id
stringlengths 30
36
| source
stringclasses 1
value | format
stringclasses 1
value | text
stringlengths 5
878k
|
|---|---|---|---|
no-problem/0001/astro-ph0001335.html | ar5iv | text | # Properties of Gamma-Ray Burst Classes
## Introduction
GRB spectral and temporal properties overlap, providing a continuum of burst characteristics. Some of this overlap is intrinsic in nature, while much is due to instrumental and observational biases. In addition to this overlap, there is clustering indicative of classes within the parameter space defined by GRB attributes. In particular, there are two long-recognized GRB classes cline74 ; kouveliotou93 based on duration (divided at roughly 2 seconds) and spectral hardness. A statistically significant third class has been identified using statistical clustering analysis mukherjee98 .
Can effects attributable to a source population be separated from instrumental effects? To answer this, we have applied computer science pattern recognition algorithms to learn why bursts cluster in some parameter spaces. For this analysis, we have used the supervised decision tree classifier C4.5 quinlan86 . Supervised classifiers establish rules for previously identified patterns, and must be trained by representative class members.
## Analysis
The three GRB classes identified by statistical clustering techniques mukherjee98 can be found from three significant classification attributes; 50 to 300 keV fluence, T90 duration, and HR321 hardness ratio (the fluence in the 100 to 300 keV band divided by the fluence in the 25 to 100 keV band). The properties of the three classes in terms of these attributes are demonstrated in Table 1.
C4.5 was trained on the three GRB classes using five fluences, two durations, three peak fluxes, and three hardness ratios. C4.5 produced a decision tree containing IF THEN ELSE branches for placing each GRB in the appropriate class; these branches were pruned to remove branches containing less than four GRBs. Rules were then generated for each class based on the pruned branches. C4.5 identifies outliers with poorly defined rules that often contain few GRBs. Statistical methods find that outliers are not closely bound to the class (cluster) centers. C4.5 rules identified a number of GRBs as having peculiar hardness ratios; these resulted from large individual channel fluence errors. The GRBs with the largest 10% relative errors (error divided by measurement) were subsequently removed from the database. The remaining 3B GRBs were reclassified using C4.5; the resulting rules were used to classify 4B Catalog GRBs and thus increase the database size.
### Class 3 Spectral Hardnesses
C4.5 verified that the three GRB classes resulted primarily from the attributes of spectral hardness, duration, and fluence. With the larger classification database, the dependence on spectral hardness could be examined in terms of the spectral fitting parameters $`\alpha `$, $`\beta `$, and E<sub>peak</sub> band93 . Using only these three attributes, C4.5 was able to accurately classify most of the 4B GRBs. The rules generated by C4.5 were able to cleanly separate Class 2 from Class 1, but could not delineate Class 3 from Class 1 (85% of Class 3 bursts were assigned to Class 1).
Upon further examination, Class 3 GRBs were found to have E<sub>peak</sub> values similar to Class 1 bursts of the same 1024 ms peak flux (Figure 1). The correlation between E<sub>peak</sub> and peak flux has been interpreted as cosmological redshift mallozzi95 .
### Class 3 Fluences and Durations
Since at least one of the three defining characteristics of Class 3 actually represents a data correlation, we hypothesized that Class 3 GRBs actually belong to Class 1. We decided to see if Class 3 fluences and durations could be explained in terms of Class 1 attributes. This could be the case if some instrumental or sampling bias made Class 1 GRBs appear to be shorter and fainter than they should be.
Figure 2 is a plot of fluence vs. 1024 ms peak flux for each of the three GRB classes, and is limited to GRBs detected when BATSE had one homogeneous set of trigger criteria. There are distinct regions outside of which no GRBs are found. GRBs with 1024 ms peak fluxes less than 0.2 photons cm<sup>-2</sup> second<sup>-1</sup> are not detected, since this is below BATSE’s minimum detection threshold. GRBs do not have fluences less than what would be found in their time-integrated 1024 ms peak fluxes, since this is the shortest timescale on which this peak flux can be measured.
Figure 3 overlays $`\mathrm{log}`$(T90) contours for Class 1 GRBs on the fluence vs. 1024 ms peak flux space. The contours demonstrate that GRBs can be modeled as a series of pulses, with pulses containing most of the fluence and interpulse separations primarily defining the duration. Most Class 2 bursts are single-pulsed events as measured on the 1024 ms timescale. This helps define the characteristics of the third distinct region outside of which no GRBs are found: high fluence, faint Class 1 GRBs are missing, whereas low fluence faint, Class 1 GRBs are present. Since a bias favoring detection of GRBs with few photons over those with many photons seems unlikely, we suspect a bias capable of underestimating fluence relative to peak flux.
We have dimmed a number of bright GRBs to where they just trigger in order to study their measured properties as they fade into background. Each burst’s peak flux is dimmed, and the time history is “noisified” with a Poisson background. The peak flux and fluence are then re-measured. These actions have been performed ten times on five bright bursts with a range of temporal structures.
One problem quickly became apparent during the analysis: the time interval bounding the fluence measurement (the fluence duration hakkila99 ) strongly influenced the amount of fluence measured. If the same fluence duration interval was used for undimmed and dimmed measurements, then the fluence-to-peak flux ratio did not change as a GRB was dimmed. If, however, the fluence duration interval shortened to account for faint pulses disappearing into the background and becoming unrecognizable, then the fluence-to-peak flux ratio decreased as the burst dimmed (see Figure 4). This bias becomes stronger near the trigger threshold.
Fluence durations taken from BATSE Catalogs provide supportive evidence for this mechanism. The durations used to calculate fluence of faint Class 1 GRBs are shorter than those of bright Class 1 GRBs hakkila99 .
## Conclusions
A mechanism exists whereby some Class 1 (Long) GRBs can develop Class 3 (Intermediate) characteristics via a combination of the hardness intensity relation and the fluence duration bias. Faint Class 1 GRBs are most likely to develop Class 3 characteristics, but it is possible for even bright GRBs with appropriate time histories and spectral features to develop these characteristics. Class 3 (Intermediate) GRBs do not therefore appear to represent a separate source population, although they cluster in the duration, fluence, hardness, attribute space. Class 2 (Short) GRBs do appear to represent a separate source population. We were unable to find a mechanism by which faint Class 1 GRBs could develop Class 2 characteristics.
GRB population studies can benefit from use of AI classifiers. There are many other attributes developed by the community that could be included for future study. To this end, we are designing a web-based AI tool for GRB classification haglin99 that includes supervised and unsupervised AI classifiers roiger99 . |
no-problem/0001/hep-ph0001032.html | ar5iv | text | # 1 Introduction
## 1 Introduction
A general-purpose generator in high-energy physics should address a number of physics aspects, such as:
* the matrix elements for a multitude of hard subprocesses of interest,
* the convolution with parton distributions to obtain the hard-scattering kinematics and cross sections,
* resonance decays that (more or less) form part of the hard subprocess (such as $`W`$, $`Z`$, $`t`$ or $`h`$),
* initial- and final-state QCD and QED showers (or, as an alternative, higher-order matrix elements, including a consistent treatment of virtual-correction terms),
* multiple parton–parton interactions,
* beam remnants,
* hadronization,
* decay chains of unstable particles, and
* general utility and analysis routines (such as jet finding).
Furthermore, one must be prepared for unexpected or less conventional effects, that could modify the assumed behaviour: the strong-interaction dynamics in QCD remains unsolved and thereby unpredictable in an absolute sense.
The PYTHIA 6.1 program was released in March 1997, as a merger of JETSET 7.4, PYTHIA 5.7 and SPYTHIA . It covers all of the above areas. The current subversion is PYTHIA 6.136, which contains over 50,000 lines of Fortran 77 code. The code, manuals and sample main programs may be found at
http://www.thep.lu.se/$``$torbjorn/Pythia.html .
The two other programs of a similar scope are HERWIG
http://hepwww.rl.ac.uk/theory/seymour/herwig/
and ISAJET
ftp://penguin.phy.bnl.gov/pub/isajet .
For parton-level processes, many more programs have been written. The availability of several generators provides for useful cross-checks and a healthy competition. Since the physics of a complete hadronic event is very complex and only partially understood from first principles, one should not prematurely converge on one single approach.
## 2 PYTHIA 6.1 Main News
Relative to previous versions, main news in PYTHIA 6.1 include
* a renaming of the old JETSET program elements to begin with PY, therefore now standard throughout,
* new SUSY processes and improved SUSY simulation relative to SPYTHIA, and new PDG codes for sparticles,
* new processes for Higgs (including doubly-charged in left–right symmetric models), technicolor, …,
* several improved resonance decays, including an alternative Higgs mass shape,
* some newer parton distributions, such as CTEQ5
* initial-state showers matched to some matrix elements,
* new options for final-state gluon splitting to a pair of $`c/b`$ quarks and modified modelling of initial-state flavour excitation,
* an energy-dependent $`p_{\mathrm{min}}`$ in multiple interactions,
* an improved modelling of the hadronization of small-mass strings, of importance especially for $`c/b`$, and
* a built-in package for one-dimensional histograms (based on GBOOK).
Some of these topics will be further studied below. Other improvements, of less relevance for $`\overline{p}p`$ colliders, include
* improved modelling of gluon emission off $`c/b`$ quarks in $`e^+e^{}`$,
* colour rearrangement options for $`W^+W^{}`$ events,
* a Bose-Einstein algorithm expanded with new options,
* a new alternative baryon production scheme ,
* QED radiation off an incoming muon,
* a new machinery to handle real and virtual photon fluxes, cross sections and parton distributions , and
* new standard interfaces for the matching to external generators of two, four and six fermions (and of two quarks plus two gluons) in $`e^+e^{}`$.
The current list of over 200 different subprocesses covers topics such as hard and soft QCD, heavy flavours, DIS and $`\gamma \gamma `$, electroweak production of $`\gamma ^{}/Z^0`$ and $`W^\pm `$ (singly or in pairs), production of a light or a heavy Standard Model Higgs, or of various Higgs states in supersymmetric (SUSY) or left–right symmetric models, SUSY particle production (sfermions, gauginos, etc.), technicolor, new gauge bosons, compositeness, and leptoquarks.
Needless to say, most users will still find that their particular area of interest is not as well addressed as could be wished. In some areas, progress will require new ideas, while lack of time is the limiting factor in others.
## 3 Matching to Matrix Elements
The matrix-element (ME) and parton-shower (PS) approaches to higher-order QCD corrections both have their advantages and disadvantages. The former offers a systematic expansion in orders of $`\alpha _s`$, and a powerful machinery to handle multiparton configurations on the Born level, but loop calculations are tough and lead to messy cancellations at small resolution scales. Resummed matrix elements may circumvent the latter problem for specific quantities, but then do not provide exclusive accompanying events. Parton showers are based on an improved leading-log (almost next-to-leading-log) approximation, and so cannot be accurate for well separated partons, but they offer a simple, process-independent machinery that gives a smooth blending of event classes (by Sudakov form factors) and a natural match to hadronization. It is therefore natural to try to combine these descriptions, so that ME results are recovered for widely separated partons while the PS sets the subjet structure.
For final-state showers in $`Z^0q\overline{q}`$, such solutions are the standard since long , e.g. by letting the shower slightly overpopulate the $`q\overline{q}g`$ phase space and then using a Monte Carlo veto technique to reduce down to the ME level. This approach easily carries over to showers in other colour-singlet resonance decays, although the various relevant ME’s have not all been implemented in PYTHIA so far.
A similar technique is now available for the description of initial-state radiation in the production of a single colour-singlet resonance, such as $`\gamma ^{}/Z^0/W^\pm `$ . The basic idea is to map the kinematics between the PS and ME descriptions, and to find a correction factor that can be applied to hard emissions in the shower so as to bring agreement with the matrix-element expression. Some simple algebra shows that, with the PYTHIA shower kinematics definitions, the two $`q\overline{q}^{}gW^\pm `$ emission rates disagree by a factor
$$R_{q\overline{q}^{}gW}(\widehat{s},\widehat{t})=\frac{(\mathrm{d}\widehat{\sigma }/\mathrm{d}\widehat{t})_{\mathrm{ME}}}{(\mathrm{d}\widehat{\sigma }/\mathrm{d}\widehat{t})_{\mathrm{PS}}}=\frac{\widehat{t}^2+\widehat{u}^2+2m_W^2\widehat{s}}{\widehat{s}^2+m_W^4},$$
which is always between $`1/2`$ and 1. The shower can therefore be improved in two ways, relative to the old description. Firstly, the maximum virtuality of emissions is raised from $`Q_{\mathrm{max}}^2m_W^2`$ to $`Q_{\mathrm{max}}^2=s`$, i.e. the shower is allowed to populate the full phase space. Secondly, the emission rate for the final (which normally also is the hardest) $`qqg`$ emission on each side is corrected by the factor $`R(\widehat{s},\widehat{t})`$ above, so as to bring agreement with the matrix-element rate in the hard-emission region. In the backwards evolution shower algorithm , this is the first branching considered.
The other possible $`𝒪(\alpha _s)`$ graph is $`qgq^{}W^\pm `$, where the corresponding correction factor is
$$R_{qgq^{}W}(\widehat{s},\widehat{t})=\frac{(\mathrm{d}\widehat{\sigma }/\mathrm{d}\widehat{t})_{\mathrm{ME}}}{(\mathrm{d}\widehat{\sigma }/\mathrm{d}\widehat{t})_{\mathrm{PS}}}=\frac{\widehat{s}^2+\widehat{u}^2+2m_W^2\widehat{t}}{(\widehat{s}m_W^2)^2+m_W^4},$$
which lies between 1 and 3. A probable reason for the lower shower rate here is that the shower does not explicitly simulate the $`s`$-channel graph $`qgq^{}q^{}W`$. The $`gq\overline{q}`$ branching therefore has to be preweighted by a factor of 3 in the shower, but otherwise the method works the same as above. Obviously, the shower will mix the two alternative branchings, and the correction factor for a final branching is based on the current type.
The reweighting procedure prompts some other changes in the shower. In particular, $`\widehat{u}<0`$ translates into a constraint on the phase space of allowed branchings.
Our published comparisons with data on the $`W`$ $`p_{}`$ spectrum show quite a good agreement with this improved simulation . A worry was that an unexpectedly large primordial $`k_{}`$, around 4 GeV, was required to match the data in the low-$`p_W`$ region. However, at that time we had not realized that the data were not fully unsmeared. The required primordial $`k_{}`$ is therefore likely to drop by about a factor of two .
It should be noted that also other approaches to the same problem have been studied recently. The HERWIG one requires separate treatments in the hard- and soft-emission regions . Another, more advanced PYTHIA-based one , also addresses the next-to-leading order corrections to the total $`W`$ cross section, while the one outlined above is entirely based on the leading-order total cross section. There is also the possibility of an extension to Higgs production , which is rather less trivial since already the leading-order cross section $`ggH`$ contains a QCD loop.
Summarizing, we now start to believe we can handle initial- and final-state showers, with next-to-leading-order accuracy, in cases where these can be separated by the production of colour singlet resonances — even if it should be realized that much work remains to cover the various possible cases. That still does not address the big class of QCD processes where the initial- and final-state radiation does not factorize. Possibly, correction factors to showers could be found also here. Alternatively, it may become necessary to start showers from given parton configurations of varying multiplicity and with virtual-correction weights, as obtained from higher-order ME calculations. So far, PYTHIA only implements a way to start from a given four-parton topology in $`e^+e^{}`$ annihilation, picking one of the possible preceding shower histories as a way to set constraints for the subsequent shower evolution . This approach obviously needs to be extended in the future, to allow arbitrary parton configurations. Even more delicate will be the consistent treatment of virtual corrections , where much work remains.
## 4 Charm and Bottom Hadronization
Significant asymmetries are observed between the production of $`D`$ and $`\overline{D}`$ mesons in $`\pi ^{}p`$ collisions, with hadrons that share some of the $`\pi ^{}`$ flavour content very much favoured at large $`x_F`$ in the $`\pi ^{}`$ fragmentation region . This behaviour was qualitatively predicted by PYTHIA; in fact, the predictions were for somewhat larger effects than seen in the data. The new data has allowed us to go back and take a critical look at the uncertainties that riddle the heavy-flavour description . Many effects are involved, and we here constrain ourselves to only mentioning one.
A hadronic event is conventionally subdivided into sets of partons that form separate colour singlets. These sets are represented by strings, that e.g. stretch from a quark end via a number of intermediate gluons to an antiquark end. Three string mass regions may be distinguished for the hadronization.
* Normal string fragmentation. In the ideal situation, each string has a large invariant mass. Then the standard iterative fragmentation scheme works well. In practice, this approach can be used for all strings above some cut-off mass of a few GeV.
* Cluster decay. If a string is produced with a small invariant mass, maybe only two-body final states are kinematically accessible. The traditional iterative Lund scheme is then not applicable. We call such a low-mass string a cluster, and consider it separately from above. In recent program versions, the modelling has now been improved to give a smooth match on to the standard string scheme in the high-cluster-mass limit.
* Cluster collapse. This is the extreme case of the above situation, where the string mass is so small that the cluster cannot decay into two hadrons. It is then assumed to collapse directly into a single hadron, which inherits the flavour content of the string endpoints. The original continuum of string/cluster masses is replaced by a discrete set of hadron masses. Energy and momentum then cannot be conserved inside the cluster, but must be exchanged with the local neighbourhood of the cluster. This description has also been improved.
In general, flavour asymmetries are predicted to be smaller for bottom than for charm, and smaller at higher energies (except possibly at very large rapidities). One can therefore not expect any spectacular manifestations at the Tevatron. However, other nontrivial features do not die out as fast, like a non-negligible systematic shift between the rapidity of a heavy quark and that of the hadron produced from it . The possibility of such effects should be considered whenever trying to extract any physics from heavy flavours.
## 5 Multiple Interactions
Multiple parton–parton interactions is the concept that, based on the composite nature of hadrons, several parton pairs may interact in a typical hadron–hadron collision . Over the years, evidence for this mechanism has accumulated, such as the recent direct observation by CDF . The occurences with two parton pairs at reasonably large $`p_{}`$ just form the top of the iceberg, however. In the PYTHIA model, most interactions are at lower $`p_{}`$, where they are not visible as separate jets but only contribute to the underlying event structure. As such, they are at the origin of a number of key features, like the broad multiplicity distributions, the significant forward–backward multiplicity correlations, and the pedestal effect under jets.
Since the perturbative jet cross section is divergent for $`p_{}0`$, it is necessary to regularize it, e.g. by a cut-off at some $`p_{\mathrm{min}}`$ scale. That such a regularization should occur is clear from the fact that the incoming hadrons are colour singlets — unlike the coloured partons assumed in the divergent perturbative calculations — and that therefore the colour charges should screen each other in the $`p_{}0`$ limit. Also other damping mechanisms are possible . Fits to data typically give $`p_{\mathrm{min}}2`$ GeV, which then should be interpreted as the inverse of some colour screening length in the hadron.
One key question is the energy-dependence of $`p_{\mathrm{min}}`$; this may be relevant e.g. for comparisons of jet rates at different Tevatron energies, and even more for any extrapolation to LHC energies. The problem actually is more pressing now than at the time of our original study , since nowadays parton distributions are known to be rising more steeply at small $`x`$ than the flat $`xf(x)`$ behaviour normally assumed for small $`Q^2`$ before HERA. This translates into a more dramatic energy dependence of the multiple-interactions rate for a fixed $`p_{\mathrm{min}}`$.
The larger number of partons also should increase the amount of screening, however, as confirmed by toy simulations . As a simple first approximation, $`p_{\mathrm{min}}`$ is assumed to increase in the same way as the total cross section, i.e. with some power $`ϵ0.08`$ that, via reggeon phenomenology, should relate to the behaviour of parton distributions at small $`x`$ and $`Q^2`$. Thus the new default in PYTHIA is
$$p_{\mathrm{min}}=(1.9\mathrm{GeV})\left(\frac{s}{1\mathrm{TeV}^2}\right)^{0.08}.$$
## 6 Interconnection Effects
The widths of the $`W`$, $`Z`$ and $`t`$ are all of the order of 2 GeV. A Standard Model Higgs with a mass above 200 GeV, as well as many supersymmetric and other Beyond the Standard Model particles would also have widths in the multi-GeV range. Not far from threshold, the typical decay times $`\tau =1/\mathrm{\Gamma }0.1\mathrm{fm}\tau _{\mathrm{had}}1\mathrm{fm}`$. Thus hadronic decay systems overlap, between a resonance and the underlying event, or between pairs of resonances, so that the final state may not contain independent resonance decays.
So far, studies have mainly been performed in the context of $`W`$ pair production at LEP2. Pragmatically, one may here distinguish three main eras for such interconnection:
* Perturbative: this is suppressed for gluon energies $`\omega >\mathrm{\Gamma }`$ by propagator/timescale effects; thus only soft gluons may contribute appreciably.
* Nonperturbative in the hadroformation process: normally modelled by a colour rearrangement between the partons produced in the two resonance decays and in the subsequent parton showers.
* Nonperturbative in the purely hadronic phase: best exemplified by Bose–Einstein effects.
The above topics are deeply related to the unsolved problems of strong interactions: confinement dynamics, $`1/N_\mathrm{C}^2`$ effects, quantum mechanical interferences, etc. Thus they offer an opportunity to study the dynamics of unstable particles, and new ways to probe confinement dynamics in space and time , but they also risk to limit or even spoil precision measurements.
A key gauge is the interconnection impact on $`W`$ mass measurements at LEP2. Perturbative effects are not likely to give any significant contribution to the systematic error, $`\delta m_W\text{ }\stackrel{<}{}\text{ }5`$ MeV . Colour rearrangement is not understood from first principles, but many models have been proposed to model effects , and a conservative estimate gives $`\delta m_W\text{ }\stackrel{<}{}\text{ }40`$ MeV. For Bose–Einstein again there is a wide spread in models, and an even wider one in results, with about the same potential systematic error as above . The total QCD interconnection error is thus below $`m_\pi `$ in absolute terms and 0.1% in relative ones, a small number that becomes of interest only because we aim for high accuracy.
A study of $`e^+e^{}t\overline{t}bW^+\overline{b}W^{}b\overline{b}\mathrm{}^+\nu _{\mathrm{}}\mathrm{}^{}\overline{\nu }_{\mathrm{}}^{}`$ near threshold gave a realistic interconnection uncertainty of the top mass of around 30 MeV, but also showed that slight mistreatments of the combined colour and showering structure could blow up this error by a factor of ten . For hadronic top decys, errors could be much larger.
The above numbers, when applied to hadronic physics, are maybe not big enough to cause an immediate alarm. The addition of a coloured underlying event — with a poorly-understood multiple-interaction structure as outlined above — has not at all been considered so far, however, and can only make matters worse in hadronic physics than in $`e^+e^{}`$. This is clearly a topic for the future, where we should be appropriately humble about our current understanding, at least when it comes to performing precision measurements.
QCD interconnection may also be at the root of a number of other, more spectacular effects, such as rapidity gaps and the whole Pomeron concept , and the unexpectedly large rate of quarkonium production .
## 7 The Future: On To C++
Finally, a word about the future. PYTHIA continues to be developed. On the physics side, there is a need to increase the support given to different physics scenarios, new and old, and many areas of the general QCD machinery for parton showers, underlying events and hadronization require further improvements, as we have seen.
On the technical side, the main challenge is a transition from Fortran to C++, the language of choice for Run II (and LHC). To address this, the PYTHIA 7 project was started in January 1998, with L. Lönnblad as main responsible. A similar project, but more ambitious and better funded, is now starting up for HERWIG, with two dedicated postdoc-level positions and a three-year time frame.
For PYTHIA, what exists today is a strategy document , and code for the event record, the particle object, some particle data and other data base handling, and the event generation handler structure. All of this is completely new relative to the Fortran version, and is intended to allow for a much more general and flexible formulation of the event generation process. The first piece of physics, the string fragmentation scheme, is being implemented by M. Bertini, and is nearing completion. The subprocess generation method is being worked on for the simple case of $`e^+e^{}Z^0q\overline{q}`$. The hope is to have a “proof of concept” version soon, and some of the current PYTHIA functionality up and running by the end of 2000. It will, however, take much further effort after that to provide a program that is both more and better than the current PYTHIA 6 version. It is therefore unclear whether PYTHIA 7 will be of much use during Run II, except as a valuable exercise for the future. |
no-problem/0001/astro-ph0001105.html | ar5iv | text | # Clustering of galaxies at faint magnitudes
## 1 Introduction
The angular correlation function of galaxies $`w(\theta )`$ is still the most commonly used tool for investigating the evolution of clustering at high redshift . While a more direct measurement of the two point correlation is significantly more powerful (see e.g. Le Fèvre et al. 1996), its computation requires deep magnitude limited redshift surveys, which currently are not feasable at the faintest reachable magnitude limits. By contrast, the angular correlation function can easily be computed from photometric galaxy catalogs alone. With the availability of the Northern Hubble Deep Fields (HDF-N, Williams et al. 1996), the angular correlation function has been computed to an R magnitude of 29 (Villumsen et al. 1996, hereafter VFC, Colley et al. 1996). These studies suggest that the amplitude of the correlation function continuously decreases with the magnitude limit of the sample over a magnitude range of more than 10 magnitudes down to the faintest magnitude limits probed so far. The amplitude at an angular separation of one arcsec, $`w(1^{\prime \prime })`$, seems to follow closely a power law $`w(1^{\prime \prime })10^{0.27R}`$(e.g. Brainerd et al. 1995).
Despite the apparent simplicity of computing the angular correlation function, a significant controversy has arisen about the magnitude limits at which the correlation function flattens. Measuring the correlation amplitude at a separation of 1 arcmin, Brainerd & Smail (1998) found that the correlation amplitude reaches a minimum for samples with limiting I magnitude of about 23 and stays flat for even deeper magnitude limited samples. On the other hand, the deeper HDF-N measurements at a separation of 1 arcsec found a continuously decreasing clustering amplitude down to the faintest magnitude limits. This discrepancy could be due to a number of different reasons such as biased field selection, different redshift selection through differences in the bandpasses, systematic errors in the determination of the clustering amplitude (e.g. due to gradients in the sensitivity of the detector), or simply random fluctuations in the correlation amplitudes. However, it could also mean a true discontinuity of the shape and/or amplitude of the correlation function at a magnitude of $`I26`$ or $`R27`$.
Before suggesting the existence of a discontinuity of the correlation amplitudes, several tests of the results should be carried out. First of all, both the ground based results and the HDF-N results should be verified by independent samples. Secondly, any apparent discontinuity should be tested with a single sample which covers the relevant magnitude range. Up to now, available ground based have not been not deep enough to overlap with the HDF results, which are only available for magnitudes fainter than $`R26`$. The current work addresses both of these points. We have used the southern HDF field to obtain an independent verification of the previous HDF-N results. In addition, we have used the “flanking fields” of the Northern and Southern HDFs (Williams et al. 1998) to derive a catalog over a larger area and thus allowing us to compute the $`w(\theta )`$ at brighter magnitudes from an HDF-like sample. Finally, we have used a sample of galaxies detected in a deep ground based image, the so-called ESO NTT deep field (Arnouts et al. 1999) to compute $`w(\theta )`$ from ground-based data at faint magnitudes. The combined data sets allow us to derive correlation amplitudes from both HST and ground based data at identical magnitude bins, and thereby search for discontinuities in the correlation amplitudes as a function of magnitude.
In Sect. 2, we present the results from the HDF fields which include the Northern HDF, the Southern HDF and the flanking fields. In Sect. 3, we present new results from the ESO NTT deep field. In Sect. 4, we investigate the correlation amplitude as a function of magnitude by combining the new data with previous estimates in the literature. Finally, in Sect. 5, we discuss the results and present our conclusions.
## 2 HDF samples
### 2.1 HDF South : The galaxy catalog
We used the “drizzled” HDF-S images distributed by the Space Telescope – European Coordinating Facility archive. The galaxy catalog was generated using the SExtractor program (Bertin & Arnouts 1996). The catalog of Clements & Couch (1996) was used by VFC to compute $`w(\theta )`$ from the HDF-N images. In order to be able to compare our results with the results obtained from the HDF-N we use the exact same extraction parameters as input to SExtractor as those described in Clements & Couch. We use a minimum object extraction area of 30 pixels and a detection threshold of 1.3$`\sigma `$ above the background. As a detection image we use a master frame calculated as the sum of the combined images taken with the F606W and F814W filters, which are similar to the R and I passbands respectively. As a detection filter we use a top-hat filter of 30 connected pixels. We checked the reliability of the extraction parameters by comparing the master frame with the output object image, which SExtractor optionally provides. We could not detect any tendency to find spurious objects in the vicinity of bright objects. The useful part of the combined mosaic frames are different between the HDF-N and HDF-S because the WFPC2 fields of view are slightly rotated relative to the mosaiced image which is aligned with the north/south direction. Therefore, we re-defined for each of the WF2, WF3 and WF4 CCDs useful areas from visual inspection of the master frame. Fig.1 shows the chosen areas. Only objects detected within the shown boundaries were used in the analysis. In order to excluded stars from the galaxy catalog we set an upper limit to the neural network star parameter in the output SExtractor catalog of 0.98, which excluded 20 objects. This is consistent with the number of stars expected for a field with the galactic latitude and depth as the HDF-S. Also we exclude saturated stars found by visual inspection from the galaxy catalog. There are not very bright stars in the HDF-S so we did not mask out any regions within the area shown in Fig. 1. We chose to use the Automatic Aperture Magnitudes (AAM) instead of the Corrected Isophotal Magnitudes (CIM), since the latter proved unstable. As well AAM as CIM are intended to give an estimate of the total flux of an object. The AAM is measured using an elliptical aperture with minor axis $`b=2.5r_1ϵ`$ and major axis $`a=2.5\frac{r_1}{ϵ}`$, where $`r_1`$ is the first moment of the light distribution and $`ϵ`$ is the ellipticity.
The resulting number of galaxies in the galaxy catalog is 1346 with 456 in the WF2 field, 530 in the WF3 field and 360 in the WF4 field.
As zero-points for the photometry we use those released by the STScI with the WFPC2 HDF-S data for the VEGAMAG system. In the following we shall refer to the F606W and F814W as R and I respectively. Fig. 2 compares the galaxy counts for HDF-N and HDF-S. It is seen that the distributions are similar. In particular, the magnitude bins at which the number counts drop significantly due to incompleteness in the two samples are almost identical. The basic properties of the galaxy catalogs from the two fields as well as from other fields to be discussed below, are listed in Table 1.
### 2.2 Correlation function from the HDF south catalog
We extract from the galaxy catalog eight R-magnitude-limited samples with limiting magnitudes from R=25.5 to R=29.0 in 0.5 mag steps. As for the HDF-N, galaxies brighter than R=23 are excluded.
The angular correlation function $`w(\theta )`$ is estimated using the optimal estimator described by Landy & Szalay (1993) :
$$w(\theta )=\frac{DD2DR+RR}{RR},$$
(1)
where DD is the number of galaxy-galaxy pairs, DR the number of galaxy-random pairs and RR the number of random-random pairs at the separation $`\theta `$. We generate a set of 32000 random points distributed according to the Poisson distribution. The number of galaxy-random pairs and random-random pairs at a given magnitude limit are normalized to the total number of galaxy-random and random-random pairs respectively. The errors are calculated from Poisson statistics.
Since the mean density of galaxies has to be estimated from the sample itself, the integral of the correlation function over the survey area is forced to zero. This reduces the correlation $`w(\theta )`$ function by an amount $`C`$ given by the integral :
$$C=\frac{1}{\mathrm{\Omega }^2}𝑑\mathrm{\Omega }_1𝑑\mathrm{\Omega }_2w(\theta ),$$
(2)
where $`\mathrm{\Omega }`$ is the solid angle of the survey area. $`C`$ is commonly referred to as the ’integral constraint‘. Assuming that $`w(\theta )`$ is a power law,
$$w(\theta )=A\theta ^{\gamma +1},\gamma =1.8,$$
(3)
then $`C=0.078(0.077,0.080)A`$ for the used parts of the WF2(WF3,WF4) area shown in Fig.1.
Fig.3 shows the observed $`w(\theta )`$ for the eight magnitude limits. The error bars represent 1 $`\sigma `$ Poisson errors. The amplitude of $`w(\theta )`$ is determined by fitting
$$w(\theta )=A\theta ^{\gamma +1}C,\gamma =1.8,$$
(4)
which takes into account the integral constraint.
### 2.3 Flanking fields
The high resolution deep images of HDF fields are uniquely suited to investigate the correlation function at the faintest magnitudes. However, their small field of view includes too few brighter galaxies to provide useful overlap with ground based measurements of the correlation function. This situation can be improved by taking advantage of the HDF “flanking fields”. These fields are contiguous to the HDF fields proper and have been taken during the HDF observing campaigns. We use the two deepest of the HDF-N flanking fields and the nine HDF-S flanking fields.
The HDF proper fields were chosen to avoid bright galaxies. It could be argued that this choice biases the measured correlation amplitude (Brainerd & Smail 1998). As the flanking fields have been chosen with proximity to the HDF fields as the only criterion, no bias related to the HDF proper field is present.
The flanking fields were only imaged with the F814W filter. There is a total of 14 fields. Three of the southern fields contained very bright stars. Rather than masking out these stars we decided not to use these frames in the analysis. Catalogs from the remaining 11 fields were generated in the same manner as for the HDF-S field. The average number counts as a function of magnitude are shown in the upper plot in Fig. 2. For magnitudes up to an I magnitude of about 25 the number per area is consistent with the one in the deep fields. The larger area therefore increases the sample in the brighter magnitude bins by almost an order of magnitude. The relevant parameters are again summarized in Table 1.
Finally, we measure the two point correlation function in the three bins $`23<`$I$`<24.5`$, $`23<`$I$`<25.0`$, and $`23<`$I$`<25.5`$ in the same way as described in Sect. 2.2.
### 2.4 The correlation function from the combined Hubble Deep Fields
In Fig.4 we compare the HDF-S R band results with the results derived by VFC for the HDF-N. The triangles with 1$`\sigma `$ error bars show the results for the HDF-S and the squares with 1$`\sigma `$ error bars show the results from the HDF-N. For all magnitude bins the measurements from the two fields agree within 1$`\sigma `$. We conclude that any differences are due to random fluctuations, and therefore combine all the available data from the HDF-N and HDF-S projects to obtain our final estimate of the correlation amplitudes from the HDF fields. In Fig.5 we perform the same comparison for the I band, this time also including the measurements from the flanking fields. The measurements are in all bins consistent within 1$`\sigma `$, and we again combine all the available data from the HDF-N and HDF-S projects including the Flanking fields to obtain our final estimate of the I band correlation amplitudes from the HDF fields.
For the R band we determine $`w(\theta )`$ in the 8 magnitude bins R$`<`$25.5, R$`<`$26 …, R$`<`$29. In these bins the number counts are not significantly influenced by incompleteness. For the I band we determine $`w(\theta )`$ in the 8 magnitude bins I$`<`$24.5, I$`<`$25,…,I$`<`$28. The flanking fields are here included in the measurement of $`w(\theta )`$ in the first three bins.
The determination of $`w(\theta )`$ and the power law fit is done as described in Sect. 2.2. We detect a positive correlation amplitude in all 8 bins in both R and I.
## 3 The ESO-NTT deep field
In order to verify that the correlation amplitudes we measured in the HST WFPC2 frames are not in some way due to properties of the WFPC2 instrument or HDF observing or reduction procedure, it is desirable to compare to a catalog of galaxies derived from deep high-resolution ground based observations. The ESO-NTT field (Arnouts et al. 1999), for which the images are publicly available, is suitable for this purpose. A catalog of galaxies for this field has been provided by S. D’Odorico at the ESO website at http://www.eso.org/ndf. The field provides deep multi-colour imaging taken with good sampling of the PSF. The depth of the images is about 26.7 in R and 26.3 in I. The effective seeings of the co-added images are about 0.7-0.8 arcsec.
We determine the amplitude of $`w(\theta )`$ in the magnitude intervals $`22`$$`24.5`$, $`22`$$`25.0`$ and $`22`$$`25.5`$ for both the I and R filters. We use an area of 4.4 arcmin<sup>2</sup> confined by the pixel values $`40<`$x,y$`<1020`$ (chosen to avoid edge-effects). For this area we derive an integral constraint of $`C=0.0468A`$. Due to the small size of the NTT SUSI Deep Field we cannot determine the amplitude of $`w(\theta )`$ very precisely. Nevertheless, we do detect a positive correlation signal in all but one of the magnitude bins. The measured amplitudes of $`w(\theta )`$ are in all bins consistent with the measurements in the two HDFs within 1$`\sigma `$ (see Fig. 6 and Fig. 7).
Stellar contamination of the sample can reduce the measured correlation amplitude. For a sample with a fraction $`f_{\mathrm{star}}`$ of galaxies, the measured correlation amplitude is reduced by a factor $`(1f_{\mathrm{star}})^2`$. Fortunately, the excellent seeing allows a firm galaxy/star discrimination. We follow Arnouts et al. in classifying as stars all objects with SExtractor classifier larger than 0.9 in the I band. This procedure removes about 5% of the objects in the field. For objects fainter than 24 the SExtractor classifier is not efficient, so we assume conservatively that about 5% of the objects used to measure the correlation amplitude are stars. In fact, the star counts are shallower at faint magnitudes than that of galaxies (Reid et al. 1996, their Fig. 1), which means that the star fraction at $`I>24`$ must be smaller than for $`I<24`$. Therefore, a conservative upper limit of the error introduced by stellar contamination of the catalog is 10%, which is significantly smaller than the random uncertainties.
## 4 The correlation amplitude as a function of magnitudes
The correlation amplitudes presented above overlap neither in magnitude nor scale with most such measurements from ground based data. Investigations of the correlation amplitude as a function of magnitudes which use these data therefore require knowledge of the power law index $`\delta `$ in order to compare amplitudes measured at different separations. The HST data presented here are consistent with the usual $`\delta =0.8`$ power law for angular separations between 1 arcsec and 1 arcmin, but do not rule out steeper of flatter slopes. On the other hand, ground based data have been used to derive the shape of the correlations function at separations between 0.5 and 5 arcminutes (e.g. Postman et al. 1998), but it is not clear whether the slopes can be extrapolated to smaller separations. In the following investigation, we have adopted a slope of 0.8 in most cases. The uncertainty in the slope is probably the largest source of uncertainty in the comparison of measurements from different surveys.
Fig. 6 shows the measured amplitudes of $`w(\theta )`$ at 1 arcsec as a function of the limiting magnitudes of surveys in the R band from measurements by Brainerd et al. (1995), Roche et al. (1993), Couch et al. (1993) and Woods & Fahlman (1997). Where necessary (data from Roche et al. and Couch et al.), we have scaled original correlation amplitudes quoted in each paper to the amplitude at 1 arcsec by assuming a power law for the correlation function with $`\delta =0.8`$. Similarly, Fig. 7 shows the measured amplitudes of $`w(\theta )`$ at 1 arcsec as a function of the limiting magnitudes of the surveys in the I band. The measurements of Postman et al. (1998), Brainerd & Smail (1998), Woods & Fahlman (1997), Benoist et al. (1999), Benoist et al. (in prep.) and Lidman & Peterson (1996) are included. Here, all the amplitudes except those from Woods & Fahlman (1997) and Postman et al. (1998) were extrapolated to a separation of 1 arcsec with $`\delta =0.8`$. For the Postman et al. (1998) measurements, the data at 1 arcmin and 0.5 arcmin were extrapolated using the slope measured by the authors in the range 0.5 arcmin – 5 arcmin.
Both in Fig. 6 and Fig. 7, we show theoretical predictions of the expected correlation amplitudes. The models are identical to those shown by VFC, but with the assumed redshift distribution in the faint magnitude bins derived from the photometric redshift distribution of HDF-N galaxy by Fernández-Soto et al. (1999). The median redshifts adopted at magnitudes brighter than 26 in the R band are the same as those used by VFC. The models are specified by a present clustering length $`r_0`$ and and evolution parameter $`ϵ`$ (see VFC for details). The models shown have a present clustering length of $`r_0=4.0h^1`$ Mpc and assume no evolution of clustering (dotted line, $`ϵ`$=0), linear evolution of clustering (solid line, $`ϵ`$=0.8) or non-linear evolution of clustering (dashed line, $`ϵ`$=1.6).
Several data sets shown in Fig. 6 and Fig. 7 are in conflict with each other at the 5-7$`\sigma `$ level. This reflects the fact that the main source of errors are systematic errors, while the error bars in most cases only include statistical errors. In particular at faint magnitudes, systematic uncertainties are hard to quantify. The disagreements appear to be larger in the I-band. This might be due to the fact that in the I band, variations in the effective filter bandpasses result in relatively large differences in the redshift distribution of extracted galaxy samples. Also, the limiting magnitude on the abscissa is somewhat depending on the limiting magnitudes at the bright end. However, as the galaxy counts ($`N`$) increases steeply with the limiting magnitude at the faint end (mag<sub>lim</sub>, $`\mathrm{log}N0.6`$ mag<sub>lim</sub>), the uncertainty due to different limiting magnitudes is small (the difference in the median magnitude of two galaxy samples with limiting magnitude 25 at the faint end and limiting magnitudes at the bright end of 20 and 24 is only about 0.5mag). More likely, however, the disagreement is related to the extrapolation from 1 arcmin scales to 1 arcsec scales. In Postman et al. (1998, their Table 3) it can be seen that the power $`\delta `$ seem to decline at $`I>21`$ to smaller values than the canonical value 0.8 ($`\delta =0.7`$ at $`I=22`$ and $`\delta =0.5`$ at $`I=23`$), especially at the smallest scales 0.5 arcsec $`<\theta <`$ 5 arcmin. A similar trend was noted by Benoist et al. (1999). Brainerd & Smail (1998) find at best fitting value of $`\delta =0.7`$ at $`I>24.5`$. A change in slope from $`\delta =0.8`$ to $`\delta =0.5`$ amounts to factor of 3.4 difference in the extrapolation from 1 arcmin scales to 1 arcsec scales. Due to these uncertainties, our current knowledge of the amplitude and shape of the correlation function in the range 1 arcmin $`<\theta <`$ 1 arcsec at the faintest magnitude limits reachable from the ground, I=23–25, is unfortunately still poor.
## 5 Summary and conclusions
From the data presented above we can draw the following conclusions.
* The clustering amplitudes measured in the HDF-N field have been confirmed with the independent sample from the Southern HDF. This rules out that the low amplitudes measured in the HDF-N field are due to some cosmological fluctuations. The HDF-N field seems to provide a fair sample of high-redshift galaxies.
* The clustering in the HDF flanking fields is consistent with the one measured from ground based samples at the bright end of the magnitude range, and with the ones measured in the HDFs at the faint end. This provides additional support to the view that the HDF fields represent indeed a fair view of the high-redshift universe.
* The measured correlation amplitudes in the ESO-NTT galaxy sample in magnitude bins which overlap those of the HDF sample are consistent with the HDF measurements. This observation verifies that correlations measured with WFPC2 are not due to some instrumental effects.
* The results plotted in Fig. 6 show a continuous decline of the amplitude of the correlation function in R. We do not confirm the Brainerd & Smail (1998) detection of a flattening in the I band at I=22–23.
* There is some indication of a flattening of the correlation amplitudes at the faintest levels ($`R27`$ or $`I26`$). Using models for the redshift distribution, VFC interpreted this flattening as evidence for linear evolution of the clustering of a galaxy population which at present has a correlation length of about 4$`h^1`$Mpc. The data presented in this paper are still in good agreement with this interpretation.
* Interpretation of Fig. 6 and 7 is difficult given the significant disagreement between data points from different studies. This disagreement could be caused by a change of $`\delta `$ as a function of magnitude and scale. To reach a final understanding of the two-point correlation function data compiled in this work, it therefore appears to be essential first to understand the detailed behavior of the shape of the correlation function.
## Acknowledgments
We wish to thank C. Benoist and B. Thomsen for helpful discussions and C. Benoist for giving giving us access to some of his results from the EIS data prior to publication. We thank S. D’Odorico for making available the catalog from the NTT Deep Field. We thank the referee, T. Brainerd, for comments which significantly improved the manuscript. This work made use of the ESO/ST-ECF Science Archive Facility. |
no-problem/0001/nlin0001066.html | ar5iv | text | # References
## Abstract
The Volterra and Toda chains equations are considered. A class of special reductions for these equations are derived.
Reductions of the Volterra and Toda chains
Andrei K. Svinin<sup>1</sup><sup>1</sup>1E-mail:svinin@icc.ru
Institute for System Dynamics and Control Theory
Siberian Branch of Russian Academy of Sciences
P.O. Box 1233, 664033 Irkutsk, Russia
The Volterra and Toda chains are known to have numerous applications. In particular, they appear in connection with continuum two-boson KP hierarchies and their discrete symmetries (or auto-Bäcklund transformations). It is worthing also to mention the works , where integrable chains serve as discrete symmetries of differential evolution equations.
In this note we establish finite-dimensional systems arising in process of special reductions of Volterra and Toda lattices. Translations on the lattices, as in the case of two-boson KP hierarchies, yield groups of discrete symmetries for these finite-dimensional systems.
Consider the Volterra lattice in the form
$$\dot{r}(i)=r(i)\left\{r(i1)r(i+1)\right\},i𝐙.$$
(1)
For any fixed $`n𝐍`$, impose a constraint in the form of the following algebraic equations:
$$r(i)+\mathrm{}+r(i+n1)=r(i1)r(i)\mathrm{}r(i+n),i𝐙.$$
(2)
We are going to show that infinite collection of equations (1) restricted by relations (2) are equivalent to finite-dimensional systems complemented by auto-Bäcklund transformations (ABT).
Fix any $`i=i_0𝐙`$ and define finite collection of functions $`𝐲=(y_1(t),\mathrm{},y_{n+1}(t))`$ by
$$y_1=r(i_0),y_2=r(i_0+1),\mathrm{},y_{n+1}=r(i_0+n).$$
Taking into account (2), we easy extract finite-dmensional systems
$$\begin{array}{c}\dot{y}_1=\frac{y_1+\mathrm{}+y_n}{y_2\mathrm{}y_{n+1}}y_1y_2,\hfill \\ \dot{y}_k=y_k(y_{k1}y_{k+1}),k=2,\mathrm{},n,\hfill \\ \dot{y}_{n+1}=y_ny_{n+1}\frac{y_2+\mathrm{}+y_{n+1}}{y_1\mathrm{}y_n}.\hfill \end{array}$$
(3)
With any solution of Volterra chain constrained by (2), for different $`i_0`$, we should have respectively different solutions of the system (3). If the constraint (2) is consistent with Volterra chain (1) then the shift $`i_0i_0+1`$ must yield invertible ABT for the system (3). Define “new” functions
$$\stackrel{~}{y}_1=r(i_0+1),\stackrel{~}{y}_2=r(i_0+2),\mathrm{},\stackrel{~}{y}_{n+1}=r(i_0+n+1).$$
From (2), we easy obtain
$$\stackrel{~}{y}_1=y_2,\mathrm{},\stackrel{~}{y}_n=y_{n+1},\stackrel{~}{y}_{n+1}=\frac{y_2+\mathrm{}+y_{n+1}}{y_1y_2\mathrm{}y_{n+1}}.$$
(4)
By straightforward but tedious calculations, one can verify that if $`(y_1(t),\mathrm{},y_{n+1}(t))`$ is some solution of (3) then, by virtue (4), $`(\stackrel{~}{y}_1(t),\mathrm{},\stackrel{~}{y}_{n+1}(t))`$ also will be solution of (3). So we can conclude that the mapping (4), for the system (3), serve as discrete symmetry. It is easy to deduce inverse to (4). We have
$$y_1=\frac{\stackrel{~}{y}_1+\mathrm{}+\stackrel{~}{y}_n}{\stackrel{~}{y}_1\stackrel{~}{y}_2\mathrm{}\stackrel{~}{y}_{n+1}},y_2=\stackrel{~}{y}_1,\mathrm{},y_{n+1}=\stackrel{~}{y}_n.$$
One defines
$$a_0(i)=r(2i1)r(2i),a_1(i)=r(2i2)r(2i1).$$
These relations are known to realize the map from the solution space of (1) to that of Toda chain equations ,
$$\dot{a}_0(i)=a_1(i+1)a_1(i),\dot{a}_0(i)=a_1(i)\left\{a_0(i)a_0(i1)\right\},i𝐙.$$
(5)
Now observe that, for even $`n`$, replacing in (2) $`i2i1`$, these equations can be rewritten in the form
$$a_0(i)\mathrm{}a_0(i+m1)=a_1(i)a_1(i+1)\mathrm{}a_1(i+m),i𝐙$$
(6)
where $`m=n/2`$.
It is naturally to suppose that constraint (6) also will lead to finite-dimensional systems. To write Toda lattice (5) in more familiar form, one introduces the functions $`\{u_i(t),i𝐙\}`$ by
$$a_0(i)=\dot{u}_i,a_1(i)=e^{u_{i1}u_i}.$$
After that we have
$$\ddot{u}_i=e^{u_{i1}u_i}e^{u_iu_{i+1}},i𝐙.$$
(7)
The constraint (6) becomes
$$\dot{u}_i+\mathrm{}+\dot{u}_{i+m1}=e^{u_{i1}u_{i+m}},i𝐙.$$
(8)
Observe that, in terms of $`\tau `$-function defined through
$$e^{u_i}=\frac{\tau _i}{\tau _{i+1}},$$
the equation (8) takes simple form. Namely, we have bilinear equation
$$D_t\tau _i\tau _{i+m}=\tau _{i1}\tau _{i+m+1}.$$
Recall that Hirota’s $`D`$-operator is defined by
$$D_t^lfg=(_{t_1}_{t_2})^lf(t_1)g(t_2)|_{t_1=t_2=t}.$$
By analogy with Volterra lattice situation, introduce finite collection of functions $`𝐪=(q_1(t),\mathrm{},q_{m+1}(t))`$ identifying
$$q_1=u_{i_0},q_2=u_{i_0+1},\mathrm{},q_{m+1}=u_{i_0+m}.$$
Taking into account the relation (8), we easy obtain finite-dimensional systems
$$\begin{array}{c}\ddot{q}_1=(\dot{q}_1+\mathrm{}+\dot{q}_m)e^{q_{m+1}q_1}e^{q_1q_2},\hfill \\ \ddot{q}_k=e^{q_{k1}q_k}e^{q_kq_{k+1}},k=2,\mathrm{},m,\hfill \\ \ddot{q}_{m+1}=e^{q_mq_{m+1}}(\dot{q}_2+\mathrm{}+\dot{q}_{m+1})e^{q_{m+1}q_1}.\hfill \end{array}$$
(9)
Define
$$\stackrel{~}{q}_1=u_{i_0+1},\stackrel{~}{q}_2=u_{i_0+2},\mathrm{},\stackrel{~}{q}_{m+1}=u_{i_0+m+1}.$$
From (8), we easy deduce the relations
$$\stackrel{~}{q}_1=q_2,\mathrm{},\stackrel{~}{q}_m=q_{m+1},\stackrel{~}{q}_{m+1}=q_1\mathrm{ln}\left[\dot{q}_2+\mathrm{}+\dot{q}_{m+1}\right].$$
(10)
By straightforward calculations, one can check that transformation (10), for any $`m𝐍`$, realize ABT for corresponding system. It is easy to write down inverse to (10). We have
$$q_1=\stackrel{~}{q}_{m+1}+\mathrm{ln}\left[\dot{\stackrel{~}{q}}_1+\mathrm{}+\dot{\stackrel{~}{q}}_m\right],q_2=\stackrel{~}{q}_1,\mathrm{},q_{m+1}=\stackrel{~}{q}_m.$$
We summarize above in the following statement.
Proposition 1. The equations of Volterra and Toda lattices restricted, respectively, by relations (2) and (8) are equivalent to the system (3) and (9) complemented by ABT (4) and (10).
It is easy to get the relations connecting the solution spaces of the systems (3) and (9), in the case when $`n=2m`$. We have
$$y_{2k1}+y_{2k}=\dot{q}_k,k=1,\mathrm{},m,$$
$$y_{2m+1}+\frac{y_2+\mathrm{}+y_{2m+1}}{y_1y_2\mathrm{}y_{2m+1}}=\dot{q}_{m+1},$$
$$y_{2k}y_{2k+1}=e^{q_kq_{k+1}},k=1,\mathrm{},m.$$
Observe that any system (9) can be cast into Lagrangian formalism.
Proposition 2. The systems (9) admit Lagrangian formulation with appropriate Lagrangian
$$=\underset{i<j}{}\dot{q}_i\dot{q}_j\underset{i=1}{\overset{m}{}}e^{q_iq_{i+1}}(\frac{1}{2}\dot{q}_1+\underset{j=2}{\overset{m}{}}\dot{q}_j+\frac{1}{2}\dot{q}_{m+1})e^{q_{m+1}q_1}.$$
Proof. Calculate Euler derivatives of $``$
$$E_{q_i}=\frac{}{q_i}\frac{d}{dt}\left(\frac{}{\dot{q}_i}\right).$$
We have
$$E_{q_1}=\underset{j=2}{\overset{m+1}{}}\ddot{q}_je^{q_1q_2}+\left(\underset{j=2}{\overset{m+1}{}}\dot{q}_j\right)e^{q_{m+1}q_1},$$
$$E_{q_k}=\underset{jk}{}\ddot{q}_je^{q_kq_{k+1}}+e^{q_{k1}q_k}+$$
$$+(\dot{q}_{m+1}\dot{q}_1)e^{q_{m+1}q_1},k=2,\mathrm{},m,$$
$$E_{q_{m+1}}=\underset{j=1}{\overset{m}{}}\ddot{q}_j+e^{q_mq_{m+1}}\left(\underset{j=1}{\overset{m}{}}\dot{q}_j\right)e^{q_{m+1}q_1},$$
Equating $`E_{q_i}`$ to zero results in the system of the form
$$\underset{j=1}{\overset{m+1}{}}A_{ij}\ddot{q}_j=f_i(q_k,\dot{q}_k)$$
(11)
where the matrix $`A=(A_{ij})`$ consists of zeros in main diagonal and units elsewhere. It is easy to see that $`A`$ is nondegenerate. Moreover it can be checked that $`A^1`$ consists of $`(2m1)/m`$’s in main diagonal and $`1/m`$’s elsewhere. One can verify that substituting right-hand sides of the system (9) in (11) gives identity. So we can conclude that (9) and (11) are equivalent. $`\mathrm{}`$
It is simple exercise, using Legendre transformation, to cast any system (9) in Hamiltonian setting. In the simplest case $`m=1`$, (9) becomes
$$\ddot{q}_1=\dot{q}_1e^{q_2q_1}e^{q_1q_2},\ddot{q}_2=e^{q_1q_2}\dot{q}_2e^{q_2q_1}.$$
(12)
Its ABT reads
$$\stackrel{~}{q}_1=q_2,\stackrel{~}{q}_2=q_1\mathrm{ln}(\dot{q}_2).$$
(13)
Note two first integrals for the system (12). They are
$$P=\dot{q}_1\dot{q}_2e^{q_2q_1},E=\dot{q}_1\dot{q}_2+e^{q_1q_2}.$$
It can be checked that $`P`$ and $`E`$ are invariant with respect to ABT (13).
Lagrangian in this case looks as
$$=\dot{q}_1\dot{q}_2e^{q_1q_2}\frac{1}{2}(\dot{q}_1+\dot{q}_2)e^{q_2q_1}.$$
We easy obtain
$$p_1=\frac{}{\dot{q}_1}=\dot{q}_2\frac{1}{2}e^{q_2q_1},p_2=\frac{}{\dot{q}_2}=\dot{q}_1\frac{1}{2}e^{q_2q_1},$$
$$=\dot{q}_1\frac{}{\dot{q}_1}+\dot{q}_2\frac{}{\dot{q}_2}=(p_1+\frac{1}{2}e^{q_2q_1})(p_2+\frac{1}{2}e^{q_2q_1})+e^{q_1q_2}$$
It is easy to check that the functions $``$ and $`P=p_1+p_2`$ are in involution with respect to standard Poisson bracket. So we can conclude that the system (12) is integrable due to Liouville’s theorem. It is naturally to suppose that all systems (9) are completely integrable in the sense of Liouville’s theorem . |
no-problem/0001/hep-ph0001280.html | ar5iv | text | # Preasymptotic effects in beauty decays11footnote 1Talk given by B. Melić at The 3ʳᵈ International Conference on B Physics and CP Violation, Taipei, Taiwan, December 3-7, 1999
## Acknowledgments
B.M. would like to thank H.Y. Cheng and W.S. Hou for invitation to participate in this very stimulating conference. This work was supported by the Ministry of Science and Technology of the Republic of Croatia under Contract No. 00980102. |
no-problem/0001/gr-qc0001005.html | ar5iv | text | # 1 Introduction
## 1 Introduction
The new symmetric property during the time and space of us who propose by this paper is equal to the symmetric property of the already known principle of general relativity. It is a symmetric property with big scale as to be. This symmetric property is caused by the mathematical law in the purity than, the physical law, the symmetric property about the dealing with the summation of the tensor. In the following chapter, it considers this thing and it proves a symmetric property about the dealing with the summation of the tensor first and it makes a new symmetric property during our time and space clear using this next.
## 2 One Proposal about the nature during our time and space
First, it considers about the symmetric property about the dealing with the summation of the position tensor.
When showing the ingredient $`x^1,x^2,x^3,x^4`$ of the position vector during our time and space in $`x^\mu `$ , it decides to show the summation of these ingredients as follows.
$$X^\mu =\underset{\mu =1}{\overset{4}{}}x^\mu $$
(1)
Incidentally, because each of the ingredients are independence, they may use either of $`X^\mu ,x^\mu `$ when showing a position vector but suppose that it uses $`X^\mu `$ in this paper.
The (2.1) formula can be shown as follows when differentiating (2.1) formula first here with ingredient $`x^1`$ and doing the operation to integrate again next about $`x^1`$ in $`x^4`$.
$$X^\mu =x^1,X^\mu =x^2,X^\mu =x^3,X^\mu =x^4$$
(2)
Incidentally, in the (2.2) formula, it used the nature that the ingredient $`x^1,x^2,x^3,x^4`$ of the vector are independence each other.
By the way, it finds that it is possible to treat more in the same way as the scalar by position vector $`X^\mu `$ than the (2.2) formula here. Even if it replaces a scalar in the formula with the vector when the scalar of the vector differentiates, the logical contradiction doesn’t occur to the equation and if using this thing, it finds the fact that it is possible to express in the following formula.
$$dX^\mu =\frac{dX^\mu }{dX^\nu }dX^\nu $$
(3)
This formula is the differential calculus relational expression among two position vectors.
If using this style, two position tensors $`X^\mu ,X^\lambda `$ can be expressed by the following relational expression.
$$X^\mu =\frac{dX^\mu }{dX^\nu }X^\nu $$
(4)
$$X^\lambda =\frac{dX^\lambda }{dX^\tau }X^\tau $$
(5)
Therefore, the relational expression to this summation of two position tensors is from (2.4), the (2.5) formula.
$$X^\mu +X^\lambda =\frac{dX^\mu }{dX^\nu }X^\nu +\frac{dX^\lambda }{dX^\tau }X^\tau =\frac{dX^\mu }{dX^\nu }X^\nu +\frac{dX^\lambda }{dX^\nu }X^\nu $$
(6)
From the inner product of the vector and the distribution law about the outer product.
$$\left[\frac{d\left(X^\mu +X^\lambda \right)}{dX^\nu }\right]X^\nu =\left[\frac{dX^\mu }{dX^\nu }+\frac{dX^\lambda }{dX^\nu }\right]X^\nu =\frac{dX^\mu }{dX^\nu }X^\nu +\frac{dX^\lambda }{dX^\nu }X^\nu $$
(7)
Incidentally, the vector in this place is a tensor.
If using this, the (2.6) formula is
$$X^\mu +X^\lambda =\frac{d\left(X^\mu +X^\lambda \right)}{dX^\nu }X^\nu $$
(8)
If comparing with the (2.4) formula, it finds that it is possible to treat the summation $`X^\mu +X^\lambda `$ of the position tensor in the same of one tensor.
From in the same way on the other hand, (2.4), the (2.5) formula of the relational expression to the product of two position tensors
$$X^\mu X^\lambda =\frac{dX^\mu }{dX^\nu }\frac{dX^\lambda }{dX^\tau }X^\nu X^\tau $$
(9)
If comparing with the (2.4) formula, the product $`X^\mu X^\lambda `$ of the position tensor is different from the case with summation and it isn’t possible to be treated in the same of one tensor by it.
Here, it decides to examine (2.8) formula in detail. When first, transform (2.8) formula and differentiate both sides in $`X^\nu `$
Incidentally, this treatment in the case is the same with the scalar.
$$\frac{dX^\nu }{dX^\nu }=\frac{d^2X^\nu }{d\left(X^\mu +X^\lambda \right)dX^\nu }\left(X^\mu +X^\lambda \right)+\frac{dX^\nu }{d\left(X^\mu +X^\lambda \right)}\frac{d\left(X^\mu +X^\lambda \right)}{dX^\nu }$$
(10)
Because it is a clause of 1st of the right side, $`\frac{dS}{d\left(X^\mu +X^\lambda \right)}\left(X^\mu +X^\lambda \right)=0`$ and moreover both sides are scalar quantity together and can change left side suffix $`\nu `$ into $`\mu `$
$$\frac{dX^\mu }{dX^\mu }=\frac{d\left(X^\mu +X^\lambda \right)}{d\left(X^\mu +X^\lambda \right)}$$
(11)
It finds that there is not hinderance even if $`\mu \lambda `$ meets actually here because the formula of either of $`\mu =\lambda ,\mu \lambda `$ of (2.11) stands up and treats as $`\mu =\lambda `$ in case of treatment.
In other words, actually, the summation of the position tensor can be simplier treated.
Incidentally, it is possible to say that the symmetric property of $`\mu =\lambda `$ exists to the treatment of the summation $`X^\mu +X^\lambda `$ of the position tensor about this thing, too.
By the way, it considers about how a summation with the position tensor which is a place of the substantial adaptation of the symmetric property in case of this treatment next is expressed by the equation.
The summation of the position tensor which shows a position on the time and space which is different here in the future decides to show the suffix of the tensor as the uppercase of the English letter.
Also, it decides to show the summation of all position tensors which compose our whole time and space specifically in $`X^I`$.
First, if limiting relation to the relation of the position tensors, the relation that it is possible that it is possible to have some position tensor $`X^\mu `$ is limited to the relation among all position tensors except $`X^\mu `$.
This thing can be shown in $`X^\mu =f\left(\overline{X^\mu }\right)`$ if deciding to represent $`X^\mu `$ and the position tensor which is independence as being $`\overline{X^\mu }`$.
Because it isn’t clear here about the concrete form of $`f\left(\overline{X^\mu }\right)`$, the definite relational expression about $`X^\mu `$ can not be gotten.
However, because the case where the relational expression can be fixed as prima facie only by the only mathematical request is only one, it makes clear about this case.
Because $`\overline{X^I}`$ doesn’t exist, it becomes $`\overline{X^I}=0`$ to only $`X^I`$.
$`X^I=f\left(0\right)`$ here, therefore, because $`X^I`$ forms no kind of connection, the following formula can be set by it from $`X^I=f\left(\overline{X^I}\right)`$.
$$X^I=0$$
(12)
Here, such nature is only pure mathematical request like the (2.12) formula because it is not and can not form an equation to the summation of the position tensor except $`X^I`$. Therefore, the place of the substantial adaptation of the symmetric property in case of previously shown treatment is only the relational expression which was concerned with $`X^I`$.
Next, it considers about making a symmetric property in case of treatment which was previously shown to the (2.12) formula be adaptable actually.
First, it takes the summation about the position tensor to be treating by the (2.8) formula and it decides to show this in $`X^J`$.
$$X^\mu +X^\lambda +X^\nu =X^N+X^\nu =X^J$$
(13)
The symmetric property in case of treatment which was previously shown here can be handled as follows if adaptable.
$$X^J=X^\mu +X^\nu $$
(14)
This thing shows that it is possible to treat by making all tensors except $`X^\nu `$ the same in $`X^\mu `$ in case of treatment of $`X^J`$.
By the way, next, it thinks of $`X^J`$ and it thinks of extending in $`X^I`$.
First the relation between $`X^J`$ and $`X^I`$
$$X^I=X^J+X^M$$
(15)
If therefore, substitute (2.13) formula
$$X^I=X^N+X^M+X^\nu =X^K+X^\nu $$
(16)
Here, because it thinks that $`X^N`$ changed to $`X^K`$ only in case of $`X^I`$ and that the discussion to the (2.11) formula stands up in case of $`X^K`$, the symmetric property in case of previously shown treatment can be more handled as follows if adaptable than (2.13) formula.
$$X^I=X^\mu +X^\nu $$
(17)
Therefore, $`X^I`$ can be $`X^J`$ in the same way treated and in case of treatment of $`X^I`$, all tensors except $`X^\nu `$ can be treated by making them the same in $`X^\mu `$.
By the dealing which shows (2.12) formula therefore, in (2.17)
$$X^\mu +X^\nu =0$$
(18)
This thing shows that it is possible to treat by making all tensors except $`X^\nu `$ the same in $`X^\mu `$ in all position tensors which compose our whole time and space.
The introduction of this symmetric property will have made dealing with the position tensor of us handy as much as the surprise.
A symmetric property in case of treatment of these us in the time and space is called a binary law in the future and the tensor to be considering this rule is called BINOR.
## 3 Discussion
It made that the new symmetric property, the binary law, existed newly in our time and space at the thing except the symmetric property of the principle of general relativity which is already known in this paper clear. The introduction of this symmetric property will have made dealing with the position tensor of us handy as much as the surprise.
Moreover, it decided to call the tensor which satisfies this binary law this back BINOR.
Incidentally, it plans to make the nature which is peculiar to BINOR clear with the continuing separate sheet of paper to this back this. |
no-problem/0001/hep-ph0001199.html | ar5iv | text | # Cosmological Aspects of Heterotic M-theory
## 1 Introduction
As it was well established by Hořava and Witten , the strong coupling limit of the $`E_8\times E_8`$ heterotic string theory can be described by $`d=11`$ supergravity (SUGRA). This theory can be compactified on a manifold with boundaries, usually expressed as $`X\times S^1/Z_2`$, where $`X`$ is the $`d=6`$ Calabi–Yau manifold and $`S^1/Z_2`$ is the so-called eleventh segment. Altogether the picture we get is that of two walls (so-called hidden and observable) that interact through gravity.
The relevant parameters in this theory are $`V`$, the volume of the $`d=6`$ manifold, and $`\pi \rho `$, the length of the eleventh segment. Among the attractive features of this theory, it is worth mentioning the fact that gauge coupling unification is now entirely natural , and the scale at which the gauge couplings meet, $`10^{16}`$ GeV, is reconciled with the $`d=4`$ Planck scale, $`M_\mathrm{P}10^{18}`$ GeV. For this to happen, the process of compactification must occur in the order $`d=11d=5d=4`$. In other words, in ordet to fit the phenomenologically preferred values for $`\alpha _{\mathrm{GUT}}`$, $`M_{\mathrm{GUT}}`$ and $`M_\mathrm{P}`$, given by
$`\alpha _{\mathrm{GUT}}`$ $`=`$ $`(4\pi )^{2/3}\kappa ^{4/3}V^1,`$
$`M_{\mathrm{GUT}}`$ $`=`$ $`V^{1/6},`$ (1)
$`M_\mathrm{P}`$ $`=`$ $`\kappa ^1\sqrt{\pi \rho V},`$
we need $`\pi \rho (4\times 10^{15}\mathrm{GeV})^1`$ and $`V(3\times 10^{16}\mathrm{GeV})^6`$. $`\kappa ^2`$ is the $`d=11`$ gravitational coupling and its corresponding Planck scale is given by $`M_{11}=\kappa ^{2/9}`$.
Within this framework we are interested in studying phenomena such as supersymmetry (SUSY) breaking and the cosmological evolution of the moduli fields, which occur at scales below the compactification one. Therefore we shall concentrate on the $`d=4`$ effective SUGRA theory from now on.
## 2 Moduli stabilization
We shall discuss the stabilization of moduli as a previous step to address the question of their cosmological evolution. In order to do that, let us notice that the values of $`V`$ and $`\pi \rho `$ are determined by the chiral superfields $`S`$ (the dilaton) and $`T`$ (the modulus). More precisely the real parts of these chiral fields are given by
$`S_\mathrm{R}`$ $``$ $`\kappa ^{4/3}VO(\alpha _{\mathrm{GUT}}^1),`$ (2)
$`T_\mathrm{R}`$ $``$ $`\kappa ^{2/3}\pi \rho V^{1/3}O(\alpha _{\mathrm{GUT}}^1).`$
Note that, in the weakly coupled heterotic case, $`S_\mathrm{R}O(\alpha _{\mathrm{GUT}}^1)O(20)`$, the same as here, whereas $`T_\mathrm{R}O(1)`$ (always in $`M_\mathrm{P}`$ units).
To study the dynamical behaviour of these fields, and whether they acquire the desired vacuum expectacion values (VEVs), we analyse the scalar potential
$$V=e^K\left\{(W_i+K_iW)(K_i^j)^1(\overline{W}^j+K^j\overline{W})3|W|^2\right\},$$
(3)
where $`W(S,T)`$ is the superpotential and $`K(S_\mathrm{R},T_\mathrm{R})`$ is the Kähler potential. The sub (super) indices indicate derivatives of the functions with respect to the (conjugate) fields. Together with the gauge kinetic functions, $`f_a`$, $`W`$ and $`K`$ determine the SUGRA Lagrangian.
To be more precise, we shall assume a particular source for SUSY breaking, which is gaugino condensation in the hidden wall . This is, so far, the most promising mechanism for breaking SUSY at the right scale in order to give rise to an acceptable phenomenology. It assumes the existence of a strong-type interaction in the hidden sector of the theory which, below a certain scale $`\mathrm{\Lambda }`$, triggers the formation of gaugino condensates, $`\lambda \lambda `$, and the breakdown of SUSY. More precisely,
$$W\lambda \lambda \underset{i=1}{\overset{N}{}}C_ie^{\alpha _if_i},$$
(4)
where the sum runs over all the condensing groups in the hidden sector ($`G_H=G_1\times G_2\times \mathrm{}\times G_N`$), $`\alpha _i`$ are proportional to the 1-loop $`\beta `$-function coefficients associated to each condensing group, $`C_i`$ are also related to each group’s characteristics and $`f_i`$ are the already-mentioned gauge kinetic functions. In particular, in the hidden sector $`f_i=(Sn_iT/2)/4\pi `$, with $`n_i`$ being model dependent coefficients and, in the observable wall, $`f_{\mathrm{obs}}=(S+n_{\mathrm{obs}}T/2)/4\pi `$.
The Kähler potential is given by the expression
$$K=K_0+K_{\mathrm{np}},$$
(5)
where $`K_0=\mathrm{ln}(2S_\mathrm{R})3\mathrm{ln}(2T_\mathrm{R})`$ is the tree level piece and $`K_{\mathrm{np}}`$ stands for M-theoretic non-perturbative effects. For the latter we will use a specific ansatz .
Now that we have defined all the functions we need to proceed with the analysis of the scalar potential, Eq. (3). We have first of all confirmed the results of Choi et al. (obtained with a slightly different ansatz for $`K_{\mathrm{np}}`$), namely that for one condensate only the weakly coupled minimum exists. With two condensates ($`W=C_1e^{\alpha _1f_1}+C_2e^{\alpha _2f_2}`$), however, things become more interesting. It is possible to find plenty of examples of condensing groups for which both moduli are fixed at the desired VEVs and SUSY is broken at the right scale (i.e. the gravitino mass, $`m_{3/2}`$, is of order 1 TeV). In fact, in order to understand the vacuum structure a bit better, it is convenient to redefine the fields $`S`$ and $`T`$ as follows:
$``$ The combinations
$`\mathrm{\Phi }_\mathrm{r}^{}`$ $``$ $`S_\mathrm{R}{\displaystyle \frac{n_1\alpha _1n_2\alpha _2}{2(\alpha _1\alpha _2)}}T_\mathrm{R},`$ (6)
$`\mathrm{\Phi }_\mathrm{i}^{}`$ $``$ $`S_\mathrm{I}{\displaystyle \frac{n_1\alpha _1n_2\alpha _2}{2(\alpha _1\alpha _2)}}T_\mathrm{I}={\displaystyle \frac{4\pi ^2k}{\alpha _1\alpha _2}}`$
are fixed by the interplay between condensates, very much in the same way as the racetrack mechanism worked in the weakly coupled heterotic case (note, from the second equation, that the condensates at the minimum, i. e. for $`k`$ odd, are in opposite phase).
$``$ The orthogonal combinations are potentially flat
$`\mathrm{\Phi }_\mathrm{r}^+`$ $``$ $`{\displaystyle \frac{n_1\alpha _1n_2\alpha _2}{2(\alpha _1\alpha _2)}}S_\mathrm{R}+T_\mathrm{R},`$ (7)
$`\mathrm{\Phi }_\mathrm{i}^+`$ $``$ $`{\displaystyle \frac{n_1\alpha _1n_2\alpha _2}{2(\alpha _1\alpha _2)}}S_\mathrm{I}+T_\mathrm{I}.`$
In fact we can easily check that these combinations of the fields do not appear in the superpotential. Therefore the potential flatness of $`\mathrm{\Phi }_\mathrm{r}^+`$ will be lifted by the presence of the Kähler potential, which depends on $`S_\mathrm{R}`$, $`T_\mathrm{R}`$, whereas $`\mathrm{\Phi }_\mathrm{i}^+`$ remains totally flat. This is a completely new feature associated to M-theory models, which depends entirely on the structure of the gauge kinetic functions $`f_a`$.
Let us proceed to analyse cases with more than two condensates. There the definitions of $`\mathrm{\Phi }_{\mathrm{r},\mathrm{i}}^{}`$ and $`\mathrm{\Phi }_{\mathrm{r},\mathrm{i}}^+`$ can be easily generalized to many condensates, and it is also possible to show that the flat direction will only remain so if and only if all the $`n_i`$ coefficients that enter the definition of the gauge kinetic functions, $`f_i`$, are the same. This is shown in Fig. 1 where we plot the scalar potential as a function of $`\mathrm{\Phi }_\mathrm{i}^\pm `$, for $`\mathrm{\Phi }_\mathrm{r}^\pm `$ fixed, in the case of two (Fig. 1a) and three (Fig. 1b) condensates.
As we can see, the flat direction of Fig. 1a is lifted by the presence of a third condensate with $`n_3n_1=n_2`$. This opens up new scenarios from the cosmological point of view, as we are about to see.
## 3 Cosmological Evolution and Moduli Problem
Let us then study the possible cosmological role of the $`\mathrm{\Phi }_{\mathrm{r},\mathrm{i}}^\pm `$ fields.
$``$ $`\mathrm{\Phi }_\mathrm{r}^{}`$ has an exponential-type potential. In fact it is the analogous of the dilaton $`S_\mathrm{R}`$ in the weakly coupled heterotic string and will therefore suffer from the same problems . The potential is too steep just before the minimum and the field tends to roll past it towards infinity. It can be, however, stabilized in the presence of a dominating background but, in any case, it is totally unsuitable as an inflaton.
$``$ $`\mathrm{\Phi }_\mathrm{i}^{}`$ has a steep sinusoidal potential, analogous to that of $`S_\mathrm{I}`$ in the weakly coupled heterotic case. Again, this field is not a good candidate for an inflaton.
On the other hand, the fields $`\mathrm{\Phi }_{\mathrm{r},\mathrm{i}}^+`$ have more promising potentials, as it was pointed out above. It is then worth studying their evolution equations in the presence of an expanding Universe
$`K_{S\overline{S}}(\ddot{S}+3H\dot{S})+K_{SS\overline{S}}\dot{S}^2+{\displaystyle \frac{V}{\overline{S}}}=0,`$ (8)
$`K_{T\overline{T}}(\ddot{T}+3H\dot{T})+K_{TT\overline{T}}\dot{T}^2+{\displaystyle \frac{V}{\overline{T}}}=0,`$
where $`H^2=\frac{1}{3}K_{S\overline{S}}\dot{S}\dot{\overline{S}}+\frac{1}{3}K_{T\overline{T}}\dot{T}\dot{\overline{T}}+\frac{V}{3}`$ is the Hubble constant. It is crucial to note here that both $`S`$ and $`T`$ have non-minimal kinetic terms, i.e. $`_{kin}=K_{S\overline{S}}D_\mu SD^\mu \overline{S}+K_{T\overline{T}}D_\mu TD^\mu \overline{T}`$. This is what introduces the new terms in the evolution equations (8).
Let us start with the $`\mathrm{\Phi }_\mathrm{r}^+`$ direction. It is easy to show that the potential along this direction behaves as an inverse power-law, i.e. $`V(\mathrm{\Phi }_\mathrm{r}^+)=A/(\mathrm{\Phi }_\mathrm{r}^+)^n`$, at least along the slope to the left of the minimum, which is the relevant one in terms of the evolution. By solving the evolution equations we have checked that the field inflates for a few e-folds before reaching its minimum. These kinds of inflationary scenarios are denoted as ‘intermediate’ inflation in the literature . An important thing to notice is that, if the field $`\mathrm{\Phi }_\mathrm{r}^+`$ were canonically normalized, then this inverse power-law potential would have given enough e-folds of inflation.
Finally we turn to the remaining direction to be analysed, namely $`\mathrm{\Phi }_\mathrm{i}^+`$: as we had said before, in the two-condensate case this is a totally flat direction, however when we introduce a third condensate with $`n_3n_1=n_2`$, things change. It can be shown that the potential in this latter case is a sinusoidal one,
$$V=V_0[1+\mathrm{cos}(a\mathrm{\Phi }_\mathrm{i}^++b)],$$
(9)
where $`a`$ and $`b`$ are functions of $`n_3`$. It is easy to see that the closer $`n_3`$ is to $`n_1`$ and $`n_2`$ the flatter the potential is. Therefore we have a way of controlling the flatness of the potential, which can lead us to successful examples of inflation. A couple of them are shown in Fig. 2, where we plot $`\mathrm{\Phi }_\mathrm{i}^+`$ as a function of the number of e-folds $`N`$. It is clear then that a combination of the imaginary parts of the dilaton $`S`$ and modulus $`T`$ fields, what we have denoted as $`\mathrm{\Phi }_\mathrm{i}^+`$, can be a suitable inflaton in the context of heterotic M-theory. These are examples of the so-called natural inflation .
Let us finish this section by discussing very briefly the status of the moduli problem within these models. This arises when we have very weakly interacting particles with VEVs of the order of $`M_\mathrm{P}`$ and light masses, of the order of $`m_{3/2}`$. If these relics decay, they must do it before nucleosynthesis in order not to ruin its predictions. This imposes a lower bound on their masses of $`10`$ TeV. On the other hand, if these particles are stable, their oscillations should not overclose the Universe. This sets an upper bound of $`10^{24}`$ eV on their masses. In the weakly coupled heterotic string case, where all moduli masses were of order $`1`$ TeV, these bounds were obviously difficult to fulfil .
We have calculated these masses for the present models. In general, we obtain that, along the $`\mathrm{\Phi }_{\mathrm{r},\mathrm{i}}^{}`$ directions, the corresponding masses are of the order of $`10^3m_{3/2}`$, well above the lower bound for decaying particles; along $`\mathrm{\Phi }_\mathrm{r}^+`$ they are of the order of $`10m_{3/2}`$, which may be just enough to save the bound, and for $`\mathrm{\Phi }_\mathrm{i}^+`$ they are very small and dependent on how close $`n_3`$ is to $`n_1`$, $`n_2`$. For example, for $`\mathrm{SU}(3)_{M=0}\times \mathrm{SU}(4)_{M=8}\times \mathrm{SU}(2)_{M=0}`$ with $`n_1=n_2=1`$ we find $`m_{\mathrm{\Phi }_\mathrm{i}^+}1.2\times 10^3m_{3/2}`$ for $`n_3=0.98`$ and $`m_{\mathrm{\Phi }_\mathrm{i}^+}1.5\times 10^4m_{3/2}`$ for $`n_3=0.995`$. There is therefore a conflict with the upper bound for stable particles, and the particle excess should be washed away with a period of thermal inflation .
## 4 Conclusions
We have studied the dynamics of the two typical M-theory moduli, namely $`\pi \rho `$ and $`V`$ in terms of $`S`$, $`T`$. In the presence of non-perturbative corrections to the Kähler potential, and gaugino condensation as the source of SUSY breaking, the scalar potential presents very interesting features.
$``$ It depends on $`S`$, $`T`$ essentially through $`f_a=Sn_aT/2`$, motivating a redefinition of fields to $`\mathrm{\Phi }_{\mathrm{r},\mathrm{i}}^{}`$ and $`\mathrm{\Phi }_{\mathrm{r},\mathrm{i}}^+`$.
$``$ $`\mathrm{\Phi }_{\mathrm{r},\mathrm{i}}^{}`$ behave similarly to the dilaton of weakly coupled heterotic string theory: unsuitable as inflatons.
$``$ $`\mathrm{\Phi }_\mathrm{r}^+`$ would be flat in the absence of $`K`$: it has an inverse power-law potential, which gives very little intermediate inflation.
$``$ $`\mathrm{\Phi }_\mathrm{i}^+`$ is totally flat for two condensates and can have a sinusoidal potential for three or more: very promising inflaton (many e-folds of natural inflation).
$``$ There might be a moduli problem associated to this almost flat direction. These lighter moduli can be diluted with a small period of weak scale inflation.
## Acknowledgments
We thank the organizers for their hospitality. The work of TB is supported by PPARC. |
no-problem/0001/cond-mat0001106.html | ar5iv | text | # Transport properties and point contact spectra of 𝑁𝑖_𝑥𝑁𝑏_{1-𝑥} metallic glasses
## Abstract
Bulk resistivity and point contact spectra of $`Ni_xNb_{1x}`$ metallic glasses have been investigated as functions of temperature (0.3-300 K) and magnetic field (0-12 T). Metallic glasses in this family undergo a superconducting phase transition determined by the $`Nb`$ concentration. When superconductivity was suppressed by a strong magnetic field, both the bulk sample $`R(T)`$ and the point contact differential resistance curves of $`Ni_xNb_{1x}`$ showed logarithmic behavior at low energies, which is explained by a strong electron - “two level system” coupling. We studied the temperature, magnetic field and contact resistance dependence of $`Ni_{44}Nb_{56}`$ point-contact spectra in the superconducting state and found telegraph-like fluctuations superimposed on superconducting characteristics. These R(V) characteristics are extremely sensitive detectors for slow relaxing ”two level system” motion.
Amorphous metallic alloys have been the subject of extensive investigations over the last two decades because of their surprising transport, magnetic and superconducting properties. To explain the most unusual low-temperature physical properties of glassy materials the concept of two-level tunneling systems (TLS) was suggested (for a review see ). According to a theoretical model , the electron scattering processes on TLS result in the commonly occurring logarithmic temperature dependence of the resistivity. On the other hand the resistivities of some amorphous alloys ($`Ni_xNb_{1x}`$, $`Fe_xAu_{1x}`$) were claimed not to exhibit any logarithmic feature at low temperatures and therefore these alloys were regarded as a different class of metallic glasses.
Point contact (PC) spectroscopy offers a very sensitive method to investigate scattering processes in conducting materials, since back scattering of electrons in a PC causes a noticeable change in the current through the PC and the R(V) characteristics give quantitative information about the energy dependence of the electron scattering processes on quasiparticles (TLS, phonons, magnons, etc.,). In particular, the strong electron-TLS coupling gives rise to a peak in the PC differential resistance around $`V=0`$. This phenomenon is usually referred to as the zero bias anomaly (ZBA). Experiments on metallic PCs containing nonequilibrium defects as well as on metallic glass PCs demonstrated the existence of ZBA suggested by the theory .
The present study was motivated by the expected absence of the logarithmic peak in $`Ni_xNb_{1x}`$, and was aimed at investigating TLS scattering in these metallic alloys with the help of PC spectroscopy. The $`Ni_xNb_{1x}`$ metallic glasses (MG) are also interesting from another point of view, since a superconducting ground state develops at high enough $`Nb`$ concentration . We examined $`Ni_{44}Nb_{56}`$ and $`Ni_{59}Nb_{41}`$ by measurements on bulk samples investigating the temperature and magnetic field dependence of resistivity and by PC spectroscopy based on break junction technique . This technique permits very stable PC in the resistance range 1-200 $`\mathrm{\Omega }`$ to be made by breaking the sample in ultra high vacuum and then forming the contact between the freshly fractured atomically clean surfaces. Due to the relatively large resistivity of $`Ni_xNb_{1x}`$ MG these contacts were basically in the Maxwell limit , where the contact resistance is calculated as $`R_{PC}=\rho /d`$ ($`\rho `$ being the electrical resistivity, $`d`$ the contact diameter).
Figure 1 presents the results of bulk sample measurements. We found that $`Ni_{44}Nb_{56}`$ has a superconducting transition at 1.8 K, while the resistance of $`Ni_{59}Nb_{41}`$ starts to decrease only below 700 mK indicating a superconducting transition just outside of the temperature range of the measurements (300 mK). For $`Ni_{59}Nb_{41}`$ we observed logarithmic behavior between 700mK and 25K . It is hard to confirm the logarithmic character for $`Ni_{44}Nb_{56}`$ because of the interfering presence of the superconducting fluctuations. In our experiments these fluctuations start at 8 K, where the $`R(T)`$ curve splits from that of the non-superconducting sample. This broad range of fluctuations is in good quantitative agreement with theoretical calculations and experiments on other amorphous superconductors . The inset in Fig. 1 shows a similarly broad magnetic field region of superconducting fluctuations (1.5-2.6 T) at constant temperature (300 mK). After suppressing the superconductivity by applying a magnetic field of 12 T the temperature dependence of the resistivity in both samples shows a clear logarithmic behavior up to 25 K. In this logarithmic region a small but noticable magnetic resistance is observed. Above the temperature range shown in Fig. 1 both samples exhibited decreasing resistivity with increasing temperature. The normal state resistivity at 50 K was $`\rho 2.5\mu \mathrm{\Omega }m`$ for $`Ni_{44}Nb_{56}`$ and $`\rho 1.6\mu \mathrm{\Omega }m`$ for $`Ni_{59}Nb_{41}`$.
The PC spectra measurements showed a common behavior of $`dV/dI(V)`$ for all contacts in the resistance range $`R_{PC}=130\mathrm{\Omega }`$ (corresponding to the contact diameter $`d`$ = 2000 - 60 nm) which means rather good reproducibility of results for different samples. For $`R_{PC}>60\mathrm{\Omega }`$ ($`d<`$ 30 nm) individual features start to prevail and the $`dV/dI(V)`$ curves for the samples of the same resistance may differ significantly. This sets the length scale of the material inhomogeneity to $``$30 nm. This value is in agreement with the small angle neutron scattering measurements performed on these metallic glasses where inhomogeneity was found on the length scale of $``$18 nm. The regime of electron flow in point contacts depends on the relationship between the contact diameter $`d`$ and the elastic $`(l_{el})`$ and the inelastic $`(l_{in})`$ mean free paths . Due to the large resistivity of $`Ni_xNb_{1x}`$ MG, the transition to the thermal limit ($`l_{el},l_{in}d`$) for low ohmic contacts occurs at small voltage bias (because of the strong energy dependence of the inelastic mean free path). In the thermal regime the excess electron energy is dissipated inside the contact, which results in the increase of the point contact temperature with respect to the bias voltage : $`T_{PC}^2=T_{bath}^2+V^2/4L`$, where L is the Lorenz number. This equation relates R(V) measurements done by PC technique to the temperature dependence of the resistivity.
Figure 2 shows the differential resistance of $`Ni_{59}Nb_{41}`$ junctions. The low ohmic contacts (curve 1) exhibit clear logarithmic peaks in the voltage region of 1-12 mV, as presented on the enlarged scale for a $`4\mathrm{\Omega }`$ junction. According to the above equation and calculating by the standard Lorenz number, the bias voltage 12 meV corresponds to $`T_{PC}=38`$ K in the thermal regime, which is somewhat higher than the border of the logarithmic region ($`25`$ K) in the bulk sample measurements of $`R(T)`$. This difference is due to the high resistivity of the material: the voltage drop in the vicinity of the contact is comparable with the voltage drop over the bulk part of the sample, which shifts the logarithmic region towards higher voltages. The decrease in the contact resistance between 0 and 50 mV is comparable to that for the bulk samples in the temperature range 5-160 K.
The $`dV/dI(V)`$ dependences for high ohmic $`Ni_{59}Nb_{41}`$ junctions show step-like singularities at high biases (curve 2, Fig. 2), which can be repeatedly reproduced for the same contact but vary in amplitude and position for different samples. The origin of these high bias anomalies is the subject of ongoing investigations.
The PC characteristics of the superconducting $`Ni_{44}Nb_{56}`$ MG below the critical temperature are quite different from these in ordinary superconductors and can be understood only qualitatively.
Figure 3 shows the PC differential resistance and I-V curves of $`Ni_{44}Nb_{56}`$ at different contact resistances. The junctions with small normal-state resistance ($`1.5\mathrm{\Omega }`$) present conventional $`IV`$ curves (Fig. 3b) of a current driven contact with a clear voltage jump above a certain critical current value and with excess current at high voltages. The evaluation of the excess current for these low resistance junctions shows, that the normal resistance - excess current product is constant giving the close-to-BCS value of $`3.2\pm 0.2`$ for $`2\mathrm{\Delta }/k_BT_c`$. For higher resistances the $`IV`$ curves are smeared, $`R_NI_{exc}`$ vanishes and an increasing residual resistance is observed at zero bias. We found that this residual resistance increases rapidly for decreasing contact diameter and may differ significantly for contacts with the same $`R_N`$. The transition between the jump-like curves and the smeared ones is also sample dependent, varying between $`12\mathrm{\Omega }`$. These phenomena can be understood qualitatively in terms of percolation-like superconductivity. In large enough contacts the current can find continuous superconducting percolation paths between the two electrodes, but below a certain contact diameter no such paths exist any more, and the current must flow through normal regions as well. In this case the residual resistance is determined by the fraction of normal and SC regions along the current paths, which explains the strongly contact-dependent behavior. The characteristic width of percolation paths is most probably close to the material inhomogeneity scale of 18 nm . This size scale of percolation is in agreement with the value of coherence length ($`10`$ nm) calculated from $`H_{c2}`$.
As Fig. 4 shows, the step-like $`IV`$ curve is smeared under the influence of magnetic field as well, but the zero bias resistance and the excess current remains constant up to 1.2 T. We believe that this smearing follows from the vortex dynamics at high current densities in the contact area: the resistance caused by vortices is superimposed on the step-like zero-field $`IV`$ curve. The $`B=1.6`$ T and $`2`$ T curves are already within the fluctuation region of $`H_{c2}`$ (see inset in Fig. 1), thus one obtains completely different $`IV`$ curves with larger zero bias resistance and small excess current. Going above $`H_{c2}`$, at $`B=5`$ T we regain the positive logarithmic peak attributed to electron scattering on two level systems (see inset in Fig. 4).
Recording the differential resistance curves of $`Ni_{44}Nb_{56}`$ as the function of temperature, we found that the transition is broadening by decreasing contact diameter. Similar behavior was observed by Naidyuk et al. in superconducting heavy fermion point contacts .
The differential resistance of some contacts displays large fluctuations around zero bias (Fig. 5). The plot of this noise as the function of time shows that the resistance is switching between two (or more) discrete states on the time-scale of seconds. This slow two level fluctuation is not sensitive to magnetic fields up to 1.5 T and decreases rapidly at larger fields. In a similar fluctuation was superimposed on the logarithmic ZBA in $`Fe_{80}B_{20}`$ and $`Fe_{32}Ni_{36}Cr_{14}P_{12}B_6`$ metallic glasses. This fluctuation was explained as the effect of slowly moving defects influencing electron-TLS coupling. In $`Ni_{44}Nb_{56}`$ contacts the motion of such relatively large defects can result in shutting down one of the percolation paths and suppressing superconductivity in a sizeable part of the constriction. These fluctuations were only observed in relatively small contacts ($`d200`$ nm). In such small areas only a few percolation paths are present, which explains that shutting down one of them has an observable effect. It makes the superconducting characteristics an extremely sensitive detector for the slow relaxing TLS motion.
In conclusion we demonstrated that in contrast to earlier observations both the bulk resistivity and the PC differential resistance of amorphous $`Ni_xNb_{1x}`$ alloys exhibit low-energy logarithmic behavior which is characteristic of electron scattering on the fast relaxing TLSs in full accordance with the Vladar-Zawadowski model . In $`Ni_{59}Nb_{41}`$ we found reproducible structures in the point contact spectra at high biases and higher ohmic contacts. We also studied the unusual features of superconducting $`Ni_{44}Nb_{56}`$ contacts which can be explained by a percolation type of superconductivity, heating effects in the normal phase with increasing bias and the influence of slow configurational changes close to the contact.
We acknowledge E. Sváb, A. Zawadowski and I.K. Yanson for useful discussions. This work was supported by the Nederlandse Organisatie voor Wetenschappelijk Onderzoek (NWO), the Stichting Fundamenteel Onderzoek der Materie (FOM) and by the Hungarian National Science Foundation under grant No. T026327. |
no-problem/0001/astro-ph0001370.html | ar5iv | text | # A Third Star in the T Tauri System
## 1 Introduction
Since its discovery in 1981 (Dyck, Simon, & Zuckerman 1982) the T Tauri Infrared Companion (T Tauri IRC, or T Tauri S) has been the subject of numerous observational and theoretical studies. These have revealed an object with a total luminosity which appears comparable to that of the visible northern star T Tauri N, but which radiates primarily at infrared wavelengths longward of 2 $`\mu `$m, with a V-K color $`\mathrm{¿}\mathrm{}12.7`$ (Stapelfeldt et al. 1998b). Physical interpretations have ranged from a lower-mass (and therefore less evolved and more deeply embedded) pre-main sequence star (Dyck, Simon, & Zuckerman 1982), to a true protostar (Bertout 1983), or a low-luminosity embedded accretion disk (“mini-FUor”; Ghez et al. 1991). Alternatively, it may be a normal T Tauri star and coeval with its primary T Tauri N, but which suffers stronger extinction due either to a special viewing geometry with respect to the surrounding diffuse material (e.g., Calvet et al. 1994; van Langevelde et al. 1994) or because it is experiencing an episode of enhanced accretion, perhaps tied to its orbital phase (Koresko, Herbst, & Leinert 1997).
This Letter presents the results of new near-infrared speckle holographic imaging of the T Tauri IRC. The IRC is found to be a close double with an angular separation of 0″.05, corresponding to a sky-projected linear separation of 7 AU at the 140 pc distance to the Taurus star-forming region. The two objects which make up the IRC are designated T Tauri Sa and T Tauri Sb. They are indistinguishable from point sources at the 0″.05 diffraction-limited resolution of the telescope, suggesting that they are probably stars. This would make T Tauri at least an hierarchical triple system. In addition to these three stars, the system contains the radio source T Tauri R (Schwartz et al. 1984) and a possible protostar 30″ to the southwest (Weintraub et al. 1999).
## 2 Observations and Results
The new holography data were taken in the CH4 (2.19 – 2.34 $`\mu `$m FWHM) filter at the 10 m Keck 1 telescope on 15 December 1997 using the Near-Infrared Camera (Matthews & Soifer 1994). The observation and data reduction techniques were similar to those described in Koresko et al. (1999). The NIRC Image Converter (Matthews et al. 1996) produced a magnified pixel spacing of 0″.02, approximately Nyquist-sampling the diffraction limit at 2 $`\mu `$m. The observations consisted of 500 exposures of the visual binary, with integration times of 0.15 sec for each frame. This short exposure time partially “froze” the atmospheric seeing, so that the point-spread function (PSF) consisted of distinguishable speckles. The 0″.7 separation of the binary was large enough compared to the typical seeing that the PSFs could be separated fairly cleanly in most of the frames.
Individual frames were calibrated in the standard way by subtracting mean sky frames, dividing by flatfield images, and “fixing” bad pixels. A model was computed for the “bleed” signal which extended along the readout direction, and this was subtracted from the calibrated frame. For each frame, a $`32\times 32`$ pixel subframe centered on the IRC was extracted, and a similar subframe centered on the primary served as a measurement of the instantaneous point-spread function (PSF). In 26 of the frames, the instantaneous seeing was too poor for this procedure to cleanly separate the stars, and these frames were rejected. For the remaining 474 frames, the Fourier power spectrum of the PSF frames, and the cross-spectrum (i.e., the Fourier transform of the cross-correlation) of the masked frame with the unmasked frame, were computed. If the primary star is unresolved, then in principle the ratio of the cross-spectrum to the PSF frame’s power spectrum is the Fourier transform of the diffraction-limited image. In practice, it was first necessary to correct for noise-bias terms in both the cross-spectrum and the power spectrum.
The cross-spectrum and the PSF power were accumulated over the whole series of good frames. An averaged Fourier power spectrum and Fourier phase for the IRC were reconstructed from them, and frequencies along the u-axis, which were corrupted by a small amount of flux from one star leaking into the subframe containing the other, were “fixed” by setting their values to those of their neighbors off the axis. An image was reconstructed from the fixed power and phase with the use of an apodizing function to suppress high-frequency noise. The apodizing function chosen was the product of a Gaussian and a Hanning function. It produced a final image resolution of 49 mas (FWHM). The Fourier components and the image were then rotated to standard orientation. They are presented in Figure 1.
