results disagree with those of Bond et al based on the method of images, and agree with those of Inoue et al based on the finite element method.
If the universe is flat, but dominated by a cosmological constant, could the toroidal universe models be saved by contributions from the the line of sight integral in (1)? Our preliminary investigation suggests that they may, but not so much in the power spectrum department as in the breaking of global isotropy. For example, a flat cubical three-torus with side length equal to half the horizon distance has a relative likelihood of $`0.26`$ when $`\mathrm{\Omega }_\mathrm{\Lambda }=0`$ and $`0.25`$ when $`\mathrm{\Omega }_\mathrm{\Lambda }=0.7`$. The real difference between these two models can be seen in Figure 5. Because the ISW effect mixes together different modes sampled at different points, it helps to hide the egregious breaking of global isotropy that lead to the matter dominated versions of these models being ruled out.
Returning to the compact hyperbolic models, we want to see if the ISW effect ruins the the matched circle test for non-trivial topology. The matched “circles in the sky” occur wherever the surface of last scatter self-intersects. Since the surface of last scatter is a 2-sphere, the intersections occur along circles. We see two copies of each circle of intersection, centered at different points on the sky. The portion of the microwave temperature coming from the surface of last scatter will be identical around each circle. However, the ISW contribution will be uncorrelated.
Taking one realization of the Weeks universe (Figure 6), we find the temperatures around a pair of matched circles (Figure 7a) and see that the match is poor. However, the ISW effect only operates on large angular scales, so we filter Figure 6 to remove all power below $`\mathrm{}=21`$. The temperature match for the filtered sky is shown in Figure 7b. The correlation coefficient increases from $`0.29`$ to $`0.92`$ after filtering out modes with $`\mathrm{}20`$. The matched circle pairs will persist until the visible universe is simply connected, which occurs at around $`\mathrm{\Omega }_o=0.95`$ for most models in the SnapPea census. If we do live in a small universe, the Microwave Anisotropy Probe will find matched circles in the sky when it starts collecting data in 2001.
This work grew out of earlier collaborations and extensive discussion with Glenn Starkman and Jeff Weeks. Financial support was provided by NASA through their funding of the Microwave Anisotropy Probe satellite mission http://map.gsfc.nasa.gov/.
|
no-problem/9906/astro-ph9906009.html
|
ar5iv
|
text
|
# How many 𝜆 Boo stars are binaries ?
## 1 Introduction
The class of $`\lambda `$ Boo stars still presents poorly defined characteristics, and this more than 50 years after the discovery of the prototype member $`\lambda `$ Boo by Morgan et al (1943) who noted the abnormally weak metal lines of this A0 dwarf star.
The properties that should define a $`\lambda `$ Boo star are not clearly established; the proposed spectroscopic criteria are usually based on the weakness of metal lines, especially of the Mg II 4481, compared with what is expected from the hydrogen line type, while C, N, O and S have nearly solar abundances. The kinematic behaviour should allow to distinguish these stars from the metal poor A-type Horizontal Branch stars. Moderate to high projected rotational velocities are usually found among $`\lambda `$ Boo stars, although some exceptions have been recently identified (e.g. HD 64491 and HD 74873 selected by Paunzen & Gray 1997).
The result of the vague definitions of these non-evolved metal underabundant stars is well reflected by the variety of opinions existing at present about the members of this class. The metal abundances obtained up to now reveal a high scatter from star to star.
Details on the evolution with time of the $`\lambda `$ Boo definition are summarized in Faraggiana & Gerbaldi (1998); the not clearly defined properties of these stars are, at least partially, responsible of the various hypotheses proposed to explain the $`\lambda `$ Boo phenomenon as well as of the uncertainty on the age attributed to these objects, which spans from that of stars not yet on the Main Sequence to that of old objects descending from contact binary systems.
The present paper reviews the characteristics of the members of this class according to recent compilations and discusses the effect of duplicity on a composite spectrum as source of misclassification for some of these candidates.
In a modern astrophysical perspective, age, position in the HR diagram and chemical abundances are the key quantities which describe a class of stars. The purpose of introducing a class of stars is to help identify a common underlying phenomenology. For $`\lambda `$ Boo stars the aim is to find the common factor which can explain the observed chemical peculiarities; to be meaningful this must explain a statistically significant sample of stars. Bearing this in mind, it is clear that any classification scheme which does not rely on abundance criteria is unlikely to be helpful. It is probable that as high accuracy abundance data accumulate, we will have to revise our concept of $`\lambda `$ Boo stars and probably reach a more physical definition.
## 2 The $`\lambda `$ Boo candidates
Many stars were classified as $`\lambda `$ Boo in the past. The catalogue of Renson et al (1990) includes over 100 candidates. Many of these turned out to be misclassified and the whole sample results too heterogeneous. We selected stars classified as $`\lambda `$ Boo in recent papers based on modern data, hoping to extract a more homogeneous sample. They should be considered $`\lambda `$ Boo candidates, since for many of them further analysis to check whether they match any given $`\lambda `$ Boo definition is still required. The candidates we selected, with the exception of three of them, have been listed in at least one of the following papers: Abt & Morrell(1995; hereafter AM), Paunzen et al (1997; hereafter CC), Gray (1999; hereafter G). The exceptions are: HD 290492 and HD 90821 which were classified as $`\lambda `$ Boo by Paunzen & Gray (1997); HD 105759 for which the G classification is unpublished.
Both methods and scope differ among the three papers (i.e. AM,CC,G). However, the degree of reliability of each of them is difficult to define. Promising candidates are found in all lists, although Gray is probably the most reliable source, because the author (Gray & Garrison 1987, 1989a,1989b; Garrison & Gray 1994; GG in Table 1) has classified a large sample of “normal” and “standard” stars using the same methods.
AM is a study of stellar v $`\mathrm{sin}i`$ of 1700 A-type stars of the Bright Star Catalogue (Hoffleit & Warren 1994) (BSC). On the basis of their available spectra (photographic spectra of dispersion 39 Å$`\mathrm{mm}^1`$) they give a classification for each star. Some are classified as $`\lambda `$ Boo. <sup>1</sup><sup>1</sup>1AM classification has been criticized by the referee (Dr. E. Paunzen) because the standards of Morgan et al. (1978) defined for a dispersion of 125 A mm<sup>-1</sup> are used, in spite of the higher dispersion available. In fact since most of the $`\lambda `$ Boo stars have $`v\mathrm{sin}i>150`$ km/s a higher dispersion does not improve the results. However, AM note that, even if this procedure may in fact lead to an inaccuracy in the spectral type of 0.27 subclasses they could see faint lines better.
CC, on the contrary, is a paper specifically aimed at $`\lambda `$ Boo stars. Their working definition is “Pop I hydrogen burning A-type stars, which are, except of C, N, O and S, metal poor”. The catalogue contains 45 objects and includes stars with a $`T_{\mathrm{eff}}`$ as low as 6500 K. The stars should have the same characteristics of those of $`\lambda `$ Boo itself and establish a homogeneous group of $`\lambda `$ Boo stars. According to the criteria adopted by the authors, the log g must be consistent with a Main Sequence evolutionary status, but the photometrically derived log g is less than 3.6 for 9 of the 45 selected objects. Accurate abundances for some elements are available for less than half of these stars and the abundances of the key elements C,N,O and S for an even smaller number of objects. Keeping in mind our remarks on the importance of abundances in establishing the $`\lambda `$ Boo status, this status still has to be confirmed for over one half of the stars in CC.
G is also specifically devoted to $`\lambda `$ Boo stars. It is based on his own accurate classification for all but one star (for which the abundance analysis has been performed by Stürenburg (1993; hereafter St93)). This classification is based on the comparison with a set of a previously selected sample of standard stars with various v$`\mathrm{sin}i`$ values. This list must be considered as the most reliable source of $`\lambda `$ Boo candidates because it is based on a large set of homogeneous data.
Since the $`\lambda `$ Boo nature rests ultimately on a peculiar abundance pattern, only an accurate abundance analysis based on high quality data and covering a broad spectral range can allow to either retain or reject these candidates.
The list assembled in such a way comprises 89 objects, 9 of which are in common with CC, G and AM; further 7 further objects are in common between CC and G, but were not observed by AM. Thus out of 89 candidates there is concordance among different classificators at most in 16 cases, i.e. less than 20 %. We interpret this poor agreement as evidence of the subjective classification criteria adopted by each author.
We intend to examine the properties of these stars with the aim at investigating:
i) if a homogeneous group can be selected;
ii) if one of the hitherto proposed theories about the origin of the $`\lambda `$ Boo phenomenon is able to explain all the observed peculiarities of these stars.
## 3 Spectral characteristics
Abundance analyses have been performed only for a few $`\lambda `$ Boo stars. Baschek and Searle (1969) analyzed 5 stars through the curve of growth method and rejected 2 of them from the $`\lambda `$ Boo class. Venn and Lambert (1990) made a new modern analysis of the 3 stars retained as $`\lambda `$ Boo by the previous authors. St93 determined the metal abundances for 15 stars, 2 of which were in common with those considered in the two previous papers. Heiter et al. (1998) analyzed 3 stars, 2 of which are in common with previous authors. Paunzen et al (1999) derived non-LTE abundances of C and O for 16 and 22 stars, respectively. To these analyses of the visual spectra we can add the semiquantitative study of the UV spectra of 10 $`\lambda `$ Boo candidates (3 of which are not included in our Table 1) made by Baschek and Slettebak (1988).
The comparison of the abundances obtained so far shows the erratic behaviour of the abundance anomalies in different stars: an emblematic example is the peculiarly slight over-abundance of Mg found by St93 in HD 38545 and the almost solar Mg abundance found in HD 193281.
Solar abundances of the elements C,N,O and S is a property that requires to be further proved; in fact it is essentially based on the Venn and Lambert (1990) study of 3 stars, but we stress that the measured lines of these elements refer to the neutral stage and lie in the red or near infrared spectral range so that a possible contamination by a cool companion star would increase the abundances of C, N, O and S. The other semiquantitative studies by Baschek and Slettebak (1988) and by Andrillat et al. (1995) are not conclusive, since in these two papers opposite conclusions are drawn in these two papers for what concerns the C,O and S elements (N does not appear in the Andrillat et al. paper).
The Paunzen et al (1999) paper strengthens our remarks on C, N, O and S abundance pattern. In fact, the mean \[C/H\] abundance results -0.37 $`\pm `$ 0.27, and the comparison of O abundances derived previously, show the inconsistency of those derived from lines in UV, optical and near-IR regions. The extreme cases of HD 141851 and HD 204041 with \[C/H\]$`=0.81`$ pose some doubts on the notion that C is solar in $`\lambda `$ Boo stars. Moreover, for HD 204041 even the LTE \[C/H\]$`=0.75`$ is not far from the \[Fe/H\]$`=0.95`$ found by St93, but we recall that a conclusive abundance pattern can be obtained only by analyzing all elements with the same model, i.e. with the same parameters $`T_{\mathrm{eff}}`$, log g, microturbulence, vsini and abundances.
The large survey by St93 does not include any abundance determination of N, O and S; an almost solar abundance is derived for C, but no details on the measured lines are given; considering the paucity of C lines in the spectral ranges analyzed by St93, quite possibly most of the results rely mainly on the CI 4932 Å line. We note also that a similar, almost solar abundance is derived for the Mg, which is usually the key element for the $`\lambda `$ Boo classification, in one third of his sample stars. Another peculiar element behaviour is that of Na whose abundance is found to be lower than the solar one in 5 stars, almost solar in 2 stars and higher than the solar one in 6 stars.
The present knowledge of the Balmer line profiles of the $`\lambda `$ Boo stars is based on the studies by Gray (1988) and by Iliev and Barzova (1993a, 1993b, 1998). Gray pointed out peculiar Balmer profiles in some $`\lambda `$ Boo stars from classification dispersion spectra. In the spectra of some stars he noted Balmer lines with broad wings and weak core, and an inconsistency of the $`\beta `$ index with the luminosity class based on the extent of the hydrogen-line wings. These problems were not present in other $`\lambda `$ Boo stars . Thus, according to the appearence of the Balmer lines, he divided the $`\lambda `$ Boo stars into two distinct classes: NHL (Normal Hydrogen-Line profile) and PHL (Peculiar Hydrogen-Line profile). Subsequent studies by Iliev and Barzova based on high dispersion photographic spectra confirmed and quantified this peculiarity. This dicothomy is therefore observationally well established, but we still lack a theoretical interpretation for it.
## 4 Kinematics
Given the metal-weakness of $`\lambda `$ Boo stars, it is proper to wonder whether some of them are truly metal-poor stars with halo kinematics. In Table 1 the kinematics of the stars observed by Hipparcos and with known RV has been computed from proper motions and radial velocity as described in Johnson & Soderblom (1987). A left-handed system was used in which $`U`$ is directed towards the Galactic anticenter, $`V`$ in the direction of Galactic rotation and $`W`$ towards the north Galactic pole. In Fig. 1 we plot the rotational velocity $`V`$ versus $`\sqrt{(U^2+W^2)}`$, which may be taken as a measure of kinetic energy not associated to rotation. All velocities are heliocentric, so that stars with $`V`$ around 0 $`\mathrm{kms}^1`$ and small $`\sqrt{(U^2+W^2)}`$ are qualified as disk members. All the $`\lambda `$ Boo stars pass this test, in agreement with Gómez et al (1998). This outcome was expected as a consequence of the brightness of the stars for which velocity data are available; out of the 14 sample stars weaker than V=8, eight were not observed by Hipparcos and 5 lack a measured RV. Thus, among the faint stars of our sample, we could compute the space velocities only for HD 101108. In Fig.2 we note that HIP 5321 (HD 6870) and HIP 47752 (HD 84123), shown as black dots, present a marginal deviation from disklike kinematics.
The first star has been classified as Population II blue straggler (Bond & MacConnell, 1971) and has been identified as a member of the $`\sigma `$ Pup group, which is similar to the globular cluster 47 Tuc (Rodgers, 1968, Eggen 1970a, 1970b), which has metallicity \[Fe/H\]$`=0.71`$ (Da Costa & Armandroff 1990). This star also presents a UV excess in the (U-B) colour, similar to that of Pop II stars, which is not found in other $`\lambda `$ Boo candidates. However, the nature of this star is still matter of debate. Paunzen et al (1999) consider it a $`\lambda `$ Boo star. Their O abundance of \[O/H\]$`=+0.05`$ could be reconciled with an extreme case of $`\alpha `$ element enhancement as found in Pop II stars. However, their C value \[C/H\]=+0.14 prompt interpretation as Pop I. Clearly an elucidation of its nature must rely on the determination of the abundaces of C,N,O, $`\alpha `$ elements and iron-peak elements using the same stellar atmospheric parameters. We note here that the abundances of C and O have been obtained by Paunzen et al (1999) by assuming very different values of $`v\mathrm{sin}i`$ (128 and 200 km s<sup>-1</sup>, respectively).
HD 84123 belongs to the cooler ($`T_{\mathrm{eff}}`$= 6900 K) $`\lambda `$ Boo candidates selected only by CC and has peculiar characteristics. The UBV colours indicate that it does not belong to the MS, as confirmed by the photometrically derived log g=3.5; the UV magnitudes measured by the TD1 experiment (Thompson et al. 1978) do not fit those computed with these parameters at 2365 and 2740 Å.
## 5 Atmospheric Parameters
Photometric indices of the $`uvby\beta `$ system are examined in order to determine the atmospheric parameters $`T_{\mathrm{eff}}`$ and log g from the calibration of the colour indices by Moon and Dworetsky (1985) (MD).
### 5.1 Colour excess
The observed indices are taken from the catalogue of Mermilliod et al. (1997). For 5 stars ( HD 5789, HD 171948, HD 177120, HD 192424 and HD 198161) they are not available and for HD 184190 only the $`\beta `$ index was measured; for the remaining 82 stars the colour were dereddened using the UVBYLIST code of Moon (1985). In fact, 30 stars in Table 1 have a visual magnitude weaker than 6.5 so that we cannot neglect a possible influence of reddenning. In 36 % of the stars the colour excess has a negative value, which is twice the occurrence found in the sample of 71 bright dwarf A0 stars recently analyzed by Gerbaldi et al (1999) with the same procedure. The most negative colour excess among this latter sample was of -0.02, found for one star only, while for the other stars with negative colour eexcess it was equal to -0.01. Among the objects of the present sample the stars with negative colour excess equal to or less than -0.02 are at least 10. The UVBYLIST code requires that the stars are assigned to one of six possible groups, depending on spectral type and colours; for several stars the choice of the appropriate group is ambiguous owing to inconsistencies of the colour indices and spectral type. We decided to give less importance to the spectral type and rely more on the unreddened $`\beta `$ index. For the still remaining doubtful cases an unambiguous choice between the parameters derived with different choices cannot be performed without further information; for the most doubtful cases the various possible choices have been retained.
Since the MD procedure has been established for solar abundance stars, we investigated if the derived colour excess is related to the lower than solar abundances, attributed to these stars by analyzing, with the same procedure, the theoretical colour indices computed for \[M/H\] = - 0.5 and \- 1.0 (Castelli 1998). The colour indices computed for abundances 10 times lower than solar produce a spurious colour excess up to 0.03 for $`T_{\mathrm{eff}}`$=8000 K, but never a negative value. We thus conclude that a negative value of E(b-y) cannot be directly related to metal underabundances.
Strange enough the value of E(b-y) is not correlated to the V magnitude which, for these non evolved stars, is expected to be roughly related to their distance (Fig. 2).
A high value of v sin i is expected to affect the photometric (Collins & Smith 1985) and spectroscopic (Cranmer & Collins 1993; Collins & Truax 1995) characteristics of A-type stars. The inspection of the stars for which AM measured the v sin i has not revealed any correlation between the anomalous negative colour excess and the v sin i value.
We observe also that for 12 stars the photometric measures refer to two components of a binary system and thus require some sort of disentangling from the influence of the companion.
### 5.2 $`T_{\mathrm{eff}}`$ and log g
The parameters $`T_{\mathrm{eff}}`$ and log g have been derived from the MD TEFFLOGG code according to the above described choice of colour excess. The table of these parameters is available from the authors.
The stars cover a broad domain of the HR diagram around the prototype $`\lambda `$ Boo; in fact $`T_{\mathrm{eff}}`$ spans from 6500 (HD 4158 and HD 106223) to 14500 K (HD 22470) and log g from 2.87 (HD 108283) to 4.50 (HD 294253). We note that the 8 stars with $`T_{\mathrm{eff}}`$ lower than 7100 K belong to the CC list and the 5 stars with $`T_{\mathrm{eff}}`$ higher than 10500 K are from AM; a systematic shift in $`T_{\mathrm{eff}}`$ probably exists between these two sources. However, the bulk of the stars in the two papers lies in the temperature range they share.
We conclude that before proceeding further with the discussion of the region of the HR diagram covered by these stars it is necessary to examine their photometric properties and their peculiarities.
### 5.3 Colour-magnitude diagram
For the stars with a Hipparcos measured parallax we may readily build a colour-magnitude diagram. Undereddened colours are used, the only “faint” star being HD 105058, which is expected to be unreddened (Faraggiana et al 1990). This is displayed in Fig. 3, where the isochrones of solar composition and ages of 200 Myr and 500 Myr of the Padova group (Bertelli et al. 1994) are also shown. We note the inconsistency between the position in the HR diagram and the log g values derived from $`uvby\beta `$ photometry and by using the MD procedure; this point deserves further analysis. Two facts are clear from this comparison: 1) most of the $`\lambda `$ Boo stars lie outside the main sequence; 2) a range in ages between 0.2 and 1 Gyr is necessary to explain the observed dispersion in the colour magnitude diagram, if we consider these stars in post-main sequence phase. The alternative hypothesis that all these stars are very young objects in the late phase of their pre-main sequence evolution is highly improbable; in fact, according to the scenario proposed by Turcotte and Charbonneau (1993) the accretion of gas, but not of dust, should occur in less than 10<sup>6</sup> years and the number of bright $`\lambda `$ Boo candidates would imply that the star formation process is still very active in the solar neighbourhood, which would imply a large number of MS B stars in a similar volume. A fraction of pre-main sequence stars cannot either be excluded or demonstrated on the basis of only these data.
## 6 Peculiarities of $`\lambda `$ Boo candidates
We performed a systematic search of known binaries among the $`\lambda `$ Boo candidates of Table 1.
HD 38545 and HD 141851 are visual binaries whose speckle interferometric measurements are given by McAlister et al (1993). Given the small separation of the pairs, ground-based spectra will always be the combined spectrum of these binaries; for the first star, Hipparcos data demonstrate that the effect of the companion is surely not negligible. Starting from the analysis of the H<sub>γ</sub> profile, Faraggiana et al (1997) were able to show that a third star, HD 111786, is indeed a binary system and this finding was supported by the identification of the lines (in the visual, but not in the UV range below 2000 Å) of the cooler companion, which are narrow if compared to those of the primary star, this happens quite likely because the companion is seen pole on. In order to determine abundances, the spectra of the two components must be disentangled.
Twelve out of 89 stars of our sample are known to be visual binaries according to the Mermilliod et al. (1997) catalogue. For these binaries only the combined colour indices have been measured and therefore they cannot be safely used to derive the atmospheric parameters of the primary component. In order to evaluate the influence of the companion star, we extracted the angular separation and the magnitudes for the components A and B from the Washington Visual Double Star Catalog (Worley & Douglass 1997; hereafter WDS); these data are given in columns 10 and 11 of Table 1 for the objects with an angular separation of less than 10 arcsec. The contamination on the observed spectra depends on the luminosity difference of the components and on the slit width of the spectrograph on the sky; for several stars the observed spectra are expected to be affected by the companion star and, in some cases (HD 290492, HD 38545, HD 47152, HD 141851, HD 153808, HD 159082, HD 160928, HD 170000 and HD 225218), a composite spectrum cannot be avoided with observations from ground instruments unless it can be demonstrated that the companion luminosity is much weaker.
The effect of the secondary star should underlie the discordant classifications proposed for HD 225218, either B9 III or A3 Vs; moreover its UV magnitudes, as determined by the TD1 experiment (Thompson et al. 1978), would suggest a much lower reddening than that derived from the Strömgren photometry, i.e. E(b-y)=0.03, a value more coherent with the stellar visual magnitude.
The two stars HD 141851 and HD 149303, being X-ray sources (Hünsch et al. 1998), are expected to have cool companions.
Significant discrepancies in the magnitude and photometric colours of the AB system HD 193281 are reported in the literature (BSC).
The ESA Hipparcos Catalogue (1997) allowed to complete information on some stars and to add new binaries; these data are collected in columns 8 and 9 of Table 1.
Further known binaries are:
$``$ The already quoted spectroscopic binary HD 111786 and HD 84948 and HD 171948. The 4 components of the last two binaries are all $`\lambda `$ Boo stars, according to Paunzen et al (1998); however, we note that the Mg does not show underabundance higher than the other metals and that the abundances of the key elements C,N,O and S are not given.
$``$ HD 142703 is a suspected occultation double (BSC).
$``$ HD 79108 is a suspected SB (BSC), but we could not retrive any further information on the possible influence of the companion on photometric and spectroscopic data.
$``$ The oxygen spectrum indicates that HD 149303 is an SB system (Paunzen et al 1999).
We also note the inconsistent classifications assigned to HD 22470 i.e. a $`\lambda `$ Boo star (AM) or He-weak with variable intensity of the SiII 4128-30 doublet (Gray 1988). The explanation of the peculiarities has been given by the Hipparcos detection of its duplicity (see columns 8 and 9 of Table 1). Similar remarks apply to two other binaries, HD 47152 and HD 170000, which have been both classified, as single objects, either Ap or $`\lambda `$ Boo .
Two more spurious $`\lambda `$ Boo candidates are HD 130158 which is in reality an Ap Si-$`\lambda `$4200 star (Gray 1988) and HD 159082 which is a B9 Hg-Mn (see, for example, Hubrig & Mathys 1996).
The peculiarity of HD 108283, which has a very high c<sub>1</sub> value and therefore a derived log g which is the lowest of our sample stars, but also a very high v sini value, deserves further analysis before being assigned to the $`\lambda `$ Boo class; Hauck et al. (1998) rejected it from the $`\lambda `$ Boo class.
We conclude that from ground and space observations close duplicity which is able to affect the observed spectrum, has been already observed or suspected for 24 $`\%`$ or 33 $`\%`$ (if the stars classified U by Hipparcos are included) of the stars of Table 1 and that some spurious $`\lambda `$ Boo candidates are present.
## 7 Predicted composite spectra
In the abundance analyses $`\lambda `$ Boo stars have been generally considered as single stars. The exception is the analysis of the two SB2 stars HD 84948 and HD 171948 by Paunzen et al (1998). In the present section we consider the influence of a possible companion on model parameters and on derived abundances and discuss the ability to pick out the already known binaries on the basis of spectra. This has a direct bearing on the issue whether binarity may lead to significant systematic errors in the abundance analysis and classification of $`\lambda `$ Boo stars.
In a binary system, if the angular separation of the two components is too small to be detected, and the two stars have a similar mean to high projected rotational velocity, the spectral lines of the components are not resolved in the observed spectrum and duplicity is not easily detected spectroscopically. If the two components have significantly different parameters, the duplicity should appear when a large spectral range is covered by observations, the contribution of each star being different at different wavelengths. This implies that there are doubts on binarity, the analysis of spectral ranges of only a few hundred Å may not be sufficient to establish or reject the binarity.
If a binary star is analyzed in the classical 3500-5500 Å range, the average values of $`T_{\mathrm{eff}}`$ and log g are derived from the combined light; synthetic spectra are computed from the adopted model atmosphere and the best fit with observations is obtained by adjusting the microturbulence value and the chemical abundances.
In Fig. 4 we illustrate the result of the comparison of a composite spectrum with two single-star spectra. We note that in these spectra the H<sub>γ</sub> profile is practically the same, but most of the metal lines simulate a metal underabundance in the composite spectrum. The hydrogen profile, in this case, would be considered as NHL. Metal underabundances would be obtained from the combined spectrum if analyzed by disregarding its binary nature. A similar result on the apparent metal underabundances of the Am triple system, $`\pi `$ Sgr, if analyzed as a single object, has been obtained by Lyubimkov & Samedov (1987).
To illustrate how a PHL profile can be produced by the combination of 2 single profiles, we constructed the composite H<sub>γ</sub> line given in Fig. 5 and compared it with the single spectra computed with $`T_{\mathrm{eff}}`$=7000 K, log g=4.0, which fits the wings, and with $`T_{\mathrm{eff}}`$=9750 K, log g=3.0, which fits the core of the composite spectrum. All these profiles refer to solar abundance models. The example has been constructed in such a way as to enhance at most the effect which is observed in PHL stars and is not intended to simulate any really observed object.
## 8 Discussion
In the previous sections we stressed the fact that the $`\lambda `$ Boo candidates of Table 1 constitute a non homogeneous group. By adding the informations spread in the literature, we summarize that among $`\lambda `$ Boo stars:
\- some but not all show PHL profiles (Gray 1988, 1998);
\- some but not all have an IR excess (2 stars in Sadakane & Nishida 1986; 2 stars in Cheng et al 1992; 1 star in Grady et al 1996; 2 more in King, 1994);
\- some but not all have a shell surrounding the star which is detected by the presence of narrow circumstellar components of Ca II K line (Holweger and Rentzsch-Holm 1995) and of metal lines (Hauck et al 1995 and 1998; Andrillat et al 1995);
\- some but not all show the UV broad absorption feature centered on $`\lambda `$1600 Å (Baschek et al 1984; Faraggiana et al 1990; Holweger et al 1994). We recall that the combined effect of the stellar flux drop and the lowering IUE sensitivity is responsible for the mostly underexposed IUE spectra of the middle and late A-type stars, so that the $`\lambda `$1600 Å could be searched only among the hottest $`\lambda `$ Boo candidates;
\- for three of them an abundance pattern similar to that of the ISM has been derived (Venn & Lambert 1990).
On the basis of the few abundance coherent analyses available, several hypotheses have been made on the age of these stars:
i) very young stars which have not reached the main sequence (Waters et al 1992; Gerbaldi et al 1993; Holweger & Rentzsch-Holm 1995).
ii) dwarfs in the middle of their life on the main sequence, with an age of 10<sup>7</sup>-10<sup>9</sup> years (Iliev & Barzova, 1995);
iii) quite old objects representing a merger of binaries of W UMa type (Andrievsky 1997).
The only property common to all $`\lambda `$ Boo stars of Table 1 is the weakness of most metal lines, which is also confirmed by the negative $`\mathrm{\Delta }`$a values (a photometric index measuring the blanketing in the region $`\lambda `$5200 Å) measured by Maitzen & Pavlovski (1989a and 1989b) for the stars they observed.
Taking this common characteristic of the group as a starting point, we inspect the possible causes that may produce a weak-lined spectrum.
The classical explanation of metal underabundances, for stars belonging to Population I, is related to the existence of single stars with peculiar atmospheres in which some elements are depleted by different amounts.
Disturbing is that the abundance pattern is not the same in the $`\lambda `$ Boo candidates analyzed up to now; one particular element shows different abundance peculiarities in different stars, so that it is difficult, at present, to establish the average chemical composition of $`\lambda `$ Boo stars and thus to elaborate a theory explaining the phenomenon.
Venn & Lambert (1990) formulated the hypothesis that the $`\lambda `$ Boo phenomenon could be the result of accretion of gas but not of dust from circumstellar or interstellar material. The only modern and detailed analyses available at present are the two papers by Venn & Lambert (1990) and by St93. We compared the metal abundances derived for the $`\lambda `$ Boo stars by these authors with those of the ISM as given by Savage & Sembach (1996). This comparison shows that the similarity is only marginal; in the first place there is a large difference from star to star in the abundance of any given element, unlike what happens in the ISM, in the second place the highest underabundances in the ISM are those of Ca and Ti while in most $`\lambda `$ Boo candidates it is that of Mg.
Among possible explanations of the $`\lambda `$ Boo phenomenon one should also take into account the possibility that these stars indeed belong to a metal-weak population. The kinematic data imply that $`\lambda `$ Boo stars belong to the disk population, which has indeed a metal-poor population in the range -0.5$`<`$\[Fe/H\]$`<`$-1.0 and possibly the thick disk has a metal-weak tail at much lower metallicities (Beers & Sommer-Larsen 1995). We must examine the two possible cases that $`\lambda `$ Boo stars are either Main Sequence (MS) and therefore relatively young, or on the Horizontal Branch and therefore old.
The main argument to reject the hypothesis that $`\lambda `$ Boo stars are metal-poor MS stars is that while in $`\lambda `$ Boo stars Mg, Si, Ca and Ti are among the most underabundant elements, in metal-poor stars the even-Z light elements, synthesized by $`\alpha `$ capture processes, show an increasing enhancement over iron with decreasing metallicity, reaching a 0.4 - 0.5 dex enhancement at \[Fe/H\]$`=1.0`$.
To distinguish Blue Horizontal Branch (BHB) stars from $`\lambda `$ Boo stars on the basis of spectroscopic properties alone, is not trivial. However, the BHB population in the solar neighbourhood would be uncomfortably large if most $`\lambda `$ Boo belonged to this class. A further argument against the BHB hypothesis is that most $`\lambda `$ Boo stars are characterized by mean to high projected rotational velocities, while BHB stars are all slow rotators. All fast rotators may be thus rejected as BHB stars. Although the possibility could be still considered open for slowly rotating $`\lambda `$ Boo stars (e.g. HD 64491 and HD 74873 (Paunzen & Gray (1997) and a few others proposed by AM), their number is very small. We recall that low v$`\mathrm{sin}i`$ stars may be either intrinsically slow rotators or fast rotators seen at high inclination.
From the foregoing discussion we reject the hypothesis of membership of the class to a metal-poor population and do not discuss it any further.
A completely different origin of weak metal lines is that produced by stellar duplicity. Examples of how a composite spectrum, which is the average of two actual components of not very dissimilar spectral type, can be classified as Mg-weak is given by Corbally (1987) for HD 27657 and HD 53921. Corbally remarks also that the AB spectrum of HD 41628 ”is close to imitating a $`\lambda `$ Boo star, but the A5 Balmer line class is a compromise between A7 V strength and A3 V wings… example of two normal parent spectra producing a peculiar composite”.
The effect of veiling in the spectrum of a binary with components not very dissimilar from one another (M=2 and 1.4 solar masses) has been investigated in detail by Lyubimkov (1992). Most of his analysis, devoted to Am stars, refers to composite spectra (computed for 4 selected evolutionary phases) obtained by combining two spectra for which solar abundances are adopted only for elements lighter than Ti. A general apparent underabundance of these elements is derived by his computations when the original duplicity is neglected, in agreement with the weak metal lines obtained by our example plotted in Fig. 4.
According to the data collected in the previous sections, 11 stars of our original sample are doubles with an angular separation smaller than 1.2 arcsec, 3 stars are SB2 and 4 are probably non-single, according to the Hipparcos data. In conclusion, for 18/89= 20 $`\%`$ of our sample stars duplicity must be examined in further detail before determining atmospheric abundances.
Grenier et al. (1999) in their radial velocity study of a sample of B to F stars, included in the Hipparcos catalogue, obtained spectra for 16 stars of our sample. Of these 12 are suspected, probable or established binaries, only 4 of these are among the 18 known binaries previously mentioned. If all of them will be confirmed to be binaries the percentage will raise to 29% . We note also that 11 of these stars are in common with the G list and 8 are classified as PHL by him.
If we apply the present knowledge to the 15 stars analyzed by St93, we see that 2 of them are SB2 (HD 38545 and HD 111786), for HD 198160 and HD 198161 the atmospheric parameters, derived from the combined photometric indices, require the hypothesis that the two stars are strictly the same so that the same $`T_{\mathrm{eff}}`$ and log g can be adopted. The duplicity of these stars requires to be further examined in order to determine accurate single elements abundances. Furthermore, the variability of the 5 variable stars must be examined to assess that its amplitude does not affect the photometrically derived atmospheric parameters.
High S/N spectroscopic data of spectral regions in which not severely blended features are present are necessary to discriminate between ”veiling” (spectral lines when they retain the breadth of their temperature type, but are shallower than normal (Corbally 1987)) which indicates a composite spectrum and normal profiles with weak intensities, which are sign of real metal underabundances. Such discrimination, however, becomes extremely difficult when the observed spectrum is characterized by broad and weak metal lines as in most $`\lambda `$ Boo candidates.
What we can expect in a composite spectrum of two similar A-type stars are Balmer lines broader than those of the single components by an amount which depends on the relative RV of the components, so simulating a star with a higher log g value when compared to computed spectra or intrinsically very high for an early A-type star as it may be the case of HD 294253 (the parameters derived by MD programs are $`T_{\mathrm{eff}}`$=10370 K log g=4.50). Moreover, the composite Balmer line profile will present a flat inner core which depends on the difference of the two radial velocities as well as a global profile which may be different from what is expected from the dominating broadenings: Doppler core and linear Stark wings.
For the stars recognized to be double by speckle observations, and not observed by the Hipparcos satellite, the extraction of luminosity ratios from speckle data will be fundamental to better define the character of the two components. Algorithms to extract luminosity ratios from speckle data have been developed, but these techniques are still limited (Sowell & Wilson 1993). If the luminosity of the companion is large enough (of the order of 25% of the total luminosity), the veiling may not be neglected; in fact the metallic lines will appear weaker, thus leading to an underestimate of the metallicity.
The IR colours could also prove to be powerful diagnostic tool for the presence of cooler companions. A cool companion of HD 111786 was predicted by its photometry in the J,H and K bands by Gerbaldi (1990) on the basis of the discrepancy with the (B-V) value and was ascribed to a probable cool companion.
The foregoing discussion leads us to formulate the hypothesis that a considerable fraction of $`\lambda `$ Boo candidates are in fact binaries. This is supported by the large fraction of binaries recently discovered among $`\lambda `$ Boo stars either through the speckle technique or by the Hipparcos experiment. Also the high number of stars with a ”blue” colour excess supports that our binarity hypothesis is at the origin of distorted energy distributions and of not coherent uvby$`\beta `$ indices of several $`\lambda `$ Boo candidates. The PHL phenomenon cannot be easily explained if the stars are single, however its explanation becomes trivial if the stars are binary, as has been demonstrated in the case of the stars HD 38545 and HD 111786, classified PHL by Gray (1988, 1998), which indeed turned out to be binaries. The apparently erratic abundance patterns pose serious problems to the accretion hypothesis, but again it may be easily reconciled in the case of binary stars.
The fact that some of the stars are binaries does not exclude the possibility that chemical peculiarities are actually present in their atmospheres. However their quantification requires that the binarity is properly accounted for.
## 9 Conclusion
We have shown that each author has his own definition and list of $`\lambda `$ Boo candidates and these lists only partly overlap. Until all the classification schemes converge into a single with a physical basis there is little hope of understanding the $`\lambda `$ Boo phenomenon.
Recent observations, mainly by speckle interferometry and by the Hipparcos satellite, have detected the binary nature of several $`\lambda `$ Boo candidates and other candidates have been recognized on the basis of high resolution spectra.
We make the hypothesis that abundance anomalies are, at least partly, due to the effect of veiling in a composite spectrum and that other still undetected binaries are likely to be present among the objects collected in Table 1. Distorted and uncertain colours (e.g. stars with negative colour excess and reddened bright stars) and peculiar Balmer line profiles are reasons for suspecting duplicity.
The photometrically derived atmospheric parameters of close visual binaries refer to the average photometric indices of the components and an abundance analysis based on them requires that the two components form a twin pair or have very different luminosities. Moreover, for some of these binaries the angular separation is such that a composite spectrum cannot be avoided.
###### Acknowledgements.
We thank Fiorella Castelli for helpful discussions and suggestions and for having put at our disposal the update version of the Kurucz BINARY program. Use was made of the SIMBAD data base, operated at the CDS, Strasbourg, France. Grants from MURST 40$`\%`$ and 60$`\%`$ are acknowledged.
|
no-problem/9906/cond-mat9906215.html
|
ar5iv
|
text
|
# Internal Spatiotemporal Stochastic Resonance in a Microscopic Surface Reaction Model
## Abstract
We show the existence of internal stochastic resonance in a microscopic stochastic model for the oscillating CO oxidation on single crystal surfaces. This stochastic resonance arises directly from the elementary reaction steps of the system without any external input. The lattice gas model is investigated by means of Monte Carlo simulations. It shows oscillation phenomena and mesoscopic pattern formation. Stochastic resonance arises once homogeneous nucleation in the individual surface phases (reconstructed and non-reconstructed) is added. This nucleation is modelled as a noise process. As a result, synchronization of the kinetic oscillations is obtained. Internal stochastic resonance may thus be an internal regulation mechanism of extreme adaptability.
PACS numbers: 05.40.Ca, 05.10.Ln, 05.45.-a, 05.45.Xt
The term stochastic resonance (SR) is given to the somewhat counter-intuitive phenomenon that in a non-linear system a weak signal can be amplified by the assistance of noise. It has been introduced in 1981 by Benzi et al. in the context of a study about the periodically recurrent ice ages. Over the last two decades it has continuously attracted increasing attention and was shown to occur in many systems in biology, chemistry, and physics. Generally systems showing SR are described in a formal mathematical way using phenomenological macroscopic equations of the mean field type including (i) a bistable system with an activation barrier or some sort of threshold, (ii) a weak coherent input, and (iii) a strong external noise which helps to overcome the activation barrier. These macroscopic equations in a sense are able to describe many different systems (because stochastic resonance is a general phenomenon which occurs in many natural systems) and have been used in the description of stochastic resonance phenomena in Nd-YAG lasers homogeneous as well as heterogeneous chemical reactions, bistable quantum systems or the Lotka-Volterra model. More complex systems (e.g. a summing network of excitable units, sheep populations, two-dimensional excitable media showing spatio-temporal pattern formation, sensory systems in crayfish or in the visual cortex, or neuron-like systems) are generally modeled via Langevin equations or the Fitzhugh-Nagumo model.
In addition a few special systems have been investigated in a more general manner via a macroscopic mathematical description, e.g. an autonomous oscillating system, a system in the limit of weak noise, a system showing stochastic multiresonance, and non-dynamical systems with both internal and external noise. Note that in the latter case the internal noise is modeled in the same way as an external noise and that it is only regarded as a general internal noise without specifying the physical background. Computer simulations performed to date consider coupled neurons or general threshold devices, which are mesoscopic models, i.e. the microscopic physical picture is again neglected.
Our present model system is very unusual in the research on SR. It gives for the first time a microscopic description of the phenomenon of internal stochastic resonance and demonstrates its physical reasons on the microscopic (atomic) length scale. It is based on stochastic transitions, each with a clear physical meaning. Without noise, the system exhibits a spatially extended heterogeneous stable state with inherent local oscillations. More important, the noise is not an external input but corresponds to a physically realistic internal nucleation process. Because of the clear physical picture on the microscopic (atomic) level our model is of course specialized and cannot describe a large variety of different systems. But on the other hand the results of this model and the conclusions which can be drawn are very general ones and suggest that internal SR may be the reason for inherent synchronization and cooperative phenomena in many physical, chemical and biological systems.
We consider a slightly modified version of a previously presented model for the catalytic CO+1/2 O<sub>2</sub> reaction on Pt single crystal surfaces, which shows different types of kinetic oscillations in agreement with experimental results. The model involves CO adsorption, desorption and diffusion, dissociative O<sub>2</sub> adsorption and two surface phases (reconstructed and non-reconstructed) which form and propagate governed by the coverage with CO. The details are given below. An extended version of the model for the CO+NO reaction on Pt(100) is able to describe the experimentally observed transition into chaotical behavior via the Feigenbaum route. Our model follows the well known model by Ziff, Gulari and Barshad (ZGB model) and is investigated by means of Monte Carlo (MC) simulations. The Pt(110) surface of the catalyst is represented by a square lattice of side length $`L`$ and lattice constant $`a=1`$. From experiment it is well known that kinetic oscillations are closely connected with the $`\alpha \beta `$ reconstruction of the Pt(110) surface, where $`\alpha `$ and $`\beta `$ denote the $`1\times 2`$ and the $`1\times 1`$ surface phase, respectively. In our model CO is able to adsorb onto a free surface site with rate $`y_{\mathrm{CO}}=y`$ and to desorb from the surface with rate $`k`$, independent of the surface phase the site belongs to. O<sub>2</sub> adsorbs dissociatively onto two nearest neighbor (NN) sites with different sticking coefficients $`s`$ onto the two phases ($`s_\alpha =0.5`$, $`s_\beta =1`$). Therefore we get the oxygen adsorption rates $`y_\mathrm{O}^\alpha =1y`$ and $`y_\mathrm{O}^\beta =2(1y)`$ for the $`\alpha `$ and $`\beta `$ phase, respectively. For O<sub>2</sub> adsorption directly at the phase border where one site belongs to the $`\alpha `$ and the other one to the $`\beta `$ phase the geometric mean of these adsorption rates is used. In addition, CO is able to diffuse with rate $`D`$ via hopping onto a vacant NN site. The CO+O reaction occurs, if CO hops to a site which is covered by O and the reaction product CO<sub>2</sub> desorbs immediately from the surface. All these processes are associated with the above kinetic transition rates of the stochastic model which therefore determine the relative speed of the individual reaction processes. In the present study we use $`y=0.51`$, $`D=100`$, and $`k=0.1`$ as standard values because CO diffusion is by far the fastest process. For details see refs. .
The $`\alpha \beta `$ phase transition is modeled as a linear phase border propagation. Consider two NN surface sites in the state $`\alpha \beta `$. The transition $`\alpha \beta \alpha \alpha `$ ($`\alpha \beta \beta \beta `$) occurs if none (at least one) of these two sites is occupied by CO. This phase border propagation mechanism mimicks the growth of the $`\beta `$ phase because of the larger adsorption energy of CO on the $`1\times 1`$ phase than on the $`1\times 2`$ phase. The individual phases are stable or metastable. The direct transition from a globally homogeneous $`\alpha `$ phase into a homogeneous $`\beta `$ phase (or vice versa) is impossible; the activation barrier is infinite. The stability of the individual phases depends on the chemical coverage $`\mathrm{\Theta }_i`$ of species $`i`$ on the surface of the catalyst. For $`\mathrm{\Theta }_{\mathrm{CO}}<0.3`$ the $`\alpha `$ phase, for larger values the $`\beta `$ phase is stable. The coverages of CO and O vary in the course of the reaction because of the different sticking coefficients of O<sub>2</sub> on the two surface phases. Starting with a heterogeneous distribution of the $`\alpha `$ and $`\beta `$ phase the activation barrier for the surface phase transition is finite, but the transition into a globally homogeneous phase does not occur because of the finite surface phase propagation velocity, i.e. the $`\alpha `$ or $`\beta `$ phases cannot grow to macroscopic islands. Therefore the oscillations remain local, interfere and cancel each other on sufficiently large surfaces. The system exists in a heterogeneous, dynamically stable state with oscillations, which are locally synchronized by CO diffusion but disappear on the macroscopic length scale for large lattices. The $`\alpha `$ and $`\beta `$ phase, however, build almost homogeneous islands on a mesoscopic length scale.
The nucleation is modeled as a spontaneous $`\alpha \beta `$ or $`\beta \alpha `$ transition of a single site, completely independent of its neighbors or the particle adsorbed onto this site. Therefore this nucleation is a homogeneous process which corresponds to a weak noise which generates dynamic defects on the surface. The term weak noise is used because the nucleation rate $`\gamma [10^6,10^2]`$ is for the relevant values several orders of magnitude smaller than the other transition rates which are of the order of $`10^1`$ to $`10^3`$. The defects grow or vanish via the $`\alpha \beta `$ reconstruction depending on their chemical neighborhood, i.e. the presence or absence of CO. It has to be emphasized that all phenomena such as local oscillations, growth and decline of the heterogeneously distributed surface phase islands, quasiperiodical and chaotical behavior exist even without the consideration of the nucleation process. Because of the very small nucleation rate, which introduces only very few defects into the existing heterogeneous surface state, one might suppose that the defects could have no influence on the oscillating system, but we shall demonstrate that the nucleation process may have a profound influence on the system behavior leading to cooperative phenomena and the synchronization of the local oscillations.
In comparison with the standard problem of SR there exist several major differences. (i) We have a two-phase system with an additional chemical coverage instead of a bi-stable system. In addition none of the two phases becomes homogeneous. Both coexist in a dynamically stable heterogeneous state. (ii) No external coherent input is considered in our model. The local oscillations originate from the kinetic definition of the model itself and are very small on a macroscopic, global length scale. These internal oscillations can be seen as a substitute of the commonly used coherent input. The macroscopic synchronization of the local oscillations then corresponds to a large output signal. (iii) The source of noise is an internal physical process of the system, the homogeneous nucleation, which breaks the mesoscopic homogeneity of the surface phase islands. The noise process is therefore not an addition to a periodic input but is independent of the oscillations. In most previous studies a weak coherent input is coupled with a strong external noise. (iv) The model is investigated via MC simulations which correspond most closely to a hierarchy of master equations with all correlations included and then mapped onto a finite lattice. The simulation procedure contains additional noise by its very nature but such a noise does not lead to SR. We may characterize the structure of our model by saying that practically all processes which give rise to SR (except for the particle flux to the surface) are internal to the system.
As can be seen in fig. 1 the nucleation (noise) has a strong influence on the system behavior. Without or with small nucleation rates $`\gamma <510^5`$ only local oscillations exist. The amplitude of the global oscillations vanishes for simulations on large lattices. With increasing nucleation rate (strength of the noise) the local oscillations are synchronized on a macroscopic scale and almost reach the theoretical maximum of $`S(\omega )=0.5`$ ($`\mathrm{\Theta }_\beta `$ varies between 0 and 1). This holds also for larger lattices up to $`L=4096`$ which is the current limit for our simulations. Further increase of the noise decreases this synchronization until at a nucleation rate of $`\gamma =10^1`$ the system is completely governed by strong noise. This behavior is the fingerprint of SR.
In addition the noise has an influence on the frequency of the oscillations (see fig. 2). The normal frequency of the system is $`\omega _0`$, which can only be observed in simulations on small lattices. For nucleation rates $`\gamma >510^5`$ the phenomenon of SR occurs. The nucleation forces the system to oscillate with a different frequency $`\omega `$ starting at $`\omega 2/3\omega _0`$ for $`\gamma =510^5`$. With increasing nucleation rate the frequency increases as well up to a value of about $`\omega 2\omega _0`$ at $`\gamma =510^2`$. This increase in the frequency is based on the increasing number of dynamic phase defects which grow very fast and accelerate the corresponding phase transition. If the number or density of defects becomes too large the oscillating behavior breaks down. The shape of the curve in fig. 2 in the interval $`\gamma [10^5,10^1]`$ is in very good agreement with the one obtained in the study of a general autonomous system by Haken et al..
The underlying mechanism of this internal stochastic resonance effect is unexpectedly simple. During the growth and decline of the individual surface phases only a few very small residual phase islands remain in domains where the local synchronization leads to large amplitudes in the phase oscillations. These residual islands are spatially separated at a mean distance $`R_\mathrm{r}`$. Adsorbate diffusion is well known to synchronize individual surface domains within the socalled synchronization lentgh $`\xi \sqrt{DT}`$, where $`D`$ is the diffusion rate and $`T`$ is the time for one oscillation period. Without nucleation $`\xi <R_\mathrm{r}`$ holds and only locally synchronized oscillations exist. In addition, the residual islands can only combine into a homogeneous phase if the second condition $`R_\mathrm{r}VT`$ for the phase border propagation is fulfilled. But this condition is also violated because $`R_\mathrm{r}>VT`$ holds. Nucleation and subsequent growth now leads to new small phase islands (see fig. 3). The mean distance between the individual phase islands decreases to $`R_\mathrm{n}`$, for which $`R_\mathrm{n}<\xi <R_\mathrm{r}`$ and $`R_\mathrm{n}VT`$ holds for proper nucleation rates. Therefore the separated phase islands can now be connected via island growth and synchronized via CO diffusion. This results in macroscopic synchronized oscillations.
The existence of the surface phase nucleation is well known but has not been investigated experimentally yet, in contrast to the initial growth of small surface phase domains. The nucleation rate should depend on the temperature $`\gamma =\gamma (T)`$, but it is almost impossible to achieve an isolated variation of the nucleation rate under experimental conditions because all other parameter such as CO desorption and CO diffusion also strongly depend on the temperature. It might thus be very difficult to experimentally verify the mechanism behind the SR phenomenon in our model but there should be other systems where it is feasible.
The nucleation of dynamic surface defects as an internal process generates globally synchronized oscillations in our CO+O<sub>2</sub>/Pt(110) model reaction system via SR. Moreover, because noise is always present in real systems, this type of internal SR should be a very general phenomenon and it may be the reason for cooperative phenomena and internal synchronization via noise processes in many physical, chemical, and biological systems where is has not been investigated experimentally or even suspected to be present to date. This especially holds for systems which exhibit inherent oscillations which are synchronized on macroscopic length scales. In this case often noise effects are supposed to be negligible, but as shown above they can also be the origin of those cooperative phenomena. We believe that SR may thus be an internal regulation mechanism of extreme adaptability. This conclusion is drawn to search for internal SR in a variety of experiments, because to date experiments designed to study the role of internal noise have been inconclusive.
Financial support from the Deutsche Forschungsgemeinschaft, the Volkswagen–Stiftung and the Fonds der chemischen Industrie is gratefully acknowledged.
|
no-problem/9906/hep-ex9906028.html
|
ar5iv
|
text
|
# Future Polarised DIS Fixed Target Experiments
## 1 INTRODUCTION
The spin structure of the nucleon has been investigated in polarised deep inelastic scattering by a series of experiments at CERN, SLAC and DESY . These experiments were initiated by the discovery of the EMC in 1987 that the contribution of the quark helicities to the proton spin is much smaller than expected originally . The new experiments confirmed the original finding of the EMC also for the neutron that the singlet axial vector current matrix element is about 1/2 to 1/3 of the predicted value of 0.6 .
Up to now only the contribution of quark helicities to the nucleon spin was studied. Further insight into the spin structure of the nucleon can be gained by investigating the gluonic contribution and the contribution of angular momentum. In addition a measurement of the decomposition of the quark contribution into the different quark flavours will yield a deeper understanding of the nucleon spin puzzle.
There is a whole list of topics which need more detailed studies:
* The flavour decomposition of the polarised quark distributions can be extracted via the measurement of semi-inclusive asymmetries.
* The gluon polarisation can be measured with the help of open charm production or high $`p_\mathrm{T}`$ hadron pairs.
* Polarised fragmentation functions and spin transfer can be studied by measuring the $`\mathrm{\Lambda }`$ polarisation in the current and target fragmentation region.
* A new topis is the study of transversity in scattering on a transversely polarised target using the Collins effect. First signal were presented by SMC and HERMES during this workshop .
* The total angular momentum of quarks might be investigated via deeply virtual Compton scattering.
Further topics on the list refer to the measurement of off-forward parton distributions, vector meson production etc.
The common feature of all these new measurements is the need to detect one or several hadrons in addition to the scattered lepton.
## 2 MEASUREMENT OF $`\mathrm{\Delta }G`$
Up to now the gluon polarisation has been investigated by indirect methods using NLO analyses of structure function data . They indicate that integral $`\mathrm{\Delta }G`$ is positive and of the order of 1 at $`Q^2=1`$ GeV<sup>2</sup> with fairly large errors while the functional form of $`\mathrm{\Delta }G(x)`$ is completely unknown although there are some prejudices that $`\mathrm{\Delta }G(x)`$ is largest around $`x0.1`$.
The cleanest channel for a direct measurement of $`\mathrm{\Delta }G`$ in polarised DIS is the photon gluon fusion process (PGF) as it depends in leading order on the gluon distribution (see fig. 1). In this context several methods are being discussed to extract $`\mathrm{\Delta }G/G`$:
* Open charm procduction:
$$\gamma \mathrm{N}\mathrm{c}\overline{\mathrm{c}}\mathrm{X}\mathrm{D}^0\mathrm{X}$$
PGF is signalled by the dectection of charmed particles in the final state, especially by $`\mathrm{D}^0`$ and $`\mathrm{\Lambda }_\mathrm{c}`$ (close to threshold). The $`\mathrm{D}^0`$’s are reconstructed via their decay to e.g. $`\mathrm{K}\pi `$ or $`\mu \mathrm{K}\pi `$. This process should yield a clear signal directly related to $`\mathrm{\Delta }G`$. Here, the hard scale is given by $`2\mathrm{m}_\mathrm{c}`$.
The cross section for open charm production $`\sigma ^{\gamma \mathrm{N}\mathrm{c}\overline{\mathrm{c}}}`$ is large for quasi real photons and $`\mathrm{\Delta }\sigma ^{\gamma \mathrm{N}\mathrm{c}\overline{\mathrm{c}}}/\sigma ^{\gamma \mathrm{N}\mathrm{c}\overline{\mathrm{c}}}`$ is largest for photon energies between 30 and 80 GeV. The asymmtry $`a_{\mathrm{LL}}`$ for the hard subprocess $`\gamma \mathrm{g}\mathrm{c}\overline{\mathrm{c}}`$ is about 1 at threshold ($`2\mathrm{m}_\mathrm{c}`$) while $`a_{\mathrm{LL}}=1`$ for large energies. In this case a positive gluon polarisation will leaad to a negative photon nucleon asymmetry:
$$A_{\gamma \mathrm{N}}^{\mathrm{c}\overline{\mathrm{c}}}=a_{\mathrm{LL}}\mathrm{\Delta }G/G.$$
* Hidden charm production:
$$\gamma \mathrm{N}\mathrm{c}\overline{\mathrm{c}}\mathrm{X}\mathrm{J}/\psi \mathrm{X}$$
Here the photon gluon process is signalled by the production of a $`\mathrm{J}/\psi `$ which is identified by its decay into a muon pair. While this is a very clean experimental signal the cross section is reduced considerably compared to open charm production. Moreover the relation of the signal to $`\mathrm{\Delta }G/G`$ has to be done via the colour singlet or the colour octet model, a question which is not yet settled in unpolarised DIS.
* High $`p_\mathrm{T}`$ hadron pairs:
$$\gamma \mathrm{N}\mathrm{q}\overline{\mathrm{q}}\mathrm{X}2\mathrm{j}\mathrm{e}\mathrm{t}\mathrm{s}\mathrm{X}$$
This third method tries to select all PGF events not only the $`\mathrm{c}\overline{\mathrm{c}}`$ production. The transverse momenta of the produced jets give the necessary hard scale.
At the moderate energies of fixed target experiments jets are not available but one can use fast hadrons instead . Selecting oppositely charged high $`p_\mathrm{T}`$ hadron pairs will enhance PGF events, but there is a considerable background especially from the QCD Compton process. Thus the measured asymmetry is given by
$$A_{\mathrm{L}L}^{\mathrm{H}H}a_{\mathrm{LL}}^{\mathrm{PGF}}\frac{\mathrm{\Delta }G}{G}\frac{\sigma ^{\mathrm{P}GF}}{\sigma ^{\mathrm{t}ot}}+a_{\mathrm{LL}}^{\mathrm{COM}}\frac{\mathrm{\Delta }u}{u}\frac{\sigma ^{\mathrm{C}OM}}{\sigma ^{\mathrm{t}ot}}.$$
The hard asymmtries $`a_{\mathrm{LL}}^{\mathrm{PGF}}`$ and $`a_{\mathrm{LL}}^{\mathrm{COM}}`$ are large in the $`Q^2`$ range of the fixed target experiments and have opposite signs.
In addition one has to investigate the contribution due to resolved photons. Thus, the results from this method will be dominated by systematic effects due to large background subtractions. A first attempt to use the method was presented by the HERMES collaboration during this workshop .
## 3 FACILITIES
Up to now the experiments concentrated on inclusive measurements of the spin structure functions $`g_1`$ and $`g_2`$. This era comes to an end with the present E155X experiment at SLAC where data are being taken for a precise measurement of $`g_2^{\mathrm{p},\mathrm{d}}`$. This effort will be continued at Jefferson Lab , where a high statistics measurement of the large $`x`$ behaviour of $`g_1`$ and $`g_2`$ is being planned using a polarised $`{}_{}{}^{3}\mathrm{He}`$ target.
Several facilities will be available during the next years to measure semi-inclusive properties in polarised DIS:
* HERMES at DESY, which is in full swing measuring semi-inclusive asymmetries, has started a large upgrade program to attack several of the questions listed above.
* The COMPASS experiment is being setup at the CERN 100–200 GeV muon beam and will start data taking in 2000 focussing in the beginning on a measurement of $`\mathrm{\Delta }G`$.
* At MAMI in Mainz, ELSA in Bonn and Jefferson Lab measurements of the GDH sumrule and the generalized GDH sumrule will be continued.
In addition there are plans for future high luminosity maschines where a continuation and extension of the present spin program will be feasible, e.g. the ELFE proposal at CERN or DESY to study polarised DIS and the APPOLON at ELFE and the SLAC real photon beam proposal to investigate photoproduction.
In the following I will concentrate on the HERMES upgrade and the COMPASS experiment.
## 4 HERMES UPGRADE PROGRAM
The main aims of the upgrade program are
* Particle identification in the full hadron momentum range, i.e. pion, kaon and proton separation,
* Enlarged muon acceptance and improved muon identification,
* Electron acceptance at very small scattering angles,
* Enlarged hadron acceptance covering also negative $`x_\mathrm{F}`$.
The first item was attacked with the installation of a dual RICH . Due to the combination of an aerogel with a gas radiator $`\pi /\mathrm{K}/\mathrm{p}`$ separation is achieved in the full momentum range up to 20 GeV. To yield high precision measurements of the Cerenkov rings the RICH is being red by an array of photomultiplier.
The RICH is already installed and was succesfully operated in 1998. Currently particle identification is being implemented in the analysis chain.
The new muon filter system has been installed during the last shutdown . It consists out of an iron absorber at the end of the spectrometer followed by a scintillator hodoscope.
The enlarged muon acceptance (for scattering angles above 170 mrad) will be made available by using tracks passing part of the magnet yoke. During the shutdown in May 1999 additional scintillators will be installed covering the region between 140 and 270 mrad behind the spectrometer magnet.
With a forward quadrupole spectrometer the electron acceptance will be extended to smaller scattering angles. This is especially important for quasireal photoproduction events. Up to now only 10% of the scattered electrons were detected by the luminosity monitor. The new spectrometer will add another 16%.
For this, quadrupoles with larger apertures were installed in the last long shutdown. The electrons will be measured using a small vertical driftchamber installed inside the quadrupole. After the successfull test of a prototype chamber the system will be installed in May 1999.
The next topic is the enlarged hadron acceptance . For this purpose a wheel of silicon detectors is being constructed to be positioned right after the target cell. This will enlarge the acceptance to $`x_\mathrm{F}<0`$. Monte Carlo simulations showed that this improvement is especially important for measuring the $`\mathrm{\Lambda }`$ decay products. The installation of the system will start in May 1999.
The last project on the list is the recoil detector. It will consist out of a layer of double sided silicon detectors positioned below the target cell. It will be used to measure recoil particles from the target to identify diffractive events and measure tagged structure functions. This year a prototype detector was tested successfully. The installation of the full system is forseen for 2001 provided funding is available.
This upgrade will allow
* To study the flavour decomposition in more detail, e.g. measure strange quark polarisation $`\mathrm{\Delta }s(x)/s(x)`$.
* A measurement of the gluon polarisation with several methods. Using open charm production and the $`\mathrm{D}_0`$ decay into $`\pi \mathrm{K}`$ and $`\mu \pi \mathrm{K}`$ a precision of $`\mathrm{\Delta }G/G`$ of 0.44 rsp. 0.40 can be reached with a luminosity of 80 pb<sup>-1</sup>. The measurement of hidden charm yields $`\delta (\mathrm{\Delta }G/G)`$ of about 0.69. In addition the enlarged hadron acceptance will improve the measurement via high $`p_\mathrm{T}`$ hadron pairs.
* The measurement of the $`\mathrm{\Lambda }`$ polarisation in the current fragmentation region will yield a significant measurement of the polarised fragmentation function $`\mathrm{\Delta }D_\mathrm{u}^\mathrm{\Lambda }/D_\mathrm{u}^\mathrm{\Lambda }`$, while the spin correlation for strange quarks will be studied in the target fragmentation region.
* After 2000, measurements with a transversely polarised target will allow studies of azimuthal spin asymmetries to extract tranversity distributions.
## 5 THE COMPASS EXPERIMENT
Currently, the COMPASS experiment is being setup at the CERN M2 muon beam line to study polarised deep inelastic muon nucleon scattering. In addition a hadron program is planned e.g. to study charmed baryons and search for glue balls .
Compared to the previous muon experiment from SMC COMPASS will have an increased muon beam intensity of $`210^8\mu `$/14.4 s with 100–200 GeV and 80% polarisation. Together with two oppositely polarised target cells of 60 cm length filled with <sup>6</sup>LiD or NH<sub>3</sub> a luminosity of about 2 fb<sup>-1</sup> per year can be reached.
The spectrometer is optimized for large hadron acceptance and particle identification (see fig. 2). To achieve acceptance up to $`\pm 180`$ mrad a new target solenoid is being constructed with a large opening minimizing the multiple scattering for hadron tracks at large angles.
The spectrometer itself consists out of two stages. Each stage has a dipole spectrometer magnet surrounded by tracking chambers followed by a RICH detector, an electromagnetic and a hadronic calorimeter and a muon filter system.
A special feature of the muon beam is the 10% halo of muon tracks surrounding the muon beam up to a radius of 0.5 to 1 m. In addition, scattered muons have to be detected very close and in the muon beam for the measurement of quasireal photoproduction. Thus the tracking system has to be split into three regions.
The beam and scattered muons with very small angles will be measured by scintillating fiber hodoscopes. Small angle tracking will be performed with micromega chambers (stage 1) and GEM detectors (stage 2) while drift (stage 1), proportional (stage 2) and straw chambers will be used for the large angle tracking.
For the first year of data taking the detector will not be complete, especially the RICH in stage 2, the electromagnetic calorimeter electronics and part of the large angle tracking will be missing. Thus the measurements will concentrate on $`\mathrm{\Delta }G/G`$ via quasireal photoproduction.
Using the above mentioned luminosity 82k charm events are expected with 1.2 $`\mathrm{D}^0`$ per charm event. The $`\mathrm{D}^0`$’s will be reconstructed via their $`\pi \mathrm{K}`$ decays. Due to MCS in the target it is not possible to reconstruct the decay vertex, thus the large combinatorial background has to be reduced by strict cuts on the $`\mathrm{D}^0`$ kinematics. This should result in 900 charm events/day with a S:B of 1:3.9. Within 1.5 y with a <sup>6</sup>LiD target a precision of $`\delta A_{\gamma \mathrm{N}}^{\mathrm{c}\overline{\mathrm{c}}}0.05`$ could be reached translating to $`\delta (\mathrm{\Delta }G/G)0.14`$ at $`x_\mathrm{g}=0.14`$. Using e.g. additional decay channels or $`\mathrm{D}^{}`$ tagging will improve the results.
Alternatively high $`p_\mathrm{T}`$ hadron pairs will be used to extract the gluon polarisation . To reduce the background due to the QCD Compton process a series of cuts (opposite charge, high $`p_\mathrm{T}>1`$ GeV, opposite azimuth, $`p_\mathrm{T}`$ balance, positive $`x_\mathrm{F}`$) have to be applied resulting in a good signal to background ratio of about 1:1. Due to the suppression of strangeness in the fragmentation process K<sup>+</sup>K<sup>-</sup> pairs yield an even cleaner signal.
The gluon polarisation can be studied in the range $`0.04<x_\mathrm{g}<0.2`$ for 200 GeV muon energy. With 1 y of data taking a precision of $`\delta (\mathrm{\Delta }G/G)0.05`$ should be achievable. The error of the gluon polarisation will then be dominated by systematic effects for this analysis.
With the described spectrometer, especially with the full setup, all topics discussed in the introduction can be investigated like the flavour decomposition of the quark helicity distributions and polarised fragmentation functions. With a tranversely polarised target azimuthal asymmetries will be measured to extract transversity distributions. The possibility to study deeply virtual Compton scattering is currently being investigated.
## 6 SUMMARY
During the next years a rich experimental program is going on in fixed target polarised DIS. Experiments at DESY and CERN will do detailed studies of semi-inclusive processes to unravel more of the details of the nucleon spin structure.
|
no-problem/9906/math9906011.html
|
ar5iv
|
text
|
# Some Concepts in List Coloring
## 1 Introduction
We consider finite, undirected simple graphs. For necessary definitions and notations we refer the reader to standard texts such as .
By a $`k`$–list assignment $`L`$ to a graph $`G`$ we mean a map which assigns to each vertex $`v`$ of $`G`$ a set $`L(v)`$ of size $`k`$. A list coloring for $`G`$ from $`L`$, or an $`L`$–coloring for short, is a proper coloring $`c`$, in which for each vertex $`v`$, $`c(v)`$ is chosen from $`L(v)`$. A graph $`G`$ is called $`k`$–choosable if it has a list coloring from any $`k`$–list assignment to it. The minimum number $`k`$ for which $`G`$ is $`k`$–choosable is called the list chromatic number of $`G`$ and is denoted by $`\chi _{\mathrm{}}(G)`$. In the following theorem all $`2`$–choosable graphs are characterized. Before we state the theorem it should be noted that the core of a graph is a subgraph which is obtained by repeatedly deleting a vertex of degree 1, until no vertex of degree 1 remains.
###### Theorem A
. A connected graph is $`2`$–choosable, if and only if its core is either a single vertex, an even cycle, or $`\theta _{2,2,2r}`$, for some $`r1`$.
A graph $`G`$ is called uniquely $`k`$–list colorable, or U$`k`$LC for short, if it admits a $`k`$–list assignment $`L`$ such that $`G`$ has a unique $`L`$–coloring. This concept was introduced by Dinitz and Martin and independently by Mahdian and Mahmoodian ( and ). A characterization of uniquely $`2`$–list colorable graphs follows.
###### Theorem B
. A graph $`G`$ is not U$`2`$LC if and only if each of its blocks is either a cycle, a complete graph, or a complete bipartite graph.
It is easy to see that for each graph $`G`$ there exists a number $`k`$ such that $`G`$ is not U$`k`$LC. The minimum $`k`$ with this property is called the m–number of $`G`$ and is denoted by $`\text{m}(G)`$. It is shown in that every planar graph has m–number at most $`5`$, and it is asked about the existence of planar graphs with m–number equal to $`5`$. We study uniquely list colorable graphs in Section 2, where we prove that every triangle–free uniquely $`(k+1)`$–colorable graph is uniquely $`k`$–list colorable. We also show that every planar graph has m–number at most $`4`$, so the answer to that question in is negative.
In Section 3 we introduce list critical graphs and characterize $`3`$–list critical graphs. Finally we pose a conjecture about list critical graphs which is shown to be equivalent to the list coloring conjecture.
## 2 Uniquely list colorable graphs
In one can find several examples of U$`k`$LC graphs, for some arbitrary positive integer $`k`$. In the following lemma we also introduce a class of U$`k`$LC graphs. In this way we relate uniquely list colorable graphs to uniquely colorable graphs.
###### Lemma 1
. Let $`G`$ be a uniquely colorable graph with chromatic number $`k+1`$, and $`c`$ be its unique $`(k+1)`$–coloring with color classes $`C_1,\mathrm{},C_{k+1}`$. If for each $`ik+1`$, $`|C_i|i1`$, then $`G`$ is a uniquely $`k`$–list colorable graph.
###### Proof.
We proceed by induction on $`k`$ and prove that there exists a $`k`$–list assignment to such graph $`G`$ using exactly $`k+1`$ colors, which induces a unique list coloring. For $`k=1`$ the result obviously holds. Let $`G`$ be a uniquely $`(k+1)`$–colorable graph as in the statement and $`k2`$. By induction $`GC_{k+1}`$ admits a $`(k1)`$–list assignment $`L^{}`$ which induces a unique list coloring and uses colors $`1,\mathrm{},k`$. For each $`vV(G)C_{k+1}`$, assign the list $`L(v)=L^{}(v)\{k+1\}`$ to $`v`$, and since $`|C_{k+1}|k`$, it is possible to assign some lists to $`C_{k+1}`$ such that $`_{vC_{k+1}}L(v)=\{k+1\}`$. Now it is easy to see that $`L`$ is the desired list assignment. $`\mathrm{}`$
It is shown in that for every $`k3`$, in a triangle–free uniquely $`k`$–colorable graph, each color class has at least $`k+1`$ vertices. Using this result, we obtain the following theorem.
###### Theorem 1
. Every triangle–free uniquely $`(k+1)`$–colorable graph is uniquely $`k`$–list colorable.
On the other hand in it is shown that for each $`k2`$, there exists a uniquely $`k`$–colorable graph with arbitrary large girth. So by theorem above, for each $`k`$, there exists a U$`k`$LC graph with arbitrary large girth.
We need here a definition which is a generalization of the concept of a U$`k`$LC graph.
###### Definition 1
. Let $`G`$ be a graph and $`f`$ a be function from $`V(G)`$ to $``$. An $`f`$–list assignment $`L`$ to $`G`$ is a list assignment in which $`|L(v)|=f(v)`$ for each vertex $`v`$. The graph $`G`$ is called to be uniquely $`f`$–list colorable, or U$`f`$LC for short, if there exists an $`f`$–list assignment $`L`$ for it such that $`G`$ has a unique $`L`$–coloring.
By definition above, if $`G`$ is a U$`f`$LC graph, where $`f(v)=k`$ for each vertex $`v`$ of $`G`$, then $`G`$ in fact is a U$`k`$LC graph. To prove the next theorem, we need a relation which is proved in Truszczyński and states that if $`G`$ is a uniquely $`k`$–colorable graph, then $`e(G)(k1)n(G)\left(\genfrac{}{}{0pt}{}{k}{2}\right).`$
###### Theorem 2
. If $`G`$ is a U$`f`$LC graph, then
$$\underset{vV(G)}{}f(v)n(G)+e(G).$$
###### Proof.
Suppose that $`L`$ is an $`f`$–list assignment to $`G`$ using colors $`1,2,\mathrm{},t`$, such that $`G`$ has a unique $`L`$–coloring. We construct a uniquely $`t`$–colorable graph $`G^{}`$ as follows. Let $`V(G)=\{v_1,\mathrm{},v_n\}`$ and $`K_t`$ is a complete graph on the vertex set $`\{w_1,\mathrm{},w_t\}`$. Now for $`G^{}`$ consider the union of $`G`$ and $`K_t`$ and add edges $`v_iw_j`$ where $`1in`$, $`1jt`$, and $`jL(v_i)`$.
Consider a $`t`$–coloring $`c`$ of $`G^{}`$. Without loss of generality we can assume that $`c(w_i)=i`$ for each $`1it`$. Since $`G`$ has a unique $`L`$–coloring, by construction of $`G^{}`$, $`c`$ is the only $`t`$–coloring of $`G^{}`$. So $`G^{}`$ is a uniquely $`t`$–colorable graph. On the other hand $`G^{}`$ has $`n(G)+t`$ vertices, and $`e(G)+\left(\genfrac{}{}{0pt}{}{t}{2}\right)+_{vV(G)}(tf(v))`$ edges. Therefore as mentioned above, we have
$$e(G)+\left(\genfrac{}{}{0pt}{}{t}{2}\right)+\underset{vV(G)}{}(tf(v))(n(G)+t)(t1)\left(\genfrac{}{}{0pt}{}{t}{2}\right)$$
and after simplification we obtain the result. $`\mathrm{}`$
A natural question which arises here is that whether or not equality holds in Theorem 2? In the following proposition we give a positive answer to this question.
###### Proposition 1
. For every graph $`G`$, there exists $`f:V(G)`$ such that $`G`$ is U$`f`$LC and $`_{vV(G)}f(v)=n(G)+e(G)`$.
###### Proof.
We proceed by induction on the number of vertices of $`G`$. For $`n(G)=1`$ the statement is obvious. Consider a graph $`G`$ with $`n(G)2`$ and a vertex $`v`$ of $`G`$. By induction there exists $`f^{}:V(Gv)`$ and an $`f^{}`$–list assignment $`L^{}`$ to $`Gv`$ such that $`Gv`$ has a unique $`L^{}`$–coloring, and $`_{wV(Gv)}f(w)=n(Gv)+e(Gv)`$. Consider a color $`a`$ which is not used by $`L^{}`$, and define a list assignment $`L`$ to $`G`$ as follows.
$$L(w)=\{\begin{array}{cc}a\hfill & \mathrm{for}w=v\hfill \\ L^{}(w)\{a\}\hfill & \mathrm{for}wN(v)\hfill \\ L^{}(w)\hfill & \mathrm{otherwise}.\hfill \end{array}$$
It is easy to verify that $`G`$ has a unique $`L`$–coloring and that we have $`_{vV(G)}|L(v)|=n(G)+e(G)`$. $`\mathrm{}`$
Although the proposition above shows that in Theorem 2 equality may hold, but it seems that if $`f(v)=k`$ for each vertex $`v`$, equality does not hold and we have $`e(G)>(k1)n(G)`$.
By definition every graph $`G`$ for $`k=\text{m}(G)1`$ is U$`k`$LCṠo by Theorem 2 we have the following.
###### Theorem 3
. For a graph $`G`$ let $`\overline{d}(G)`$ denote the average degree of $`G`$, i.e. $`\overline{d}(G)=2e(G)/n(G)`$. Then
$$\text{m}(G)\frac{\overline{d}(G)}{2}+2.$$
For example suppose that $`G`$ is a bipartite graph. We have $`\overline{d}(G)n(G)/2`$ so Theorem 3 implies $`\text{m}(G)n(G)/4+2`$. This bound can be improved to a logarithmic bound as we will show in Theorem 4, but first we need a lemma.
Let $`L`$ be a $`k`$–list assignment to a graph $`G`$ such that $`G`$ has a unique $`L`$–coloring $`c`$. For each vertex $`v`$ of $`G`$, all the elements of $`L(v)\{c(v)\}`$ must appear in $`N(v)`$, so if we denote by $`c_N(v)`$ the set of colors appearing in $`N(v)`$, then $`|c_N(v)|k1`$. In the following lemma we state a stronger result.
###### Lemma 2
. Suppose that $`G`$ is a U$`k`$LC graph, and $`L`$ is a $`k`$–list assignment to $`G`$ such that $`G`$ has a unique $`L`$–coloring $`c`$ with color classes $`C_1,\mathrm{},C_t`$ such that $`c(C_i)=\{i\}`$. There exist at least $`k1`$ classes containing a vertex $`v`$ with $`|c_N(v)|k`$.