The power spectrum and phase show the striping patterns characteristic of a binary star in a roughly hexagonal region containing spatial frequencies below the diffraction limit of the hexagonal Keck primary mirror. Models were fit to the Fourier components in order to derive the brightness ratio, separation, and position angle of the IRC binary. The fitting was done separately for the power and phase. The quality of a fit was estimated by visual inspection of a display of the ratio of the measured power to the model power, and of the difference between the measured phase and the model phase. Limiting parameter values were those for which the binary-star striping patterns began to be apparent. The estimates resulting from the power and phase fits are consistent with each other and have comparable uncertainties. The fainter companion, designated T Tauri Sb, is found to lie 53 $`\pm `$ 9 mas ($`7`$ AU) from its neighbor T Tauri Sa along position angle 225 $`\pm `$ 8 deg, and to have only 0.09 $`\pm `$ .02 times its brightness. The total flux in the IRC was found to be 0.32 times that from T Tauri N. If the K-band (2.2 $`\mu `$m) magnitude of the system taken as a whole is 5.4 (Rydgren, Schmelz, & Vrba 1982) then applying the brightness ratios measured in the CH4 filter implies K-band magnitudes of 5.7, 7.0, and 9.6 for T Tauri N, T Tauri Sa, and T Tauri Sb, respectively.
## 3 Discussion
There are two reasons to believe that the double structure seen in the IRC represents a pair of stars and not, e.g., the two bright scattering lobes at the poles of a circumstellar disk seen nearly edge-on (e.g., Koresko 1998; Wood et al. 1998). The first is that they are compact enough to appear pointlike at the 0″.05 resolution of the holographic data. The second is that the position angle of the line joining them differs by $`45`$ deg from that of the North-South outflow associated with the IRC (Solf & Bohm 1999); if the IRC were a single star surrounded by a disk, one would expect the disk’s polar axis to be parallel to the jet.
The presence of a second stellar component in the IRC suggests a simple explanation for the small disk mass implied by the nondetection of the IRC in the submillimeter interferometric measurements made by Hogerheijde et al. (1997) and Akeson et al. (1998), which constrain the mass of any disk in the IRC to be no more than 3$`\times 10^3`$ M. A binary companion is expected to truncate a circumstellar disk to a radius $`\frac{1}{3}`$ of the binary separation (Lin & Papaloizou 1993; Artymowicz & Lubow 1994), which would suggest a maximum disk radius of only $`2`$ AU in the T Tauri IRC. Although poorly constrained by observations, most disk models assume surface density profiles which place the majority of the mass at larger radii (e.g., Beckwith et al. 1990).
It is clear that both of the stars in the IRC suffer very strong extinction. Visible-light imaging using the Hubble Space Telescope indicates that the IRC as a whole has $`\mathrm{V}>19.6`$ (Stapelfeldt et al. 1998b). Even assigning this V-band magnitude to T Tauri Sb would give it a V-K color of 10 mag, making it much redder than any stellar photosphere, and any other assumption about the origin of the visible light would require the bluer of the two IRC stars to be redder still. An extinction of $`\mathrm{A}_\mathrm{V}35`$ would be required to redden a normal stellar photosphere to match the near-IR color of the IRC (Koresko, Herbst, & Leinert 1997).
The origin of the large extinction to the T Tauri IRC has been the subject of much recent speculation. It is not obviously surrounded by an optically-thick scattering envelope as is the IRC orbiting Haro 6–10, another T Tauri star in the Taurus cluster (Koresko et al. 1999). One possibility is that the T Tauri IRC lies behind the disk associated with T Tauri N (e.g., van Langevelde et al. 1994). In this picture, the IRC could be intrinsically quite similar to T Tauri N, and its unusual observational properties a result of by its special viewing geometry. As noted by Akeson et al. (1998), although the radius of the T Tauri N disk appears smaller in their submillimeter images than the distance to the IRC, submillimeter imaging cannot rule out the existence of a more diffuse outer disk such as that proposed by Hogerheijde et al. (1997).
However, this simple picture by itself cannot completely account for the strange properties of the IRC. The T Tauri IRC is one of only two known pre-main sequence sources of nonthermal, circularly polarized radiation at centimeter wavelengths (Phillips et al. 1993; Skinner & Brown 1994), the other being the Class 1 protostar IRS 5 in the Corona Australis “Coronet Cluster” (Feigelson, Carkner, & Wilking 1998). This observation hints at the action of some unusual energetic process, perhaps accretion-driven, involving strong magnetic fields.
The North-South jet associated with the IRC lies only $`11`$ deg from the plane of the sky (Solf & Bohm 1999). This jet, which has apparently been traced over a distance of $`1.5`$ pc in a giant Herbig-Haro flow (Reipurth, Bally, & Devine 1997), provides independent evidence for rapid accretion. Its axis is nearly perpendicular to the jet from T Tauri N, whose axis lies close to the line of sight and has an East-West sky-projected direction. A variable accretion rate in a luminous disk has been proposed to account for a 2-magnitude brightening seen during the period from 1987-1991 (Ghez et al. 1991).
The observations presented here are not sensitive the IRC’s jet, so it is not clear which of the stars in the IRC is responsible for driving it. The orientation of the IRC’s jet close to the plane of the sky suggests that the jet source is likely to be surrounded by a disk viewed nearly edge-on. Integration of the density profile of the model disk described by Wood et al. (1998) along a line of sight 11 deg from the disk plane, and extending from the star to a disk cutoff radius of 2 AU, shows that the large extinction required for the IRC could easily be produced with reasonable parameter values, even given the small mass implied by the submillimeter maps. In this picture, both of the stars in the IRC would need to be surrounded by nearly edge-on disks to account for the large extinction they suffer.
The example of HK Tauri (Stapelfeldt et al. 1998; Koresko 1998) suggests the possibility that the observed near-infrared light may emerge via scattering in the diffuse upper regions of such a disk, rather than simply being highly-reddened light directly from the stellar photosphere. The maximum vertical thickness of such a disk would need to be small compared to the 0″.05 resolution of the holographic observations, which is plausible if the disk has been truncated. In this picture, the shape of the observed spectral energy distribution may not be representative of the of the true extinction to the star. Changes in a disk which processes stellar photons via a combination of extinction and scattering might explain how the IRC was able to vary in brightness by a nearly wavelength-independent factor of 5 between 1.65 and 10 $`\mu `$m (Ghez et al. 1991). Detailed radiative-transfer calculations will be needed to test this possibility.
If T Tauri Sa and T Tauri Sb are indeed stars and have masses $`1\mathrm{M}_{}`$, then their small separation will result in an orbital period of only $`10`$ yr, making the orbital motion readily detectable with the holography technique on timescales of a few years or less. Because the speed of the orbit of the IRC’s stars around each other should be much larger than the speed of the IRC as a whole around T Tauri N, it may be possible to use T Tauri N as an astrometric reference and thereby derive an estimate for the ratio of the masses of the IRC stars without fully solving for the orbit. A search for orbital motion will be the subject of an upcoming paper.
## 4 Conclusions
The nature of the T Tauri IRC remains a mystery despite extensive observational studies by many workers. That it is a binary whose separation is small compared to the fiducial size of a pre-main sequence disk offers a simple explanation for the relatively small dust mass required by recent submillimeter observations. But the fundamental question of its evolutionary status remains unanswered: Is the IRC a “normal” young star coeval with the optically-visible T Tauri N and simply observed under special circumstances, or a more exotic object with a different evolutionary status?
It is a pleasure to thank M. Kuchner for his assistance with the observations, and R. Akeson and G. Blake for useful discussions and comments on an early version of the text. Data presented herein were obtained at the W.M. Keck Observatory, which is operated as a scientific partnership among the California Institute of Technology, the University of California and the National Aeronautics and Space Administration. The Observatory was made possible by the generous financial support of the W.M. Keck Foundation. This research was supported by the National Aeronautics and Space Administration. |
no-problem/0001/astro-ph0001331.html | ar5iv | text | # The Diffusive Galactic GeV/TeV Gamma-Ray Background: Sources vs. Transport
## Introduction
The observations of the diffuse Galactic $`\gamma `$-ray emission can be described rather well by a suitable model for the diffuse interstellar gas, GCR, and photon distributions (e.g. Hunter et al. 1997a). However, above 1 GeV the observed average diffuse $`\gamma `$-ray intensity, foremost in the inner Galaxy, $`300^{}<l<60^{}`$, $`|b|10^{}`$, exceeds the model prediction significantly. As far as the energetic particles are concerned, there are at least two possible explanations for this discrepancy (e.g. Weekes et al. 1997; Hunter et al. 1997b, and references therein). The high-energy $`\gamma `$-ray excess may indicate that the GCR spectrum observed in the local neighborhood is not representative of the diffuse CR population in the Galactic disk. An unresolved distribution of CR sources is the other possibility. Since the $`\gamma `$-ray emission is the product of the energetic particle intensity on the one hand, and of the gas density or the photon density, on the other, it is of course possible that deviations from the above model assumptions for these latter densities across the Galaxy can also lead to changes in the observed energy spectrum of the diffuse gamma rays. We shall not discuss such deviations here. We shall rather evaluate the contribution of the sources, assumed to be the ensemble of Supernova Remnant (SNR) shells, following a recent calculation by Berezhko & Völk (1999). We shall also consider the transport of the particles from the same sources out of the Galaxy to naturally increase with decreasing Galactic radius (Breitschwerdt et al., 1991). We shall leave aside the possibility of new sources of particles, not known in the neighborhood of the Solar system.
## Gamma rays from the ensemble of SNRs
Since at best a handful of shell SNRs could be argued to have been detected up to now in gamma rays, we shall ignore their discrete contributions and consider the CR sources to be spatially averaged over the volume $`V_g=2.5\times 10^{66}`$ cm<sup>3</sup> of the Galactic gas disk, with a radius of 10 kpc and a thickness of 240 pc. The corresponding gas mass is $`M_g=4\times 10^9M_{}`$ (Dickey & Lockman, 1990). The source input rate in the form of energetic particle energy equals $`\nu _{SN}\delta E_{SN}`$, where we take $`\nu _{SN}=1/30yr`$, $`E_{SN}=10^{51}`$ erg. The efficiency per SNR is $`\delta <1`$. The total number of localized SNRs which still contain their shock accelerated CRs, called here the source CRs (SCRs), is given by $`N_{SN}=\nu _{SN}T_{SN}`$, where $`T_{SN}`$ is their assumed life time, i.e. the time until which they can confine the accelerated particles in their interior. Thus $`N_{SN}`$ is dominated by the population of old SNRs. We estimate $`T_{SN}10^5`$ yr. After the time $`T_{SN}`$ the SCRs rather quickly become part of the ordinary GCRs that presumably occupy a large Galactic residence volume uniformly.
### Acceleration model
We assume the overall SCR number inside a single SNR to be given by a power law spectrum $`N_{SCR}dEϵ^{\gamma _{SCR}}dE`$ in energy $`ϵ`$ in the relativistic range.
Averaged over the disk volume, the spatial density $`n_{SCR}(ϵ)`$ of SCRs is given $`n_{SCR}(ϵ)=N_{SCR}(ϵ)N_{SN}/V_g`$, with energy density $`e_{SCR}=N_{SN}\delta E_{SN}/V_g`$. In terms of $`e_{SCR}`$, we have
$$n_{SCR}(ϵ)=\frac{n_0^{SCR}(\gamma _{SCR}1)}{mc^2}\left(\frac{ϵ}{mc^2}\right)^{\gamma _{SCR}}$$
(1)
and
$$n_0^{SCR}=\frac{(\gamma _{SCR}2)e_{SCR}}{(\gamma _{SCR}1)mc^2},$$
(2)
for $`\gamma _{SCR}>2`$. The same expressions hold for the GCRs, given $`e_{GCR}`$ and $`\gamma _{GCR}`$ .
For the SCR we may quite possibly have $`\gamma _{SCR}=2`$, and then
$$n_0^{SCR}=\frac{e_{SCR}}{mc^2\mathrm{ln}(ϵ_{max}/mc^2)},$$
(3)
where $`ϵ_{max}10^5mc^2`$ is the maximum SCR energy.
The $`\pi ^0`$-decay prodution rate is given by
$$Q_\gamma (ϵ)=Z_\gamma \sigma _{pp}cN_gn(ϵ),$$
(4)
(Drury et al. 1994), which leads to the ratio $`R=Q_\gamma ^{SCR}/Q_\gamma ^{GCR}`$ of the $`\gamma `$-ray production rates due to SCRs and GCRs, given by
$`R(ϵ_\gamma )`$ $`=`$ $`{\displaystyle \frac{Z_\gamma ^{SCR}N_{SN}\delta E_{SN}}{Z_\gamma ^{GCR}(\gamma _{GCR}2)\mathrm{ln}(ϵ_{max}/mc^2)V_ge_{GCR}}}`$ (5)
$`\times `$ $`\zeta \left({\displaystyle \frac{ϵ_\gamma }{mc^2}}\right)^{\gamma _{GCR}2},`$
where $`\zeta `$ is the ratio $`N_g^{SCR}/N_g^{GCR}`$, $`N_g^{SCR}`$ is the mean source gas density, and $`N_g^{GCR}`$ denotes the average gas density in the disk.
With $`\delta =0.2`$, $`e_{GCR}2\times 10^{12}`$ erg/cm<sup>3</sup> for the relativistic part of the GCRs, and $`\gamma _{GCR}=2.75`$ which results in $`Z_\gamma ^{SCR}/Z_\gamma ^{GCR}=10`$ (Drury et al. 1994), we obtain
$$R(ϵ_\gamma )=0.16\zeta \left(\frac{T_{SN}}{10^5\text{yr}}\right)\left(\frac{ϵ_\gamma }{1\text{GeV}}\right)^{0.75},$$
(6)
for $`\gamma _{SCR}=2`$.
The total $`\gamma `$-ray spectrum measured from an arbitrary Galactic disk volume is then expected to be
$$\frac{dN^\gamma }{dϵ_\gamma }=\frac{dN_{GCR}^\gamma }{dϵ_\gamma }[1.4+R(ϵ_\gamma )],$$
(7)
where the additional factor 0.4 is introduced to approximately take into account the contribution of GCR electron component to the diffuse $`\gamma `$-ray emission at GeV energies, and where $`\frac{dN_{GCR}^\gamma }{dϵ_\gamma }`$ is taken from the paper by (e.g. Hunter et al. 1997b).
### ”Leaky Box”-type model
We can derive very similar results from a leaky box-type balance equation
$$\frac{n_{GCR}(ϵ)}{\tau _c}=\frac{N_{SCR}(ϵ)}{V_c(ϵ)}\nu _{SN},$$
(8)
where $`V_c(ϵ)`$ is the energy-dependent residence volume occupied by GCRs that reach the gas disk during their constant mean residence time $`\tau _c3\times 10^7`$ yrs in $`V_c(ϵ)`$. In the case of an extended Galactic Halo, $`V_c(ϵ1\mathrm{G}\mathrm{e}\mathrm{V})V_g`$ (Ptuskin et al. 1997). Using eq. (4) we can write
$$\frac{n_{SCR}}{n_{GCR}}=\frac{V_cT_{SN}}{V_g\tau _c}=\frac{T_{SN}}{\tau _g}.$$
(9)
The GCR residence time in the disk volume
$$\tau _g=\tau _cV_g/V_c=\frac{xV_g}{vM_g}$$
(10)
can be derived from the measured grammage $`x=14v/c(ϵ/4.4\text{GeV})^{0.60}`$ g/cm<sup>2</sup>, for $`ϵ>4.4`$ GeV, and $`x=14v/c`$ g/cm<sup>2</sup>, for $`ϵ<4.4`$ GeV (Engelman et al. 1990).
At relativistic energies $`ϵ>mc^2`$, the GCR spectrum and the overall SCR spectrum $`N_{SCR}ϵ^{\gamma _{SCR}^{}}`$ are connected by the relation
$$\gamma _{SCR}^{}=\gamma _{GCR}0.6=2.15.$$
(11)
Taking $`\gamma _{SCR}=2.15`$, which leads to $`Z_\gamma ^{SCR}/Z_\gamma ^{GCR}=7.5`$ (Drury et al. 1994), we obtain for $`ϵ_\gamma 4.4`$ GeV:
$$R(ϵ_\gamma )=0.06\zeta \left(\frac{T_{SN}}{10^5\text{yr}}\right)\left(\frac{ϵ_\gamma }{1\text{GeV}}\right)^{0.6}$$
(12)
(Berezhko & Völk, 1999).
The question is, of course, whether the SN confinement time $`T_{SN}`$ is time dependent. Probably this dependence is $`T_{SN}(ϵ)=t_0(ϵ/ϵ_{max})^5`$, where $`t_0`$ is the sweep-up time when the SNR enters the Sedov phase and the shock speed begins to decrease with time. For average ISM parameters $`t_010^3`$ yr.
## Results including the SCRs
In Fig. 1 we show the measurements by Hunter et al. (1997a) and our two estimates for the total $`\gamma `$-ray emission, from GCRs plus SCRs. They demonstrate that the SCR contribution for the acceleration model exceeds the leaky box values for all energies. The reason is that for our empirical model the acceleration efficiency for the relativistic part of the spectrum is only $`\delta 0.08`$. This is probably due to the fact that the mean injection efficiency at the SNR shock is lower than the values typically assumed for a parallel shock by a factor of a few. We take the lower value for the $`\gamma `$-ray emission in Fig.1 as the most reliable estimate for the expected diffuse $`\gamma `$-ray emission, including the SCRs. Nevertheless the SCR distribution, which is about 10 percent at GeV energies, becomes dominant beyond 100 GeV, and exceeds the GCR emission at 1 TeV by almost a factor of 10. It would be very interesting to detect the diffuse Galactic $`\gamma `$-ray emission at 1 TeV in order to test this prediction.
Until now we have only discussed the $`\gamma `$-ray emission from hadronic SCRs. In fact, there are many reasons to assume that electrons are equally well accelerated in SNRs, even if their injection into the shock acceleration process is much less well understood. The inverse Compton emission by SCR electrons can be comparable with the hadronic emission, even though it does not contribute at TeV energies. For a more detailed discussion we refer to the paper of Berezhko & Völk (1999).
## Transport effects
The models used to fit the $`\gamma `$-ray data from, say, EGRET assume a GCR energy spectrum that is uniform throughout the Galaxy. This tacitly assumes that the GCR transport properties leading to the escape from the Galaxy are everywhere the same. However that needs not be the case, and in fact is almost certainly not true. The dynamical processes leading to GCR escape depend on the strength of the regular magnetic field and on its fluctuation characteristics, as well as on the CR pressure, and the gravitational field. An example is the formation of Parker bubbles which remove the enclosed CRs through their boyant rise into the Halo and ultimately into the Intergalactic Medium. Another example which we wish to discuss here in some more detail, involves the Galactic Wind which is partly driven by the GCRs themselves (e.g. Breitschwerdt et al., 1991, 1993; Zirakashvili et al., 1996). In fact, the wind velocity perpendicular to the disk - in z-direction - is much larger in the central regions of the Galaxy than at larger radii, through the radial variation of the Galactic gravitational field alone (see Fig. 2). This implies that for a given particle energy the boundary seperating the dominantly diffusive transport perpendicular to the Galactic disk near the disk from the dominantly convective transport at greater halo heights moves down in direction to the Galactic midplane in the inner Galaxy. Since the GCR diffusion coefficient increases with energy, the position of this boundary will depend on energy. As shown by Ptuskin et al. (1997), the energy spectrum of the GCRs is typically $`E^{1.9}`$ in the convection region compared to the standard spectrum $`E^{2.7}`$ in the diffusive confinement region of volume $`V_c`$ discussed in subsecion 2.1. A line of site that intersects this boundary will therefore receive gamma rays from two regions of very different GCR energy spectra, emitting correspondingly harder spectra than does the diffusive confinement region alone. Qualitatively this implies a hardening of the truly diffuse $`\gamma `$-ray spectrum with Galactic longitude towards the inner Galaxy, for given latitude. However, the effect will disappear for high enough energies when the convective zone does no more extend into regions of significant gas density.
Thus, in contrast to the contribution of the sources, this transport effect looses importance at high energies.
It remains to work out this effect quantitatively. But its very existence illustrates the interest we should attach to the measurements of the diffuse Galactic $`\gamma `$-ray emission over an as wide as possible range of energies.
## Conclusions
The foregoing discussion shows that there are at least two mechanisms of basic physical interest that contribute to a deviation of the diffuse Galactic $`\gamma `$-ray emission spectrum from what would be expected from CR observations in the Solar vicinity. The contribution from the SCRs is an inevitable one and is essentially sufficient to explain the data at least for the inner Galaxy; it should be part of the $`\gamma `$-ray emission model to begin with. Clearly this does not rule out effects from potentially existing new populations of CRs, especially electrons, or the influence of an increased strength, for instance, of the Interstellar radiation field. This is particularly true for high Galactic latitudes.
Acknowledgements The work on the role of the CR sources, summarized in the first part of the paper, has been done jointly with E.G. Berezhhko. I am also indebted to V.S. Ptuskin for a discussion on the effects of the Galactic Wind on the diffuse $`\gamma `$-ray emission. |
no-problem/0001/nucl-th0001053.html | ar5iv | text | # 1 Introduction
## 1 Introduction
In hard particle production in proton-nucleus ($`pA`$) collisions, the effect of the nuclear medium manifests itself in a cross section enhancement at high transverse momentum relative to a simple superposition of the proton-proton ($`pp`$) production cross section. The purpose of this contribution is to study the Cronin effect, and to look for an explanation in terms of the dense nuclear medium that the projectile proton has to traverse in $`pA`$ collisions. The presence of the medium leads to an increase of the width of the transverse momentum distribution of partons. The transverse momentum distribution of partons is recently also utilized in the description of dilepton data at SPS.
## 2 Transverse momentum distribution in $`pp`$ collisions
The E609 collaboration extracted a width of $`k_T=(0.9\pm 0.2)`$ GeV/c for the transverse momentum distribution of partons in the proton from dijet events at 400 GeV/c. We use $`k_T`$ as an energy dependent parameter to obtain a satisfactory description of hard-particle production cross sections in the center-of-mass energy range 18 GeV $`\sqrt{s}`$ 65 GeV in $`pp`$ collisions. The transverse momentum distribution enters in the perturbative QCD (pQCD) calculation of the hard-particle production cross section via the factorization-theorem based convolution
$`E{\displaystyle \frac{d\sigma ^{pp}}{d^3p}}`$ $`=`$ $`{\displaystyle \underset{abcd}{}}{\displaystyle 𝑑x_1𝑑x_2d^2k_{T1}d^2k_{T2}f_{a/p}(x_1,Q^2)g_1(\stackrel{}{k}_{T1})f_{b/p}(x_2,Q^2)g_2(\stackrel{}{k}_{T2})}`$ (1)
$`{\displaystyle \frac{d\sigma }{d\widehat{t}}}(abcd){\displaystyle \frac{D_c(z_c,\widehat{Q}^2)}{\pi z_c}},`$
where the connection to the observed incoming and outgoing particles is established through the parton distribution functions (PDFs) $`f(x,Q^2)`$ and fragmentation functions (FFs) $`D(z,\widehat{Q}^2)`$. The value of $`z_c`$ is fixed by energy-momentum conservation. The transverse momentum distribution is denoted by $`g(\stackrel{}{k}_T)`$, and as an approximation, is taken to be a Gaussian: $`g(\stackrel{}{k}_T)=\mathrm{exp}(k_T^2/k_T^2)/\pi k_T^2`$. Here, $`k_T^2`$ is the 2-dimensional width of the $`k_T`$ distribution and it is related to the average transverse momentum of one parton as $`k_T^2=4k_T^2/\pi `$.
The cross section of the hard subprocesses in Eq. (1) is to be calculated in pQCD. In principle, the order in the strong coupling constant, to which this is done, is a matter of decision based on the convergence of the series. At next-to-leading order (NLO), the cross section can be viewed as
$$\frac{d\sigma ^{NLO}}{d\widehat{t}}=\frac{d\sigma _{Born}^{NLO}}{d\widehat{t}}+\frac{d\sigma _{corr}^{NLO}}{d\widehat{t}}=K\frac{d\sigma _{Born}^{NLO}}{d\widehat{t}},$$
(2)
where the factor $`K`$ may depend on both, the center-of-mass energy and the transverse momentum, and represents an approximate way to take into account corrections beyond the Born term. For sufficiently high energies and transverse momenta $`K=const.`$ is a good approximation. We adapted $`K=2`$ in the present calculation. For the PDFs we used the MRST98 set, which is a NLO parameterization and incorporates an intrinsic $`k_T`$. The scales are fixed, $`\mathrm{\Lambda }_{\overline{MS}}(n_f=4)=300`$ MeV and $`Q=\widehat{Q}=p_T/2`$. An NLO parameterization was used for the FFs, too, thus assuring the consistency of Eq. (1). The energy dependence of $`k_T^2`$ and the quality of the agreement with hard $`\pi ^0`$ and $`\gamma `$ production data in $`pp`$ collisions is displayed in Ref. . Note that a satisfactory description of the pion spectra does not appear possible in NLO without transverse-momentum smearing.
## 3 Enhanced transverse momentum in $`pA`$ collisions
The Cronin effect in $`pA`$ collisions can be understood phenomenologically in terms of an increased $`k_T^2`$ according to
$$k_T^2_{pA}=k_T^2_{pp}+Ch_{pA}(b).$$
(3)
Here, $`h_{pA}(b)`$ describes the number of nucleon-nucleon collisions at impact parameter $`b`$, which contribute to the transverse momentum broadening of the partons in the projectile. The particle-producing hard collision is clearly not counted, but it is also not obvious that all other possible collisions contribute. We therefore use the term effective to distinguish the transverse-momentum broadening collisions. Each effective collision imparts a transverse momentum squared $`C`$ on average.
If it is assumed that the projectile proton suffers soft collisions, which do not change the projectiles identity as a proton, but enhance the width of the transverse momentum distribution of its partons via random interactions, each imparting some transverse momentum, then all collisions of the projectile at impact parameter $`b`$ are effective. This can be expressed by using $`h_{pA}^{all}(b)=\nu _A(b)1`$, where $`\nu _A(b)=\sigma _{NN}t_A(b)`$ is the collision number at impact parameter $`b`$, with $`\sigma _{NN}`$ being the total nucleon-nucleon cross section. However, using this prescription, the coefficient $`C`$ turns out to be target dependent.
Furthermore, it is necessary to use values as high as $`C1`$ GeV<sup>2</sup> with this prescription, no longer characteristic of soft physics. Thus it is natural to assume that the projectile proton suffers a more drastic change while traveling through the dense matter of the target nucleus. In particular, we assume that the projectile proton will break up in a hard or semi-hard collision. As a consequence of this violent event, partonic degrees of freedom become appropriate. The enhanced width of the transverse momentum distribution reflects this change of the physically relevant degrees of freedom, and gets “measured” in the collision that breaks the proton apart.
We studied scenarios where only a fraction of the possible $`\nu _A(b)1`$ collisions is effective in the sense of nucleon-nucleon collisions. In the limiting case, referred to as the saturating prescription, we allow at most one effective collision using a smoothed step function $`h_{pA}^{sat}(b)`$, defined as
$$h_{pA}^{sat}(b)=\{\begin{array}{cc}0\hfill & \text{ if }\nu _A(b)<1\text{ }\hfill \\ \nu _A(b)1\hfill & \text{ if }1\nu _A(b)<2\text{ }.\hfill \\ 1\hfill & \text{ if }2\nu _A(b)\text{ }\hfill \end{array}$$
(4)
The saturated Cronin factor is denoted by $`C^{sat}`$. By the saturating prescription we limit our consideration to at most one effective transverse momentum broadening collision in the above sense.
The left panel of Fig 1. shows the data/theory ratio on a linear scale for the $`pW\pi ^0X`$ reaction with (full line) and without (dashed line) the saturated Cronin enhancement, using the value $`C^{sat}=1.2`$ GeV<sup>2</sup>, with the data from Ref. . In the left panel we illustrate that the same value of $`C_{sat}`$ provides a satisfactory description of recent photon production data in p + Be collisions in the same energy region. Since the effect of the nuclear medium on the projectile proton should not depend on the outgoing particle, we consider the common value of $`C_{sat}`$ a strength of our model.
Our $`C^{sat}`$ is determined at $`\sqrt{s}30`$ GeV. Its value may of course depend on the center-of-mass energy. This dependence needs to be mapped out in a wide energy range up to $`\sqrt{s}=200`$ GeV in order to make the model useful for RHIC predictions. On the other hand, we expect $`C^{sat}`$ to be independent of the target and of the produced hard particle. Systematic experimental studies to test these predictions would be welcome.
## 4 Summary
We reported on a phenomenological model to understand the Cronin effect. The enhancement of hard particle production cross sections in $`pA`$ collisions relative to scaled $`pp`$ results was connected to an increase in the width of the transverse momentum distribution of the projectile as it traverses the dense nuclear medium of the target. Physically rather different pictures for the enhancement of the transverse momentum width were studied. The extreme saturating prescription gives reasonable agreement with available data at the present level of the calculation.
## 5 Acknowledgement
This work was supported in part by DOE grant DE-FG02-86ER40251 and Hungarian grant OTKA-T032796. Partial support from the Domus Hungarica program of the Hungarian Academy of Sciences is gratefully acknowledged. |
no-problem/0001/astro-ph0001138.html | ar5iv | text | # A Preliminary Visual Orbit of BY Draconis
## 1 Introduction
BY Draconis (HDE 234677, Gl 719) is a well-studied nearby ($``$ 15 pc), multiple stellar system containing at least three objects. The A and B components form a short-period (6 d) late-type (K6 Ve – K7 Vvar) binary system, whose spectroscopic orbit is well known (Bopp & Evans (1973), hereafter BE73; Vogt & Fekel (1979), hereafter VF79; and Lucke & Mayor (1980), hereafter LM80). BY Dra is the prototype of a class of late-type flare stars characterized by photometric variability due to star spots, rapid rotation, and Ca II H and K emission lines. Like the BY Dra system itself, a large fraction ($`>`$ 85%, Bopp & Fekel (1977); Bopp et al. (1980)) of BY Dra stars are known to be in short-period binary orbits. The rapid rotation of BY Dra A (period 3.83 d) that gives rise to the spotting and photometric variability is consistent with pseudosynchronus rotation with the A – B orbital motion (Hut (1981); Hall (1986)), but pseudosynchronization is disputed by Glebocki & Stawikowski (1995, 1997) who assert asynchronous rotation and roughly 30 misalignment of the orbital/rotational angular momentum vectors.
Despite the fact that the BY Dra A and B components are nearly equal mass, the system exhibits a significant brightness asymmetry in spectroscopic studies (VF79, LM80). VF79 attributes this to the hypothesis that A and B components of BY Dra are pre-main sequence objects (VF79, Bopp et al. (1980)), and are still in the contraction phase. VF79 argues for physical sizes of the A component in the range of 0.9 – 1.4 R, based primarily on rotation period and $`v\mathrm{sin}i`$ considerations. LM80 concur with the A component physical size argument from their $`v\mathrm{sin}i`$ measurements, estimating a 1.2 R size for a $`\mathrm{sin}i`$ 0.5 (presuming rotation/orbit spin alignment with pseudosynchronization, and our orbital inclination from Table 2). However, they continue by pointing out that if the A component macroturbulance were significantly larger than solar, then the $`v\mathrm{sin}i`$ measurements and the component diameters they are based on are biased high. If the pre-main sequence interpretation is correct, the BY Dra components are additionally interesting as an examples of the transition region between pre and zero-age main sequence states.
BY Dra was detected as a hierachical triple system through common proper motion measurements of a BY Dra C component by Zuckerman et al. (1997); they find the C component separated by 17” from the A – B pair. Zuckerman’s photometry on the C component is consistent with an M5 main-sequence interpretation ($`V`$ \- $`K`$ $``$ 6.2); assuming all three stars are coeval this clearly poses problems for the VF79 pre-main sequence hypothesis for the A and B components. At a projected physical separation of approximately 260 AU from the A – B binary, the putative low-mass C component would have negligible dynamical influence on the A – B binary motion. Further, the Hipparcos catalog (ESA (1997)) implies the presence of at least one additional component as it lists BY Dra as having a circular photocentric orbital solution with 114 d period. This 114 d period is previously unreported in spectroscopic studies, and if correct it is difficult to understand why this motion was not previously detected.
Herein we report on a preliminary determination of the BY Dra A – B system visual orbit from near-infrared, long-baseline interferometric measurements taken with the Palomar Testbed Interferometer (PTI). PTI is a 110-m $`H`$ (1.6$`\mu `$m) and $`K`$-band (2.2 $`\mu `$m) interferometer located at Palomar Observatory, and described in detail elsewhere (Colavita et al. (1999)). PTI has a minimum fringe spacing of roughly 4 milliarcseconds (10<sup>-3</sup> arcseconds, mas) in $`K`$-band at the sky position of BY Dra, allowing resolution of the A – B binary system.
## 2 Observations
The interferometric observable used for these measurements is the fringe contrast or visibility (squared, $`V^2`$) of an observed brightness distribution on the sky. The reader is refered to one of the previous papers in this series (Boden et al. 1999a ; Boden et al. 1999b ; Boden et al. (2000)) for descriptions of the $`V^2`$ observables for binary stars.
BY Dra was observed in conjunction with objects in our calibrator list (Table 1) by PTI in $`K`$-band ($`\lambda 2.2\mu `$m) on 21 nights between 23 June 1999 and 21 October 1999, covering roughly 20 periods of the system. Additionally, BY Dra was observed by PTI in $`H`$-band ($`\lambda 1.6\mu `$m) on 13 September, 2 October, and 20 October 1999. BY Dra, along with calibration objects, was observed multiple times during each of these nights, and each observation, or scan, was approximately 130 sec long. For each scan we computed a mean $`V^2`$ value from the scan data, and the error in the $`V^2`$ estimate from the rms internal scatter. BY Dra was always observed in combination with one or more calibration sources within $``$ 10 on the sky. Table 1 lists the relevant physical parameters for the calibration objects. We have calibrated the $`V^2`$ data by methods discussed in Boden et al. (1998). Our 1999 observations of BY Dra result in 67 calibrated visibility measurements (50 in $`K`$-band, 17 in $`H`$-band). One notable aspect of our BY Dra observations is that its high declination (51) relative to our Palomar site (33 latitude) puts it at the extreme Northern edge of the delay line range on our N-S baseline (Colavita et al. (1999)), implying extremely limited $`uv`$ coverage on BY Dra. To our PTI visibilities we have added 44 double-lined radial velocity measurments: 14 from BE73, seven from VF79, and 23 CORAVEL measurements from LM80.
## 3 Orbit Determination
As in previous papers in this series (Boden et al. 1999a ; Boden et al. 1999b ; Boden et al. (2000)) the estimation of the BY Dra visual orbit is made by fitting a Keplerian orbit model directly to the calibrated (narrow-band and synthetic wide-band) $`V^2`$ and RV data on BY Dra; because of the limited $`uv`$ coverage in our data derivation of intermediate separation vector models is impossible. The fit is non-linear in the Keplerian orbital elements, and is therefore performed by non-linear least-squares methods with a parallel exhaustive search strategy to determine the global minimum in the chi-squared manifold. The reader is refered to the previous papers for futher details on our orbit estimation procedures.
Figure 1 depicts the relative visual orbit of the BY Dra system, with the primary component rendered at the origin, and the secondary component rendered at periastron. We have indicated the phase coverage of our $`V^2`$ data on the relative orbit with heavy lines; our data samples most phases of the orbit well, leading to a reliable orbit determination. The apparent inclination is very near the estimate given by VF79 based on the primary rotation period, assumed size, and assumption of parallel orbital/rotational angular momentum alignment. The orbit is seen approximately 30 from a face-on perspective, which makes physical parameter determination difficult (§5).
Figure 2 illustrates comparisons between our PTI $`V^2`$ and archival RV data and our orbit model. In Figure 2a six consecutive nights of $`K`$-band visibility data and visibility predictions from the best-fit model are shown, inset with fit residuals along the bottom. That there are only 1 – 3 data points in each of the nights follows from the brief time each night that BY Dra is simultaneously within both delay and zenith angle limits. Figure 2b gives the phased archival RV data and model predictions, inset with a histogram of RV fit residuals. The fit quality is consistent with previous PTI orbit analyses.
Spectroscopic orbit parameters (from VF79 and LM80) and our visual and spectroscopic orbit parameters of the BY Dra system are summarized in Table 2. We give the results of separate fits to only our $`V^2`$ data (our “$`V^2`$-only Fit” solution), and a simultaneous fit to our $`V^2`$ data and the archival double-lined radial velocities – both with component diameters constrained as noted above. All uncertainties in parameters are quoted at the one sigma level. We see good statistical agreement between all the derived orbital parameters with the exception of the LM80 period estimate.
## 4 Comparisons With Hipparcos Model
The Hipparcos catalog lists a circular photocentric orbital solution for BY Dra with a 114 day period (ESA (1997)), presumably in addition to the well-established 6 d period A – B motion. As noted above it is difficult to reconcile this hypothesis with the quality of the existing short-period spectroscopic orbit solutions from VF79 and LM80. However if the A – B system indeed did have a companion with this period, unlike BY Dra C it would lie within PTI’s 1” primary beam, and if sufficiently luminous it would bias the visibility measurements used in our BY Dra A – B visual orbit model. We see no indications of this in our orbital solutions; the quality of our BY Dra visual orbit solution is consistent with our results on other systems. But it remains possible we have misinterpreted our $`V^2`$ data in the binary star model fit, and we are motivated to consider the 114-d periodicity hypothesis in the archival RV data.
We note that the Hipparcos model is a photocentric orbit, and therefore calls for the A – B system to exhibit a reflex motion with radius $``$ 0.05 AU at the putative distance of BY Dra. The 114-d orbit hypothesis is at high inclination (113), and therefore would produce a radial velocity semi-amplitude for the A – B system barycenter $``$ 3.95 km s<sup>-1</sup>. This value is large compared to fit residuals observed by VF79 and LM80 in their spectroscopic orbital solutions (and by ourselves in the joint fit with our visibilities; Table 2), suggesting the 114-d motion hypothesis is unlikely.
To quantify this issue we have considered Lomb-Scargle periodogram analyses of the LM80 BY Dra radial velocity data. We have chosen to use the LM80 data because it is the more precise sample, yielding an rms residual of roughly 0.5 km s<sup>-1</sup> in our A – B orbit analysis. Figure 3 gives periodograms of the LM80 primary and secondary radial velocity data. First, Figure 3a gives a periodogram of the primary and secondary RV data, with probability of false alarm levels (P<sub>fa</sub>) as noted. As expected, both component lines exhibit a significant periodicity at the observed A – B orbit frequency of approximately (6 d)<sup>-1</sup> (indicated by vertical line). In the same plot we sample the frequency of the 114-d orbit hypothesis, and no comparable periodicity is evident on the scale of the A – B motion.
Presuming the 114-d motion hypothesis might be superimposed on the 6-d A – B orbit, in Figure 3b we give a periodogram of the LM80 primary and secondary RV fit residuals to the 6-d hypothesis fit. In Figure 3b we have adjusted to range of the periodogram to finely sample the frequency range around the (114 d)<sup>-1</sup> hypothesis (indicated by vertical line). No significant periodicity is noted at this or any other frequency. The LM80 RV dataset spans roughly 400 days, so a 4 km s<sup>-1</sup> amplitude periodicity in this dataset should have been evident in our analysis. Given these considerations, it seems unlikely that our $`V^2`$ measurements and A – B orbit model are affected by the presence of a third luminous body.
## 5 Discussion
The combination of the double-lined spectroscopic orbit and relative visual orbit allow us to estimate the BY Dra A – B component masses and system distance. However, because of the nearly face-on geometry of the A – B orbit, the accuracy of our inclination estimate, and consequently the component mass estimates and system distance estimate, is relatively poor. The low-inclination geometry is particularly difficult for astrometric studies because the astrometric observable ($`V^2`$ in this case) becomes highly insensitive to small changes in the inclination Euler angle. Table 3 lists the physical parameters we estimate from our Full-Fit orbit solution. The mass estimates are in a reasonable range for stars of this type but are formally crude – roughly 24% 1-$`\sigma `$ uncertainties, as is the system distance estimate – roughly 8% 1-$`\sigma `$. In both cases the dominant error is the inclination; to improve these mass and distance estimates by a factor of two we must improve the inclination estimate by a factor of two. The orbital parallax with its large error is in 1-$`\sigma `$ agreement with the Hipparcos triginometric determination of 60.9 $`\pm `$ 0.75 mas, but we note that the Hipparcos solution was derived jointly with the 3 mas, 114-d orbital hypothesis that we believe to be suspect (§4).
Table 3 also gives component absolute magnitudes and $`V`$ \- $`K`$ color indices derived from archival broad-band photometry, our $`H`$ and $`K`$ component relative magnitudes, the LM80 $`V`$ relative magnitude, and our system distance estimate. The component color indices are unaffected by errors in the system distance, and seem to be in good agreement with the color indices expected from stars in this mass range and the classical spectral typing of the BY Dra system. The absolute magnitudes are somewhat uncertain from the relatively poor distance estimate, but suggest a modest discrepancy with the mass-luminosity models of Baraffe et al (1998, BCAH98). As depicted in Figure 4, both components appear roughly 0.5 – 1 mag brighter at $`V`$ and $`K`$ than the BCAH98 models evaluated at our model masses would predict. The cause for this discrepancy could be abnormally large component sizes suggested by VF79, or the component masses are underestimated in our analysis. Clearly we need to improve our orbit model so as to further constrain the BY Dra component parameters.
To assess the VF79 component size/pre-main sequence hypothesis, the most interesting measurement of the BY Dra A – B system would be unequivocal measurements of the component diameters. Unfortunately our $`V^2`$ data are as yet insufficient to determine these diameters independently. Canonical sizes of 1.0 R and 0.8 R (indicated by model effective temperatures and our IR flux ratios) yield angular diameters of approximately 0.6 and 0.5 mas for the A and B components respectively. At these sizes neither the $`H`$ nor $`K`$-band fringe spacings of PTI sufficiently resolve the components to independently determine component sizes. Consequently we have constrained our orbital solutions to these 0.6 and 0.5 mas model values. Our data does in fact prefer the slightly smaller primary component diameter to the larger 1.2 R size implied by VF79 and LM80 $`v\mathrm{sin}i`$ and rotational period measurements. However either primary model diameter is possible with the expected systematic $`V^2`$ calibration errors. Additional data we will collect in the coming year (see below) may well place interesting upper limits on the component sizes, but unambiguous resolution of the BY Dra A and B components will have to wait for a longer baseline infrared interferometer; most likely the CHARA array currently under construction on Mt. Wilson <sup>1</sup><sup>1</sup>1http://www.chara.gsu.edu/CHARAArray/chara\_array.html
The prospects for improving the BY Dra orbital model in general, and the orbital inclination estimate in particular are reasonable. To acheive 10% mass determinations on the BY Dra components we will need to reduce the inclination uncertainty to approximately 1.5, which implies a rough quadrupling of the PTI visibility data on BY Dra, or an improvement of our $`V^2`$ measurement precision by a factor of two on a data set of similar size to that already collected. Both are plausible; PTI is expected to be in normal operation for at least another season, and hardware upgrades to the PTI fringe camera are planned and should provide significant improvements in sensitivity.
The work described in this paper was performed at the Infrared Processing and Analysis Center, California Institute of Technology, and the Jet Propulsion Laboratory under contract with the National Aeronautics and Space Administration. Interferometer data were obtained at Palomar Observatory using the NASA Palomar Testbed Interferometer, supported by NASA contracts to the Jet Propulsion Laboratory. Science operations with PTI are conducted through the efforts of the PTI Collaboration (http://huey.jpl.nasa.gov/palomar/ptimembers.html), and we acknowledge the invaluable contributions of our PTI colleagues. We further acknowledge I.N. Reid (UPenn) for the compilation of mass-luminosity data on Henry & McCarthy and Andersen objects. This research has made use of the Simbad database, operated at CDS, Strasbourg, France. |
no-problem/0001/cond-mat0001120.html | ar5iv | text | # Fractional calculus and continuous-time finance
## 1 Introduction
The importance of random walks in finance has been known since the seminal work of Bachelier which was completed at the end of the XIXth century, nearly a hundred years ago. The ideas of Bachelier were further carried out by Mandelbrot , who introduced the concept of Lévy flights and stable distributions in finance, and by the MIT school of Samuelson .
Although it was well-known that the distribution of returns or of logarithmic returns approximately followed a stable law, there was a barrier to the application of these concepts in the financial practice. Indeed, stable distributions have non-finite variance, and this leads to many mathematical difficulties (for a discussion on this point the reader is referred to chapter 3 of Merton’s book ). Therefore, in mainstream finance, both theoreticians and practitioners prefer to use the more tractable continuous Wiener process instead of discontinuous Lévy flights. A way of overcoming these difficulties has been provided by empirical studies suggesting the use of truncated Lévy flights, characterized by probability density distributions with finite moments .
In financial markets, not only prices and returns can be considered as random variables, but also the waiting time between two transactions varies randomly. So far, a large part of the financial practice is based on daily price changes. However, a company specialized in intra–day transactions and high–frequency data analysis, Olsen & Associates, has published various working papers related to the time behaviour of tick-by-tick data (see, for instance, on fractional time ref. and on mean first passage time ref. ).
The purpose of this paper is to present a rather general phenomenological theory of tick-by-tick dynamics in financial markets. Many well-known aspects, such as the Lévy scaling form of ref. , follow as particular cases of the theory. The theory fully takes into account the non-Markovian and non-local character of financial time series. Predictions on the long-time behaviour of the waiting-time probability density are presented. Finally, a more general scaling form is given, based on the solution of the fractional diffusion equation.
The paper is divided as follows. In Sec. 2, we discuss the relevance of continuous-time random walks in finance by explicitly performing a mapping from financial data to random walks. In Sec. 3, we present the master equation and we show that it reduces to the fractional diffusion equation in the hydrodynamic limit (corresponding to a long jump-observation scale and long observation times) if some simple scaling assumptions on the jump and waiting-time probability densities hold true. Sec. 4 is devoted to the solutions of the fractional diffusion equation and their natural scaling properties. Finally, in Sec. 5, we point out the main conclusions and outline the direction for future work.
As a final remark, let us stress that the theory of continuous-time random walks is well developed , and its relation to the fractional diffusion equation and fractional calculus has been recently discussed by various authors . However, as far as we know, these concepts have not yet been applied to finance in the form we present here.
## 2 Continuous-time random walk in finance
The price dynamics in financial markets can be mapped onto a random walk whose properties are studied in continuous, rather than discrete, time . Here, we shall perform this mapping, pioneered by Bachelier and fully exploited by Samuelson and his school , in a rather general way.
As a matter of fact, there are various ways in which to embed a random walk in continuous time. Here, we shall base our approach on the so-called continuous-time random walk (henceforth abbreviated as CTRW) in which time intervals between successive steps are random variables, as discussed by Montroll and Weiss
Let $`S(t)`$ denote the price of an asset or the value of an index at time $`t`$. In a real market, prices are fixed when demand and offer meet and a transaction occurs. In this case, we say that a trade takes place. In finance, returns rather than prices are considered. For this reason, in the following we shall take into account the variable $`x(t)=\text{log}S(t)`$, that is the logarithm of the price. Indeed, for a small price variation $`\mathrm{\Delta }S=S(t_{i+1})S(t_i)`$, the return $`r=\mathrm{\Delta }S/S(t_i)`$ and the logarithmic return $`r_{log}=log[S(t_{i+1})/S(t_i)]`$ virtually coincide.
As we mentioned before, in financial markets, not only prices can be modelled as random variables, but also waiting times between two consecutive transactions vary in a stochastic fashion. Therefore, the time series $`\{x(t_i)\}`$ is characterised by $`\phi (\xi ,\tau )`$, the joint probability density of jumps $`\xi _i=x(t_{i+1})x(t_i)`$ and of waiting times $`\tau _i=t_{i+1}t_i`$. The joint density satisfies the normalization condition $`𝑑\xi 𝑑\tau \phi (\xi ,\tau )=1`$.
Montroll and Weiss have shown that the Fourier-Laplace transform of $`p(x,t)`$, the probability density function, pdf, of finding the value $`x`$ of the price logarithm (which is the diffusing quantity in our case) at time $`t`$, is given by:
$$\stackrel{~}{\widehat{p}}(\kappa ,s)=\frac{1\stackrel{~}{\psi }(s)}{s}\frac{1}{1\stackrel{~}{\widehat{\phi }}(\kappa ,s)},$$
(1)
where
$$\stackrel{~}{\widehat{p}}(\kappa ,s)=_0^+\mathrm{}𝑑t_{\mathrm{}}^+\mathrm{}𝑑x\text{e}^{st+i\kappa x}p(x,t),$$
(2)
and $`\psi (\tau )=𝑑\xi \phi (\xi ,\tau )`$ is the waiting time pdf.
Let us now consider the situation in which the waiting time and the size of the step are independent. In this case the joint density function, $`\phi `$, can be factorized, namely written as the product of a “spatial” part and a temporal part: $`\phi (\xi ,\tau )=\lambda (\xi )\psi (\tau )`$. Here $`\lambda (\xi )`$ is the probability for a displacement $`\xi `$ in each single step (transition probability density). Now, the normalization condition for the transition pdf: $`𝑑\xi \lambda (\xi )=1`$ must be added to that for the probability density of the waiting time $`𝑑\tau \psi (\tau )=1`$.
As a consequence we get:
$$\stackrel{~}{\widehat{p}}(\kappa ,s)=\frac{1\stackrel{~}{\psi }(s)}{s}\frac{1}{1\widehat{\lambda }(\kappa )\stackrel{~}{\psi }(s)}=\frac{\stackrel{~}{\mathrm{\Psi }}(s)}{1\widehat{\lambda }(\kappa )\stackrel{~}{\psi }(s)},$$
(3)
where $`\widehat{\lambda }(\kappa ),`$ the Fourier transform of the transition probability density, is usually called the structure function of the random walk and $`\stackrel{~}{\mathrm{\Psi }}(s)=(1\stackrel{~}{\psi }(s))/s`$ is the Laplace transform of
$$\mathrm{\Psi }(t)=_t^{\mathrm{}}\psi (t^{})𝑑t^{}=1_0^t\psi (t^{})𝑑t^{}.$$
(4)
$`\mathrm{\Psi }(t)`$ is the survival probability at the initial point position ($`t_0=0`$) . $`_0^t\psi (t^{})𝑑t^{}`$ represents the probability that at least one step is taken in the interval ($`0,t`$), hence $`\mathrm{\Psi }(t)`$ is the probability that the diffusing quantity does not change during the time interval of duration $`t`$ after a jump .
According to Weiss , $`\mathrm{\Psi }(t)`$ can be viewed as the probability that the duration of a given interval between successive steps is strictly greater than $`t`$ and is the peculiar function needed to specify the probability of the displacement at time $`t^{}+t`$ in a CTRW, where $`t^{}`$ is the instant of the last jump. The waiting-time pdf is related to $`\mathrm{\Psi }(t)`$ by the formula: $`\psi (t)=d\mathrm{\Psi }(t)/dt`$.
Let us finally remark that, in general, the CTRW is a non-Markovian model , as at any time one has to know the value of the diffusing quantity as well as the time at which the last step took place in order to predict the further course of the walk. The non-Markovian property arises because the time of the previous step does vary and could be even $`t=0`$, so that the complete history of the process must be taken into account at all times. The only Markovian version of the CTRW is the one in which the waiting time pdf, $`\psi (\tau )`$, is a negative exponential:
$$\psi (\tau )=\frac{1}{T}\text{exp}(\tau /T),$$
where $`T`$ is the average time between successive steps. Only for this form of the density, the probability that a step of the random walk will take place in $`(t,t+dt)`$ is $`dt/T,`$ as $`dt0,`$ independent of the time at which the immediately preceding step occurred. This is not true of any other form of $`\psi (\tau ).`$
## 3 Master equations and fractional diffusion
The master equation governing the probability density profile in a CTRW can be derived by inverting the Fourier-Laplace transform in eq. (3). Rewriting (3) as
$$\stackrel{~}{\widehat{p}}(\kappa ,s)=\stackrel{~}{\mathrm{\Psi }}(s)+\stackrel{~}{\psi }(s)\widehat{\lambda }(\kappa )\stackrel{~}{\widehat{p}}(\kappa ,s)$$
we obtain
$$p(x,t)=\delta _{x\mathrm{\hspace{0.17em}0}}\mathrm{\Psi }(t)+_0^t𝑑t^{}\psi (tt^{})_{\mathrm{}}^+\mathrm{}𝑑x^{}\lambda (xx^{})p(x^{},t^{}).$$
(5)
This form of the master equation is quoted, e.g., in Klafter et al. and Hilfer and Anton . However, equivalent forms can be found in the literature. The following form shows the non-local and non-Markovian character of the CTRW :
$$\frac{}{t}p(x,t)=_0^t𝑑t^{}\varphi (tt^{})\left[p(x,t^{})+_{\mathrm{}}^+\mathrm{}𝑑x^{}\lambda (xx^{})p(x^{},t^{})\right];$$
(6)
here, the kernel $`\varphi (t)`$ is defined through its Laplace transform
$$\stackrel{~}{\varphi }(s)=\frac{s\stackrel{~}{\psi }(s)}{1\stackrel{~}{\psi }(s)}.$$
The above equations allow to compute $`p(x,t)`$ from the knowledge of the jump pdf $`\lambda (\xi )`$ and of the waiting-time pdf $`\psi (\tau )`$. In principle, both these quantities are empirically accessible from high-frequency market data, even if, recently, within the physics community, emphasis has been given to the jump pdf .
The time-evolution equation for $`p(x,t)`$ has a remarkable limit, if some scaling conditions on the structure function and on the waiting time pdf are verified.
Let us assume the following scaling behaviour in the hydrodynamic limit (long-jump scale and long observation times):
$$\widehat{\lambda }(\kappa )1|\kappa |^\alpha ,\kappa 0,0<\alpha 2,$$
(7)
and
$$\stackrel{~}{\psi }(s)1s^\beta ,s0,0<\beta 1.$$
(8)
The above approximations are consistent with the following explicit expressions for the Fourier and Laplace transforms:
$$\widehat{\lambda }(\kappa )=\text{exp}(|\kappa |^\alpha ),0<\alpha 2,$$
(9)
and
$$\stackrel{~}{\psi }(s)=\frac{1}{1+s^\beta },0<\beta 1.$$
(10)
We note that eq. (9) represents the characteristic function for the symmetric Lévy stable pdf of index $`\alpha ;`$ for $`0<\alpha <2`$ the pdf decays like $`|x|^{(\alpha +1)}`$ as $`|x|\mathrm{},`$ for $`\alpha =2`$ the Gaussian pdf is recovered.
From eq. (10), we observe that
$$\stackrel{~}{\mathrm{\Psi }}(s)=\frac{1\stackrel{~}{\psi }(s)}{s}=\frac{s^{\beta 1}}{1+s^\beta },0<\beta 1,$$
(11)
so that the survival probability turns out to be
$$\mathrm{\Psi }(t)=E_\beta (t^\beta ),0<\beta 1,$$
(12)
where
$$E_\beta (t^\beta )=\underset{n=0}{\overset{\mathrm{}}{}}(1)^n\frac{t^{\beta n}}{\mathrm{\Gamma }(\beta n+1)}$$
is the Mittag-Leffler function of order $`\beta `$ . Thus the pdf for the waiting time is
$$\psi (t)=\frac{d}{dt}\mathrm{\Psi }(t)=\frac{d}{dt}E_\beta (t^\beta ),0<\beta 1,$$
(13)
which is in agreement with the expression obtained in in terms of the generalized Mittag-Leffler function in two parameters.
For $`0<\beta <1`$ the Mittag-Leffler function $`E_\beta (t^\beta )`$ is known to be, for $`t>0`$, a completely monotonic function of $`t`$, decreasing from 1 (at $`t=0`$) to 0 like $`t^\beta `$ as $`t\mathrm{}`$ . As a consequence the pdf for the waiting time is strictly positive and monotonically decreasing to zero like $`t^{(\beta +1)}.`$ For $`\beta =1`$ the Mittag-Leffler function reduces to $`\text{exp}(t)`$ and we recover from eqs. (1113) the Markovian CTRW.
If we insert eqs. (7) and (8) into eq. (1), we get the limiting relation:
$$s^\beta \stackrel{~}{\widehat{p}}(\kappa ,s)+|\kappa |^\alpha \stackrel{~}{\widehat{p}}(\kappa ,s)=s^{\beta 1}.$$
(14)
Inverting eq. (14), we obtain the time-evolution equation for $`p(x,t)`$ in the hydrodynamic limit. If $`0<\beta 1`$ and $`0<\alpha 2`$, we have, for $`xR`$:
$$\frac{^\beta p(x,t)}{t^\beta }=\frac{^\alpha p(x,t)}{|x|^\alpha }+\frac{t^\beta }{\mathrm{\Gamma }(1\beta )}\delta (x),(t>0).$$
(15)
In eq. (15), we have introduced the fractional derivatives $`^\beta /t^\beta `$ and $`^\alpha /|x|^\alpha `$ defined as the inverse Laplace and Fourier transforms of $`s^\beta `$ and $`|\kappa |^\alpha `$, respectively . Fractional derivatives are non-local operators belonging to the larger class of pseudo-differential operators , which allow power-law effects. In particular, the “time” operator in eq. (15) is the Riemann-Liouville fractional derivative of order $`\beta `$ defined as (if $`0<\beta <1`$):
$$\frac{d^\beta }{dt^\beta }f(t)=\frac{1}{\mathrm{\Gamma }(1\beta )}\frac{d}{dt}\left\{_0^t\frac{f(\tau )}{(t\tau )^\beta }𝑑\tau \right\},$$
whereas the “jump” operator is the Riesz fractional derivative of order $`\alpha `$ which, if $`0<\alpha <2`$ can be represented as :
$$\frac{d^\alpha }{d|x|^\alpha }f(x)=\mathrm{\Gamma }(1+\alpha )\frac{\mathrm{sin}(\alpha \pi /2)}{\pi }_0^{\mathrm{}}\frac{f(x+\xi )2f(x)+f(x\xi )}{\xi ^{1+\alpha }}𝑑\xi .$$
Finally, let us mention that eq. (14) was derived by Weiss and by Afanas’ev and co-workers . Moreover, the above derivation of eq. (15) was implicit in a paper by Fogedby and was explicitly presented by Compte and by Saichev and Zaslavsky .
## 4 Lévy flights and scaling of solutions
We start this section with the analysis of a particular case of eq. (15), the limit $`\beta 1`$, where we have (in the weak sense) :
$$\underset{\beta 1}{lim}\frac{t^\beta }{\mathrm{\Gamma }(1\beta )}=\delta (t),$$
and eq. (15) becomes equivalent to the following initial value problem:
$$\frac{p(x,t)}{t}=\frac{^\alpha p(x,t)}{|x|^\alpha },p(x,0)=\delta (x).$$
(16)
The Cauchy problem (16) can be solved by Fourier-transforming both sides of the equation with respect to $`x`$. After integrating and inverse Fourier-transforming, one gets:
$$p(x,t)=\frac{1}{t^{1/\alpha }}L_\alpha \left(\frac{x}{t^{1/\alpha }}\right),$$
(17)
where $`L_\alpha (u)`$ is the Lévy standardized probability density function:
$$L_\alpha (u)=\frac{1}{2\pi }_{\mathrm{}}^+\mathrm{}\text{e}^{iqu|q|^\alpha }𝑑q.$$
(18)
Taking the limit $`\beta 1`$ in eq. (15) corresponds to considering independent time increments. Continuous-time random walks whose pdf $`p(x,t)`$ is given by eq. (17) are called symmetric Lévy flights or better symmetric $`\alpha `$-stable Lévy processes . In 1963, analysing the scaling properties of financial time-series, Mandelbrot found that the empirical pdf $`p(x,t)`$ could be well fitted by the Lévy density function (17) with $`\alpha =1.7`$. As we mentioned in the introduction, the main difficulty in dealing with the Lévy distribution is that its moments diverge. For $`0<\alpha <2`$, the only bounded finite moments have index $`\gamma `$ satisfying $`1<\gamma <\alpha `$. For this reason, the results of Mandelbrot were well-known but not much used in mainstream quantitative finance . The recent empirical analysis of Mantegna and Stanley suggests that truncated Lévy flights should be used instead, as good models for financial price dynamics . Koponen introduced a class of truncated Lévy flights, which was successively generalized by Boyarchenko and Levendorskii . However, all these studies somehow neglected the waiting-time pdf.
In the general case, the Cauchy problem of eq. (15) can be solved by the same technique used above. There is, however, a mathematical subtlety. In order to give a meaning to the Cauchy problem, the Riemann-Liouville operator must be replaced by the Caputo fractional derivative of order $`\beta `$ :
$$\frac{d^\beta }{dt^\beta }f(t)=\frac{1}{\mathrm{\Gamma }(1\beta )}\frac{d}{dt}\left\{_0^t\frac{f(\tau )}{(t\tau )^\beta }𝑑\tau \right\}\frac{t^\beta }{\mathrm{\Gamma }(1\beta )}f(0).$$
Now, the solution is:
$$p(x,t)=\frac{1}{t^{\beta /\alpha }}W_{\alpha ,\beta }\left(\frac{x}{t^{\beta /\alpha }}\right),$$
(19)
and $`W_{\alpha ,\beta }(u)`$ is the following scaling function:
$$W_{\alpha ,\beta }(u)=\frac{1}{2\pi }_{\mathrm{}}^+\mathrm{}\text{e}^{iqu}E_\beta (|q|^\alpha )𝑑q,$$
(20)
where $`E_\beta `$ the Mittag-Leffler function of order $`\beta `$ and argument $`z=|q|^\alpha `$
Further empirical studies on high-frequency financial data may reveal the scaling form (20), if the waiting-time pdf satisfies the asymptotics (8).
## 5 Conclusions and outlook
In this paper, we argued that the continuous-time random walk (CTRW) is a good phenomenological model for high-frequency price dynamics in financial markets, as, in general, this dynamics is non-Markovian and/or non-local.
CTRW naturally leads to the so-called fractional diffusion equation in the hydrodynamic limit if some scaling properties of the waiting time pdf $`\psi (\tau )`$ and of the jump pdf $`\lambda (\xi )`$ hold true in that limit. This point needs a further discussion. Indeed, the scaling regime of eqs. (8) and (7) breaks down for very large jumps. For this reason, truncated Lévy flights have been introduced . Preliminary investigations on high frequency financial data show that a similar problem is present for the waiting time pdf. Nevertheless, we can view the fractional diffusion equation (15) as a model for approximating the true behaviour of returns in financial markets.
In the region where the dynamics is well approximated by eq. (15), we expect the following scaling for the waiting-time pdf (see the discussion in Sec. 3):
$$\psi (\tau )\tau ^\mu ,$$
(21)
where $`\mu =\beta +1`$ varies in the range $`1<\mu <2`$. Consequently, the more complex scaling form (20) should hold true.
Empirical analyses on market high-frequency data will be necessary in order to verify these predictions. In any case, we expect that the concepts of CTRW and of fractional calculus will be of help in practical applications such as option pricing, as they provide an intuitive background for dealing with non-Markovian and non-local random processes.
In this paper, the mathematical apparatus has been kept to a minimum, the interested reader will find full mathematical details in a forthcoming paper .
## Acknowledgements
This work was partially supported by the Italian INFM and INFN, and by the Research Commission of the Free University in Berlin. R. G. is grateful to the Italian ”Istituto Nazionale di Alta Matematica” for supporting his visit in Italy. E. S. wishes to thank CERN for its wonderful library open 24 hours a day: an ideal place to study in a quiet atmosphere after shifts at the colliders. Discussions on high-frequency financial data with Marco Raberto inspired this work. |
no-problem/0001/astro-ph0001474.html | ar5iv | text | # LiBeB Production and Associated Astrophysical Sites
## 1. Introduction
A turning point in the theory of the origin and evolution of light elements has been the observation of a linear relationship between both Be and B and Fe in metal poor halo stars (Rebolo, Molaro & Beckman 1988; Duncan et al 1992, 1997; Gilmore et al 1992, Boesgaard & King 1993; Ryan, Norris, Bessel & Deliyannis 1994, Primas et al 1999). But recently, the debate has taken a complex turn, due to a modification of the O-Fe correlation indicated by the data of Israelian et al (1998) and Boesgaard et al (1999). This revision is however not universally admitted, and the debate is still open (e.g. Fulbright & Kraft 1999; Gustafsson 1999; Reetz 2000). Since the last International Cosmic Ray Conference (Barring 2000) and the most recent LiBeB meeting (Ramaty et al 1999), the situation has not ceased to evolve.
Prior to the proposal of change of the O-Fe relation, the situation was the following: in order to explain the observed proportionality of the Be/H ratio to Fe/H (itself taken proportional to O/H in halo stars as implied at that time by the data and by the SNII calculated yields), it was necessary to invoke a primary production mechanism (i.e. a production rate of LiBeB independent of the interstellar metallicity), driven by the break up in flight of C and O colliding with H and He in the ISM (for reviews see Vangioni-Flam et al 1999b, 2000a,b).
Now the reality of this primary component is questioned on the basis of the ”new” O-Fe correlation. Concerning oxygen, however, the situation is not settled. The abundances derived from the forbidden (OI) line, which is certainly the most accurate source when it is not too weak, suggest a plateau but measurements of near IR OH band in dwarfs and subgiants lead to a rising trend with decreasing Fe/H. In contrast, Fulbright & Kraft (1999) have analysed in great details the (OI) spectral region in the two metal poorest stars of the Israelian et al sample and have found a lower O/Fe ratio. As stressed for instance by Pagel (1999), both methods, indeed, have their drawbacks and technical difficulties: the OH bands are subject to uncertainties in UV continuum absorption (Balachandran & Bell 1998) and effective temperature, while the forbidden lines are so weak at low metallicity that the determination of the continuum becomes problematic.
So the situation is wide open and one is inclined to propose a different metallicity index, less ambiguous than O. The closer element to O whose abundance is widely measured is Mg. It has been chosen, due to its various advantages, both observational and theoretical, as the reference element to follow galactic evolution by Thomas et al (1998), Fuhrmann (1998) and Shigeyama & Tsujimoto (1998). Indeed i) Mg seems easier to measure than O and ii ) Mg and O are coproduced in SNII explosions, which are the main sources of these elements (Woosley & Weaver 1995, Thielemann, Nomoto & Hashimoto, 1996). Moreover the Mg yield is independent of the mass cut, and does not significantly depend on metallicity (Umeda et al 2000). In the light of existing data, the Mg/Fe vs Fe/H correlation is rather flat up to Fe/H = -1, like that of other alpha elements (Mc Williams 1997, Pagel & Tautvaisiene, 1995). Taken at face value, (assuming Mg proportional to O, on nucleosynthetic grounds, neglecting the peculiar behaviour of the most massive stars which are marginal in the chemical evolution budget), these data indicate the need of a primary component, then we are back to the previous situation (Cassé et al 1995; Vangioni-Flam et al 1996, 1998; Ramaty et al 1996). Thus, on this sole basis, a purely secondary origin of Be in the halo (Fields & Olive 1999a) driven by the standard GCR seems inadequate. However, the situation is not completely settled.
The outline of this paper is the following: in section 2 we recall the basic production mechanisms of LiBeB, in section 3 we decline the astrophysical agents and sites, in section 4, we compare the various models designed, in section 5 we propose key observations to remove the present ambiguities.
## 2. Nucleosynthesis of LiBeB
### 2.1. Thermal production and destruction
Thermal nucleosynthesis in the Big Bang produces negligible amounts of Be and B. Only <sup>7</sup>Li is synthesized in significant amounts. Moreover, LiBeB do not survive stellar temperatures except in a thin surface layer where they are observed, reflecting the interstellar composition inherited by the star at birth. <sup>7</sup>Li is however thought to be produced by AGB stars (Abia et al 1993) and novae (Hernanz et al 1996) and also by SNII through neutrino spallation of carbon. This neutrino spallation in carbon shells of type II supernovae is expected to produce also <sup>11</sup>B, but the yields are sensitive to the assumed temperature (energy) of the neutrinos, which is uncertain (see Hartmann et al 1999 for a review).
### 2.2. Non Thermal production
Nuclear spallation, i.e. the break up of medium heavy elements by collisions with protons and alphas remains the leading production process of light elements in the cosmos (Meneguzzi, Audouze & Reeves 1971). In principle, all isotopes of interest are generated either by the interaction of fast p and alphas on CNO at rest in the ISM, or conversely by the interaction of fast C and O (principally) with ambient H and He, supplemented by the alpha +alpha reaction giving rise exclusively to Li isotopes. (e.g. Reeves 1994).
The cross sections are well measured (Read & Viola 1985; Webber et al 1990 a, b), and have been updated recently by Ramaty et al (1997). The hierarchy of the cross sections reflects that of the abundances of the light nuclei in nature. In a collision between a proton and an oxygen nucleus, the probability of production of <sup>11</sup>B is higher than that of <sup>10</sup>B which is itself higher than that of <sup>9</sup>Be. Thus, it is not surprising that the abundances of these three isotopes go in declining order. This is a stricking example of a direct application of nuclear physics in the understanting of natural abundances. Note that the peaks of the cross sections lie at low energy especially that of alpha + alpha (Read & Viola 1985). Thus low energy particles (about 10 MeV/n) have to be inserted carefully in the treatment of the problem. Note also that at low energy, where the alpha + alpha reaction is operating at full strength, <sup>7</sup>Li and <sup>6</sup>Li are produced in comparable amounts (1.5), which is at variance with the <sup>7</sup>Li/<sup>6</sup>Li ratio observed in meteorites (12.5). A stellar source of pure <sup>7</sup>Li has to be invoked to explain this high value (see above).
The production rate, in the most general case, is function i) of the number density of the target nuclei ($`N_T`$) and their composition, ii) of the flux of the projectiles ($`\mathrm{\Phi }`$) and iii) of the cross section averaged over the energy spectrum in the interaction region :
$`dN(L)/dt=N_T<\sigma >\mathrm{\Phi }`$.
If p and alpha are the projectiles and CNO the targets, one deals with a ”secondary” production, thus:
$`dN(L)/dt=N_{CNO}<\sigma >\mathrm{\Phi }_{p,\alpha }`$.
It is assumed (quite reasonably) that the flux of energetic particles is proportional to the supernova rate since they are thought to be the main agents of acceleration through the shock waves they produce. Due to the fact that SNII are also the main O producers, one expects that:
$`\mathrm{\Phi }(t)prop.d(SN(t))/dtprop.d(O/H)/dt`$.
On the other hand, the CNO abundance cumulated in the ISM up to time t, is proportional to the total number of SN having exploded from 0 to t. Summarizing:
$`d(Be/H)/dtprop.(O/H)d(O/H)/dt`$ or after integration Be/H prop. $`(O/H)^2`$.
In contrast, if C, O are the projectiles and H, He the targets, a primary production arises governed by the equation:
$`dN(L)/dt=N_{HHe}<\sigma >\mathrm{\Phi }_{CO}`$
Now with the same hypothesis than above, and considering that the target abundances (H, He) do not evolve significantly:
$`d(Be/H)/dtprop.d(O/H)/dtandBe/Hprop.O/H`$
## 3. Astrophysical agents and sites
### 3.1. Galactic Cosmic Rays (GCR)
The energy spectrum of GCR is directly observed above, say, 1GeV/n. Below, it is deduced from various indirect observations (Strong & Moskalenko 1999). It is reasonably well explained by the diffuse shock wave acceleration mechanism (Blandford & Ostriker 1978; Jones & Ellison 1991; Ellison et al 1997). The observed composition is extrapolated back to the sources thanks to a classical propagation model.
It is edifying to compare the elemental and isotopic source composition to that of other materials of known abundances (the solar system for instance, or that of the supernova ejecta computed with stellar models). Indeed SNIa alone do not fit O/Fe, Ne/Fe, Mg/Fe neither the s process elements, whereas, SNII alone do not fit Fe/Co/Ni, neither the s process elements. (see e.g. Meyer 1996, 2000). Indeed, the nucleosynthetic origin of the groups (Mg, Si, Ca, Fe, Ni), (Sr, Zr, Ba, Ce), (Pt peak, actinides) are all different (explosive burning, s-process and r process, respectively) and their production sites are also different, thus the solar mix is a complex mixture of all that, historitically built up, and individual sources are unlikely to lead to the GCR source abundances (since they are similar to that of the solar system). Note that in this example we have chosen only refractory elements that are not supposed to be affected by selective effects (see below).
A stricking fact, of great importance for our purpose, is the similarity between the isotopic composition of the GCRs and of the solar system (Connell & Simpson 1997; Stone et al 1998; Wiedenbeck et al 1999), which is a mixture of the products of generations and generations of stars of different masses indicating at face value that GCR are nuclei accelerated out of a normal reservoir and not strange, exotic, stars or objects.
In recent years, grains have been considered to play a central role in explaining the pecularities of the GCR source composition (Meyer et al 1997; Ramaty et al 1997). It is assumed that grain debris are more efficiently accelerated than elements in the gas phase and grain models have progressively replaced the traditional two step acceleration mechanism (injection by flare stars and acceleration by supernova shocks) in which grains were undesirable (Cassé & Goret 1978, Meyer 1985; Silberberg & Tsao 1990).
However the situation is not absolutely settled (Shapiro 1999, Silberberg et al 2000; Cassé & Vangioni-Flam 2000, in preparation). Anyway, among the grain supporters themselves, there are divergent views concerning the origin and nature of the grains of interest. In one camp, they are supposed interstellar (Meyer and coworkers) and on the other they are supposed to contain fresh products of nucleosynthesis (Ramaty and colleagues). This has a strong bearing on the primary or secondary character of the LiBeB production mechanism. If the grains that are impacted by the accelerating shock waves have the ISM composition, then, the resulting CR composition should reflect that of the ISM which is the true reservoir of CR particles and, as it is H, He dominated, the process is secondary.
Ramaty and coworkers (see Ramaty et al 2000 and references therein) assume that GCR originate from grains loaded with freshly synthesized nuclei, C and O that are released by SNII and accelerated by shock waves in galactic superbubbles (SB). Subsequently, they interact with the surrounding interstellar medium to give LiBeB in a primary way. However, this proposal has to face the following objections: in the superbubble (SB), high temperature context, according to the observed trend (Cardelli 1996; Savage & Sembach, 1996; Jenkins et al 1998; Howk, Savage & Fabian 1999), grains have little chances to remain intact. Only their refractory cores would survive in the hot phase. Silicon, for instance would be significantly evaporated (see fig 5 in Savage & Sambach 1996) with respect to Fe, and the Si/Fe ratio in the accelerated particles should be different than solar contrary to what is deduced from the GCR observations. Thus grains in SB are unlikely to account, in a detailed manner, for the present GCRS composition. Moreover, the <sup>34</sup>S/<sup>32</sup>S ratio, derived at the cosmic ray sources poses also a problem (Cassé & Vangioni-Flam 2000 in preparation). Finally, the grain mechanism does not go without gas (H, He) acceleration and thus particle accelerated in superbubbles cannot induce a purely primary process all time long and as the Galaxy evolves, the secondary process becomes more and more important. In this case, the transition primary-secondary will take place at about \[Fe/H\] about -1 (taking an average superbubble metallicity of about 5 times solar, Higdon et al 1998). Thus a purely primary component seems unlikely.
On this basis we are tempted to conclude that standard GCR, as traditionally thought, act as a secondary source of LiBeB, and if a primary component is made necessary by the data, it should be different from the standard GCR one. It could come from SB’s under the condition that this SB component is confined to low energy not to spoil the observed CGR composition.
### 3.2. Superbubble Accelerated Particles (SAP)
Superbubbles gather a great number of massive stars which explode as type II supernovae, enriching the surrounding medium in fresh products of nucleosynthesis, and among them, oxygen. The shock wave and turbulence sutained by a given supernova, accelerate the material enriched by the previous ones (Bykov & Fleisshman 1992; Bykov 1995; Bykov et al 2000). The energy spectrum obtained could depart significantly from the GCR one, depending on the detail of the mechamism and more precisely on the escape time of the fast particles, which is poorly known (Klepach et al 1999). As said previously, an energy cut off should be imposed to avoid contamination of the (observed) GCR abundances. This energy cut off is still a free parameter and we can choose it in order to avoid energetic problems (Ramaty et al 1996), i.e. in the range 30-300 MeV/n. Then, all in all, superbubbles appear to be the best agents of a primary production process of LiBeB (Parizot & Drury 1999; Bykov et al 2000). Admittedly, this theoretical proposal remains to be firmly (experimentally) substantiated by X ray and gamma-ray line observations (see the proceedings edited by Ramaty et al 1999 for a general discussion). The search for a non thermal low energy component as the one once claimed to be discovered by the COMPTEL experiment but discarded afterwards (Bloemen et al 1999) is, in our minds, one of the major objectives of the European INTEGRAL satellite to be lauched in 2001. For the time being, we can only speculate on the composition and the spectrum of this hypothetical (but somewhat necessary) component. In the halo phase, the composition is taken as representative of the ejecta of massive low metallicity stars. It is highly enriched in O w.r.t present CRs. (Woosley & Weaver 1995). This composition however is expected to vary in time due to metallicity dependent mass loss rate (Vangioni-Flam et al 1997).
## 4. Production and evolution of LiBeB
### 4.1. Observed correlations
Turning to observations, we are confronted to a certain ambiguity (as mentionned in the introduction). According to the choice of the conversion between Fe and O abundances we get different conclusions. Relying on the Israelian et al (1998) and Boesgaard et al (1999) relation we get Be proportional to $`O^{1.7}`$, i.e. nearly secondary, but taking the conventional O-Fe relation, strengthened by the Mg data, we get Be proportional to O, i.e. purely primary, at least in the halo phase. Thus the O/Fe behaviour at low metallicity is a central issue, once again. Furthermore non LTE effects on B (Kiselman 1999) make the situation even more complex (see also Primas 2000). Thus we must rely on a independent argument to answer the question: is a primary component really necessary? This argument of energetic nature has been essentially developed by Ramaty et al (1996, 1997, 2000). To briefly summarize: it seems that a primary component is required by both the Mg-Fe relation (more secure than O-Fe) and by energetic arguments.
### 4.2. The three evolutionary models
Concerning the galactic evolutionary models, there exists three types of them i) a pure secondary standard GCR based on the variable O/Fe ratio (Fields & Olive 1999a) ii) a pure primary GCR from SB, which, suprisingly is still valid with flat or variable O/Fe (Ramaty et al 2000) iii) an hybrid model (Standard GCR + SAP based on a flat O/Fe ratio (Vangioni-Flam et al 1998) or based on a variable O/Fe ratio (Fields et al 2000).
### 4.3. Energetic requirements
The number of atoms of Be produced per erg injected per supernova is promoted to the role of a selection criterion for the theoretical models (both primary and secondary). Stated differently, are there enough SNII and are they sufficiently efficient to produce all the Be observed in halo stars in the primary and/or secondary cases? According to Ramaty et al (2000) the pure secondary standard GCR produce BeB at high energy cost in the early Galaxy. SAP or primary GCR are much more economical. Thus a very plausible solution is that a primary component appears first, and a secondary component takes over afterwards, at a metallicity (O/H ratio) which remains to be determined precisely. The position of the break depends on the Fe-O correlation used. Relying on the new one, and using the analysis based on the IRFM data, (Fields et al 2000) one finds a break point at about \[O/H\] = -1.6. On the other hand, if the ancient correlation is chosen, motivated by the magnesium data, the break point (if any) is at higher \[O/H\] (about -1). Anyway, the existence of a primary component in the early stages of the evolution of the Galaxy seems mandatory, irrespective of the abundance data used, and this is a strong conclusion. Definitively, a primary component is required to fulfill the energetic constraint. What is the nature of this primary component? In our opinion, once again, it is distinct from GCR (assumed by Ramaty et al 2000 to originate from SBs), since, as said previously, we do not think that it is possible to identify SAP and GCR, as Higdon et al (1998) did, due to their different inferred composition.
Thus, we support the view that a low energy component is at work, complementing GCR to produce its lot of LiBeB, specifically in the halo phase. Is this component still active? Nuclear gamma ray line astronomy will say. Thus our hopes are related to the INTEGRAL satellite. Our conclusion is that an hybrid model combining both a primary and a secondary components reconcile the abundance observations and energy requirement. The two components dominate sequentially. Of course the primary component, related to SAP would be overwhelming in the halo phase and is afterwards dominated by the secondary one, which constitutes the standard GCR. The SAP component plays a major role in the LiBeB production when the galactic gas is almost devoid of medium elements i.e. in the early Galaxy. Now, what is the ultimate reservoir of GCR? Grains in the ISM or stellar surfaces? This point is left to a future discusssion.
## 5. Conclusion
Ambiguities on abundance data preclude definitive conclusion: is O/Fe flat or not at low Z? For the time being, the answer to this question depends on the observer to whom it is posed. Hopefully the debate will clarify in the next years. In the mean time we propose to rely on magnesium which is used as a secure metallicity index by a growing number of people. Special care should be taken to make NLTE corrections on Fe and B at low Z. A primary process probably is made necessary by energetic requirements. We endorse the view that two different components are responsible for the synthesis and evolution of LiBeB in the Galaxy. This hybrid model invoking the operation of both GCR (extracted by flares from stellar surfaces and/or grain debris in the ISM) subsequently accelerated by shock waves and a primary component of lower energy coming probably from superbubbles fulfil (or at least do not violate) all the composition constraints on i) the present cosmic radiation ii) the light element abundances in stars of all metallicities, including <sup>6</sup>Li (Vangioni-Flam et al 1999a; Fields & Olive 1999b).
What to do next? Obviously measure. One would like idealy to get simultaneously the abundances of <sup>6,7</sup>Li, Be, B, O, Mg and Fe in the same star, this for many halo members of various metallicities, which does not seems out of reach of a dedicated VLT program (Cayrel, private communication). The solution of the LiBeB riddle is definitly in the hands of observers. The best way to reveal the presence of a low energy flux of C and O in the Galaxy related supperbubbles is the observation of broad gamma ray lines arising from their excitation and deexcitation in flight. The best hope to detect them in Vela, Orion and other star forming regions is offered by the european INTEGRAL satellite to be launched around 2001.
Aknowledgements
We thank R. Cayrel and C. Furhmann for having drawn pure attention on the reliability of using magnesium as a metallicity index. We thank also Andrei Bykov and Vladimir Ptuskin for illuminating discussions on superbubbles.
## References
Abia, C., Isern, J. & Canal, R. 1993, A&A, 275, 96
Balachandran, S. & Bell, R.A. 1998, Nature, 392, 791
Barring, M. 2000, Summary-Rapporteur Volume of The 26th International Cosmic ray Conference, ed. B.L. Dingus (AIP, New-York). astro-ph/9912058
Blandford, R.D. & Ostriker, J.P. 1978, ApJ, 221, L29
Bloemen, H. et al 1999, in ”The Extreme Universe”, 3rd INTEGRAL Workshop, Astrophysical Letters and Communications, vol. 38, 349
Boesgaard, A.M. & King, J.R. 1993, AJ, 106, 2309
Boesgaard, A.M., King, J.R., Deliyannis, C.P. & Vogt, S.S. 1999, AJ, 117, 492
Bykov, A.M. 1995, Space Sci. Rev., 74, 397
Bykov, A.M. & Fleishman, G.D. 1992, MNRAS, 15, 269
Bykov, A.M., Gustov, M.Y. & Petrenko, M.V. 2000, preprint
Cardelli, J.A. 1996, Science, 265, 209
Cassé, M. Lehoucq, R. & Vangioni-Flam, E. 1995, Nature, 373, 38
Cassé, M. & Goret, Ph. 1978, ApJ, 221, 703
Connel, J.J. & Simpson, J.A. 1997, ApJ, 475, L61
Duncan, D., Lambert, D.L. & Lemke, M. 1992, ApJ, 401, 584
Duncan, D. et al 1997, ApJ, 488, 338
Ellison, D., Drury, L.O’C. & Meyer, J.P. 1997, ApJ, 487, 197
Fields, B.D. & Olive, K.A. 1999a, ApJ, 516, 797
Fields, B.D. & Olive, K.A. 1999b, New Astron., 4, 255
Fields, B.D., Olive, K.A., Vangioni-Flam, E. & Cassé, M. 2000, astro-ph/9911320
Fuhrmann, K. 1998, A&A, 338, 161
Fulbright , J.P. & Kraft, R.P. 1999, AJ, 118, 527
Gilmore, J. et al 1992, Nature, 357, 379
Gustafsson, B. 1999, in ”Chemical evolution from zero to high redshift”, eds: J.R. Walsh & M.R. Rosa, Springer, 1
Hartmann, D., Myers, J. , Woosley, S., Hoffman, R. & Haxton, H. 1999, in ”LiBeB, Cosmic rays, and related X-and gamma-rays”, edts: Ramaty et al. ASP Conf. Ser. vol 171, 235
Hernanz, M., José, J., Coc, A. & Isern, J. 1996, ApJ, 465, L27
Higdon, B., Lingenfelter, R.E. & Ramaty, R. 1998, ApJ, 309, L33
Howk, J.C., Savage, B.D. & Fabian, D. 1999, ApJ, 525, 253
Israelian, G., Garcia-Lopez, R.J. & Rebolo, R. 1998, ApJ, 507, 805
Jenkins, E.J. et al 1998, ApJ, 492, L147
Jones, F.C. & Ellison, D.C. 1991, Sp. Sci. Rev., 58, 259
Klepach, E.G., Psuthskin, V.S. & Zirakashvili, V.N. 1999, Astroparticle Physics, in press
Kiselman, D. 1999, in”LiBeB Cosmic Rays, and Related X-Ray and Gamma-rays”, Edts Ramaty et al, ASP Conf. Ser. vol 171, 85
McWilliam, A. 1997, ARA&A, 35, 503
Meneguzzi, M., Audouze, J. & Reeves, H. 1971, A&A, 15, 337
Meyer, J.P. 1985, ApJS, 57, 173
Meyer, J.P. 1996, in ”The Sun and beyond”, 2nd Rencontre du Vietnam, Eds: Tran Than Van et al, Edts frontières, 27
Meyer, J.P., Drury, L.O’C. & Ellison, D.C. 1997, ApJ, 488, 730
Meyer, J.P. 2000, ISSI Workshop, Kluwer Academics Publishers, in press
Pagel, B.E.J & Tautvaisiene, G. 1995, MNRAS, 276, 505
Pagel, B.E.J. 1999, astro-ph/9911204
Parizot, E. & Drury, L.O’C. 1999, A&A, 346, 339
Primas, F., Duncan, D.K., Peterson, R.C. & Thorburn, J.A. 1999, A&A, 313, 545
Primas, F. 2000, this conference
Ramaty, R., Kozlovsky, B. & Lingenfelter, R.E. 1996, ApJ, 456, 525
Ramaty, R., Kovlowsky, B., Lingenfelter, R.E. & Reeves, H. 1997, ApJ, 488, 730
Ramaty, R., Vangioni-Flam, E., Cassé, M. & Olive, K.A. 1999, Proceedings ”LiBeB, Cosmic Rays, and Related X-and Gamma-Rays, ASP Conf. Ser., vol. 171
Ramaty, R., Scully, S., Lingenfelter, R.E. & Kozlowsky, B. 2000, astro-ph/9909021, ApJ, in press
Read, S. & Viola, R. 1985, Atomic Data Nucl. Data tables, 31, 359
Rebolo, R., Molaro, P. & Beckman, J.E. 1988, A&A 192, 192
Reetz, J. 2000, in ”Galaxy Evolution: Connecting the distant Universe with Local Fossil Record”, Eds M. Spite and N. Crifo, Astroph. Spc. Sci., in press
Reeves, H. 1994, Rev. Mod. Phys., 66, 193
Ryan, S.G., Norris, I., Bessel, M. & Deliyannis, C. 1994, ApJ, 388, 184
Savage, B.D. & Sembach, K.R. 1996, ARA&A, 34, 279
Shapiro, M. 1999, in ”LiBeB, Cosmic rays, and related X-and Gamma-Rays, Edts Ramaty el al, ASP Conf. Ser. vol. 171, 138
Shigeyama, T. & Tsujimoto, T. 1998, ApJ, 507, L139
Silberberg, R. & Tsao, C.H. 1990, ApJ, 352, L49
Silberberg, R. Tsao, C.H. & Barghouty, A.F. 2000, in the 26th International Cosmic Ray Conference eds B.L. Dingus AIP (New York), OG.3.1.06, in press
Stone, E. et al 1998, Spc. Sci. Rev., 96, 285
Strong, A. & Moskalenko, I. 1999, in ”LiBeB, Cosmic rays, and related X-and Gamma-Rays, Edts Ramaty el al, ASP Conf. Ser. vol. 171, 162
Thomas, D., Greggio, L. & Bender, R. 1998, MNRAS, 296, 119
Thielemann, F.K., Nomoto, K. & Hashimoto, M. 1996, ApJ, 460, 408
Umeda, H., Nomoto, K. & Nakamura, W. 2000, to appear in the proceedings of the MPA/ESO conference ”The first stars” eds A.Wess et al, Springer , astro-ph/9912248
Vangioni-Flam, E., Cassé, M., Fields, B. & Olive, K. 1996, ApJ, 468, 199
Vangioni-Flam, E., Cassé, M. & Ramaty, R. 1997, in ” The Transparent Universe”, 2nd INTEGRAL Workshop, ESA, SP382, 123
Vangioni-Flam E., Ramaty, R., Olive, K.A. & Cassé, M. 1998, A&A, 337, 714
Vangioni-Flam, E. Cassé, M., Cayrel, R., Audouze, J., Spite, M. & Spite F. 1999a, New Astronomy, vol. 4, no 4, 245
Vangioni-Flam, E., Ramaty, R., Cassé, M. & Olive, K.A. 1999b, in ”LiBeB, Cosmic-Ray and related X-and gamma-rays”, Edts E. Ramaty et al, ASP Conf. Ser. vol. 171, 268
Vangioni-Flam, E. & Cassé, M. 2000a, in ”Galaxy Evolution: Connecting the distant Universe with the local fossil Record” Edts M. Spite, F. Crifo, Ap&SS, in press
Vangioni-Flam, E. Cass é, & Audouze, J. 2000b, Physics Report, in press
Webber, W.R., Kish, J.C. & Schrier, D.A. 1990a, Phys. Rev. C 41, 520
Webber, W.R., Kish, J.C. & Schrier, D.A. 1990b, Phys. Rev. C 41, 547
Wiedenbeck, M.E. et al 1999, ApJ, 523, L61
Woosley, S.E., & Weaver, T.A. 1995, APJS, 101, 181 |
no-problem/0001/math0001074.html | ar5iv | text | # Exactness and the Novikov Conjecture
## 1. Introduction
Let $`X`$ be a discrete metric space with metric $`d`$. A function $`f`$ from $`X`$ to a separable Hilbert space $``$ is a uniform embedding if there exist non-decreasing proper functions $`\rho _\pm :[0,\mathrm{})[0,\mathrm{})`$ such that
$`\rho _{}(d(x,y))f(x)f(y)\rho _+(d(x,y)),\text{for all }x\text{}yX.`$
The Strong Novikov Conjecture states that the assembly map on K-theory,
$$\mu :K_{}(B\mathrm{\Gamma })K_{}(C_r^{}(\mathrm{\Gamma })),$$
is injective. Answering a question of Gromov, Yu proved the following theorem .
###### Theorem 1.1.
Let $`\mathrm{\Gamma }`$ be a finitely presented discrete group. If $`\mathrm{\Gamma }`$ is uniformly embeddable in a Hilbert space, then $`\mathrm{\Gamma }`$ satisfies the Strong Novikov conjecture.
This is currently the weakest general hypothesis implying the Novikov conjecture. It is conceivable, however, that there exist groups which are not uniformly embeddable in a Hilbert space but which nevertheless satisfy the Novikov conjecture. At present there are no such examples known.
From another direction, there is the question of whether all finitely generated discrete groups are exact, . Recall that a discrete group $`\mathrm{\Gamma }`$ is exact if its reduced $`C^{}`$-algebra, $`C_r^{}(\mathrm{\Gamma })`$, is an exact $`C^{}`$-algebra. That is, given the exact sequence of $`C^{}`$-algebras
$$0IBB/I0$$
the sequence
$$0I_{min}C_r^{}(\mathrm{\Gamma })B_{min}C_r^{}(\mathrm{\Gamma })(B/I)_{min}C_r^{}(\mathrm{\Gamma })0$$
is also exact.
The Novikov conjecture and Exactness question appear to have little in common other than that they both involve properties which might be possessed by all finitely presented groups. However, there is a link provided by results of Roe-Higson and Yu . Combined these results state that if $`\mathrm{\Gamma }`$ acts amenably, in the topological sense , on a compact space then it is uniformly embeddable in a Hilbert space, and hence satisfies the Novikov conjecture. On the other hand, it is an easy observation that this condition also implies that $`\mathrm{\Gamma }`$ is exact. Thus, the same hypothesis yields both properties.
We note that Gromov has asserted the existence of finitely presented groups that are not uniformly embeddable . This follows from his assertion that there exists a finitely presented group whose Cayley graph contains a sequence of expanding graphs , together with his observation that, when viewed as a discrete metric space, a sequence of expanding graphs is not uniformly embeddable. On the other hand, it follows simply from a result of Voiculescu that the uniform algebra of such a metric space is, in general, not exact. Based on this it seems likely that Gromov’s examples of non-uniformly embeddable groups will in general fail to be exact.
The purpose of this note is to study the relationship between uniform embeddability and exactness. We state the main result, leaving precise definitions for later in the paper. We need the uniform Roe algebra, $`UC^{}(\mathrm{\Gamma })`$, sometimes called the “rough” algebra, introduced by Roe, . It is isomorphic to the reduced cross product $`C(\beta \mathrm{\Gamma })_r\mathrm{\Gamma }`$. The left regular representation provides an inclusion of $`C_r^{}(\mathrm{\Gamma })`$ into $`(l^2(\mathrm{\Gamma }))`$, and in fact into $`UC^{}(\mathrm{\Gamma })`$. Recall that a group $`\mathrm{\Gamma }`$ is exact if this inclusion is a nuclear embedding of $`C_r^{}(\mathrm{\Gamma })`$ into $`(l^2(\mathrm{\Gamma }))`$ . We modify this condition by requiring that the inclusion be a nuclear embedding of $`C_r^{}(\mathrm{\Gamma })`$ into $`UC^{}(\mathrm{\Gamma })`$. Note that most classes of groups which are know to be exact, including word hyperbolic groups, discrete subgroups of connected Lie groups, Coxeter groups, etc., actually satisfy this stronger condition.
###### Theorem 1.2.
Let $`\mathrm{\Gamma }`$ be a finitely presented group. If the inclusion of the reduced $`C^{}`$-algebra $`C_r^{}(\mathrm{\Gamma })`$ into the uniform Roe algebra $`UC^{}(\mathrm{\Gamma })`$ is a nuclear map, then $`\mathrm{\Gamma }`$ is uniformly embeddable in a Hilbert space, and hence satisfies the Novikov conjecture.
It is natural to consider other refinements of exactness which can be obtained by replacing $`UC^{}(\mathrm{\Gamma })`$ by other subalgebras of $`(l^2(\mathrm{\Gamma }))`$. This will be discussed in a future paper.
## 2. Approximate units and negative type functions
In this section we will assemble some of the facts needed for the results in Section 3. In particular, we will establish an analog of a theorem of Akemann-Walter, .
A complex-valued function $`f`$ on the set $`X\times X`$ is said to be positive definite if, for any $`n1`$,
$`{\displaystyle \underset{ij}{}}\overline{z}_if(x_i,x_j)z_j0,\text{for all }x_1,\mathrm{},x_nX\text{ and }z_1,\mathrm{},z_n.`$
A real-valued function $`h`$ on $`X\times X`$ is of negative type if
* $`h(x,x)=0`$ for all $`xX`$,
* $`h(x,y)=h(y,x)`$ for all $`x`$, $`yX`$, and
* $`_{ij}a_ih(x_i,x_j)a_j0`$, for all $`x_1,\mathrm{},x_nX`$ and $`a_1,\mathrm{},a_n`$ satisfying $`_ja_j=0`$.
It will be convenient to have the following notation. If $`X`$ is a metric space and $`A`$ is a subspace, then $`C_0(X;A)`$ will be the set of functions which tend to zero off of $`A`$. That is, $`fC_0(X;A)`$ if for any $`\epsilon >0`$ there is an $`R>0`$ such that $`|f(x)|<\epsilon `$ if $`d(x,A)>R`$.
Suppose now that $`X`$ is a discrete metric space. Consider the ideal $`C_0(X\times X;\mathrm{\Delta })l^{\mathrm{}}(X\times X)`$ , where $`\mathrm{\Delta }`$ denotes the diagonal of $`X\times X`$. A sequence $`f_nl^{\mathrm{}}(X\times X)`$ satisfies $`f_nff0`$ for all $`fC_0(X\times X;\mathrm{\Delta })`$ if and only if $`f_n1`$ uniformly on any set of the form $`B_\mathrm{\Delta }(R)=\{(x,y):d(x,y)<R\}`$. Finally, we say that a complex-valued function $`f`$ on $`X\times X`$ is metrically proper if it satisfies that for any $`C>0`$ there is an $`R>0`$ such that $`|f(x,y)|>C`$ if $`d(x,A)>R`$.
The following result is a generalization of \[1, Theorem 10\] from the case of groups to that of equivalence relations.
###### Theorem 2.1.
Let $`X`$ be a discrete metric space. There exists an approximate unit for $`C_0(X\times X;\mathrm{\Delta })`$ consisting of positive definite functions if and only if there exists a metrically proper negative type function on $`X\times X`$.
###### Proof.
Let $`\varphi `$ be a metrically proper negative type function on $`X\times X`$. By a generalization of Schoenberg’s Theorem, , the function $`e^{t\varphi (x,y)}`$ is positive definite for any $`t0`$. Since $`\varphi `$ is metrically proper, one has, for any $`t`$, $`e^{t\varphi (x,y)}C_0(X\times X;\mathrm{\Delta })`$. On the other hand one also has $`lim_{t0}e^{t\varphi (x,y)}1=0`$ uniformly on $`B_\mathrm{\Delta }(R)`$ for any $`R>0`$. Thus, $`\mathrm{\Phi }_t=e^{t\varphi }`$ provides the approximate unit for $`C_0(X\times X;\mathrm{\Delta })`$ consisting of positive definite functions.
For the converse, let $`u_\lambda `$ be an approximate unit consisting of positive definite functions. Since $`u_\lambda 1`$ uniformly on $`B_R(\mathrm{\Delta })`$, there exists an $`R`$ and $`\lambda _0`$ so that $`u_\lambda >0`$ if $`d(x,y)<R`$ and $`\lambda >\lambda _0`$. One may thus adjust the approximate unit so that $`u_\lambda (x,x)=1`$ for all $`xX`$. Now, exactly as in \[1, Theorem 10\] one extracts a sequence $`u_{\lambda _i}`$ such that the function $`_nRe(1u_{\lambda _n})2^n`$ converges to the required metrically proper negative type function. ∎
We next recall the result of Yu, , relating metrically proper negative type functions to uniform embeddings in a Hilbert space.
###### Theorem 2.2.
The metric space $`X`$ is uniformly embeddable in a Hilbert space if and only if there exists a metrically proper negative type function on $`X\times X`$.
Combining these two results we obtain
###### Theorem 2.3.
The following are equivalent for the countable discrete metric space $`X`$.
* $`X`$ is uniformly embeddable in a Hilbert space.
* There is a metrically proper negative type function on $`X\times X`$.
* There is an approximate unit for $`C_0(X\times X;\mathrm{\Delta })`$ consisting of positive definite functions.
In Section 3 we will discuss the relation of this to the Haagerup property for the groupoid $`\beta \mathrm{\Gamma }\mathrm{\Gamma }`$.
## 3. Exactness
In this section we restrict $`X`$ to be a finitely presented group with a length function determined by a finite, symmetric set of generators. The length function, $`l`$, determines a right invariant metric via $`d(s,t)=l(st^1)`$. The quasi-isometry type of $`(\mathrm{\Gamma },d)`$ is independent of the choice of generators. We next recall the definition of the uniform Roe algebra associated to $`(\mathrm{\Gamma },d)`$.
Consider the set of $`A:\mathrm{\Gamma }\times \mathrm{\Gamma }`$ satisfying
* there exists $`M>0`$ such that $`|A(s,t)|M`$, for all $`s`$, $`t\mathrm{\Gamma }`$
* there exists $`R>0`$ such that $`A(s,t)=0`$ if $`d(s,t)>R`$
Each such $`A`$ defines a bounded operator on $`l^2(\mathrm{\Gamma })`$ via the usual formula for matrix multiplication:
$$A\xi (s)=\underset{r\mathrm{\Gamma }}{}A(s,r)\xi (r),\text{for }\xi l^2(\mathrm{\Gamma }).$$
These will be referred to as finite width operators. The collection of finite width operators is a $``$-subalgebra of $`(l^2(\mathrm{\Gamma }))`$. The uniform Roe algebra of $`\mathrm{\Gamma }`$, denoted $`UC^{}(\mathrm{\Gamma })`$, is the closure of the $``$-algebra of finite width operators. It is a $`C^{}`$-algebra The quasi-isometry class of $`(\mathrm{\Gamma },d)`$ determines $`UC^{}(\mathrm{\Gamma })`$, which is therefore independent of the choice of generators.
Every $`t\mathrm{\Gamma }`$ acts on $`l^2(\mathrm{\Gamma })`$ by the left regular representation. The action of $`t\mathrm{\Gamma }`$ is represented by the matrix $`A`$ defined by $`A(s,r)=1`$ if and only if $`s=tr`$. Clearly, $`t\mathrm{\Gamma }`$ acts as a finite width operator. Thus, $`[\mathrm{\Gamma }]UC^{}(\mathrm{\Gamma })`$, and we have
$`C_r^{}(\mathrm{\Gamma })UC^{}(\mathrm{\Gamma })(l^2(\mathrm{\Gamma })).`$
Recall that if a unital \*-homomorphism, $`T:𝒜`$, between unital $`C^{}`$-algebras is nuclear then there is a net $`T_\lambda :𝒜`$ of finite rank, unital, completely positive linear maps such that $`lim_\lambda T_\lambda (x)T(x)=0`$ for all $`x𝒜`$. It was shown by Kirchberg that a unital $`C^{}`$-algebra $`𝒜`$ is exact if and only if every non-degenerate, faithful representation of $`𝒜`$ on a Hilbert space $``$ provides a nuclear embedding of $`𝒜`$ into $`B()`$. In particular, a discrete group $`\mathrm{\Gamma }`$ is exact if and only if the inclusion of $`C_r^{}(\mathrm{\Gamma })`$ into $`(l^2(\mathrm{\Gamma }))`$ given by the left regular representation is a nuclear embedding. The main theorem of this section states that if one restricts the range of the nuclear embedding a little bit, then this strengthened form of exactness implies the uniform embeddability of $`\mathrm{\Gamma }`$.
###### Theorem 3.1.
Let $`\mathrm{\Gamma }`$ be a finitely generated discrete group. If the inclusion $`C_r^{}(\mathrm{\Gamma })UC^{}(\mathrm{\Gamma })`$ is a nuclear map then $`\mathrm{\Gamma }`$ is uniformly embeddable in a Hilbert space (and hence satisfies the Novikov conjecture).
###### Proof.
By Theorem 2.3 it is sufficient to produce an approximate unit for $`C_0(X\times X;\mathrm{\Delta })`$ consisting of positive definite functions. This will be obtained using nuclearity of the inclusion.
There is a general procedure to associate to a linear map $`T:C_r^{}(\mathrm{\Gamma })(l^2(\mathrm{\Gamma }))`$ a function $`u:\mathrm{\Gamma }\times \mathrm{\Gamma }`$ given by the formula
$`u(s,t)=\delta _s,T(st^1)\delta _t,`$
where $`\delta _t`$ denotes the characteristic function of the element $`t\mathrm{\Gamma }`$. Note that if $`T`$ is bounded then $`ul^{\mathrm{}}\mathrm{\Gamma }\times \mathrm{\Gamma }`$. The correspondence
$$\{T:C_r^{}(\mathrm{\Gamma })(l^2(\mathrm{\Gamma }))\}\{u:\mathrm{\Gamma }\times \mathrm{\Gamma }\}$$
has the following properties:
* if $`T`$ is unital and completely positive then $`u`$ is positive definite, and
* if $`T:C_r^{}(\mathrm{\Gamma })UC^{}(\mathrm{\Gamma })`$ has finite rank then $`uC_0(\mathrm{\Gamma }\times \mathrm{\Gamma };\mathrm{\Delta })`$.
Further if $`T_\lambda :C_r^{}(\mathrm{\Gamma })(l^2(\mathrm{\Gamma }))`$ is a net of bounded linear maps with associated functions $`u_\lambda `$ then
* if $`T_\lambda (x)x0`$, for all $`xC_r^{}(\mathrm{\Gamma })`$, then $`u_\lambda 1`$ uniformly on $`B_R(\mathrm{\Delta })`$ for all $`R`$.
We verify these properties below, but for now observe that together they imply the desired result. Assuming nuclearity of the inclusion of $`C_r^{}(\mathrm{\Gamma })`$ into $`UC^{}(\mathrm{\Gamma })`$ we obtain unital completely positive maps $`T_\lambda :C_r^{}(\mathrm{\Gamma })UC^{}(\mathrm{\Gamma })`$ as above. It follows immediately from the properties above that the associated functions $`u_\lambda l^{\mathrm{}}(\mathrm{\Gamma }\times \mathrm{\Gamma })`$ form the desired approximate unit..
We now turn to the verification of (i)–(iii), beginning with (i). Let $`s_1,\mathrm{},s_n\mathrm{\Gamma }`$ and $`z_1,\mathrm{},z_n`$. Define an element of $`=l^2(\mathrm{\Gamma })`$ by $`\xi =(z_1\delta _{s_1},\mathrm{},z_n\delta _{s_n})`$ and an operator on $`l^2(\mathrm{\Gamma })`$ by the $`n\times n`$ matrix $`A=[A_{ij}]M_n((l^2(\mathrm{\Gamma })))`$ where
$`A_{ij}=T(s_js_i^1).`$
A direct calculation shows
$`{\displaystyle \underset{i,j}{}}\overline{z}_iu(s_i,s_j)z_j`$ $`=`$ $`{\displaystyle \underset{i,j}{}}\overline{z}_i\delta _{s_j},T(s_js_i^1)\delta _{s_i}=\xi ,A\xi _{}.`$
Thus, it suffices to show that $`A`$ is positive operator on $``$. However, since $`T`$ is completely positive, this will follow from the fact that the $`n\times n`$ matrix $`B=[B_{ij}]M_n((l^2(\mathrm{\Gamma })))`$, where $`B_{ij}=s_js_i^1`$ defines a positive operator on $``$. This is equivalent to the assertion that
$`{\displaystyle \underset{ij}{}}(s_js_i^1)f_i,f_j=(s_1^1f_1,\mathrm{},s_N^1f_n)^20`$
for all $`f_1,\mathrm{},f_nl^2(\mathrm{\Gamma })`$, which is straightforward.
We now prove (ii). Since $`T`$ has finite rank there exist finitely many $`f_iC_r^{}(\mathrm{\Gamma })^{}`$ and $`S_iUC^{}(\mathrm{\Gamma })`$ such that $`T=f_iS_i`$. Since $`u`$ depends (conjugate) linearly on $`T`$ it is sufficient to consider the rank one case where $`T(s)=f(s)S`$. In this case,
$`|u(s,t)|=|\delta _s,T(st^1)\delta _t|=|f(st^1)||\delta _s,S\delta _t|f_{C_r^{}(\mathrm{\Gamma })^{}}|\delta _s,S\delta _t|,`$
and it suffices show that for all $`\epsilon >0`$ there exists $`R>0`$ such that $`|\delta _s,S\delta _t|<\epsilon `$ provided $`d(s,t)R`$.
At this point the requirement that $`SUC^{}(\mathrm{\Gamma })`$ is needed. Let $`S^{}`$ be a finite width operator such that $`SS^{}<\epsilon `$. Then we have
$`|S\delta _t,\delta _s|SS^{}+|S^{}\delta _t,\delta _s|,`$
and for large enough $`R`$, $`d(s,t)>R`$ forces the last term to be zero. The result follows.
We conclude the proof by verifying (iii). Consider
$`u(s,t)1`$ $`=`$ $`\delta _s,T(st^1)\delta _t\delta _s,\delta _s`$
$`=`$ $`\delta _s,T(st^1)\delta _t\delta _s`$
$`=`$ $`\delta _s,(T(st^1)st^1)\delta _t.`$
Thus, if we have a family $`T_\lambda `$, it follows that
$`|u_\lambda (s,t)1|T_\lambda (st^1)st^1.`$
To verify the uniform convergence on sets of the form $`B_R(\mathrm{\Delta })`$ note that $`d(s,t)<R`$ implies that $`st^1`$ lies in a bounded subset of $`\mathrm{\Gamma }`$, hence only a finite number of such products are possible. Thus, by taking $`\lambda `$ sufficiently large the right side can be made as small as necessary. This completes the proof. ∎
## 4. Approximate units and the Haagerup property
The results of Section 2 can be used to directly relate the existence of an approximate unit of positive definite functions to the Haagerup property for the transformation groupoid $`\beta \mathrm{\Gamma }\mathrm{\Gamma }`$. Here $`\beta \mathrm{\Gamma }`$ is the Stone-Cech compactification of $`\mathrm{\Gamma }`$ and $`\mathrm{\Gamma }`$ acts on $`\beta \mathrm{\Gamma }`$ on the right by extending right translation.
This requires extending the notion of positive definite and negative type functions to groupoids. This has been done by Tu \[12, Section 3.3\] in the following way. We specialize to the case of a transformation groupoid $`X\mathrm{\Gamma }`$, defined as above, where $`X`$ is a compact space on which $`\mathrm{\Gamma }`$ acts on the right. A complex-valued function $`\phi `$ on $`X\mathrm{\Gamma }`$ is positive definite if
$`{\displaystyle \underset{ij}{}}\overline{z}_i\phi (xs_i,s_i^1s_j)z_j0,\text{for all }xX\text{}s_1,\mathrm{},s_n\mathrm{\Gamma }\text{ and }z_i,\mathrm{},z_n.`$
A real-valued function $`\psi `$ on $`X\mathrm{\Gamma }`$ is of negative type if
* $`\psi (x,e)=0`$ for all $`xX`$,
* $`\psi (xs,s^1t)=\psi (xt,t^1s)`$, for all $`xX`$, $`s`$, $`t\mathrm{\Gamma }`$,
* $`_{ij}a_i\psi (xs_i,s_i^1s_j)a_j0`$, for all $`xX`$, $`s_1,\mathrm{},s_n\mathrm{\Gamma }`$ and $`a_1,\mathrm{},a_n`$ satisfying $`_ja_j=0`$.
###### Definition 4.1.
The transformation groupoid $`X\mathrm{\Gamma }`$ has the Haagerup property if there exists a proper, negative type function $`\psi :X\mathrm{\Gamma }`$.
If $`X\mathrm{\Gamma }`$ has the Haagerup property, then it admits a proper affine action on a field of Hilbert spaces, . This latter property, in the case of groups, is call a-T-menability. We may now state the main result of this section.
###### Theorem 4.1.
Let $`\mathrm{\Gamma }`$ be a discrete group. The following are equivalent:
* $`\mathrm{\Gamma }`$ is uniformly embeddable in a Hilbert space.
* The groupoid $`\beta \mathrm{\Gamma }\mathrm{\Gamma }`$ has the Haagerup property.
* $`C_0(\beta \mathrm{\Gamma }\mathrm{\Gamma })`$ admits an approximate unit of positive definite functions.
###### Proof.
There is an equivalence of groupoids, $`\alpha :\mathrm{\Gamma }\mathrm{\Gamma }\mathrm{\Gamma }\times \mathrm{\Gamma }`$ given by $`\alpha (s,t)=(s,st)`$. Here $`\mathrm{\Gamma }\times \mathrm{\Gamma }`$ is the trivial groupoid. The inverse of $`\alpha `$ is $`\beta (s,t)=(s,s^1t)`$. These maps define correspondences, $`\alpha ^{}:l^{\mathrm{}}(\mathrm{\Gamma }\times \mathrm{\Gamma })C_b(\beta \mathrm{\Gamma }\mathrm{\Gamma }):\beta ^{}`$ between functions on $`\mathrm{\Gamma }\times \mathrm{\Gamma }`$ and $`\beta \mathrm{\Gamma }\mathrm{\Gamma }`$ via $`\alpha ^{}(f)(s,t)=f(s,st)`$ and $`\beta ^{}(g)(s,t)=g(s,s^1t)`$. Note that $`\alpha ^{}(f)`$, initially defined on $`\mathrm{\Gamma }\times \mathrm{\Gamma }\beta \mathrm{\Gamma }\mathrm{\Gamma }`$, extends by continuity to $`\beta \mathrm{\Gamma }\mathrm{\Gamma }`$ since $`f\alpha (,t)`$ is bounded for each fixed $`t\mathrm{\Gamma }`$.
It is easy to check that $`\alpha ^{}`$ and $`\beta ^{}`$ are inverses and provide a bijection between $`l^{\mathrm{}}(\mathrm{\Gamma }\times \mathrm{\Gamma })`$ and $`C_b(\beta \mathrm{\Gamma }\mathrm{\Gamma })`$. These maps have the following properties which are direct consequences of the definitions.
* A function $`fl^{\mathrm{}}(\mathrm{\Gamma }\times \mathrm{\Gamma })`$ is metrically proper if and only if $`\alpha ^{}(f)`$ is a proper function on $`\beta \mathrm{\Gamma }\mathrm{\Gamma }`$.
* The map $`\alpha ^{}`$ takes the ideal $`C_0(\mathrm{\Gamma }\times \mathrm{\Gamma };\mathrm{\Delta })`$ to the ideal $`C_0(\beta \mathrm{\Gamma }\mathrm{\Gamma })`$.
* A net $`\{u_\lambda \}`$ is an approximate unit for $`C_0(\mathrm{\Gamma }\times \mathrm{\Gamma };\mathrm{\Delta })`$ if and only if $`\alpha ^{}(f)(u_\lambda )`$ is an approximate unit for $`C_0(\beta \mathrm{\Gamma }\mathrm{\Gamma })`$.
It remains to note that $`\alpha ^{}`$ preserves positive definite and negative type functions. This also follows in a straightforward way from the above formulas.
Now the result follows from Theorem 2.3. ∎
## 5. Remarks
A finitely generated discrete group $`\mathrm{\Gamma }`$ is strongly exact if the inclusion of $`C_r^{}(\mathrm{\Gamma })`$ into $`UC^{}(\mathrm{\Gamma })`$ given by the left regular representation is a nuclear map, that is, if $`\mathrm{\Gamma }`$ satisfies the hypethesis of Theorem 3.1.
1. If a discrete group $`\mathrm{\Gamma }`$ has the property that there is a nuclear embedding of $`C_r^{}(\mathrm{\Gamma })`$ into $`B(l^2(\mathrm{\Gamma }))`$ then the inclusion given by the left regular representation is also nuclear. It is possible that this inclusion is also a nuclear map into $`UC^{}(\mathrm{\Gamma })`$. In other words, it is possible that every exact group is strongly exact. If this is indeed the case then one would deduce that an exact group satisfies the Novikov conjecture.
2. One may consider algebras $`A`$ satisfying
$$C_r^{}(\mathrm{\Gamma })AUC^{}(\mathrm{\Gamma })$$
and impose the requirement that the inclusion of $`C_r^{}(\mathrm{\Gamma })`$ into $`A`$ be a nuclear map; if $`A=UC^{}(\mathrm{\Gamma })`$ then $`\mathrm{\Gamma }`$ is strongly exact, whereas if $`A=C_r^{}(\mathrm{\Gamma })`$ then $`C_r^{}(\mathrm{\Gamma })`$ is nuclear. One obtains a family of conditions interpolating between strong exactness and nuclearity. In the case $`A=C_r^{}(\mathrm{\Gamma })`$ the procedure employed above for constructing a proper negative type function on $`\mathrm{\Gamma }\times \mathrm{\Gamma }`$ actually yields an invariant one, which descends to $`\mathrm{\Gamma }`$ showing that $`\mathrm{\Gamma }`$ has the Haagerup property. This gives an alternate account of the result of Beki, Cherix and Valette . |
no-problem/0001/astro-ph0001498.html | ar5iv | text | # Current Issues
## Introduction
A star lives its life, from its birth out of a loose assembly of molecular gas till the time when it makes its quietus as a white dwarf, neutron star or black hole, fighting gravity. Although it eventually looses the fight (unless it manages to make a Type Ia supernova), it does not concede graciously. Time and again it finds itself transitioning from a metastable equilibium to a state of lower energy on a dynamical timescale. Similar principles govern the evolution of galaxies where there can be runaway formation of massive stars or episodic accretion onto the central, massive black hole. These are “Cosmic Explosions” - the topic of this workshop.
When I was first asked to introduce “Current Issues”, I thought the title curiously apposite because it is ultimately currents \- electrical, weak (charged and neutral) and (with some license) strong - that are responsible for these impulsive releases of energy. Out of the many astrophysical sites that could have been included on the program, the organizers have chosen to concentrate on a few of the most interesting ones, that I shall consider in turn \- the solar corona, accretion disks surrounding young stellar objects, novae, supernovae, “hypernovae” and jets. As I am neither competent nor patient enough to describe these in any detail, I have chosen to list some recent advances in our observational and theoretical understanding in each case and to pose a few questions some of which may already have answers which I hope subsequent speakers will provide. In view of the large range of topics reviewed I cannot hope to give a representative or even a useful bibliography, and so I shall give none and defer to subsequent speakers.
## Solar and Stellar Flares
The combined observations of the YOHKOH, SOHO and TRACE satellites are transforming our view of the solar corona and, consequently, of the surface activity of other stars. In particular they have given us an appreciation of the dynamics of magnetic field lines as they are gently shuffled by underlying convective motions. The whole region above the photosphere is permeated by a magnetic carpet which is re-woven every couple of days. The solar prominences and coronal arches, prominent in X-ray images are just the regions where the plasma happens to be hottest and, contrary to what might have been thought, the magnetic field is weakest. This magnetic activity is intrinsically dissipative and this keeps the corona at million degree temperatures and launches the solar wind.
The quiet solar wind appears to be a simple and quasi-steady flow at least at high latititude (as measured by the Ulysses spacecraft) with poloidal and toroidal magnetic field components declining as $`r^2,r^1`$, respectively. By contrast, the equatorial outflow appears to be dominated by unsteady coronal mass ejections. The equatorial current sheet is naturally unstable and develops its characteristic “ballerina skirt” sector structure. (Perhaps something similar has been observed by Chandra in the Crab Nebula.) Solar physics has much to teach us about accretion disks, where the underlying motions are much faster and, necessarily, supersonic. It should be no surprize that they are often accompanied by hot coronae, that dissipate a large fraction of the gravitational energy release, and powerful outflows.
My list of questions includes:
* What are the true laws of astrophysical MHD? Traditional, global MHD has been based upon analytic solutions of the equations of conservation of mass, momentum and flux under conditions of high symmetry ignoring dissipation. However, real MHD is heavily influenced by the microphysical behavior of current sheets, tiny reconnecting regions, shock fronts etc in much the same way that hydrodynamic flows are beholden to boundary layers. Perhaps there are simple, phenomenological rules which can reconcile these two approaches.
* What is a solar flare? We know of many examples of magnetostatic configurations that can be slowly altered until they become unstable and release a large amount of magnetic energy. However we do not understand which of these are most likely to occur in practice and what is the partition of the release of energy between local heating and the bulk kinetic energy that drives outgoing shock waves. (Similar questions exist in earthquake studies.)
* What is the structure of shock fronts? Simulation and in situ measurement has greatly improved our understanding of collisionless shocks. It appears that thermal electrons are commonly transmitted with sub-equipartition energies, as is also found to be the case with supernova shock waves. The detailed plasma physics still eludes us, though. This issue is related to the question of the injection of suprathermal ions into the first order Fermi acceleration process that appears to be responsible for producing most Galactic cosmic rays.
* What determines the energy and length scales that dominate coronal heating? The form of this dissipation appears to be primarily reconnection and to be dominated by frequent “nanoflares”, (although this conclusion is still controversial). This realization has, in turn, stimulated analysis of new modes of magnetic reconnection. Still, it is the occasional giant flare that commands our observational attention and provides the most detailed diagnostics.
* How is the solar wind launched? The observed coronal temperature is insufficient to give the gas its 700-800 km s<sup>-1</sup> outflow speed as measured by Ulysses. This leaves hydromagnetic wave acceleration as the prime suspect. Understanding the acceleration and stability of the solar wind is highly relevant to the study of jets.
## Young Stellar Objects
Accretion disks and bipolar outflows appear to be a standard feature of star formation. The optical jets can propagate through the interstellar medium over distances more than $`10`$ pc, quickly polluting it with magnetic field and metals (an observation of some cosmological importance). However, this outflow is not steady. In particular, thermal instability of the accretion disk produces “FU Ori” outbursts where perhaps $`0.01`$ M of gas are expelled with comparatively high speed over a decade or so every ten thousand years. Smaller scale explosions create the “Herbig-Haro” objects which are presumably traveling forward-reverse shocks. These are sometimes observed in matched pairs, one in each jet confirming that they originate at the disk. The morphological similarity to, for example, the knots in the M87 jet is clear.
Some important questions include:
* How much of the mass in the original protostellar disk accretes onto the central protostar and how much is lost in the form of a wind or jet? A related question is how much of the angular momentum is removed in this manner as opposed to being transported radially in the disk to large radii where it can, supposedly be extracted by large tidal torques.
* How much mass and energy is associated with the FUOR outbursts and how much is transported in the long intervals between outbursts? (It is not clear how far one can push the dynamical analogy but similar questions have been raised in trying to understand the Galactic microquasar GRS 1915+105.)
* How are the optical jets collimated and confined? Magnetic collimation is commonly invoked, but even here, several alternatives have been discussed. The field may be primarily vertical near the disk and shape the outflow through magnetic pressure. Alternatively, the field may have a significant radial component so that the jet can be launched centrifugally so that the hoop stress associated with toroidal field may be largely responsible for the collimation. A third possiblity, that has been discussed, is that the magnetic field be mostly toroidal near the disk and wound up like a coiled spring so that it can push the gas away vertically. Observations of protostellar outflows have as good a chance as those of any jets of measuring the magnetic structure. Ultimately the flow must be confined laterally at large cylindrical radius. It is not clear whether this is achieved by the ram pressure of infalling gas or through the application of a quasi-static thermal pressure.
* What is the dynamical structure of Herbig-Haro objects and what can they tell us about the explosions that cause them? Forward-reverse shocks arise naturally if the velocity with which the jet is launched varies so that the faster moving gas overtakes the slower outflow and forms a shock. The pressure behind this shock front may be sufficient to form a reverse shock giving a characteristic dynamical structure. The high pressure inter-shock gas will expand transversely, weakening the shock strengths. Again, observations should help us to understand what is really happening.
## Novae
Classical novae, by contrast, are thermonuclear explosions which arise when hydrogen-rich gas from a companion accumulates on the surface of a C-O or O-Ne-Mg white dwarf and then detonates, initially uncontrollably, under degenerate conditions. The energy release per nucleon is enough to heat the gas above the Fermi temperature, causing it to expand, and then above the escape energy. As several of the nuclear reactions involve weak interactions that take place on timescales that are long compared with the dynamical timescales, the ejected gas is believed to contain many prominent radioactive species that can act as monitors of stellar activity.
X-ray novae involve similar processes occuring on the surface of a neutron star. Here, the reactions occur much faster but the gravitational potential well is so deep that the gas cannot escape using its own thermal pressure. (It may be expelled by radiation pressure, however.) Naturally, this burning will not occur uniformly over the surface of the star and rotational modulation of the X-ray emission was predicted and is observed. (Curiously, the rotational frequency is observed to vary slightly, which may be due to elevation of the X-ray photosphere by radiation pressure with approximate conservation of angular momentum.)
My personal question list for novae is:
* What can be learned by observing radioactive nuclei and positron annihilation from classical novae? Novae are prime targets for missions like INTEGRAL and HESSI that promise to open up the new field of MeV spectroscopy. We need to go beyond mere detection of radioactive nuclei and use measurements of line strengths and widths to learn about the underlying explosion.
* What is the status of the beat-frequency model of QPOs? This posited that the neutron star was rotating with a period similar to that of the inner regions of the accretion disk and that the observed, varying frequencies were a beat rather than a fundamental. The first part of this hypothesis has been vindicated, but I wonder about the evidence for the second part?
* Are any QPO modes due to neutron star osciallations? The problem here is that some of the modes that had been attributed to neutron star oscillations are also found in black hole systems. (Neutron star modes can only provide the clock because the energies associated with them are necessarily quite small.)
* Can we measure the neutron star mass-radius relation? One of the best ways for high energy astrophysics to repay its immense debt to nuclear physics is to measure the equation of state of cold nuclear matter (in contrast to the hot nuclear matter that will be explored by heavy ion colliders). This may be possible through measuring the gravitational redshift of atomic and nuclear lines from the surface of neutron stars, however it is not clear what it will take to do this in practice.
## Supernovae
Supernovae are once again at center stage. In cosmology, Type Ia explosions have been modeled empirically as one parameter standard candles, and if this is the case, they suggest that the universe is entering a (second?) epoch of inflationary expansion. This is a remarkable discovery, if true. In addition, there is circumstantial evidence that at least some types of $`\gamma `$-ray bursts are associated with supernovae, both through the suggested identification of GRBs with star forming regions and the possible discovery of supernova light curves in a few instances. For both lines of research to advance, it is imperative to develop a far better understanding of the physics and the astrophysics of supernova explosions.
The blast waves that result from these explosions are not always well-described by Sedov point explosions in uniform media. Even if the energy release were fairly isotropic (and there are several reasons for suspecting that it is not) the external medium is likely to be anisotropic. The beautiful images of $`\eta `$ Car and SN 1987a, the former being an accident waiting to happen and the latter being one that we are still witnessing, explain why so many mature supernova remnants are quite non-circular despite having essentially isobaric interiors. These supernova remnants are excellent laboratories for studying particle acceleration and magnetic field amplification at shock fronts that provide a bridge between heliospheric studies and more energetic phenomena associated with AGN and GRBs. Non-relativistic shocks behave quite differently from relativistic shocks and so it is fortunate that we have plerions like the Crab Nebula and classic remnants like Tycho so close to home to study.
The questions:
* How important are Type Ia supernova evolutionary corrections? The big concern, as always with cosmographic studies of the expansion of the universe, is whether or not we are confusing kinematics with physical evolution. This is particularly troubling here because there is no commonly agreed identity for the progenitors of these explosions and, I believe, no consensus yet on the reason for the “Phillips” correction although some promising suggestions have been made. There are internal consistency checks and some of these have already been satisfied but more will be needed before we can sign off on the result.
* How do we classify supernovae observationally? I doubt that I am alone in not understanding the spectroscopic and physical distinctions between the various types of supernova Type Ibc, Type IIn etc. I hope we can have a primer on the subject here.
* What are Type Ia supernovae anyway? Single degenerate and double degenerate models have their advocates. Likewise for detonation of Chandrasekhar mass CO white dwarf versus an off center explosion in a lighter star with a helium envelope.
* When do Type II supernovae form black holes as opposed to neutron stars and what are the associated rates? This question is timely because Chandra has just discovered a point source inside Cas A, and as of now, the odds are about evens for it being a black hole or a neutron star.
## Hypernovae
Gamma ray bursts continue to amaze. There has been direct verification that the long duration bursts are located at cosmological distances through the measurement of redshifts. (The same is probably true for the short duration bursts, although HETE2 is probably going to be necessary to verify this.) This leads to an impressively broad range of isotropic burst energies, from $`10^6`$ M$`{}_{}{}^{}c_{}^{2}`$ in the case of GRB 980425 to $`2`$ M$`{}_{}{}^{}c_{}^{2}`$ for GRB 990123 - hardly standard candles (though this has not prevented some from trying to use them for cosmography). GRBs are now widely interpreted as optically thick fireballs created with large entropies per baryon, like the universe itself. The actual $`\gamma `$-ray emission, lasting for up to a few minutes, is commonly thought to be produced by internal shocks in the expanding ejecta and this accounts for the great heterogeneity in observed $`\gamma `$-ray burst time profiles. The ninth magnitude optical burst, seen by ROTSE from GRB 990123 (with an isotropic energy $`10^3`$ of the total) may be caused by a reverse shock. Studying the afterglows is proving to be interesting in its own right, for what it has to say about the behavior of relativistic shocks, as a probe of the environment in which the burst occurs, as a measure of the explosion energy and as an indicator of beaming. Broken power law spectra are observed and these have been variously interpreted as being due to cut-offs in the electron distribution function, radiative cooling and self-absorption.
A recent development is the circumstantial evidence for the association of GRBs 970228, 980326, 980425 with supernovae, albeit of different types. If the association is also with young stars, then GRBs will be invaluable probes of the early universe and galaxy formation. Another somewhat more secure story is that soft $`\gamma `$-ray repeaters are “magnetars”. That is to say, their outbursts are magnetically powered and originate on the surfaces of young neutron stars with surface magnetic fields $`B10^{14}`$ G.
It is hard to limit the number of questions in this subject.
* Are GRBs beamed? In my view, although the jet hypothesis is eminently reasonable and fits in with some source models, especially collapsars, we are really only interpreting occasional steepening in the light curves in this manner, rather than seeing the clear evidence that was provided by VLBI in the case of AGN. The argument that the bursts must be beamed, otherwise they would have energies in excess of a stellar rest mass, reminds me of a similar argument in favor of them being local!
* Are there $`\gamma `$-ray quiet afterglows? These are surely a prediction of beaming. At present the observational constraints are surprisingly poor.
* Is magnetic field amplified at external shocks? We know from observations of young supernova remnants that relativistic protons and electrons are accelerated at non-relativistic shock fronts and that thermal electrons are transmitted with temperatures below the equipartition value. However, in a source like Cas A, it appears that the magnetic field only becomes strong in the interaction zone between the shocked interstellar medium and the explosion debris. (It is noteworthy that even as impressive a radio source as Cas A is four orders of magnitude under-luminous relative to a homogenous, maximally emitting synchrotron source with the same total pressure.) This can make a big difference. When Chris McKee and I computed the nonthermal emission that would be observed from decelarating, relativistic blast waves, we assumed that the magnetic field is just compressed along with the gas in passing through the relativistic shock front. In this case, $`ϵ_{\mathrm{mag}}`$, the ratio of the magnetic to total energy density is only $`2v_A^2/c^210^9`$ in the interstellar medium, where $`v_A`$ is the Alfvén speed ahead of the shock front. More recent calculations, that are applied specifically to GRB afterglows, generally assume that $`ϵ_{\mathrm{mag}}10^2`$ which is necessary to fit the fluxes (although the scaling laws are unchanged). For this reason and because the best studied GRB afterglow, GRB 970508, shows no sign of a mildly relativistic transition in the particle acceleration efficiency, I still suspect that the afterglow emission originates well downstream from the outer shock.
* Are GRBs really associated with supernovae? The late time light curve seen in GRB 980326 could have a different explanation. In particular, as Ann Esin and I have been considering, it fits rather well with scatttering of the initial optical burst by dust just outside the sublimation radius. This occurs typically at a distance $`10^{18}`$ cm. As refractory dust has high albedo and is forward scattering, the characteristic delay is plausibly a few months, as observed.
* How easy is it to have an ultrarelativistic jet emerge from inside a collapsing star? Entrainment of gas might easily occur and prevent the flow from attaining bulk Lorentz factors $`300`$.
* Are the “cyclotron” lines real and the $`300`$ keV “breaks” generic and, if so, can they be formed in ultrarelativistic outflows? Existing explanations seem a little contrived.
* What is the underlying physical mechanism for creating the fireball? Most models now seem to involve black holes, magnetic field and wishful thinking. The problem is hard. The main challenge is to amplify the magnetic field fast enough to make an electromagnetic bomb. Fortunately there are new ingredients in the strongly curved spacetime around a black hole or a pair of orbiting neutron stars. The orbits can precess differentially at near relativistic speed and this can lead to an extremely rapid field growth - faster than conventional dynamos and, indeed, faster than exponential. This, in turn, induces $`10^{22}`$ V EMFs which cannot be shorted out and accelerate pairs directly.
## Jets
The black hole model of AGN has been vindicated observationally, and, beyond all reasonable doubt, most normal galaxies, like our own contain central black holes with masses in the million to billion solar mass range. There are also at least nine well measured compact object masses in excess of $`2.5`$ M that are surely also black holes. There are promising, but so far, less compelling indications that some of these holes spin rapidly.
Jet are a common, though not universal, accompaniment of accretion which suggests that they are involved in carrying off some of the energy and angular momentum released by the infalling gas. However, we do not understand how they are formed or even if there is a universal mechanism at work. We do know that magnetic field must grow to dynamically significant levels in accretion disks and most jet formation models now involve magnetic field except perhaps when the mass accretion rate greatly exceeds the Eddington rate.
Our understanding of the emission from jets has also advanced. The discovery of rapidly variable GeV and TeV $`\gamma `$-rays from blazars shows that they can be extremely luminous. Radio astronomers have been mapping the smoke not the fire. The standard radio synchrotron model is also under assault. There is increasing evidence that compact components have brightness temperatures well in excess of the inverse Compton limit, even allowing for plausible Lorentz factors. Large degrees of circular polarization are also being reported. All of this suggests that some alternative, possibly coherent emission mechanism is at work at least in the compact cores. (Note, that it is not sufficient to explain how high brightness radio emission is emitted. It is also necessary to explain how it is transmitted out of the nucleus when there are many potential non-linear scattering mechanisms which will degrade the brightness temperature.)
In an impressive display of the power of VLBI, the radio astronomers have been able to show that the M87 jet is collimated within $`100m`$. If relativistic jets are powered by the black hole itself or the gas flow around the black hole then their energy has to be carried initially in some form other than electron-positron pairs, which are subject to catastrophic radiative losses. Electromagnetic Poynting flux is the prime suspect. In other words, relativistic jets are starting to resemble pulsars. (Contrariwise, the Crab pulsar now appears to form a pair of “jets”).
My final list of questions is:
* Are AGN jets hyper-relativistic? Radio jets exhibit bulk Lorentz factors $`\mathrm{\Gamma }10`$; $`\gamma `$-ray burst models have taken us over the psychological hurdle to $`\mathrm{\Gamma }300`$ which might just account for the reported radio variability of jets under the synchrotron model were it not to imply steady $`\gamma `$-ray burst level powers in AGN jets. Hence the appeal to coherent processes. Before we solve this problem, though, we must identify the jet working substance and if, and where, Poynting flux is transformed to plasma. (This last is still an interesting question in the case of the Crab pulsar wind.)
* Are jets better approximated as episodic or steady? Traditionally, we have modeled jets as stationary flows upon which have been imprinted perturbative disturbances which form shock fronts - the emitting elements. However, GRS 1915+105 suggest a quite different model - jets as a sequence of small explosions, perhaps associated with intermittent flow in the accretion disk, that expand into and keep open an evacuated channel. (YSO jets offer support to both views.)
* How are jets collimated? Ordered and disordered, poloidal and toroidal field have all been proposed for launching and collimating AGN jets from disks, just as with the YSO jets. 3D MHD global simulations are becoming increasingly ambitious and ever more relevant.
* Are relativistic jets powered by the spin energy of the hole or the binding energy of the accreting gas? The former seems more likely to form ultrarelativistic outflows; the latter may, on average, release more power. General relativistic numerical simulations are starting to guide our intuition.
* Why are there no gamma-ray megabursts? GRBs are thought to be associated with the birth of stellar black holes and produce powers of up to $`10^7c^5/G`$ for $`10^6m`$. If massive black holes are formed with masses $`10^6`$ M at a rate of several per year, we might expect to see megabursts with similar powers but lasting for months. We don’t. Perhaps, instead, massive black holes grow from much smaller holes, which, themselves, might be relics of the first generation of stars which may have masses $`10^310^4`$ M at $`z30`$.
* How much do AGN contribute to the luminosity density of the universe? This is a closely related question. The measurement of the far infared background, the discovery of hard X-ray emission from some Seyfert galaxies and the spectrum of the X-ray background all point to AGN power being a significant fraction of the stellar luminosity density. If so, then there are probably implications for galaxy formation and development. For example, elliptical galaxies may result when a black hole grows rapidly and early in the life of the galaxy so that it is capable of blowing away late infalling gas before it can form a disk.
## Connections
As I hope this brief introduction has brought out, there are strong inter-connections between our studies of these different types of cosmic explosion. Accretion disk coronae can look quite like their solar counterpart. Novae have some dynamical similarities to miniature supernovae whose remnants, in turn, behave quite like aging $`\gamma `$-ray burst afterglows. Similar electromagnetic processes are at work around pulsars and black holes. $`\gamma `$-ray bursts themselves have some similarities, at least radiatively, with the early universe. And so on.
## Acknowledgements
I acknowledge support under NASA grant 5-2837 and NSF grant AST 99-00866 and Re’em Sari for comments. |
no-problem/0001/astro-ph0001540.html | ar5iv | text | # The Number of Publications Used as a Metric of the NOAO WIYN Queue Experiment
## 1 Introduction
The 3.5-m Wisconsin-Indiana-Yale-NOAO (WIYN) telescope was dedicated on 1994 October 15, and shared-risk observing began in 1995 March. NOAO’s share of the time is 40%, and nearly all of this has been carried out in “queue” mode, where the observations from highly ranked proposals are placed in a queue and executed during nights assigned to the queue program. The observations are carried out by highly experienced professionals, who are extremely familiar with the instrumentation, without the direct assistance of the proposing astronomer. A small fraction of the NOAO time is scheduled out in “classical mode”, with the observers present at the telescope. The time allocated to the university consortium members (roughly 60%) is all carried out in classical mode.
The goal of the NOAO WIYN queue experiment was eloquently described by Silva & De Young (1996) as an empirical test of “the hypothesis that in the face of a high over-subscription rate, the science throughput of WIYN can be maximized by executing the most highly ranked science programs first, completing datasets in a timely manner, allowing a larger range of program lengths, and matching the observing program to the observing conditions on an observation-by-observation basis.”
The WIYN queue has often been described as an “experiment” at least in part because other observatories are considering scheduling some or all of their time in this mode, and NOAO staff have felt that what we can learn from the WIYN queue will be useful to others. In an era that sees both the proliferation of very large ($``$8 m) telescopes, but ever-tightening financial resources, observatories are scrambling to understand how to maximize their scientific return.
Queue observing offers a variety of theoretical advantages, as nicely summarized by Mountain (1996) and Boroson et al. (1998). For very highly ranked programs that require rare conditions, queue observing may be the only practical way to acquire such data. Queue observing naturally allows synoptic observations, and such scheduling easily accommodates target-of-opportunity requests, such as optical follow-ups of gamma-ray bursts or supernovae. Furthermore, as instrumentation becomes more complex, queue observing carried out by dedicated observers may result in more efficient use of telescope time than if the observations were carried out by visitors who uses the equipment only occasionally. This contention is partially supported by evidence that observers collect less data on the first night of an observing run than on subsequent nights (Bohannan 1998).
However, there are obvious down-sides to the queue mode. The astronomer is not present at the telescope, and therefore cannot make real-time decisions concerning the data. Serendipity is eliminated, as are the risky programs many of us have snuck in during gaps in our main observing program, and which have sometimes led to the more interesting results. Some of us suspect, rightly or wrongly, that we could better carry out our own observations. And, there is not the same strong sense of “data ownership” that comes with having carried out the observations ourselves: the memory of a night may provide details that are relevant to the interpretation of the reduced data, as well as providing an emotional impetus for seeing the project through to its completion.
There is also a non-negligible expense of running a queue, which is off-set to some degree by the smaller support required for visiting astronomers.
Boroson (1996) has described a simulation program that can be used to test how successfully programs are completed in a queue mode vs. a classical mode, using Monte Carlo sampling of characteristic observing conditions (weather, seeing) for the site. Boroson et al. (1998) used this simulation program comparing queue mode and classical scheduling for two actual semesters (1997) of WIYN programs, concluding that queue scheduling at WIYN has led to a significant gain in efficiency and scientific effectiveness.
Now that the queue experiment has run for several years, we thought it would be worth examining the gain using some real-world measure. As emphasized by Boroson et al. (1998), much of the argument about observing modes can be emotional. We seek some metric that we can use to test the hypothesis enunciated above that the queue observing mode leads to significant improvement in the science throughput. One such simple metric is the number of refereed papers published. This may not be as meaningful in its long-term impact on astronomy as, say, the number of important new discoveries, but at least it has the advantage of being quantifiable, and, if the experimental and control samples are well matched, equitable and fair.
We choose to compare the number of papers produced by the WIYN queue to the following two controls, each with its advantages and disadvantages:
1. The number of papers produced by observations made over the same time period with the Mayall 4-m telescope.
2. The number of papers produced by observations made over the same time period by non-queue use of WIYN; i.e., primarily the time used by the consortium universities.
The first comparison has the primary advantage that both the 4-m and WIYN proposals have undergone similar scrutiny by the same time allocation committees (TACs), which often consider such factors as the past track-record of the proposers as well as the scientific excellence of the proposals. Thus proposers to the 4-m and WIYN will feel similar pressures to publish in a timely manner, and the feasibility of the proposals has been carefully evaluated. Users of the university time may choose to undertake longer-term projects, leading ultimately to more important results, but not processing the same rapid turn around from observing to publication. We offer the second comparison as there may be differences in the actual on-sky performance of the two telescopes that would affect the results: the 4-m is a mature telescope, possibly with fewer teething problems, than the newer WIYN.
If the queue leads to significantly higher scientific throughput, then we expect that the number of papers published using data obtained via the queue should be significantly greater than those produced by the control samples, after normalization on the basis of the number of scheduled nights.
## 2 The Data Set
All of the 1998 and 1999 issues of the main US astronomy journals were examined for papers which used 4-m and/or WIYN observations. The complete list of 135 papers is given in Table A1 of the Appendix.
In order to make a fair comparison, we restricted ourselves to only those papers for which the data were obtained in semester “1996B” or later (i.e., after 1996 August 1). This was the third full semester of WIYN queue time, and the first semester in which both the imager and fiber positioner were fully functional. (A non-linearity problem with the S2KB imager chip was discovered and fixed during the 1996A semester, and a mechanical problem which compromised the positioning accuracy of the Hydra fiber positioner was fixed in 1996 March.)
We list in Table 1 the number of papers published during 1998 and 1999. Six papers used both 4-m and WIYN data; we chose to count each of these papers separately for both telescopes, depending upon the date in which the data were obtained for the telescope under consideration; i.e., if the data for WIYN was obtained in 1996B or later, but the 4-m data was obtained prior to 1996, it would count as a WIYN publication but not as a 4-m paper. There were six papers in our list in which the data collected were such a minor component of the paper that we chose not to count the paper at all; only one of these used the WIYN queue, and in that case the data had been published previously by the original proposers.
## 3 Results
### 3.1 Comparison of the WIYN and the 4-m
In order to make a valid comparison, we must first take into account that not as much time is scheduled for the WIYN queue as for the 4-m. We expect the answer is about 40%, as NOAO receives 40% of the time on WIYN, and almost all of this goes to the queue. However, the 4-m is shut down during July and August, while WIYN continues to operate; on the other hand, there are more engineering nights scheduled at WIYN. One could use the total number of clear hours spent observing as the normalization, but these data are hard to extract reliably. Instead, we took the final observing schedules for semesters 1996B, 1997A, 1997B, 1998A, and 1998B, and simply counted the number of nights assigned to the WIYN queue, and to science operations at the 4-m. (For the latter, we included half-night instrument “checkout” nights, as much of this time is typically returned to the observers scheduled on the second half; full-night “check” nights and engineering nights were excluded. We excluded all engineering nights scheduled at WIYN, although occasionally queue observations are obtained during such time.) The numbers of nights so scheduled for the WIYN queue and for the 4-m are 260 and 656 respectively; i.e., the number of nights scheduled to the WIYN queue turned out to be 39.6% of the nights scheduled at the 4-m.
If the hypothesis described above is correct, we would expect the number of publications based upon WIYN queue data to be significantly greater than 40% of those produced by the 4-m. Instead, we find in Table 1 that there were only 9 papers produced by WIYN queue data as opposed to 34 papers produced by the 4-m; i.e., 26%. Thus there are actually 1.5 times fewer papers published (on a per night basis) based on queue WIYN data relative to those based on 4-m data. This comparison does not support the hypothesis of greater science throughput by the WIYN queue.
Can we rule out the hypothesis given the small number statistics? If we assume the simplest model that a 1$`\sigma `$ uncertainty in the number of publications $`N`$ is simply the $`\sqrt{N}`$, then the 1$`\sigma `$ error on the 0.26 ratio of WIYN to 4-m publications is 0.13. What does it mean for there to be a “significant” enhancement in the scientific throughput? Boroson et al. (1998) discuss how their simulation predicts this will depend upon program type, TAC grade, and so on, and that overall about 2.5 times as many programs will be completed by queue observing than with classical observing. We take here a more conservative approach: certainly a 50% increase (a factor of 1.5) would be cause for celebration. Were this enhancement present, we would expect there to be 1.5 $`\times `$ 39.6% = 59.4% as many WIYN queue papers as 4-m papers. We observe 0.26$`\pm `$0.13 We thus can reject such an increase at a +2.5$`\sigma `$ level; i.e., with a 99.3% confidence.<sup>1</sup><sup>1</sup>1The rejection probability corresponding to +2.5$`\sigma `$ was found by
$$1.00.5\times (1.0A_G(x\mu /\sigma )),$$
where $`A_G`$ is the integral probability of the normal distribution with a mean of $`\mu `$ and a standard deviation of $`\sigma `$; see, for example, Fig. C-2 in Bevington (1969).
### 3.2 Comparison of Queue vs. Non-Queue Time at WIYN
Of the 731 nights scheduled for science at WIYN during 1996B through 1998B, we find that 260 nights were scheduled for queue observations (35.6%), 27 nights were scheduled for NOAO classical observations (3.7%), and 444 as university time (60.7%). If queue observing produced a significantly higher scientific throughput, we would expect significantly more than 36% of the papers produced by WIYN data to be based on data obtained with the queue. Instead, of the 28 total WIYN papers in our sample, 9 (32%) were produced from queue data. This is essentially the same fraction of time on WIYN used by the queue (36%), and therefore does not suggest that queue provides a significant advantage.
While the data fail to offer any support for the hypothesis, at what level can we reject the claim, given our limited statistics? Using the same argument as above that we would hope for a factor of 1.5 enhancement over the non-queue publication rate, we can ask at what level can we exclude the queue publications amounting to 1.5$`\times `$ 35.6% = 53.4% of the total. The uncertainty in our ratio 0.32 ratio is 0.17. Thus we can exclude a 50% enhancement at the +1.3$`\sigma `$ level; i.e., with an 89.8% confidence.
Nevertheless, it is clear that queue observing does fare better in this comparison than it did in comparison to the 4-m control, although still failing to produce a higher number of publications. Several explanations come to mind. One possibility is that the 4-m simply operates more efficiently than WIYN (at least in the time period when most of the data were acquired), and that it was thus easier to obtain usable data at the 4-m. It is possible that review of queue proposals by an outside TAC leads to a higher publication rate than time used by the universities, who have a preallocated amount of time, which is divided up internally. (As suggested earlier, the university time may be spent on longer-term programs than the NOAO portion.) Finally, the 4-m supports a wider complement of instrumentation (such as infrared imaging and spectroscopy) than WIYN, which plausibly provides greater coverage of astronomical disciplines and thus involvement in a wider variety of publications.
Although the numbers are small, the very high publication rate for NOAO time that is scheduled classically at WIYN suggests that it may be the TAC process rather than the telescope or instrumentation which explains why the queue does better in this comparison than it does in comparison to the 4-m: 14% of the WIYN papers were produced by the small (3.7%) time allocated to non-university classical observing. The classically scheduled NOAO time undergoes the same rigorous review as the queue proposals, and thus is under the same pressure to publish rapidly.
## 4 Discussion
Arguably, the WIYN queue has been as well run as it is possible for any queue to be. A survey carried out of astronomers who had proposed for queue time suggests that people were very satisfied with the quality of the data they received (Boroson et al. 1998); some might expect that maintaining data quality to be the hardest part of a queue. Yet the evidence so far fails to support the suggestion that queue observing leads to a higher scientific throughput, at least as measured by the number of publications. Why does this differ from the dramatic predictions of simulations that suggest that a much higher percentage of programs should be completed by the queue mode?
We have read through the papers based upon the WIYN queue data and have several observations of our own to offer. First, let us consider the advantage that queue offers in providing easy “target of opportunity” (TOO) observations. Of the full set of 11 papers (ignoring the 1996B cutoff), four rely on the TOO advantage of queue for optical followup of gamma-ray bursts (Galama et al. 1998) or supernovae (Jha et al. 1999; Perlmutter et al. 1999; and Riess et al. 1998). Although WIYN played a role in these important studies, our examination of these papers suggests that it was a relatively minor role, with the majority of the data coming from elsewhere. For instance, there are considerably more data from the CTIO 4-m (which is classically scheduled) than from WIYN in the Riess et al. (1998) study. Inspection of these papers suggest that there is no lack of ways for large groups to acquire such data. The number of authors on these four papers range from 17 to 42, and with a large number of participants being a reflection of the degree (and method?) of telescope access. Thus TOO use of WIYN may not be more significant simply because there are other ways of obtaining such data.
One of the other purported advantages for queue observing is the ability to take advantage of particularly good conditions, and indeed some programs may not be completed any other way. However, this advantage is larger the greater the range of conditions. For instance, if the frequency histogram of delivered image quality (DIQ) is very sharply peaked, then queue offers less of an advantage, as all programs will obtain something like the median seeing. At WIYN the median DIQ (at R) is 0.8 arcseconds, and 0.6 arcsecond or better images are achieved 18% of the time (Green 1999). Of the 11 queue papers listed in Table A1, Armandroff, Jacoby, & Davies (1999) is one of the clearest examples of taking advantage of the queue to obtain the best DIQ. The study utilized sub-arcsecond conditions (0.8 arcsec at B, 0.6 arcsec at V, and 0.7 arcsec at I) for deep imaging of a newly discovered dwarf member of the Local Group, Andromeda VI, after confirming its nature using imaging at the 4-m. Nevertheless, these DIQ values are not all that different than the median values.
However, it may be that the sociological issues raised in the introduction dominate. The use of queue may reduce the sense of “data ownership,” and given situations of “data saturation,” we are more likely to publish the data more rapidly if we have acquired them ourselves. The use of “queue mode” on HST has been perceived as being highly successful, although a meaningful control sample is hard to find for comparison; however, one important difference comes to mind, namely that observing time (to US proposers) usually comes with grants, providing a financial incentive to produce results rapidly, coupled with a 1-year proprietary period for unique data. An additional consideration is that HST supplies the user with fully reduced data, unlike WIYN, which provides basic calibration data and requested standard observations, but which does not attempt a “pipe-line” reduction. However, our own experience with HST data is that customized reductions are often needed in order to provide the data most meaningful for a particular application.
Finally, it may be that we simply have not been sufficiently patient. As is evident from the 4-m publications, only one-third of the 4-m papers in the past two years relied purely on “new” data (i.e., all data obtained in the past 3.5 years). While our control samples explicitly took this into account, we are nevertheless comparing numbers that are on the the tails of the distribution of how quickly data finds its way into the literature. This may be particularly true if the datasets from the WIYN queue were to be larger than that in the control samples, or if they take longer to reduce. Current plans call for discontinuing the WIYN queue at the end of semester 2000A, but continuing to provide some synoptic and target of opportunity service observing beyond that. It will be interesting to re-examine the literature five years from now using data obtained in 1996B-2000A as the selection criterion.
We note that the quantity we would most like to measure is “quality”, but this is of course harder to do in an objective manner. Citation rates might provide one means, but not enough time has past for these to be meaningful. Counting the number of papers is some measure of the “output” of a telescope, but it is not necessarily the best; it does have the advantage of being objective and reproducible, qualities usually assumed to be desirable in any experiment.
Nevertheless, our results suggest that it may benefit observatories to evaluate their queue programs using some external measure, such as the number of publications, if suitable controls can be defined.
Helmut Abt, Dave De Young, David Sawyer, Dave Silva, and Sidney C. Wolff were kind enough to provide thoughtful comments on the manuscript. We also benefited conversations with Taft Armandroff, Bruce Bohannan, and Abi Saha on the issues of queue observing.
## 5 Appendix
In Table A1 we present the list of papers published in the Astronomical Journal, the Astrophysical Journal (Parts 1 and 2), and the Publications of the Astronomical Society of the Pacific during 1998 and 1999 that used data from the 4-m and/or WIYN. We list the dates of the first data obtained (from the relevant telescope). Often this information was directly obtained from the paper, but in many cases we had to contact the authors, or inspect the observing schedule or list of queue programs to determine the actual data or semester. |
no-problem/0001/quant-ph0001057.html | ar5iv | text | # Discussion: Byrne and Hall on Everett and Chalmers
## Abstract
Byrne and Hall (1999) criticized the argument of Chalmers (1996) in favor of the Everett-style interpretation. They claimed to show “the deep and underappreciated flaw in any Everett-style interpretation”. I will argue that it is possible to interpret Chalmers’s writing in such a way that most of the criticism by Byrne and Hall does not apply. In any case their general criticism of the many-worlds interpretation is unfounded. The recent recognition that the Everett-style interpretations are good (if not the best) interpretations of quantum mechanics has, therefore, not been negated.
School of Physics and Astronomy
Raymond and Beverly Sackler Faculty of Exact Sciences
Tel Aviv University, Tel-Aviv 69978, Israel.
1. Introduction. It is probably impossible to present an interpretation of quantum mechanics in unambiguous way without writing equations. Chalmers’s presentation of Everett-style interpretation also can be understood in different ways. Instead of equations Chalmers used some technical jargon of quantum theory, however, some words like “substates” have no clear meaning even for physicists. Byrne and Hall (BH) interpreted Chalmers’s jargon in a way which leads to contradictions. In this note I will argue that by taking a more positive approach, one can see in Chalmers’s writing a consistent (although not necessarily very persuasive) argument.
In the second part of their paper BH claimed to show not only that Chalmers has failed to establish his Everett-inspired interpretation, but that “anything resembling it should not be taken seriously”. Their first point is of a general character: if the spaces of states in two theories are identical but the dynamics is not, it is not obvious that the interpretation of these states in the two theories must be identical too. BH point out that this is the situation regarding the interpretation of quantum states in the orthodox and the Everett interpretations. I will argue that although their general argument is correct, its application is not. There is enough similarity between the dynamics that makes the identification plausible. The second point of BH is that the Everett-style interpretation has less “substantive content” than the orthodox interpretation. This is because in the Everett (many-worlds) interpretation there is no counterpart of “outcome probabilities”, the concept of the orthodox interpretation associated with a system in a superposition of eigenstates of some variable. I will argue that the definition of the probability of an outcome in the framework of the many-worlds interpretation which I recently proposed solves this difficulty and makes this BH criticism obsolete.
The organization of this note is as follows. In Section 2 I will adopt the BH interpretation of Chalmers and will show (in a different from BH way) how it leads to a contradiction. In Section 3 I propose an alternative interpretation of Chalmers’s writing which leads to a consistent argument. In Section 4 I critically analyze the general arguments of BH against the Everett-style interpretations. Finally, in Section 5 I summarize my defense of the many-worlds interpretation.
2. Byrne and Hall interpretation and a contradiction in the Chalmers argument. The central thesis of Chalmers quoted by BH is the principle of organizational preservation under superposition:
> OPUS“If a computation is implemented by a system in a maximal physical state $`P`$, it is also implemented by a system in a superposition of $`P`$ with orthogonal physical states.”(Chalmers, 350)
Consider a simple model: a computer which performs calculations in a classical way. If at time $`t_0`$ the computer receives a classical input (a particular punching of its keyboard), then it evolves in time is such a way that it is always in a “classical” state. This means that all the registers of the computer at all times are in some definite states (exited or not exited) i.e., not in a superposition of excited and not excited. Suppose that $`P`$ corresponds to a computation of a square of a number 5, while $`Q`$ corresponds to a computation of a square of a number 10. Denote $`|P(t)`$ a quantum state of the computer at time $`t`$ performing the calculation of the square of 5, while $`|Q(t)`$ a quantum state of the computer at time $`t`$ performing the calculation of the square of 10. In the two computations at any time the registers must be in different states, therefore, $`|P(t)`$ is orthogonal to $`|Q(t)`$. Thus, according to OPUS the computer in a quantum state
$$|R_+(t)1/\sqrt{2}(|P(t)+|Q(t)),$$
(1)
also implements computation of the square of 5. The quantum state
$$|R_{}(t)1/\sqrt{2}(|P(t)|Q(t)),$$
(2)
is orthogonal to $`|R_+(t)`$. BH read Chalmers in such a way that OPUS can be applied to $`|R_+(t)`$ and $`|R_{}(t)`$, i.e., that the superposition $`1/\sqrt{2}(|R_+(t)|R_{}(t))`$ also implements computation of the square of 5. But,
$$\frac{1}{\sqrt{2}}(|R_+(t)|R_{}(t))=\frac{1}{2}[(|P(t)+|Q(t))(|P(t)|Q(t))]=|Q(t).$$
(3)
The state $`|Q(t)`$ corresponds to the computation of the square of 10. It corresponds to the punching of a different input, it has different registers activated during the calculation, it has different output. Clearly, it does not implement computation of the square of 5.
Applying this direct reading of Chalmers, BH reached somewhat different contradiction which lead them to reject Chalmers’s approach.
3. An alternative interpretation of Chalmers. It is possible to read Chalmers in another way such that the contradictions of the type described in the previous section do not arise. Let us make the following modification of the OPUS principle:
> OPUS“If a computation is implemented by a system in a maximal physical state $`P`$ which is not a superposition, it is also implemented by a system in a superposition of $`P`$ with orthogonal physical states”(Chalmers, 350)
This modified principle can be applied to $`P`$ and $`Q`$, but it cannot be applied to $`R_+`$ and $`R_{}`$ and, therefore, one cannot reach the contradiction described above as well as the contradictions described by BH.
One might see that OPUS is what Chalmers actually had in mind even though he did not say it explicitly. Indeed, another way to see the difference between OPUS (as read by BH) and OPUS is that in the latter it is required that $`P`$ corresponds to a single experience.<sup>1</sup><sup>1</sup>1In principle, the quantum state corresponding to a particular experience have a nonzero overlap with quantum states corresponding to other experiences due to the tails of quantum waves which must exist because of the uncertainty principle. But these overlaps are so small that they can be neglected in the discussion. Chalmers’s first definition of the OPUS principle is:
> If the theory predicts that a system in a maximal physical state $`P`$ gives rise to an associated maximal phenomenal state $`E`$, then the theory predicts that a system in a superposition of $`P`$ with some orthogonal physical states will also give rise to $`E`$. (Chalmers, 349)
The word “associated” hints that Chalmers meant that there is only one experience (“phenomenal state $`E`$” in Chalmers’s notation) corresponding to physical state $`P`$.
In fact, BH saw a possibility of reading OPUS as OPUS. The “(Version of) OPUS” described in their section 5.2.3 is essentially OPUS. They rejected this because they understood that Chalmers denies the existence of preferred basis. BH are correct in their criticism that without preferred basis there is no way to distinguish between quantum state which is a “superposition” and a state which is not a “superposition”. Thus, the modification of OPUS to OPUS cannot be done without assuming preferred basis.
We can read Chalmers in such a way that we do not run into inconsistency: Chalmers only objects to the claim that the mathematical formalism of quantum mechanics, i.e. the Schrödinger equation, leads to preferred basis. He cannot object to the existence of preferred basis, but he views it as arising from his theory of consciousness. This reading of Chalmers is justified by the following quotations:
> Everett assumes that a superposed brain state will have a number of distinct subjects of experience associated with it, but he does nothing to justify this assumption. It is clear that this matter depends crucially on a theory of consciousness. A similar suggestion is made by Penrose (1989): “… a theory of consciousness would be needed before the many-worlds view can be squared with what one actually observes” (348)
> … last three strategies are all indirect strategies, attempting to explain the discreteness of experience by explaining an underlying discreteness of macroscopic reality. An alternative strategy is to answer the question about experience directly. (349)
The main difficulty which BH see in putting together the principle of organization invariance together with OPUS follows from the same misinterpretation of Chalmers. If there is no preferred basis then they have reasons to say:
> … perceptual experience is (more or less) entirely illusory. When you seem to see a voltmeter needle pointing to ‘10’ your perceptual experience is probably veridical: the needle (if, indeed, we can sensibly speak of such a thing) is not pointing to ‘10’ or anywhere else.
However, accepting preferred basis, even if it is defined by the concept of experience itself, resolves the difficulty: the pointer does point to ‘10’ and in addition, in parallel worlds, to other values too.
Chalmers claims that his independently motivated theory of consciousness predicts that even in the world which is in a giant superposition there are subjects who experience a discrete world. He bases his argument on “the claim that consciousness arises from implementation of an appropriate computation.” Taking the model of a simple computer presented above, we can follow (at least approximately) his proof on p. 350. Projection of the superposed state on “the hyperplane of $`P`$” might mean projection of the quantum state of the computer in a “superposed” state at the initial time on the state corresponding to the input of calculating square of the number 5 which leads to quantum states of the various registers at later times corresponding to this calculation. The parallel between the calculation and experience yields the desired result, but accepting this parallel is relying on our experience. So, if we read Chalmers as BH do, that he claims to deduce “what the world is like if the Schrödinger equation is all” without the guide of our experience, then they have a valid criticism. However, Chalmers admits that Schrödinger equation cannot be all:
> … the only physical principle needed in quantum mechanics is the Schrödinger equation, and the measurement postulate and other basic principles are unnecessary baggage. To be sure, we need psychophysical principles as well, but we need those principles in any case, and it turns out that the principles that are plausible on independent grounds can do the requisite work here. (350-351)
I feel that these “independent grounds” are connected with our experience in a stronger way than one might imagine reading Chalmers. But this fact cannot lead to rejection of this approach as BH claim.
4. Byrne and Hall against any Everett-style interpretation. BH start their argument by pointing out that the orthodox quantum theory and the Everett interpretation formally defined on the same “family of state spaces” and that the difference is only in dynamics. Then they say that because of the difference in dynamics it does not follow that the quantum state corresponding to a particular experience in the orthodox theory will correspond to the same belief (if at any) in the framework of the Everett theory.
This might be considered as a criticism of Chalmers if one reads him saying that Everett theory predicts what our experiences should be, but usually this connection is postulated in Everett-style theories. There is a strong motivation for this postulate. The orthodox theory is defined only on a (tiny) part of the space of all quantum states: macroscopic quantum systems cannot be in a “superposition states”. The dynamics of the allowed states between quantum measurements is identical to the dynamics of the quantum states in the Everett theory. Let us discuss the example analyzed by BH at the end of p.385. When a state $`\varphi `$ is a state of an observer who has the belief that the measurement outcome was “up” in the orthodox theory, the dynamics will tell that she will write “up” in her lab-book. The dynamics of the state $`\varphi `$ in the Everett theory leads to the same action. This justifies considering $`\varphi `$ to be a “belief vector” in the Everett theory too.
BH proceed with their criticism claiming that Everett’s interpretation has less of “substantiative content” because when a quantum system is in a superposition of eigenstates with different eigenvalues of some quantity M, the orthodox interpretation associates probabilities to the various outcomes, while the Everett theory does not.
It is true that there is a difficulty with the concept of probability in the framework of the Everett-style interpretation. The Everett theory is a deterministic theory and it does not have a genuine randomness of the collapse of the orthodox interpretation. A deterministic theory might have the concept of ignorance probability, but it is not easy to find somebody who is ignorant of the result of a quantum experiment: it is senseless to ask what is the probability that an observer will obtain a particular result, because she will obtain all results for which there are a non-zero probabilities according to the orthodox approach. It seems also senseless to ask what is the probability of the observers in various branches (these are persons with the same name and the same memories about events which took place before the measurements, but who live in different branches corresponding to the different outcomes) to obtain various results, since obviously the probability to obtain the result “$`𝐌=m_i`$” in branch “$`j`$” is 1 if $`i=j`$ and it is 0 if $`ij`$. These are not the quantum probabilities we are looking for.
Nevertheless, there is solution for this difficulty (Vaidman 1998, 2000). The splitting into various branches occurs usually before the time when the observers in these branches become aware of the outcome of the measurement. (To ensure this we may ask the observer to keep her eyes close during the measurement.) Thus, an observer in each branch is ignorant about the outcome of the measurement and she can (while any external person cannot!) define the the ignorance probability for the outcome of the measurement. She will do so using standard probability postulate: the probability of an outcome is proportional to the square of the amplitude of the corresponding branch. Moreover, since observers in all these branches have identical concept of ignorance probability and since they all are descendents of the observer who performed the experiment, we can associate probability for an outcome of a measurement for this observer in the sense that this is the ignorance probability of her descendents in various branches.
The fact that I have used a probability postulate here does not spoil the argument: I had to show that substantive content of Everett interpretation is not less than that of the orthodox interpretation. The latter has the probability postulate as well. What was done here (and what was not trivial from the beginning) is presenting a way which allows to define probability in the frame of the many-worlds interpretation.
The last argument of BH relies on their claim that Everett-style interpretation lacks “statistical algorithm”. Since the ignorance probability defined above generates the same statistical algorithm as the the orthodox theory, this argument does not hold either.
5. Conclusions. The main claim of BH is “that any Everett-style interpretation should be rejected”. The basis of their argument is the observation that neither Chalmers nor anybody else can answer the question: “What the world is like if the Shrödinger equation is all?” It is true that this question is much more difficult to answer in the framework of the Everett-style interpretation relative to interpretations which do not have multitude of worlds. “The world is everything which exist” is not a valid definition. Moreover, the Shrödinger equation itself cannot define the concept of a “world”. The world is the concept defined by conscious beings and it requires the analysis of the mind-body connection. Chalmers’s theory of consciousness provides an answer. One might argue how substantial his answer is, but even if there is no a detailed answer to this question today, one cannot reject the Everett interpretation. It suffices that Everett’s theory is consistent with what we see as our world. It is so superior to the alternatives from the physics point of view, because it avoids randomness and action at a distance in Nature (e.g., see Vaidman 2000), that it is still preferable in spite of the fact that it is less satisfactory from the philosophical point of view. Therefore, even if BH were able to point out a the difficulty in obtaining the interpretation out of the “bare theory” this would not be enough for rejecting the Everett interpretation. Moreover, I have argued that the BH have not presented persuasive arguments showing the difficulty. Their first argument is that it is not obvious that the correspondence between quantum states and classical properties in the orthodox quantum mechanics can be transformed as it is to the Everett interpretation. This argument does not take into account the similarity in dynamics which justifies the identification. Their other arguments rely on the well known difficulty in the interpretation of probability in the many-worlds interpretation disregard a recently proposed solution of this difficulty (Vaidman, 1998).
In summary, BH were not able to show a flaw in Everett-style interpretations. The temptation to appeal to the philosophy of mind in interpreting quantum mechanics, in particular, the idea that a theory of mind might help rescue from the difficulties with standard interpretation is still very attractive. Indeed, the Everett-style interpretation which says that physics is described in full by the Schrödinger equation is the most satisfactory from the physics point of view. What is left is to complete Chalmers’s work, i.e. to elaborate the connection between the quantum state evolving according to the Schrödinger equation and our experience.
This research was supported in part by grant 471/98 of the Basic Research Foundation (administered by the Israel Academy of Sciences and Humanities).
REFERENCES
Byrne, A. and Hall, N. (1999) “Chalmers on Consciousness and Quantum Mechanics”, Philosophy of Science 66: 370-390.
Chalmers, D. J. (1996) The Conscious Mind, New York: Oxford University Press.
Vaidman, L. (1998) ‘On Schizophrenic Experiences of the Neutron or Why We should Believe in the Many-Worlds Interpretation of Quantum Theory’, International Studies in the Philosophy of Science 12, 245-261.
Vaidman, L. (2000) “The Many-Worlds Interpretation of Quantum Theory”, Stanford Encyclopedia of Philosophy (temporarily in http://www.tau.ac.il/$``$vaidman/mwi/mwst1.html). |
no-problem/0001/cs0001025.html | ar5iv | text | # Computational Geometry Column 38
## Acknowledgements
I thank E. Althaus, N. Amenta, M. Bern, T. Dey, D. Eppstein, J. Giesen, and K. Mehlhorn for their comments. |
no-problem/0001/cond-mat0001398.html | ar5iv | text | # Epitaxy and Magnetotransport of Sr2FeMoO6 thin films
## Abstract
By pulsed-laser deposition epitaxial thin films of Sr<sub>2</sub>FeMoO<sub>6</sub> have been prepared on (100) SrTiO<sub>3</sub> substrates. Already for a deposition temperature of 320C epitaxial growth is achieved. Depending on deposition parameters the films show metallic or semiconducting behavior. At high (low) deposition temperature the Fe,Mo sublattice has a rock-salt (random) structure. The metallic samples have a large negative magnetoresistance which peaks at the Curie temperature. The magnetic moment was determined to 4 $`\mu _B`$ per formula unit (f.u.), in agreement with the expected value for an ideal ferrimagnetic arrangement. We found at 300 K an ordinary Hall coefficient of $`6.01\times 10^{10}`$ m<sup>3</sup>/As, corresponding to an electronlike charge-carrier density of 1.3 per Fe,Mo-pair. In the semiconducting films the magnetic moment is reduced to 1 $`\mu _B`$/f.u. due to disorder in the Fe,Mo sublattice. In low fields an anomalous holelike contribution dominates the Hall voltage, which vanishes at low temperatures for the metallic films only.
The observation of colossal magnetoresistance in the half-metallic perovskite manganites has led to an intense research on ferromagnetic oxides. Recently, a large room temperature (RT) magnetoresistance was found in Sr<sub>2</sub>FeMoO<sub>6</sub> (SFMO) , a material belonging to the class of double-perovskites ($`AA^{}BB^{}`$O<sub>6</sub>). Depending on the metal ion radius the $`B`$ and $`B^{}`$ ions arrange in a random or ordered fashion. For the latter a layered or rock-salt structure is observed. The high Curie temperature and the high spin-polarization render these materials attractive as part of magnetic field sensors, e.g. in magnetic tunnel junctions. The results reported lately on epitaxial thin film preparation on SrTiO<sub>3</sub> (STO) substrates are not consistent. Metallic as well as semiconducting behavior was found. Also the magnetic saturation moments were smaller than expected suggesting disorder in the rock-salt arrangement. Our goal was to find the preparation parameters for metallic, fully ordered films. We investigate in this letter in detail the differences between epitaxial metallic and semiconducting films with respect to their structural, magnetic and magnetotransport behavior, including Hall effect. We prepared a whole series of samples, but will discuss exhaustively two samples A and B which mark all the general differences.
SFMO thin films were prepared by pulsed laser ablation in an oxygen partial pressure of $`10^110^7`$ Torr or in argon atmosphere of $`10^1`$ Torr from a stoichiometric target on (100) STO substrates. During deposition the substrate temperature $`T_\mathrm{D}`$ was constant with values covering the range from 300C to 950C. Crystal structure investigations were performed using a two-circle and a four-circle X-ray diffractometer. The magnetic properties were determined with a SQUID magnetometer. With a Mireau interferometer we evaluated the film thicknesses to typically 100 nm. By standard photolithographic methods the samples were patterned to a 3 mm wide and 8 mm long bridge. The longitudinal resistivity was measured by the standard four-point technique with a DC current. Below RT a standard superconducting magnet and above RT a cryostat with a furnace in a RT bore were used. The procedure for Hall effect measurements of the patterned samples is described in detail elsewhere.
The chemical composition of the films presented here was identical to the nominal composition of the target (Sr<sub>2</sub>FeMoO<sub>6±δ</sub>) as determined by Rutherford backscattering on a reference sample on MgO. Although the dimension of the crystallographic unit cell is $`\sqrt{2}a_0\times \sqrt{2}a_0\times 2a_0`$, where $`a_0`$ is the lattice parameter for a single $`AB`$O<sub>3</sub> perovskite ($`a_0`$ 4 Å), we use the larger cell doubled in all directions to underline the symmetry of the $`B,B^{}`$ rock-salt arrangement. The films have a (00$`l`$) orientation perpendicular to the STO plane. A segregation into clusters of compositions SrMoO<sub>3</sub> ($`2a_0=7.950`$ Å) and SrFeO<sub>3</sub> ($`2a_0=7.738`$ Å) should be visible in X-ray diffraction either as severe peak broadening for small clusters or as peak splitting for large clusters. Both effects are not observed. The in-plane orientation, film axes parallel to substrate axes, was checked by $`\varphi `$-scans of the symmetry equivalent {224} reflections.
Both deposition temperature $`T_\mathrm{D}`$ and oxygen partial pressure play a crucial role for phase formation and epitaxy. Growth of SFMO films at oxygen partial pressures above $`10^1`$ Torr was not possible irrespective of the substrate temperature, but a polycrystalline, yellow, insulating phase formed. Lower oxygen partial pressure during deposition stabilized formation of the SFMO phase. In the following we report the influence of the substrate temperature on phase formation and epitaxy for a series of films deposited at a very low oxygen partial pressure. This was realized either in flow of pure oxygen at a pressure of $`10^5`$ Torr or in flow of pure argon (99.996%) at a total pressure of $`10^1`$ Torr. At $`T_\mathrm{D}=320^{}`$C there is a sharp phase boundary for the epitaxial thin film growth. For lower temperatures the same insulating yellow phase was observed. At 320C up to the highest temperatures achievable with our heater of 950C the samples are single phase, dark and a high degree of $`a,b`$ and $`c`$-axis orientation is achieved. Annealing these samples in an oxygen partial pressure of more than $`10^1`$ Torr leads again to the polycrystalline yellow phase. Already at 320C perfect epitaxial growth is obtained with rocking curve width $`\mathrm{\Delta }\omega `$ of the (004) reflections below 0.04. Compared to results gained on other related perovskites as manganites or high $`T_c`$ superconductors, this is an astonishingly sharp crossover and low epitaxy temperature. A lowering of the epitaxy temperature will be expected for a perfect in-plane lattice match of film and substrate. However, this is not the case here. The bulk material is cubic with a doubled perovskite unit cell. It was refined in the space group Fm3m with a lattice constant of 7.897 Å. For the sample A, prepared at $`T_\mathrm{D}=320^{}`$C all axes, $`a,b`$ and $`c`$, are elongated to 7.972 Å and 8.057 Å, respectively. The lattice mismatch to the substrate is 2%.
With a higher $`T_\mathrm{D}`$ the $`c`$-axis length decreases continuously while the rocking curve widths $`\mathrm{\Delta }\omega `$ of the (004) reflection remain around 0.05 for substrate temperatures $`320T_\mathrm{D}910^{}`$C, as can be seen in Fig. 1.
For higher deposition temperatures the rocking curves broaden to an angular spread of 0.23 at 930C (sample B). At this deposition temperature we obtained an in-plane lattice constant of 7.876 Å and an out-of-plane constant of 7.896 Å. Therefore the cell volume of sample B is decreased significantly by 4.5% compared to sample A. Investigations of the reciprocal lattice show clear differences in the {111} reflections of the two films. They are present in sample B but absent in sample A. Intensity in this reflection is generated by the Fe,Mo ordered rock-salt arrangement. The intensity ratio of the (111) and (004) reflection in sample B is close to the expected value in the tetragonal I4/mmm structure. In contrast the absence of this reflection in sample A indicates random $`B,B^{}`$ site occupation resulting in an enlarged cell volume. The detailed dependence of the (111) intensity on $`T_\mathrm{D}`$ will be the subject of a future study. In the following we show that the differences in the crystal structure go along with respective differences in the transport properties.
The temperature coefficient of the resistivity depends on deposition conditions. For deposition temperatures $`320T_\mathrm{D}920^{}`$C the temperature coefficient is negative, i.e. semiconductorlike behavior.
The transport properties for these semiconducting films are described in detail elsewhere. Here we give in short the values for one of these films, sample A, whose crystallographic properties are discussed above. Its resistivity increases almost two orders of magnitude while cooling from RT down to 10 K. No significant MR is visible in this sample which marks the lowest possible deposition temperature for epitaxial film growth. With increasing deposition temperatures the ratios between the RT resistivities and the respective low temperature resistivities ($`T=10`$ K) increase, as is shown in Fig. 1. At 930C finally the ratio is larger than unity, i.e. a metallic behavior occurs. The transition from a negative to positive temperature coefficient of the resistivity is connected with the appearance of the (111) reflection signaling $`B,B^{}`$-site ordering. Figure 2 presents the resistivity (left axis) of the metallic sample as a function of temperature in zero field and in high magnetic field of 8 T. A positive temperature coefficient exists in the temperature range from 50 K to 500 K. Below 50 K a small increase can be observed with a residual resistivity at 4 K of 360 $`\mu \mathrm{\Omega }`$cm. The magnetic ordering shows up in the temperature dependence of the resistivity as a small anomaly in the zero field curve near the Curie temperature $`T_C`$. The anomaly vanishes by applying a high magnetic field. This temperature dependence is similar to the Sr-doped manganites which have the highest $`T_C`$ within the manganites.
Two different regimes with a high magnetoresistance MR, defined by MR=$`[\rho (T,\mu _0H=0\mathrm{T})\rho (T,\mu _0H=8\mathrm{T})]/\rho (T,\mu _0H=0\mathrm{T})`$ (right axis) can be distinguished. At very low temperatures the MR increases supposedly due to grain boundary effects. The peak at 380 K in the neighbourhood of the Curie temperature $`T_C`$ is probably due to the suppression of spin-fluctuations by the external magnetic field. These results are similar to those of polycrystalline manganites. The magnetic field dependence of the resistivity below and above (multiplied by a factor 10) $`T_C`$ is shown in Fig. 3.
Both curves have opposite curvatures. The MR is for small fields proportional to $`H^2`$ for $`T>T_C`$ while for $`T<T_C`$ a linear dependence is found in agreement with the symmetry considerations for the manganites by Snyder et al.
The Hall effect in a ferromagnetic material is described by
$$\rho _{xy}=R_HB+R_A\mu _0M$$
(1)
with the the magnetization $`M`$ and the ordinary and anomalous Hall coefficients $`R_H`$ and $`R_A`$, respectively. We evaluate in this paper the high field regime where the magnetization is close to saturation and $`\mathrm{d}\rho _{xy}/\mathrm{d}B\mathrm{d}\rho _{xy}/\mathrm{d}(\mu _0H)`$. We measured the Hall voltage $`U_{\mathrm{H}all}`$ with a current of 1 mA at constant temperatures $`T=4,150,300`$ K in magnetic fields up to 8 T. The results for sample B are shown in Fig. 4. In this figure the symbol sizes are larger than the experimental errors. With increasing temperature a steep increase of $`U_{\mathrm{H}all}`$ in low fields can be seen. At 1 T a maximum exists and at higher fields a linear negative slope is visible. This behavior is typical for ferromagnets where the anomalous Hall effect dominates due to the change in magnetization in low fields. At higher fields the magnetization saturates. Then according to Eq. 1 the anomalous Hall contribution is constant and the ordinary Hall effect is apparent. In contrast to the manganites $`R_A`$ is holelike and $`R_H`$ electronlike. The same sign correlation as in SFMO was also observed in iron and ferromagnetic iron alloys. In the following we focus on the discussion of the linear slopes $`\mathrm{d}\rho _{xy}/\mathrm{d}(\mu _0H)`$ in the high-field regime, given only by the ordinary Hall contribution.
We find at RT a Hall coefficent of $`6.01\times 10^{10}`$ m<sup>3</sup>/As, corresponding to $`n_e=1.3`$ electrons per $`B,B^{}`$-pair in a single-band model. With decreasing temperature $`n_e`$ decreases slightly to 0.9 electrons per f.u. The zero field mobility at 4 K and RT is calculated with these values of the charge-carrier density to 232 mm<sup>2</sup>/Vs and 100 mm<sup>2</sup>/Vs, respectively. A second metallic sample has almost the same charge-carrier concentration. However, in the semiconducting films a higher $`n_e`$ of 4.1 electrons per $`B,B^{}`$-pair is measured and it increases with decreasing temperature. The increase of the nominal charge-carrier density in the semiconducting sample compared to the metallic one and the high value shows, that a single band model is not appropriate for this compound and band structure effects dominate the Hall effect. Nevertheless, we give the nominal charge-carrier concentration per f.u. for easier comparison with other reports. The values gained from the metallic sample B are similar to single crystal data by Tomioka et al. At RT we find an anomalous Hall coefficient $`R_A`$=$`1.56\times 10^8`$ m<sup>3</sup>/As which vanishes for our metallic SFMO films at very low temperatures. This is expected by theory for an undistorted ferromagnetic state. SQUID magnetization measurements confirm this picture showing full saturation near 4.0 $`\mu _B`$/f.u. The nonvanishing anomalous Hall contribution in our semiconducting SFMO thin films and the single crystal data is an indication for a distorted ferrimagnetic state, which is indicated also by the reduced magnetizations of these samples. The values of the saturation magnetization $`M_S^{\mathrm{e}xp}`$ are 1.0 $`\mu _B`$/f.u. for the semiconducting SFMO film and 3.2 $`\mu _B`$/f.u. for the metallic single crystal. The reduction of the magnetic moment is due to $`B,B^{}`$-site disorder causing antiferromagnetic Fe-O-Fe bonds. Mößbauer measurements suggest a Fe<sup>3+</sup>(S=5/2) configuration and with Mo<sup>5+</sup>(S=1/2) one expects in a ferrimagnetic arrangement a value of 4 $`\mu _B`$/f.u. In Fig. 5 we show the hysteresis loop at 10 K for the metallic sample B.
The saturation magnetization $`M_S^{\mathrm{e}xp}`$ is close to the ideal value of $`M_S^{\mathrm{t}heo}=4`$ $`\mu _B`$/f.u.
In Table I crystal structure, magnetotransport, and magnetization data of the samples A and B are listed to emphasize the differences between low temperature deposited SFMO thin films and their high temperature deposited counterparts.
Concluding we prepared epitaxial Sr<sub>2</sub>FeMoO<sub>6</sub> thin films with a high degree of epitaxy by pulsed laser deposition. Already for a substrate temperature during fabrication of 320C epitaxial growth is achieved. Depending on deposition parameters we found a strong variation in cell volume and degree of order of the $`B,B^{}`$-site occupation, resulting in metallic or semiconducting behavior of the resistivity. For the metallic samples the negative magnetoresistance peaks in the neighbourhood of the Curie temperature. The magnetic moment is equal to the expected value for a ferrimagnetic arrangement of the Fe and Mo cations. At low temperatures the anomalous Hall effect vanishes, indicating an undistorted ferrimagnetic state in agreement with magnetization measurements. The ordinary and anomalous Hall coefficients are negative and positive, respectively. Both coefficients have reversed sign compared to the colossal magnetoresistive manganites. We found a charge-carrier concentration of 1.3 electrons per f.u. at room temperature.
The authors thank P. Latorre Carmona for experimental assistance, P. Gütlich from Institut für Anorganische Chemie und Analytische Chemie for usage of the magnet cryostat with RT access and G. Linker from Forschungszentrum Karlsruhe for the Rutherford backscattering analysis of the film stoichiometry. This work was supported by the Deutsche Forschungsgemeinschaft through Project No. JA821/1-3 and the Materialwissenschaftlichen Forschungszentrum (MWFZ) Mainz. |
no-problem/0001/astro-ph0001120.html | ar5iv | text | # Sunyaev-Zeldovich effect from hydrodynamical simulations: maps and low order statistics
## I Introduction
Cosmic microwave background (CMB) photons propagating through the universe are scattered by hot electrons along their path. The effect of this scattering on the photon distribution was first described by Sunyaev and Zeldovich and is called the Sunyaev-Zeldovich (SZ) effect. The effect conserves the number of photons, but changes their energy distribution. Low frequency photons in the Rayleigh-Jeans regime gain energy on average and are moved into the high frequency part of the Planck distribution, with a zero crossing at 217GHz in the nonrelativistic case. The amplitude of the effect is proportional to the product of electron temperature and density, or pressure, and is thus dominated by hot dense structures such as clusters.
The SZ effect is nowadays routinely detected by a number of instruments such as BIMA, Diabolo, OVRO, PRONAOS, Ryle, SEST, SuZie and Viper (see refs. for review). All of these have so far concentrated their efforts on known clusters, but future experiments will also observe blank fields of the sky with an effort to isolate this effect. Among these are the proposed one square degree survey and a number of CMB experiments, for which SZ may be a significant source of fluctuations that needs to be separated from the primary signal. The Planck satellite for example, with its combination of frequency coverage, all-sky map and angular resolution, should be able to identify several thousand SZ sources .
The SZ effect has a number of potentially observable consequences in addition to the one descibed above on the known clusters. First, the mean SZ distortion can be observed through the deviation of the photon spectrum from the Planck distribution. No such distortion has been detected by the COBE/FIRAS data yielding the upper limit on Comptonization parameter $`y`$ (defined below) of $`y<1.5\times 10^5`$ . This is a constraint that has to be satisfied by any viable cosmological model. As we will show below cluster abundance normalized models explored here typically satisfy this constraint.
Second, most of the CMB experiments measure fluctuations, for which the most relevant quantity is their power spectrum. Again no clear detection of SZ has been reported so far, although this is not unexpected given that the main source of fluctuations on degree scales probed so far are primary CMB fluctuations generated at the time of recombination. The next generation of CMB experiments is beginning to detect fluctuations at smaller angular scales. With higher sensitivities and broader frequency coverage SZ may become detectable even in random patches of the sky surveyed by these experiments. There are two related questions that one would like to address. First is the overall level of SZ fluctuations and whether SZ may be a significant source of CMB foreground contamination. Second is the power of SZ to distinguish between cosmological models using power spectrum or any other statistics, such as counting of sources or nongaussian signatures.
There are two main theoretical approaches that can be used when calculating the SZ effect. One is analytical and is based on the Press-Schechter approximation. In this approach all the mass in the universe is distributed into virialized halos. Gas in these halos is assumed to follow a prescribed distribution in density and temperature (usually isothermal with a $`\beta `$ density profile). One then integrates over the mass function and over the redshifts to calculate the first moment of the distribution (the $`y`$ parameter). A similar calculation also yields the uncorrelated contribution to the second moment to which one must also add the correlations between the halos . This approach involves a number of approximations that must be verified with numerical simulations. For example, some fraction of the gas may not be in virialized halos, but may be residing in more diffuse filaments or at the outskirts of the halos. There may be significant temperature and density fluctuations of the gas inside the halos. Furthermore, N-body simulations that have examined the accuracy of the Press-Schechter mass function have shown it to be a reasonable approximation for massive halos, but to overpredict the number of halos at lower masses. The resulting spectrum may be very sensitive to the details of the mass function distribution.
The alternative approach adopted in this paper is to use numerical simulations to calculate the SZ effect. Our approach is based on moving mesh hydrodynamical simulation code and we use ray tracing through the simulation box to generate maps of SZ. This is the approach that was adopted previously by . Most of these early studies were of low resolution, but nevertheless showed that SZ could be an important effect on the CMB. Recently generated SZ maps using SPH simulations of higher resolution, although they did not calculate the CMB power spectra. An alternative method based on $`3d`$ power spectra was developed in and was recently adopted to the MMH code as used in this paper by . We compare our results to the previous work where possible.
## II Method
We use moving mesh hydro (MMH) code developed by one of us . The code is a hybrid between Eulerian and Lagrangian grid based hydrodynamic methods. By deforming the grid in the dense regions along potential flow lines it provides a ten fold increase in resolution compared to fixed grid Eulerian codes, while maintaining regular grid conditions everywhere . The code has been succesfully parallelized on shared memory systems and has a low computational cost per grid cell. A $`256^3`$ run takes around 2-4 days of wall clock time on 32 processors.
We ran about a dozen simulations varying cosmological parameters, box sizes and particle/mesh dimensions to check for various numerical effects. Our main models are $`\tau CDM`$, $`\mathrm{\Lambda }CDM`$ and $`OCDM`$. Their parameters are
* $`\tau CDM`$: flat model with $`\mathrm{\Omega }_m=1`$, $`\mathrm{\Omega }_\mathrm{\Lambda }=0`$, $`\sigma _8=0.56`$, $`\mathrm{\Omega }_bh=0.034`$ and $`\mathrm{\Omega }_mh=0.25`$.
* $`\mathrm{\Lambda }CDM`$: flat model with $`\mathrm{\Omega }_m=0.37`$, $`\mathrm{\Omega }_\mathrm{\Lambda }=0.63`$, $`\sigma _8=0.8`$, $`\mathrm{\Omega }_bh=0.034`$ and $`\mathrm{\Omega }_mh=0.25`$.
* $`OCDM`$: open model with $`\mathrm{\Omega }_m=0.37`$, $`\mathrm{\Omega }_\mathrm{\Lambda }=0.0`$, $`\sigma _8=0.8`$, $`\mathrm{\Omega }_bh=0.034`$ and $`\mathrm{\Omega }_mh=0.25`$.
We ran several simulations varying box sizes from 50$`h^1`$Mpc to 200$`h^1`$Mpc to verify the effects of mass and force resolution on small scales and lack of power on large scales. We also varied mass and scale resolution, using mesh sizes between $`128^3`$ and $`256^3`$ and typically placing 1-8 dark matter particles per mesh cell. Our largest simulations are $`\mathrm{\Lambda }CDM`$ and $`OCDM`$ models with 100$`h^1`$Mpc box, $`256^3`$ mesh and $`256^3`$ dark matter particles.
During the simulation we store 2-d projections through the 3-d box at every conformal time step that corresponds to a light crossing time through the box. The projections are made alternatively in $`x`$, $`y`$ and $`z`$ direction to minimize the repetition of the same structures in projection. We store projections of SZ, kinetic SZ, gas and dark matter density. For SZ we store the projection
$$\mathrm{\Delta }y=\frac{k_B\sigma _T}{m_ec^2}n_eT_ea\mathrm{\Delta }\chi $$
(1)
where $`\sigma _T`$ is the Thomson cross-section, c the speed of light, $`m_e`$ the electron mass, $`\mathrm{\Delta }\chi `$ the comoving width of the box, $`a`$ the expansion factor, $`n_e`$ electron number density and $`T_e`$ electron temperature, both of which are obtained from the output of the MMH hydro code (note that we ignore relativistic corrections, which are only relevant for rare hot clusters). Our 2-d maps are $`1024^2`$ for $`256^3`$ and $`512^2`$ for $`128^3`$. We have verified that this preserves all the information content by comparing the results to the higher resolution projection. Number of stacked projections depends on the box size and ranges between 25 (for $`\tau CDM`$ with 200$`h^1`$Mpc box) to 150 (for $`\mathrm{\Lambda }CDM`$ with 50$`h^1`$Mpc box size).
After the simulation is completed we use the 2-d projections to make maps of SZ. We stack together the maps of SZ separated by the width of simulation box, randomly choosing the center of each box (note that use of periodic boundary conditions guarantees there is no edge for any of the maps). We then project these maps onto a map constant in angular size. In principle the angular scale of the projection should be determined by the angular scale of the simulation box at the initial $`z`$, since any larger scale produces repetition of the same structures in the map. However, most of the structures in the maps are coming from low $`z`$ and these cover very little volume of the simulation box; hence they do not repeat itself even if we increase the angular scale beyond the size of the box at the highest redshift. Typically we project up to $`z430`$ using oversampling between 1-4 without any noticeable artifacts in the maps. We checked for possible artifacts in the power spectra by comparing weak lensing maps produced in the same way to the analytic predictions . The agreement was very good in all models. For SZ it should be even better, since SZ power spectrum is much more dominated by isolated sources and, as will be shown below, large scale correlations do not dominate even on large scales.
A given simulation can be used not only for the simulated parameters, but also for models with lower $`\sigma _8`$. To do this one can simply relabel an earlier time output as being one at a later time by changing the redshift of projection
$$z^{\mathrm{new}}=(z^{\mathrm{old}}+1)\sigma _8^{\mathrm{new}}/\sigma _8^{\mathrm{old}}1,$$
(2)
where superscript old stands for original simulation and new for the new one. Once the redshifts are relabeled one can project over the 2-d projections using the nearest projection to the required $`z`$. For a flat model with no cosmological constant this scaling preserves all cosmological parameters (except $`\sigma _8`$). This scaling is particularly useful to obtain the dependence on $`\sigma _8`$ in a flat model independent of other parameters. We verified its accuracy by running two simulations: one with $`\sigma _8=1`$ but using the rescaling of equation 2 to obtain $`\sigma _8=0.56`$ model, and another with $`\sigma _8=0.56`$ at the final output, and found very good agreement between the two. The simulation with $`\sigma _8=1`$ was then used to predict $`\sigma _8`$ dependence of the $`y`$ parameter and SZ power spectrum.
It is also easy to verify the effect of the baryon density in the limit $`\mathrm{\Omega }_b\mathrm{\Omega }_m`$. In this limit the baryons are just a tracer of dark matter potential and do not dynamically couple to it. This fails in the cores of the halos, which are however already not correctly modelled in these adiabatic simulations without cooling and non-thermal energy injection. In this approximation SZ scales linearly with baryon density times Hubble constant $`\mathrm{\Omega }_bh`$. We only show results that are normalized to BBN nuclesynthesis constraint $`\mathrm{\Omega }_bh^2=0.018`$, but results for other values of $`\mathrm{\Omega }_bh`$ can be easily obtained using this scaling. Incidentally, one could also use transformation of scale to map a given model into another family of models, this time varying the shape and the amplitude of the input power spectrum. By combining this transformation with the time transformation above one can find a family of models which are normalized to $`\sigma _8`$ today and have a different shape of the power spectrum. The trade-off in this case is a loss in force and mass resolution. Some of the maps produced by the simulations can be found at http://feynman.princeton.edu/$``$uros/sz.html.