###### Proof.
Without loss of generality suppose that for $`\mathrm{}k1`$, $`C_{\mathrm{}}`$ contains no vertex $`v`$ with $`|c_N(v)|k`$. Assume that $`uC_{k1}`$, $`i=c(v_0)`$, and $`jL(v_0)\{1,\mathrm{},k1\}`$. Suppose that $`G_{ij}`$ is the subgraph of $`G`$ induced on $`C_iC_j`$. Since for each vertex $`v`$ of the component of $`G_{ij}`$ containing $`v_0`$ we have $`|c_N(v)|=k1`$, it is implied that $`i,jL(v)`$. So we can interchange the colors $`i`$ and $`j`$ in this component to obtain a new $`L`$–coloring for $`G`$. This contradiction completes the proof. $`\mathrm{}`$
It is shown in that every non–$`k`$–choosable bipartite graph has more than $`2^{k1}`$ vertices. So by applying Lemma 2, we deduce the following theorem.
###### Theorem 4
. Let $`G`$ be a bipartite graph. Then $`\text{m}(G)2+\mathrm{log}_2n(G).`$
###### Proof.
Suppose that $`L`$ is a $`k`$–list assignment to $`G`$ such that $`G`$ has a unique $`L`$–coloring $`c`$. By Lemma 2, $`G`$ has a vertex $`v_0`$, such that there are at least $`k`$ colors appeared at $`N(v_0)`$ in $`c`$. Let $`G^{}`$ be the graph obtained from $`G`$, by duplicating $`v_0`$, i.e. adding a new vertex $`w`$ to $`G`$ and joining it to $`N(v_0)`$. Now assign to $`w`$ a list containing $`k`$ of the colors appeared at $`N(v_0)`$ in $`c`$, and the list $`L(v)`$ to each other vertex $`v`$ of $`G^{}`$. It is clear that $`G^{}`$ is a bipartite graph and it has no coloring from these lists, so it is not $`k`$–choosable. Hence $`n(G^{})>2^{k1}`$ vertices. This implies that $`n(G)2^{k1}`$, and so $`k1+\mathrm{log}_2n(G)`$. Now we obtain the desired relation by setting $`k=\text{m}(G)1`$. $`\mathrm{}`$
In the remainder of this section we state some consequences of Theorems 2 and 3.
It is well known that a planar graph with $`n`$ vertices has at most $`3n6`$ edges. So the following theorem is an immediate consequence of Theorem 3.
###### Theorem 5
. For every planar graph $`G`$ we have $`\text{m}(G)4`$.
By Lemma 1 the planar graph shown in Figure 1 is a U$`3`$LC graph for which the inequalities in Theorem 3 and Theorem 5 turn to be equalities.
Furthermore we know that a triangle–free planar graph $`G`$, has at most $`2n(G)4`$ edges. So Theorem 3 implies that each triangle–free planar graph $`G`$ has m–number at most $`3`$. In the following proposition a stronger result is obtained.
###### Proposition 2
. If a plane graph has at most $`7`$ triangular faces, then $`\text{m}(G)3`$.
###### Proof.
Consider a U$`3`$LC plane graph $`G`$ with $`n`$ vertices, $`e`$ edges, $`f`$ faces, and $`t`$ triangular faces. We have $`2e4(ft)+3t=4ft`$, and by Euler formula $`f=2n+e`$, so $`t84n+2e`$. On the other hand Theorem 2 implies that $`e2n`$. So $`t8`$, as desired. $`\mathrm{}`$
The following conjecture is about the structure of U$`3`$LC planar graphs which is motivated by the proposition above.
###### Conjecture 1
. Every U$`3`$LC planar graph has $`K_4`$ as a subgraph.
For another application of Theorem 2, we study line and total versions of uniquely list coloring.
A graph $`G`$ is called to be uniquely $`k`$–list edge colorable, if $`L(G)`$ is a uniquely $`k`$–list colorable graph. The edge m–number of $`G`$ is defined to be $`\text{m}(L(G))`$, and is denoted by $`\text{m}^{}(G)`$. It is straightforward to see that for each graph $`G`$, $`\overline{d}(L(G))\mathrm{\Delta }(L(G))2\mathrm{\Delta }(G)2`$. So using Theorem 3 we deduce the following.
###### Theorem 6
. For every graph $`G`$, we have $`\text{m}^{}(G)\mathrm{\Delta }(G)+1`$ and if $`\text{m}^{}(G)=\mathrm{\Delta }(G)+1`$ then $`G`$ is a regular graph.
Note that in Theorem 6 it is shown that if $`G`$ is not a regular graph, then $`\text{m}(G)\mathrm{\Delta }(G)`$. So in this case $`\text{m}(G)\chi (G)`$.
## 3 List critical graphs
In this section we introduce a concept of list critical graphs and we state some results concerning it.
###### Definition 2
. A graph $`G`$ is called $`\chi _{\mathrm{}}`$–critical if for each proper subgraph $`H`$ of it we have $`\chi _{\mathrm{}}(H)<\chi _{\mathrm{}}(G)`$.
We sometimes refer to a $`\chi _{\mathrm{}}`$–critical graph $`G`$ as a $`k`$–list critical graph, where $`k=\chi _{\mathrm{}}(G)`$. It can easily be verified that the only connected $`2`$–list critical graph is $`K_2`$, odd cycles are $`3`$–list critical, and the complete graph $`K_k`$ is $`k`$–list critical.
Obviously every graph $`G`$ contains a $`\chi _{\mathrm{}}`$–critical subgraph $`H`$ such that $`\chi _{\mathrm{}}(H)=\chi _{\mathrm{}}(G)`$, and by an argument similar to critical graphs, $`\delta (G)\chi _{\mathrm{}}(G)1`$. On the other hand there exists some differences between critical graphs and list critical graphs. For example it is well known that every critical graph is $`2`$–connected. In Figure 2 we have given an example of a $`3`$–list critical graph which is not $`2`$–connected.
In the next theorem $`3`$–list critical graphs are characterized.
###### Theorem 7
. A graph is $`3`$–list critical if and only if it is either an odd cycle, two even cycles with a path joined them, $`\theta _{r,s,t}`$ where $`r`$, $`s`$, $`t`$ have the same parity, and at most one of them is $`2`$, or $`\theta _{2,2,2,2r}`$ where $`r1`$.
###### Proof.
By use of Theorem A, it is easy to see that all the graphs listed in the statement are $`3`$–list critical.
For the converse suppose that $`G`$ is a $`3`$–list critical graph. If $`G`$ is $`2`$–connected, by a theorem of Whitney $`G`$ has an ear decomposition $`K_2P^1\mathrm{}P^q`$. If $`q4`$, deleting an edge of $`P^q`$, yields a non–$`2`$–choosable graph, which contradicts the $`3`$–list criticality of $`G`$. So $`q3`$ and we consider the following three cases.
* If $`q=1`$, $`G`$ is a cycle, and so it is an odd cycle.
* If $`q=2`$, $`G=\theta _{r,s,t}`$. In this case by deleting each edge of $`G`$, we obtain a graph whose core is a cycle, and since this cycle must be even, the numbers $`r`$, $`s`$, and $`t`$ have the same parity. Now if at least two of $`r`$, $`s`$, and $`t`$ are equal to $`2`$, we have $`\chi _{\mathrm{}}(G)=2`$, a contradiction.
* The last case is $`q=3`$. By deleting each edge of $`G`$ we obtain a graph whose core is a $`\theta _{r,s,t}`$, and since this graph must be $`2`$–choosable, we have $`r=s=2`$ and $`t`$ is an even number. Now by case analysis, it is easy to see that $`G=\theta _{2,2,2,2\mathrm{}}`$.
On the other hand if $`G`$ is not $`2`$–connected, we consider two end–blocks $`B_1`$ and $`B_2`$ of $`G`$. Since $`\delta (G)2`$ each of $`B_1`$ and $`B_2`$ has a cycle. So $`G`$ has a subgraph $`H`$ which is composed of two edge–disjoint cycles joined to each other by a path (possibly of length zero). We know that $`\chi _{\mathrm{}}(H)=3`$, and so by $`\chi _{\mathrm{}}`$–criticality of $`G`$, $`G`$ has no edge outside $`H`$, i.e. $`G=H`$. Hence $`G`$ satisfies the statement. $`\mathrm{}`$
Suppose that $`G`$ is a $`k`$–list critical graph, and $`L`$ is a $`k`$–list assignment to $`G`$. Consider a vertex $`v`$ in $`G`$ and a color $`aL(v)`$. Assign to each vertex $`u`$ in $`Gv`$ the list $`L(u)\{a\}`$. Since $`Gv`$ is $`(k1)`$–choosable, it has a coloring from the assigned lists, and one can extend this coloring to an $`L`$–coloring of $`G`$ by assigning the color $`a`$ to $`v`$. So there exists an $`L`$–coloring for $`G`$ in which $`v`$ takes $`a`$.
As mentioned in the previous paragraph, every $`k`$–list critical graph has at least $`k`$ colorings from each $`k`$–list assignment so every $`k`$–list critical graph has m–number at most $`k`$.
A graph $`G`$ is called to be edge $`k`$–choosable, if the graph $`L(G)`$ is $`k`$–choosable, and the list chromatic index of $`G`$ written $`\chi _{\mathrm{}}^{}(G)`$ is defined to be $`\chi _{\mathrm{}}(L(G))`$. As in the case of defining $`\chi ^{}`$–critical graphs, one can define a $`\chi _{\mathrm{}}^{}`$–critical graph $`G`$ to be a graph in which for each proper subgraph $`H`$, $`\chi _{\mathrm{}}^{}(H)<\chi _{\mathrm{}}^{}(G)`$. We recall here the well known List Coloring Conjecture (LCC), which first appeared in print in .
###### Conjecture
. Every graph $`G`$ satisfies $`\chi _{\mathrm{}}^{}(G)=\chi ^{}(G)`$.
Suppose that $`G`$ is a counterexample to the LCC with minimum number of edges. So for each edge $`uv`$ of $`G`$ we have $`\chi _{\mathrm{}}^{}(Guv)=\chi ^{}(Guv)`$, and since $`\chi ^{}(Guv)\chi ^{}(G)<\chi _{\mathrm{}}^{}(G)`$, we conclude that $`\chi _{\mathrm{}}^{}(Guv)=\chi _{\mathrm{}}^{}(G)1`$. This means that $`G`$ is a $`\chi _{\mathrm{}}^{}`$–critical graph and therefore $`\chi _{\mathrm{}}^{}`$–critical graphs may be useful to attack the LCC.
In the study of $`\chi _{\mathrm{}}^{}`$–critical graphs we have lead to the following conjecture.
###### Conjecture 2
. Every $`\chi _{\mathrm{}}^{}`$–critical graph is $`\chi ^{}`$–critical.
###### Proposition 3
. The conjecture above is equivalent with the LCC, while its converse is implied by the LCC.
###### Proof.
It is straight forward to check that the list coloring conjecture implies Conjecture 2 and its converse. On the other hand suppose that Conjecture 2 is true, and $`G`$ is a counterexample to the list coloring conjecture with minimum number of edges. As mentioned above $`G`$ is $`\chi _{\mathrm{}}^{}`$–critical, and by Conjecture 2, it is $`\chi ^{}`$–critical. By removing an arbitrary edge $`uv`$ from $`G`$ we obtain a graph for which the list coloring conjecture holds. So $`\chi _{\mathrm{}}^{}(Guv)=\chi ^{}(Guv)`$, and this means that $`\chi _{\mathrm{}}^{}(G)1=\chi ^{}(G)1`$, a contradiction. $`\mathrm{}`$
In it is proved that every bipartite multigraph fulfills the LCC. On the other hand we know that the only bipartite $`\chi ^{}`$–critical graphs are stars. So the following theorem is implied by a similar argument as in the previous paragraph, the only bipartite $`\chi _{\mathrm{}}^{}`$–critical graphs are stars.
## Acknowledgements
The authors wish to thank Professor E. S. Mahmoodian and R. Tusserkani for their helpful comments. They are indebted to the Institute for Studies in Theoretical Physics and Mathematics (IPM), Tehran, Iran for their support.
eslahchi@karun.ipm.ac.ir
ghebleh@karun.ipm.ac.ir
hhaji@karun.ipm.ac.ir
|
no-problem/9906/chao-dyn9906032.html
|
ar5iv
|
text
|
# Experimental vs. Numerical Eigenvalues of a Bunimovich Stadium Billiard – A Comparison
## I Introduction
Quantum manifestations of classical chaos have received much attention in recent years and for the semiclassical quantization of conservative chaotic systems, two-dimensional billiard systems provide a very effective tool . Due to the conserved energy of the ideal particle propagating inside the billiard’s boundaries with specular reflections on the walls, the billiards belong to the class of Hamiltonian systems with the lowest degree of freedom in which chaos can occur and this only depends on the given boundary shape. Such systems are in particular adequate to study the behavior of the particle in the corresponding quantum regime, where spectral properties are completely described by the stationary Schrödinger equation. The spectral fluctuations properties of such systems were investigated both analytically and numerically. It has been found that these properties coincide with those of the ensembles of random matrix theory (RMT) having the proper symmetry if the given system is classically non-integrable. For time-reversal invariant systems, to which group the here investigated billiards belong, the relevant ensemble is the Gaussian orthogonal ensemble (GOE).
In the last decades this subject was dominated by theory and numerical simulations. About seven years ago experimentalists have found effective techniques to simulate quantum billiard problems with the help of macroscopic devices. Due to the equivalence of the stationary Schrödinger equation and the classical Helmholtz equation in two dimensions one is able to model the billiard by a similarly shaped electromagnetic resonator .
Theoretical predictions assume to have always an ideal system with a perfect geometry whereas experiments have been performed with real systems. Real microwave cavities – in particular the here used superconducting ones – are usually not machined by the most accurate technique, e.g. milling from a solid block. They are cut from niobium sheet material which is shaped and afterwards electron beam welded. Finally they are chemically etched to clean their inner surface. Their final inner geometry is not accessible for a direct measurement so that their exact geometric properties are not known. In the case of the investigated Bunimovich stadium billiard , where $`\gamma =a/R=1`$ (see inset of Fig. 1), the following properties are especially crucial: The radius of curvature of the boundary does not change abruptly but smoothly at the transition from the straight $`a`$ to the circular section of the boundary . Another point is that the angle at the corners is not exact 90 degrees.
The paper is about the question to which extent a comparison of experimental data with theoretical predictions for such billiards is meaningful. Therefore we want to compare a numerical simulation (calculation of eigenvalues) for a $`\gamma =1`$ stadium billiard with a measurement of a real superconducting microwave cavity (measurement of eigenvalues) by studying the statistical properties of the two sequences of eigenvalues.
The paper is organized as follows. In Sec. II the experimental set-up and the measurement of the eigenfrequencies are described and in Sec. III the numerical calculations. The comparison of both sets of eigenvalues is shown in Sec. IV by analyzing their spectral fluctuations.
## II Experiment
Experimentally we have investigated a two-dimensional microwave cavity which simulates a two-dimensional quantum billiard. Here we present results based on measurements using a superconducting niobium cavity, having the shape of a quarter Bunimovich stadium billiard. The billiard has been desymmetrized to avoid superpositions of several independent symmetry classes . Its inner dimensions are $`a=R=20`$ cm corresponding to $`\gamma =a/R=1`$, see inset of Fig. 1, and it has a height of $`d=0.7`$ cm, that guarantees a two-dimensionality up to a frequency of $`f_{max}=c/2d=21.4`$ GHz ($`c`$ denotes the speed of light).
As in previous investigations the measurement of the $`\gamma =1`$ stadium billiard has been carried out in a LHe-bath cryostat. This experimental set-up is very stable concerning temperature and pressure fluctuations. The cavity has been put into a copper box which was covered by liquid helium, so that a constant temperature of 4.2 K inside the resonator was guaranteed during the whole measurement. The box has also been evacuated to a pressure of $`10^2`$ mbar to eliminate effects of the dielectric gas inside the cavity. We were able to excite the cavity in the frequency range of $`45`$ MHz $`<f<20`$ GHz in 10 kHz-steps using four capacitively coupling dipole antennas sitting in small holes on the niobium surface, see inset of Fig. 1. These antennas penetrated only up to a maximum of 0.5 mm into the cavity to avoid perturbations of the electromagnetic field inside the resonator. Using one antenna for the excitation and either another one or the same one for the detection of the microwave signal, we are able to measure the transmission as well as the reflection spectra of the resonator using an HP8510B vector network analyzer. In Fig. 1 a typical transmission spectrum of the billiard in the range between 10.5 and 11.5 GHz is shown. The signal is given as the ratio of output power to input power on a logarithmic scale. The measured resonances have quality factors of up to $`Q=f/\mathrm{\Delta }f10^7`$ and signal-to-noise ratios of up to $`S/N60`$ dB which made it easy to separate the resonances from each other and detect also weak ones above the background. As a consequence of using superconducting resonators, all the important characteristics like eigenfrequencies and widths can be extracted with a very high accuracy . A detailed analysis of the raw spectra yielded a total number of 955 resonances up to 20 GHz. To reduce the possibility of missing certain modes with a rather weak electric field vector at the position of the antennas, the measurements were always performed with different combinations of antennas. Thereby the number of missed modes is dramatically reduced below three to five in a typical case of a measurement of about one thousand eigenfrequencies. For the $`\gamma =1`$ stadium billiard investigated here the 955 detected eigenmodes agrees exactly with the expected number calculated from particular geometry of the cavity with the help of Weyl’s formula (see Sec. IV below). The measured sequence of frequencies extracted from the experimental spectra forms the basic set of the statistical investigations in Sec. IV.
## III Numerical Calculations
### A Theory
The problem that we have to solve is to find the eigenvalues of the Dirichlet problem in a billiard, i.e. to solve the following equations:
$$\mathrm{\Delta }\mathrm{\Psi }(\stackrel{}{r})+k^2\mathrm{\Psi }(\stackrel{}{r})=0\text{ for }\stackrel{}{r}𝒟$$
(1)
$$\mathrm{\Psi }(\stackrel{}{r})=0\text{ for }\stackrel{}{r}𝒟$$
(2)
In order to solve these equations we search for a solution in form of an expansion of regular Bessel functions, so that we write
$$\mathrm{\Psi }(\stackrel{}{r})=\mathrm{\Psi }(r,\theta )=\underset{l=L}{\overset{L}{}}a_lJ_l(kr)e^{il\theta }$$
(3)
This expansion is obviously a solution of the first equation, so that we have to solve it under the boundary conditions Eq. (2). On the billiard boundary, which may be parameterized by the curvilinear abscissa $`s`$, the expansion Eq. (3) becomes
$$\mathrm{\Phi }(s)=\mathrm{\Psi }(r(s),\theta (s))=\underset{l=L}{\overset{L}{}}a_lJ_l(kr(s))e^{il\theta (s)}$$
(4)
which is a periodic function of $`s`$. This periodic function may be decomposed in a Fourier series whose coefficients are given by
$$C_n(k)=\frac{1}{2\pi }_0^{}𝑑s\mathrm{\Phi }(s)e^{2in\pi s/}=\underset{l=L}{\overset{L}{}}a_lC_{n,l}(k)$$
(5)
with
$$C_{n,l}(k)=_0^{}𝑑se^{2in\pi s/}J_l(kr(s))e^{il\theta (s)}$$
(6)
To satisfy the boundary condition Eq. (2), one may impose the equivalent conditions that all Fourier coefficients $`C_n(k)`$ are zero, at least for $`LnL`$, so that we are left with the following linear system in the $`a_l`$:
$$\underset{l=L}{\overset{L}{}}a_lC_{n,l}(k)=0,LnL$$
(7)
For this homogeneous system to have a non trivial solution, one has the following set of equations:
$$D(k)=det[C_{n,l}(k)]=0$$
(8)
Thus one is left with the problem of constructing the matrix $`C_{n,l}(k)`$ and finding the zeroes of its determinant $`D(k)`$. Before going further, we would like to point out the advantages of this method as compared to the well known collocation method. In our method, one may vary independently the number of boundary points used to evaluate the integrals $`C_{n,l}(k)`$ and the number of partial waves used in the expansion, whereas this is not so in usual boundary methods. (For other methods we refer the reader e.g. to Refs. ). Furthermore, it appears clearly that if the billiard is close to a circle, the matrix is nearly diagonal so that its determinant is easy to compute numerically, whereas this is not true for plane wave decompositions. Finally, in contrast with the Green’s method, one may always deal with real matrices by decomposing on sine and cosine rather than on exponentials (for time reversal invariant systems). However, one should note that this method does not work for billiards whose boundary consists of several distinct curves (for instance Sinaï billiards).
### B Practical
In practice, the application of the above algorithm depends on the problem one has to solve. For instance, a particular choice of the origin of the coordinates may simplify notably the evaluation of the matrix: For the stadium, a good choice would be the symmetry center of the billiard, so that one may separate easily different symmetry classes. With this choice, the expansion given in Eq. (3) become for odd-odd symmetry:
$$\mathrm{\Psi }(\stackrel{}{r})=\mathrm{\Psi }(r,\theta )=\underset{n=1}{\overset{N}{}}a_nJ_{2n}(kr)sin(2n\theta )$$
(9)
Therewith one tabulates the function $`D(k)`$ in an interval of $`k`$ such that the number of partial waves needed is constant in this interval:
$$k_{min}kk_{max}$$
(10)
with
$$k_{min}R_{max}=L\text{ and }k_{max}R_{max}=L+2$$
(11)
In this equation, $`R_{max}`$ is the greatest distance of the boundary from the origin, in our case $`R_{max}=R+a`$. The number of coefficients $`a_n`$ may then be taken as
$$N=L/2+dN$$
(12)
where $`dN`$ typically ranges from 0 to 3 or 4. Usually, the positions of the eigenvalues depend very little on this parameter.
The integrals
$$C_{n,m}(k)=_0^{}𝑑ssin(n\pi s/)J_{2m}(kr(s))sin(2m\theta (s))$$
(13)
are evaluated using points regularly spaced along the boundary. As soon as their spacing is such that the fastest varying phase in Eq. (13) changes by less than $`\pi `$ in a given step, the evaluation of these integrals is accurate enough to give the position of the zeroes of $`D(k)`$. In other words, the step $`\mathrm{\Delta }s`$ used in evaluating Eq. (13) should be such that the following condition is verified:
$$\frac{N\pi \mathrm{\Delta }s}{}+\frac{2N\mathrm{\Delta }s}{d_{min}}\pi ,$$
(14)
where $`d_{min}`$ is the smallest distance of the boundary from the origin, in our case $`R`$. With these ingredients, the precise computation of the levels of a rather regular billiards is fast, and does not require powerful computers. Typically, for the stadium, the computation of the 1000 first levels takes a few hours run on a personal computer, the dimension of the matrices used being smaller than 100. The precision obtained is much better than $`1/100th`$ of the average spacing, which is enough for our purpose.
## IV Spectral fluctuations
In the following section we want to discuss the statistical properties of the data obtained for the $`\gamma =1`$ stadium billiard as described in Secs. II and III. After a short summary of the general concepts of analyzing spectral fluctuations in Sec. IV A, we present the results for the comparison of the measured and calculated data in Sec. IV B. Finally, in Sec. IV C we compare the influence of a special family of orbits in different Bunimovich stadium billiards.
### A Theoretical background
From the measured resp. numerically simulated eigenvalue sequences (“stick spectrum”) the spectral level density $`\rho (k)=_i\delta (kk_i)`$ is calculated ($`k`$ is the wave number, $`k=2\pi /cf`$) and a staircase function $`N(k)=\rho (k^{})𝑑k^{}`$ is constructed which fluctuates around a smoothly varying part, defined as the average of $`N(k)`$. Usually this smooth part $`N^{smooth}(k)`$ is related to the volume of the classical energy-allowed phase-space. For the 2D billiard at hand with Dirichlet boundary conditions it is given by the Weyl-formula
$$N^{Weyl}(k)\frac{A}{4\pi }k^2\frac{C}{4\pi }k+\mathrm{const}.$$
(15)
where $`A`$ is the area of the billiard and $`C`$ its perimeter. The constant term takes curvature and corner contributions into account. Higher order terms are not relevant for the present analysis. The remaining fluctuating part of the staircase function $`N^{fluc}(k)=N(k)N^{Weyl}(k)`$ oscillates around zero. While Eq. (15) does not contain any information regarding the character of the underlying classical dynamics of the system, the fluctuating part does.
In order to perform a statistical analysis of the given eigenvalue sequence independently from the special size of the billiard, the measured resp. calculated spectrum is first unfolded , i.e. the average spacing between adjacent eigenmodes is normalized to one, using Eq. (15). This proper normalization of the spacings of the eigenmodes then leads to the nearest neighbour spacing distribution $`P(s)`$, from now on called NND, the probability of a certain spacing $`s`$ between two adjacent unfolded eigenfrequencies. To avoid effects arising from the bining of the distribution, we employ here the cumulative spacing distribution $`I(s)=P(s)𝑑s`$.
To uncover correlations between nonadjacent resonances, one has to use a statistical test which is sensitive on larger scales. As an example we use the number variance $`\mathrm{\Sigma }^2`$ originally introduced by Dyson and Mehta for studies of equivalent fluctuations of nuclear spectra . The $`\mathrm{\Sigma }^2`$-statistics describes the average variance of a number of levels $`n(L)`$ in a given interval of length $`L`$, measured in terms of the mean level spacing, around the mean for this interval, which is due to the unfolding equal to $`L`$,
$$\mathrm{\Sigma }^2(L)=\left(n(L)n(L)_L\right)^2_L=n^2(L)_LL^2.$$
(16)
Furthermore, to characterize the degree of chaoticity in the system, the spectra are analyzed in terms of a statistical description introduced in a model of Berry and Robnik which interpolates between the two limiting cases of pure Poissonain and pure GOE behavior for a classical regular or chaotic system, respectively. The model introduces a mixing parameter $`q`$ which is directly related to the relative chaotic fraction of the invariant Liouville measure of the underlying classical phase space in which the motion takes place ($`q=0`$ stands for a regular and $`q=1`$ for a chaotic system).
The chaotic features of a classical billiard system are characterized by the behavior of the orbits of the propagating point-like particle. The quantum mechanical analogue does not know orbits anymore but only eigenstates, i.e. wave functions and corresponding eigenenergies. Thus, the individual and collective features of the eigenstates must reflect the behavior of the classical orbits. The semiclassical theory of Gutzwiller assumes that a chaotic system is fully determined through the complete set of its periodic orbits. The influence of the isolated periodic orbits of a billiard is most instructively displayed in the Fourier transformed (FT) spectrum of the eigenvalue density $`\rho ^{fluc}(k)=dN^{fluc}(k)/dk`$, i.e.
$$\stackrel{~}{\rho }^{fluc}(x)=_{k_{min}}^{k_{max}}e^{ikx}\left[\rho (k)\rho ^{Weyl}(k)\right]𝑑k,$$
(17)
with $`[k_{min},k_{max}]`$ being the wave number interval in which the data are taken.
### B Results
In this section we present the results of the analysis of our data using the techniques described in Sec. IV A. We start with a direct one-to-one comparison of both, experimentally and numerically obtained data sets, and perform first a comparison with the sequences of the unfolded eigenvalues. To obtain the unfolded eigenvalues, we fit the Weyl-formula, Eq. (15), onto our measured resp. numerically simulated spectral staircase $`N(k)`$. Doing this, one obtains the following parameter of Eq. (15) for the measured data: an area $`A_{fit,exp}=(710.52\pm 4.12)\mathrm{cm}^2`$ and a perimeter $`C_{fit,exp}=(113.68\pm 2.16)\mathrm{cm}`$, which are very close to the design values $`A_{design}=714.15\mathrm{cm}^2`$ and $`C_{design}=111.41\mathrm{cm}`$. Within the given uncertainties the respective coefficients $`A`$ and $`C`$ agree fairly well. From the numerical data which were obtained from a $`\gamma =1`$ stadium billiard with area $`A=4\pi `$ the coefficient $`C`$ can be calculated. The fit of the Weyl-formula to the numerical data yields values for coefficients $`A`$ and $`C`$ within a few per mille of the calculated coefficients. From this we conclude that due to the not well known mechanical imperfections of the billiard the uncertainties in the average properties of the spectrum expressed through the coefficients $`A`$ and $`C`$ of the Weyl-formula are larger in the experiment than in the numerical simulations.
With the unfolding indicated above the different geometrical properties of the real and ideal billiards are removed and a direct comparison of the first 770 eigenvalues becomes possible. Computing the difference of the experimental and the numerical data, $`ϵ_{exp}ϵ_{num}`$, and plotting this difference over the unfolded experimental eigenvalues, the upper part of Fig. 2 is obtained. The curve fluctuates around zero, which is an indication for complete spectra, i.e. no mode is missing. The appearing oscillations in the curve are random and reflect the fact of how accurate a certain eigenfrequency could be determined experimentally. If only one eigenvalue in the measured sequence is artificially removed, the curve shows a clearly observable step of height 1 at the position where the eigenvalue was dropped. This is demonstrated in the lower part of Fig. 2, where the $`427^{th}`$ mode has been removed in the experimental spectrum. More information can not be extracted from this direct comparison since systematic errors and imperfection have been removed through the unfolding procedure. Therefore we want to concentrate in the following on the statistical properties of the presented complete sequences of data.
In the next step, we extract the fluctuating part of the staircase function $`N^{fluc}(k)=N(k)N^{Weyl}(k)`$ from the measured resp. calculated eigenvalue sequences (Fig. 3). As expected oscillations around zero are seen, also indicating no missing modes.
In the upper part of Fig. 3 a strong enhancement of the amplitude of the fluctuations as a function of wave number $`k`$ – a characteristic feature of regular systems – is observed as well as a periodic gross structure with spacing of $`15.7\mathrm{m}^1`$ (750 MHz). This behavior of $`N^{fluc}(k)`$ is caused by the well known family of marginally stable periodic orbits, which bounce between the two straight segments of the billiard, the so-called bouncing ball orbits (bbo), with length $`2r`$. The cumulative level density $`N(k)`$ shows therefore periodic oscillation with a fixed period of $`2\pi /2R=15.7`$ m<sup>-1</sup> around the value given by the Weyl-formula. These observations can be described by the semiclassical expression of the contribution of the bbo to the spectrum which reads
$$N^{bbo}(k)=\frac{a}{r}\left(\underset{1nX}{}\sqrt{X^2n^2}\frac{\pi }{4}X^2+\frac{1}{2}X\right)$$
(18)
with $`X=(kr)/\pi `$. This formula (for a generalization of it in three dimensions see ) reproduces the mean behavior of the experimental data as shown in the upper part of Fig. 3. After subtraction of this smooth correction in addition to the expression of Weyl the proper fluctuating part of the level density is obtained, which is plotted in the lower part of Fig. 3. Naturally the same result is obtained for the data from the numerical calculations.
To determine the degree of chaoticity of the investigated $`\gamma =1`$ stadium billiard we next calculate the cumulative nearest neighbour spacing distribution $`I(s)`$ for the fluctuating part of the staircase function corrected by the Weyl and the bbo terms. In Fig. 4 the dashed curve shows the density for a certain spacing $`s`$ of two adjacent unfolded eigenmodes for the experimental and the numerical data. Beside the data also the two limiting cases, the Poisson- and the GOE-distribution, are displayed. Using the ansatz of Berry-Robnik one obtains a mixing parameter $`q=0.97\pm 0.01`$ for the experimental and $`q=0.98\pm 0.02`$ for the numerical data. As stated in the Bunimovich stadium billiard should be fully chaotic, which is expressed through $`q`$ being very close to unity within the uncertainties of the fitting procedure.
For the $`\mathrm{\Sigma }^2`$-statistics which measures long-range correlations the effects of the bbo’s are strikingly visible, see Fig. 5. Their presence influences the rigidity of the spectrum for large values of length $`L`$. A proper handling of these orbits as described for the cumulative NND changes the form of the curve towards the expected GOE-like behavior. In the upper part of Fig. 5 the experimental data ($`\mathrm{}`$) and the numerical data ($`\mathrm{}`$) are displayed for the $`\mathrm{\Sigma }^2`$-statistics before removing the contribution of the bbo’s. Comparing the $`\mathrm{\Sigma }^2`$-statistics of the numerical and the experimental data, no significant difference between both data sets can be seen. In contrast to this a small difference between both is extracted when the bbo contribution is removed (lower part of Fig. 5). However, in both cases the $`\mathrm{\Sigma }^2`$-statistics follow very closely the GOE prediction up to $`L3.5`$, where the distribution saturates. This is in a very good agreement with the theoretical predicted value according to of $`L_{max}3.8`$. This length $`L=L_{max}`$ also refers to the shortest periodic orbit, which is in the present billiard the bbo. Why is $`\mathrm{\Sigma }^2(L)`$ different for $`L>L_{max}`$ for the two data sets? This might be due to the fact that the fabricated microwave cavity is a real system with small but existing mechanical imperfections, e.g. a slight non-parallelity of the straight segments of the cavity, which should have its effect on the periodic orbits.
After comparing the statistical measures of the two investigated $`\gamma =1`$ stadium billiards, we now consider their periodic orbits which are given through the Fourier transformed spectrum of the eigenvalue density $`\rho ^{fluc}(k)`$, see Eq. (17). The lengths of the classical periodic orbits (po) correspond to the positions and their stability roughly spoken to the height of the peaks in the spectrum. In Fig. 6 the mod-squared of the Fourier transformed of $`\rho ^{fluc}(k)`$ for the experimental (upper part) and the numerical (lower part) data are compared. The numerical length spectrum is scaled to the experimental one in such a way that the lowest bbo occurs at the same location in the two Fourier spectra. Then the bbo’s, because of their dominating nature in the spectra, were removed . Now a direct comparison between experiment and simulation becomes possible. It can be seen in Fig. 6 that beside the positions of the peaks (lengths of the po) also their heights (stability of the po) are almost identical in both spectra.
Let us finally in this subsection return to the bouncing ball orbits. A detailed analysis of those in the experimental billiard, which have lengths $`x=nR`$ ($`n=2,4,6,\mathrm{}`$) shows that the distance $`R`$ between the straight lines of the microwave resonator is not exactly $`R=20`$ cm as specified in the construction but rather $`R=(19.92\pm 0.05)`$ cm (average over the first six recurrences of the bbo’s). That means that the two straight segments of the billiard are not exactly parallel. The reasons for this deviation ($`\mathrm{\Delta }R/R=4\times 10^3`$) from the design value are on the one hand a finite mechanical tolerance during the fabrication ($`3\times 10^3`$) and on the other hand effects coming from thermal contractions at low temperatures ($`1\times 10^3`$). This small mechanical imperfection is also the reason for the higher saturation level of the $`\mathrm{\Sigma }^2`$-statistics in the experimental data in comparison to the numerical one after removing the bbo contribution (lower part of Fig. 5).
### C Influence of the bouncing ball orbits in different Bunimovich stadiums
In this section we discuss finally the influence of the bouncing ball orbits in the length spectra and in the level statistics of different Bunimovich stadium billiards. Therefore we compare the $`\gamma =1`$ stadium discussed so far with an also measured superconducting $`\gamma =1.8`$ stadium billiard ($`a=36`$ cm and $`R=20`$ cm), already presented in . For this comparison we use the first 1060 eigenfrequencies up to a frequency of 18 GHz.
The interesting features for comparing the two stadiums are the amplitudes of the peaks for the bbo and not the peaks of the unstable po’s, since they naturally have to differ. In Fig. 7 the length spectra of the $`\gamma =1`$ and the $`\gamma =1.8`$ stadium billiards are shown. It is clearly visible that the peaks of the bbo’s have different heights or amplitudes in the two investigated stadiums. Because the height of the peaks corresponds to the stability of the periodic orbits, these differences can be easily explained. For the $`\gamma =1.8`$ stadium the geometrical area on which the bbo’s exist is almost twice the area of the $`\gamma =1`$ billiard. This correlation between the area of the rectangular part of the billiards with radius $`R`$ fixed and the amplitude of the peaks for the bbo in the FT has been first pointed out in . This effect is also expressed through the ratio $`a/R`$ in Eq. (18).
The effects of the bbo’s are furthermore visible in the statistical measures, as pointed out in the previous section. For the nearest neighbour spacing distribution we have found a Berry-Robnik mixing parameter $`q=0.97\pm 0.02`$ for the experimental $`\gamma =1`$ billiard before and $`q=0.98\pm 0.01`$ after removing the bbo contribution. But for the $`\gamma =1.8`$ stadium, presented in , the situation differs. There we have a mixing parameter $`q=0.87\pm 0.03`$ for the billiard with bbo’s and $`q=0.97\pm 0.02`$ without them. Using the $`\mathrm{\Sigma }^2`$-statistics to investigate the long range correlations, the effect of the bbo’s for stadiums of different $`\gamma `$ values becomes even more obvious. The $`\mathrm{\Sigma }^2`$-statistics for the $`\gamma =1.8`$ stadium, including bbo’s, increases more strongly, see upper part of Fig. 8 ($``$), than for the $`\gamma =1`$ billiard in the lower part of the same Figure ($`\mathrm{}`$). After removing the bbo contribution one notices saturation sets in at the same $`L_{max}4`$ for both stadiums. In other words, the shortest po is the same in both stadiums as it should be. The results which have been found are thus in excellent agreement as predicted in which state that the influence of the bbo’s is in the $`\gamma =1`$ stadium billiard is very small, so that it can be characterized as being the most chaotic stadium.
## V Conclusion
In this paper we have shown two different methods, an experimental and a numerical one, to obtain eigenvalue sequences of a $`\gamma =1`$ Bunimovich stadium billiard. A direct comparison of these two data sets clearly reveals that both sets are complete and that no eigenmode is missing. Informations, e.g. accuracy of the measured resp. simulated eigenvalues can not be obtained from this test. Therefore we analyzed the statistical properties of the billiards. Using the cumulative nearest neighbour spacing distribution $`I(s)`$ both data sets show the same predicted GOE behavior after removing the contribution of the bouncing ball orbits. The same result is obtained from the $`\mathrm{\Sigma }^2`$-statistics, by which we investigated the long range correlations of the eigenvalues. Only the saturation level of the experimental data at large intervals is somewhat above of the level of the numerical data, because the bbo’s can not be removed exactly. The reason can be found in the length spectra of the data which uncovers that the lengths of the bbo’s in the experiment slightly deviate from their expected values due to the design of the cavity. The cavity contracts at low temperatures and its mechanical fabrication has only been possible within certain tolerances. On the other hand we have shown that such small imperfections of the real system do not have any influence on the results of the statistical analysis. Furthermore, we have investigated the influence of the bbo in two different stadium billiards. They have a clearly different effect on the respective statistical measures, confirming in particular that the $`\gamma =1`$ stadium is even more chaotic than the $`\gamma =1.8`$ one.
The comparison of the experimental and the numerical eigenvalues was performed to show how accurate the results of the statistical analysis depend on the given method of providing eigenvalues. To obtain informations about simple two-dimensional billiards, like the presented $`\gamma =1`$ stadium billiard, numerical calculations sometimes have an advantage, e.g. if one is interested only in eigenvalues. To simulate experiments involving eigenfunctions like in is a different matter. Furthermore, for problems where one is interested in billiards with scatters inside, billiards with fractal boundaries or three-dimensional billiards, etc. the experiment clearly offers a very convenient way to obtain large sets of eigenvalues quickly.
## VI Acknowledgement
For the precise fabrication of the niobium resonators we thank H. Lengeler and the mechanical workshop at CERN/Geneva. This work has been supported partly still by the Sonderforschungsbereich 185 “Nichtlineare Dynamik” of the Deutsche Forschungsgemeinschaft (DFG) and by the DFG contract RI 242/16-1.
|
no-problem/9906/cond-mat9906341.html
|
ar5iv
|
text
|
# Multiple quantum phases in artificial double-dot molecules
## Abstract
We study coupled semiconductor quantum dots theoretically through a generalized Hubbard approach, where intra- and inter-dot Coulomb correlation, as well as tunneling effects are described on the basis of realistic electron wavefunctions. We find that the ground-state configuration of vertically-coupled double dots undergoes non-trivial quantum transitions as a function of the inter-dot distance $`d`$; at intermediate values of $`d`$ we predict a new phase that should be observable in the addition spectra and in the magnetization changes.
Semiconductor quantum dots (QDs) are nano- or mesoscopic structures that can be regarded as ‘artificial atoms’ because of the three-dimensional carrier confinement and the resulting discrete energy spectrum. They are currently receiving great attention because they can be designed to study and exploit new physical phenomena: On the one hand, the nature and scale of electronic confinement allows the exploration of regimes that are not accessible in conventional atomic physics; on the other hand, they lead to novel devices dominated by single- or few-electron effects .
One of the important challenges at this point is to understand the fundamental properties of coupled quantum dots, the simplest structures that display the interactions controlling potential quantum-computing devices. Here also the interdot coupling can be tuned through external parameters, far out of the regimes known in ‘natural molecules’ where the ground-state interatomic distance is dictated by the nature of bonding. We expect that new phenomena will occur in these ‘artificial molecules’ (AMs) when the relative importance of Coulomb interaction and single-particle tunneling is varied. Through the study of such phenomena, coupled dots may become a unique laboratory to explore electronic correlations.
In this paper we analyze ground and excited few-particle states of realistic double quantum dots (DQDs) using a theoretical scheme that fully takes into account intra- and inter-dot many-body interactions. We focus on strongly confined vertically-coupled DQDs, and show that, for a given number of electrons, $`N`$, the ground state configuration undergoes non-trivial quantum transitions as a function of the inter-dot distance $`d`$; we identify specific ranges of the DQD parameters characterizing a ‘coherent’ molecular phase, where tunneling effects dominate, and a phase where the inter-dot interaction is purely electrostatic. At intermediate values of $`d`$ we predict a new phase that should be observable by means of transport experiments in the addition spectra and in the magnetization changes.
From the theoretical point of view, a major difficulty is that we cannot make use of perturbative schemes in the calculation of the DQD many-body ground state, since we want to investigate all inter-dot coupling regimes. We thus write the many-body hamiltonian $`H`$ within a generalized Hubbard (GHH) approach , and chose a basis set formed by suitable single-particle wavefunctions localized on either dot ($`i=1,2`$), and characterized by orbital quantum numbers, $`\alpha `$ (to be specified below), and by the spin quantum number, $`\sigma `$. On this basis the many-body hamiltonian is
$`\widehat{H}`$ $`=`$ $`{\displaystyle \underset{i\alpha \sigma }{}}\stackrel{~}{\epsilon }_\alpha \widehat{n}_{i\alpha \sigma }t{\displaystyle \underset{\alpha \sigma ij}{}}\widehat{c}_{i\alpha \sigma }^{}{}_{}{}^{}\widehat{c}_{j\alpha \sigma }+{\displaystyle \frac{1}{2}}{\displaystyle \underset{i\alpha \beta \sigma }{}}U_{\alpha \beta }\widehat{n}_{i\alpha \sigma }\widehat{n}_{i\beta \sigma }`$ (2)
$`+{\displaystyle \frac{1}{2}}{\displaystyle \underset{i\alpha \beta \sigma }{}}\left(U_{\alpha \beta }J_{\alpha \beta }\right)\widehat{n}_{i\alpha \sigma }\widehat{n}_{i\beta \sigma }+{\displaystyle \underset{\alpha \beta \sigma \sigma ^{}}{}}\stackrel{~}{U}_{\alpha \beta }\widehat{n}_{1\alpha \sigma }\widehat{n}_{2\beta \sigma ^{}}.`$
The first two addenda are the single-particle on-site and hopping term respectively: $`\stackrel{~}{\epsilon }_\alpha `$ are the single-particle energies, $`t`$ the tunneling parameter; $`\widehat{n}`$ are the occupation numbers and $`\widehat{c}`$ ($`\widehat{c^{}}`$) the creation (destruction) operators. The third and fourth terms account for intra-dot Coulomb interaction between electrons with antiparallel and parallel spins, respectively: $`U`$ and $`J`$ are the intra-dot Coulomb and exchange integrals. Finally, the last term represents the inter-dot Coulomb coupling, $`\stackrel{~}{U}`$ being the inter-dot Coulomb integrals. These integrals are in turn expressed in terms of single-particle states:
$$U_{\alpha \beta }=\frac{e^2\left|\varphi _{i\alpha }\left(𝐫\right)\right|^2\left|\varphi _{i\beta }\left(𝐫^{}\right)\right|^2}{\kappa _r\left|𝐫𝐫^{}\right|}d𝐫d𝐫^{};$$
$$\stackrel{~}{U}_{\alpha \beta }=\frac{e^2\left|\varphi _{1\alpha }\left(𝐫\right)\right|^2\left|\varphi _{2\beta }\left(𝐫^{}\right)\right|^2}{\kappa _r\left|𝐫𝐫^{}\right|}d𝐫d𝐫^{};$$
$$J_{\alpha \beta }=\frac{e^2\varphi _{i\alpha }^{}{}_{}{}^{}\left(𝐫\right)\varphi _{i\beta }^{}{}_{}{}^{}\left(𝐫^{}\right)\varphi _{i\alpha }\left(𝐫^{}\right)\varphi _{i\beta }\left(𝐫\right)}{\kappa _r\left|𝐫𝐫^{}\right|}d𝐫d𝐫^{};$$
with $`\varphi _{i\alpha }\left(𝐫\right)`$ full three-dimensional single-particle wavefunctions (obtained within the usual envelope-function approximation), $`e`$ electronic charge and $`\kappa _r`$ relative dielectric constant.
Here we shall consider a specific type of nanostructures, namely vertically coupled cylindrical DQDs ; the ingredients entering the Hamiltonian can therefore be defined explicitly. For simplicity, $`V(r)`$ can be assumed to be separable in the $`xy`$ and $`z`$ components; the profile is taken to be parabolic in the $`xy`$ plane, with confinement energy $`\mathrm{}\omega `$, and a symmetric double quantum well (QW) along $`z`$ . The eigenstates of the $`xy`$ harmonic potential are the usual Fock-Darwin states $`|\alpha =|(n,m)`$ ($`n`$, $`m`$ radial and angular quantum numbers) . Along $`z`$, the QW thickness is such that only the lowest eigenstates (symmetric, $`|s`$, and antisymmetric $`|a`$) are relevant for the low-energy spectrum. From these, we construct a complete set of states that are localized on either dot (see inset of Fig. 1): $`|1=\left(|s+|a\right)/\sqrt{2}`$ and $`|2=\left(|s|a\right)/\sqrt{2}`$. The basis we use is therefore the direct product $`|i\alpha \sigma =|i|\alpha |\sigma `$. The single-particle energies are $`\stackrel{~}{\epsilon }_\alpha =\epsilon _\alpha +(ϵ_s+ϵ_a)/2`$, and the tunneling parameter is $`t=(ϵ_aϵ_s)/2`$, with $`ϵ_s`$, $`ϵ_a`$ double-well eigenenergies, and $`\epsilon _\alpha `$ oscillator energies.
Using the above expressions, the three-dimensional Coulomb integrals and the tunneling parameter are computed directly from the single-particle states for each sample. Fig. 1 shows the result for $`t`$, $`U_{\alpha \beta }`$ and $`\stackrel{~}{U}_{\alpha \beta }`$ \[$`\alpha =\beta =(0,0)`$\] calculated for a GaAs/AlAs DQD ($`\kappa _r=`$ 12.98, effective mass 0.065 $`m_e`$) with $`\mathrm{}\omega =`$ 10 meV, as a function of the interdot distance (barrier width) $`d`$. Note that already around $`d=5`$ nm the inter-dot Coulomb integral exceeds the single-particle hopping parameter.
To obtain the many-body energies and eigenstates, the Hamiltonian $`\widehat{H}`$ is then diagonalized exactly for each value of $`N`$ (total number of electrons) on each configuration subspace labeled by the quantum numbers $`S`$ ($`z`$-component of the total spin) and $`M`$ ($`z`$-component of the total angular momentum).
This approach has two main advantages: First, it allows us to solve the many-body problem consistently in the different coupling regimes, from the limit where $`t`$ dominates over the Coulomb integrals to the opposite limit where $`t`$ is negligible. Second, it can provide quantitative predictions for given DQD structures, since it contains no free parameters and uses realistic ingredients ($`t`$, $`U_{\alpha \beta }`$, $`\stackrel{~}{U}_{\alpha \beta }`$, $`J_{\alpha \beta }`$) calculated for each nanostructure .
Fig. 2 shows the calculated ground-state energies $`E_N`$ of correlated $`N`$-particle states as a function of the inter-dot distance $`d`$ . The three lowest excited states are also shown for comparison. As expected, when $`d`$ is large, the system behaves as two isolated QDs. With decreasing $`d`$, some of the many-body excited states, favoured by Coulomb interactions, become lower in energy. For $`N3`$ a single quantum phase transition occurs, below which the new ground state is a molecular-like state. For $`N>3`$, two successive transitions take place, at $`d=d_b`$ and $`d=d_a`$, and an intermediate non-trivial phase is predicted to occur in the range $`d_a<d<d_b`$. Note that this phase is stable in a relatively large range of $`d`$ values, which depends on the number of electrons. An accurate determination of $`d_a`$ and $`d_b`$ requires the correct inclusion of all inter-dot coupling terms, including the inter-dot Coulomb integrals. If the latter are neglected, all quantum transitions occur for smaller values of the inter-dot distance, and the $`d`$-range of the intermediate phase is underestimated significantly; had also the other many body terms been neglected, the intermediate phase would disappear leaving a simple crossover from a molecular to an atomic like regime at $`d=da=db`$.
To understand the nature of the different phases it is useful to examine the many-body states in terms of the single Slater determinants that contribute to each of them. Here we discuss explicitly the 4-electron case with the help of the insets in Fig. 2, but the same reasoning can be followed for the other cases. Both in the case of very small and very large interdot distances the ground state can be essentially described in terms of a single Slater determinant: for large values of $`d`$ ($`d>d_b`$), the relevant configurations for the ground state have two electrons in the lowest level of each isolated dot; the $`|s`$ and $`|a`$ extended ‘molecular’ orbitals derived from the lowest ‘atomic’ states are of course almost degenerate and both filled with two electrons. In the opposite limit ($`d<d_a`$), by expanding the localized atomic orbitals in terms of molecular orbitals we recognize that the $`|s`$ state derived from the lowest atomic state is filled with two electrons, while the corresponding $`|a`$ molecular state is empty. The two remaining electrons occupy the next bonding molecular orbitals —derived from the higher $`p_x`$ and $`p_y`$ levels— with parallel spin, in such a way that $`S`$ is maximized. This is the manifestation of Coulomb interaction which leads to Hund’s rule for molecules.
In these two extreme phases the single particle picture is essentially correct, provided that the appropriate basis set (either localized or extended orbitals) is used. In the intermediate phase, $`d_a<d<d_b`$, this is no longer true and the ground state is a mixture of different Slater determinants in any basis set . In this sense again the intermediate phase exhibits an intrinsic many-body character. Coulomb direct and exchange terms, responsible for the selection of the global quantum numbers, determine a new ground state configuration, where both $`S`$ and $`M`$ are maximized.
We obtain a clear evidence of the different electronic distribution in the three quantum phases by calculating the spin-dependent electronic pair correlation function, defined as
$$g_{\sigma ^{},\sigma ^{\prime \prime }}(\rho ,z^{},z^{\prime \prime })=𝑑𝐑\widehat{\mathrm{\Psi }}_\sigma ^{}^{}(𝐫^{})\widehat{\mathrm{\Psi }}_{\sigma ^{\prime \prime }}^{}(𝐫^{\prime \prime })\widehat{\mathrm{\Psi }}_{\sigma ^{\prime \prime }}(𝐫^{\prime \prime })\widehat{\mathrm{\Psi }}_\sigma ^{}(𝐫^{}),$$
where $`𝐑=(\rho ^{}+\rho ^{\prime \prime })/2`$ and $`\rho =(\rho ^{}\rho ^{\prime \prime })`$. Here $`\rho ^{}=(x^{},y^{})`$, $`\rho ^{\prime \prime }=(x^{\prime \prime },y^{\prime \prime })`$ are the in-plane spatial coordinates of the electrons in the pair, and $`z^{}`$ and $`z^{\prime \prime }`$ are their coordinates along $`z`$, that will be kept fixed at the center of either QD, i.e. in $`z_1`$ or $`z_2`$. $`𝐫^{}=(\rho ^{},z^{})`$, $`𝐫^{\prime \prime }=(\rho ^{\prime \prime },z^{\prime \prime })`$. $`\sigma ^{}`$, $`\sigma ^{\prime \prime }`$ are the spin variables and can assume the values $``$ or $``$. In Fig. 3 we plot for example $`g_,(x,y,z^{},z^{\prime \prime })`$ for a double QD with $`N=6`$ electrons, at three values of the inter-dot distance corresponding to different quantum phases. Here $``$ represents the minority spins. For $`d<d_a`$ (left column), it is indeed apparent that the pair correlation function is the same when both electrons are in the same dot \[$`z^{}=z^{\prime \prime }=z_1`$, panel (a)\] or on different dots \[$`z^{}=z_1`$ and $`z^{\prime \prime }=z_2`$, panel (b)\]. In this sense, the system behaves ‘coherently’. For $`d>d_b`$ (right column), the maps indicate that the probability of finding two $``$-electrons in the same dot is negligible, consistently with the picture of isolated dots. In the intermediate phase $`d_a<d<d_b`$ this is no longer the case: the electronic wavefunctions extend over both dots, and the $`g_,`$ pair correlation functions are very different depending on the location of both electrons in the same or in different dots. Note that transitions between different electronic configurations vs. $`d`$ were recently identified theoretically also for classical coupled dots : We find that, for the small values of $`N`$ considered here, the number of distinct phases and the spatial distribution of electrons (as reflected in their correlation functions) is drastically modified by quantum effects.
The above findings are expected to be observable experimentally. First, the calculated magnetic-field dependent addition spectra $`A(N)`$ present clear signatures of the phase transitions described above, as illustrated in Fig. 4. Here $`A(N)=E(N)E(N1)`$ is obtained from the many-body ground state energies of the N- and (N-1)-electron systems, on the basis of single-particle states calculated in the presence of the external magnetic field $`B`$. The behaviour shown in the left, central, and right panels is representative of the three phases. Secondly, the changes in the magnetization induced by one electron addition —also accessible experimentally — are expected to follow a different pattern in each phase (see the different sequence of quantum numbers in Fig. 2).
A large experimental effort is presently devoted to transport experiments in double QDs. In most cases, the dots are obtained by gating a two-dimensional electron gas (lateral confinement), and their coupling can be tuned through a gate voltage . Indeed, a set of experiments has recently demonstrated a clear transition to a ‘coherent’ state with increasing coupling between dots . Outside this strong coupling regime, however, experiments performed in the lateral geometry have sofar evidenced classical interdot capacitance effects, probably owing to the size of dots . On the other hand, transport experiments are now available on DQD structures with strong lateral confinement fabricated by combined growth and etching techniques . The advantage is that the number of electrons in the structure is limited, while an accurate control on the inter-dot coupling is still possible by designing samples with appropriate barrier thickness. Both aspects are important to enhance many-body effects and to explore intermediate coupling regimes. Recent experimental work on these structures has focused on the weak-coupling regimes . We hope that further investigations will be stimulated by the present work, since the relevant transitions are now predicted quantitatively and should occur in a range of parameters that is accessible to state-of-the-art experiments. We expect that such studies will bring new insight into electron-electron interaction effects in coupled quantum nanostructures.
We acknowledge helpful discussions with C. Calandra and G. Goldoni. This work was supported in part by INFM PRA99-SSQI, by the EC under the TMR Network “Ultrafast Quantum Optoelectronics”, and by the MURST-40% program “Physics of Nanostructures”.
|
no-problem/9906/hep-ph9906447.html
|
ar5iv
|
text
|
# May 1999 SINP/TNP/99-22 Determination of cosmological parameters: an introduction for non-specialistsInvited talk at the “Discussion meeting on Recent Developments in Neutrino Physics”, held at the Physical Research Laboratory, Ahmedabad, February 2–4, 1999.
## 1 Introduction
In this conference on Neutrino Physics, I have been asked to talk about the determination of cosmological parameters. The reason for this, obviously, is the potential importance of neutrinos for cosmology. They can serve as a component for the dark matter of the universe. On the other hand, many important constraints on neutrino properties are derived from various cosmological considerations. The precise values of such constraints are determined by various cosmological parameters like the Hubble parameter, the density of matter in the universe, etc. In the light of this relationship, it is important to know the cosmological parameters accurately. In this talk, I will outline the methods for the determination of cosmological parameters. Since I am not a specialist in this field, at best I can hope that this exposition will be useful to other non-specialists.
The outline of the paper is the following. In Sec. 2, I define the cosmological parameters. There are, in fact, three of them — the Hubble parameter, and the present density of matter energy and vacuum energy in the universe. In Sec. 3, I show how the age of the universe depends on these parameters. Then in Sec. 4, I show how the evolution of the universe depends on the values of the cosmological parameters. An understanding of the evolution is essential in formulating strategies for the cosmological parameters. Some such strategies are discussed in general terms in Sec. 5. At the end of this section, we indicate the outline for the rest of the paper.
## 2 Defining the cosmological parameters
In cosmology, we start with the the assumption of a homogeneous and isotropic universe, which is described by the Friedman-Robertson-Walker (FRW) metric:
$`ds^2=dt^2a^2(t)\left[{\displaystyle \frac{dr^2}{1kr^2}}+r^2\left(d\theta ^2+\mathrm{sin}^2\theta d\varphi ^2\right)\right].`$ (2.1)
Here, $`r`$ is a dimensionless co-ordinate distance and $`a(t)`$ is an overall scale parameter. The physical distance between two points depends on the co-ordinate distance and the scale parameter, in a manner which will be discussed later. The parameter $`k`$, if it is non-zero, can always be adjusted to have unit magnitude by adjusting the definition of $`r`$. Thus, the possible values of $`k`$ are given by
$`k=0,\pm 1.`$ (2.2)
Usually, we think of the universe to be open if $`k<0`$, for which it expands forever. Contrarily, if $`k>0`$, the universe is closed, i.e., it is destined to recollapse at some time in the future. The borderline case, $`k=0`$, is called a flat universe. We will see that these statements are correct only if there is no cosmological constant. From the general consideration of homogeneity and isotropy, however, the cosmological constant can be present, and these notions get modified, as we will see.
In presence of a cosmological constant $`\mathrm{\Lambda }`$, the equations of motion are given by
$`R_{\mu \nu }{\displaystyle \frac{1}{2}}g_{\mu \nu }R=8\pi GT_{\mu \nu }+\mathrm{\Lambda }g_{\mu \nu }.`$ (2.3)
Here, $`R_{\mu \nu }`$ is the Ricci tensor which is defined through the metric of Eq. (2.1), and $`R=R_{\mu \nu }g^{\mu \nu }`$. Although in general Eq. (2.3) gives 10 equations, here most of them are identical because the metric is homogeneous and isotropic. In fact, one gets only two independent equations. One of them is given by
$`\left({\displaystyle \frac{\dot{a}}{a}}\right)^2={\displaystyle \frac{8\pi G}{3}}\rho +{\displaystyle \frac{\mathrm{\Lambda }}{3}}{\displaystyle \frac{k}{a^2}},`$ (2.4)
where $`\rho `$ is the energy density of matter. This equation is valid for all time. Applying it to the present time, we can write
$`H_0^2={\displaystyle \frac{8\pi G}{3}}\rho _0+{\displaystyle \frac{\mathrm{\Lambda }}{3}}{\displaystyle \frac{k}{a_0^2}},`$ (2.5)
where $`H_0`$ is the value of $`\dot{a}/a`$ at the present time, and the subscript ‘0’ on $`\rho `$ and $`a`$ denote the values of these parameters at the present time. We now introduce the dimensionless parameters
$`\mathrm{\Omega }_m`$ $``$ $`{\displaystyle \frac{8\pi G}{3H_0^2}}\rho _0,`$
$`\mathrm{\Omega }_\mathrm{\Lambda }`$ $``$ $`{\displaystyle \frac{\mathrm{\Lambda }}{3H_0^2}},`$
$`\mathrm{\Omega }_k`$ $``$ $`{\displaystyle \frac{k}{a_0^2H_0^2}}.`$ (2.6)
These definitions enable us to rewrite Eq. (2.5) in the following form:
$`1=\mathrm{\Omega }_m+\mathrm{\Omega }_\mathrm{\Lambda }+\mathrm{\Omega }_k.`$ (2.7)
This shows that, among the three parameters introduced in Eq. (2.6), we can take only $`\mathrm{\Omega }_m`$ and $`\mathrm{\Omega }_\mathrm{\Lambda }`$ to be independent. $`\mathrm{\Omega }_k`$ is not. In addition, $`H_0`$ is another parameter. These are the three cosmological parameters, and the aim of this lecture is to show how they may be determined.
Before proceeding, I would like to make one comment. Note that all these parameters are defined using values of physical quantities at the present time. One may tend to think that because of that, the present time $`t_0`$ should also be added to the list of parameters. But we will show that in fact $`t_0`$ is not independent.
## 3 The age of the universe
So far, we have discussed only one of the independent equations that arise among the set given in Eq. (2.3). In order to proceed, we need the other. This is in fact equivalent to the statement of conservation of matter, which means that the quantity $`\rho a^3`$ is constant over time, or
$`\rho a^3=\rho _0a_0^3.`$ (3.1)
From now on, it is more convenient to use dimensionless variables instead of $`a`$ and $`t`$. We define
$`y{\displaystyle \frac{a}{a_0}},\tau H_0(tt_0).`$ (3.2)
Using these variables, we can rewrite Eq. (2.4) in the following form:
$`\left({\displaystyle \frac{dy}{d\tau }}\right)^2`$ $`=`$ $`{\displaystyle \frac{y^2}{H_0^2}}\left[{\displaystyle \frac{8\pi G}{3}}{\displaystyle \frac{\rho _0}{y^3}}+{\displaystyle \frac{\mathrm{\Lambda }}{3}}{\displaystyle \frac{k}{y^2a_0^2}}\right]`$ (3.3)
$`=`$ $`{\displaystyle \frac{1}{y}}\mathrm{\Omega }_m+y^2\mathrm{\Omega }_\mathrm{\Lambda }+\mathrm{\Omega }_k`$
Eliminating $`\mathrm{\Omega }_k`$ now using Eq. (2.7), we obtain
$`\left({\displaystyle \frac{dy}{d\tau }}\right)^2=1+\left({\displaystyle \frac{1}{y}}1\right)\mathrm{\Omega }_m+\left(y^21\right)\mathrm{\Omega }_\mathrm{\Lambda },`$ (3.4)
or
$`d\tau ={\displaystyle \frac{dy}{\sqrt{1+\left(\frac{1}{y}1\right)\mathrm{\Omega }_m+\left(y^21\right)\mathrm{\Omega }_\mathrm{\Lambda }}}}.`$ (3.5)
If there was a big bang, $`y`$ was zero at the time of the bang, i.e., at $`t=0`$. On the other hand, $`y=1`$ now, by definition. Integrating Eq. (3.5) between these two limits, we obtain
$`H_0t_0`$ $`=`$ $`{\displaystyle _0^1}{\displaystyle \frac{dy}{\sqrt{1+\left(\frac{1}{y}1\right)\mathrm{\Omega }_m+\left(y^21\right)\mathrm{\Omega }_\mathrm{\Lambda }}}}.`$ (3.6)
This is the equation which shows that the age of the universe is not independent, but rather is determined by $`H_0`$, $`\mathrm{\Omega }_m`$ and $`\mathrm{\Omega }_\mathrm{\Lambda }`$.
Conventionally, one does not use the dimensionless parameter $`y`$, but rather uses the red-shift parameter $`z`$, defined by
$`1+z{\displaystyle \frac{a_0}{a}}={\displaystyle \frac{1}{y}}.`$ (3.7)
Using this variable, Eq. (3.5) becomes
$`d\tau ={\displaystyle \frac{dz}{1+z}}{\displaystyle \frac{1}{\sqrt{(1+z)^2(1+\mathrm{\Omega }_mz)z(2+z)\mathrm{\Omega }_\mathrm{\Lambda }}}},`$ (3.8)
so that Eq. (3.6) can be written in the following equivalent form:
$`H_0t_0`$ $`=`$ $`{\displaystyle _0^{\mathrm{}}}{\displaystyle \frac{dz}{1+z}}{\displaystyle \frac{1}{\sqrt{(1+z)^2(1+\mathrm{\Omega }_mz)z(2+z)\mathrm{\Omega }_\mathrm{\Lambda }}}}.`$ (3.9)
In Fig. 1, we have shown contours of equal values of $`H_0t_0`$ for different values of $`\mathrm{\Omega }_m`$ and $`\mathrm{\Omega }_\mathrm{\Lambda }`$. For a fixed value of $`\mathrm{\Omega }_\mathrm{\Lambda }`$, the figure shows that $`H_0t_0`$ decreases for increasing values of $`\mathrm{\Omega }_m`$. This is because with more matter, the force of gravity is larger, and the initial bang slows down in less time. On the other hand, for fixed $`\mathrm{\Omega }_m`$, the age increases with increasing $`\mathrm{\Omega }_\mathrm{\Lambda }`$. And finally, note that the contour lines for large values of $`H_0t_0`$ appear very closer and closer, and approaches asymptotically the line for infinite age. Beyond this line, there is a region in the parameter plane for which the age integral diverges. This is the upper left part of the plot. Later we will discuss what sort of evolution does this part represent.
## 4 Evolution of the universe
In order to discuss the evolution of the universe, let us not integrate Eq. (3.5) all the way to the initial singularity, but rather to any arbitrary time $`t`$. This gives
$`H_0(tt_0)={\displaystyle _0^y}{\displaystyle \frac{dy^{}}{\sqrt{1+\left(\frac{1}{y^{}}1\right)\mathrm{\Omega }_m+\left(y^21\right)\mathrm{\Omega }_\mathrm{\Lambda }}}}.`$ (4.1)
Equivalently, using the red-shift variable, we can write
$`H_0(t_0t)`$ $`=`$ $`{\displaystyle _0^z}{\displaystyle \frac{dz^{}}{1+z^{}}}{\displaystyle \frac{1}{\sqrt{(1+z^{})^2(1+\mathrm{\Omega }_mz^{})z^{}(2+z^{})\mathrm{\Omega }_\mathrm{\Lambda }}}}.`$ (4.2)
The numerical results of this integration has been shown in Fig. 2 for different values of the pair $`(\mathrm{\Omega }_m,\mathrm{\Omega }_\mathrm{\Lambda })`$.
Some discussion of these evolution plots are in order. We have shown the plots in six different panels, with some overlaps between them. The first panel represents different examples of universes with $`\mathrm{\Omega }_\mathrm{\Lambda }=0`$, i.e., no cosmological constant. We know that in this case, the universe closes and collapses to zero volume in the future if $`\mathrm{\Omega }_m>1`$. This shows clearly in the curve D, for which $`\mathrm{\Omega }_m=3`$. For curve C, the collapsing is not clear in the diagram, but that’s because we have not plotted for large enough values of the time variable. For curve B, $`\mathrm{\Omega }_m=0.1`$, and this universe expands forever. Curve A has $`\mathrm{\Omega }_m=1`$. Since $`\mathrm{\Omega }_\mathrm{\Lambda }=0`$, this implies that $`\mathrm{\Omega }_k=0`$, i.e., the universe is flat.
In the second panel, we show more flat universes, i.e., universes for which $`k=0`$, or alternatively $`\mathrm{\Omega }_m+\mathrm{\Omega }_\mathrm{\Lambda }=1`$. We see that, subject to this condition, the larger the value of $`\mathrm{\Omega }_\mathrm{\Lambda }`$, the longer is the past history of the universe. The extreme example is obtained for $`\mathrm{\Omega }_\mathrm{\Lambda }=1`$, shown as curve F. Although this means $`\mathrm{\Omega }_m=0`$ which is unrealistic, since we know there must be some matter in the universe, otherwise who is writing this article and who is going to read it? But in any case, this choice is instructive, and it shows that in this case, the scale parameter started in the infinite past with a zero value, and expanded very slowly for most of the past history of the universe, until recently when it started to grow.
In the third panel, we show universes with $`k>0`$, i.e., $`\mathrm{\Omega }_m+\mathrm{\Omega }_\mathrm{\Lambda }>1`$. One of these curves, viz. D, has been encountered in the first panel already. This one has $`\mathrm{\Omega }_m=3`$ and $`\mathrm{\Omega }_\mathrm{\Lambda }=0`$, and it is destined for a recollapse in the future. If we increases $`\mathrm{\Omega }_\mathrm{\Lambda }`$ a little bit, viz. to 0.1, we still obtain a recollapsing universe, as shown by the curve G. However, if we take a much larger value of $`\mathrm{\Omega }_\mathrm{\Lambda }`$, we find a evolution curve like that shown as H. Here, the universe expands forever, despite the fact that $`k>0`$, or $`\mathrm{\Omega }_k<0`$. This is contrary to our intuition obtained from the case of $`\mathrm{\Omega }_\mathrm{\Lambda }=0`$ universes.
Similar conflict is encountered in the fourth panel, which shows open universes, i.e., universes with $`\mathrm{\Omega }_m+\mathrm{\Omega }_\mathrm{\Lambda }<1`$. The curve B, shown earlier in the first panel, shows an evergrowing universe, and so does curve I. But curve J shows a recollapsing universe.
In view of these, let us adopt a more general classification of the different kinds of universes that we have seen. We will keep calling a universe open or closed depending on whether $`\mathrm{\Omega }_k`$ is positive or negative. On the other hand, a universe will be called elliptic if it recollapses in the future, and hyperbolic if it is evergrowing. For $`\mathrm{\Omega }_\mathrm{\Lambda }=0`$, open universes are necessarily hyperbolic and closed universes are elliptic. But for non-zero $`\mathrm{\Omega }_\mathrm{\Lambda }`$, we can have open hyperbolic, open elliptic, closed hyperbolic or closed elliptic universes.
Before we move on to the next section, we need to discuss the two lower panels of Fig. 2. In the lower left panel, we have repeated curve F which was shown in the second panel of flat universes. As we said, in this case the universe spends an infinite amount of time for which it hardly expands. Curve K gives another example of this kind. For different values of $`\mathrm{\Omega }_m`$ and $`\mathrm{\Omega }_\mathrm{\Lambda }`$, the initial scale parameter is different. These are examples of loitering universes. Curve L shows a universe which is very close to the loitering case. It stays at a particular value of the scale parameter for a long time, but not for infinite amount of time.
If the value of $`\mathrm{\Omega }_\mathrm{\Lambda }`$ is increased further from those of any of the loitering universe cases in the previous panel, we find bouncing universes, as shown in the last panel. In these cases, the universe started with an infinite scale factor, shrunk to a minimum value at some time in the past and is expanding now. This expansion will go on forever. These are the cases for which the age integral in Eq. (3.6) diverges, and is represented by the empty upper left corner of the age plots of Fig. 1.
All these panels have been summarized in Fig. 3, where we have shown which values of $`\mathrm{\Omega }_m`$ and $`\mathrm{\Omega }_\mathrm{\Lambda }`$ give open or closed universes, as well as hyperbolic or elliptical universes. As we see clearly in this plot, the identification of closed universes with elliptical ones works only for $`\mathrm{\Lambda }=0`$, and so does the identification of open universes with hyperbolic ones.
## 5 Strategies
We have seen, with many examples, the nature of the evolution of the universe for various values of the cosmological parameters. Now the big question is: how to determine which one is our own universe?
One thing is easy to say, that we do not live in a bouncing universe. The reason is that these universes have a minimum value of $`a/a_0`$, i.e., a certain maximum value for the red-shift parameter $`z`$, say $`z_{\mathrm{max}}`$. This value satisfies the relation
$`z_{\mathrm{max}}^2(z_{\mathrm{max}}+3){\displaystyle \frac{2}{\mathrm{\Omega }_m}}.`$ (5.1)
We know there exists quasars at $`z>4`$. So, if we make the reasonable estimate that $`\mathrm{\Omega }_m>0.02`$, these cosmologies are ruled out. However, this still leaves us with a vast region in the $`\mathrm{\Omega }_m`$-$`\mathrm{\Omega }_\mathrm{\Lambda }`$ parameter region, and we need to set up a strategy to proceed.
The general (and obvious) strategy is the following. We need to find some observable quantity which depends on the cosmological parameters, determine the value of that observable. That will give us the values for the cosmological parameters. Our original question now shifts to the following one: what is a good observable for this purpose?
We have already encountered one quantity which depends on the cosmological parameters, viz., the age of the universe. Unfortunately, it cannot serve our needs very well. The reason is that one cannot really determine $`t_0`$ from any direct observations. No matter how old an object is found in the universe, that will not determine $`t_0`$, but rather only put a lower limit on it. So we will use estimates of $`t_0`$ only as a check for whatever strategy we adopt.
Another quantity that depends on $`\mathrm{\Omega }_m`$ and $`\mathrm{\Omega }_\mathrm{\Lambda }`$ is the lookback time for any object, which is $`t_0t`$ that appears in, e.g., Eq. (4.2). We look at a certain object in the sky. We find its red-shift, which can be done very accurately. If we now have some way of knowing the time at which the light that we observe was emitted from the object, we are through. But this cannot be done very well. By looking at the composition of the object and by comparing it with some evolution model, we can probably estimate the age, but it will depend on the evolution model, and so the method is somewhat uncertain.
Therefore one usually relies on the measurement of distances of objects as a function of their redshifts for the purpose of determining cosmological constants. In order to explain the process, first we need to find out how the physical distance of an object depends on $`z`$ for different values of $`\mathrm{\Omega }_m`$ and $`\mathrm{\Omega }_\mathrm{\Lambda }`$. This issue is taken up in the next section. Following that, in Sec 7, we will discuss some methods of measurements of distances, and summarize how they determine the cosmological constants. The importance of these determinations for particle physics is given in Sec. 8.
## 6 Distance vs redshift relation
A light ray traces a null geodesic, i.e., a path for which $`ds^2=0`$ in Eq. (2.1). Thus, a light ray coming to us satisfies the equation
$`{\displaystyle \frac{dr}{dt}}={\displaystyle \frac{\sqrt{1kr^2}}{a}},`$ (6.1)
where $`r`$ is the dimensionless co-ordinate distance introduced in Eq. (2.1). Using Eqs. (3.7) and (3.8), we can rewrite it as
$`{\displaystyle \frac{dr}{\sqrt{1+\mathrm{\Omega }_ka_0^2H_0^2r^2}}}=(1+z){\displaystyle \frac{dt}{a_0}}={\displaystyle \frac{1}{H_0a_0}}{\displaystyle \frac{dz}{\sqrt{(1+z)^2(1+\mathrm{\Omega }_mz)z(2+z)\mathrm{\Omega }_\mathrm{\Lambda }}}},`$ (6.2)
where on the left side, we have replaced $`k`$ by $`\mathrm{\Omega }_k`$ using the definition of Eq. (2.6). Integration of this equation determines the co-ordinate distance as a function of $`z`$:
$`H_0a_0r(z)={\displaystyle \frac{1}{\sqrt{|\mathrm{\Omega }_k|}}}\mathrm{sinn}\left[\sqrt{|\mathrm{\Omega }_k|}{\displaystyle _0^z}{\displaystyle \frac{dz^{}}{\sqrt{(1+z^{})^2(1+\mathrm{\Omega }_mz^{})z^{}(2+z^{})\mathrm{\Omega }_\mathrm{\Lambda }}}}\right],`$ (6.3)
where “sinn” means the hyperbolic sine function if $`\mathrm{\Omega }_k>0`$, and the sine function if $`\mathrm{\Omega }_k<0`$. If $`\mathrm{\Omega }_k=0`$, the sinn and the $`\mathrm{\Omega }_k`$’s disappear from the expression and we are left only with the integral.
The physical distance to a certain object can be defined in various ways.<sup>1</sup><sup>1</sup>1See, e.g., Sec. 3.3 of Ref. , or Sec. 19.9 of Ref. . For what follows, we will need what is called the “luminosity distance” $`\mathrm{}`$, which is defined in a way that the apparent luminosity of any object goes like $`1/\mathrm{}^2`$. This is related to the co-ordinate distance by
$`\mathrm{}(z)=a_0^2r(z)/a(z)=(1+z)a_0r.`$ (6.4)
Thus,
$`H_0\mathrm{}(z)={\displaystyle \frac{1+z}{\sqrt{|\mathrm{\Omega }_k|}}}\mathrm{sinn}\left[\sqrt{|\mathrm{\Omega }_k|}{\displaystyle _0^z}{\displaystyle \frac{dz^{}}{\sqrt{(1+z^{})^2(1+\mathrm{\Omega }_mz^{})z^{}(2+z^{})\mathrm{\Omega }_\mathrm{\Lambda }}}}\right].`$ (6.5)
In Fig. 4, we plot $`H_0\mathrm{}(z)`$ for various choices of $`\mathrm{\Omega }_m`$ and $`\mathrm{\Omega }_\mathrm{\Lambda }`$ to show the general nature of the variation.