## III Press-Schechter predictions
The Compton parameter $`y=𝑑y`$ is given as a projection over the electron pressure along the line of sight (equation 1). This can be reexpressed as a line of sight integral over the density weighted temperature of electrons,
$$y=\frac{k_B\sigma _T\overline{n_e}}{m_ec^2}a^2(1+\delta _g)T_e𝑑\chi ,$$
(3)
where $`\delta _g`$ is the gas overdensity and $`\overline{n_e}`$ mean electron density today.
In Press-Schechter (PS) picture all matter in the universe is divided into halos of a given mass. The mass distribution is specified by the halo density mass function $`dn(M)/dM`$. This can be written as
$$\frac{dn(M)}{dM}dM=\frac{\overline{\rho }}{M}f(M)dM,$$
(4)
where $`\overline{\rho }`$ is the mean matter density of the universe. The function $`f(M)`$ denotes the fraction of mass in halos of mass $`M`$. It can be expressed in units in which it has a universal form independent of power spectrum or redshift if we express it as a function of peak height $`\nu =[\delta _c(z)/\sigma (M)]^2`$, where $`\delta _c`$ is the value of a spherical overdensity at which it collapses at $`z`$ ($`\delta _c=1.68`$ for Einstein-de Sitter model) and $`\sigma (M)`$ is the rms fluctuation in spheres that contain on average mass $`M`$ at initial time, extrapolated using linear theory to $`z`$ . For scale free spectra
$$\nu f(\nu )=\frac{M^2}{\overline{\rho }}\frac{dn}{dM}\frac{d\mathrm{ln}M}{d\mathrm{ln}\nu }.$$
(5)
The actual form given by Press & Schechter is $`\nu f(\nu )=(\nu /2\pi )^{1/2}e^{\nu /2}`$, although modified versions of this form that fit better N-body simulations have been proposed . Direct hydrodynamic simulations have shown good agreement with our form 5 .
The virial temperature of the halo in the spherical collapse model is only a function of virial mass and is given by $`k_BT_e=qM^{2/3}`$. The conversion factor from mass to temperature $`q`$ is approximately unity if mass is expressed in units of $`10^{13}h^1M_{}`$ and $`T`$ in keV. Density weighted temperature is given by
$$(1+\delta )T=qf(\nu )𝑑\nu M^{2/3}.$$
(6)
This gives $`(1+\delta )T=0.3`$keV today in a $`\mathrm{\Lambda }CDM`$ model with $`\mathrm{\Omega }_m=0.37`$ and $`\sigma _8=0.8`$, with other models giving comparable numbers. Further assuming that gas traces dark matter, ie $`\delta _g=\delta `$, we can calculate the integrated Compton $`y`$ parameter from equation 3. This gives $`y=2\times 10^6`$ for this model. Flat models give a factor of 2 lower value because of a more rapid evolution of clusters with $`z`$. This reduces $`y`$ relative to a low density model assuming the cluster abundance today is the same.
Using equation 6 we may also explore the dependence of $`y`$ on $`\sigma _8`$. For this we need to relate $`\sigma (M)`$ to $`M`$. Linear power spectrum at the cluster scale can be approximated as a power law $`P(k)\sigma _8^2k^n`$. Since rms variance $`\sigma ^2(M)`$ scales as $`k^3P(k)`$ one finds $`k\nu (\sigma _8^2)^{2/(n+3)}`$. Mass goes as $`M\overline{\rho }R^3k^3`$ for which we find $`M^{2/3}k^2\sigma _8^{\frac{4}{3+n}}`$. On cluster scales we have $`2<n<1`$, hence $`(1+\delta )T\sigma _8^{24}`$. The $`y`$ dependence on $`\sigma _8`$ depends on the projection of $`(1+\delta )T`$ along the line of sight (equation 3), but gives qualitatively similar result.
One can also compute power spectrum using the PS model. To do this one can first calculate the 3-d power spectrum of density weighted temperature $`P_{(1+\delta )T}(k)`$ and then project it along the line of sight using the Limber’s equation,
$$C_l=32\pi ^3\left[\frac{k_B\sigma _T\overline{n_e}}{m_ec^2}\right]^2P_{(1+\delta )T}(k=l/r,a)\frac{d\chi }{a^4r^2},$$
(7)
where $`r`$ is the comoving angular distance to $`\chi `$, given by $`\chi `$, $`R\mathrm{sinh}(\chi /R)`$ or $`R\mathrm{sin}(\chi /R)`$ in a flat, open or closed universe with curvature R, respectively. The above expression is valid in Rayleigh-Jeans limit where $`\mathrm{\Delta }T/Tj(x)y=2y`$. This can easily be rescaled to another frequency using the spectral function $`j(x)=x(e^x+1)(e^x1)^14`$ with $`x=h\nu /k_BT`$ (which in the limit $`x0`$ gives $`j=2`$).
In the PS formalism there are two contributions to the pressure power spectrum
$$P(k)=P^P(k)+P^{hh}(k)$$
(8)
The first term $`P^P(k)`$ arises from the correlations within the single halo. This term contribution to the 3-d power spectrum is given by
$$P_{(1+\delta )T}^P(k)=\frac{q^2}{(2\pi )^3}f(\nu )𝑑\nu \frac{M}{\overline{\rho }}M^{4/3}|y(k)|^2,$$
(9)
where $`y(k)`$ is the Fourier transform of the halo profile normalized to unity on large scales ($`k0`$), assumed here for simplicity to be independent of $`M`$. One power of $`M`$ above is given by mass pair weighting and $`M^{4/3}`$ is given by the square of the temperature of the halo. This term is heavily weighted toward rare massive clusters, resulting in a very strong Poisson term compared to the halo-halo correlation term discussed below. As shown in this term dominates over the halo-halo correlation even on large scales, where it behaves as a white noise. In the large scale limit ($`k0`$, $`y(k)=1`$) the scaling with $`\sigma _8`$ is given by $`P_{(1+\delta )T}^P\sigma _8^{14/(3+n)}\sigma _8^{714}`$. This will again be modified somewhat by the integration over the redshift, but does provide qualitative understaing for the strong $`\sigma _8`$ dependence seen in the simulations described below.
Second term is the contribution from halos correlated with one another. On large scales these cluster according to the linear power spectrum $`P_{\mathrm{lin}}(k)`$, except that they can be biased relative to the dark matter. The halo bias can be either larger than unity for halos more massive than the nonlinear mass or less than unity for those below that. An expression that fits N-body simulations reasonably well was given in (see also for a modification relevant for lower mass halos)
$$b(\nu )=1+\frac{\nu 1}{\delta _c}.$$
(10)
The halo-halo contribution to the density weighted $`T`$ power spectrum is given by
$`P_{(1+\delta )T}^{hh}(k)`$ $`=`$ $`P_{\mathrm{lin}}(k)\left[q{\displaystyle f(\nu )𝑑\nu M^{2/3}(\nu )b(\nu )y(k)}\right]^2`$ (11)
$``$ $`P_{\mathrm{lin}}(k)\left[b_{(1+\delta )T}(1+\delta )T\right]^2,`$ (12)
where $`b_{(1+\delta )T}`$ is the pressure weighted bias, defined as
$$b_{(1+\delta )T}\frac{f(\nu )𝑑\nu M^{2/3}(\nu )b(\nu )y(k)}{f(\nu )𝑑\nu M^{2/3}(\nu )y(k)}.$$
(13)
For $`\mathrm{\Lambda }CDM`$ model we find in the large scale limit ($`y=1`$) $`b_{(1+\delta )T}=3`$, hence SZ halos are significantly biased relative to the dark matter. This is because the $`TM^{2/3}`$ weighting weights preferentially towards the large halo masses which are biased. The bias decreases towards smaller scales where $`y(k)`$ suppression for finite $`k`$ is more important for larger more massive halos and the dominant contribution shifts towards smaller, unbiased or antibiased halos. Despite this large bias the halo correlation term is small relative to the halo term even on large scales and Press-Schechter model predicts that SZ power spectrum does not trace large scale structure on large scales.
If SZ does not trace LSS, the large scale bias $`b`$ cannot be measured from its power spectrum. Cross-correlating SZ with weak lensing or galaxy map is a more promising way to obtain this large scale bias. This is because the Poisson term is given by $`M^{5/3}`$ weighting,
$$P_{(1+\delta )T,\delta }^P(k)=\frac{q}{(2\pi )^3}f(\nu )𝑑\nu \frac{M}{\overline{\rho }}M^{2/3}|y(k)|^2,$$
(14)
and is so less dominated by rare objects than SZ power spectrum with $`M^{7/3}`$ weighting. The correlated contribution to the cross-correlation between density weighted $`T`$ and dark matter density is given by
$`P_{(1+\delta )T,\delta }^{hh}(k)`$ $`=`$ $`P_{\mathrm{lin}}(k)\left[q{\displaystyle f(\nu )𝑑\nu M^{2/3}b(M)}\right]\left[{\displaystyle f(\nu )𝑑\nu b(M)}\right]`$ (15)
$`=`$ $`P_{\mathrm{lin}}(k)b_{(1+\delta )T}(1+\delta )T,`$ (16)
where the second integral above is the mass weighted bias, which is by definition unity. Numerical comparison between the two terms indeed shows that the are comparable on large scales. To obtain the SZ-weak lensing or galaxy cross-correlation power spectrum we again use this 3-d power spectrum in Limber’s equation 7 with the appropriate window function for the dark matter or galaxy projection.
## IV Simulation results
### A Mean Compton parameter
We first present the results on the mean $`y`$ parameter. The value for this parameter is $`1.0\times 10^6`$ for $`\tau CDM`$ in 100$`h^1`$Mpc simulations with $`128^3`$ mesh, $`2.4\times 10^6`$ for $`OCDM`$ in 100$`h^1`$Mpc with $`256^3`$ mesh and $`2.0\times 10^6`$ for $`\mathrm{\Lambda }CDM`$ with 100$`h^1`$Mpc with $`256^3`$ mesh or 50$`h^1`$Mpc with $`128^3`$ mesh. For the latter $`y`$ decreases by roughly 20% if the resolution is decreased to $`128^3`$ mesh, suggesting that small halos not resolved in larger box simulations contribute a significant fraction to its value. This is also confirmed by Press-Schechter type calculations . Non-thermal energy injection and cooling may provide additional corrections to the results of the adiabatic simulations. For example, include feedback in their simulations and find significantly higher density averaged temperature (1keV) than our simulations (0.3keV), although other simulations that also include heating do not (G. Bryan, private communication). The mean $`y`$ parameter is therefore sensitive to the thermal history of the gas.
Our results are still one order of magnitude below the current experimental limits from COBE/FIRAS $`y<1.5\times 10^5`$ , so it is unlikely that this constraint will play a major role in distinguishing between the models that differ in the history of energy injection in the universe. On the other hand mean $`y`$ parameter does put a constraint on the $`\sigma _8`$. This scaling of $`y`$ with $`\sigma _8`$ can be obtained from the same simulation using the method described in previous section. We find $`y\sigma _8^{34}`$, which is in a good agreement with the Press-Schechter predictions. To violate the FIRAS limits one needs to increase $`\sigma _8`$ by a factor of 2 over the cluster abundance value. COBE normalized standard CDM with $`\sigma _8=1.3`$ would be problematic. Our results are a factor of 2 lower from those in for the two low density models (once we account for the difference in $`\mathrm{\Omega }_bh`$), while we are in a good agreement for $`\tau CDM`$ model. This disagreement does not seem to be explained by limited mass resolution on small scales, since our simulations have higher mass and force resolution, yet predict lower values for $`y`$ for the low density models. We agree well with results in .
### B SZ power spectrum
Next we explore the power spectra of SZ in the Rayleigh-Jeans regime. We first investigate the effects of resolution. These are shown in figure 1 for different parameters of the simulation for $`\mathrm{\Lambda }CDM`$ model. Shown are $`256^3`$ 100$`h^1`$Mpc simulation, $`128^3`$ 100$`h^1`$Mpc simulation, $`128^3`$ 200$`h^1`$Mpc simulation and $`128^3`$ 50$`h^1`$Mpc simulation. They all agree on large scales, where the shape is close to the white noise model, characteristic of a power spectrum dominated by rare sources. On small scales 200$`h^1`$ simulation begins to loose power for $`l>1000`$ compared to other simulations. Comparison between $`128^3`$ and $`256^3`$ 100$`h^1`$Mpc shows that the two are in good agreement on scales up to $`l1500`$, beyond which $`128^3`$ begins to loose power. Similar conclusion is obtained by comparing 50$`h^1`$Mpc and 100$`h^1`$Mpc both with $`128^3`$. The figure shows that a 50$`h^1`$Mpc box is sufficient for the power spectrum on large scales. This indicates that the large scale power spectrum is not dominated by very massive clusters above $`10^{15}M_{}`$ which do not form in such small boxes, but rather by less massive $`10^{1415}M_{}`$ clusters, which happen to be nearby in the projection along the line of sight. This conclusion is also confirmed by the Press-Schechter calculation, where the brightest sources correspond to this mass range. On small scales the 50$`h^1`$Mpc box has a power spectrum comparable to $`256^3`$ 100$`h^1`$Mpc simulation with the same mass and force resolution. Very little small scale power therefore arises from mode-mode coupling with scales larger than $`50h^1`$Mpc. Overall 100$`h^1`$Mpc $`256^3`$ simulation resolves the power spectrum between $`100<l<5000`$, while $`128^3`$ simulations of the same size are sufficient between $`100<l<2000`$.
On large scales the power spectrum approaches white noise in slope. This is an indication that the power spectrum is dominated by the rare bright sources in the map and not by the correlations between them, which would give a much shallower slope. The same behaviour is seen also in other cosmological models, shown in figure 2. While $`OCDM`$ gives similar predictions to $`\mathrm{\Lambda }CDM`$ model, $`\tau CDM`$ is significantly lower, just like in the case of mean $`y`$ parameter. Low density $`\mathrm{\Lambda }CDM`$ predicts $`3\times 10^7`$ fluctuations around $`l100`$, which rises to 3$`\times 10^6`$ at $`l5000`$. Comparison with primary CMB in figure 2 shows that the SZ is unlikely to contaminate CMB power spectrum for MAP, while Planck and smaller scale experiments should be able to measure the SZ power spectrum. Our results are in good agreement with , although we note that our $`OCDM`$ spectrum is somewhat higher than theirs, in better agreement with the PS calculations.
Figure 3 shows field to field variations in the power spectrum for $`2^{}\times 2^{}`$ maps. It shows significant fluctuations on large scales. This is another indication that the power spectrum on large scales is dominated by rare bright sources. On smaller scales the power spectrum is more robust. The main contribution on those scales is from smaller more abundant halos, so there is less sampling variance. Note that as the map size increases the chance of finding a bright source in it increases as well, so the median power spectrum on large scales grows with the map size. This is equivalent to the power spectrum after the brightest sources have been removed. The power spectrum in a typical few degree map should correspond to the power spectrum with a few thousand brightest sources removed. As shown in this can reduce the power spectrum on large scales significantly and should be kept in mind when comparing our results to other predictions. For example, power spectra in are based on all the power without the removal of bright sources, which should in general give a higher amplitude on large scales.
Figure 4 shows the change in the power spectrum as a function of $`\sigma _8`$ for $`\tau CDM`$ model. The amplitude of the power spectrum is very sensitive to $`\sigma _8`$. Doubling this parameter changes the power spectrum by 2 orders of magnitude. Fitting to a power law we find $`C_l\sigma _8^7`$. This steep dependence on $`\sigma _8`$ can be understood with Press-Schechter formalism developed in §3 and is caused by a rapid increase in number of bright sources as a function of $`\sigma _8`$, and is in contradiction to the estimates of . One can also see a crude estimate of the scaling relation by noting that the $`C_l<T>\delta ^2`$ and the temperature is a function of the non-linear length scale $`r_{\mathrm{NL}}`$, i.e. $`T(r_{\mathrm{NL}}H_0)^2`$ . For CDM like spectra near the non-linear mass scale we have $`r_{\mathrm{NL}}\sigma _8^{12}`$, giving $`C_l\sigma _8^{68}`$, similar to what we find in simulations and to the PS arguments. This indicates that care must be exercised when extracting cosmological parameters from SZ maps, since similar differences arise between low and high density cosmological models (figure 2). A small increase in $`\sigma _8`$ of the order of 20% changes the spectrum as much as does changing the density from $`\mathrm{\Omega }_m=1`$ to $`\mathrm{\Omega }_m=0.3`$. Unless we are confident that we know the local value of $`\sigma _8`$ to better than this accuracy we cannot use SZ to infer the density of the universe. Similar argument also explains the redshift dependence of SZ power spectrum, which is very strong. Most of the contribution to the power spectrum comes from $`z<1`$. This is also true in $`OCDM`$ and $`\mathrm{\Lambda }CDM`$, although the $`z`$ dependence is less steep there.
### C Non-Gaussian signatures
Since we have 2-d maps of SZ we may use these to study non-Gaussian signatures in SZ. There are two reasons why to investigate non-Gaussian signatures of SZ. First is that such information can be used to determine the cosmological model. Example of this is the SZ luminosity function, where one identifies SZ sources and plots the number density as a function of the flux in SZ . The abundance of such sources in the limit where they are unresolved has been made by the Press-Schechter formalism or peak-patch formalism . It has been shown that the number density of sources is strongly sensitive to the density parameter, similar to the mean $`y`$ and power spectrum statistics discussed above. However, the sources have some internal structure, they may be clustered, may contain substructure and may not be spherical, all of which complicates such analytic approaches and they need to be verified using the simulations. This is of particular interest to the proposed surveys of small fields centered on a random portion of the sky which will not in general detect massive clusters, but rather a collection of smaller sources. Simulations such as these presented here are necessary for investigation of the non-Gaussian effects in such random portions of the sky.
Second reason to study non-Gaussian signatures is that they may provide additional leverage in separating the SZ from the primary CMB, at least under the assumption that primary CMB is Gaussian. Even when one does not have sufficient frequency coverage to distinguish the two components on the basis of their different frequency dependence one can use non-Gaussian signature of SZ to estimate its contribution to the power spectrum. The only assumption in this procedure is that at a given smoothing scale there is a strong correlation between the non-Gaussian signature and its second moment, so that the latter can be estimated from the former.
It was shown above that the 2-point statistics is dominated by rare bright sources and so is very noisy on large scales. This will be even worse if one considers higher order statistics such as skewness and kurtosis. For this reason we concentrate here on the one point distribution function (pdf) of SZ smoothed at a given angular scale. These are shown in figure 5 for a $`2^{}\times 2^{}`$ field in $`\mathrm{\Lambda }CDM`$ model with several smoothing radii, all smoothed with a top-hat window. As shown in figure 5 the pdf is approximately lognormal on small scales, in the sense that there is an excess of large $`y`$ decrements caused by bright rare sources. The pdf can be transformed into an approximate Gaussian if plotted against $`\mathrm{log}(y)`$. For larger smoothing angles the pdf becomes narrower and approaches a Gaussian, so it becomes more difficult to distinguish it from primary CMB on this basis.
### D SZ-weak lensing and galaxy cross-correlation
SZ map can be cross-correlated with other maps, such as weak lensing, galaxy or X-ray maps. The first two are sensitive to projected density and may not be so strongly dominated by rare bright objects, while the latter traces projected $`\rho ^2T^{1/2}`$ and is even more strongly dominated by rare objects than SZ. Examples of weak lensing maps are those reconstructed from shape distortions of $`z1`$ galaxies or from CMB distortions at $`z1100`$. Example of projected galaxy maps are those from APM, SDSS or from numerous degree field surveys. They can be parametrized by the mean galaxy redshift. In the case of SDSS one can use photometric information to weight the galaxies according to their distance to optimize the signal to noise of the cross-correlation. Even more promising are smaller and deeper surveys, such as those used for weak lensing studies with several hundred thousand galaxies over a degree area.
A useful quantity to compare the cross-correlation between different maps is to compute the cross-correlation coefficient
$$\mathrm{Corr}(l)=\frac{C_{SZ,X}(l)}{[C_{SZ}(l)C_X(l)]^{1/2}}.$$
(17)
Figure 6 shows the cross-correlation coefficient for weak lensing map reconstructed from background galaxies at $`z=0.5,1,1100`$. The cross-correlation coefficients for all 3 cases is between 0.4 and 0.6 and only weakly depends on scale. Similar results are also obtained if one cross-correlates SZ with a galaxy catalog with a mean redshift of $`z0.20.5`$. This result shows that both SZ and weak lensing maps are dominated by relatively nearby objects, which is why the cross-correlation coefficient does not significantly decrease as the redshift of background galaxies is decreased. This is good news for shallow surveys such as SDSS, which should be able to detect the cross-correlation signal when compared to MAP or Planck maps.
The amplitude of the cross-correlations can also be compared to the analytic predictions developed in §3 and we find good agreement between the two. The predictions have also been made in the case of weak lensing of CMB by . Our results are several times lower on scales between $`100<l<3000`$. A more detailed comparison shows that the assumed bias in the two models is comparable ($`b34`$) and the difference is mostly caused by lower density weighted temperature in our simulations (0.3keV), consistent with PS, compared to their assumption (1keV). Density weighted temperature is sensitive to the thermal injection from stars and supernovae and it could be significantly higher than in our simulations, which neglect this effect. However, such heating of the gas would shift the weight in SZ to cooler halos (below 1keV), which are not biased or are even antibiased. So non-thermal increase in density weighted temperature also requires a decrease in bias and the two effects nearly cancel out, leading to a smaller senitivity of the cross-correlation power spectrum on the density weighted temperature than one would naively assume. Unfortunately this also implies that the expected signal to noise of SZ and CMB cross-correlation will be significantly lower than predicted and becomes only marginal using our results.
## V Discussion and Conclusions
Using hydrodynamical simulations we have produced maps of SZ and analyzed some of their low order statistics. The mean Compton parameter was found to be in the range $`\overline{y}12.5\times 10^6`$, an order of magnitude below current FIRAS limits. Only if one increases $`\sigma _8`$ by a factor of 2 over the cluster abundance constraint does one violate the FIRAS limit, something which is clearly excluded based on cluster abundance data. FIRAS limits will therefore not play a major role in constraining the cosmological models.
The power spectra were found to be noisy and white noise like on large scales, indicating that they are dominated by uncorrelated bright sources. The amplitude at $`l1000`$ is $`13\times 10^6`$ and does not represent a major source of foreground to primary CMB on large scales ($`l<1000`$). On smaller scales SZ power spectrum should be detectectable and should dominate over CMB for $`l>2000`$. On these scales the predictions become sensitive to the thermal energy injection into the gas and simulations that ignore such effects become unreliable. Conversely, power spectrum of SZ on small scales should give us important information on the thermal history of the gas in small halos. Note that the dominant contribution to the power spectrum comes from relatively nearby structures with $`z<1`$, so SZ will not provide information on gas history at high $`z`$.
On scales where SZ is not negligible compared to primary CMB one can use the non-Gaussian signatures of SZ to estimate its contribution to the power spectrum. We have shown that on arcminute scales the one point distribution function is well approximated as a log-normal, differing significantly from a normal distribution. This signature can be used to estimate the contrubution of SZ to the power spectrum even in the absence of multifrequency information.
Further insight into the thermal history of the gas can be obtained by cross-correlating SZ with other maps that trace large scale structure, such as weak lensing or projected galaxy map. This can also provide information on correlations between groups and clusters on large scales and their bias relative to dark matter. The cross-correlation coefficient is quite high, of order 0.5, across a wide range of scales even when the redshift distribution of galaxies or lensing mass peaks well below $`z1`$. This strong correlation provides further incentive to planned SZ surveys on random patches of the sky, since even if sensitivity is not sufficient to detect SZ directly, one may be able to detect it through cross-correlation with weak lensing or galaxy maps, thus providing information on the thermal state of the gas in the universe.
We thank Eiichiro Komatsu, Alexandre Refregier and David Spergel for useful correspondence. U.S. ackowledges the support of NASA grant NAG5-8084. Computing support from the National Center for Supercomputing Applications is ackowledged. |
no-problem/0001/cond-mat0001343.html | ar5iv | text | # Clogging Time of a Filter
## Abstract
We study the time until a filter becomes clogged due to the trapping of suspended particles as they pass through a porous medium. This trapping progressively impedes and eventually stops the flow of the carrier fluid. We develop a simple description for the pore geometry and the motion of the suspended particles which, together with extreme-value statistics, predicts that the distribution of times until a filter clogs has a power-law long-time tail, with an infinite mean clogging time. These results and its consequences are in accord with simulations on a square lattice porous network.
PACS Numbers: 47.55.Kf, 83.70.Hq, 64.60.Ak, 05.40.+j
In this letter, we investigate the time required for a filter to clog. In a typical filtration process, a dirty fluid is “cleaned” by passing it through a porous medium to remove the suspended particles. The medium enhances filtering efficiency by increasing both the available filter surface area for trapping suspended particles, as well as the exposure time of the suspension to the active surfaces. Such a mechanism is the basis of water purification, air filtration, and many other separation processes. As suspended particles become trapped, the fluid permeability of the medium gradually decreases, and eventually the filter becomes clogged. Determining the time dependence of this clogging is basic to predicting when a filter is no longer useful, either because of reduced throughput or reduced filtration efficiency, and should be discarded.
We develop a minimalist model which provides an intuitive understanding for the clogging of a filter. Our analytical results are based on representing the clogging of a porous medium by the clogging of a single parallel array of pores which are blocked in decreasing size order (Fig. 1). This geometrical reduction is based on previous filtration studies which showed that, when particles and pores are of comparable size, clogging preferentially occurs in the upstream end of the network. A single parallel array of pores represents the ultimate limit of this gradient-controlled process.
The approximation of size-ordered blocking is based on the commonly-used picture that a particle has a probability proportional to the relative flux to enter a given unblocked pore from among the outgoing pores at a junction . For Poiseuille flow, this flux is proportional to $`r^4`$, where $`r`$ is the pore radius. Size-ordered blocking arises if the exponent 4 in this flow-induced pore entrance probability is replaced by $`\mathrm{}`$. For a broad distribution of pore radii, this is not a drastic approximation. This ordering also provides a direct correspondence between the radius of the currently-blocked pore and the time until it is blocked. We then use extreme value statistics to compute the radius distribution of the last few unblocked pores, together with the connection between the radius of the currently blocked pore and the network permeability, to determine the distribution of filter clogging times. The predictions of this simple modelling are in good agreement with Monte Carlo simulations of clogging on a square lattice porous network.
In our simulations of clogging, we consider a lattice porous medium whose bonds represent pores. We assume Poiseuille flow, in which the fluid flux through a bond of radius $`r_i`$ is proportional to $`r_i^4p`$, where $`p`$ is the local pressure gradient when a fixed overall pressure drop is imposed. Dynamically neutral suspended particles traverse the medium in accordance with the local flow and the flow-induced entrance probability at each junction. A crucial feature is that particles are injected at a finite rate which is proportional to the overall fluid flux, corresponding to a fixed non-zero density of suspended particles in the fluid. This is in distinction to many previous studies of filtration, where particles were injected singly and tracked until trapping occurred; this corresponds to an arbitrarily dilute suspension. While useful for understanding the percolation process induced by clogging, the time evolution of filtration necessitates the consideration of a finite-density suspension.
For the trapping mechanism we adopt size exclusion, in which a particle of radius $`r_{\mathrm{particle}}`$ is trapped within the first bond encountered with $`r_{\mathrm{bond}}<r_{\mathrm{particle}}`$. We assume that a trapped particle blocks a pore completely and permanently. The difference between partial and total blockage in a single trapping event appears to be immaterial for long-time properties. The overall trapping rate is then controlled by the relative sizes of particles and bonds. For simplicity and because it occurs in many types of porous media, we consider the Hertz distribution of pore and particle radii, respectively,
$$b(r)=2\alpha re^{\alpha r^2},\mathrm{and}p(r)=2re^{r^2},$$
(1)
with the ratio between the average bond and particle radii $`s=1/\sqrt{\alpha }`$ a basic parameter which determines the nature of the clogging process.
First consider the case where pores are larger than particles ($`s1`$). Many particles must be injected before a sufficiently large particle arises which can block the largest pore. Since the permeability of the system remains nearly constant when only a few pores are blocked, the time between successive particle injection events during this initial stage is nearly constant. Once the largest pores are blocked, it takes many fewer particles to block the remaining smaller pores and clogging proceeds more quickly. Thus the clogging time is dominated by the initial blockage events.
To estimate the clogging time in this large-pore limit, let us find the number of injection events before encountering a particle large enough to block the largest bond. The radius of this bond is given by the criterion
$$_{r_{\mathrm{max}}}^{\mathrm{}}2\alpha re^{\alpha r^2}𝑑r=\frac{1}{w},$$
(2)
that one bond out of $`w`$ has has radius $`r_{\mathrm{max}}`$ or larger. This gives $`r_{\mathrm{max}}^{(b)}=s\sqrt{\mathrm{ln}w}`$. Similarly, the largest particle out of $`N`$ has radius $`r_{\mathrm{max}}^{(p)}=\sqrt{\mathrm{ln}N}`$. Thus $`N_1w^{s^2}`$ particles typically need to be injected before one occurs which is large enough to block the largest bond. Continuing this picture sequentially, the radius of the $`k^{\mathrm{th}}`$-largest bond is given by Eq. (2), but with $`1/w`$ replaced by $`k/w`$; this gives $`r_k=s\sqrt{\mathrm{ln}(w/k)}`$. Therefore $`N_k(w/k)^{s^2}`$ particles need to be injected before the $`k^{\mathrm{th}}`$-largest bond is blocked. Since $`N_k`$ decreases rapidly with $`k`$, the initial blockage events control the clogging time $`T`$, whose lower bound is given by $`TN_1/w>w^{s^21}`$. The factor of $`1/w`$ arises because the particle injection rate is proportional to $`w`$ for a fixed-density suspension and a fixed pressure drop. However, there is considerable particle penetration in the large-pore limit, so that the equivalence to the clogging of a parallel bond array is not really appropriate and the bound on $`T`$ is quite crude.
The latter case where pores are smaller than particles ($`s1`$) is simpler, as each particle injection event typically leads to the clogging of the first pore entered. Now the initial pores are blocked quickly, while later pores are blocked more slowly because the overall flow rate decreases significantly near the end of the clogging process. We argue that clogging is dominated by the time of these later blockage events. First, let us determine how the flow rate varies as the last few bonds get blocked. Since only the smallest bonds remain open near clogging, the permeability is determined by these smallest radii. We estimate the radius of the $`k^{\mathrm{th}}`$ smallest bond from
$$_0^{r_k}2\alpha re^{\alpha r^2}𝑑r=\frac{k}{w},$$
(3)
which gives $`r_k=s\sqrt{k/w}`$. The permeability of a parallel bundle of these $`k`$ smallest pores is then
$$\kappa (k)=\underset{j=1}{\overset{k}{}}r_j^4s^4\underset{j=1}{\overset{k}{}}(j/w)^2s^4k^3/w^2,$$
(4)
while the initial system permeability (obtained by setting $`k=w`$ above) is simply $`\kappa (w)=s^4w`$.
Since the overall fluid flow is proportional to the permeability for a fixed pressure drop, the time increment $`t_k`$ between blocking the $`(k1)^{\mathrm{st}}`$-smallest and $`k^{\mathrm{th}}`$-smallest pore scales as
$$t_k=\frac{\kappa (w)}{w\kappa (k)}\frac{w^2}{k^3}.$$
(5)
Here the factor $`w`$ in the denominator again accounts for a particle injection rate proportional to $`w`$. The clogging time $`T`$ is now dominated by the time to block the smallest bond, so that $`T>t_1w^2`$.
Thus the relative pore and particle radii drives a transition in the clogging process. For $`s^2>3`$, clogging is dominated by the initial blockage events and $`T>w^{s^21}`$, while for $`s^2<3`$, the last events dominate, leading to $`T>w^2`$. For the small-pore limit, we may carry the analysis further and obtain the distribution of clogging times. This distribution gives a divergent mean clogging time; nevertheless, suitably defined moments of the clogging time distribution scale as $`w^2`$.
We find the clogging time distribution in terms of the radius distribution of the smallest bond, since this bond ultimately controls clogging in a single parallel bond array. For the Hertz distribution, the probability that a given bond has a radius greater than or equal to $`r`$, $`B_>(r)`$, is
$$B_>(r)=_r^{\mathrm{}}2\alpha re^{\alpha r^2}𝑑r=e^{\alpha r^2}.$$
(6)
Then the radius distribution of the smallest bond from among $`w`$, $`S_w(r)`$, is given by
$$S_w(r)=wb(r)B_>(r)^{w1}=2\alpha wre^{\alpha wr^2}.$$
(7)
The first equality expresses the fact that one of the $`w`$ bonds has (smallest) radius $`r`$, with probability $`b(r)`$, and the other $`w1`$ must have radii larger than $`r`$.
From the basic connections between permeability, pore radius, and time scale (Eqs. (4) and (5)), and the fact that the clogging time is dominated by $`t_1`$, we deduce
$$Tt_1\frac{\kappa (w)}{w\kappa (1)}=\frac{s^4}{r_1^4},$$
(8)
while the clogging time distribution, $`P_w(T)`$, is directly related to the smallest bond radius distribution through $`P_w(T)dT=S_w(r)dr`$. Using Eqs. (7) and (8), we obtain the basic result (independent of system length)
$$P_w(T)\frac{w}{T^{3/2}}e^{w/T^{1/2}}.$$
(9)
The power law should apply in the time range $`w^2<t<w^2N^2`$. The former corresponds to the size of the typical smallest pore in a single realization, $`s/\sqrt{w}`$, while the latter corresponds to the smallest pore from among $`N`$ realizations of the system, $`s/\sqrt{Nw}`$. The essentially singular short-time cutoff arises from those realizations where the smallest bond happens to be anomalously large. Other coincident particle and bond radius distributions lead to qualitatively similar forms for $`P_w(T)`$, but with a different exponent value. For example, if $`S_w(r)r^\mu `$ as $`r0`$, then the long-time exponent in $`P_w(T)`$ is $`(\mu +5)/4`$.
We test our predictions by Monte Carlo simulations of the motion and trapping of suspended particles in a lattice porous medium. Each bond (with unit length) corresponds to a pore and the sites represent pore junctions. The algorithm is event-driven to bypass the physical slowing down of the flow as clogging is approached. At an intermediate stage, there are a finite number of particles in the network, consistent with a unit pressure gradient and a constant-density suspension. Particles are defined to be always at lattice sites and each (including newly injected particles) evolves by either: (a) moving to the next “downstream” site, (b) blocking the downstream bond it has entered, or (c) remaining stationary. The probability of each of these possibilities is defined so that the process corresponds to particles moving, on average, at the local velocity.
To achieve this, each particle at a junction selects an outgoing bond $`i`$, with entrance probability equal to the fractional flux into this bond. In an update, a particle attempts to move to the next junction, via the pre-selected bond, with probability proportional to $`v_i/v_{\mathrm{max}}`$, where $`v_i`$ is the local bond velocity and $`v_{\mathrm{max}}`$ is the instantaneous largest particle velocity in the system. If the attempt occurs, the particle either moves to the next downstream site, or, if the particle is too large, blocks the bond. After a single update of all particles, the time is incremented by $`\mathrm{\Delta }t=1/v_{\mathrm{max}}`$ to ensure that, on average, a particle moving along bond $`i`$ has speed $`v_i`$.
After all particles undergo their move attempts, $`𝒩\varphi \mathrm{\Delta }t`$ new particles are injected at the upstream end of the network, where $`\varphi `$ is the overall fluid flux. This ensures the correct absolute velocity for each particle. Since the time increment systematically increases as bonds get blocked, the particle injection rate progressively slows. Our approach is distinct from earlier multiparticle simulations of filtration, where the injection rate was decoupled from the overall flow.
In Fig. 2(a), we plot the clogging time distribution from simulations on the square lattice and the bubble model of the same spatial extent. The latter is a series of parallel bond arrays with perfect mixing at each junction; this idealized system accounts for geometrical aspects of clogging . The agreement between the two distributions is remarkably good, suggesting that the schematic evolution proposed in Fig. 1 is quantitatively correct. The tails of the two distributions in Fig. 9(a) are well-fit by a power law with exponent $`3/2`$. As a further test of Eq. (9), we plot, in scaled units, data for the clogging time distribution for systems of various widths $`w`$ (Fig. 2(b)) and compare to the functional form in Eq. (9).
An important consequence of the power-law tail in the clogging time distribution is a transition in the corresponding moments. These are
$$t^k=_0^{\mathrm{}}tP_w(t)𝑑t_{w^2}^{N^2w^2}wt^{k3/2}𝑑t,$$
(10)
where the approximation replaces the short-time cutoff at $`tw^2`$ in Eq. (9) by the lower limit in the integral. Consequently
$$M_k(w)t^k^{1/k}\{\begin{array}{cc}w^2N^{21/k}\hfill & k>1/2\text{;}\hfill \\ w^2(\mathrm{ln}N)^2\hfill & k=1/2\text{;}\hfill \\ w^2\hfill & k<1/2\text{.}\hfill \end{array}$$
(11)
Thus the mean clogging time diverges when the number of realizations is infinite, leading to large sample to sample fluctuations in the clogging time, while the moments $`M_k`$ with $`k<1/2`$ should be well behaved. The transitions in the moments is illustrated in Fig. (3), where $`M_k`$ behaves erratically for $`k=1`$ and 1/2, but then appears to grow as $`w^2`$ for $`k1/3`$.
The behavior of $`M_k`$ raises the question of when is a filter no longer useful. Waiting until complete clogging is impractical because of large fluctuations in the clogging time and eventual poor filter performance. It would be more useful to operate a filter only until the permeability decays to a (small) fraction of its initial value, such that reasonable flow and trapping efficiency are maintained, while minimizing fluctuations in this threshold time. We provide some preliminary results about these questions. Simulations indicate that the time to reach permeability fraction $`f`$, $`t(f)`$, is proportional to $`w^{2/3}f^{1/2}`$ for $`10^{10}f10^3`$, with the distribution of these fractional times progressively broadening for decreasing $`f`$ (Fig. 4). Thus waiting until the permeability decays to a fixed fraction can provide at least a reliable criterion for when a filter should be discarded. A consequence of $`t(f)w^{2/3}f^{1/2}`$ is that the permeability decays as $`1/t^2`$ over a wide range. Because of this rapid decrease, a practical filter should also have pores typically larger than particles to have a reasonable lifetime. These issues and understanding the efficiency of a filter throughout its useful life are currently under investigation.
We thank P. L. Krapivsky for many helpful discussions and a critical manuscript reading, as well as grants NSF DMR9978902 and ARO DAAD19-99-1-0173 for financial support. |
no-problem/0001/quant-ph0001009.html | ar5iv | text | # Untitled Document
New strategy for suppressing decoherence in quantum computation
Miroljub Dugić
Department of Physics, Faculty of Science, Kragujevac, Yugoslavia
Abstract: Controlable strong interaction of the qubit’s bath with an external system (i.e. with the bath’s environment) allows for choosing the conditions under which the decoherence of the qubit’s states can be substantially decreased (in a certain limit: completely avoided). By ”substantially decreased” we mean that the correlations which involve the bath’s states prove negligible, while the correlations between the qubit’s and the environment’s states can be made ineffective during a comparatively long time interval. So, effectively, one may choose the conditions under which, for sufficiently long time interval, the initial state of ”qubit + bath” remains unchanged, thus removing any kind of the errors. The method has been successfully employed in the (simplified) model of the solid-state-nuclear quantum computer (proposed by Kane).
1. Introduction
The issue of decoherence in quantum computation is one of the central subjects in both fundamentals and practical realizability of the quantum computers.
Here we propose a new strategy (method) for substantial suppression of decoherence in quantum computers for arbitrary type of the errors. As opposite to the existing approaches/methods , our strategy is simple both conceptualy and mathematicaly.
Physicaly, the central idea of our approach relies on the following assumptions: (i) extension of the composite system ”qubit + bath (Q+B)” with the bath’s environment, so obtaining the new composite system ”qubit + bath + (bath’s) environment (Q+B+E)”, and (ii) assumption of existence of robust states of the bath which can be selected (and kept (approximately) unchanged) by controlable strong (quantum-measurement-like) interaction of the bath with its environment.
Given the points (assumptions) (i) and (ii), the standard stationary perturbation theory points out the next possibiity: with a proper choice of the initial state of the bath, one may obtain substantial suppression of decoherence: the correlations involving the bath’s states are ”(virtually) arbitrarily small”, while for comparatively long time interval the initial state of Q+B remains unchanged, thus removing any kind of the errors!
We give a brief account of the issues naturally appearing in this context, appointing (without details) that the method works in the (simplified) model of the solid-state-nuclear quantum computer.
2. General background
We employ the standard assumptions of the general decoherence theory : (a) initially, the Q+B+E system is decoupled, and (b) the dominant terms are the interaction-Hamiltonians, $`\widehat{H}_{QB}+\widehat{H}_{BE}`$. Then we apply the standard algebra , based on the direct use of the unitary time-evolution of the composite system Q+B+E.
One should remind that the standard models of the theory of decoherence (including the considerations in quantum computation) deal with the composite system Q+B, where the interaction $`\widehat{H}_{QB}`$ usually is of the type (of the separable kind ):
$$\widehat{H}_{QB}=c\underset{p,q}{}\gamma _{pq}\widehat{P}_{Qp}\widehat{\mathrm{\Pi }}_{Bq},$$
$`(1)`$
where $`\widehat{P}_{Qp}`$ and $`\widehat{\mathrm{\Pi }}_{Bq}`$ represent the projectors onto the corresponding subspaces of the Hilbert state-spaces of Q, and of B, respectively; $`c`$ represents the coupling constant.
Then the decoherence is loosely (but sufficiently) presented by the (time) decrease of the off-diagonal elements of the Q’s ”density matrix”:
$$\rho _{Qpp^{}}=C_pC_p^{}^{}z_{pp^{}}(t),$$
$`(2a)`$
where the ”correlation amplitude” $`z_{pp^{}}(t)`$” :
$$z_{pp^{}}(t)=\underset{q}{}p_q\mathrm{exp}(ıct(\gamma _{pq}\gamma _{p^{}q})/\mathrm{}),\underset{q}{}p_q=1;$$
$`(2b)`$
notice the dependence on the eigenvalues of $`\widehat{H}_{QB}`$.
It is important to note that in the context of the, so-called, macroscopic considerations , the states enumerated by the indeces $`p,p^{}`$ \- which are the elements of the ”pointer basis” - are considered to be robust under the influence of the environment. I.e., most of the basic assumptions necessary for the ”transition from quantum to classical” presuppose invariance of the ”pointer basis” under the transformations generated by the interaction Hamiltonian. In idealized form it reads:
$$\widehat{H}_{QB}|\mathrm{\Psi }_p_Q|\chi _B=|\mathrm{\Psi }_p_Q|\chi ^{}_B,$$
$`(3)`$
where $`|\mathrm{\Psi }_p_Q`$ is an element of the ”pointer basis” - i.e., of the set of states bearing the (semi)classical character \[2-4\] of an open quantum system.
3. The strategy
We extend the system Q+B by the bath’s environment (E): thus dealing with the new composite system Q+B+E.
The methodology distingusihed in Section 2 points to the following assumptions: (i) initially, the system is in uncorrelated state $`|\mathrm{\Psi }_Q|0_B|\chi _E`$, and (ii) the unitary-evolution operator $`\widehat{U}_{QBE}`$ can be written as:
$$\widehat{U}_{QBE}\mathrm{exp}(ıt(\widehat{H}_{QB}+\widehat{H}_{BE})/\mathrm{}).$$
$`(4)`$
These standard assumptions are extended by the following crucial assumptions: (A) The interaction Hamiltonian $`\widehat{H}_{BE}`$:
$$\widehat{H}_{BE}=C\underset{i,j}{}\kappa _{ij}\widehat{P}_{Bi}\widehat{\mathrm{\Pi }}_{Ej},$$
$`(5)`$
(compare to eq.(1)) is the dominant term, thus making $`\widehat{H}_{QB}`$ the perturbation ($`C`$ is the coupling constant), and (B) The initial state $`|0_B`$ can be chosen such that one may state (cf. eq. (3)):
$$\widehat{H}_{BE}|0_B|\chi _E=|0_B|\chi ^{}_E.$$
$`(6)`$
Then one may employ the standard stationary perturbation theory .
3.1 The perturbation theory employed
The basic idea of the perturbation theory is presented by the following expressions::
$$\widehat{H}_{BE}|\mathrm{\Phi }_n_{QBE}=E_n^{(0)}|\mathrm{\Phi }_n_{QBE},$$
$`(7)`$
$$(\widehat{H}_{BE}+\widehat{H}_{QB})|\mathrm{\Psi }_n_{QBE}=E_n|\mathrm{\Psi }_n_{QBE},$$
$`(8)`$
where in the limit $`c0`$ one has: $`|\mathrm{\Psi }_n_{QBE}|\mathrm{\Phi }_n_{QBE}`$, and $`E_nE_n^{(0)}`$. (Remind that: the coupling constant $`c`$ is given and the above limit should not be literaly understood (it is here for the formal completeness of the considerations); what we shall further need is the ratio of the coupling constants $`c/C`$, where the limit $`c/C0`$ is legitimate.)
As it directly follows from eq.(5), the states $`|\mathrm{\Phi }_{QBE}`$ can be chosen as $`|pij|p_Q|i_B|j_B`$. Then, in accordance with (7) and (8) one may write for the normalized eigenstates:
$$|\mathrm{\Psi }_n_{QBE}|\mathrm{\Psi }_{pij}_{QBE}=(1ϵ_{pij}^2)^{1/2}|pij+ϵ_{pij}|\chi _{pij},$$
$`(9)`$
where $`pij|\chi _{pij}=0`$ and $`\chi _{pij}|\chi _{pij}=1`$, and
$$E_nE_{pij}=E_{pij}^{(0)}+\lambda _{pij}C\kappa _{ij}+\lambda _{pij}.$$
$`(10)`$
Notice: the corrections of the $`\widehat{H}_{BE}`$’s eigenstates and eigenvalues are $`ϵ_{pij}|\chi _{pij}`$ and $`\lambda _{pij}`$, respectively, while eqs.(9, 10) are exact!
Then from eq. (4), (9), (10), one directly obtains:
$$\widehat{U}_{QBE}\widehat{U}_1+\widehat{U}_2,$$
$`(11a)`$
where
$$\widehat{U}_1=\underset{(pij)}{}\mathrm{exp}(ıtE_{pij}/\mathrm{})(1ϵ_{pij}^2)|pijpij|,$$
$`(11b)`$
where the sum runs over the different combinations of the indeces ”p, i, j”. Bearing in mind that $`\widehat{U}_1+\widehat{U}_2=1`$, it is a matter of straightforward algebra to prove that:
$$1=n_1+n_2,n_1=\mathrm{\Psi }|\widehat{U}_1^{\mathrm{}}\widehat{U}_1|\mathrm{\Psi },$$
$`(12)`$
and
$$n_1(1ϵ_{max}^2)^2,$$
$`(12b)`$
where $`ϵ_{max}`$ is the maximal value of $`ϵ`$s defined by eq.(9).
So, applying $`\widehat{U}_{QBE}`$ onto the initial state $`|\mathrm{\Psi }_Q|0_B|\chi _E`$, one obtains:
$$\widehat{U}_{QBE}|\mathrm{\Psi }_Q|0_B|\chi _E\underset{(pij)}{}C_p\alpha _i\beta _j|pij+O(ϵ_{max}),$$
$`(13)`$
where $`C_p=p|\mathrm{\Psi }`$, $`\alpha _i=i|0`$, $`\beta _j=j|\chi `$, and the bases $`\{|i_B\}`$ and $`\{|j_E\}`$ diagonalize $`\widehat{H}_{BE}`$.
With the choice (cf. above point (ii)) presented by eq.(6), $`\alpha _i=\delta _{ii_{}}`$, one obtains:
$$\widehat{U}_{QBE}|\mathrm{\Psi }_Q|0_B|\chi _E\underset{(pj)}{}C_p\beta _j\mathrm{exp}(ıtE_{p0j}/\mathrm{})|p0j+O(ϵ_{max}).$$
$`(14)`$
Now, since
$$E_{p0j}=C\kappa _{0j}+\lambda _{p0j},$$
$`(15)`$
and
$$|\lambda _{p0j}|(1ϵ_{p0j}^2)^{1/2}|p0j|\widehat{H}_{QB}|p0j|+$$
$$+|ϵ_{p0j}|(1ϵ_{p0j}^2)^1|p0j|\widehat{H}_{QB}|\chi _{p0j}|$$
$`(16)`$
one obtains the main result of this paper:
if one may choose $`|0_B`$ so as the maximal value ($`\lambda _{max}`$) of $`|\lambda _{poj}|`$s can be very small, then one may speak of the substantial suppression of decoherence.
Actually, since it can be estimated that $`ϵ_{max}c/C`$, one may say that the correlations between Q and B (and E) can be considered arbitrarily small, the occurrence of the errors which come together with the change of B’s state represent substantially rare events: the total probability of these errors not exceding the order of $`(c/C)^2`$, where $`C`$ is virtually arbitrary. Now, for $`\lambda _{max}`$ very small, one may write:
$$\widehat{U}_{QBE}|\mathrm{\Psi }_Q|0_B|\chi _E\underset{(pj)}{}C_p|p_Q$$
$$|0_B\underset{j}{}\beta _j\mathrm{exp}(ıtC\kappa _{0j}/\mathrm{})|j_E$$
$$|\mathrm{\Psi }_Q|0_B\underset{j}{}\beta _j\mathrm{exp}(ıCt\kappa _{0j}/\mathrm{})|j_E.$$
$`(17)`$
for at least the time interval $`\tau `$:
$$\tau (\lambda _{max}/\mathrm{})^1.$$
$`(18)`$
I.e. during this time interval, the correlations between the states of Q and E do not become effective. (Notice: the situation with this regard is even much better, for the correlations are ”driven” by the second term on the r.h.s. of eq.(16)!)
3.2 Physical interpretation
The method directly points to the next protocol for avoiding the errors in quantum computation:
First, an effective, quantum-measurement-like action of the environment on the bath should be performed, which serves for preparing a robust initial state $`|0_B`$. Once such a state is obtained, the interaction between B and E should be strenghtened and prolonged in time, for at least the interval $`\tau `$. If the interaction is sufficiently strong ($`c/C1`$), then, for the proper choice of the initial state of the bath, the above algebra guaranties that for the interval of the order of $`\tau `$ (eq.(18)), the initial state of Q+B would appear unchanged, which is sufficient for preparing the computations in the time intervals much shorter than $`\tau `$. The same procedure should be repeated for each calculation step.
Being virtually arbitrary, $`ϵ_{max}`$ and $`\lambda _{max}`$ allow for substantial suppression of decoherence of the qubit’s states, all the effects of decoherence referring to substantially rare events ($`c/C1`$), or falling far beyond the interval $`\tau `$ ($`|0_B`$ such that $`\lambda _{max}1`$).
Needless to say, in the limits $`ϵ_{max}0`$ and $`\lambda _{max}0`$, the r.h.s. of eq.(17) is exact!
4. Discussion
The strategy can be elaborated along the following lines:
(1) Generalizations concerning the interactions of the qubits themselves, likewise the self-Hamiltonians of Q, B, and E.
(2) Existence of unknown part of the bath; i.e., that E interacts with $`B_1`$, but not with $`B_2`$ (the real bath then would be $`B=B_1+B_2`$)
(3) Extension of the (well known) bath $`B`$, if it is not sufficiently ”macroscopic” as to provide us with the robust states
(4) Existence of the common bath for all the qubits
(5) Avoiding the interaction of Q and E in the realistic situations
(6) Considerations of the ”classical environment(E)” , inculding the ”mixed” initial states of B and E.
As regards the points (1)-(4), the results are encouraging.
The work is in progress as regards the points (5,6).
The method has been successfully employed in the simplified model of the solid-state-nuclear quantum computer proposed by Kane. The details will be presented elsewhere.
5. Conclusion
The decoherence in quantum computers can be, at least in principle, suppressed. The idea is to properly, strongly ”press” the qubit(s)’s bath, and to produce: the stochastic change of the initial state of Q+B represents an improbable event (in the limit $`c/C0`$ it is stochastically impossible), and for sufficiently long time interval this state remains unchanged. So, one may say that we use decoherence (on B), to combat decoherence (on Q).
In practical realizations one should try to choose the interaction of B and E which should be very strong ($`Cc`$), and such the initial state of B so as to one may state $`|p0j|\widehat{H}_{QB}|p0j|1`$. These choices are really a matter of the particular model ! The preliminarysuccess with the solid-state-nuclear computer is encouraging.
After the interval $`\tau `$, the correlations between Q and E become effective, leading to decoherence presented by:
$$\rho _{Qpp^{}}=C_pC_p^{}^{}z_{pp^{}}(t),$$
where (compare to eq.(2a)):
$$z_{pp^{}}(t)=\underset{j}{}|\beta _j|^2\mathrm{exp}(ıt(\lambda _{p0j}\lambda _{p^{}0j})/\mathrm{})$$
P. Zanardi and M. Rasetti, Phys. Rev. Lett. 79 (1997) 3306; A.M. Steane, Phys. Rev. Lett. 77 (1996) 793; L. Viola and S. Lloyd, Phys. Rev. A58 (1998) 2733; and references therein
W.H. Zurek, Phys. Rev. D 26, 1862 (1982)
M. Dugić, Physica Scripta 53, 9 (1996); Physica Scripta 56, 560 (1997)
W.H. Zurek, Prog. Theor. Phys. 89, 281 (1993)
W.H. Zurek, Phys. Today, October 1991, p. 26
R. Omnés, 1994. ”The Interpretation of Quantum Mechanics”, Princeton University Press, Princeton, New Jersey
A. Messiah, ”Quantum Mechanics”, North-Holland Publishing Co., Amsterdam, 1961 (Vol. II, Chap. XVI.17); T. Kato, Prog. Theor. Phys. 4, 154 (1949); C. Bloch, Nucl. Phys. 6, 329 (1958)
P. Grigolini, ”Quantum mechanical irreversibility and measurement”, World Scientific, Singapore, 1993
B.E. Kane, Nature 323, 153 (1998) |
no-problem/0001/astro-ph0001545.html | ar5iv | text | # The Distance to the Large Magellanic Cloud
## 1. Distance to the LMC – Controversy and New Determinations
The Hubble constant, $`H_0`$, is one of the most important cosmological parameters. There are two major paths to determine $`H_0`$. The more elegant one is non-local and based on observations of the high-redshift Universe. Modeling of gravitational lensing of quasars and observations of the Cosmic Microwave Background both belong to this group of methods. The second path goes through a determination of distances and recession velocities of objects with motions dominated by the Hubble linear expansion. The distances, which are harder to measure than recession velocities, are determined based on a distance-ladder approach. The Magellanic Clouds, and especially the Large Magellanic Cloud (LMC), play a major role in this treatment. Almost all of the extragalactic distance scale is known only relative to the LMC (Madore et al. 1999). Therefore, it is absolutely essential to establish a reliable distance to the LMC ($`d_{\mathrm{LMC}}`$). This can be achieved only with an understanding of the systematics inherent in the standard candles used for the distance determinations.
For many years now there has been a division between the so called “short” and “long” distance scales to the LMC. Currently, the measured values of $`d_{\mathrm{LMC}}`$ span a range of over 25% (see e.g., Feast & Catchpole 1997; Stanek, Zaritsky, & Harris 1998).
Distances are measured based on a general formula
$$\mu _{\mathrm{LMC}}=XM_XA_X$$
(1)
where $`X`$ is an apparent magnitude, $`M_X`$ is an absolute magnitude, and $`A_X`$ is an extinction in the band of the observations. The uncertainties in all ingredients present in equation (1) may compromise the final answer. Here I am going to concentrate on three new or revised methods that cluster consistently around a short distance to the LMC:
Paczyński & Stanek (1998) pointed out that clump giants should constitute an accurate distance indicator. Udalski et al. (1998b) and Stanek et al. (1998) applied the clump method and found a very short distance to the LMC ($`\mu _{\mathrm{LMC}}\genfrac{}{}{0pt}{}{{}_{}{}^{}{}_{_<}{}^{}}{{}_{}{}^{}{}_{}{}^{^{}}}18.2`$). In response, Cole (1998) and Girardi et al. (1998) suggested that clump giants are not standard candles and that their absolute $`I`$ magnitudes, $`M_I(\mathrm{RC})`$, depend on the metallicity and age of the population. Udalski (1998a, 1998b) rejected this criticism by showing that the metallicity dependence is at a low level of about $`0.1`$ mag/dex, and that $`M_I(\mathrm{RC})`$ is approximately constant for cluster ages between 2 and 10 Gyr. Recent developments (see e.g., Popowski 2000 for a short review) suggest that the absolute character of $`M_I(\mathrm{RC})`$ is a major systematic uncertainty in this method.
Popowski & Gould (1999) determined the absolute magnitude of RR Lyrae stars, $`M_V(\mathrm{RR})=0.71\pm 0.07`$ at $`[\mathrm{Fe}/\mathrm{H}]=1.6`$, from the statistical parallax, cluster kinematics and trigonometric parallax methods. When this result is coupled with the LMC RR Lyrae photometry of Udalski et al. (1999) and Walker (1992), one obtains $`\mu _{\mathrm{LMC}}18.30\pm 0.08`$. The value of $`M_V(\mathrm{RR})`$ remains the main uncertainty of this determination.
Guinan et al. (1998) solved the eclipsing binary HV2274 and obtained various stellar parameters and the distance to the LMC. The spectra used for this purpose did not extend far enough toward long wavelengths, and the $`B`$ and $`V`$ photometry was needed to break the degeneracy between the reddening and the shape of the extinction curve. With Udalski et al. (1998c) photometry, Guinan et al. (1998) obtained $`\mu _{\mathrm{LMC}}=18.30\pm 0.07`$. Application of Nelson et al. (2000) photometry would result in $`\mu _{\mathrm{LMC}}=18.40\pm 0.07`$. The reddening constitutes a major uncertainty.
In two out of three cases the absolute magnitudes of distance indicators are under debate. Due to a huge number of possible environments, it is very hard to prove the standard character of a given candle. However, it should be possible to check whether other stellar characteristics of a candle behave in a predictable fashion. To follow this suggestion, I will concentrate on the stars in the Galactic bulge.
## 2. The Mystery of Anomalous Colors in the Galactic Bulge
Paczyński (1998) tried to explain why the clump giants in the Baade’s Window have $`(VI)_0`$ colors which are approximately $`0.2`$ mag redder than in the solar neighborhood. He could not find any satisfactory answer. Stutz, Popowski, & Gould (1999) found a corresponding effect for the Baade’s Window RR Lyrae stars, which have $`(VI)_0`$ redder by about 0.17 than their local counterparts. The similar size of the color shift in RR Lyrae stars and clump giants suggests common origin. The bulge RR Lyrae stars and clump giants both burn Helium in their cores, but similarities end there. RR Lyrae stars pulsate, clump giants do not. RR Lyrae stars are metal-poor, clump giants are metal-rich. RR Lyrae stars are a part of an axisymmetric stellar halo, whereas clump giants form a bar. Stutz et al. (1999) suggested that the very red $`(VI)_0`$ of the bulge RR Lyrae stars might have resulted from an unusual abundance of $`\alpha `$ elements. Why should a clump population which emerged in a different formation process share the same property?
The presence of the same type of color anomaly for different types of stars suggests that the effect might be unrelated to the physics of those stars. The investigated RR Lyrae and clump giants share two things in common. First, photometry of both types of stars comes from the OGLE, phase-I, project. Indeed, Paczyński et al. (1999) showed that OGLE-I $`V`$-magnitudes are 0.021 mag fainter, and $`I`$-magnitudes 0.035 mag brighter than better calibrated OGLE-II magnitudes. Therefore, the correct $`(VI)`$ colors should be 0.056 bluer. Additionally, the new $`(VI)_0`$ from the more homogeneous Baade’s Window clump is bluer than Paczyński’s & Stanek’s (1998) color, even when reduced to OGLE-I calibration. As a result, the $`(VI)_0`$ anomaly shrinks and the remaining shift amounts to $`0.11`$ for both the RR Lyrae stars and clump giants. Second, Paczyński (1998) and Stutz et al. (1999) use the same extinction map (Stanek 1996) and the same coefficient of conversion from visual extinction $`A_V`$ to a color excess $`E(VI)`$. The absolute values of the $`A_V`$s are probably correct, because the zero-point of the extinction map was determined from the $`(VK)`$ color, and $`A_V/E(VK)`$ is very close to 1. However, $`R_{VI}=A_V/E(VI)`$ is not as secure, and has a pronounced effect on the obtained color.
The value and variation of $`R_{VI}`$ was thoroughly investigated by Woźniak & Stanek (1996). The essence of the Woźniak & Stanek (1996) method to determine differential extinction is an assumption that regions of the sky with a lower surface density of stars have higher extinction. This is quite a natural expectation, as far as the density of the underlying true population of stars does not depend on location. However, it is not obvious a priori how to convert a certain density of stars to an amount of visual extinction. Therefore, Woźniak & Stanek (1996) used clump giants to calibrate their extinction. To make a calibration procedure completely unbiased would require, among other things, that the $`V`$-magnitudes of clump giants do not depend on their color \[here $`(VI)_0`$\], that reddened and unreddened clump giants be drawn from the same parent population, and that clump giants were selected without any assumption about $`R_{VI}`$. None of those is true (for details see Popowski 2000).
Because the smaller selective extinction coefficient is not excluded by the current studies, I will assume $`R_{VI}=2.1`$ to match the $`(VI)_0`$ colors of the bulge with the ones in the solar neighborhood. The color is a weak function of \[Fe/H\], so this procedure is justified because the metallicities of the bulge and solar neighborhood are similar. The change in $`R_{VI}`$ from 2.5 to 2.1 will decrease the $`I`$-mag extinction by 0.11 mag, and increase the clump-based distance to the Galactic center by the same amount.
## 3. Recalibration of Clump Giant Stars
How do the bulge results bear on the distance to the LMC? The better photometry from Paczyński et al. (1999) and the modification of $`R_{VI}`$ influence the relative RR Lyrae and clump distances to the Galactic center. Thus, the $`M_I(\mathrm{RC})`$ – \[Fe/H\] relation for clump giants used in the LMC, which was calibrated with respect to the baseline provided by RR Lyrae stars (Udalski 1998a), changes. Assuming linearity of $`M_I(\mathrm{RC})`$ – \[Fe/H\] and making some small adjustments to \[Fe/H\] used by Udalski (1998a), I find:
$$M_I(\mathrm{RC})=0.23+0.19[\mathrm{Fe}/\mathrm{H}],$$
(2)
with a slope 0.10 mag/dex steeper than the original result. Such adjustment increases the best clump giant estimate from $`\mu _{\mathrm{LMC}}=18.18\pm 0.06`$ to $`\mu _{\mathrm{LMC}}=18.27\pm 0.07`$. When distance estimates from the red clump, RR Lyrae stars and eclipsing binary HV2274 are combined, one obtains $`\mu _{\mathrm{LMC}}=18.31\pm 0.04`$. The $`1\sigma `$ uncertainty of this determination is only a formal error. The systematic errors are likely to dominate the true uncertainty. However, if the methods presented do not suffer from severe biases, then a distance to the LMC as long as 52-55 kpc ($`\mu _{\mathrm{LMC}}`$ 18.6-18.7) is highly disfavored by the current results.
## References
Cole, A. A. 1998, ApJ, 500, L137
Feast, M. W. & Catchpole, R. M. 1997, MNRAS, 286, L1
Girardi, L., Groenewegen, M. A. T., Weiss, A., & Salaris, M. 1998, MNRAS, 301, 149
Guinan, E. F., Fitzpatrick, E. L., Dewarf, L. E., et al. 1998, ApJ, 509, L21
Madore, B. F., Freedman, W. L., Silbermann, N., et al. 1999, ApJ, 515, 29
Nelson, C. A., Cook, K. H., Popowski, P., & Alves, D. A. 2000, AJ, to appear in March 2000 issue \[astro-ph/9910172\]
Paczyński, B. 1998, Acta Astron., 48, 405
Paczyński, B. & Stanek, K. Z. 1998, ApJ, 494, L219
Paczyński, B., Udalski, A., Szymański, M., Kubiak, M., Pietrzyński, G., Soszyński, I., Woźniak, P., & Żebruń, K., 1999, Acta Astron., 49, 319
Popowski, P. 2000, ApJ, 528, L9
Popowski, P. & Gould, A. 1999, in Post-Hipparcos Cosmic Candles, ed. A. Heck & F. Caputo (Dordrecht: Kluwer), 53 \[astro-ph/9808006\]
Stanek, K. Z. 1996, ApJ, 460, L37
Stanek, K. Z., Zaritsky, D., & Harris, J. 1998, ApJ, 500, L141
Stutz, A., Popowski, P., & Gould, A. 1999, ApJ, 521, 206
Udalski, A. 1998a, Acta Astron., 48, 113
Udalski, A. 1998b, Acta Astron., 48, 383
Udalski, A., Pietrzyński, G., Woźniak, P., Szymański, M., Kubiak, M., & Żebruń, K. 1998a, ApJ, 509, L25
Udalski, A., Szymański, M., Kubiak, M., Pietrzyński, G., Woźniak, P., & Żebruń, K. 1998b, Acta Astron., 48, 1
Udalski, A., Szymański, M., Kubiak, M., Pietrzyński, G., Soszyński, I., Woźniak, P., & Żebruń, K. 1999, Acta Astron., 49, 201 \[astro-ph/9907236\]
Walker, A. R. 1992, ApJ, 390, L81
Woźniak, P. & Stanek, K. Z. 1996, ApJ, 464, 233
## Discussion
Darragh O’Donoghue: Your presentation is a little more than a party political propaganda talk. It will help now if you were to investigate why advocates of the long distance scale such as Feast & Catchpole (1997) are wrong.
Piotr Popowski: It is enough if I prove that I am right, and not that they are wrong. We have three reliable methods which consistently cluster around a short distance to the LMC of about 45 kpc, and there is another potentially reliable method or standard candle (Cepheids) which gives a different result. I think that we should investigate what is going on with Cepheids (it is the best if this is done by Cepheid people).
David Laney: The statistical arguments by Feast & Catchpole have been demonstrated to be correct by Koen & Laney and, using Monte Carlo simulations, by Pont. The error bar from the Monte Carlo simulations is closer to 0.15 than 0.1, however.
Piotr Popowski: This point has been raised by someone in the audience. No disagreement here, so I made no comment about it.
Giuseppe Bono: Two comments: 1. In a recent investigation by Romaniello et al. (1999) based on HST data of red clump stars, they found a distance modulus for LMC which seems to support the long distance scale. 2. Evolutionary models suggest that the RR Lyrae luminosity decreases with increasing metallicity.
Géza Kovács: When you talk about distance of the LMC based on RR Lyrae stars you should also consider RRd stars. Applying them as distance indicators for the LMC yields $``$ 18.5 mag for the distance modulus (Kovács & Walker 1998). This is in agreement with the Cepheid B–W distance scale.
Piotr Popowski: The preliminary results of reanalysis of RRd stars from the MACHO group (David Alves and I are involved in this work) indicate that RRd stars’ distance to the LMC is short as well. |
no-problem/0001/nlin0001030.html | ar5iv | text | # Incomplete approach to homoclinicity in a model with bent-slow manifold geometry
## 1 Introduction
Number of autonomous dynamical systems exhibit complex bifurcation sequences with alternate periodic-chaotic states in control parameter space. The chaotic states are usually predominant mixture of the periodic states occurring on either side of the chaotic states. These periodic states are characterized by combinations of relatively large amplitude excursions and small amplitude near harmonic oscillations of the trajectories and have been referred to as mixed mode oscillations (MMOs) in the literature. The MMOs and the accompanying complex bifurcation sequences have been observed in models and experiments in various fields of chemical kinetics , electrochemical reactions , biological systems , and in many other physical systems.
Both numerical as well as analytical studies have been carried out extensively to the explain the origin of the MMOs and the complex bifurcation sequences these systems exhibit. Even though the origin of complex bifurcation sequences and the accompanying MMOs may depend on the particular system under study, almost all proposed mechanisms suggest that these bifurcation portraits are the artefact of global bifurcations of the system . Investigations into the global nature of the bifurcations of the MMOs and the complex bifurcation sequences have shown that homoclinic bifurcations may be relevant to a wide variety of systems which display the MMOs. Shilnikov has shown that if a dynamical system possesses a homoclinic orbit which is bi-asymptotic to a saddle focus type of equilibrium set satisfying the Shilnikov condition, then there are countably infinite number of periodic solutions in the vicinity of this homoclinic orbit. The analysis also shows that in the vicinity of this homoclinic orbit, complex bifurcation sequence can be expected in the phase portrait .
Another approach to explain the complex bifurcation sequences has been through Gavrilov - Shilnikov scenario. It has been shown that systems having homoclinic tangencies to periodic solutions possess quasi-random dynamics and MMO like behavior in the control parameter space. Each of these scenarios are characterized by bifurcation diagrams obtained from the stability analysis of the homoclinic orbit and by the corresponding scaling relations involved in the approach to the homoclinicity. Apart from these studies on the homoclinic bifurcations of the continuous time systems, attempts have also been made to study the MMOs through discrete maps .
In order to understand the numerically obtained Poincare maps from such model systems and those obtained from experiments, attempts have been made to derive map structure starting from local analysis. Such attempts have been reasonably successful in the sense that the features predicted by the derived maps agree with the features of the numerically obtained Poincare maps. However, due to the very nature of global bifurcations, there is no easy way of classifying the entire complex bifurcation sequences of these systems. Hence numerical evidence plays a crucial role in classifying the dynamics of these systems.
Yet another approach to the understanding of these complex bifurcation sequences and the MMOs is based on the analysis of the structure of the slow manifold of the system of equations. A standard slow manifold structure that has been used in the study of MMOs is the $`S`$shaped structure wherein the upper and lower pleats are attractive and the middle branch is repulsive. By appropriately locating the fixed point on the upper or lower pleat, the origin of MMOs has been explained. In this construction of slow manifold, Shilnikov’s criterion is satisfied if the direction of approach of the fast variable is transverse to the slow manifold containing the fixed point. However, in many situations, including the system under study, the approach to the saddle fixed point by the fast variable is not transversal. Depending on the nature of the approach of the fast variable to the fixed point, a classification for the MMOs has been suggested by Koper et. al. as type-I and type-II corresponding to the tangential approach and transversal approach (to the slow manifold) respectively. It has also been suggested that the occurrence of MMOs and incomplete approach to homoclinicity is related to the presence of Hopf bifurcation close to the fold of the S-shaped slow manifold structure .
In this paper, we analyze the complex dynamical behavior of a model which has been introduced in the context of a type of plastic instability called the Portevin - Le Chatelier effect. The model exhibits a rich variety of dynamics such as period bubbling, doubling and the complex bifurcation sequences. Two distinct features of the model are the atypical nature of the relaxation oscillation and the MMOs. The latter exhibits partial features of Shilnikov, and Gavrilov-Shilnikov scenario and shows a incomplete approach to the homoclinic point. Our effort is focussed in understanding this issue in the context of our model. To begin with we study the nature and the origin of the relaxation oscillations by analyzing the geometry of the slow manifold which controls the relaxation oscillations. We show that the underlying cause of the relaxation oscillations is due to the atypical bent geometry of the slow manifold. This feature forms the basis of further analysis of the nature of the MMO sequences, and the incomplete approach to homoclinic bifurcation. The paper is organized as follows. In section II, for the sake of completeness, we start with a brief introduction to the phenomenon followed by a description of the model. Section III contains a detailed analysis of the complex bifurcation sequences exhibited by the model in the plane of two physically interesting parameters. Section IV contains a discussion on the origin of relaxation oscillations using the geometry of the slow manifold. Using this, we explain the origin of the MMOs and derive a scaling relation involving the maximum number of periodic orbits allowed for given value of the control parameter. The analysis helps us to understand the cause of the incomplete approach to homoclinicity. We conclude the paper with discussion and conclusions in section V.
## 2 A Dynamical Model for Jerky Flow
Since the model is rooted in the area of plastic instability, for the sake of completeness, we start with a brief introduction to the phenomenon. The Portevin-Le Chatelier (PLC) effect is a plastic instability manifesting when specimens of metallic alloys are deformed under tensile deformation. Under normal conditions, the stress-strain curve is smooth. However, repeated yield drops occur when the material parameters are in the regime of instability. Each of the load drops is related to the formation and propagation of dislocation bands . The PLC effect (or the jerky flow) is seen in several metallic alloys such as commercial aluminium, brass, alloys of aluminium and magnesium . The phenomenon is observed only in a window of strain rates and temperature. It is generally agreed that the microscopic cause of the instability is due to the interaction of dislocations with mobile point defects. This leads to the negative strain rate characteristic of the yield stress. The basic idea was formulated by Cottrell few decades ago. However, this model and its extensions do not deal with the time dependent nature intrinsic to the phenomena.
The first dynamical description was attempted Ananthakrishna and coworkers several years ago . The basic idea of the model is that most of the generic features of the PLC effect stem from nonlinear interactions between defect populations. The model in its original form does introduce spatial dependence of specific nature. However, further analysis of the model ignores the details of spatial inhomogeneous structure. The model consists of three types of dislocations and some transformations between them. The model has proved to be very successful in that it could explain most of the experimentally observed features such as the existence of bounds on the strain rate for the PLC effect to occur, the negative strain rate sensitivity, etc.,. Several aspects of the model has been investigated .
A few comments may be in order regarding the spatial aspect of the PLC effect. The nature of spatial terms that should be introduced in the description of the PLC effect has been a controversial topic. However, there is some consensus that the double cross slip mechanism plays an important role in the spatial aspect. In the above model, justification for ignoring the spatial inhomogeneous structure and considering only the temporal aspects of the phenomenon is that the variables (dislocation densities) correspond to the collective degrees of freedom of the spatially extended systems. But, if one were to be interested in spatial aspects directly, we refer the reader to an improved version of the model. Further work to improve the model by including the nonlocal effects of immobile density along with the cross slip term is under active investigation.
One important prediction of the model is that the phenomena should be chaotic in a certain regime of applied strain rate. This prediction has been verified by analyzing stress signals obtained from single and polycrystalline samples . Further, the number of degrees of freedom required for a dynamical description of the phenomenon estimated from the analysis was found to be four or five consistent with that envisaged in the model. Moreover, since the physical system is spatially extended, this reduction to few degrees of freedom does suggest that these modes correspond to collective degrees of freedom of the participating defects. From this point of view, dealing with the temporal aspect appears to be justified. Therefore, it is natural to investigate the chaotic behavior of the model in its own right. Some preliminary results on the chaotic aspects of the model have been published earlier .
The model consists of mobile dislocations and immobile dislocations and another type which mimics the Cottrell’s type, which are dislocations with clouds of solute atoms . Let the corresponding densities be $`N_m`$, $`N_{im}`$ and $`N_i`$, respectively. The rate equations for the densities of dislocations are:
$`\dot{N}_m`$ $`=`$ $`\theta V_mN_m\beta N_m^2\beta N_mN_{im}+\gamma N_{im}`$ (1)
$`\alpha _mN_m`$
$`\dot{N}_{im}`$ $`=`$ $`\beta N_m^2\beta N_{im}N_m\gamma N_{im}+\alpha _iN_i,`$ (2)
$`\dot{N}_i`$ $`=`$ $`\alpha _mN_m\alpha _iN_i.`$ (3)
The overdot, here, refers to the time derivative. The first term in Eq. (1) is the rate of production of dislocations due to cross glide with a rate constant $`\theta `$. $`V_m`$ is the velocity of the mobile dislocations which in general depends on some power of the applied stress $`\sigma _a`$. The second and third term refer to annihilation or immobilization processes. The fourth term represents the remobilization of the immobile dislocations due to stress or thermal activation (see the loss term $`\gamma N_{im}`$ in Eq. 2). The last term represents the immobilization of mobile dislocations either due to solute atoms or due to other pinning centers. $`\alpha _m`$ refers to the concentration of the solute atoms which participate in slowing down the mobile dislocations. Once a mobile dislocation starts acquiring solute atoms we regard it as a new type of dislocation, namely the Cottrell’s type $`N_i`$. This process is represented as a gain term in Eq. (3). As they acquire more and more solute atoms they will slow down and eventually stop the dislocation entirely. At this point, they are considered to have transformed to $`N_{im}`$. This process has been represented by the loss term in Eq. (3) and a gain term in Eq. (2).
These equations should be dynamically coupled to the machine equations describing the rate of change of the stress developed in the sample. This is given by
$`\dot{\sigma _a}=\kappa (\dot{ϵ_a}B_0N_mV_m),`$ (4)
where $`\kappa `$ is the effective modulus of the system, $`\dot{ϵ}_a`$ is the applied strain rate, $`V_m`$ is the velocity of mobile dislocations and $`B_0`$ is the Burgers vector. These equations can be cast into a dimensionless form by using the scaled variables
$`x=N_m\left({\displaystyle \frac{\beta }{\gamma }}\right),y=N_{im}\left({\displaystyle \frac{\beta }{\theta V_0}}\right),`$
$`z=N_i\left({\displaystyle \frac{\beta \alpha _i}{\gamma \alpha _m}}\right),\tau =\theta V_0t,\text{ and }\varphi =\left({\displaystyle \frac{\sigma _a}{\sigma _0}}\right).`$
Using the power law dependence $`V_m=V_0(\frac{\sigma _a}{\sigma _0})^m`$, Eqs. (1-3) and (4) can be rewritten as
$`\dot{x}`$ $`=`$ $`\varphi ^mxaxb_0x^2xy+y,`$ (5)
$`\dot{y}`$ $`=`$ $`b_0\left(b_0x^2xyy+az\right),`$ (6)
$`\dot{z}`$ $`=`$ $`c(xz),`$ (7)
$`\dot{\varphi }`$ $`=`$ $`d\left(e\varphi ^mx\right),`$ (8)
Here $`a=(\alpha _m/\theta V_0),b_0=(\gamma /\theta V_0),c=(\alpha _i/\theta V_0)`$, $`\kappa =(\theta \beta \sigma _0d/\gamma B_0)`$ and $`e=(\dot{ϵ}_a\beta /B_0V_0\gamma )`$. For these set of equations there is only one steady state which is stable. There is a range of the parameters $`a,b_0,c,d,m`$ and $`e`$ for which the linearized equations are unstable. In this range $`x,y,z`$ and $`\varphi `$ are oscillatory.
Among these physically relevant parameters, we report here the behavior of the model as a function of most important parameters namely the applied strain rate $`e`$ and the velocity exponent $`m`$. We use $`e`$ as the primary control parameter for the analysis. The values of other parameters are kept fixed at $`a=0.7,b_0=0.002,c=0.008,d=0.0001`$ and $`k=1.0`$. The present choice of parameters does not necessarily correspond to a realistic experimental situation, although there is a range of allowed values. As can be verified these equations exhibit a strong volume contraction in the four dimensional phase space. We note that there are widely differing time scales corresponding to $`a,b_0,c`$ and $`d`$ (in the decreasing order) in the dynamics of the model. For this reason, the equations are stiff and numerical integration routines were designed specifically to solve this set of equations. We have used a variable order Taylor series expansion method as the basic integration technique with coefficients being determined using a recursive algorithm. Most of the bifurcation analyses were performed using these indigenous routines. AUTO software package was used exclusively for two parameter continuation of bifurcation points.
## 3 Summary of bifurcation exhibited the model
In an attempt to understand the complex bifurcation sequences exhibited by the model, we start with an outline of the gross features of the phase diagram in the $`(m,e)`$ plane shown in Fig. 1. In our discussion, we consider $`m`$ to be the unfolding parameter. For values of $`m>m_d6.8`$, the equilibrium fixed point of the system of equations is stable. At $`m=m_d`$, we have a degenerate Hopf bifurcation as a function of $`e`$. For values less than $`m_d`$, we have a back-to-back Hopf bifurcation. The periodic orbit connecting these two Hopf bifurcations is referred to as the Principal Periodic Orbit (PPO). The dynamics of the system is essentially bounded by these two Hopf bifurcations. In Fig. 1, the broken line represents the Hopf bifurcation and the dotted lines correspond to the first three successive period doubling bifurcations leading to period 2, 4 and 8 orbits. The region in between the first period two and the Hopf bifurcation line exhibits monoperiodic relaxation like oscillations.
The PPO for most of the parameter plane $`(m,e)`$ is born through a subcritical Hopf bifurcation leading to relaxation like oscillations. However, the narrow region between the Hopf bifurcation line (corresponding to large values of $`e`$) and the period two regime is characterized by small amplitude nearly symmetric coplanar limit cycles. The complex bifurcation sequences, characterized by alternate periodic-chaotic sequences seen in the parameter space, are roughly indicated by the hatched region. Since the part of the hatched region extends beyond the outermost period doubling line (large values of $`e`$ and small values of $`m`$), both the MMOs and the small amplitude monoperiodic limit cycle solutions coexist in this region. (See also Fig. 3.) A codimension two bifurcation point in the form of a cusp (shown as a filled diamond) at $`(e_c,m_c)`$ formed by merging of the locus of two saddle node periodic orbits (represented by bold lines) of the PPO is also displayed in Fig. 1. Apart from these bifurcations, we failed to detect any other bifurcation or equilibrium set in the phase space. We will deal with each of these regions in detail below.
Bifurcation diagrams have been obtained by plotting the maxima of any one of the variables $`x`$,$`y`$,$`z`$ or $`\varphi `$ as a function of the control parameters ($`e,m`$). We have mostly shown the bifurcation diagrams in the variable $`x`$. This choice enables a good visual representation of the bifurcation diagram since the maxima of the $`x`$ variable is quite large compared to other variables. Based on the nature of the bifurcation sequences, the parameter space can be broadly grouped into two regions, viz. $`m2.0`$ and $`m<2.0`$, and we will discuss the changes as a function of $`e`$ fixing $`m`$ at a particular value.
### 3.1 Region $`m2.0`$
We briefly summarize the results for this region. For values of $`m>2.16`$, the bifurcation diagrams are characterized by an incomplete period doubling cascades followed by reverse period doubling bifurcations, displayed as nested bubbles of periodic states (see Fig. 2). As $`m`$ decreases, the number of periodic bubbles nested in the structure increases as $`2^n`$, with $`n\mathrm{}`$, culminating in chaos. Just below $`m=2.16`$, the disjoint chaotic bubbles collide with each other forming an extended attractor which has been referred to as ’bubble bursting’ in the literature. Similar features have been observed with $`m`$ as the control parameter keeping $`e`$ fixed at an appropriate value. The rates of the period doubling (PD) bifurcations as well as the reverse period doubling bifurcations with respect to $`e`$ and $`m`$ fall close to the value of the Feigenbaum’s constant for quadratic unimodal maps, namely $`\delta _F=4.66`$. When the value of $`m`$ reaches a critical value $`m=2.11`$, a period three cycle is born through a saddle node bifurcation and has the largest width (in $`e`$) for $`m2.0`$ in this regime.
### 3.2 Region $`m<2.0`$
For $`m<2.0`$, the system exhibits qualitatively different behavior compared to $`m>2.0`$, in the sense that higher order MMOs emerge gradually as $`m`$ is decreased. In this region, the system exhibits complex bifurcation sequences or the alternate periodic-chaotic sequences which are characteristic to this system. The stable periodic orbits in the bifurcation sequence typically exhibit MMO nature and they are labelled by $`L^s`$, where $`L`$ is the number of large amplitude loops and $`s`$ is the number of small amplitude loops of the periodic orbit. These MMOs are heralded by the creation of the period three ($`L^s:1^2`$) region after the PD cascade to chaos in the bifurcation plot. To illustrate the nature of bifurcations in this region, we fix the unfolding parameter at $`m=1.8`$ and discuss the bifurcation with respect to $`e`$.
Figure 3 shows the bifurcation sequence with alternating chaotic and periodic states along with the higher order periodic isolas (isolated bifurcation curves). The unstable periodic orbits are shown by dashed lines. In the case of isolas, we have shown only the largest amplitude isola for any given periodicity. The first instability of the PPO through a PD bifurcation opens up a period doubled orbit having a large parameter width in $`e`$. This feature persists for the entire $`m_c<m<2.0`$ regime. As in the case of $`m=2.0`$, a period three isola is born through a saddle node (SN) bifurcation from the chaotic attractor. In Fig. 3, stable periodic orbits are shown to be bounded between PD bifurcation (filled circle) and the SN bifurcation (filled triangle). The sequential way SN and PD bifurcations arrange themselves to form an isola can be easily understood by considering the behavior of the Floquet eigenvalue of the periodic orbit. The period three orbit is born in a SN bifurcation accompanied by the disappearance of the first chaotic window attractor. At this value of $`e`$, Floquet eigenvalue is at $`+1`$ creating a pair of stable and unstable period three orbits. As $`e`$ is increased, the eigenvalue of the unstable periodic orbit increases beyond $`+1`$ while the eigenvalue corresponding to the stable orbit keeps decreasing and crosses $`1`$ resulting in a PD bifurcation. Due to the isola structure, further increase in $`e`$ makes this eigenvalue cross back $`1`$ thus restabilizing the period three orbit. For any further increase in $`e`$, the Floquet eigen values of the stable and unstable orbit merge again at $`+1`$ and vanish in another SN bifurcation to complete the isola structure. The next higher order isola is also created through a SN resulting in the disappearance of the chaotic attractor born from the destabilisation of stable period three orbit. Higher order periodic orbits (isolas) are formed in a similar way. Note that these isolas are independent of the PPO. We refer to this sequence of periodic orbits of this form as the principal period adding sequence (PPAS) or the principal Farey sequence. In the MMO notation, the PPAS can be written as $`L^s:1^n`$ where $`n=2,3,4,\mathrm{}`$. As the periodicity increases, the width of the isolas in $`e`$ decreases. Since the isolas are independent of the PPO, any change in the stability of the PPO has no effect on the nature of the sequence. This is evident from the bifurcation diagram where higher periodic orbits (isolas) are seen even after stability of the PPO is reestablished. The restabilisation of the PPO is through a reverse period doubling cascade from chaos. This chaotic segment formed from the reverse period doubling of the PPO expands in an interior crisis due to its collision with an unstable periodic orbit of the next higher period isola (period three in the case of $`m=1.8`$). This crisis point, shown by an arrow, marks the lower boundary of the multistability region in the bifurcation diagram.
The inset of Fig. 1 shows the expanded region of interest of the phase diagram in the $`(m,e)`$ plane. As can be seen, the locus of the SN bifurcations corresponding to the MMOs are distinct and higher period SN bifurcation curves cross the lower ones resulting in the period $`n`$ isola extending beyond the period ($`n1`$) isola. Moreover, the region where SN bifurcation curves overlap with the region of the period doubling curves, multistability regions in the parameter space $`(m,e)`$ are created. This is clearly seen in Fig. 3 where bifurcation diagram for $`m=1.8`$ is shown. (See also Fig. 4.) Typically, the same mechanism as described for the case of $`m=1.8`$ operates for the period adding sequences in the region $`2.0>m>m_c`$, where $`m_c1.1`$ is the value of $`m`$ at the cusp point. As $`m`$ decreases from $`m=2.0`$, higher number of stable periodic windows are accommodated with concomitant decrease in the width of the chaotic regimes separating the periodic windows. The arithmetically increasing periods of the orbits going from left to right form an incomplete period adding sequence with decreasing widths for higher order periodic windows. These features are shown in the bifurcation diagrams for $`m=1.8`$, and $`m=1.2`$ in Fig. 3 and Fig. 4 respectively.
Below $`m_c`$, the bifurcation diagram is even more interesting and rich. Here, we outline the features related to the Farey states. For the case $`m=1.0`$, only three principal Farey states denoted by $`L^s`$, $`s=1,2`$ and $`3`$ survive, as shown in Fig. 5. The well developed sub-Farey sequences are also shown in the inset of Fig. 5. The sub-Farey states created go from right to left in contrast to the principal Farey states (see Fig. 5). All these sub-Farey sequences culminate in a SN bifurcation. While in the first bifurcation of the PPO (SN bifurcation), the transition is from $`1^0\mathrm{}^1`$, the mid region of parameter accommodates both the large amplitude and small amplitude solutions with nearly equal weights. Towards the end region of $`e`$, we find no fine structure typical of the first two principal chaotic windows.
## 4 Mechanism of relaxation oscillations
One characteristic feature of the dynamics of the system is its strong relaxation nature. This is seen even in the case of the mono-periodic solutions emerging from the Hopf bifurcation for small values of $`e`$. This feature, of course persists for other regions of the $`(m,e)`$ plane where the MMO type of oscillations are also seen. These two aspects are interrelated and are a result of the structure of the slow manifold as we will show. Since our system does not follow the known homoclinic scenarios, we look for a new mechanism for the MMOs based on the mechanism proposed for the relaxation oscillations.
### 4.1 Relaxation oscillations
Relaxation oscillations are highly nonlinear oscillations with large amplitude excursions of the fast variable. These oscillations arise as a consequence of the existence of a fast time scales compared to the time scales of other variables in the dynamics of the system. The relaxation oscillations have been an intense area of research in the context of biological rhythms. The relaxation oscillations that manifest in the model under study is a type of relaxation oscillation wherein the fast variable takes on large values for a short time after which it assumes small values of the same order of magnitude as that of the slow variables. The time spent by the fast variable in the part of phase space where the amplitude of the fast variable is small is a substantial portion of the period of the orbit. It is this type of relaxation oscillation that is dominantly seen in our system, even though other types of relaxation oscillations are also seen for certain other regimes of the $`(m,e)`$ plane. We shall refer to this type of relaxation oscillation as pulsed type relaxation (PTR) oscillation. A typical plot of $`x(t)`$ is shown in inset of Fig. 6 for $`e=200.0`$ and $`m=1.2`$.
In the case when two disparate time scales are present in the dynamics, using multiple scale perturbative analysis, Baer and Ernaux have shown that Hopf bifurcation can lead to relaxation type of oscillations. They have shown that nearly sinusoidal solutions born out of the Hopf bifurcation change over to relaxation like oscillations in a small region of the value of the control parameter. In such a case, the crossover to relaxation oscillations is confined to the slow manifold around the fixed point. As we will see, the nature of the relaxation oscillation in our model is very different from that discussed by Baer and Ernaux. Here, it suffices to say that the PTR is a result of the evolution wherein the trajectories visiting the slow manifold region around the fixed point are pushed out to another part of the slow manifold away from the fixed point where trajectories spend considerable fraction of its period.
To understand the nature of the relaxation oscillations, we first study the structure of the slow manifold ($`S`$) and the behavior of the trajectories visiting different regions of $`S`$. Consider the slow manifold given by
$`\dot{x}=g(x,y,\varphi )`$ $`=`$ $`b_0x^2+x\delta +y=0`$ (9)
with $`\delta =\varphi ^mya`$. Here, the slow variables $`y`$ and $`\varphi `$ (and therefore $`\delta `$) are regarded as parameters. Further, as we will see below, it is simpler to deal with the structure of the slow manifold in terms of the $`\delta `$ instead of both $`y`$ and $`\varphi `$. Then, the physically allowed solution of the above equation is
$$x=\frac{\delta +\sqrt{\delta ^2+4b_0y}}{2b_0}$$
(10)
where $`\delta `$ can take on both positive and negative values. Noting that $`b_0`$ is small and therefore $`\delta ^24b_0y`$, two distinct cases arise corresponding to $`\delta >0`$ and $`\delta <0`$ for which $`x\delta /b_0`$ and $`xy/\delta `$ respectively. Further, since the slow variable $`\varphi `$ and $`y`$ take on values of the order of unity, the range of $`\delta =\delta (y,\varphi )`$ is of the same order as that of $`\varphi `$ and $`y`$ (as is evident from Figs. 6 and 7). Thus, we see that $`xy/\delta `$ is small and $`x\delta /b_0`$ is large. For values around $`\delta =0`$ and positive, we get $`x\left(y/b_0\right)^{1/2}`$.
Let $`S_1`$ denote the region of slow manifold values of $`x`$ corresponding to $`\delta >0`$ and $`S_2`$ the region of slowmanifold values of $`x`$ for $`\delta <0`$. The bent-slow manifold structure along with the two portions of the slow manifold are shown by a bold lines in the $`(x,\delta )`$ plane in Fig. 6. A local stability analysis for points on $`S_1`$ and $`S_2`$ shows that $`g/x=\delta 2b_0x`$ is negative. This implies that the rate of growth of $`x`$ is damped and hence these regions, $`S_1`$ and $`S_2`$ will be referred to as attracting. In Fig. 6, we have shown a trajectory corresponding to a mono-periodic relaxation oscillation ($`m=1.2`$ and $`e=200.0`$) by a thin line. As can be seen, the trajectory spends most of the time on $`S_1`$ and $`S_2`$. For points below the line $`2b_0x=\delta `$ ($`\delta >0`$), $`g/x>0`$ implying a positive rate of growth of the $`x`$ variable and hence we call this region as repulsive or ’unstable’ (shaded region of Fig. 6). We stress here that this region is not a part of the slow manifold. Even then, the trajectory starting on $`S_2`$ does continue in the direction of increasing $`\delta `$ beyond $`\delta =0`$. Once the trajectory is in this region, it moves up rapidly in the $`x`$ direction (due to the ‘unstable’ nature) until it reaches $`x=\delta /2b_0`$ line, thereafter, the trajectory quickly settles down on to the $`S_1`$ part of the slow manifold due to the fact $`g/x`$ becomes negative. As the trajectory descends on $`S_1`$ approaching $`S_2`$, we see that the trajectory deviates away from $`S_1`$. This happens when the value of $`x`$ is such that $`2b_0x<\delta `$, i.e., $`g/x>0`$. Thus, points on $`S_1`$ satisfying this condition are locally unstable. (For points in this neighborhood $`\delta 0.2,x50`$.) Thus, the trajectory makes a jump from $`S_1`$ to $`S_2`$ in a short time. This roughly explains the origin of the relaxation oscillation in terms of the reduced variables $`\delta `$ and $`x`$.
The actual dynamics is in a higher dimensional space and a proper understanding will involve the analysis of the movement of the trajectory in the appropriate space. Moreover, quite unlike the standard $`S`$ shaped manifold with upper and lower attracting pleats with the repulsive (unstable) branch, in our model, both branches of the bent-slow manifold are connected, and there is no repulsive branch of the slow manifold. Thus, the mechanism of jumping of the orbit from $`S_2`$ to $`S_1`$ is not clear. In order to understand this, consider a 3-d plot of the trajectory shown in Fig. 7. Retaining the same notation for the 3-d regions of the slow manifold as that used for the $`x\delta `$ plane, regions $`S_1`$ and $`S_2`$ are shown in Fig. 7. As can be seen, the region $`S_2`$ corresponding to small values of $`x`$ lies more or less on the $`y\varphi `$ plane and the region $`S_1`$ corresponding to large values of $`x`$ is nearly normal to the $`y\varphi `$ plane due to the large $`b_0^1`$ factor. (Note that the scales of $`y`$ and $`\varphi `$ are the smallest for the system. ) Regions $`S_1`$ and $`S_2`$ are demarcated by the ‘fold curve’ which lies in the $`y\varphi `$ plane and is given by $`\delta =\varphi ^mya=0`$. As in the case of $`x\delta `$ plane, in 3-d space also, the rapidly growing nature of the trajectory seen in the approximate region below the surface of $`2b_0x=\varphi ^mya`$ and lying to right of the ‘fold curve’ is due to $`g/x>0`$.
The principal features of the relaxation oscillations that we need to explain are: a) very slow time scale for evolution on $`S_2`$, b) fast transition from $`S_2`$ to $`S_1`$ and c) evolution on $`S_1`$. As mentioned in the introduction, these are related to the slow - fast time scales. In order to understand this, we shall analyze Eqs. (6) and (8) by recasting them in terms of $`\delta `$. In the whole analysis it would be helpful to keep in mind the range of values of $`x,y,z`$ and $`\varphi `$ (shown in Figs. 6 and 7), in particular, their values as the trajectory enters and leaves $`S_1`$. Consider rewriting Eq. (6) valid on the slow manifold $`S`$ in terms of $`\delta `$:
$$\dot{y}=b_0(x\delta xy+az).$$
(11)
The idea is to study this equation along with Eq.(8) in specific regions of the phase space to understand the general features of the flow, viz., on $`S_2`$, just outside $`S_2`$, and on $`S_1`$. The presence of the $`z`$ variable in Eq. (11) poses some problems. However, it is possible to get a rough estimate of the magnitude of $`z`$ and the relative changes in the values of $`z`$ which is all that will be needed for our further discussions. To see this, consider Eq. (7) from which we see that $`z`$ follows $`x`$. Further, Once the trajectory moves out of $`S_2`$, $`x`$ changes rapidly and therefore the value of $`z`$ increases (in a relatively short interval of time), reaching its maximum value, $`z_{max}`$, just before the trajectory returns to $`S_2`$. When the trajectory is on $`S_2`$, since $`xy/|\delta |`$, from Eq. (7) we see that the value of $`z`$ is slowly decreasing (with a time constant $`c^1`$) starting from $`z_{max}`$, reaching its minimum value, say $`z_{min}`$, around the time when the trajectory leaves $`S_2`$. In other words, the magnitude of $`z`$ is maximum when the trajectory enters $`S_2`$ and minimum when it leaves $`S_2`$. Further, we note that $`x=y/\delta x_0=z_0e/2`$ and $`z`$ oscillates around its equilibrium value $`z_0`$. Thus, the values of $`z_{max}`$ and $`z_{min}`$ are larger than the range of allowed values of $`y`$. (Note that this is also consistent with the fact that the time scale of $`z`$ is larger than that for $`y`$ and $`\varphi `$.)
Consider the behavior of Eq. (11) on $`S_2`$. Using the values of $`xy/\left|\delta \right|`$, we get
$$\dot{y}=b_0\left[y\frac{y^2}{\left|\delta \right|}+az\right].$$
(12)
By noting that on $`S_2`$, $`z`$ decreases from $`z_{max}`$ to $`z_{min}`$, we see that there is a range of small values of $`y`$ for which $`\dot{y}>0`$ and for relatively larger values of $`y`$, $`\dot{y}<0`$. Thus, $`y`$ clearly has a turning point on $`S_2`$ beyond which $`y`$ decreases.
Next, consider the changes in $`\varphi `$. Using the value of $`x=y/|\delta |`$ on $`S_2`$ in Eq. (8), we find that $`e`$ is much larger than $`\varphi ^my/\left|\delta \right|`$, since these variables are of the order of unity. Thus, $`\varphi `$ increases linearly, at a rate close to $`de<<1`$. Considering the fact that $`\dot{x}0`$ for the entire interval the trajectory is on $`S_2`$, the time scale of evolution of the trajectory is entirely controlled by the two slow time scales of $`y`$ and $`\varphi `$. This roughly explains the behavior of the trajectory on $`S_2`$.
Now consider the behavior of Eq. (11) in a small region just outside $`\delta =0`$. Using $`x\left(y/b_0\right)^{1/2}`$ valid for $`\delta 0`$, we get,
$$\dot{y}=b_0\left[\left(\frac{y}{b_0}\right)^{1/2}(\delta y)+az\right]b_0\left[\frac{y^{3/2}}{b_0^{1/2}}+az\right].$$
(13)
We are interested in investigating the behavior of Eq. (13), for $`z=z_{max}`$, appropriate as the trajectory approaches $`S_2`$ and $`z=z_{min}`$, appropriate as the trajectory leaves $`S_2`$. Consider the first choice $`(z=z_{max})`$ corresponding to the trajectory as it approaching $`S_2`$ from outside ($`\delta >0`$). Then, using the value of $`b_0`$, an order of magnitude calculation shows that there is a range of small values of $`y`$ for which $`\dot{y}>0`$. This implies that $`y`$ grows for small $`y`$, meaning that the trajectory moves towards $`S_2`$. Now consider using $`z=z_{min}`$ corresponding to the situation when trajectory has left $`S_2`$. Similar estimation shows that there is a range of (relatively larger) values of $`y`$ for which $`\dot{y}<0`$. This implies that $`y`$ decreases for relatively large values of $`y`$, meaning that the trajectory is moving away from $`S_2`$. ( Note that for this case, there may or may not be a range of $`y`$ for which $`\dot{y}>0`$.) Thus, in both cases the directions of growth of $`y`$ for small and large $`y`$ just outside $`S_2`$ are consistent with the behavior of $`y`$ just inside $`S_2`$. (See Fig. 7.)
Now, consider Eq. (8) with $`x\left(y/b_0\right)^{1/2}`$ valid for the region $`\delta `$ positive but small. Then,
$$\dot{\varphi }=d\left[e\varphi ^m\left(\frac{y}{b_0}\right)^{1/2}\right].$$
(14)
Keeping in mind the order of magnitude of $`b_0`$, and the fact that $`y`$ and $`\varphi `$ are of the order of unity, the magnitude of $`\varphi ^m\left(y/b_0\right)^{1/2}`$ is seen to be larger than its value on $`S_2`$. Note that a quick order of magnitude calculation shows that there are values of $`y`$ and $`\varphi `$ such that $`\varphi ^m\left(y/b_0\right)^{1/2}`$ is of the order of $`e`$ which implies that $`\varphi `$ is about to decrease and therefore is near its maximum. Moreover, if anything, $`\varphi ^mx`$ in Eq. (8) increases as the trajectory tends to moves out of $`S_2`$, since $`\dot{x}x\delta `$ just outside $`S_2`$. This implies $`\varphi `$ will eventually decrease.
Combining the results on $`\dot{y}`$ and $`\dot{\varphi }`$ for regions just outside and inside the ‘fold’, we see that the trajectory enters $`S_2`$ in the region corresponding to small values of $`y`$ and $`\varphi `$, and makes an exit for relatively larger values of $`\varphi `$ and $`y`$ (compared to their values as the trajectory enters $`S_2`$). Finally, we can see that just to the right of $`\delta =0`$ line, $`\dot{x}x\delta `$, with $`\delta `$ very small, which suggests that the time constant is small. Thus, the growth of $`x`$ is slow in the neighborhood of $`\delta =0`$, and is tangential to the $`S_2`$ plane even in the ‘unstable’ region. However, once the trajectory moves away from $`\delta =0`$, the growth of the trajectory is controlled by $`g/x`$ and hence the time scale of growth of $`x`$ is of the order of $`\delta ^1`$ which is of the order of unity. This essentially explains why the trajectory tends to move into the ‘unstable’ region and grows rapidly.
Once in the ‘unstable’ region, the value of $`x`$ continues to grow in this region of the phase space as can be seen from Eq. (8) until the value of $`x`$ is such that $`\varphi ^mx=e`$ is satisfied. Beyond this value of $`\varphi `$, $`\dot{\varphi }`$ is negative. Thus, the trajectory leaving $`S_2`$ eventually falls onto the $`S_1`$ part of the slow manifold. We can again evaluate $`\dot{y}`$ and $`\dot{\varphi }`$ just as the trajectory reaches $`S_1`$. Using $`x\delta /b_0`$ in Eq. (11), we find
$$\dot{y}=b_0\left[\frac{\delta }{b_0}(\delta y)+az\right],$$
(15)
The sign of $`\dot{y}`$ is determined by the factor ($`\delta y`$) at the point where the trajectory reaches $`S_1`$. To see the relative magnitudes of $`\delta `$ and $`y`$, consider obtaining an equation for $`\delta `$ starting from $`g(x,y,\delta )=0`$. Differentiating this and using $`\dot{x}=0`$ on for the slow manifold, we get
$$\dot{\delta }=\frac{b_0}{x}\left[x(\delta y)+az\right].$$
(16)
Using $`x\delta /b_0`$ on $`S_1`$, we see that $`y`$ is a fast variable compared to $`\delta `$. Thus, in this interval of time, we could take $`\dot{y}=0`$, i.e.,
$$\delta =y\left(\frac{b_0az}{\delta }\right).$$
(17)
Since all these variables $`\delta `$, $`y`$ and $`z`$ are positive on $`S_1`$, we see that $`y>\delta `$. (Note the factor $`b_0az/\delta `$ is small.) Using this in Eq. (15) we see that $`y`$ decreases. Now, consider the equation for $`\varphi `$. Using $`x\delta /b_0`$ on $`S_1`$, we get
$$\dot{\varphi }=d\left[e\frac{\varphi ^m\delta }{b_0}\right].$$
(18)
Noting the value of $`b_0`$, we see that $`\dot{\varphi }`$ will be negative when the trajectory reaches $`S_1`$. The time scale of evolution of $`y`$ in Eq. (15) is of the order of unity while that of $`\varphi `$ is $`d/b_0`$. These time scales are relatively fast. (These statements are true only as the trajectory hits $`S_1`$.) Moreover, since $`x`$ is a fast variable, the changes in $`x`$ component dominates the descent of the trajectory. Finally, as the trajectory approaches $`S_2`$, $`g/x`$ becomes positive and the trajectory jumps from $`S_1`$ to $`S_2`$. Combining these results, we see that the trajectory moves towards the region of smaller values of $`y`$ and $`\varphi `$ entering $`S_2`$ in a region of small values of $`y`$ and $`\varphi `$.
In summary, the sequential way the orbit visits various parts of the phase space is as follows. The trajectory enters $`S_2`$ part of the slow manifold in regions of small $`y`$ and $`\varphi `$ making an exit along $`S_2`$ for relatively large $`\varphi `$ and $`y`$. Thereafter, the trajectory moves through the ‘unstable’ part of the phase space before falling onto the $`S_1`$ and quickly descends on $`S_1`$. This completes the cyclic movement of the trajectory and explains the geometrical feature of the trajectory shuttling between these two parts of the manifold and the associated time scales.
Now, the question that remains to be answered is $``$ do the trajectories always visit both $`S_1`$ and $`S_2`$ or is there a possibility that the trajectory remains confined to $`S_1`$ ? It is clear that if the former is true, relaxation oscillations with large amplitude will occur and if the latter is true, the oscillations are likely to be of small amplitude. Here, we recall that the coordinates of the saddle focus fixed point are $`x_0=z_0e/2`$ which is much larger than the values of $`x`$ on $`S_2`$ ($`y/\left|\delta \right|`$). Thus, the fixed point located on the $`S_1`$ will be close to the ‘fold’ at the first Hopf bifurcation, $`e=e_f`$, since the latter occurs at small values of $`e`$ ($`e_f5`$). Due to the unstable nature of the fixed point, the trajectories spiralling out are forced onto the $`S_2`$ part of the manifold resulting in relaxation oscillation. This point has been illustrated by considering the example of a period eleven orbit for $`m=1.2`$ and $`e=267.0`$ shown in Fig. 8. As is clear from this diagram, the small amplitude oscillations are located on the $`S_1`$. As the small amplitude oscillations grow, the relaxation nature does not manifest until the orbit crosses over to $`S_2`$. To the best of the authors knowledge, the mechanism suggested here for pulsed type relaxation oscillations is new.
The above feature of the trajectories continuing in the same direction of the slow manifold ($`S_2`$) well into the ‘unstable’ part of the phase space is somewhat similar to canard solutions where the trajectories tend to follow the slow manifold well into the repulsive part of the slow manifold before jumping to a attracting branch . The differences, however, are clear. While in canard solutions, the trajectory tends to move along the repulsive part of the slow manifold before jumping to the attracting branch of the slow manifold, in our case, the trajectory leaves the slow manifold and moves into the ‘unstable’ part of the phase space which is not a part of the slow manifold.
### 4.2 Mixed Mode Oscillations
We now consider the origin of the MMO sequences in our model. Global bifurcation scenarios are known to be relevant to the MMOs and in the introduction, we briefly mentioned two of the possible global bifurcation scenarios which display MMO like sequences. These scenarios are based on the homoclinic contacts of an equilibrium set like the saddle focus fixed point and saddle periodic orbit for the Shilnikov and Gavrilov-Shilnikov scenarios respectively. Each of these scenarios are characterized by the bifurcation diagrams obtained from the stability analysis of the homoclinic orbits and by the corresponding scaling relations involved in the approach to homoclinicity .
First, we consider the similarities of the behavior of our model with the characteristic features of the Shilnikov scenario. In three dimension, the Shilnikov criterion is stated in terms of the two possible combinations of the dimensions for the invariant manifolds of the saddle-focus; the unstable manifold is two dimensional and stable manifold is one dimensional and vice versa. For these two cases, the presence of an homoclinic orbit is given by the condition $`|\rho /\lambda |<1`$, where the eigen values of the fixed point are given by $`\rho \pm i\omega ,\lambda `$, where $`\rho >0,\lambda >0`$ for the first case and $`\rho <0,\lambda <0`$ for the latter. The analysis of the Shilnikov scenario shows that in the neighborhood of the homoclinic point, the parameter space is organized such that the period of the principal periodic orbit tends to infinity as the parameter approaches the value corresponding to the homoclinic point. In our case, the system is four dimensional, with the unstable manifold of the fixed point characterized by a pair of complex eigenvalues $`\rho \pm i\omega `$ ($`\rho >0`$) and the stable manifold by two eigenvalues $`\lambda _1<0`$ and $`\lambda _2<0`$. Here, $`\lambda _1`$ stays close to zero and $`\lambda _2`$ is substantially negative. Thus, in our case, the criterion $`|\rho /\lambda |<1`$ refers to $`\left|\rho /\lambda _1\right|<1`$. We find that this condition is satisfied only in a small region just prior to the disappearance of the PPO in a Hopf bifurcation. A typical plot of the eigenvalues for $`m=1.2`$ is shown in Fig. 9, where we have also shown the phase $`\omega `$. Even though $`\left|\rho /\lambda \right|<1`$ is not satisfied over large portion of $`e`$ and $`m`$, we do see that the period of the periodic orbits tend to increase as $`e`$ is increased ($`m<2.0`$) which is typical of the Shilnikov scenario. A plot showing the period (of the superstable orbits) verses the deviation from estimated homoclinic point($`e^{}`$) is displayed in Fig. 10 for $`m=1.4`$. (We have plotted points from period three onwards.) Here, we have taken the value of $`e^{}`$ to be the value of $`e`$ for the onset of the last observed periodic orbit with period $`12`$ ($`e=247.63`$). It must be stated that we do not face any difficulty in locating any of the periodic orbits upto the period $`12`$. However, we are unable to detect the next period which we interpret as an incomplete approach to the true homoclinic point. Here, it must be mentioned that incomplete approach to homoclinicity is quite common. We stress that even in the region where Shilnikov criterion is obeyed, we do not observe homoclinic orbit.
One other feature which is usually seen in the Shilnikov scenario is that the reinjection of the trajectory in to the neighborhood of the fixed point is along the direction of the fast variable after which the trajectories tend to stay around the saddle focus fixed point. In our model, since $`x`$ is the fastest variable, it also acts as a reinjection direction. However, a closer examination shows that the spiraling in of the orbit towards the fixed point is along the $`z`$ direction which is the next fastest variable. This is evident in Fig. 11 where a typical trajectory is shown. Even more dominant feature of the Shilnikov scenario is that the successive bifurcations should be connected by the PPO. This, however, is not true in our case as we have seen earlier, since the isolas which form the period adding sequence are distinct from the PPO. This feature is clear from Fig. 3 for $`m=1.8`$. In fact, this feature of the isolas being distinct from the PPO is more like that of the Gavrilov-Shilnikov scenario which requires the presence of an unstable periodic orbit which we failed to detect in the entire parameter region wherein nontrivial dynamics is present. This rules out the possible presence of any homoclinic bifurcation due to the saddle periodic orbit. Thus, we see that our model has partial features of both these scenarios.
The above discussion suggests that the origin of the MMOs in our model is likely to be different from the two scenarios. In order to understand the mechanism of the MMOs in our model, we will use the information on the nature of the relaxation oscillations. We first note that the fixed point $`(x_0,y_0,z_0,\varphi _0)`$ is on the $`S_1`$ part of the slow manifold and moves up on $`S_1`$ as $`e`$ is increased. Both $`x_0`$ and $`z_0`$ have a near linear dependence on $`e`$ namely $`x_0=z_0\frac{e}{2}`$, while $`y_0`$ and $`\varphi _0`$ are practically constant. Since the fixed point is unstable, any orbit in its neighborhood will locally expand along its unstable directions. Thus, we expect to get insight into the mechanism of the MMOs by studying the rate of expansion of such orbits. In order to understand the mechanism operating in our model, let us consider a periodic orbit of $`L^s:1^{10}`$ type shown in Fig. 12. If the orbit has reached the neighborhood of the fixed point, any orbit on $`S_1`$ should spiral out with a local dynamics determined by the linearized eigenvalues around the fixed point. Within this approximation, the orbit expands at rate $`exp[2\pi \rho /\omega ]`$ per rotation around the fixed point. Assuming a linearized behavior for $`n`$ rotations ( for fixed values of $`e`$ and $`m`$), we get
$$r_n/r_1=\mathrm{exp}[2\pi (\rho /\omega )(n1)]$$
(19)
where $`r_n`$ is the distance measured from the fixed point after $`n`$ rotations along a fixed direction in the unstable manifold of the fixed point. Here, $`r_1`$ is the value of $`r_n`$ for $`n=1`$. Since Eq. (19) is based on linearized approximation, the values of $`r_n`$ obtained from the phase plots will be in general different due to the influence of nonlinearities. For this reason, we will first study the region of validity of Eq. (19). We note that the unstable manifold is nearly in $`xz`$ plane and in the neighborhood of the fixed point, the major contribution to $`r_n`$ comes from $`x`$ and $`z`$. Thus, it would be sufficient to consider $`x(t)`$ (or $`z(t)`$) in place of $`r_n`$ and in particular, we will use the minima or maxima to analyze the small amplitude oscillations of the periodic orbit using Eq. (19). Consider the plot shown in Fig. 12. We note that the time interval between the successive minima or maxima of $`x(t)`$ can be taken to correspond to one rotation of the orbit. We shall denote the the deviations of the $`n`$-th minimum from the fixed point value, $`x_0(e)x_n^{min}`$, by $`x_n^{}`$. In Fig.12, $`x(t)`$ and $`z(t)`$ are plotted along with their fixed point values ($`x_0=z_0=137.82`$). ( We have also shown regions of $`x(t)`$ corresponding to regions $`S_1`$ and $`S_2`$ of the slow manifold.) The decreasing nature of the maxima values of the amplitude of $`x(t)`$ for the first few cycles is a reflection of the fact that the orbit has not reached the neighborhood of the unstable manifold of fixed point lying on $`S_1`$ (see Fig. 11). It is only later (in time) that the expanding nature of the oscillations manifest, seen as the increase in the magnitude of the successive maxima values of $`x(t)`$. Thus, the value of $`x_1^{}`$ read off from $`x(t)`$ (in place of $`r_1`$) will contain contributions arising from reinjection mechanism. Hence, identifying $`x_1^{}`$ as representing the value of the first minimum of $`x(t)`$ would be an incorrect, if one wishes to use Eq. (19). Thus, $`x_1^{}`$ has to be estimated by extrapolating the values of $`x_n^{}`$ using values of $`n`$ where nonlinearity plays an insignificant role, i.e., $`n>3`$ for the case of Fig.12. We denote this extrapolated value by $`x_1^{}`$. In addition, as the amplitude grows as a function of $`n`$, one should also expect that the linear dynamics breaks down. Thus, for larger $`n`$ values, we should again see the effect of nonlinearity. In fact, this feature shows up as a decrease (though marginal) in the time interval between successive minima for higher $`n`$ as can be verified from Fig. 12 (seen between the ninth and the tenth peaks).
Now, we attempt to estimate the changes in the magnitude of the small amplitude oscillations located on $`S_1`$ as a function of $`e`$ and estimate at what value of $`e`$ the trajectory hits the ‘fold’ between $`S_1`$ and $`S_2`$. We denote the distance of the fixed point $`(x_0(e),y_0(e),z_0(e),\varphi _0(e))`$ from the ‘fold’ given by $`D=((x_0(e)x_0(e_f))^2+(y_0(e)y_0(e_f))^2+(z_0(e)z_0(e_f))^2+(\varphi _0(e)\varphi _0(e_f))^2)^{\frac{1}{2}}`$, where $`(x_0(e_f),y_0(e_f),z_0(e_f),\varphi _0(e_f))`$ refers to the value of the fixed point at the first Hopf bifurcation ($`e=e_f`$). Further, we note that the fixed point is close to the ‘fold’ at the first Hopf bifurcation and noting $`x_0=z_0\frac{e}{2}`$ with a very weak dependence of $`y_0`$ and $`\varphi _0`$ on $`e`$. Thus, in one dimension where we are dealing with $`x`$ variable alone, we can take $`Dx_0(e)x_0(e_f)`$. (Note that position of ‘fold’ is insensitive to $`e`$.) Using the fact that $`x_0(e_f)`$ is small, we get $`Dx_0(e)`$. Using this, we will attempt to find the maximum value of $`n`$ allowed for which the condition $`x_n^{}>D`$ is satisfied. For this, we need to know the dependence of $`\rho `$ and $`\omega `$ on $`e`$ which decides the rate of growth of the small amplitude oscillations as a function of $`e`$. In the range of the MMO sequences that we are interested, $`\rho `$ decreases and $`\omega `$ increases nearly linearly as can be seen from Fig. 9. Thus, we take $`\rho =\rho _0m_\rho e`$ and $`\omega =\omega _0+m_\omega e`$, where, $`m_\rho `$ and $`m_\omega `$ are the corresponding slopes. These are evaluated numerically as the best fit for the region of interest of $`e`$. The fit yields $`\rho _0=0.10433,m_\rho =0.0003632,\omega _00`$ and $`m_\omega =0.0004114`$ for the case shown in Fig. 12. Using this in Eq. (19) gives
$$\frac{x_n^{}}{x_1^{}}=\mathrm{exp}[2\pi (\rho /\omega )(n1)]=\mathrm{exp}\{[k_1+k_2/e](n1)\}$$
(20)
Here $`k_1`$ and $`k_2`$ are functions of $`\rho _0`$, $`m_\omega `$ and $`m_\rho `$. This equation can be interpreted as a scaling form for the small amplitude oscillations of stable periodic orbits (i.e., for a fixed $`n`$) as a function of $`e`$. Now, consider the set of all stable periodic orbits of the form $`L^s`$ for a given value of $`m`$. Then, for each of these $`L^s`$ orbits, the values of $`n`$ ranges upto $`s`$. Since the magnitude of small amplitude oscillations of these periodic orbits depends $`e`$, we can plot $`lnx_n^{}`$ versus $`1/e`$, where we have used $`x_n^{}`$ values corresponding to the $`n`$-th minimum of a periodic orbit. The plots of $`lnx_n^{}`$ as a function of $`e^1`$ for $`n=4`$ to $`7`$ are shown in Fig. 13. In the figure, the lowest curve (+) corresponds to $`n=2`$ and other successive higher curves refer to $`n=3`$ upwards. The lowest band within the $`n=2`$ curve corresponds to a periodic orbit of the form $`L^{10}`$ and successive bands have decreasing $`s`$ values. (The gaps correspond to chaotic bands between successive periodic orbits.) It is clear that the plots are linear and the slopes of the curves corresponding to $`n=4,5,6`$ and 7 show an increasing trend, increasing in multiples of $`1530`$ ($`k_2`$). To illustrate the presence of nonlinearity, we have also shown plots of $`lnx_n^{}`$ for $`n=2`$ and 3. One can easily notice that these two curves deviate from linearity considerably for large $`e`$. In addition, we see that the slopes of these two lines are not in multiples of $`k_2`$. This suggests that it should be possible to collapse all the curves for $`n>3`$ onto a single curve. Noting that the slopes of curves are in multiples of $`k_2`$, $`lnx_n^{}/n`$ renders them parallel and the preexponential factor which satisfies Eq.(20), denoted by $`x_1^{}`$ can be easily determined. The value of $`x_1^{}`$ so obtained can be now used in
$$\frac{x_n^{}}{x_1^{}}=\mathrm{exp}[2\pi (\rho /\omega )(n1)]$$
(21)
to estimate the maximum number of small amplitude oscillations allowed for any given value of $`e`$ before the size of the orbit (under the linear approximation) is large enough to hit the ‘fold’. This is determined by the value of $`n=n_c`$ at which $`x_n^{}=x_{n_c}^{}>De/2`$.
In the inset of Fig. 13, we have shown that the curves for $`n=4`$ to $`7`$ can be collapsed on to a single curve. The value of $`x_1^{}`$ obtained by extrapolating the curves $`ln(x_n^{}(e))/n`$, for $`n=7`$ to $`4`$ is 5.617. We have verified that the $`n=2`$ curve and to a lesser extent $`n=3`$ curve deviates from the collapsed curve reflecting that nonlinearity corresponding to reinjection is dominant for these two cases. This also implies that the orbits do not approach the fixed point close enough that the linearized eigenvalues could be useful. Using the value of $`\rho =0.0073`$ and $`\omega =0.1348`$ for $`e=267.0`$, we can now estimate the value of small period oscillations for $`n=11`$ is 171.0 which is larger than $`e/2`$. (Here $`e=267.0`$.) This means that a maximum of nine small amplitude oscillations are allowed beyond the first minimum at this value of $`e=267.0`$. This is consistent with what is seen in Fig. 12. We have verified that this method of estimating the maximum number of allowed small amplitude oscillations for any given value of $`e`$ works very well as long as $`n>3`$ for $`m=1.2`$ and other values of $`m`$ as well. For $`n=2`$ and 3, the value of $`e`$ at which these orbits disappear shifts to much lower values that what actually observed.
A little reflection on the above results reveal the cause of incomplete approach to the homoclinic point. Recall that the growth of small amplitude oscillations is controlled by $`2\pi \rho /\omega `$. We have seen that this quantity depends inversely on $`e`$. Thus, the arithmetically increasing number of small amplitude oscillations accommodated on $`S_1`$ ( without the trajectory crossing over to $`S_2`$) is a direct result of ‘softening’ of $`2\pi \rho /\omega `$ as a function of $`e`$. In other words, for every small amplitude oscillation accommodated on $`S_1`$, $`e`$ changes by a fixed amount commensurate with the softening rate. Thus, the number of small amplitude oscillations that can be accommodated in the allowed interval of $`e`$ in the bifurcation plot will be limited. This also implies that the approach to homoclinic point can at best be asymptotic due to finite rate of softening of $`2\pi \rho /\omega `$, with the asymptotic nature manifesting only in the limit $`k_20`$. Clearly, these results are valid under the assumption that the contribution from nonlinear terms to the growth of the small amplitude oscillations is not strong, which is substantiated by the numerical evidence that the estimated number of small amplitude oscillation allowed for a particular value of $`e`$ agrees with what is numerically observed.
It may be worth pointing out here that even though the analysis given here is for MMOs of the kind $`1^s`$, it is clear that they can be easily generalized to $`L^s`$ kind of MMOs. It must be mentioned here that the ‘softening’ of $`2\pi \rho /\omega `$ is a result of the global constraint in our model, namely, the back-to-back Hopf bifurcation. Thus, we see that the apparent homoclinic scenario exhibited by the model system is completely new. We have shown that this feature coupled with the mechanism of the relaxation oscillations operating in the model gives rise to features of MMOs common to both Shilnikov and Gavrilov-Shilnikov scenario.
### 4.3 Discussion and Conclusions
Some comments may be in order here about the bent-slow manifold structure of the system. The relaxation oscillations seen in this system differs qualitatively from that seen in systems with the $`S`$shaped slow manifold. The major difference in the structure is that while in the $`S`$ shaped manifold, there are two attractive pleats separated by a repulsive part, in our case, both pieces $`S_1`$ and $`S_2`$ of the bent-slow manifold are attractive and are connected continuously. This aspect coupled with the fact that there is no repulsive part in the bent-slow manifold as in the $`S`$\- shaped manifold suggests that the mechanism causing jumps between the $`S_1`$ and $`S_2`$ is very different. The only similarity is that the number of jumps ( fast transitions) accomplished by the trajectories from one part of the slow manifold ($`S_2`$) to another ($`S_1`$) and vice versa is two as in the case of $`S`$shaped manifold. During the jump from $`S_2`$ to $`S_1`$, the trajectory tends to move out of the slow manifold into unstable phase space by sticking to the direction of of motion on $`S_2`$. Here the motion is accelerated due to unstable nature of the phase space. This has some similarity to canard type of solutions but the comparison is superficial since the unstable part of the phase space to which the trajectory moves is not a part of the slow manifold. There is another difference namely the operative time scales in the dynamics on the slow manifold. The dynamics on $`S_2`$ is slow as it is controlled by the slow variables $`y`$ and $`\varphi `$, since $`\dot{x}0`$ for the entire interval of time the trajectory is on $`S_2`$. On the other hand, on $`S_1`$, the time dependence of a trajectory is largely controlled by the fast variable $`x`$.
We have also analyzed the effect of bent-manifold structure on mixed mode oscillations and the incomplete homoclinic scenario. We have shown that in the case of bent-manifold structure, the approach of the fast variable towards the fixed point is along the slow manifold itself eventhough the eventual approach is along the $`z`$ direction. This is in direct contrast with the $`S`$shaped structure where the direction of the jump of the fast variable is transverse to the slow manifold containing the fixed point as in the Rossler’s $`S`$shaped slow manifold. There is another possibility as pointed out by Koper et al. In this case, the plane of relaxation oscillations is parallel to the plane of nearly harmonic small amplitude oscillations. These authors suggest that this type of dynamics may be responsible for the incomplete approach to homoclinicity since the approach to the fixed point is along the attracting pleat of the $`S`$shaped manifold. However, in both cases, irrespective of the location of the fixed point, the nature of the relaxation oscillations remain the same. Moreover, in both these cases, the small amplitude oscillations(near harmonic), as well as the large amplitude (relaxation type) oscillations are well characterized. However, the nature of intermediate amplitude oscillations are not so well understood. It has been suggested in the literature that these are related to canard type of solutions. The latter type of oscillations result from ‘sticking’ of the trajectory to the repelling part of the $`S`$shaped slow manifold before jumping to the attracting pleat of the slow manifold. In our case, although the oscillations have a superficial similarity with canard type of solutions, it is the ’sticking’ of the trajectories in the same direction of $`S_2`$ well into the unstable phase space, coupled with the fact that there is no inherent constraint relating $`S_1`$ to $`S_2`$ in the manifold structure that appears to lead to jumps of all sizes. As an illustration, we have shown a plot of the trajectory for $`m=1.8`$ and $`e=190.0`$ in the $`x\delta `$ plane (Fig.14). It is clear that while the small amplitude oscillations are located in the neighborhood of the fixed point, the intermediate amplitude oscillations result from the trajectory ‘sticking’ to the direction of the $`S_2`$ plane and moving into the ‘unstable’ part of the phase space by varying amounts each time the trajectory leaves $`S_2`$.
Although the bent-manifold structure is characteristic of our systems, we believe similar structure is likely to be seen in many other models and experimental systems. Particularly, in chemical kinetics where only binary collisions are permitted, models are expected to involve only quadratic or biquadratic nonlinearities. Such models are promising candidates to exhibit the bent-slow manifold structure. In experimental systems, the dominant signature to look for would be the trapping of fast variable at small values over a substantial portion of its period followed by sharply peaked pulse like behavior.
From the above discussion, we see that the bent-slow manifold structure is at the root of understanding of the pulsed type of relaxation oscillation and the MMOs. We note here that our analysis is completely local since we have used the linearized eigenvalues around the fixed point. For the same reason, the scaling relation obeyed by the small amplitude oscillations (Eq. 21 and Fig. 13) is found to be valid only where the influence of nonlinearity to the growth of these small amplitude oscillations (as a function of $`e`$) is minimal. In spite of this, the scaling relation so obtained forms the basis for estimating the maximum number of small period oscillations permitted for a given $`e`$ thus explaining the origin of MMOs in the model.
Here, we mention that the relaxation oscillations arising out of the atypical bent-slow manifold structure is directly related to a dominant characteristic of the PLC effect, namely, the negative strain rate sensitivity of the flow stress. The analysis of time scales involved in the relaxation oscillations has been useful in understanding the origin of the negative strain rate sensitivity of the flow stress which is reported elsewhere .
In summary, we have analyzed the dynamics of a model for a type of plastic instability due to Ananthakrishna and coworkers with particular attention to the complex dynamics exhibited by the model. We have shown that the nature of relaxation oscillations and the MMO sequences exhibited by the model is atypical. We have proposed a new mechanism for the relaxation oscillations based on the bent-slow manifold structure of the model. Using this we have explained the origin of the MMOs. We have further shown that a crucial role in organizing the dynamics is played by the physical constraint, namely, the stress oscillations are seen only in a window of strain rates. (Indeed, the model has been devised to be consistent with this experimental feature.) This constraint translates to back-to-back Hopf bifurcation in the model leading to the ‘softening’ of the eigenvalue of the saddle fixed point that controls the small amplitude oscillation. It is this finite rate of softening that is responsible for the incomplete approach to the homoclinic bifurcation.
## 5 Acknowledgements
The authors would like to thank Dr.T.M. John who was involved in the early stages of the work and one of the authors (SR) wishes to thank the theory group, Materials Research Center, IISc, for helpful discussions. |
no-problem/0001/astro-ph0001322.html | ar5iv | text | # 1 Line equivalent widths (Å) and veilings.
|
no-problem/0001/hep-ex0001057.html | ar5iv | text | # 1 Introduction
## 1 Introduction
Measurement of Cherenkov ring image requires two-dimensional photon information such as x and y coordinates as RICH and DIRC do . With the use of a quartz bar as a Cherenkov radiator and also a light-guide like the DIRC counter , a combination of Time-Of-Propagation (TOP) of Cherenkov photons to a bar-end and their emission angles at the bar-end also provide the ring image information. Here we briefly describe the principle of such a device, named TOP-counter, ( its detail is cited in ) and explain some results of its R&D works. The specific aspect of this counter is its compactness relying upon a horizontal focussing approach described below. We intend to develop this counter in a bid to upgrade the BELLE pid detector.
Figure 1 illustrates a side view of Cherenkov photons propagating a quartz bar. TOP is inversely proportional to z(quartz-axis direction)-component of the light-velocity, which produces TOP differences of, for instance, about 100 ps or more for normal incident 4 GeV/c $`K`$ and $`\pi `$ at 2 m long propagation. The TOP difference is a function of photon’s horizontal (x-z plane) emission angle ($`\mathrm{\Phi }`$). Time measurement for a single photon with a 100 ps resolution provides 1 $`\sigma `$ separation, and therefore expected number of 30 photons in this case give us, briefly speaking, a factor of $`\sqrt{30}`$ times higher separation. Furthermore, a detection of backward-going (BW) photons reflected at the other-end as seen in the figure enhances, in principle, the separation by an another factor of $`\sqrt{2}`$ for normal incident particle. As is easily noticed, the TOP measurement inevitably includes also the Time-Of-Flight (TOF) from an interaction-point to the TOP counter both of whose difference between $`K`$ and $`\pi `$ could have the same sign with each other in most of the cases. Adding the TOF information therefore helps the separation, as a result, TOP is hereafter defined as TOP+TOF.
In order to estimate the achievable separability of TOP counter, we optimized its parameters as illustrated in Fig. 2, where the butterfly-shaped horizontal focussing mirror with an arc radius of 250 mm was designed to have the $`\mathrm{\Phi }`$-aparture of $`\pm `$45<sup>o</sup> and dispersion of d$`\mathrm{\Phi }`$/dx=0.5<sup>o</sup>/1 mm. Root-mean-square of the focussed accuracy is $`\mathrm{\Delta }x\pm `$ 0.4 mm. The bar and mirrors are made of synthetic optical quartz with refractive index (n) of 1.47 at $`\lambda `$=390 nm. These counters are supposed to be placed at 1 m radially away from the interaction point of KEKB-BELLE to form a cylindrical structure.
## 2 Expected separability
There are three dominant contributions to TOP measurement: (1) Chromatic effect of Cherenkov lights, (2) aberration effect of the focussing mirror and (3) transit time spread (TTS) of photomultiplier tube (PMT). Since BELLE-CDC (Central Drift Chamber) provides precise enough track information such as position, angle and momentum for particles, track relating ambiguity is about 10 ps which is much smaller than the above three contributions. As a necessary item to be considered, TDC start-signal is assumed to have a 25 ps uncertainty in the calculation.
It is worth mentioning that the quartz bar-thickness produces a harmless effect on the measurement, since the variation of a sum of a particle’s travel time in a quartz bar up to Cherenkov radiation point and the photon propagation time to the bar-end is about 20 ps or less for particles of any incident angle. Consequently, this contribution is also minute comparing to the others. The width of the crystal bar, on the other hand, is effectualy nullified, in principle, due to the horizontal focus, and in practice, within the achievable focussing accuracy. This is the reason to choose the horizontal instead of the vertical focus, otherwise the ring image would grow dim by a finite size of the bar-width. Resultantly, both the finite sizes of thickness and width now can be disregarded, therefore we do not need any lengthly image projection to nullify the bar cross-section.
A PMT (Hamamatsu, R5900U-00-L16: linear-arry 16-anodes) is used for R&D works without magnetic field (B). Its specific parameters are: Surface area of 30$`\times `$30 mm<sup>2</sup>, sensitive area of 16$`\times `$15 mm<sup>2</sup>, the anode size of 0.8 mm-wide with 1.0 mm-pitch and 15 mm-long, quantum efficiency QE of 20-25%, gain of 2$`\times `$10<sup>6</sup>, risetime of 0.6 ns, and TTS of $`\sigma `$=70-80 ps. Specific modification of L16 and development of a PMT (R6135MOD-L24: Fine-mesh 24-anodes) operable under a magnetic field are being proceeded in cooperation with Hamamatsu Co. For the latter PMT, a position resolution of better than 0.5 mm is achieved under B$`>`$0.2 TG and TTS of $`\sigma `$=130 ps is currently realized under B$`<`$0.6 TG.
Calculated TOP differences between 4 GeV/c $`K`$ and $`\pi `$ and the above-mentioned three contributions are illustrated for two cases in Fig. 3, where TTS is set as 80 ps to include other small uncertainties such as the start-signal. When the particle incident polar angle ($`\theta _{\mathrm{inc}}`$) gets around or smaller than 40<sup>o</sup>, TOP difference reverses its sign against the TOF difference, as seen in Fig. 3(b), and the separability power reduces a bit. While the expected number of detectable forward-going (FW) photons is at an average 35 and 115 at (a) and (b), respectively, only the early arrived photons at the individual anodes are taken into account for the time measurement. When the BW photons are also regarded for detection, they come more than 15 ns later than FW photons which are widely separated enough for measurement to take place and for distinguishing between each other.
As a sample of simulation study, Fig. 4(a) shows a Log-Likelihood distribution in a case of the FW photon detection for 4 GeV/c $`K`$ and $`\pi `$ with $`\theta _{\mathrm{inc}}`$=90<sup>o</sup>. Resulted separability is $`S(=\sqrt{2\mathrm{\Delta }ln})`$=5.7. Over-all expected $`\pi /K`$ separability is shown in Fig. 4(b) in the case of BELLE configuration. High momentum limit is indicated by a thick line for the pions in $`B\pi \pi `$ decay. It is found that $`S>`$5 is achieved at any barrel region of $`\theta _{\mathrm{inc}}`$=30<sup>o</sup>-130<sup>o</sup>.
## 3 Beam Test
A test counter of 1 m long quartz bar was constracted with the structure as described in Fig. 2 but an absorptive filter, instead of a reflection mirror, for BW photons at the bar-end is prepared. Six L16 PMTs (96 anode channels in total) were attached at the mirror. Since the photoelectron detection efficiency of L16 PMT is about 1/2 and an effective mirror surface coverage by six PMTs with our configuration is approximately 40%, the total photon detection efficiency, besides PMT’s QE, is nearly 20%. The above photon insensitive area, most of which is the structual space of PMT, would reflect the photons and resultantly hit other wrong anodes. To avoid this phenomena, absorptive filters were inserted in front of such areas. Measurement was performed using $`\pi ^{}`$ beam at KEK-PS.
First, beam was tuned to normally hit the counter at $`L`$=0.02 m. Recorded data are shown in Fig. 5. Single photon peak is clearly seen in ADC spectrum. Besides Cherenkov photons, two small contributions of knock-on electrons and reflected photons are found on TDC spectrum. Resultant time resolution over all 96 channels is about $`\sigma `$=85 ps, as plotted in Fig. 6. Since the chromatic contribution can be ignored at this configuration, the resulted resolution is dominated by TTS of L16 PMT.
Next, beam position was moved to $`L`$=1 m and three different momenta of 1.1. 2 and 4 GeV/c were set. Expected number of fired anodes was around 6, while we observed 6.3 at an average including both the contributions from the knock-on electron and reflected photons for individual three different momenta. Cherenkov ring image is clearly observed as a function of $`\mathrm{\Phi }`$-angle, as seen in Fig. 7. In order to extract the resolution in this case, a simple tricky analysis had to be applied, because the beam divergence defined by trigger counters were not sufficiently small enough as expected at the BELLE detector system to make its contribution ineffectual. That is, the triggered samples are required to have a signal at a certain channel, for example, 27th channel, within the first 150 ps part of the measured raw time distribution of 350 ps (FWHM). This bias would restrict the beam divergence somehow but not explicitly. Thus obtained resolutions are plotted in Fig. 6: Fairly good agreement with the expectation can be seen. The parabolic rise of the calculated resolution at large $`\mathrm{\Phi }`$ is due to the aberration effect of the mirror rather than the chromatic contribution of the Cherenkov light at $`L`$=1 m case.
## 4 Summary
TOP counter is quite compact and has high separability. Due to the horizontal focussing and thin radiator thickness, the size of quartz bar’s cross-section can be disregarded so that it does not need a large standoff projection space such as DIRC. It is still at an early R&D stage and needs more essential studies as mentioned below.
First, confirmation of basic TOP behavior, especially the performance at $`L`$= 1 m or longer distances, should be done using tracking chambers at next beam test.
Increasing the detected number of photons is the most important issue and two approaches for enlarging the sensitive area are being examined: One way is to use a light-guide, and the other way is to develop L16 PMT. When a way to succesfully collect sufficient number of photons is established, TOP counter can be used as a real detector under certain experimental condition such as, for instance, fixed target experiment with no magnetic field. It needs much less space comparing to Gas Cherenkov counter, and can be configured to make the counter normal to incident particles so that the separability is enhanced by detecting both the FW and BW photons.
In order to utilize the TOP counter as the next BELLE pid detector, the second most important issue is to develop a single photon and position sensitive, high time-resolving detector operational under a magnetic field of 1.5 TG. R&D work of L24 PMT is being earnestly proceeded so that a successful outcome can be within our grasp in the near future.
This work is supported by Grant-in-Aid for Scientific Reasearch on Priority Areas (Physics of CP violation) from the Ministry of Education, Science, and Culture of Japan. |
no-problem/0001/astro-ph0001433.html | ar5iv | text | # Theoretical Implications of Diffuse Non-Thermal Emission from Clusters of Galaxies
## 1. Introduction
The space between galaxies in clusters of galaxies is filled with a hot and dilute thermal gas, which is the main component of the intra-cluster medium (ICM). Radio astronomical observations gave the first evidence for the occurrence of diffuse non-thermal phenomena in the ICM with the detection of the radio halo of the Coma cluster (Willson 1970). Recently more non-thermal signatures were detected from Coma and a number of other clusters with the observations of extreme ultra-violet (EUV) emission (Lieu et al. 1996) and high energy X-ray (HEX) emission (Fusco-Femiano et al. 1999) in excess of the thermal one. The interpretation of these fluxes is still not yet settled, nor are the different classes of diffuse radio emission completely understood. We give an overview on current interpretations and theories. This should outline the physical insight we will gain from understanding the currently strongly increasing number of non-thermal detections of cluster of galaxies (Feretti 1999).
## 2. Diffuse Radio Emission from Clusters of Galaxies
Radio observations are the most sensitive probes of non-thermal processes in clusters of galaxies, and especially low frequency radio astronomy plays a crucial role since the radio spectra are steep. The diffuse radio sources can be divided into two classes: cluster radio halos and cluster radio relics<sup>1</sup><sup>1</sup>1‘Cluster radio relics’ should not be confused with dying radio galaxies, which are also called ‘radio relics’. For this reason Luigina Feretti proposed to call them ‘peripheral halos’, and Ron Ekers proposed ‘radio flotsam’. Since all three names have disadvantages I decided to stick to the old one..
The morphology of the radio halos is similar to that of the X-ray gas. They are unpolarized and occur preferentially in clusters which also show evidence for ongoing or recent cluster merger events. Cluster radio relics are irregularly shaped, more peripherally located with respect to the cluster center and sometimes linearly polarized. Both, cluster relics and halos, seem to be connected to cluster merger, but their natures are different.
## 3. Cluster Radio Relics
There is increasing evidence that the cluster relics appear at locations where shock waves of an ongoing merger event, or from the accretion flow onto the cluster are located. In the cases of Abell 2256 and Abell 1367 temperature substructures of the hot ICM gas could be detected (Briel & Henry 1994; Donnelly et al. 1998), which support the presence of a shock wave at the location of cluster relics in these clusters. For Abell 2256, Abell 3667 and also the Coma cluster numerical simulations of merger events were satisfactorily fitted to the X-ray data, which also supports the shock wave-relic connection (Roettiger, Burns, & Pinkney 1995; Roettiger, Burns, & Stone 1999).
The shock wave probably accelerates electrons (and also protons) via Fermi acceleration. The accelerated electrons illuminate the cluster magnetic fields by synchrotron radiation leading to the cluster relic. The magnetic fields get compressed and therefore aligned with the shock plane, causing the radio emission to be polarized if seen edge on. Thus, the observed properties of cluster radio relics, such as as degree and direction of polarization, surface luminosity, peripheral position etc., can be understood (Enßlin et al. 1998; Roettiger, Burns, & Stone 1999).
But shock waves are common due to the high merging fraction of cluster of galaxies, whereas cluster relics are only rarely found in clusters. This could be a surface brightness problem, or the indication of a second necessary condition (in addition to a shock wave) for the occurrence of a cluster relic.
The cluster relic 1253+275 in Coma might tell us what this second condition is: its radio emission is morphological connected to the radio tails of the galaxy NGC 4789 (Giovannini, Feretti, & Stanghellini 1991). Following the radio tails of that galaxy from the head to the cluster relic, one sees clear evidence for spectral steepening of the electron population in the radio spectral index, and a flattening at the location of the relic. This suggests that it is the radio plasma the galaxy releases which actually flashes up at the shock wave and not the normal inter-galactic medium (IGM). Radio plasma has the ideal composition for this: evidently magnetic fields and a pre-accelerated electron population are present. It is not necessary for a cluster radio relic, as it is the case for 1253+275, that the source of the radio plasma is nearby or visible, since also old invisible radio plasma sitting in the IGM could be re-activated by shock waves.
## 4. Radio Ghosts
Active radio galaxies fill large volumes in the IGM with radio plasma. This plasma becomes rapidly invisible to radio telescopes due to inverse Compton (IC) and synchrotron energy losses of the relativistic electrons. Afterwards it might form an invisible, but possibly important phase of the IGM. Due to the difficult detectability the name ‘radio ghosts’ for patches of this remnant radio plasma<sup>2</sup><sup>2</sup>2More nomenclature: ‘Since the term ‘ghost’ does not fit the context from the viewpoint of Hindu philosophy, Gopal Krishna proposed ‘radio fossil’, drawing upon the Indian mythological tale of lady Ahilya (ref: the epic Ramayana, circa 2000 B.C.). Turned into a fossil while meditating, Ahilya was revived to life at the touch of Lord Rama’s feet, as the Lord passed her way on a pre-ordained track during his 14-year exile in the forest. Likewise, ‘radio fossils’ can spring back to a radiant life upon passage of a shock wave. was proposed (Enßlin 1999).
After injection of the radio plasma it will expand or contract until it reaches pressure equilibrium with the surrounding medium. Its pressure is given by that of the confined relativistic particles and the magnetic fields, assumed to be in rough energy equipartition. Therefore magnetic field energy density in radio ghosts should be lower, but of the order of the thermal energy density of the environment. Subsonic turbulence in this environment, which has an energy density below the thermal energy density, is therefore not strong enough to overcome the magnetic elastic forces of the radio ghost. Sonic or super-sonic turbulence, which is e.g. expected in giant merger events of cluster of galaxies, can ‘shred’ the ghost into smaller pieces. The size of such pieces will be comparable to the eddy size of the turbulence. Since a typical turbulent spectrum has less energy density on smaller scales there is a length-scale below which the turbulence is not able to overcome magnetic forces. Turbulent erosion of radio ghosts should stop at this length-scale, leaving small-scale patches of still unmixed old radio plasma.
Radio ghosts are practically invisible as long as their electron population remains at low energies. But if the population is re-accelerated the ghost becomes radio luminous again. This can happen when the ghost is dragged into a large-scale shock wave, e.g. in a merger event of clusters of galaxies or at the accretion shock where the matter is falling onto a cluster. The emission region is expected to be irregularly shaped, and should exhibit linear polarization due to the compression of the magnetic fields in the shock. Such regions are indeed observed at peripheral locations of a few clusters of galaxies: the cluster radio relics. Further observable signatures of radio ghosts are discussed in Enßlin (1999), Medina Tanco & Enßlin (2000), and Enßlin & Kaiser (2000).
## 5. Cluster Radio Halos
The origin of the relativistic electrons and magnetic fields in the ICM, which are necessary for the synchrotron process which produces the radio halo, is not known. The magnetic fields could be primordial (Olinto 1997), injected by galactic winds (Kronberg, Lesch, & Hopp 1999; Völk & Atoyan 1999), or by radio galaxies (Daly & Loeb 1990; Okoye & Onuora 1996), produced in shock waves of the large scale structure formation (Kang, Rachen, & Biermann 1997), amplified in the turbulence from galaxy motion (Roland 1981; de Young 1992; Sánchez-Salcedo, Brandenburg, & Shukurov 1998) and cluster formation (Roettiger, Stone, & Burns 1999; Dolag, Bartelmann, & Lesch 1999), or of another unknown origin. Harris & Miley (1978) and Komissarov (1985) proposed that radio halos are superpositions of large numbers of relic radio galaxis.
There are also many possible sources for the relativistic electrons. Fig. 1 gives an overview about the current theories and their dependencies. In the following the different parts of the diagram are explained and references to the literature are given:
The energy source of all non-thermal processes in clusters should be either the kinetic energy of matter falling onto clusters, or the outflows from galaxies. The latter can be divided in galactic winds, which are strongest for starburst galaxies, and ejection of radio plasma from an AGN. All these processes can produce shock waves and inject turbulence into the ICM, and therefore produce conditions where Fermi mechanisms accelerate particles (Jaffe 1977; Schlickeiser, Sievers, & Thiemann 1987; Roettiger, Burns, & Loken 1993; Tribble 1993; Völk, Aharonian, & Breitschwerdt 1996; Deiss et al. 1997; Eilek & Wetherall 1999; Wetherall & Eilek 1999).
The non-thermal particles could be accelerated out of the thermal pool (Liang 1999), or they were produced by radio galaxies (Jaffe 1977; Rephaeli 1977; 1979; Valtaoja 1984). In the latter case, they had to escape out of the radio plasma, which is difficult since the necessary diffusion across magnetic fields is a extremely slow process. Turbulence in the ICM can strongly increase that diffusion coefficient, and especially shortly after mayor merger events there exists a short time window when particles might be able to escape (Enßlin 1999).
The non-thermal electron population produces bremsstrahlung, inverse Compton and synchrotron emission, which allows us to detect its presence, but also leads to energy losses which have to be compensated by re-acceleration (Jaffe 1977; Giovannini et al. 1993; Brunetti et al. 1999) or injection of fresh electrons. If the radio electrons are injected and not accelerated in-situ, the sources have to be distributed throughout the cluster volume. Otherwise, if only a few located sources produce the electrons, cooling times of the electrons are much shorter than plausible travel times from the source to the boundaries of the observed radio halos (Jaffe 1977; but see Holman, Ionson, & Scott 1979; Ragot & Kirk 1997). A promising injection mechanism is therefore secondary particle production from hadronic interactions of relativistic protons with the background gas (Dennison 1980; Vestrand 1982; Blasi & Colafrancesco 1999):
$`p+p`$ $``$ $`2N+\pi ^\pm `$
$`\pi ^\pm `$ $``$ $`\mu ^\pm +\nu _\mu /\overline{\nu }_\mu e^\pm +\nu _e/\overline{\nu }_e+\nu _\mu +\overline{\nu }_\mu `$
The lifetime of relativistic protons in the ICM is of the order of the Hubble time, or larger (Berezinsky, Blasi, & Ptuskin 1997). Thus they are able to travel large distances from their sources before they release their energy. The production of electrons via charged pions has to be accompanied by gamma ray production via neutral pions (Vestrand 1982; Enßlin et al. 1997; Colafrancesco & Blasi 1998; Blasi 1999):
$`p+p`$ $``$ $`2N+\pi ^o`$
$`\pi ^o`$ $``$ $`2\gamma `$
Thus clusters with radio halos might have gamma-ray halos, which would be, if detected, a direct proof for a hadronic origin of radio halos.
## 6. The Electron Spectrum
The electron spectrum is formed by various injection, cooling and acceleration processes, which should leave their fingerprints on it. Fig. 2 shows a compilation of the electron spectrum in the Coma cluster. Note, that the EUV and the HEX upper limits actually result from detections of excess emissions (compared to the thermal X-ray emission at these energies; Lieu et al. 1996; Fusco-Femiano et al. 1999), so that some of the upper limits actually could be data points.
The EUV-excess could result from a low temperature gas (Lieu et al. 1996). But this is problematic, since one would expect the presence of resonance lines, which are not detected (Dixon, Hurwitz, & Ferguson 1996). Therefore an IC origin of the excess is very likely. The scattered photon field could be the starlight of the galaxies, which could explain the narrow radial profile of the emission, but this would require a very high relativistic electron energy density in Coma (Enßlin, Lieu, & Biermann 1999). Probably the excess is due to IC scattered cosmic microwave background (CMB) photons (Hwang 1997; Enßlin & Biermann 1998; Bowyer & Berghöfer 1998), which would require a strong electron population around 150 MeV, not unreasonable since cooling is slowest at this energy (Sarazin & Lieu 1998). In order to connect this data point with the radio electrons a huge step in the electron spectrum is required, if the Faraday rotation determined magnetic field strength of $`6\mu `$G (Feretti et al. 1995) in Coma is correct. An electron population which was accelerated e.g. in a violent merger event in the past would have produced such a step due to cooling (Enßlin et al. 1999; Sarazin 1999; Atoyan & Völk 1999). If the magnetic fields are inhomogeneous, the required step can be further reduced, since cooling produces a electron distribution spatially anti-correlated with the magnetic fields, thus hiding the EUV producing electrons from radio observations.
The HEX-excess can not result from IC scattered starlight photons, since this would violate other limits to the electron spectrum (see Fig. 2). In principle it could result from IC scattered CMB photons, but this would require magnetic field strength of $`0.16\mu `$G since otherwise the synchrotron emission of the responsible electrons would exceed the observed radio halo of Coma (Fusco-Femiano et al. 1999). This is in strong contradiction to Faraday rotation measurements. Therefore it is very likely, that a supra-thermal electron population produces the HEX excess by bremsstrahlung (Enßlin et al. 1999; Sarazin & Kempner 1999). Such a population is expected due to turbulent particle acceleration of thermal electrons (Dogiel 1999; Liang 1999; Blasi 2000), and it can be confirmed with future CMB telescopes due to its non-thermal IC signature (Enßlin & Kaiser 2000).
### Acknowledgments.
I like to thank the scientific and the local organizing committee for the invitation to this stimulating conference. I am grateful for financial travel support from the International Astronomical Union (IAU) and the Deutsche Forschungsgemeinschaft (DFG). During this journey I enjoyed the friendly hospitality at the National Center for Radio Astrophysics (NCRA) in Pune, at the Inter University Centre for Astronomy and Astrophysics (IUCAA) in Pune, and the Raman Research Institute (RRI) in Bangalore.
## References
Atoyan, A. M., & Völk, H. J., 1999, ApJ, in press, astro-ph/9912557
Berezinsky, V. S., Blasi, P., & Ptuskin, V. S. 1997, ApJ, 487, 529
Blasi, P., & Colafrancesco, S. 1999, Astroparticle Physics, 12, 169
Blasi, P. 1999, ApJ, 525, 603
Blasi, P. 2000, ApJ, in press, astro-ph/0001344
Bowyer, S., & Berghöfer, T. W. 1998, ApJ 506, 502
Briel, U.G., & Henry, J.P. 1994, Nature, 372, 439
Brunetti, G., Setti, G., Feretti, L., & Giovannini G. 1999, submitted to MNRAS
Colafrancesco, S., & Blasi, P. 1998, Astroparticle Physics, 9, 227
Daly, R.A., & Loeb, A. 1990, ApJ, 364, 451
Deiss, B. M., Reich, W., Lesch, H. & Wielebinski, R. 1997, A&A, 321, 55
Dennison, B. 1980, ApJ, 239, 93
de Young, D. S. 1992, ApJ, 386, 464
Dixon, W. V., Hurwitz, M., & Ferguson, H. C. 1996, ApJ 469, L77
Dogiel, V. A. 1999, in Ringberg<sup>3</sup>, 259
Dolag, K., Bartelmann, M., & Lesch, H. 1999, A&A, 348, 351
Donnelly, R. H., Markevitch, M., Forman, W., Jones, C., David, L. P., Churazov, E., & Gilfanov, M. 1998, ApJ, 500, 138
Jaffe, W. J. 1977, ApJ, 212, 1
Eilek, J., & Wetherall, J. C. 1999, in Ringberg<sup>3</sup><sup>3</sup>3Ringberg Workshop on ‘Diffuse Thermal and Relativistic Plasma in Galaxy Clusters’, Eds: H. Böhringer, L. Feretti, P. Schuecker, MPE Report No. 271, 1999, 249
Enßlin, T. A., Biermann, P. L., Kronberg, P. P., & Wu, X.-P. 1997, ApJ, 477, 560
Enßlin, T. A., & Biermann, P. L. 1998, A&A, 330, 90
Enßlin, T. A., Biermann, P. L., Klein, U., & Kohle, S. 1998, A&A, 332, 39
Enßlin, T. A., Lieu, R., & Biermann, P. L. 1999, A&A, 344, 409
Enßlin, T. A. 1999, in Ringberg<sup>3</sup>, 275, astro-ph/9906212
Enßlin, T. A., & Kaiser, C. R. 2000, A&A submitted, astro-ph/0001429
Feretti, L., Dallacasa, D., Giovannini, G., & Tagliani, A. 1995, A&A, 302, 680
Feretti, L. 1999, these proceedings
Fusco-Femiano, R., dal Fiume, D., Feretti, L., Giovannini, G., Grandi, P., Matt, G., Molendi, S., & Santangelo, A. 1999, ApJ, 513, L21
Giovannini, G., Feretti, L., & Stanghellini, C. 1991, A&A, 252, 528
Giovannini, G., Feretti, L., Venturi, T., Kim, K. -T. & Kronberg, P. P. 1993, ApJ, 406, 399
Harris, D. E., & Miley, G. K. 1978, A&AS, 34, 117
Holman, G. D., Ionson, J. A. & Scott, J. S. 1979, ApJ, 228, 576
Hwang, C.-Y. 1997, Science 278, 1917
Kang, H., Rachen, J. P., & Biermann, P. L. 1997, MNRAS, 286, 257
Komissarov, S. S. 1985, Soviet Astronomy, 29, 382
Kronberg, P. P., Lesch, H., Hopp, U. 1999, ApJ 511, 56
Liang, H. 1999, in Ringberg<sup>3</sup>, 33
Lieu, R., Mittaz, J. P. D., Bowyer, S., Breen, J. O., Lockman, F. J., Murphy, E. M., & Hwang, C. -Y. 1996, Science, 274, 1335
Medina-Tanco, G., Enßlin, T. A., 2000, in preparation
Okoye, S. E. & Onuora, L. I. 1996, MNRAS, 283, 1047
Olinto, V.A., 1997, in 3rd RESCEU International Symposium on “Particle Cosmology”, University of Tokyo, 1997, astro-ph/9807051
Ragot, B. R., & Kirk, J. G. 1997, A&A, 327, 432
Rephaeli, Y. 1977, ApJ, 212, 608
Rephaeli, Y. 1979, ApJ, 227, 364
Roettiger, K., Burns, J., & Loken, C. 1993, ApJ, 407, L53
Roettiger K., Burns, J. O., Pinkney J. 1995, ApJ, 453, 634
Roettiger, K., Stone, J. M., & Burns, J. O. 1999, ApJ, 518, 594
Roettiger, K., Burns, J. O., & Stone, J. M. 1999, ApJ, 518, 603
Roland, J. 1981, A&A, 93, 407
Sánchez-Salcedo, F. J., Brandenburg, A., & Shukurov, A. 1998, Ap&SS, 263, 87
Sarazin, C. L., & Lieu, R. 1998, ApJ, 494, L177
Sarazin, C. L. 1999, ApJ, 520, 529
Sarazin, C. L., & Kempner, J. C. 1999, ApJ, in press, astro-ph/9911335
Schlickeiser, R., Sievers, A., & Thiemann, H. 1987, A&A, 182, 21
Tribble, P. C. 1993, MNRAS, 263, 31
Valtaoja, E. 1984, A&A, 135, 141
Vestrand, W. T. 1982, AJ, 87, 1266
Völk, H. J., Aharonian, F. A., & Breitschwerdt, D. 1996, Space Sci. Rev., 75, 279
Völk, H. J., & Atoyan, A.M. 1999, in Ringberg<sup>3</sup>, 99
Wetherall, J.C., & Eilek, J. 1999, in Ringberg<sup>3</sup>, 255
Willson, M. A. G. 1970, MNRAS, 151, 1 |
no-problem/0001/cond-mat0001042.html | ar5iv | text | # Replica-exchange molecular dynamics simulation for supercooled liquids
## Abstract
We investigate to what extend the replica-exchange Monte Carlo method is able to equilibrate a simple liquid in its supercooled state. We find that this method does indeed allow to generate accurately the canonical distribution function even at low temperatures and that its efficiency is about 10-100 times higher than the usual canonical molecular dynamics simulation.
If a liquid is cooled to a temperature close to its glass transition temperature, its dynamical properties show a drastic slowing-down. At the same time, a crossover from highly unharmonic liquid-like behavior to harmonic solid-like behavior is expected in its static (thermodynamic) properties at a certain temperature $`T_K`$, the Kauzmann temperature . Very recently the value of $`T_K`$ of simple model liquids have been determined analytically and numerically and some possibilities of a thermodynamic glass transition at $`T_K`$ have been discussed. Although the values of $`T_K`$ obtained with the different methods are consistent with each other, it was necessary for the numerical calculations of $`T_K`$ to extrapolate high temperature data ($`T0.45`$) of the liquid and disordered solid branches of the configurational entropy $`S(T)`$ down to significantly lower temperatures ($`T_K0.3`$). With a guide of an analytic prediction for liquids, $`S(T)T^{0.4}`$ , and for harmonic solids, $`S(T)\mathrm{log}T`$, a crossing of the two branches has been found and used to calculate $`T_K`$. However, to make those observations more reliable, very accurate calculations of thermodynamics properties are necessary in the deeply supercooled regime, which is difficulat since the typical relaxation times of the system are large.
In recent years, several efficient simulation algorithms have been developed to generate canonical distributions also for complex systems. Examples are the multi-canonical , the simulated tempering , and the replica-exchange (RX) methods. Although these methods were originally developed for Ising-type spin systems, their applications to any off-lattice model by use of Monte Carlo or molecular dynamics simulations are rather straightforward . However, it has been found that the application of some of these algorithms to supercooled liquids or structural glasses is of only limited use . The main motivation of the present paper is to test the efficiency of the RX method, which seems to be in many cases the most efficient algorithm, to the case of highly supercooled liquids .
The system we study is a two-component ($`AB`$) Lennard-Jones mixture, which is a well characerized model system for supercooled simple liquids. The total number of particles is $`N=1000`$, and they interact via the (truncated and shifted) potential $`\varphi _{\alpha \beta }(r_{ij})=4ϵ_{\alpha \beta }[(\sigma _{\alpha \beta }/r_{ij})^{12}(\sigma _{\alpha \beta }/r_{ij})^6]`$, where $`r_{ij}`$ is the distance between particles $`i`$ and $`j`$, and the interaction parameters are $`\alpha ,\beta A,B`$, $`ϵ_{AA}=1`$, $`ϵ_{AB}=1.5`$, $`ϵ_{BB}=0.5`$, $`\sigma _{AA}=1`$, $`\sigma _{AB}=0.8`$, and $`\sigma _{BB}=0.88`$. Other simulation parameters and units are identical as in . The time step $`\mathrm{\Delta }t`$ for numerical integration is $`0.018`$.
The algorithm of our replica-exchange molecular dynamics (RXMD) simulation is essentially equivalent to that of Ref. , and therefore we summarize our simulation procedure only briefly. (i) We construct a system consisting of $`M`$ noninteracting subsystems (replicas), each composed of $`N`$ particles, with a set of arbitrary particle configurations $`\{𝒒_1,\mathrm{},𝒒_M\}`$ and momenta $`\{𝒑_1,\mathrm{},𝒑_M\}`$. The Hamiltonian of the $`m`$-th subsystem is given by
$$H_m(𝒑_m,𝒒_m)=K(𝒑_m)+\mathrm{\Lambda }_mE(𝒒_m),$$
(1)
where $`K`$ is the kinetic energy, $`E`$ is the potential energy, and $`\mathrm{\Lambda }_m\{\lambda _1,\mathrm{},\lambda _M\}`$ is a parameter to scale the potential. (ii) A MD simulation is done for the total system, whose Hamiltonian is given by $`=_{m=1}^MH_m`$, at a constant temperature $`T=\beta _0^1`$ using the constraint method . Step (ii) generates a canonical distribution $`P(𝒒_1,\mathrm{},𝒒_M;\beta _0)=\mathrm{\Pi }_{m=1}^MP(𝒒_m;\mathrm{\Lambda }_m\beta _0)\mathrm{exp}[\beta _0_{m=1}^M\mathrm{\Lambda }_mE(𝒒_m)]`$ in configuration space . (iii) At each time interval $`\mathrm{\Delta }t_{RX}`$, the exchange of the potential scaling parameter of the $`m`$-th and $`n`$-th subsystem are considered, while $`\{𝒒_1,\mathrm{},𝒒_M\}`$ and $`\{𝒑_1,\mathrm{},𝒑_M\}`$ are unchanged. The acceptance of the exchange is decided in such a way that it takes care of the condition of detailed balance. Here we use the Metropolis scheme, and thus the acceptance ratio is given by
$$w_{m,n}=\{\begin{array}{cc}1,\hfill & \mathrm{\Delta }_{m,n}0\hfill \\ \mathrm{exp}(\mathrm{\Delta }_{m,n}),\hfill & \mathrm{\Delta }_{m,n}>0,\hfill \end{array}$$
(2)
where $`\mathrm{\Delta }_{m,n}=\beta _0(\mathrm{\Lambda }_n\mathrm{\Lambda }_m)(E(𝒒_m)E(𝒒_n))`$. (iv) Repeat steps (ii) and (iii) for a sufficient long time. This scheme leads to canonical distribution functions $`P(E;\beta _i)`$ at a set of inverse temperatures $`\beta _i=\lambda _i\beta _0`$. To make a measurement at an inverse temperature $`\beta _l`$ one has to average over all those subsystems $`(i1,\mathrm{},M)`$ for which we have (temporarily) $`\beta _l=\lambda _i\beta _0`$. Usual canonical molecular dynamics (CMD) simulations are realized if we skip step (iii).
In the present simulation, we take $`M=16`$, $`\beta _0=0.45^1`$, $`\lambda _i=10.0367(i1)`$ and thus cover a temperature range $`0.45T1`$. Exchange events are examined only between subsystems that have scaling parameters $`\lambda _i`$ and $`\lambda _{i+1}`$ that are nearest neighbors; the events with $`i=1,3,5,\mathrm{}`$ or $`i=2,4,6,\mathrm{}`$ are repeated alternatively every $`\mathrm{\Delta }t_{RX}`$ intervals. We find that the highest average acceptance ratio for this type of move is $`0.186`$ for the exchange of $`\lambda _1`$ and $`\lambda _2`$, and the lowest is $`0.027`$ for $`\lambda _{15}`$ and $`\lambda _{16}`$. Although these values can be made more similar by optimizing the different gaps between $`\lambda _i`$ and $`\lambda _{i+1}`$ for a fixed choice of $`\lambda _1`$ and $`\lambda _M`$, only small improvements were obtained by such a simple optimization in our case. We also note that the choice of $`\mathrm{\Delta }t_{RX}`$ strongly affects the efficiency of the RX method; $`\mathrm{\Delta }t_{RX}`$ should be neither too small or too large . We used $`\mathrm{\Delta }t_{RX}=10^3\mathrm{\Delta }t`$, a time which is a bit larger than the one needed for a particle to do one oscillation in its cage, and data are accumulated for $`0t5\times 10^6\mathrm{\Delta }t`$ after having equilibrated the system for the same amount of time. At the beginning of the production run, the subsystems were renumbered so that at $`t=0`$ we had for all $`m`$ $`\mathrm{\Lambda }_m=\lambda _m`$.
In Fig. 1(a), we show the time evolution of the subsystems in temperature space. One can see that the subsystems starting from the lowest ($`m=1`$) and the highest ($`m=16`$) temperature explore both the whole temperature space from $`i=1`$ to $`16`$. Fig. 1(b) presents the mean squared displacements (MSD)
$$\mathrm{\Delta }R^2(t)=|𝒒_m(t)𝒒_m(0)|^2/N$$
(3)
for the RXMD (with $`m=1`$) and for the CMD performed at $`T=(\lambda _1\beta _0)^1=0.45`$. From this figure we recognize that, due to the temperature variation in the RXMD method, the system moves very efficiently in configuration space, while in the CMD the system is trapped in a single metastable configuration for a very long time. If one uses the MSD to calculate an effective diffusion constant, one finds that this quantity is around 100 times larger in the case of the RXMD than in the CMD case, thus demonstrating the efficiency of the former method.
Fig. 2 shows the canonical distribution function of the total potential energy at the different temperatures,
$$P_i(E)P(E;\lambda _i\beta _0),$$
(4)
obtained by a single RXMD simulation. For adjacent temperatures the corresponding distribution functions should have enough overlap to obtain a reasonable exchange probabilities and hence can be used to optimize the efficiency of the algorithm. Further use of these distribution functions can be made by using them to check whether or not one has indeed equilibrated the system. Using the reweighting procedure , it is in principle possible to calculate the canonical distribution functions
$$P_i(E;\lambda _j\beta _0)=\frac{P_i(E)\mathrm{exp}[(\lambda _i\lambda _j)\beta _0E]}{𝑑E^{}P_i(E^{})\mathrm{exp}[(\lambda _i\lambda _j)\beta _0E^{}]}$$
(5)
at a new temperature $`T_j=(\lambda _j\beta _0)^1`$ from any $`P_i(E)`$. Note that in equilibrium the left hand side should be independent of $`i`$ to within the accuracy of the data.
In Fig. 3 we plot different $`P_i(E;\lambda _4\beta _0)`$, using as input the distributions $`P(E;\lambda _i\beta _0)`$ for $`1i8`$, obtained from RXMD (a) and CMD (b) simulations. (Both simulations extended over $`8.7\times 10^4`$ time units.) We see that in the case of the RXMD the different distributions $`P_i`$ fall nicely on top of each other in the whole energy range, thus giving evidence that the system is indeed in equilibrium. In contrast to this, the different distributions of the CMD, Fig 3b, do not superimpose at low energies (=low temperatures), thus demonstrating the lack of equilibration. This can be seen more clealy by comparing Fig. 3(c) and (d), where $`P_i(E;\lambda _1\beta _0)`$ is plotted.
Fig. 4 shows the temperature dependence of the potential energy $`E(T)`$ obtained from RXMD simulations via
$$E(T_j)=𝑑E^{}P(E^{};\lambda _j\beta _0)E^{}.$$
(6)
For the sake of comparison we have also included in this plot data from CMD with the same length of the production run as well as data from CMD simulations which were significantly longer (about one order of magnitude) . The solid line is a fit to the RXMD results with the function $`E(T)=E_0+AT^{0.6}`$, a functional form suggested by analytical calculations . One can see that RXMD and CMD results coincide at higher temperatures, but deviations become significant at low temperatures (see Inset). Furthermore, we see that the present RXMD results agree well with CMD data of the longer simulations.
As a final check to see whether the RXMD is indeed able to equilibrate the system also at low temperatures, we have calculated the temperature dependence of the (constant volume) heat capacity $`C_v(T)`$ via the two routes
$`C_v(T)`$ $`=`$ $`E(T)/T`$ (7)
$`=`$ $`(E^2E^2)/T^2,`$ (8)
and plot the results in Fig. 5. Again we see that within the accuracy of our data the two expressions give the same answer, thus giving evidence that the system is indeed in equilibrium.
Summary: We have done replica-exchange molecular dynamics and canonical molecular dynamics simulations for a binary Lennard-Jones mixture in order to check the efficiency of the replica-exchange method for a structural glass former in the strongly supercooled regime. We find that at low temperatures the RXMD is indeed significantly more efficient than the CMD, in that the effective diffusion constant of the particles is around 100 times larger in the RXMD. However, accurate simulations are still difficult for $`T<0.45`$ even with RXMD. Finding an optimal choice of $`M`$, $`\{\lambda _1,\mathrm{},\lambda _M\}`$, and $`\mathrm{\Delta }t_{RX}`$ may be important in order to allow simulations also for $`T<0.45`$ within reasonable computation times. Furthermore it might be that the efficiency of RXMD improves even more if one uses it below the critical temperature of mode-coupling theory , since there is evidence that below this temperature the nature of the energy landscape is not changing anymore .
The authors acknowledge the financial support from the DFG through SFB 262. RY acknowledges the Grants in Aid for Scientific Research from the Ministry of Education, Science, Sports and Culture of Japan and thanks Prof. B. Kim for valuable discussions. Calculations have been performed at the Human Genome Center, Institute of Medical Science, University of Tokyo. |
no-problem/0001/hep-th0001055.html | ar5iv | text | # Discrete Torsion, AdS/CFT and Duality
## 1 Introduction
Orbifold singularities have been the object of much attention in string theory, as they provide solvable examples of conformal field theories and correspond to degenerate limits of Calabi-Yau families of vacua. Early on it was noticed that on calculating the spectrum of an orbifold there is an inherent ambiguity in the phases chosen for the twisted sector. This ambiguity, discrete torsion, is classified by the group $`H^2(\mathrm{\Gamma },U(1))`$, where $`\mathrm{\Gamma }`$ is the orbifold group.
Discrete torsion has been analyzed in the context of deformation theory in where it was noticed that discrete torsion singularities cannot be geometrically resolved, and that they provide totally distinct branches for compactifications of string theory. As a parameter, discrete torsion can be naturally associated to the $`NSNS`$ 2-form in string theory, and thus it is natural to interpret it in terms of gerbes.
More recently, it has been shown that discrete torsion can be analyzed in terms of D-branes at the orbifold by taking into account projective representations of the orbifold group. This is also needed for Matrix theory setups, as in . In this way one builds quiver theories for the modified orbifold which have a totally different content than the theories without discrete torsion. This analysis has been extended to the boundary state formalism in .
A new tool which to date has not been exploited to analyze this situation is to make use of the AdS/CFT correspondence . In particular, the large $`N`$ limit of D3 branes at an orbifold was discussed in , but discrete torsion not considered. In this paper we analyze orbifolds with discrete torsion within the AdS/CFT correspondence.
The first task is to describe the physics of a particular type of orbifold with $`\mathrm{\Gamma }=_m\times _n`$. We find the low energy field theory decsription of D3-branes at the singularity with discrete torsion for all possible choices of discrete torsion. This is done in Sections 2,3. The theories obtained are supersymmetric four-dimensional field theories of the brane-box type, with a modified superpotential. The orbifold can be recovered by the moduli space of vacua, which depends on the superpotential. This calculation is done in Section 4. We finish the field theory study in Section 5 where we compute the spectrum of chiral primaries of the CFT at the origin.
Secondly, in Section 6 we study the large $`N`$ theories associated to the orbifold with discrete torsion. The first fact that we need to consider is that they lie in the moduli space of deformations of more standard quiver diagrams, as can be argued along the lines of . By following the deformations in supergravity, we are able to find a duality of type IIB string theories at weak coupling compactified on very different spaces. This is an example of mirror symmetry for RR backgrounds.
Subgroups of $`\mathrm{\Gamma }`$ leave 2-planes fixed, which descend to fixed circles in the large $`N`$ limit. By a careful analysis of the string theory at the orbifold one can find the correct boundary conditions for the corresponding twisted sector states. These states are massless and hence survive in the supergravity limit. Thus discrete torsion is encoded in the near-horizon geometry, through different choices of boundary conditions for twisted fields. One can compare these results with the field theory from Section 5, and agreement is found. As a by-product we obtain geometric information on the orbifold with discrete torsion.
The analysis shows that if one considers resolutions by blowups of the classical geometry, on going around the fixed circles they are affected by monodromy. The monodromy encodes the discrete torsion, and shows that the topology of the $`S^5/\mathrm{\Gamma }`$ is very different for the cases with and without discrete torsion. In particular, with a non-trivial monodromy one finds torsion classes in the homology, and then the discrete torsion parameter is the B-field flux around this homology two-cycle. This difference in topology also explains why the deformation theory of the orbifolds with and without discrete torsion are so different.
Section 7 deals with new dualities for large $`N`$ field theories. Two theories with the same field content at different values of marginal couplings have the same moduli space of flat directions and the same spectrum of chiral primary operators, and, furthermore, the coupling that distinguishes them does not appear in planar diagrams. The duality acts discontinuously on the space of couplings by a permutation of the roots of unity.
## 2 The orbifold
We analyze orbifolds of the Type IIB superstring of the form $`^3/\mathrm{\Gamma }`$ with $`\mathrm{\Gamma }=_m\times _n`$, and whose generators act on the coordinates by
$`e_1:(z_1,z_2,z_3)(\alpha z_1,\alpha ^1z_2,z_3)`$ (1)
$`e_2:(z_1,z_2,z_3)(z_1,\beta z_2,\beta ^1z_3)`$ (2)
with<sup>1</sup><sup>1</sup>1Throughout, we use the notation $`\omega _ke^{2\pi i/k}`$. $`\alpha =\omega _m`$ and $`\beta =\omega _n`$.
Because of the choice of action of the orbifold, we preserve $`N=2`$ supersymmetry in four dimensions, which can be broken to $`N=1`$ by the addition of D-branes at the singularity.
The discrete torsion parameter, is a ‘bilinear’ element $`ϵH_2(\mathrm{\Gamma },U(1))`$ such that
$`ϵ(g_1,g_2g_3)`$ $`=`$ $`ϵ(g_1,g_2)ϵ(g_1,g_3)`$ (3)
$`ϵ(g_1,g_2)`$ $`=`$ $`\overline{ϵ}(g_2,g_1)`$ (4)
$`ϵ(g_1,g_1)`$ $`=`$ $`1`$ (5)
for $`g_1,g_2,g_3\mathrm{\Gamma }`$. As $`\mathrm{\Gamma }`$ has two generators, $`ϵ`$ is completely determined by the number
$$\eta =ϵ(e_1,e_2)$$
(6)
It is easy to see that the solution of these equations is as follows. If $`m,n`$ are relatively prime, there is no discrete torsion. If we denote $`p=gcd(m,n)`$, then we find $`H_2(\mathrm{\Gamma },U(1))=_p`$. This includes the oft-studied special case $`m=n`$, where clearly $`p=n`$. The discrete torsion is thus determined by $`\eta =\omega _p^r`$, where $`r=0,1,\mathrm{},p1`$. Let $`s`$ be the smallest non-zero number such that $`\eta ^s=1`$. It is the integer $`s`$ which determines the physics, and for each value of $`s`$ the orbifolds will behave differently; for example, they will have a different resolution of singularities.
For later use, it will be useful to define a common root of unity $`\lambda `$, such that $`\lambda ^{mn}=1`$. This phase will appear in the superpotential of the low energy gauge theory. It will also be useful to define $`q=lcm(m,n)`$, the least common multiple of $`m,n`$.
## 3 D-branes on the orbifold
We want to locate D3-branes at the orbifold. The construction for the low energy effective action follows . In order to incorporate the discrete torsion parameter into the field theory, one has to consider all possible projective representations of the orbifold group with the same cocycle. That is, one considers representations where
$$\gamma (e_1)\gamma (e_2)=\eta \gamma (e_1e_2)$$
(7)
It is easy to see that given a projective representation $`R`$ of the group $`\mathrm{\Gamma }`$ and a non-projective representation $`\chi `$, the representation $`R\chi `$ is a projective representation of $`\mathrm{\Gamma }`$ with the same discrete torsion parameter as $`R`$. By tensoring $`R`$ with all possible representations of $`\mathrm{\Gamma }`$ we can obtain all possible projective representations of $`\mathrm{\Gamma }`$.
We can thus build the analog of the regular representation of $`\mathrm{\Gamma }`$ by considering $`R=_idim(R_i)R_i`$, with $`R_i`$ all the unitarily inequivalent projective representations of $`\mathrm{\Gamma }`$, and out of the regular representation we should obtain the quiver diagram representing a D-brane which is outside the singularity. Any other D-brane configuration can be obtained by choosing the representation to be non-regular . String consistency imposes anomaly cancellation on the configuration.
One irreducible projective representation of $`\mathrm{\Gamma }`$ is given in terms of $`s\times s`$ matrices
$$\gamma (e_1)=\text{diag}(1,\eta ^2,\eta ^4,\mathrm{},\eta ^{2(s1)}),\gamma (e_2)=\left(\begin{array}{ccccc}0& 1& 0& \mathrm{}& 0\\ 0& 0& 1& \mathrm{}& 0\\ \mathrm{}& \mathrm{}& \mathrm{}& \mathrm{}& \mathrm{}\\ 1& 0& 0& \mathrm{}& 0\end{array}\right)$$
(8)
Any other irreducible representation is given by multiplying $`\gamma (e_1)`$ by an $`m^{th}`$ root of unity $`\alpha ^k`$ and $`\gamma (e_2)`$ by an $`n^{th}`$ root of unity $`\beta ^{\mathrm{}}`$. This is equivalent to multiplying the representation $`R_i`$ by a character of the orbifold group, and thus producing another projective representation of the group, which is also irreducible. Two representations labeled by $`(k,\mathrm{})`$ and $`(k^{},\mathrm{}^{})`$ are equivalent if $`k=k^{}modm/s`$ and $`\mathrm{}=\mathrm{}^{}modn/s`$, as can be seen by decomposing the representation into either characters of $`_m`$ or $`_n`$.
As all the irreducible projective representations are $`s`$-dimensional, we obtain the regular representation by taking all possible different projective representations with the same cocycle, multiplied by the dimension of the representation.
$$R=_{k<m/s,\mathrm{}<n/s}sR_k\mathrm{}$$
(9)
The total dimension of $`R`$ is $`mn`$, which is the order of the group, $`|\mathrm{\Gamma }|`$.
The gauge theory living on the D-branes is given by a quiver diagram with $`\frac{m}{s}\times \frac{n}{s}`$ nodes, one for each representation. This theory is dual to an elliptic brane-box, as in , a part of which appears in Figure 1.
The nodes indicate the different $`(k,\mathrm{})`$ gauge groups corresponding to each irreducible representation. To each of them we associate the gauge group $`U(N_k\mathrm{})`$. For the regular representation, all of the gauge groups have equal rank, in particular $`N_k\mathrm{}=s`$. The arrows represent chiral multiplets transforming in the $`(N,\overline{N})`$ of the groups it connects. The nodes are connected by multiplications by specific characters of $`\mathrm{\Gamma }`$.
Let us refer to the chiral multiplets corresponding to the horizontal lines as $`\varphi _{1k\mathrm{}}`$, the vertical ones as $`\varphi _{3k\mathrm{}}`$ and the diagonal ones as $`\varphi _{2k\mathrm{}}`$. To fully specify the theory, one also needs to give the superpotential. As brane box diagrams dual to theories with discrete torsion and without discrete torsion coincide, they can only differ in the superpotential .
The superpotential is of the form
$$\sqrt{2}g\mathrm{tr}(\varphi _1\varphi _2\varphi _3\lambda ^s\varphi _1\varphi _3\varphi _2)$$
(10)
A cubic term appears for each small triangle in the quiver, with a phase difference, depending on the orientation of the given triangle. Notice that it is the phase $`\lambda `$ that appears in the superpotential, instead of $`\eta `$. To see that this is correct, consider rephasing all of the chiral fields by arbitrary phases. This transformation would induce superpotential terms with a set of couplings $`\lambda _C`$ ($`\lambda _A`$) for (anti-)clockwise triangles.
$$\sqrt{2}g\mathrm{tr}(\lambda _A\varphi _1\varphi _2\varphi _3\lambda _C\varphi _1\varphi _3\varphi _2)$$
(11)
There is one combination of couplings which is invariant under field redefinitions, and hence cannot be eliminated. It is the quotient
$$\frac{_{\mathrm{\Delta }_C}\lambda _C}{_{\mathrm{\Delta }_A}\lambda _A}=\eta $$
(12)
and turns out to equal the discrete torsion. In this way, the discrete torsion is encoded in the theory. With our choice of conventions, $`\lambda _C=\lambda ^s,\lambda _A=1`$. We have chosen the couplings so that they all are the same, and therefore the quantum symmetry is easier to construct. We could just as well have made all the couplings equal except at one triangle, where we would insert the phase $`\eta `$. Notice that when $`m=n`$ with only one node in the diagram, this reduces to the case considered in .
It should be possible to deduce the form of the superpotential directly through string computations. In particular, we wish to show that the diagrams in Figure 2 differ essentially in phase.
In the field theory, there is a vacuum diagram which encodes the phase $`\eta `$, and can thus be traced back to the superpotential (10). It uses each of the cubic vertices exactly once, and thus is proportional, in field theory, to the ratio of phases (12) (or its inverse, depending on chirality). When drawn on top of the quiver, it has the topology of a net of hexagons (one for each node), as shown in Figure 3.
In string theory, we would evaluate this using projected fields $`\varphi _{a,inv}=_{g\mathrm{\Gamma }}r_a(g)\gamma ^1(g)\varphi _a\gamma (g)`$. The diagram is equivalent to a genus one worldsheet with $`mn/s^2`$ holes (one for each node in the quiver). Thus the result is a product of traces, one for each node labelled $`i,j`$, each of the form
$$\mathrm{tr}_{R_{ij}}\gamma ^1(g_{ij,1})\gamma (g_{ij,2})\gamma ^1(g_{ij,3})\gamma (g_{ij,4})\gamma ^1(g_{ij,5})\gamma (g_{ij,6})$$
(13)
where the group elements are paired, viz $`g_{i,j,3}=g_{i+1,j,6}^1,g_{i,j,4}=g_{i+1,j+1,1}^1,g_{i,j,5}=g_{i,j+1,2}^1`$. This result is the analogue of the single node computation of Ref. .
It should be possible to reduce the product of traces by implementing the relation (7) and using character formulas. Furthermore, the results given later in the paper are fully consistent with the form of the superpotential, (10).
## 4 The moduli space of a regular D-brane
In order to verify that we have the correct field theory, we should consider the moduli space of vacua for a regular D-brane. In particular, one should have a three dimensional component where the D-brane lives away from the singularities, which corresponds to the total orbifold space minus the fixed lines. When the D-brane approaches one of the fixed lines, or the fixed point, the D-brane can fractionate, and the dimension of the moduli space jumps.
From the orbifold point of view, the moduli space is described by six variables with three relations. Only the $`\mathrm{\Gamma }`$-invariant polynomials in the variables are good functions on the space. This polynomial ring is generated by $`x=z_1^m`$, $`y=z_3^n`$, $`z=z_2^q`$, $`u=(z_1z_2)^n`$, $`v=(z_2z_3)^m`$ and $`w=z_1z_2z_3`$, where $`q`$ is the least common multiple of $`m,n`$. There are three relations<sup>2</sup><sup>2</sup>2Note that when $`m=n=q`$, the variables $`u,v`$ are redundant, and the system reduces to just $`f_1=0`$.
$`f_1`$ $`=`$ $`x^{q/m}y^{q/n}zw^q=0`$ (14)
$`f_2`$ $`=`$ $`vxw^m=0`$ (15)
$`f_3`$ $`=`$ $`uyw^n=0`$ (16)
and we have three complex lines of singularities where two out of the three $`x,y,z`$ are equal to zero. These are the lines which are fixed by some elements of the orbifold.
From the gauge theory, we need to construct the equivalent of variables $`x`$, $`y`$, $`z`$, $`u`$, $`v`$, $`w`$ in terms of gauge invariant polynomials of the chiral fields. As is well known from Matrix theory , one should hope to replace the equations (14)–(16) by matrix versions. If the matrices are made to commute, then we can identify variables $`x,y,z,u,v,w`$ for individual branes by their eigenvalues. In particular, if we want branes in the bulk, we should require that $`x,y,z`$ be all invertible (in this case as matrices).
Referring to the quiver, the natural gauge invariant coordinates are given by traces of closed loops in the diagram. Matrix coordinates will be given by taking the same closed loops without the trace, as they represent matrices whose eigenvalues are gauge invariant quantities, since gauge transformations act by similarity transformations.
Let us consider the variables $`\stackrel{~}{x}_j=_{i=1}^{m/s}\varphi _{1ij}`$, and $`\stackrel{~}{y}_j=_{i=1}^{n/s}\varphi _{3ji}`$, and similarly $`\stackrel{~}{z}_j`$ along diagonals. It is easy to see that all the $`\stackrel{~}{x}_i`$ are isospectral up to phases, the same being true for all the $`\stackrel{~}{y}_j`$, under the assumption that all the chiral matrices that appear in the diagram are invertible and satisfy the $`F`$-constraints. The proof goes by making use of moves which relate different matrices, and by application of the Cayley-Hamilton theorem. An example of a move is shown in Figure 4.
The simple moves are given by solving the $`F`$ constraints for a single chiral multiplet. Under such a move, we get a phase given by the ratio of the two associated triangle couplings in the superpotential.
The eigenvalues of the matrix $`\stackrel{~}{x}_j`$ are determined by the roots of the characteristic polynomial of $`\stackrel{~}{x}_j`$. As $`\stackrel{~}{x}_j`$ satisfies its own characteristic polynomial, we have an identity $`P(\stackrel{~}{x}_j)=0`$, which we can use as a matrix equation. We can now change a row by multiplying by $`\varphi _2`$ at the end. As we want a matrix equation that begins and ends on the same node, we can complete it by multiplication by the rest of the triangle that returns the operator to the starting node. We can now make use of the moves to shift the polynomial equation one step down, as shown in the Figure 5.
Indeed, apart from a global phase in the roots, the two characteristic polynomials are the same, and therefore the two matrices have the same content. If we multiply by the triangle operator and use the $`F`$-term constraints, we can move the polynomial equation along the rows, and hence all of the $`\stackrel{~}{x}_j`$ have the same roots (modulo phases), and therefore they are equivalent when considered as matrix variables. Similarly for all the $`\stackrel{~}{y}_k`$ and $`\stackrel{~}{z}_k`$. As a result, we need only consider the matrices $`\stackrel{~}{x}_1`$, $`\stackrel{~}{y}_1`$, $`\stackrel{~}{z}_1`$.
Now, using the moves, we can also see using (12) that
$$\stackrel{~}{x}_1\stackrel{~}{y}_1=\eta \stackrel{~}{y}_1\stackrel{~}{x}_1$$
(17)
This means that if $`\stackrel{~}{y}_1`$ is invertible, then the eigenvalues of $`\stackrel{~}{x}_1`$ are only well defined modulo $`\eta `$, and hence the trace of $`\stackrel{~}{x}_1`$ vanishes, so $`\stackrel{~}{x}_1`$ alone is not a good matrix variable to describe the moduli space. On the other hand, it shows us also that the matrices must be of order $`s\times s`$ in order to find a representation of the algebraic equation (17). This is exactly the size of the matrices in the regular representation of the quiver.
Now we can take $`x=\stackrel{~}{x}_1^s`$, and similarly $`y=\stackrel{~}{y}_1^s`$ and $`z=\stackrel{~}{z}_1^s`$. These matrices will commute amongst themselves, and they have $`s`$ identical eigenvalues. The matrix $`w`$ is constructed from any small triangle in the graph, and it is easy to see that $`w`$ commutes with all $`x,y,z`$. $`w`$ is defined up to a phase depending only on the orientation of the triangle.
One can then see that the matrices $`x,y,z,w`$ satisfy eq. (14) by making use of the $`F`$-constraint moves. As the matrices $`x,y,z,w`$ are all $`s\times s`$ and proportional to the identity, we can identify the eigenvalues with the coordinates $`x,y,z,w`$ of the algebraic variety which describes the orbifold. The two different choices of $`w`$ differ by a root of unity which cancels in the equation, as the order of $`\lambda ^s`$ divides $`q`$. Similar analyses can be used to construct the constraints (15)–(16), using the zigzags $`u_j=_{i=1}^{n/s}\left((\varphi _1\varphi _2)_{ij}\right)^s`$ and $`v_i=_{j=1}^{m/s}\left((\varphi _2\varphi _3)_{ij}\right)^s`$.
For a total number of branes which is $`N`$ times that of the regular representation, we get mutually diagonalizable matrices whose eigenvalues have a common multiplicity of $`s`$ each. Hence, the spectrum of regular $`D3`$ branes gives a moduli space which is a symmetric product of the moduli space of a single regular brane. Namely
$$_N=_1^N/S_N$$
(18)
Notice that if we relax the invertibility of the matrices, then the branes can only fractionate at the fixed complex lines, as we need two out of the three matrices $`x,y,z`$ to be non-invertible in order for the previous arguments to fail, which is just as expected.
Consider now adding fractional $`D3`$ branes. The configuration is allowed only if the gauge theory anomaly vanishes, that is, if the fractional D-brane is constructed by combinations of effective $`N=2`$ directions . The effective $`N=2`$ directions are associated with the fixed complex lines of subgroups of $`\mathrm{\Gamma }`$, on which fractional branes can reside because the orbifold singularity is not isolated. In principle, one can see a brane moving in the Coulomb branch only if it is a linear combination of $`N=2`$ effective directions, and thus is automatically anomaly free. This will be important when we compare with the supergravity analysis.
## 5 Spectrum of chiral operators
In order to make the comparison with supergravity later on, we need to calculate the spectrum of single trace chiral primaries .
The quantum symmetry generators $`Q_1`$, and $`Q_2`$, which correspond to the group of characters of $`\mathrm{\Gamma }`$ act by
$`Q_1`$ $`:`$ $`\varphi _{1,i,j}\alpha \varphi _{1,i+1,j}`$ (21)
$`\varphi _{2,i,j}\alpha ^1\varphi _{2,i+1,j}`$
$`\varphi _{3,i,j}\varphi _{3,i+1,j}`$
$`Q_2`$ $`:`$ $`\varphi _{1,i,j}\varphi _{1,i,j+1}`$ (24)
$`\varphi _{2,i,j}\beta \varphi _{2,i,j+1}`$
$`\varphi _{3,i,j}\beta ^1\varphi _{3,i,j+1}`$
and they only act by shifts on the gauge fields. It is obvious that they are invariances of the field theory as long as all gauge couplings and Yukawa couplings are equal, which is the orbifold point. This identification of the quantum symmetry is necessary in order to define the twist charge of operators in the field theory. In particular, untwisted operators correspond to quantum symmetry singlets. The field theory also has a $`U(1)^3`$ symmetry. One of the $`U(1)`$’s is the $`R`$-symmetry and the other $`U(1)`$’s twist the field $`\varphi _1,\varphi _2,\varphi _3`$ by global phases.
Untwisted single-trace chiral primary operators are given schematically by
$$O_{abc}=\mathrm{tr}\varphi _1^a\varphi _2^b\varphi _3^c$$
(25)
for any closed loop in the quiver. In order for the operator to be a conformal primary, it must be such that dragging one field around and using the cyclicity of the trace induces no phase. The true operator is a linear combination of all of the possible terms with the same structure that are related by the $`F`$ constraints. The other linear combinations will correspond to chiral fields which are not primary.
For a primary to be untwisted, it must satisfy a more stringent constraint. It is straightforward to show that untwisted primaries may be written as
$$\mathrm{tr}P(\varphi _1^m,\varphi _2^q,\varphi _3^n,\varphi _1\varphi _2\varphi _3,(\varphi _1\varphi _2)^n,(\varphi _2\varphi _3)^m)$$
(26)
where $`P`$ is any monomial. The point is that the variables in $`P`$ all transform without phase under the quantum symmetry and commute with one another, so that they are not constrained by the presence of discrete torsion. These monomials are in one-to-one correspondence with monomials in $`x`$, $`y`$, $`z`$, $`u`$, $`v`$, $`w`$ of the last section. These operators are in fact superconformal primaries, and additional conformal primaries may be constructed by inserting factors of $`W_\alpha ^2`$.
Twisted states come in families of the form
$$𝒪_1^a=\mathrm{tr}\varphi _1^{am/s},$$
(27)
for some integer $`a`$, in multiplicity $`n/s`$. Similar twisted operators occur in $`\varphi _2`$ and $`\varphi _3`$ as well. These are operators where one closes loops horizontally, vertically or diagonally. It is a non-trivial fact that these operators exhaust the list of twisted superconformal primaries. Note that the operator with $`a`$ a multiple of $`s`$ is untwisted.
## 6 Supergravity duals
### 6.1 Massless Twisted States and Monodromy
The supergravity duals of the previous D-brane configurations are obtained by looking at the near-horizon geometry of the $`D3`$-branes at the orbifold, and they will correspond to geometries of the type $`AdS_5\times S^5/\mathrm{\Gamma }`$, where $`\mathrm{\Gamma }`$ acts by isometries of the $`S^5`$. Our first task is to understand how the orbifold encodes the discrete torsion.