## 7 Measurement of distances
### 7.1 General remarks
So let us now ask how one can measure the distance of an object. In a sense, this is the most important problem of observational cosmology. Older methods of measuring distances used ladder techniques . This means that, upto a certain distance, one particular method was used, and this was used to calibrate another method which could be used for larger distances. This process was carried on to higher and higher rungs of the ladder. In order to get to distances large enough to distinguish between different universes, one had to go through several rungs of the ladder, and accordingly there were too many uncertainties in the measurement.
Recently various other methods have been developed and employed to measure the cosmological parameters. These methods avoid using the ladder technique and are therefore expected to provide much better estimates of distances. Some of these are listed below.
Gravitational lensing. Different images of an object from a gravitational lens are formed by light rays coming through different directions, and therefore traversing different path lengths. If the original object has some periodicity in the luminosity, the periodicity of the images will depend on the path length. This provides a way of determining distances . So far, data is very scarce.
Sunayev-Zeldovich effect. Photons from the cosmic microwave background, when passing through a galaxy, get scattered. As a result, they gain in energy. Thus, looking at the direction of a galaxy, the microwave background radiation does not look thermal. Rather, there is some depletion of the low energy photons and some gain of the high energy ones. The amount of this depletion depends on the size of the galaxy. Using this, one can measure the actual sizes of the galaxies. The apparent size observed would then provide a measure for the distance . The data obtained until now show a lot of scatter, and we will not discuss them here.
Microwave background anisotropies. The evolution of density perturbations depend on the values of the cosmological parameters. Therefore, the anisotropies in the microwave background radiation should be predictable in terms of the values of $`\mathrm{\Omega }_m`$ and $`\mathrm{\Omega }_\mathrm{\Lambda }`$. With very accurate determinations of these anisotropies in the near future, this method will probably turn out to be the best probe for the cosmological parameters.
Supernova 1a. In this method, it is assumed that any supernova of type 1a has the same absolute luminosity. The measurement of the apparent luminosity therefore provides a measure for the distance. This will be discussed in more detail in the remaining part.
### 7.2 Results from Supernova 1a measurements
Supernova of type 1a are identified by the nature of their light curve, i.e., the variation of their intensity with time, as shown in Fig. 5. In addition, the spectral analysis reveals lines for heavy elements like magnesium and silicon.
Supernovas of this type are believed to occur due of the merger of two white dwarf stars whose masses are very close to the Chandrasekhar limit. Since the masses entering into the explosion is roughly the same for any supernova 1a, it is reasonable to assume that the intrinsic (or absolute) luminosity of any supernova of this type is the same. Thus, if one measures the apparent luminosity at the maximum of the light curve, this should scale as the inverse square of the distance of the supernova, where the distance is defined to be the luminosity distance. Astronomers use ‘magnitudes’ to denote luminosities, which is a logarithmic representation. Thus, the apparent magnitude at the maximum should satisfy the relation
$`m_{\mathrm{max}}\mathrm{log}\mathrm{}^2.`$ (7.1)
We have already shown in Fig. 4 how $`\mathrm{log}\mathrm{}`$ varies with $`z`$ for different values of $`\mathrm{\Omega }_m`$ and $`\mathrm{\Omega }_\mathrm{\Lambda }`$. So the strategy must be simple. One should plot the observed values of the distance vs $`z`$, and determine which pairs of values of $`\mathrm{\Omega }_m`$ and $`\mathrm{\Omega }_\mathrm{\Lambda }`$ give the best fit for the observed points.
There is only one small point to settle before embarking on this program. The vertical axis of Fig. 4 gives the logarithmic values of $`H_0\mathrm{}`$, not of $`\mathrm{}`$. Thus, unless we know what $`H_0`$ is, we cannot proceed. However, the thing to note is that for small values of $`z`$, the plots are independent of $`\mathrm{\Omega }_m`$ and $`\mathrm{\Omega }_\mathrm{\Lambda }`$. Thus, if one determines the red-shifts and apparent magnitudes of nearby supernovas, that should determine the Hubble parameter. Using this value then, one can go over to larger values of $`z`$ and determine $`\mathrm{\Omega }_m`$ and $`\mathrm{\Omega }_\mathrm{\Lambda }`$.
The part of the task at low $`z`$ was extensively done a few years ago , and the data are shown in Fig. 6. The value of the Hubble parameter obtained from this data is:
$`H_0=(63.1\pm 3.4\pm 2.9)\mathrm{km}\mathrm{s}^1\mathrm{Mpc}^1.`$ (7.2)
Using this value, one can determine the distances of farther supernovas. This has been started in 1998 under the team effort called the ‘Supernova Cosmology Project’ (SCP). They measured the redshift and the effective magnitude of 42 supernovas and published their result, which is shown in Fig. 6. Superimposed on their data are the results expected from different combinations of $`\mathrm{\Omega }_m`$ and $`\mathrm{\Omega }_\mathrm{\Lambda }`$. Apart from an overall normalization, these are the curves as shown in Fig. 4. One can now determine which values of $`\mathrm{\Omega }_m`$ and $`\mathrm{\Omega }_\mathrm{\Lambda }`$ fit the data sufficiently well, and the results of the analysis of the SCP has been shown in Fig.7.
Clearly, the fits show that the cosmological constant is non-zero. In fact, if we stick to flat universes, which are denoted by the diagonal straight line on this plot, the best value that the SCP advocate is
$`\mathrm{\Omega }_m0.22,\mathrm{\Omega }_\mathrm{\Lambda }0.78.`$ (7.3)
In general, however, we obtain a region in the plane, and the regions allowed at different confidence levels are shown on the plot.
In Fig. 1, we have seen how these contours look like. In general, one obtains contours of equal $`H_0t_0`$. Assuming a value of $`H_0`$, one can draw contours of equal age, which have also been superposed on the plot of Fig. 7. The value of $`H_0`$ assumed is inspired by the results shown in Eq. (7.2), and is indicated in the figure and the caption. The best fit that the SCP advocate is:
$`t_0=(14.9\pm 1)\times 10^9\mathrm{yr},`$ (7.4)
for flat universes. If one does not assume a flat universe, the best fit for the age is
$`t_0=(14.5\pm 1)\times 10^9\mathrm{yr}.`$ (7.5)
Of course, both these values for $`t_0`$ are for the value of $`H_0`$ indicated in the figure, which is the central value quoted in Eq. (7.2).
## 8 Implications
What do these result mean for particle physics? Various particles, like neutrinos, can contribute only to $`\mathrm{\Omega }_m`$. The central value of the Hubble parameter, taken literally, tells us that the critical density of the universe is
$`\rho _c=4\mathrm{keV}/\mathrm{cm}^3.`$ (8.1)
Moreover, taking the best values advocated in Eq. (7.3), we obtain that the matter density in the present universe is
$`\rho _0900\mathrm{eV}/\mathrm{cm}^3.`$ (8.2)
With the standard number density of neutrinos of about 110/cm<sup>3</sup>, the maximum allowed mass for light stable neutrinos is about 8 eV. One can similarly go through various other constraints on neutrino properties derived from cosmology, and find that they become much more stringent than the values usually quoted.
And this is not all. As we already pointed out, the growth of density perturbations have different characteristics in a universe with non-zero $`\mathrm{\Lambda }`$. These considerations also put extra constraints on different types of possible dark matter. This is a subject which is only beginning to be investigated .
|
no-problem/9906/hep-ph9906518.html
|
ar5iv
|
text
|
# Diffractive Interactions: Theory SummaryTalk given at the 7ᵗʰ International Workshop on Deep Inelastic Scattering and QCD (DIS 99), Zeuthen, Germany, 19–23 April 1999, to appear in Nucl. Phys. B (Proc. Suppl.)
## 1 Diffraction in DIS
There has been an ongoing effort in the last years to describe diffractive DIS within the framework of QCD. Progress has been made in understanding the connections between different approaches, and we have had presentations about modelling the nonperturbative input needed in a QCD description.
Soper recalled that to leading twist accuracy, i.e. up to corrections in powers of $`1/Q^2`$, the inclusive diffractive cross section factorises into a hard photon-parton scattering subprocess and diffractive parton distributions. This factorisation has been shown to hold to all orders in perturbative QCD and may be regarded on the same footing as the corresponding theorem for the inclusive DIS cross section. Diffractive parton distributions are defined through quark and gluon field operators in a similar way as ordinary parton distributions or fragmentation functions; consequences of this are that they are process independent, and that their dependence on the factorisation scale is governed by the usual DGLAP equations.
Two tasks follow from this: to test experimentally where and how well these theory predictions are satisfied, and to measure the diffractive parton densities as a source of information on the nonperturbative physics at work in diffraction.
Two models for diffractive parton densities at a low starting scale have been presented. Soper et al. have evaluated them for a small-size hadron, coupling to a heavy quark-antiquark pair, which can be calculated perturbatively. Under the hypothesis that the basic features of the answer survive in the nonperturbative regime relevant for a proton target the result is then compared with data on $`F_2^D`$, and does indeed show the correct qualitative behaviour. In a similar spirit Mueller’s dipole approach to the BFKL pomeron, which is based on heavy onium scattering, is being used by Peschanski et al. for the description of $`F_2^{}`$ and $`F_2^D`$.
Hebecker et al. have taken the opposite extreme of a very large hadron and modelled diffractive and non-diffractive parton distributions of the proton in their semiclassical approach. Note that this description is originally formulated in a frame where the target proton is at rest: the fast-moving $`\gamma ^{}`$ splits into $`q\overline{q}`$ or $`q\overline{q}g`$ partonic states which are scattered in the colour field of the target. The expression of the amplitude obtained in this way can however be re-interpreted in the Breit frame: to leading accuracy in $`1/Q^2`$ it displays factorisation into a hard photon-parton scattering and diffractive or non-diffractive parton densities, including their QCD evolution. In a simple model for the colour field of a large target Hebecker et al. obtain parton distributions in fair agreement with the data on $`F_2^{}`$ and $`F_2^D`$.
It is remarkable that two rather opposite model assumptions of a very small and a very large hadron give results that have several similarities. One is the large amount of gluons compared with quarks in the diffractive distributions, and the other the behaviour of these distributions at small and large parton momentum fraction $`z`$ with respect to the Pomeron momentum $`x_{IP}p`$:
$$\frac{dq(z)}{dx_{IP}dt}\stackrel{z0}{}\text{const},\frac{dq(z)}{dx_{IP}dt}\stackrel{z1}{}(1z)$$
(1)
for quarks and
$$\frac{dg(z)}{dx_{IP}dt}\stackrel{z0}{}\frac{1}{z},\frac{dg(z)}{dx_{IP}dt}\stackrel{z1}{}(1z)^2$$
(2)
for gluons. Notice that such gluon distributions gently fall off as $`z1`$ and do not correspond to a “super-hard” gluon.
Wüsthoff has pointed out that the wave function of a $`\gamma ^{}`$ splitting into $`q\overline{q}`$ or $`q\overline{q}g`$ appears in these calculations, as well as in the evaluation of two-gluon exchange by Bartels et al. , in earlier work by Nikolaev , and in the dipole approach to BFKL . It is the idea underlying the BEKW parametrisation that the kinematic factors provided by the photon wave functions control the behaviour of $`F_2^D`$ at the endpoints $`\beta 0`$ and $`\beta 1`$. For the leading twist part this is related with the endpoint behaviour of the diffractive parton densities; in particular a gluon density going like $`(1z)^n`$ for $`z1`$ corresponds to a $`(1\beta )^{n+1}`$ behaviour at $`\beta 1`$ for the contribution of boson-gluon fusion to $`F_2^D`$. It would be interesting to understand in more detail to which extent the perturbative physics of the photon wave function can account for the $`\beta `$-dependence of $`F_2^D`$ and to which extent this dependence reflects the nonperturbative dynamics of gluons in the proton.
In the comparison of theory with experiment the large-$`z`$ behaviour of the diffractive gluon density still remains to be understood; a reflection of this is the existence of two solutions in the BEKW fit to the H1 data , one corresponding to a very “hard” gluon, the other to a fairly “soft” one.
An important part of the programme to investigate diffractive parton densities, and to constrain their shapes, is to look at other processes where diffractive factorisation is expected to hold. With densities obtained from an analysis of $`F_2^D`$ data it is indeed possible to describe diffractive dijet production in $`\gamma ^{}p`$ collisions, and in $`\gamma p`$ collisions in the region where they are dominated by the direct, pointlike component of the photon .
What the implications are of the H1 and ZEUS data on diffractive charm production is too early to say and will have to be clarified. It has been emphasised in the discussions that if this process is compared with the results of two-gluon exchange calculations then both the $`c\overline{c}`$ and $`c\overline{c}g`$ final states must be taken into account, the latter being important at small $`\beta `$. Note also that the $`c\overline{c}`$ final state is not included in a description based on diffractive parton densities (unless one introduces a diffractive charm quark distribution).
Further information can be expected from studies of the diffractive final state. At this point it is important to remember that part of the final state configurations is not included in the leading-twist description with diffractive parton distributions. An example is a $`q\overline{q}`$-pair with large relative transverse momentum, originating from a longitudinally polarised $`\gamma ^{}`$. Its importance for the analysis of $`F_2^D`$ at large $`\beta `$ has been stressed , in particular because of its influence on the scaling violation pattern. Bartels has reported on work to calculate the $`q\overline{q}g`$ final state in the two-gluon exchange picture for a wider part of phase space than where it is known so far, namely for configurations where quark and antiquark do not balance in transverse momentum and where the gluon is not approximately collinear with the initial proton. One motivation of this study is that for configurations with only high-$`p_T`$ partons in the diffractive system one may expect a steeper energy dependence than for the inclusive cross section as a manifestation of hard pomeron dynamics.
Williams et al. have investigated the restrictions on the diffractive system imposed by a rapidity gap cut in the HERA frame, following earlier work by Ellis and Ross. They find that no effect is to be expected with presently used values of $`\eta _{\mathrm{𝑚𝑎𝑥}}`$, but advocate to use data with stronger cuts as a means to study the structure of the final state in a way that is sensitive to the diffractive mechanism.
## 2 Leading baryons in DIS
The concept of diffractive factorisation can be extended to non-diffractive production of a leading proton or neutron; in fact diffractive parton densities are a special case of fracture functions , which describe semi-inclusive particle production in the target fragmentation region. The data on leading baryon production do indeed indicate that this factorisation is satisfied, and in particular show that in DIS leading baryon production is a leading twist phenomenon, as is diffraction. It is further found that, when the cross section is integrated over a certain range of momentum fraction for the leading baryon, its dependence on Bjorken-$`x`$ follows that of the inclusive cross section, which supports the idea of limiting fragmentation .
## 3 The energy dependence and the BFKL pomeron
The same similarity in energy dependence has been observed some time ago in the diffractive regime, and led Whitmore et al. to compare diffractive quantities (integrated over a certain range in $`x_{IP}`$ at fixed $`x`$) with inclusive ones at the level of parton densities . It should also be noticed that a number of models can actually make simultaneous predictions for $`F_2^D`$ and $`F_2^{}`$ , and one may hope that more will be learnt from confronting the dynamics of inclusive and diffractive DIS.
While for the dependence of $`F_2^D`$ on $`Q^2`$, and to a lesser degree on $`\beta `$, a number of predictions can be made in QCD and a certain convergence between theory approaches has been achieved, it is fair to say that the energy or $`x_{IP}`$-dependence is still poorly understood.
Applying the ideas of Regge phenomenology to diffractive parton densities one arrives at the Ingelman-Schlein proposal . In this scenario diffractive parton distributions factorise into a flux factor $`f_{IP/p}`$ and parton distributions $`F_{q,g/IP}`$ of the pomeron, cf. Fig. 1 (a),
$$\frac{dq,dg(z,x_{IP},t)}{dx_{IP}dt}=f_{IP/p}(x_{IP},t)F_{q,g/IP}(z,t),$$
(3)
where the flux factor can be obtained from the Regge phenomenology of soft hadronic reactions. A slightly more general ansatz with a sum over contributions from the pomeron and various reggeons turns out to work rather well as analyses of the data for $`F_2^D`$ and for leading baryons show . As is well known the pomeron intercept extracted in diffractive DIS is larger than the one found in hadron-hadron scattering and in photoproduction. It remains to be understood how such a Regge description can go together with the observed similarity in the energy dependence of $`F_2^{}`$ and $`F_2^D`$ mentioned above.
It may be worthwhile to notice that if at some factorisation scale $`Q_0^2`$, large enough to use DGLAP evolution, the $`x_{IP}`$-dependence of the diffractive parton distributions factorises as in (3), then this dependence remains unchanged when one evolves to higher $`Q^2`$, and with it the energy dependence of the leading twist part of $`F_2^D`$ . This is very much in contrast to the $`x`$-dependence in $`F_2`$, which is made steeper by evolution.
Yet another kind of factorisation is Regge factorisation (or $`k_T`$ factorisation in the context of perturbative QCD) of the entire diffractive process $`\gamma ^{}pXp^{}`$ into pomeron exchange and impact factors, describing the transitions $`\gamma ^{}X`$ and $`pp^{}`$, cf. Fig. 1 (b). It should be remembered that this factorisation goes beyond leading twist, as the $`\gamma ^{}X`$ impact factor contains more than the leading power in $`1/Q^2`$. The role of a large scale $`Q^2`$ here is to introduce perturbative QCD dynamics into the process.
To probe the pomeron in a dynamical situation where it is as much dominated by hard physics as possible, two types of ”gold plated” processes are being discussed: high energy $`\gamma ^{}\gamma ^{}`$ collisions, where the pomeron couples to a hard scale at both ends, and diffraction at high $`t`$, where a large momentum is transferred across the $`t`$-channel.
Royon et al. have studied the total $`\gamma ^{}\gamma ^{}`$ cross section in the dipole BFKL approach. Choosing similar virtualities of the two photons should provide a means to separate BFKL from DGLAP dynamics (which describes the evolution between two different momentum scales), something which cannot be achieved in the proton structure functions $`F_2^{}`$ or $`F_2^D`$. It may be worth noting that the leading order BFKL prediction for $`\gamma ^{}\gamma ^{}X`$ is excluded by the L3 data . In a phenomenological estimate of NLO corrections Royon et al. find that the effect of BFKL resummation over bare two-gluon exchange could be seen at a linear collider, while the discriminating power of LEP2 is marginal.
Theory aspects of high-$`t`$ diffraction in $`ep`$ or hadron-hadron collisions have been discussed by Forshaw . He emphasised that within the BFKL resummation even a moderately large momentum transfer across the gluon ladder is very efficient in suppressing the dangerous infrared regions in the phase space integrations, where the results of perturbation theory become doubtful. He also pointed out the virtues of large-$`t`$ diffractive production of $`J/\mathrm{\Psi }`$, light mesons or real photons, processes which do not suffer from the difficulties of hadronisation effects and rapidity gap survival encountered in forward high-$`p_T`$ jet production, and according to estimations give observable event rates in large parts of phase space. Inclusive high-$`t`$ diffraction $`\gamma pXY`$, defined by the largest rapidity separation in the event has been discussed by Cox .
On the theoretical side progress has been reported in understanding the structure of the BFKL pomeron. Kotsky has verified that the impact factors occurring in the BFKL equation at NLO satisfy a condition of self-consistency for the reggeisation of the gluon. At the level of LO accuracy the subject of multi-pomeron couplings and unitarity corrections has received much interest. Ewerz recalled that in order to satisfy the unitarity bound the leading $`\mathrm{log}(1/x)`$ approximation has to be relaxed to take into account diagrams with more than 2 reggeised gluons in the $`t`$-channel. Investigating the transitions between 2 and up to 6 gluons he found a structure consistent with the unitarity of an underlying effective field theory at high energy, based on conformal invariance, which remains to be formulated. In the dipole approach Peschanski has obtained several exact results for multi-pomeron vertices, noting that the corresponding pomeron configurations may be phenomenologically accessible in the triple Regge regime of $`ep`$ diffraction and for various types of gap-jet events at the Tevatron. Gay Ducati presented an evolution equation for $`F_2`$ which incorporates unitarity corrections and contains the Gribov-Levin-Ryskin equation as a limiting case. This equation can be derived within the dipole pomeron approach, where it corresponds to parton recombination in the colour dipole cascade initiated by the virtual photon.
## 4 Diffraction in $`p\overline{p}`$ and $`ep`$ collisions
The confrontation of rapidity gap events at the Tevatron with those at HERA provides an opportunity to learn about the interplay between hard and soft dynamics and about the transition from partons to hadrons in reactions with a hard scale.
While in diffractive DIS there is a factorisation theorem for the inclusive cross section and factorisation is expected to hold for hard diffractive $`\gamma p`$ processes where they are dominated by the pointlike component of the photon, there are theory arguments that in hadron-hadron diffraction factorisation should break down due to interactions between the spectator partons of the participating hadrons. This expectation is borne out in the comparison between HERA and Tevatron data: the presentations by Whitmore and the experiments consolidate previous statements that with diffractive parton distributions which fit the inclusive $`F_2^D`$ and diffractive jet production data at HERA (and which are thus required to contain a significant amount of gluons) predictions for Tevatron processes come out far too big. There are also hints for factorisation breaking in diffractive jet photoproduction in the region of $`x_\gamma `$ where the hadronic component of the photon becomes important .
A way of quantifying the phenomenon of factorisation breaking is the concept of gap survival probability, according to which a potential rapidity gap left by a hard subprocess is filled by hadrons produced in collisions between spectator partons. Gotsman presented a model for the survival probability, where in particular a strong dependence of gap survival on the total energy in the reaction is found.
The experimental observation that the transverse energy spectra in double diffractive, single diffractive and non-diffractive jet production at the Tevatron look very similar supports the picture that in all cases one has to do with the same hard partonic subprocesses, and that it is mainly soft interactions which may or may not destroy the rapidity gap, while not modifying the large-$`E_T`$ spectrum of the jets.
A particular implementation of the idea that hard diffractive and non-diffractive events can be described by a perturbative subprocess and nontrivial dynamics of hadronisation which determines whether there will be a rapidity gap or not, is the soft colour interactions model. Ingelman showed that in this model a number of processes both at the Tevatron and at HERA can be fairly well described, without incurring the huge discrepancies in rates of the factorisation ansatz. The physics assumption underlying soft colour interactions is a rearrangement, before hadronisation, of the colour strings between partons due to their interaction with a colour background field. Ingelman further presented an alternative mechanism based on re-interactions among the strings themselves, with the hypothesis that these interactions tend to minimise the phase space area “swept out” by the strings. This rather simple model is able to give a reasonable description of $`F_2^D`$.
## 5 Light meson production
Exclusive vector meson production has long been a major source of information in diffractive physics. Within perturbative QCD it has been shown that in the Bjorken limit of large $`Q^2`$ at fixed $`x`$ and $`t`$, and for longitudinal polarisation of the initial photon, the amplitude for $`\gamma ^{}pMp^{}`$ factorises into a skewed parton distribution in the proton, a hard parton scattering and the distribution amplitude of the meson $`M`$, cf. Fig. 2. The amplitude for transverse photons should be power suppressed by $`1/Q`$. How far one is from the asymptotic regime can thus in particular be studied with polarisation observables. The data on the ratio $`R=\sigma _L/\sigma _T`$ of cross sections for longitudinal and transverse photons have indicated for some time that there is a substantial amount of transverse cross section even at a $`Q^2`$ of $`10\text{ GeV}^2`$ or more, and the measurements of the full decay angular distributions provide a wealth of information about nonleading twist phenomena and the physics of the photon-$`\rho `$ transition in a perturbative regime. In this sense their importance goes well beyond the statement that $`s`$-channel helicity is not conserved to an accuracy better than some 10%. The simplest descriptions of the $`\rho `$, be it through a distribution amplitude, where the relative transverse momentum between quark and antiquark is integrated out, or as a nonrelativistic bound state where a constituent $`q\overline{q}`$-pair equally shares the meson momentum, are both too simple to describe this process beyond a 10% accuracy, and the data on the $`\rho `$ polarisation density matrix strongly indicate that the inclusion of transverse momentum of the $`q\overline{q}`$-pair is essential. Three implementations of this, using rather different frameworks and physics assumptions, have been presented by Kirschner, Nikolaev and Royen . Within the present experimental errors they can all account for the data, in particular for the pattern of $`s`$-channel helicity violation, but they differ among themselves to an extent that it may in the future be possible to learn which are the adequate degrees of freedom in the transition from $`\gamma ^{}`$ to $`\rho `$.
In diffractive $`\rho `$-production at large $`t`$, already mentioned in connection with the search for the perturbative pomeron , the photon-$`\rho `$ transition can again be studied in a region where one should be able to describe it within perturbative QCD. Also here polarisation observables can provide important insight into the dynamics, and it will be interesting to understand the results on the $`\rho `$ polarisation at high $`t`$ shown by Crittenden .
Polyakov has shown that the leading twist description of $`\rho `$-production can be extended to the production of a $`\pi ^+\pi ^{}`$-pair with invariant mass $`M_{\pi \pi }^2Q^2`$, be it on or off the $`\rho `$-peak. The corresponding generalised distribution amplitude describes how a pion pair is formed out of a fast-moving $`q\overline{q}`$-pair, and is related by crossing with the parton distributions of the pion. It offers a new way to look at the distribution amplitude of a resonance, and also shows that in the factorising regime it is not necessary to extract a “$`\rho `$-signal” from the pion invariant mass spectrum in order to study, say, the $`x_{IP}`$\- and $`t`$-dependence of the process and the physics of skewed parton distributions.
The cornerstone of understanding diffraction in QCD is the gluon exchange picture, which has as a natural implication that an odderon should exist as the negative charge conjugation partner of the pomeron. No experimental evidence for this object exists so far. Exclusive production of a pseudoscalar instead of a vector meson offers a way to look for odderon exchange in $`ep`$ collisions. Using the description of high-energy scattering developed by the Heidelberg group, which accounts quite well for data on elastic hadron-hadron scattering and exclusive $`\rho `$-production, Berger has presented an estimate of the cross section for photoproduction of a $`\pi ^0`$, going along with proton dissociation. He found rates that should make it possible to discover the odderon at HERA.
## 6 Heavy meson production and skewed parton distributions
Since the first data on exclusive $`\mathrm{{\rm Y}}`$ photoproduction have been presented a year ago there have been improvements in the theory of this process, for which at the time predictions were far below the measured cross sections (while with the same model assumptions $`J/\mathrm{\Psi }`$-production could be described rather well). A number of simplifying assumptions had to be refined, and the presentations by McDermott and Teubner showed that while differing in details both groups obtain cross sections in fair agreement with the data. Points of debate are mainly the choice of factorisation scale in the gluon distribution, and the question of how good a nonrelativistic approximation is for the $`\mathrm{{\rm Y}}`$ wave function (in addition, Teubner et al. also give a result based on parton-hadron duality). An effect which increases the cross section compared with the ”naive” result is the inclusion of the real part of the scattering amplitude, whose importance is related with the very steep energy dependence of the skewed gluon distribution at the high factorisation scale provided by the $`\mathrm{{\rm Y}}`$ mass. Perhaps even more spectacular is the effect of the “skewedness” in the gluon distribution, i.e. the difference between the momentum fractions $`x_1`$ and $`x_2`$ of the two exchanged gluons (cf. Fig. 2), which is fixed at $`x_1x_2=(M_V^2+Q^2)/W^2`$ for the production of a meson with mass $`M_V`$. Estimations find that due to the large mass of the $`\mathrm{{\rm Y}}`$ the effect of this asymmetry amounts to a factor of 2 to 3 in the cross section, compared to just approximating the skewed gluon distribution by the ordinary one, where $`x_1=x_2`$.
Important theory progress has been made in the understanding of skewed parton distributions, in particular concerning their evolution (cf. also the presentations in the spin working group of this workshop). Martin has given arguments why for small $`x_1`$ and $`x_1x_2`$ the effect of the asymmetry between $`x_1`$ and $`x_2`$ at a low factorisation scale becomes more and more washed out as one evolves to large scales, so that the asymmetry there becomes increasingly dominated by the dynamics of the evolution. One thus expects to obtain a good approximation of skewed distributions at a high factorisation scale by evolving them from a low scale, approximating the skewed distributions at the starting scale by the ordinary ones in an appropriate manner. Different ways of performing this approximation have been presented by Martin and by Golec-Biernat . This procedure then relates skewed and ordinary distributions in a nontrivial but controlled way.
In the small-$`x`$ regime the measurement of skewed parton (mainly gluon) distributions may in such a way be used to obtain information on the ordinary gluon distribution. For larger $`x`$, where also the asymmetry $`x_1x_2`$ is larger, one can expect that the skewed quantities (now mainly the quark distributions) will contain nonperturbative information on the proton structure that cannot be obtained from the ordinary ones. The kinematics at HERMES allows one to study this regime, and the first comparison of exclusive $`\rho `$-production data with an estimate based on skewed quark distributions is encouraging for the applicability of this description at lower energies.
Freund has emphasised that the most direct information on skewed distributions may be obtained in deeply virtual Compton scattering, $`\gamma ^{}p\gamma p`$. On one hand there is no second nonperturbative unknown like a meson wave function, and on the other hand the interference of Compton scattering with the Bethe-Heitler process in $`epe\gamma p`$ offers a possibility to measure the new distributions at *amplitude* level. In particular the different ways how the skewed distributions enter in the real and imaginary parts of the Compton amplitude contains valuable information. The interference term is accessible through an azimuthal asymmetry, and even more directly through the asymmetries in the beam lepton charge ($`e^+`$ vs. $`e^{}`$) or polarisation.
With $`\mathrm{{\rm Y}}`$-production providing probably the first evidence for nontrivial effects of skewedness, it can be hoped that future data on various processes will enable us to make use of skewed parton distributions as an additional tool to study hadron structure.
## Acknowledgements
It is a pleasure to thank many participants of our working group for discussions, my co-conveners for the pleasant collaboration, and the organisers for the smooth running of this workshop. I am grateful to M. McDermott for making available the “diffractive DIS convention summary” , on which the present appendix is based. Special thanks are due to W. Buchmüller for many discussions, and to T. Teubner for a careful reading of the manuscript.
## Appendix
This appendix gives some commonly used notation in diffractive DIS. Four-momenta are defined in Fig. 3.
* General DIS variables:
$`Q^2`$ $`=`$ $`q^2=(kk^{})^2`$
$`W^2`$ $`=`$ $`(p+q)^2`$
$`x`$ $`=`$ $`{\displaystyle \frac{Q^2}{2pq}}={\displaystyle \frac{Q^2}{W^2+Q^2m_p^2}}`$
$`s`$ $`=`$ $`(p+k)^2`$
$`y`$ $`=`$ $`{\displaystyle \frac{qp}{kp}}={\displaystyle \frac{W^2+Q^2m_p^2}{sm_p^2}}`$
* Diffractive DIS variables:
$`t`$ $`=`$ $`(pp^{})^2`$
$`M_X^2`$ $`=`$ $`(pp^{}+q)^2`$
$`M_Y^2`$ $`=`$ $`p^2`$
$`x_{IP}`$ $`=`$ $`{\displaystyle \frac{(pp^{})q}{pq}}={\displaystyle \frac{M_X^2+Q^2t}{W^2+Q^2m_p^2}}`$
$`\beta `$ $`=`$ $`{\displaystyle \frac{Q^2}{2(pp^{})q}}={\displaystyle \frac{Q^2}{M_X^2+Q^2t}}={\displaystyle \frac{x}{x_{IP}}}`$
It is also common to write $`\xi `$ instead of $`x_{IP}`$.
* Diffractive structure functions:
$`{\displaystyle \frac{d^4\sigma (epeXY)}{dxdQ^2dx_{IP}dt}}={\displaystyle \frac{4\pi \alpha _{em}^2}{xQ^4}}`$
$`\times \left[\left(1y+{\displaystyle \frac{y^2}{2}}\right)F_2^{D(4)}{\displaystyle \frac{y^2}{2}}F_L^{D(4)}\right]`$
$$F_2^{D(4)}=F_T^{D(4)}+F_L^{D(4)}$$
* $`t`$-integrated diffractive structure functions:
$`F_i^{D(3)}(x_{IP},\beta ,Q^2)=`$
$`{\displaystyle _{|t|_{\mathrm{𝑚𝑖𝑛}}}^{|t|_{\mathrm{𝑚𝑎𝑥}}}}d|t|F_i^{D(4)}(x_{IP},\beta ,Q^2,t)`$
with $`i=2,T,L`$. Here $`|t|_{\mathrm{𝑚𝑖𝑛}}`$ is the lower kinematic limit of $`|t|`$ and $`|t|_{\mathrm{𝑚𝑎𝑥}}`$ has to be specified.
|
no-problem/9906/astro-ph9906279.html
|
ar5iv
|
text
|
# On matter-antimatter separation in open relativistic material system
## 1 Formulation of the problem
A matter-antimatter mix is generally expected to annihilate. However, under certain conditions a trend might appear, which leads to a matter-antimatter space separation. The problem was discussed elsewhere (see, for example ). In the present work an original approach is discussed. Our formulation of the problem is as follows.
* An open system of material objects and radiation is considered in a flat space-time. There is no phisical cause for preference in choosing inertial reference frames. Thus, the system is assumed to be baryon symmetric, uniform and isotropic on average in any reference frame. It leads to specific requirement for a matter coordinate-momentum distribution, as shown further.
* Material objects are characterized by a mass distribution (in a unit volume) $`N(m)`$, which is the same for matter and antimatter. Baryon charge is conserved. In addition, the system is characterized by a uniform coordinate-momentum distribution (in a unit volume).
$$f(\mathrm{x},\mathrm{p})=Consd\mathrm{x}d\mathrm{p},$$
(1)
where $`\mathrm{x}`$, $`\mathrm{p}`$ — space and momentum 3-vectors.
* All known physical interactions are allowed between system constituents.
The above formulation of the problem reflects the idea of a generalized matter transport equation. In the present work a treatment of the problem will be as much simplified as possible to concentrate our attention on the question of matter-antimatter separation. We do not specify the physical nature of material objects. Any concrete astrophysical picture of a matter structure might be embedded, if needed, into the frame under consideration. In general, an object is meant to be a free microscopic or macroscopic particle, free solid body or a gravitationally linked system (particle cloud, multi-body association, galaxy etc). We assume that a free object is made of either matter or antimatter. Hence, an object is characterized by its mass at rest and baryonic type, its internal evolution being ignored. For the purpose of this work it seems sufficient to consider object-object interaction in general terms of a random collision followed by a formation of a compound system, which may disintegrate into new objects with channel probabilities of fragmentation, annihilation, and merge processes. We assume that matter-antimatter annihilation results in gamma radiation being a source of a following pair production. It means that annihilated matter is regenerated in the “first group” of a mass distribution. The “first group” may be referred, in principle, to elementary particles as a “seed matter” and plays a role of a source term in a kinetic equation.
Next comments concern relativistic properties of an open system. First of all, we need to explain the term “relativistic material system”. A uniform isotropic matter distribution in a flat space is characterized by a uniform coordinate-momentum distribution function (1). It is Lorentz-invariant, that is unchanged if measured in any inertial reference frame. Such a state of matter can be treated as a “maximal chaos” in terms of Bayesian approach for description of relativistic gas . Actually, this is an approximation of non-interacting particles. It is worth noting that the relativistic gas model has nothing in common with Friedmann-Lemaitre expanding universe model , which treats matter like dust-like matter-made particles with small relative velocities in an observer’s vicinity. The expanding universe itself plays a role of the absolute reference frame. In our approach an opportunity of introducing the absolute reference frame is denied. However, the above approximation of non-interacting particles leads to an infinite energy density. In a more rigorous relativistic gas model one has to take into account that physical processes locally observed under conditions of an open system should be characterized by retarded casual connections with the rest of the space, and inertial systems must be referred to a limited space-time volume depending on how precise they are needed to be defined. Hence, a “realistic” coordinate-momentum distribution must be characterized by some space-energy correlation and must have a smooth cut-off at however high energy range, or an effective temperature parameter. Now any local energy density will be found limited and invariant in a broad set of inertial systems (the higher temperature, the broader set). We assume that the temperature is well above the threshold needed for pair production process being effective.
Under above conditions annihilation and pair production processes have to be balanced, for there is no energy dissipation in the open system. In other words, the system is expected to be in a self-sustained baryon symmetric state. If a mass distribution function is not degenerated into the “first source group” one can state that a matter-antimatter space separation takes place. The mass distribution function should be found from a material balance (kinetic) equation.
## 2 Model kinetic equation
Let a mass distribution function for like matter objects (in a unit volume) be given in a mass group form for matter $`m_i`$ or antimatter $`m_i^{}`$:
$$N_i=N(m_i)orN_i^{}=N(m_i^{})$$
(2)
with a total mass conserved:
$$M=m_iN_iM^{}=m_i^{}N_i^{}M=M^{}.$$
(3)
Next we use so called one velocity approximation, that is an equation being averaged over momentum distribution. Let us introduce a generation rate $`G_i`$ of matter-made objects in a group $`i`$ in a result of matter-matter $`jk`$ type object collisions. If widths of groups are narrow enough we may ignore interactions inside groups.
$$G_i=\underset{j}{}\underset{k}{}<N_jN_kv\sigma (m_j,m_k)K([m_j,m_k]m_i)>,$$
(4)
here the brackets $`<>`$ symbolize averaging over momentum distribution, $`v`$ is a relative velocity, $`\sigma `$ is an object-object collision cross-section, and $`K([m_j,m_k]m_i)`$ is a “channel function”, describing a probability of an object to appear in a group $`i`$ in a result of decay of compound system \[$`m_j`$,$`m_k`$\]. Masses are conserved due to a proper K-function normalization.
Similarly, an object generation rate $`G_i^{}`$ resulting in matter-antimatter type collision (\* symbolize “antimatter participant”, as before) may be given:
$$G_i^{}=\underset{j}{}\underset{k^{}}{}<N_jN_k^{}v\sigma (m_j,m_k^{})K([m_j,m_k^{}]m_i)>.$$
(5)
In this case, the K-function is normalized to ensure a correct mass balance, annihilated mass in a form of gamma-radiation being taken into account.
Choosing a proper summation rule we come to expressions for a removal rate $`R_i`$ and $`R_i^{}`$ in matter-matter and matter-antimatter types of collision, correspondingly.
$$R_i=\underset{j}{}\underset{k}{}<N_iN_jv\sigma (m_i,m_j)K([m_i,m_j]m_k)>$$
(6)
$$R_i^{}=\underset{j}{\overset{}{}}\underset{k}{}<N_iN_j^{}v\sigma (m_i,m_j^{})K([m_i,m_j^{}]m_k)>$$
(7)
As is said above, an annihilation rate $`G_\gamma `$ is followed by a matter recreation in the first group, $`G_1(m_r)`$, masses of radiation $`m_\gamma `$ and recreated matter $`m_r`$ being in balance:
$$G_1(m_r)=G_\gamma =\underset{j}{}\underset{k^{}}{}<N_jN_k^{}v\sigma (m_j,m_k^{})K([m_j,m_k^{}]m_\gamma )>.$$
(8)
Finally, we have a stationary kinetic equation:
$$G_i+G_i^{}+R_i+R_i^{}=\delta _{1i}G_1(m_r)(i=1,2,3,\mathrm{}),$$
(9)
where $`\delta _{1i}=1`$ for $`i=1`$, and $`\delta _{1i}=0`$ for $`i1`$.
Due to a symmetry of physical properties of matter and antimatter one can get the same equation starting considering generation/removal rates for antimatter-made objects. The equation (9) may be written in an integral form. It is easy to see that the equation has a non-trivial solution if a merge channel providing a transfer from a mass group $`i`$ to a mass group $`i+1`$ is open. Since microscopic and macroscopic merging processes are known from conventional Physics we may conclude that the above equation has a physical meaning.
## 3 Discussion
To demonstrate a solution we realized the Monte Carlo method for K-functions reduced to delta-functions. In other words, the fragmentation process has been ignored. We checked that it does not appreciably influence the form of solution though a computing time significantly increases. Obviously, a solution depends on a collision cross-section as a function of an object mass. Four variants were studied: constant cross-section, cross-section, proportional to the mass of a target object, cross-section proportional to both the mass of target and the mass of incident object, finally cross-section proportional to the squared mass of the target object. The last two variants were expected to have only slightly different solutions. The variants seem to include typical types of mass-dependent cross-sections of body-body physical interactions. However, concrete energy-dependent cross-sections for specific interactions may differ significantly in magnitude in different mass intervals. We do not know to what extent the solutions are influenced by “one velocity approximation”. This important question needs a special study on open relativistic matter system properties, their dependence on energy density, in particular. Besides, taking into account an object internal evolution might be important also if the results were used for their speculative applications concerning astrophysical problems . All these questions are out of the scope of the present work.
The main conclusion we would like to emphasize is that an open steady-state isotropic system consisted of baryon symmetric uniformly distributed matter should be characterized by a relativistic momentum distribution with an effective high temperature parameter. Subsequently, a space matter-antimatter separation takes place resulted from random interactions within the system. The possible scale of separation may be seen from the statistical assessments of stationary mass distribution functions in Fig. 1. The picture gives us the idea of a solution of a simple model kinetic equation describing the above system: the solution may be approximated by a function $`1/m^n`$ with $`n`$ ranging between 2 and 3.5, depending on model parameters. Apparently, there might be a trend of decreasing $`n`$ for asymptotic solution. The latter should be found by analytical method. The result shows that an open baryon symmetric matter system reveals a structure containing matter or antimatter-made objects of any size. There is no material limit for their construction because of the system being unlimited in space. The bigger size (age) objects have, the less their population is. Further computer simulations with more sophisticated models might show more physically meaningful features of the baryon symmetric open system.
|
no-problem/9906/astro-ph9906123.html
|
ar5iv
|
text
|
# Radiative Levitation in Hot Horizontal Branch Stars
## 1 Introduction
Grundahl et al. (1999) recently suggested that radiative levitation of heavy elements to supersolar abundances occurs for globular cluster HB stars hotter than about $`11,500`$ K. This suggestion was based on the following evidence:
* Grundahl et al. obtained Strömgren $`u`$, $`uy`$ CMDs of 14 globular clusters and found evidence for a ubiquitous “jump” at about $`T_{\mathrm{eff}}`$ = $`11,500\mathrm{K}`$, in the sense that stars hotter than this temperature are about 0.25 mag brighter in Strömgren $`u`$ than predicted by HB models. This jump is not present in ultraviolet CMDs of globular clusters obtained with HST or the Ultraviolet Imaging Telescope (UIT), which suggests that it is due to an atmosphere effect (causing a redistribution in the flux), rather than to a change in the bolometric luminosity. (The presence of a luminosity jump is difficult to discern at wavelengths longer than Strömgren $`u`$ because the HB becomes “vertical” in the CMD, and the effects of changes in luminosity cannot be distinguished from changes in temperature.)
* As summarized by Moehler (1999), the derived gravities of HB stars with $`11,500\mathrm{K}<T_{\mathrm{eff}}<20,000`$ K are consistently found to be about 0.2 dex lower than predicted by canonical HB models. Grundahl et al. performed simple experiments with Kurucz model atmospheres which suggested that both the gravity anomaly and the brightening in Strömgren $`u`$, could be explained if the HB photosphere had a supersolar metallicity, rather than the metallicity of the cluster.
* Hot HB stars in globular clusters are observed to show strong depletions of helium (e.g. Moehler et al. 1997). The early theoretical work of Michaud et al. (1983) suggested that if the HB atmosphere is stable enough to allow for the gravitational settling of helium, then overabundances of heavy elements by factors of $`10^310^4`$ might be expected due to radiative levitation.
* Helium-depleted field HB stars hotter than $`11,500\mathrm{K}`$ show unusual abundance patterns, and have higher iron abundances than observed in cooler HB stars. In particular, in the well-studied field HB star Feige 86 (Castelli et al. 1997), the elements lighter than sulfur are depleted (with the exception of phosphorus, with \[P/H\] = +1.8), while iron-peak elements are slightly supersolar (\[Fe/H\] = +0.4), and the heavy metals are strongly overabundant (e.g. \[Au/H\] = +4.0). Glaspey et al. (1989) reported an overabundance of iron by a factor of 50 (and a helium depletion) in a hot HB star in NGC 6752 (CL 1083) with $`T_{\mathrm{eff}}=16,000`$ K.
With the exception of the single star in NGC 6752 studied (at low S/N) by Glaspey et al., all the evidence for an iron enhancement in globular cluster HB stars discussed by Grundahl was indirect. However, independent of Grundahl et al. work, Behr et al. (1999) were using the Keck HIRES echelle spectrograph to study abundances in 13 hot HB stars in M13. Their results provide striking direct evidence for radiative levitation in globular cluster hot HB stars, and for abundance patterns similar to those observed in Feige 86. The iron abundances in the M13 HB stars hotter than $`11,500`$ K are about $`+2.0`$ dex higher than in stars cooler than the jump temperature. Phosphorus is also enhanced (\[P/H\] = +1.0), but the magnesium abundances (\[Mg/H\] $`1.5`$) show no change across the jump temperature. Subsequently, Moehler et al. (1999) used medium ($`2.6`$ Å) resolution spectroscopy to show the presence of a similar jump to supersolar iron abundances (with Mg remaining at the cluster abundance) in NGC 6752. In addition, Moehler et al. were able to explicitly show that most – though not all – of the discrepancy between the derived gravities and canonical models could be removed if the Balmer lines were analyzed using appropriately metal-rich atmospheres. We hope to further explore the appropriate model atmospheres for hot HB stars using our Cycle 8 HST program (8256) to obtain STIS ultraviolet spectra of nine HB stars in NGC 6752 spanning the temperature range of $`10,000\mathrm{K}<T_{\mathrm{eff}}<24,000`$ K.
The sample of stars with abundances determined by Behr et al. or Moehler et al. does not include any stars hotter than 20,000 K. Peculiar abundance patterns are known to exist in the sdB stars (e.g. Lamontagne et al. 1987), and pulsation studies indicate the presence of radiative levitation of iron within the envelope (Charpinet et al. 1997). However, the photometric discrepancy with canonical models in Strömgren $`u`$ discussed by Grundahl et al. decreases for $`T_{\mathrm{eff}}`$$`>`$ 20,000 K, and the field sdB stars do not show the strong overabundances seen in cooler (and lower gravity) HB stars, such as Feige 86. Thus, we suggest that the most dramatic effects of radiative levitation occur in the temperature range $`11,500\mathrm{K}<T_{\mathrm{eff}}<20,000`$ K.
The presence of strong abundance anomalies in hot HB stars complicates the derivation of accurate luminosities, gravities, and masses, and the analysis of the integrated ultraviolet spectra of old, metal-poor stellar populations. But these abundance anomalies also mean that the HB of globular clusters will likely provide a superb laboratory for studying radiative levitation and diffusion processes in the outer atmospheres of hot stars. The HB stars in a globular cluster have known initial abundances, and provide a populous sample for studying the effects of temperature, initial metallicity, and rotation on the photospheric abundances resulting from radiative levitation. (The uniformity of the Strömgren $`u`$ jump indicates that the effects of radiative levitation must be rapid compared to the HB lifetime of about $`10^8`$ yr.)
In analogy to the HgMn stars (e.g. Leckrone et al. 1999), the rough abundance pattern observed in hot HB stars can be understood as being due to saturation of radiative forces in the more abundant lighter elements. However other effects reported by Behr et al. and Moehler et al. have no ready explanation, including the increased helium depletion with increasing $`T_{\mathrm{eff}}`$, and in particular, the abruptness of the transition to supersolar iron abundances at 11,500 K. An important clue might be provided by the observation by Behr et al. of a low rotation ($`v\mathrm{sin}i`$$`<6`$ km s<sup>-1</sup>) in stars hotter than the jump, since rotationally-induced turbulence can inhibit diffusion processes. But this only moves the problem one level deeper, since the origin of the abrupt change in rotational velocities would still be unknown.
## 2 Implications
What are the implications of the discovery of radiative levitation for the outstanding problems in the studies of globular cluster HB stars? First, the discovery of radiative levitation has no direct implications for the age or distance scale of globular clusters, since hot HB stars have not been used as absolute calibrators. Similarly, the discovery of radiative levitation in hot HB stars does not answer the question of why some HB stars are hot, or, more generally, on the origin of the HB morphology (e.g. the second parameter problem).
Another outstanding problem in HB studies is the origin of “gaps” in the temperature distribution of HB stars, and in this case, radiative levitation might have a contributing role for any gap located near $`11,500`$ K (such as the G1 gap discussed by Ferraro et al. 1998). In certain bandpasses, the sudden change in photospheric abundances near this temperature might shift the positions of stars along the HB. For example, in a $`y`$, $`uy`$ diagram, the brightening in Strömgren $`u`$ with the onset of radiative levitation could induce a gap in the HB distribution at $`11,500`$ K.
Although the dominant implications of radiative levitation are for the HB stellar atmosphere, in principle, the changes in the radial chemical profile could alter the bolometric luminosities and lifetimes computed in HB interior models. This effect is likely to be small, although Seaton (1997) does warn that the surface abundance changes patterns in the HgMn stars are probably a manifestation of radiative diffusion processes deep in the stellar envelope, which (through the modified opacities) can alter the stellar structure.
Hot HB stars are one of the few potential ultraviolet sources in an old stellar population, and thus the discovery of radiative levitation has significant implications for the ultraviolet spectral synthesis of old, metal-poor systems. The implications are less for a metal-rich population for two reasons: first, the metallicity enhancement due to radiative levitation is less pronounced for metal-rich stars, and second, fewer hot HB stars are expected in a metal-rich system, since the HB is expected to bifurcate into either very hot ($`>20,000`$ K) or cool HB stars (Dorman, Rood, & O’Connell 1995). Spectral synthesis models of the ultraviolet upturn in elliptical galaxies computed in either the metal-poor or the mixed metallicity scenarios discussed in Yi et al. (1999), should probably be performed using metal-rich atmospheres for metal-poor HB stars with $`T_{\mathrm{eff}}>11,500`$ K. In particular, the use of metal-rich model atmospheres might help suppress the model flux in the 1800 – 2500 Å spectral region, and improve the agreement of the metal-poor models shown in Yi et al. with the observed spectra of ellipticals.
Finally, the empirical finding of Behr et al. (1999) and Moehler et al. (1999) that Mg abundances are unaltered by diffusion processes suggests that Mg is the most reliable abundance indicator for field hot HB stars. For example, while most the of the abundances derived in Feige 86 by Castelli et al. (1997) have been modified by diffusion processes, the magnesium abundance (\[Mg/H\] = –0.64) probably provides a good measure of the stellar metallicity.
###### Acknowledgements.
I thank my collaborators on this topic, including M. Catelan, F. Grundahl, T. Lanz, S. Moehler, C. Proffitt, and A. Sweigart
|
no-problem/9906/astro-ph9906233.html
|
ar5iv
|
text
|
# Bispectrum speckle interferometry of the Orion Trapezium stars: detection of a close (33 mas) companion of Θ¹Ori C Based on data collected at the SAO 6 m telescope in Russia.
## 1 Introduction
The Orion Nebula cluster is one of the most prominent and nearby ($`D450`$ pc) star forming regions (for a review see Genzel & Stutzki 1989). Its core contains a very dense cluster of young ($`1\times 10^6`$ yr) stars (cf. Herbig & Terndrup 1986; McCaughrean & Stauffer 1994; Hillenbrand 1997). The Trapezium ($`\mathrm{\Theta }^1\mathrm{Ori}`$ ABCD), the system of the four most massive and luminous O-type and early B-type stars, is located in the center of the cluster. The strong stellar wind and the ionizing radiation of $`\mathrm{\Theta }^1\mathrm{Ori}`$ C has strong effects on the surrounding cloud material (Bally et al. 1998; see also Richling & Yorke 1998).
Petr et al. (1998; P98 hereafter) presented the results of 130 mas resolution near-infrared speckle holographic observations of the Trapezium cluster core, in which they could detect sub-arcsecond companions of the two Trapezium stars $`\mathrm{\Theta }^1\mathrm{Ori}`$ A and $`\mathrm{\Theta }^1\mathrm{Ori}`$ B. Simon et al. (1999) reported the detection of an additional, very faint companion of $`\mathrm{\Theta }^1\mathrm{Ori}`$ B. In this paper we present the first near-infrared bispectrum speckle interferometry observations with diffraction-limited resolution of 57 mas in the $`H`$-band and 76 mas in the $`K`$-band.
## 2 Speckle observations and results
The speckle interferograms of $`\mathrm{\Theta }^1\mathrm{Ori}`$ A, B, C, and D were obtained with the 6 m telescope at the Special Astrophysical Observatory (SAO) in Russia on Oct. 14, 1997 ($`H`$-band) and Nov. 3, 1998 ($`K`$-band). Diffraction-limited images were reconstructed from the speckle data using the bispectrum speckle interferometry method (Weigelt 1977; Lohmann et al. 1983). The modulus of the object Fourier transform was determined with the speckle interferometry method (Labeyrie 1970). The speckle transfer functions were derived from speckle interferograms of the unresolved star $`\mathrm{\Theta }^1\mathrm{Ori}`$ E. Figures 1 and 2 show the power spectra and images of $`\mathrm{\Theta }^1\mathrm{Ori}`$ A, B, and C. The observational parameters and the properties of the resolved stars are summarized in Table 1. The flux ratios were determined by fitting cosine functions to the power spectra.
$`\mathrm{\Theta }^1Ori`$ A: The companion A<sub>2</sub> of the primary star $`\mathrm{\Theta }^1\mathrm{Ori}`$ A<sub>1</sub> (detected by P98) is clearly visible in our images.
$`\mathrm{\Theta }^1Ori`$ B: The two companions $`\mathrm{\Theta }^1\mathrm{Ori}`$ B<sub>2,3</sub> are clearly resolved, confirming the detection by P98. In our images we cannot see the new faint component detected by Simon et al. (1999) because it is just below our detection limit.
$`\mathrm{\Theta }^1Ori`$ C: Our power spectra and images of $`\mathrm{\Theta }^1\mathrm{Ori}`$ C show a companion C<sub>2</sub> with a separation of $`(33\pm 5)`$ mas from the primary C<sub>1</sub> ($`H`$-band; see Tab. 1). This is the first detection of a companion of $`\mathrm{\Theta }^1\mathrm{Ori}`$ C.
$`\mathrm{\Theta }^1Ori`$ D: We find no indication for a companion of D.
From the average surface density of stars in the Trapezium cluster reported by Simon et al. (1999) we estimate that the probability of finding a star with $`K<10`$ within $`1^{\prime \prime }`$ from a given position is $`<1\%`$. This suggests that the visual companions we observe actually are companions of the respective primaries and not only chance projections of unrelated stars.
## 3 Stellar properties of the companions
In order to obtain information about the physical properties of the companions from our speckle results, we have used the photometric and spectroscopic data and stellar parameters compiled by Hillenbrand (1997) and Hillenbrand et al. (1998). From the known system magnitudes and the flux ratios determined in the speckle images we have computed the $`K`$-band magnitudes and the $`HK`$ colors of the speckle companions. These data can be used to estimate the luminosity and the effective temperature of the stars: the $`K`$-band magnitude yields the stellar luminosity as a function of the stellar temperature (using the compilation of intrinsic $`VK`$ colors and bolometric corrections of Kenyon & Hartmann 1995), and the $`HK`$ color can be transformed into the stellar temperature. The resulting parameters can then be employed to estimate stellar masses and ages (cf. Fig. 3 and Table 2).
For $`\mathrm{\Theta }^1\mathrm{Ori}`$ C<sub>2</sub> we find $`K=5.95\pm 0.11`$ and $`HK=0.24\pm 0.10`$. Since the extinction<sup>1</sup><sup>1</sup>1We assume that the extinction to the companion is the same as to the primary. Although one might expect the companion, being a very young stellar object, to be surrounded by circumstellar material which might cause additional extinction, we believe that the strong radiation and wind of $`\mathrm{\Theta }^1\mathrm{Ori}`$ C would have dispersed any diffuse material in its immediate vicinity very quickly. of $`A_V=1.8`$ corresponds to $`E(HK)=0.11`$ (cf. Rieke & Lebofsky 1985), the error ranges of the dereddened magnitude and color are $`K_0=[5.675.83]`$ and $`(HK)_0=[0.03\mathrm{\hspace{0.17em}0.24}]`$. Since the magnitude range defines a band in the HR-diagram and the color range corresponds to a range of temperatures of $`[35507000]`$ K, these data define the grey shaded band in Fig. 3. The comparison of this band with theoretical PMS tracks suggests that the companion is a very young intermediate- or low-mass ($`M6M_{}`$) star.
For $`\mathrm{\Theta }^1\mathrm{Ori}`$ A<sub>2</sub> a similar computation (using the $`H`$-band flux ratio from P98) yields $`K_0=[7.187.25]`$ and $`(HK)_0=[0.01\mathrm{\hspace{0.17em}0.09}]`$. The corresponding band in the HRD suggests a mass of $`M35M_{}`$.
The lack of photometric $`H`$-band data for the components of $`\mathrm{\Theta }^1\mathrm{Ori}`$ B prevents us from placing them into the HRD. Nevertheless, we can derive upper limits for the masses from the $`K`$-band flux ratios if we assume that the stars lie at or above, but not below, the main-sequence. The corresponding limits are $`M<5M_{}`$ for B<sub>2</sub>, $`M<3.5M_{}`$ for B<sub>3</sub>, and $`M<2M_{}`$ for B<sub>4</sub> (the faint companion detected by Simon et al. 1999).
## 4 Multiplicity of the massive Trapezium stars
Several recent studies (P98; Simon et al. 1999) have concordantly found that the binary frequency of the low-mass stars in the Orion nebula cluster (ONC) is comparable to that of solar type field stars, which is about 60% with a median number of companions per primary of about 0.5 (c.f. Duquennoy & Mayor 1991; Fischer & Marcy 1992). The binary frequency of O-type and early B-type field stars seems to be similar (cf. Abt 1983; Mason et al. 1998).
Our detection of $`\mathrm{\Theta }^1\mathrm{Ori}`$ C<sub>2</sub> increases the number of known companions to the four Trapezium stars to 7. The average number of at least 1.75 companions per primary among the high-mass Trapezium stars is clearly higher than the corresponding number for the low-mass stars in the ONC. A similar trend has been found by Abt et al. (1991) and Morrell & Levato (1991): most of the spectroscopic binaries in the ONC are among the O- and early B-type stars, and much less frequent among the later B- and A-type stars. This finding suggests different formation mechanisms for the high-mass and low-mass multiple systems. This is consistent with the recent results of Bonnell et al. (1998) who assumed that high-mass stars form through accretion-induced collisions of protostars. Their theory predicts that close binary systems should be very common among the massive stars. This is supported by our results.
|
no-problem/9906/cond-mat9906261.html
|
ar5iv
|
text
|
# Phase Diagram of Integer Quantum Hall Effect
\[
## Abstract
The phase diagram of integer quantum Hall effect is numerically determined in the tight-binding model, which can account for overall features of recently obtained experimental phase diagram with direct transitions from high plateau states to the insulator. In particular, the quantum Hall plateaus are terminated by two distinct insulating regimes, characterized by the Hall resistance with classic and quantized values, respectively, which is also in good agreement with experiments. The physical origin of this peculiar phase diagram is discussed.
today \]
The previously proposed theoretical global phase diagram (GPD) of the quantum Hall (QH) effect predicts that only the $`\nu =1`$ QH liquid state is adjacent to the insulator, while higher QH plateau states do not neighbor with the insulator as schematically illustrated in Fig. 1(a). As a result, a direct transition from higher QH plateau ($`\nu >1`$) states to the insulator is prohibited. However, a series of recent experimental measurements have indicated a phase diagram with qualitatively different topology. Most recently, it has been established experimentally by Hilke et al. that direct transitions from $`\nu =1,2,\mathrm{},7`$ to the insulating phase can all take place as clearly shown in Fig. 1(b). These experiments have challenged the basic theoretical understanding of QH systems at weak magnetic field.
These QH liquid-insulator transitions have also exhibited distinct properties in different regimes. For example, the Hall resistance $`\rho _{xy}`$ remains quantized at $`h/e^2`$ in the transition region between the $`\nu =1`$ QH state and the insulator at strong magnetic field or low Landau-level (LL) filling factor $`n_\nu `$, even though the longitudinal resistance $`\rho _{xx}`$ already increases almost by one order of magnitude above the value $`h/e^2`$ in the same region denoted as II in Fig. 1(b). By contrast, at weak magnetic field or higher $`n_\nu `$, $`\rho _{xy}`$ near the critical region of the $`\nu >1`$ QH states to insulator transition becomes $`n_\nu `$-dependent and is very close to the classic value $`B/nec`$ rather than the quantized value. In fact, such a classic behavior of $`\rho _{xy}`$ has been found to persist well into an insulating regime designated by I in Fig, 1(b), which suggests that there are two distinct insulating regimes surrounding the QH liquid states as opposed to the single one.
Direct transitions between the $`\nu >1`$ QH states and the insulator have been already found in the tight-binding model (TBM) based on numerical calculations. But the relevance of such a lattice model to the experiment is still controversial, since the strength of the magnetic field usually cannot be reduced weak enough to directly simulate the realistic situation within the numerical capacity. Therefore, it is particularly important to identify the overall phase diagram and corresponding transport properties in such a model in order to understand the underlying physics and establish a real connection with the experiments.
In this Letter, we obtain, for the first time, a numerical “global phase diagram” for the integer QH effect based on the TBM and the results are summarized in Fig. 1(c). The topology of the phase diagram is strikingly similar to the experimental one shown in Fig. 1(b), and in particular the insulating phase is indeed divided into two regimes: In a strong disorder and low magnetic field region (Insulator I), the Hall resistance follows a classic value while the longitudinal resistance shows insulating behavior; in a weak disorder and high magnetic field region (Insulator II), we find that $`\rho _{xy}`$ remains at the quantized value $`h/e^2`$ near the transition region even when $`\rho _{xx}`$ increases up to $`8h/e^2`$. Both are in good agreement with the aforementioned experiments. Finally we provide a physical interpretation of the nature of the present non-float-up phase diagram based on the calculation of the equilibrium edge current.
The phase diagram in Fig. 1(c) can be determined by following the trace of extended levels by continuously tuning the disorder strength or magnetic field $`B`$. The position of each extended level forms a boundary which separates a given QH plateau state from another QH state or the insulating phase, and can be identified by the peak (which is sample size independent) of the density of states carrying nonzero Chern number, calculated based on the TBM Hamiltonian $`H=_{<ij>}e^{ia_{ij}}c_i^+c_j+H.c.+_iw_ic_i^+c_i`$, which is characterized by two parameters: the magnetic flux per plaquette $`\varphi =_{\mathrm{}}a_{ij}=2\pi /M`$ and the disorder strength $`W`$ of the random potential $`w_i`$: $`|w_i|<W/2`$. The result shown in Fig. 1(c) is calculated at $`M=64`$. In the Chern number calculation, the sample size is up to $`64\times 64`$. The position of an extended level can be equally identified by the peak of the longitudinal conductance $`\sigma _{xx}`$, which coincides with the Chern number result, but this latter method has an advantage as it can be applied to much weaker magnetic fields. At $`M=384`$, the sample size in calculating $`\sigma _{xx}`$ is up to $`L_x=200`$ and $`L_y=10^3M`$ using transfer matrix method. The phase diagram in Fig. 1(c) remains essentially the same as we continuously change the magnetic flux from $`M=8`$ to $`384`$. Note that $`W_c`$ (which depends on the magnetic field B) is the critical value at which the last QH plateau state disappears and the system becomes an insulator.
The similarity between the numerical phase diagram \[Fig. 1(c)\] and the experimental one \[Fig. 1(b)\] is striking. Similar experimental phase diagram is also obtained earlier in Ref.. Several detailed features in Fig. 1(c) are worth mentioning. Firstly, starting from the strong-magnetic-field insulator II and reducing B continuously at a fixed electron density, we obtain numerically a dashed curve $`A`$ shown in Fig. 1(c) which cuts through different phases with a transition pattern $`01230`$. \[As B is reduced, $`W/W_c(B)`$ increases due to the $`B`$-dependence of $`W_c`$\]. Such a scan curve should be equivalent to a constant gate voltage $`V_G`$ line in Fig. 1(b) as the fixed $`V_G`$ means both the disorder strength and electron density are constants. Secondly, one can clearly see that all the higher extended levels as boundaries separating different QH states are almost vertical lines in Fig. 1(c) which do not “float up” much in terms of the LL filling number $`n_\nu `$ at increasing disorder strength. The same non-float-up picture also unequivocally shows in the experimental phase diagram of Fig. 1(b). Only the lowest one which defines the boundary between $`\nu =1`$ and the insulator floats up, also in agreement with an earlier experiment.
The scan curve A in Fig. 1(c) connects insulating regimes at two ends. Let us first focus on the insulating region I which has a boundary neighboring with the high-plateau QH states. Fig. 2 shows the calculated $`\rho _{xx}`$ and $`\rho _{xy}`$ versus $`n_\nu `$ at fixed $`W`$ \[parallel to the scan line B illustrated in Fig. 1(c)\]. The prominent feature in this region is that $`\rho _{xy}`$ follows the classic behavior (the dashed curve in Fig. 2 denotes $`\frac{1}{n_\nu }\frac{h}{e^2}=\frac{B}{nec}`$): In Fig. 2(a), the magnetic filed is fixed at $`M=32`$ while different disorder strengths are considered. Even though $`\rho _{xx}`$ in the inset grows with $`W`$ very quickly, $`\rho _{xy}`$ is insensitive to disorders and remains close to the classical value at $`W=4,5,6`$($`W_c=3.5`$). The finite size effect of $`\rho _{xy}`$ is shown in Fig. 2(b) at $`M=16`$. By increasing the sample length $`L`$ from $`16`$ to $`64`$, one sees that $`\rho _{xy}`$ converges to the dashed curve (the classic value) very quickly whereas $`\rho _{xx}`$ keeps increasing monotonically with the sample size. It is noted that both $`\sigma _{xx}`$ and $`\sigma _{xy}`$ calculated here are for square samples $`L\times L`$ using Landauer and Kubo formula, respectively, and more than $`2000`$ disorder configurations are taken at $`L=64`$ and even more at smaller sample sizes.
Such a classic behavior of $`\rho _{xy}`$ has been extensively observed experimentally in weak magnetic field regime. $`\rho _{xy}1/n_\nu `$ in fact still holds at the critical point between the QH liquid and insulating regime I as previously shown experimentally and numerically. Since LLs are effectively coupled together at weak field and strong disorders, we believe that this phenomenon reflects the fact that the regime I is basically an Anderson insulator: $`\rho _{xy}`$ is always unrenormalized and remains at the classic value. In other words, the insulator I in Fig. 1(c) should continuously evolve into the well-known Anderson insulator at zero magnetic field without changing the classic behavior of $`\rho _{xy}`$ while $`\rho _{xx}`$ is always divergent in the thermodynamic limit at zero temperature.
Now we consider the insulating regime II in Fig. 1(c). Along the scan line C in Fig. 1(c), the results of $`\rho _{xx}`$ and $`\rho _{xy}`$ are presented in Fig. 3(a). It shows that $`\rho _{xy}`$ remains at quantized value $`h/e^2`$ while $`\rho _{xx}`$ arises almost an order of magnitude from the critical value at $`n_{\nu c}`$ into the insulator region. This is in contrast to the aforementioned classic behavior $`\rho _{xy}=\frac{1}{n_\nu }h/e^2`$ in the regime I. Such a quantized $`\rho _{xy}`$ exists in the whole critical region along the boundary between the $`\nu =1`$ QH state and the insulator. The open circles in Fig. 1(c) at $`W/W_c=1`$ lies very close to the boundary of the two insulating regimes as indicated in our numerical calculations (how two regimes exactly cross over will need a more careful study which is beyond the scope of the present paper). It is noted that in the transition region where $`\rho _{xy}=h/e^2`$ is observed, both $`\sigma _{xx}`$ and $`\sigma _{xy}`$ satisfy a one parameter scaling law, which suggests that it is a consequence related to quantum phase transition.
The experimental results of $`\rho _{xx}`$ and $`\rho _{xy}`$ are presented for comparison in Fig. 3(b). It shows that the range of $`n_\nu `$ for the quantized $`\rho _{xy}`$ and the corresponding values of $`\rho _{xx}`$ are very close to our numerical ones. Here the temperature dependence of the experimental data can be translated into the $`L`$-dependence of our numerical results at $`T`$=0 through a dephasing length $`L_{in}`$. To further compare with the experiments, the calculated $`\rho _{xx}`$ as a scaling function of the relative LL filling number $`\delta n_\nu =n_\nu n_{\nu c}`$, i.e., $`\rho _{xx}=f(\delta n_L/\nu _0)`$, is shown in the inset of Fig. 3(a), where $`\nu _0=c_0(L/l_0)^{1/x}`$, $`x=2.3`$, and $`l_0`$ is the magnetic length ($`c_0`$ is a dimensionless constant). The experimental data (from Fig. 3 of Ref.) are also plotted in the inset using the $`T`$-dependent $`\nu _0`$ and an excellent agreement over a wide range of the scaling variable: $`2<\delta n_\nu /\nu _0<2`$ is clearly shown.
Here we note that in some experiments whether the quantum critical regime is reached is still controversial and there is an alternative explanation for the quantized $`\rho _{xy}`$ regime in which interactions may play a crucial role for a non-scaling behavior of the transport coefficients. An important distinction between such an interaction case and the present theory is that in the former case $`\rho _{xy}`$ is always well quantized in the insulating regime while the quantization of $`\rho _{xy}`$ in Fig. 3(a) is mainly confined around the critical point $`n_{\nu c}`$ with $`\rho _{xx}<10`$ $`h/e^2`$ and $`\rho _{xy}`$ eventually will start to distinctly grow with $`\rho _{xx}`$ as $`\rho _{xx}`$ further increases. Further experimental measurement in this regime may help to clarify this issue.
Finally, we would like to discuss a key physical distinction between the numerical phase diagram in Fig. 1(c) and the GPD in Fig. 1(a). In the latter case, all the QH boundaries eventually float up to $`n_\nu \mathrm{}`$ at $`B0`$ with the LL plateau structure in between remaining basically unchanged. But in both the TBM and the experiments, those vertical $`\nu >1`$ QH boundaries \[see Fig. 1(c)\] do not markedly “float up” in $`n_\nu `$ with increasing $`W`$ or reducing $`B`$ such that each LL plateau in between never floats away: only the width of the $`\nu `$-th QH plateau is reduced and vanishes eventually at the $`\nu 0`$ transition boundary. It then results in direct transitions and two insulating regimes in Fig. 1(c). To confirm this picture, we calculate the so-called equilibrium edge current which is proportional to $`n/B|_{E_f}`$ ($`n`$ is the electron density and $`E_f`$ is the Fermi energy) and is $`L`$ independent. The results (which are $`B`$-independent) are present in Fig. 4 in which the peaks determine the centers of QH plateaus. Indeed, such a quantity is continuously reduced with increasing $`W`$ and eventually diminishes at $`W_c`$, but its peak positions at $`W<W_c`$ never move away which clearly indicates that the recovery of a Andersen insulator at strong disorder in the integer QH system is due to the destruction of the plateaus instead of a float-up of the whole QH structure towards $`n_\nu \mathrm{}`$.
To summarize, we have determined a numerical phase diagram of the integer QH state for the first time based on the TBM. The topology of such a phase diagram is remarkably similar to the experimental one obtained recently for the QH system. Two kinds of insulating regimes surrounding the QH plateau phase are identified whose transport properties, characterized by the classic and quantized values of the Hall resistance, respectively, are also in good agreement with the experiments. The nature of such a phase diagram can be understood as a continuous narrowing and collapsing of the QH plateaus which are pinned around discrete LL filling numbers without floating away.
Acknowledgments \- The authors would like to acknowledge helpful discussions with R. N. Bhatt, S. V. Kravchenko, X.-G. Wen, L. P. Pryadko, A. Auerbach, and especially P. Coleridge and M. Hilke who also provided us their experimental data prior to publication. This work is supported by the State of Texas through ARP Grant No. 3652707 and Texas Center for Superconductivity at University of Houston.
Fig. 1 The phase diagram in disorder - $`1/n_\nu `$ plane: (a) Theoretic globle phase diagram predicted in Ref.; (b) Experimental one in Ref. (c) The present numerical result. Note that the scan line A in (c) corresponds to a constant $`V_G`$ line in (b) (see text).
Fig. 2 Hall resistance $`\rho _{xy}`$ (in units of $`h/e^2`$) as a function of $`n_\nu `$ along the scan line B in Fig. 1(c). The dashed curve represents the classic value of $`1/n_\nu (h/e^2)=B/nec`$. (a) $`\rho _{xy}`$ at different disorder strength $`W`$’s. The inset: $`\rho _{xx}`$ vs. $`n_\nu `$. (b) The finite size effect of $`\rho _{xy}`$.
Fig. 3 (a) The longitudinal resistance $`\rho _{xx}`$ and Hall resistance $`\rho _{xy}`$ (in units of $`h/e^2`$) versus $`n_\nu `$ along the scan line C in Fig. 1(c) at $`M=8`$. The inset: The scaling function $`\rho _{xx}=f(\delta n_\nu /\nu _0)`$ obtained from the numerical calculation ($`+`$) and the experimental measurement (Fig. 3 of )($``$). (b) Experimental data at different temperatures (note that $`\rho _{xx}`$ is in units of $`\rho _c=1.73h/e^2`$ according to Ref.).
Fig. 4 The equilibrim edge current $`n/B|_{E_f}`$ vs. $`n_\nu `$. It indicates that the QH plateaus are pinned at integer $`n_{\nu }^{}{}_{}{}^{}s`$ until their destruction by disorder.
|
no-problem/9906/hep-th9906220.html
|
ar5iv
|
text
|
# Untitled Document
PUPT-1874 hep-th/9906220
Tachyon Stabilization in the AdS/CFT Correspondence
Igor R. Klebanov
Joseph Henry Laboratories
Princeton University
Princeton, New Jersey 08544
USA
Abstract
We consider duality between type 0B string theory on $`AdS_5\times S^5`$ and the planar CFT on $`N`$ electric D3-branes coincident with $`N`$ magnetic D3-branes. It has been argued that this theory is stable up to a critical value of the ‘t Hooft coupling but is unstable beyond that point. We suggest that from the gauge theory point of view the development of instability is associated with singularity in the dimension of the operator corresponding to the tachyon field via the AdS/CFT map. Such singularities are common in large $`N`$ theories because summation over planar graphs typically has a finite radius of convergence. Hence we expect transitions between stability and instability for string theories in AdS backgrounds that are dual to certain large $`N`$ gauge theories: if there are tachyons for large AdS radius then they may be stabilized by reducing the radius below a critical value of order the string scale.
June 1999
1. Introduction
It has been proposed \[1,,2,,3,,4\] that certain 4-dimensional non-supersymmetric large $`N`$ gauge theories are dual to backgrounds of the type 0 string theory . This work was motivated by the recently discovered relations between type IIB strings and superconformal gauge theories on $`N`$ coincident D3-branes \[6,,7,,8,,9,,10\] as well as by Polyakov’s suggestion \[11\] (building on his earlier work ) that the type 0 string theory is a natural setting for extending this duality to non-supersymmetric gauge theories. Since then a number of further studies of D-branes in type 0 theories \[13,,14,,15\] and the associated dualities \[16,,17,,18,,19,,20,,21,,22,,23\] have appeared.
It is well known that taking the low energy limit on a stack of $`N`$ D3-branes in flat space and comparing it to the corresponding limit of the 3-brane classical solution suggests that the $`𝒩=4`$ supersymmetric $`SU(N)`$ gauge theory is dual to type IIB strings on $`AdS_5\times S^5`$ \[8,,9,,10\]. Generalizations of this duality to other 4-dimensional gauge theories follow if the stack of D3-branes is placed on a transverse space $`Y_6`$ which is a cone over an Einstein manifold $`X_5`$ \[24,,25,,26,,27,,28,,29,,30\]. Then the the CFT on the D3-branes is dual to type IIB string theory on $`AdS_5\times X_5`$. An interesting set of dualities of this sort is found if $`Y_6`$ is an orbifold $`R^6/\mathrm{\Gamma }`$ where $`\mathrm{\Gamma }`$ is a discrete subgroup of $`SO(6)`$ \[25,,26\]. The corresponding theories on the branes are conformal if the orbifold group acts appropriately on the gauge indices . Duality between such CFT’s and type IIB backgrounds of the form $`AdS_5\times S^5/\mathrm{\Gamma }`$ is thus an orbifold of the original duality between the $`𝒩=4`$ theory and the $`AdS_5\times S^5`$ background. Remarkably, a large class of such CFT’s is non-supersymmetric in the large $`N`$ (planar) limit \[25,,26\]. A phenomenological application of non-supersymmetric CFT’s was recently proposed in but later criticized in .
Alternatively, one may construct gauge theory/string dualities by stacking D3-branes in string models other than type IIB. Type 0 theories are attractive in this regard but they have a well-known and seemingly fatal flaw in that their spectrum contains a tachyon of $`m^2=2/\alpha ^{}`$ which is not removed by the non-chiral GSO projection $`(1)^{F+\stackrel{~}{F}}=1`$. In it was suggested, however, that the tachyon may be stabilized in appropriate type 0 backgrounds which are relevant to gauge theory applications.