Because of our choice of group action, some orbifold elements will leave invariant three fixed circles of the $`S_5`$, which descend from the fixed planes of the orbifold. Along these circles, we have locally a singularity of the type $`^2/_k`$, where the $`_k`$ is abelian and generated by the element that fixes the point, that is, we will have a fibration of the $`A_{k1}`$ singularities along these circles, which can be locally resolved by $`k1`$ blowup parameters into a bouquet of $`k`$ spheres which intersect according to the extended Dynkin diagram of the $`A_{k1}`$ group with one relation among these cocycles.
The twisted states are associated to elements of the group $`\mathrm{\Gamma }`$. They can survive in the supergravity limit if they correpond to massless particles. It is known that zero modes for blowup modes are not present in sufficient numbers to smooth out the geometry, so it is not possible to study these models within classical geometry; we must turn to string theory for detailed computations. Massless twisted states have support at the fixed points of the group element they are associated with. Thus at each fixed circle we have massless states in the string theory which survive the low energy limit, and are associated with the group elements that fix the circle.
By going around the fixed circle on a closed loop, we are actually performing a twist by the group elements which don’t fix the circle, because the fixed circle is an invariant subspace of the other elements of the orbifold group. The circle is of radius $`k^1`$.
For the twisted strings that live at the orbifold circles, going around the loop one picks up a phase equal to the discrete torsion of the cycle acting on the group element to which the twisted state corresponds. This is the discrete torsion phase in the partition function. In our case, it sets the boundary conditions for the massless twisted sector states.
Consider for example the singularity of $`z_2=z_3=0`$ and the $`g=e_2^{\mathrm{}}`$ twisted sector corresponding to the subgroup which fixes the circle. When we set the boundary conditions for the twisted field, we get
$$𝒪_g(t+2\pi hR/m)=ϵ(g,e_1^h)𝒪_g(t)$$
(28)
for $`t`$ the geodesic coordinate along this circle measured in units of $`R`$, the radius of the AdS space. This circle is of size $`m^1`$, and we wind around it $`h`$ times. Winding once around is the same as twisting by $`e_2`$ in the unorbifolded covering space.
The blowup spheres are related to the discrete Fourier transform of the twisted states, and the action of the discrete torsion phase (which one can associate to the quantum symmetry of the local orbifold) becomes an automorphism of the Dynkin diagram which preserves the intersection form and the orientation. Thus the geometry differs from the standard $`S^5/\mathrm{\Gamma }`$ space in that the singularities have monodromy of the resolving spheres. This is a special example of monodromies that have been studied in , which are related to canonical singularities .
The equation (28) then becomes for the fields on the basis of the spheres
$$\widehat{𝒪}_i(t+2\pi hR/m)=\widehat{𝒪}_{i+a(m/s)h}(t)$$
(29)
The precise discrete torsion value is encoded in the possible integer values of $`a`$, which are relatively prime to $`s`$. Let us take $`h=1=a`$ for convenience. Then on going around the fixed circle the monodromy acts as shown in Figure 6.
It is worth pointing out that although we start with $`k1`$ twisted sectors, we have a bouquet of $`k`$ shrunken spheres. The relation in homology makes one linear combination of the local fields untwisted, and thus does not belong to the twisted string states. The counting of twisted states is the same in both approaches.
Details of the exact calculation of the spectrum of masses are given in the Appendix A. The masses are given exactly by the tree level mixing induced by the RR and NSNS backgrounds. It is not difficult to see that the spectrum of masses of chiral twisted state operators gives rise to conformal fields of the appropriate dimensions and twist charge as the ones predicted by the orbifold low energy effective field theory. In fact, one can see that the monodromy of the spheres makes the fields periodic with period $`2\pi Rs/m`$, as opposed to $`2\pi R/m`$, so the circles look $`s`$ times larger than the geometric circle, and the quantization of the masses of the states gives a factor in the mass which is proportional to $`ms^1R^1`$ as opposed to the factor of $`mR^1`$ coming just from the size of the circle (c.f. eq. (27)). This is exactly the spectrum of states predicted by the field theory too.
Notice also that the value of $`a`$ is irrelevant for this calculation, as it amounts to a different permutation of the spheres when going around the singularity. This is also in accord with the field theory. This is the first check on the duality.
We should also point out that the spectrum of D-branes that can leave the singularity is also equal in both theories (supergravity and field theory), as any fractional D-brane that is seen in the bulk of supergravity is either untwisted, or consists of a collection of D-brane states that are fixed at the orbifold circles.
### 6.2 Torsion Classes in Homology
Notice that the singular manifold thus described has torsion on homology two-cycles. Upon monodromy on the circle, different spheres are identified.
On the singularity that we have just studied, there are $`m/s`$ equivalence classes in homology at the singularity. They can be represented as the average cycles
$$\stackrel{~}{C}_i=\frac{1}{s}[C_i+C_{i+m/s}+C_{i+2m/s}+\mathrm{}]$$
(30)
and they intersect according to the $`A_{m/s1}`$ Dynkin diagrm.
The relation in homology is given by
$$\underset{i}{}C_i=0=s\underset{i}{}\stackrel{~}{C}_i$$
(31)
so calling $`T=_i\stackrel{~}{C}_i`$ we can prove that $`sT=0`$, but $`T0`$ as a homology cycle. Indeed, if we wrap a brane around the $`T`$ cycle, the brane is locally made of $`m/s`$ spheres, and is not allowed to leave the circle of singularities, since it can be made to correspond to a root of the $`A_{m1}`$ lattice, and therefore it can not be deformed away from the singular cycle. This argument works for $`t`$ copies of the cycle for all $`t<s`$. Thus the torsion cycle is of order $`s`$.
This is seen also in the field theory, because one cannot solve the $`F`$ constraints for $`\varphi _1,\varphi _2,\varphi _3`$ all different from zero, unless one has rank $`s`$ on each of the gauge groups in the quiver.
This establishes a direct correspondence between fractional regular branes and the torsion classes of the manifold. The existence of a torsion class is consistent with the absence of blowup modes in the string spectrum. As the fractional $`Dp`$ branes also carry an anomalous $`D(p2)`$ brane charge (because the orbifold has constant B field along each of the blowup cycles), one can have a BPS torsion class with finite tension.
## 7 New $`N=1`$ Dualities
Notice that the quivers we have obtained are identical to quivers of standard type without discrete torsion. They only differ in the superpotential, and in fact, they lie in the moduli space of deformations of the standard quiver theories . For a single-node quiver, the field theory is a marginal deformation of the $`N=4`$ theory
$$W=\sqrt{2}g\mathrm{tr}\varphi _1[\varphi _2,\varphi _3]\sqrt{2}g(\lambda ^s1)\mathrm{tr}\varphi _1\varphi _2\varphi _3$$
(32)
Alternatively, this may be rewritten in terms of $`f_{abc}`$ and $`d_{abc}`$ as appears in Ref. . In the case of multinode quiver theories, the superpotential is a marginal deformation of standard $`N=1`$ quiver theories.
Thus, with the AdS/CFT dictionary in hand, these field theory deformations act by geometrically deforming the theory which started with $`S^5`$, or $`S^5/\mathrm{\Gamma }`$. In particular, we expect that the marginal deformation corresponds to replacing the 5-sphere with a deformed version, and also turning on 2-form fields.
In supergravity it has been argued that one cannot observe some of these deformations . However, as we deform the superpotential, the conditions for a field being primary change, as the moduli space of couplings is curved. In the field theory this is seen because conformal invariance imposes restrictions on the couplings.
The moduli space of the theory changes when we follow these deformations. In particular, we have to find solutions of the $`F`$-constraints, which take the form
$$\varphi _1\varphi _2=\omega \varphi _2\varphi _1$$
(33)
For generic $`\omega `$, there is no solution to these matrix equations with $`\varphi _1\varphi _20`$ at finite $`N`$, and therefore the classical moduli space disappears. But on the other hand, we can elevate the previous equation to non-commutative geometry as in , if we let $`\varphi _i`$ be operators acting on an infinite dimensional Hilbert space. In this case we would be forced onto the large $`N`$ field theory from the start.
As a marginal perturbation, $`\lambda `$ would seem to be a continuous parameter. However, properties of the theory change discontinuously as we change $`\lambda `$. There is a direct analogy here to the $`\theta `$-angle of six-dimensional gauge theories, which are also related to non-commutative geometry. For values of $`\lambda `$ which are roots of unity, the classical moduli space has solutions for finite matrices, and it is exactly in this case where we have the dual theory with the orbifold.
Thus, for each of these theories under the conditions for the Maldacena conjecture to be reduced to supergravity, we have two types of supergravity duals. Namely, $`AdS_5\times \stackrel{~}{S}_5`$ where we have a deformed sphere $`\stackrel{~}{S}^5`$ and $`AdS_5\times S^5/\mathrm{\Gamma }_d`$, where we use the subscript $`d`$ to indicate the presence of discrete torsion. (This is the extreme case where the quiver theory has only one node, but one can restate it for the general case by writing the extra quotients properly.) These theories are mirror to one another, even at finite $`N`$.
At infinite $`N`$, there are additional field theory dualities relating different values of $`\lambda `$. On $`AdS_5\times S^5/\mathrm{\Gamma }_d`$ we have various choices of discrete torsion parameter. But it can easily be seen that neither the chiral ring nor the classical moduli space depends on this choice, once $`s`$ is fixed. Moreover one sees that the discrete torsion parameter can only be measured in chiral correlation functions involving twist fields that have support at different singularities. In the supergravity approximation, in order for twist fields to interact, they should be able to be at the same spacetime point. Thus all of these amplitudes are suppressed because the twisted strings have to extend far away along the sphere in order to interact.
On the field theory side, these amplitudes vanish in the free field limit, as they would involve, for example, the OPE of $`\mathrm{tr}(\varphi _1^k)\mathrm{tr}(\varphi _2^k)\mathrm{tr}(\varphi _3^k)`$. Notice that if the free field limit is the correct way to calculate three point functions (just as in $`AdS_5\times S^5`$) then the discrete torsion parameter disappears in the supergravity regime. Indeed, discrete torsion appears only in a non-planar diagram, as in Figure 3.
It would be natural then to identify all of these $`CFT`$ theories as dual pairs. They meet all of the requirements of $`N=1`$ duality, as they have the same moduli space of flat directions and the same spectrum of chiral primaries. At large $`N`$ they also have the same set of three point correlation functions.
This duality would act by exchanging the fundamental roots of unity. It is not continous on the parameter $`\lambda `$, but it is a well known mathematical object: it is the Galois group of the algebraic extension $`(\eta )`$. In our case this Galois group can also be made to act on the algebra generated by the matrices $`x,y,z,u,v,w`$ if all of their components are algebraic numbers, but the significance of this curiosity is at best obscure.
## 8 Conclusions
We have analyzed some Abelian orbifolds with discrete torsion from the D-brane perspective. Our results show that the simple cases give rise to deformations of known theories of the brane box type, which differ only in the superpotential. This result extends the work of . We tested the field theory by reproducing the orbifold as the moduli space of D-branes and by calculating the spectrum of chiral primaries.
In analyzing the field theory with the AdS/CFT correspondence, we discovered a new geometric picture for the orbifolds with discrete torsion. Indeed we found that as the fixed planes of singularities of the orbifold become fixed circles in the AdS/CFT correpondence, the study of massless twisted states on these fixed circles revealed that the blowup modes of the singularities have monodromy. The resolved geometry then has a different topology than the standard orbifold, and provides a geometrical understanding of the discrete torsion. Also, the obstructions to resolving the Calabi-Yau singularities may be traced back to the existsence of monodromy of the blowup modes. We have explicitly checked this picture by calculating the spectrum of chiral primaries in the supergravity/string theory and comparing with the field theory.
The AdS/CFT correspondence can then be used to prove new examples of duality. Indeed, as the theories with discrete torsion are seen to be given by deformations of standard orbifolds, one can follow these deformations at the supergravity level. This provides new examples of mirror symmetry, as both string theory descriptions are weakly coupled.
In the large $`N`$ limit we have also found new dualities. These dualities appear because the spectrum of operators in the supergravity is insensitive to the discrete torsion parameter, and the three-point functions of chiral twisted states that could measure this parameter vanish in this limit. The duality acts discontinuosly on the parameters defining the theory by permuting the different roots of unity.
It would be interesting to get a better understanding of the deformation theory and resolution of singularities within the D-brane language along the lines of , and to extend these results for other marginal and relevant deformations of the field theory. We would also like to extend the discrete torsion models to other geometries, wuch as the conifold. This is work in progress .
###### Acknowledgments.
Partially supported by the United States Department of Energy, grant DE-FG02-91ER40677 and an Outstanding Junior Investigator Award.
## Appendix A Supergravity spectrum of chiral operators
In this appendix we set up the calculation of the supergravity spectrum of states in the orbifolds with discrete torsion. Our construction is done following the background normal coordinate expansion for Ramond-Ramond background type of calculations explained in . These results coincide with supergravity calculations obtained in in cases where smooth geometry exists, but generalizes to spaces which have less local symmetry. For simplicity we will study the local $`^2/_2`$ fixed circles (only because it is easier to write the twisted vertex operators). The results generalize immediately to any other orbifolds.
The analysis is completely standard. The spin field $`S^\alpha `$ is written in terms of $`Spin(5,1)\times Spin(4)`$ spin-fields: $`S^\alpha \{S^A\mathrm{\Sigma }^a,S_A\mathrm{\Sigma }^{\dot{a}}\}`$. Conventions for operator products may be found in the Appendix B.
### A.1 Twisted States of the $`^2/_2`$ Orbifold
At the fixed circle, we have a $`^2/_2`$ orbifold. The twisted sector massless strings are given by the following operators.
In the $`NSNS`$ sector, we find
$$𝒪_{NS}^{ab}\mathrm{\Sigma }^a\stackrel{~}{\mathrm{\Sigma }}^be^\varphi e^{\stackrel{~}{\varphi }}\sigma _6\sigma _7\sigma _8\sigma _9e^{ikX}$$
(34)
These transform as $`(\mathrm{𝟏},(\mathrm{𝟏}+\mathrm{𝟑},\mathrm{𝟏}))`$ under $`Spin(5,1)\times Spin(4)`$. and correspond to the $`NS`$ $`\theta `$-angle, $`\mathrm{\Theta }_{NS}`$ and the triplet of blowup modes.
In the $`RR`$ sector, we find
$$𝒪_{R,AB}S_A\stackrel{~}{S}_Be^{\varphi /2}e^{\stackrel{~}{\varphi }/2}\sigma _6\sigma _7\sigma _8\sigma _9e^{ikX}$$
(35)
which transform as $`(\mathrm{𝟔}+\mathrm{𝟏𝟎},(\mathrm{𝟏},\mathrm{𝟏}))`$. The $`\mathrm{𝟔}`$ is the field strength of the $`RR`$ $`\theta `$-angle, $`_\mu \mathrm{\Theta }_R`$.
### A.2 The Background
The background that we are interested in must be a singlet under both $`Spin(4)`$ and $`Sp(2)Spin(5)Spin(5,1)`$. The $`RR`$ part is then determined as follows. The $`\mathrm{𝟔}`$ of $`Spin(5,1)`$ has one $`Sp(2)`$ singlet, while the $`\mathrm{𝟏𝟎}`$ has none. We thus find that the $`RR`$ background vertex is
$$V_{RR}h\left(J_{AB}ϵ_{ab}S^A\mathrm{\Sigma }^a\stackrel{~}{S}^B\stackrel{~}{\mathrm{\Sigma }}^b+\alpha J^{AB}ϵ_{\dot{a}\dot{b}}S_A\mathrm{\Sigma }^{\dot{a}}\stackrel{~}{S}_B\stackrel{~}{\mathrm{\Sigma }}^{\dot{b}}\right)e^{\varphi /2}e^{\stackrel{~}{\varphi }/2}$$
(36)
where $`h`$ is a normalization constant, $`\alpha `$ is a fixed phase and $`J`$ is the antisymmetric $`Sp(2)`$ invariant.
First, let us note that two $`RR`$ background insertions give a log divergence when they are close together
$$V_{RR}(z)V_{RR}(z^{})16h^2\mathrm{ln}\epsilon \left(\psi ^\mu \stackrel{~}{\psi }_\mu \psi ^M\stackrel{~}{\psi }_M\right)e^\varphi e^{\stackrel{~}{\varphi }}$$
(37)
The signature here is $`(5,5)`$, and this divergence corresponds to a contribution to the graviton $`\beta `$-function. The other contribution may be computed directly within the normal coordinate expansion, and takes the form
$$\frac{1}{4}\mathrm{log}\epsilon R_{ij}\psi ^i\stackrel{~}{\psi }^je^\varphi e^{\stackrel{~}{\varphi }}$$
(38)
Thus we fix the normalization of the $`RR`$ background
$$hR=\frac{1}{2\sqrt{2}}$$
(39)
where $`R`$ is the radius of the $`AdS_5`$ metric.
### A.3 Mixing and Masses
Here we compute the masses of twisted eigenstates. In the presence of the $`RR`$ background, there is operator mixing at lowest order. In particular, we find mixing between $`\mathrm{\Theta }_{NS}`$ and $`\mathrm{\Theta }_R`$, as follows. Using the notations (34),(35), these singlets have vertex operators of the form
$`𝒪_{NS}^{(1)}`$ $``$ $`\mathrm{\Theta }_{NS}ϵ_{ab}𝒪_{NS}^{ab}`$ (40)
$`𝒪_R^{(1)}`$ $``$ $`_\mu \mathrm{\Theta }_R(\mathrm{\Gamma }^\mu )^{AB}𝒪_{R,AB}`$ (41)
Mixing is induced through the Fischler-Susskind mechanism – there are logarithmic divergences between the background $`V_{RR}`$ and these operators.
Consider the equation of motion for the $`RR`$ singlet. We find mixing through
$$𝒪_{NS}^{(1)}(z)V_{RR}(z^{})2h\mathrm{\Theta }_{NS}J_{AB}k/^{AC}k/^{BD}\mathrm{log}\epsilon 𝒪_{R,CD}(z^{}).$$
(42)
Thus we find an equation of motion
$$\frac{1}{2}k^2_\mu \mathrm{\Theta }_R(\mathrm{\Gamma }^\mu )^{CD}+2hJ_{AB}k/^{AC}k/^{BD}\mathrm{\Theta }_{NS}=0$$
(43)
and thus mass terms
$$\frac{n^2}{R^2}\mathrm{\Theta }_R+4ih\frac{n}{R}\mathrm{\Theta }_{NS}$$
(44)
Similarly, mixing for the $`NSNS`$ singlet is induced through
$$𝒪_R^{(1)}(z)V_{RR}(z^{})h_\mu \mathrm{\Theta }_RJ_{AB}(\mathrm{\Gamma }^\mu )^{AB}\mathrm{log}\epsilon ϵ_{ab}𝒪_{NS}^{ab}(z^{})$$
(45)
Putting these results together, we have a mass matrix
$$[m^2]=\left(\begin{array}{cc}(\frac{k}{R})^2& 8ih(\frac{k}{R})\\ +4ih(\frac{k}{R})& (\frac{k}{R})^2\end{array}\right)$$
(46)
where $`k`$ is the momentum along the fixed circle in units of $`R`$.
Within the AdS-CFT correspondence, we expect that the eigenvalue masses correspond to the dimensions of operators in the boundary CFT:
$$\mathrm{\Delta }=2+\sqrt{4+m^2R^2}$$
(47)
Given the result 46, we find a tower of states
$$\mathrm{\Delta }_1=2+|k\pm 2|$$
(48)
where $`k`$.
Similiarly, one obtains for the triplets
$$\mathrm{\Delta }_3=2+|k|$$
(49)
The interesting part of this calculation comes from the origin of the mass shift, which can be traced to the $`\sigma `$-model coupling proportional to $`R_{abcd}\psi ^a\psi ^b\overline{\psi }^c\overline{\psi }^d`$, and thus for spaces where $`R`$ is not symmetric one should be able to observe the effect here.
It is clear that these calculations are independent of the fact that we had a $`^2/_2`$ orbifold. The structure of states is the same for any orbifold group. The different orbifolds $`^2/_n`$ will have instead $`n1`$ twisted sectors.
The singlet states are chiral. Because of the monodromy of the circle $`k=am/s`$, and the dimensions of the operators match the $`\mathrm{tr}(\varphi _i^{(am/s)})`$, with multiplicities. The other operators are related to $`\mathrm{tr}(W^2\varphi _i^{(am/s)})`$.
## Appendix B Conventions for spin fields
Under $`Spin(5,1)\times Spin(4)Spin(9,1)`$, the spin fields decompose as $`S^\alpha \{S^A\mathrm{\Sigma }^a,S_A\mathrm{\Sigma }^{\dot{a}}\}`$. Here $`A`$ is an $`Spin(5,1)`$ spinor index and $`a,\dot{a}`$ are $`Spin(4)=SU(2)\times SU(2)`$ indices. We denote vector indices by $`\mu ,\nu ,\mathrm{}`$ for $`Spin(5,1)`$ and $`M,N,\mathrm{}`$ for $`Spin(4)`$.
Operator products are
$`\mathrm{\Sigma }^a(z)\mathrm{\Sigma }^b(0){\displaystyle \frac{ϵ^{ab}}{z^{1/2}}}`$ (50)
$`\mathrm{\Sigma }^{\dot{a}}(z)\mathrm{\Sigma }^{\dot{b}}(0){\displaystyle \frac{ϵ^{\dot{a}\dot{b}}}{z^{1/2}}}`$ (51)
$`\mathrm{\Sigma }^a(z)\mathrm{\Sigma }^{\dot{a}}(0)(\mathrm{\Gamma }_M)^{a\dot{a}}\psi ^M(0)`$ (52)
$`\mathrm{\Sigma }^{\dot{a}}(z)\mathrm{\Sigma }^a(0)(\mathrm{\Gamma }_M)^{\dot{a}a}\psi ^M(0)`$ (53)
$`\psi ^M(z)\mathrm{\Sigma }^a(0)\mathrm{\Sigma }^a(z)\psi ^M(0){\displaystyle \frac{1}{z^{1/2}}}(\mathrm{\Gamma }^M)^{a\dot{a}}\mathrm{\Sigma }_{\dot{a}}(0)`$ (54)
$`\psi ^M(z)\mathrm{\Sigma }^{\dot{a}}(0)\mathrm{\Sigma }^{\dot{a}}(z)\psi ^M(0){\displaystyle \frac{1}{z^{1/2}}}(\mathrm{\Gamma }^M)^{\dot{a}a}\mathrm{\Sigma }_a(0)`$ (55)
The matrices $`\mathrm{\Gamma }^M`$ have Clifford normalization. Similarly
$$S^A(z)S_B(0)S_A(z)S^B(0)\frac{\delta _B^A}{z^{3/4}}$$
(56)
$$S^A(z)S^B(0)\frac{1}{z^{1/4}}(\mathrm{\Gamma }_\mu )^{AB}\psi ^\mu (0)$$
(57)
$$S_A(z)S_B(0)\frac{1}{z^{1/4}}(\mathrm{\Gamma }_\mu )_{AB}\psi ^\mu (0)$$
(58)
$$\psi ^\mu (z)S^A(0)\frac{1}{z^{1/2}}(\mathrm{\Gamma }^\mu )^{AB}S_B(0)$$
(59)
$$\psi ^\mu (z)S_A(0)\frac{1}{z^{1/2}}(\mathrm{\Gamma }^\mu )_{AB}S^B(0)$$
(60)
$$S^A(z)\psi ^\mu (0)\frac{1}{z^{1/2}}S_B(0)(\mathrm{\Gamma }^\mu )^{BA}$$
(61)
$$S_A(z)\psi ^\mu (0)\frac{1}{z^{1/2}}S^B(0)(\mathrm{\Gamma }^\mu )_{BA}$$
(62)
where $`(\mathrm{\Gamma }_\mu )_{AB}=J_{AC}(\mathrm{\Gamma }_\mu )^{CD}J_{DB}`$, with $`J`$ the antisymmetric $`Sp(2)`$ invariant. |
no-problem/0001/astro-ph0001411.html | ar5iv | text | # Search for Astronomical Sites in Developing Countries and their Preventive Protection
## 1 Introduction
The knowledge of short time-scale variation stars has made some progress due to campaigns of non-stop observations with simple manual telescopes and photometers, conducted simultaneously around the world. During the last two decades, the automation of telescopes (remote-controlled or robotic telescopes) opened the way to study the variability in the entire HR diagram (e.g. Henry, 1999) and mainly at places where simultaneous variations of different characteristic times from hours to years are discovered (AGB, RGB stars, etc.). To follow these permanent or non-permanent variations during months or years, and to understand their origin, networks of robotic telescopes seem to be nowadays the most appropriate technology.
The first discoveries of the visible counterparts of gamma-ray bursts and of many new NEO demonstrate that the networks of robotic telescopes are/will be powerful in many scientific fields (e.g. ROBONET, GNAT, TORUS, NORT, etc., as described in Querci and Querci, 1999).
Consequently, sites for networks have to be implemented on various longitudes (continents) and in the two hemispheres. Excellents sites (Hawaii, North of Chile, South Pole, etc.) or some very good ones (Canary Islands, South Africa, India, Uzbekistan, etc.) are already at work. Are we sure that other excellent sites do not exist elsewhere? In Developing Countries (hereafter DCs) for example?
Astronomy and Space Science could be a contribution to the development of DCs, as already seen, some decades ago, in Canary Islands, Chile, and so on.
## 2 A way to Select Sites in DCs
A world-wide preliminary map of mean nebulosity (at 0.55 $`\mu `$m) was obtained from 12-year meteorological archives and with 250 km square meshes (Querci and Querci, 1998a,b).
$``$ A first step should be a cross correlation analysis between a worldwide map of high mountain summits (altitude: 2400-3200 m or more) and a worldwide map of small-mesh meteorological archives (2 to 5 km) on cloudiness, humidity, sand winds and light pollution. It could permit to obtain 20 to 30 new meteorological excellent and/or very good sites adapted to optical, IR, or millimetric observations.
$``$ A second step should be a detailed analysis of the local atmospheric turbulence for these 20 to 30 pre-selected sites by a seeing-monitor or by a grating scale-monitor (G.S.M.) technique (Martin et al., 1994). The registration of the parameters L<sub>0</sub>, the wavefront outer scale, r<sub>0</sub>, the Fried parameter, $`\tau `$, the speckle lifetime, and of the isoplanetism angle for each of these sites should permit to select finally 8 to 10 sites besides those already classified as high quality observing sites.
$``$ A third step should be the development of the cooperation and the analysis of the local facilities to implement and to maintain robotic telescopes and their equipment.
The two last steps could be a way to introduce Astronomy and Space Science in DCs through robotic equipment and the analysis of variable objects supported by hydrodynamical calculations.
Preliminary works on large meshes are in progress in many countries. Moreover, analysis with small meshes are in progress on High Atlas (Morocco) and on Lebanese border mountains (Syria)(private communication).
## 3 A Protection for the Selected Sites
In many DCs, Astronomy and Space Science are not developed at all, and the search for sites is ignored. Consequently, potential sites might be polluted and lost for science in the future.
In a few DCs, collaboration with astronomically-developed countries are under progress. So, the site prospecting and the preventive site protection are now taken into account by national scientific authorities, contributing to the scientific and technical development of the country.
## 4 Conclusion
The prospecting and the preventive protection of potentially future astronomical sites are very important for Astronomy and Space Science in the next century. These sites of which excellent ones are in DCs, have been suggested from a world-wide mean annual nebulosity map.
We take the opportunity of this IAU/UN symposium for asking the questions: could such a prospecting and preventive protection be promoted
* by each DC individually?
* by some DCs or new astronomical countries grouped together inside regional astronomical organizations such as the Arab Union for Astronomy and Space Science (AUASS)?
* by international astronomical organizations such as European Southern Observatory (ESO), Cerro Tololo Inter-American Observatory (CTIO), etc.?
* by the International Astronomical Union?
* by the UN Office for Outer Space Affairs?
The encouragement and the help of international organizations would be certainly decisive for DCs. At the scientific benefit of themselves as well as of the international community. |
no-problem/0001/nlin0001037.html | ar5iv | text | # Hamiltonian mappings and circle packing phase spaces
## I Introduction
Nonlinear dynamical systems offer a rich and seemingly endless variety of behaviour. In this paper we introduce three Hamiltonian mappings with phase space structures which resemble circle packings bullet ; bullet2 ; herrmann ; parker ; keen . A circle packing is a set with empty interior and whose complement is the union of disjoint open circular discs. Circle packings are often constructed using geometrical or group theoretical methods, for example limit sets of Kleinian groups maskit , but here we may have uncovered them in the context of Hamiltonian mechanics. Each of the three mappings acts on a two-dimensional manifold with constant curvature, defining one of three different phase space geometries: spherical (positive curvature), planar (zero curvature) or hyperbolic (negative curvature). In each case an infinite number of stable periodic orbits coexist and seem to ‘pack’ the manifold with circular resonances. Some analytical data has been found for these maps including the location of a large number of periodic orbits, but there remain many open questions, the most critical being whether the circular resonances densely pack the phase space. That is, whether the residual set not covered by the resonances has empty interior. If this is the case then we have indeed found new examples of circle packings. In sections II, III, and IV we introduce, respectively, the planar, hyperbolic and spherical mappings, and include all analytical findings. Finally, in Section V we discuss some open questions concerning these maps.
## II Planar Map
The Hamiltonian which generates our mapping on the Euclidean plane is
$$H(x,p,t)=\frac{1}{2}\omega (x^2+p^2)+\mu |x|\underset{n=\mathrm{}}{\overset{\mathrm{}}{}}\delta (tn),$$
where $`x`$, $`p`$ and $`t`$ are position, momentum and time respectively, and, $`\omega [0,2\pi )`$ and $`\mu 0`$ are parameters. In the planar case $`\mu `$ may be set to unity by rescaling $`x`$ and $`p`$. The mapping which takes $`(x,p)`$ from just before a kick to one period later is
$$\left[\begin{array}{c}x^{n+1}\\ p^{n+1}\end{array}\right]=F\left[\begin{array}{c}x^n\\ p^n\end{array}\right]=\left[\begin{array}{cc}\mathrm{cos}\omega & \mathrm{sin}\omega \\ \mathrm{sin}\omega & \mathrm{cos}\omega \end{array}\right]\left[\begin{array}{c}x^n\\ p^n\mu s^n\end{array}\right]$$
(1)
where $`s^n\text{sgn}x^n`$ ($`\text{sgn}x`$ is the signum function with the convention $`\text{sgn}\mathrm{\hspace{0.17em}0}=0`$). Between kicks the Hamiltonian is that of a simple harmonic oscillator and all phase space points rotate clockwise about the origin through an angle of $`\omega `$. The effect of the kick is to add a position dependent shift in the momentum of $`\mu \text{sgn}x`$. If $`x0`$ the tangent mapping, $`F(x,p)/(x,p)`$, is simply a linear rotation with eigenvalues $`e^{\pm i\omega }`$. Thus if an orbit is to have nonzero Lyapunov exponents, such as an unstable periodic orbit, we need at least one iterate of the orbit on the line $`x=0`$. However, in this instance the tangent mapping is undefined. We overcome this problem simply by defining an unstable orbit to be one with a point on $`x=0`$, and all other orbits stable.
If we let $`z=x+ip`$ then we can rewrite the mapping as
$`z^{n+1}`$ $`=`$ $`Fz^n`$
$`=`$ $`e^{i\omega }(z^ni\mu s^n).`$
Suppose that $`z^0`$ is a periodic point with period $`n`$. That is,
$`z^0`$ $`=`$ $`z^n`$
$`=`$ $`e^{in\omega }z^0i\mu {\displaystyle \underset{k=0}{\overset{n1}{}}}e^{i(kn)\omega }s^k.`$
Solving for $`z^0`$ we obtain
$$z^0=\frac{i\mu }{1e^{in\omega }}\underset{k=0}{\overset{n1}{}}e^{ik\omega }s^k.$$
(2)
If $`z^0`$ is a stable periodic point then $`s^k=\pm 1`$, which simply states whether the iterate $`z^k`$ is on the right or left side of the complex plane. The unstable periodic orbits must have at least one point on $`x=0`$. That is, $`s^k=0`$ for some $`k=0\mathrm{}n1`$. Note that if $`n\omega `$ is an integer multiple of $`2\pi `$ then equation (2) is invalid. In these cases one finds that there are an infinite number of period $`n`$ orbits. But when this is not the case, each periodic orbit is uniquely determined by the sequences $`\{s^k=0,\pm 1\}_{k=0\mathrm{}n1}`$, with at most two $`s^k`$ being zero (see the Lemma below). By cycling a particular sequence we obtain the $`n`$ iterates of the periodic orbit. Although each periodic orbit is uniquely represented by a sequence, not every sequence represents a periodic orbit. Hence, we still need to find which sequences are legitimate. This task can be simplified by noting that $`F`$ can be decomposed into the two involutions $`I_1`$ and $`I_2`$, where
$`I_1z`$ $`=`$ $`z^{}i\mu \text{sgn}x`$
$`I_2z`$ $`=`$ $`e^{i\omega }z^{}`$
$`I_{1}^{}{}_{}{}^{2}`$ $`=`$ $`I_{2}^{}{}_{}{}^{2}=1`$
$`I_1FI_1`$ $`=`$ $`I_2FI_2=F^1`$
and
$$F=I_2I_1.$$
It is well-known that whenever a mapping admits such a decomposition there exist symmetric periodic orbits having points which lie on the fixed lines of each involution pina . The fixed line of an involution $`I`$ is the set
$$\text{Fix}(I)\{z|Iz=z\}.$$
Theorem lamb : Let $`F`$ be an invertible mapping and $`I`$ an involution ($`I^2=1`$) satisfying $`IFI=F^1`$. Let $`z\text{Fix}(I)`$ and suppose that $`z`$ is a periodic point of $`F`$ with least period $`n`$. Then $`\text{Fix}(I)`$ contains no other points of the periodic orbit if $`n`$ is odd, or exactly one other, $`F^{n/2}z`$, if $`n`$ is even.
Consider the unstable periodic orbits. From our definition, an unstable orbit must have at least one point lying on $`x=0`$. Noting that $`\text{Fix}(I_1)=\{x+ip|x=0\}`$ and putting $`I=I_1`$ in the above theorem we obtain the following Lemma:
Lemma : Let $`\{s^k\}_{k=0\mathrm{}n1}`$ be the sequence defining an unstable periodic orbit of least period $`n`$. If $`n`$ is odd then the sequence contains exactly one zero, $`s^0=0`$ say, and if $`n`$ is even the sequence contains exactly two zeros, $`s^0=0`$ and $`s^{n/2}=0`$.
We now know that all unstable periodic orbits have one or two points on the fixed line of the involution $`I_1`$ ($`x=0`$). Many of the stable periodic orbits have points on the fixed line of the second involution $`I_2`$, which is given by the equation $`x=p\mathrm{tan}\omega /2`$. In this case one can also show that the orbits with odd period have one point on this line while the orbits with even period have two. One cannot show, however, that the stable periodic orbits must have point on this line, indeed, some do not. There is a second decomposition of the mapping, $`F=\stackrel{~}{I}_2\stackrel{~}{I}_1`$, where $`\stackrel{~}{I}_1=I_1`$ has the fixed line $`p=\mu /2\text{sgn}x`$ and $`\stackrel{~}{I}_2=I_2`$ has the fixed line $`p=x\mathrm{tan}\omega /2`$. We are unsure as to whether every stable periodic orbit has a point on one of the above fixed lines. However, numerical investigations seem to suggest that they do. For the case $`\omega =\pi (\sqrt{5}1)`$ we have checked all $`2^{41}2`$ of the possible sequences for a stable periodic orbit of period $`40`$ and found only those with points on these lines. Hence we tentatively conjecture that all stable periodic orbits have one or two points on at least one of the fixed lines of $`I_1`$, $`\stackrel{~}{I}_1`$ and $`\stackrel{~}{I}_2`$.
Proposition : If $`\{s^k\}_{k=0\mathrm{}n1}`$ is the sequence defining a periodic orbit of least period $`n`$ with $`z^0`$ being the first point, then
1. $`s^k=s^{nk}`$ if $`z^0\text{Fix}(I_1)`$,
2. $`s^k=s^{nk1}`$ if $`z^0\text{Fix}(I_2)`$,
3. $`s^k=s^{nk}`$ if $`z^0\text{Fix}(\stackrel{~}{I}_1)`$,
4. $`s^k=s^{nk1}`$ if $`z^0\text{Fix}(\stackrel{~}{I}_2)`$,
for $`k=1\mathrm{}n1`$.
Proof : We will only prove the first case. The others are similar. We know that $`F^nz^0=z^0`$ and $`I_1z^0=z^0`$, since $`z^0\text{Fix}(I_1)`$. The iterate $`z^k`$, $`1k<n`$, is another point of the periodic orbit and
$`I_1z^k`$ $`=`$ $`I_1F^kz^0`$
$`=`$ $`I_1I_1F^kI_1z^0`$
$`=`$ $`F^kz^0`$
$`=`$ $`F^{nk}z^0`$
$`=`$ $`z^{nk}.`$
Thus $`I_1z^k`$ is also a point of the periodic orbit, and since $`I_1`$ changes the sign of $`x`$ the above equation gives the formula $`s^k=s^{nk}`$. $`\mathrm{}`$
The phase space portrait for $`\omega =\pi (\sqrt{5}1)`$ and $`\mu =1`$ is shown in Fig. 1. The unstable orbits are in black and consist of unstable periodic and aperiodic motion. They were generated by plotting all images and preimages of the vertical line $`x=0`$. The circular discs are resonances and are packed entirely with stable orbits. At the center of each of these lies a stable periodic orbit. The positions of a large number of periodic orbits have been found analytically. For want of a better term, we will call these ‘first order’ periodic orbits. All of the resonances visible in Fig. 1 are of first order. Only on greater levels of magnification does one encounter resonances of higher order.
The planar mapping has a stable first order periodic orbit of period $`n`$ if there exists an integer $`m`$ such that $`1mn`$, $`\mathrm{gcd}(n,m)=1`$, and
$$\frac{m1}{n}<\frac{\omega }{2\pi }<\frac{m}{n}\text{if }n\text{ is even,}$$
or
$$\frac{m1/2}{n}<\frac{\omega }{2\pi }<\frac{m}{n}\text{if }n\text{ is odd.}$$
These orbits are born near the origin for the smaller value of $`\omega `$ and then are destroyed at infinity when $`\omega /2\pi =m/n`$. The sequences defining these orbits are
$$s^k=\text{sgn}\left(\mathrm{sin}2\pi \frac{km+1/4}{n}\right)$$
(3)
$`k=0\mathrm{}n1`$. The $`n`$ iterates of each periodic orbit are found by inserting the $`n`$ cycles of the corresponding sequence into (2). The orbits with odd period come in pairs. The second of each pair is found by negating every term of the sequence. All these orbits, apart from those of period 1 and 2, are created with a corresponding unstable periodic orbit. These unstable orbits exist for the same $`\omega `$ as the stable and have sequences
$$s^k=\text{sgn}\left(\mathrm{sin}2\pi \frac{km}{n}\right).$$
(4)
There is also an unstable period 1 orbit at the origin which exists for all $`\omega `$.
The first order periodic orbits were found by observing that at infinity the kick has no effect and the map is simply a linear rotation through an angle of $`\omega `$. Consequently, when $`\omega /2\pi =m/n`$ there will be period $`n`$ orbits at infinity and these orbits must have sequences given by (3) or (4), depending on their stability. Further investigation reveals that they were born near the origin at a smaller value of $`\omega `$ when points of the stable periodic orbit intersected the line $`x=0`$. Hence one can use (2) to find out when this occurred and obtain the above results.
All these orbits revolve around the origin under iteration, and in general, the orbits for which $`m/n`$ is larger are closer to the origin since they are destroyed at infinity for a greater value of $`\omega `$. In Fig. 2 and Fig. 3 we have plotted, respectively, the positions of all stable and unstable first order periodic orbits. The values of $`\omega `$ and $`\mu `$ are the same as in Fig. 1. The radii of the circular resonances enclosing each of the stable first order periodic orbits can also been found. This may be done by noting that every stable periodic orbit must have an iterate with its circular resonance tangent to the line $`x=0`$. The radii is then given by the $`x`$ position of that iterate. These reduce to
$$r_n=\frac{1}{2}\mathrm{tan}\frac{\omega n}{4}\text{if }n\text{ is even,}$$
or
$$r_n=\frac{1}{2}\mathrm{tan}\left(\frac{\omega n}{2}+\frac{\pi }{2}\right)\text{if }n\text{ is odd.}$$
All iterates of an orbit inside a resonance can be written as
$$z^k=z_p^{k\text{mod}n}+(z^0z_p^0)e^{ik\omega }$$
(5)
where $`z_p^0`$ is the period $`n`$ point at the center of the resonance containing $`z^0`$, and
$$|z^0z_p^0|<r_n.$$
If $`\omega /2\pi `$ is irrational then the orbit iterates in a quasiperiodic circular motion about the central periodic orbit. However, when $`\omega /2\pi `$ is rational, that is, $`\omega /2\pi =p/q`$ where $`p`$ and $`q`$ are integers with $`\mathrm{gcd}(p,q)=1`$, then $`z^{nq}=z^0`$. Hence the orbit is periodic. In this case each image of the vertical line $`x=0`$ is a collection of parallel line segments at one of only $`q`$ different possible angles. Consequently, the resonances are all polygons which tile the phase plane. The phase space portrait for $`\omega /2\pi =5/8`$ and $`\mu =1`$ is shown in Fig. 4. Here the phase plane is tiled with octagons of decreasing size. In Fig. 5 where $`\omega /2\pi =3/5`$ the phase plane is tiled with pentagons and decagons.
## III Hyperbolic Map
The hyperbolic map is derived from the Hamiltonian
$$H(𝐊,t)=\omega K_3+\mu |K_1|\underset{n=\mathrm{}}{\overset{\mathrm{}}{}}\delta (tn),$$
where $`𝐊=(K_1,K_2,K_3)=(x_2p_3x_3p_2,x_3p_1+x_1p_3,x_1p_2x_2p_1)`$ is the Minkowski 3-vector for a particle confined to a pseudospherebalazs , normalized such that $`𝐊`$ lies on the hyperboloid
$$1+K_{1}^{}{}_{}{}^{2}+K_{2}^{}{}_{}{}^{2}=K_{3}^{}{}_{}{}^{2},K_3>0.$$
(6)
The evolution of $`𝐊`$ under the above Hamiltonian is governed by the equations
$$\dot{K_i}=\{K_i,H\},\{K_1,K_2\}=K_3,\{K_2,K_3\}=K_1,\{K_3,K_1\}=K_2,$$
where $`\{,\}`$ are the Poisson brackets. The mapping which takes $`𝐊`$ from just before a kick to one period later is
$`\left[\begin{array}{c}K_1^{n+1}\\ K_2^{n+1}\\ K_3^{n+1}\end{array}\right]=F\left[\begin{array}{c}K_1^n\\ K_2^n\\ K_3^n\end{array}\right]`$ $`=`$ $`\left[\begin{array}{ccc}\mathrm{cos}\omega & \mathrm{sin}\omega & 0\\ \mathrm{sin}\omega & \mathrm{cos}\omega & 0\\ 0& 0& 1\end{array}\right]\left[\begin{array}{ccc}1& 0& 0\\ 0& \mathrm{cosh}\mu s^n& \mathrm{sinh}\mu s^n\\ 0& \mathrm{sinh}\mu s^n& \mathrm{cosh}\mu s^n\end{array}\right]\left[\begin{array}{c}K_1^n\\ K_2^n\\ K_3^n\end{array}\right]`$
$``$ $`\text{F}(s^n)𝐊^n`$
where $`s^n\text{sgn}K_1^n`$.
The unstable orbits are defined to be those with at least one point with $`K_1=0`$. In a similar fashion to the planar case we can label all periodic orbits with sequences $`\{s^k=0,\pm 1\}_{k=0\mathrm{}n1}`$. Then the position of the periodic point must be the solution of
$$𝐊=\text{F}(s^{n1})\text{F}(s^{n2})\mathrm{}\text{F}(s^0)𝐊,$$
(8)
which lies on the hyperboloid (6). By using the previous theorem and decomposing $`F`$ into the two involutions,
$`I_1𝐊`$ $`=`$ $`(K_1,K_2\mathrm{cosh}(\mu \text{sgn}K_1)+K_3\mathrm{sinh}(\mu \text{sgn}K_1),K_2\mathrm{sinh}(\mu \text{sgn}K_1)+K_3\mathrm{cosh}(\mu \text{sgn}K_1)),`$
$`I_2𝐊`$ $`=`$ $`(K_1\mathrm{cos}\omega K_2\mathrm{sin}\omega ,K_1\mathrm{sin}\omega +K_2\mathrm{cos}\omega ,K_3),`$
where $`F=I_2I_1`$, we can show that the lemma and proposition in the previous section are also true in the hyperbolic case. The second pair of involutions for the hyperbolic map are
$`\stackrel{~}{I}_1𝐊`$ $`=`$ $`(K_1,K_2\mathrm{cosh}(\mu \text{sgn}K_1)K_3\mathrm{sinh}(\mu \text{sgn}K_1),K_2\mathrm{sinh}(\mu \text{sgn}K_1)+K_3\mathrm{cosh}(\mu \text{sgn}K_1)),`$
$`\stackrel{~}{I}_2𝐊`$ $`=`$ $`(K_1\mathrm{cos}\omega +K_2\mathrm{sin}\omega ,K_1\mathrm{sin}\omega K_2\mathrm{cos}\omega ,K_3),`$
with $`F=\stackrel{~}{I}_2\stackrel{~}{I}_1`$. Hence the mapping has symmetry lines on the hyperboloid (6) of $`K_1=0`$, $`K_1=K_2\mathrm{tan}\omega /2`$, $`K_2=K_3\mathrm{tanh}(\mu /2\text{sgn}K_1)`$, and $`K_2=K_1\mathrm{tan}\omega /2`$ which are the fixed lines of the involutions $`I_1`$, $`I_2`$, $`\stackrel{~}{I}_1`$, and $`\stackrel{~}{I}_2`$, respectively.
To display the hyperbolic phase space we will use the conformal disk model ratcliffe where all angles are Euclidean. Hence all circles will look like Euclidean circles, but some will appear larger than others of the same periodic orbit because the model distorts distances. The mapping
$$x=\frac{K_1}{1+K_3},y=\frac{K_2}{1+K_3},$$
takes all points on the hyperboloid (6) into the disk $`x^2+y^2<1`$ in the Euclidean plane. The phase space portraits for $`\omega =\pi (\sqrt{5}1)`$, $`\mu =0.04`$ and $`\mu =0.2`$ are shown in Fig. 6 and Fig. 7, respectively. On comparing Fig. 6 with Fig. 1 we can see that the hyperbolic map resembles the planar map when $`\mu `$ is small. In fact, under a suitable transformation of variables the hyperbolic map can be approximated by the planar map when close to the origin and $`\mu `$ is small enough. Consequently, the first order periodic orbits of the hyperbolic map exist for the same $`\omega `$ as in the planar map with the exception that they are destroyed at infinity ($`x^2+y^2`$=1) at a smaller value of $`\omega `$ dependent on $`\mu `$. The hyperbolic mapping has a stable first order periodic orbit of period $`n`$ if there exists an integer $`m`$ such that $`1mn`$, $`\mathrm{gcd}(n,m)=1`$, and
$$\frac{m1}{n}<\frac{\omega }{2\pi }<f_{mn}(\mu )\frac{m}{n}\text{if }n\text{ is even,}$$
or
$$\frac{m1/2}{n}<\frac{\omega }{2\pi }<f_{mn}(\mu )\frac{m}{n}\text{if }n\text{ is odd.}$$
The sequences defining these orbits are given by (3), and the $`k`$-th iterate is found by inserting the $`k`$-th cycle of the sequence into (8) and solving on the hyperboloid (6). Again, the orbits with odd period come in pairs. The second of each pair is found by negating the sequence. The corresponding unstable orbits have sequences given by (4). The functions $`f_{mn}(\mu )`$ could not be found analytically except in the special cases of $`(m,n)=(1,1)`$ where
$$f_{11}(\mu )=1\frac{1}{\pi }\mathrm{arcsin}\left(\mathrm{tanh}\frac{\mu }{2}\right),$$
and $`(m,n)=(1,2),(1,4),(3,4)`$ where
$$f_{mn}(\mu )=\frac{m}{n}\frac{1}{\pi }\mathrm{arcsin}\left(\mathrm{sin}\frac{\pi }{n}\mathrm{tanh}\frac{\mu }{2}\right).$$
These can be found by explicitly solving (8) for the position of each periodic orbit. A much simpler way to find these orbits of low period is to make use of the fact that some iterates will lie on the symmetry lines. Although the functions $`f_{mn}(\mu )`$ are generally unknown, the task of finding all first order periodic orbits is still quite simple. One simply replaces the inequality $`\omega /2\pi <f_{mn}(\mu )`$ with the weaker inequality $`\omega /2\pi <m/n`$ together with the condition that the nontrivial solutions of (8) (with (3) or (4)) have $`K_{3}^{}{}_{}{}^{2}K_{1}^{}{}_{}{}^{2}K_{2}^{}{}_{}{}^{2}=t^2>0`$ (the case of $`f_{mn}(\mu )<\omega /2\pi <m/n`$ corresponds to $`K_{3}^{}{}_{}{}^{2}K_{1}^{}{}_{}{}^{2}K_{2}^{}{}_{}{}^{2}=t^2`$).
In the previous section we found that when $`\omega /2\pi `$ is rational the phase plane of the planar map is entirely filled with polygons. This does not occur in the hyperbolic case. The tangent mapping to the planar map is the same linear rotation matrix for both the left and right half of the phase plane. Consequently, the local rotation about every stable periodic orbit is $`\omega `$ (see Eq. (5)), and if $`\omega /2\pi `$ is rational then this rotation is itself periodic. The hyperbolic map, however, has different tangent mappings for the left and right half of the hyperbolic plane. Hence the local rotation about each stable periodic orbit will be different. However we can still choose $`\omega `$ and $`\mu `$ such that the rotation about one particular orbit is periodic. When this occurs we find that each resonance of the periodic orbit forms a polygon. For example, by choosing $`\mu =0.5`$ and setting $`\omega =\pi +\mathrm{arccos}\mathrm{tanh}^2\mu /2=4.652..`$ we find that the resonances enclosing each of the two period 1 orbits are hyperbolic squares (see Fig. 8). Each point inside these squares is a period 4 orbit, rotating clockwise exactly $`\pi /2`$ radians at each iteration.
## IV Spherical Map
The spherical map is given by the Hamiltonian
$$H(𝐉,t)=\omega J_3+\mu |J_1|\underset{n=\mathrm{}}{\overset{\mathrm{}}{}}\delta (tn),$$
where $`(𝐉)_i=J_i=ϵ_{ijk}x_jp_k`$ $`(i=1,2,3)`$ are the three components of angular momentum for a particle confined to a sphere, normalized such that
$$J_{1}^{}{}_{}{}^{2}+J_{2}^{}{}_{}{}^{2}+J_{3}^{}{}_{}{}^{2}=1.$$
(9)
The evolution of $`𝐉`$ under the above Hamiltonian is governed by the equations
$$\dot{J_i}=\{J_i,H\},\{J_i,J_j\}=ϵ_{ijk}J_k.$$
The mapping which takes $`𝐉`$ from just before a kick to one period later is
$`𝐉^{n+1}=\left[\begin{array}{c}J_1^{n+1}\\ J_2^{n+1}\\ J_3^{n+1}\end{array}\right]`$ $`=`$ $`\left[\begin{array}{ccc}\mathrm{cos}\omega & \mathrm{sin}\omega & 0\\ \mathrm{sin}\omega & \mathrm{cos}\omega & 0\\ 0& 0& 1\end{array}\right]\left[\begin{array}{ccc}1& 0& 0\\ 0& \mathrm{cos}\mu s^n& \mathrm{sin}\mu s^n\\ 0& \mathrm{sin}\mu s^n& \mathrm{cos}\mu s^n\end{array}\right]\left[\begin{array}{c}J_1^n\\ J_2^n\\ J_3^n\end{array}\right]`$
$``$ $`\text{F}(s^n)𝐉^n`$
where $`s^n\text{sgn}J_1^n`$.
The unstable orbits are defined to be those with at least one point for which $`J_1=0`$. Again, we can label all periodic orbits with sequences $`\{s^k=0,\pm 1\}_{k=0\mathrm{}n1}`$. Then the position of the periodic point must be the solution of
$$𝐉=\text{F}(s^{n1})\text{F}(s^{n2})\mathrm{}\text{F}(s^0)𝐉,$$
(23)
on the unit sphere (9) with $`\text{sgn}J_1=s^0`$. By decomposing $`F`$ into the two involutions,
$`I_1𝐉`$ $`=`$ $`(J_1,J_2\mathrm{cos}(\mu \text{sgn}J_1)J_3\mathrm{sin}(\mu \text{sgn}J_1),J_2\mathrm{sin}(\mu \text{sgn}J_1)+J_3\mathrm{cos}(\mu \text{sgn}J_1)),`$
$`I_2𝐉`$ $`=`$ $`(J_1\mathrm{cos}\omega J_2\mathrm{sin}\omega ,J_1\mathrm{sin}\omega +J_2\mathrm{cos}\omega ,J_3),`$
where $`F=I_2I_1`$, we can show that the lemma and proposition of Section II hold in this case. The second pair of involutions for the spherical map are
$`\stackrel{~}{I}_1𝐉`$ $`=`$ $`(J_1,J_2\mathrm{cos}(\mu \text{sgn}J_1)+J_3\mathrm{sin}(\mu \text{sgn}J_1),J_2\mathrm{sin}(\mu \text{sgn}J_1)+J_3\mathrm{cos}(\mu \text{sgn}J_1)),`$
$`\stackrel{~}{I}_2𝐉`$ $`=`$ $`(J_1\mathrm{cos}\omega +J_2\mathrm{sin}\omega ,J_1\mathrm{sin}\omega J_2\mathrm{cos}\omega ,J_3),`$
with $`F=\stackrel{~}{I}_2\stackrel{~}{I}_1`$. Hence the mapping has symmetry lines on the unit sphere (9) of $`J_1=0`$, $`J_1=J_2\mathrm{tan}\omega /2`$, $`J_2=J_3\mathrm{tan}(\mu /2\text{sgn}J_1)`$, and $`J_2=J_1\mathrm{tan}\omega /2`$ which are the fixed lines of the involutions $`I_1`$, $`I_2`$, $`\stackrel{~}{I}_1`$, and $`\stackrel{~}{I}_2`$, respectively.
We will only display the eastern hemisphere $`(J_1>0)`$ of the spherical phase space. Under the transformation $`(J_1,J_2,J_3)(J_1,J_2,J_3)`$ the mapping (IV) remains invariant, hence the phase space structure in the the western hemisphere will be symmetrical to the eastern under this reflection. The mapping
$$x=\frac{J_2}{1+J_1},y=\frac{J_3}{1+J_1},$$
takes all points on the eastern hemisphere into the disk $`x^2+y^2<1`$ in the Euclidean plane. All circles in the spherical phase space are transformed to Euclidean circles, but some will appear larger than others of the same periodic orbit because the mapping distorts distances. The phase space portraits for $`\omega =\pi (\sqrt{5}1)`$, $`\mu =0.02`$ and $`\mu =\pi (\sqrt{5}1)`$ are shown in Fig. 9 and Fig. 10, respectively. For small $`\mu `$ the southern hemisphere can be approximated by the planar map, while the northern hemisphere can be approximated by the planar map under the transformation $`\omega 2\pi \omega `$ (compare Fig. 9 with Fig. 1). Hence we would expect there to be a stable first order periodic orbit of period $`n`$ if there exists an integer $`m`$ such that $`1mn`$, $`\mathrm{gcd}(n,m)=1`$, and
$$\frac{m1}{n}<\frac{\omega }{2\pi }<\frac{m}{n}\text{or}\frac{m1}{n}<1\frac{\omega }{2\pi }<\frac{m}{n}\text{if }n\text{ is even,}$$
or
$$\frac{m1/2}{n}<\frac{\omega }{2\pi }<\frac{m}{n}\text{or}\frac{m1/2}{n}<1\frac{\omega }{2\pi }<\frac{m}{n}\text{if }n\text{ is odd,}$$
together with some condition on $`\mu [0,2\pi )`$. Again, the sequences defining these orbits are given by (3), while the corresponding unstable orbits have sequences given by (4). The position of each periodic orbit is found by solving (23). The condition on $`\mu `$ for the existence of each periodic orbit could not be found analytically except in the special cases of $`n=1,2,3,4,6.`$ The orbits of period $`n=1,2,4`$ exist for all $`\mu `$ (and $`\omega `$). The period $`3`$ orbits exist for
$$\mathrm{cos}\mu >\frac{\mathrm{cos}\omega }{1\mathrm{cos}\omega }$$
while the period $`6`$ orbits exist for
$$\mathrm{cos}\mu >\frac{\mathrm{cos}\omega }{1+\mathrm{cos}\omega }.$$
For orbits of higher period one can simply just ignore any condition imposed on $`\mu `$. After solving (23) we then need to check whether we have actually found a periodic orbit. All first order periodic orbits can be found in this manner. As for the hyperbolic mapping, some choices of $`\omega `$ and $`\mu `$ will produce polygonal resonances.
## V Discussion and Conclusion
We are left with the question as to whether the resonances of the stable periodic orbits densely pack the phase space. Recall that if $`\omega /2\pi `$ was rational in the planar map then the phase plane was found to be tiled with polygons. There exist trivial examples of the tiling when $`\omega /2\pi =0,1/2,1/4,3/4,1/3,2/3,1/6,5/6`$. In these cases there are no periodic orbits of higher order. When $`\omega /2\pi =1/4,3/4`$ the phase plane is tiled with a grid of squares. A similar situation occurs when $`\omega /2\pi =1/3,2/3,1/6,5/6`$. Then the phase plane is tiled with triangles and hexagons. In all of these cases the unstable set (closure of the set of all images and preimages of the vertical line $`x=0`$) has zero Lebesgue measure and is one dimensional. For other rationals something different occurs. When $`\omega /2\pi =1/8,3/8,5/8,7/8`$ the phase plane is tiled with octagons of decreasing size in a selfsimilar fashion (see Fig. 4). Using selfsimilarity one can show that the unstable set has zero measure with the Hausdorff dimension hausdorff of $`\mathrm{log}3/\mathrm{log}(1+\sqrt{2})=1.246..`$. When $`\omega /2\pi =1/5,2/5,3/5,4/5,1/10,3/10,7/10,9/10`$ the phase plane is tiled with pentagons and decagons (see Fig. 5) and the unstable set has the Hausdorff dimension of $`\mathrm{log}6/\mathrm{log}(2+\sqrt{5})=1.241..`$. For other rationals the phase plane is tiled with many different types of polygons in patterns of great complexity (see Fig. 11). We were unable to find the dimension of the unstable set for other rationals, but it would be safe to conjecture that they also have zero measure. Hence one could assume that whenever $`\omega /2\pi `$ is rational the unstable set has empty interior and the phase plane is densely packed with polygonal resonances. However when $`\omega /2\pi `$ is irrational we are uncertain as to whether the phase plane will be densely packed with circles. But this seems likely since any irrational can be approximated by a sequence of rationals. Hence a phase plane portrait for $`\omega /2\pi `$ irrational can be approximated by a sequence of dense polygonal packings. As we get closer to the irrational new smaller polygons are created and the number of sides of each existing polygon increases until they approximate circles. This argument is quite naive however and a mathematical proof is needed.
In the hyperbolic and spherical cases there are no polygonal packings and the above argument does not apply. Whether these mappings produce circle packings is also an open question. Consider the case when $`\omega =\mu =\pi (\sqrt{5}1)`$ in the spherical map (Fig. 10). Only the first order periodic orbits of periods 1, 2 and 4 exist. The largest circle is a resonance of a period 2 orbit, the second largest is of period 1, while the two smaller circles are of period 4. All smaller resonances are of higher order. Part of the phase portrait is shown in detail in Fig. 12. At this level of magnification one can see that the layout of the high order resonances is highly irregular and there seems to be no selfsimilarity. This is a great departure from the simple polygonal packings in Fig. 4 and Fig. 5 where the selfsimilarity is immediately apparent. One can also see that if the unstable set has zero measure then its fractal dimension would be extremely close to two. Indeed, in Fig. 12 the unstable set looks more likely to be a set of positive measure.
It is worthwhile to point out that phase space structures similar to that in the planar mapping can be produced in the sawtooth standard map
$`x^{n+1}`$ $`=`$ $`x^n+k\text{saw}(y^n)`$
$`y^{n+1}`$ $`=`$ $`y^n+x^{n+1}`$
if $`k=4\mathrm{sin}^2(\theta /2)`$, where $`(x,y)[0,1)^2`$ and $`\text{saw}(y)=yy1/2`$. Numerical investigations have led Ashwin ashwin to conjecture that the closure of the set of images of the discontinuity ($`y=0`$) has positive Lebesgue measure whenever $`\theta /\pi `$ is irrational. If this conjecture proves true we also expect it to hold for our planar map (when $`\omega /2\pi `$ is irrational) as the two mappings seem closely related. We should also point out that the polygonal tilings of the planar map also arise in ‘polygonal dual billiards’ vivaldi ; tabachnikov1 ; tabachnikov2 .
Regardless of whether we have discovered new examples of circle packings, these deceptively simple mappings demonstrate that phase space structure of stunning complexity can arise in Hamiltonian dynamics. As we have shown, a large class of periodic orbits can be found. Perhaps some form of periodic orbit quantization might be realizable. If nothing else, their aesthetic beauty is an admirable quality. |
no-problem/0001/hep-lat0001008.html | ar5iv | text | # CERN/TH-99-376 CPT-99/PE.3916 LAPTH-Conf-771/99 A numerical treatment of Neuberger’s lattice Dirac operator11footnote 1Talk given by K.J. at the “Interdisciplinary Workshop on Numerical Challenges to Lattice QCD”, Wuppertal, August 22-24,1999
## 1 Lattice formulation of QCD
Today, we believe that the world of quarks and gluons is described theoretically by quantum chromodynamics (QCD). This model shows a number of non-perturbative aspects that cannot be adequately addressed by approximation schemes such as perturbation theory. The only way to evaluate QCD, addressing both its perturbative and non-perturbative aspects at the same time, is lattice QCD. In this approach the theory is put on a 4-dimensional Euclidean space-time lattice of finite physical length $`L`$, with a non-vanishing value of the lattice spacing $`a`$. Having only a finite number of grid points, physical quantities can be computed numerically by solving a high-dimensional integral by Monte Carlo methods, making use of importance sampling.
The introduction of a lattice spacing regularizes the theory and is an intermediate step in the computation of physical observables. Eventually, the regularization has to be removed and the value of the lattice spacing has to be sent to zero to reach the target theory, i.e. continuum QCD. In fact, in conventional formulations of lattice QCD , the introduction of the lattice spacing renders the theory on the lattice somewhat different from the continuum analogue and a number of properties of the continuum theory are only very difficult and cumbersome to establish in the lattice regularized theory. One of the main reasons for this difficulty is that in conventional lattice QCD the regularization breaks a particular symmetry of the continuum theory, which plays a most important role there, namely chiral symmetry.
However, the last few years have seen a major breakthrough in that we now have formulations of lattice QCD that have an exact lattice chiral symmetry . In this approach, many properties of continuum QCD are preserved even at a non-vanishing value of the lattice spacing . This development followed the rediscovery of the so-called Ginsparg–Wilson (GW) relation which is fulfilled by any operator with the exact lattice chiral symmetry of . It is not the aim of this contribution to discuss the physics consequences of the GW relation. We have to refer the interested reader to reviews about these topics. Here we would like to discuss the numerical treatment of a particular lattice operator that satisfies the GW relation, namely Neuberger’s solution . This solution has a complicated structure and is challenging to implement numerically. Thus, the large theoretical advantage of an operator satisfying the GW relation must be weighed against the very demanding computational effort required to implement it.
This contribution is organized as follows. After discussing Neuberger’s lattice Dirac operator we want to show how we evaluated the operator in our practical application and what kind of improvements we found to accelerate the numerical computations. For alternative ideas for improvements, see the contributions of H. Neuberger and A. Borici to this workshop. We finally give some estimates of the computational expense of using Neuberger’s operator.
## 2 Neuberger’s lattice Dirac operator
The operator we have used acts on fields (complex vectors) $`\mathrm{\Phi }(x)`$ where $`x=(x_0,x_1,x_2,x_3)`$ and the $`x_\mu ,\mu =0,1,2,3`$, are integer numbers denoting a 4-dimensional grid point in a lattice of size $`N^4`$ with $`N=L/a`$. The fields $`\mathrm{\Phi }(x)`$ carry in addition a “colour” index $`\alpha =1,2,3`$ as well as a “Dirac” index $`i=1,2,3,4`$. Hence, $`\mathrm{\Phi }`$ is a $`N^434`$ complex vector.
In order to reach the expression for Neuberger’s operator we first introduce the matrix $`A`$
$$A=1+s\frac{a}{2}\left\{\gamma _\mu \left(_\mu ^{}+_\mu \right)a_\mu ^{}_\mu \right\},$$
(1)
where $`_\mu `$ and $`_\mu ^{}`$ are the nearest-neighbour forward and backward derivatives, the precise definition of which can be found in the Appendix. The parameter $`s`$ is to be taken in the range of $`|s|<1`$ and serves to optimize the localization properties of Neuberger’s operator, which is then given by
$$D=\frac{1}{a}\left\{1A\left(A^{}A\right)^{1/2}\right\}.$$
(2)
Through the appearance of the square root in eq. (2), all points on the lattice are connected with one another, giving rise to a very complicated, multi-neighbour action. However, the application of $`D`$ to a vector $`\mathrm{\Phi }`$ will only contain applications of $`A`$ or $`A^{}A`$ on this vector. Since these matrices are sparse, as only nearest-neighbour interactions are involved, we will never have to store the whole matrix.
In the computation of physical quantities, the inverse of $`D`$, applied to a given vector, is generically needed. Hence one faces the problem of having to compute a vector $`X=D^1\eta `$, with $`\eta `$ a prescribed vector (the “source”) as required by the particular problem under investigation. Fortunately, a number of efficiently working algorithms for computing $`X=D^1\eta `$ are known, such as conjugate gradient, BiCGstab, or variants thereof . In conventional approaches to lattice QCD an operator $`\stackrel{~}{D}`$ is used that is very similar to the matrix $`A`$ in eq. (1). Computing the vector $`\stackrel{~}{X}=\stackrel{~}{D}^1\eta `$ requires a number $`n_{\mathrm{iter}}`$ of iterations of some particular method, say BiCGstab. Employing Neuberger’s operator $`D`$ in computing $`X=D^1\eta `$, it turns out that the number of iterations needed is of the same order of magnitude as when using $`\stackrel{~}{D}`$. At the same time, in each of these iterations, the square root has to be evaluated. When this is done by some polynomial approximation, it is found that the required degree of this polynomial is roughly of the same order as the number of iterations needed for computing the vector $`X`$. Hence, with respect to the conventional case, the numerical effort is squared and the price to pay for using the operator $`D`$ is high.
On the other hand, any solution of the Ginsparg–Wilson relation gives us a tool by which particular problems in lattice QCD can be studied, which would be extremely hard to address with conventional approaches. It is for these cases that the large numerical effort is justified, but clearly, we would like to have clever ideas coming from areas such as Applied Mathematics, to decrease the numerical expense or even overcome this bottleneck.
## 3 Approximation of $`\left(A^{}A\right)^{1/2}`$
For computing the square root that appears in eq. (2), we have chosen a Chebyshev approximation by constructing a polynomial $`P_{n,ϵ}(x)`$ of degree $`n`$, which has an exponential convergence rate in the interval $`x[ϵ,1]`$. Outside this interval, convergence is still found but it will not be exponential. The advantages of using this Chebyshev approximation are the well-controlled exponential fit accuracy as well as the possibility of having numerically very stable recursion relations to construct the polynomial, allowing for large degrees. In order to have an optimal approximation, it is desirable to know the lowest and the highest eigenvalue of $`A^{}A`$. A typical example of the eigenvalues of $`A^{}A`$ is shown in fig. 1, where we show the 11 lowest eigenvalues as obtained on a number of configurations using the Ritz functional method .
There is a wide spread and very low-lying eigenvalues appear. Choosing $`ϵ`$ to be the value of the lowest of these eigenvalues would result in a huge degree $`n`$ of the polynomial $`P_{n,ϵ}`$. We therefore computed $`\mathrm{O}(10)`$ lowest-lying eigenvalues of $`A^{}A`$ as well as their eigenfunctions and projected them out of the matrix $`A^{}A`$. The approximation is then only performed for the matrix with a reduced condition number, resulting in a substantial decrease of the degree of the polynomial. In addition, we computed the highest eigenvalue of $`A^{}A`$ and normalized the matrix $`A`$ such that $`A^{}A1`$.
Since our work , aiming at the physical question of spontaneous chiral symmetry breaking in lattice QCD, has been one of the first of its kind, we wanted to exclude possible systematic errors and demanded a very high precision for the approximation to the square root:
$$XP_{n,ϵ}(A^{}A)A^{}AP_{n,ϵ}(A^{}A)X^2/2X^2<10^{16}$$
(3)
where $`X`$ is a gaussian random vector. In our practical applications we fixed this precision beforehand and set $`ϵ`$ to be the $`11`$th lowest eigenvalue of $`A^{}A`$. This then determines the degree of the polynomial $`n`$ and hence our approximation $`D_n`$ to the exact Neuberger operator $`D`$. We checked that the precision we required for the approximation of the square root is directly related to the precision by which the GW relation itself is fulfilled. Choosing $`n`$ such that the accuracy in eq. (4) is reached results in
$$\left[\gamma _5D_n+D_n\gamma _5D_n\gamma _5D_n\right]X^2/X^210^{16}.$$
(4)
In addition, we find that the deviations from the exact GW relation decrease exponentially fast with increasing $`n`$.
## 4 The inverse of Neuberger’s operator
As mentioned above, in physics applications a vector $`D^1\eta `$ has to be computed, with $`\eta `$ a prescribed source vector. Not only is the computation of this vector very costly, there also appears to be a conceptual problem: in inspecting the lowest eigenvalue of $`D_n^{}D_n`$, very small eigenvalues are often found as shown in fig. 2. These very small eigenvalues belong to a given chiral sector of the theory, i.e. their corresponding eigenfunctions $`\chi `$ are eigenfunctions of $`\gamma _5`$ with $`\gamma _5\chi =\pm \chi `$. In fact, these modes play an important physical role as they are associated with topological sectors of the theory .
As far as the practical applications are concerned, it is clear that in the presence of such a small eigenvalue, the inversion of $`D_n`$ will be very costly, as the condition number of the problem is then very high. In order to address this problem, we followed two strategies:
* We compute the lowest eigenvalue of $`D_n^{}D_n`$ and its eigenfunction (using again the Ritz functional method ) and if it is a zero mode –in which case it is also a zero mode of $`D_n`$– we project this mode out of $`D_n`$ and invert only the reduced matrix; this is then well conditioned, as the very small eigenvalues appear to be isolated. In this strategy, the knowledge of the eigenfunction must be very precise and an accuracy of approximating the square root as indicated in eq. (3) is mandatory.
* Again we determine the lowest eigenvalue of $`D_n^{}D_n`$ and the chirality of the corresponding zero mode, if there is any. We then make use of the fact that $`D_n^{}D_n`$ commutes with $`\gamma _5`$. This allows us to perform the inversion in the chiral sector without zero modes. In this strategy, the accuracy demanded in eq. (3) could be relaxed and this strategy, which essentially follows ref. , is in general much less expensive than following strategy $`(i)`$.
However, even adopting strategy $`(ii)`$, solving the system $`D_nX=\eta `$ is still costly. We therefore tried two ways of improving on this. We first note that instead of solving
$$\left[1A/\sqrt{A^{}A}\right]X=\eta $$
(5)
we can equally well solve
$$\left[A^{}\sqrt{A^{}A}\right]X=A^{}\eta .$$
(6)
In practice, however, we found no real advantage in using the formulation of eq. (6). We have further considered two acceleration schemes.
Scheme (a)
We choose two different polynomials (now approximating $`\sqrt{A^{}A}`$ and not the inverse) $`P_{n,ϵ}`$ and $`P_{m,ϵ}`$, $`m<n`$, such that
$$P_{n,ϵ}=P_{m,ϵ}+\mathrm{\Delta }$$
(7)
with $`\mathrm{\Delta }`$ a “small” correction. Then we have
$`\left[A^{}P_{n,ϵ}\right]^1`$ $`=`$ $`\left[A^{}P_{m,ϵ}\mathrm{\Delta }\right]^1`$ (8)
$``$ $`\left[1+\left(A^{}P_{m,ϵ}\right)^1\mathrm{\Delta }\right]\left(A^{}P_{m,ϵ}\right)^1.`$
This leads us to the following procedure of solving $`D_nX=\eta `$:
* first solve
$$\left(A^{}P_{m,ϵ}\right)Y=\eta ;$$
(9)
* then solve
$$\left(A^{}P_{m,ϵ}\right)X_0=\eta +\mathrm{\Delta }Y;$$
(10)
* use $`X_0`$ as a starting vector to finally solve
$$\left(A^{}P_{n,ϵ}\right)X=\eta .$$
(11)
The generation of the starting vector $`X_0`$ in steps (1) and (2) is only a small overhead. In fig. 3 we plot the relative residuum $`ϵ_{\mathrm{stop}}^2=D_nX\eta ^2/X^2`$ as a function of the number of applications of $`D_n`$. In this case $`n=100`$ and $`m=30`$. We show the number of applications of the matrix $`D_n`$ for the case of a random starting vector (dotted line) and the case where $`X_0`$ was generated according to the above procedure (solid line). The gain is of approximately a factor of two.
Scheme (b)
In the second approach, we use a sequence of polynomials to solve $`D_nX=\eta `$. To this end we first solve
$$\left[A^{}P_{m_1,ϵ}\right]X_1=A^{}\eta $$
(12)
by choosing a polynomial $`P_{m_1,ϵ}`$ and a stopping criterion for the solver $`ϵ_{\mathrm{stop}}^{(1)}`$ such that
$$m_1<n,ϵ_{\mathrm{stop}}^{(1)}>ϵ_{\mathrm{stop}}.$$
(13)
The value of $`ϵ_{\mathrm{stop}}^{(1)}`$ is chosen such that it is roughly of the same order of magnitude as the error that the polynomial of degree $`m_1`$ itself induces. The solution $`X_1`$ is then used as a starting vector for the next equation, employing a polynomial $`P_{m_2,ϵ}`$ and stopping criterion $`ϵ_{\mathrm{stop}}^{(2)}`$ with
$$m_1<m_2<n,ϵ_{\mathrm{stop}}^{(1)}>ϵ_{\mathrm{stop}}^{(2)}>ϵ_{\mathrm{stop}}.$$
(14)
This procedure is then repeated until we reach the desired polynomial $`P_{n,ϵ}`$ and stopping criterion $`ϵ_{\mathrm{stop}}`$ to solve the real equation
$$\left[A^{}P_{n,ϵ}\right]X=A^{}\eta .$$
(15)
As for scheme (a), we gain a factor of about two in the numerical effort. We finally remark that some first tests using the scheme proposed in resulted in a similar performance gain as the two schemes presented above.
In table 1 we give a typical example of the expense of a simulation following strategy $`(ii)`$. We list both the cost of computing the lowest eigenvalue of $`D_n^{}D_n`$ in terms of the number of iterations to minimize the Ritz functional and the number of iterations to solve $`D_nX=\eta `$. In both applications, a polynomial of degree $`n`$ is used to approximate the square root. The numbers in table 1 indicate that a quenched calculation, employing Neuberger’s operator, leads to a computational cost that is comparable with a dynamical simulation using conventional operators.
## 5 Conclusions
The theoretical advance that an exact chiral symmetry brings to lattice gauge theory is accompanied by the substantial increase in numerical effort that is required to implement operators satisfying the GW relation. Thus, while the Nielsen–Ninomiya theorem has been circumvented, the “no free lunch theorem” has not. Whether alternative formulations, such as domain wall fermions, can help in this respect remains to be seen.
## Appendix
We give here the explicit definitions needed in eq. (1). The forward and backward derivatives $`_\mu ,_\mu ^{}`$ act on a vector $`\mathrm{\Phi }(x)`$ as
$`_\mu \mathrm{\Phi }(x)`$ $`=`$ $`{\displaystyle \frac{1}{a}}\left[U(x,\mu )\mathrm{\Phi }(x+a\widehat{\mu })\mathrm{\Phi }(x)\right]`$
$`_\mu ^{}\mathrm{\Phi }(x)`$ $`=`$ $`{\displaystyle \frac{1}{a}}\left[\mathrm{\Phi }(x)U(xa\widehat{\mu },\mu )^1\mathrm{\Phi }(xa\widehat{\mu })\right],`$
where $`\widehat{\mu }`$ denotes the unit vector in direction $`\mu `$. The (gauge) field $`U(x,\mu )SU(3)`$ lives on the links connecting lattice points $`x`$ and $`x+a\widehat{\mu }`$ and acts on the colour index $`\alpha =1,2,3`$ of the field $`\mathrm{\Phi }`$. Finally, the Dirac matrices $`\gamma _\mu ,\mu =0,1,2,3`$ are hermitean $`4\times 4`$ matrices acting on the Dirac index $`i`$ of the field $`\mathrm{\Phi }`$. Their explicit form is given by
$$\gamma _\mu =\left(\begin{array}{cc}0& e_\mu \\ e_\mu ^{}& 0\end{array}\right)$$
(16)
with
$$e_0=1,e_k=i\sigma _k$$
(17)
and
$$\sigma _1=\left(\begin{array}{cc}0& 1\\ 1& 0\end{array}\right)\sigma _2=\left(\begin{array}{cc}0& i\\ i& 0\end{array}\right)\sigma _3=\left(\begin{array}{cc}1& 0\\ 0& 1\end{array}\right).$$
(18)
With the choice of the $`\gamma `$ matrices given above, the matrix $`\gamma _5=\gamma _0\gamma _1\gamma _2\gamma _3`$ is diagonal and given by
$$\gamma _5=\left(\begin{array}{cc}1& 0\\ 0& 1\end{array}\right).$$
(19)
We finally note that whenever repeated indices appear, they are summed over. |
no-problem/0001/nlin0001063.html | ar5iv | text | # Spiral Turbulence: From the Oxidation of CO on Pt(110) to Ventricular Fibrillation
## I Introduction
Spiral waves are found in a wide variety of physical, chemical, and biological systems. These include convection patterns in cylindrical cells , patterns in the Belousov-Zhabotinsky chemical reactions , the oscillations of cyclic-AMP in aggregating Dictyostelium Discoidium amoebae , the oxidation of carbon monoxide (CO) on surfaces of platinum single crystals , calcium waves in the cell cytoplasm of Xenophus oocytes, and in ventricular fibrillation which is perhaps the most dangerous form of cardiac arrhythmia. A variety of mathematical models have been used to describe these phenomena; we refer the reader to the review by Cross and Hohenberg for a recent discussion of many such models . The models range from cellular automata to deterministic partial differential equations, such as the Complex Ginzburg Landau equation (for oscillatory chemical reactions near onset) and equations in the FitzHugh-Nagumo class (for CO oxidation and ventricular fibrillation). We concentrate on the FitzHugh-Nagumo types of equations here. Our description is not very technical since the talk on which this article is based was aimed at an audience comprising scientists from different disciplines. We begin with a brief overview of phenomena followed by an introduction to equations in the FitzHugh-Nagumo class. We then present some results for the specific models suggested for the oxidation of CO on Pt(110) and ventricular fibrillation including our recent studies .
Many chemical reactions display spontaneous spatial self-organisation if the reactors are not stirred and the reactants are fed at a constant rate. Among these reactions the oxidation of CO on Pt(110) has been studied in great detail . The experimental setup consists of a cell containing a Pt wafer kept at a constant temperature. The reactants are fed into the cell at a constant rate and the products removed. Spatial structures in the concentration fields of the reactants have been observed by using photoelectron emission microscopy (PEEM). For a picture of the growth of a spiral wave on such a Pt(110) wafer see Ref. .
A human heart consists of four chambers - the two upper atria and two lower ventricles. Heart activity is driven by periodic signals emitted by the cells of the sinus node situated in the upper right atrium. These cells initiate the heartbeat cycle by emitting a pulse of electrical activity. This activity travels as a wave across the atria and stimulates the nerves of the atrioventricular node (AVN). From the AVN the signal travels along the His bundle and into the ventricles via the Purkinje fibre network which extensively innervates the ventricular wall. The ventricles now contract thus completing the beat cycle. Cardiac arrhythmia is caused when the signal from the sinus node does not propagate in the normal manner. Especially dangerous is ventricular fibrillation, an irregular pulsation of the ventricles, which is a major cause of death in the industrialised world . It has been conjectured for many years that ventricular fibrillation is associated with the formation of spiral waves (or their three-dimensional analogues called scroll waves) on the walls of the ventricles. Experimental evidence for such spiral waves has been increasing with the advent of voltage-sensitive dyes and advanced imaging techniques . Figure 2 shows such a spiral wave on the ventricle of a canine heart. Such spirals have also been seen to break up.
Both the oxidation of CO on Pt (110) and ventricular fibrillation have been the focus of many theoretical studies. These are motivated partially by a desire to understand the phenomena seen in experiments and partially to elucidate the general properties of systems displaying such spatiotemporally chaotic, statistical steady states. The remainder of this paper is organised as follows: Section 2 consists of an introduction to FitzHugh-Nagumo equations and the models we consider. Section 3 describes some results for the model of CO oxidation on Pt(110). Section 4 describes some results for the model of ventricular fibrillation. Section 5 ends with concluding remarks.
## II Models
FitzHugh-Nagumo-type equations belong to the general class of reaction-diffusion equations. These are of the form
$$\frac{U_i}{t}=f_i(U_j)+D_{ij}^2U_j;$$
(1)
where the $`f_i`$ are nonlinear functions of the $`U_j`$’s and $`D_{ij}`$ are elements of a matrix of diffusion constants. The FitzHugh-Nagumo-type system used by Hildebrand et. al. for the oxidation of CO on Pt(110) is
$`{\displaystyle \frac{u}{t}}`$ $`=`$ $`^2u{\displaystyle \frac{1}{\epsilon }}u(u1)(u(v+b)/a),`$ (2)
$`{\displaystyle \frac{v}{t}}`$ $`=`$ $`f(u)v;`$ (3)
here the fields $`u`$ and $`v`$ are related to the CO coverage and the surface reconstruction, $`a,b,`$ and $`\epsilon `$ are control parameters. Physically, $`\epsilon `$ is proportional to the ratio of the rate constant for change in surface structure to the rate constant for the oxidation of adsorbed CO on the surface. Further, $`f(u)=0\text{ if }u<\frac{1}{3}\text{ , }f(u)=16.75u(u1)^2\text{ if }\frac{1}{3}u<1\text{, and }f(u)=1\text{ if }u1.`$ The form of $`f(u)`$ leads to the production of $`v`$ only above a certain threshold value of $`u`$. This is an effective model with $`u`$ and $`v`$ related to the reactant coverages and the surface reconstruction. We refer the reader to Ref. for further details and the specific form for $`f(u)`$, which has been found to model the CO oxidation on Pt(110) experiments well. We have studied Model (3) with both periodic and Neumann (no-flux) boundary conditions. Here we restrict ourselves to Neumann B.C.’s for ease of comparision with model (5).
The Panfilov-Hogeweg model for ventricular fibrillation that we consider is
$`{\displaystyle \frac{e}{t}}`$ $`=`$ $`^2ef(e)g,`$ (4)
$`{\displaystyle \frac{g}{t}}`$ $`=`$ $`ϵ(e,g)(keg);`$ (5)
where $`e`$ is the transmembrane potential and $`g`$ is a recovery variable corresponding to the amount of openness of the ion channels . Also, $`f(e)=C_1e`$ if $`e<e_1`$; $`f(e)=C_2e+a`$ if $`e_1ee_2`$; $`f(e)=C_3(e1)`$ when $`e>e_2`$, and $`ϵ(e,g)=ϵ_1`$ when $`e<e_2`$; $`ϵ(e,g)=ϵ_2`$ when $`e>e_2`$, and $`ϵ(e,g)=ϵ_3`$ when $`e<e_1`$ and $`g<g_1`$. Physically $`ϵ(e,g)^1`$ determines the refractory time period of this model of ventricular muscle. Since the ventricles are electrically insulated from the atria, we impose Neumann boundary conditions.
## III CO oxidation on Pt(110)
For the model of CO oxidation on Pt(110) (3), Hildebrand et. al. , find that there are many different statistical steady states. These include states with no waves (N), states with flat shrinking waves (F), states with rigidly rotating spirals (S), states with meandering spirals (M) and turbulent states with creation and annihilation of spirals (T1 and T2). These authors have found a stability diagram or ”phase diagram” in the $`b\epsilon `$ plane for $`a=0.84`$, and have calculated the spatial autocorrelation function of the density of spiral cores $`n(\stackrel{}{x},t)n(\stackrel{}{x}+\stackrel{}{s},t+\tau )`$. They suggest that the change in the spiral dynamics on passing from T1 to T2 is like a ”liquid-gas” transition. We have simulated model (3) by discretising it on a square lattice and using a variable stepsize fourth-fifth order Runge-Kutta integrator for time stepping . In some cases we also use a fast integration scheme proposed by Barkley . We have used Neumann boundary conditions to compare with model (5). We have investigated the local dynamics, i.e., the dynamics obtained by plotting $`u(\stackrel{}{x_n},t_n)`$ versus $`v(\stackrel{}{x_n},t_n),n=1,2,\mathrm{}`$ for fixed spatial location $`\stackrel{}{x_n}`$ and $`t_n=n\mathrm{\Delta }`$ where $`\mathrm{\Delta }`$ is a sampling interval which we choose to be $`=0.1`$ time units. Such local phase portraits show that, in the states S, M, and T1, the system displays oscillatory behaviour; a representative phase portrait is shown in Fig. (2) for T1.
The trajectory in the local phase portrait loops around the fixed point $`(u_{},v_{})(0.66,0.484)`$ of equation (3) without the $`^2`$ term. We define the phase $`\varphi (𝐱,t)\mathrm{tan}^1((v(𝐱,t)v_{})/(u(𝐱,t)u_{}))`$ and note that it winds by $`2\pi `$ around the cores of spiral defects . Thus it can be used to obtain the defect density $`\rho `$ .
## IV Ventricular Fibrillation
It is believed that ventricular fibrillation occurs because of the spontaneous breakdown of contraction waves in the heart muscle leading to a state with many rotating spirals . Physiological abnormalities do not seem to play an important role in this arrhythmia in a sizeable fraction of cases. Thus it is interesting to study models in which spiral breakup appears spontaneously without introducing spatial inhomogenieties in the diffusion constant, for example. For the model (5), Panfilov et. al. find that, for the parameter values $`e_1=0.0026,e_2=0.837,C_1=20,C_2=3,C_3=15,a=0.06,k=3,ϵ_{1}^{}{}_{}{}^{1}=75,ϵ_{2}^{}{}_{}{}^{1}=1,g_1=1.8`$ and $`0.5<ϵ_{3}^{}{}_{}{}^{1}<10`$, an initial condition in the shape of a broken wave gives rise to states containing many spirals. They have also calculated an electrocardiogram (ECG) numerically for this model in the turbulent state. In this section we elucidate the spatiotemporal nature of the chaotic behaviour of this system. In Ref. it has been noted that the breakdown of spirals in models (3) and (5) is qualitatively similar. We elaborate on this similarity below. The state with broken spirals in model (5) is similar to the state T1 in model (3) as can be seen by comparing Figs. (5) and (3).
We have simulated model (5) using finite-differencing in space with a forward euler algorithm for time stepping. Most of our simulations are done with a system size of $`256\times 256`$ with a spatial grid step of $`dx=0.5`$ and a time step of $`dt=0.022`$. Physically, one time unit is equal to 5 milliseconds, while one spatial unit is equal to 1 millimeter. We use the parameter values given above. We have calculated a local phase portrait for this model, i.e. a plot of $`e(\stackrel{}{x}_n,t_n)`$ vs $`g(\stackrel{}{x}_n,t_n)`$ for a fixed spatial location $`\stackrel{}{x}_n`$. This is shown in Fig. (6). Note that this local phase portrait is similar to that in the T1 state of the CO oxidation model (3).
We find that for a system of size $`50\times 50`$ the lifetime $`\tau _t`$ of the turbulent state is $`300`$ time units. After this time the entire field configuration goes to zero for various initial conditions. For a system of size $`100\times 100`$ $`\tau _t950`$ time units. We conjecture that $`\tau _t`$ grows with system size; we are investigating this scaling of $`\tau _t`$. The typical size of a human heart corresponds to a system size of $`600\times 600`$ so the size dependence of $`\tau _t`$ is of obvious interest. We note in passing that this qualitative trend of $`\tau _t`$ increasing with system size is in accord with the well-known phenomenon that only the larger mammals have heart attacks.
To characterise chaos in model (5), we have calculated the maximum Lyapunov exponent $`\lambda _{max}`$ for a system of size $`128\times 128`$. The maximum Lyapunov exponent measures the divergence of nearby trajectories in the system. To calculate it we use standard methods . In the transient turbulent state of model (5), $`\lambda _{max}`$ saturates to a value of $`0.2`$. This is shown in Fig (7). Note that when this transient decays, the maximum Lyapunov Exponent decreases to negative values.
## V Conclusions
Spiral patterns are seen in a wide variety of natural systems . Here we have concentrated on two such systems which are also excitable media. The chaotic states of these systems display the breakup of large spirals into smaller ones in the states T1 and T2. The qualitative similarity of such spiral breakup in models (3) and (5) has been mentioned briefly earlier . Here we have elucidated this similarity by comparing the spiral patterns (Figs. 3 and 5) in the state T1 of model (3) and the chaotic state of model (5); local phase portraits (Figs. 2 and 6) are also similar. We have characterised the chaos in model (5) by computing the maximum Lyapunov exponent $`\lambda _{max}`$ (Fig. 7) as we have done for model (3) elsewhere .
Our study shows that there is considerable similarity between the behaviour of model (3) in the state T1 and model (5) for the parameter values given above. It should be possible to exploit this similarity to gain new insights about these two excitable media (e.g., by doing experiments on the oxidation of CO on Pt(110) which might be hard to do on a human heart). Of course some caution must be exercised while doing this for the two models are different in some respects: the chaotic state of model (5) is a transient for the system sizes we have studied, whereas the analogous state for model (3) seems to be a statistical steady state; furthermore, if the $`^2`$ terms are dropped in both models, then model (3) has an extra unstable fixed point \[at $`(u,v)=(u_{},v_{})`$\], i.e., the reaction kinetics are different in the two models. We hope our work will stimulate experimental studies that will try to elucidate the similarities and differences between spiral turbulence in the oxidation of CO on Pt(110) and ventricular fibrillation. Since ventricular fibrillation in cardiac muscle is fatal , any insight gained about it by studying similar systems is of great interest.
We thank CSIR and JNCASR for support and SERC (IISc) for computational facilities. |
no-problem/0001/astro-ph0001260.html | ar5iv | text | # Evidence for Stellar Streaming in the Cores of Elliptical Galaxies: A Kinematic Signature of Mergers?
## 1 Photometric Properties of Elliptical Galaxies
It is generally accepted that elliptical galaxies are the merger product of smaller mass systems (Toomre 1977). While elliptical galaxies populate a “fundamental plane” in their global scaling properties (e.g., Dressler et al. 1987; Djorgovski & Davis 1987), they also appear to be broadly composed of two families, giants and dwarfs (e.g., Kormendy & Djorgovski 1989). These distinct properties are often assumed to be a result of luminous ellipticals forming from lower luminosity systems through a merger process. Recently, Faber et al. (1997) have shown that the photometric profiles of the cores of elliptical galaxies can also be broadly classified into two groups, those with “cores” and those with “cusps”. Systems with cusps have surface brightness profiles which continue to rise into their inner-most regions. In contrast, systems with cores have an inner region of almost constant surface brightness with a well-defined radius where the profile breaks to form the standard $`R^{1/4}`$ profile. Core profiles are found exclusively within giant ellipticals while those systems with photometric cusps are invariably lower luminosity systems. Faber et al. (1997) suggested that the distended cores of the higher luminosity systems could be the result of a kinematic heating of the stellar population supplied by massive BHs which survive the merger process. In this scenario, these massive black holes may produce detectable stellar “wakes” as they traverse the background sea of relatively low mass stars.
We have found evidence for non-Gaussian velocity fields within the cores of giant ellipticals. In §2 we describe our observational data and analysis procedures. In §3 we introduce a “structure index” in order to quantify the kinematic structure we find. We present evidence that the structure index is strongly correlated with the absolute magnitude of the galaxy. We discuss some possible interpretations of this phenomena in §4 and §5 contains a summary of our results.
## 2 Spectroscopy of Elliptical Galaxies
High signal-to-noise, medium resolution spectroscopy of several hundred elliptical galaxies has been obtained with the WIYN 3.5-m telescope on Kitt Peak as part of an extensive investigation of the fundamental plane of early-type galaxies. The sample presented in this paper includes the brighter elliptical galaxies in the Virgo and Coma clusters. The Virgo data were acquired with “DensePak” in order to sample these nearby systems over similar spatial scales as do 2-3 arcsec fibers with more distant galaxies. DensePak is an integral-field unit consisting of a $`7\times 13`$ array of 3 arcsec diameter fibers which sample a $`30\times 45`$ arcsec region on the sky (Barden, Sawyer & Honeycutt 1998) . A second fiber bundle consisting of 96, robotically positioned, 2-arcsec diameter fibers (Hydra) was used to acquire spectroscopy of the Coma galaxies.
The exposure times using DensePak were typically 30-min for each Virgo galaxy while the Coma data constituted 4 1-hour exposures with Hydra. The signal-to-noise was typically about 50 per pixel. Both fiber bundles feed a versatile, bench-mounted spectrograph enabling a wide variety of configurations. We chose the standard f/6 paraboloid collimator with the refractive f/1.43 “red” camera and the 860 l/mm grating in second order to produce a reciprocal dispersion of 19.9 Å/mm or 0.477 Å/pixel with the Tektronics 2048 $`\times `$ 2048, thinned CCD (T2KC). This configuration produced a FWHM of 2.5 pixels as measured from the comparison lines corresponding to an instrumental velocity resolution (1 $`\sigma `$) of 29 km s<sup>-1</sup>. We chose the spectral region from $`\lambda \lambda `$ 4813Å $``$ 5793Å including H$`\beta `$, Mg-b, and Fe + CaI. Comparison spectra of a CuAr lamp were taken periodically over the course of the night. Flat field calibration was done on each night using a quartz lamp in order to correct for the fiber-to-fiber sensitivity variations and to define the extraction regions for the individual spectra. Observations of several “super metal-rich” K-giant stars (Faber et al. 1985) were obtained through different fibers over the course of each night. In the case of the DensePak data, we acquired these template stars both in and out of focus in order to test for artifacts introduced by fiber-to-fiber variations in the spectrograph. We also acquired exposures of the twilight sky on each night. The only differences discernible between any of the template and/or twilight exposures were very small and negligible focus variations along the spectrograph slit.
All the spectra were processed and extracted using the “DoHydra” script within IRAF. The flat-field exposures were used to define the apertures for optimal extraction. CuAr comparison spectra were extracted and fitted to provide the wavelength calibration. Sky spectra were extracted and co-added. For the Hydra data we allocated 25 fibers to the sky but in the case of DensePak only 4 fibers are available for sky. The calibrated and extracted spectra were saved in a “multispec” format for later analysis. We began by computing the median intensities of each of the DensePak spectra. From these we calculated the intensity-weighted, spatial centroid of the galaxy on the DensePak array. Given the projected position of each fiber on the sky, the spectra were binned over various spatial scales to create summed object spectra. The spectra of the template K giant stars were extracted and cross-correlated with each object spectra using the FXCor task within IRAF. This task is based upon the algorithm of Tonry & Davis (1979). The cross-correlation is performed in Fourier space but any model fitting is done in real space. One advantage of this approach is that the resulting profile is easily compared to a model, in our case a Gaussian, and any deviations are readily apparent.
## 3 Results
Upon examination of the cross-correlation profiles of the Coma data we were immediately struck by the presence of strong irregularities in the profiles of the most luminous systems. The DensePak data for Virgo were found to show similar features. We show some selected examples in Figure 1 along with the best-fit Gaussian models. The Coma data were taken over the course of several nights and the data from each night was reduced independently. We found that the features present in the cross-correlation peak repeated from night-to-night. We tried several different K-giant stars as templates but could not find any detectable differences in the cross-correlation profiles. For the DensePak data nearly all the Virgo galaxies were centered on the same fiber of the array. We did find that when the DensePak data were binned over successfully larger radii the amplitude of the deviations decreased rapidly. Over scales of $`R<15`$ arcsec, or about 1 kpc, the cross-correlation profiles became highly Gaussian. While it is possible that small asymmetries in the profiles could be introduced by line-strength variations it seems implausible that such variations can produce the dramatic substructure seen in Figure 1. We therefore conclude that the features we find in the cross-correlation profiles are real and due to distinct kinematic components in the velocity field of these galaxies.