The type 0 models have a doubled set of RR fields compared to their type II cousins, hence they also possess twice as many D-branes . For example, since type 0B spectrum has an unrestricted 4-form gauge potential, there are two types of D3-branes: those that couple electrically to this gauge potential, and those that couple magnetically. Very importantly, the weakly coupled spectrum of open strings on type 0 D-branes does not contain tachyons after the GSO projection $`(1)^{F_{open}}=1`$ is implemented \[1,,11,,33\]. Thus, gauge theories living on such D-branes do not have obvious instabilities. This suggests via the gauge field/string duality that the bulk tachyon instability of type 0 theory may be cured as well \[1,,4\].
Let us review an argument for the tachyon stabilization in the simplest setting, which is the stack of $`N`$ electric and $`N`$ magnetic D3-branes . For such a stack the net tachyon tadpole cancels so that there exists a classical solution with $`T=0`$. In fact, since the stack couples to the selfdual part of the 5-form field strength, the type 0B 3-brane classical solution is identical to the type IIB one. Taking the throat limit suggests that the low-energy field theory on $`N`$ electric and $`N`$ magnetic D3-branes is dual to the $`AdS_5\times S^5`$ background of type 0B theory and is therefore conformal in the planar limit . This theory is the $`U(N)\times U(N)`$ gauge theory coupled to 6 adjoint scalars of the first $`U(N)`$, 6 adjoint scalars of the second $`U(N)`$, and fermions in the bifundamental representations – 4 Weyl fermions in the $`(𝐍,\overline{𝐍})`$ and 4 Weyl fermions in the $`(\overline{𝐍},𝐍)`$ (the $`U(1)`$ factors decouple in the infrared). While such a theory does not correspond to placing type IIB D3-branes at an orbifold point, it is nevertheless a $`Z_2`$ projection of the $`𝒩=4`$ $`U(2N)`$ gauge theory . The $`Z_2`$ is generated by $`(1)^{F_s}`$, where $`F_s`$ is the fermion number, together with conjugation by $`\left(\begin{array}{cc}I& 0\\ 0& I\end{array}\right)`$ where $`I`$ is the $`N\times N`$ identity matrix. This is related to the fact that type 0 string theories may be viewed as $`(1)^{F_s}`$ orbifolds of the corresponding type II theories . In it was pointed out that, since $`(1)^{F_s}`$ is an element of the center of the $`SU(4)`$ R-symmetry, this $`Z_2`$ projection of the $`𝒩=4`$ theory belongs to the class studied in \[25,,26\].
The fact that the non-supersymmetric AdS/CFT duality considered in is a $`Z_2`$ quotient of the $`𝒩=4`$ duality \[4,,18\] lends additional credence to its validity. In particular, general arguments presented in guarantee that the field theory is conformal in the planar limit and that planar correlators of untwisted vertex operators coincide with those of the $`𝒩=4`$ theory. Since this CFT does not appear to have any instabilities at weak coupling, it was argued in that type 0B string theory on $`AdS_5\times S^5`$ is stable for sufficiently small radius. This provides a simple AdS/CFT argument in favor of tachyon stabilization; it is also an example of how gauge theory may be used to make predictions about the string theory dual to it.
We should note that some important aspects of this particular $`Z_2`$ quotient are related to the twisted sector of the orbifold and are not covered by the analysis in . To analyze some of the physics it will be important that this theory is dual to type 0B rather than type IIB background. Therefore, in the usual gravity limit where the radius of $`AdS_5\times S^5`$ becomes large in string units and space becomes locally flat, this background is unstable . From the CFT point of view it means that the limit of infinite ‘t Hooft coupling, which is commonly discussed in the AdS/CFT context, does not make sense. The AdS/CFT correspondence applied to this particular case implies that, as $`\lambda =g_{YM}^2N`$ is increased, a transition should happen at a critical value $`\lambda _c`$ . In this paper we address the origin of this transition from the field theory point of view and argue that it is due to a singularity of a sum over planar diagrams. In non-supersymmetric large $`N`$ field theories such singularities are expected to occur on general grounds .<sup>1</sup> In view of this fact, which is also supported by evidence from exactly solvable large $`N`$ models , it may seem surprising that certain non-supersymmetric planar CFT’s appear to make sense at infinite ‘t Hooft coupling . A possible explanation is that in these theories $`\lambda _c`$ is negative. We identify the field theory quantity which we expect to become singular at $`\lambda _c`$: it is the dimension of the operator which corresponds to the tachyon field via the AdS/CFT map. Using the methods of we show that this operator is
$$𝒪_T=\frac{1}{4}\mathrm{Tr}F_{\alpha \beta }^2\frac{1}{4}\mathrm{Tr}G_{\alpha \beta }^2+\frac{1}{2}\mathrm{Tr}(D_\alpha X^i)^2\frac{1}{2}\mathrm{Tr}(D_\alpha Y^i)^2+\mathrm{},$$
where $`F_{\alpha \beta }`$ and $`X^i`$ are the field strength and the 6 adjoint scalars of $`SU(N)_1`$ while $`G_{\alpha \beta }`$ and $`Y^i`$ are the corresponding objects of $`SU(N)_2`$. Adding such an operator to the action creates a difference between the gauge couplings of $`SU(N)_1`$ and $`SU(N)_2`$. In non-supersymmetric field theory this is not expected to be an exactly marginal deformation. By analyzing 2-loop beta functions we show that the operator (1.1) indeed picks up an anomalous dimension of order $`\lambda ^2`$. Higher loops graphs will correct the dimension $`\mathrm{\Delta }`$ as well, and it is conceivable that $`\mathrm{\Delta }(\lambda )`$ develops a singularity at some critical value $`\lambda _c>0`$ as suggested by the AdS considerations.
2. Anomalous dimension of the tachyon operator
The standard relation between scalar operator dimension in a conformal field theory and mass in $`AdS_5`$ is \[9,,10,,30\]
$$\mathrm{\Delta }=2\pm \sqrt{4+m^2L^2}.$$
This relation is definitely valid in the gravity limit where $`L^2/\alpha ^{}=\sqrt{2\lambda }1`$. Applying this relation to the case of type 0 tachyon, $`m^2=2/\alpha ^{}`$, we immediately find that for large $`\lambda `$ $`\mathrm{\Delta }`$ is complex, hence the CFT is unstable. From the AdS point of view, the stability bound that mass-squared must exceed $`4/L^2`$ is violated for sufficiently large $`L`$. For $`\lambda `$ of order 1 (i.e. for AdS radius of order the string scale) there may be various corrections to the relation (2.1). Even if the overall form of this relation remains the same, $`L^2/\alpha ^{}`$ should be treated as some unknown function of $`\lambda `$, and effective $`m^2`$ might have the form
$$m^2=\frac{2}{\alpha ^{}}+\frac{a_1}{L^2}+\frac{a_2\alpha ^{}}{L^4}+\mathrm{}$$
This forces us to proceed qualitatively.
To begin with let us simply continue the gravity relation to all $`\lambda `$ without any change. Then we find
$$\mathrm{\Delta }(\lambda )=2\pm \sqrt{42\sqrt{2\lambda }}.$$
The dimension is real for sufficiently small $`\lambda `$, and the unitarity bound $`\mathrm{\Delta }>1`$ forces us to pick the positive branch of the square root. This simplified “model” for $`\mathrm{\Delta }(\lambda )`$ implies that $`\mathrm{\Delta }(0)=4`$ which, as we will see, is the correct answer. However, obviously (2.1) cannot be exact: it gives an expansion in powers of $`\sqrt{\lambda }`$ while in the gauge theory it has to proceed in integer powers of $`\lambda `$ (we will see that the leading anomalous dimension is $`O(\lambda ^2)`$). We expect various corrections to this “model” such as those discussed above. For example, tachyon interactions with the R-R 5-form field strength, which have the form $`F_5^2T^2`$ , give rise to a positive $`a_1`$ in (2.1). One effect of this shift is to push $`\lambda _c`$ to a higher value, but we cannot rule out the presence of other similar corrections. Our qualitative picture is quite insensitive to details however: $`\mathrm{\Delta }(\lambda )`$ is real for $`\lambda <\lambda _c`$ but has a branch cut for $`\lambda >\lambda _c`$ which signals an instability. In the following we will use perturbative field theory arguments to support this scenario.
The primary question is: what is the operator in the CFT on $`N`$ electric and $`N`$ magnetic D3-branes that couples to the tachyon. The tachyon coupling to a stack of $`N`$ electric D3-branes is \[1,,3,,13\]
$$T_3d^4xk_e(T)(N+\frac{1}{4}\mathrm{Tr}F_{\alpha \beta }^2+\frac{1}{2}\mathrm{Tr}(D_\alpha X^i)^2+\mathrm{}),$$
where $`k_e(T)=1+\frac{1}{4}T+O(T^2)`$ and $`T_3`$ is the tension of an electric D3-brane. The coupling to a stack of $`N`$ magnetic D3-branes is instead \[1,,3,,13\]
$$T_3d^4xk_m(T)(N+\frac{1}{4}\mathrm{Tr}G_{\alpha \beta }^2+\frac{1}{2}\mathrm{Tr}(D_\alpha Y^i)^2+\mathrm{}),$$
where $`k_m(T)=1\frac{1}{4}T+O(T^2)`$. Adding these two actions, we find that the tachyon coupling to a stack of $`N`$ electric and $`N`$ magnetic branes is
$$d^4xT(\frac{1}{4}\mathrm{Tr}F_{\alpha \beta }^2\frac{1}{4}\mathrm{Tr}G_{\alpha \beta }^2+\frac{1}{2}\mathrm{Tr}(D_\alpha X^i)^2\frac{1}{2}\mathrm{Tr}(D_\alpha Y^i)^2+\mathrm{})$$
leading to the expression (1.1) for the gauge invariant operator corresponding to the tachyon field via the AdS/CFT map. This operator is odd under the interchange of the two $`SU(N)`$’s which is related to the fact that the tachyon is one of the twisted states from the point of view of orbifolding type IIB string theory by $`(1)^{F_s}`$. Indeed, the tachyon belongs to the $`(NS,NS)`$ sector of type 0B theory which is not present in the type IIB spectrum. In other $`Z_2`$ orbifold theories it was also found that twisted states couple to vertex operators odd under the interchange of the factors of the gauge group \[38,,28\]. Fields from the untwisted sector, on the other hand, couple to operators that are even under the interchange of the gauge groups: for example, the dilaton couples to
$$\frac{1}{4}\mathrm{Tr}F_{\alpha \beta }^2+\frac{1}{4}\mathrm{Tr}G_{\alpha \beta }^2+\mathrm{}.$$
In order to study the anomalous dimension of the operator (1.1) at weak coupling, we may consider slightly unequal gauge couplings for $`SU(N)_1`$ and $`SU(N)_2`$. Then the coupling constant for this operator is
$$T=\frac{1}{\lambda _1}\frac{1}{\lambda _2},$$
where $`\lambda _i=g_i^2N`$, and $`i=1,2`$. The planar (leading order in $`N`$) renormalization group equations to two-loop order are
$$\mu \frac{}{\mu }\frac{1}{\lambda _1}=a(\lambda _1\lambda _2)+O\left[(\lambda _1\lambda _2)^2\right],$$
$$\mu \frac{}{\mu }\frac{1}{\lambda _2}=a(\lambda _2\lambda _1)+O\left[(\lambda _1\lambda _2)^2\right],$$
and we find $`a=(2\pi )^4`$. Due to the vanishing of the one-loop beta function there is no constant term on the right-hand side of the equations. The two-loop contributions vanish for equal couplings, as shown explicitly in , but they no longer vanish for $`\lambda _1\lambda _2`$.
Taking the difference of the two equations, and working to first order in $`T`$, we have
$$\mu \frac{T}{\mu }=\frac{\lambda ^2}{8\pi ^4}T.$$
Solving this we have
$$T=T_0\left(\frac{\mu }{\mathrm{\Lambda }}\right)^{\frac{\lambda ^2}{8\pi ^4}}.$$
Since $`T`$ grows in the UV (for $`\mu \mathrm{\Lambda }`$) the associated operator (1.1) is irrelevant to this order: its dimension is
$$\mathrm{\Delta }(\lambda )=4+\frac{\lambda ^2}{8\pi ^4}+O(\lambda ^3).$$
Based on the AdS/CFT correspondence, we conjecture that further corrections to $`\mathrm{\Delta }(\lambda )`$ will produce a function which develops a singularity at $`\lambda _c>0`$. We do not expect direct calculations of the anomalous dimension to be feasible beyond $`O(\lambda ^3)`$ which may be extracted from 3-loop beta functions. However, it might be possible to put bounds on the general $`O(\lambda ^n)`$ terms and prove the existence of a critical point at $`\lambda _c>0`$.
To summarize this paper, we have suggested that finite radius of convergence of planar graphs, which is a well-known phenomenon in large $`N`$ field theories, has interesting implications in view of the AdS/CFT correspondence. Namely, if there is an operator whose dimension is real up to a critical value of the ‘t Hooft coupling but becomes complex beyond that value, then the dual AdS phenomenon is that a tachyon present in the large radius (gravity) limit is stabilized for sufficiently small radii. It is difficult to study such backgrounds with string-scale curvature directly, but on the dual CFT side perturbative calculations are quite tractable. The perturbative stability of the CFT thus provides an argument for stabilization of tachyons (this is one of the few results to date where field theory has been used to make new predictions about string theory), and finiteness of the radius of convergence of planar graphs suggests how instability may develop at sufficiently strong coupling. We have discussed one specific example which arises in the context of type 0 theory but we hope that these phenomena are quite general.
Acknowledgements
I am grateful to E. Silverstein, A. Tseytlin and especially J. Minahan for useful discussions and comments. This work was supported in part by the NSF grant PHY-9802484 and by the James S. McDonnell Foundation Grant No. 91-48.
References
relax I.R. Klebanov and A.A. Tseytlin, “D-Branes and Dual Gauge Theories in Type 0 Strings,” Nucl. Phys. B546 (1999) 155, hep-th/9811035. relax J. Minahan, “Glueball Mass Spectra and Other Issues for Supergravity Duals of QCD Models,” hep-th/9811156. relax I.R. Klebanov and A.A. Tseytlin, “Asymptotic Freedom and Infrared Behavior in the Type 0 String Approach to Gauge Theory,” Nucl. Phys. B547 (1999) 143, hep-th/9812089. relax I.R. Klebanov and A.A. Tseytlin, “Non-supersymmemtric CFT from Type 0 String Theory,” J. High Energy Phys. 9903 (1999) 015, hep-th/9901101. relax L. Dixon and J. Harvey, “String theories in ten dimensions without space-time supersymmetry”, Nucl. Phys. B274 (1986) 93; N. Seiberg and E. Witten, “Spin structures in string theory”, Nucl. Phys. B276 (1986) 272; C. Thorn, unpublished. relax I.R. Klebanov, “World volume approach to absorption by nondilatonic branes,” Nucl. Phys. B496 (1997) 231, hep-th/9702076; S.S. Gubser, I.R. Klebanov, and A.A. Tseytlin, “String theory and classical absorption by three-branes,” Nucl. Phys. B499 (1997) 217, hep-th/9703040. relax S.S. Gubser and I.R. Klebanov, “Absorption by branes and Schwinger terms in the world volume theory,” Phys. Lett. B413 (1997) 41, hep-th/9708005. relax J. Maldacena, “The Large N limit of superconformal field theories and supergravity,” Adv. Theor. Math. Phys. 2 (1998) 231, hep-th/9711200. relax S.S. Gubser, I.R. Klebanov, and A.M. Polyakov, “Gauge theory correlators from noncritical string theory,” Phys. Lett. B428 (1998) 105, hep-th/9802109. relax E. Witten, “Anti-de Sitter space and holography,” Adv. Theor. Math. Phys. 2 (1998) 253, hep-th/9802150. relax A.M. Polyakov, “The Wall of the Cave,” hep-th/9809057. relax A.M. Polyakov, “String theory and quark confinement,” Nucl. Phys. B (Proc. Suppl.) 68 (1998) 1, hep-th/9711002. relax M.R. Garousi, “String Scattering from D-branes in Type 0 Theories”, hep-th/9901085. relax M. Billó, B. Craps and F. Roose, “On D-branes in Type 0 String Theory,” hep-th/9902196. relax K. Zarembo, “Coleman-Weinberg Mechanism and Interaction of D3-branes in Type 0 String Theory, hep-th/9901106; A. Tseytlin and K. Zarembo, “Effective Potential in Non-supersymmetric $`SU(N)\times SU(N)`$ Gague Theory and Interactions of Type 0 D3-branes,” hep-th/9902095. relax G. Ferretti and D. Martelli, “On the construction of gauge theories from non critical type 0 strings,” hep-th/9811208. relax J. Minahan, “Asymptotic Freedom and Confinement from Type 0 String Theory,” hep-th/9902074. relax N. Nekrasov and S. Shatashvili, “On non-supersymmetric CFT in four dimension,” hep-th/9902110. relax M. Costa, “Intersecting D-branes and black holes in type 0 string theory,” hep-th/9903128. relax M. Alishahiha, A. Brandhuber and Y. Oz, “Branes at Singularities in Type 0 String Theory,” hep-th/9903186. relax O. Bergman and M. Gaberdiel, “Dualities of Type 0 Strings,” hep-th/9906055. relax A. Armoni and B. Kol, “Non-supersymmetric Large $`N`$ Gauge Theories from Type 0 Brane Configurations,” hep-th/9906081. relax R. Blumenhagen, A. Font and D. Lust, “Non-Supersymmetric Gauge Theories from D-Branes in Type 0 String Theory,” hep-th/9906101. relax M. Douglas and G. Moore, “D-branes, quivers, and ALE instantons,” hep-th/9603167. relax S. Kachru and E. Silverstein, “4d conformal field theories and strings on orbifolds,” Phys. Rev. Lett. 80 (1998) 4855, hep-th/9802183. relax A. Lawrence, N. Nekrasov and C. Vafa, “On conformal field theories in four dimensions,” Nucl. Phys. B533 (1998) 199, hep-th/9803015. relax A. Kehagias, “New Type IIB Vacua and Their F-Theory Interpretation,” hep-th/9805131. relax I.R. Klebanov and E. Witten, “Superconformal field theory on threebranes at a Calabi-Yau singularity,” Nucl. Phys. B536 (1998) 199, hep-th/9807080; S.S. Gubser, “Einstein manifolds and conformal field theories,” hep-th/9807164. relax D.R. Morrison and M.R. Plesser, “Non-Spherical Horizons, I,” hep-th/9810201. relax I.R. Klebanov and E. Witten, “AdS/CFT Correspondence and Symmetry Breaking,” hep-th/9905104. relax P. Frampton, “ADS/CFT String Duality and Conformal Gauge Theories,” hep-th/9812117; P. Frampton and C. Vafa, “Conformal Approach to Particle Phenomenology,” hep-th/9903226. relax C. Csaki, W. Skiba and J. Terning, “Beta Functions of Orbifold Theories and the Hierarchy Problem,” hep-th/9906057. relax O. Bergman and M. Gaberdiel, “A Non-supersymmetric Open String Theory and S-Duality,” Nucl. Phys. B499 (1997) 183, hep-th/9701137. relax M. Bershadsky, Z. Kakushadze and C. Vafa, “String expansion as large N expansion of gauge theories”, Nucl. Phys. B523 (1998) 59, hep-th/9803076; M. Bershadsky and A. Johansen, “Large N limit of orbifold field theories,” Nucl. Phys. B536 (1998) 141, hep-th/9803249. relax G. ‘t Hooft, “On the Convergence of Planar Diagram Expansions,” Comm. Math. Phys. 86 (1982) 449. relax E. Brezin, C. Itzykson, G. Parisi and J. Zuber, Comm. Math. Phys. 59 (1978) 35. relax P. Breitenlohner and D.Z. Freedman, “Stability in gauged extended supergravity”, Ann. Phys. 144 (1982) 249. relax S. Gukov, “Comments on N=2 AdS Orbifolds,” Phys. Lett. B439 (1998) 23, hep-th/9906081.
|
no-problem/9906/math9906150.html
|
ar5iv
|
text
|
# Flows on solenoids are generically not almost periodic
## Introduction
For any compact symplectic manifold $`M`$ of dimension at least $`4`$, L. Markus and K. R. Meyer demonstrate in \[MM\] that the space $`^k\left(M\right)`$ of all $`C^k`$ ($`k4`$) Hamiltonians on $`M`$ contains a generic subset $`𝔐_\mathrm{\Sigma }`$ such that for each Hamiltonian $`dH^\mathrm{\#}`$ in $`𝔐_\mathrm{\Sigma }`$ and for each solenoid $`\mathrm{\Sigma }_N`$ there exists a minimal set for the flow induced by $`dH^\mathrm{\#}`$ that is homeomorphic to $`\mathrm{\Sigma }_N`$. They leave open the question of whether the flows on these solenoids are almost periodic (\[MM\], p. 90).
Corresponding to a sequence of natural numbers $`N=(n_1,n_2,\mathrm{})`$ with $`n_j2`$ for each $`j`$ there is the solenoid $`\mathrm{\Sigma }_N`$ which is the inverse limit of the inverse sequence $`\{X_j,f_j\}_{j=1}^{\mathrm{}}`$ with factor space $`X_j=S^1=\{\mathrm{exp}\left(2\pi it\right)t[0,1)\}`$ for each $`j`$ and bonding maps $`f_j\left(z\right)=z^{n_j}`$
$$S^1\stackrel{n_1}{}S^1\stackrel{n_2}{}S^1\stackrel{n_3}{}\mathrm{}\mathrm{\Sigma }_N=\{z_j_{j=1}^{\mathrm{}}\underset{i=1}{\overset{\mathrm{}}{}}S^1z_j=f_j\left(z_{j+1}\right)\text{ for }j=1,2,\mathrm{}\}.$$
Here $`_{i=1}^{\mathrm{}}S^1`$ is the compact topological group with the group operation (written $`\mathrm{`}\mathrm{`}+\mathrm{"}`$) given by factor–wise multiplication and with the metric
$$d(x_j_{j=1}^{\mathrm{}},y_j_{j=1}^{\mathrm{}})=\underset{j=1}{\overset{\mathrm{}}{}}\frac{1}{2^j}\left|x_jy_j\right|.$$
This metric also serves as a metric for the subgroup $`\mathrm{\Sigma }_N`$. There is then the continuous (but not bicontinuous) isomorphism $`\pi _N:C_N`$ onto the $`\mathrm{\Sigma }_N`$–arc component $`C_N`$ of the identity $`e=(1,1,\mathrm{})`$ given by
$$t\mathrm{exp}\left(2\pi it\right),\mathrm{exp}\left(\frac{2\pi it}{n_1}\right),\mathrm{},\mathrm{exp}\left(\frac{2\pi it}{n_1\mathrm{}n_j}\right),\mathrm{},$$
and we have the family of linear flows $`\mathrm{\Lambda }_N`$ on $`\mathrm{\Sigma }_N`$:
$`\mathrm{\Lambda }_N`$ $`=`$ $`\{\varphi _N^\alpha :\times \mathrm{\Sigma }_N\mathrm{\Sigma }_N\alpha \}`$
$`\varphi _N^\alpha (t,x)`$ $`=`$ $`\pi _N\left(\alpha t\right)+x`$
and any almost periodic flow on $`\mathrm{\Sigma }_N`$ is equivalent ($`C^0`$ conjugate) to some $`\varphi _N^\alpha \mathrm{\Lambda }_N`$ \[C\]. (A *flow* is a continuous group action of $`(,+)`$.)
Given any point $`x\mathrm{\Sigma }_N`$ the map $`\pi _N+x`$ sending $`t\pi _N\left(t\right)+x`$ parameterizes the arc component of $`x`$, which will coincide with the trajectory of $`x`$ for any non–singular flow $`\varphi `$ on $`\mathrm{\Sigma }_N`$. For $`k=0,1,\mathrm{},`$ $`\mathrm{}`$ and $`0\alpha 1`$, we consider a flow $`\varphi :\times \mathrm{\Sigma }_N\mathrm{\Sigma }_N`$ to be $`C^{k+1+\alpha }`$ if the associated “vector field”
$$v_\varphi :\mathrm{\Sigma }_N;x\frac{d\left[\left(\pi _N+x\right)^1\varphi (t,x)\right]}{dt}_{t=0}$$
exists and satisfies the conditions:
1. for $`j=1,\mathrm{},k`$ the function $`v_\varphi ^j:\mathrm{\Sigma }_N;x{\displaystyle \frac{d^j\left[\left(\pi _N+x\right)^1\varphi (t,x)\right]}{dt^j}}_{t=0}`$ is continuous \[$`v_\varphi ^1=v_\varphi `$\]
2. for each $`x\mathrm{\Sigma }_N`$ the function $`;tv_\varphi ^k\left(x+\pi _N\left(t\right)\right)`$ is $`\alpha `$–Hölder (see below) if $`\alpha >0`$ and $`k<\mathrm{}`$.
We then endow the space $`C^{k+\alpha }\left(N\right)`$ of vector fields $`v_\varphi `$ on $`\mathrm{\Sigma }_N`$ stemming from $`C^{k+1+\alpha }`$ flows $`\varphi `$ with the metric
$$d_k(v_\varphi ,v_\psi )\stackrel{def}{=}\underset{j=1}{\overset{k}{}}\underset{x\mathrm{\Sigma }_N}{\mathrm{max}}\left|v_\varphi ^j\left(x\right)v_\psi ^j\left(x\right)\right|\text{ for }1k<\mathrm{}$$
$$\text{and }d_{\mathrm{}}(v_\varphi ,v_\psi )\stackrel{def}{=}\underset{r=1}{\overset{\mathrm{}}{}}\frac{2^rd_r(v_\varphi ,v_\psi )}{1+d_r(v_\varphi ,v_\psi )}.$$
If $`\mathrm{\Sigma }_N`$ is embedded in some compact $`C^k`$ manifold so that the flow $`\varphi _N^1`$ extends to a $`C^k`$ flow on the manifold, then this metric induces the same topology as the restriction of the Whitney topology (see \[MM\], pp. 34–35 for a description of the Whitney topology). We are interested in the non–singular flows on $`\mathrm{\Sigma }_N`$ and so will work with $`C^{k+\alpha }\left(N+\right),`$ the subspace of $`C^{k+\alpha }\left(N\right)`$ consisting of positive $`v_\varphi `$. (Whenever $`\varphi `$ is non–singular the map $`v_\varphi `$ must be either positive or negative since each trajectory is dense and the intermediate value theorem yields a singularity if there is a change of signs, and so we consider positive $`v_\varphi `$ without loss of generality.)
If $`p_j:\mathrm{\Sigma }_NS^1`$ denotes the projection onto the $`j^{th}`$ factor $`x_k_{k=1}^{\mathrm{}}x_j`$ and if $`\widehat{N}`$ denotes the subgroup of $`(,+)`$ generated by
$$\{\frac{1}{n_1},\frac{1}{n_1n_2},\mathrm{},\frac{1}{n_1n_2\mathrm{}n_j}j=1,2,\mathrm{}\}$$
and if for $`r={\displaystyle \frac{m}{n_1n_2\mathrm{}n_j}}\widehat{N}`$ $`\chi _r:\mathrm{\Sigma }_NS^1`$ denotes the homomorphism sending $`x\left(p_{j+1}\left(x\right)\right)^m,`$ then $`\mathrm{\Xi }\left(N\right)=\{\chi _rr\widehat{N}\}`$ is the group of characters of $`\mathrm{\Sigma }_N`$. Given a continuous function $`f:\mathrm{\Sigma }_N`$, the theorem of Peter and Weyl guarantees that $`f`$ can be uniformly approximated by finite linear combinations of characters $`\chi _r`$ (see, e.g., \[P\]), and since $`\chi _{\frac{m}{n}}\pi _N\left(t\right)=\mathrm{exp}\left(2\pi imt/n\right)`$ the map $`f_e:;`$ $`tf\left(\pi _N\left(t\right)\right)`$ is the uniform limit of periodic maps and is thus *limit periodic* by definition (see, e.g., \[Be\] 1§6), which is a special type of almost periodic function. It then follows that the Bohr–Fourier series
$$f_e\left(t\right)\underset{r\widehat{N}}{}f_r\chi _r\pi _N\left(t\right)=\underset{r\widehat{N}}{}f_r\mathrm{exp}\left(2\pi irt\right)$$
when appropriately ordered converges uniformly to $`f_e`$ provided $`f_e`$ is $`\alpha `$–Hölder for some $`0<\alpha 1`$:
$$\underset{t}{sup}\left|f_e\left(t+\delta \right)f_e\left(t\right)\right|<C\delta ^\alpha \text{ }\left(\text{for some }C>0\text{ and all }\delta >0\right)\text{ (see }\text{[Be]}\text{ 1§8).}$$
Whenever we are dealing with such a function we will assume without comment that the series is so arranged that $`f=\underset{r\widehat{N}}{}f_r\chi _r`$.
We shall show that in each space $`C^{k+\alpha }\left(N+\right)`$ the collection of vector fields $`v_\varphi `$ corresponding to flows $`\varphi `$ which are not almost periodic contains a dense $`G_\delta `$ provided $`k1`$ or $`\alpha =1`$ (in other words: $`v_\varphi `$ is Lipschitz).
## 1. Proof of the Main Result
Fix a space $`C^{k+\alpha }\left(N+\right)`$ with $`k1`$ or $`\alpha =1`$. As already mentioned, a flow $`\varphi `$ on $`\mathrm{\Sigma }_N`$ is almost periodic exactly when there is a homeomorphism $`h^{}:\mathrm{\Sigma }_N\mathrm{\Sigma }_N`$ providing an equivalence of $`\varphi `$ with some linear flow $`\varphi _N^\alpha ^{}`$, $`h^{}\varphi (t,x)\varphi _N^\alpha ^{}(t,h^{}\left(x\right))`$. Any such $`h^{}`$ is homotopic to an automorphism $`a`$ of $`\mathrm{\Sigma }_N`$, and the automorphism $`a^1`$ will in turn equate $`\varphi _N^\alpha ^{}`$ with a linear flow $`\varphi _N^\alpha `$ (see \[S\] and \[C\]). And so $`\varphi `$ is almost periodic exactly when there is a homeomorphism $`h=a^1h^{}`$ homotopic to the identity providing an equivalence between $`\varphi `$ and a linear flow $`\varphi _N^\alpha `$.
We now translate the existence of such an $`h`$ into a more useful form, so assume $`\varphi `$ is almost periodic and that such an $`h`$ exists. Each $`x\mathrm{\Sigma }_N`$ belongs to the section $`𝒮_x\stackrel{def}{=}p_1^1\left(p_1\left(x\right)\right)`$ (which is a Cantor set), and the time it takes the linear flow $`\varphi _N^\alpha `$ to return to $`𝒮_x`$ is the constant $`{\displaystyle \frac{1}{\alpha }}`$ since
$$p_1\left(\varphi _N^\alpha (t,x)\right)=p_1\left(\pi _N\left(t\alpha \right)+x\right)=p_1\left(\pi _N\left(t\alpha \right)\right)p_1\left(x\right)=\mathrm{exp}\left(2\pi it\alpha \right)p_1\left(x\right)$$
and the smallest positive value of $`t`$ satisfying $`\mathrm{exp}\left(2\pi it\alpha \right)=1`$ is $`{\displaystyle \frac{1}{\alpha }}`$. Since $`1/x`$ is $`C^{\mathrm{}}`$ on any closed interval not containing $`0`$, $`\lambda \left(x\right)\stackrel{def}{=}{\displaystyle \frac{1}{v_\varphi \left(x\right)}}=\underset{r\widehat{N}}{}\lambda _r\chi _r\left(x\right)`$ is $`C^{k+\alpha }`$ and the $`\varphi `$–return time $`\tau `$ of $`x`$ to the section $`𝒮_x`$ is given by
$`\tau \left(x\right)`$ $`=`$ $`{\displaystyle _0^1}\lambda \left(x+\pi _N\left(t\right)\right)𝑑t`$
$`=`$ $`\lambda _0+{\displaystyle \underset{r\widehat{N}\left\{0\right\}}{}}{\displaystyle \frac{\lambda _r}{2\pi ir}}\chi _r\left(x+\pi _N\left(1\right)\right){\displaystyle \underset{r\widehat{N}\left\{0\right\}}{}}{\displaystyle \frac{\lambda _r}{2\pi ir}}\chi _r\left(x\right)`$
$``$ $`\lambda _0+{\displaystyle \underset{r\widehat{N}\left\{0\right\}}{}}{\displaystyle \frac{\lambda _r}{2\pi ir}}\left[\chi _r\left(\pi _N\left(1\right)\right)1\right]\chi _r\left(x\right)={\displaystyle \underset{r\widehat{N}}{}}\tau _r\chi _r\left(x\right).`$
Now $`h`$ is homotopic to the identity and so we have the map $`\delta \underset{r\widehat{N}}{}\delta _r\chi _r:\mathrm{\Sigma }_N`$
$$\delta \left(x\right)\stackrel{def}{=}\text{the unique time }t\text{ satisfying }h\left(x\right)=\varphi _N^\alpha (t,x)$$
since $`h\left(x\right)`$ must lie in the same path component as $`x`$. Since $`h`$ provides a flow equivalence, the $`\varphi _N^\alpha `$–return time to the $`\varphi _N^\alpha `$–section $`h\left(𝒮_x\right)`$ for $`h\left(x\right)`$ is $`\tau \left(x\right)`$. We can now express the constancy of the $`\varphi _N^\alpha `$–return time to $`𝒮_x`$ as
$$\delta \left(x\right)+\tau \left(x\right)\delta \left(x+\pi _N\left(1\right)\right)=\frac{1}{\alpha }$$
or
$$\tau \left(x\right)\frac{1}{\alpha }=\delta \left(x+\pi _N\left(1\right)\right)\delta \left(x\right).$$
We also have
$$\delta \left(x+\pi _N\left(1\right)\right)\delta \left(x\right)\underset{r\widehat{N}\left\{0\right\}}{}\left[\chi _r\left(\pi _N\left(1\right)\right)1\right]\delta _r\chi _r\left(x\right)$$
leading to the conditions
$$\delta _r=\frac{\tau _r}{\left[\chi _r\left(\pi _N\left(1\right)\right)1\right]}=\frac{\lambda _r}{2\pi ir}\text{ for }r\widehat{N}\left\{0\right\}$$
and
$$\lambda _0=\tau _0=\frac{1}{\alpha }.$$
(There is no restriction on $`\delta _0`$ since we may follow $`h`$ by translations of elements in $`C_N`$ and obtain other flow equivalences homotopic to the identity.)
We then have the limit periodic function
$$\delta _e\left(t\right)\underset{r\widehat{N}}{}\delta _r\mathrm{exp}\left(2\pi irt\right)=\delta _0+\underset{r\widehat{N}\left\{0\right\}}{}\frac{\lambda _r}{2\pi ir}\mathrm{exp}\left(2\pi irt\right).$$
According to Bohr’s theorem on the integral of an almost periodic function (see \[B\] §68–69), the integral $`F\left(T\right)=_0^Tf\left(t\right)𝑑t`$ of a limit periodic function $`f\left(t\right)\underset{r\widehat{N}}{}f_r\mathrm{exp}\left(2\pi irt\right)`$ is almost periodic if and only if $`F`$ is bounded, in which case
$$F\left(t\right)\underset{r\widehat{N}}{}F_r\mathrm{exp}\left(2\pi irt\right)=F_0+\underset{r\widehat{N}\left\{0\right\}}{}\frac{f_r}{2\pi ir}\mathrm{exp}\left(2\pi irt\right).$$
Comparing the Fourier–Bohr series of $`\delta _e`$ with that of $`\lambda _e`$, we see that the bounded function $`\delta _e\left(t\right)`$ represents an integral of $`\lambda _e\left(t\right)\lambda _0`$. Moreover, whenever $`\lambda _e\left(t\right)\lambda _0`$ has a bounded integral we are able to construct a map $`\delta `$ as above, which in turn allows us to construct an equivalence $`h`$. Notice that by its construction $`\delta `$ will be as smooth as $`\lambda `$. This gives us the following result.
###### Theorem 1.1.
The flow $`\varphi C^{k+\alpha }\left(N+\right)`$ with vector field $`v_\varphi `$ is almost periodic if and only if $`\lambda _e\left(t\right)\lambda _0`$ has a bounded integral, where $`\lambda _e\left(t\right)={\displaystyle \frac{1}{v_\varphi \pi _N\left(t\right)}}=\underset{r\widehat{N}}{}\lambda _r\mathrm{exp}\left(2\pi irt\right)`$.
If $`f\left(t\right)`$ is periodic with Fourier series $`\underset{r\widehat{N}}{}f_r\mathrm{exp}\left(2\pi irt\right)`$, a necessary and sufficient condition that it have a periodic integral is that
$$f_0=M\left\{f\right\}\stackrel{def}{=}\underset{T\mathrm{}}{lim}\frac{1}{T}_0^Tf\left(t\right)𝑑t=0\text{ (see }\text{[B]}\text{ §68),}$$
and since any function $`\lambda _e\left(t\right)\lambda _0`$ as above can be approximated (relative to $`d_k`$) arbitrarily closely by a periodic “partial series” (possibly containing infinitely many terms) of its Fourier–Bohr series and since $`M\left\{f\right\}=0`$ for any such partial series $`f`$, we obtain the following.
###### Corollary 1.2.
The collection of almost periodic flows in $`C^{k+\alpha }\left(N+\right)`$ is dense.
(We are making use of the fact that the function $`v_\varphi 1/v_\varphi =\lambda `$ is a homeomorphism of $`C^{k+\alpha }\left(N+\right).`$) We postpone the existence of flows for given functions until the next section.
Now we examine the functions $`\lambda =1/v_\varphi `$ corresponding to flows $`\varphi `$ which are not almost periodic. First, we note
$$\left|M\left\{f\right\}M\left\{g\right\}\right|\underset{T\mathrm{}}{lim}\frac{1}{T}_0^T\left|f\left(t\right)g\left(t\right)\right|𝑑tsup\left|f\left(t\right)g\left(t\right)\right|.$$
And so $`d_1(\lambda \lambda _0,\mu \mu _0)2d_1(\lambda ,\mu )`$. Now for $`n=1,2,\mathrm{}`$ we define the sets
$$U_n\stackrel{def}{=}\{\lambda \text{ there is a }T_n\text{ with }\left|_0^{T_n}\left(\lambda _e\left(t\right)\lambda _0\right)𝑑t\right|>n\}$$
and claim that $`U_n`$ is open for $`n=1,2,\mathrm{}`$ . So suppose then that $`\lambda U_n`$ with
$$\delta =\left|_0^{T_n}\left(\lambda _e\left(t\right)\lambda _0\right)𝑑t\right|n>0.$$
Now if $`d_1(\lambda ,\mu )d_k(\lambda ,\mu )<{\displaystyle \frac{\delta }{3\left|T_n\right|}}`$ we have
$$\left|_0^{T_n}\left(\lambda _e\left(t\right)\lambda _0\right)𝑑t_0^{T_n}\left(\mu _e\left(t\right)\mu _0\right)𝑑t\right|\left|_0^{T_n}\frac{2\delta }{3\left|T_n\right|}𝑑t\right|<\delta $$
and so $`\mu `$ is also in $`U_n`$, demonstrating that $`U_n`$ is open. And so the $`G_\delta `$ $`_{n=1}^{\mathrm{}}U_n`$ is the collection of $`\lambda =1/v_\varphi `$ corresponding to flows $`\varphi `$ which are not almost periodic.
It then remains to show that $`_{n=1}^{\mathrm{}}U_n`$ is dense. Let $`\lambda =1/v_\varphi `$ be given. We need to find an arbitrarily close function which corresponds to a flow which is not almost periodic. Consider for $`m=1,2,\mathrm{}`$ the $`C^{\mathrm{}}`$ maps $`\mathrm{\Sigma }_N`$
$$\rho _m\left(x\right)\stackrel{def}{=}\underset{j=m}{\overset{\mathrm{}}{}}\frac{1}{n_1\mathrm{}n_j}\chi _{\frac{1}{n_1\mathrm{}n_j}}\left(x\right).$$
Bohr’s theorem shows that $`\rho _m\pi _N\left(t\right)`$ has an unbounded integral since Parseval’s equation for almost periodic functions would fail for the Fourier–Bohr series of a bounded $`_0^T\rho _m\pi _N\left(t\right)𝑑t`$, implying that at least one of $`\text{ Re}\left(\rho _m\pi _N\left(t\right)\right)`$ and $`\text{Im}\left(\rho _m\pi _N\left(t\right)\right)`$ too has an unbounded integral. (Notice that both the real and imaginary parts of a function with mean value $`0`$ also have mean value $`0`$.) If $`\lambda _e\left(t\right)\lambda _0`$ has unbounded integral, there is nothing to prove; and if not, we can choose $`\lambda +\text{Re}\rho _m`$ and $`\lambda +\text{Im}\rho _m`$ positive and as close to $`\lambda `$ as desired by choosing $`m`$ large enough since $`\left|\rho _m\left(x\right)\right|{\displaystyle \frac{1}{2^{m1}}}`$ (and similarly for the derivatives), and at least one of these two functions (say $`\lambda ^m`$) will be such that $`\lambda ^m\pi _N`$ will have an unbounded integral when its mean value $`\lambda _0`$ is subtracted.
###### Corollary 1.3.
The collection of flows in $`C^{k+\alpha }\left(N+\right)`$ which are not almost periodic is a dense $`G_\delta `$.
## 2. Realization of a flow for a given vector field
Let $`v`$ be a positive Lipschitz function $`\mathrm{\Sigma }_N`$. We seek a flow $`\varphi _v`$ on $`\mathrm{\Sigma }_N`$ which has $`v`$ as its vector field. For $`n2`$ there is a $`C^{\mathrm{}}`$ flow $`\varphi `$ on the tube $`\mathrm{\Pi }_0=B^{2n1}\times S^1`$ ($`B^{2n1}`$ is the closed unit ball of $`^{2n1}`$) satisfying:
1. $`\varphi `$ has a homeomorphic copy $`𝔖_N`$ of $`\mathrm{\Sigma }_N`$ as a limit set
2. $`d\psi /dt1`$ along all the orbits of $`\varphi ,`$ where $`0\psi <1`$ parameterizes the $`S^1`$ factor of $`\mathrm{\Pi }_0`$ (see \[MM\], p. 87).
The limit set $`𝔖_N`$ is realized as the nested intersection of compact tubes $`\mathrm{\Pi }_j`$, $`j=0,1,2,\mathrm{}`$ , satisfying for $`j=0,1,\mathrm{}`$ :
1. $`\mathrm{\Pi }_j`$ is homeomorphic to $`\mathrm{\Pi }_0`$ and
2. $`\mathrm{\Pi }_{j+1}`$ encircles $`\mathrm{\Pi }_j`$$`n_{j+1}`$ times.
Therefore, $`\varphi `$ restricted to $`𝔖_N`$ has the constant return time of $`1`$ to the sections
$$𝒮_t\stackrel{def}{=}\{(x,\psi )B^{2n1}\times S^1𝔖_N\psi =t\}$$
and is thus equivalent to the linear flow $`\varphi _N^1`$ on $`\mathrm{\Sigma }_N`$ via a homeomorphism, say $`h`$, where without loss of generality $`\psi (h^1(e))=0`$. Associated with the flow $`\varphi `$ is a vector field $`𝔳`$. We shall obtain the desired flow $`\varphi _v`$ by a time change of the flow $`\varphi `$: we shall multiply the vectors of $`𝔳`$ by a function to change their lengths but not their directions and then use $`h`$ to obtain $`\varphi _v`$. (Here we are measuring lengths of vectors by the lengths of their projections onto the $``$ factor in the tangent bundle corresponding to the $`S^1`$ factor in $`\mathrm{\Pi }_0`$.)
Setting $`\psi _0=\psi `$, for $`j=1,2,\mathrm{}`$ let $`0\psi _j<n_1\mathrm{}n_j`$ parameterize the $`S^1`$ factor of $`\mathrm{\Pi }_j`$ in such a way that for $`(x,\psi _j)=(x,\psi _{j1})`$ in $`\mathrm{\Pi }_j`$ we have $`\psi _j=\psi _{j1}\left(\text{mod}1\right)`$ and $`\psi _j(h^1(e))=0`$. (We could use $`\varphi `$–sections transverse to the elliptic periodic orbits $`\gamma _0,\gamma _1,\mathrm{}`$ for example (see \[MM\], p. 87).) For $`j=0,1,2,\mathrm{}`$ let $`s_j\left(t\right)`$ be a purely periodic Lipschitz partial sum of $`v_e\left(t\right)=\underset{j\mathrm{}}{lim}s_j\left(t\right)=\underset{r\widehat{N}}{}v_r\mathrm{exp}\left(2\pi irt\right)`$, where the period of $`s_0`$ is $`1`$ and the period of $`s_j`$ is $`n_1\mathrm{}n_j`$ for $`j=1,2,\mathrm{}`$ (see \[Be\] 1§8). We then form for $`j=0,1,2,\mathrm{}`$ the following functions $`\tau _j:\mathrm{\Pi }_0`$; $`\tau _0\left((x,\psi _0)\right)=s_0\left(\psi _0\right)`$$`\tau _1\left((x,\psi _0)\right)=\tau _0\left((x,\psi _0)\right)`$ for $`(x,\psi _0)`$ in $`\mathrm{\Pi }_0𝒯_1`$ (where $`𝒯_1`$ is a tube satisfying $`\mathrm{\Pi }_1𝒯_1\mathrm{\Pi }_0`$) and $`\tau _1\left((x,\psi _0)\right)`$ is $`\tau _0\left((x,\psi _0)\right)`$ gradually changed in $`𝒯_1\mathrm{\Pi }_1`$ until finally $`\tau _1\left((x,\psi _1)\right)=s_1\left(\psi _1\right)`$ for $`(x,\psi _1)\mathrm{\Pi }_1`$ . Continue in the same manner so that $`\tau _{j+1}\left((x,\psi _0)\right)=\tau _j\left((x,\psi _0)\right)`$ for $`(x,\psi _0)`$ in $`\mathrm{\Pi }_0𝒯_{j+1}`$ (where $`𝒯_{j+1}`$ is a tube satisfying $`\mathrm{\Pi }_{j+1}𝒯_{j+1}\mathrm{\Pi }_j`$) and $`\tau _{j+1}\left((x,\psi _0)\right)`$ is $`\tau _j\left((x,\psi _0)\right)`$ gradually changed in $`𝒯_{j+1}\mathrm{\Pi }_{j+1}`$ until finally $`\tau _{j+1}\left((x,\psi _{j+1})\right)=s_{j+1}\left(\psi _{j+1}\right)`$ for $`(x,\psi _{j+1})\mathrm{\Pi }_{j+1}`$. And then with $`\tau \left((x,\psi _0)\right)\stackrel{def}{=}\underset{j\mathrm{}}{lim}\tau _j\left((x,\psi _0)\right)`$ we have the function to obtain the desired time change of $`𝔳`$.
This also shows that for arbitrarily small time changes we can alter a generic class of Hamiltonian flows to obtain flows which are not Lyapunov stable on solenoidal minimal sets (see \[NS\] V§8). By adjusting the coefficients of the $`\rho _m`$ as in the previous section, we even have flows which have points on the same orbit arbitrarily close that “lap” one another relative to these tubes. However, it is not clear what happens if we restrict ourselves to Hamiltonian flows. To make matters even more complicated, it is not clear that these solenoids persist as limit sets under small perturbations (see \[AM\] 8§5).
|
no-problem/9906/cond-mat9906192.html
|
ar5iv
|
text
|
# Excitonic photoluminescence in symmetric coupled double quantum wells subject to an external electric field
## I Introduction
Because of their technical importance and unique physical properties, semiconductor quantum wells have been the subject of intensive research since their first fabrication in 1973. A brief review of the main achievements and a number of representative references can be found, e.g., in books and . The tunnel-coupled quantum states in DQWs are very sensitive to both electric and magnetic fields, and changes induced by these fields have been intensively studied by PL, PL excitation, and photoconductivity .
A symmetric coupled DQW consists of two identical quantum wells separated by only a thin barrier. The energy levels of the coupled QWs split owing to the interwell tunneling. In the flat-band condition $`(F=0)`$, the eigenfunctions of the DQW have defined symmetries. In this situation, only transitions between electron and hole states of the same symmetry are optically allowed. The ground-state wave function is symmetric, while that of the first excited state is antisymmetric. When an electric field is applied to the DQW, the wave functions become predominantly localized in one well only. Transitions allowed in the flat-band case evolve into spatially indirect (interwell) transitions, which shift linearly in energy as a function of the applied field. On the other hand, transitions forbidden in the flat-band case become allowed and represent spatially direct (intrawell) transitions. Two factors determine the optical transition intensities: (a) the overlap integral of the electron and hole single-particle wave functions, and (b) the population of the energy subbands for a given temperature. In order to describe the optical transitions correctly one has to account for the Coulomb interaction between electrons and holes. We use a two-particle wave function composed of the electron and hole single-particle wave functions multiplied by a function of their relative positions to get an exciton binding energy.
The purpose of this report is to demonstrate that our quite simple theory, applied to the PL results obtained from our specifically designed sample, provides excellent quantitative agreement with regard to both the energetic position of the PL peaks and their shapes and intensities as a function of applied electric fields. These results represent clearly a further improvement compared to the previously reported results .
## II Theory
The calculation starts from the envelope-function excitonic Hamiltonian
$$H_{ex}=H_{oe}+H_{oh}+H_{2D}+U+E_g,$$
(1)
where the respective electron and hole single-particle terms read
$$H_{o\nu }=\frac{\mathrm{}^2}{2}\frac{}{z_\nu }\frac{1}{m_\nu (z_\nu )}\frac{}{z_\nu }+V_\nu (z_\nu )q_\nu Fz_\nu ,$$
(2)
$`\nu =\{e,h\}`$ denotes an electron or hole, and $`q_\nu `$ is the respective charge. $`m_\nu (z_\nu )`$ and $`V_\nu (z_\nu )`$ are the respective electron or hole $`z_\nu `$ dependent effective masses and confining potentials of the DQW structure. The electric field $`F`$ is applied along the $`z`$ axis parallel to the growth direction. $`H_{2D}`$ represents the kinetic energy in the $`xy`$ plane. The electron-hole exciton interaction is included by means of the Coulomb term $`U`$, and the GaAs bandgap energy $`E_g`$ completes the total energy $`H_{ex}`$.
Eigenenergies and eigenfunctions of the single-particle 1D Hamiltonian $`H_{o\nu }`$ are found by using linear combinations of analytical functions sin$`(\xi )`$ and cos$`(\xi )`$ (wells, $`F=0`$), Airy functions Ai$`(\xi )`$ and Bi$`(\xi )`$ (wells, $`F>0`$), or exp$`(\pm \xi )`$ (barriers), and by matching the wave function amplitudes and their derivatives divided by the effective masses at each interface. Representative results of such a calculation made for a DQW structure with $`L_w=18`$ ML, $`L_b=4`$ ML ($``$ 5.09 nm, and 1.13 nm, respectively), and for $`F=30`$ kV/cm are schematically plotted in Fig. 1.
The first two single-particle eigenfunctions $`\phi _1^\nu (z_\nu )`$, $`\phi _2^\nu (z_\nu )`$ of $`H_{o\nu }`$ are used to construct a set of basis functions for the variational calculation of the exciton states as a single product of one electron and one hole wave function multiplied by a function of relative electron hole positions :
$$\chi _{ij}=N_{ij}\phi _i^e(z_e)\phi _j^h(z_h)\mathrm{exp}\left(\frac{\sqrt{\rho ^2+\alpha _{ij}(z_ez_h)^2}}{R_{ij}}\right),$$
(3)
where the normalization factors $`N_{ij}`$ and the variational parameters $`R_{ij},\alpha _{ij}`$ are, in general, different for each exciton. While the $`z`$ coordinates of both carriers are specified absolutely $`(z_e,z_h)`$, only the relative distance between them projected onto the $`xy`$ plane $`(\rho )`$ is relevant. We minimize the total energy \[Eq. (1)\] for each exciton separately and obtain four wave functions {$`\chi _{11},\chi _{12},\chi _{21},\chi _{22}`$}.
In our calculation, we neglect the heavy- and light-hole bands mixing and assume strictly parabolic dispersion relations. Also, the mixing of the exciton states by the Coulomb interaction is not taken into account because of its small effect on the narrow DQW system. The GaAs/AlGaAs/AlAs material parameters and their compositional dependence is taken as in .
The $`\chi _{ij}`$s are used to evaluate the optical oscillator strength (see, e.g., Ref. ) and the electron-hole overlap integral $`F_{ij}(0)`$, which is given by
$$F_{ij}(0)=\underset{\mathrm{}}{\overset{\mathrm{}}{}}\chi _{ij}\text{(z}\text{e}\text{=z,z}\text{h}\text{=z,}\rho \text{=0)}𝑑z=N_{ij}\underset{\mathrm{}}{\overset{\mathrm{}}{}}\phi _i^e(z)\phi _j^h(z)𝑑z.$$
(4)
The normalized exciton PL intensity spectrum is then calculated as
$`I(E)`$ $`=`$ $`{\displaystyle \frac{\underset{i,j=1}{\overset{2}{}}L_{ij}(E)|F_{ij}(0)|^2e^{\beta E_{ij}}}{\underset{i,j=1}{\overset{2}{}}|F_{ij}(0)|^2e^{\beta E_{ij}}}},`$ (5)
$`L_{ij}(E)`$ $`=`$ $`{\displaystyle \frac{\beta \mathrm{\Delta }_{ij}}{\pi }}{\displaystyle \underset{0}{\overset{\mathrm{}}{}}}{\displaystyle \frac{e^{\beta E^{}}dE^{}}{(EE_{ij}E^{})^2+\mathrm{\Delta }_{ij}^2}},`$ (6)
$`\beta =1/k_BT`$, $`k_B`$ is the Boltzmann constant and $`T`$ the temperature. The convolution form $`L_{ij}(E)`$ expresses the normalized line-shape of the $`i,j`$th transition, which is determined by a thermal distribution of the excitonic kinetic energy and slightly diffused by a Lorenz function. The width $`\mathrm{\Delta }_{ij}=0.5`$ meV is used in every case.
## III Experimental
Our PL experiments were performed on a sample grown by MBE at a temperature of 600<sup>o</sup>C on a semi-insulating GaAs substrate oriented in the direction. The growth started with a 500 nm wide n-doped (Si, $`2\times 10^{18}`$ cm<sup>-3</sup>) GaAs layer, followed by a 300 nm wide n-doped (Si, $`1.4\times 10^{18}`$ cm<sup>-3</sup>) GaAlAs layer. The Al content in the GaAlAs layers was always 0.3. After this the following sequence of GaAlAs layers was grown: 500 nm intrinsic, 5 nm p-$`\delta `$-doped (C, $`3\times 10^{17}`$ cm<sup>-3</sup>) layer, and 100 nm intrinsic. On top of this separating layer, a sequence of three symmetric DQWs with 4 MLs ($``$ 1.13 nm) wide pure AlAs central barriers in between were grown, employing growth interruptions of 20 s at each heterointerface. The well widths are 35 MLs, 26 MLs, and 18 MLs ($``$ 10 nm, 7.5 nm, and 5 nm), respectively. The DQWs are separated by a 100 nm wide GaAlAs layer in each case. The growth then continued with another sequence of GaAlAs layers: 100 nm intrinsic, 5 nm n-$`\delta `$-doped (Si, $`4\times 10^{17}`$ cm<sup>-3</sup>), and 500 nm intrinsic. On the top a 300 nm wide p-doped (C, $`1.4\times 10^{18}`$ cm<sup>-3</sup>) GaAlAs layer and a 20 nm wide p-doped (C,$`2\times 10^{18}`$ cm<sup>-3</sup>) GaAs cap layer were grown.
The p-i-n configuration of the sample allows us to apply a bias voltage $`U_{pn}`$ by means of selective Ohmic contacts to the p-doped layers on the top and to the n-doped layers at the bottom. Usually, the bands in any p-i-n structure are tilted and application of forward bias is required to flatten them. We, therefore, used a special sample design comprising $`\delta `$-doped layers inside the intrinsic region of the structure in order to screen the built-in electric field. Thus, the flat-band regime is obtained almost at $`U_{pn}=0`$ V, avoiding high dark current present at higher forward bias. The measured devices of the size $`250\mu `$$`\times 250\mu `$m were defined photolithographically and mesa-isolated. Detailed results will only be presented for the narrow DQW with $`L_w=5`$ nm and $`L_b=1.13`$ nm, for which the application of our simple theoretical model is justified.
Figure 2 shows the PL spectra of the sample for various bias voltages. The sample was cooled in a closed-cycle cryostat and excited by a Ti:sapphire laser pumped by an Ar<sup>+</sup>-ion laser. We used an excitation power of $`100`$ mW/cm<sup>2</sup> at a photon energy of 1722 meV (below the bandgap energy of Ga<sub>0.7</sub>Al<sub>0.3</sub>As). The emitted luminescence was analyzed by a monochromator with 0.6 m focal length and a 1200 grooves/mm grating, and detected by a cooled CCD camera. We applied reverse bias up to $`U_{pn}=8`$ V with the maximum current of only $`10\mu `$A. An estimation of a dissipated power and a negligible shift in position of the spectrum measured with $`U_{pn}=+2`$ V proved that Ohmic heating of the sample for both reverse and forward biased junction is insignificant.
## IV Discussion
The sample was designed as a single (not multiple) DQW structure in order to minimize the well width fluctuations. For the same reason, the growth interruptions were employed, and a pure AlAs barrier instead of an Al<sub>x</sub>Ga<sub>1-x</sub>As one was chosen to improve its homogeneity and to avoid barrier height fluctuations. The broadening of the direct exciton peak of $`4`$ meV (see Fig. 2) and even lower at 10 K ($`2`$ meV, not shown) indicates a quite good quality of the sample. Furthermore, the exciton peaks are not split due to large area monolayer fluctuations of the well widths, a phenomenon which is frequently observed on samples grown with interruptions.
The $`\delta `$-doped layers shielding the DQWs within the p-n junction are effective in case of an excitation below the Ga<sub>0.7</sub>Al<sub>0.3</sub>As bandgap energy, when the carriers are excited exclusively in the wells. The flat band condition is found almost at zero bias voltage. On the other hand, when using a He-Ne laser excitation above bandgap energy, the $`\delta `$-doped layers are neutralized and become inactive. Consequently, a forward bias corresponding to the bandgap energy is needed in order to overcome the built-in field and to establish the flat-band regime.
The experimental spectra plotted in Fig. 2 were analyzed to gain maximum data for the theoretical procedure. In the calculation, we kept the parameters $`L_w`$ and $`L_b`$ constant and varied only $`T`$ and $`F`$ for optimization. The temperature $`T=45`$ K of the sample was obtained by a comparison of the measured and calculated relative peak intensities of the spectra in the linear field range. The field $`F`$ was deduced by means of tracing the energy distance between the first two exciton transitions, $`\chi _{11}`$ and $`\chi _{12}`$. A very good agreement of the calculated and observed transition energies is represented by Fig. 3. Notice that the optimized $`T`$ provides an excellent correspondence in the peak positions. The absolute positions of all the peaks including the resonant splitting of the symmetric-antisymmetric states were obtained without any free parameter. The small difference in the intensities and positions of the theoretical lines $`1`$ meV in comparison with the experiment can be attributed to (i) the approximations used in the theory and (ii) small deviations from the intended structure of the sample.
The inset of Fig. 4 shows the deduced field $`F`$ as a function of the bias voltage applied to the sample. In the case of the spectrum close to flat-band regime ($`U_{pn}=0V`$), the corresponding field ($`F=1.6`$ kV/cm) was determined from a linear part of this dependence. As anticipated, the dependence is linear within a limited voltage interval ($`3VU_{pn}2V`$), and saturates for higher reverse bias. The linear regime of the band tilting is in very good agreement with $`F`$ calculated directly from the sample design (dotted line in the inset of Fig. 4). A small discrepancy can be explained as an effect of a non-homogeneous distribution of the electric field over the three DQW systems, or as a result of a too simple description of the single wave functions in the applied variational method.
Figure 4 shows the spectra calculated for the studied DQW structure for the optimized temperature and the fields corresponding to individual PL curves plotted in Fig. 2. The spectra are magnified in accordance with the measured integral intensities $`N(U_{pn})`$ (see the inset of Fig. 2) to allow a comparison with the measured PL. The optimized temperature seems to be reasonable taking into account that the sample in a closed-cycle cryostat is cooled by a cold finger and placed on a rather low thermally conductive substrate. The close concert of the experimental and calculated PL spectra in the low field interval ($`3VU_{pn}2V`$) is evident. By increasing reverse bias the observed indirect transition diminishes. This results from a low stability of the lowest energy $`\chi _{11}`$ exciton which becomes dissociated by electron tunneling outside the DQW. The highest transition $`\chi _{22}`$ appears near the flat-band case only. It manifests itself in Fig. 4 in the spectrum calculated for $`F=1.6`$ kV/cm as the small peak at the energy of 1626 meV close to the $`\chi _{21}`$ transition. In the experiment, these transitions merge into one peak observed at 1623 meV. Out of the flat-band case, $`\chi _{22}`$ is suppressed due to a weak population of the higher levels and the lower value of the electron-hole overlap integral $`F_{22}(0)`$ \[Eq. (4)\]. With respect to the good agreement of relative intensities of the spatially direct and indirect transitions we can conclude that the overlap integrals are well estimated and the particle distribution is described by our wave functions in a satisfactory way.
## V conclusion
In this paper we have concentrated on the photoluminescence of narrow symmetric coupled double quantum wells, where the energy difference between excitonic states is high enough to allow neglecting of band mixing effects. The inclusion of exciton interaction is unavoidable for a correct description of the electron-hole states and positioning of optical transitions. We have shown that the simple model describes nearly perfectly all dominant features of the optical transitions, namely relative intensities, absolute positions of the individual transitions, and the resonant splitting of the symmetric-antisymmetric states.
## ACKNOWLEDGMENTS
This work was supported by the Friedrich-Alexander-Universität Erlangen-Nürnberg, by the Grant Agency of the Czech Republic (Grant No. 202/95/1533 and No. 202/98/0085), and by the Ministry of Education of the Czech Republic (Grant No. VS-97113). R. Grill acknowledges support from the Alexander von Humboldt Foundation (AvH, Bonn, Germany).
|
no-problem/9906/chao-dyn9906031.html
|
ar5iv
|
text
|
# Hamiltonian for a restricted isoenergetic thermostat
## Abstract
Nonequilibrium molecular dynamics simulations often use mechanisms called thermostats to regulate the temperature. A Hamiltonian is presented for the case of the isoenergetic (constant internal energy) thermostat corresponding to a tunable isokinetic (constant kinetic energy) thermostat, for which a Hamiltonian has recently been given.
Thermostats are modifications to the equations of motion of a classical system to simulate thermal interaction of a system with the environment. The Nosé-Hoover thermostat is used to simulate fluctuations in energy of an equilibrium system corresponding to the canonical ensemble of statistical mechanics, and the Nosé-Hoover and Gaussian thermostats, among others, are used to remove heat from a system driven by external forces into a nonequilibrium stationary state . There has been a recent interest in thermostatted equations of motion, focussed on the symplectic structure of the equations of motion, and the related pairing of the Lyapunov exponents. Both a Hamiltonian and pairing of Lyapunov exponents are known for Nosé-Hoover and Gaussian isokinetic (GIK: constant kinetic energy) thermostats . Numerical evidence against pairing (and hence the existence of a Hamiltonian) are discussed in for the GIK thermostat applied to shearing systems and in for the Gaussian isoenergetic (GIE: constant internal energy) thermostat. The latter paper does, however show that in a special case of the GIE thermostat, involving one rather than two arbitrary potentials, the Lyapunov exponents are paired. The purpose of this short note is present a Hamiltonian for this case.
The GIE thermostat has equations of motion of the form
$`{\displaystyle \frac{d𝐱_i}{dt}}={\displaystyle \frac{𝐩_i}{m_i}},`$ $`{\displaystyle \frac{d𝐩_i}{dt}}={\displaystyle \frac{\mathrm{\Phi }^{(ext)}}{𝐱_i}}{\displaystyle \frac{\mathrm{\Phi }^{(int)}}{𝐱_i}}\alpha 𝐩_i,`$ (1)
$`\alpha `$ $`=`$ $`{\displaystyle \frac{_i\frac{𝐩_i}{m_i}\frac{\mathrm{\Phi }^{(ext)}}{𝐱_i}}{_i𝐩_i𝐩_i/m_i}},`$ (2)
where $`\mathrm{\Phi }^{(ext)}`$ is the external driving potential, $`\mathrm{\Phi }^{(int)}`$ the interparticle potentials, and $`\alpha `$ is the thermostat term which ensures that the equations conserve internal energy $`E=_i𝐩_i^2/(2m_i)+\mathrm{\Phi }^{(int)}`$. The equations reduce to no thermostat when $`\mathrm{\Phi }^{(ext)}=0`$ and to GIK when $`\mathrm{\Phi }^{(int)}=0`$. A more general example of a limit involving only one arbitrary potential is the case $`\mathrm{\Phi }^{(ext)}=\gamma \mathrm{\Phi }`$, $`\mathrm{\Phi }^{(int)}=(1\gamma )\mathrm{\Phi }`$, leading to the equations
$$\frac{d𝐱_i}{dt}=\frac{𝐩_i}{m_i},\frac{d𝐩_i}{dt}=\frac{\mathrm{\Phi }}{𝐱_i}+\gamma \frac{_i\frac{𝐩_i}{m_i}\frac{\mathrm{\Phi }}{𝐱_i}}{_i𝐩_i𝐩_i/m_i}𝐩_i,$$
(3)
which conserve energy $`E=_i𝐩_i^2/(2m_i)+(1\gamma )\mathrm{\Phi }`$. Here, $`\gamma `$ effectively controls the strength of the thermostat from no thermostat ($`\gamma =0`$), to the GIK thermostat ($`\gamma =1`$). For any $`\gamma `$ the Lyapunov exponents are paired , suggesting the existence of a Hamiltonian.
Following the GIK case , the conservation law is enforced by setting the numerical value of the Hamiltonian equal to the conserved energy, assigned the value zero by a shift in the potential energy. This allows the kinetic energy term in the denominator of (2) to be replaced by minus the potential energy (note $`\mathrm{\Phi }<0`$),
$$\alpha =\frac{\gamma }{2(1\gamma )}\underset{i}{}\frac{𝐩_i}{m_i}\frac{}{𝐱_i}\mathrm{ln}|\mathrm{\Phi }|.$$
(4)
Another aspect of a Hamiltonian description of thermostatted systems is that in the physical variables $`(𝐱,𝐩)`$ there is a phase space volume contraction rate proportional to $`\alpha `$, while in the canonical variables $`(𝐱,𝝅)`$ phase space volume is conserved. This means that $`𝝅`$ must be greater than $`𝐩`$ by a factor equal to $`\mathrm{exp}\left(\alpha 𝑑t\right)=|\mathrm{\Phi }|^{\gamma /(2(1\gamma ))}`$. Multiplying the zero energy by an arbitrary power of $`|\mathrm{\Phi }|`$ we have
$$H_\beta (𝐱,𝝅,\lambda )=|\mathrm{\Phi }|^{\frac{\gamma }{1\gamma }+\beta }\underset{i}{}\frac{𝝅_i^2}{2m_i}+(1\gamma )\mathrm{\Phi }|\mathrm{\Phi }|^\beta ,$$
(5)
which, combined with the constraint $`H_\beta =0`$ and the identifications $`dt=|\mathrm{\Phi }|^{\gamma /(2(1\gamma ))+\beta }d\lambda `$ and $`𝐩_i=|\mathrm{\Phi }|^{\gamma /(2(1\gamma ))}𝝅_i`$ leads to the equations of motion, (3). Interesting cases are $`\beta =\gamma /(2(1\gamma ))`$ for which there is no time scaling, $`\beta =0`$ has a certain simplicity, $`\beta =\gamma /(1\gamma )`$ yields the familiar form of kinetic plus potential energy, and $`\beta =1`$ for which the Hamiltonian is that of a geodesic in a conformally flat space, see .
The author is grateful for discussions with W. G. Hoover.
|
no-problem/9906/cond-mat9906357.html
|
ar5iv
|
text
|
# Coherent dynamics of Bose-Einstein condensates in high-finesse optical cavities
## Abstract
We study the mutual interaction of a Bose-Einstein condensed gas with a single mode of a high-finesse optical cavity. We show how the cavity transmission reflects condensate properties and calculate the self-consistent intra-cavity light field and condensate evolution. Solving the coupled condensate-cavity equations we find that while falling through the cavity, the condensate is adiabatically transfered into the ground state of the periodic optical potential. This allows time dependent non-destructive measurements on Bose-Einstein condensates with intriguing prospects for subsequent controlled manipulation.
PACS number(s): 3.75.Fi, 3.65.Bz, 42.50.Vk, 42.50.Gy
Since the first experimental realisations of Bose-Einstein condensation in dilute gases , the properties and possible applications of condensates in various situations have been investigated. Most recently, attention has been drawn to the study of condensates in optical lattices as have been intensely used in the context of laser cooling and trapping of clouds of noninteracting atoms . But whereas the occupation of the lattice sites in an optical molasses for a cloud of laser cooled atoms at best is one atom per ten wells, the atomic densities found in a condensate allow for multiple occupation of each single well, which gives rise to a variety of new phenomena.
In this Letter we investigate the case of a condensate falling through a driven high-finesse optical cavity. The strong coupling of the condensed atoms to the cavity mode changes the resonance frequency of the cavity which hence is shifted into or out of resonance with the driving field. Consequently, the intracavity field intensity is modified and this can easily be measured by detecting the cavity output field. We show that according to the collective nature of the condensate this gives a measurable effect even for very low field intensities and for detunings from the atomic resonance frequency so large that the spontaneous scattering of photons is negligible. The proposed system should allow us to perform non-destructive measurements on the condensate. Similar systems have been used recently to predict amplification of matter waves and the appearance of dressed condensates .
Let us first introduce in more detail our model system, which is similar to that used recently to study the effect of a dynamically changing cavity field on the motion of a single atom .
We consider a Bose condensate consisting of $`N`$ two-level atoms of resonance frequency $`\omega _a`$ and spontaneous decay rate $`\mathrm{\Gamma }`$ falling through an optical cavity. The atom-cavity coupling is
$$g(x,t)=g_0\mathrm{cos}(kx)e^{(v_zt)^2/(2w^2)},$$
(1)
for a cavity mode in the form of a standing wave in the longitudinal direction and a Gaussian with waist $`w`$ transversally. The condensate is assumed to fall with constant velocity $`v_z`$, meaning that we neglect the transverse light forces on the atoms and the gravitational acceleration in the interaction region. Two further assumptions have been made at this point. First, the spatial extension of the condensate has to be small compared to the waist $`w`$ of the cavity in order to allow a quasi one-dimensional treatment. Second, we assume that the induced resonance frequency shift of the cavity is much smaller than the longitudinal mode spacing, so that we can restrict the model to a single longitudinal mode with wavenumber $`k`$.
The cavity with resonance frequency $`\omega _c`$ and cavity decay rate $`\kappa `$ is externally driven by a laser of frequency $`\omega `$ with pump amplitude $`\eta `$ and is treated classically, that is, the intra-cavity field is described by a (complex) field amplitude $`\alpha `$.
As we are interested in the limit where the condensate is not destroyed by the light field, we will assume a large detuning $`\mathrm{\Delta }_a=\omega \omega _a\mathrm{\Gamma }`$ of the driving laser from the atomic resonance such that the saturation parameter $`s=g_0^2/\mathrm{\Delta }_a^21`$. Moreover, we want the cavity decay to dominate over the spontaneous decay of all atoms and thus impose the condition
$$\kappa N\mathrm{\Gamma }s.$$
(2)
In this realistically achievable limit we are not only allowed to adiabatically eliminate the excited state of the atoms but also to completely omit the effect of atomic decay.
Hence we obtain the equation of motion for the field amplitude
$$\dot{\alpha }(t)=\left[i\mathrm{\Delta }_ciNU(\widehat{x},t)\kappa \right]\alpha (t)+\eta $$
(3)
where $`\mathrm{\Delta }_c=\omega \omega _c`$ is the cavity-pump detuning, $`U(x,t)=g(x,t)^2/\mathrm{\Delta }_a`$ the optical potential per photon, and “$`\mathrm{}`$” denotes the expectation value taken with respect to the condensate wave function $`|\psi (t)`$ at this time. This term describes the action of the condensate on the cavity field: the refractive index of the condensate shifts the resonance frequency of the cavity by an amount of $`NU(\widehat{x},t)`$. (With very high finesse optical cavities this effect has already been observed even for a single atom , that is, for $`N=1`$.) If we require that this effect significantly changes the intra-cavity field intensity in order to yield a measurable difference in the cavity output, then the maximum frequency shift must be of the order of or larger than the cavity line width $`\kappa `$, which implies
$$Ng_0^2/\mathrm{\Delta }_a\kappa .$$
(4)
From this we obtain an order of magnitude for the required detuning $`\mathrm{\Delta }_a`$ which we insert into Eq. (2) to obtain the following condition for the cavity parameters:
$$\frac{Ng_0^2}{\mathrm{\Gamma }\kappa }1.$$
(5)
The condensate wave function itself obeys a nonlinear Schrödinger equation (known as the Gross-Pitaevskii equation) with the Hamiltonian
$$H=\frac{\widehat{p}^2}{2m}+|\alpha (t)|^2U(\widehat{x},t)+Ng_{coll}|\psi (\widehat{x},t)|^2$$
(6)
where the last term describes two-particle collisions between the condensed atoms and is related to the s-wave scattering length $`a`$ by $`g_{coll}=4\pi \mathrm{}^2a/m`$. This Hamiltonian, together with Eq. (3) for the cavity field, forms a set of coupled nonlinear equations describing the dynamics of the compound system formed by the condensate and the optical cavity.
In this Letter we will only consider the special case where the cavity decay rate $`\kappa `$ is much larger than the oscillation frequency of bound atoms in the optical potential of the cavity. In this limit the intra-cavity field amplitude follows adiabatically the condensate wavefunction and hence at any time is given by
$$\alpha (t)=\frac{\eta }{\kappa i[\mathrm{\Delta }_cNU(\widehat{x},t)]}.$$
(7)
Thus, the light intensity of the cavity output, which is proportional to $`|\alpha |^2`$, provides information about the condensate wavefunction. In the following we will investigate this effect in certain special parameter limits.
Let us first consider the simple case where the cavity field is weak enough and the interaction time $`\tau =w/v_z`$ is short enough such that the condensate wavefunction remains essentially unperturbed (flat on the length scale of an optical wavelength). In this limit the cavity field can be evaluated analytically by inserting the frequency shift per atom
$$U(\widehat{x},t)=\frac{U_0}{2}e^{(v_zt)^2/w^2}$$
(8)
into Eq. (7), where we have introduced $`U_0=g_0^2/\mathrm{\Delta }_a`$. In Fig. 1 the resulting mean cavity photon number $`I=|\alpha |^2`$ is plotted as a function of time for different atom numbers in the condensate or, equivalently, for different optical potentials.
For the parameters chosen in Fig. 1, the empty cavity is in resonance with the driving field but is shifted out of resonance by the presence of the condensate. Hence the condensate is detected by the absence of light, which further reduces the action of the cavity onto the condensate. Therefore the cavity provides a non-perturbative method of detection. The maximum resolution of the detection is limited, however, by the cavity waist $`w`$ and is not good enough to detect fine structures such as condensate interference fringes. It might be useful, however, to measure the output of an atom laser as has recently been realised experimentally . Note also that this detection scheme only relies on the density of the condensate, not on its coherence, and thus in principle works as well with an incoherent cloud of atoms.
Let us now consider the opposite limit of a condensate falling very slowly through the cavity. We will find that under such conditions the condensate is adiabatically transfered into the lowest bound state of the optical potential and hence is strongly localized. In this case we must use the position and time dependent optical potential
$$U(\widehat{x},t)=U_0\mathrm{cos}^2(k\widehat{x})e^{(v_zt)^2/w^2},$$
(9)
so that the condensate at all times feels a periodic optical potential with periodicity $`\lambda /2`$. The condensate wavefunction is conveniently described in terms of Bloch states
$$\psi _{n,q}(x)=e^{iqx}\varphi _{n,q}(x)$$
(10)
where the functions $`\varphi _{n,q}`$, $`n0`$, are periodic and the Bloch momentum $`\mathrm{}q`$ is confined to the interval $`[\mathrm{}k,\mathrm{}k]`$. Note that the coherent interaction of the condensate with the cavity light only couples Bloch states of different $`n`$ but leaves states of different $`q`$ decoupled and thus $`q`$ is a conserved quantity.
We are interested in the adiabatic limit corresponding to small transverse velocities $`v_zw/\tau _R`$, where $`\tau _R`$ is the inverse of the recoil frequency $`\omega _R=\mathrm{}k^2/(2m)`$. In this limit the coherent time evolution associated with a potential of the form of Eq. (9) maps the Bloch energy bands $`n`$ onto the free space momentum intervals $`[(n+1)\mathrm{}k,n\mathrm{}k]`$ and $`[n\mathrm{}k,(n+1)\mathrm{}k]`$. This phenomenon is known and has been exploited in the context of laser cooling . The same effect can be used in our model to transfer a falling condensate with a transverse momentum distribution confined in $`[\mathrm{}k,\mathrm{}k]`$ into the lowest energy band ($`n=0`$) of the optical potential inside the cavity. However, the situation is more complex here than for the case of laser cooling for two reasons. First, the light intensity itself depends on the condensate wavefunction and hence on time. Second, the condensate wavefunction obeys the nonlinear Schrödinger equation. Hence, for any given z-position of the condensate (any given time) the lowest energy state has to be found by self-consistently solving Eq. (7) and the Gross-Pitaevskii equation.
We show in Fig. 2 the time evolution of the intra-cavity field intensity $`I`$ and the overlap $`P_0`$ of the condensate wavefunction with the lowest energy band as the condensate is falling through the cavity without atomic collisions ($`g_{coll}=0`$). For simplicity we assume that the initial condensate wavefunction is the free momentum state of zero momentum. The conservation of the Bloch momentum $`q`$ then ensures that at any time only the Bloch states with $`q=0`$ are populated and the overlap with the lowest energy band is given by
$$P_0=|\psi |\varphi _{n=0,q=0}|^2.$$
(11)
For a condensate falling with a velocity $`v_z=w/\tau _R`$ the transfer of the free wavefunction into the optical potential is not adiabatic (especially not on entering and leaving the cavity) and the occupation of the ground state drops to about 90%. Hence there is a significant occupation of excited modes and the corresponding spatial oscillation of the condensate is reflected in the oscillation of the cavity output. For $`v_z=w/(3\tau _R)`$, however, nearly all of the population is transfered into the lowest bound state and accordingly the oscillations are suppressed.
For the parameters of Fig. 2, $`U_0>0`$ and the condensate is attracted to the nodes of the light field. Hence, the lowest bound state is localized at these positions which leads to a much reduced coupling of the condensate to the cavity and correspondingly to a much smaller frequency shift of the cavity resonance. This can be easily seen by comparing the results for the cavity field with the solid line of Fig. 1, which is taken for the same parameters but a condensate falling so fast that the wavefunction does not have the time to change due to the presence of the optical potential wells.
Figure 3 shows the effects of collisions on the intracavity photon number in the adiabatic regime ($`v_zw/\tau _R`$). The curves are derived from the self-consistent solution of Eq. (7) and the Gross-Pitaevskii equation for different values of $`g_{coll}`$. For increasing $`g_{coll}`$ the atoms in the condensate increasingly repell each other, counteracting the confining effect of the optical potential. The wavefunction becomes broader and the coupling of the condensate to the cavity stronger, which in turn leads to larger shifts in the cavity resonance frequency and a reduced cavity field intensity. Hence, the decrease in the cavity output provides a direct measure of the atom-atom interaction within the condensate. Note that this could be used for in situ measurements of Feshbach resonances, if one manipulates the s-wave scattering length by applying a magnetic field .
Finally, we plot in Fig. 4 the energy of the self-consistent ground state and of the two lowest excitations. The latter are calculated in lowest order perturbation theory, that is, using the self-consistent ground state for the collisional term of the Hamiltonian (6). The energy difference between these states is the one which is responsible for the oscillation of the cavity output in the case of non-perfect adiabatic transfer of the condensate wavefunction into the lowest bound state as shown previously in Fig. 2.
For $`g_{coll}=0`$ the ground state is strongly localized at the antinodes of the cavity leading to a large intracavity field intensity. Hence the lowest three eigenstates are well approximated by harmonic oscillator states and the frequency difference between the ground state and the first excited state is the same as the frequency difference between the first and the second excited state. For increasing $`g_{coll}`$ the internal (collisonal) energy of the condensate increases. Simultaneously the effective potential (optical potential plus collisional term) for the excited modes changes its shape from a harmonic oscillator to a potential well with a nearly flat bottom. This is due to the well-known fact that the ground state of the Gross-Pitaevskii equation becomes proportional to the negative confining potential if the collisional term dominates over the kinetic energy term in the Hamiltonian (Thomas-Fermi limit). Hence the excited modes are those of a potential which looks more like a box in this limit and the level spacing increases between more highly excited modes. However, as the wavefunction broadens with increasing $`g_{coll}`$ the cavity field intensity decreases and above a certain value ($`Ng_{coll}28\omega _R`$ for the parameters in Fig. 4) no bound state exists in the optical potential. Hence the eigenfunctions above this critical value resemble free space momentum states which leads to the significant change in the behaviour of the spectrum of excitation energies.
In conclusion we have shown that a high-quality optical cavity provides a powerful tool to investigate properties of Bose-Einstein condensates in a non-destructive way. For slow initial velocities the condensate is adiabatically transfered into the self-consistent ground state of the optical potential, which contains ample information on condensate properties.
In addition, the system suggests many interesting applications. Using cavities with a decay rate $`\kappa `$ of the order of the condensate vibrational frequencies in the potential, the finite response time of the cavity implies a damping (or amplification) of the condensate oscillations as it has been shown for cooling and trapping of a single atom . By changing the intensity and/or the frequency of the driving laser depending on the cavity output one gets a handle for the controlled non-destructive in situ manipulation of the condensate wavefunction. Similarly, changing the magnitic field during the condensate passage should allow a direct measurement of the effective scattering length and the relaxation dynamics.
This work was supported by the United Kingdom EPSRC, the Austrian Science Foundation FWF (P13435), and the European Comission TMR network (FRMX-CT96-0077).
|
no-problem/9906/gr-qc9906004.html
|
ar5iv
|
text
|
# Best Approximation to a Reversible Process in Black-Hole Physics and the Area Spectrum of Spherical Black Holes
## I introduction
Can the assimilation of a test particle by a black hole be made reversible in the sense that all changes in the black-hole parameters can be undone by another suitable process ? This seemingly naive question goes deep into the basic laws of black-hole physics. A classical theorem of Hawking’s says that black-hole surface area cannot decrease. Hence, any physical process which increases the horizon area is obviously (classically) irreversible. The answer to the above question was given by Christodoulou (later generalized by Christodoulou and Ruffini for the case of charged point particles ) almost three decades ago. The assimilation of a (point) particle is reversible if it is injected at the horizon from a turning point of its motion. In such a case the black-hole surface area is left unchanged and the changes in the other black-hole parameters (mass, charge and angular-momentum) can be undone by another suitable (reversible) process.
However, as was pointed out by Bekenstein in his seminal work the limit of a point particle is not a legal one in quantum theory. As a concession to quantum theory Bekenstein ascribes to the particle a finite proper radius $`b`$ while continuing to assume, in the spirit of Ehrenfest’s theorem, that the particle’s center of mass follows a classical trajectory. Bekenstein has shown that the assimilation of the finite size neutral particle inevitably causes an increase in the horizon area. This increase is minimized if the particle is captured when its center of mass is at a turning point a proper distance $`b`$ away from the horizon :
$$(\mathrm{\Delta }\alpha )_{min}=2\mu b,$$
(1)
where the “rationalized area” $`\alpha `$ is related to the black-hole surface area $`A`$ by $`\alpha =A/4\pi `$ and $`\mu `$ is the rest mass of the particle. For a point particle $`b=0`$ and one finds $`(\mathrm{\Delta }\alpha )_{min}=0`$. This is Christodoulou’s result for a reversible process. However, a quantum particle is subjected to quantum uncertainty. According to Bekenstein’s analysis, a relativistic quantum particle cannot be localized to better than its Compton wavelength (This claim certainly is not correct when the particle has (locally measured) energy greater than its mass. In this paper we give a different and rigorous argument which leads to a lower bound on the increase in the black-hole surface area). Thus, $`b`$ can be no smaller than $`\mathrm{}/\mu `$. From here one finds a lower bound on the increase in the black-hole surface area due to the assimilation of a (neutral) test particle
$$(\mathrm{\Delta }\alpha )_{min}=2\mathrm{}.$$
(2)
It is easy to check that the reversible process of Christodoulou and Ruffini and the lower bound Eq. (2) of Bekenstein are valid only for non-extremal black holes. Thus, for non-extremal black holes there is a universal (i.e., independent of the black-hole parameters) minimum area increase as soon as one allows quantum nuances to the problem. This fact is used as one of the major arguments in favor of a uniformly spaced area spectrum for quantum black-holes .
The universal lower bound Eq. (2) derived by Bekenstein is valid only for neutral particles . In this paper we analyze the assimilation of a quantum (finite size) charged particle by a Reissner-Nordström black hole and show that the fundamental lower-bound on the increase in the black-hole surface area is smaller than the value given by Bekenstein for neutral particles.
## II Assimilation of a charged particle by a Reissner-Nordström black-hole
The major goal of this paper is to calculate the (inevitable) minimal increase in black-hole surface area caused by the assimilation of a particle of rest mass $`\mu `$, charge $`e`$ and proper radius $`b`$. We are interested in the area increase ascribable to the particle itself, as contrasted with any increase incidental to the process of bringing the particle to the black-hole horizon . For example, gravitational radiation emitted by the particle or by any device which might had lowered it into the hole will also cause an increase in the area. In addition, electromagnetic radiation emitted during the process of bringing the charged particle to the horizon will also result in an increase in the black-hole surface area. In this paper, as in the seminal work of Bekenstein , we ignore these incidental effects and concentrate on the inevitable increase in the black-hole surface area caused by the captured particle all by itself.
The external gravitational field of a spherically symmetric object of mass $`M`$ and charge $`Q`$ is given by the Reissner-Nordström metric
$$ds^2=\left(1\frac{2M}{r}+\frac{Q^2}{r^2}\right)dt^2+\left(1\frac{2M}{r}+\frac{Q^2}{r^2}\right)^1dr^2+r^2d\mathrm{\Omega }^2.$$
(3)
The black hole’s (event and inner) horizons are located at
$$r_\pm =M\pm (M^2Q^2)^{1/2},$$
(4)
The equation of motion of a charged particle on the Reissner-Nordström background is a quadratic equation for the conserved energy $`E`$ of the particle
$$r^4E^22eQr^3E+e^2Q^2r^2\mathrm{\Delta }(\mu ^2r^2+p_{\varphi }^{}{}_{}{}^{2})(\mathrm{\Delta }p_r)^2=0,$$
(5)
where $`\mathrm{\Delta }`$ is given by
$$\mathrm{\Delta }=r^22Mr+Q^2=(rr_{})(rr_+).$$
(6)
The quantities $`p_\varphi `$ and $`p_r`$ are the conserved angular momentum of the particle and its covariant radial momentum, respectively. It is useful to express this last quantity in terms of the physical component (in an orthonormal tetrad) $`P\mathrm{\Delta }^{1/2}rp^r`$ .
In order to find the change in the black-hole surface area caused by an assimilation of a point particle one should first solve Eq. (5) for $`E`$ and then evaluate it at the horizon $`r=r_+`$ of the black-hole. As was pointed in Refs. and , this increase is minimized (actually vanishes) if the particle is captured from a turning point. How would the non-zero proper radius $`b`$ of the particle (which is an inevitable feature of the quantum theory) change this scenario ? First, we note that in the spirit of the Ehrenfest’s theorem we continue to assume that the particle’s center of mass follows a classical path. Second, as was pointed out by Bekenstein, regardless of the manner in which the particle arrives at the horizon, it must acquire its parameters ($`E`$ and $`p_\varphi `$) while every part of it is still outside the horizon, i.e., while it is not yet part of the black hole . Thus, the motion of the particle’s center of mass at the moment of capture should be described by Eq. (5). Third, in order to generalize the results given in Refs. and to the case of a finite size particle one should evaluate $`E`$ at $`r=r_++\delta (b)`$, where $`\delta (b)`$ is determined by
$$_{r_+}^{r_++\delta (b)}(g_{rr})^{1/2}𝑑r=b.$$
(7)
In other words, $`r=r_++\delta (b)`$ is a point a proper distance $`b`$ outside the horizon. Integrating Eq. (7) one finds
$$\delta (b)=(r_+r_{})sinh^2\left(\frac{b}{2r_+}\right)[1+O(b/r_+)].$$
(8)
Since we consider the case $`br_+`$ we may replace this expression by
$$\delta (b)=(r_+r_{})\frac{b^2}{4r_{+}^{}{}_{}{}^{2}}.$$
(9)
The conserved energy $`E`$ of a particle having a physical radial momentum $`P`$ at $`r=r_++\xi `$ (where $`\xi r_+`$) is given by Eq. (5)
$`E`$ $`=`$ $`{\displaystyle \frac{eQ}{r_+}}+{\displaystyle \frac{\sqrt{(\mu ^2+P^2)r_+^2+p_\varphi ^2}(r_+r_{})^{1/2}}{r_+^2}}\xi ^{1/2}\left\{1+O\left[\xi /(r_+r_{})\right]\right\}`$ (11)
$`{\displaystyle \frac{eQ}{r_+^2}}\xi \left\{1+O\left[\xi /(r_+r_{})\right]\right\}.`$
This expression (for $`P=0`$) is actually the effective potential (gravitational plus electromagnetic plus centrifugal) for given values of $`\mu ,e`$ and $`p_\varphi `$. It is clear that it can be minimized by taking $`p_\varphi =0`$ (which also minimize the increase in the black-hole surface area. This is also the case for neutral particles ). However, $`P^2`$ cannot be said to vanish because of Heisenberg quantum uncertainty principle. For $`eQ>0`$ the effective potential has a maximum located at
$$\xi ^{}=\frac{(r_+r_{})(\mu ^2+P^2)r_+^2}{4e^2Q^2}.$$
(12)
The assimilation of the particle results in a change $`dM=E`$ in the black-hole mass and a change $`dQ=e`$ in the black-hole charge. Using the first-law of black-hole thermodynamics
$$dM=\mathrm{\Theta }d\alpha +\mathrm{\Phi }dQ,$$
(13)
where $`\mathrm{\Theta }=\frac{1}{4}(r_+r_{})/\alpha `$ and $`\mathrm{\Phi }=Qr_+/\alpha `$, one finds
$$d\alpha _{min}(s,\mu ,e,b)=\frac{4(\mu ^2+P^2)^{1/2}r_+}{(r_+r_{})^{1/2}}\delta (b)^{1/2}\frac{4eQ}{r_+r_{}}\delta (b),$$
(14)
which is the minimal area increase for given values of the black-hole parameters $`r_+`$ and $`Q`$ ($`s`$ stands for these two parameters) and the particle parameters $`\mu ,e,b`$ and $`P`$.
In order to be captured by the black-hole the particle has to be over the potential barrier. There are two distinct cases that should be treated separately: for particles satisfying the relation $`\delta (b)\xi ^{}`$ the area increase is minimized if $`bP`$ is minimized. However, the limit $`bP0`$ is not a legal one in the quantum theory. According to Bekenstein’s analysis the particle cannot be localized to better than its Compton wavelength $`\mathrm{}/\mu `$ (We will discuss the validity of this assumption below. In this paper we shall use a more rigorous argument to provide a lower bound on the product $`bP`$). On the other hand, particles satisfying the inequality $`\delta (b)>\xi ^{}`$ cannot be captured from a turning point of their motion. In order to overcome the potential barrier and be captured by the black hole they must have (at least) an energy $`E(\xi ^{})`$.
Let us consider the first case $`\delta (b)\xi ^{}`$. Substituting Eq. (9) into Eq. (14) one finds
$$d\alpha _{min}(s,\mu ,e,b)=2(\mu ^2+P^2)^{1/2}b\frac{eQb^2}{r_+^2}.$$
(15)
According to Bekenstein’s original analysis one may minimize this expression by minimizing the value of $`b`$ (and setting $`P^2=0`$ at the turning point). However, the claim used in that a particle cannot be localized to within less than its Compton wavelength certainly is not correct when the particle has (locally measured) energy greater than its mass. The locally measured energy (by a static observer) of a particle near the horizon of a black hole can be arbitrarily large. In addition, at the turning point the physical radial momentum $`P^2`$ cannot be said to vanish, but must be replaced by its uncertainty $`(\delta P)^2`$ . In other words, according to Heisenberg quantum uncertainty principle the particle’s center of mass cannot be placed at the horizon with accuracy better than the radial position uncertainty $`\mathrm{}/(2\delta P)`$.
Using the restriction $`\delta (b)\xi ^{}`$ one finds
$$|e|\frac{(\mu ^2+P^2)^{1/2}r_+^2}{|Q|b}.$$
(16)
Thus, the minimal area increase is given by
$$d\alpha _{min}(s,\mu )=[\mu ^2+(\delta P)^2]^{1/2}b,$$
(17)
which, according to Heisenberg quantum uncertainty principle, yields (for $`\delta P\mu `$)
$$d\alpha _{min}(s,\mu )=\mathrm{}/2.$$
(18)
Next, we consider the assimilation of particles which satisfy the relation $`\delta (b)>\xi ^{}`$. These particles cannot be captured from a turning point of their motion. In order to be captured by the black hole they must have a minimal energy of
$$E_{min}=E(\xi ^{})=\frac{eQ}{r_+}+\frac{(\mu ^2+P^2)(r_+r_{})}{4eQ}.$$
(19)
Using the first-law of black-hole thermodynamics Eq. (13) one finds that the increase in the black-hole surface area is given by
$$d\alpha _{min}(s,\mu ,e)=\frac{(\mu ^2+P^2)r_+^2}{eQ}.$$
(20)
What physics prevents us from using particles which make expression (20) as small as we wish ? Or, in other words, what physics prevents us from recovering Christodoulou’s reversible process $`\mathrm{\Delta }A=0`$ ? The answer is Schwinger-type charge emission (vacuum polarization) . We must remember that the the black hole may discharge itself through a Schwinger-type emission. The critical electric field $`\mathrm{\Xi }_c`$ for pair-production of particles with rest mass $`\mu `$ and charge $`e`$ is given by
$$\mathrm{\Xi }_c=\frac{\pi \mu ^2}{e\mathrm{}}.$$
(21)
This order of magnitude can easily be understood on physical grounds; Schwinger discharge is exponentially suppressed unless the work done by the electric field on the virtual pair of (charged) particles in separating them by a Compton wavelength is of the same order of magnitude (or more) of the particle’s mass. Thus, assuming the existence of elementary particles with mass $`\mu `$ and charge $`e`$, a spherical black hole of charge $`Q`$ and radius $`r_+`$ (whose electric field near the horizon is $`\mathrm{\Xi }_+=Q/r_+^2`$) may be considered as quasi-static only if it satisfies the relation $`\mathrm{\Xi }_+\mathrm{\Xi }_c`$, or equivalently
$$|e|\frac{\pi \mu ^2r_+^2}{|Q|\mathrm{}}.$$
(22)
Substituting this into Eq. (20) one finds (for $`\mu ^2P^2`$)
$$d\alpha _{min}(s,\mu )=\frac{\mathrm{}}{\pi },$$
(23)
which is now the fundamental lower bound on the increase in the black-hole surface area. We note that this lower bound is universal in the sense that it is independent of the black-hole parameters $`M`$ and $`Q`$.
## III Summary and Discussion
We have studied the assimilation of a charged particle by a Reissner-Nordström black hole. The capture of a particle necessarily results in an increase in the black-hole surface area. The minimal area increase equals to $`4\mathrm{}`$. We note that this value is smaller than the value given by Bekenstein for neutral particles. Thus, this process is a better approximation to a reversible process in the context of black-hole physics.
As was pointed out by Bekenstein (for neutral particles) the underling physics which excludes a completely reversible process is the Heisenberg quantum uncertainty principle. However, for charged particles it must be supplemented by another physical mechanism - a Schwinger-discharge of the black hole. Without this mechanism one could have reached the reversible limit. It is interesting that the lower bound found here is of the same order of magnitude as the one given by Bekenstein, even-though they emerge from different physical mechanisms.
The universality of the fundamental lower bound (i.e., its independence on the black-hole parameters $`M`$ and $`Q`$) is a further evidence in favor of a uniformly spaced area spectrum for spherical quantum black-holes (see Ref. ). Moreover, the universal value $`\mathrm{\Delta }A_{min}=4\mathrm{}`$ is in excellent agreement (to within a factor of $`ln2`$) with the area spacing predicted by Mukhanov and Bekenstein and (to within a factor (of order unity) of $`ln3`$) with the area spacing predicted by Hod .
It should be recognized that the precise value of the fundamental lower bound Eq. (23) can be challenged. This lower bound follows from Eq. (21) which can only be interpreted as the critical electric field to within factors of few. Nevertheless, the new and interesting observation of this paper is the role of Schwinger pair production in providing an important limitation on the minimal increase in black-hole surface area.
ACKNOWLEDGMENTS
It is a great pleasure to thank Prof. Tsvi Piran for reading the manuscript. This research was supported by a grant from the Israel Science Foundation.
|
no-problem/9906/nucl-ex9906006.html
|
ar5iv
|
text
|
# 1 Introduction
## 1 Introduction
Correlations of identical particles emitted with similar velocities are commonly used as a tool to measure the space-time development of the emission process in particle and heavy-ion collisions. The technique is most often used for bosons, for which the effect of quantum (Bose-Einstein) statistics is the only source of correlations (photons), or plays a dominant role (pions and kaons). Other effects, in particular those due to the Coulomb interaction for pairs of charged particles, are usually treated as a correction. In the case of protons, final state interactions, due to strong and Coulomb forces, typically dominate the effect of quantum statistics and determine the form of the correlation function. A negative correlation effect, due to Fermi-Dirac statistics and Coulomb repulsion, competes with the positive (attractive) correlation due to the strong interaction, giving rise to a characteristic “dip-peak” structure (minimum at zero, maximum at $``$ 20 MeV/$`c`$ proton momentum in the pair rest frame) in the two-proton correlation function. The height of the correlation peak decreases as the proton source size increases. All effects are strongly sensitive to the space-time parameters of the emission process .
Correlations of protons emitted with small relative momenta were observed for the first time in $`\pi ^{}+Xe`$ interactions at 9 GeV/$`c`$ . Currently, two-proton correlations are used extensively in low energy (of order 10-1000 MeV/$`c`$) heavy-ion physics where the sizes measured for small proton momenta are comparable to or exceed the size of the larger (usually target) nucleus. A similar result was obtained at much higher CERN SPS energies in the target fragmentation region, where a two-proton correlation analysis yielded source sizes compatible with the sizes of the target nuclei . A decrease of the measured size with increasing proton momenta is reported in many papers however, and for very different projectile energies . Several authors have also reported a dependence on collision impact parameter in low energy heavy-ion interactions . The observed trends in the measured sizes may be related to differences in the proton emission time , and may also reflect the reaction dynamics, which can produce correlations between the momentum and position of the emitted particles .
In spite of extensive analysis of pion and kaon correlations in relativistic heavy-ion collisions, data for two-proton correlations are scarce. This paper reports the first measurement of proton-proton correlations at midrapidity at CERN SPS energies. The origin of protons emitted in relativistic heavy-ion collisions differs from that of lighter particles. Protons are constituent parts of the projectile and target nuclei, unlike pions and kaons which are produced in the collision. The different dynamics of proton emission may therefore be reflected in different source sizes. At the SPS, the expected net baryon-free region at midrapidity for the highest collision energies is not yet achieved. There is still significant stopping, with the associated rapidity shifts of projectile and target nucleons leading to an increase in the proton density at midrapidity . Consequently, the ratio of antiprotons to protons at midrapidity is approximately 0.1 . The relatively high degree of stopping implies that secondary interactions of pions, kaons and other produced particles with nucleons are important, and may influence the size of the proton emission region. The relation between the proton source size, the dimensions of the colliding nuclei, and the corresponding source sizes for pions and kaons at midrapidity for CERN SPS energies is still an open question. In this context, the results of the two-proton correlation analysis described here provide complementary information to existing data from two-boson correlations which can be useful in understanding the underlying dynamics of the collisions .
The paper presents the results of a proton-proton correlation analysis for collisions of two different projectiles, protons and sulphur nuclei, with a lead target. In the first case, the projectile consists of a single proton, and the measured pairs contain mainly protons from the target or protons produced in the interaction. In the second case, the correlated pairs contain both projectile and target participants as well as produced protons. The relative role of these different sources will depend on the impact parameter of the collisions, i.e. the centrality selection. This dependence is examined by an analysis of the correlation effects as a function of charged particle multiplicity in the collision. Finally, the experimental correlation functions are compared to predictions obtained from both a Gaussian source model and from the RQMD event generator, version 1.08 .
## 2 Experiment
The NA44 detector is described in . The spectrometer is optimized to measure single- and two-particle distributions of charged hadrons produced near midrapidity in $`pA`$ and $`AA`$ collisions. The momentum acceptance is $`\pm `$ 20% of the nominal momentum setting, and only particles of a fixed charge sign are detected for a given spectrometer setting.
The data used for this analysis were taken at the 6 GeV/$`c`$ setting with the spectrometer axis at 44 mrad with respect to the incident beam. The corresponding acceptance for protons is 2.35 $`<y<`$ 2.70 and 0.0 $`<p_T<`$ 0.66 GeV/$`c`$ ($`p_T`$ 230 MeV/$`c`$). Scintillator hodoscopes are used for tracking, and also provide time-of-flight information with $`\sigma _{\mathrm{TOF}}<`$ 100 ps. The momentum resolution $`\delta p/p`$ is approximately 0.2%. Two threshold Cerenkov detectors in conjunction with the time-of-flight (mass-squared) information provide particle identification. Particle contamination is less than 1% for the data shown here. An interaction trigger is provided by two rectangular scintillator paddles sitting downstream of the target and covering approximately 1.3 $`<\eta <`$ 3.5. Offline, a silicon pad multiplicity detector with 2$`\pi `$ azimuthal coverage over the range 1.5 $`<\eta <`$ 3.3 is used for event characterization.
Target thicknesses of 10mm and 5mm were used for the $`p+Pb`$ and $`S+Pb`$ data, corresponding to approximately 5.9% and 5.0% interaction lengths respectively. Multiple scattering in the targets, combined with the spectrometer momentum resolution, lead to a resolution in $`k^{}`$, the particle momentum in the rest frame of the pair, of $`\sigma (\delta k^{})`$ 9 MeV/$`c`$. The experimental trigger required an interaction in the target, at least two tracks in the spectrometer, and a veto on pions (and lighter particles) in the spectrometer acceptance. No centrality selection was made in the trigger, however the two-track requirement itself biases the data towards high-multiplicity events. Offline track reconstruction was followed by track quality cuts, including rejection of close-by tracks, and particle identification cuts to select proton pairs. Additional cuts to reject multiple interactions in the target (pileup) and require clean events were made. The final event samples are approximately 8k pairs for the $`p+Pb`$ data and 15k pairs for the $`S+Pb`$ interactions.
## 3 Correlation Analysis
The experimental correlation function is constructed from identified two-proton events. Due to limited statistics, only a one-dimensional analysis has been performed. The correlation function is calculated as:
$$C(k^{})=\frac{N_{corr}(k^{})}{N_{uncorr}(k^{})}$$
(1)
where $`k^{}`$ is the particle momentum in the rest frame of the pair: $`k^{}=\frac{\sqrt{Q^2}}{2}`$, $`Q\{q_0,\stackrel{}{q}\}=p_1p_2`$ and $`p_1,p_2`$ are the particle 4-momenta. The correlated, $`N_{corr}`$, and uncorrelated, $`N_{uncorr}`$, proton pairs are taken from the same and different events respectively. The stability of the resulting correlation function under the applied cuts is studied by varying each cut (such as the mass-squared selection, and the hodoscope multiplicity) individually. In all cases, the correlation function data points are within one standard deviation of their nominal values. The stability with respect to different sampling procedures for producing the mixed-event distribution is also tested, and introduces no systematic deviations.
The experimental correlation functions presented here do not contain any correction factors. Such corrections are often made in boson correlation measurements to take account of the effects of finite detector acceptance and momentum resolution, residual correlations in the background distribution , and the Coulomb interaction, thereby recovering the “ideal” correlation function in which correlations are due purely to quantum statistics. Rather than correct the experimental distributions, theoretical correlation functions based on model predictions have been generated, and modified to include each of these effects. The results of these simulations are then compared directly to the measured correlation functions.
The theoretical correlation function is generated by selecting proton pairs from a given source model, calculating the associated weight due to quantum statistics and final state interactions (strong and Coulomb) , and propagating the particles through a realistic simulation of the experimental detector, including all instrumental and acceptance effects. A full calculation of Coulomb effects is used rather than the Gamow factor approximation. Data cuts are then made in exactly the same way as for the experimental data. For this analysis, two models have been used as inputs for the simulation: a Gaussian source model and the RQMD event generator.
In the case of the Gaussian model, a Gaussian density distribution $`\rho (\stackrel{}{r},t)=\frac{1}{(2\pi )^2r_o^3\tau }e^{\frac{\stackrel{}{r}^2}{2r_o^2}\frac{t^2}{2\tau ^2}}`$ is assumed for both the spatial and temporal extent of the source . It is assumed that protons are emitted by sources moving in the longitudinal direction with the velocities of the proton pairs (the Longitudinally Co-moving Source, LCMS). In this way the fast longitudinal motion of proton sources at mid-rapidity is taken into account. Because the rapidity acceptance in this measurement is narrow (2.35 $`<y<`$ 2.70), the results assuming the LCMS should not differ significantly from the results assuming a fixed frame of reference centered at about $`y`$ = 2.5. Since a one-dimensional analysis is performed here, a spherically-symmetric source parametrized by equal $`r_o`$ radius and $`\tau `$ time parameters is assumed. The RMS radius in this parametrization is given by $`\sqrt{3}r_o`$.
In order to compare with RQMD predictions, the theoretical correlation function for pairs of protons from generated events is calculated , and the appropriate weighting for different impact parameter collisions is taken into account by matching the multiplicity distributions of the data and model. This matching shows that, in contrast to the $`p+Pb`$ data where only the highest multiplicity events are selected by the trigger, the $`S+Pb`$ data are only weakly biased toward central collisions.
In calculating correlation functions for comparison to data, three factors are particularly important: the admixture of indirect protons coming from hyperon (mainly $`\mathrm{\Lambda }\pi ^{}p`$) decays, the acceptance and resolution of the experiment, and the residual correlations arising in the single particle background distributions.
Weak decays of strange baryons are a significant source of protons and contribute to the yields measured in the NA44 spectrometer. The influence of the admixture of indirect protons on the shape of correlation function has been studied using data from the RQMD and Venus (v5.21) event generators, combined with a detailed simulation (GEANT) of the detector. The two models give similar results: about 22% of protons measured in the spectrometer come from the decay of $`\mathrm{\Lambda }`$’s in both $`p+Pb`$ and $`S+Pb`$ collisions, and cannot be distinguished from direct protons. This contribution has only a weak $`p_T`$ dependence, and is greater at low $`p_T`$. In order to take into account this non-correlated contribution to coincident pairs, a fraction of 22% of “decay” protons is included when calculating the correlation function, giving about 39% of uncorrelated pairs in the two-particle sample. These pairs are assigned a weight of unity in the correlation function calculation. This contribution significantly reduces the magnitude of the observed correlation effect, but does not change the general shape of the correlation function.
The acceptance and resolution of the NA44 spectrometer for protons is estimated using a detailed simulation of the detector, including systematic effects introduced by the track reconstruction procedure for close-by pairs of particles. The $`k^{}`$ acceptance for proton pairs is convoluted with the detector resolution and the rather complicated shape of the two-proton correlation function, and gives rise to an asymmetric distortion of the correlation peak. This “smearing” is particularly important for small $`k^{}`$, where the expected depletion due to Coulomb and statistical effects is masked by the experimental resolution.
Residual correlations arise in the mixing procedure because the single particle distributions forming the reference (background) sample are in fact projections of the measured two-particle distribution, which includes the correlations. This effect is taken into account by calculating and applying weights which correspond to the mean value of the correlation function for all combinations of a given particle with all others . This residual correlation modifies slightly the height of the correlation peak, and induces a slope for large values of $`k^{}`$.
All three effects are included in the simulation procedure. In Fig. 1 they are demonstrated separately in order to illustrate their particular features and the relative significance of each of them.
## 4 Results
The experimental proton-proton correlation functions for $`p+Pb`$ and $`S+Pb`$ collisions are presented in Fig. 2. The correlation functions have been normalized such that $`C(k^{})=1`$ for $`80<k^{}<160`$ MeV/$`c`$. The peak observed in the region of small $`k^{}`$ values ($``$ 20 MeV/$`c`$) can be attributed to the combined effect of the final state interaction and Fermi-Dirac statistics. It reflects the space-time properties of the proton emission process. The two correlation functions have qualitatively the same shape, but the peak is much more pronounced for the $`p+Pb`$ data than for the $`S+Pb`$, indicating that protons are emitted from a smaller source in the case of $`p+Pb`$ collisions. Also shown are the correlation functions and associated source sizes from the Gaussian model obtained with a minimum-$`\chi ^2`$ fit over the range $`0<k^{}<160`$ MeV/$`c`$. The fits yield $`r_o=\tau `$ = 1.42$`{}_{0.05}{}^{}{}_{}{}^{+0.04}`$ fm ($`\chi ^2/N_{DF}`$ = 10.3/10) for $`p+Pb`$ and $`r_o=\tau `$ = 2.65$`{}_{0.09}{}^{}{}_{}{}^{+0.09}`$ fm ($`\chi ^2/N_{DF}`$ = 9.6/10) for the $`S+Pb`$ data. (Note that the $`r_o`$ values have to be multiplied by $`\sqrt{3}`$ to obtain the corresponding RMS radii.) Fits assuming $`\tau `$ = 0 give $`r_o`$ parameters which are equal within errors to those obtained with $`\tau =r_o`$. This is a consequence of the weak time dependence of the final state interaction effects for small particle velocities ; for these data the mean velocity of the proton pairs in the LCMS frame is $``$ 0.24$`c`$.
These results have been obtained assuming that the fraction of measured protons from $`\mathrm{\Lambda }`$ decay is that given by the simulations described above (22%). While the RQMD and Venus models give consistent results for the fraction of protons from hyperon decays in the NA44 acceptance, there is a factor of order two discrepancy between the available experimental data on $`\mathrm{\Lambda }`$ production yields in $`S+A`$ collisions at CERN SPS energies . The RQMD prediction is roughly midway between the two measurements. Using the experimental data as an estimate of the uncertainty on the $`\mathrm{\Lambda }`$ yield (approx. $`\pm `$ 40%), and hence on the fraction of protons from $`\mathrm{\Lambda }`$ decays, gives $`r_o=\tau `$ = 1.42$`{}_{0.17}{}^{}{}_{}{}^{+0.16}`$ fm for $`p+Pb`$ and $`r_o=\tau `$ = 2.65$`{}_{0.17}{}^{}{}_{}{}^{+0.19}`$ fm for $`S+Pb`$. The RMS radii of the proton and the sulphur and lead nuclei are 0.88 fm, 3.26 fm and 5.50 fm respectively . In the frame of this geometrical interpretation, the measured source sizes are greater than the size of the projectile but smaller than that of the target.
In order to analyze the centrality dependence of the correlation effect, both the $`p+Pb`$ and $`S+Pb`$ data have been divided into three multiplicity bins, based on the charged-particle multiplicity measured by the silicon detector. The binning is chosen to give roughly equal proton-pair statistics in each sub-sample. The correlation functions corresponding to the three multiplicity selections are shown in Fig. 3 for $`p+Pb`$ and $`S+Pb`$. A systematic decrease of the correlation effect with increasing multiplicity is observed, corresponding to an increase in the size of the emitting volume. The $`r_o`$ values indicate the scale of these changes.
The comparison of the data with the predictions of RQMD is shown in Fig. 4. In the case of $`p+Pb`$ collisions, the experimental data are well reproduced by the model. For $`S+Pb`$, the model gives a somewhat larger correlation effect than is observed in the data. Fitting the $`S+Pb`$ RQMD prediction with the same Gaussian model (and same fraction of protons from $`\mathrm{\Lambda }`$ decay) as used for the data yields $`r_o=\tau `$ = 2.27$`{}_{0.02}{}^{}{}_{}{}^{+0.02}`$ fm, compared to $`r_o=\tau `$ = 2.65$`{}_{0.17}{}^{}{}_{}{}^{+0.19}`$ for the data.
## 5 Discussion
A simple, one-dimensional description with Gaussian co-moving sources is used to quantify the correlation effects. In introducing the co-moving source, it is assumed that particles with similar longitudinal velocities are emitted from nearby space-time points, such as occurs when fast longitudinal expansion takes place. In this case, the correlation function effectively measures only a part of the space-time extent of the emission volume. For the $`S+Pb`$ system, the multiplicity cuts select different impact parameter collisions, and hence different sizes of the overlapping region between projectile and target nucleons. This is reflected in the observed centrality dependence of the correlation functions. Comparison of the measured source sizes with the RMS radii of the projectile and the target is also in qualitative agreement with such an interpretation. Note that in the target fragmentation region the sizes measured by two-proton correlations reflect rather the size of the target nucleus . In the case of $`p+Pb`$ collisions, the number of projectile participants is always equal to one and the measured size can be related to the extent of the hot region along the path of the proton in the lead nucleus. The source sizes in this case are much smaller than in the case of the sulphur projectile.
These data can be compared to correlation results obtained for other particle types near midrapidity and in similar $`p_T`$ intervals, which assume the same type of emitting source. For both systems studied ($`p+Pb,S+Pb`$), the proton source sizes appear to be smaller than those of pions, especially in the case of the sulphur projectile: for $`p+Pb`$, $`r_{Ts}(\pi ^+)`$ = 2.00 $`\pm `$ 0.25 fm while for $`S+Pb`$, $`r_{Ts}(\pi ^+)`$ = 4.15 $`\pm `$ 0.27 fm ($`p_T`$ 150 MeV/$`c`$ in both cases) . The $`r_o`$ values for protons are similar to the corresponding kaon source sizes: for $`p+Pb`$, $`r_{Ts}(K^+)`$ = 1.22 $`\pm `$ 0.76 fm while for $`S+Pb`$, $`r_{Ts}(K^+)`$ = 2.55 $`\pm `$ 0.20 fm ($`p_T`$ 240 MeV/$`c`$ in both cases) . (Note that for the three-dimensional boson source parameters, multiplication by a factor of $`\sqrt{3}`$ is also necessary to obtain RMS sizes). The Gaussian source model used in these analyses describes sources for which the momenta and positions of the emitted particles are independent in the LCMS. Collision dynamics may, however, induce correlations between the measured source size and the transverse momentum of the particles. In particular, a decrease of the measured size with increasing transverse mass $`m_T`$ (where $`m_T=\sqrt{p_T^2+m^2}`$ ), observed in common for pions and kaons, may indicate the collective expansion of an equilibrated system formed during the collision . The smaller source sizes for the two-proton data ($`m_T`$ 970 MeV) compared to the two-boson data are in qualitative agreement with such an $`m_T`$ dependence.
The contribution of indirect protons coming from hyperon decays is also a parameter in the description of the experimental results, and can influence the values of the extracted source sizes. The experimental uncertainty on the $`\mathrm{\Lambda }`$ yield is described above, and results in a systematic error of approximately 0.2 fm on the extracted $`r_o`$ parameters. Of the other hyperons, Monte Carlo studies show that only the $`\mathrm{\Sigma }^+`$ also contributes to the proton yield measured in the NA44 spectrometer. RQMD and Venus predict that the number of detected protons from $`\mathrm{\Sigma }^+`$ decay is at most 30% and 10%, respectively, of that from $`\mathrm{\Lambda }`$ decay - less than the experimental uncertainty on the $`\mathrm{\Lambda }`$ yield itself. Possible uncertainty in the contribution of indirect protons cannot significantly alter the conclusion on the size of the proton source compared to the pion source however, nor on the centrality dependence of the observed sizes. In order to produce the change in the measured correlation functions between the lowest and highest multiplicity bins assuming that the difference is due entirely to indirect protons implies that more than 50% of detected protons come from hyperon decays. However, an enhanced emission of hyperons with respect to that predicted by the RQMD model in the case of $`S+Pb`$ collisions may be responsible for the difference between the model prediction and the results of measurements.
## 6 Conclusion
These results demonstrate the first attempt to include baryon-baryon correlations in the analysis of the space-time dynamics of particle emission at midrapidity at CERN SPS energies. The measured source sizes increase with the mass of the projectile and with the collision multiplicity. Within the context of a Gaussian source model, the proton radius parameters are smaller than the size of the target nucleus but larger than the sizes of the projectiles. Comparing to other particle types, the proton size parameters are smaller than those for pions, and similar to those for kaons, measured in the same collisions. (Note that collision dynamics may induce a dependence of the measured source sizes on the momentum of the emitted particles.) Good agreement with the results of RQMD (v1.08) simulations is seen for $`p+Pb`$ collisions. For $`S+Pb`$ collisions, the measured correlation effect is somewhat weaker than that predicted by the model simulations.
## 7 Acknowledgements
The NA44 Collaboration wishes to thank the staff of the CERN PS-SPS accelerator complex for their excellent work. We thank the technical staff at CERN and the collaborating institutes for their valuable contributions. We are also grateful for the support given by the Science Research Council of Denmark; the Japanese Society for the Promotion of Science, and the Ministry of Education, Science and Culture, Japan; the Science Research Council of Sweden; the US Department of Energy and the National Science Foundation. We thank Dr. H. Sorge for the use of the RQMD code; Dr. K. Werner for the Venus code; and Drs. R. Lednicky and S. Pratt for making their correlation codes available to us, and for fruitful discussions.
|
no-problem/9906/astro-ph9906198.html
|
ar5iv
|
text
|
# Frequency resolved spectroscopy of Cyg X-1: fast variability of the Fe 𝐾_𝛼 line.
## 1 Introduction
Over the last several years it has become widely accepted that the Galactic X-ray binaries exhibit Compton reflection features in their spectra.
Qualitatively, the reflection component in the X-ray spectra of these systems can be described as a broad “hump” at energies 20–30 keV, fluorescent Fe line at energies 6.4–6.7 keV (depending on the ionization state of the reflecting medium) and an absorption edge at the energy $``$7.1 keV (Basko, Sunyaev & Titarchuk, (1974), George & Fabian, (1991)). For all geometries which do not obscure the direct primary continuum from the observer, the detected energy spectra consists of this direct component (approximately a power law in the 3–13 keV energy band) and the reflection features. For standard cosmic abundances the expected equivalent width of the fluorescent Fe line is $``$100–200 eV (for a source above semi–infinite slab of neutral matter).
The geometry and mutual location of the source of primary continuum and reflecting medium should affect both the equivalent width of the fluorescent line and the character of it’s variability. In particular the finite size of the reflector implies that the time variations of the reflected radiation should be smeared out on the time scales corresponding to the light crossing time of the reflector. In addition, a time lags between different emission components might apprear. Alternatively the geometrically different region of the main energy release zone (where the primary continuum is produced) may have different efficiency for production of the reflection component. As a result timing properties of the reprocessed component may be linked to the properties of the selected region of this zone.
## 2 Observations and data analysis
For our analysis we chose publicly available data of the Rossi X-ray Timing Explorer observations P10238 performed between Mar. 26, 1996 and Mar. 31, 1996 with a total exposure time $``$70 ksec (we used only observations were all 5 PCU were turned on). For the frequency resolved spectral analysis we used PCA data in the “Generic Binned” mode, having $`\frac{1}{64}`$sec$``$16 msec time resolution in 64 energy channels covering the whole PCA energy band (B\_16ms\_64M\_0\_249). For analysis of the averaged spectrum we used the “Standard Mode 2” data, having twice as many energy channels in the energy band of our interest (3–13 keV) than the “Generic Binned” mode.
For the obtained spectra we constructed the response matrix using the standard tasks of FTOOLS 4.2 package with PCA RMF v3.5 (Jahoda 1999). The comparison of the timing mode spectra with that obtained from “Standard Mode 2” showed, that they do not differ more than by 1–1.5%. The background was calculated using “Q6” model ($`VLE`$-based model, preferable for bright sources does not work for these observations). However, the PCA background is negligible for the time averaged spectra of Cyg X-1 in the energy band 3–13 keV. The PCA background contribution to the frequency resolved spectra at the frequencies 0.03–30 Hz is even less important.
## 3 Frequency resolved spectral analysis
We calculated the Power Density Spectra (PDSs) in each energy channel for every 128 sec time segments using the standard Fast Fourier Transform (FFT) procedure and adopting the normalization of Miyamoto et al. (1991):
$$P_j=2|a_j|^2/N_\gamma R$$
$$a_j=\underset{k=1}{\overset{2^m}{}}x_ke^{i\omega _jt_k}$$
where $`t_k`$ is the time label of the $`k`$th bin, $`x_k`$ is the number of counts in this bin, $`N_\gamma `$ is the total number of photons and $`R`$ is the average count rate in the segment ($`R=N_\gamma /T`$, $`T`$ – total duration of the light curve segment). This PDS normalization ((rms/mean)<sup>2</sup>/Hz) allows one to obtain squared fractional $`rms`$ by integrating the power over the frequencies of interest. Then the obtained power spectra were averaged over the light curve segments and the logarithmic frequency bins. For every obtained frequency bin the frequency dependent spectra were constructed according to the formula:
$$S(E_i,f_j)=R_i\sqrt{P_i(f_j)\mathrm{\Delta }f_j}=\sqrt{\frac{2|a_{ij}|^2}{T}\mathrm{\Delta }f_j}$$
here $`S(E_i,f_j)`$ is the count rate of the spectrum on the frequency $`f_j`$ in the energy channel $`E_i`$. Since the light curves of Cyg X-1 in the different energy bands almost perfectly coherent and have practically the same form of the PDS (see e.g. Nowak et al. 1999), the spectra obtained using the procedure above can be used to describe the real energy spectrum of the source X-ray flux variations on the different time scales. The above expession also can be easily adapted to describe the energy spectrum of different PDS spectrum components.
## 4 Results
The energy spectra of Cyg X–1 in several narrow frequency ranges between 0.03 and 30 Hz are shown in Fig. 1. The change of the spectral shape with the decrease of the characteristic time scale can be clearly seen. The relative amplitude of the reflection features, in particular that of the broad line at $``$ 6–7 keV and the “smeared edge” above $`7`$ keV, is apparently decreasing with the increase of frequency and vanishes above $`20`$ Hz. Along with becoming less “wiggled” the spectra become harder as frequency increases.
In order to quantify these effects we fit frequency resolved spectra in 16 frequency bins covering 0.03–32 Hz range in the 3–13 keV energy band with a simple two component model consisting of a power law and a Gaussian line with the fixed energy $`E_{\mathrm{line}}=6.4`$ keV and the width FWHM=0.1 keV (that is comparable with the PCA energy resolution at this energy). The best fit parameters are given in Table 1. The dependence of the line equivalent width upon frequency is shown in Fig. 2. As is seen from the Fig. 2, the equivalent width of the line starts decreasing above $`1`$ Hz and falls by a factor of 2 at $`10`$ Hz. No significant flux in the line was detected above 15 Hz with the 2$`\sigma `$ upper limit on the equivalent width in the 15–32 Hz frequency range of $``$40 eV.
Variations of the parameters of the spectral model, in particular the change of the line centroid to 6.0 and 6.7 keV and increase of the line width to 1.0 keV, do not change the general trend. These variations, however, affect the particular values of the equivalent width and, to some extent, the shape of the curve in Fig. 2. The change of the line centroid energy to the 5 keV or 9 keV, on the contrary, results in disappearance of the effect.
Finally we should mention that similar behaviour of the frequency resolved spectra was detected in Cyg X-1 observations 20175-01-01-00, 20175-01-02-00 and in all available observations of the set P30157 (01–10), performed in 1997 but with lower significance.
## 5 Discussion
A time averaged energy spectrum of Cyg X-1 (e.g. Ebisawa et al. (1992), Gierlinski et al. (1997)) as well as frequency resolved spectra at low frequency (Fig.1) show clear deviation from a single slope power law spectrum. This deviation is commonly ascribed to the reflection from an optically thick media located in the vicinity of the production site of primary hard X–ray radiation. In a simple case of reflection from an optically thick cold neutral medium the main reflection features are well known (Basko, Sunyaev & Titarchuk, (1974), George & Fabian, (1991)) – a narrow unshifted iron $`K_\alpha `$ line at 6.4 keV with an equivalent width of $`100`$ eV, an absorption edge at $`7.1`$ keV and a reflected continuum peaked at $`2030`$ keV. However, the real spectra of compact X–ray binaries show considerably more complicated behavior. In particular the centroid energy of the line is often different from 6.4 keV, the line width in many cases is as large as $`1`$ keV and a broad “smeared edge” above $``$ 7–8 keV is observed instead of a sharp absorption edge at $`7.1`$ keV (Ebisawa et al. (1992)). This departure from a simple reflection model hints at a complicated ionization state and/or motion (e.g. Keplerian motion in the disk) of the reflecting media.
It is obvious, that the simple spectral model used for fitting the data in the previous section is neither adequate nor completely justified from the physical point of view. The apparent line centroid energy and width vary with frequency. In particular the best fit centroid energy shifts below 6.0 keV and the width of the line exceeds 1 keV as frequency increases. However, due to the lack of a satisfactory realistic model of reflection in an X–ray binary, especially applicable to the frequency resolved spectral data, we restricted ourselves to the simple two component model described in the previous section. The main purpose of using this spectral model was to quantify the major effect which is clearly seen in Fig.1, namely the suppression of the reflection features in the energy spectrum with the increase of frequency.
The results of the analysis presented in this paper show that the relative amplitude of the reflected component variations is lower on the short time scales ($``$ 50–100 msec), than that of the primary X-ray radiation. Similar effect was found Oosterbroek et al. (1996) for GS 2023+338 but the characteristic time scale in this case was much longer, $``$200 sec. GS 2023+338 is peculiar in many ways. In particular the source exhibited strong ($`10^{23}`$ cm<sup>-2</sup>) and highly variable on the time scales of $`10^3`$ sec low energy absorption (e.g. Zycki, Done & Smith (1999)) and, most importantly, extremely strong Fe line with equivalent width up to 1.4 keV (Oosterbroek et al. (1996)), that could mean that geometry of the reprocessing media in GS2023+338 is different from that in Cyg X–1. Recently we found suppression of the reflected component variations in the low spectral state of another black hole candidate GX339–4 (Revnivtsev, Gilfanov & Churazov, 1999). In this case the characteristic time scale ($``$50–100 msec) was similar to that in Cyg X–1. We therefore can tentatively conclude that such a behavior might be a common feature of the black hole binaries in the low spectral state.
A most straightforward explanation of this effect would be in terms of a finite light crossing time of the reflector $`\tau _{\mathrm{refl}}l_{\mathrm{refl}}/c`$, caused by a finite spatial extend $`l_{\mathrm{refl}}`$ of the reflecting media. From Fig.2 one can see that the equivalent width of the iron line drops by a factor of two at the frequency $`f_{1/2}10`$ Hz. This value gives us a rough estimate of the characteristic response time of the reflector $`\tau _{\mathrm{refl}}1/2\pi f_{1/2}15`$ msec and the characteristic size of the reflecting media $`l_{\mathrm{refl}}5\times 10^8`$ cm which would correspond to $`150R_\mathrm{g}`$ for a $`10M_{\mathrm{}}`$ black hole. We note however that if the primary continuum originates within the $`10R_\mathrm{g}`$ sphere in the inner zone of the accretion disk, then only a small fraction of the emitted hard radiation has the possibility to be reflected from the accretion disk regions with $`R150R_\mathrm{g}`$ (in the case of a flat disk) and it can hardly provide observed $`100`$ eV equivalent width of the iron line. Therefore the assumption that high frequency variations of the reflected component are caused by the finite light crossing time of the reflector implies interesting constraints on the geometry of the source of primary continuum and reflector.
An alternative explanation might be that the short time scale, $``$ 50–100 msec, variations appear in geometrically different, likely inner, part of the accretion flow and give a rise to significantly weaker, if any, reflected emission than the longer time scale events, presumably originating in the outer regions. This might be caused, for instance, by a smaller solid angle of the reflector as seen by the short time scale events and/or due to screening of the reflector from the short time scale events by the outer parts of the accretion flow.
###### Acknowledgements.
This research has made use of data obtained through the High Energy Astrophysics Science Archive Research Center Online Service, provided by the NASA/Goddard Space Flight Center.
|
no-problem/9906/gr-qc9906091.html
|
ar5iv
|
text
|
# 1 Monopole and dyon solutions for Λ=-0.01 and 𝑣=1. (a) Monopoles: (𝑎,𝑏)=(0,0.001) and (0,0.005). (𝑤,𝑚) at 𝑟=∞ are (0.339,0.034) and (-0.878,0.191), respectively. (b) Dyons: (𝑎,𝑏)=(0.003,0.001) and (0.002,0.0005). (𝑤,𝑢,𝑚) at 𝑟=∞ are (0.031,0.080,0.099) and (0.421,0.064,0.056), respectively.
Revised. 21 September 1999 To appear in Phys Rev Lett
Stable monopole and dyon solutions in the Einstein-Yang-Mills theory in asymptotically anti-de Sitter Space
Jefferson Bjoraker and Yutaka Hosotani
School of Physics and Astronomy, University of Minnesota
Minneapolis, MN 55455, U.S.A.
## Abstract
A continuum of new monopole and dyon solutions in the Einstein-Yang-Mills theory in asymptotically anti-de Sitter space are found. They are regular everywhere and specified with their mass, and non-Abelian electric and magnetic charges. A class of monopole solutions which have no node in non-Abelian magnetic fields are shown to be stable against spherically symmetric linear perturbations.
Soliton and black hole solutions to the Einstein-Yang-Mills (EYM) equations have generated considerable interest this past decade -. In flat space there can be no static soliton solution in the pure Yang-Mills theory . Even unstable static solutions cannot exist. The presence of gravity provides attractive force which can bind non-vanishing Yang-Mills fields together. Such static solutions to the EYM equations have been found in asymptotically Minkowski or de Sitter space. They are all purely magnetic. In asymptotically Minkowski space-time the electric components are forbidden in static solutions . All of these EYM solitons and black holes were shown to be unstable, where the number of unstable modes is equal to twice the number of times the magnetic component of the gauge field crosses the axis .
If the spacetime is modified to include the cosmological constant, the no-go theorems forbidding the electric components of the gauge fields fail, thus, permitting dyon solutions. We have found that in asymptotically de Sitter spacetime, a non-zero electric component to the gauge fields causes $`\sqrt{g}`$ to diverge at the cosmological horizon, thus excluding dyon solutions.
In this letter, we present new monopole and dyon solutions in asymptotically anti-de Sitter (AdS) space. They are regular everywhere. In asymptotically AdS space there are solutions with no cosmological horizon and dyons solutions are allowed. Furthermore, we have found a continuum of solutions where the gauge fields have no nodes, corresponding to stable monopole and dyon solutions.
AdS space has attracted huge interest recently. The BTZ black holes in three-dimensional AdS space provide valuable information about black hole thermodynamics and quantum gravity. In four dimensions linearly stable black hole solutions have been found in asymptotically AdS space. The correspondence between four-dimensional super-conformal field theory and type IIB string theory on AdS<sub>5</sub> has been established. In this Letter we shall show that even in a simple Einstein-Yang-Mills system stable monopole and dyon solutions exist in four-dimensional asymptotically AdS space, which we believe leads to further understanding of rich structure of quantum field theory in AdS space as well as profound implications to the evolution of the early universe.
We start with the most general expression for the spherically symmetric SU(2) gauge field in the singular gauge :
$$A=\frac{1}{2e}\left\{u\tau _3dt+\nu \tau _3dr+(w\tau _1+\stackrel{~}{w}\tau _2)d\theta +(\mathrm{cot}\theta \tau _3+w\tau _2\stackrel{~}{w}\tau _1)\mathrm{sin}\theta d\varphi \right\},$$
(1)
where $`u`$, $`\nu `$, $`w`$ and $`\stackrel{~}{w}`$ depend on $`r`$ and $`t`$. The regularity at the origin imposes the boundary conditions $`u=\nu =0`$ and $`w^2+\stackrel{~}{w}^2=1`$ at $`r=0`$. Under a residual $`U(1)`$ gauge transformation $`S=e^{i\tau _3z(t,r)/2}`$, $`(u,\nu )(u+\dot{z},\nu +z^{})`$ and $`w+i\stackrel{~}{w}e^{iz}(w+i\stackrel{~}{w})`$. The spherically symmetric metric is written as
$$ds^2=\frac{Hdt^2}{p^2}+\frac{dr^2}{H}+R^2(d\theta ^2+\mathrm{sin}^2\theta d\varphi ^2),$$
(2)
where $`H`$, $`p`$, and $`R`$ are functions of $`t`$ and $`r`$, in general.
We first look for static regular soliton solutions in the $`\nu =0`$ gauge. Since the Yang-Mills equations imply $`\stackrel{~}{w}=Cw`$ where $`C`$ is some constant, another gauge transformation allows us to set $`C=0`$, or $`\stackrel{~}{w}=0`$. Without loss of generality one can choose $`w(0)=1`$. $`u`$ can be interpreted as the electric part of the gauge fields and $`w`$ as the magnetic part. The Schwarzshild gauge $`R=r`$ is taken for the metric. The coupled static EYM equations of motion are
$`\left({\displaystyle \frac{H}{p}}w^{}\right)^{}`$ $`=`$ $`{\displaystyle \frac{p}{H}}u^2w{\displaystyle \frac{w}{p}}{\displaystyle \frac{(1w^2)}{r^2}}`$ (3)
$`\left(r^2pu^{}\right)^{}`$ $`=`$ $`{\displaystyle \frac{2p}{H}}w^2u`$ (4)
$`p^{}`$ $`=`$ $`{\displaystyle \frac{2v}{r}}p\left[(w^{})^2+{\displaystyle \frac{u^2w^2p^2}{H^2}}\right]`$ (5)
$`m^{}`$ $`=`$ $`v\left[{\displaystyle \frac{(w^21)^2}{2r^2}}+{\displaystyle \frac{1}{2}}r^2p^2(u^{})^2+H(w^{})^2+{\displaystyle \frac{u^2w^2p^2}{H}}\right].`$ (6)
Here $`H(r)=12m(r)/r\mathrm{\Lambda }r^2/3`$ and $`v=G/4\pi e^2`$. $`\mathrm{\Lambda }`$ is the cosmological constant. We are looking for a solution which is regular everywhere and has finite ADM mass $`m(\mathrm{})`$.
The presence of the cosmological constant affects the behavior of the soliton solutions significantly. For $`\mathrm{\Lambda }=0`$, both $`H`$ and $`p`$ approach a constant value as $`r\mathrm{}`$. Consequently $`w^21`$ from (3) and (6). For $`\mathrm{\Lambda }0`$, $`w`$ need not have an asymptotic value $`\pm 1`$. It is known that Eq. (4) implies the absence of solutions with $`u(r)0`$ in the $`\mathrm{\Lambda }=0`$ case . A similar argument applies to the $`\mathrm{\Lambda }>0`$ case in which there appears a cosmological horizon at $`r=r_h`$; $`H(r_h)=0`$. At the horizon either $`u`$ or $`w`$ must vanish to have regular solutions to Eqs. (3)-(6). Eq. (4) implies that $`u(r)`$ is either identically zero or a monotonic function of $`r`$ for $`r<r_h`$. Hence, $`w(r_h)=0`$ and $`u(r_h)0`$ if $`u^{}(0)0`$. However, Eq. (3) implies
$$\frac{Hw^{}}{pw}|_{r_1}^{r_h}=_{r_1}^{r_h}𝑑r\left\{\frac{H}{p}\left(\frac{w^{}}{w}\right)^2+\frac{pu^2}{H}+\frac{(1w^2)}{pr^2}\right\}$$
(7)
provided $`w(r)>0`$ for $`r_1r<r_h`$. The r.h.s. of (7) diverges whereas the l.h.s. remains finite if $`w(r_h)=0`$ and $`u(r_h)0`$. This establishes that $`u(r)=0`$ in the de Sitter space. All solutions are purely magnetic. Suppose that $`0<w^21`$ for $`0rr_h`$ and take $`r_1=0`$ in (7). As $`w^{}(0)=0`$, the l.h.s. vanishes whereas the r.h.s. is negative definite. This implies that $`w`$ satisfying $`w^21`$ for $`rr_h`$ must vanish somewhere between 0 and $`r_h`$. It has been known that in all solutions in asymptotically de Sitter space $`w(r)`$ has at least one node.
The situation is quite different in asymptotically AdS space ($`\mathrm{\Lambda }<0`$). $`H(r)`$ is positive everywhere. All equations are consistently solved if $`u(r)u_0+(u_1/r)+\mathrm{}`$ and $`w(r)w_0+(w_1/r)+\mathrm{}`$ for large $`r`$. There is no restriction on the value of $`u_0`$ or $`w_0`$. Furthermore $`w(r)`$ can be nodeless.
Solutions are classified by the ADM mass, $`M=m(\mathrm{})`$, electric and magnetic charges, $`Q_E`$ and $`Q_M`$. From the Gauss flux theorem
$$\left(\begin{array}{c}Q_E\\ Q_M\end{array}\right)=\frac{e}{4\pi }𝑑S_k\sqrt{g}\left(\begin{array}{c}F^{k0}\\ \stackrel{~}{F}^{k0}\end{array}\right)$$
(8)
are conserved, but are also gauge-dependent. With the ansatz in the singular gauge
$$\left(\begin{array}{c}Q_E\\ Q_M\end{array}\right)=\left(\begin{array}{c}u_1p_0\\ 1w_0^2\end{array}\right)\frac{\tau _3}{2}$$
(9)
where $`p(r)=p_0+(p_1/r)+\mathrm{}`$ etc. If $`(u,w,m,p)`$ is a solution, then $`(u,w,m,p)`$ is also a solution. Dyon solutions come in a pair with $`(\pm Q_E,Q_M,M)`$.
The solutions to Eqs. (3) to (6), for $`\mathrm{\Lambda }<0`$, are evaluated numerically. The procedure is to solve these equations at $`r=0`$ in terms of two free adjustable parameters $`a`$ and $`b`$ and ‘shoot’ for solutions with the desired asymptotical behavior. The behavior of solutions near the origin are
$`w(r)`$ $`=`$ $`1br^2`$ (10)
$`m(r)`$ $`=`$ $`v(2b^2+{\displaystyle \frac{1}{2}}a^2)r^3`$ (11)
$`p(r)`$ $`=`$ $`1(4b^2+a^2)vr^2`$ (12)
$`u(r)`$ $`=`$ $`ar+{\displaystyle \frac{a}{5}}(2b+{\displaystyle \frac{1}{3}}\mathrm{\Lambda }+2v(a^2+4b^2))r^3.`$ (13)
Purely magnetic solutions (monopoles) are found by setting $`a=0`$ ($`u=0`$). Varying the initial condition parameter $`b`$, a continuum of monopole solutions were obtained, which are similar to the black hole solutions found in ref. , but are regular in the entire space. $`w`$ crosses the axis an arbitrary number of times depending on the value of the adjustable shooting parameter $`b`$. Typical solutions are displayed in fig. 1.
The behavior of $`m`$ and $`p`$ is similar to that of the asymptotically de Sitter solutions previously considered . In contrast, as shown in fig. 1, there exist solutions where $`w`$ has no nodes. These solutions are of particular interest because they are shown to be stable against linear perturbations.
If the adjustable shooting parameter $`a`$ is chosen to be non-zero, we find dyon solutions. As shown in fig. 1, the electric component, $`u`$, of the YM fields starts at zero and monotonically increases to some finite value. The behavior of $`w`$, $`m`$ , $`H`$, and $`p`$ is similar to that in the monopole solutions.
Again we find a continuum of solutions where $`w`$ crosses the axis an arbitrary number of times depending on $`a`$ and $`b`$. Similarly there exist solutions where $`w`$ does not cross the axis.
Solutions are found for a continuous set of the parameters $`a`$ and $`b`$. This is in sharp contrast to the $`\mathrm{\Lambda }0`$ case, in which only a discrete set of solutions are found. For some values of $`a`$ and $`b`$, solutions blow up, or the function $`H(r)`$ crosses the axis and becomes negative. One example of solutions near the critical value \[$`(a,b)=(0.01,0.69)`$\] is displayed in fig. 2. $`H(r)`$ becomes very close to zero at $`r1`$. It has $`(Q_E,Q_M,M)(0.015,0.998,0.995)`$.
In fig. 3 $`M`$ is plotted as a function of $`Q_M`$ for monopole solutions. The behavior of the solutions near $`b=0.7`$ needs more careful analysis, although we did find that when $`b>0.7`$, all the solutions appeared to have a horizon.
We have found that dyon solutions cover a good portion of the $`Q_E`$-$`Q_M`$ plane. There are solutions with $`Q_M=0`$ but $`Q_E0`$, which has non-vanishing $`w(r)`$. In the shooting parameter space $`(a,b)`$, these solutions correspond not exactly, but almost to a universal value for $`b0.0054`$. See fig. 4. We have not understood why it should be so.
It has been shown that the BK solutions and the de Sitter-EYM solutions are unstable . In contrast, the AdS black hole solutions were shown to be stable for $`u=0`$. We shall show that the monopole solutions without nodes presented in this paper are stable against spherically symmetric linear fluctuations.
In examining time-dependent fluctuations around monopole solutions it is convenient to work in the $`u(t,r)=0`$ gauge. Solutions have non-vanishing $`w(r)`$, $`p(r)`$, and $`H(r)`$, but $`\stackrel{~}{w}(r)=\nu (r)=0`$. Linearized equations for $`\delta w(t,r)`$, $`\delta \stackrel{~}{w}(t,r)`$, $`\delta \nu (t,r)`$, $`\delta p(t,r)`$, and $`\delta H(t,r)`$ have been derived in the literature . Fluctuations decouple in two groups. $`\delta w(t,r)`$, $`\delta p(t,r)`$, and $`\delta H(t,r)`$ form even-parity perturbations, whereas $`\delta \stackrel{~}{w}(t,r)`$ and $`\delta \nu (t,r)`$ form odd-parity perturbations. The linearized equations imply that $`\delta p(t,r)`$, and $`\delta H(t,r)`$ are determined by $`\delta w(t,r)`$, and $`\delta \stackrel{~}{w}(t,r)`$ by $`\delta \nu (t,r)`$.
$`\beta (t,r)=r^2p\delta \nu /w=e^{i\omega t}\beta (r)`$ satisfies $`\left\{(d/d\rho )^2+U_\beta (\rho )\right\}\beta =\omega ^2\beta `$ where
$$U_\beta =\frac{H}{r^2p^2}(1+w^2)+\frac{2}{w^2}\left(\frac{dw}{d\rho }\right)^2$$
(14)
and $`d\rho /dr=p/H`$. The range of $`\rho `$ is finite: $`0\rho \rho _{\mathrm{max}}`$. Eq. (14) is of the form of the Schrödinger equation on a one-dimensional interval. When $`w(r)`$ in the monopole solution has no node ($`w>0`$ for all $`r`$), then $`U_\beta >0`$ is a smooth potential so that $`\beta `$ is a smooth function of $`\rho `$ in the entire range satisfying $`\beta \beta ^{}=0`$ at $`\rho =0`$ and $`\rho _{\mathrm{max}}`$. The eigenvalue $`\omega ^2`$ is positive-definite, i.e. the solution is stable against odd-parity perturbations. If $`w(r)`$ has $`n`$ nodes, i.e. $`w(r_j)=0`$ ($`j=1,\mathrm{},n`$), the potential $`U_\beta `$ develops $`(\rho \rho _j)^2`$ singularities. The solution $`\beta (\rho )`$ to (14) is no longer regular at $`\rho =\rho _j`$ so that the positivity of the differential operator $`d^2/d\rho ^2`$ is not guaranteed. Indeed, Volkov et al. have proven for the BK solutions that there appear exactly $`n`$ negative eigenmodes ($`\omega ^2<0`$) if $`w`$ has $`n`$ nodes . Their argument applies to our case without modification. One concludes that the solutions with nodes in $`w`$ are unstable against parity-odd perturbations.
Similarly parity-even perturbations $`\delta w(t,r)=e^{i\omega t}\delta w(r)`$ satisfies the Schrödinger equation with a potential
$$U_w=\frac{H(3w^21)}{p^2r^2}+4v\frac{d}{d\rho }\left(\frac{Hw^2}{pr}\right).$$
(15)
$`U_w(\rho )`$ is not positive definite, but is regular in the entire range $`0\rho \rho _{\mathrm{max}}`$. We have solved the Schrödinger equation for $`\delta w`$ numerically for typical monopole solutions. The potential is displayed in fig. 5 for the solutions with $`(a,b)=(0,0.001)`$ and $`(0,0.005)`$. The former has no node in $`w`$, while the latter has one node. The asymptotic value $`w(\mathrm{})`$ is 0.339 or $`0.878`$, respectively. The lowest eigenvalue $`\omega ^2`$ is found to be 0.028 or 0.023. Hence, these solutions are stable against parity-even perturbations. Note that in the $`\mathrm{\Lambda }<0`$ case some of the $`n=1`$ solutions are stable against parity-even perturbations, while in the $`\mathrm{\Lambda }0`$ case they are unstable.
In the present paper a continuum of new monopole and dyon solutions to the EYM equations in asymptotically AdS space have been found. There are solutions with no node in the magnetic component $`w(r)`$ of the $`SU(2)`$ gauge fields. The monopole solutions with no node in $`w`$ have been shown to be stable against spherically symmetric perturbations. The stability of those solutions with non-zero electric fields is currently under investigation. As the monopole and dyon solutions are found in a continuum set, the dyon solutions without nodes in $`w(r)`$ are also expected to be stable.
The existence of stable monopole and dyon configurations may have tremendous consequences in cosmology if the early universe ever was in the AdS phase. Stable solutions exist only with a negative cosmological constant. The existence of the boundary in the AdS space must be playing a crucial role. The connection to the AdS/CFT correspondence is yet to be explored. A more thorough analysis of the solutions with varying $`\mathrm{\Lambda }`$ as well as blackhole solutions will be presented in separate publications.
Acknowledgments
This work was supported in part by the U.S. Department of Energy under contracts DE-FG02-94ER-40823 and DE-FG02-87ER40328.
|
no-problem/9906/physics9906039.html
|
ar5iv
|
text
|
# 1 Introduction
## 1 Introduction
Sonoluminescence, the phenomenon of emission of light from small bubble occurring during ultrasonic excitation, has been known for more than half a century. In this phenomenon an intense standing wave increases the pulsation of a bubble of gas trapped at a velocity node to attain sufficient amplitude so as to emit picosecond flash of light. The analysis of the dynamics of a small bubble or cavity in a fluid dates back to the work of Lord Rayleigh at the beginning of this century. A large number of publications followed in subsequent decades, including the studies of oscillating bubbles by Plesset , Eller $`\&`$ Crum , Flynn, Lauterborn, Prosperetti, and many others. Experiments indicate that the collapse is remarkably spherical, and that water is the best fluid for SL. Some noble gas is essential for stable SL, and that the light intensity increases as ambient temperature is lowered. The theory of light emitting mechanism is still open, but the traditional scenario is supersonic bubble collapse launching an imploding shock wave which ionizes the bubble contains so as cause it to emit. Bremsstrahlung radiation is the best theory for SL phenomenon. In the theory of radiation by Bremsstrahlung as the bubble collapses, produced shock wave ionized the gas and cause it to radiate. The imploding shock wave concentrated at the center of bubble, through this concentration, shock wave warmed the gas until will receive to the center of bubble and explode, through this explosion gas which was compressed just behind the imploding shock front now finds itself in front of the shock front again. As the shock passes through these particles a second time, there is another burst of heating and maximum temperature reached by the exploding shock wave . The mean idea of our work is that, if in the imploding regime the plasma cools by radiation then in the exploding epoch, shock front can not warm particles again to produce intensive radiation. Thus by comparing cooling time with dynamical time for collapsing, we can say that if cooling time is less than collapsing time then plasma inside bubble will cool and the exploding shock wave can not warm this gas again. This constraint gives us some useful parameter for SL, therefore, we can explain results of some experiments such as why the light intensity increases as ambient temperature is lowered.
## 2 Collapse Mechanism and radiation
Explanation of the light-emitting mechanism of SL naturally seeks to interpret the featureless spectrum in terms of emission from a hot spot, for example black body radiation, if the radiation and matter are near to equilibrium, or Bremsstrahlung from accelerating unbound electrons if the light-emitting region is hot enough to be ionized yet sufficiently rarefied so as to be transparent to radiation. The shock wave model provides an extra stage of energy focusing by assuming that supersonic inward collapse of the bubble wall launches a shock wave into the bubble’s interior. This shock can run through the already compressed gas inside the bubble, increasing its amplitude and speed as it focuses towards the origin. There is now a surface of radius $`R_s`$ (the radius of the shock front), within the bubble of radius $`R`$. The similarity solution is obtained by assuming that the shock radius takes this form :
$$R_s=A(t)^\alpha ,$$
(1)
(Guderley 1942) where time is measured from the time of the convergence of the shock, and A is the ”launch” condition of the shock, which couples the shock to the bubble motion. The similarity solution yields an exponent $`\alpha `$ of 0.72 in air and 0.69 for noble gases. Since the exponent is less that unity, the Mach number of shock as goes to the origin approaches infinity. The bubble wall is collapsing at the speed of sound when it is passing through its ambient radius. In this case (Barber et al) showed that:
$$R_s=R_0(\frac{t}{t_0})^\alpha ,$$
(2)
where
$$t_0=\alpha \frac{R_0}{c_0},$$
(3)
In terms of Mach number
$$M=\left|\frac{t_0}{t}\right|^{1\alpha }.$$
(4)
one can define dynamical time or collapsing time according to:
$`t_{dy}`$ $`=`$ $`{\displaystyle \frac{R(t)}{\frac{dR(t)}{dt}}},`$ (5)
$`t_{dy}`$ $`=`$ $`{\displaystyle \frac{\alpha R_0}{c_0}}({\displaystyle \frac{1}{M}})^{\frac{1}{1\alpha }}.`$ (6)
where $`t_{dy}`$ is the time scale for the shock to have a radius smaller than $`R_s`$. As the shock wave approaches the center the temperature immediately behind the imploding shock front increases by this factor $`\frac{T}{T_0}=M^2`$. When the shock wave converges to the origin it explodes from the origin with the same similarity solution. Thus the gas which was compressed just behind imploding shock front now finds itself in the front of the shock front again. As the shock passes through these particles a second time, there is another burst of heating and the maximum temperatures reached by the exploding shock wave now increases by the factor $`\frac{T}{T_0}=M^4`$.