In an attempt to quantify the degree of substructure in the cross-correlation profiles we defined a “structure index” (S). Using the Coma data taken over different nights as a guide we adopted criteria as to what constituted a significant deviation from the best-fit Gaussian profile. Deviations below the full width at half maximum (FWHM) of the profile were ignored as were any deviations which were centered on the profile peak. We then simply counted the deviations, both negative and positive, from the best fitting Gaussians. The value of S was estimated by both of us and any discrepancies were resolved. Figure 2 shows the value of S for each galaxy plotted against its absolute magnitude in B. The apparent B magnitudes were taken from the RC3 (de Vaucouleurs et al. 1991) or computed from our unpublished I-band photometry assuming a mean $`BI`$ color of 2.35. For the Virgo sample we assumed an average distance modulus of 31.0 and for the Coma sample we assumed an average distance modulus of 34.9. The spectroscopic and photometric data will be presented in greater detail at a later date. Figure 2 shows a clear trend of S increasing with the absolute magnitude of the galaxy for those systems with $`M_B20.0`$. Lower luminosity systems have smoother, Gaussian-like velocity fields (i.e., low S). Both the Virgo and Coma samples show the trend of increasing S with luminosity and a similar “transition luminosity”. There is a hint that the lower luminosity systems in Coma may show more structure than do comparable systems in Virgo but the inherent uncertainty in our index (S) prevents a more definitive conclusion.
## 4 Discussion and Conclusions
The kinematic structure we find within the cores of luminous elliptical galaxies is inconsistent with the smooth, Gaussian profiles expected from classical models for elliptical galaxies (e.g., Binney & Tremaine 1987). One possible explanation is that it is produced by “stellar streaming”. However, there may be a problem with the time scale. If we estimate the stellar orbital periods within the cores of luminous ellipticals as their scale ($``$ 10 kpc) divided by the velocity dispersion ($`300`$ km/sec), we obtain a typical period of $`10^7`$ years. Since phase mixing is expected to dilute coherence over only a few orbital periods we estimate the lifetime of such stellar streams to be only a few $`\times 10^7`$ years. For comparison the typical crossing time of the galaxies within these clusters is much longer, $`10^9`$ years. As a result, it seems unlikely that galaxy-galaxy encounters are sufficiently frequent to maintain the kinematic structure which we find.
While there may be alternative scenarios which could produce kinematic structure, we favor a merger scenario. Faber et al. (1997) recently suggested that the existence of multiple Black Holes (BH) could explain the distinction between the photometric profiles of elliptical galaxies. The distended cores for the most luminous systems can be maintained if the BHs presumed to be present in the nuclei of lower luminosity galaxies (e.g., van der Marel 1999) survive the merger process and then heat the stellar distribution through gravitational encounters (e.g., Quinlan & Hernquist 1997). Faber et al. (1997) found that the “break radius” which separates the inner core profile and the outer $`R^{1/4}`$ profile correlated strongly with galaxy luminosity; the most luminous systems had the largest cores. Interestingly, the luminosity at which we find the kinematic structure to increase is essentially the same as that found by Faber et al. (1997) for the transition between ellipticals with cores and those with cusps. We interpret this as further evidence that these two phenomena are related. In the context of their model, the kinematic structure which we find might be explained as the “stellar wakes” produced by these BHs as they traverse through the background distribution of stars. We speculate that the increased structure we find within the most luminous systems could reflect a larger number of merger events for these galaxies. If so, an investigation of this phenomena in field environments might be warranted. Furthermore, the development of the NGST could provide the opportunity of investigating this phenomena at redshifts of $``$ 0.5 and perhaps allow the merger history of luminous elliptical galaxies to be quantified.
## 5 Summary
We find evidence for kinematic structure within the cores of luminous elliptical galaxies. Lower luminosity systems show smooth, Gaussian velocity fields while the higher luminosity systems show a more irregular velocity field. The structure within luminous systems is most prominent within 100 pc of the center with smooth Gaussian velocity fields more characteristic of the integrated light within 1 kpc. We define a structure index which is found to be well correlated with the absolute magnitude of the galaxy. We favor an interpretation of this phenomena as resulting from coherent stellar streaming and suggest that it could be a relic signature of the merger history of luminous elliptical galaxies. There have been recent suggestions that luminous ellipticals may harbor several massive Black Holes which heat the stars and produce the extended luminosity profiles found in these systems. In this scenario, the kinematic substructure reported here could be interpreted as the “stellar wakes” produced by these Black Holes.
## 6 Acknowledgments
We wish to thank Sam Barden (NOAO), Dave Sawyer (WIYN), and Kent Honeycutt (Indiana) for the development of DensePak. Robin Tripoli assisted with some of the observations described in this paper. |
no-problem/0001/hep-ph0001138.html | ar5iv | text | # GLUON RADIATION IN TOP PRODUCTION AND DECAY11footnote 1Presented at the International Workshop on Linear Colliders, Sitges, Spain, April 28-May 5, 1999
## 1 Monte Carlo Calculation
A high energy linear electron-positron collider will provide an excellent environment for producing and studying the top quark, free of the large QCD backgrounds produced in hadron colliders. An important QCD effect which does occur at $`e^+e^{}`$ colliders, however, is the radiation of gluons from the produced top quark and/or its decay products. Radiated gluons appear as additional jets in top events which can complicate event identification and top mass reconstruction.
In this talk we examine gluon radiation in the production and decay ($`tWb`$) of top quark pairs at high energies; see also . At lepton colliders, there is no gluon radiation from the initial state, but the top quarks and their daughter $`b`$ quarks can emit gluons.<sup>2</sup><sup>2</sup>2We neglect radiation from decay products of the $`W`$ bosons. We perform a Monte Carlo calculation of the process
$$e^+e^{}\gamma ^{},Z^{}t\overline{t}(g)bW^+\overline{b}W^{}g.$$
(1)
We compute exact matrix elements of the contributing diagrams with all spin correlations and the bottom mass included. We keep the finite top width $`\mathrm{\Gamma }_t`$ in the top quark propagator and include all interferences between diagrams, and we use exact kinematics in all parts of the calculation.
Radiation from the top quarks can occur as part of the top production process (before the top quarks go on shell) or as part of the decay process (after the top go on shell). Radiation from the $`b`$ quarks is always part of top decay. For reconstructing the top quark mass (to identify top events and to measure $`m_t`$), whether to include the gluon in the reconstruction is determined by whether the radiated gluon is part of the production or decay process. If it is emitted in the production stage, we obtain the top mass from the $`W`$ and $`b`$ momenta only ($`m_t^2=p_{Wb}^2`$), but if it is part of the decay, we must include it in the reconstruction ($`m_t^2=p_{Wbg}^2`$).
In practice this distinction cannot be made absolutely in an experiment, but it is useful to make it in the calculation. There is a single Feynman diagram that contributes to both production- and decay- stage radiation from the top quark, and the corresponding matrix element contains two propagators:
$$ME\left(\frac{1}{p_{Wbg}^2m_t^2+im_t\mathrm{\Gamma }_t}\right)\left(\frac{1}{p_{Wb}^2m_t^2+im_t\mathrm{\Gamma }_t}\right).$$
(2)
We can rewrite the product as
$$\frac{1}{2p_{Wb}p_{Wbg}}\left(\frac{1}{p_{Wb}^2m_t^2+im_t\mathrm{\Gamma }_t}\frac{1}{p_{Wbg}^2m_t^2+im_t\mathrm{\Gamma }_t}\right),$$
(3)
which separates the production and decay contributions according to where the terms in parentheses peak. The cross section in turn contains separate production and decay contributions.<sup>3</sup><sup>3</sup>3It also contains interference terms, which in principle confound the separation but in practice are small.
## 2 Results
We show in Figure 1 the fraction of the total cross section due to production stage emission, in events with an extra gluon, as a function of the minimum gluon energy. The solid histogram is for center of mass energy 1 TeV, and the dashed histogram for 500 GeV. The main effect is due to kinematics — both curves fall off with increasing gluon energy because of a reduction in the phase space for gluon emission, and the production fraction is always smaller for the lower collision energy. In both cases decay-stage radiaion dominates for all gluon energy thresholds.
We now consider mass reconstruction. Figure 2 shows top invariant mass distributions with and without the extra gluon included. In both cases there is a clear peak at the correct value of $`m_t`$. In the left-hand plot, where the gluon is not included in the reconstruction, we see a low-side tail due to events where the gluon was radiated in the decay. Similarly, in the right-hand plot we see a high-side tail due to events where the gluon was radiated in association with production, and was included when it should not have been.
The narrowness of the peaks and the length of the tails in Figure 2 suggests that an invariant mass cut would be useful to separate the two types of events. We can in fact do even better using cuts on the angle between the gluon and the $`b`$ quarks. This works because although there is no collinear singularity for radiation from massive quarks, the distribution of gluons radiated from $`b`$ quarks peaks close to the $`b`$ direction. Such gluons are emitted in decays. We should note that an important reason the cuts are so effective is that we work at the parton level. In an experiment, hadronization and detector effects are likely to significantly reduce the resolution.
Finally we turn to the interference between the production- and decay-stage radiation, which though small is interesting for its sensitivity to the top width $`\mathrm{\Gamma }_t`$ (about 1.5 GeV in the Standard Model). The interference between the two propagators shown above can be thought of as giving rise to two overlapping Breit-Wigner resonances. The peaks are separated roughly by the gluon energy, and each curve has width $`\mathrm{\Gamma }_t`$. Therefore when the gluon energy becomes comparable to the top width, the two Breit-Wigners overlap and interference can be substantial. In constrast, if the gluon energy is much larger than $`\mathrm{\Gamma }_t`$, overlap and hence interference is negligible. Hence the amount of interference serves as a measure of the top width.
This effect was discussed in using the soft gluon approximation, and Figure 3 shows that the effect remains in the exact calculation. There we plot the distribution in the angle between the emitted gluon and the top quark for gluon energies between 5 and 10 GeV and with $`\mathrm{cos}\theta _{tb}<0.9`$. The center-of-mass energy is 750 GeV. The histograms show the decomposition into the various contributions. The negative solid histogram is the production-decay interference, and we see that not only is it substantial, it is also destructive. That means that the interference serves to suppress the cross section. If the top width is increased, the interference is larger, further suppressing the cross section. Although the sensitivity does not suggest a precision measurement, it is worth noting that the top width is difficult to measure by any means, and it is the total width that appears here.
In summary, we have presented preliminary results from an exact parton-level calculation of real gluon radiation in top production and decay at lepton colliders, with the $`b`$ quark mass and finite top width, as well as all spin correlations and interferences included. We have indicated some of the issues associated with this gluon radiation in top mass reconstruction and top width sensitivity in the gluon distribution. Further work is in progress.
## Acknowledgments
Work supported in part by the U.S. Department of Energy, under grant DE-FG02-91ER40685 and by the U.S. National Science Foundation, under grant PHY-9600155.
## References |
no-problem/0001/hep-ph0001305.html | ar5iv | text | # 1 Limits on 𝜆 (in meters) that can be obtained (using unitarity) from DAΦNE, BABAR, CPLEAR, NA31 and E731 for the cases 𝐸^∗=𝑝⃗²/𝐸, 𝐸^∗=𝐸, 𝐸^∗=|𝑝⃗| and 𝐸^∗=𝑀.
CERN-TH/2000-034
hep-ph/0001305
January 2000
Testing $`\kappa `$-Poincaré with neutral kaons
Giovanni AMELINO-CAMELIA<sup>1</sup><sup>1</sup>1Marie Curie Fellow of the European Union (address from February 2000: Dipartimento di Fisica, Universitá di Roma “La Sapienza”, Piazzale Moro 2, Roma, Italy) and Franco BUCCELLA<sup>2</sup><sup>2</sup>2On leave of absence from Dipartimento di Scienze Fisiche, Universitá di Napoli, Mostra D’Oltremare, Pad. 19, Napoli, Italy.
Theory Division, CERN, CH-1211, Geneva, Switzerland
ABSTRACT
In recent work on experimental tests of quantum-gravity-motivated phenomenological models, a significant role has been played by the so-called “$`\kappa `$” deformations of Poincaré symmetries. Sensitivity to values of the relevant deformation length $`\lambda `$ as small as $`510^{33}m`$ has been achieved in recent analyses comparing the structure of $`\kappa `$-Poincaré symmetries with data on the gamma rays we detect from distant astrophysical sources. We investigate violations of CPT symmetry which may be associated with $`\kappa `$-Poincaré in the physics of the neutral-kaon system. A simple estimate indicates that experiments on the neutral kaons may actually be more $`\lambda `$-sensitive than corresponding astrophysical experiments, and may already allow to probe values of $`\lambda `$ of order the Planck length.
It has been recently realized that observations of gamma rays from distant astrophysical sources can be used to set stringent bounds on the length parameter that characterizes “$`\kappa `$-Poincaré”, one of the most studied scenarios for the dimensionful quantum deformation of Poincaré symmetries, which has been developed most notably in Refs. . This realization has attracted significant attention because the level of sensitivity of planned gamma-ray observatories (see, e.g., Refs. ) should be sufficient to explore values of the length parameter as small as the Planck length ($`L_p1.610^{35}m`$). Only very few other experimental contexts can achieve this type of “Planckian sensitivity”, and such sensitivity levels with respect to $`\kappa `$-Poincaré are especially meaningful in light of the fact that it appears likely that the space-time foam of quantum gravity would induce a dimensionful (although, of course, not necessarily “$`\kappa `$”) deformation of Poincaré symmetries.
We are here interested in exploring the consequences of the fact that the $`\kappa `$-Poincaré deformation would also induce a corresponding deformation of CPT invariance. There is a wide literature on the idea that ordinary CPT invariance might be violated by quantum-gravity effects (see, e.g., Refs. ), but usually the relevant quantum-gravity scenarios are not developed to the point of allowing a definite prediction for the effects of violation of ordinary CPT invariance, and one is led to the use of multi-parameter phenomenological approaches. As observed in Ref. , in the $`\kappa `$-Poincaré framework it appears plausible that one should arrive at a definite prediction for all CPT violating effects parametrized only by the single length parameter $`\lambda `$ characterizing<sup>3</sup><sup>3</sup>3Note that in some of the related literature the length parameter we denote by $`\lambda `$ is written as $`\mathrm{}\kappa ^1`$ (see, e.g., Refs. ) or $`\mathrm{}cE_{QG}^1`$ (see, e.g., Refs. ). Following Ref. , we adopt the $`\lambda `$ notation to emphasize the forefront role played by the $`\kappa `$-Minkowski space-time in our considerations. In particular, a requirement of duality with respect to $`\kappa `$-Minkowski singles out the realization of $`\kappa `$-Poincaré which is considered here and in Refs. . the underlying non-commutative “$`\kappa `$-Minkowski” space-time
$$[x^j,t]=ı\lambda \frac{x^j}{c},[x^j,x^k]=0$$
(1)
where $`c`$ is the conventional speed-of-light constant and $`j,k=1,2,3`$.
Such a single-parameter description of CPT violation within the $`\kappa `$-Poincaré framework however still requires the development of several mathematical tools on the non-commutative $`\kappa `$-Minkowski space-time. Since it appears likely that a long time will be necessary for these technical developments to mature, in this Letter we present a simple heuristic estimate of $`\kappa `$-Poincaré-induced CPT violation in the neutral-kaon system. As already emphasized in Ref. , the neutral-kaon system, with its delicate balance of mass/length scales, can provide a natural context for setting bounds on $`\lambda `$ through tests of CPT invariance. It is in fact well established that observable properties of the neutral-kaon system are extremely sensitive to any deviation from ordinary CPT invariance.
Our heuristic estimate is based on the $`\kappa `$-deformed dispersion relation
$$c^4M^2=\frac{\mathrm{}^2c^2}{\lambda ^2}\left(e^{\frac{\lambda E}{\mathrm{}c}}+e^{\frac{\lambda E}{\mathrm{}c}}2\right)c^2\stackrel{}{p}^2e^{\frac{\lambda E}{\mathrm{}c}}E^2c^2\stackrel{}{p}^2+\frac{\lambda cE}{2\mathrm{}}\stackrel{}{p}^2,$$
(2)
where $`\mathrm{}`$ is the Planck constant, $`M`$, $`\stackrel{}{p}`$ and $`E`$ respectively denote the mass, the momentum and the energy of the particle, and, since available data already imply $`\lambda <510^{33}m`$, on the right-hand side we used the fact that in all contexts of possible interest the particles will satisfy $`E\mathrm{}c/\lambda `$. Clearly, solving for $`E`$ in Eq. (2) one does not recover the ordinary result (with its traditional two solutions of equal magnitude and opposite sign); instead, one finds that the two solutions $`E_+`$, $`E_{}`$ are given by
$$E_\pm \frac{\lambda c}{2\mathrm{}}\stackrel{}{p}^2\pm \sqrt{c^4M^2+c^2\stackrel{}{p}^2}.$$
(3)
Since it is anyway quite natural for quantum gravity to violate CPT invariance and Eq. (3) provides not exactly opposite solutions for $`E_+`$ and $`E_{}`$, as an heuristic $`\kappa `$-Poincaré-inspired argument we conjecture that in the phenomenological Hamiltonian describing neutral-kaon dynamics in presence of CPT violation within otherwise ordinary quantum mechanics<sup>4</sup><sup>4</sup>4The possibility of CPT violation outside quantum mechanics has also been considered in the literature , and it is not implausible that it might turn out to be the proper way to describe CPT violation in the $`\kappa `$-Poincaré framework; however, for the present preliminary and heuristic analysis we shall only consider the possibility of CPT violation within quantum mechanics.
$$H=\left(\begin{array}{cc}(M+\frac{1}{2}\delta M)\frac{1}{2}i(\mathrm{\Gamma }+\frac{1}{2}\delta \mathrm{\Gamma })& M_{12}^{}\frac{1}{2}i\mathrm{\Gamma }_{12}^{}\\ M_{12}\frac{1}{2}i\mathrm{\Gamma }_{12}& (M\frac{1}{2}\delta M)\frac{1}{2}i(\mathrm{\Gamma }\frac{1}{2}\delta \mathrm{\Gamma })\end{array}\right),$$
(4)
the parameter $`\delta M`$ could take a value of order
$$|\delta M|\frac{E_+E_{}}{E_++E_{}}2M\frac{\lambda c}{\mathrm{}}\frac{\stackrel{}{p}^2M}{\sqrt{c^4M^2+c^2\stackrel{}{p}^2}}.$$
(5)
We also make an heuristic estimate of the parameter $`\delta \mathrm{\Gamma }`$ using the observation that a difference $`\delta M`$ in the masses would induce a corresponding difference for the rate into two pions, which gives the main contribution to the width of the opposite-strangeness kaons. The amplitude is proportional to $`M^2M_\pi ^2`$ (it should vanish in the SU(3) limit )
$$\mathrm{\Gamma }\frac{\sqrt{M^24M_\pi ^2}(M^2M_\pi ^2)^2}{M^2}$$
(6)
and (if $`\delta \mathrm{\Gamma }`$ is exclusively due to $`\delta M`$) this leads to the estimate
$$\frac{|\delta \mathrm{\Gamma }|}{\mathrm{\Gamma }}\left(\frac{3M^47M^2M_\pi ^28M_\pi ^4}{M^45M^2M_\pi ^2+4M_\pi ^4}\right)\frac{\delta M}{M}4\frac{\delta M}{M},$$
(7)
which according to (5) corresponds to
$$|\delta \mathrm{\Gamma }|4\frac{\lambda c}{\mathrm{}}\frac{\stackrel{}{p}^2\mathrm{\Gamma }}{\sqrt{c^4M^2+c^2\stackrel{}{p}^2}}.$$
(8)
Based on the structure of our estimate (7), the fact that $`\mathrm{\Gamma }/M1`$ and the fact that in precision measurements on the neutral-kaon system $`\delta M`$ competes with the quantity $`M_{K_L}M_{K_S}3.510^{15}GeV`$ and $`\delta \mathrm{\Gamma }`$ competes with $`\mathrm{\Gamma }7.410^{15}GeV`$ we conclude that the bounds on $`\lambda `$ that can be derived from our estimate (8) of $`\delta \mathrm{\Gamma }`$ are necessarily much less stringent than the corresponding bounds obtainable from our estimate (5) of $`\delta M`$. We restrict our attention to (5) in the following.
A prominent feature of our estimate (5) is that, for given $`|\lambda |`$, it predicts a $`|\delta M|`$ which is an increasing function of $`|\stackrel{}{p}|`$, quadratic in the non-relativistic limit and linear in the ultra-relativistic one. Therefore among experiments achieving comparable $`\delta M`$ sensitivity the ones studying more energetic kaons are going to lead to more stringent bounds on $`\lambda `$.
Meson factories in $`e^+e^{}`$ rings, such as DA$`\mathrm{\Phi }`$NE , can provide strong tests of CPT . In particular, the KLOE experiment is expected to reach sensitivity to values of $`\delta M`$ around $`310^{18}GeV`$, which, using our estimate (5) and the fact that the kaons have momenta of about $`110MeV`$ at the $`\varphi `$ resonance, corresponds to sensitivity to values of $`|\lambda |`$ around $`610^{32}m`$.
Of course, the argument for $`\kappa `$-Poincaré-induced CPT violation we have considered for the neutral-kaon system would also apply to the other analogous neutral-meson systems and, despite the lower precision reachable for $`\delta M_B`$ at beauty factories , sensitivity comparable to the one of KLOE could be achieved as a result of the boost with respect to the laboratory, which is a consequence of the asymmetric setup of these factories. At BaBar electrons with momenta $`9GeV`$ colliding with positrons with momenta $`3GeV`$ produce B’s with momenta $`2.8GeV`$ and this, for the expected sensitivity to $`\delta M_B510^{15}GeV`$, provides sensitivity to values of $`|\lambda |`$ around $`1.410^{31}m`$.
A better sensitivity is given by the CPLEAR experiment , which studied the neutral kaons produced in the reactions
$$p\overline{p}K^+\pi ^{}\overline{K}^0,p\overline{p}K^{}\pi ^+K^0,$$
(9)
with the strangeness of the neutral kaon being tagged as the opposite of the corresponding charged kaon. They find $`|\delta M|<7.510^{18}GeV`$ and this corresponds to the bound $`|\lambda |<1.210^{32}m`$, taking into account that the CPLEAR reactions happen at rest so that the neutral-kaon energy is around $`2M_{proton}/3`$, which corresponds to a momentum of order 0.4 GeV. By neglecting $`\delta \mathrm{\Gamma }`$ and considering only the contribution of the $`\pi \pi `$ decay channel to the Bell-Steinberger unitarity relationship
$$\left[\frac{1}{2}(\mathrm{\Gamma }_{K_S}+\mathrm{\Gamma }_{K_L})+i(M_{K_L}M_{K_S})\right]<K_S|K_L>=\mathrm{\Sigma }_fA^{}(K_Sf)A(K_Lf)$$
(10)
CPLEAR obtains the bound $`|\delta M|<4.410^{19}GeV`$ and this corresponds to the bound $`|\lambda |<710^{34}m`$.
Very stringent bounds on $`\lambda `$ can be derived from the data of the NA31 and the E731 experiments by assuming the approximate equality
$$M_{\overline{K}^0}M_{K^0}2\frac{(M_{K_L}M_{K_S})|\eta |\left(\frac{2}{3}\varphi ^++\frac{1}{3}\varphi ^{00}\varphi _0\right)}{sin\varphi _0}$$
(11)
where the phases $`\varphi ^+`$ and $`\varphi ^{00}`$ are defined by
$$\frac{A(K_L\pi ^{+(0)}\pi ^{(0)})}{A(K_S\pi ^{+(0)}\pi ^{(0)})}=|\eta ^{+(00)}|e^{i\varphi ^{+(00)}},$$
(12)
and $`\varphi _0`$ is the superweak phase
$$\varphi _0=tg^1\frac{2(M_{K_L}M_{K_S})}{\mathrm{\Gamma }_{K_S}\mathrm{\Gamma }_{K_L}}=43.5^o\pm 0.1^o,$$
(13)
which also follows from unitarity by considering only the $`\pi \pi `$ decay channel on the right-hand side of Eq. (10) and neglecting any other source of CPT violation. The experimental results
$$\varphi ^+=46.9\pm 2.2^o,\varphi ^{00}=47.1\pm 2.8^o,$$
(14)
and
$$\varphi ^+=42.4\pm 1.4^o,\varphi ^{00}\varphi ^+=1.6\pm 1.2^o,$$
(15)
lead to the bounds $`|\delta M|<510^{18}GeV`$ and $`|\delta M|<2.510^{18}GeV`$ respectively. Taking into account that the kaon beam has an average momentum of $`100GeV/c`$ in NA31 and ranges from $`40GeV/c`$ to $`150GeV/c`$ in E731 (which we take to correspond to a reference momentum of $`80GeV/c`$) one obtains the bounds $`|\lambda |<210^{35}m`$ from NA31 data and
$$|\lambda |<1.210^{35}m$$
(16)
from E731.<sup>5</sup><sup>5</sup>5The fact that the sample of the NA31 and E731 data is rather homogenous (both measuring the phases $`\varphi ^+`$ and $`\varphi ^{00}`$ and using kaons with momenta of roughly the same order) can provide motivation for obtaining a limit on $`\lambda `$ by combining the results of the two experiments. By properly taking into account the slightly different kaon-momentum scales of the two experiments we find the combined limit $`|\lambda |<710^{36}m`$. The fact that the central values of the NA31 and of the E731 data would correspond to values of $`\lambda `$ with opposite signs is one of the factors that contribute to rendering the combined limit significantly (roughly a factor 2) more stringent than (16). This “conditional” bound on $`\lambda `$ (conditional in the sense that it relies on an heuristic analysis which might turn out to be unreliable when tested within a more rigorous study) obtained by analyzing the neutral-kaon system goes even beyond the Planck length and is significantly more stringent than the previous bound $`\lambda <510^{33}m`$, which was obtained by comparing data on the gamma rays we detect from distant astrophysical sources with the structure of the $`\kappa `$-Poincaré deformed dispersion relation. Our analysis therefore provides motivation for more rigorous analyses of the implications of $`\kappa `$-Poincaré for the neutral-kaon system; in fact, if our heuristic estimate was confirmed by such more rigorous studies one could conclude that precision measurements on the validity of CPT in the neutral-kaon system are a better probe of the $`\kappa `$-Poincaré deformation than the astrophysical observations previously considered in the literature.
This possible “competition” between astrophysical and neutral-kaon-related bounds on $`\lambda `$ is made possible by the fact that, as already emphasized in Ref. , the $`\kappa `$-Poincaré framework can lead to phenomenological models in which the magnitude of all new effects could be related directly and calculably to the single parameter $`\lambda `$, while in other quantum-gravity-motivated formalisms the evaluation of the magnitude of the effects directly from the original theory turns out to be too difficult and one can only make phenomenological models to parametrize the magnitude of the effects, with independent parametrizations for each of the effects.
Looking beyond our relation (5), which is only a conjecture in the framework of $`\kappa `$-Poincaré, and also in light of the fact that $`\kappa `$-Poincaré is of course only one specific candidate for quantum-gravity-deformed symmetries, it is worth observing that in any quantum-gravity theory predicting a non-vanishing $`\delta M/M`$ and predicting an effect that is linear in a characteristic length scale $`\lambda `$ (to be possibly identified with the Planck length) one would be able to write a relation of the type $`\delta M/M\lambda E^{}/\mathrm{}`$, where $`E^{}`$ carries dimensions of an energy (from now on we set $`c=1`$ to simplify formulas). Our conjecture within the $`\kappa `$-Poincaré framework corresponds to the case $`E^{}=\stackrel{}{p}^2/E`$, but in general $`E^{}`$ will be given by some combination of $`M`$, $`|\stackrel{}{p}|`$ and $`E`$. For the simple possibilities $`E^{}=E`$, $`E^{}=|\stackrel{}{p}|`$, $`E^{}=M`$, in addition to the possibility $`E^{}=\stackrel{}{p}^2/E`$ we already considered, we report in Table 1 the limits on $`\lambda `$ which can be obtained (using unitarity) from the quoted experiments. Notice that, apart from the case $`E^{}=M`$ (where the best bound is set by CPLEAR), E731 always sets the best bound on $`\lambda `$ and this bound is more than two orders of magnitude better than the astrophysical bound $`\lambda <510^{33}m`$ (the kaons of E731 and NA31 are relativistic and therefore do not distinguish between the scenarios $`E^{}=E`$, $`E^{}=|\stackrel{}{p}|`$ and $`E^{}=\stackrel{}{p}^2/E`$, as it is evidently also true for the gamma rays used for the astrophysical bound). Even the bounds set by CPLEAR are more stringent than the astrophysical bound.
In closing, we observe that the study of neutral kaons and other neutral mesons might also prove useful for the exploration of some of the open conceptual issues associated with deformations of spacetime symmetries, and in particular with $`\kappa `$-Poincaré. Most notably, above our estimates were obtained by considering the momentum of the particles in the laboratory frame of reference (which also made sense for the comparison with the astrophysical bounds where a corresponding assumption was made), but the frame dependence inherently associated with $`\kappa `$-Poincaré renders this choice quite significant. One could for example imagine that the “$`\kappa `$-Poincaré-preferred frame” would be set by the particular state of the quantum-gravity foam realized in the experiment. If effects of the type here considered were eventually detected it seems likely that the controlled environment of neutral-meson experiments might prove more useful than astrophysical experiments for the investigation of these delicate issues.
It is our pleasure to greatfully acknowledge conversations on these topics with Giuseppe Nardulli and Holger Nielsen. |
no-problem/0001/cond-mat0001252.html | ar5iv | text | # Direct Coulomb and Exchange Interaction in Artificial Atoms
## Abstract
We determine the contributions from the direct Coulomb and exchange interactions to the total interaction in semiconductor artificial atoms. We tune the relative strengths of the two interactions and measure them as a function of the number of confined electrons. We find that electrons tend to have parallel spins when they occupy nearly degenerate single-particle states. We use a magnetic field to adjust the single-particle state degeneracy, and find that the spin-configurations in an arbitrary magnetic field are well explained in terms of two-electron singlet and triplet states.
The addition of a single electron charge to a quantum box costs a certain energy, which is responsible for Coulomb blockade in electron transport . Also a change in spin is associated with a certain change in energy, e.g.exchange energy is gained when electrons are added with parallel spins as compared to anti-parallel spins. Depending on the system, a large total spin (ferromagnetic filling) or a minimum total spin value (anti-ferromagnetic filling) is favored. In semiconductor quantum dots alternate spin filling as well as spin-polarized filling have been reported. Here, we study vertical quantum dots which have well-defined single-particle states. When these states are separated by a large energy, $`\mathrm{\Delta }E`$, an anti-ferromagnetic filling is favored. For small $`\mathrm{\Delta }E`$, a ferromagnetic filling is observed, which is in line with Hund’s first rule from atomic physics. We use a magnetic field, $`B`$, to tune $`\mathrm{\Delta }E`$($`B`$) allowing us to alter the spin filling. We first discuss a simple model that describes filling of two single-particle states with two interacting electrons. Fig. 1(b) shows two, spin-degenerate single-particle states with energies $`E_a`$ and $`E_b`$ crossing each other at $`B=B_0`$. The ground state (GS) energy, $`U`$(1), for one electron occupying these states, equals $`E_a`$ for $`B`$ $`<`$ $`B_0`$ and $`E_b`$ for $`B>B_0`$ (thick line in Fig. 1(b)). For two electrons we can distinguish four possible configurations with either total spin $`S`$ $`=0`$ (spin-singlet) or $`S`$ = 1 (spin-triplet). (We neglect the Zeeman energy difference between $`S_z=1,0`$ and 1.) The corresponding energies, $`U_i`$(2, $`S`$) for $`i=1`$ to 4, are given by: $`U_1(2,0)=2E_a+C_{aa},U_2(2,0)=2E_b+C_{bb}`$, $`U_3(2,1)=E_a+E_b+C_{ab}|K_{ab}|`$, $`U_4(2,0)=E_a`$ $`+E_b+C_{ab}`$. Here, $`C_{ij}`$ $`(i,j=a,b)`$ is the direct Coulomb (DC) energy between two electrons occupying states with energies $`E_i`$ and $`E_j`$, and $`K_{ab}`$ is the exchange (EX) energy ($`K_{ab}`$ $`<0`$) between two electrons occupying $`E_a`$ and $`E_b`$ with parallel spins . The experiments below measure the electrochemical potential defined for a two electron system as $`\mu (2)U(2)U(1)`$. For each $`U_i(2)`$ we obtain the potentials: $`\mu _i(2)=U_i(2)E_a`$ for $`B`$ $`<`$ $`B_0`$ and $`\mu _i(2)=U_i(2)E_b`$for $`B`$ $`>B_0`$ (see Fig. 1(c)).
The GS has $`S=0`$ away from $`B_0`$. Near $`B_0`$, the lowest energy is $`\mu _3(2)`$, such that here $`S=1`$. The downward cusp in the thick line identifies this spin-triplet region. The transition in the GS from $`S=0`$ to 1 and $`S`$ $`=1`$ to 0, respectively, occurs when $`\mu 1=\mu 3`$ for $`B`$ $`<B_0`$ and when $`\mu _2=\mu _3`$ for $`B`$ $`>`$ $`B_0`$. We define two energies, $`\mathrm{\Delta }_1`$ and $`\mathrm{\Delta }_2`$, to characterize the size of the downward cusp in the GS at $`B`$ $`=B_0`$: $`\mathrm{\Delta }_1=\mu _1\mu _3=C_{aa}C_{ab}+|K_{ab}|`$, $`\mathrm{\Delta }_2=\mu _2\mu _3=C_{bb}C_{ab}+|K_{ab}|`$, $`\mathrm{\Delta }_1\mathrm{\Delta }_2=C_{aa}C_{bb}`$. Our semiconductor quantum dot (see Fig. 1(a)) has the shape of a two-dimensional disk . For detecting the GSs, we set the source-drain voltage, $`V`$, to a small value. The GSs and excited states (ESs) are both measured when $`V`$ is set to a value sufficiently greater than the excitation energy . For small $`V`$, a series of current peaks results from changing the number of electrons in the dot, $`N`$, one-by-one . The position of a current peak for the transition from $`N1`$ to $`N`$ measures the GS electrochemical potential $`\mu `$($`N`$). The sample is cooled down to about 100 mK. Fig. 2(a) shows the evolution of current peaks with magnetic field for $`N=7`$ to 16.
Features associated with a parabolic confining potential are all observed such as a shell structure . The large spacing for $`N`$ = 12 at $`B=0`$ T can be seen in Fig. 2(a) (see double arrow) and marks the complete filling of the first three shells. The pairing between neighboring peaks indicates anti-parallel spin filling of a single orbital state by two electrons. Modifications to this pairing are observed for the peaks labeled by $`\mathrm{}`$ at 0 T, and in each of the dashed ovals connecting pairs of peaks at non-zero field. These are all signatures of Hund’s first rule; i.e. spin-polarized filling. Note that the Zeeman effect is negligible in this experiment . We show expansions of the evolution of the $`N=8`$ and $`N=24`$ peaks in Fig. 2(b). The downward cusps are clearly seen. The dashed lines form a parallelogram, from which we obtain parameters $`\mathrm{\Delta }_1`$ and $`\mathrm{\Delta }_2`$. To compare the two-electron model with larger electron numbers, we assume that other states are far away in energy so that they can be neglected. Then, the downward cusps should occur for higher $`even`$-electron numbers, whereas they should be absent for $`odd`$-electron numbers. This is clearly observed in the ovals in Fig. 2(a). For instance, the $`B`$-field dependence of the 9th peak compares well to the thick line in Fig. 1(b) and the $`B`$-field dependence of the 10th peak compares well to the thick line in Fig. 1(c). Other pairs of even and odd numbered peaks show the same behavior.
This justifies our assumption so that we can simplify the many-electron system to just one or two electrons. More detailed agreement is obtained by measuring the excitation spectrum . Fig. 3 shows a $`dI/dV_g`$ plot, taken for $`V=2`$ mV. This larger voltage opens a sufficiently wide transport window between the Fermi levels of the source and drain, that both the GS and first few ESs can be detected. The GS and ESs for $`N=7`$ to 9 can be assigned from the magnetic field dependence of the dark blue lines. Solid red lines highlight the GSs whereas the ESs are indicated by dashed red lines. TThe set of GS and ES lines for $`N`$ $`=7`$ shows a single crossing similar to that in Fig. 1(b). The spectrum for $`N`$ $`=8`$ compares well to Fig. 1(c) and we can clearly distinguish the parallelogram formed by the GS and first ES. The downward cusp in the GS for $`N`$ $`=8`$ (labeled $`\mathrm{}`$) is at a slightly higher $`B`$-field than the upward cusp in the first ES (labeled $`\mathrm{}`$). This asymmetry implies that $`\mathrm{\Delta }_1>`$ $`\mathrm{\Delta }_2`$, i.e. $`Caa>Cbb`$. The same type of asymmetry is always observed along the dashed line in Fig. 2(a), implying that $`Caa>C_{bb}`$ for all $`N`$. Note that the GS for $`N`$ $`=9`$ shows an upward cusp (labeled by $`\mathrm{}`$) quite similar in form to the first ES in the spectrum for $`N`$ $`=8`$. This implies that the filling of the ninth electron is closely linked to the configuration of the $`N`$ $`=8`$ first ES . As illustrated in Fig. 2(b), we can derive the experimental values for $`\mathrm{\Delta }_1`$ and $`\mathrm{\Delta }_2`$ for different $`N`$. These values are plotted in Fig. 4. We find that $`\mathrm{\Delta }_1`$ is larger than $`\mathrm{\Delta }_2`$ for all $`N`$, again implying that $`Caa>Cbb`$. As $`N`$ increases from 6 to 12, $`\mathrm{\Delta }_1`$ first increases and then slowly decreases, whilst $`\mathrm{\Delta }_2`$ slightly decreases.
To calculate DC and EX energies we now need to specify the confining potential of the disk. In earlier work it was shown that the lateral confinement is well described by a parabolic potential with cylindrical symmetry . The eigenfunctions with eigenenergies, $`E_{n,l}`$, in this potential are known as Fock-Darwin (FD) states :
$$E_{n,l}=\frac{l}{2}\mathrm{}\omega _c+(n+\frac{1}{2}+\frac{1}{2}|l|)\mathrm{}\sqrt{4\omega _0^2+\omega _c^2}$$
(1)
where $`n=0,1,2,\mathrm{}`$ is the radial quantum number and $`l=0,\pm 1,\pm 2,\mathrm{}`$ is the quantum number for angular momentum. $`\mathrm{}\omega _0`$ is the lateral confining energy and $`\mathrm{}\omega _c`$ $`=eB/m^{}`$ is the cyclotron energy. Each FD-state is spin degenerate. At $`B=0`$ T the FD-spectrum has sets of states with increasing degeneracy (see Fig. 2(c)). This degeneracy is lifted on increasing $`B`$, but as $`B`$ is increased further new crossings can occur. The last crossing is always a crossing between just two FD-states. The up-going state is always $`(n,l)=(0,1)`$, whereas the down-going state, $`(0,l>1)`$, has an increasing angular momentum for states with increasing energy. (The relation with Fig. 1 is: $`E_a`$ $`=E_{0,1}`$ and $`E_b=E_{0,l>1}`$.) Note that the last crossings also correspond to the dashed line in Fig. 2(a). From the electron distributions of the FD-states we calculate the DC and EX energies for two electrons occupying two degenerate states. We take $`\mathrm{}\omega _0`$ $`=2`$ meV as deduced from earlier experiments and obtain $`\mathrm{\Delta }_1`$ and $`\mathrm{\Delta }_2`$. The dashed curves in Fig. 4 show $`\mathrm{\Delta }_1`$ and $`\mathrm{\Delta }_2`$ when we neglect screening of the interactions within the dot by electrons in the leads and in the gate. In this case the Coulomb potential falls off as $`1/r`$, where $`r`$ is the distance between the electrons . For the solid curves we have approximated the screening effects by replacing the Coulomb potential by exp$`\{r/d\}/r`$. We have taken $`d`$ $`=10`$ nm which is roughly the thickness of the tunnel barriers. Fig. 4 shows that screening considerably reduces $`\mathrm{\Delta }_1`$ to values much closer to the experimental values. Screening also removes the minimum in $`\mathrm{\Delta }_2`$, which is also in better agreement with the experiment. Since the average radius of the wavefunctions increases with angular momentum, two electrons are closer together when they both occupy (0,-1) compared to when they both occupy ($`0,l=N/21`$) for even $`N`$ $`>4`$ (or $`l>1`$), so the DC interaction is stronger in the former. This explains our observation ($`\mathrm{\Delta }_1>`$ $`\mathrm{\Delta }_2`$) $`C_{aa}>C_{bb}`$ for all $`N`$. The overlap between $`different`$ wavefunctions, (0,-1) and ($`0,l=N/21`$), decreases for even $`N`$ $`>4`$ (or $`l>1`$). This results in a decrease in both $`C_{ab}`$ and $`|K_{ab}|`$ with $`N`$. It then follows that $`\mathrm{\Delta }_1`$ increases until it saturates at a value equal to $`C_{aa}`$. The gradual decrease of experimental $`\mathrm{\Delta }_1`$ for $`N`$ $`>12`$ is probably related to the decrease in the lateral confinement with $`N`$ and thus the decrease in $`C_{aa}`$. We finally discuss the interaction effects for the $`N`$ $`=4`$, 8 and 14 peaks near $`B`$ $`=0`$ T (the $`N=8`$ and 14 peaks are labeled n in Fig. 2(a)). These correspond to the GS electrochemical potentials for adding the second electron to the second, third and fourth shells, respectively. The inset to Fig. 5 demonstrates the resemblance to the model of Fig. 1(c) for $`N=8`$ near $`B`$ $`=0`$ T. Comparing these data to the FD-spectrum, we assign the states such that: $`E_a`$ $`=E_{0,2}`$ and $`E_b`$ $`=E_{0,2}`$. Likewise, for $`N`$ $`=4`$ we have $`E_a`$ $`=`$ $`E_{0,1}`$ and $`E_b`$ $`=`$ $`E_{0,1}`$ and for $`N`$ $`=`$ 14 we have $`E_a=E_{0,3}`$ and $`E_b=E_{0,3}`$ . Note that these states correspond to wavefunctions with a complete overlap. Also, for $`B`$ $`=B_0=0`$ T the two crossing states have the same orbital symmetry implying $`\mathrm{\Delta }_1=\mathrm{\Delta }_2=|K_{ab}|`$; i.e. only EX effects contribute to the downward cusp .
We derive $`|`$$`K_{ab}`$$`|`$as illustrated in the inset to Fig. 5. The obtained EX energy quickly becomes smaller for higher lying shells. For comparison we also show the calculated screened and unscreened values. The screened case provides the best quantitative agreement for our realistic choices of the confining energy and the screening distance. Our general model provides a clear identification of effects due to EX and DC interactions. More advanced calculations support our analyses . An important simplification is the reduction of a many-electron system to just two interacting electrons. The type of spin filling in many nearly degenerate levels near $`B`$ $`=0`$ T in larger electron boxes remains an interesting open issue. We thank G. Bauer, M. Danoesastro, M. Eto, R. van der Hage, T. Honda, J. Janssen, T. Oosterkamp, H. Tamura, and T. Uesugi for their help and useful discussions. S. T. and L.P.K. acknowledge financial support from the Specially Promoted Research, Grant-in-Aid for Scientific Research, from the Ministry of Education, Science and Culture in Japan, from the Dutch Organization FOM, from the NEDO program NTDP-98, and from the EU via a TMR network. |
no-problem/0001/hep-th0001016.html | ar5iv | text | # Acknowledgments
## Acknowledgments
This work has been supported by the European Union TMR programs FMRX-CT96-0012 Integrability, Non-perturbative Effects, and Symmetry in Quantum Field Theory and ERBFMRX-CT96-0090 Beyond the Standard model as well as by the Spanish grants AEN96-1655 and AEN96-1664. |
no-problem/0001/hep-ph0001078.html | ar5iv | text | # Drell-Yan proton-deuteron asymmetry and polarized light-antiquark distributions
## 1 Introduction
Nowadays, flavor asymmetry of light-antiquark distributions is an established fact in unpolarized distributions . At first, the unpolarized $`\overline{u}/\overline{d}`$ asymmetry was suggested by the NMC finding of the Gottfried-sum-rule violation. Then, NA51 and E866 collaborations investigated the $`\overline{u}/\overline{d}`$ ratio by measuring Drell-Yan proton-deuteron asymmetry. Their results clearly showed that the $`\overline{u}`$ and $`\overline{d}`$ distributions are different from each other. In particular, E866 data revealed detailed $`x`$ dependence of the $`\overline{d}/\overline{u}`$ ratio. Furthermore, semi-inclusive deep inelastic scattering data which were measured by HERMES also showed the flavor asymmetry.
On the other hand, the flavor asymmetry in polarized distributions is totally unknown at this stage although there are some model predictions. Our research purpose is to study the flavor asymmetry in more detail. At this stage, the longitudinally-polarized parton distributions are mainly investigated by measuring the spin structure function $`g_1`$. However, $`g_1`$ data are not enough to find the flavor asymmetry. It may be possible to get the information about the flavor asymmetry from semi-inclusive deep inelastic scattering data which were measured by SMC and HERMES. However, the precision of the present data is not enough to determine whether there exists the flavor asymmetry although the results show the tendency that the $`\mathrm{\Delta }\overline{u}\mathrm{\Delta }\overline{d}`$ becomes positive. Situation is more serious for another polarized distributions, namely transversity distributions, since they cannot be measured by inclusive deep inelastic scattering and $`W`$-production processes because of the chiral-odd property. The $`W`$ production is expected to provide important information about the flavor asymmetry in the unpolarized and longitudinally-polarized distributions. In the light of the present situation, we should study other independent processes to get the detailed information on the flavor asymmetry and to determine the major mechanism for creating the asymmetry. In this study, we investigate the method in which we use the polarized proton-deuteron (pd) Drell-Yan process with pp Drell-Yan .
In Sec. 2, model studies are explained on the light-antiquark flavor asymmetry in longitudinally-polarized and transversity distributions $`\mathrm{\Delta }_{(T)}\overline{u}\mathrm{\Delta }_{(T)}\overline{d}`$. Here, $`\mathrm{\Delta }_{(T)}`$ denotes $`\mathrm{\Delta }`$ or $`\mathrm{\Delta }_T`$ for the longitudinal-polarized or transversity distribution, respectively. Then, we discuss the relation between the polarized Drell-Yan proton-deuteron (p-d) asymmetry which is defined by the ratio of the polarized pd Drell-Yan cross section to the proton-proton (pp) one $`\mathrm{\Delta }_{(T)}\sigma _{pd}/2\mathrm{\Delta }_{(T)}\sigma _{pp}`$ and the polarized flavor asymmetry in Sec. 3. Furthermore, we show the numerical results for the flavor asymmetry effects on the Drell-Yan p-d asymmetry in Sec. 4. Conclusions are given in Sec. 5.
## 2 Theoretical predictions on the polarized flavor asymmetry
In this section, we briefly introduce the present status of the model studies on the flavor asymmetry $`\mathrm{\Delta }_{(T)}\overline{u}\mathrm{\Delta }_{(T)}\overline{d}`$. First, as one of the origins of the flavor asymmetry, there is a perturbative-QCD contribution. Next-to-leading-order (NLO) $`Q^2`$ evolution gives rise to the difference between the $`\overline{u}`$ and $`\overline{d}`$ distributions even if the flavor-symmetric distributions are used at initial $`Q^2`$. However, its effect is not so large as to reproduce the measured asymmetry in the unpolarized distributions if the $`Q^2`$ evolution is calculated in the perturbative $`Q^2`$ range. Therefore, we expect that dominant effects come from non-perturbative mechanisms.
We have been studying the flavor asymmetry in the polarized distributions by using typical models for the unpolarized $`\overline{u}/\overline{d}`$ asymmetry, and we have been also investigating its effect on the Drell-Yan spin asymmetry $`A_{TT}`$ . One of the typical models is a meson-cloud model. In this model, we calculate the meson-nucleon-baryon (MNB) process in which the initial nucleon splits into a virtual meson and a baryon, then the virtual photon from lepton interacts with this meson. Since the lightest vector meson is $`\rho `$ meson, we investigate a $`\rho `$-meson contribution to the polarized flavor asymmetry. We take into account $`\mathrm{\Delta }`$, in addition to the proton, as a final state baryon, and all the possible $`\rho `$NB processes are considered. Among them, the dominant contribution comes from the process with $`\rho ^+`$ meson. Because the $`\rho ^+`$ has a valence $`\overline{d}`$ quark, this mechanism contributes to the $`\overline{d}`$ excess over $`\overline{u}`$. Note that the $`\rho `$-meson contributions to the flavor asymmetry are also studied in Ref. 5.
Another typical model in the unpolarized case is Pauli-exclusion-principle model. Although this mechanism does not seem to explain the whole $`\overline{u}/\overline{d}`$ asymmetry, it is still worth while discussing the polarized asymmetry. This model for the polarized case has already studied in Ref. 6. According to the SU(6) quark model, each quark-state probabilities in the spin-up proton are given by $`u^{}=5/3`$, $`u^{}=1/3`$, $`d^{}=1/3`$, and $`d^{}=2/3`$, respectively. Since the probability of $`u^{}`$ ($`d^{}`$) is much larger than that of $`u^{}`$ ($`d^{}`$), it is more difficult to create $`u^{}`$ ($`d^{}`$) sea than $`u^{}`$ ($`d^{}`$) sea because of the Pauli-exclusion principle. Then, if we assume that the magnitude of the exclusion effect is the same as the one in the unpolarized case, $`(u_s^{}u_s^{})/(u_v^{}u_v^{})=(d_su_s)/(u_vd_v)`$ and a similar equation for $`d_s^{}d_s^{}`$, the magnitude of $`\mathrm{\Delta }\overline{u}`$ and $`\mathrm{\Delta }\overline{d}`$ become $``$0.13 and $`+`$0.05, respectively. In this way, we find the $`\mathrm{\Delta }\overline{u}/\mathrm{\Delta }\overline{d}`$ flavor asymmetry from this mechanism.
We numerically calculate the flavor-asymmetric distribution $`\mathrm{\Delta }_{(T)}\overline{u}\mathrm{\Delta }_{(T)}\overline{d}`$ by using the above model results and actual initial distributions. As a result, we find that both model predictions have similar tendency that the $`\mathrm{\Delta }_{(T)}\overline{u}\mathrm{\Delta }_{(T)}\overline{d}`$ becomes negative. Furthermore, the meson contribution seems to be smaller than that of the exclusion model. In addition to these mechanisms, there are also some model studies on the polarized flavor asymmetry . To distinguish these mechanisms, we need detailed experimental information on the $`\mathrm{\Delta }_{(T)}\overline{u}`$ and $`\mathrm{\Delta }_{(T)}\overline{d}`$ asymmetry. This is the reason for investigating the possibility of finding the flavor asymmetry by the polarized pd Drell-Yan data.
## 3 Polarized proton-deuteron Drell-Yan process and flavor asymmetry
Recently, a formalism of the polarized pd Drell-Yan process was completed in Refs. 8 and 9. They showed that there are additional spin asymmetries compared with a spin-1/2 hadron reaction. These new spin asymmetries are related to new spin structure in a spin-1 hadron and one of major purposes to investigate the polarized pd Drell-Yan process is to study this new spin structure.
Here, we briefly comment on this topic although it is not the major purpose in this work. The spin structure of the spin-1/2 hadron is relatively well investigated by measuring the structure function $`g_1`$. On the other hand, there is no experimental study for the spin structure of the spin-1 hadron. In fact, we know that the new spin structure, namely, the polarized tensor distributions can be investigated by measuring the tensor structure function $`b_1`$ in the deep inelastic scattering with unpolarized lepton off the polarized deuteron. The $`b_1`$ has not be measured at this stage, but it is expected to be measured at HERMES in future. In the pd Drell-Yan, on the other hand, parton-model analysis suggests that only three spin asymmetries $`A_{LL}`$, $`A_{TT}`$, and $`A_{UQ_0}`$ remain finite and other asymmetries vanish in the leading-twist level. In these asymmetries, the $`A_{UQ_0}`$ is a new one in the spin-1 hadron. In the parton model, this spin asymmetry is expressed by
$$A_{UQ_0}=\frac{_ae_a^2\left[q_a(x_1)\delta \overline{q}_a^d(x_2)+\overline{q}_a(x_1)\delta q_a^d(x_2)\right]}{_ae_a^2\left[q_a(x_1)\overline{q}_a^d(x_2)+\overline{q}_a(x_1)q_a^d(x_2)\right]},$$
(3.1)
where, $`\delta q_a^d`$ and $`\delta \overline{q}_a^d`$ represent the quark and antiquark tensor distributions in the deuteron. The subscript $`a`$ represents quark flavor, and $`e_a`$ is the corresponding quark charge. Therefore, we can study the tensor distributions also by measuring this spin asymmetry. In particular, $`A_{UQ_0}`$ has an advantage to investigate the antiquark tensor distributions in comparison with the deep inelastic scattering. This topic is very interesting but we do not discuss in this paper. The details are discussed in Refs. 8 and 9, so that the interested reader may read these papers.
For investigating the flavor asymmetry, we use the results for the spin asymmetries $`A_{LL}`$ and $`A_{TT}`$. Because of the existence of the tensor distribution, it was not clear whether the polarized pd Drell-Yan cross sections are expressed by the same forms as the pp ones. References 8 and 9 revealed this point. From their analysis, the difference between the longitudinally-polarized pd Drell-Yan cross sections is given by
$`\mathrm{\Delta }\sigma _{pd}`$ $`=`$ $`\sigma (_L,1_L)\sigma (_L,+1_L)`$ (3.2)
$``$ $`{\displaystyle \underset{a}{}}e_a^2\left[\mathrm{\Delta }q_a(x_1)\mathrm{\Delta }\overline{q}_a^d(x_2)+\mathrm{\Delta }\overline{q}_a(x_1)\mathrm{\Delta }q_a^d(x_2)\right],`$
where the $`_L`$, $`+1_L`$, and $`1_L`$ represent the longitudinal polarization and $`\sigma (pol_p,pol_d)`$ represents the cross section with the proton and deuteron polarizations, $`pol_p`$ and $`pol_d`$. The $`\mathrm{\Delta }q_a^d`$ and $`\mathrm{\Delta }\overline{q}_a^d`$ are the longitudinally-polarized quark and antiquark distributions in the deuteron. The momentum fractions are given by $`x_1=\sqrt{\tau }e^{+y}`$ and $`x_2=\sqrt{\tau }e^y`$ in the case of small $`P_T`$. Here, the $`\tau `$ is defined by $`\tau =M_{\mu \mu }^2/s`$ with dimuon mass $`M_{\mu \mu }`$ and the dimuon rapidity is given by $`y=(1/2)\mathrm{ln}[(E^{\mu \mu }+P_L^{\mu \mu })/(E^{\mu \mu }P_L^{\mu \mu })]`$. In the same way, the transversely-polarized cross-section difference is given by
$`\mathrm{\Delta }_T\sigma _{pd}`$ $`=`$ $`\sigma (\varphi _p=0,\varphi _d=0)\sigma (\varphi _p=0,\varphi _d=\pi )`$ (3.3)
$``$ $`{\displaystyle \underset{a}{}}e_a^2\left[\mathrm{\Delta }_Tq_a(x_1)\mathrm{\Delta }_T\overline{q}_a^d(x_2)+\mathrm{\Delta }_T\overline{q}_a(x_1)\mathrm{\Delta }_Tq_a^d(x_2)\right],`$
where the $`\varphi `$ is the azimuthal angle of a polarization vector. The $`\mathrm{\Delta }_Tq`$ and $`\mathrm{\Delta }_T\overline{q}`$ are quark and antiquark transversity distributions.
The pp Drell-Yan cross sections are given simply by replacing the distributions in the deuteron in Eqs. (3.2) and (3.3) by the ones in the proton. We use these equations for investigating the flavor asymmetry in the polarized distributions. To study the flavor asymmetry, we define the Drell-Yan proton-deuteron (p-d) asymmetry $`R_{pd}`$ by
$`R_{pd}`$ $``$ $`{\displaystyle \frac{\mathrm{\Delta }_{(T)}\sigma _{pd}}{2\mathrm{\Delta }_{(T)}\sigma _{pp}}}`$ (3.4)
$`=`$ $`{\displaystyle \frac{_ae_a^2\left[\mathrm{\Delta }_{(T)}q_a(x_1)\mathrm{\Delta }_{(T)}\overline{q}_a^d(x_2)+\mathrm{\Delta }_{(T)}\overline{q}_a(x_1)\mathrm{\Delta }_{(T)}q_a^d(x_2)\right]}{2_ae_a^2\left[\mathrm{\Delta }_{(T)}q_a(x_1)\mathrm{\Delta }_{(T)}\overline{q}_a(x_2)+\mathrm{\Delta }_{(T)}\overline{q}_a(x_1)\mathrm{\Delta }_{(T)}q_a(x_2)\right]}}.`$
First, we show the behavior of $`R_{pd}`$ in the large $`x_F`$ ($`=x_1x_2`$) limit. Because sea-quark distributions in the proton become smaller than other distributions in this limit, the proton sea-quark terms in the numerator and denominator of Eq. (3.4) can be ignored. In our analysis, we neglect the nuclear effects in the deuteron and assume the isospin symmetry. Then, the distributions in the deuteron can be written in terms of the distributions in the proton as
$$\begin{array}{ccc}\mathrm{\Delta }_{(T)}u^d=\mathrm{\Delta }_{(T)}u+\mathrm{\Delta }_{(T)}d,& \mathrm{\Delta }_{(T)}d^d=\mathrm{\Delta }_{(T)}d+\mathrm{\Delta }_{(T)}u,& \mathrm{\Delta }_{(T)}s^d=2\mathrm{\Delta }_{(T)}s,\\ \mathrm{\Delta }_{(T)}\overline{u}^d=\mathrm{\Delta }_{(T)}\overline{u}+\mathrm{\Delta }_{(T)}\overline{d},& \mathrm{\Delta }_{(T)}\overline{d}^d=\mathrm{\Delta }_{(T)}\overline{d}+\mathrm{\Delta }_{(T)}\overline{u},& \mathrm{\Delta }_{(T)}\overline{s}^d=2\mathrm{\Delta }_{(T)}\overline{s}.\end{array}$$
(3.5)
However, for a precise comparison with future experimental data, the nuclear corrections should be properly included. Using these relations, $`R_{pd}`$ becomes
$$R_{pd}(x_F1)=1\frac{[\mathrm{\hspace{0.17em}4}\mathrm{\Delta }_{(T)}u_v(x_1)\mathrm{\Delta }_{(T)}d_v(x_1)][\mathrm{\Delta }_{(T)}\overline{u}(x_2)\mathrm{\Delta }_{(T)}\overline{d}(x_2)]}{8\mathrm{\Delta }_{(T)}u_v(x_1)\mathrm{\Delta }_{(T)}\overline{u}(x_2)+2\mathrm{\Delta }_{(T)}d_v(x_1)\mathrm{\Delta }_{(T)}\overline{d}(x_2)},$$
(3.6)
where $`x_11`$ and $`x_20`$. Because of the $`\mathrm{\Delta }_{(T)}\overline{u}\mathrm{\Delta }_{(T)}\overline{d}`$ factor, the ratio $`R_{pd}`$ simply becomes one if the distribution $`\mathrm{\Delta }_{(T)}\overline{u}`$ is equal to $`\mathrm{\Delta }_{(T)}\overline{d}`$. If we assume that the valence-quark distributions satisfy $`\mathrm{\Delta }_{(T)}u_v(x1)\mathrm{\Delta }_{(T)}d_v(x1)`$, Eq. (3.6) can be written in a more simple form as
$`R_{pd}(x_F1)`$ $`=`$ $`1\left[{\displaystyle \frac{\mathrm{\Delta }_{(T)}\overline{u}(x_2)\mathrm{\Delta }_{(T)}\overline{d}(x_2)}{2\mathrm{\Delta }_{(T)}\overline{u}(x_2)}}\right]_{x_20}`$ (3.7)
$`=`$ $`{\displaystyle \frac{1}{2}}\left[\mathrm{\hspace{0.17em}1}+{\displaystyle \frac{\mathrm{\Delta }_{(T)}\overline{d}(x_2)}{\mathrm{\Delta }_{(T)}\overline{u}(x_2)}}\right]_{x_20}.`$
From this equation, it is clear that $`R_{pd}`$ becomes larger (smaller) than one if the $`\mathrm{\Delta }_{(T)}\overline{u}`$ distribution is negative as suggested by recent parametrizations and if the $`\mathrm{\Delta }_{(T)}\overline{u}`$ distribution is larger (smaller) than the $`\mathrm{\Delta }_{(T)}\overline{d}`$.
Next, we discuss the behavior of $`R_{pd}`$ in another limit, namely $`x_F1`$. In this limit, the ratio $`R_{pd}`$ becomes
$$R_{pd}(x_F1)=\frac{[\mathrm{\hspace{0.17em}4}\mathrm{\Delta }_{(T)}\overline{u}(x_1)+\mathrm{\Delta }_{(T)}\overline{d}(x_1)][\mathrm{\Delta }_{(T)}u_v(x_2)+\mathrm{\Delta }_{(T)}d_v(x_2)]}{8\mathrm{\Delta }_{(T)}\overline{u}(x_1)\mathrm{\Delta }_{(T)}u_v(x_2)+2\mathrm{\Delta }_{(T)}\overline{d}(x_1)\mathrm{\Delta }_{(T)}d_v(x_2)},$$
(3.8)
where $`x_10`$ and $`x_21`$. If we assume $`\mathrm{\Delta }_{(T)}u_v(x1)\mathrm{\Delta }_{(T)}d_v(x1)`$, the ratio becomes
$$R_{pd}(x_F1)=\frac{1}{2}\left[\mathrm{\hspace{0.17em}1}+\frac{\mathrm{\Delta }_{(T)}\overline{d}(x_1)}{4\mathrm{\Delta }_{(T)}\overline{u}(x_1)}\right]_{x_10}.$$
(3.9)
In this equation, we find the extra factor 4 in comparison with the equation for $`x_F1`$ limit. Therefore, $`R_{pd}`$ in this limit is not as sensitive to the $`\mathrm{\Delta }_{(T)}\overline{u}/\mathrm{\Delta }_{(T)}\overline{d}`$ asymmetry as the one in the large-$`x_F`$ limit although we can investigate the flavor asymmetry also in this limit. From Eq. (3.9), if the $`\mathrm{\Delta }_{(T)}\overline{u}`$ distribution is equal to the $`\mathrm{\Delta }_{(T)}\overline{d}`$, $`R_{pd}`$ becomes $`5/8=0.625`$ and if the $`\mathrm{\Delta }_{(T)}\overline{u}`$ distribution is larger (smaller) than the $`\mathrm{\Delta }_{(T)}\overline{d}`$, the ratio becomes larger (smaller) than this value.
From these analyses, we find that we can investigate the flavor asymmetry by measuring the polarized pd Drell-Yan process and taking the pd and pp cross-section ratio. In particular, $`R_{pd}`$ in the large-$`x_F`$ and small-$`x_F`$ regions are useful to find a signature for the flavor asymmetry.
## 4 Numerical results
We numerically calculate the Drell-Yan p-d asymmetry $`R_{pd}`$ by using a recent parametrization. In this section, we show the numerical results and discuss the flavor asymmetry effect on the $`R_{pd}`$ . As a initial distributions, we use the 1999 version of the LSS (Leader-Sidorov-Stamenov) parametrization for the longitudinally-polarized distributions. The LSS99 distributions are given at $`Q^2`$ = 1 GeV<sup>2</sup> by assuming SU(3) flavor-symmetric sea. For the transversity distributions, we simply assume that they are the same as the longitudinal distributions at $`Q^2`$= 1 GeV<sup>2</sup> by considering the quark-model predictions. Furthermore, we take center-of-mass energy $`\sqrt{s}=50`$ GeV and dimuon mass $`M_{\mu \mu }=5`$ GeV. Although there are some model predictions for the flavor asymmetry in the polarized distributions as explained in Sec. 2, we simply take the $`\mathrm{\Delta }_{(T)}\overline{u}/\mathrm{\Delta }_{(T)}\overline{d}`$ ratio as
$$r_{\overline{q}}\frac{\mathrm{\Delta }_{(T)}\overline{u}}{\mathrm{\Delta }_{(T)}\overline{d}}=0.7,1.0,\mathrm{or}1.3,$$
(4.1)
at $`Q^2`$=1 GeV<sup>2</sup> in the following analysis. The initial distributions with these $`r_{\overline{q}}`$ are evolved to those at $`Q^2=M_{\mu \mu }^2`$ by leading-order (LO) evolution equations. We use the FORTRAN programs which are provided in Ref. 11 for calculating the $`Q^2`$ evolution of the longitudinally-polarized and transversity distributions. Then, the pd/pp Drell-Yan cross-section ratio $`R_{pd}`$ is calculated for each $`r_{\overline{q}}`$. The results are shown in Fig. 1.
The solid and dashed curves represent the longitudinal and transverse results, respectively. As clearly shown by this figure, the flavor-symmetric ($`r_{\overline{q}}`$ = 1.0) results become one in the large-$`x_F`$ limit and 0.625 in the small-$`x_F`$ limit as discussed in the previous section. The results with the flavor asymmetry deviate from the flavor-symmetric ones. In particular, the deviations are conspicuous in the large-$`x_F`$ region. From these results, we find that the $`R_{pd}`$ in the large-$`x_F`$ region is very useful for finding the flavor asymmetry in the polarized distributions. Furthermore, we also find that there is almost no difference between the longitudinal and transverse results in this kinematical range if the initial distributions are identical.
In Fig. 2, we show NLO evolution results. We evolve the same LSS99 distributions at $`Q^2`$ = 1 GeV<sup>2</sup> to those at $`Q^2=M_{\mu \mu }^2`$ by NLO evolution equations. The results are almost the same as the LO ones. However, there is slight deviation. For example, the ratio $`R_{pd}`$ in the large-$`x_F`$ region is slightly different from one although antiquark distributions are flavor symmetric at $`Q^2`$ = 1 GeV<sup>2</sup>. It is because the NLO evolution gives rise to the flavor asymmetry. Although such perturbative-QCD effect is not so large in this kinematical region as clearly shown by this figure, we should include the NLO contributions for a precise analysis.
Next, we discuss the dependence on the center-of-mass energy $`\sqrt{s}`$. The results for the longitudinal ratio at the RHIC energies $`\sqrt{s}`$ = 200 and 500 GeV are shown in Fig. 3. The solid, dashed, and dotted curves indicate the results at $`\sqrt{s}`$ = 50, 200, and 500 GeV, respectively. The calculated results are almost the same as those at $`\sqrt{s}`$ = 50 GeV in the large- and small-$`x_F`$ regions. However, the ratio in the intermediate-$`x_F`$ region has much dependence on the c.m. energy and becomes a steeper function of $`x_F`$ as $`\sqrt{s}`$ increases. From these results, we find that the $`R_{pd}`$ in the intermediate-$`x_F`$ region is sensitive to the details of the distributions.
Finally, we show the dependence on the used parametrizations. In our analysis, we use the LSS99 distributions. In order to show the parametrization dependence, we employ the GRSV96 and the Gehrmann-Stirling set A (GS-A) distributions. The ratios $`R_{pd}`$ calculated with these initial distributions are shown in Fig. 4. The solid, dashed, and dotted curves represent the results with LSS99, GRSV96, and GS-A parametrizations. As shown in this figure, there is not so much difference between the LSS99 and GRSV results. However, the GS-A results has very different behavior especially in the intermediate-$`x_F`$ region. Because the GS-A antiquark distribution is positive in the large-$`x`$ region but becomes negative in the small-$`x`$ region, while the LSS99 and GRSV ones are negative in the whole $`x`$, the denominator of the $`R_{pd}`$ becomes zero at certain $`x`$ and the ratio goes to infinity at such $`x`$ points. Therefore, the $`R_{pd}`$ in the intermediate-$`x_F`$ region is especially useful for determining the detailed $`x`$ dependence of the polarized antiquark distributions.
From these numerical analyses, we find that the $`R_{pd}`$ is very valuable for investigating the details of the antiquark distributions. At this stage, there is no proposal for the polarized pd Drell-Yan experiment. However, we think that there are possibilities at FNAL, HERA, and RHIC.
## 5 Conclusions
We have studied the Drell-Yan proton-deuteron asymmetry $`R_{pd}`$ which is defined by pd and pp cross-section ratio $`\mathrm{\Delta }_{(T)}\sigma _{pd}/2\mathrm{\Delta }_{(T)}\sigma _{pp}`$. Using the formalism for the polarized pd Drell-Yan process, we have shown that the $`R_{pd}`$ is very useful for finding the light-antiquark flavor asymmetry in the polarized distributions ($`\mathrm{\Delta }_{(T)}\overline{u}/\mathrm{\Delta }_{(T)}\overline{d}`$), especially in the large-$`x_F`$ region. We have also shown the dependence on the center-of-mass energy and the used parametrizations. As a result, we have found that the $`R_{pd}`$ in the intermediate-$`x_F`$ region is valuable for determining the detailed $`x`$ dependence of the polarized antiquark distributions. Our results are important particularly for the transversity distributions for which $`W`$ production does not provide information on the $`\mathrm{\Delta }_T\overline{u}/\mathrm{\Delta }_T\overline{d}`$ asymmetry because of their chiral-odd nature.
## Acknowledgements
This work was partly supported by the Grant-in-Aid for Scientific Research from the Japanese Ministry of Education, Science, and Culture. M.M. was supported by a JSPS Research Fellowship for Young Scientists. |
no-problem/0001/astro-ph0001329.html | ar5iv | text | # Iron Gradients in Cooling Flow Galaxies and Groups
## 1. Introduction
One important aspect of the problem of galaxy formation is to understand the connection between the history of star formation and the metal enrichment history of the extended hot gaseous component (e.g., Cavaliere, Giacconi, & Menci 2000). Recent studies have argued that most of the metals produced by star-forming galaxies at high redshifts exist in the hot gaseous halos of galaxies and proto-clusters (e.g., Pettini 1999, 2000). By assuming that the hot gaseous components of nearby rich clusters are a fair representation of the local universe as a whole, Renzini (1997, 2000) has argued that the cluster Fe abundances indicate a prompt initial enrichment of the early universe.
Hence, X-ray observations may be vital for obtaining a complete picture of the history of star formation of the universe In massive elliptical galaxies the metals injected by previous generations of stars do not escape the gravitational field, and thus the hot gas in these systems records the star formation history via the heating and enrichment of the hot gas by supernovae (e.g., Ciotti et al 1991; David, Forman, & Jones 1991; Loewenstein & Mathews 1991). These systems typically fall into the class of cooling flows (e.g., Fabian 1994).
Unfortunately, reliable constraints on the metal enrichment in such systems are lacking because the metal abundances in the hot gas in cooling flow galaxies and groups determined from previous X-ray studies have yielded conflicting results with most authors finding very sub-solar Fe abundances (see Buote 2000a and references therein). These low Fe abundances are generally less than the stellar values which implies little or no enrichment from Type Ia supernovae and thus an IMF much flatter than that of the Milky Way (e.g., Arimoto et al 1997; Renzini 1997; Wyse 1997). In contrast, in our recent analyses of the ASCA data accumulated within $`r3\mathrm{}`$-$`5\mathrm{}`$ of the brightest ellipticals and groups we find that the Fe abundances are approximately solar and that the previous inferences of very sub-solar Fe abundances using ASCA data are the result of a bias arising from assuming the gas to be isothermal in the presence of significant temperature gradients (Buote & Fabian 1998; Buote 1999, 2000a).
To investigate the possible role of Fe abundance gradients on these measurements we have re-examined the ROSAT PSPC data of the brightest galaxies and groups. The PSPC allows for spatially resolved spectral analysis on a half-arcminute scale, and interesting constraints on both the temperatures and Fe abundances are possible because the $`1`$ keV temperatures of these systems are well matched to the bandpass of the PSPC. Moreover, analysis of the PSPC data is much less subject to the “Fe Bias” (Buote, 2000a) which plagues analysis of the ASCA data (see §5.1).
We selected for re-analysis all of the galaxies and groups possessing published temperature profiles from ROSAT (Forman et al 1993; Ponman & Bertram 1993; David et al 1994; Trinchieri et al 1994; Rangarajan et al 1995; Kim & Fabbiano 1995; Irwin & Sarazin 1996; Jones et al 1997; Trinchieri, Fabbiano, & Kim 1997; Mulchaey & Zabludoff 1998) with the exception of the Pegasus I group (Trinchieri et al, 1997) because of the low S/N. All of these systems have rising temperature profiles and multitemperature ASCA spectra usually attributed to cooling flows.
All but one of these ROSAT studies determined the temperatures and Fe abundances by fitting a single temperature component to the spectrum in an annulus on the sky. However, each annulus on the sky contains the projection of emission from larger radii which can confuse interpretation because of the substantial temperature gradients. Consequently, we have developed our own deprojection code to obtain three-dimensional spectral parameters.
The paper is organized as follows. In §2 we present the observations and discuss the data reduction. The deprojection procedure is described in §3 and the results for the temperature and Fe abundance profiles are given in §4.2; results for the column densities appear in Buote (2000b, hereafter PAPER1; 2000c, hereafter PAPER3). Discussion of the results and our conclusions are presented in §5.
## 2. ROSAT Observations and Data Reduction
We obtained ROSAT PSPC data from the public data archive maintained by the High Energy Astrophysics Science Archive Research Center (HEASARC). The properties of the observations are listed in Table 1. Except where noted below, these data were reduced using the standard FTOOLS (v4.2) software according to the procedures described in the OGIP Memo OGIP/94-010 (“ROSAT data analysis using xselect and ftools”), the WWW pages of the ROSAT Guest Observer Facility (GOF) (see http://heasarc.gsfc.nasa.gov/docs/rosat), and Plucinsky et al (1993).
The events files of each observation were cleaned of after-pulse signals by removing all events following within 0.35 ms of a precursor. To minimize the particle background contribution only events with Master Veto Rate less than 170 ct s<sup>-1</sup> were selected (Plucinsky et al, 1993). We corrected the Pulse Invariant (PI) bins of each data set for spatial and long-term temporal gain variations using the most up-to-date calibration files. For the ftool pcecor, which corrects for the variation in the linearity of the PSPC response, we used the in-flight calibration data for the correction.
From visual inspection of the light curve of an observation we identified and removed time intervals of significant enhancements in the count rate in order to obtain a flat distribution of count rate versus time. Such short-term enhancements are typically the result of scattered light from the Sun, auroral X-rays, and enhanced charged particle precipitation (Snowden et al, 1994). The raw and filtered exposure times are listed in Table 1.
The particle background spectrum for an observation was obtained by following the instructions in Plucinsky et al (1993). We developed our own software to perform this task because we identified serious errors in the ftools implementation of the Plucinsky et al procedure (i.e., program pcparpha). For our data sets the particle background rate is alway much less than that of the diffuse background.
For each observation we obtained a background spectrum from source-free regions far away from the center of the field (typically distances of $`45\mathrm{}`$-$`50\mathrm{}`$). By using a local background estimate residual contamination from solar X-rays and any other long-term background enhancements (Snowden et al, 1994) are fully accounted for in the ensuing spectral analysis. However, the extended emission of the galaxies and groups do contribute at some level to the flux even at these large radii.
To assess the contribution of galaxy and group emission within the background regions we fit the background spectrum with a model after subtracting the particle background. We represent the cosmic X-ray background by a power law and the Galactic emission by two thermal components following Chen et al (1997). To account for emission from the galaxy or group we include another thermal component with variable temperature. Each component is modified by the Galactic hydrogen column listed in Table 1.
This composite model provides a good fit to the background spectrum for each system. In every case the additional thermal component is required with temperatures ranging from 0.3-1.6 keV. In 8 out of 10 cases this additional component contributes significantly only to energies above $`0.5`$ keV which is consistent with emission from the galaxy or group in question. For these systems we subtract this temperature component from the background when performing spectral analysis (see §4.1.4), and the background rates listed in Table 1 reflect this subtraction. The effect of excluding this extra component in the background on the source spectral parameters is only significant for a few systems, and is noticeable only in the outermost one or two radial bins (see §4.3).
For NGC 5044 and 5846 we obtained $`T0.3`$ keV for the extra thermal component and found that it contributed very significantly to the emission in both the soft and hard energy channels. In these cases the extra component probably represents a combination of emission from long-term enhancements in the background (Snowden et al, 1994) with any residual emission from the groups. As a result, we did not subtract out the extra thermal component from the background of these systems.
For NGC 4472 we confirm the finding by Forman et al (1993) that a background spectrum taken from regions $`40\mathrm{}`$-$`50\mathrm{}`$ from the field center tends to over-subtract energies below $`0.3`$ keV probably because of contamination from unresolved point sources. To compensate for this we took the background from regions just inside the inner ring of the PSPC where the PSF is much smaller which allows us to avoid point sources more effectively. We confine the region to the north of the center of NGC 4472 where the galaxy emission is lower (Irwin & Sarazin, 1996).
## 3. Deprojection Method
The method we use to deproject the X-ray data is the well-established technique pioneered by Fabian et al (1981). This method is non-parametric in that no functional form is assumed for the spatial distribution of the X-ray emission or for any of the associated spectral quantities (temperature, abundances etc.). After first assuming a specific geometry one begins by determining the emission in the bounding annulus and then works inwards by subtracting off the contributions from the outer annuli. For the case of spherical symmetry the volume emission density is related to the surface brightness according to the simple geometric formula given by Kriss, Cioffi, & Canizares (1983).
The assumption of hydrostatic equilibrium is usually incorporated into this deprojection procedure in order to obtain the mass and mass deposition rate as well as the spectral parameters (Fabian et al 1981; Sarazin 1986; Arnaud 1988). In the interest of generality we do not make this additional assumption. Hence, the principal assumption in our version of the deprojection procedure is that of spherical symmetry which means that our derived spectral parameters should be considered spherically averaged quantities for systems that have significant ellipticity.
A thorough, up-to-date discussion of this deprojection method is given by McLaughlin (1999), and we refer the reader to that paper for the relevant equations. Since McLaughlin is interested in the globular cluster distribution in M87 he formulates the deprojection algorithm in terms of the number of globular clusters, $`𝒩(R_{i1},R_i)`$, located between radii $`R_{i1}`$ and $`R_i`$ on the sky. To express his equations in terms of quantities relevant for X-ray analysis we simply associate $`𝒩(R_{i1},R_i)`$ with the X-ray flux, $`F_\mathrm{x}(R_{i1},R_i)`$, in erg cm<sup>-2</sup> s<sup>-1</sup>.
We have developed our own code to implement the deprojection algorithm and have verified it using synthetic XMM data obtained by using the quicksim software (Snowden, 1999). Some aspects of the deprojection analysis require special mention which we now address.
### 3.1. Edge Effect
The deprojection algorithm assumes there is no source emission outside of the bounding annulus, $`R_m`$. This is generally not the case, and if the exterior emission is not accounted for then the volume emission density of the outermost annuli will be overestimated. Nulsen & Böhringer (1995) provide an analytic correction factor for the flux of the bounding annulus in the limit of annuli with zero width. Since, however, the width of the bounding annulus tends to be large because of decreasing S/N with increasing radius, and since any exterior emission also projects into interior annuli, we follow McLaughlin (1999) and compute a separate correction factor, $`f(R_{i1},R_i)`$, for each annulus arising from emission exterior to $`R_m`$.
The emissivity of this exterior emission projected into an annulus $`(R_{i1},R_i)`$ is taken to be proportional to $`f(R_{i1},R_i)F_\mathrm{x}(R_{m1},R_m)`$; i.e. the spectral shape of the emission exterior to $`R_m`$ is assumed to be the same as that for the outermost annulus. Moreover, to compute $`f(R_{i1},R_i)`$ a model for the spatial profile for the exterior emission is required. We assume that the volume X-ray emission density is a power law outside of $`R_m`$ which is a good approximation for galaxies and groups since the X-ray emission at large radii is generally well described by a $`\beta `$ model $`(r^{6\beta })`$ with $`\beta 0.5`$-$`0.7`$ (e.g., Mulchaey & Zabludoff 1998). Our results are not very sensitive to these assumptions because our chosen bounding annuli are quite wide so that $`f1`$ for $`\beta `$ over this range.
The formulae for $`f(R_{i1},R_i)`$ for the cases $`\beta =1/2`$ and $`\beta =2/3`$ are given by equations A7 and A8 in McLaughlin (1999). All results presented in §4 are for the case $`\beta =2/3`$ since, as mentioned above, the results are quite insensitive to this choice. Both the numerator and denominator of McLaughlin’s equation A8 evaluate to zero for $`R_{i1}=0`$ which is problematical for numerical computations. Fortunately, the limit is well-behaved, and in Appendix A we give the limiting equation.
### 3.2. Error Estimation
The deprojected X-ray emission in any annulus interior to the bounding annulus depends on the previous results obtained for adjacent exterior annuli. Since the deprojection procedure introduces correlations between annuli it is natural to employ Monte Carlo simulations to estimate the statistical uncertainties on the spectral parameters. Hence, after we have obtained best-fitting models for all desired annuli on the sky, we simulate a new data set for each annulus by using these models as templates. Then we deproject these simulated data and obtain parameters in the same manner as done for the actual data.
For each observation we determine the errors on each parameter from 100 Monte Carlo simulations. In most cases we found that the simulated parameters are reasonably symmetrically distributed about a central peak. Consequently, we define confidence limits for a parameter with respect to the median value from the 100 simulations. For example, after sorting the 100 values for a given parameter into ascending order we define the 68% confidence limits to be given by the 16th and 84th values; the 90% limits are given by the 5th and 95th values, etc.
This scheme for defining confidence limits has its limitations. In some cases when the parameter has a large error the simulated distribution is very flat so that the median is not obviously the best reference point from which to define confidence limits. In such instances the 68% confidence limits defined as above do not always enclose the best-fitting value. However, we do find that in all cases the best-fitting values are enclosed with the $`90\%`$ confidence limits indicating that our definitions are not very unreasonable, and thus we apply the above definitions of confidence levels in every case for consistency and convenience.
### 3.3. Radial Parameter Fluctuations
It is well known from optical studies that the deprojection of the luminosity distribution of an elliptical galaxy yields a jagged profile where the departures from a smooth profile are of the same magnitude or larger than the statistical errors on the photometry (e.g., Binney, Davies, & Illingworth 1990). A variety of factors contribute to the high-frequency noise responsible for the jagged profile such as the incomplete removal of point sources. An analogous situation occurs for the deprojection of X-ray data.
In actuality the situation is worse for the X-ray case because one desires the profiles of several spectral parameters in addition to the luminosity, but the statistical noise is much greater while the ability to remove contaminating sources is usually much worse than in the optical. As a result, the radial fluctuations in parameter values can sometimes be sufficiently large so that the values get sent off to obscure regions of parameter space never to return. Such fluctuations do not agree with the parameters obtained from fitting the X-ray spectra without deprojection.
Our preferred means to eliminate such fluctuations is to increase the S/N by increasing the widths of the annuli. This method is very effective and has the great advantage that no theoretical prejudice is forced onto the data (other than the desire for the 2D and 3D profiles to be qualitatively similar). Although doing this removes most of the serious parameter fluctuations in our ROSAT observations, in many cases there is still one annulus with divergent parameter values.
Hence, to keep any remaining fluctuations in check we smooth the derived parameter distributions in a manner related to linear regularization (e.g., Press et al 1992). The problem with linear regularization is that one restricts the point-to-point variation in the parameters with a pre-conceived model. For example, Finoguenov & Ponman (1999) assume the smooth profiles of the temperature and abundances to be linear functions of logarithmic radius. (The amount of smoothing applied also depends on an additional weighting factor – see Finoguenov & Ponman). However, as we show in §5.2 these choices usually lead to very biased parameter profiles.
Our method of regularizing the parameters is not standard since we do not add an extra term to the $`\chi ^2`$ equation because it cannot be done in the current version of xspec (Arnaud, 1996). Instead we perform an ex post facto regularization by simply restricting the available range for the temperatures and Fe abundances at a given radius according to the desired amount of smoothing. This method is simpler to implement and, we believe, allows more transparent control over the smoothing process.
To insure that we do not over-smooth our parameter profiles we do the following: (1) We always compare the results obtained from analysis with and without deprojection and require that the two cases do not differ qualitatively. (2) We only regularize the temperatures and Fe abundances and always vary the amount of smoothing for each observation until we are satisfied that significant bias is not introduced. In most cases we found that restricting the absolute values of the radial logarithmic derivatives in the temperature and Fe abundance to be less than 1 and 1.5 respectively worked well.
### 3.4. $`\chi ^2`$ Issues
All spectral fitting was performed with the software package xspec (Arnaud, 1996) using the $`\chi ^2`$ method implemented in its standard form. The weights for the $`\chi ^2`$ method are computed in each PI bin assuming gaussian statistics. To insure that the weights are valid in this assumption we regrouped the PI bins for each source spectrum so that each group has at least 30 counts. The extraction of spectra is discussed in the following section.
A possible concern with the interpretation of $`\chi ^2`$ for goodness-of-fit is that the deprojection procedure causes the emission at inner radii to depend on that from larger radii. However, the model representing the projected emission within a given annulus is taken to be exact; i.e. statistical errors from the projected emission model are not included in the deprojection method, and therefore the statistical errors in each PI bin remain those of the 2D annulus which are gaussian. We have verified the validity of using $`\chi ^2`$ for measuring goodness-of-fit under the standard assumption of gaussian statistics by comparing to results obtained from 2D analysis without deprojection: in all cases investigated we find that when a large improvement in $`\chi ^2`$ is obtained from the deprojection analysis a qualitatively consistent large improvement is also obtained with the 2D analysis.
## 4. Spectral Analysis
### 4.1. Preliminaries
#### 4.1.1 Extraction of Annular Spectra
For each system we extracted spectra in concentric circular annuli located at the X-ray centroid (computed within a $`2\mathrm{}`$ radius) such that for each annulus the width was $`1\mathrm{}`$ and the background-subtracted counts was larger than some value chosen to minimize uncertainties on the spectral parameters for each system while maintaining as many annuli as possible. Data with energies $`0.2`$ keV were excluded to insure that the PSF was $`<1\mathrm{}`$ FWHM. For our on-axis sources $`99\%`$ of the PSF at 0.2 keV is contained within $`R=1\mathrm{}`$ (Hasinger et al, 1995).
For systems possessing more than one observation (see Table 1) we extracted the source and background spectra separately from each observation and then added them together. In all such cases the observations occurred after the October 1991 gain change so the Redistribution Matrix File (RMF), which specifies the channel probability distribution for a photon, is the same for all of them. Since, however, the detector location of the annuli are in general slightly different for the different observations because of slight aspect differences, we averaged their respective Auxiliary Response Files (ARFs) which contain the information on the effective area as a function of energy and detector location. Any background sources that were identified by visual examination of the image were masked out before the extraction.
#### 4.1.2 Models
Since the X-ray emission of the galaxies and groups in our sample is dominated by hot gas, we use coronal plasma models as the basic component of our spectral models. We use the MEKAL plasma code which is a modification of the original MEKA code (Mewe, Gronenschild, & van den Oord 1985; Kaastra & Mewe 1993) where the Fe L shell transitions crucial to the X-ray emission of ellipticals and groups have been re-calculated (Liedahl et. al, 1995); the superior performance of the MEKAL model over the Raymond-Smith code for spectral analysis of elliptical galaxies we have previously discussed in detail in Buote (1999). Because of the limited energy resolution of the PSPC, we focus on a “single-phase” description of the X-ray emission in which a single temperature component exists at each (three-dimensional) radius. As we discuss below in §5.1 we do not expect a significant “Fe Bias” resulting from the single-phase analysis. We did investigate multiphase models of the hot gas such as a two-temperature plasma and a constant-pressure cooling flow (Johnstone et al, 1992), but interesting constraints were not obtained as explained below in §4.3; see §3.1.1 of Buote (2000a) for further description of such models.
We account for absorption by our Galaxy using the photo-electric absorption cross sections according to Balucińska-Church & McCammon (1992). Although Arabadjis & Bregman (1999) point out that the He cross section at 0.15 keV is in error by 13%, since we analyze $`E>0.2`$ keV we find that our fits do not change when using the Morrison & McCammon (1983) cross sections which have the correct He value. The absorber is modeled as a uniform screen at zero redshift with solar abundances. The hydrogen column density of the absorber is generally allowed to be a free parameter to indicate any additional absorption due to, e.g., intrinsic absorbing material, calibration errors, etc. We refer to this as the standard absorber model. (In PAPER3 we discuss more complex absorption models.)
As discussed in PAPER1 and PAPER3 we also find it useful to consider absorption due to an oxygen edge which we represent by the simple parameterization, $`\mathrm{exp}\left[\tau (E/E_0)^3\right]`$ for $`EE_0`$, where $`E_0`$ is the energy of the edge in the rest frame of the galaxy or group and $`\tau `$ is the optical depth; i.e. $`\tau `$ represents an absorbing screen located at the source redshift but placed in front of the source. Partial covering models are discussed in PAPER3.
Finally, data with energies between 0.2 and 2.2 keV were included in the analysis.
#### 4.1.3 Meteoritic Solar Abundances
Ishimaru & Arimoto (1997) have pointed out that the accepted value for the solar Fe abundance relative to H is approximately $`3.24\times 10^5`$ by number. This value is often called the meteoritic value since it was originally obtained for meteorites, but it also agrees with recent measurements from the solar photosphere. This is to be contrasted with the old photospheric value of Fe/H of $`4.68\times 10^5`$ (Anders & Grevesse, 1989) that is widely used. We have decided to use the correct “meteoritic” Fe abundance for this and subsequent papers; in xspec we use the abundance table of Feldman (1992). However, when comparing to previous results we shall always take care to consider the factor of 1.44 between the different Fe/H values.
Iron is the only abundance that we allow to be a free parameter. All other elemental abundances are tied to iron in their solar ratios; i.e. the values of the other abundances vary with Fe in their fixed solar ratios. In a sense we are actually fitting a metallicity, but since the Fe L-shell lines dominate all other lines in the PSPC spectrum for the $`1`$ keV plasmas of bright galaxies and groups, the fitted metallicity is almost entirely determined by Fe. Hence, we shall always quote our results as Fe abundances.
#### 4.1.4 Scaling the Background to the Source Position
The observed background spectrum obtained from regions far away from the source in general covers a different amount of detector area, requires a different detector response, and if taken from a different observation can have a different exposure time. Within xspec the effect of different areas and exposure times for source and background are taken into account but not different detector responses. Background spectra taken near the edge of the PSPC field suffer from vignetting which significantly reduces the count rate with respect to the source spectra near the field center. This effect is also energy dependent, and thus the spectral shape of the background is altered as well.
To properly scale the background spectrum to the source position we perform the following simple procedure. First, as explained in §2 we fit a model to the background spectrum. Let us denote the flux of this model for an energy, $`E`$, by, $`F_E(r_b,A_b,t_b)`$, where $`r_b`$ represents the detector position, $`A_b`$ the background area on the detector, and $`t_b`$, the exposure time. The corresponding flux in a given PI channel, $`i`$, predicted by this model is then,
$$F_i(r_b,A_b,t_b)=\underset{E}{}F_E(r_b,A_b,t_b)\mathrm{RMF}_{Ei}(r_b)\mathrm{ARF}_E(r_b).$$
The model, $`F_E`$, is independent of the response, and thus the expected background flux in channel, $`i`$, at the source position is simply,
$`F_i(r_s,A_s,t_s)=`$
$`{\displaystyle \underset{E}{}}\left(F_E(r_b,A_b,t_b){\displaystyle \frac{A_st_s}{A_bt_b}}\right)\mathrm{RMF}_{Ei}(r_s)\mathrm{ARF}_E(r_s),`$ (1)
where $`r_s`$, $`A_s`$, and $`t_s`$ are respectively the position, area, and exposure time for the source.
In the general case one would prefer to use the actual background data rather than a best-guess model. To do this simply scale the real background spectrum by the factors $`F_i(r_s,A_s,t_s)/F_i(r_b,A_b,t_b)`$ for each PI channel. These ratios are typically very insensitive to the detailed shape of the input model spectrum.
This scaling procedure was followed for NGC 5044 and 5846. For the other systems we needed to subtract a thermal component due to the extended source emission as discussed in §1. Since for these systems the background estimate is necessarily defined by a model we used equation (1) for the scaling.
### 4.2. Results
We plot in Figures 1-4 the temperature and Fe abundance profiles obtained from the deprojection analysis according to the number of annuli for which useful constraints on the parameters were obtained. This categorizes the systems essentially according to the S/N of the data. Fe abundance profiles are shown for the cases where the column density of the standard absorber was (1) fixed at the Galactic value and (2) treated as a free parameter. (The temperature profiles are very similar in each case, and thus for clarity of presentation we show only the case for fixed Galactic column density.) For both the temperature and Fe abundance profiles we also show the results obtained with and without an extra oxygen edge at 0.532 keV (rest frame). The profiles of column density and edge optical depth are given in PAPER1 for NGC 1399 and 5044 and in PAPER3 for all the systems.
The deprojected temperature profiles are all very similar to the 2D profiles obtained in many previous studies with the ROSAT PSPC (Forman et al 1993; Ponman & Bertram 1993; David et al 1994; Trinchieri et al 1994; Rangarajan et al 1995; Kim & Fabbiano 1995; Irwin & Sarazin 1996; Jones et al 1997; Trinchieri et al 1997; Mulchaey & Zabludoff 1998; Buote 1999). As expected, our 3D profiles display a slightly steeper rise from the central minimum to the maximum at $`r50`$-100 Mpc in very good agreement with the 3D profiles obtained by Finoguenov and co-workers for several systems using both ROSAT and ASCA data (Finoguenov et al 1999; Finoguenov & Ponman 1999; Finoguenov & Jones 2000). The temperature profiles also agree very well with two-temperature models of ASCA data (Buote 1999, 2000a).
The models with an oxygen edge tend to have lower temperatures at small radii. This can occur because lower temperature gas produces stronger oxygen lines which are absorbed by the oxygen edge. Although in most cases both models give fits of similar quality, the edge model provides a significantly better fit for some of the systems (see PAPER1 and PAPER3).
For the four systems with the highest S/N data in our sample (NGC 507, 1399, 4472, and 5044) the Fe abundance profiles obtained from models (without an oxygen edge) where the column density is treated as a free parameter are very much larger than for all models with fixed Galactic column density. Even after accounting for the differences between meteoritic and photospheric solar abundances, the Fe abundances implied by these models greatly exceed and are very inconsistent with the Fe abundances obtained from all ASCA studies. They are, however, consistent with previous ROSAT studies where the column density was treated as a free parameter (e.g., Forman et al 1993; David et al 1994).
For these systems the models with variable column density give Fe abundances that increase rapidly with decreasing radius, but the column densities (except for NGC 507) also decrease well below the Galactic values as $`r0`$ (PAPER1 and PAPER3). Since the column densities at large radii are consistent with the Galactic values this decrease at small radii cannot be due to errors in the background subtraction. Rather, the sub-Galactic columns and huge Fe abundances that are inconsistent with ASCA signals some inadequacy in the emission model. As it turns out (see Figures 1-4) except for NGC 5044 the variable-column density models with an oxygen edge give Fe abundances (and columns) that are very consistent with the fixed-column models. This gives additional support for the edge model discussed in PAPER1 and PAPER3 (though see §5.3 of PAPER3).
In most of the other systems the variable-column-density models also tend to predict sub-Galactic columns at small radii though at much lower significance. These sub-Galactic columns are also accompanied by systematically larger Fe abundances than the fixed-Galactic case, though again the edge models usually give consistent abundances for the free- and fixed-column density cases. Hence, the Fe abundance profiles which typically agree best with previous ASCA analyses are those obtained from models with fixed Galactic column density (with or without an edge), though variable column models with an oxygen edge yield comparable results in most cases. Since, however, the evidence for intrinsic oxygen absorption is strong for half of these systems (PAPER1 and PAPER3) we believe the models with an intrinsic oxygen edge and a standard absorber with $`N_\mathrm{H}=N_\mathrm{H}^{\mathrm{Gal}}`$ to be the most physical of the models investigated (i.e. crosses and dotted diamonds in middle column of Figures 1 \- 4).
From inspection of Figures 1-4 it is readily apparent that most of the systems have Fe abundance profiles that decrease as a function of radius. Typically within $`r10`$ kpc the Fe abundance ranges from $`Z_{\mathrm{Fe}}1Z_{\mathrm{}}(\mathrm{several})Z_{\mathrm{}}`$ and decreases to $`Z_{\mathrm{Fe}}0.5`$$`Z_{\mathrm{}}`$for $`r50100`$ kpc. We now discuss the individual systems in groups defined by the total number of annuli with interesting constraints. When comparing our Fe abundances to previous studies we implicitly account for projection effects and the different plasma codes and solar abundances used (§4.1.3).
#### 4.2.1 Systems with 7 Annuli
In Figure 1 we display the results for the three systems where the spectral parameters are well determined in seven annuli. These observations thus generally correspond to the highest S/N data in our sample. In each case the models with an oxygen edge exhibit significant negative Fe abundance gradient, though in NGC 1399 the key evidence occurs in the central radial bin.