## 3 Cooling Condition for Radiation
The temperature increase inside the bubble by shock wave ionizes the region. The free electrons released by the heating will accelerate and radiate light as they collide with the ions. The Bremsstrahlung radiation generated per second per volume is equal to :
$$\frac{dE}{dtdv}=\frac{16z^2e^6n^2}{3m^2c^3h}(\frac{2mkT}{\pi })^{1/2}g_B1.5\times 10^{27}n^2T^{1/2}g_B.$$
(7)
where $`e`$ and $`m`$ are the electron charge and mass, $`n`$ the density of free electrons and ions, $`c`$ the speed of light, and $`g_B`$ the Gount factor. In the temperature of T, the energy density of this plasma is equal to:
$$\frac{dE}{dv}=\frac{3}{2}nkT.$$
(8)
So we can introduce cooling time for this plasma
$`t_{cool}`$ $`=`$ $`{\displaystyle \frac{\frac{dE}{dv}}{\frac{dE}{dtdv}}}.`$ (9)
$`t_{cool}`$ $`=`$ $`{\displaystyle \frac{9m^2c^3h}{32z^2e^6}}({\displaystyle \frac{\pi }{3mk}})^{1/2}n^1T^{1/2}.`$ (10)
$`{\displaystyle \frac{0.26}{g_B}}\times 10^{12}n^1T^{1/2}.`$ (11)
As the bubble collapses the value of $`n`$ changes with the radius, by using the dynamical equation of shock wave we can interpret the above density in terms of initial density and radius of gas in the bubble as:
$$n=n_0(\frac{R_0}{R_s})^3.$$
(12)
By using dynamical equation for shock wave, cooling time scale can be obtained in terms of Mach number
$$t_{cool}=\frac{0.26}{g_B}\frac{10^{12}}{z^2n_0}(\frac{1}{M})^{\frac{3\alpha }{1\alpha }}T^{1/2}.$$
(13)
Now we compare this time scale with dynamical time scale for the shock wave. If $`t_{cool}>t_{dy}`$ then for the imploding shock wave, as the shock wave implodes toward the origin, plasma cannot cool thus adiabatic condition holds. So in the explosion epoch temperature can rise sufficiently to produce an extensive radiation. Now if $`t_{cool}<t_{dy}`$, then as the imploding shock wave implodes towards the center of bubble by the radiation mechanism, plasma cools and temperature can not rise in such a way that exploding shock can produce an intensive radiation. As mentioned before, temperature behind the shock wave rises by a factor of $`\frac{T}{T_R}=M^2`$, where $`T(R)`$ is the temperature of the bubble when its radius is $`R`$. For the constraint $`t_{cool}>t_{dyn}`$, temperature of gas inside the bubble rises with an adiabatic compressing:
$$T(R)=T_0(\frac{R_0}{R})^{3(\gamma 1)},$$
(14)
where $`T_0`$ is the ambient temperature of bubble and $`R_0`$ is the size of the bubble when it is in a static mechanical equilibrium. By using shock wave dynamics, the above equation can be obtained in terms of Mach number, so we have:
$$T(R)=T_0M^{\frac{3}{2}\frac{\alpha }{1\alpha }(\gamma 1)}.$$
(15)
By substituting Eqs.(6 & 13) in the adiabatic condition for imploding shock wave, constraint condition obtain in the term of ambient temperature, Mach number, initial density and initial radius and constants of gas
$$R_0<\frac{0.26}{g_B}\frac{c_010^{12}}{\alpha zn_0}M^{\frac{3\alpha \gamma 11\alpha +4}{2(1\alpha )}}T_0^{1/2},$$
(16)
where $`c_0`$ is the velocity of sound and $`n_0`$ is the initial density of gas inside the bubble. From experiment we have $`n_04.16\times 10^{19}\frac{par}{cm^3}`$ and $`g_B=1.2`$. Measurements indicate that the bubble wall is collapsing at more than 4 times the ambient speed of sound in air. So the above inequality can be written this way:
$$R_0<0.258\times 10^4T_{0}^{}{}_{}{}^{1/2}.$$
(17)
As we said temperature rises by two processes:
i. By shock wave $`T=M^2T(R)`$.
ii. By adiabatic collapsing of bubble$`T(R)=M^{(\gamma 1)(\frac{3\alpha }{1\alpha })}T_0`$.
The factor of increase in temperature in the above processes is $`T=M^2M^{(\gamma 1)(\frac{3\alpha }{1\alpha })}T_0`$. For air, we have $`T=1144\times T_0`$ where $`T_0`$ is the ambient temperature, If we consider room temperature for ambient temperature, the temperature of bubble rises up to $`10^5`$. In the Bremsstrahlung radiation we considered all of our gases to be ionized. Verification of that statement can be obtained by Saha’s equation
$$\frac{q^2}{1q}=2.4\times 10^{21}T^{3/2}exp(\frac{\chi }{kT})\frac{1}{n}$$
(18)
where $`q=\frac{n_e}{n}`$ is the degree of ionization, $`\chi `$ is the ionization potential, and the pre factor $`T`$ is given in Kelvin. If we put the given parameters for SL in Saha’s equation for air, we find that $`n_e=n_i=n`$. This is the reason for considering completely ionized gas Bremsstrahlung formula. The inequality curve of Eq.(17) gives us a constraint for producing radiation in SL (Fig). On the other hand Bradley et al (1994) show the extreme sensitivity of SL to external parameters such as the water temperature. They show that as the water temperature decreases from $`40^{}`$ to $`1^{}`$, the intensity of light emission increases by a factor of over $`200`$ . At about $`0^{}`$ the purple light emitted by the bubble is so bright that one can see it by an unaided eye, but at $`40^{}`$ the SL is barely visible even in a darkened room. According to Bradley et al (1994) by light scattering technique initial radius of bubble obtained respect to ambient temperature. Comparing constraint curve for SL with the experimental relation between ambient radius of bubble and temperature (Fig), we see that above about $`25^{}`$ constraint for SL breaks down and confirms the experiment results of Bradley et al (1994).
## 4 Acknowledgment
I would like to thank Dr.Rasool Sadighi and Dr.Kamal Seied yaghobi for usefull discussion and comments.
## 5 Figure Caption
Bar lines indicated the experimental measurement between ambient temperature and initial radius of bubble. On the other hand there is a constraint line form Eq.(17), in such a way that below this line flashes of light can produce in SL. About $`25^{}`$ experimental bars cross our line and causes to SL break downs. This confirms the results of Bradley et at (1994)
|
no-problem/9906/astro-ph9906181.html
|
ar5iv
|
text
|
# Neural Networks for Spectral Analysis of Unevenly Sampled Data
## 1 Introduction
Periodicity analysis of unevenly collected data is a relevant issue in several scientific fields. Classical spectral analysis methods are unsatisfactory to solve the problem. In this paper we present a neural network based estimator system which performs well the frequency extraction from unevenly sampled signals. It uses an unsupervised Hebbian nonlinear neural algorithm to extract the principal components of the signal auto-correlation matrix, which, in turn, are used by the MUSIC frequency estimator algorithm to extract the frequencies , , . We generalize this method to avoid an interpolation preprocessing step, which generally adds high noise to the signal, and improve the system performance by using a new stop criterion to avoid overfitting problems. The experimental results are obtained comparing our methodology with the others known in literature (see , , , , , ).
## 2 Evenly and unevenly sampled data
In what follows, we assume $`x`$ to be a physical variable measured at discrete times $`t_i`$. $`x(t_i)`$ can be written as the sum of the signal $`x_s`$ and random errors $`R`$: $`x_i=x(t_i)=x_s(t_i)+R(t_i)`$. The problem we are dealing with is how to estimate fundamental frequencies which may be present in the signal $`x_s(t_i)`$ , , .
If $`X`$ is measured at uniform time steps (even sampling) there are a lot of tools to effectively solve the problem which are based on Fourier analysis , , . These methods, however, are usually unreliable for unevenly sampled data . For instance, the typical approach of resampling the data into an evenly sampled sequence, through interpolation, introduces a strong amplification of the noise which affects the effectiveness of all Fourier based techniques which are strongly dependent on the noise level.
To solve the problem of unevenly sampled data, we consider two classes of spectral estimators:
Spectral estimators based on Fourier Trasform (Least Squares methods);
Spectral estimators based on the eingevalues and eingevectors of the covariance matrix (Maximum Likelihood methods).
Classic Periodogram , , , Lomb’s Periodogram , Scargle’s Periodogram , DCDFT are the methods of the first class that we use, while MUSIC , , and ESPRIT belong to second class.
The methods based on the covariance matrix are more recent and have great potentiality. Starting by this consideration, we develop a method based on the MUSIC estimator. It is compared with classic methods to highlight the results.
## 3 The Neural Estimator
In the last years several papers dealed with learning in PCA neural nets , , , , finding advantages, problems and difficulties of such neural networks. In what follows we shall use a robust hierarchical learning algorithm
$`𝐰_{k+1}(i)`$ $`=`$ $`𝐰_k(i)+\mu _kg(y_k(i))𝐞_k(i),`$ (1)
$`𝐞_k(i)`$ $`=`$ $`𝐱_k{\displaystyle \underset{j=1}{\overset{i}{}}}y_k(j)𝐰_k(j)`$ (2)
where $`𝐰_k(i)`$ is the weight vector of the $`ith`$ output neuron at step $`k`$, $`y_k(i)`$ is the corresponding output, $`\mu _k`$ is the learning rate and $`g(t)=\mathrm{tanh}\left(\alpha t\right)`$ is learning function because it has been experimentally shown that it is the best performing one in our problem , .
Our neural estimator (ne) can be summarized as follows:
* Preprocessing: calculate and subtract the average pattern to obtain zero mean process with unity variance.
Interpolate input data if it is the case.
* Initialize the weight matrix and the other neural network parameters;
* Input the $`kth`$ pattern $`𝐱_k=[x(k),\mathrm{},x(k+N+1)]`$ where $`N`$ is the number of input components.
* Calculate the output for each neuron $`y(j)=𝐰^T(j)𝐱_ii=1,\mathrm{},p`$.
* Modify the weights $`𝐰_{k+1}(i)=𝐰_k(i)+\mu _kg(y_k(i))𝐞_k(i)i=1,\mathrm{},p`$.
* If convergence test is true then goto STEP 8.
* $`k=k+1`$. Goto STEP 3.
* End.
* Frequency estimator: we use the frequency estimator MUSIC. It takes as input the weight matrix columns after the learning. The estimated signal frequencies are obtained as the peak locations of the function of following equation , , :
+ $`P_{MUSIC}={\displaystyle \frac{1}{M_{i=1}^M|𝐞_f^H𝐰(i)|^2}}`$
where $`𝐰(i)`$ is the $`i`$th neural network weight vector after learning, and $`𝐞_f^H`$ is the pure sinusoidal vector. In the case of an interpolation preprocessing $`𝐞_f^H=[1,e_f^{j2\pi f},\mathrm{},e_f^{j2\pi f(L1)}]^H`$. In the generalization to non interpolated input data, $`𝐞_f^H=[1,e_f^{j2\pi ft_0},\mathrm{},e_f^{j2\pi ft_{(L1)}}]^H`$ where $`\{t_0,t_1,\mathrm{},t_{\left(L1\right)}\}`$ are the first $`L`$ components of the temporal coordinates of the uneven signal.
When $`f`$ is the frequency of the $`i`$th sinusoidal component, $`f=f_i`$, we have $`𝐞=𝐞_i`$ and $`P_{MUSIC}\mathrm{}`$. In practice we have a peak near and in corrispondence of the component frequency. Estimates are related to the highest peaks
Furthermore, to optimize the performance of the PCA neural networks, we stop the learning process when $`_{i=1}^p|𝐞_f^H𝐰(i)|^2<M`$ $`f`$, so avoiding overfitting problems. In fact leaving the stop condition used in the ne causes to the ne to find periodicities not present in the signal, while the new condition preserves it from this problem. A simple example is illustrated in figure 2 where we can see how the frequency identification varies depending on the stop condition (see next section for signal information). In fact, without the early stopping (see figure 2.b), $`\mathrm{\Delta }f`$ after $`100`$ epochs remains at a value between $`0.20`$ and $`0.25`$ and we cannot know when the system reaches the best performance. The new stopping criterion, instead, permits to $`\mathrm{\Delta }f`$ to have a final value about $`0.0`$ just after $`50`$ epochs (see figure 2.a).
## 4 Experimental Results
Many experiments on synthetic and real signals were made, and in this paper we present the results obtained with one specific real signal, which highlights the main features of our problem.
The real signal is related to the Cepheid SU Cygni . The sequence was obtained with the photometric technique UBVRI and the sampling made from June to December 1977. The light curve is composed by 21 samples, and has a period of $`3.8^d`$, as shown in figure 1.
The first experiment is concerning the interpolation. In this case we apply three different methods by using the Signal Processing Mathlab <sup>@</sup> Toolbox: linear, cubic and spline, because they are quite simple and the most used ones. In figure 3 there is a plotting of the interpolating functions and the frequency estimates obtained by the ne with the spline interpolated signal as input.
In this case, the parameters of the ne are: $`N=10`$, $`p=2`$, $`\alpha =20`$, $`\mu =0.001`$. The estimate frequency interval is $`[0(1/JD),0.5(1/JD)]`$. The estimated frequency without interpolation is $`0.260`$ (1/JD).
A comparison is made with the other methods cited in a previous section and the experimental results are shown in figure 4 and in table 1. Only the Lomb’s Periodogram is in agreement with the right periodicity, but showing some spurious peaks. Furthermore, if we enlarge the frequency window for the two best performing methods, while the ne continues to work well, the Lomb’s periodogram does not work at all as illustrated in figure 5.
## 5 Concluding Remarks
In this paper we have illustrated an improved technique based on PCA neural Networks and MUSIC to estimate the frequency of unevenly sampled data. It has been shown that it obtains good results on real data (here we used the SU Cygni light curve) compared with other well-known methods. In fact, it obtains a good estimate of the signal frequency also with few unevenly sampled inputs, it reduces the noise problems related to input data interpolation, it optimizes the convergence by introducing an early stopping criterion, and, finally, it is more resistant to the dimension of the frequency windows.
Future research lines regard the introduction of genetic algorithms to optimize the weight initialization of the PCA neural networks and to use filters to extract and identify one frequency at each time when dealing with multi-frequency signals.
|
no-problem/9906/hep-ph9906281.html
|
ar5iv
|
text
|
# Theoretical Constraints on the Vacuum Oscillation Solution to the Solar Neutrino Problem
## 1 Introduction
There are three main explanations of the solar neutrino flux deficits, requiring oscillations of electron neutrinos into other species. Namely, the small and large angle MSW solutions, and the vacuum oscillation (VO) solution. In this paper we focus on the latter, which requires the relevant mass splitting and mixing angle in the range
$`5\times 10^{11}\mathrm{eV}^2<`$ $`\mathrm{\Delta }m_{sol}^2`$ $`<1.1\times 10^{10}\mathrm{eV}^2,`$
$`\mathrm{sin}^22\theta _{sol}`$ $`>`$ $`0.67.`$ (1)
On the other hand, Superkamiokande observations of atmospheric neutrinos require neutrino oscillations (more precisely $`\nu _\mu \nu _\tau `$ oscillations if we do not consider oscillations into sterile species) driven by a mass splitting and a mixing angle in the range
$`5\times 10^4\mathrm{eV}^2<`$ $`\mathrm{\Delta }m_{atm}^2`$ $`<10^2\mathrm{eV}^2,`$
$`\mathrm{sin}^22\theta _{atm}`$ $`>`$ $`0.82.`$ (2)
Let us remark the enormous hierarchy of mass splittings<sup>2</sup><sup>2</sup>2It has been pointed out that disregarding Cl data on solar neutrinos, vacuum oscillations with much larger mass splitting could account for the solar anomaly. between the different species of neutrinos, $`\mathrm{\Delta }m_{sol}^2\mathrm{\Delta }m_{atm}^2`$, which is apparent from eqs.(1, 2).
It has been argued that the extreme tinyness of $`\mathrm{\Delta }m_{sol}^2`$ in this scenario could be related to some continuous or discrete symmetry at high energy . However, independently of the origin of the small splittings, it must be required that their size is not spoiled by radiative corrections, the dominant part of which can be accounted by integrating the renormalization group equations (RGEs) between the scale at which the effective mass matrix is generated and low energy. The aim of this paper is to analyze under which circumstances this is in fact the case. As a result, we obtain important theoretical restrictions on the VO scenario.
Let us introduce now some notation. We define the effective mass term for the three light (left-handed) neutrinos in the flavour basis, $`\nu ^T=(\nu _e,\nu _\mu ,\nu _\tau )`$, as
$`={\displaystyle \frac{1}{2}}\nu ^T_\nu \nu +\mathrm{h}.\mathrm{c}.`$ (3)
The mass matrix, $`_\nu `$, is diagonalized in the usual way, i.e. $`_\nu =V^{}DV^{}`$, where $`D=\mathrm{diag}(m_1e^{i\varphi },m_2e^{i\varphi ^{}},m_3)`$ and $`V`$ is a unitary ‘CKM’ matrix, relating flavour to mass eigenstates
$`\left(\begin{array}{c}\nu _e\\ \nu _\mu \\ \nu _\tau \end{array}\right)=\left(\begin{array}{ccc}c_2c_3& c_2s_3& s_2e^{i\delta }\\ c_1s_3s_1s_2c_3e^{i\delta }& c_1c_3s_1s_2s_3e^{i\delta }& s_1c_2\\ s_1s_3c_1s_2c_3e^{i\delta }& s_1c_3c_1s_2s_3e^{i\delta }& c_1c_2\end{array}\right)\left(\begin{array}{c}\nu _1\\ \nu _2\\ \nu _3\end{array}\right).`$ (4)
Here, $`s_i`$ and $`c_i`$ denote $`\mathrm{sin}\theta _i`$ and $`\mathrm{cos}\theta _i`$, respectively, and in the rest of the paper we will neglect CP-violating phases. In the following, we label the mass eigenstates $`\nu _i`$ in such a way that $`|\mathrm{\Delta }m_{12}^2|<|\mathrm{\Delta }m_{23}^2||\mathrm{\Delta }m_{13}^2|`$, where $`\mathrm{\Delta }m_{ij}^2m_j^2m_i^2`$ ($`m_{\nu _3}^2`$ is thus the most split eigenvalue). Consequently, $`\mathrm{\Delta }m_{sol}^2`$, $`\theta _{sol}`$ correspond to $`\mathrm{\Delta }m_{12}^2`$, $`\theta _3`$, while $`\mathrm{\Delta }m_{atm}^2`$, $`\theta _{atm}`$ correspond to $`\mathrm{\Delta }m_{23}^2\mathrm{\Delta }m_{13}^2`$, $`\theta _1`$ respectively. In our notation $`\mathrm{\Delta }m_{sol}^2`$, $`\mathrm{\Delta }m_{atm}^2`$ denote always the “experimental” splittings of eqs.(1, 2), while $`\mathrm{\Delta }m_{12}^2`$, $`\mathrm{\Delta }m_{23}^2`$ denote the computed splittings once the radiative corrections are incorporated. Concerning the $`\theta _2`$ angle, according to the most recent combined analysis of SK $`+`$ CHOOZ data (last paper of ref. ) it is constrained to have low values, $`\mathrm{sin}^22\theta _2<0.36(0.64)`$ at 90% (99%) C.L.
We assume along the paper that the effective mass matrix for the left-handed neutrinos, $`_\nu `$, is generated at some high energy scale, $`\mathrm{\Lambda }`$, by some unspecified mechanism. Below $`\mathrm{\Lambda }`$, we consider two possibilities: either the effective theory is the Standard Model (SM) or the minimal supersymmetric Standard Model (MSSM) with unbroken $`R`$parity. In the first case, the lowest dimension operator producing a mass term of this kind is uniquely given by
$`{\displaystyle \frac{1}{4}}\kappa \nu ^T\nu HH+\mathrm{h}.\mathrm{c}.`$ (5)
where $`\kappa `$ is a matricial coupling and $`H`$ is the ordinary (neutral) Higgs. Obviously, $`_\nu =\frac{1}{2}\kappa H^2`$. The effective coupling $`\kappa `$ runs with the scale from $`\mathrm{\Lambda }`$ to $`M_Z`$, with a RGE given by
$`16\pi ^2{\displaystyle \frac{d\kappa }{dt}}=\left[3g_2^2+2\lambda +6Y_t^2+2\mathrm{T}\mathrm{r}𝐘_𝐞^{}𝐘_𝐞\right]\kappa {\displaystyle \frac{1}{2}}\left[\kappa 𝐘_𝐞^{}𝐘_𝐞+(𝐘_𝐞^{}𝐘_𝐞)^T\kappa \right],`$ (6)
where $`t=\mathrm{log}\mu `$, and $`g_2,\lambda ,Y_t,𝐘_𝐞`$ are the $`SU(2)`$ gauge coupling, the quartic Higgs coupling, the top Yukawa coupling and the matrix of Yukawa couplings for the charged leptons, respectively. The last term of eq.(6) is the most important one for our purposes, since it modifies the texture of $`_\nu `$. It is worth noticing that the modification of a mass eigenvalue is always proportional to the mass eigenvalue itself. In the MSSM case, things are very similar but with an important difference. Namely, the term that modifies the texture in the RGEs has the same form as in eq.(6) but with coefficient $`+1`$ instead of $`\frac{1}{2}`$. Moreover, in the MSSM the $`𝐘_𝐞`$ couplings are $`1/\mathrm{cos}\beta `$ larger than the SM ones. All this implies that the effect of the RGEs in the supersymmetric case is $`2/\mathrm{cos}^2\beta =2(1+\mathrm{tan}^2\beta )`$ times larger (for $`\mathrm{tan}\beta =2`$ this already represents one order of magnitude). It should be mentioned that in the supersymmetric case there are two stages of running: from $`\mathrm{\Lambda }`$ to $`M_{SUSY}`$ with the MSSM RGEs, and from $`M_{SUSY}`$ to $`M_Z`$ with the SM ones (the latter is normally much less important than the former).
In order to study the quantitative effect of the RGEs on the mass splittings and mixing angles, it is convenient to consider separately the following three possible scenarios
$`\mathrm{A}`$ $`:`$ $`|m_3||m_{1,2}|,`$
$`\mathrm{B}`$ $`:`$ $`|m_1||m_2||m_3|,`$ (7)
$`\mathrm{C}`$ $`:`$ $`|m_1||m_2||m_3|.`$
In case A, radiative corrections are generically not dangerous. The reason is that, as stated before, the mass eigenvalues renormalize proportionally to themselves, i.e. $`\mathrm{\Delta }_{RGE}m_i=(K_0+K_i)m_i`$, where $`K_0`$ is the universal contribution for all the eigenvalues and $`|K_i|1`$. Thus, unless $`m_{1,2}^2\mathrm{\Delta }m_{12}^2`$, the running cannot spoil the initial smallness of the solar mass-splitting (this is in particular the case of a hierarchical spectrum $`m_1^2m_2^2m_3^2`$). Roughly speaking, the mass splittings generated radiatively get larger than the allowed range of eq. (1) for $`m_{1,2}^210^4`$ eV<sup>2</sup>, although the precise value depends on several details, in particular on the values of the mixing angles.
On the other hand, case B \[cosmologically relevant for $`m_i=𝒪`$(eV)\], has been shown to be inconsistent with the VO solution in refs. . Namely the mass splittings $`\mathrm{\Delta }m_{ij}^2`$ generated through the running are several orders of magnitude larger than the required VO splitting, even for $`\mathrm{\Lambda }`$ very close to $`M_Z`$. According to the previous discussion, the supersymmetric case works even worse. The only way-out would be an extremely artificial fine-tuning between the initial values of the mass splittings (and mixing angles) and the effect of the RG running, something clearly unacceptable.
Finally, the impact of the radiative corrections on a spectrum of the type C has not been considered yet in the literature. Since in this case the large $`\mathrm{\Delta }m_{sol}^2\mathrm{\Delta }m_{atm}^2`$ hierarchy forces $`m_{1,2}^2\mathrm{\Delta }m_{12}^2`$, one can expect important radiative effects. The analysis of this scenario is performed in section 2, where we study in two separate subsections the possibilities that $`m_1`$ and $`m_2`$ have equal or opposite signs. The conclusions are presented in section 3.
## 2 The case $`m_1^2m_2^2m_3^2`$
As explained in the Introduction, radiative corrections play an important rôle when $`m_1^2m_2^2m_3^2`$, case in which the mass splitting relevant for solar oscillations is the one between the heavier neutrinos. In this framework, radiative corrections can actually make $`\mathrm{\Delta }m_{12}^2\mathrm{\Delta }m_{sol}^2`$ in contradiction with observations. This effect will be stronger the heavier is the overall neutrino mass scale: the most conservative case thus corresponds to $`m_1^2m_2^2\mathrm{\Delta }m_{atm}^2`$ and $`m_3^20`$ (masses smaller than this cannot accommodate atmospheric oscillations of the required frequency).
The rationale is then the following: at some high-energy scale $`\mathrm{\Lambda }`$ one assumes that new physics generates a dimension-5 operator leading to non-zero neutrino masses and fixes $`_\nu (\mathrm{\Lambda })`$ such that, with good approximation $`m_1^2=m_2^2`$ and $`m_3^2=0`$. The most important radiative corrections to this tree-level masses are proportional to $`\mathrm{ln}(\mathrm{\Lambda }/M_Z)`$ and can be included using standard RG techniques, that is, running $`_\nu `$ down from $`\mathrm{\Lambda }`$ to $`M_Z`$ using the relevant RGEs. The latter depends on what is the effective theory below $`\mathrm{\Lambda }`$. As we said, we consider two cases: SM and MSSM, and the RGEs relevant for these two effective theories can be found e.g. in ref. .
The analytical integration of the RGEs is straightforward in the leading-log approximation and the additional simplification that $`m_30`$ at all scales permits us to concentrate on the two other masses alone. The results are qualitatively different depending on the relative sign between $`m_1`$ and $`m_2`$ and we consider the two cases separately in the next subsections.
### 2.1 $`m_1m_2`$
After integration from $`\mathrm{\Lambda }`$ to $`M_Z`$, the radiatively corrected $`_\nu (M_Z)`$ has eigenvalues which in first approximation are given by
$$\begin{array}{cc}m_1=& m_\nu ,\hfill \\ m_2=& m_\nu \left[1+2ϵ_\tau (1c_1^2c_2^2)\right],\hfill \\ m_3=& 0.\hfill \end{array}$$
(8)
These expressions include the leading-log radiative corrections to the mass differences and are obtained under the approximation that the initial 1-2 mass splitting is zero. In eq.(8), the family-universal renormalization effect (not important for our discussion) has been absorbed in $`m_\nu `$, which is fixed to give the proper value for $`\mathrm{\Delta }m_{31}^2\mathrm{\Delta }m_{atm}^2`$. The $`\theta _1`$, $`\theta _2`$, $`\theta _3`$ angles have been kept as free parameters. Our numerical results are always obtained integrating numerically the RGEs and confirm that the analytical expressions we write represent an excellent approximation. For our numerics we choose both the lower and upper limits of the allowed range for $`\mathrm{\Delta }m_{atm}^2`$, thus fixing $`m_\nu =2.2\times 10^2\mathrm{eV}`$ or $`0.1\mathrm{eV}`$ (corresponding to $`\mathrm{\Delta }m_{31}^2=5\times 10^4\mathrm{eV}^2`$ and $`10^2\mathrm{eV}^2`$, respectively).
The parameter $`ϵ_\tau `$ depends on what is the effective low-energy theory below $`\mathrm{\Lambda }`$ :
$`ϵ_\tau ={\displaystyle \frac{h_\tau ^2}{32\pi ^2}}\mathrm{log}{\displaystyle \frac{\mathrm{\Lambda }}{M_Z}}(\mathrm{SM}),`$ (9)
$`ϵ_\tau ={\displaystyle \frac{h_\tau ^2}{32\pi ^2}}\left[{\displaystyle \frac{2}{\mathrm{cos}^2\beta }}\mathrm{log}{\displaystyle \frac{\mathrm{\Lambda }}{M_{SUSY}}}+\mathrm{log}{\displaystyle \frac{M_{SUSY}}{M_Z}}\right](\mathrm{MSSM}),`$ (10)
where $`M_{SUSY}`$ sets the mass scale for the supersymmetric spectrum (we take $`M_{SUSY}1\mathrm{TeV}`$). As usual, the size of $`ϵ_\tau `$ grows logarithmically with the scale of new physics $`\mathrm{\Lambda }`$ (a conservative estimate we often make is to choose a low value $`\mathrm{\Lambda }=1\mathrm{TeV}`$). Also, for sufficiently large $`\mathrm{\Lambda }/M_Z`$, the size of $`ϵ_\tau `$ is enhanced by a factor $`2/\mathrm{cos}^2\beta =2(1+\mathrm{tan}^2\beta )`$ in the MSSM with respect to the SM (already a factor 10 for $`\mathrm{tan}\beta =2`$) so that radiative corrections are more important in this case.
The typical size of $`ϵ_\tau `$ is $`8\times 10^7`$ ($`8\times 10^6`$) for $`\mathrm{\Lambda }=10^3\mathrm{GeV}`$ ($`\mathrm{\Lambda }=10^{12}\mathrm{GeV}`$) in the SM and $`8\times 10^5`$ for $`\mathrm{\Lambda }=10^{12}\mathrm{GeV}`$ in the MSSM with $`\mathrm{tan}\beta =2`$. According to Eq. (8) this would lead to
$`{\displaystyle \frac{\mathrm{\Delta }m_{sol}^2}{m_\nu ^2}}={\displaystyle \frac{\mathrm{\Delta }m_{sol}^2}{\mathrm{\Delta }m_{atm}^2}}=4ϵ_\tau (1c_1^2c_2^2),`$ (11)
too large compared with the observed value unless there is a cancellation in $`(1c_1^2c_2^2)`$, which requires $`\mathrm{sin}^22\theta _{1,2}0`$. This is far from the best-fit values mentioned in the Introduction. Choosing $`\mathrm{sin}^22\theta _11`$ and $`\mathrm{sin}^22\theta _20`$ we must conclude that $`\mathrm{\Delta }m_{12}^2`$ turns out to be too large for vacuum oscillations of solar neutrinos.
The precise results are given in Figure 1, which shows the predicted $`\mathrm{\Delta }m_{12}^2`$ in $`\mathrm{eV}^2`$ (solid lines) as a function of $`c_1^2c_2^2`$ for the SM case with $`\mathrm{\Lambda }=10^3\mathrm{GeV}`$ and $`\mathrm{\Delta }m_{atm}^2=5\times 10^4\mathrm{eV}^2`$ (lower curve) and $`10^2\mathrm{eV}^2`$ (upper curve). The experimental constraints on $`\theta _{1,2}`$ leave open the windows $`0c_1^2c_2^20.142`$ and $`0.232c_1^2c_2^20.71`$, as indicated. The neutrino mass splitting required by VO solar oscillations is marked by the horizontal dotted lines. As was clear from the previous discussion, there is no overlapping between the $`\mathrm{\Delta }m_{12}^2`$ predicted and the $`\mathrm{\Delta }m_{sol}^2`$ required. Indeed, $`\mathrm{\Delta }m_{12}^2`$ is always much larger than the allowed range. In the MSSM (or for larger $`\mathrm{\Lambda }`$) the situation is even worse because in both cases $`\mathrm{\Delta }m_{12}^2`$ increases significantly in the way discussed above.
Let us turn in more detail to the mixing angles in this scenario. At the scale $`\mathrm{\Lambda }`$ one has some mixing angles $`\theta _i`$ which will be different in general at the scale $`M_Z`$ after radiative corrections to $`_\nu `$ have been included. At the same level of approximation as in Eqs. (8), the eigenvectors of the perturbed neutrino mass matrix are of the form
$`V_1^{}={\displaystyle \frac{1}{\sqrt{\alpha ^2+\beta ^2}}}(\alpha V_1+\beta V_2),V_2^{}={\displaystyle \frac{1}{\sqrt{\alpha ^2+\beta ^2}}}(\beta V_1+\alpha V_2),V_3^{}=V_3,`$ (12)
where $`V_i`$ are the eigenvectors corresponding to $`_\nu (\mathrm{\Lambda })`$
$`V_1=\left(\begin{array}{c}c_2c_3\\ c_1s_3s_1s_2c_3\\ s_1s_3c_1s_2c_3\end{array}\right),V_2=\left(\begin{array}{c}c_2s_3\\ c_1c_3s_1s_2s_3\\ s_1c_3c_1s_2s_3\end{array}\right),V_3=\left(\begin{array}{c}s_2\\ s_1c_2\\ c_1c_2\end{array}\right).`$ (22)
From this, we deduce that the relationships between $`\theta _i(M_Z)`$ and $`\theta _i(\mathrm{\Lambda })`$ are
$$\begin{array}{cc}\mathrm{sin}^22\theta _1(M_Z)=& \mathrm{sin}^22\theta _1(\mathrm{\Lambda }),\hfill \\ \mathrm{sin}^22\theta _2(M_Z)=& \mathrm{sin}^22\theta _2(\mathrm{\Lambda }),\hfill \\ \mathrm{sin}^22\theta _3(M_Z)=& \mathrm{sin}^22\theta _3(\mathrm{\Lambda })+\frac{2}{(1+r)^2}\left[\sqrt{r}(1r)\mathrm{sin}4\theta _3(\mathrm{\Lambda })+2r\mathrm{cos}4\theta _3(\mathrm{\Lambda })\right],\hfill \end{array}$$
(23)
where $`r\alpha ^2/\beta ^2`$. In leading-log approximation we have
$`{\displaystyle \frac{\alpha }{\beta }}{\displaystyle \frac{s_1c_3+c_1s_2s_3}{s_1s_3c_1s_2c_3}},`$ (24)
with all angles evaluated at the scale $`\mathrm{\Lambda }`$.
If we substitute this in (23) we find the simpler expression
$`\mathrm{sin}^22\theta _3(M_Z)={\displaystyle \frac{\mathrm{sin}^2\theta _2\mathrm{sin}^22\theta _1}{(1\mathrm{cos}^2\theta _1\mathrm{cos}^2\theta _2)^2}}.`$ (25)
For the bimaximal mixing case ($`s_20`$, $`c_1s_11/\sqrt{2}`$) we end up with $`\mathrm{sin}^22\theta _3(M_Z)0`$, which is not acceptable (observations require $`\mathrm{sin}^22\theta _30.67`$).
In conclusion, the scenario $`m_1m_2m_3`$ is very contrived from the theoretical point of view. It is not natural to expect in this framework the values of mass splittings and mixing angles which are suggested by experiment. As mentioned in the Introduction, the only way-out would be an extremely artificial fine-tuning between the initial values of the mass splittings (and mixing angles) and the effect of the RG running. If one insists on this possibility, starting for example with $`\mathrm{\Delta }m_{atm}^2=5\times 10^4\mathrm{eV}^2`$, $`s_20`$, $`c_1s_11/\sqrt{2}`$, $`\mathrm{\Lambda }=10^3`$ GeV (a conservative choice for the fine-tuning problem), one is forced to take the initial mass splitting and mixing angle within the narrow ranges $`|m_1^2m_2^2|(1.82\pm 0.02)\times 10^7\mathrm{eV}^2`$ and $`\theta _3\pi /2\pm 5.5\times 10^3`$ in order to compensate the effect of the RGEs and reproduce the required pattern of masses and mixings at $`M_Z`$ \[these numbers cannot be extracted from the previous eq.(11), as in this case the approximation of initial degenerate eigenvalues does not hold\]. One cannot certainly expect such a conspiracy between totally unrelated effects. If one slightly separates from these narrow ranges the low-energy mass splitting would be much larger than the required one. Of course, as $`\mathrm{\Delta }m_{atm}^2`$ or $`\mathrm{\Lambda }`$ are raised, or one goes to the supersymmetric case, the fine-tuning becomes much stronger.
Finally, it is interesting to note that for sizeable values of the cut-off ($`\mathrm{\Lambda }\stackrel{>}{_{}}10^{12}\mathrm{GeV}`$) and/or a supersymmetric scenario, the values of $`\mathrm{\Delta }m_{12}^2`$ are naturally 1-3 orders of magnitude larger than those represented in Fig.1, falling in the small-angle MSW range ($`3\times 10^6\mathrm{eV}^2<\mathrm{\Delta }m_{sol}^2<10^5\mathrm{eV}^2`$). This is appealing since, as has been discussed in this section, starting with $`\theta _{1,2}`$ mixing angles in agreement with experiment ($`s_20`$, $`c_1s_11/\sqrt{2}`$) the RGEs drive $`\mathrm{sin}^22\theta _3(M_Z)0`$, independently of its initial high-energy value, see eq.(23). This is exactly what is needed for a successful small-angle MSW solution to the solar neutrino problem.
### 2.2 $`m_1m_2`$
In this case, the neutrino mass eigenvalues at $`M_Z`$ are, in leading-log approximation
$$\begin{array}{cc}m_1=& m_\nu \left[1+2ϵ_\tau (s_1s_3c_1s_2c_3)^2\right],\hfill \\ m_2=& m_\nu \left[1+2ϵ_\tau (s_1c_3+c_1s_2s_3)^2\right],\hfill \\ m_3=& 0,\hfill \end{array}$$
(26)
with $`ϵ_\tau `$ as given by Eqs. (9) and (10). The mixing angles are, in first approximation, equal at $`M_Z`$ and $`\mathrm{\Lambda }`$. We fix again $`m_\nu \sqrt{\mathrm{\Delta }m_{atm}^2}`$. In order not to spoil the size of the required solar mass splitting, the radiative corrections should generate $`\mathrm{\Delta }m_{12}^2\mathrm{\Delta }m_{sol}^2`$. The prediction from (26) is
$`{\displaystyle \frac{\mathrm{\Delta }m_{12}^2}{m_\nu ^2}}={\displaystyle \frac{\mathrm{\Delta }m_{12}^2}{\mathrm{\Delta }m_{atm}^2}}=4ϵ_\tau \left[(\mathrm{cos}^2\theta _1\mathrm{sin}^2\theta _2\mathrm{sin}^2\theta _1)\mathrm{cos}2\theta _3\mathrm{sin}2\theta _1\mathrm{sin}\theta _2\mathrm{sin}2\theta _3\right].`$ (27)
Getting a sufficiently small number for this quantity requires some (in general delicate) correlation between the mixing angles, in such a way that
$`\mathrm{tan}2\theta _3{\displaystyle \frac{\mathrm{cos}^2\theta _1\mathrm{sin}^2\theta _2\mathrm{sin}^2\theta _1}{\mathrm{sin}\theta _2\mathrm{sin}2\theta _1}}.`$ (28)
It is remarkable that the bimaximal values of the mixing angles ($`\mathrm{sin}^22\theta _1\mathrm{sin}^22\theta _31`$ and $`\mathrm{sin}^22\theta _20`$) do satisfy (28).
Figures 2a,b show, for the SM case, the regions in the plane $`(\mathrm{sin}^22\theta _2,\mathrm{sin}^22\theta _3)`$ where the correlation (28) takes place, giving $`\mathrm{\Delta }m_{12}^21.1\times 10^{10}\mathrm{eV}^2`$ (the upper limit on $`\mathrm{\Delta }m_{sol}^2`$). The width of these allowed regions is controlled by $`\mathrm{\Delta }m_{sol}^2/(ϵ_\tau \mathrm{\Delta }m_{atm}^2)`$. The larger $`ϵ_\tau \mathrm{\Delta }m_{atm}^2`$ (or the smaller $`\mathrm{\Delta }m_{sol}^2`$), the thinner these regions get \[because a more delicate cancellation must take place in (28)\]. In figure 2a we have fixed $`\mathrm{sin}^22\theta _1=1`$ and $`\mathrm{\Delta }m_{atm}^2=5\times 10^4\mathrm{eV}^2`$, and we give the allowed areas for the two choices $`\mathrm{\Lambda }=10^3\mathrm{GeV}`$ (thick region, delimited by the outermost lines) and $`10^{12}\mathrm{GeV}`$ (thin region). If we choose $`\mathrm{\Delta }m_{atm}^2=10^2\mathrm{eV}^2`$ instead, the two regions would shrink significantly and will be somewhere inside the thin region shown for $`\mathrm{\Lambda }=10^{12}\mathrm{GeV}`$. The dashed lines delimit the allowed region for the two mixing angles $`\theta _2`$ and $`\theta _3`$ ($`0\mathrm{sin}^22\theta _20.64`$ and $`0.67\mathrm{sin}^22\theta _31`$). Figure 2b corresponds to the case $`\mathrm{sin}^22\theta _1=0.82`$ (the lower experimental limit) and the same values of other parameters as in figure 2a. The results are similar except for a shift towards smaller $`\mathrm{sin}^22\theta _2`$ values in the region of interest. Note in particular that the upper limit $`\mathrm{sin}^22\theta _20.64`$ is never reached in this scenario. We see that in the most conservative case, with $`\mathrm{\Delta }m_{atm}^2=5\times 10^4\mathrm{eV}^2`$ and $`\mathrm{\Lambda }=10^3\mathrm{GeV}`$, a significant portion of parameter space could accommodate a $`\mathrm{\Delta }m_{12}^2`$ of the right order of magnitude (including the bimaximal mixing solution). It is interesting to note that inside this region, starting with degenerate $`m_1,m_2`$ would lead to a correct $`\mathrm{\Delta }m_{sol}^2`$ at low energy, thus providing a dynamical origin for this small number. Notice however that as soon as $`\mathrm{\Delta }m_{atm}^2`$ or $`\mathrm{\Lambda }`$ are raised the required fine-tuning becomes much stronger. This occurs in particular if the lower bound $`\mathrm{\Delta }m_{atm}^2=5\times 10^4`$ that we have used is increased according to the analyses of the most recent data .
The situation is worse in the MSSM case. Roughly speaking, for $`\mathrm{tan}\beta =3`$ radiative corrections are 20 times larger than in the SM (with the same $`\mathrm{\Lambda }`$). The cancellation between mixing angles in (28) is thus much more delicate in the supersymmetric case, as expected.
## 3 Conclusions
The vacuum oscillation (VO) solution to the solar neutrino problem requires an extremely small mass splitting, $`\mathrm{\Delta }m_{sol}^2\stackrel{<}{_{}}10^{10}\mathrm{eV}^2`$. We have studied in this paper under which circumstances this smallness (whatever its origin) is or is not spoiled by radiative corrections, in particular by the running of the renormalization group equations (RGEs) between the scale at which the effective neutrino mass matrix is generated ($`\mathrm{\Lambda }`$) and low energy. We consider the cases where the effective theory below $`\mathrm{\Lambda }`$ is the Standard Model (SM) or the Minimal Supersymmetric Standard Model (MSSM). The results depend dramatically on the type of neutrino spectrum. In particular, if $`m_1^2m_2^2m_3^2`$, radiative corrections are always relatively small and do not cause any significant change in the splittings. On the other hand, if $`m_1^2m_2^2m_3^2`$, radiative corrections always induce mass splittings that are several orders of magnitude larger than the required $`\mathrm{\Delta }m_{sol}^2`$. Hence, this type of spectrum is not plausible for the VO solution. The only way-out would be an extremely artificial fine-tuning between the initial values of the mass splittings (and mixing angles) and the effect of the RG running, something clearly unacceptable.
Most of the paper is devoted to the third possible type of spectrum, $`m_1^2m_2^2m_3^2`$, which requires $`m_{1,2}^2\mathrm{\Delta }m_{atm}^2`$ (or larger). Here again, the radiatively generated splittings are in general too large, making the scenario unnatural. As a general rule, this gets worse as $`\mathrm{\Delta }m_{atm}^2`$ or $`\mathrm{\Lambda }`$ grow. Also, the supersymmetric scenario works worse than the SM one, especially as $`\mathrm{tan}\beta `$ incresases. More precisely, if $`m_1`$ and $`m_2`$ have equal signs, the RGE-induced splittings are always too large, even for the most favorable case. In addition, the solar mixing angle is driven by the RGEs to very small values, $`\mathrm{sin}^22\theta _3(M_Z)0`$, which is incompatible with the VO solution. It is however worth noticing that such a small angle is what is needed for a successful small-angle MSW solution to the solar neutrino problem. Moreover, for $`\mathrm{\Lambda }\stackrel{>}{_{}}10^{12}\mathrm{GeV}`$ and/or for the MSSM scenario the values of $`\mathrm{\Delta }m_{12}^2`$ may fall naturally in the small-angle MSW range, $`10^5\mathrm{eV}^2`$.
If $`m_1`$ and $`m_2`$ have opposite signs, the results are analogous, but now the splitting generated by the RGEs can vanish if the mixing angles are correlated in a particular way (which remarkably is always satisfied by the exact bimaximal case). This correlation or tuning of parameters is acceptable in the SM scenario, provided the cut-off scale $`\mathrm{\Lambda }`$ is not much larger than $`1\mathrm{TeV}`$ and if $`\mathrm{\Delta }m_{atm}^2`$ is in the low side of its experimentally preferred range ($`5\times 10^4\mathrm{eV}^2`$). Interestingly, this could provide a dynamical origin for the smallness of $`\mathrm{\Delta }m_{sol}^2`$. For larger $`\mathrm{\Lambda }`$ and/or $`\mathrm{\Delta }m_{atm}^2`$ (or equivalently, for the MSSM scenario) radiative corrections grow in size and the required tuning of mixing angles becomes quickly unacceptable. This occurs in particular if the lower bound $`\mathrm{\Delta }m_{atm}^2=5\times 10^4`$ that we have used is increased according to the most recent data analyses .
In conclusion, apart from the mentioned small windows, a completely hierarchical spectrum of neutrinos (i.e. as the spectrum of quarks and charged leptons), $`m_1^2m_2^2m_3^2`$, seems to be the only plausible one for the VO solution to the solar neutrino problem.
## Acknowledgements
This research was supported in part by the CICYT (contract AEN98-0816). A.I. and I.N. thank the CERN Theory Division for hospitality.
|
no-problem/9906/math9906142.html
|
ar5iv
|
text
|
# 1 Introduction
## 1 Introduction
Let $`M`$ be a complex manifold of complex dimension $`n`$ and $`\text{Aut}(M)`$ the group of its holomorphic automorphisms. The group $`\text{Aut}(M)`$ is a topological group with the natural compact-open topology. If $`M`$ is Kobayashi-hyperbolic (e.g., if $`M`$ is a bounded domain in complex space), then $`\text{Aut}(M)`$ is in fact a finite-dimensional real Lie group whose topology agrees with the compact-open topology \[Ko1\], \[Ko2\].
We are interested in the problem of characterizing hyperbolic complex manifolds by the dimensions of their automorphism groups. Let $`M`$ be such a manifold. It is known (see \[Ka\], \[Ko1\]) that $`\text{dim}\text{Aut}(M)n^2+2n`$ and, if $`M`$ is connected and $`\text{dim}\text{Aut}(M)=n^2+2n`$, then $`M`$ is holomorphically equivalent to the unit ball $`B^n\mathrm{}^n`$. In \[IK\] we obtained the following result.
###### THEOREM 1.1
Let $`M`$ be a connected hyperbolic manifold of complex dimension $`n2`$. Then the following holds:
(a) If $`\text{dim}\text{Aut}(M)n^2+3`$, then $`M`$ is biholomorphically equivalent to $`B^n`$.
(b) If $`\text{dim}\text{Aut}(M)=n^2+2`$, then $`M`$ is biholomorphically equivalent to $`B^{n1}\times \mathrm{\Delta }`$, where $`\mathrm{\Delta }`$ is the unit disc in $`\mathrm{}`$.
In this paper we completely characterize hyperbolic manifolds with automorphism groups of dimension $`n^2+1`$. In \[GIK\] we observed that the automorphism group of a hyperbolic Reinhardt domain in $`\mathrm{}^n`$ cannot have dimension $`n^2+1`$. Therefore, it has been our expectation that there should be very few manifolds with $`\text{dim}\text{Aut}(M)=n^2+1`$. In fact, we have known of only one example of a manifold with such an automorphism group dimension; this occurs in dimension $`n=3`$.
Example. Consider the 3-dimensional Siegel space (the symmetric bounded domain of type $`(III_2)`$):
$$S:=\{(z_1,z_2,z_3)\mathrm{}^3:EZ\overline{Z}\text{is a positive-definite matrix}\},$$
where
$$Z:=\left(\begin{array}{cc}z_1& z_2\\ z_2& z_3\end{array}\right),$$
and $`E`$ is the $`2\times 2`$ identity matrix. The automorphism group of this domain is isomorphic to $`Sp_4(\mathrm{})/\{\pm \text{Id}\}`$ and has dimension $`10=3^2+1`$ (see, e.g., \[S\]).
In this paper we prove the following theorem that shows that the Siegel space $`S`$ is indeed the only possibility.
###### THEOREM 1.2
Let $`M`$ be a connected hyperbolic manifold of complex dimension $`n2`$. Then the following holds:
(a) If $`n3`$, then $`\text{dim}\text{Aut}(M)n^2+1`$.
(b) If $`n=3`$ and $`\text{dim}\text{Aut}(M)=10`$, then $`M`$ is holomorphically equivalent to the Siegel space $`S`$.
Note that, for $`n=1`$, all manifolds with positive-dimensional automorphism groups are known explicitly \[FK\]. In particular, if $`\text{dim}\text{Aut}(M)=3`$, then $`M`$ is equivalent to $`\mathrm{\Delta }`$, and there does not exist a hyperbolic manifold $`M`$ with $`\text{dim}\text{Aut}(M)=2`$.
There is probably no hope to obtain a complete classification of all $`n`$-dimensional hyperbolic manifolds with automorphism groups of dimension $`n^2`$ or smaller. Indeed, in \[GIK\] we classified all hyperbolic Reinhardt domains in $`\mathrm{}^n`$ whose automorphism groups have dimension $`n^2`$, and even in this special case the classification is rather large and non-trivial.
For the sake of completeness of our exposition, and because some of the ideas originating there will recur in the proof of Theorem 1.2, we reproduce the proof of Theorem 1.1 in Section 2. We prove Theorem 1.2 in Section 3.
Before proceeding, we would like to thank W. Kaup for helpful discussions and for showing us useful references.
## 2 Proof of Theorem 1.1
For the proof of Part (a) of Theorem 1.1 we need the following lemma.
###### Lemma 2.1
Let $`G`$ be a Lie subgroup of the unitary group $`U(n)`$ and let $`G^c`$ be its connected component of the identity. Suppose that $`\text{dim}Gn^22n+3`$, $`n2`$, $`n4`$. Then either $`G=U(n)`$, or $`G^c=SU(n)`$. For $`n=4`$ this list must be augmented by subgroups of $`U(4)`$ whose Lie algebras are isomorphic to $`\mathrm{}sp_{2,0}`$.
Proof of Lemma 2.1: Since $`G`$ is compact, it is completely reducible (see, e.g., \[VO\]) and thus is isomorphic to a direct product $`G_1\times \mathrm{}\times G_k`$; here $`G_j`$ for each $`j`$ is a compact subgroup of $`U(n_j)`$, $`_{j=1}^kn_j=n`$, and $`G_j`$ acts irreducibly on $`\mathrm{}^{n_j}`$. Since $`\text{dim}G_jn_j^2`$ and $`\text{dim}Gn^22n+3`$, it follows that $`k=1`$, i.e. $`G`$ acts (complex) irreducibly on $`\mathrm{}^n`$.
Let $`gu_n`$ be the Lie algebra of $`G`$ and $`g^{\mathrm{}}gl_n`$ its complexification. It then follows that $`g^{\mathrm{}}`$ acts irreducibly and faithfully on $`\mathrm{}^n`$. Therefore by a theorem of É. Cartan (see, e.g., \[GG\]), $`g^{\mathrm{}}`$ is either semisimple or is the direct sum of a semisimple ideal $`h`$ and $`\mathrm{}`$, where $`\mathrm{}`$ acts on $`\mathrm{}^n`$ by multiplication. Clearly the action of the ideal $`h`$ on $`\mathrm{}^n`$ is irreducible and faithful.
Suppose first that $`g^{\mathrm{}}`$ is semisimple, and let $`g^{\mathrm{}}=g_1\mathrm{}g_m`$ be its decomposition into the direct sum of simple ideals. It then follows (see, e.g., \[GG\]) that the representation of $`g^{\mathrm{}}`$ is the tensor product of some irreducible faithful representations of $`g_j`$. Let $`n_j`$ denote the dimension of the representation of $`g_j`$, $`j=1,\mathrm{},m`$. Then $`n=n_1\mathrm{}n_m`$ and $`\text{dim}_{\mathrm{}}g_jn_j^21`$, $`n_j2`$ for $`j=1,\mathrm{},m`$. It is now not difficult to prove the following claim.
> Claim: If $`n=n_1\mathrm{}n_m`$, $`m2`$, $`n_j2`$ for $`j=1,\mathrm{},m`$, then $`_{j=1}^mn_j^2n^22n`$.
It follows from the claim that $`m=1`$, i.e., $`g^{\mathrm{}}`$ is simple. The minimal dimensions of irreducible faithful representations of complex simple Lie algebras are well-known (see, e.g., \[VO\]). In the table below $`V`$ denotes representations of minimal dimension.
| $`g`$ | $`\text{dim}V`$ | $`\text{dim}g`$ |
| --- | --- | --- |
| $`sl_k`$ $`k2`$ | $`k`$ | $`k^21`$ |
| $`o_k`$ $`k7`$ | $`k`$ | $`\frac{k(k1)}{2}`$ |
| $`sp_{2k}`$ $`k2`$ | $`2k`$ | $`2k^2+k`$ |
| $`e_6`$ | 27 | 78 |
| $`e_7`$ | 56 | 133 |
| $`e_8`$ | 248 | 248 |
| $`f_4`$ | 26 | 52 |
| $`g_2`$ | 7 | 14 |
Since $`\text{dim}_{\mathrm{}}g^{\mathrm{}}n^22n+3`$, it follows that $`g^{\mathrm{}}sl_n`$. Since $`g`$ is a compact algebra, we get that $`g=su_n`$ (see \[VO\]) and therefore $`G^c=SU(n)`$ (note here that if $`g`$ is a subalgebra in $`u_n`$ and $`g`$ is isomorphic to $`su_n`$, then $`g`$ coincides with $`su_n`$, i.e., it consists exactly of matrices with zero trace).
Suppose now (for the second case) that $`g^{\mathrm{}}=h\mathrm{}`$, where $`h`$ is a semisimple ideal in $`g^{\mathrm{}}`$. Then, repeating the above argument for $`h`$ and taking into account that $`\text{dim}hn^22n+2`$, we conclude that $`hsl_n`$ for $`n4`$. Therefore, for $`n4`$, $`g^{\mathrm{}}=gl_n`$ and hence $`g=u_n`$, which implies that $`G=U(n)`$.
For $`n=4`$, either $`hsl_4`$ or $`hsp_4`$. We thus find that either $`g=u_4`$ (in which case $`G=U(4)`$), or $`g\mathrm{}sp_{2,0}`$.
The lemma is proved. $`\mathrm{}`$
Proof of Theorem 1.1, Part (a): Let $`pM`$ and let $`I_p`$ denote the isotropy group of $`p`$ in $`\text{Aut}(M)`$. Since the complex dimension of $`M`$ is $`n`$, the real dimension of any orbit of the action of $`\text{Aut}(M)`$ on $`M`$ does not exceed $`2n`$, and therefore we have $`\text{dim}I_pn^22n+3`$.
Consider the isotropy representation $`\alpha _p:I_pGL(T_p(M),\mathrm{})`$:
$$\alpha _p(f):=df(p),fI_p.$$
The mapping $`\alpha _p`$ is a continuous group homomorphism (see, e.g., Lemma 1.1 of \[GK\]) and thus is a Lie group homomorphism (see \[Wa\]). Since $`I_p`$ is compact (see \[Ko1\]), there is a positive-definite Hermitian form $`h_p`$ on $`T_p(M)`$ such that $`\alpha _p(I_p)U_{h_p}(n)`$, where $`U_{h_p}(n)`$ is the group of complex linear transformations of $`T_p(M)`$ preserving the form $`h_p`$. We choose a basis in $`T_p(M)`$ such that $`h_p`$ in this basis is given by the identity matrix.
By \[E\] and \[Ki\], the mapping $`\alpha _p`$ is one-to-one. Further, since $`\text{dim}I_pn^22n+3`$, we see that $`\alpha (I_p)`$ is a compact subgroup of $`U_{h_p}(n)`$ of dimension at least $`n^22n+3`$. We are now going to use Lemma 2.1.
Assume first that $`n4`$. Then we have that either $`\alpha _p(I_p)=U_{h_p}(n)`$, or $`\alpha _p(I_p)^c=SU_{h_p}(n)`$ (the latter denotes the subgroup of $`U_{h_p}(n)`$ consisting of matrices with determinant 1). The groups $`U_{h_p}(n)`$ and $`SU_{h_p}(n)`$ act transitively on the unit sphere in $`T_p(M)`$ and thus act transitively on directions in $`T_p(M)`$ (see \[GK\] and \[BDK\] for terminology). Since $`M`$ is non-compact (because the dimension of $`\text{Aut}(M)`$ is positive—see \[Ko1\]), the main result of \[GK\] and its generalization in \[BDK\] applies. Thus $`M`$ is biholomorphically equivalent to $`B^n`$ (and therefore the possibility $`\alpha _p(I_p)^c=SU_{h_p}(n)`$ is in fact not realizable).
Suppose now that $`n=4`$. If we have that either $`\alpha _p(I_p)=U_{h_p}(4)`$, or else $`\alpha _p(I_p)^c=SU_{h_p}(4)`$ for some $`pM`$, then by the above argument $`M`$ is equivalent to $`B^4`$. Suppose now that the Lie algebra of $`\alpha _p(I_p)`$ is isomorphic to $`\mathrm{}sp_{2,0}`$ for every $`pM`$. Then $`\text{dim}\alpha _p(I_p)=11`$ for any $`p`$. Since $`\text{dim}\text{Aut}(M)19`$, we have that in fact $`\text{dim}\text{Aut}(M)=19`$, and thus $`M`$ is homogeneous. Therefore, by \[N\], \[P-S\], $`M`$ is biholomorphically equivalent to a Siegel domain $`D\mathrm{}^4`$ of the first or second kind. Further, we note that any representation $`\varphi :sp_{2,0}gl_4`$ is conjugate to the standard embedding of $`sp_{2,0}`$ into $`gl_4`$ by an element from $`GL(4,\mathrm{})`$ (to see this, one can extend $`\varphi `$ to a 4-dimensional representation of the complex Lie algebra $`sp_4`$ and notice that such a representation is unique up to conjugation by elements of $`GL(4,\mathrm{})`$). Therefore $`\varphi (sp_{2,0})`$ contains an element $`X`$ such that $`\mathrm{exp}(X)=\text{id}`$, and thus $`\alpha _p(I_p)`$ contains $`\text{id}`$ for any $`p`$. Hence the domain $`D`$ is in fact symmetric. It now follows from the explicit classification of symmetric Siegel domains (see \[S\]) that in fact there is no symmetric Siegel domain in $`\mathrm{}^4`$ with automorphism group of dimension equal to 19. This concludes the proof of Part (a).
For the proof of Part (b) we will need the following lemma (which follows from the proof of Lemma 2.1).
###### Lemma 2.2
Let $`U_h(n)`$ be the group of linear transformations of a complex $`n`$-dimensional space $`V`$ that preserve a positive-definite Hermitian form $`h`$ on $`V`$, and let $`G`$ be a Lie subgroup of $`U_h(n)`$ with $`\text{dim}Gn^22n+2`$, $`n2`$, $`n4`$. Then either $`G=U_h(n)`$, or $`G^c=SU_h(n)`$, or $`V`$ splits into a sum of 1- and $`(n1)`$-dimensional $`h`$-orthogonal complex subspaces $`V^1`$ and $`V^2`$ such that $`G=U_{h^1}(1)\times U_{h^2}(n1)`$, where $`h^j`$ is the restriction of $`h`$ to $`V^j`$. For $`n=4`$, there is the additional possibility that $`G`$ can be any subgroup of $`U_h(4)`$ with Lie algebra isomorphic to either $`sp_{2,0}`$ or $`\mathrm{}sp_{2,0}`$.
Proof of Theorem 1.1, Part (b): We will use the notation from the proof of Part (a) above. Let $`pM`$ and $`I_p`$ be the isotropy group of $`p`$ in $`\text{Aut}(M)`$. Then we have $`\text{dim}I_pn^22n+2`$ and thus $`\alpha _p(I_p)`$ is a subgroup of $`U_{h_p}`$ of dimension at least $`n^22n+2`$. We now use Lemma 2.2. If, for some $`pM`$, we have that either $`\alpha _p(I_p)=U_{h_p}(n)`$ or $`\alpha _p(I_p)^c=SU_{h_p}(n)`$, then $`\alpha _p(I_p)`$ acts transitively on directions in $`T_p(M)`$. Hence, as in the proof of Part (a), $`M`$ is biholomorphically equivalent to $`B^n`$; this is impossible since $`\text{dim}\text{Aut}(M)=n^2+2`$.
Further suppose that, for any point $`pM`$, $`T_p(M)`$ splits into the sum of 1- and $`(n1)`$-dimensional $`h_p`$-orthogonal complex subspaces $`V_p^1`$ and $`V_p^2`$ such that $`\alpha _p(I_p)=U_{h_p^1}(1)\times U_{h_p^2}(n1)`$. In particular, $`\text{dim}I_p=n^22n+2`$ for all $`pM`$ and therefore $`M`$ is homogeneous. Then, by \[N\], \[P-S\], $`M`$ is biholomorphically equivalent to a homogeneous Siegel domain $`D`$ of the first or second kind. Since $`\alpha _p(I_p)`$ contains the transformation $`\text{id}`$ for all $`pM`$, the domain $`D`$ is in fact symmetric. The theorem for $`n4`$ now follows from the explicit classification of symmetric Siegel domains (see \[S\]).
Suppose now that $`n=4`$ and that, for some point $`pM`$, the Lie algebra of $`\alpha _p(I_p)`$ is isomorphic to either $`sp_{2,0}`$ or to $`\mathrm{}sp_{2,0}`$. In the proof of Part (a) we noted that any embedding of $`sp_{2,0}`$ into $`gl_4`$ is conjugate by an element of $`GL(4,\mathrm{})`$ to the standard one. Therefore $`\alpha _p(I_p)`$ contains a subgroup that is conjugate by an element of $`GL(4,\mathrm{})`$ to $`Sp_{2,0}`$. Since $`Sp_{2,0}`$ acts transitively on the sphere of dimension 7, we get that $`\alpha _p(I_p)`$ acts transitively on directions in $`T_p(M)`$ and therefore, as in the proof of Part (a), $`M`$ is biholomorphically equivalent to the unit ball which is impossible.
The theorem is proved. $`\mathrm{}`$
Remark. The argument in the last paragraph in the proof of Part (b) could also be used in the proof of Part (a) for the case $`n=4`$ without reference to the classification theory of symmetric domains. For hyperbolic Reinhardt domains, Theorem 1.1 was obtained by a different argument in \[GIK\].
## 3 Proof of Theorem 1.2
We will use the notation from the proof of Theorem 1.1. Suppose first that $`n=2`$. Since $`\text{dim}\text{Aut}(M)=5`$, for any point $`pM`$ we have $`\text{dim}\alpha _p(I_p)1`$. It follows from Lemma 2.2 that, if $`G`$ is a positive-dimensional closed subgroup of $`U(2)`$, then one of the following holds:
(i) $`G=U(2)`$;
(ii) $`G^c=SU(2)`$;
(iii) $`\mathrm{}^2`$ splits into a sum of two 1-dimensional orthogonal subspaces $`V^1`$ and $`V^2`$, such that $`G=U_{h^1}(1)\times U_{h^2}(1)`$, where $`h^j`$ is the restriction of the standard Hermitian form on $`\mathrm{}^2`$ to $`V^j`$;
(iv) $`G`$ is 1-dimensional.
If, for some $`pM`$, $`\alpha _p(I_p)`$ is as in (i) or (ii), then $`M`$ is holomorphically equivalent to $`B^2\mathrm{}^2`$, which is impossible since $`\text{dim}\text{Aut}(M)=5`$.
If, for every $`pM`$, $`\alpha _p(I_p)`$ is as the group $`G`$ in (iii) (i.e., $`T_p(M)`$ splits into a sum of subspaces $`V^1`$ and $`V^2`$ as in (iii)), then, for every $`pM`$, $`\alpha _p(I_p)`$ contains the element $`\text{id}`$. Since $`M`$ is hyperbolic, it is a complex metric Banach manifold (when equipped with the Kobayashi metric). It now follows from the results of \[V\] (see Theorem 17.16 in \[U\]) that $`M`$ is homogeneous, which is again impossible.
If, for every $`pM`$, $`\alpha _p(I_p)`$ is 1-dimensional, then $`M`$ is homogeneous and hence is equivalent to a bounded homogeneous domain in $`\mathrm{}^2`$ \[N\]. Therefore, $`M`$ is equivalent to either $`B^2`$ or $`\mathrm{\Delta }^2`$. But such an equivalence is impossible since $`\text{dim}\text{Aut}(B^2)=8`$ and $`\text{dim}\text{Aut}(\mathrm{\Delta }^2)=6`$.
We now assume that $`M=M_1M_2`$, with $`M_1,M_2\mathrm{}`$, and, for $`pM_1`$, $`\alpha _p(I_p)`$ is as the group $`G`$ in (iii) and, for $`pM_2`$, $`\alpha _p(I_p)`$ is 1-dimensional. Fix $`p_0M_2`$ and let $`\mathrm{\Omega }`$ be the orbit of $`p_0`$ under $`\text{Aut}(M)^c`$. Then $`\mathrm{\Omega }`$ is a homogeneous subdomain of $`M`$. Therefore, $`\mathrm{\Omega }`$ is holomorphically equivalent to either $`B^2`$ or $`\mathrm{\Delta }^2`$. Consider the restriction map $`\varphi :\text{Aut}(M)^c\text{Aut}(\mathrm{\Omega })`$:
$$\varphi :ff|_\mathrm{\Omega }.$$
Clearly, $`\varphi `$ is continuous and hence a Lie group homomorphism. It is also one-to-one by the uniqueness theorem. Therefore $`\text{Aut}(\mathrm{\Omega })`$ contains a (not necessarily closed) subgroup $`H`$ of dimension 5 that acts transitively on $`\mathrm{\Omega }`$. Since $`M_1\mathrm{}`$, $`H`$ contains a subgroup isomorphic to $`U(1)\times U(1)`$. It now follows from the explicit formulas for the automorphism groups of $`B^2`$ and $`\mathrm{\Delta }^2`$ that such a subgroup $`H`$ in fact does not exist. This proves Part (a) for $`n=2`$.
Suppose now that $`n=3`$. If $`\text{dim}\text{Aut}(M)=10`$ then, for any $`pM`$, we have $`\text{dim}\alpha _p(I_p)4`$. It follows from Lemma 2.2 that, if $`G`$ is a closed subgroup of $`U(3)`$ of dimension at least 4, then one of the following holds:
(i) $`G=U(3)`$;
(ii) $`G^c=SU(3)`$;
(iii) $`\mathrm{}^3`$ splits into a sum of a 1- and 2-dimensional orthogonal subspaces $`V^1`$ and $`V^2`$ respectively such that $`G=U_{h^1}(1)\times U_{h^2}(2)`$, where $`h^j`$ is the restriction of the standard Hermitian form on $`\mathrm{}^3`$ to $`V^j`$;
(iv) $`G`$ is 4-dimensional.
If, for some $`pM`$, $`\alpha _p(I_p)`$ is as in (i) or (ii) then, as above, $`M`$ is holomorphically equivalent to $`B^3\mathrm{}^3`$, which is impossible since $`\text{dim}\text{Aut}(M)=10`$.
If, for every $`pM`$, $`\alpha _p(I_p)`$ is as the group $`G`$ in (iii), then, by \[V\], as in the case $`n=2`$ above, $`M`$ is homogeneous which is impossible.
If, for every $`pM`$, $`\alpha _p(I_p)`$ is 4-dimensional, then $`M`$ is homogeneous and hence is holomorphically equivalent to one of the following domains: $`B^3`$, $`B^2\times \mathrm{\Delta }`$, $`\mathrm{\Delta }^3`$, or the Siegel space $`S`$. Among these domains only $`S`$ has automorphism group of dimension 10.
We now assume that $`M=M_1M_2`$, with $`M_1,M_2\mathrm{}`$, and, for $`pM_1`$, we suppose that $`\alpha _p(I_p)`$ is as the group $`G`$ in (iii) and, for $`pM_2`$, $`\alpha _p(I_p)`$ is 4-dimensional. As in the case $`n=2`$ above, this implies that there exists a subdomain $`\mathrm{\Omega }M`$ such that $`\text{Aut}(\mathrm{\Omega })`$ contains a subgroup $`H`$ of dimension 10 that acts transitively on $`\mathrm{\Omega }`$ and that contains a subgroup isomorphic to $`U(1)\times U(2)`$. It now follows from the explicit formulas for the automorphism groups of $`B^3`$, $`B^2\times \mathrm{\Delta }`$ and $`S`$ that such a subgroup $`H`$ in fact does not exist. This proves Part (b).
Let now $`n4`$. Since $`\text{dim}\text{Aut}(M)=n^2+1`$, for any $`pM`$ we have $`\text{dim}\alpha _p(I_p)n^22n+1`$. It follows from Lemma 2.2 that, if $`G`$ is a closed subgroup of $`U(n)`$ of dimension at least $`n^22n+1`$, then one of the following holds:
(i) $`G=U(n)`$;
(ii) $`G^c=SU(n)`$;
(iii) $`\mathrm{}^n`$ splits into a sum of a 1- and $`(n1)`$-dimensional orthogonal subspaces $`V^1`$ and $`V^2`$ respectively such that $`G=U_{h^1}(1)\times U_{h^2}(n1)`$, where $`h^j`$ is the restriction of the standard Hermitian form on $`\mathrm{}^n`$ to $`V^j`$;
(iv) $`\text{dim}G=n^22n+1`$;
and, for $`n=4`$, there is one more possibility:
(v) The Lie algebra of $`G`$ is isomorphic to either $`sp_{2,0}`$ or $`\mathrm{}sp_{2,0}`$.
If, for some $`pM`$, $`\alpha _p(I_p)`$ is as in (i), (ii) or (v), then, as above, $`M`$ is holomorphically equivalent to $`B^n\mathrm{}^n`$, which is impossible since $`\text{dim}\text{Aut}(M)=n^2+1`$.
If, for every $`pM`$, $`\alpha _p(I_p)`$ is as the group $`G`$ in (iii), then by \[V\], $`M`$ is homogeneous which is impossible.
If, for every $`pM`$, $`\alpha _p(I_p)`$ is $`n^22n+1`$-dimensional, then $`M`$ is homogeneous and hence, by \[N\], \[P-S\], is equivalent to a Siegel domain of the first or second kind in $`\mathrm{}^n`$. We now need the following proposition which is of independent interest as it gives an alternative proof of Theorem 1.1 and a proof of Theorem 1.2 for Siegel domains in $`\mathrm{}^n`$, $`n4`$.
###### Proposition 3.1
Let $`U\mathrm{}^n`$, $`n4`$, be a Siegel domain of the first or second kind. Suppose that $`\text{dim}\text{Aut}(U)n^2+1`$. Then $`U`$ is holomorphically equivalent to either $`B^n`$ or $`B^{n1}\times \mathrm{\Delta }`$.
Proof of Proposition 3.1: The domain $`U`$ has the form
$$U=\{(z,w)\mathrm{}^{nk}\times \mathrm{}^k:\text{Im}wF(z,z)C\},$$
(3.1)
where $`1kn`$, $`C`$ is an open convex cone in $`\mathrm{}^k`$ not containing an entire line and $`F=(F_1,\mathrm{},F_k)`$ is a $`\mathrm{}^k`$-valued Hermitian form on $`\mathrm{}^{nk}\times \mathrm{}^{nk}`$ such that $`F(z,z)\overline{C}\{0\}`$ for all non-zero $`z\mathrm{}^{nk}`$.
We will first show that $`k2`$. It follows from \[KMO\] that
$$\text{dim}\text{Aut}(U)4n2k+\text{dim}g_0(U).$$
Here $`g_0(U)`$ is the Lie algebra of all vector fields on $`\mathrm{}^n`$ of the form
$$X_{A,B}=Az\frac{}{z}+Bw\frac{}{w},$$
where $`Agl_{nk}`$, $`B`$ belongs to the Lie algebra $`g(C)`$ of the group $`G(C)`$ of linear automorphisms of the cone $`C`$, and the following holds:
$$F(Az,z)+F(z,Az)=BF(z,z),$$
(3.2)
for $`z\mathrm{}^{nk}`$. By the definition of Siegel domain, there exists a positive-definite linear combination $`R`$ of the components of the Hermitian form $`F`$. Then, for a fixed matrix $`B`$ in formula (3.2), the matrix $`A`$ is determined at most up to a matrix that is Hermitian with respect to $`R`$. Since the dimension of the algebra of matrices Hermitian with respect to $`R`$ is equal to $`(nk)^2`$, we have
$$\text{dim}g_0(U)(nk)^2+\text{dim}g(C),$$
and thus the following holds
$$\text{dim}\text{Aut}(U)4n2k+(nk)^2+\text{dim}g(C).$$
(3.3)
###### Lemma 3.2
We have
$$\text{dim}g(C)\frac{k^2}{2}\frac{k}{2}+1.$$
Proof of Lemma 3.2: Fix a point $`x_0C`$ and consider its stabilizer $`G_{x_0}(C)G(C)`$. The stabilizer is compact since it leaves stable the bounded open set $`C(x_0C)`$ and therefore we can assume that it is contained in the group $`O(k,\mathrm{})`$. The group $`O(k,\mathrm{})`$ acts transitively on the sphere $`S(|x_0|)`$ of radius $`|x_0|`$ in $`\mathrm{}^k`$, and the stabilizer $`H_{x_0}`$ of the point $`x_0S(|x_0|)`$ under this action is isomorphic to $`O(k1,\mathrm{})`$. Since $`G_{x_0}H_{x_0}`$, we have
$$\text{dim}G_{x_0}\text{dim}H_{x_0}=\frac{k^2}{2}\frac{3k}{2}+1,$$
and therefore
$$\text{dim}g(C)\frac{k^2}{2}\frac{k}{2}+1.$$
The lemma is proved. $`\mathrm{}`$
It now follows from (3.3) and Lemma 3.2 that
$$\text{dim}\text{Aut}(U)\frac{3k^2}{2}k\left(2n+\frac{5}{2}\right)+n^2+4n+1.$$
(3.4)
It is easy to check that the right-hand side in (3.4) is strictly less than $`n^2+1`$ for $`n4`$ and $`k3`$. Therefore, $`k2`$.
If $`k=1`$, the domain $`U`$ is equivalent to $`B^n`$. Suppose that $`k=2`$. Without loss of generality we can assume that the first component $`F_1`$ of the $`\mathrm{}^2`$-valued Hermitian form $`F`$, is positive-definite. We will show that the second component $`F_2`$ has to be proportional to $`F_1`$. Indeed, if $`F_2`$ is not proportional to $`F_1`$, then in formula (3.2) the matrix $`A`$ is determined by the matrix $`B`$ up to transformations that are : (1) Hermitian with respect to the positive-definite form $`F_1`$; (2) Hermitian with respect to some other Hermitian form which is not proportional to $`F_1`$. The dimension of the algebra of matrices satisfying conditions (1) and (2) does not exceed $`(n2)^22`$, and therefore
$$\text{dim}g_0(U)(n2)^22+\text{dim}g(C).$$
Hence
$$\text{dim}\text{Aut}(U)n^22+\text{dim}g(C),$$
which together with Lemma 3.2 implies
$$\text{dim}\text{Aut}(U)n^2,$$
which is a contradiction. Thus, $`F_2`$ is proportional to $`F_1`$. Therefore, $`U`$ is holomorphically equivalent to one of the following domains:
$$U_1:=\{(z,w)\mathrm{}^{n2}\times \mathrm{}^2:\text{Im}w_1|z|^2>0,\text{Im}w_2>0\},$$
or
$$U_2:=\{(z,w)\mathrm{}^{n2}\times \mathrm{}^2:\text{Im}w_1|z|^2>0,\text{Im}w_2|z|^2>0\}.$$
The domain $`U_1`$ is equivalent to $`B^{n1}\times \mathrm{\Delta }`$. We will show that $`\text{dim}\text{Aut}(U_2)<n^2+1`$. Let $`g(U_2)`$ be the Lie algebra of $`\text{Aut}(U_2)`$. By \[KMO\], $`g(U_2)`$ is a graded Lie algebra:
$$g(U_2)=g_1(U_2)g_{1/2}(U_2)g_0(U_2)g_{1/2}(U_2)g_1(U_2),$$
where $`\text{dim}g_1(U_2)=2`$, $`\text{dim}g_{1/2}(U_2)=2(n2)`$, and $`g_0(U_2)`$ is described by (3.2). It is clear from (3.2) that $`\text{dim}g_0(U_2)=n^24n+5`$. Further, $`g_{1/2}(U_2)`$ and $`g_1(U_2)`$ admit explicit descriptions (see \[S\]). These descriptions imply that
$$g_{1/2}(U_2)=\{0\},g_1(U_2)=\{0\},$$
and therefore
$$dimg(U_2)=n^22n+3<n^2+1.$$
The proposition is proved. $`\mathrm{}`$
We will now finish the proof of Theorem 1.2. It follows from Proposition 3.1 that, if, for every $`pM`$, $`\alpha _p(I_p)`$ is $`n^22n+1`$-dimensional, then $`M`$ has to be equivalent to either $`B^n`$ or $`B^{n1}\times \mathrm{\Delta }`$ which is impossible since the dimensions of the automorphism groups of these domains are bigger than $`n^2+1`$.
We now assume that $`M=M_1M_2`$, with $`M_1,M_2\mathrm{}`$, and, for $`pM_1`$, we suppose that $`\alpha _p(I_p)`$ is as the group $`G`$ in (iii) and, for $`pM_2`$, $`\alpha _p(I_p)`$ is $`n^22n+1`$-dimensional. As in the cases $`n=2,3`$ above, this implies that there exists a subdomain $`\mathrm{\Omega }M`$ such that $`\text{Aut}(\mathrm{\Omega })`$ contains a subgroup $`H`$ of dimension $`n^2+1`$ that acts transitively on $`\mathrm{\Omega }`$ and that contains a subgroup isomorphic to $`U(1)\times U(n1)`$. It now follows from the explicit formulas for the automorphism groups of $`B^n`$ and $`B^{n1}\times \mathrm{\Delta }`$ that such a subgroup $`H`$ in fact does not exist. This proves Part (a) for $`n4`$.
The theorem is proved. $`\mathrm{}`$
Centre for Mathematics and Its Applications
The Australian National University
Canberra, ACT 0200
AUSTRALIA
E-mail address: Alexander.Isaev@anu.edu.au
Department of Mathematics, Campus Box 1146
Washington University in St. Louis
One Brookings Drive
St. Louis, Missouri 63130
USA
E-mail address: sk@math.wustl.edu
|
no-problem/9906/quant-ph9906123.html
|
ar5iv
|
text
|
# Disentangling Nonlocality and Teleportation
## 1 Introduction
Entangled states in quantum theory can be used both to demonstrate that quantum theory is not a local hidden variable theory and to teleport quantum states from one place to another . The fact that entangled states display nonlocality has led people to assume that teleportation, which requires the use of entangled states, is essentially a nonlocal phenomena. This may be true of quantum teleportation, but, as will be shown in this paper, it is not necessarily true of teleportation in general. To prove this a toy physical theory will be described which has the following properties: (a) It is local, (b) it does not allow cloning of states, and (c) it allows teleportation of states.
The question of whether teleportation is essentially nonlocal has been raised in two papers. Popescu showed that certain mixed states called Werner states which have been shown by Werner not to violate any Bell inequalities can nevertheless be used to teleport quantum states with a greater fidelity than could be achieved without any entanglement. More recently, Braunstein has shown how teleportation is possible with continuous variables. He points out that it is possible to consider only states having positive Wigner functions and that the measurements he considers can all be functions of $`q`$ and $`p`$. If we are restricted to such measurements on a state with a positive Wigner function then Bell has showed that it is possible to construct a local hidden variable model. However, it is possible to consider different states and different measurements, and then there will be nonlocality (see ).
In general, it is very difficult to differentiate between two physical concepts within the same physical theory. If that physical theory has both of the corresponding properties then these properties are likely to be manifest at the same time making it difficult to disentangle them. However, by considering alternative physical theories, we can hope to show that two physical concepts are distinct
It is the no-cloning theorem which makes quantum teleportation interesting. In standard classical physics it is possible in principle to make measurements on a system which establish a complete description of the state without disturbing the system. This complete description of the state can then be used to build another system in the same state. Hence, a possible teleportation scenario would be that such a complete measurement is made on a classical system, the resulting information is sent to another location, and a new copy is created. The original copy could be destroyed and this would seem like teleportation. However, there is no reason to destroy the original. On the other hand, if there is a no-cloning theorem which says it is impossible to make a copy of a system when the state of the system is unknown then there can be no dilemma about whether to destroy the original. It must have been destroyed if there is now a copy somewhere else. Hence we will take the existence of a no-cloning theorem to be part and parcel of what we mean by teleportation. Since classical physics as it stands does not have a no-cloning theorem, we will invent an alternative physical theory which does and which is explicitly local. This theory is only intended to illustrate the central point of this paper (it does not correspond to any real physical system).
A number of real optical experiments have been performed to demonstrate quantum teleportation . Since these are real experiments the detectors are less than 100% efficient. Taking advantage of this fact Risco-Delgado has shown that it is possible to build a local hidden variable model which agrees with the results the first two of these experiments (he has not considered the third). However, this model does not support perfect teleportation and it is not it clear whether a no-cloning theorem can be derived. Hence, they are not useful for our present purpose. Nevertheless, they played a role in motivating the approach taken here.
## 2 A toy local theory
We will now state the seven postulates of the toy local theory.
### 2.1 State description postulate
A system consists of particles and each particle can be in one of four states labeled $`0,1,2,3`$.
A system of $`N`$ particles has the state $`X=(x_1,x_2,\mathrm{},x_n,\mathrm{},x_N)`$ where $`x_n\{0,1,2,3\}`$ is the state of the $`n`$th particle. The number of possible states $`X`$ is $`4^N`$.
### 2.2 State measurement and preparation postulates
There exist measurement apparatusses which can be used to extract information about the state of a system (though, due to postulates 4 and 5 below, only partial information can be extracted). A given measurement, $`A`$, will have $`R`$ outcomes labeled $`r=0,1,2,\mathrm{},R1`$. Associated with the $`r`$th outcome is the set of possible states
$$A_r=\{X_{r1},X_{r2},\mathrm{},X_{rL}\}$$
A given $`X`$ does not appear in more than one such set. I.e. the sets $`A_r`$ are disjoint. Furthermore, each possible $`X`$ must appear in one of the sets $`A_r`$.
If the state of the system is $`X`$ then the outcome of a measurement of $`A`$ will be $`r`$ where $`XA_r`$. Since $`X`$ only appears in one of the sets $`A_r`$ there can only be one outcome for a given $`X`$.
After a measurement of $`A`$ having outcome $`r`$ the state of the system will be $`X_r`$ with probability $`\frac{1}{L}`$ (i.e. the state is selected randomly from the set $`A_r`$).
¿From postulates 3 and 4 we see that if a measurement is repeated on a system the outcome will be the same. Postulate 4 has the consequence that a measurement will, in general, disturb the state of a system. This limits the amount of information that can be extracted about the state of a system by repeated measurements. The set $`A_r`$ can be represented by the matrix
$$A_r=\left(\begin{array}{cccc}x_1^1& x_2^1& \mathrm{}& x_N^1\\ x_1^2& x_2^2& \mathrm{}& x_N^2\\ \mathrm{}& \mathrm{}& \mathrm{}& \mathrm{}\\ x_1^L& x_2^L& \mathrm{}& x_N^L\end{array}\right)$$
The entries in the $`l`$th row are the elements of $`X_{rl}`$ and the $`n`$th column pertains to the $`n`$th particle. For $`L=5`$ a typical column might be
$$\begin{array}{c}2\hfill \\ 0\hfill \\ 0\hfill \\ 2\hfill \\ 2\hfill \end{array}$$
(1)
The sets $`A_r`$ can be chosen in any way which is consistent with the constraints imposed in the postulate 2 and the following additional constraints. The sets $`A_r`$ must be such that (i) in each column there are at least two different values of $`x`$ and (ii) each value which occurs must occur in the column for at least 25% of the entries.
To illustrate postulate 5 we see that the column in (1) is ok because (i) there are at least two different values, in this case 0 and 2, and (ii) because each value which occurs does occur for at least $`\frac{1}{4}`$ of the entries, in this case the ratios are $`\frac{2}{5}`$ and $`\frac{3}{5}`$. Postulate 5, like postulate 4, limits the amount of information that can be extracted by measurement. If all the entries in a particular column were 0 then, if that outcome were recorded, we could be certain that the state of the corresponding particle is 0 both before and after the measurement. However, this is not possible. Postulate 5 implies that there must be at least a 25% chance that the state is something else.
### 2.3 State Manipulation postulates
It is possible to manipulate the particles which make up a system though only in accordance with the following rules.
The particle can be moved around in space (though at sub-luminal speeds). A joint measurement can only be made on two or more particles if they in the same place.
This postulate is necessary since we are constructing a local theory.
There exist apparatusses which make it possible to manipulate each individual particle without knowing its state in such a way that if the particle is in the state $`x`$ it will go to the state $`U(x)`$ where $`U`$ is a one to one function.
These manipulations are taken to be analogous to unitary transformations in quantum theory. We will be particularly interested in the operation $`U_k`$ where
$$U_k(x)=(x+k)mod4$$
(2)
We will call such operations rotations.
## 3 Examples of measurements
### 3.1 One particle
Consider making measurements on a single particle. In this case $`N=1`$ so there will only be one column in the matrix representation of the sets $`A_r`$. One possible measurement is defined by the matrices
$$A_0=\left(\begin{array}{c}0\\ 1\end{array}\right)A_1=\left(\begin{array}{c}2\\ 3\end{array}\right)$$
We see that both $`A_0`$ and $`A_1`$ are consistent with postulates 2 and 5. Imagine that Alice is given a single particle which has $`x=0`$ but that she does not know what its state is. She could perform the measurement described by these matrices. Since $`x=0`$ she will get outcome 0 corresponding to $`A_1`$ by postulate 3. She will now know that the original state was either 0 or 1 but she will not know which. After this measurement the new state of the particle could be either 0 or 1 (with even probabilities) by postulate 4 but Alice will not know which. She could repeat the measurement many times and each time she would get the same outcome. The state $`x`$, is in some sense, a hidden variable since a particle cannot be prepared with a known $`x`$. The preparable states of the system are combinations such as \[50% of 0 and 50% of 1\]. An alternative measurement is defined by the matrices
$$B_0=\left(\begin{array}{c}1\\ 2\end{array}\right)B_1=\left(\begin{array}{c}3\\ 0\end{array}\right)$$
If Alice now gives the particle to Bob and he measures $`B`$ there is a 50% chance he will get outcome 0 and a 50% chance he will get outcome 1. If he gets outcome 0 then since the state Alice gave him was known to be \[50% of 0 and 50% of 1\] they can conclude that the state of the particle Alice gave him was $`x=1`$. However, the state of the system after Bob’s measurement will be \[50% of 1 and 50% of 2\] corresponding to $`B_0`$. Furthermore they do not know whether $`x=1`$ was the state of the system that was originally given to Alice (in our example it was not).
### 3.2 Two particles
The state of two particles is represented by $`X=(x_1,x_2)`$. We will be interested in only one possible measurement that can be made on two particles. This is described by the matrices
$$B_0=\left(\begin{array}{cc}0& 0\\ 1& 1\\ 2& 2\\ 3& 3\end{array}\right)B_1=\left(\begin{array}{cc}0& 3\\ 1& 0\\ 2& 1\\ 3& 2\end{array}\right)$$
(3)
$$B_2=\left(\begin{array}{cc}0& 2\\ 1& 3\\ 2& 0\\ 3& 1\end{array}\right)B_3=\left(\begin{array}{cc}0& 1\\ 1& 2\\ 2& 3\\ 3& 0\end{array}\right)$$
(4)
These matrices satisfy the rules in postulates 2 and 5. This measurement on two particles will play the same role the Bell measurement plays in quantum teleportation. Furthermore, it can be used to prepare the analogue of the four Bell states. Imagine the two particles are initially in some unknown state. Now a $`B`$ measurement is performed. If the outcome is 0 corresponding to $`B_0`$ then the state will be given by one of the rows in the $`B_0`$ matrix. Looking at this matrix we see that the two particles will become correlated and their state will be $`(y,y)`$ where $`y`$ is unknown. Similar remarks apply to the other outcomes. For the outcome $`r`$ corresponding to $`B_r`$ the state will be $`(y,yrmod4)`$. Any of these states can be changed into any other by applying the appropriate rotation to one particle using the operation defined in (2).
## 4 A no-cloning theorem
As stressed earlier, a no-cloning theorem is essential to give meaning to teleportation. We will now prove that within our toy local theory it is not possible to produce clones of particles in unknown states. We will take an operational approach to cloning. Thus, imagine that Peter prepares a particle in some state by making a measurement on it. As a result of his measurement he might know that the state is something like \[50% of 0 and 50% of 1\]. He now gives this particle to Alice. Alice does not know what measurement Peter used to prepare the particles. Alice is challenged to give two particles back to Peter which are, so far as Peter can establish, in the same state as the original. One of these two particles is to be regarded as the original and the other as the clone. Peter will then repeat the same measurement he used to prepare the particles to test how well Alice has done. If he gets the same outcome as he did originally but now for both particles then Alice has passed the test. This test is repeated a large number of times. Alice must pass each time to convince Peter that she can produce clones. We will now prove that this is impossible – Alice must fail this test sometimes.
Assume Peter prepares the state by measuring $`P`$ where
$$P_0=\left(\begin{array}{c}0\\ 1\end{array}\right)P_1=\left(\begin{array}{c}2\\ 3\end{array}\right)$$
He could, of course, have chosen any other measurement. Assume further that he gets outcome 0. This means that the state is, from his point of view, \[50% of 0 and 50% of 1\]. Lets assume that actually the state of the particle is 1 (though Peter will not know this). Peter passes the particle on to Alice. Alice receives a particle in state 1 (though she does not know its state is 1) but no information about how it was prepared. Imagine now that on another occasion that Peter prepares the state by measuring $`P^{}`$ where
$$P_0^{}=\left(\begin{array}{c}0\\ 3\end{array}\right)P_1^{}=\left(\begin{array}{c}1\\ 2\end{array}\right)$$
and he gets outcome 1. This means that the prepared state, from Peter’s point of view, is \[50% of 1 and 50% of 2\]. We can assume that the state of the particle is actually 1 (again Peter will not know this). Peter passes the particle on to Alice. Again, Alice receives a particle in state 1 (though she does not know this) and no information about how it was prepared. From Alice’s point of view these two cases are identical. The only way she can hope to pass the test in all such cases is to prepare two particles in the state 1.
To do this let us imagine that she has $`N1`$ particles. When she receives the particle from Peter she has $`N`$ particles. She will perform some operations on the particle and then give two particles to Peter who will perform his test. There are only two types of operation she can perform on the particles. She can perform measurements on all the particles as described in postulates 2–5 and she can perform manipulations on individual particles as described by postulate 7. It is clear that the manipulations described in postulate 7 cannot help since they only act on individual particles and effectively do nothing more than relabel the states. Hence, assume Alice makes a measurement $`C`$. This measurement will have a certain number of outcomes. In the particular case we are discussing we want to produce two particles in the state 1. Hence, for at least one outcome, $`r_1`$ say, the matrix $`C_{r_1}`$ must have at least two 1’s in at least one row (say the first row) and the positions of these 1’s in this row must be known by Alice (since she must know which particles to give to Peter). Assume that the 1’s are in the first two columns (so that, in the case of obtaining this outcome, Alice will give Peter particles 1 and 2). Then the matrix will have the form
$$C_{r_1}=\left(\begin{array}{ccc}1& 1& \mathrm{}\\ u& v& \mathrm{}\\ \mathrm{}& \mathrm{}& \mathrm{}\end{array}\right)$$
(5)
In the first column at least 25% of the entries must be different to 1 by postulate 5. Assume that $`u`$ is one such entry ($`u1`$). If the outcome is $`r_1`$ then after the measurement the state will randomly selected from one of the rows in $`A_{r_1}`$. However, at least 25% of these are of the form $`(u,v,\mathrm{})`$ where $`u1`$. Hence, when Alice gives the first two particle to Peter, they will sometimes fail the test because they will not always be in the state $`(1,1)`$. This proves that Alice cannot successfully clone 100% of the time.
## 5 Teleportation in the toy local theory
To perform teleportation two particles, 2 and 3, are prepared in a Bell-like state as described in section (3.2). This state is of the form $`(y,y)`$ where $`y`$ is unknown. The two particles separate from the place of preparation. Particle 2 goes towards Alice, and particle 3 goes towards Bob. Alice is given a particle in an unknown state $`x`$. She takes this particle which we will call particle 1 together with particle 2 and performs the Bell-like measurement $`B`$ defined in (3,4). She will get some outcome $`r`$ for this measurement. If the outcome is $`r=0`$ we see from the matrix $`B_0`$ in equation (3) that $`x`$ for particle 1 is equal to $`y`$ for particle 2. In general, if she gets outcome $`r`$ then we see from the matrices $`B_r`$ that $`y=xrmod4`$. Particle 3 is in the state $`y`$. Alice sends the information $`r`$ to Bob and he can perform the manipulation $`U_r`$ defined in (2) on particle 3. After this its state will be $`x`$. This means that, without knowing the state of particle 1, we have successfully transferred its state onto particle 3. This is teleportation. The state of particle 1 is completey randomised after the Bell-like measurement and so the analogue with quantum teleportation is very strong.
## 6 Discussion
We have shown that it is possible to have a local theory which has a no-cloning theorem and supports teleportation. The particular model here provides some insight into what is essential for teleportation. What happens in this model is the following. The state of particle 1 is probed at a ‘microscopic’ level by particle 2 in a way that no ‘macroscopic’ measuring apparatus could probe it. It is this that makes it possible to extract full information about the state of particle 1 without running into the no-cloning theorem. This information is divided between the measurement result $`r`$ which Alice and Bob can know and the state of particle 3 which they cannot know and which was originally classically correlated with the ‘probe’ particle 2. The original state can then be reconstructed in an analogous way to the way it is reconstructed in quantum teleportation. Nonlocality plays no role in this process.
Given this model, we must not assume that nonlocality is playing an essencial role in quantum teleportation, though it may be the case that it is. There are two reasons for believing that nonlocality may play an important role in quantum teleportation: (1) As has been pointed out by Bennett , the amount of information needed to specify a general qubit is much greater than the two bits of information which is classically communicated during quantum teleportation. One might speculate that, when a qubit is teleported, the extra information is being carried by the nonlocal properties of the entangled state. On the other hand, it is not possible to extract more classical information from a qubit than the two classical bits communicated during teleportation and so there must remain questions about the reality of the quantum information apparently transmitted during teleportation. (2) If one qubit of an entangled pair is teleported it is possible to obtain a violation of Bell’s inequalities between the second qubit of this pair and the teleported qubit. Hence, the teleportation machine must be able to convey what ever quality is necessary for this nonlocal correlation. However, we cannot necessarily assert on the basis of this fact that nonlocality plays a role in quantum teleportation. It is possible that the extra information which establishes the nonlocal correlations is only transmitted in the process of measuring the quantities in Bell’s inequalities, and is not transmitted in the teleportation process.
Even if it is eventually concluded that quantum teleportation is a nonlocal process, examples like that of Popescu suggest that there may be a regime in which imperfect teleportation is happening, but no nonlocality is in involved. In such a regime teleportation would still circumvent the no-cloning theorem by probing the system to be teleported at a microscopic level in much the same way as the toy local theory considered in this paper. The Werner states considered by Popescu have since been shown to be nonlocal if more complicated measurements are considered than those originally considered by Werner (Popescu ). If two-particle quantum states were found which are local and yet useful in teleportation then we could conclude that the type of process discussed in this paper also happens in quantum theory. States having bound entanglement are obvious candidates to consider. However, while it seems likely, it has not yet been shown that such states will always satisfy Bell type inequalities. Furthermore, it is not known whether such states could be useful for teleportation although results derived so far are negative .
Whatever conclusions may eventually be drawn about nonlocality in quantum teleportation, it has been established in this paper that teleportation in theories which have a no-cloning theorem can be an entirely local phenomenon.
|
no-problem/9906/astro-ph9906054.html
|
ar5iv
|
text
|
# 1 Introduction:
## 1 Introduction:
Several sources, Active Galactic Nuclei (AGNs) and Super Nova Remnants (SNRs) are identified as TeV gamma-ray sources so far by the ground based Cherenkov telescopes. TeV gamma-ray observations of AGNs give us an opportunity to investigate the environment near the central massive black hole in the AGN and the particle acceleration in the jet. In addition, detail study of the energy spectrum of the TeV gamma rays from different distance sources makes it possible to measure the density of inter-galactic infrared photons. With these reasons, the study of TeV gamma-ray emission from AGNs is very important. The nearest X-ray selected BL Lacs, Mrk421 (z=0.031) and Mrk501(z=0.034) are identified as TeV gamma-ray sources.
Using the Utah Seven Telescope Array, we have observed Mrk421 and Mrk501 intensively since early 1997. We have started the survey observation of XBL Lacs since the autumn of 1997 just after the detection of flares of TeV gamma-ray emission from Mrk501 in 1997. Survey observations were carried out more than 700 hrs so far. Some results from the survey observation in 1997, 1998 and 1999 are reported in this paper.
## 2 Experiment:
Table 1 shows observation time under good weather conditions for each sources. We have selected nearby XBL Lacs with the redshift less than 0.2 in the northern hemisphere (Stecker et al. 1996). Observation targets list of the day was made every night by choosing the objects which pass thorough near the zenith. According to the target list, all telescopes track the same target. We adopted the tracking method called as “Raster Scan”. In this mode, the tracking center of the telescopes scans the square region of $`\pm 0.5^{}`$ in right ascension and declination coordinate centered on the target with a cycle of 48minutes. After the observation with one raster scan cycle, the data is analyzed immediately. If we find unusual event excess at the source position, the observation is continued for another raster scan cycle. If there is no excess, all telescopes begin to track next target as in schedule. These data are transfered to Japan through the Internet for detail analysis.
Optimization of the cut values for image parameters was carried out using the data of Mrk501 obtained in 1997. Threshold energy and mode energy of the detector for gamma rays are estimated as 600 GeV and 900GeV, respectively. By using the optimized cut values, we estimate by the Monte Carlo simulation that 40% of gamma rays are picked up and 99.8% of the background cosmic rays are rejected.
## 3 Results:
We have analyzed about 700 hours observation data of AGNs under the good weather condition(corresponds to 2000 hrs $`\times `$ telescopes exposure). We have detected clear signals from Mrk501 and Mrk421. Possible detection of gamma-ray emissions from 1ES1959 is reported in OG.2.1.21. We could not see clear signals from other sources except for these three, however, detail analysis is still going on. In this report, the results of Mrk501 and Mrk421 will be presented.
Figure 1 shows time variation of gamma-ray intensity from Mrk501 in 1997 and 1998. The observation time was 360 hrs $`\times `$ tels and 215 hrs $`\times `$ tels, respectively. Mrk501 was very active in 1997 and the intensity of the gamma rays varied with about 24 days periodicity (see OG.2.1.17). The activity of Mrk501 became low in Sep. 1997. From the observation in 1998, we could not see any clear flares, however, most of data points deviate in the positive direction as shown in figure 1. By adding all data observed in 1998, we obtained 6 sigma excess. Two dimensional excess map from the Mrk501 for 1998 data is shown in Figure 3. We can conclude that the activity of Mrk501 became lower in Sep. 1997, however, the gamma-ray emission still continued until Sep. 1998. The Mrk501 should be monitored continuously to investigate the cause of the variability of the VHE gamma-ray emission.
The observation of Mrk421 were carried out for 25 nights from April to May in 1998, under good weather condition. Total observation period was 56 hrs. Among these observations, the most clear excess of 5.3 sigma was found in 1.3 hrs observation on 30th of April 1998(2450933 MJD). $`\alpha `$ distribution of the data in this flare is shown in Figure 3.
As reported in the previous paper (Hayashida et al 1998), the energy spectrum of Mrk501 in 1997 showed interesting feature, the spectral index became steeper around several TeV. If we assume that this lack of higher energy gamma rays is due to the interaction with the inter-galactic infrared photons, it is expected that the similar shape appears for the spectrum of Mrk501(98) and Mrk421 because of the similar redshifts of Mrk501 and Mrk421. In Figure 4, integral spectra of Mrk421(30/Apr/98), Mrk501(97) and Mrk501(98) are compared. The indices of the integral spectrum between 900 GeV and 3 TeV are -1.54 for Mrk501(97), -1.83 for Mrk501(98) and -1.81 for Mrk421(30/Apr/98). Following this figure, only the spectrum for Mrk501(97) is harder at low energy and softer at high energy than those for Mrk421 and Mrk501(98).
Acknowledgments
This work is supported in part by the Grants-in-Aid for Scientific Research (Grants #0724102 and #08041096) from the Ministry of Education, Science and Culture. The authors would like to thank the people at Dugway for the help of observations.
References
Stecker, F.W., De Jager, O.C.,& Salamon, M.H., 1996, ApJ 473, L75
Hayashida, N. et al, 1998, ApJ 504, L71
Nishikawa, D. et al, 1999,Proc.26<sup>th</sup> ICRC(Salt Lake city,1999)OG.2.1.17
|
no-problem/9906/cond-mat9906122.html
|
ar5iv
|
text
|
# THERMALLY ASSISTED TUNNELING: AN ALTERNATIVE MODEL FOR THE THERMOLUMINESCENCE PROCESS IN CALCITE
## I Introduction
Calcite is a mineral found in many geological formations. Its thermoluminescence (TL) has been studied, mainly due to its application in geological and fossil dating. Calcite can also be used for ionizing and ultraviolet radiation dosimetry.
Investigations on calcite have shown different TL characteristics depending on the impurity content and on the genesis of the sample. It was also observed on artificially grown CaCO<sub>3</sub> crystals that the type and concentration of impurities modify both TL sensitivity and glow curve shape.
Important contributions have been made to the understanding of the TL mechanism of calcite since the pioneering works of Kolbe and Smakula and Medlin in the sixties. Studying the influence of manganese concentration, and by comparison of the emission spectrum of calcite with the emission of the others lattices including manganese, Medlin found that the emission band centered in 615nm, corresponding to a transition from the first excited to the fundamental state of Mn<sup>2+</sup> ion occupying a Ca<sup>2+</sup> site. Recent works showed that the shape of the emission spectrum changes from a broad band without structure to a well-resolved narrow band when the temperature of the TL peak increases and when the concentration of the Mn decrease. They also concluded that the orange emission is related to Mn impurity. Calderon at al suggested that the holes are trapped by the impurities and electrons are trapped at dislocations forming (CO<sub>3</sub>)<sup>3-</sup> . The irradiation with UV light was found to modify the intensity of the TL peaks suggesting that the trapped carriers can be phototransfered from deep to shallow traps . Concerning the kinetics of the detrapping process, it was found that the isothermal decay follow a t<sup>-1</sup> law which could not be accounted for a single trap model. Many workers explained this results assuming that the process are due to a distribution in the activation energies. On the other hand, Visocekas et al suggested that, for the peaks observed on their samples below room temperature, this behavior could be due to a tunnelling process.
In the present work we investigate the effect of gamma and ultraviolet radiations on the thermoluminescence of Brazilian calcite from Miranda, MS. Based on a correlation between TL and electron spin resonance (ESR) and on the photoinduced TL decay, we suggest an explanation for the appearance of the three peaks between room temperature and 450<sup>o</sup>C, and propose an alternative model for the thermoluminescence process.
## II Experimental
Brazilian calcite from Miranda, MS used in this work was powdered and sieved and the grains with diameters between 74$`\mu `$m and 174$`\mu `$m were selected. Chemical analysis by atomic absorption and X-ray fluorescent spectroscopy techniques detected the following impurities: Si($`<`$0.1%), Al($`<`$0.1%), Fe($`<`$0.05%), Ti($`<`$0.05%), Mg($`<`$0.23%), Na($`<`$0.03%) and K($`<`$0.05%). Electron spin resonance measurements revealed the presence of manganese. Samples were exposed to <sup>60</sup>Co gamma rays from 50Gy to 500Gy. Ultraviolet irradiation was performed at a distance of 0.5 m from a high pressure 400W mercury lamp. A home made TL reader was used to heat 4 mg powder samples from 30<sup>o</sup>C to 400<sup>o</sup>C in a nitrogen atmosphere, at a heating rate between 1<sup>o</sup>C/s to 3<sup>o</sup>C/s. We performed a series of monochromatic TL measurements with wavelength varying from 450 to 750nm in steps of 10nm. Data were stored in a microcomputer and the 3D isometric plots were constructed via software after correcting each of monochromatic curves to the wavelength response of the system.
ESR experiments were performed with a BRUKER ER 200D equipment, also on powdered samples at room temperature, in the X-band and 1.0 mW microwave power (below saturation). The peak to peak height of the first derivative of the signal was considered to be proportional to the number of paramagnetic centres.
## III Results
The glow curve of untreated sample (without any treatment in laboratory) consists of two peaks at 245 and 320<sup>o</sup>C when heated at a rate of 1.8<sup>o</sup>C/s. After a laboratory dose of gamma rays, the intensity of these peaks increases and another peak, around 150<sup>o</sup>C, appears. Hence for simplicity these peaks are labeled as peaks A, B, and C as their temperatures increase.
Figure 1 shows the isometric plot and the contour plot of the TL spectra of the irradiated samples. The three main peaks present a broad emission band centered at 615 nm. Besides this band, peaks A and B present a very weak emission band at 480nm, not usually reported. Comparing the TL spectra around 615nm at the peak temperatures we can see that peak C displays a fine-structure with at least 4 emission lines in this region while peak A are mainly composed by the main emission at 615nm. These results agree with previous results. Treated samples (with a heat treatment at 400<sup>o</sup>C for 1 hour followed by a gamma irradiation) show a similar glow curve, without changes in the peak positions.
The growth and the saturation of the peak heights with the exposure to gamma rays are shown in figure 2. The effect of post-irradiated thermal treatment at different temperatures shows that the decrease of TL intensity is accompanied by the dislocation of the TL peak maxima to higher temperatures indicating that the TL processes can not be accounted for a first order kinetic. Figure 3 shows the effect of 10 min post-irradiation annealing at different temperatures. We can see that as the temperature of post-irradiation thermal treatment increases, the intensity of the TL peaks decrease. Peak A vanish first followed by peak B and finally by peak C.
The effect of ultraviolet (UV) irradiation on the thermoluminescent properties was studied using both untreated (without any treatment in the laboratory) and treated samples (with a heat treatment at 400<sup>o</sup>C for 1 hour followed by gamma irradiation). When treated samples are irradiated with UV all the three peaks decrease and the peak maxima shift to higher temperatures (Figure 4), a similar shifts is produced by thermal bleaching. To check if any isothermal reduction of the peak was happening during the UV illumination, the temperature of the samples was monitored. We found that during the UV illumination the temperature is around 30<sup>o</sup>C. Measurements of the isothermal decay of the TL peaks revealed that, considering the time intervals used in these experiments, the isothermal decay at this temperature can be neglected. Figure 5 shows the variation of the inverse of TL intensity as a function of the illumination time ”t ” where we can see that TL intensity of the three peaks are proportional to t<sup>-1</sup>.
In figure 6 we show the result of the intensities of the ESR signals as a function of the radiation dose. The ESR measurements were performed at room temperature and it is possible to see that increasing the gamma ray doses, the well known signals of the Mn<sup>2+</sup> ion present in the samples decrease. At the same time three other signals labelled 2, 3 and 5, with g<sub>2</sub>=2.0036, g<sub>3</sub>=2.0032, and g<sub>5</sub>=2.0006 typical of free radicals, increase. The effect of the post-irradiation thermal treatment at different temperatures during 10 minutes on the labeled ESR signals are shown in figure 7. The increase of temperature of post irradiation heat treatments causes the increase of the Mn<sup>2+</sup> signal and decrease firstly of the signal 3, at the 75-150<sup>o</sup>C range, followed by the signal 2, at 200-350<sup>o</sup>C, and finally the signal 5, starting at 280<sup>o</sup>C (Figure 7). Thermal treatment at 400<sup>o</sup>C during 1 hour eliminate all the 2, 3 and 5 signals produced by the irradiation.
## IV Discussion
At present, the plain understanding of thermoluminescent emission requires the identification of the luminescence and the trapping centers and the indication of the charge carrier transference and recombination processes. The results presented on this work allow us to confirm the explanation presented by other researches and point out an alternative model to the charge carrier transference.
From figure 1 we can see that the three peaks have the same emission band centered on 615nm. This result combined with the result that the ESR signal related to Mn<sup>2+</sup> decrease with the irradiation dose and is recovered if the sample is heated, (see figures 6 and 7), clearly establish that the Mn is the hole trap formed upon irradiating the crystal and that during the heating it is both the recombination and the luminescence center. These results agree with the results found by other authors that pointed out that the emission band centered in 615nm is due to the <sup>4</sup>G $``$ <sup>6</sup>S transition of the Mn<sup>2+</sup> ion.
ESR signals 3 (g<sub>3</sub> = 2.0036), 2 (g<sub>2</sub> = 2.0032) have been associated to the (CO<sub>3</sub>)<sup>3-</sup> while signal 5 (g<sub>5</sub> = 2.0006) is associated with (CO<sub>2</sub>)<sup>-</sup>. In our experiments signals 3 and 2 bleach at different temperature ranges suggesting two different types of (CO<sub>3</sub>)<sup>3-</sup> on the crystal lattice. The decay of the free radical ESR signals with increasing heat treatment temperature points out to an unambiguous correlation between the A, B and C TL peaks and 3, 2 and 5 ESR signals. From this result, combined with the growth of these signal as a function of the radiation dose described in figure 6, we can conclude that the (CO<sub>3</sub>)<sup>3-</sup> and the (CO<sub>2</sub>)<sup>-</sup> are the electron trapping centers.
The results presented by Medlin showed that the isothermal decay of the phosphorescence of calcite crystals follows a t<sup>-1</sup> law. Medlin and Pagonis et al attributed this behavior to a recombination processes with a distribution in the energies of the electron traps.
The T<sub>m</sub>xT<sub>stop</sub> method, initially proposed by McKeever as a tool for distinguishing among different kinetics and mechanisms, was performed in natural calcite by Lima et al. Their results clearly indicated that the distribution in activation energies can not account for the TL peaks in calcite.
In the present work UV light was used to release the trapped electrons and generates the electron-hole recombination process (figures 4 and 5). The results obtained from these measurements show that the decay of the gamma-ray induced TL peaks due to exposure to UV light is inversely proportional to the illumination time, as we can see from figure 5.
The detrapping rate, in this experiment, is proportional to the intensity of the UV light but the mechanism of the charge recombination and the light emission should be the same if we employed thermal activation. In the former, a distribution of traps can not explain the t<sup>-1</sup> law also observed.
Taking into account all these results we are proposing that the charge transfer mechanism is processed via a thermally assisted tunnelling from the carbonate groups, as described earlier, and the Mn<sup>3+</sup>. This process can explain the decay of the TL peak intensity with the UV illumination time, found in the present work, the isothermal decay of the phosphorescence in calcite, found by Medlin and Pagonis et al., and the T<sub>m</sub>xT<sub>stop</sub> results, found by Lima et al. This proposal is also in agreement with the observation that the TL peaks in calcite are not accompanied by any thermally stimulated conductivity peaks, as it was showed by Medlin, and that the emission of light during the recombination process is generated in a ”localized-type ” mechanism.
If the recombination of charges between the electron and the hole traps are via a tunnelling process, both centers should be close enough to enable the tunnelling process to take place. Hence, we suggest that during the irradiation of the samples the (CO<sub>3</sub>)<sup>3-</sup> and (CO<sub>2</sub>)<sup>-</sup> centers are formed in the vicinity of a Mn<sup>2+</sup> substituting for a Ca<sup>2+</sup> of the calcite matrix. The two different (CO<sub>3</sub>)<sup>3-</sup> observed in the ESR measurements can be attributed to carbonate groups at nearest-neighbor and at next-nearest-neighbor positions of the Mn dopant.
## V Conclusion
In conclusion, the results observed in the present work clearly establish the role of the (CO<sub>3</sub>)<sup>3-</sup> and (CO<sub>2</sub>)<sup>-</sup> ions as the main trapping centres and the role of the Mn<sup>2+</sup> as the recombination and the luminescence centre in the TL processes in calcite. The observed TL peaks can be explained as follows: during irradiation, Mn<sup>2+</sup> ions change to Mn<sup>3+</sup>, losing one electron that is captured in a (CO<sub>3</sub>)<sup>2-</sup>, forming the (CO<sub>3</sub>)<sup>3-</sup> ion or the (CO<sub>2</sub>)<sup>-</sup> ion. The (CO<sub>3</sub>)<sup>3-</sup> can be formed either in a nn or in a nnn position to the Mn<sup>3+</sup> dopant generating two different (CO<sub>3</sub>)<sup>3-</sup> ions. Upon heating, the electron recombines with the hole left in the Mn<sup>3+</sup>, after a thermally assisted tunnelling from a trap close to the manganese. The resulting Mn<sup>2+</sup> is formed in an excited state, whose decay is responsible for the emitted light during the TL process.
|
no-problem/9906/gr-qc9906085.html
|
ar5iv
|
text
|
# 1 Introduction
## 1 Introduction
Dilatonic Cosmological models have been recently investigated in great detail in Cosmology and Supergravity .Earlier the role of spin-torsion as a source of repulsive gravity have been investigated in detail by Gasperini and Cosimo Stornaiolo in his PhD.thesis .More recently we have investigate the role of torsion on inflating defects .Also recently Maroto and Shapiro have investigated a cosmological model of de Sitter type with dilatons and torsion in Higher-order gravity .In this letter we solve the Einstein-Cartan dilaton cosmology equations of gravity for the case of a de Sitter model where the equation of state given a priori.The first is a massless dilatonic case.The three others deal with massive dilatons in three distinct cases.The first and second is the radiation era case for torsion kinks and massive potential.The last one handles with the dust of spinning particles with Cartan torsion .The temperature fluctuation is computed by using the nearly flat spectrum produced during inflation by gravitational waves and the solution obtained in our radiation era model that shows that the dilaton mass equals the Hubble constant.The resul is compared with COBE data.The spin-torsion fluctuation is also computed in a similar way.
## 2 Spin-torsion effects in Dilatonic Inflationary Cosmology
Let us now consider the cosmological metric for the spatially flat section to be given by
$$ds^2=dt^2a^2(dx^2+dy^2+dz^2)$$
(1)
The Einstein-Cartan equations are given by
$$H^2=\frac{8\pi G}{3}(\rho _{eff}2\pi G\sigma ^2)$$
(2)
and
$$\dot{H}+H^2=\frac{4\pi G}{3}(\rho _{eff}+3p_{eff}8\pi G\sigma ^2)$$
(3)
These two equations combined produced the following conservation equation
$$(\dot{\rho _{eff}}2\pi G(\dot{\sigma ^2}))+3H(\rho _{eff}+p_{eff}4\pi G\sigma ^2)=0$$
(4)
here $`H=\frac{\dot{a}}{a}`$ ,and $`\sigma ^2=<S_{ijk}S^{ijk}>`$ where $`S_{ijk}`$,and $`(i,j=0,1,2,3)`$,represents the spin tensor and the spins are considered ramdomly oriented since at the very early Universe due to the very high temperatures and broken symmetries is not possible to have polarized matter and $`\rho _{eff}`$,and $`p_{eff}`$ are given respectively by
$$\rho _{eff}=\dot{\varphi }^2+V(\varphi )$$
(5)
and
$$p_{eff}=\dot{\varphi }^2V(\varphi )$$
(6)
we notice that by imposing the ”stiff” matter equation of state $`p_{eff}=\rho _{eff}`$ where $`\varphi `$ is the dilaton potential and $`V`$ is the dilaton energy potential,implies that $`V(\varphi )`$ vanishes, which represents massless dilatons.Now let us equate equations (3) and (4),this yields
$$\dot{\varphi }^2=\pi G\sigma ^2$$
(7)
Use of the conservation equation yields
$$\dot{\varphi }^2=2\pi G\sigma ^2+e^{6Ht}$$
(8)
Comparison of these two last equations yields an expression for the spin-torsion density in terms of time
$$\sigma ^2=\frac{1}{\pi G}e^{6Ht}$$
(9)
Substitution of (9) into the expression for the Hawking-Penrose convergence condition
$$R_{ik}u^iu^k=3(H^2)=16\pi G(\dot{\varphi }^22\pi G\sigma ^2)=16\pi G\sigma ^2<0$$
(10)
shows that it is violated and a repulsive gravity is obtained.Cosmological perturbation densities can then be computed and the spin-torsion density determined.The computation without dilatons have been recently addressed by D.Palle .In the next section we perform these computations for the cases of radiation and hadron eras.
## 3 De Sitter Inflation in the Radiation Era and spin-torsion density primordial fluctuations
In this section we shall consider the equation of state for the radiation era $`p_{eff}=\frac{1}{3}\rho _{eff}`$.This condition yields
$$\dot{\varphi }^2+2V(\varphi )=0$$
(11)
This equation is interesting because can be solved for different potentials like torsion kinks and for the mass potential $`V(\varphi )=\frac{m^2}{2}\varphi ^2`$ for the massive dilaton.By substituting this last expression into equation (11) yields the following dilaton field equation
$$\dot{\varphi }=m\varphi $$
(12)
where we choose the minus sign for later convenience.The dilaton solution is
$$\varphi (t)=ae^{mt}$$
(13)
Nevertheless equation (11) should be matched with the well known dilaton equation
$$\ddot{\varphi }+3H\dot{\varphi }=\frac{V(\varphi )}{\varphi }$$
(14)
which solution is
$$\varphi =e^{\frac{3H}{2}t}$$
(15)
The spin-torsion density does not appear in this equation since it does not interact with the dilaton field.With this matching we obtain $`m=\frac{3H}{2}`$ and the constant $`a=1`$.Therefore dilaton mass can be expressed in terms of the Hubble constant. Equating equations (2) and (3) yields
$$2(\rho _{eff}2\pi G\sigma ^2)=(\rho _{eff}+3p_{eff}8\pi G\sigma ^2)$$
(16)
and substitution of the radiation equation of state into equation (14) yields
$$\rho _{eff}=3\pi G\sigma ^2$$
(17)
Substitution of (15) into equation (4) yields
$$\dot{\sigma ^2}+8H\sigma ^2=0$$
(18)
which yields
$$\sigma ^2=e^{8Ht}$$
(19)
and $`\rho _{eff}=3V(\varphi )`$.From (15) yields
$$V(\varphi )=\pi G\sigma $$
(20)
From expression (18) and (19) we note that the dilaton vanishes at recent epochs of the Universe as well as the spin-torsion density and no spin-torsion dominated inflation over the dilatonic inflation exists.Also this fact is also in agreement with the dificulty in observing torsion effects on cosmology from the classical point of view. Similar aspects were discussed recently by Maroto and Shapiro . Note that the dust case is very similar since equation (11) is the same except that the factor of two in front of the potential should be dropped.Let us now examine that radiation era for a torsion kink potential $`V(\varphi )=\alpha \varphi ^4`$.In this case equation (11) changes to $`\dot{\varphi }^2+\alpha \varphi ^4=0`$ which has a simple solution $`\varphi =(\alpha ^{\frac{1}{2}}t)^1`$.We would like to mention that also in the radiation era the repulsive gravity is in effect since the violation of the geodesic convergence condition is maintained in this case.From the COBE data we know that $`\frac{\delta T}{T}=10^5`$.The spectrum of gravitational waves with a temperature fluctuation of the order $`\frac{\delta T}{T}=\frac{H}{m_{Pl}}`$ and in our model there is a relation between the dilatonic mass and the Hubble constant thus $`H=\frac{2}{3}m=10^4m_{Pl}`$ the temperature fluctuation yields the value $`\frac{\delta T}{T}=10^4`$ which agrees with the value observed by COBE.From formula (19) and the density of temperature based on the gravitational wave spectrum one must infer that $`H=\frac{ln\sigma ^2}{8t}`$.Substitution of this value into the gravitational wave spectrum density we obtain $`\frac{\delta T}{T}=\frac{ln\sigma ^2}{8m_{Pl}t}=10^5`$ by the COBE data ,this in turn implies that $`\sigma ^2=e^{8.10^{5t}}`$.At the inflation era $`(t=10^{35s})`$ the value of the spin-torsion density becomes $`\sigma ^2=e^{10^{45}s}`$.This is an extremely big value if we think that today’s Universe possess a spin-torsion density of $`e^{10^{15billionyears}}`$!.
## Acknowledgements
I am very much indebt to Prof.M.Gasperini,Prof.Shapiro,Prof.R.Ramos and Prof.H.P. de Oliveira for helpful discussions on the subject of this paper. Financial support from CNPq. and UERJ (FAPERJ) is gratefully acknowledged.
|
no-problem/9906/cond-mat9906422.html
|
ar5iv
|
text
|
# Simulation of I-V Hysteresis Branches in An Intrinsic Stack of Josephson Junctions in High 𝑇_𝑐 Superconductors
## 1 Introduction
I-V characteristics of the high T<sub>c</sub> superconductor Bi<sub>2</sub>Sr<sub>2</sub>Ca<sub>1</sub>C<sub>2</sub>O<sub>8</sub> (BSCCO) shows a strong hysteresis, producing many branches . Many experiments have been reported to investigate properties of hysteresis branches , and it has turned out that there is rich physics in properties of I-V characteristics. Substructures are found in hysteresis branches, and are discussed in relation with phonon effects . The d-wave effect is also discussed in higher voltage region, where the quasi-particle tunneling current becomes important .
The present paper is concerned to the origin of hysteresis jumps forming the main branches. Instead of attributing the origin of hysteresis branches to inhomogeneity of critical currents, there are several theoretical attempts to regard its origin as an intrinsic property of strong anisotropy in high T<sub>c</sub> superconductors. High T<sub>c</sub> superconductors have a layered crystal structure. Specially, BSCCO has a strong anisotropy. Then a model of multi-layered Josephson junctions, which has an alternating stack of superconducting and insulating layers, is considered as a proper model to describe electronic properties of those systems. One of main differences from the case of a single junction is the appearance of an interlayer coupling among phase-differences. Since the thickness of the superconducting layer is of the order of 3-6Å, influence among adjacent junctions is not negligible.
To explain the main branch structure, the following mechanisms are proposed. A mechanism of the interlayer coupling induced by the charging effect at the superconducting layer has been proposed by one of the author (T. K.) , and a branch structure of I-V characteristics is obtained . It is pointed out in ref. that a nonequilibrium effect in superconducting layers also mediates an inter-layer coupling, and I-V characteristics are discussed . When the magnetic field is applied, the inductive coupling becomes important, and its effects have been discussed .
In this paper, we investigate in detail the scheme of hysteresis jumps appearing in main branches without magnetic filed. Since the inductive coupling among adjacent junctions does not exist, the charging effect at superconducting layers becomes important. A charging at a superconducting layer induces the change of electric fields in the neighboring insulating layers and induces an inter-layer coupling, since the time-derivative of the phase-difference is related with the voltage. We will show that the origin of the multi-branch structure is the nonlinear effect among phase-differences through the charging effect. Various disturbances, such as dissipation in insulating layers, dissipation in superconducting layers, noise, a surface proximity effect affect schemes of hysteresis jumps. We will investigate how the schemes of hysteresis jumps are affected by the above mentioned disturbances. The dissipation in the superconducting layer will be introduced phenomenologically through the charge relaxation. Schemes of hysteresis jumps are obtained by increasing applied current gradually up to certain value and then decreasing. Obtained equations show that the resistive dissipation in the insulating layer works to damp the motion of phase-difference in each junction, while the charge relaxation in the superconducting layer works to damp relative motions of phase differences among neighboring junctions. This fact affects the way of formation of hysteresis jumps in the process of increasing and decreasing applied current. It will turn out that the dissipation in superconducting layer and noise enhance the occurrence of jumps in both increasing and decreasing current, although the main branch structure is determined by the coupling through the charging effect. By comparing the result of numerical simulation with the experiment of a small stack , we will show validity of the present model.
In the next section, the necessary formula for calculation is summarized. The dissipation in superconducting layers is introduced phenomenologically. In Sect. 3, results of numerical simulation are presented. The origin of hysteresis jumps is investigated. It will be shown that, at a position of a jump, a stationary solution changes its form. We present a comparison with the experiment of ref. and show validity of the charging mechanism. Sect. 4 is devoted the conclusion.
## 2 Josephson Junction Stacks
In this section we summarize the multi-Josephson junction model of high T<sub>c</sub> superconductors . Let us consider the $`N+1`$ superconducting layers, numbered from $`0`$ to $`N`$. We denote the gauge invariant phase difference of the $`(l1)`$-th and $`l`$-th superconducting layer by $`\phi (l)`$, and voltage by $`V(l)`$. The widths of the insulating and superconducting layers are denoted by $`D`$ and $`s`$, respectively. At the edges, the effective width of the superconducting layer may be extended due to the proximity effect into attached lead metals. The widths of the $`0`$-th and $`N`$-th superconducting layers are denoted by $`s_0`$ and $`s_N`$, respectively.
We assume that physical quantities are spatially homogeneous on each layer. This assumption is applicable in cases of no applied magnetic field and small sample size in the a-b direction. According to the procedure presented in ref. , we have the following equation for the total current $`J`$ flowing through each junction as
$$\frac{J}{J_c}=j_c(l)\mathrm{sin}\phi (l)+\beta \frac{V(l)}{V_0}+\frac{1}{\omega _p}\frac{}{t}\frac{V(l)}{V_0}.$$
(2.1)
The current $`J`$ is normalized by the critical current $`J_c`$, $`j_c(l)=J_c(l)/J_c`$ with $`J_c(l)`$ being the critical current for the $`l`$-th junction. The time $`t`$ is normalized by the inverse of the plasma frequency, $`\omega _p`$ = $`\sqrt{\frac{2e}{\mathrm{}}\frac{4\pi DJ_c}{ϵ}}`$, with $`ϵ`$ being dielectric constant of the insulating layer. The voltage $`V(l)`$ is normalized by $`V_0=\frac{\mathrm{}\omega _p}{2e}`$. The parameter $`\beta `$ is given by $`\beta =\frac{\sigma V_0}{J_cD}`$. In this paper we assume
$$j_c(l)=1\text{for all }l,$$
(2.2)
for simplicity. The relation between time-derivative of phases and voltages at junctions is given by
$$\frac{1}{\omega _p}\frac{}{t}\phi (l)=\underset{l^{}=1}{\overset{N}{}}A_{ll^{}}\frac{V(l^{})}{V_0}.$$
(2.3)
The matrix $`A`$ is given by
$$A=\left(\begin{array}{cccccc}1+\alpha (1+\frac{s}{s_0})& \alpha & 0& \mathrm{}& & \\ \alpha & 1+2\alpha & \alpha & 0& \mathrm{}& \\ 0& \alpha & 1+2\alpha & \alpha & 0& \mathrm{}\\ & & \mathrm{}& & & \\ \mathrm{}& 0& & & \alpha & 1+\alpha (1+\frac{s}{s_N})\end{array}\right),$$
(2.4)
where
$$\alpha =\frac{\mu ^2ϵ}{sD}$$
(2.5)
with $`\mu ^2`$ being given by the relation between the charge density and scalar potential $`A_0`$ as
$$4\pi \rho =\mu ^2(A_0+\frac{\mathrm{}}{2e}\frac{}{t}\varphi ),$$
(2.6)
where $`\varphi `$ is the phase of the order parameter defined on the superconducting layer. In the present analysis, we assume that the current flows only along the c-axis, say, the z-direction. Since the current $`j_z`$ in superconducting layers is given by
$$\frac{4\pi }{c}j_z=\lambda ^2(A_z\frac{\mathrm{}c}{2e}\frac{}{z}\varphi ),$$
(2.7)
the current conservation law leads to
$$\mu ^2\frac{1}{c}\frac{}{t}A_0+\lambda ^2\frac{}{z}A_z=0,$$
(2.8)
under the condition
$$\left(\mu ^2(\frac{1}{c}\frac{}{t})^2\lambda ^2\left(\frac{}{z}\right)^2\right)\varphi =0.$$
(2.9)
The gauge condition (2.8) is called phason gauge , and is determined in such a way that the response of the current and charge to the space-time variation of phase appears in a gauge invariant form. It is shown in ref. that a nonequilibrium effect caused by the tunneling current induces the dissipation effect in $`\mu ^2`$. Since the superconducting layer is thin, the current also receives effects from boundaries between superconducting and insulating layers, which may also works as dissipation to the superconducting current. When the effect of the d-wave superconductivity is considered, one expects also certain dissipation due to the presence of the gap-less region. Such effects have not been much studied yet. Then we introduce phenomenologically the dissipation in superconducting layer in a Lorentz form as
$$\mu ^2=\mu _0^2\frac{1}{1i\omega \tau }.$$
(2.10)
This corresponds to the following phenomenological form of charge relaxation,
$$\frac{}{t}\rho =\frac{1}{\tau }(\rho +\frac{\mu _0^2}{4\pi }(A_0+\frac{\mathrm{}}{2e}\frac{}{t}\phi )).$$
(2.11)
By considering the expression $`\alpha `$ in Eq. (2.5), we replace $`\alpha `$ by
$$\alpha \alpha (1i\omega \tau )$$
(2.12)
with $`\alpha `$ being given by Eq. (2.5) replaced $`\mu ^2`$ by $`\mu _0^2`$. From Eqs. (2.1) and (2.3), we have
$$\frac{1}{\omega _p^2}\frac{^2}{t^2}\phi (l)=\underset{l^{}}{}(A_{ll^{}}+A_{1ll^{}}\tau \frac{}{t})(\frac{J}{J_c}j_c(l^{})\mathrm{sin}\phi (l^{}))\beta \frac{1}{\omega _p}\frac{}{t}\phi (l),$$
(2.13)
where $`A`$ is the matrix $`A`$ in Eq. (2.4) and $`A_1`$ is given as
$$A_1=\left(\begin{array}{cccccc}\alpha (1+\frac{s}{s_0})& \alpha & 0& \mathrm{}& & \\ \alpha & 2\alpha & \alpha & 0& \mathrm{}& \\ 0& \alpha & 2\alpha & \alpha & 0& \mathrm{}\\ & & \mathrm{}& & & \\ \mathrm{}& 0& & & \alpha & \alpha (1+\frac{s}{s_N})\end{array}\right).$$
(2.14)
As the above derivation shows, the charging effect at superconducting layers is the main mechanism inducing the inter-layer coupling through the electromagnetic interaction, when the spatial homogeneity in the a-b plane is satisfied. A charge at a superconducting layer modifies electric fields in the neighboring insulating layers, and the change of electric fields affects the time-derivative of phase differences. Phonon effect considered in ref. may be introduced through the frequency dependence of the dielectric constant $`ϵ`$, which is not considered in this paper. The equation (2.13) means that the dissipation in the insulating layer works as a damping of phase motions at each junction, and that the dissipation in the superconducting layer works as a damping of the relative phase motion. In the next section, we present results obtained by numerical simulation of Eq. (2.13).
## 3 Results of Numerical Simulation
We have performed numerical simulations of Eq. (2.13) by use of the 5-th order of the predictor corrector method. In all calculation, we choose the parameter $`\alpha =1.0`$, $`\beta =0.2`$. The time step $`dt`$ is chosen as $`\omega _pdt=1.0\times 10^3`$.
The parameter $`\beta `$ is related to the conductivity of the insulating layer. Although it works to lead a system into a stationary state and determines the slope of the I-V characteristics, its effect to a scheme of hysteresis jumps is not sensitive in this parameter range. The parameter $`\alpha `$ plays an important role to determine hysteresis branches due to non-linear effects among phase-differences. The value of $`\alpha `$ in BSCCO was estimated in ref. as $`1<\alpha <3`$ and, by use of $`ϵ=25`$ for La<sub>2-x</sub>Sr<sub>x</sub>CuO<sub>4</sub>, the value of $`\alpha =2.7`$ was obtained. There is a discussion that $`ϵ`$ for BSCCO may be smaller, and from a microscopic theory, $`\alpha `$ is estimated much smaller value as 0.2 . Also it was reported that the equal spacing of I-V branches observed in experiments of BSCCO is reproducible, with a smaller choice of $`\alpha 1`$ . Here we use a typical value $`\alpha =1.0`$.
By use of Eqs. (2.3) and (2.1), the relation for obtaining a voltage for each junction is written as
$$\frac{1}{\omega _p}\frac{}{t}\phi (l)=\underset{l^{}}{}((A\tau \beta A_1)_{ll^{}}\frac{V(l)}{V_0}+\tau A_{1ll^{}}(\frac{J}{J_c}j_c(l^{})\mathrm{sin}\phi (l^{}))).$$
(3.1)
The average of voltage $`\overline{V}(l)`$ is given by
$$\overline{V}(l)=\frac{1}{T_{max}T_{min}}_{T_{min}}^{T_{max}}𝑑tV(l).$$
(3.2)
The total DC-voltage $`V`$ is obtained by
$$V=\underset{l=1}{\overset{N}{}}\overline{V}(l).$$
(3.3)
In Fig. 1, I-V characteristics is presented for $`N=10`$, $`s_0/s=s_N/s=2`$, $`\tau \omega _p=0.1`$. The current $`J`$ is gradually increased with the step $`dJ/J_c=0.01`$ up to $`J/J_c=2.0`$, and then gradually decreased. The I-V curve shows the jump at $`J/J_c=1.0`$ and, after two jumps, it linearly increases up to $`J/J_c=2.0`$. In the current decreasing process, there appear three jumps when the current becomes smaller than $`J/J_c=0.4`$.
In Fig. 2, we show the branch structure in the I-V characteristics corresponding to Fig. 1. This is obtained by decreasing the current when a jump occurs in the current increasing process, and by increasing the current when a jump occurs in the current decreasing process. It should be noted that the positions of jumps labeled by 1 and 5 lie on the same branches, while the positions 2 and 4 are close but lie on the different branches.
In order to understand this branch structure, we show, in Fig. 3, the voltage distributions on each junction just after the jumps. Fig. 3(a) is for the current increasing process and Fig. 3(b) is for the current decreasing process. For the line labeled 1, two junctions at the edges have higher voltage. The solution shows that the phases with higher voltages increases approximately linearly in time, corresponding to the phase rotating motion, while other junctions have oscillating phase motions. In fact, when the time average of $`\frac{\phi (l)}{t}`$ is plotted in Fig. 4, junctions with rotating and oscillating phases are clearly identified. As can be seen in Fig.3, the number of junctions with rotating phase increases as 2, 4, and 6 due to the symmetric arrangement. The number of phase rotating junctions is two for the voltage lines labeled by 1 and 5, four for 2 and 4. The lines 1 and 5 shows the same pattern of voltage distribution and therefore form the same branch in Fig.3, while the line 2 and 4 show the different pattern, though the number of the phase rotating junction is same.
With this analysis, we see that the hysteresis jumps are associated with the change of the distribution pattern of rotating phase motions. If the $`l`$-th junction has a rotating phase, the time average of $`\frac{\phi (l)}{t}`$ is constant and that of $`\mathrm{sin}(\phi (l))`$ is zero. If the $`l`$-th junction has an oscillating phase, the time average of $`\frac{\phi (l)}{t}`$ is zero and that of $`\mathrm{sin}(\phi (l))`$ is constant. Let us consider that $`N_R(0)`$ junctions have rotating phases. Considering the above facts, we can obtain the equation to determine the voltage-current relation from the time-average of Eqs. (2.1) and (2.13) as
$$\beta \frac{V}{V_0}=N\frac{J}{J_c}\underset{l_0}{}j_c(l_0)<\mathrm{sin}\phi (l_0)>,$$
(3.4)
$$\underset{l_0^{}}{}A_{l_0l_0^{}}j_c(l^{})<\mathrm{sin}\phi (l_0^{})>=a(l_0)\frac{J}{J_c},$$
(3.5)
where $`<\mathrm{}>`$ indicates time average, $`l_0`$ and $`l_0^{}`$ are the labels of junctions with oscillating phases,
$$a(1)=1+\alpha \frac{s}{s_0},a(N)=1+\alpha \frac{s}{s_N},\text{otherwise}a(l)=1.$$
(3.6)
As is shown in Fig. 3, even if the number of phase-rotating junctions is same, slightly different branches are formed according to the pattern of distribution of rotating phases. That is, branches are grouped by the number of phase-rotating junctions, $`N_R`$, and different patterns of distribution of phase-rotating junctions lead slightly shifted branches. The Eqs. (3.4) and (3.5) are solved in a form
$$\beta \frac{V}{V_0}=n\frac{J}{J_c}.$$
(3.7)
In Fig. 5, we present $`n`$ versus $`N_R`$, the number of rotating phases. The small dots are for all possible patterns. The edge junctions mostly rotate, and rotating phases have mostly oscillating phases in the next neighbors. Taking into account these facts, we have the larger circles with the restrictions, 1) pattern is symmetric, 2) two edges have rotating phases and 3) there are no patterns with more than three successive rotating phases. From the figure we can see that branches appears roughly with equal spacing but that those at higher voltage distribute denser. In the numerical simulation, it seems that only special patterns are realized. The simulated result in Fig. 2 gives $`n=3.21`$ for $`N_R=2`$, $`n=6.21`$ and $`6.47`$ for $`N_R=4`$, and $`n=8.59`$ for $`N_R=6`$. What stability conditions should be imposed in addition to Eqs. (3.4) and (3.5) is still an open question.
In order to see how the scheme of hysteresis jump changes with various conditions, we show, in the following, structures of branches in I-V characteristics. In these analyses, the current is increased up to $`J/J_c=2`$ and decreased to zero. Branches are picked up when a voltage jump occurs. Fig. 6 is for finite junction stack $`N=10`$, and $`s_0=s_N=2.0`$ is chosen. The parameters are $`\alpha =1.0`$, $`\beta =0.2`$. In Fig. 6, (a) is for $`\omega _p\tau =0.0`$ without noise, (b) is for $`\omega _p\tau =0.1`$ without noise, (c) is for $`\omega _p\tau =0.0`$ with noise, and (d) is for $`\omega _p\tau =0.1`$ with noise. The noise is introduced through the small randomness of the time derivative of the phase at each time step, $`\delta (\frac{1}{\omega _p}\frac{\phi (l)}{t})`$ = $`1.0\times 10^4`$. Fig. 7 is for N=10 with the periodic boundary condition. Fig. 7(a) is for $`\omega _p\tau =0.0`$ without noise, (b) is for $`\omega _p\tau =0.1`$ without noise, (c) is for $`\omega _p\tau =0.0`$ with noise, and (d) is for $`\omega _p\tau =0.1`$ with noise.
When a noise is introduced, possible patterns are not necessarily symmetric, so that the number of branches increased. In addition, an instability to cause a hysteresis jump is much easily enhanced. The effect of $`\tau `$ also enhance hysteresis jumps, which may be understood from the fact that it works as damping of the relative motion. Specially, it enhances hysteresis jumps in the current increasing process.
In the periodic cases, this tendency is more enhanced as can be seen from comparison with Fig. 7(a) and others. In a long stack, where surface effects are less, we can say that the dissipation effect in the superconducting layer or noise play an important role to induce hysteresis jumps.
In Fig. 8, we present the comparison with our numerical result and the experiment of ref. . We choose $`N=8`$ from their arrangement of the sample. In order to explain large extension of hysteresis branches in the current increasing process, we assume that the critical currents $`J_c`$ are reduced at the edges. We choose $`J_c=0.47meV`$, since the numerical analysis shows that the $`J_c`$ is situated roughly at the middle of the second branch. From the data, we have $`J_c(0)/J_c=0.4`$, $`J_c(N)/J_c=0.9`$ and other $`J_c(l)/J_c=1.0`$. The parameter $`\alpha `$ is chosen 1.0, by assuming $`s=6`$Å, $`D=6`$Å, $`\mu =`$2Åand $`ϵ=10`$. In ref. , it is reported that the longitudinal plasma energy is of order of $`0.5meV`$. Then $`V_0`$ is $`0.25meV`$. The $`\beta `$ is estimated from the branch 4; that is, the branch 4 has four rotating phases and is on the line passing $`J=0.9mA`$, $`V=60meV`$. In Eq. (3.7) we choose $`n=6`$ from the result of Fig. 5, and $`\beta `$ is estimated as $`\beta =0.05`$. Since we do not introduce any other parameters, we set $`s_0/s=1.0`$, $`s_N/s=1.0`$. The parameter $`\omega _p\tau `$ is chosen as $`0.5`$, which is obtained by trial and error to produce as many branches as possible. The inset is from ref. . The labels of branches correspond to those in the inset. The numbers in the brackets indicate the number of rotating phases. The values of J, at which hysteresis jumps occur, fit very well with experimental values of the inset. Hysteresis jumps are induced when certain instability develops in the non-linear equation of (2.13). Though each junction, except for one at edges, has the same $`J_c`$, a critical $`J`$ for each branch is different and is controlled mainly by $`\alpha `$. The agreement of this critical $`J`$-values observed in Fig. 8 supports the validity of the present charging mechanism. In Fig. 8 certain branches in the increasing current are missing. How hysteresis jumps occurs depends of effects of disturbances. There may be certain noises to make development of instability easier. The length of branches depends on choice of parameter. In decreasing current, there are a few discrepancies. In numerical calculation, the I-V curve is straight, while the experiment shows a curvature in the low current region. In addition, the positions of jumps are different. Such discrepancy may relate effects of d-wave superconductivity, such as the presence of the quasi-particle tunneling, for example, but it is beyond the discussion of the present work.
## 4 Conclusion
In this paper we have studied the I-V characteristic branches in high T<sub>c</sub> superconductors by the multi-layered Josephson junction model. We have shown that the charging effect in superconducting layers causes an inter-layer coupling among phase-differences and that hysteresis jumps are induced as a result of non-linear effect among phase-differences. The appearance of hysteresis branches is intrinsic in high T<sub>c</sub> superconductors. On this aspect, we note the experimental report that, even in a sample with negligible $`J_c`$-inhomogeneity, a branch structure is observed .
Due to the weak interlayer coupling arising from the charging effect, equations for phase-differences form a non-linear coupled differential equation. Its solution is classified by the number of rotating phases. The change of its number occurs intrinsically in this nonlinear equation, which leads to formation of I-V characteristic branches. Although the number of rotating phases is same, different patterns of distribution of rotating phases lead slightly shifted branches. It is still an open question and a future problem which patterns are more easily realized among many possible patterns.
In this system, the resistive dissipation in the insulating layer works as damping of phase motion in each junction and the dissipation caused by charge relaxation in superconducting layers works as damping of relative phase motions. The effects of dissipation, noise, and surface proximity to structure of hysteresis jumps in I-V characteristics are investigated. The dissipation in superconducting layer and noise enhance the occurrence of jumps in both increasing and decreasing current.
We have presented the comparison with the experiment in Fig. 8. The agreement of the critical $`J`$ for each hysteresis branch shows validity of the charging mechanism as the origin of the inter-layer coupling.
Finally, we comment on effects of the d-wave superconductivity. The fact that the Josephson current is given by the phase-difference is not modified in the d-wave superconductivity with a single component. Therefore, the main conclusions in the present paper are not modified. As was pointed out in the text, the d-wave superconductivity gives, due to the presence of a gapless region, an additional contribution to dissipation in superconducting layers and additional quasi-particle tunneling current even in the low voltage region. Effects of such contributions may be interesting future problems.
Acknowledgements
This work was supported by Special Research Grant in Faculty of Engineering, Seikei University. One of the authors (F. W.) was supported by Ikuei scholarship in Japan.
Figure captions
Fig.1 I-V characteristics Parameters are $`\alpha =1.0`$, $`\beta =0.2`$, $`N=10`$, $`\frac{s_0}{s}=\frac{s_N}{s}=2.0`$ and $`\omega _p\tau =0.1`$.
Fig.2 I-V characteristics and branch structure Parameters are $`\alpha =1.0`$, $`\beta =0.2`$, $`N=10`$, $`\frac{s_0}{s}=\frac{s_N}{s}=2.0`$ and $`\omega _p\tau =0.1`$.
Fig.3 Voltage distribution Parameters are $`\alpha =1.0`$, $`\beta =0.2`$, $`N=10`$, $`\frac{s_0}{s}=\frac{s_N}{s}=2.0`$ and $`\omega _p\tau =0.1`$. (a) is for increasing current, and (b) is for decreasing current.
Fig.4 Average $`\frac{1}{\omega _p}\frac{\phi }{t}`$ distribution Parameters are $`\alpha =1.0`$, $`\beta =0.2`$, $`N=10`$, $`\frac{s_0}{s}=\frac{s_N}{s}=2.0`$ and $`\omega _p\tau =0.1`$. (a) is for increasing current, and (b) is for decreasing current.
Fig.5 Slope $`n`$ vs. the number of rotating phases $`N_R`$ Slope $`n`$ is defined by $`\beta \frac{V}{V_0}=n\frac{J}{J_c}`$. Parameters are $`\alpha =1.0`$, $`N=10`$, $`\frac{s_0}{s}=\frac{s_N}{s}=2.0`$. Small dots correspond to all possible patterns, while larger circles are with restrictions presented in the text.
Fig.6 I-V characteristics and branch structure for a finite stack Parameters are $`\alpha =1.0`$, $`\beta =0.2`$, $`N=10`$, $`\frac{s_0}{s}=\frac{s_N}{s}=2.0`$. (a)$`\omega _p\tau =0.0`$ without noise, (b)$`\omega _p\tau =0.1`$ without noise, (c)$`\omega _p\tau =0.0`$ with noise, and (d)$`\omega _p\tau =0.1`$ with noise.
Fig.7 I-V characteristics and branch structure for a periodic stack Parameters are $`\alpha =1.0`$, $`\beta =0.2`$, $`N=10`$, $`\frac{s_0}{s}=\frac{s_N}{s}=2.0`$. (a)$`\omega _p\tau =0.0`$ without noise, (b)$`\omega _p\tau =0.1`$ without noise, (c)$`\omega _p\tau =0.0`$ with noise, and (d)$`\omega _p\tau =0.1`$ with noise.
Fig.8 I-V characteristics Inset is from ref. . The labels of the branches are identified with those of the inset. The numbers in the bracket are the numbers of rotating-phases.
|
no-problem/9906/hep-th9906182.html
|
ar5iv
|
text
|
# 1 Introduction
## 1 Introduction
In string theory, when considered as a framework for unifying gravity and quantum mechanics, the fundamental strings are naturally thought of as Planck size objects. At much lower energies, such as the typical weak or strong interaction scales, the strings have lost all their internal structure and behave just as ordinary point-particles. The physics in this regime is therefore accurately described in terms of ordinary local quantum field theory, decoupled from the planckian realm of all string and quantum gravitational physics. The existence of this large separation of scales has often been used as an argument against the potential predictive power or even the potential reality of string theory. However, as we will argue in this letter, there are a number of recent insights and ideas that suggest a rather more optimistic scenario, in which string physics may have visible consequences at much lower energies scales than assumed thus far. The two particular recent developments that we wish to combine are:
(i) The AdS/CFT correspondence
The basic example here is the proposed duality between four-dimensional $`𝒩`$ =4 SYM theory and type II B string theory on $`AdS_5\times S^5`$ geometry
$$ds^2=e^{2y/R}\mathrm{𝑑𝑠}_4^2+dy^2+R^2d\mathrm{\Omega }_5^2$$
(1)
Here $`\mathrm{𝑑𝑠}_4^2=dx_{_{//}}^2`$ denotes the flat four-dimensional metric and
$$R^2=\alpha ^{}\sqrt{4\pi Ng_s}$$
(2)
with $`g_s=g_{ym}^2`$ the type IIB string coupling . In this AdS/CFT dictionary, the coordinate $`y`$ needs to be thought of as parametrizing the four-dimensional scale: two SYM excitations related by a scale transformation
$$x_{_{//}}e^\lambda x_{_{//}}$$
(3)
translate on the AdS-side into two excitations concentrated around different locations in the $`y`$-direction related by a translation
$$yy+\lambda R.$$
(4)
This map thus provides a “holographic” projection of the physics of the gauge theory (which can be thought of as living on the AdS-boundary) onto the one higher-dimensional AdS space.
Truncating the AdS theory to $`y`$-values larger (or smaller) than some finite value $`y=y_0`$ amounts to introducing an UV (or IR) cut-off in the gauge theory . In the strict continuum limit, however, the range of $`y`$-values extends over the full real axis. A consequence of this is that, while the type II theory on the AdS-space contains gravity, the dual conformal gauge theory on the boundary does not. Modes of the AdS gravitational field that extend all the way to the boundary $`y\mathrm{}`$ are not normalizable, and therefore do not fluctuate.
The metric $`ds^2`$ has the rather striking property that for sufficiently large $`y`$-values it differs from the four-dimensional brane-metric $`\mathrm{𝑑𝑠}_4^2`$ by an exponentially large “red-shift factor” $`e^{2y/R}`$ , that is, distances along the $`x_{_{//}}`$-directions are measured very differently by both metrics. A most dramatic consequence of this that the fundamental IIB strings, while still of Planck size when viewed by the full metric $`ds^2`$, can become arbitrarily large when measured in terms world-brane metric $`\mathrm{𝑑𝑠}_4^2`$, simply by moving to larger $`y`$ values. The SYM interpretation of these large IIB strings is that they represent the color-electric flux lines .
In principle this type of holographic correspondence can be generalized to less symmetric gauge theories with non-zero beta-functions and more complicated phase diagrams. The idea remains the same, namely that $`y`$-translations in effect amount to renormalization group transformations on the gauge theory side. Beta-functions or symmetry breaking or other type of phase transitions thus translate into a non-trivial $`y`$-dependence (e.g. in the form of domain wall structures) of the metric, dilaton and other fields relative to the $`AdS_5`$-background.
(ii) Compactifications with a D3-brane world
In a large class of type I string compactifications, the gauge theory degrees of freedom in our four-dimensional world may be thought of as bound to a (collection of) D3-brane(s). These 3-branes wrap our world but are otherwise localized as point-like objects somewhere inside the compactification manifold. Scenarios of this type have recently been studied from various points of view, as in particular they seem to offer some promising new avenues for addressing the gauge hierarchy problem . Besides (perhaps) allowing for the possibility of large extra dimensions , a second new aspect of these type of compactifications is that (due to the backreaction of the D3-branes) various fields, such as the dilaton and in particular the conformal factor of the four-dimensional metric, may acquire non-trivial dependence on the compact coordinates.
An interesting example of this type of geometry was recently considered by Randall and Sundrum in . In this set-up, the four-dimensional conformal factor is found to depend exponentially on the extra fifth coordinate $`y`$, precisely as in the AdS-geometry (1). A new ingredient, relative to the standard AdS situation, is that in the extra coordinate $`y`$ is chosen to run over a finite (or semi-infinite) range. As a consequence, there exists a normalizable gravitational collective mode, given by those fluctuations of the metric $`ds^2`$ that preserve the form (1) but with $`\mathrm{𝑑𝑠}_4^2`$ replaced by a (sufficiently slowly varying but otherwise) general four-dimensional metric
$$\mathrm{𝑑𝑠}_4^2=g_{\mu \nu }^{(4)}(x_{_{//}})dx_{_{//}}^\mu dx_{_{//}}^\nu .$$
(5)
Hence this collective mode behaves just as the ordinary graviton of our four-dimensional world. In addition, due to the exponential $`y`$-behavior of the graviton wave-function, a given object with five-dimensional mass $`m`$ has an effective four-dimensional mass $`m_4`$, that depends of its $`y`$-location via
$$m_4(y)=me^{y/R}$$
(6)
In the presence of this exponential red-shift factor (or “warp factor”) was argued to provide a natural explanation of the large mass hierarchies, such as between the weak and Planck scale.
Putting (i) and (ii) together:
It seems natural to look for a way of combining these two ideas. Concretely, one could ask the following two (probably equivalent) questions:
* Can the AdS/CFT type of dualities be extended to situations where the radial AdS-coordinate is effectively compactified, and does this compactifiation indeed automatically imply the presence of four-dimensional gravity on the boundary?
* Are there string compactification scenarios where a compact slice of AdS-space appears as part of the compactification geometry, and does the radial AdS coordinate then again have a “holographic” interpretation as parametrizing the RG scale of the four-dimensional theory?
In the following we will argue that both questions have a positive answer, though the first one only for a (large but) finite set of D3-brane charges $`N`$. Consequently, the compactified AdS-geometry will be a good approximation provided the string coupling $`g_s`$ is not too small compared to $`1/N`$. In the concluding section we will address some of the possible consequences of this new view on string compactification.
## 2 AdS Compactification
Let us start with considering the toy example of a four-dimensional $`𝒩`$=4 supersymmetric world described by the $`T^6`$ compactification of the type I superstring. Its low energy effective description is given by $`𝒩`$=4 super-Yang-Mills theory, with as maximal unbroken gauge symmetry group $`𝒢=SO(32)`$, coupled to supergravity. In the following we will consider the situation where this symmetry is broken down to an $`U(N)`$ sub-group with $`N16`$.
By applying $`T`$-duality this type I string theory can be equivalently described as an orientifold of type II string theory on $`T^6`$. In this representation, there are $`2^6=64`$ orientifold planes located at all the half-way points of the $`T^6`$. In addition, there are 32 D3-branes inside the $`T^6`$ which are pairwise identified under the orientifold $`𝐙_2`$-action . Hence it is possible for $`N16`$ of these D3-branes to form a small cluster inside the $`T^6`$, and in the limit where all $`N`$ coincide at the same point, the unbroken $`U(N)`$ gauge symmetry appears.
At low energies in the uncompactified world, gravity effectively decouples from the $`U(N)`$ SYM dynamics. Thus we can look for a regime of parameters in which we can apply the AdS/CFT duality map and obtain type II theory on $`AdS_5\times S^5`$ as a good dual low energy description. The main restriction (in order to be able to trust the sigma-model approximation) is that the $`AdS_5`$ radius of curvature $`R`$, given in (2), needs to be large compared to the string scale,
$$g_s>>\frac{1}{4\pi N}.$$
(7)
Hence, relative to the usual large $`N`$ context, we now have a somewhat more limited control over the approximations involved in the duality map.
Around the string scale $`L_S`$, open string effects will start to modify the SYM dynamics. Let us choose the radial AdS-coordinate $`y`$ such that this string scale $`L_S`$ gets mapped to the region around $`y0`$. Then the $`AdS_5\times S^5`$-description is a valid approximation for the low-energy regime $`y>>R`$. Around $`y0`$, on the other hand, we reach the transition region where the low energy regime meets with the high energy regime described by orientifold of type II on $`T^6`$. Our main new proposal is that these two regimes can be consistently glued together into one single dual type II string background. In the following we will assume that the size of the $`T^6`$ is bigger or of the order of the AdS-radius $`R`$ in (2).
As a first approximation, we can visualize the total target space of this new type II theory as follows. We first cut out a small ball around the D3-branes inside the $`T^6`$, such that its outside radius coincides with the $`S^5`$ radius $`R`$ in (2). Then we replace the inside of the ball by the $`y>0`$ region of $`AdS_5\times S^5`$. (See fig. 1). It seems reasonable to assume that the resulting total space can be smoothed out to obtain an exact consistent type II background, and the end result of this procedure should be equivalent to taking into account the gravitational backreaction of the D3-branes.
In a more complete treatment, we must also include the backreaction of the 64 orientifold planes. These have a negative tension equal to $`1/4`$ times the D3-brane tension, which need to be taken into account. we can write an explicit form for the background metric, by starting from the general form
$$ds^2=\frac{1}{H(x_{})^{1/2}}\mathrm{𝑑𝑠}_4^2+H(x_{})^{1/2}dx_{}^2.$$
(8)
with $`\mathrm{𝑑𝑠}_4^2=dx_{_{//}}^2`$, and where $`x_{}`$ denote the coordinates inside the $`T^6`$. For simplicity, let us consider the most symmetric example of the $`SO(32)`$ invariant point, where all 16 D3-branes coincide with their 16 $`𝐙_2`$ images at one of the orientifold fixed points, say at $`x_{}=0`$. Taking the $`T^6`$ to be an exact cube with period $`R_c`$, the harmonic function $`H(x_{})`$ then takes the form
$$H(x_{})=1+4\pi g_s(\alpha ^{})^2\left[f_D(x_{})f_O(x_{})\right]$$
(9)
where
$$f_D(x_{})=2\underset{\stackrel{}{n}𝐙^6}{}\frac{16}{|\stackrel{}{x}_{}+\stackrel{}{n}R_c|^4}$$
(10)
denotes the contribution of the 16 D3-branes and their $`𝐙_2`$-image, and
$$f_O=2\underset{\stackrel{}{n}𝐙^6}{}\frac{1/4}{|\stackrel{}{x}_{}+\frac{1}{2}\stackrel{}{n}R_c|^4}$$
(11)
that of the 64 orientifold planes located at all the half-way points inside the $`T^6`$. In the region close to the D3-branes at $`x_{}=0`$, the above geometry indeed reduces to the $`AdS_5\times S^5`$ metric (1).<sup>1</sup><sup>1</sup>1For the $`SO(32)`$ symmetric example at hand, the $`S^5`$ must in fact be replaced by its orientifold $`S^5/𝐙_2`$. We further notice that close to the other orientifold planes the function $`H`$ goes through 0; in this horizon region, however, $`\alpha ^{}`$ and non-perturbative string-corrections are expected to play an important role.
The present minimal example (of the simple $`T^6`$ orientifold compactification) can straightforwardly be generalized to less symmetric theories obtained e.g. by applying further orbifoldings. In this way one can find holographic duals to a large class of four-dimensional type I string theories, typically with a reduced number of (or even zero) supersymmetries. The total target space of the type II duals should all describe consistent compactifications of the $`AdS_5\times S^5`$ geometry. The number of such consistent AdS-compactifications, however, is constrained by the usual tadpole cancelation conditions. In particular, one needs to make sure that the RR flux $`N`$ that escapes the AdS-region gets absorbed by the appropriate number of branes and orientifold planes present in the compact outside region. Consistent solutions to this condition can be found only for a finite set of $`N`$-values. Still, relative to our minimal set-up, one can significantly extend the range of allowed values for $`N`$, up to values of the order of $`N10^3`$ or larger, by considering more general orbifold spaces. Hence in this more general setting, the restriction (7) in effect just amounts to $`g_s>>10^4`$, which leaves a quite sufficient parameter range for which we can trust our approximations.
## 3 Gravitational coupling
This geometric set-up is now indeed very similar to that of , where our $`T^6`$ region with the orientifold planes plays the role as the positive tension brane in their scenario. Several of the same conclusions apply. In particular, we must find a collective graviton mode of the exact same type as described in the Introduction, that couples precisely as the four-dimensional graviton to the boundary field theory. In our case this result was of course expected from the start, since the boundary theory is in fact the type I string theory on $`T^6\times 𝐑^4`$, that also contains a gravitational closed string sector.
The relation between the four-dimensional Planck length $`L_{pl}`$ and the 10-dimensional one $`L_{_{10}}`$ takes the usual form
$$(L_{_{10}})^8=(L_{pl})^2V_6$$
(12)
where $`V_6`$ is the appropriately measured volume of the $`T^6`$. Plugging the form (8) into the ten-dimensional action gives
$$\frac{1}{(L_{_{10}})^8}d^{10}x\sqrt{g}{}_{_{10}}{}^{}R_{_{10}}^{}=\frac{V_6}{(L_{_{10}})^8}d^4x_{_{//}}\sqrt{g}{}_{4}{}^{}R_{4}^{}$$
(13)
with
$$V_6=_{T^6}d^6x_{}H(x_{}).$$
(14)
Notice that, in spite of the fourth order pole in $`H(x_{})`$ near the branes, this integral indeed yields a finite answer.
The collective mode of the 10-dimensional metric that produces the four-dimensional graviton again has the same shape as the full metric, as given in eqn (8), with $`ds_4^2`$ a general four-metric (though sufficiently slowly varying with $`x_{_{//}}`$). The shape of the graviton wave-function the AdS-region decays with $`y`$ as $`e^{2y/R}`$. This exponential factor is reflects the fact that the coupling of matter to the graviton becomes weaker and weaker at lower and lower scales. Indeed, we can extract the strength of the coupling to matter by inserting the variation $`\delta g_{\mu \nu }^{(4)}=h_{\mu \nu }`$ of the four-metric $`ds_4^2`$ into the the ten-metric (8). The corresponding variation of the 10-d matter action reads
$$d^{10}x\sqrt{g}{}_{_{10}}{}^{}\delta g_{_{10}}^{\mu \nu }T_{\mu \nu }=d^4x_{_{//}}h^{\mu \nu }\overline{T}_{\mu \nu }$$
(15)
with
$$\overline{T}_{\mu \nu }(x_{_{//}})=d^6x_{}H(x_{})T_{\mu \nu }.$$
(16)
Inserting on the right-hand side the energy-momentum tensor (here all indices contracted with the ten-metric $`ds^2`$)
$$T_{\mu \nu }=\frac{m}{\sqrt{g_{_{10}}}}𝑑\tau \frac{\dot{x}_\mu \dot{x}_\nu }{\sqrt{\dot{x}^2}}\delta _{_{10}}(xx(\tau ))$$
(17)
of a 10-d point-particle, with mass $`m`$ located at a given location $`y_{}`$, gives for the effective 4-d energy-momentum tensor (with all indices contracted with the four-metric $`ds_4^2`$)
$$\overline{T}_{\mu \nu }=\frac{m}{H(x_{})^{1/4}}𝑑\tau \frac{\dot{x}_\mu \dot{x}_\nu }{\sqrt{\dot{x}^2}}\delta _4(x_{_{//}}x_{_{//}}(\tau ))$$
(18)
So the four-dimensional mass depends on $`x_{}`$ as
$$m_4(x_{})=\frac{m}{H(x_{})^{1/4}}.$$
(19)
This is the expected red-shift effect; close to the D3 branes it reduces to (6).
## 4 RG scale as a real extra dimension
Though familiar and standard, the appearance of both open and closed string dynamics in type I string theory is still a deep fact. The two are indeed intertwined in an intricate way, since when one starts including the open string quantum effects, there is no obviously unique way of separating the open string loop diagrams from closed string tadpoles. This very subtlety of course lies at the heart of the D-brane correspondence between gauge theory and gravity. A microscopic reconstruction of the reasoning of indeed shows that the appearance of the curved AdS-metric (1), as describing the near-horizon region close to the D3-branes, must be thought of as a cumulative consequence of open string quantum effects.
Concretely, one can imagine setting up a renormalization group flow, where via a Fischler-Susskind type mechanism, the effect of integrating out successive momentum shells in the open string channel gets absorbed into an appropriate redefinition of the world-sheet sigma-model couplings. Roughly speaking, in this procedure the slice of the AdS target space in between two values $`y_1`$ and $`y_2`$ gets created from the open string dynamics in between the corresponding energy scales. Eventually, after all open strings are integrated out, its planar diagrams have been replaced by a stretched-out closed string worldsheet, moving inside the semi-infinite AdS-throat geometry. Thus, in a quite precise sense, the AdS closed strings provide the long sought after dual representation of the gauge theory planar graphs .
If instead one runs this renormalization group flow from the string scale down to a certain finite length scale $`L`$ (for example the weak scale), the resulting target space consists of a compact slice of AdS bounded by the Planck region $`y0`$ on one side and by an effective D3-brane located at $`yR\mathrm{log}L/L_S`$ on the other. The D3-brane hosts the remaining sector of low energy open strings<sup>2</sup><sup>2</sup>2Due to the redshift, these open strings have acquired a renormalized tension $`\alpha _{eff}^{}\sqrt{Ng_s}L^2`$. The energy of a string with this tension stretched over a distance $`L`$ reproduces a force comparable to the $`1/L`$ Coulombic force of the SYM model., describing the physics at distance scales larger than $`L`$. Sub-planckian physics thus happens far inside the AdS-tube, and from this perspective the $`T^6`$-orientifold region merely acts as a kind of sounding board, providing some appropriate boundary condition in the far away planckian regime. Via high energy experiments on the D3-brane world-volume, however, one can still generate quantum fluctuations that extend to smaller values of $`y`$ and thereby probe the corresponding planckian geometry.
What new lesson can we extract from this? It is clear that, relative to the standard notion of an extra dimension, the $`y`$-direction is rather unusual. Normally one would expect that space-like separated events are independent, while here we are learning that $`y`$-translations are in effect scale transformations. Therefore, to avoid over-counting of the number of degrees of freedom, we thus indeed need to adopt the hypothesis that physics that happens far out in the $`y`$-direction is not really independent from that, say, at the boundary region $`y0`$, but rather a “holographic image” of physics at $`y0`$ happening at a corresponding scale $`L(y)=L_Se^{y/R}`$.<sup>3</sup><sup>3</sup>3Because of this holographic identification, the distinction made in between a “hidden” and “visible” brane is no longer applicable in our context. Also the (continuum) KK tower of gravitational excitations must be reinterpreted as actually representing low energy SYM degrees freedom in the boundary theory. This is the familiar dictionary of the AdS/CFT correspondence .
Though quite miraculous, in itself this holographic equivalence doesn’t teach us anything new yet, as it simply points to some redundance in the description. We could for example insist on describing all the physics as happening just at $`y0`$, and consider the holographic map as some purely mathematical equivalence. Relative to the standard AdS/CFT situation, however, we now have four-dimensional gravity as an essential new ingredient. In first approximation the gravity coupling seems quite consistent with holography, since (as seen from the previous section) it looks like to a relatively standard Kaluza-Klein dimensional reduction. The holographic reconstruction does, however, strictly amount to a non-trivial redistribution of matter, and since gravity is still in the game, in the end we do need to specify which energy-momentum distribution acts as a source for the gravitational field.
The new consequence of this holographic view of gravity, therefore, is that the renormalization group scale is promoted to a real physical extra direction. Concretely, this means that for type I string theory the gravitational energy-momentum in our four-dimensional world really spreads out into the extra $`y`$-direction, indeed similar to a holographic, or in a perhaps more accurate analogy, chromatographic image.
## 5 Concluding remarks
Finally, let us try to draw a few obvious and less obvious conclusions. We begin with
The Planck vs large N QCD string
A first striking property of this new scenario is that the planckian type II string and the low energy electric flux strings are in essence made up from one and the same object: the two are simply related by a translation in the $`y`$-direction. Similarly, in more realistic type I compactifications with QCD-like confining gauge sectors, one expects to find glueball excitations that on the type II side correspond to bound state solutions to the graviton wave-equation, localized (possibly around some domain wall structure) far inside the AdS-region. This is just like in the standard AdS/CFT correspondence. However, the new ingredient here is that the bulk gravity (used for representing the glueballs) and the gravity in the physical “boundary world” are directly linked and produced by the same type II string.
Low energy string phenomenology
The above set-up seems to open up the possibility of finding type II string compactification scenarios that in effect bypass the quark and gluon stage and immediately connect with the infra-red degrees of freedom in terms of mesons, baryons, etc. More generally, instead of the usual “top-down” approach to string phenomenology, one may contemplate a rather different “bottom-up” philosophy. Ideally, one could first try to connect string theory with low-energy physics (e.g. via standard non-compact AdS/CFT type technology and variations thereof) and then afterwards introduce gravity by trying to look for all possible consistent compactifications of this AdS-type space. Of course, there are still important technical as well as conceptual obstacles to deal with, such as supersymmetry breaking, unification, massless moduli fields, and allowing $`N`$ to be small rather than large, just to name a few.
The gauge hierarchy problem
Perhaps the deepest consequence of the above picture is that the renormalization group scale parameter of the four-dimensional world gets promoted to a real physical extra dimension. In this way all relative mass hierarchies, such as those between the weak scale and the Planck scale, are translated into relative separations in this extra direction. Since the conformal factor decays exponentially with $`y`$ with a planckian decay length, separations of say 50 or 100 times the Planck length generate scale factors as large as $`10^{15}`$ or $`10^{30}`$ . While perhaps one could argue that this gives a natural explanation of the large mass separations in our world, without giving a good reason of why string theory would necessarily choose this type of compactification geometry, one cannot claim it really solves the hierarchy problem. Instead, it seems more sensible to turn the conclusion around, and consider the existence of a large mass hierarchy as evidence supporting this type of compactification scenario as its most natural geometric realization.
The cosmological constant problem
The most vexing hierarchy problem is of course that of the cosmological constant. In essence it arises as a clash between experiment and the theoretical knowledge of general relativity and the renormalization group. There are several ingredients in the present set-up, however, such as (i) the reinterpretation of the RG scale as a holographic extra dimension, (ii) the intimate connection between the RG flow and the AdS-gravity equations of motion, (iii) the presence of a planckian size negative cosmological constant inside the AdS-region, that suggest a rather new perspective on this problem.
In this connection, it seems interesting to note that a naive AdS holographic description of the universe (as a boundary theory with finite temperature equal to that of the microwave background) identifies the total geometry with that of an AdS black hole with a horizon radius of about a millimeter, and thus with associated entropy of the order of $`10^{90}`$. A suggestive coincidence is that this horizon size is of similar magnitude as the thermal wavelength of the radiation itself, which perhaps indicates that holography may provide a link between the size and the total entropy of the universe .
Finally, we notice that this view of the universe as an AdS black hole has the striking consequence that any pair of points can be connected via a path of at most a couple of millimeter length. Indeed, one can simply first travel close to the black hole horizon, traverse the required distance in the $`x_{_{//}}`$-direction, and then go back. This does not mean, however, that one can travel at super-luminous speeds, since the travel time (as measured in our world) is delayed correspondingly by the red-shift factor.
Acknowledgements
This work is supported by NSF-grant 98-02484, a Pionier fellowship of NWO, and the Packard foundation. I would like to thank Costas Bachas, Micha Berkooz, Jan de Boer, Robbert Dijkgraaf, Savas Dimopoulos, Lisa Randall, Savdeep Sethi, Raman Sundrum, Erik Verlinde, and E. Witten for very helpful discussions.
|
no-problem/9906/nucl-th9906062.html
|
ar5iv
|
text
|
# High energy collisions of strongly deformed nuclei: An old idea with a new twist
## I Introduction
An “old idea” mentioned in the title is to select the head-on “long-long” collisions, by simply triggering on maximal number of participants $`N_p`$. Although it is kind of a folklore of the field, the only written material on it I found is a memo written by P.Braun-Munzinger. His estimates show that, due to larger A and $`deformation`$, the gain in energy density for UU over AuAu<sup>*</sup><sup>*</sup>*Below we use PbPb instead. can reach the factor 1.8. Although our study had found smaller numbers, new applications are proposed. They are mostly related with a different geometry of the collisions, the parallel one, which incorporates about the same energy density as available in central PbPb collisions, with significant deformation.
The general attitude of this study was to look how UU collisions can help to understand the existing open issues of the SPS heavy ion program. It certainly is of interest to RHIC program, although it is probably premature to discuss it now. The emphasis is made here on event selection, and other interesting options (like using targets with crystals with naturally aligned U) are not studied.
Let me start with outlining simple argument for head-on collisions. Representing U as a homogeneous ellipsoid with one long ($`R_l`$) and two short ($`R_s`$) semi-axis, one can related their ratio to deformation parameter $`\delta `$ used in nuclear physics (see e.g. )
$`{\displaystyle \frac{R_l}{R_s}}=({\displaystyle \frac{1+4\delta /3}{12\delta /3}})^{1/2}`$ (1)
For $`\delta _U.27`$ this ratio is 1.29, the basic deformation ratio to be used below.
It is convenient to think first in terms of “wounded” or “participant” nucleons first, a purely geometric concept, and only then consider real multiplicity (entropy) production. (We will follow such logic throughout the paper.) Let us thus start with comparing the density of participants per transverse area $`n_p=N_p/(\pi R_s^2)`$, for “long-long” collisions of the deformed nuclei vs the spherical one with the same A and $`R=(R_s^2R_l)^{1/3}`$. The effect only comes from reduction of the area, so
$`{\displaystyle \frac{n_p^{deformed}}{n_p^{spherical}}}={\displaystyle \frac{A^{deformed}}{A^{spherical}}}({\displaystyle \frac{R}{R_s}})^2`$ (2)
For A=238 we will use $`R_l=8.4,R_s=6.5,R=7.0fm`$, and so deformation alone increase the $`n_p`$ by 1.16. For U and Pb ($`R_{Pb}`$=6.78 fm in such model) one gets the participant density gain 1.24.
Transferring this into initial entropy density, one should recall that for (spherical) AA collisions
$`{\displaystyle \frac{dN}{dy}}(y=0)A^{1+\alpha }`$ (3)
with $`\alpha .12`$. So, assuming as usual that final multiplicity is proportional to the initial entropy density, we see that there is a correction to the simple idea that each participant nuclei gives the same (energy dependent) contribution to the spectrum. This non-zero $`\alpha `$ incorporates both (i) additional increase in multiplicity, and (ii) extra stopping (shift toward mid-rapidity): we are only interested in their combination dN/dy(y=0). Furthermore, it is natural to think that transverse dimensions of the system enter trivially here, and so these extra effects due to increased density. With this additional factor $`n_p^{3\alpha }`$ we obtain the UU/PbPb total initial density gainAt fixed time after the crossing, a la Bjorken. If more stopping means earlier equilibration, the available density is larger. $`1.24^{1+3\alpha }=1.34`$.
The main questions addressed in this paper are two-fold. One is to make some realistic estimates of the effect, not just for a particular configuration but for ensemble of events selected by some experimentally accessible criteria. The second is to outline possible applications of high energy collisions of the deformed nuclei.
## II UU collisions versus PbPb: the simulation
Simple Monte-Carlo program was written, which initializes nucleons inside nuclei and follow their paths through another one. Since we are not really interested in peripheral collisions, we did not included diffuse boundary of nuclei and used the ellipsoids described above. We also ignored probabilistic nature of the interaction, considering transverse distance between nucleons $`R<(\sigma _{in}/\pi )^{1/2}`$ to be sufficient reason to make both of them participants. So, the only source of fluctuationsAnd it is by no means assumed to be accurate account for fluctuations. are random positions of the nucleons inside the nucleus.
Spherical nuclei have only one parameter - impact parameter b - which in such classical treatment determines the mean number of participants. Deformed nuclei have in general 5 such parameters: b and 4 spherical angles $`\theta _i,\varphi _i,i=1,2`$ indicating the orientation of their longer axes at the collision moment. The main objective of the calculation is to see how well one can actually fixed those, by using experimentally available information.
Few words about our definitions. After all participant nuclei are identified, we calculated the tensor
$`T_{ij}=<x_ix_j>`$ (4)
in transverse plane, diagonalize it and find its eigenvalues $`R_+^2,R_{}^2`$. The density of participants we use below is defined as $`n_p=N_{part}/(\pi R_+R_{})`$ and deformation as $`R_+/R_{}`$.
If no selection is made, the density of the participants $`n_p`$ shown in 1 is very similar, except for the tail region. A gain in $`n_p`$ of the order of 16% seem indeed possible, with reasonable loss of statistics. However, if one introduces “centrality” cuts, the situation is different. We would define a fair “centrality cut” by restricting $`N_p>0.9(2A)`$, with corresponding A for for both cases. Triggering on large $`N_p`$ (or forward energy) one effectively eliminates spectators (and many complicated geometries).
The first striking feature one finds after such cut is that distribution of the deformations of the initial 2d ellipsoid $`R_+/R_{}`$ is very different: see figure Fig.2. The ratios as large as 1.35 are accessible, while in PbPb collisions all the “central” collisions are very spherical. The maximal deformation, not surprisingly, is of the order of deformation of U<sup>§</sup><sup>§</sup>§Note that it is comparable to what is obtained for medium b for spherical nuclei, but now reached for larger energy density and larger system. Those correspond to collisions with two long directions parallel to each other and orthogonal to the beam. It is about the same as is obtained for medium b for spherical nuclei, but at larger energy density and larger system (see below).
The joint distributions in participant density - deformation plane for (most central) UU and PbPb collisions are compared in Fig.3(a,b). The main message one can get from it is that strong correlation between the deformation and density, existing for spherical nuclei, is to some extent relaxed for UU.
How one can measure the 2d deformation of the initial conditions? The measured elliptic deformation of spectra of secondary particles, pions or nucleons, $`v_2`$, is proportional to this initial deformation (with EOS depending coefficient) and should have similar distribution. divide measured distribution over $`v_2`$ into more and less deformed.
Suppose now that one uses both control parameters, $`N_{part}`$ $`and`$ $`v_2`$. Is it really possible to select the particular geometries of the collisions we want? In Fig.4(a) one can see that it is to a significant effect correct: the $`lessdeformed`$ sample is rich in the region $`cos\theta 1`$, or in “head-on” collisions, while the $`moredeformed`$ collisions have none of them, and concentrate at small $`cos\theta `$. In Fig.4(b) one can see that the same selection of events correspond to the difference of the azimuthal angles $`\varphi _1\varphi _2`$ to be peaked around 0, or have a flat distribution, respectively.
Finally note also that the figures presented above have shown only the density of participants $`n_p`$: let us now recall the correction factor $`n_p^{3\alpha }`$ we discussed in the Introduction for the initial energy density. It can be seen as additional non-linear deformation of the axis, when going from $`n_p`$ to dN/dy. The contrast between UU and PbPb in dN/dy is larger by another 8%. Note also, that we have only discussed local quantities, like participant density $`n_p`$. For many applications the absolute size of the system is of similar and sometimes even larger importance.
## III Possible applications
### A Hard versus soft processes
The so called $`hard`$ processes include Drell-Yan dilepton production, two-jet events, heavy flavor production etc. Those are described in the first order by the parton model, and so are the simplest to treat for any geometry. For example, for head-on collisions discussed in the introduction, the output should be simply
$`\sigma _{hard}R_z^2R_t^2`$ (5)
Significant interest is related with various QCD $`corrections`$ to the parton model: some of them are related to initial state re-scattering, and some to the final state ones. Examples of the former are “shadowing” of nuclear structure functions and increase in parton $`p_t`$, the latter processes lead to the so called “jet quenching” or energy losses. The debates about their relative contributions continues: one, at one talk as a semi-joke I proposed to consider rectangular nuclei with different $`R_z,R_t`$ to separate them . Now we propose a particular realization of this idea.
How large level arm do we have, with UU collisions? For head-on collisions (discussed above) $`R_z=R_l.R_t=R_s`$, while for “parallel collisions” with maximal deformation it is the other way around, $`R_z=R_s`$ and $`one`$ of the $`R_t=R_l`$. So the ratio $`R_z/R_t`$ varies between about 1.3 and 1/1.3, or a span of 1.7. This may be enough to disentangle different mechanisms.
### B Ellipticity and EOS
In high energy collisions the shape of the “initial almond” for non-central collisions leads to enhanced “in-plane” flow, in direction of the impact parameter . It is very important because (as pointed out in (a)) it is developed $`earlier`$ than the radial one, and thus it may shed light on Equation of State (EOS) at early time. The particular issue is whether we do or do not have QGP at such time, at SPS or RHIC. Recent review of ellipticity one can find in .
Ellipticity is now measured by the asymmetry of the particle $`number`$, or $`v_i`$ harmonics defined as
$`{\displaystyle \frac{dN}{d\varphi }}={\displaystyle \frac{v_0}{2\pi }}+{\displaystyle \frac{v_2}{\pi }}\mathrm{cos}(2\varphi )+{\displaystyle \frac{v_4}{\pi }}\mathrm{cos}(4\varphi )+\mathrm{}`$ (6)
rather than asymmetry of the momentum distribution, which was used more at low energies. Furthermore, $`v_i`$ are often additionally normalized to the spatial asymmetry of the initial state (the “almond”) at the same b, $`\alpha _2=(R_+^2R_{}^2)/(R_+^2+R_{}^2)`$ in order to cancel out this kinematic factor and to see the response to asymmetry.
There are two ways to look at elliptic flow: by using collision energy dependence, or centrality dependence at fixed beam. The first method is more difficult, but it deals with fixed geometry and only slowly changing size of the system and relevant densities. For recent discussion of it see ref.3. The $`centrality`$ dependence is easier to measure, but interpretation is more complicated, because increasing b we make “almond” more elliptic, but also much smaller and thinner: eventually finite size corrections reduce pressure build-up. This is of crucial importance for the “QGP push” issue at SPS, since it should be present only for the largest densities available, and at small b the $`v_2`$ is very small and difficult to measure.
Preliminary NA49 data presented at QM99 may indeed indicate “the plasma push”, as enhancement of $`v_2`$ at small b. However much more work is clearly needed to understand this complex interplay of EOS and finite size effects.
U collisions discussed below in principle provide the means to decouple finite size and deformation issues. In particularly, the deformation of U (about 1.3) is actually enough to generate well measurable $`v_2`$ of the order of several percent, without significant loss in density.
### C $`J/\psi `$ suppression
One of the most intriguing observables, the $`J/\psi `$ suppression, is unique in its significant centrality and A-dependence. Here is not a place to discuss it in any details, but let me make few remarks about recent developments.
New NA50 data reported at QM99 have clarified the situation for the most central collisions: using now very thin target, it was found that 1996 data suffered from multiple interactions. In fact there is a significantly stronger suppression at small b. Furthermore, it seem like the two component picture, with separate $`\chi `$ and $`\psi `$ thresholds, really emerges. In view of this, it is desirable to increasing the density and/or the famous variable L: only deformed U provides an opportunity here at SPS.
In order to discriminate experimentally different ideas on the nature of $`J/\psi `$ suppression, we should be able to tell whether suppression happened quickly or need a longer time. The old idea is to study suppression dependence on $`p_t`$. Unfortunately changing $`p_t`$ we also change the kinematics: e.g. destruction by gluons or hadrons go better if $`p_t`$ grows.
Maybe better idea is to use azimuthal dependence of the suppression. Instantaneous suppression should show $`no`$ asymmetry, but if it takes few fm/c the anisotropy should show up. However, as for the elliptic flow, the problem is the initial “almond” at b$`<`$ 8 fm is not very anisotropic, and for larger b there is no anomalous suppression.
Here too the deformed U can help, providing (with trigger conditions discussed above) a variety of geometries, including the “parallel collisions”.
## IV Acknowledgements
I thank W.Henning, whose question forced me to think about this subject, and P.Braun-Munzinger and T.Ludlam for helping me retrieve the ref.1. This work is supported by US DOE, by the grant No. DE-FG02-88ER40388.
|
no-problem/9906/hep-lat9906008.html
|
ar5iv
|
text
|
# 1 Introduction and Algorithm
## 1 Introduction and Algorithm
Recent numerical investigations have revealed an interesting phase structure for the three dimensional Thirring model as a function of coupling $`g^2`$ and number of fermion species $`N_f`$. For sufficiently small $`N_f<N_{fc}`$, there is a continuous transition at $`g^2=g_c^2(N_f)`$ between a weak coupling phase in which chiral symmetry is realised in the limit $`m0`$, and a strong coupling one in which the symmetry is spontaneously broken. The critical indices characterising the transition are distinct for $`N_f=2`$ and $`N_f=4`$, suggesting that the continuum limits defined at the critical points define distinct interacting field theories. For $`N_f=6`$, however, simulations on a $`16^3`$ lattice with $`m=0.01`$ give tentative evidence for a first order chiral transition , implying both that no continuum limit exists in this case, and that $`4<N_{fc}<6`$.
Both the nature of the transition and the value of $`N_{fc}`$ are non-perturbative issues, inaccessible via either a standard perturbative expansion (which is non-renormalisable for $`d>2`$), or a $`1/N_f`$ expansion . There have, however, been analytic attempts to investigate the transition via the truncated Schwinger-Dyson (SD) equations . In the most systematic treatment , the SD equations are solved in ladder approximation in the the strong coupling limit; chiral symmetry breaking solutions are found for $`N_f<N_{fc}=128/3\pi ^24.32`$. Note that in a continuum approach the value of $`N_{fc}`$ manifests itself as the value of $`N_f`$ below which a non-trivial solution can be found, eg. by bifurcation theory. In a later paper , Sugiura extended the solution to finite $`g^2`$ for $`d(2,4)`$; for the case $`d=3`$ his predictions for the critical coupling as a function of $`N_f`$ read:
$`g^21:g_c^2`$ $`=`$ $`{\displaystyle \frac{2\pi ^2}{3}}N_f`$ (1)
$`g^21:g_c^2`$ $`=`$ $`6\left[\mathrm{exp}\left({\displaystyle \frac{1}{\omega }}\left(2\pi 4\mathrm{tan}^1\omega \right)\right)1\right]`$
with $`\omega ^2(N_f)={\displaystyle \frac{N_{fc}}{N_f}}1`$
The essentially singular behaviour seen in (1) as $`N_fN_{fc}`$ is consistent with the existence of a conformal fixed point . Using a different sequence of truncations of the SD equations, however, Hong and Park found chiral symmetry breaking solutions for all $`N_f`$, and predicted
$$g_c^2\mathrm{exp}\left(\frac{\pi ^2}{16}N_f\right),$$
(3)
which is manifestly non-perturbative in $`1/N_f`$.
The differing predictions of the SD approach motivate the use of lattice field theory methods to address this problem. In this Letter we present results of simulations performed with values of $`N_f`$ falling between the values $`N_f=2,4`$ and 6 explored in previous works, in an effort both to map out the phase diagram in more detail, and to constrain further the value of $`N_{fc}`$. To start with, let us define the lattice action:
$`S`$ $`=`$ $`{\displaystyle \frac{1}{2}}{\displaystyle \underset{x\mu i}{}}\overline{\chi }_i(x)\eta _\mu (x)\left[(1+iA_\mu (x))\chi _i(x+\widehat{\mu })(1iA_\mu (x\widehat{\mu }))\chi _i(x\widehat{\mu })\right]`$
$`+m{\displaystyle \underset{xi}{}}\overline{\chi }_i(x)\chi _i(x)+{\displaystyle \frac{N}{4g^2}}{\displaystyle \underset{x\mu }{}}A_\mu ^2(x)`$
$``$ $`{\displaystyle \underset{xyi}{}}\overline{\chi }_i(x)_{ij}[A](x,y)\chi _j(y)+{\displaystyle \frac{N}{4g^2}}{\displaystyle \underset{x\mu }{}}A_\mu ^2(x),`$
where the indices $`i,j`$ run over $`N`$ flavors of staggered lattice fermion. The $`A_\mu `$ are real auxiliary vector fields defined on the lattice links; gaussian integration over $`A_\mu `$ yields a form of the action with explicit four-fermion couplings (other variants of the lattice-regularised Thirring model are discussed in ). Analysis of the spin-flavor content of staggered lattice fermions reveals that in 3 dimensions the number of continuum four-component physical flavors is given by $`N_f=2N`$. Therefore use of the above action limits us to even $`N_f`$. If, however, the fermions are integrated out, then the resulting effective action for the $`A_\mu `$ fields reads:
$$S_{eff}=\frac{N_f}{8g^2}\underset{x\mu }{}A_\mu ^2(x)\frac{N_f}{2}\mathrm{ln}\text{det}[A].$$
(5)
The effective action, though non-local, has an analytic dependence on $`N_f`$, and hence can be employed for odd or non-integer values.<sup>1</sup><sup>1</sup>1Note that the form of (1) permits even-odd partitioning, which means that we can simulate integer powers of $`\text{det}`$, rather than $`\text{det}^{}`$, with a local action. It can be simulated using a hybrid molecular dynamics algorithm, which evolves the $`\{A\}`$ configuration through a fictitious time $`\tau `$ by a combination of microcanonical and Langevin dynamics. In the limit of timestep $`\mathrm{\Delta }\tau 0`$ the $`\{A\}`$ are distributed according to the equilibrium ensemble. In this work we have implemented the R algorithm of Gottlieb et al ; in this case the systematic errors are $`O(N^2\mathrm{\Delta }\tau ^2)`$ . We found that on a $`12^3`$ lattice a value $`\mathrm{\Delta }\tau =0.01`$ was sufficient, with a mean interval $`\overline{\tau }=1.0`$ between refreshments. Our results typically arise from averages over 500 units of $`\tau `$. Fig. 1 shows a comparison between chiral condensate $`\overline{\chi }\chi `$ data obtained for $`N_f=4`$, for various values of the bare mass $`m`$, using both the hybrid algorithm and a hybrid Monte Carlo algorithm , which in principle is free from systematic error.
In Sec. 2.1 we will present results from simulations with variable $`N_f`$ at a fixed coupling $`1/g^2=1.0`$. The phase transition is located and found to fall at a value of $`N_f`$ intermediate to the cases studied in earlier work . Fits to a power law equation of state ansatz yield critical exponents consistent with this picture. In Sec. 2.2 we present results obtained with variable $`g^2`$ at a fixed $`N_f=5`$, the goal being to determine if the transition remains continuous, or whether there is evidence for coexisting phases signalling a first order transition, as observed for $`N_f=6`$ . Our conclusions, and the resulting phase diagram, are given in Sec. 3.
## 2 Numerical Simulations
### 2.1 Fixed $`1/g^2=1.0`$
Using the algorithm as described above we performed simulations with variable $`N_f`$ on a $`12^3`$ lattice at fixed $`1/g^2=1.0`$, with bare mass $`m=0.04,0.03,0.02`$. In Fig. 2 we plot the transverse and longitudinal susceptibilities $`\chi _{t,l}`$, using the definitions given in . As described there, $`\chi _t`$ can be calculated either from the integrated two-point pion propagator, or from $`\overline{\chi }\chi `$ via the axial Ward identity: the two methods give compatible results, the latter being the less noisy. These quantities are related to the inverse square masses of respectively the pseudoscalar and scalar bound states. For $`N_f3.6`$ they are roughly equal, indicating that the scalar and pion states are degenerate, and chiral symmetry is realised. For $`N_f2.8`$, in contrast, $`\chi _t\chi _l`$, indicating that the pion has become light as expected for a pseudo-Goldstone mode in a phase of broken chiral symmetry. Fig. 2 thus offers evidence for a chiral symmetry breaking phase transition in the region $`N_f3.0`$.
Fig. 3 shows results for the order parameter $`\overline{\chi }\chi `$ as a function of $`N_f`$, together with fits to a renormalisation-group inspired equation of state of the form
$$m=At\overline{\chi }\chi ^{\delta 1/\beta }+B\overline{\chi }\chi ^\delta ,$$
(6)
where $`\delta ,\beta `$ are conventional critical exponents, and the parameter $`t`$ expresses the distance from criticality. In previous work at fixed $`N_f`$, variable $`1/g^2`$ we have defined $`t=1/g^21/g_c^2`$; here we define
$$tN_fN_f,$$
(7)
where $`N_f`$ denotes the value of $`N_f`$ at the transition. Experience from previous studies has shown that equations of state of the sort (6) provide adequate descriptions of the data in the vicinity of the transition, and that the resulting fitted exponents do not differ much from those extracted by more sophisticated finite volume scaling analyses. Also following previous work we fix $`\beta `$ by the requirement $`\delta 1/\beta =1`$, in order to stabilise the fitting procedure; we refer to the resulting four parameter fit as fit I. The results obtained by fitting data for $`N_f[3.0,3.6]`$ are given in Table 1. Also shown is a five parameter fit II over the same range for which the constraint on $`\beta `$ was relaxed.
A number of comments are in order. First, the constraint of fit I appears to be well satisfied by the data, since from fit II the fitted value of $`\delta 1/\beta =0.97(8)`$. Second, the fitted values of the exponents fall between those found for $`N_f=2`$ and $`N_f=4`$ from runs with variable $`1/g^2`$ , being numerically closer to the $`N_f=4`$ case: the results for these from comparable fits are summarised in Table 2. This supports the idea that $`g_c^2(N_f)`$ is a critical line of fixed points along which critical exponents vary smoothly, with the numerical value of the exponents independent of the direction from within the $`(g^2,N_f)`$ plane from which the critical line is approached. Third, as a note of caution, it should be noted that the criterion for the range of $`N_f`$ over which the fit was made was purely on the basis on minimising the $`\chi ^2`$ per degree of freedom, as in . Inspection of Fig. 3 suggests that the resulting fits may not describe the broken phase particularly well, and indeed fits which included smaller $`N_f`$ values resulted in systematically larger values of $`\delta `$. Our philosophy is to regard (6) merely as a phenomenological decription of the data; results from comparable fitting procedures applied to different models can thus be used to establish trends regardless of whether the true equation of state is ultimately of the form (6) or not.
### 2.2 Fixed $`N_f=5`$
With the confidence that simulations with a non-integer number of staggered flavors yield results consistent with our expectations for $`N_f3`$, we then conducted a series of simulations on $`12^3`$ lattices with similar parameters but this time with $`N_f=5`$, a value which previous studies suggest is close to $`N_{fc}`$. Our results for $`\overline{\chi }\chi `$ are shown in Fig. 4. We have included results from simulations with $`m=0.01`$, although it is likely that these are badly finite-volume affected; however, they do reveal evidence of discontinuous behaviour for $`1/g^20.5`$. Fits to the equation of state (6), with the parameter $`t`$ now given by $`1/g^21/g_c^2`$, were attempted, but were less successful than those of the previous section, in general either resulting in an unacceptably large $`\chi ^2`$ or not reaching convergence. In any case, the resulting fitted parameters are strongly dependent on the fit range, and inclusion of the $`m=0.01`$ points makes fit quality much worse. Our best results, coming from fitting data with $`m[0.04,0.02]`$, $`1/g^2[0.4,0.6]`$, are shown in Tab. 3.
The fitted values of $`\delta `$ and $`\beta `$ are consistent with the trends of Tab. 2, although it is interesting to note that the fitted value of $`\delta 1/\beta =0.6(2)`$, indicating that in contrast to previous fits the data do not satisfy the usual constraint in this case. It could be argued that this disfavours the fit, since the combined assumptions of the degeneracy of scalar and pseudoscalar bound states in the chiral limit of the symmetric phase, plus the applicability of a power-law equation of state (6), naturally predict $`\delta 1/\beta =1`$ .
To gain more insight we present the same data in the form of a Fisher plot (ie. $`\overline{\chi }\chi ^2`$ vs. $`m/\overline{\chi }\chi `$) in Fig. 5. This plot is devised to yield trajectories of constant $`1/g^2`$ which intersect the vertical axis in the chiral limit in the broken phase, and the horizontal axis in the symmetric phase. Departures from the mean field indices $`\delta =3,\beta =\frac{1}{2}`$ are revealed as curvature in the lines, and departures from $`\delta 1/\beta =1`$ are revealed by variations in the sign of the curvature between the two phases. If we ignore the $`m=0.01`$ points, which are those closest to the horizontal axis, then the plot suggests a critical $`1/g_c^20.5`$, with tentative evidence for $`\delta 1/\beta <1`$. Inclusion of the low mass points, however, suggests an accumulation of the constant coupling trajectories around a line which if continued would intercept the horizontal axis. This is similar to the Fisher plot for $`12^3`$ data for the model with $`N_f=6`$, shown in fig. 6 of , and suggests similarities between the two cases. For $`N_f=6`$, tentative evidence for tunnelling between coexisting vacua on simulations with $`m=0.01`$ on a $`16^3`$ lattice in the critical region was presented in , consistent with the chiral transition being first order. We tested this possibility by performing long $`N_f=5`$ simulations on a $`16^3`$ lattice for $`m=0.01`$ and $`1/g^2=0.45`$, 0.47 (these runs requiring $`\mathrm{\Delta }\tau =0.002`$): a time history for the latter is shown in Fig. 6.
Whilst the fluctuations are large, with excursions of $`O(50\%)`$ about the mean, there is no convincing evidence for a two-state signal. Double gaussian fits to the histograms of the binned data, which revealed coexisting states via twin peaks in , were this time statistically indistinguishable from simple single gaussians.
In summary, whilst there seems to be clear evidence for a chiral transition with $`1/g^2=0.50(5)`$, the simulations performed to date are unable to determine the order. Fits based on the assumption of a continuous transition, whilst falling into the broad trend of existing data, are unsatisfactory and fail to satisfy the constraint $`\delta 1/\beta =1`$. A search for evidence of metastability consistent with a first order transition, on the other hand, proved negative.
## 3 The Phase Diagram
In this Letter we have used a fresh simulation algorithm to explore non-integer numbers of staggered fermion flavors, and found results broadly consistent with earlier studies, namely that for $`N_f3`$ the transition is continuous with critical exponents intermediate between those of $`N_f=2`$ and $`N_f=4`$ models, and that for $`N_f=5`$ the order of the transition is difficult to determine, but shares features in common with the $`N_f=6`$ transition, believed to be first order. This change in order of the transition with increasing $`N_f`$ has also been observed in studies of $`\text{QED}_4`$ . The results also support the view that the partition function can be regarded as analytic in $`N_f`$, and that the non-locality of the fermionic action for $`N_f`$ odd has no severe consequences. In this final section we collect together these new results with the best estimates obtained from previous finite volume scaling studies in , to plot the phase diagram of the Thirring model on the $`(1/g^2,N_f)`$ plane in the chiral limit. The result is shown in Fig. 7. The shading of the points indicates the suspected order of the transition. The picture we have established is consistent with a critical line of fixed points, along which critical exponents vary smoothly, up to some $`N_{fc}5`$, whereupon the transition becomes first order.
It is work comparing our findings to the analytic approximations of Refs. and . Certainly the curvature of our critical line $`1/g_c^2(N_f)`$ is consistent with both predictions (1) and (3). A detailed comparison with the former, which predicts a conformal fixed point at $`N_f=N_{fc}4.32`$ is hampered by the difficulties in identifying the strong coupling limit in the lattice model, since there is an additive renormalisation relating $`1/g_{CONT}^2`$ to $`1/g_{LATT}^2`$ due to non-conservation of the interaction current in the lattice model . Therefore we are forced to identify $`N_{fc}`$ with that value at which the transition becomes first order, at which point no continuum limit exists for the lattice model. Our estimate for $`N_{fc}`$ is thus roughly consistent; however, the trends in the exponents extracted from the lattice studies, summarised in Table. 4 do not support a conformal fixed point. At such a point we expect the exponent $`\delta `$ to take the value 1, as for an asymptotically-free theory. The exponent $`\eta `$ is related to the anomalous dimension of the composite operator $`\overline{\chi }\chi `$ by $`\eta =d2\gamma _{\overline{\chi }\chi }`$ , which in turn is predicted at the fixed point to be given by :
$$\gamma _{\overline{\chi }\chi }=\frac{d2}{2}.$$
(8)
At a conformal fixed point $`\eta `$ should thus take the value 2.
Finally, it is interesting to note that a fit of the form
$$1/g_c^2=A+B\mathrm{exp}\left(\frac{\pi ^2}{16}N_f\right),$$
(9)
motivated by Eq. (3) adjusted to allow for the renormalisation of $`1/g^2`$, can be made to the data shown in Fig. 7; the fit is particularly successful if the $`N_f=2`$ point is excluded, in which case the resulting coefficients are $`A=0.28(2)`$, $`B=4.7(2)`$, with $`\chi ^2`$/d.o.f.=0.6. This suggests that the scenario of , in which chiral symmetry breaking is predicted for all $`N_f`$ and hence that $`N_{fc}=\mathrm{}`$, cannot be excluded by the lattice studies to date.
## Acknowledgements
This project was supported in part by the TMR-network “Finite temperature phase transitions in particle physics” EU-contract ERBFMRX-CT97-0122. We are greatly indebted to the CRT Computer Center of ENEL (PISA) for collaboration in the use of their CRAY YMP-2E. One of us (B.L.) would like to thank the Physics Department of Swansea for kindly hospitality.
|
no-problem/9906/quant-ph9906097.html
|
ar5iv
|
text
|
# References
Quantum state diffusion, measurement and second quantization
by
Ian C. Percival
Department of Physics
Queen Mary and Westfield College, University of London
Mile End Road, London E1 4NS, England
Abstract
Realistic dynamical theories of measurement based on the diffusion of quantum states are nonunitary, whereas quantum field theory and its generalizations are unitary. This problem in the quantum field theory of quantum state diffusion (QSD) appears already in the Lagrangian formulation of QSD as a classical equation of motion, where Liouville’s theorem does not apply to the usual field theory formulation. This problem is resolved here by doubling the number of freedoms used to represent a quantum field. The space of quantum fields is then a classical configuration space, for which volume need not be conserved, instead of the usual phase space, to which Liouville’s theorem applies. The creation operator for the quantized field satisfies the QSD equations, but the annihilation operator does not satisfy the conjugate eqation. It appears only in a formal role.
980615, QMW-PH-99-?? Submitted to Phys. Lett. A
1. Introduction
Quantum measurement is a physical process by which the state of a quantum system influences the value of a classical variable. The meaning of quantum measurement here includes any such process, including laboratory measurements, but also other, very different, processes, such as the cosmic rays that produced small but detectable dislocations in mineral crystals during the Jurassic era, and the quantum fluctuations in the early universe that are believed to have caused today’s anisotropies in the universal background radiation and in galactic clusters .
Since Bohr and Einstein it has been recognized that it is difficult to represent quantum measurement as a dynamical process . Quantum theories that attempt it have problems with unitarity. These include those theories which depend on quantum state diffusion (QSD), the principal subject of this letter .
According to Bohr, the result of a measurement is influenced by the conditions of the measurer. Dynamical theories of quantum measurement seek a dynamical process for this influence. Quantum measurement dynamics does not follow from the unitary dynamics of Schrödinger or Heisenberg, nor from the quantum dynamics of fields, strings or branes. In a unified physics, they must be reconciled.
The methods of quantum field theory have been used for measurement dynamics before, but this letter deals with the apparent contradiction between the principles upon which they are based. Quantum field theory is unitary, whereas quantum state diffusion is not. Further, it has long been known that quantum measurement is nonunitarity .
Here we trace the problem to the classical dynamics of a de Broglie wave, considered as a classical field, in particular to the violation of Liouville’s theorem by the measurement process. This makes it necessary to reformulate the classical dynamics of the field differently, making quantum measurement dynamics a nonlinear field theory of a special type.
The experimental consequences of this theory are the same as the standard results of quantum state diffusion when applied to the dynamics of measurement, which is indistinguishable from the results of the usual interpretation of nonrelativistic quantum theory for past and current experiments, though not necessarily for all future experiments . But the classical field theory of quantum state diffusion has some unusual features, in particular that the de Broglie wave is defined by a point in a configuration space, not in a phase space. There are canonical conjugate momenta, but their role is probably formal, rather than physical, and they are not the same as the complex conjugate amplitudes. The reason for this violation of one of the basic principles of field quantization is given in the next section.
This letter is confined to the nonrelativistic formulation of the dynamics of the single-particle de Broglie waves of QSD, first as a classical field, then as a quantized field.
2 Quantum state diffusion
QSD represents measurement dynamics as a continuous stochastic process, in which the state vector is the solution of an Itô stochastic differential equation . This is expressed in terms of a complex differential stochastic fluctuation $`\mathrm{d}\xi `$ with equal and independent fluctuations in its real and imaginary parts, so that
$$\mathrm{Md}\xi =0,\mathrm{M}(\mathrm{d}\xi )^2=0,\mathrm{M}|\mathrm{d}\xi |^2=\mathrm{d}t.$$
(1)
where $`\mathrm{M}`$ represents the mean over an ensemble.
Suppose a system with state $`|\psi `$ has Hamiltonian $`𝐇`$, and the dynamical variable $`G`$ with Hermitean operator $`𝐆`$ is being measured. According to QSD, measurement is a very rapid diffusion in state space, whose rate is given by a real factor $`c`$ Then the quantum state diffusion equation is
$$\frac{\mathrm{d}|\psi (t)}{\mathrm{d}t}=(i/\mathrm{})𝐇|\psi (t){\scriptscriptstyle \frac{1}{2}}c^2𝐆_\mathrm{\Delta }^2|\psi (t)+c𝐆_\mathrm{\Delta }|\psi (t)\frac{\mathrm{d}\xi }{\mathrm{d}t},$$
(2)
where $`𝐆_\mathrm{\Delta }=𝐆\psi (t)|𝐆|\psi (t)`$ is the shifted $`\psi `$-dependent $`𝐆`$-operator whose expectation for the current state $`|\psi (t)`$ is zero. For simplicity, we will absorb the constant $`c`$ into $`𝐆`$. The equation is nonlinear, but the norm of $`\psi (t)`$ is preserved. The stochastic coefficient $`\mathrm{d}\xi /\mathrm{d}t`$ is a highly singular function of time, whose singular properties are handled by the Itô calculus, but they need not concern us here. What is important is that it is a stochastic function of time representing complex Gaussian white noise.
For laboratory experiments, the diffusion is so fast that the state appears to jump between states on a time scale far shorter than the other time scales of the system, in particular those of the Hamiltonian. However, according to quantum state diffusion theory, this is a limiting case. For small isolated quantum systems the diffusion is so slow that it has not been detected. This is the other limiting case.
The details of quantum state diffusion as a theory of quantum measurement are given in .
3 Classical measurement dynamics of the scalar field
For simplicity, first consider a free particle with no measurement, in a one-dimensional box with a bounded energy, so that there is a finite number $`N`$ of discrete states. In energy representation, with energies $`E_j=\mathrm{}\omega _j`$, the corresponding complex amplitudes
$$\psi _j=\frac{1}{\sqrt{2}}(q_j+ip_j)$$
(3)
satisfy
$$i\dot{\psi }_j=\omega _j\psi _j,\text{so that}$$
(4)
$$\dot{q}_j=\{q_j,H\}=\omega _jp_j,\dot{p}_j=\{p_j,H\}=\omega _jq_j,\text{with}\text{ }H=\underset{j}{}\frac{\omega _j}{2}(p_j^2+q_j^2),$$
(5)
which are are Hamilton’s equations for $`N`$ oscillators with real canonically conjugate configuration and momentum coordinates $`q_j,p_j`$. Schrödinger evolution of the wave produces a unitary transformation in the state space, a generalized rotation on the unit sphere, identical to the motion of the phase point in the phase space of the oscillators. The unit sphere is an energy shell of a phase space, Liouville’s theorem is satisfied, so the phase space density for a continuous distribution of systems is conserved. Second quantization of the field amplitudes follows just as first quantization for the oscillators.
The same applies formally for the complex configuration coordinate and its canonical conjugate momentum
$$\psi _j=\frac{1}{\sqrt{2}}(q_j+ip_j)\text{and}\text{ }i\psi _j^{}=\frac{1}{\sqrt{2}}(iq_j+p_j).$$
(6)
In this representation, the equations of motion for the configuration and momentum coordinates are independent:
$$\dot{\psi }_j=i\omega _j\psi _j,\dot{\psi }_j^{}=i\omega _j\psi _j^{}\text{with}\text{ }H=\underset{j}{}\omega _j\psi _j^{}\psi _j=\underset{j}{}(i\omega _j)(i\psi _j^{})\psi _j.$$
(7)
The dynamics of quantum measurement is very different. The norm of the state is preserved, so it is confined to the unit sphere in state space, but the density of states on the surface of the unit sphere is not preserved. Measurement of the energy, for example, puts the system into one of the energy eigenstates, so a continuous distribution over the unit sphere of an ensemble of systems is reduced towards a set of at most $`N`$ points. A uniform distribution over the unit sphere in this finite-dimensional state space evolves towards a set of equal $`\delta `$-distributions at each of the energy eigenstates. The total volume of the surface of the sphere is reduced towards zero. Liouville’s theorem is violated with a vengeance, so the space of quantum states of the particle cannot be the phase space of any classical Hamiltonian system.
However, the state space can be treated as a configuration space. There is no conservation of volume in this configuration space, so a Lagrangian or Hamiltonian measurement dynamics is possible. The equations of motion for the complex configuration coordinates of the oscillators are independent of the equations for the conjugate momenta, as they are when there is no measurement, but the conjugate momenta are no longer the same as the complex conjugates of the configuration coordinates.
4 Lagrangian theory of free-field QSD
The configuration space trajectory for a time-independent dynamical system with two configuration coordinates $`q,q^{}`$ is stationary for the action integral
$$S=_{t_0}^{t_1}dtL(q,q^{})$$
(8)
Equivalent Lagrangians have action integrals that give the same equations of motion.
Before treating the QSD equations, consider a simpler classical model. The equations of motion for a dynamical system with equivalent Lagrangians
$$L=q^{}\dot{q}+q^{}f(q),L^{}=q\dot{q}^{}+q^{}f(q)$$
(9)
are
$$\dot{q}=f(q),\dot{q}^{}=q^{}f(q)/q$$
(10)
and the momenta conjugate to $`q,q^{}`$ are
$$p=L/\dot{q}=q^{},p^{}=L/\dot{q}^{}=q.$$
(11)
We can identify $`q^{}`$ and $`p`$, write them both as $`p`$, and similarly we can write $`p^{}`$ as $`q`$. The primed coordinates are then no longer needed, as in the usual canonical theory of quantum fields. But a momentum then appears in the Lagrangian, as it does in quantum field theory, but which is not normally allowed in classical dynamics. The quantum theory of a complex amplitude of a free linear field has just this form with $`f(q)=i\omega q`$, where $`q`$ is complex, and $`q^{}=q^{}`$, its complex conjugate, which is treated as an independent canonical coordinate, giving
$$L=q^{}\dot{q}+i\omega qq^{},L^{}=q\dot{q}^{}+i\omega qq^{}.$$
(12)
We can now identify $`q^{}`$, $`q^{}`$ and $`p`$ in the resultant Lagrange equations. This is consistent for this case because the equation of motion for $`q^{}`$ is just the complex conjugate of the equation of motion for $`q`$. Identifying $`q^{}`$, $`q^{}`$ and $`p`$ is a only a formal problem for the theory of quantum fields and is very convenient in practice..
However if $`f(q)cq`$, the equation of motion for $`q^{}=q^{}`$ is not the complex conjugate of the equation for $`q`$, so even if the identification is made at some initial time, it will no longer hold for later times. This is what happens for the Lagrangian theory of the wave $`\psi `$ in QSD. For QSD we start with four independent configuration coordinates $`q,q^{},q^{},q_{}^{}{}_{}{}^{}`$. The starred coordinates $`q^{},q_{}^{}{}_{}{}^{}`$ are complex conjugates of the coordinates $`q,q^{}`$, but the primed coordinates are not conjugate momenta. In QSD for a Schrödinger field, the configuration coordinates corresponding to $`q,q^{}`$ are $`\psi _j,\psi _j^{}`$.
The Lagrangian formulation of QSD for the measurement of a dynamical variable $`G`$ of a particle follows from this approach. We derive the equations for the fields themselves, not the field components. It is convenient to express the total action, which is a function of the configuration coordinates $`\psi ,\psi ^{},\psi ^{},\psi _{}^{}{}_{}{}^{}`$, as twice the real part of a complex Lagrangian $`L_\mathrm{c}`$, which depends on all the configuration coordinates except the last. Consequently $`L_\mathrm{c}^{}`$ is independent of $`\psi ^{}`$ and makes no contribution to Lagrange’s equation for $`\psi `$.
The action is
$$S=dt(L_\mathrm{c}+L_\mathrm{c}^{}),\text{where}$$
(13)
$$iL_\mathrm{c}=\mathrm{d}^3x\psi ^{}\dot{\psi }i\mathrm{d}^3x\psi ^{}H\psi +\mathrm{d}^3x\psi ^{}Q\psi \text{and}$$
(14)
$$Q=Q(\psi ^{},\psi )={\scriptscriptstyle \frac{1}{2}}\left(G\mathrm{d}^3x\psi ^{}G\psi \right)^2+\left(G\mathrm{d}^3x\psi ^{}G\psi \right)\mathrm{d}\xi /\mathrm{d}t.$$
(15)
The Lagrangian of equation (14) is the form suitable for varying with respect to $`\psi ^{}`$ and $`\psi _{}^{}{}_{}{}^{}`$. For the variation with respect to $`\psi `$ and $`\psi ^{}`$, we vary the equivalent Lagrangian obtained by partial integration.
The variation of $`S`$ with respect to $`\psi ^{}`$ is straightforward, the $`L_\mathrm{c}^{}`$ term does not contribute, and $`L_\mathrm{c}`$ gives the QSD equation for $`\psi `$, and the derivative of $`L_\mathrm{c}`$ with respect to $`\dot{\psi }`$ gives the definition of the momentum $`p_\psi `$. Together they make Hamilton’s equations for $`\psi `$:
$$\dot{\psi }=iH\psi +Q\psi ,\text{(QSD),}\text{ }p_\psi =i\psi ^{}$$
(16)
The corresponding operations with $`\psi _{}^{}{}_{}{}^{}`$ and $`\dot{\psi }^{}`$ give Hamilton’s equations for $`\psi ^{}`$ and $`\psi _{}^{}{}_{}{}^{}`$ and as the action is real, the equations are just the complex conjugates of those for $`\psi `$ and $`\psi ^{}`$.
$$\dot{\psi }^{}=iH\psi ^{}+Q^{}\psi ^{},\text{(QSD),}\text{ }p_\psi ^{}=i\psi _{}^{}{}_{}{}^{}.$$
(17)
Since these are complex conjugate equations, $`\psi ^{}`$ remains the complex conjugate wave for all time.
The equation for $`\dot{\psi }^{}`$ is given by the variation with respect to $`\psi `$. It is not nearly so simple as the QSD equation, as it involves both the $`L_\mathrm{c}^{}`$ and $`L_\mathrm{c}`$ terms. The additional terms ensure that even if initially $`\psi ^{}=\psi ^{}`$, it does not remain so for later times, as it does in the absence of measurement. In this way the (probably nonphysical) momentum space of $`\psi ^{},\psi _{}^{}{}_{}{}^{}`$ can carry away the phase space volume that is lost by the motion of the state in the configuration space of $`\psi ,\psi ^{}`$.
5 Second quantization
Before going into the analysis, we note an essential difference between this theory and the usual theory of second quantization. In both, the most important operator is the field creation operator $`\psi ^{}`$.
In the usual classical field theory, (we ignore the factors $`i`$), the complex conjugate of the field amplitude $`\psi ^{}`$ is the canonical conjugate of the amplitude $`\psi `$, and both are important for quantization. The operator $`\psi `$ is the field annihilation operator, whose properties come from its commutation relations, and these in turn are derived from the commutation relations for the conjugate momentum. $`\psi ^{}`$ is also the Hermitian conjugate operator of $`\psi `$. Both these roles are held by the same operator, which greatly simplifies the theory.
For the quantization of the QSD equations, these roles are separated. The annihilation operator corresponding to the creation operator $`\psi ^{}`$ is the conjugate momentum operator $`\psi _{}^{}{}_{}{}^{}`$, which is not the same as the Hermitean conjugate operator $`\psi `$. For convenience we denote the annihilation operator by
$$\psi _{}^{}{}_{}{}^{}=\psi ^0.$$
(18)
The quantum state of a Schrödinger field with $`n`$ particles is given by operating $`n`$ times on the vacuum with the creation operator $`\psi ^{}(t)`$. The Heisenberg equations of motion for $`\psi ^{}(t)`$ are the same as Hamilton’s equations (17) above. But because of the decoupling between the coordinate and momentum equations, only the first of these equations is important. As far as we know, the second Heisenberg equation has no physical significance. The annihilation operator $`\psi ^0(t)`$ has an important formal role in deriving Heisenberg equations, just as in ordinary quantum field theory, but since only the creation operators are needed to obtain a field from the vacuum, the annihilation operators appear to have no other physical significance than this.
Because of the decoupling of the coordinate and momentum equations of motion in the Heisenberg equations, the relatively complicated Heisenberg equation for $`\psi ^0`$ is not needed to get the physical field. However, without the momenta $`\psi ^{}`$ and $`\psi ^0`$, there would be no Liouville theorem for the classical formulation of QSD, and consequently no unitarity for the quantized field. The price of unitarity is additional fields that are not physical, as far as we know.
6 Discussion
There are two possible approaches to the quantum field theory of QSD. In the first, which is treated here, the QSD equations for the de Broglie wave are derived from a classical Lagrangian, as a prelude to the second quantization. In the second, a quantum state diffusion term is added to the second quantized equations of motion for the particle.
The physical difference between these two approaches, is that in the first approach, the Liouville’s equation is satisfied in the extended phase space of the de Broglie wave, so the quantized theory can be unitary. This is consistent with the unitarity of standard quantum field theory, but the price is the introduction of conjugate momenta that appear to play no physical role. In the second approach, any diffusion terms will destroy the unitarity of the field theory, which makes it very difficult or impossible to reconcile with the modern theory of fields, strings and branes.
|
no-problem/9906/cond-mat9906450.html
|
ar5iv
|
text
|
# Quantum Hall effect at low magnetic fields
\[
## Abstract
The temperature and scale dependence of resistivities in the standard scaling theory of the integer quantum Hall effect is discussed. It is shown that recent experiments, claiming to observe a discrepancy with the global phase diagram of the quantum Hall effect, are in fact in agreement with the standard theory. The apparent low-field transition observed in the experiments is identified as a crossover due to weak localization and a strong reduction of the conductivity when Landau quantization becomes dominant.
\]
The behavior of the quantum Hall effect (QHE) at low magnetic fields has attracted a lot of attention in recent years, both experimentally and theoretically. Of particular interest has been the fate of the critical states responsible for the transitions between the different quantized Hall plateaus and the form of the global phase diagram of the QHE . Recent experiments have been interpreted as being incompatible with the global phase diagram based on the levitation of critical states . At high magnetic fields, such that the Landau level separation is much larger than the disorder broadening of the Landau bands, these critical states are situated near the centers of the Landau bands. They separate phases with different, quantized values of the Hall conductivity. As the Hall conductivity is constant throughout these phases, the critical energies cannot just terminate when magnetic field or disorder is changed. They either can move to infinite energy or they can terminate when they intersect another critical state with the opposite change in Hall conductivity. Both of these scenarios are realized: the critical energies at the centers of the Landau bands move to infinite energy as the magnetic field becomes infinitely strong, while the termination of critical states has been observed in lattice models, where due to the band structure, states with negative Hall conductivity exist .
Since in the absence of a periodic potential, no states with negative Hall conductivity exist, Khmel’nitskii and Laughlin have argued, that the existence of critical states at high magnetic fields can only be reconciled with the absence of extended states at zero magnetic field, predicted by the scaling theory of localization in two dimensions , if the critical states float above the Fermi energy, when the magnetic field is decreased towards zero. The scaling theory of the integer QHE predicts the Hall conductivity to be quantized at $`ne^2/h`$ between the critical energies, with integer $`n`$, and to be $`(n+1/2)e^2/h`$ at the critical energies. The dissipative conductivity vanishes, except at the critical energies, where it takes on values of the order of $`e^2/h`$. In Fig. (1) the levitation of the critical states is sketched and the resulting behavior of the conductivities and resistivities is shown. Note, that it is not possible to predict the behavior of the resistivities in the phase with zero Hall conductivity, as their values depend on the way how the conductivities tend to zero.
Since the QH plateau transitions are quantum phase transitions, the discussion presented above is concerned with the phase diagram of an infinite system at zero temperature. Experimentally, only finite systems at finite temperature are accessible. In numerical simulations, also the restriction to finite systems applies. When comparing the result of experiments with the predictions of scaling theory, it is therefore imperative to consider the effects of finite temperatures and/or finite system sizes. While this was appreciated in early theoretical and experimental work , it has wandered out of the focus of much of the recent work. It is the purpose of this paper to show that the experiments on the low-field QHE can be understood within the standard scaling theory, obviating the need for a more exotic explanation. We will restrict our attention to the integer QHE and consider interaction effects only on the level of weak localization corrections.
We will start our argument by considering the appropriate starting points for a renormalization of the conductivities, the bare conductivities $`\sigma _{ij}^0`$, corresponding to short length scales or high temperatures . At high temperatures and low magnetic fields, quantum effects are negligible and the conductivities can be calculated from kinetic equations to give the Drude expressions
$`\sigma _{xx}^0`$ $`=`$ $`{\displaystyle \frac{\sigma _0}{1+(\omega _c\tau )^2}},`$ (2)
$`\sigma _{xy}^0`$ $`=`$ $`\omega _c\tau \sigma _{xx}^0,`$ (3)
with $`\sigma _0=e^2n_c\tau /m^{}=en_c\mu `$, $`\omega _c=eB/m^{}`$, and $`n_c`$, $`\tau =\mathrm{}/v_F`$, $`\mathrm{}`$, and $`\mu `$ are the carrier density, transport time, elastic mean free path, and the mobility, respectively (Fig. (2)). In terms of resistivities, the classical values are $`\rho _{xx}^0=1/\sigma _0`$, independent of the magnetic field $`B`$, and $`\rho _{xy}^0=B/en`$. Quantum effects modify these results in two ways: quantum interference leads to localization, and Landau quantization drastically modifies the density of states at strong magnetic fields. Quantum mechanically, three energy scales are relevant: the cyclotron energy $`E_B=\mathrm{}\omega _c`$, the disorder broadening of the Landau bands $`\mathrm{\Gamma }`$, and the thermal energy $`E_T=k_BT`$. The ratio of the former two depends on the strength of the magnetic field and the strength and range of the disorder and corresponds to the classical quantity $`\omega _c\tau `$, characterizing the classical effects of the magnetic field and disorder in eqs. (1). While $`E_B/\mathrm{\Gamma }`$ and $`\omega _c\tau `$ are not identical, $`\omega _c\tau =1`$ can serve as an estimate of the point where Landau level quantization becomes important. The ratio of cyclotron energy to temperature, $`E_B/E_T=\mathrm{}eB/m^{}k_BT`$, determines whether the classical expression for $`\sigma _{xx}`$ is appropriate or Landau quantization has to be taken into account. In the limit of strong magnetic fields, such that Landau level mixing can be neglected, the high temperature (or short length scale) conductivity is qualitatively well described within the self-consistent Born approximation (SCBA) . In this approximation, the conductivity vanishes at zero temperature for integer filling factors $`\nu =n_c2\pi l_c^2`$, with the magnetic length $`l_c^2=\mathrm{}/eB`$, due to a vanishing density of states. For smooth random potential, relevant to most experiments on GaAs/AlGaAs heterostructures, the peak value of $`\sigma _{xx}^0`$ in SCBA is given by $`(l_c^2/\pi d^2)(e^2/h)`$, independent of the Landau level index. $`d`$ is the range of the disorder potential (Fig. (2)). Landau level quantization can thus lead to a strong reduction in the conductivity compared to the Drude result, provided that the magnetic field is strong enough, i.e. $`\omega _c\tau >1`$.
The localizing effect of quantum interference is most important for the occurrence of the QHE and it is the key to our understanding of the temperature dependence of the resistivities. In the absence of a magnetic field, quantum interference leads to a size-dependent reduction of the conductivity
$$\sigma _{xx}(L)=\sigma _{xx}^0\frac{2e^2}{\pi h}\mathrm{log}\left(\frac{L}{\mathrm{}}\right).$$
(4)
This weak localization expression is valid for large $`\sigma _{xx}`$. While the system size dependence is logarithmically weak for small system sizes, scaling theory predicts that it will eventually lead to complete localization and vanishing conductivity. The corresponding corrections to the Hall conductivity are given by
$$\sigma _{xy}(L)=\sigma _{xy}^0\omega _c\tau \frac{4e^2}{\pi h}\mathrm{log}\left(\frac{L}{\mathrm{}}\right).$$
(5)
Again, the decrease in the Hall conductivity is the precursor of the vanishing Hall conductivity at low fields predicted by scaling theory (Fig. (1)). In terms of the resistivities, these corrections lead to a logarithmic increase in the dissipative resistivity, while the Hall resistivity remains unchanged. In addition to the disorder effects, Coulomb interactions lead to logarithmic corrections to $`\sigma _{xx}`$, but not to $`\sigma _{xy}`$ . While these effects are important for a detailed comparison with experiment, they do not change the conclusions of the present discussion and will be neglected in the following.
In the presence of a magnetic field, quantum interference effects are reduced and the system size dependence of $`\sigma _{xx}`$ becomes even weaker ,
$$\sigma _{xx}(L)=\sigma _{xx}^0\frac{1}{\pi ^2\sigma _{xx}^0}\left(\frac{e^2}{h}\right)^2\mathrm{log}\left(\frac{L}{l_c}\right).$$
(6)
This means that localization effects become strong, when the system size exceeds the localization length
$$\xi ^0=l_c\mathrm{exp}(\pi ^2\sigma _{xx}^0{}_{}{}^{2}h_{}^{2}/e^4),$$
(7)
defined by $`\sigma _{xx}(\xi ^0)=0`$. In contrast to the zero field case, the system then does not become completely localized but exhibits a series of critical energies at which the conductivity remains finite and the Hall conductivity changes by $`e^2/h`$ as shown in Fig. (1). The effect of a finite temperature can be incorporated in the present discussion by replacing the system size $`L`$ by a phase coherence length $`L_\mathrm{\Phi }`$ that diverges as the temperature tends to zero.
From scaling theory, the following scenario for the temperature or system size dependence emerges: on small length scales or at high temperatures, classical Drude theory applies. At high magnetic fields, the effects of Landau quantization become visible, when $`\pi ^2E_T/E_B1`$ . In GaAs and for $`T=4.2`$K this happens at a magnetic field of about 2T. At lower temperatures, the Drude expression for $`\sigma _{xx}^0`$ is only valid up to about $`\omega _c\tau =1`$ beyond which the SCBA result $`\sigma _{xx}^0e^2/\pi h`$ becomes appropriate. Localization effects leading to the QHE become important when the system size and phase coherence length exceed the localization length $`\xi ^0`$. If $`\sigma _0`$ exceeds $`e^2/h`$, this length scale very rapidly becomes larger than the phase coherence length in present day experiments, provided $`\omega _c\tau <1`$. However, around $`\omega _c\tau =1`$ the bare conductivity drops below $`e^2/h`$ and $`L_\mathrm{\Phi }`$ can exceed $`\xi ^0`$ at low temperatures. In particular, near integer filling factors the conductivity is very small and the crossover length $`\xi ^0`$ becomes small. The point $`\omega _c\tau 1`$ separates two regions with very different temperature behaviors: at low fields, $`\rho _{xx}`$ increases slowly with decreasing temperature, while at higher fields, $`\rho _{xx}`$ decreases, most strongly near integer filling factors, due to the onset of strong localization on the quantum Hall plateau. At the crossover point the resistivity will be only very weakly temperature dependent. Note however, that this point does not correspond to a critical point in the zero temperature phase diagram. Up to the crossover point near $`\omega _c\tau =1`$ deviations from Drude behavior are small so that near the crossing point $`\rho _{xx}=\rho _{xy}`$.
We thus find that standard scaling theory predicts the essential features of the experiments that have been interpreted as showing a low-field QH-insulator transition: A magnetic field at which $`\rho _{xx}`$ is temperature independent has been observed in various experiments . This “critical” field separates an “insulating” low-field region with weak temperature dependence from a metallic QH region with stronger temperature dependence on the high field side. At this “transition” Hall and dissipative resistivity are approximately equal . The value of the resistivities at this transition is approximately $`1/\sigma _0=1/en_c\mu `$. The density dependence of this values should thus follow the density dependence of the zero-field mobility.
It should be stressed, that the validity of this argument goes beyond the validity of the employed approximations. The physical mechanism responsible for the drastic change in the temperature dependence near $`\omega _c\tau =1`$ is the suppression of the bare conductivity at high fields due to the gaps in the density of states as a result of Landau quantization. This leads to the strong field-dependence of the crossover scale $`\xi ^0`$. For a more quantitative agreement with experiment, the bare conductivities should be evaluated in SCBA taking into account Landau level mixing and higher order corrections should be included in eq. (6).
The question arises, under which conditions the non-monotonic dependence of the Hall conductivity predicted by the levitation scenario could be observed. Khmel’nitskii and Laughlin have argued that the plateau transitions are given by the condition that the Drude Hall conductivity $`\sigma _{xy}^0`$ equals half-integer multiples of $`e^2/h`$. This implies a lower bound on $`n_c\mu `$ for the occurrence of the QHE. The maximum value of $`\sigma _{xy}^0`$ is $`\sigma _0/2`$ at $`\omega _c\tau =1`$. Thus, for $`\sigma _0<e^2/h`$ there are no plateau transitions and hence no QHE. The reentrant plateau transitions occur for $`\omega _c\tau <1`$, where Drude theory is the appropriate expression for the bare conductivity. The minimum $`\sigma _0`$ for the occurrence of the $`n=2`$ plateau is $`3e^2/h`$ and the minimum crossover length $`\xi ^0`$ at $`\omega _c\tau =1`$ is $`l_c\mathrm{exp}((3\pi /2)^2)=4.410^9l_c`$, a macroscopic quantity for magnetic fields in the Tesla range. The zero temperature phase diagram with the levitating critical states at low fields is thus of very little importance for experiments on the QHE at low magnetic fields. Even though scaling theory predicts a very different behavior at zero temperature, at all but exponentially low temperatures it predicts a linear increase of the Hall resistivity up to fields where $`\omega _c\tau 1`$ and the onset of monotonically increasing quantum Hall plateaus beyond.
The system behaves differently, when only the $`n=1`$ plateau is observable. The bare conductivity at the low-field QH-insulator transition can then be of the order of $`e^2/2h`$ and the crossover scale $`\xi ^0`$ can be microscopic. At this transition scaling behavior should be observable.
From these consideration, we are led to conclude that the recent experimental observation of a temperature independent resistivity at low magnetic fields does, in fact, not contradict the scaling theory of the QHE, but rather is an expected finite-temperature effect. We further see that experiments on the low-field behavior of QH systems reveal only very limited information on the zero-temperature quantum phase transitions. In particular, they don’t give much insight into the nature of the insulator phase below the lowest QH transition. The experiments can, however, help to improve our understanding of the finite-size and finite-temperature effects associated with weak localization.
The situation is quite different on the high magnetic field side. Here, the SCBA applies as the starting point for the renormalization of the conductivities and the bare conductivity in the lowest Landau level is less than $`e^2/h`$. Thus it is possible to reach the asymptotic scaling regime, both in experiments and in numerical simulations . The results for finite temperatures/system sizes can reliably be extrapolated by finite-size scaling. However, even here the nature of the insulating phase remains quite elusive experimentally. In order to study the insulating phase, it is necessary to go beyond the scaling region of the QH-insulator transition. Numerically, it has been found that the Hall resistivity remains quantized at $`h/e^2`$ throughout the region where scaling behavior is observed. At the high-field end of the scaling region the longitudinal resistivity was found to be up to $`16h/e^2`$ , making accurate measurements of the much smaller Hall resistivity difficult. At zero temperature the width of the scaling region shrinks to zero. The experimentally observed quantization of the Hall resistivity through the transition is thus likely to be a confirmation of scaling behavior and is no indication of the transport properties of the insulating phase at zero temperature.
In conclusion, I have discussed the behavior of the quantum Hall effect at low magnetic fields as expected from the scaling theory of the QHE. The large localization length in a magnetic field in two dimensions restricts the observability of the levitating critical states to exponentially small temperatures and exponentially large systems. At accessible temperatures and system sizes the Hall resistivity will be a monotonically increasing function of magnetic field. Near magnetic fields, such that $`\omega _c\tau 1`$, the temperature dependence of the dissipative resistivity changes from weakly increasing at low magnetic fields to decreasing at higher magnetic fields, in accordance with recent experiments. At this approximately temperature-independent point $`\rho _{xx}`$ and $`\rho _{xy}`$ are of equal magnitude.
I acknowledge stimulating discussions with Z. Wang and X.C. Xie on the quantum Hall insulator and the hospitality of the Institute for Theoretical Physics at Santa Barbara. This work was performed within the research program of the SFB 341 of the DFG and supported by the NSF at ITP.
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.