NGC 507: The evidence for a Fe abundance gradient is strong as is the large value in the central bin: 95% confidence lower limits are 1.3$`Z_{\mathrm{}}`$ and 1.1$`Z_{\mathrm{}}`$ respectively for models with and without an oxygen edge (column density fixed at Galactic). The average Fe abundance within $`r=4\mathrm{}`$ is $`1Z_{\mathrm{}}`$ in excellent agreement with the value obtained by Buote & Fabian (1998) from a two-temperature model of the ASCA data. Our Fe abundance profile also agrees reasonably well with the 2D profile obtained in the previous ROSAT study by Kim & Fabbiano (1995).
NGC 1399: The Fe abundance profile is constant for $`r3\mathrm{}`$, rises slightly for $`r1\mathrm{}`$-$`2\mathrm{}`$, and in the center has different values depending on the model. As shown in PAPER1 and PAPER3 the edge model is strongly preferred, and thus the Fe abundance is consistent with being largest at the center; note that even though the error estimate for the fixed-column-density case appears to underestimate the Fe abundance, we still estimate a 95% lower limit of 1.1$`Z_{\mathrm{}}`$which is significantly greater than at large radii. The average Fe abundance within $`r=4\mathrm{}`$ is $`1.4Z_{\mathrm{}}`$ which agrees with the lower limit obtained by Buote (1999) from a two-temperature model of the ASCA data and is in better agreement with a cooling flow model. Our Fe abundance profile for the case of variable column density and no edge agrees very well with the 2D profile of Jones et al (1997).
NGC 5044: Again there is clear evidence that the Fe abundance profile is not constant with radius. As noted earlier the models with variable column density predict large Fe abundances that are inconsistent with ASCA studies. (But they are consistent with the previous ROSAT study of David et al 1994.) For example, Buote (1999) obtains $`Z_{\mathrm{Fe}}0.9Z_{\mathrm{}}`$ within $`r5\mathrm{}`$ for a two-temperature model of the ASCA data. Since in PAPER1 and PAPER3 we also find that an oxygen edge is clearly required in the central regions, let us focus on the Fe abundance profile obtained from a model with an edge and with fixed Galactic column density. This model yields a Fe abundance of $`Z_{\mathrm{Fe}}0.7Z_{\mathrm{}}`$ for $`r3\mathrm{}`$ and $`Z_{\mathrm{Fe}}1.5Z_{\mathrm{}}`$ for smaller $`r`$.
#### 4.2.2 Systems with 5-6 Annuli
The results for systems having 5 or 6 annuli are plotted in Figure 2. Evidence for a negative Fe abundance gradient is clear for NGC 4472, marginal for NGC 5846, and non-existent for NGC 2563.
NGC 2563: Independent of the model it is clear that the Fe abundance is constant with radius with a value of $`2.0Z_{\mathrm{}}`$ when the column density is treated as a free parameter as opposed to $`1.5Z_{\mathrm{}}`$ for fixed Galactic column density. These results are consistent with the Fe abundance determined within $`r3\mathrm{}`$ by Buote (2000a) from a two-temperature model of the ASCA data.
NGC 4472: Analogously to NGC 5044 the Fe abundances obtained for the model with variable column density and no edge are very large and very inconsistent with all previous ASCA measurements. Also similar to NGC 5044 is that the edge model is strongly preferred (see PAPER3). The edge models give $`Z_{\mathrm{Fe}}1Z_{\mathrm{}}`$ for $`r>3\mathrm{}`$ and $`Z_{\mathrm{Fe}}2Z_{\mathrm{}}3Z_{\mathrm{}}`$ at smaller radii. These values are in excellent agreement with those obtained from previous 2D studies with ROSAT (Forman et al 1993; Irwin & Sarazin 1996). These Fe abundances also agree very well with the value obtained by Buote (1999) from a two-temperature model of the ASCA data.
NGC 5846: The models with variable column density show clear evidence for a negative Fe abundance gradient, but since the column densities obtained within the inner two bins are sub-Galactic we do not consider these models to be physical (§5.3 of PAPER3). The models with fixed Galactic column density with and without an edge give very similar profiles except that in the central bin the upper limit on the Fe abundance for the edge model is not very well determined. These fixed-column models suggest a negative gradient in Fe abundance, but the errors are sufficiently large so that a constant profile is not clearly excluded. These Fe abundances also agree very well with the values obtained by Buote (2000a) from multitemperature models of the ASCA data.
#### 4.2.3 Systems with 4 Annuli
The galaxies and groups for which we obtained useful constraints in four radial bins are shown in Figure 3. Although the data sets for these systems have lower S/N than the previous examples, the evidence for negative Fe abundance gradients is as strong or stronger than the others.
NGC 533: All models give $`Z_{\mathrm{Fe}}0.5Z_{\mathrm{}}`$ for $`r3\mathrm{}`$ and $`Z_{\mathrm{Fe}}1Z_{\mathrm{}}`$ within $`1\mathrm{}`$. These Fe abundances are consistent with the results for multitemperature models obtained with ASCA within $`r=3\mathrm{}`$ by Buote (2000a).
NGC 4636: In the outer bin all models give a very sub-solar value for the Fe abundance, $`Z_{\mathrm{Fe}}0.4Z_{\mathrm{}}`$. The Fe abundance rises with decreasing radius such that $`Z_{\mathrm{Fe}}1Z_{\mathrm{}}`$ in the central bin. These results are consistent with those obtained by Buote (1999) for multitemperature models of ASCA data analyzed within $`r=5\mathrm{}`$.
HCG 62: Similar to NGC 4636, all models give a very sub-solar value for the Fe abundance, $`Z_{\mathrm{Fe}}0.2Z_{\mathrm{}}`$, in the outer bin. The Fe abundance rises very sharply with increasing radius such that $`Z_{\mathrm{Fe}}2Z_{\mathrm{}}`$ in the central bin. This system appears to possess the most significant Fe abundance gradient in our sample: the 95% lower limits on the values in the central bin are 4.6$`Z_{\mathrm{}}`$ and 3.2$`Z_{\mathrm{}}`$ respectively for models with and without an edge (both with column density fixed to Galactic). The Fe abundance of $`1.4Z_{\mathrm{}}`$ obtained by Buote (2000a) using a two-temperature model of the ASCA data accumulated within $`r=3\mathrm{}`$ is consistent with our ROSAT results provided the Fe abundance in the central bin is near the estimated $`1\sigma `$ lower limits.
#### 4.2.4 Systems with 3 Annuli
Finally, in Figure 4 we show the results for the galaxy with the smallest number of annuli, NGC 4649. Similar to the 4-annuli objects, NGC 4649 is one of the most significant examples for a decrease in Fe abundance with increasing radius.
NGC 4649: For this system the edge model is clearly preferred (see PAPER3), though the Fe abundances are fairly similar for all of the models. In the outer bin $`Z_{\mathrm{Fe}}0.25Z_{\mathrm{}}`$ for all models and rises to $`Z_{\mathrm{Fe}}1.5`$ within the central bin for the edge models. Buote & Fabian (1998) obtained a Fe abundance of $`1.3Z_{\mathrm{}}`$ from a two-temperature model of the ASCA data within $`r=3\mathrm{}`$ in good agreement with our ROSAT results.
### 4.3. Caveats
(i) Spherical Symmetry: The X-ray isophotes of most of the systems are approximately circular over the regions examined. Notable exceptions are NGC 1399 and 4472 which show significant deviations from circular symmetry at large radius. For $`R5\mathrm{}`$ there is a significant N-S asymmetry in the X-ray surface brightness of NGC 1399 which Jones et al (1997) hypothesize is due to incomplete relaxation of the gas in the surrounding group. In contrast, outside of $`R3\mathrm{}`$ the X-ray isophotes of NGC 4472 flatten which Irwin & Sarazin (1996) attribute to ram pressure as the galaxy moves with respect to the Virgo cluster. Our results for these systems at large radii average over these asymmetries. It is worth remarking that although these systems have larger than solar Fe abundances at their centers, we find that at the largest radii $`Z_{\mathrm{Fe}}0.75Z_{\mathrm{}}`$ for NGC 1399 and $`Z_{\mathrm{Fe}}0.3Z_{\mathrm{}}`$ for NGC 4472; i.e. it is plausible that the small value for NGC 4472 is significantly influenced by the ambient gas of the Virgo cluster.
(ii) Single-Phase Gas: We have focused on single-phase analysis of the hot gas because in no case did we find that multiphase models of the hot gas improved the fits significantly. In the central bins of most of the systems we expect gas to be emitting over a range of temperatures because of the observed temperature gradients. However, deprojection has removed the high-temperature gas components from the centers leaving a much smaller temperature range (typically a few tenths of a keV) which cannot be distinguished by the limited energy resolution of the PSPC. (Note that even if gas is dropping out of a cooling flow, emission weighted temperatures are within 50% of the ambient value – e.g., Buote, Canizares, & Fabian 1999). These small temperature ranges also indicate that the “Fe Bias” should be unimportant (see appendix of Buote 2000a), and thus multitemperature models do not give qualitatively different Fe abundances in the central bins – see also §5.1.
(iii) Background Issues: There are two backgrounds which must be considered. First, the edge effect described in §3.1 requires one to assume a model for the emission outside of the bounding annulus. We never found any noticeable effect on the derived temperatures and abundances for reasonable choices of $`\beta `$. Only the electron density is affected and only for the outermost annuli. Second, we examined the effects of taking background estimates from different parts of the detector and from different fields nearby to a given object. We found that in most cases when the background level was over-estimated the Fe abundances tended to be larger than when the background was under-estimated. The size of the effect depends on many factors (e.g., S/N) though again we found that for reasonable background choices only the outermost annuli are affected as would be expected. (Of course such effects will propagate to smaller radii if the regularization criteria are not applied with care – see below.)
## 5. Discussion and Conclusions
Deprojection analysis of the ROSAT PSPC data of 10 cooling flow galaxies and groups reveals clear evidence for Fe abundances that decrease with radius in all but one system (NGC 2563). Typically $`Z_{\mathrm{Fe}}0.5Z_{\mathrm{}}`$ at the largest radii examined ($`r50`$-100 kpc) which increases to $`Z_{\mathrm{Fe}}1Z_{\mathrm{}}(\mathrm{several})Z_{\mathrm{}}`$ within the central radial bin ($`r10`$ kpc). In most cases the estimated uncertainties on $`Z_{\mathrm{Fe}}`$ are large within the central bin and allow for $`Z_{\mathrm{Fe}}1Z_{\mathrm{}}`$ (but never $`Z_{\mathrm{Fe}}1Z_{\mathrm{}}`$). Throughout this paper we have used the “meteoritic” solar abundances (Fe/H is $`3.24\times 10^5`$ – see Ishimaru & Arimoto 1997) which leads to Fe abundances that are a factor of 1.44 larger than previous studies that used the old “photospheric” solar value for Fe (see §4.1.3).
These 3D Fe abundance profiles are generally consistent with the original 2D ROSAT studies (Forman et al. 1993; David et al 1994; Kim & Fabbiano 1995; Irwin & Sarazin 1996; Jones et al 1997) after accounting for projection effects and for the different plasma codes and solar abundances used (though see below in §5.1). For all 10 systems the Fe abundances (and temperatures) that we have obtained from the deprojected ROSAT data agree with those we have obtained previously from analysis of the ASCA data accumulated within $`r3\mathrm{}`$-$`5\mathrm{}`$ (Buote & Fabian 1998; Buote 1999, 2000a); qualitatively similar results for some of these systems also have been recently obtained with ASCA data by Allen, Di Matteo, & Fabian (2000).
Therefore, within $`r50`$ kpc of these bright galaxies and groups the ROSAT and ASCA data clearly demonstrate that the gas is non-isothermal with approximately solar Fe abundances. We now examine whether it is the assumption of isothermality or the presence of Fe abundance gradients which accounts for the very sub-solar Fe abundances inferred by most previous ASCA (and some ROSAT ) studies (see Buote 2000a for a detailed review).
### 5.1. Fe Bias vs Fe Gradients
Previously we have discussed how fitting a single-temperature plasma model to the soft X-ray spectrum of a bright elliptical galaxy or galaxy group that actually consists of multiple temperature components can lead to a serious underestimate of the Fe abundance (Buote & Canizares 1994; Buote & Fabian 1998; Buote 1999, 2000a). In our most comprehensive treatment to date (Buote, 2000a) we provided a physical explanation of this “Fe Bias” from examination of ASCA data which we now summarize in Figure 5.
The accumulated ASCA spectra within $`r3\mathrm{}`$-$`5\mathrm{}`$ of the brightest elliptical galaxies and groups (including those in our present investigation) are well represented by two-temperature (2T) models. These 2T models are also consistent with the radial temperature gradients inferred from the ROSAT data within the same spatial regions (Buote 1999, 2000a). Moreover, we have verified that such 2T models continue to provide a good representation of the accumulated spectra generated by the ROSAT temperature profiles out to the largest radii ($`15\mathrm{}`$) investigated in our present paper whether or not the temperature rises and then falls (e.g., NGC 2563 – Figure 2) or continues to rise out to the largest radii (e.g., NGC 4472 – Figure 2). This agreement is expected because a 2T model can mimic very accurately the soft X-ray spectra of cooling flows (Buote et al, 1999).
In Figure 5 we show a simulated ASCA SIS spectrum of a 2T model with $`Z_{\mathrm{Fe}}=1Z_{\mathrm{}}`$ with temperatures and relative emission measures typical of the bright galaxies and groups. For this discussion we consider only the energies 0.55-1.5 keV to emphasize the Fe L lines and the lower bandpass limit of the SIS. Also shown in Figure 5 is the result of fitting a single-temperature (1T) model with $`Z_{\mathrm{Fe}}`$ fixed at $`1Z_{\mathrm{}}`$ to the simulated 2T spectral data. It is readily apparent that the 1T model is a poor fit to the 2T spectrum as it is too peaked near 1 keV while being deficient in emission at other energies. This behavior simply reflects the ability of the 2T model to excite Fe L lines over a wider range of energies than the 1T model. In order to force the 1T model to better fit the 2T spectrum one has to reduce the size of the peak at 1 keV which means reducing the Fe abundance. Allowing $`Z_{\mathrm{Fe}}`$ to be a free parameter results in a better (but still not good) fit and with a best-fitting value of $`Z_{\mathrm{Fe}}=0.3Z_{\mathrm{}}`$ (see Figure 5.1). It is this effect that we have termed the “Fe Bias”.
This Fe Bias is dependent on the energy resolution of the detector and is more pronounced at higher resolution which is why we have focused on the ASCA SIS data. But if the models in Figure 5.1 are folded with the relatively low resolution of the ROSAT PSPC we obtain a best-fitting value of $`Z_{\mathrm{Fe}}=0.4Z_{\mathrm{}}`$ for the 1T model, very similar to that obtained above at higher resolution with the SIS. However, it must be emphasized that this underestimate with the PSPC is achieved only if the lower limit on the bandpass is restricted to energies $`0.6`$ keV as appropriate for the SIS. If instead the lower energy limit is reduced to 0.2 keV as is appropriate for the PSPC then we obtain $`Z_{\mathrm{Fe}}0.60.7Z_{\mathrm{}}`$ for the 1T model.
This smaller underestimate occurs because while reducing the Fe abundance diminishes the line emission of the 1T model to better fit the spectral maximum near 1 keV, the continuum of the 1T model must increase to compensate for the decrease in the Fe L emission at energies near $`0.8,1.2`$ keV. It also compensates for the corresponding decrease in the strong O K$`\alpha `$ emission near 0.6 keV since O (like the other elements) varies with Fe in its solar ratio (see §4.1.3).
This raising of the continuum by the 1T model generally goes unnoticed when analyzing ASCA data for the following reason. The ASCA SIS and GIS bandpasses do not extend below the energy regions shown in Figure 5, but they do include energies up to $`10`$ keV. However, the emission at higher energies is usually accounted for by introducing an additional high-temperature bremsstrahlung component (BREM) to represent emission from discrete sources (e.g., Matsumoto et al. 1997), and thus any increase or decrease in the emission of the 1T model at energies much larger than 1 keV can be compensated by changing the normalization of the BREM component. The required amount of compensation is generally unimportant because the emission from the 1T model decreases as $`\mathrm{exp}(E/T)`$, and thus a plasma with $`T1`$ keV contributes little to the emission of a 1T+BREM model at higher energies.
However, spectral analysis with PSPC data includes lower energies (i.e. $`0.20.5`$ keV) that are dominated by continuum emission. The data at these energies restrict the attempt by the data near 1 keV to simultaneously reduce the Fe abundance and increase the continuum in order to force a better fit of a 1T model to the 2T spectrum. Hence, the Fe Bias affecting analysis of PSPC data is much less pronounced than for ASCA data because of the lower energy resolution of the PSPC, and especially because the PSPC data include energies below $`0.5`$ keV.
Our deprojection analysis of the PSPC data removes the temperature components projected from larger radii from the spectrum of a given annulus. Consequently, if the gas is single-phase (i.e., a single density and temperature at each radius) the range of temperatures within a given annular spectrum must be small after deprojection (i.e., essentially the temperature difference between the annuli that surround the annulus in question). Since each deprojected spectrum is nearly isothermal, and since the Fe Bias is already greatly reduced when analyzing PSPC data, we expect that the Fe Bias does not significantly affect our measurements of the Fe abundances from the PSPC data in this paper. It is therefore not surprising that the Fe abundances we have obtained from our deprojection analysis of the PSPC data agree only with those obtained from the ASCA studies that have removed the Fe bias by fitting multitemperature models of the ASCA spectra accumulated within substantial spatial regions ($`r3\mathrm{}5\mathrm{}`$; Buote & Fabian 1998; Buote 1999, 2000a; Allen et al 2000).
If a warm gas phase ($`T=10^510^6`$ K) also contributes to the X-ray emission of the galaxies and groups as indicated by the intrinsic absorption reported in PAPER1 and PAPER3, then the single-phase assumption needs to be reconsidered. However, the low temperature of the warm gas implies a negligible contribution to the Fe L emission and thus will not contribute to an Fe Bias. Another possibility is that the hot gas emits over a continuous range of temperatures as might be expected from an inhomogeneous cooling flow that cools at constant pressure (e.g., Johnstone et al 1992). As mentioned in §4.3 we do not find any significant change in the inferred Fe abundances if we use such a cooling flow model.
If the Fe Bias is not very important for analysis of ROSAT PSPC data, then why do some ROSAT studies find very low Fe abundances? As stated above the Fe abundances we have determined from the deprojection analysis of the PSPC data of the 10 galaxies and groups in our sample agree very well with those obtained from previous spatial-spectral analyses of these systems with the PSPC after accounting for (1) the factor of 1.44 arising from our use of the meteoritic solar abundances, (2) different plasma codes used, and (3) small differences associated with our use of a deprojection analysis (Forman et al. 1993; David et al 1994; Kim & Fabbiano 1995; Irwin & Sarazin 1996; Jones et al 1997).
Our results also agree with those of Helsdon & Ponman (2000) who fitted 1T models within large apertures (typically $`20\mathrm{}`$ radius) for several groups including NGC 533, 2563, 4636, and 5846. To make a more realistic comparison we summed up the emission models we obtained for all annuli for these systems, simulated a PSPC observation using the same exposure time, and then fitted the simulated spectrum with a 1T model like Helsdon & Ponman. The Fe abundances we obtained from this exercise are fully consistent within the $`2\sigma `$ errors for each system. We emphasize that Helsdon & Ponman find $`Z_{\mathrm{Fe}}0.6Z_{\mathrm{}}`$ (scaled to meteoritic solar) for these systems even when using very large apertures.
In contrast Mulchaey & Zabludoff (1998) obtain $`Z_{\mathrm{Fe}}0.2Z_{\mathrm{}}`$ from 1T (Raymond-Smith) fits to the PSPC spectra of NGC 2563, HCG 62, and NGC 5846 within radii of $`20\mathrm{}`$ (though $`Z_{\mathrm{Fe}}0.6Z_{\mathrm{}}`$ is obtained for NGC 533). The values for NGC 2563 and 5846 obtained by Mulchaey & Zabludoff are inconsistent with those obtained from the very similar analysis by Helsdon & Ponman (2000) and from our deprojection analysis.
A possible explanation could be associated with the background subtraction. As we noted in §4.3 the Fe abundances will be underestimated if the background is also underestimated. This effect is only noticeable when the background is underestimated by factors or $`2`$ or more. Such a large factor will result if the background spectrum is taken from the edge of the field and not corrected for vignetting before subtracting from a source region near the center of the field. As explained in §4.1.4 we always account for the different detector responses for the source and background in our analysis.
However, when ignoring the response differences we are able to reproduce the very small Fe abundances obtained by Mulchaey & Zabludoff for NGC 2563, 5846, and HCG 62. (We perform the 1T fits in the same manner as described above for our comparison to the results of Helsdon & Ponman; i.e. we include all of the annuli investigated.) Our value for NGC 533 is consistent within the large statistical error with that obtained by Mulchaey & Zabludoff whether or not we properly scale the background to the source positions. Since the Fe abundances that we have determined from the PSPC data agree with all other previous ROSAT studies except Mulchaey & Zabludoff, we are lead to conclude that they must not have corrected their background spectra for vignetting before subtracting from the source spectra.
Hence, there is no evidence from the ROSAT PSPC data that the brightest elliptical galaxies and galaxy groups have average Fe abundances that are very sub-solar – regardless of whether they have steep Fe abundance gradients. The only clear evidence for very sub-solar Fe abundances is thus derived from fits to the ASCA data assuming an isothermal gas which we have shown to give misleading results because of the Fe bias. Thus, it is the assumption of isothermality, rather than the presence of Fe abundance gradients, which accounts for the very sub-solar Fe abundances inferred by most previous ASCA studies (see Buote 2000a for a detailed review).
### 5.2. Low Fe Abundance as the Result of Over-Smoothing
The consistent picture that we have described is not supported by the work of Finoguenov and co-workers who find very sub-solar Fe abundances from deprojection analysis of the ASCA data of HCG 62, NGC 4472, 4636, 4649, 5044, and 5846 (Finoguenov et al 1999; Finoguenov & Ponman 1999; Finoguenov & Jones 1999). For HCG 62 and NGC 5044 Finoguenov & Ponman (1999) also perform a deprojection of the ROSAT data and obtain results not too dissimilar from their ASCA analysis. As noted before in §4.2 we emphasize that our temperature profiles for these systems agree well with those obtained in the Finoguenov papers.
We have previously speculated on possible reasons for the different Fe abundances obtained by Finoguenov (see §4.1 of Buote 2000a). Now equipped with our own deprojection code we can test one of those speculations: the regularization assumptions. As discussed in §3.3 Finoguenov & Ponman (1999) require that the temperatures and abundances vary approximately logarithmically with radius. However, we have shown that most of these systems, especially HGC 62, have steep Fe abundance gradients and thus these regularization criteria must be carefully checked.
In Figure 6 we plot the radial profiles of temperature and Fe abundance of HCG 62 obtained from the ROSAT deprojection analysis where we have used the old photospheric solar Fe abundances for consistent comparison with Finoguenov & Ponman. Because of the steep Fe abundance gradient we chose a slope of 2 to bound the radial logarithmic Fe abundance derivative. (We arrived at this choice after trying several different values and from comparison to results obtained without deprojection – see §3.3.) We also show the results when the maximum Fe abundance derivative is set to 1. These latter results agree well with the ASCA and ROSAT deprojection results of Finoguenov & Ponman; e.g., for $`r30`$ kpc we obtain $`Z_{\mathrm{Fe}}0.6Z_{\mathrm{}}`$ which is within the 90% limits of their ASCA result and in excellent agreement with their ROSAT value. This value is to be contrasted with $`Z_{\mathrm{Fe}}1.3Z_{\mathrm{}}`$ obtained at that radius when we set the maximum Fe abundance derivative to 2.
Therefore, we attribute the smaller Fe abundances obtained by Finoguenov & Ponman for HGC 62 to their overly restrictive regularization criteria. Interestingly, unlike their ASCA analysis of HCG 62, their ROSAT deprojection does not perform actual regularization but rather adopts simple smooth analytical models which apparently achieve the same effect. Finoguenov & Ponman do apply their regularization method to the deprojection of the ROSAT data of NGC 5044, and we can only reproduce their temperatures and small Fe abundances if we require the radial logarithmic derivatives of both the temperature and Fe abundance to be $`0.2`$! In fact, given the steep Fe abundance gradients we obtain for most of the systems analyzed in the Finoguenov papers over-smoothing probably accounts for most of the Fe abundance differences between our papers.
### 5.3. Implications and Future Work
The Fe abundance gradients of these bright, nearby cooling flow galaxies and groups reflect the history of star formation and the dynamical evolution of the hot gas in these systems. In Figure 7 we compare the Fe abundances we have measured for NGC 4472 with those predicted from a gas-dynamical metal enrichment model which assumes (1) a Galactic IMF, (2) accretion of primordial gas arising from secondary infall, and (3) a Type Ia supernova (SNIa) rate normalized to the observed value in nearby E/S0 galaxies reported by Cappellaro et al (1997). The qualitative agreement is clear, though the deviations at larger radii indicate that the ram pressure distortions discussed by Irwin & Sarazin (1996) compromise to some extent the assumptions of spherical symmetry and hydrostatic equilibrium in the gas-dynamical model.
The relative contributions of the supernovae to the Fe abundance within $`50`$ kpc are consistent with the “solar supernova proportion” defined by Renzini et al (1993): i.e., enrichment models of the Milky Way indicate that 1/4 of the Fe in the Sun arises from Type II supernovae (SNII) and the remaining 3/4 arises from SNIa. Consequently, the approximately solar Fe abundances we have found for the brightest galaxies and groups within $`r50`$ kpc are consistent with models of the chemical enrichment of the Milky way and allow for the possibility of a universal IMF (e.g., Renzini et al. 1993; Renzini 1997, 2000; Wyse 1997). The problems associated with the theoretical interpretation of the very sub-solar Fe abundances inferred from previous studies of these systems (e.g., Renzini et al. 1993; Arimoto et al. 1997; Renzini 1997) are eliminated by the approximately solar Fe abundances we have inferred from both ASCA and ROSAT data of bright galaxies and groups.
The sub-solar Fe abundances at large radii require additional explanation, such as the dilution arising from the accretion of primordial material included in the model in Figure 7. An alternative explanation for negative Fe abundance gradients as the consequence of the sedimentation of heavier elements (Fabian & Pringle 1977; see Qin & Wu 2000 for a recent discussion) seems less plausible since the drift velocities are generally similar to or less than the flow velocities in a cooling flow.
We mention that the Fe abundance at large radius ($`100`$ kpc) predicted by the model for NGC 4472 (Figure 7) arises entirely from SNII ejecta. These SNII explosions accompanied the initial burst of star formation which ended at a redshift before $`2`$. Hence, the Fe abundance measured at such large radii is a direct fossil record of the ISM abundance at high redshift. For those systems where we have measured Fe abundances at $`r100`$ kpc we find $`Z_{\mathrm{Fe}}0.5Z_{\mathrm{}}`$. If they are indeed the result of the initial SNII enrichment these large Fe abundances are consistent with the arguments that large quantities of the metals at high redshift reside in hot gaseous halos (Pettini 1999, 2000) and that there was a prompt initial enrichment of the early universe (Renzini 1997, 2000).
Finally, Chandra and XMM will substantially clarify our understanding of the metal enrichment and gas dynamics in cooling flows. Their superior spatial and spectral resolution will enable abundances to be measured on scales $`r1`$ kpc which is critical since the ROSAT data suggest that the gradients may be largest at small radii (§4.2). These new missions will also allow accurate mapping of the abundances of several $`\alpha `$ elements (e.g., Si) which is of special importance because the $`\alpha `$/Fe ratios provide the strongest constraints on the relative numbers of Type II and Type Ia supernovae. The detailed constraints on the supernovae feedback in these nearby cooling flows will be invaluable for interpreting the star formation process in galaxies at high redshift (Cavaliere et al, 2000).
It is a pleasure to thank the referee, J. Irwin, for helpful comments, F. Kelly for programming advice, and W. Mathews for permission to include Figure 7 in this paper. This research has made use of (1) data obtained through the High Energy Astrophysics Science Archive Research Center Online Service, provided by the NASA/Goddard Space Flight Center, and (2) the NASA/IPAC Extragalactic Database (NED) which is operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract with NASA. Support for this work was provided by NASA through Chandra Fellowship grant PF8-10001 awarded by the Chandra Science Center, which is operated by the Smithsonian Astrophysical Observatory for NASA under contract NAS8-39073.
## Appendix A Limiting Case for Edge Correction Factor
The edge correction factor, $`f(R_{i1},R_i)`$, given by equation A8 of McLaughlin (1999) evaluates to zero in both the numerator and denominator when $`R_{i1}=0`$. However, the limit is well-behaved there, and after applying L’Hôpital’s rule we obtain,
$$\underset{ϵ0}{lim}f(ϵ,R_i)=\frac{\left(R_{m1}+R_m\right)R_{m1}R_m}{R_i^3}\left(1+\frac{2}{\pi }\left[\frac{2R_i}{R_m}+\mathrm{cos}^1\left(\frac{R_i}{R_m}\right)\frac{R_i}{R_m}\sqrt{1\frac{R_i^2}{R_m^2}}\right]\right),$$
(A1)
where $`R_i`$ is the radius of the circle in question and $`(R_{m1},R_m)`$ are the inner and outer radii of the bounding annulus. |
no-problem/0001/astro-ph0001300.html | ar5iv | text | # The Unique Potential of SKA Radio Observations of Gamma-Ray Bursts.
## 1 Introduction
Gamma-ray bursts (GRBs) are the strongest phenomenon seen at $`\gamma `$-ray wavelengths. Since their discovery in the 1970s these events, which emit the bulk of their energy in the $`0.11.0`$ MeV range, and whose durations span milliseconds to tens of minutes, posed one of the great unsolved problems in astrophysics. Until recently, no counterparts (quiescent as well as transient) could be found and observations did not provide a direct measurement of their distance, and thereby the true energy output was unknown by several orders of magnitude. The breakthrough came in early 1997, when the Wide Field Cameras aboard the Italian-Dutch BeppoSAX satellite allowed rapid and accurate localization of GRBs. Follow-up on these positions resulted in the discovery of X-ray , optical and radio afterglows . These observations revealed that GRBs come from ‘cosmological’ distances. GRBs are by far the most luminous photon sources in the universe, with (isotropic) peak luminosities in $`\gamma `$ rays up to $`10^{52}`$ erg/s, and total energy budgets up to several $`10^{5354}`$ erg (e.g., ). The optical signal from GRB is regularly seen to be 10 magnitudes brighter (absolute) than the brightest supernovae, and once even 18 magnitudes brighter.
Here we discuss the current status of GRB afterglow observations (Sect. 2 to 6) and discuss the unique potential of SKA observations of GRBs (Sect. 7).
## 2 Relativistic blast-wave models
GRB afterglows are in good agreement with, so called, fireball-plus-relativistic blast-wave models (see for an extensive review). The basic model is a point explosion with an energy of order $`10^{52}`$ ergs, which leads to a ‘fireball’, an optically thick radiation-electron-positron plasma with initial energy much larger than its rest mass that expands ultra-relativistically. The GRB may be due to a series of ‘internal shocks’ that develop in the relativistic ejecta before they collide with the ambient medium. When the fireball runs into the surrounding medium a ‘forward shock’ ploughs into the medium and heats it, and a ‘reverse shock’ does the same to the ejecta. As the forward shock is decelerated by increasing amounts of swept-up material it produces a slowly fading ‘afterglow’ of X rays, then ultraviolet, optical, infrared, millimetre, and radio radiation.
Confirmation of the relativistic blast-wave model. Radio light curves of the afterglow of GRB 970508 show variability on time scales of less than a day, but these dampen out after one month (see Fig. 1). Interpreting this as the effect of source expansion on the diffractive interstellar scintillation a source size of roughly 10<sup>17</sup> cm was derived, corresponding to a mildly relativistic expansion of the shell .
The first X-ray and optical (but see ) afterglows show power-law temporal decays, with power-law exponents in the range 1 to 2. These afterglow light curves agree well with the predictions of the relativistic blast-wave model (e.g., ).
The broad-band afterglow spectra are also power laws (in four distinct regions); together with the observed decrease of the cooling break and the peak frequency the observations conform nicely with simple relativistic blast-wave models in which the emission is synchrotron radiation by electrons accelerated in a relativistic shock .
The brightness temperature of the GRB 990123 optical flash exceeds the Compton limit of $`10^{12}`$ K, confirming the highly relativistic nature of the GRB source .
## 3 Progenitors and the cause of the explosion
The GRB and the afterglow are produced when relativistic ejecta are slowed down; no observable radiation emerges directly from the ‘hidden engine’ that powers the GRB. Thus, in spite of all recent discoveries the origin of GRBs remains unknown (although an important link may be provided by the possible connection of GRBs to SNe). Currently popular models for the origin of GRBs are the neutron star-neutron star and neutron star-black hole mergers, white dwarf collapse, and core collapses of very massive stars (‘failed’ supernovae or hypernovae). These models can in principle provide the required energies.
SN 1998bw/GRB 980425. Galama et al. discovered a relatively rare and bright SN of type Ic within the small BeppoSAX localization of GRB 980425, and suggested that the two objects are connected. A conservative estimate of the probability of a chance coincidence of the supernova and the GRB is $`9\times 10^5`$ . In the radio, the SN rapidly brightened and became one of the most luminous radio SNe . Kulkarni et al. drew attention to the fact that the radio emitting shell in SN 1998bw must be expanding at relativistic velocities, $`\mathrm{\Gamma }\text{ }>2`$; see Fig. 1. This relativistic shock could well have produced the GRB at early times. The consequence of accepting such an association is that the $`\gamma `$-ray peak luminosity of GRB 980425 and its total $`\gamma `$-ray energy budget are much smaller (a factor of $``$ 10<sup>5</sup>) than those of ‘normal’ GRBs. GRB 980425 is thus a member of a new class of GRBs – low luminosity GRBs related to nearby SNe (SN 1998bw is at $`z=0.0085`$). Such GRBs may well be the most most frequently occuring GRBs!
GRBs and SNe. But perhaps most, if not all, GRBs are associated with supernovae. There is growing evidence linking the usual GRBs – the cosmologically located GRBs – to SNe. The most direct evidence comes from the suggestion of an underlying SN in the afterglow of GRB 980326 . The SN is revealed readily by its distinctive UV-poor spectrum against the broad-band afterglow. Also for GRB 970228 there is evidence that a supernova dominated the light curves at late times . So there may not be a dichotomy between ‘normal’ and supernova GRBs, only a gradual transition. The relation between cosmologically located GRBs like GRBs (980326, 970228) and GRB 980425/SN 1998bw is as yet unclear.
Collapsar model. All these observational developments support the collapsar model pioneered by Woosley and collaborators (see and refs therein) in which massive stars core collapse to form black holes. Energy is somehow extracted from the spinning black hole and the jets drill their way out and power the GRBs and their afterglows.
## 4 GRBs as potential probes of the high-redshift universe
Host galaxies have been seen in most optical afterglow images. The detection of \[O II\] $`\lambda `$ 3727 and Lyman $`\alpha `$ emission from some hosts indicates that these are sites of vigorous star formation. The observed connection between some GRBs and star forming regions suggests that GRBs occur at critical phases in the evolution of massive stars. If GRBs are related to the deaths of massive stars (whose total lifetime is very short), their rate is proportional to the star formation rate (SFR). In that case GRBs may very well be at very high redshifts, with $`z6`$ or greater, for the faintest bursts (e.g. ). GRBs may therefore become a powerful tool to probe the far reaches of the universe by guiding us to regions of very early star formation, and the (proto) galaxies and (proto) clusters of which they are part. The redshifts determined so far range between $`z=0.41`$ and $`z=3.42`$.
## 5 The early afterglow
The discovery of a very bright and brief optical flash coincident in time with GRB 990123 shows that the early optical signal from GRB can be some 18 magnitudes brighter than the brightest supernovae. The reverse shock could cause emission that peaks in the optical waveband and is observed only during or just after the GRB. GRB 990123 would then be the first burst in which all three emitting regions have been seen: internal shocks causing the GRB, the reverse shock causing the prompt optical flash, and the forward shock causing the afterglow .
## 6 Strongly anisotropic outflow (beaming)
An important uncertainty concerns the possible beaming of the $`\gamma `$-ray and afterglow emissions. This has an immediate impact on the burst energetics, and the nature and number of events needed to account for the observed burst rate . If the afterglow is beamed with opening angle $`\theta `$, a change of the light curve slope occurs at the time when the Lorentz factor $`\mathrm{\Gamma }`$ of the blast wave equals $`1/\theta `$. Slightly later the jet begins a lateral expansion, which causes a further steepening of the light curve. Perhaps such a transition has been observed in the optical afterglow light curve of GRB 990123 (e.g., ). A similar transition was better sampled in afterglow data of GRB 990510; optical observations of GRB 990510, show a clear steepening of the rate of decay of the light between $``$ 3 hours and several days . Together with radio observations , which reveal a similar steepening of the decline, it is found that the transition is very much frequency-independent; this virtually excludes explanations in terms of the passage of the cooling or the peak frequency, but is what is expected in case of beaming. Harrison et al. (1999) derive a jet opening angle of $`\theta =0.08`$ radians, which for this burst would reduce the total energy in $`\gamma `$ rays to $`10^{51}`$ erg.
## 7 The unique potential of SKA observations of GRB afterglows.
* Interstellar Scintillation. As discussed in Sect. 2, the observations of GRB afterglows are in good agreement with the relativistic-blast wave model. However, the expansion rate and size of the blast wave have never been observed directly. Observations of the size, expansion rate and the shape of the GRB remnant would provide a stringent test of the relativistic blast wave model. The size $`d`$ of the GRB remnant is of the order of
$$d=\gamma ct,$$
(1)
where $`\gamma `$ is the Lorentz factor of the blast wave, $`c`$ is the speed of light and $`t`$ the time in the observer’s frame. Hence, after 1 week the source size is $`2\gamma `$ light-weeks, which, at a typical redshift of $`z1`$ corresponds to an angular size of $``$ 1 microarcsecond.
Let us take GRB 970508 as an example. Radio light curves of the afterglow of this GRB show rapid variability on time scales of less than a day, but these dampen out after one month . The reduced flux density modulations, interpreted as diffractive interstellar scintillation (DISS), are caused by the expansion of the source and then imply an angular diameter of at most a few $`\mu `$arcsec. At the redshift of GRB 970508 this corresponds to a linear diameter of roughly 10<sup>17</sup> cm corresponding to a mildly relativistic expansion of the shell . Similar estimates of the source size were derived from the observed flux density for frequencies below the self-absorption frequency , and from the presence of several breaks in the spectral energy distribution of GRB 970508 . It is clear that with the current resolution and sensitivity of earth-bound Very Large Baseline Interferometry it is going to be impossible to ever obtain a direct measurement of the source size, except for the nearest GRBs like GRB 980425/SN 1998bw.
However, the example of GRB 970508 shows that indirect source size estimates of GRB remnants can be obtained by observations of interstellar scintillation (DISS and RISS). The current observations are still severely sensitivity limited but as we discuss below there can be fantastic progress with the sensitivity provided by SKA.
Strong scattering can be observed at frequencies below the transition frequency $`\nu _0`$ (typically $`\nu _0`$ 5 GHz at high galactic latitudes ). For $`\nu >\nu _0`$, the scattering is weak and the modulations scale as $`(\nu /\nu _0)^{17/12}<1`$. For $`\nu <\nu _0`$ we will see strong diffractive scintillation only if the size of the source, $`\theta _S<\theta _D=\theta _{F_0}(\nu /\nu _0)^{6/5}`$. Depending on the properties of the turbulent plasma screen we may expect to encounter such conditions in the first day(s) after the GRB event.
The other ISS parameters of interest are the decorrelation timescale $`t_{\mathrm{diff}}`$ (time for significant changes in the detected flux), and the bandwidth over which the diffractive ISS is decorrelated, $`\mathrm{\Delta }\nu =\nu _0(\nu /\nu _0)^{22/5}`$. These scale as: $`t_{\mathrm{diff}}\nu ^{1.2}`$ and $`\mathrm{\Delta }\nu _\mathrm{d}\nu ^{4.4}`$. As emphasized by Goodman all these observables carry independent information on the properties of the ISM.
Observations of the modulation index, decorrelation time scale and decorrelation bandwidth as a function of frequency, and the determination of the transition frequency $`\nu _0`$ between weak and strong scattering, in early GRB afterglows, hence provide a wealth of information on the dynamically changing size and shape of the source. In principle the scintillation tool allows us to infer a crude measure of the morphology of the radio source: is it ring-like, as in spherical blast-wave models, or jet-like, in currently popular models. In the decaying, sub- or non-relativistic phase, the source may develop double structure (approaching and receding jet components) leading to distinct patterns in the dynamically changing spectra.
The scintillation method, however, will need a proper calibration before it will release its potential. This is clearly shown by the inferred sizes of two recently discovered scintillating quasars. For the radio quasar PKS 0405-385 an angular size of $``$ 5 $`\mu `$arcsec was estimated. However, for J1819+3845, Dennett-Thorpe and de Bruyn estimate a size of $``$ 25 $`\mu `$arcsec at 5 GHz (possibly related to relatively nearby plasma turbulence). The calibration of scintillation screen properties can be provided by the observation of angularly nearby pulsars, the perfect point sources, of which SKA could easily detect about 1 per square degree.
SKA will provide the right technical specifications for these exciting and unique observations. The proposed instantaneous bandwidth: $`0.5+f/5`$ GHz and the large number of spectral channels: 10<sup>4</sup>, allow the recording of dynamic spectra of GRBs, as is now common for pulsars.
* Synchrotron self absorption. The large instantaneous bandwidth, the large number of spectral channels, and sensitiviy of SKA will also allow detailed study of the transition from optically thin to optically thick frequencies and the evolution of the shape and location of this transition. Such high quality observations will provide strong constraints on models of GRB afterglows.
* Supernova-GRBs. As discussed in Sect. 3 the most common GRBs may very well be the low luminosity GRBs like GRB 980425/SN 1998bw. The bright radio emission of such sources may easily be detected out to redshifts of $`z1`$, with SKA’s sensitivity. From the duration of the radio phase in SN 1998bw, its typical brightness of $``$ 10 mJy at GHz frequencies, and some simple order-of-magnitude extrapolations to the whole sky, we derive that at any given time there will be several tens of such fading supernova-GRB radio afterglows above a flux density of 1 $`\mu `$Jy per square degree! Most of these sources will be distant, hence small and scintillating. This is how they could be discerned. What distinguishes them from AGN, however, is that they would appear at places where previously there would have been no radio source. A survey of the sky, carried out with SKA in its first year, could provide the template, against which to pick up these new sources. (Nota bene, this is also how new radio SNe would be discovered, but they are typically orders of magnitude fainter and rarely reach the radio brightness temperatures of GRB afterglows).
* Rapid response. The early radio afterglow emission of GRBs is currently hard to observe: the sources are very faint at ages less than 1 day and the current response time, which is typically a few hours, is too slow. Soon, with the launch of the gamma-ray satellite HETE-II, the response time may be very much improved upon (positions will be available within tens of seconds of the event), but the sensitivity is expected to remain problematic for such early observations. SKA will provide the required sensitivity plus it will have this other unique capability: the possibility of very rapid response (we may, however, have to build a few SKA’s to cover both the northern and southern hemispheres, and provide 24h watch!). We envision that SKA may be triggered directly by future gamma-ray spacecraft, and then rapidly electronically steer to the location on the sky. Such observations may provide insight in the physics of the fireball at very early times, for example we may expect to detect emission from the reverse shock .
* Wide Field Surveys. SKA will be a unique instrument for surveying large areas of sky (as noted above in connection with the GRB 980425/SN 1988bw association). If the GRB luminosity function is bimodal (i.e. GRBs like 980425 versus the more distant GRBs like 970508) then such surveys may be dominated by supernova-GRBs. However, a substantial number of the more distant GRBs may be discovered too, depending on the amount of collimation into jets. If the GRB luminosity function is not bimodal, but very wide then we may also detect a substantial number of radio afterglows from intermediate luminosity GRBs ($`E_\gamma 10^{4850}`$ erg). Thus, thousands of GRBs and supernova-GRBs may be discovered from such surveys by their distinct observational characteristics (for example a self-absorbed synchrotron spectrum, observed ISS and the characteristic damping of ISS fluctuations with time due to expansion of the source, the fact that GRB afterglows are not expected to be recurrent, etc). An important aspect of a radio selected survey is its unbiased-to-dust nature. Afterglows may thus be discovered independent of an optical or even a GRB identification.
The statistics of radio afterglows may be compared with the numbers expected from specific spherical- or jet-fireball models. As the bulk Lorentz factor, $`\mathrm{\Gamma }`$ decreases with time after the event, the observer sees more and more of the emitting surface, $`\theta 1/\mathrm{\Gamma }`$. It follows that if gamma ray bursts are highly collimated, many more radio transients should be observed without associated gamma rays than with them. The ratio of expected (assuming spherical symmetry) to observed number of afterglows is thus a direct measure of the amount of collimation in GRB fireballs.
The number of radio afterglows may also be compared to that of optical afterglows (subject to obscuration by dust) to reveal possibly dusty environments (expected for massive star progenitor models).
* The high-$`z`$ universe and the cosmic star formation/death history. A GRB at a redshift $`z1`$ may easily be a mJy bright. With SKA’s senstivity such radio counterparts can be detected out to redshifts of 10 or greater. SKA will also be sufficiently sensitive that the radio emission of GRB hosts can be studied. Currently this is barely feasible. Such observations may provide information on the progenitors of GRBs (e.g., young or old stellar populations). Also, as discussed in Sect. 4 GRBs are expected to trace the star formation rate (SFR) in the universe. Vice versa by observing GRB hosts we will learn about the star formation (and star death!) history. Observations in the (sub-)mm band suggest that star formation at high redshift is dominated by dusty star burst galaxies . Estimates of the SFR of GRB host galaxies can be accurately determined by radio observations. These observations will be insensitive to dust, a crucial fact if one wants to be complete.
## References |
no-problem/0001/astro-ph0001167.html | ar5iv | text | # MILLISECOND OSCILLATIONS IN X-RAY BINARIES
## 1 Introduction
Submitted to the Annual Review of Astronomy and Astrophysics; to appear September 2000
The principal motivation for studying accreting neutron stars and black holes is that these objects provide a unique window on the physics of strong gravity and dense matter. One of the most basic expressions of the compactness of these compact objects is the short (0.1–1 msec) dynamical time scale characterizing the motion of matter under the influence of gravity near them. Millisecond variability will naturally occur in the process of accretion of matter onto a stellar-mass compact object, an insight that dates back to at least Shvartsman (1971). For example, hot clumps orbiting in an accretion disk around black holes and neutron stars will cause quasi-periodic variability on time scales of about a millisecond (Sunyaev 1973). Accreting low-magnetic field neutron stars will reach millisecond spin periods, which can be detected when asymmetric emission patterns form on the star’s surface during X-ray bursts (Radhakrishnan & Srinivasan 1984, Alpar et al. 1982; Shara 1982, Livio & Bath 1982, see also Joss 1978). These early expectations have finally been verified in a series of discoveries with NASA’s Rossi X-Ray Timing Explorer (RXTE; Bradt et al. 1993) within 2.5 years after launch on 30 December 1995.
In this review, I discuss these newly discovered phenomena and the attempts to use them to perform measurements of fundamental physical interest. I concentrate on millisecond oscillations, periodic and quasi-periodic variations in X-ray flux with frequencies exceeding 10<sup>2.5</sup> Hz, but I also discuss their relations to slower variability and X-ray spectral properties. Millisecond oscillations have so far been seen nearly exclusively from low-magnetic-field neutron stars, so these will be the focus of this review, although I shall compare their phenomenology to that of the black-hole-candidates (§6-7).
Accreting neutron stars and black holes occur in X-ray binaries (e.g., Lewin et al. 1995a). In these systems matter is transferred from a normal (’donor’) star to a compact object. Thermal X-rays powered by the gravitational potential energy released are emitted by the inner regions of the accretion flow and, if present, the neutron star surface. For a compact object with a size of order 10<sup>1</sup> km, 90% of the energy is released in the inner $``$10<sup>2</sup> km. It is with this inner emitting region that we shall be mostly concerned here. Because accreting low-magnetic-field neutron stars are mostly found in low-mass X-ray binaries (in which the donor star has a mass of $`<`$1M) these systems will be the ones we focus on.
The mass transfer usually occurs by way of an accretion disk around the compact object. In the disk the matter moves in near-Keplerian orbits, i.e., with an azimuthal velocity that is approximately Keplerian and a radial velocity much smaller than this. The disk has a radius of 10<sup>5-7</sup> km, depending on the binary separation. The geometry of the flow in the inner emitting regions is uncertain. In most models for accretion onto low-magnetic-field neutron stars (e.g., Miller et al. 1998a) at least part of the flow extends down into the emitting region in the form of a Keplerian disk. It is terminated either at the radius $`R`$ of the star itself, or at a radius $`r_{in}`$ somewhat larger than $`R`$, by for example the interaction with a weak neutron-star magnetic field, radiation drag, or relativistic effects. Within $`r_{in}`$ the flow is no longer Keplerian and may or may not be disk-like. Both inside and outside $`r_{in}`$ matter may leave the disk and either flow in more radially or be expelled. Particularly for black holes advective flow solutions are discussed where the disk terminates and the flow becomes more spherical at a much larger radius (e.g., Narayan 1997).
Whatever the geometry, it is clear that as the characteristic velocities near the compact object are of order $`(GM/R)^{1/2}0.5c`$, the dynamical time scale, the time scale for the motion of matter through the emitting region, is short; $`\tau _{dyn}(r^3/GM)^{1/2}`$$``$0.1 ms for $`r`$=10 km, and $``$2 ms for $`r`$=100 km near a 1.4M neutron star, and $``$1 ms at 100 km from a 10M black hole. So, the significance of millisecond X-ray variability from X-ray binaries is clear: milliseconds is the natural time scale of the accretion process in the X-ray emitting regions, and hence strong X-ray variability on such time scales is nearly certainly caused by the motion of matter in these regions. Orbital motion, neutron-star spin, disk- and neutron-star oscillations are all expected to happen on these time scales.
The inner flow is located in regions of spacetime where strong-field general-relativity is required to describe the motion of matter. For that reason one expects to detect strong-field general-relativistic effects in these flows, such as for example the existence of a region where no stable orbits are possible. The precise interactions between the elementary particles in the interior of a neutron star which determine the equation of state (EOS) of supra-nuclear-density matter are not known. Therefore we can not confidently predict the radius of a neutron star of given mass, or the maximum spin rate or mass of neutron stars (e.g., Cook et al. 1994). So, by measuring these macroscopic quantities one constrains the EOS and tests basic ideas about the properties of elementary particles. In summary, the main motivation for studying millisecond variations in X-ray binaries is that their properties depend on untested, or even unknown, properties of spacetime and matter.
Three different millisecond phenomena have now been observed in X-ray binaries. Historically, the first to be discovered were the twin kilohertz quasi-periodic oscillations (kHz QPOs), widely interpreted now as due to orbital motion in the inner accretion flow. Then came the burst oscillations, probably due to the spin of a layer in the neutron star’s atmosphere in near-corotation with the neutron star itself. Finally RXTE detected the first true spin frequency of an accreting low-magnetic field neutron star, the long-anticipated accreting millisecond pulsar.
In this review, I first examine the millisecond pulsar (§3), then the burst oscillations (§4) and finally the kHz QPOs (§5). We will thus be venturing from the (relatively) well-understood accreting pulsars via the less secure regions of what happens in detail on a neutron star’s surface during the thermonuclear runaway that is an X-ray burst, into the mostly uncharted territory of the innermost accretion flows around neutron stars and black holes, which obviously is “where the monsters are”, but also where the greatest rewards wait. The possibly related phenomena found in black-hole candidates (§6.1) and at lower frequencies (§6.3) are discussed next, and in §7 the kHz QPO models are summarized.
## 2 Techniques
Most of the variability measurements discussed here rely on Fourier analysis of X-ray count-rate time series with sub-millisecond time resolution (van der Klis 1989b). A quasi-periodic oscillation (QPO) in the time series stands out in the power spectrum (the square of the Fourier transform) as a broad, usually Lorentzian peak (in Fig. 5 several of such peaks can be seen), characterized by its frequency $`\nu `$ (“centroid frequency”), width $`\lambda `$ (inversely proportional to the coherence time of the oscillation) and strength (the peak’s area is proportional to the variance of the QPO signal). The variance is nearly always reported in terms of the root-mean-square of the signal expressed as a fraction of the count rate, the “fractional rms amplitude” $`r`$; the coherence often in terms of a quality factor $`Q=\nu /\lambda `$. Conventionally, to call a local maximum in a power spectrum a QPO peak one requires $`Q>2`$. Time delays between signals simultaneously detected in different energy bands are usually measured using cross-spectra (the frequency-domain equivalent of the cross-correlation function; van der Klis et al. 1987, Vaughan et al. 1994a, Nowak et al. 1998) and often expressed in terms of a phase lag (time lag multiplied by frequency).
The signal-to-noise of a broad power-spectral feature is $`n_\sigma =\frac{1}{2}I_xr^2(T/\lambda )^{1/2}`$ (van der Klis 1989b, see van der Klis 1998 for more details), where $`I_x`$ is the count rate and $`T`$ the observing time (assumed $`1/\lambda `$). Note that $`n_\sigma `$ is proportional to the count rate and to the signal amplitude squared, so that it is sufficient for the amplitude to drop by 50% for the signal-to-noise to go from, e.g., a whopping 6$`\sigma `$ to an undetectable 1.5$`\sigma `$ – i.e., if a power-spectral feature “suddenly disappears” it may have only decreased in amplitude by a factor of two.
## 3 Millisecond pulsations
An accreting millisecond pulsar in a low-mass X-ray binary has sometimes been called the “Holy Grail” of X-ray astronomy. Its discovery was anticipated for nearly 20 years, because magnetospheric disk accretion theory as well as evolutionary ideas concerning the genesis of millisecond radio pulsars strongly suggested that such rapid spin frequencies must occur in accreting low-magnetic field neutron stars (see Bhattacharya & van den Heuvel 1991). However, in numerous searches of X-ray binary time series (e.g., Leahy et al. 1983, Mereghetti & Grindlay 1987, Wood et al. 1991, Vaughan et al. 1994b) such rapid pulsars did not turn up.
More than two years after RXTE’s launch, the first, and as of this writing only accreting millisecond pulsar was finally discovered on April 13, 1998 in the soft X-ray transient SAX J1808.4$``$3658 (Fig. 1; Wijnands & van der Klis 1998a,b). The pulse frequency is 401 Hz, so this is a 2.5 millisecond pulsar. The object is nearly certainly the same as the transient that burst out at the same position in September 1996 and gave the object its name (in ’t Zand et al. 1998). As this transient showed two type 1 X-ray bursts, SAX J1808.4$``$3658 is also the first genuine bursting pulsar, breaking the long-standing rule (e.g., Lewin & Joss 1981) rule that pulsations and type 1 X-ray bursts are mutually exclusive.
The orbital period of this pulsar is 2 hrs (Fig. 2; Chakrabarty & Morgan 1998a,b). With a projected orbital radius $`a\mathrm{sin}i`$ of only 63 light milliseconds and a mass function of 3.8 10<sup>-5</sup>M, the companion star is either very low mass, or we are seeing the orbit nearly pole-on. The amplitude of the pulsations varied between 4 and 7% and showed little dependence on photon energy (Cui et al. 1998b). However, the pulsations as measured at higher photon energies preceded those measured in the 2-3 keV band by a gradually increasing time interval from 20 $`\mu `$sec (near 3.5 keV) up to 200 $`\mu `$sec (between 10 and 25 keV) (Cui et al. 1998b). These lags could be caused by Doppler shifting of emission from the pulsar hot spots, higher-energy photons being emitted earlier in the spin cycle as the spot approaches the observer (Ford 1999).
An accreting magnetized neutron star spinning this fast must have a weak magnetic field. If not, the radius of the magnetosphere $`r_M`$ would exceed the corotation radius, and matter corotating in the magnetosphere would not be able to overcome the centrifugal barrier. A simple estimate leads to upper limits on $`r_M`$ of 31 km, and on the surface field strength $`B`$ of 2–6 10<sup>8</sup> Gauss (Wijnands & van der Klis 1998b). A similarly simple estimate, involving in addition the requirement that for pulsations to occur $`r_M`$ must be larger than the radius $`R`$ of the neutron star, would set a strong constraint on the star’s mass-radius relation (Burderi & King 1998) and hence on the EOS (see also Li et al. 1999a). However, the process of accretion onto a neutron star with such a low $`B`$ is not identical to that in classical, 10<sup>12</sup>-Gauss accreting pulsars. In particular, the disk model (used in calculating $`r_M`$) is different this close to the neutron star, the disk-star boundary layer may be different, and multipole components in the magnetic field become important. Conceivably a classical magnetosphere does not even form and the 4–7%-amplitude pulsation occurs due to milder effects of the magnetic field on either the flow or the emission. Psaltis & Chakrabarty (1999) discuss these issues and conclude that $`B`$ is (1–10) 10<sup>8</sup> Gauss, which puts the source right among the other, rotation-powered, millisecond pulsars (the msec radio pulsars; Fig. 2). When the accretion shuts off sufficiently for the radio pulsar mechanism to operate, the system will likely show up as a radio pulsar. This should happen at the end of the system’s life as an X-ray binary, i.e., SAX J1808.4$``$3658 is indeed the long-sought millisecond radio-pulsar progenitor, but might also occur in between the transient outbursts (Wijnands & van der Klis 1998b). So far, radio observations have not detected the source in X-ray quiescence (Gaensler et al. 1999).
It is not clear what makes the neutron star spin detectable in SAX J1808.4$``$3658 and not (so far) in other low-mass X-ray binaries of similar and often much higher flux. Perhaps a peculiar viewing geometry (e.g., a very low inclination of the binary orbit) allows us to see the pulsations only in this system, although possible X-ray and optical modulations with binary phase make an inclination of zero unlikely (Chakrabarty & Morgan 1998b, Giles et al. 1999).
With a neutron star spin frequency that is certain and good estimates of $`r_M`$ and $`B`$, SAX J1808.4$``$3658 can serve as a touchstone in studies of low-mass X-ray binaries. Although no burst oscillations (§4) or kHz QPOs (§5) have been detected from the source, their absence is consistent with what would be expected from a standard LMXB in the same situation (Wijnands & van der Klis 1998c), and in more intensive observations during a next transient outburst such phenomena could be detected. This would strongly test the main assumptions underlying the models for these phenomena. The X-ray spectral properties (Heindl & Smith 1998, Gilfanov et al. 1998) and the slower types of variability (Wijnands & van der Klis 1998c) of the source are very similar to those of other LMXBs at low accretion rate suggesting that either the neutron stars in those systems have similar $`B`$, or the presence of a small magnetosphere does not affect spectral and slow-variability characteristics.
## 4 Burst oscillations
Type 1 X-ray bursts are thermonuclear runaways in the accreted matter on a neutron-star surface (Lewin et al. 1995b for a review). When density and temperature in the accumulated nuclear fuel approach the ignition point, the matter ignites at one particular spot, from which a nuclear burning front then propagates around the star (Bildsten 1998b for a review). This leads to a burst of X-ray emission with a rise time of typically $`<`$1 s, and a 10<sup>1</sup>–10<sup>2</sup> s exponential decay due to cooling of the neutron-star atmosphere. The total amount of energy emitted is 10<sup>39-40</sup> erg. In some bursts the Eddington critical luminosity is exceeded and atmospheric layers are lifted off the star’s surface, leading to an increase in photospheric radius of $``$10<sup>1</sup>–10<sup>2</sup> km, followed by a gradual recontraction. These bursts are called “radius expansion bursts”.
In the initial phase, when the burning front is spreading, the energy generation is inherently very anisotropic. The occasional occurrence of multiple bursts closely spaced in time indicates that not all available fuel is burned up in each burst, suggesting that in some bursts only part of the surface participates. Magnetic fields and patchy burning (Bildsten 1995) could also lead to anisotropic emission during X-ray bursts. Anisotropic emission from a spinning neutron star leads to periodic or quasi-periodic observable phenomena, because due to the stellar rotation the viewing geometry of the brighter regions periodically varies (unless the pattern is symmetric around the rotation axis). Searches for such periodic phenomena during X-ray bursts were performed by various groups (Mason et al. 1980, Skinner et al. 1982, Sadeh et al. 1982, Sadeh & Livio 1982a,b, Murakami et al. 1987, Schoelkopf & Kelley 1991, Jongert & van der Klis 1996), but claims of detections remained unconfirmed.
The first incontestable type 1 burst oscillation was discovered with RXTE in a burst that occurred on 1996 February 16 in the reliable burst source 4U 1728$``$34. An oscillation with a slightly drifting frequency near 363 Hz was evident in a power spectrum of 32 s of data starting just before the onset of the burst (Strohmayer et al. 1996a,b,c; Fig. 3). The oscilation frequency increased from 362.5 to 363.9 Hz in the course of $``$10 s.
Burst oscillations have now been detected in six (perhaps seven) different sources (Table 1). They do not occur in each burst, and some burst sources have not so far shown them at all. Sometimes the oscillations are strong for less than a second during the burst rise, then become weak or undetectable, and finally occur for $``$10 s in the burst cooling tail. This can happen even in radius expansion bursts, where after the photosphere has recontracted the oscillations (re)appear (Smith et al. 1997, Strohmayer et al. 1997a, 1998a). Some oscillations are seen only in the burst tail and not in the rise (Smith et al. 1997).
Usually, the frequency increases by 1–2 Hz during the burst tail, converging to an “asymptotic frequency” which in a given source tends to be stable ((Strohmayer et al. 1998c; Fig. 3), with differences from burst to burst of $``$0.1% (Table 1). This is of the order of what would be expected from binary orbital Doppler shifts and strengthens an interpretation in terms of the neutron star spin. Perhaps the orbital radial velocity curve can be detected in the asymptotic frequencies, but it remains to be seen if the asymptotic frequencies are intrinsically stable enough for this. Exceptions from the usual frequency evolution pattern do occur. Oscillations with no evidence for frequency evolution were observed in KS 1731$``$26 (Smith et al. 1997) and 4U 1743$``$29 (Strohmayer et al. 1997a), and in 4U 1636$``$53 a decrease in frequency was seen in a burst tail (Miller 1999b, Strohmayer 1999; Fig. 4).
In a widely (but not universally; §7.2) accepted scenario, the burst oscillations arise due to a hot spot or spots in an atmospheric layer of the neutron star rotating slightly slower than the star itself because it expanded by 5–50 m in the X-ray burst but conserved its angular momentum (Strohmayer et al. 1997a and references therein, Bildsten 1998b, Strohmayer 1999, Strohmayer & Markwardt 1999, Miller 1999b). The frequency drifts are caused by spin-up of the atmosphere as it recontracts in the burst decay. The asymptotic frequency corresponds to a fully recontracted atmosphere and is closest to the true neutron star spin frequency. From this scenario one expects a frequency drop during the burst rise, but no good evidence has been found for this as yet (Strohmayer 1999). The case of a frequency drop in the burst tail (Fig. 4) is explained by invoking additional thermonuclear energy input late in the burst, which also affects the burst profile (Strohmayer 1999).
If the oscillations are due to a stable pattern in the spinning layer, then it should be possible to describe them as a frequency-modulated, strictly coherent signal. By applying a simple exponential model to the frequency drifts, it is possible to establish coherences of up to Q$``$4000 (Strohmayer & Markwardt 1999, see also Zhang et al. 1998c, Smith et al. 1997, Miller 1999a,b). However, this is still $``$20% less than a fully coherent signal of this frequency and duration, i.e., it has not yet been possible to count the exact number of cycles in the way this can be done in a pulsar. Possibly, exact coherence recovery is feasible for these signals, but current signal-to-noise limits prevent to accomplish this as the exact frequency drift ephemeris can not be found.
The harmonic content of the oscillations is low. In 4U 1636$``$53 it is just possible, by combining data from the early stages of several bursts, to detect a frequency near 290 Hz, half the dominant one (Miller 1999a,b), suggesting that $``$290 Hz, not $``$580 Hz is the true spin frequency and that two antipodal hot spots produce the burst oscillation in this source (whose kHz peak separation is $``$250 Hz, §5). No harmonics or “subharmonics” have been seen in other sources, or in the burst tails of any source (Strohmayer & Markwardt 1999), with the possible exception of the marginal detections in the Rapid Burster (Fox et al. 1999).
The oscillation amplitudes range from $``$50% (rms) of the burst flux early in some bursts to between 2 and 20% (rms) in the tail (references see Table 1). (Note, that sometimes sinusoidal amplitudes are reported, which are a factor $`\sqrt{2}`$ larger than rms amplitudes, and that amplitudes are expressed as a fraction of burst flux, total flux minus the persistent flux before the burst. Early in the rise, when burst flux is low compared to total flux, the measured amplitude is multiplied by a large factor to convert it to burst flux fraction.) In KS 1731$``$260 (Smith et al. 1997) and 4U 1743$``$29 (Strohmayer et al. 1997a) the photon energy dependence of the oscillations was measured. Above 7 or 8 keV the amplitude was 9–18% while below that energy it was undetectable at $`<`$2–4%. 4U 1636$``$53 shows a slight variation in spectral hardness as a function of oscillation phase (Strohmayer et al. 1998). In Aql X-1 photons below 5.7 keV lag those at higher energies by roughly 0.3 msec, which may be caused by Doppler shifts (Ford 1999).
If the burst oscillations are due to hot spots on the surface, then their amplitude constrains the “compactness” of the neutron star, defined as $`R_G/R`$ where $`R`$ is the star’s radius and $`R_G=GM/c^2`$ its gravitational radius (Strohmayer et al. 1997b, 1998a, Miller & Lamb 1998). The more compact the star, the lower the oscillation amplitude, as gravitational light bending increasingly blurs the beam. In particular when the oscillations are caused by two antipodal hot spots (cf. 4U 1636$``$53, above, and possibly other sources, §5.5), and the amplitudes are high, the constraints are strong. The exact bounds on the compactness depend on the emission characteristics of the spots, and no final conclusions have been reached yet.
Modeling of the spectral and amplitude evolution of the oscillations through the burst in terms of an expanding, cooling hot spot has been succesful (Strohmayer 1997b). However, several issues with respect to this attractively simple interpretation have yet to be resolved. The presence of two burning sites as required by the description in terms of two antipodal hot spots could be related to fuel accumulation at the magnetic poles (Miller 1999a), but their simultaneous ignition seems not easy to accomplish, and obviously, in view of the frequency drifts, these sites must decouple from the magnetic field after ignition. The hot spots must survive the strong shear in layers that, from the observed phase drifts, must revolve around the star several times during the lifetime of the spots, and they must even survive through photospheric radius expansion by factors of at least several during radius-expansion bursts (which probably implies the roots of the hot spots are below the photospheric layers).
The resolution of these issues ties in with questions such as why only some burst sources show the oscillations, and why only some bursts exhibit them. This is hard to explain in a magnetic-pole accumulation scheme as the viewing geometry of the poles remains the same from burst to burst. Studies of the relation between the characteristics of the bursts and the surrounding persistent emission and the presence and character of the burst oscillations could help to shed light on these various questions.
## 5 Kilohertz quasi-periodic oscillations
The kilohertz quasi-periodic oscillations (kHz QPOs) were discovered at NASA’s Goddard Space Flight Center in February 1996, just two months after RXTE was launched (in Sco X-1: van der Klis et al. 1996a,b,c, and 4U 1728$``$34: Strohmayer et al. 1996a,b,c; see van der Klis 1998 for a historical account). Two simultaneous quasi-periodic oscillation peaks (“twin peaks”) in the 300–1300 Hz range and roughly 300 Hz apart (Fig. 5) occur in the power spectra of low-mass X-ray binaries containing low-magnetic-field neutron stars of widely different X-ray luminosity L<sub>x</sub>. The frequency of both peaks usually increases with X-ray flux (§5.4). In 4U 1728$``$34 the separation frequency of the two kHz peaks is close to $`\nu _{burst}`$4; Strohmayer et al. 1996b,c). This commensurability of frequencies provides a powerful argument for a beat-frequency interpretation (§5.1, 5.5, 7.1).
### 5.1 Orbital and beat frequencies
Orbital motion around a neutron star occurs at a frequency of
$$\nu _{orb}=\left(\frac{GM}{4\pi ^2r_{orb}^3}\right)^{1/2}1200\text{Hz}\left(\frac{r_{orb}}{15\text{km}}\right)^{3/2}m_{1.4}^{1/2},$$
and the corresponding orbital radius is
$$r_{orb}=\left(\frac{GM}{4\pi ^2\nu _{orb}^2}\right)^{1/3}15\text{km}\left(\frac{\nu _{orb}}{1200\text{Hz}}\right)^{2/3}m_{1.4}^{1/3},$$
where $`m_{1.4}`$ the star’s mass in units of 1.4M (Fig. 6). In general relativity, no stable orbital motion is possible within the innermost stable circular orbit (ISCO), $`R_{ISCO}=6GM/c^212.5m_{1.4}\text{km}`$. The frequency of orbital motion at the ISCO, the highest possible stable orbital frequency, is $`\nu _{ISCO}(1580/m_{1.4})\text{Hz}`$.
These expressions are valid for a Schwarzschild geometry, i.e., outside a non-rotating spherically symmetric neutron star (or black hole). Corrections to first order in $`j=cJ/GM^2`$, where $`J`$ is the neutron-star angular momentum have been given by, e.g., Miller et al. (1998a) and can be several 10%. For more precise calculations see Morsink & Stella (1999).
In spin-orbit beat-frequency models some mechanism produces an interaction of $`\nu _{orb}`$ at some preferred radius in the accretion disk with the neutron star spin frequency $`\nu _s`$, so that a beat signal is seen at the frequency $`\nu _{beat}=\nu _{orb}\nu _s`$. As $`\nu _{beat}`$ is the frequency at which a given particle orbiting in the disk overtakes a given point on the spinning star, it is the natural disk/star interaction frequency. In such a rotational interaction (with spin and orbital motion in the same sense) no signal is produced at the sum frequency $`\nu _{orb}+\nu _s`$ (it is a “single-sideband” interaction).
### 5.2 Early interpretations
It was immediately realized that the observed high frequencies of kHz QPOs could arise in orbital motion of accreting matter very closely around the neutron star, or in a beat between such orbital motion and the neutron-star spin (van der Klis et al. 1996a, Strohmayer et al. 1996a; some of the proposals to look for such rapid QPOs with RXTE had in fact anticipated this). A magnetospheric spin-orbit beat-frequency model (Alpar & Shaham 1985, Lamb et al. 1985) was already in use in LMXBs for slower QPO phenomena (§6.3) so when three commensurable frequencies were found in 4U 1728$``$34, a beat-frequency interpretation was immediately proposed (Strohmayer et al. 1996c). Let us call $`\nu _2`$ the frequency of the higher-frequency (the “upper”) peak and $`\nu _1`$ that of the lower-frequency kHz peak (the “lower peak”). Then the beat-frequency interpretation asserts that $`\nu _2`$ is $`\nu _{orb}`$ at some preferred radius in the disk, and $`\nu _1`$ is the beat frequency between $`\nu _2`$ and $`\nu _s`$, so $`\nu _1=\nu _{beat}=\nu _{orb}\nu _s\nu _2\nu _{burst}`$, where the approximate equality follows from $`\nu _{burst}\nu _s`$4). That only a single sideband is observed is a strong argument for a rotational interaction (§5.1). Later this was worked out in detail in the form of the “sonic point beat-frequency model” by Miller et al. (1996, 1998a; §7.1), where the preferred radius is the sonic radius (essentially, the inner edge of the Keplerian disk).
The beat-frequency interpretation implies that the observed kHz QPO peak separation $`\mathrm{\Delta }\nu =\nu _2\nu _1`$ should be equal to the neutron star spin frequency $`\nu _s`$, and should therefore be constant and nearly equal to $`\nu _{burst}`$. As we shall see in §5.5, it turned out that $`\mathrm{\Delta }\nu `$ is not actually exactly constant nor precisely equal to $`\nu _{burst}`$ (in §7 we examine how a beat-frequency interpretation could deal with this), and this triggered the development of other models for the kHz QPOs. Stella & Vietri (1998) noted that the frequency $`\nu _{LF}`$ of low-frequency QPOs (§6.3) in the 15–60 Hz range that had been known in the Z sources since the 1980’s (cf. van der Klis 1995), and that were being discovered with RXTE in the atoll sources as well is approximately proportional to $`\nu _{2}^{}{}_{}{}^{2}`$. This triggered a series of papers (Stella & Vietri 1998, 1999, Stella et al. 1999b) together describing what is now called the “relativistic precession model”, where $`\nu _2`$ is the orbital frequency at some radius in the disk and $`\nu _1`$ and $`\nu _{LF}`$ are frequencies of general-relativistic precession modes of a free particle orbit at that radius (§7.2). For a further discussion of kHz QPO models see §7. With the exception of the photon bubble model (Klein et al. 1996b; §7.3), all models are based on the interpretation that one of the kHz QPO frequencies is an orbital frequency in the disk.
### 5.3 Dependence on source state and type
Twenty sources have now<sup>1</sup><sup>1</sup>11999 October 15 shown kHz QPOs. Sometimes only one peak is detectable, but 18 of these sources have shown two simultaneous kHz peaks; the exceptions with only a single peak are the little-studied XTE J1723$``$376 and EXO 0748$``$676. Tables 2 and 3 summarize the results. There is a remarkable similarity in QPO frequencies and peak separations across a great variety of sources.
However, at a more detailed level there turn out to be differences between the different source types. The two main types are the Z sources and the atoll sources (Hasinger & van der Klis 1989, see van der Klis 1989a, 1995a,b for reviews). Z sources are named after the roughly Z-shaped tracks they trace out in X-ray color-color and hardness-intensity diagrams on a time scale of hours to days (Fig. 7). They are the most luminous LMXBs, with X-ray luminosity L<sub>x</sub> near the Eddington luminosity L<sub>Edd</sub>. Atoll sources produce tracks roughly like a wide U or C (e.g., Fig. 7) which are somewhat reminiscent of a geographical map of an atoll because in the left limb of the U the motion through the diagram becomes slow, so that the track is usually broken up by observational windowing into “islands”. The bottom and right-hand parts of the U are traced out in the form of a curved “banana” branch on a time scale of hours to a day. In a given source, the islands correspond to lower flux levels than the banana branch. Most atoll sources are in the 0.01–0.2 L<sub>Edd</sub> range; a group of four bright ones (GX 3+1, GX 9+1, GX 13+1 and GX 9+9) that is nearly always in the banana branch is more luminous than this, perhaps 0.2–0.5 L<sub>Edd</sub> (the distances are uncertain). Most timing and spectral characteristics of these sources depend in a simple way on position along the Z or atoll track. So, the phenomenology is essentially one-dimensional. A single quantity, usually referred to as “inferred accretion rate”, varying on time scales of hours to days on the Z track and the banana branch, and more slowly in the island state must govern most of the phenomenology (but see §5.4).
In all Z sources and in 4U 1728$``$34 the kHz QPOs are seen down to the lowest inferred $`\dot{M}`$ levels these sources reach. The QPOs always become undetectable at the highest $`\dot{M}`$ levels. In the atoll sources, where the count rates are higher at higher inferred $`\dot{M}`$, this can not be a sensitivity effect. In most atoll sources, the QPOs are seen in the part of the banana branch closest to the islands, i.e., near the lower left corner of the U (Fig. 7); that they are often not detected in the island state may be related to low sensitivity at the low count rates there, but in one island in 4U 0614+09 the undetected lower kHz peak is really much weaker than at higher inferred $`\dot{M}`$ (Méndez et al. 1997). No kHz QPOs have been seen in the four bright atoll sources (Wijnands et al. 1998d, Strohmayer 1998, Homan et al. 1998a), perhaps because they do not usually reach this low part of the banana branch, and in several faint LMXBs, probably also atoll sources in the island state (SAX J1808.4$``$3658; Wijnands & van der Klis 1998c and §3, XTE J1806$``$246; Wijnands & van der Klis 1999b, SLX 1735$``$269; Wijnands & van der Klis 1999c, 4U 1746$``$37; Jonker et al. 1999a, 4U 1323$``$62; Jonker et al. 1999b, 1E 1724$``$3045, SLX 1735$``$269 and GS 1826$``$238; Barret et al. 1999).
The QPO frequencies increase when the sources move along the tracks in the sense of increasing inferred $`\dot{M}`$ (this has been seen in more than a dozen sources, and no counterexamples are known). On time scales of hours to a day increasing inferred $`\dot{M}`$ usually corresponds to increasing X-ray flux, so on these time scales kHz QPO frequency usually increases with flux. When flux systematically decreases with inferred $`\dot{M}`$, as is the case in some parts of these tracks, the frequency is expected to maintain its positive correlation with inferred $`\dot{M}`$ and hence become anticorrelated to flux, and indeed this has been observed in the Z source GX 17+2 (Wijnands et al. 1997b).
So, the kHz QPOs fit well within the pre-existing Z/atoll description of LMXB phenomenology in terms of source types and states, including the fact that position on the tracks in X-ray color-color or hardness-intensity diagrams (“inferred $`\dot{M}`$”, see §5.4) and not X-ray flux drives the phenomenology.
### 5.4 Dependence of QPO frequency on luminosity and spectrum
Kilohertz QPOs occur at similar frequency in sources that differ in X-ray luminosity $`L_x`$ by more than 2 orders of magnitud, and the kHz QPO frequency $`\nu `$ seems to be determined more by the difference between average and instantaneous $`L_x`$ of a source than by $`L_x`$ itself (van der Klis 1997, 1998). In a plot of $`\nu `$ vs. L<sub>x</sub> (defined as $`4\pi d^2f_x`$ with $`f_x`$ the X-ray flux and $`d`$ the distance; Ford et al. 1999; Fig. 8) a series of roughly parallel lines is seen, to first order one line per source (but see below). In each source there is a definite relation between L<sub>x</sub> and $`\nu `$, but the same relation does not apply in another source with a different average L<sub>x</sub>. Instead, that source covers the same $`\nu `$-range around its particular average L<sub>x</sub>. This is unexplained, and must mean that in addition to instantaneous L<sub>x</sub>, another parameter, related to average L<sub>x</sub>, affects the QPO frequency (van der Klis 1997, 1998). Perhaps this parameter is the neutron star magnetic field strength, which previously, on other grounds, was hypothesized to correlate to average L<sub>x</sub> (Hasinger & van der Klis 1989, Psaltis & Lamb 1998, Konar & Bhattacharya 1999, see also Lai 1998), but other possibilities exist (van der Klis 1998, Ford et al. 1999).
A similar pattern of parallel lines, but on a much smaller scale, occurs in some individual sources. When observed at different epochs, a source produces different frequency vs. flux tracks that are approximately parallel (GX 5$``$1: Wijnands et al. 1998c; GX 340+0: Jonker et al. 1998; 4U 0614+09: Ford et al. 1997a,b, Méndez et al. 1997, van Straaten et al. 1999; 4U 1608$``$52: Yu et al. 1997, Méndez et al. 1998a, 1999; 4U 1636$``$53: Méndez 1999; 4U 1728$``$34: Méndez & van der Klis 1999; 4U 1820$``$30: Kaaret et al. 1999b; Aql X-1: Zhang et al. 1998c, Reig et al. 1999; Figs. 8 and 9). This is most likely another aspect of the well-known fact (e.g., van der Klis et al. 1990, Hasinger et al. 1990, Kuulkers et al. 1994, 1996, van der Klis 1994, 1995a) that while the properties of timing phenomena such as QPOs are well correlated with one another and with X-ray spectral shape as diagnosed by X-ray colors (and hence with position in tracks in color-color diagrams) the flux correlates well to these diagnostics only on short (hours to days) time scales and much less well on longer time scales. This is why color-color diagrams, independent of flux, are popular for parametrizing spectral variability in these sources. Similarly to other timing parameters, kHz QPO frequency correlates much better with position on the track in the color-color diagram (Fig. 9) than with flux. Correlations of frequency with parameters describing spectral shape such as blackbody flux (Ford et al. 1997b) or power-law slope (Kaaret et al. 1998) in a two-component spectral model are also much better than with flux.
Usually, all this has been interpreted by saying that apparently inferred accretion rate (§5.3) governs both the timing and the spectral properties, but not flux (e.g., van der Klis 1995a). Of course, energy conservation suggests total $`\dot{M}`$ and flux should be well correlated. Perhaps, inferred $`\dot{M}`$ is not the total $`\dot{M}`$ but only one component of it, i.e., that through the disk, while there is also a radial inflow (e.g. Fortner et al. 1989, Kuulkers & van der Klis 1995, Wijnands et al. 1996b, Kaaret et al. 1998), maybe there are large and variable anisotropies or bolometric corrections in the emission, so that the flux we measure is not representative for the true luminosity (e.g., van der Klis 1995a), or possibly mass outflows destroy the expected correlation by providing sinks of both mass and (kinetic) energy (e.g., Ford et al. 1999). The true explanation is unknown. We do not even know if the two different parallel-lines phenomena (across sources and within individual sources; Fig. 8) have the same origin. It is possible that the quantity that everything depends on is not $`\dot{M}`$, but some other parameter such as inner disk radius $`r_{in}`$.
### 5.5 Peak separation and burst oscillation frequencies
In interpretations (§5.1, 7.1) where the kHz peak separation $`\nu _2\nu _1=\mathrm{\Delta }\nu `$ is the neutron star spin frequency $`\nu _s`$ one expects $`\mathrm{\Delta }\nu `$ to be approximately constant. This is not the case (Fig. 10). In Sco X-1 (van der Klis 1996c, 1997b), 4U 1608$``$52 (Méndez et al. 1998a,b), 4U 1735$``$44 (Ford et al. 1998b), 4U 1728$``$34 (Méndez & van der Klis 1999) and marginally also 4U 1702$``$43 (Markwardt et al. 1999a) the separation $`\mathrm{\Delta }\nu `$ decreases considerably when the kHz QPO frequencies increase. Other sources may show similar $`\mathrm{\Delta }\nu `$ variations (see also Psaltis et al. 1998). In a given source the $`\nu _1`$, $`\nu _2`$ relation seems to be reproducible.
That $`\nu _{burst}`$, by interpretation close to $`\nu _s`$4), is close to $`\mathrm{\Delta }\nu `$ (or $`2\mathrm{\Delta }\nu `$) is the most direct evidence for a beat frequency interpretation of the kHz QPOs (§5.1). The evidence for this is summarized in Table 4, where the highest asymptotic burst frequency (likely to be closest to the spin frequency; §4) is compared to the largest well-measured $`\mathrm{\Delta }\nu `$ in the five sources where both have been measured. The frequency ratio may cluster at 1 and 2, but a few more examples are clearly needed; in two sources discrepancies of $``$15% occur between $`\nu _{burst}`$ and $`2\mathrm{\Delta }\nu `$. Of these five sources, 4U 1728$``$34 also has a measured $`\mathrm{\Delta }\nu `$ variation (Fig. 10). When the QPO frequency drops, $`\mathrm{\Delta }\nu `$ increases, to saturate 4% below the burst oscillation frequency (Méndez & van der Klis 1999). It certainly seems here as if the kHz QPO separation “knows” the value of the burst oscillation frequency.
### 5.6 Strong gravity and dense matter
Potentially kHz QPOs can be used to constrain neutron-star masses and radii, and to test general relativity. Detection of the predcited innermost stable circular orbit (ISCO, §5.1) would constitute the first direct detection of a strong-field general-relativistic effect and prove that the neutron star is smaller than the ISCO. This possibility has fascinated since the beginning and was discussed well before the kHz QPOs were found (Kluźniak and Wagoner 1985, Paczynski 1987, Kluźniak et al. 1990, Kluźniak & Wilson 1991, Biehle & Blandford 1993). If a kHz QPO is due to orbital motion around the neutron star, its frequency can not be larger than the frequency at the ISCO. Kaaret et al. (1997) proposed that kHz QPO frequency variations that then seemed uncorrelated to X-ray flux in 4U 1608$``$52 and 4U 1636$``$53 were due to orbital motion near the ISCO and from this derived neutron star masses of $``$2M. However, from more detailed studies (§5.4) it is clear now that on short time scales frequency does in fact correlate to flux in these sources.
The maximum kHz QPO frequencies observed in each source are constrained to a narrow range. The 12 atoll sources with twin peaks have maximum $`\nu _2`$ values in the range 1074–1329 Hz; among the Z sources there are two cases of much lower maximum $`\nu _2`$ values ($`<`$900 Hz in GX 5$``$1 and GX 340+0), while the other four fit in. Zhang et al. (1997b) proposed that this narrow distribution is caused by the limit set by the ISCO frequency, which led them to neutron star masses near 2M as well. It is in principle possible that the maximum is set by some other limit on orbital radius (e.g., the neutron star surface), or that it is caused by the ISCO, but the frequency we observe is not orbital, in which cases no mass estimate can be made.
Miller et al. (1996, 1998a) suggested that when the inner edge of the accretion disk reaches the ISCO, the QPO frequency might level off and remain constant while $`\dot{M}`$ continues rising. Later, an apparent leveling off at $`\nu _2`$=1060$`\pm `$20 Hz with X-ray count rate was found in 4U 1820$``$30 (Zhang et al. 1998b; Fig. 11, left). If this is the orbital frequency at the ISCO, then the neutron star has a mass mass of $``$2.2M and is smaller than its ISCO, and many equations of state are rejected. The leveling off is also observed as a function of X-ray flux and color (Kaaret et al. 1999b) and position along the atoll track (Bloser et al. 1999). However, the frequency vs. flux relations are known in other sources (§5.4) to be variable, and in 4U 1820$``$30 the leveling off seems not to be reproduced in the same way in all data sets (M. Méndez et al. in prep.; Fig. 11, right). It may also be more gradual in nature than originally suggested. No evidence for a similar saturation in frequency was seen in other sources, and most reach higher frequencies. Possibly, the unique aspects of the 4U 1820$``$30 system are related to this (it is an 11-min binary in a globular cluster with probably a pure He companion star; Stella et al. 1987a).
If a kHz QPO peak at frequency $`\nu `$ corresponds to stable Keplerian motion around a neutron star, one can immediately set limits on neutron star mass $`M`$ and radius $`R`$ (Miller et al. 1998a). For a Schwarzschild geometry: (1) the radius $`R`$ of the star must be smaller than the radius $`r_K`$ of the Keplerian orbit: $`R<r_K=(GM/4\pi ^2\nu ^2)^{1/3}`$, and (2) the radius of the ISCO (§5.1) must also be smaller than $`r_K`$, as no stable orbit is possible within this radius: $`r_{ISCO}=6GM/c^2<(GM/4\pi ^2\nu ^2)^{1/3}`$ or $`M<c^3/(2\pi 6^{3/2}G\nu )`$. Condition (1) is a mass-dependent upper limit on $`R`$, and condition (2) an upper limit on $`M`$; neither limit requires detection of orbital motion at the ISCO. Fig. 12 (left) shows these limits in the neutron star mass-radius diagram for $`\nu =1220`$ Hz, plus an indication of how the excluded area (hatched) shrinks for higher values of $`\nu `$. The currently highest value of $`\nu _2`$, identified in most models with the orbital frequency, is 1329$`\pm `$4 Hz (van Straaten et al. 1999), so the hardest equations of state are beginning to be imperiled by the method. Corrections for frame dragging (shown to first order in $`j`$5.1, right\] in Fig. 12) expand the allowed region. They depend on the neutron star spin rate $`\nu _s`$, and somewhat on the neutron star model, which sets the relation between $`\nu _s`$ and angular momentum. For 1329 Hz the above equations imply $`M<1.65`$M and $`R_{NS}<12.4`$ km; with corrections for a 300 Hz spin these numbers become 1.9M and 15.2 km (van Straaten et al. 1999). Calculations exploring to what extent kHz QPOs constrain the EOS have further been performed by Miller et al. (1998b), Datta et al. (1998), Akmal et al. (1998), Kluźniak (1998), Bulik et al. (1999), Thampan et al. (1999), Li et al. (1999b), Schaab & Weigel (1999) and Heiselberg & Hjorth-Jensen (1999).
If $`\mathrm{\Delta }\nu `$ is near $`\nu _s`$ then the 18 neutron stars where this quantity has been measured all spin at frequencies between $``$240 and $``$360 Hz, a surprisingly narrow range. If the stars spin at the magnetospheric equilibrium spin rates (e.g., Frank et al. 1992) corresponding to their current luminosities L<sub>x</sub>, this would imply an unlikely, tight correlation between L<sub>x</sub> and neutron-star magnetic-field strength (White & Zhang 1997; note that a similar possibility came up in the discussion about the uniformity of the QPO frequencies themselves, §5.4). White & Zhang (1997) propose that when $`r_M`$ is small, as is the case here, it depends only weakly on accretion rate (as it does in some inner disk models; cf. Ghosh & Lamb 1992, Psaltis & Chakrabarty 1999). Another possibility is that the spin frequency of accreting neutron stars is limited by gravitational radiation losses (Bildsten 1998a, Andersson et al. 1999, see also Levin 1999). If so, then gravitational radiation is transporting angular momentum out as fast as accretion is transporting it in, making these sources the brightest gravitational radiation sources in the sky. They would produce a periodic signature at the neutron star spin frequency, which would facilitate their detection.
### 5.7 Other kilohertz QPO properties
The amplitudes of kHz QPOs increase strongly with photon energy (Fig. 13). In similar X-ray photometric bands the QPOs tend to be weaker in the more luminous sources, with 2–60 keV amplitudes ranging from as high as 15% (rms) in 4U 0614+09, to typically a few % (rms) at their strongest in the Z sources. At high energy amplitudes are much higher (e.g., 40% rms above 16 keV in 4U 0614+09; Méndez et al. 1997). Fractional rms usually decreases with inferred $`\dot{M}`$ (e.g., van der Klis et al. 1996b, 1997b, Berger et al. 1996) but more complex behavior is sometimes seen (e.g., di Salvo et al. 1999). The measured widths of QPO peaks are affected by variations in centroid frequency during the measurement, but typical values in the Z sources are 50–200 Hz. In the atoll sources, the upper peak usually has a width similar to this, although occasionally peaks as narrow as 10 or 20 Hz have been measured (eg. Wijnands & van der Klis 1997, Wijnands et al. 1998a,b) but the lower peak is clearly much narrower. It is rarely as wide as 100 Hz (Méndez et al. 1997) and sometimes as narrow as 5 Hz (e.g., Berger et al. 1996, Wijnands et al. 1998b). Various different peak width vs. inferred $`\dot{M}`$ relations have been seen, but there seems to be a tendency for the upper peak to become narrower as its frequency increases.
A strong model constraint is provided by the time lags between kHz QPO signals in different energy bands (§2). Time-lag measurements require very high signal-to-noise ratios (Vaughan et al. 1997), and have mostly been made in the very significant lower peaks observed in some atoll sources. Finite lags of 10–60$`\mu `$sec occur in these peaks (Vaugan et al. 1997, 1998, Kaaret et al. 1999a, Markwardt et al. 1999b; see Lee & Miller 1998 for a calculation of Comptonization lags relevant to kHz QPOs). Contrary to the initial report, the low-energy photons lag the high-energy ones (these are “soft lags”) by increasing amounts as the photon energy increases (Fig. 13). The lags are of opposite sign to those expected from the inverse Compton scattering thought to produce the hard spectral tails of these sources (e.g., Barret & Vedrenne 1994), and correspond to light travel distances of only 3–20 km. From this it seems more likely that the lags originate in the QPO production mechanism than in propagation delays. Markwardt et al. (1999b) reported a possible hard lag in an atoll-source upper peak.
## 6 Correlations with low-frequency timing phenomena and with other sources
### 6.1 Black hole candidates
The only oscillations with frequencies exceeding 10<sup>2.5</sup> Hz (§1) known in black hole candidates (BHCs) are, marginally, the 100–300 Hz oscillations in GRO J1655$``$40, and XTE J1550$``$564, and those reported very recently in Cyg X-1 and 4U 1630$``$47. An oscillation near 67 Hz observed in GRS 1915+105 is usually discussed together with these QPOs, although it is not clear that it is related. Usually, the phenomenology accompanying these high-frequency QPOs (spectral variations, lower-frequency variations) is complex.
The 67 Hz QPO in GRS 1915+105 (Morgan et al. 1997, Remillard & Morgan 1998) varies by only a few percent in frequency when the X-ray flux varies by a factor of several. It is relatively coherent, with Q usually around 20 (but sometimes dropping to 6; Remillard et al. 1999a), has an rms amplitude of about 1%, and the signal at high photon energy lags that at lower energy by up to 2.3 radians (Cui 1999). The 300 Hz QPO in GRO J1655$``$40 (Remillard et al. 1999b) was seen only when the X-ray spectrum was dominated by a hard power law component. This feature was relatively broad (Q$``$4) and weak (0.8% rms), and did not vary in frequency by more than $``$30 Hz. The 185–285 Hz QPO in XTE J1550$``$564 (Remillard et al. 1999c) shows considerable variations in frequency (Homan et al. 1999b, Remillard et al. 1999c). It is seen in similar X-ray spectral conditions as the 300 Hz QPO in GRO J1655$``$40, has similarly low amplitudes and 3$`<`$Q$`<`$10. Recent reports indicate that QPOs in this frequency range also occur in Cyg X-1 and 4U 1630$``$47 (Remillard 1999a,b).
The fact that the 300 and 67 Hz oscillations were constant in frequency (different from anything then known to occur in neutron stars) triggered interpretations where these frequencies depend mostly on black-hole mass and angular momentum and only weakly on luminosity, such as orbital motion at the ISCO (Morgan et al. 1997), Lense-Thirring precession there (Cui et al. 1998c, see also Merloni et al. 1999) or trapped-mode disk oscillations (Nowak et al. 1997). However, the variations in frequency of the QPO in XTE J1550$``$564 have cast some doubt on the applicability of such models.
As is well known, strong similarities exist with respect to many spectral and timing phenomena between low-magnetic-field neutron stars and BHCs (e.g., van der Klis 1994a,b; §6.3). While the 100–300 Hz oscillations may be related to the kHz QPOs observed in neutron stars (Psaltis et al. 1999a; see §6.3), there could also be a relation with the recently reported relatively stable QPO peaks near 100 Hz in 4U 0614+09 and 4U 1728$``$34 (van Straaten et al. 1999, di Salvo et al. 1999), which are clearly distinct from kHz QPOs. More work clearing up the exact phenomenology and more observations of black-hole transients leading to more examples of high frequency QPOs are clearly needed.
### 6.2 Cen X-3
The detection of QPO features near 330 and 760 Hz in the 4.8 s accreting pulsar Cen X-3 was recently reported by Jernigan et al. (1999). This is the first report of millisecond oscillations from a high-magnetic-field ($``$10<sup>12</sup> Gauss) neutron star. The QPO features are quite weak. Jernigan et al. (1999) carefully discuss the instrumental effects, which are a concern at these low power levels and interpret their results in terms of the photon bubble model (§7.3).
### 6.3 Low-frequency phenomena
Low-frequency ($`<`$100 Hz) QPOs have been studied in accreting neutron stars and black hole candidates since the 1980’s, mostly with the EXOSAT and Ginga satellites (see van der Klis 1995a). Two different ones were known in the Z sources, the 6-20 Hz so-called normal and flaring-branch oscillation (NBO; Middleditch & Priedhorsky 1986) and the 15–60 Hz so-called horizontal branch oscillation (HBO; van der Klis et al. 1985). The frequency of the HBO turned out to correlate well to those of the kHz QPOs (Fig. 15; van der Klis et al. 1997b, Wijnands et al. 1997b,1998a,c, Jonker et al. 1998, 1999c), and the same is true for the NBO in Sco X-1 (van der Klis et al. 1996b). Broad power-spectral bumps and, rarely, low-frequency QPOs were know in atoll sources as well (Lewin et al. 1987, Stella et al. 1987b, Hasinger & van der Klis 1989, Dotani et al. 1989, Makishima et al. 1989, Yoshida et al. 1993). With RXTE, QPOs similar to HBO are often seen (Strohmayer et al. 1996b, Wijnands & van der Klis 1997, Wijnands et al. 1998b, Homan et al. 1998a) and their frequencies also correlate well to kHz QPO frequency (Fig. 15; Stella & Vietri 1998, Ford & van der Klis 1998, Markwardt et al. 1999a, van Straaten et al. 1999, di Salvo et al. 1999, Boirin et al. 1999). It is not sure yet wether these atoll QPOs are physically the same as HBO in Z sources but this seems likely.
It is possible that these correlations arise just because all QPO phenomena depend on a common parameter (e.g. inferred $`\dot{M}`$, §5.4), but Stella & Vietri (1998) proposed that their origin is a physical dependence of the frequencies on one another (§7.2). In their relativistic precession model the HBO and the similar-frequency QPOs in the atoll sources are the same phenomenon, and their frequency $`\nu _{LF}`$ is predicted to be proportional to $`\nu _{2}^{}{}_{}{}^{2}`$. This is indeed approximately true in all Z sources except Sco X-1 (Psaltis et al. 1999b), as well as in the atoll sources (references above). Psaltis et al. (1999b) argue that a combination of the sonic-point and magnetospheric beat-frequency models can explain these correlations as well (§7.1).
Additional intriguing correlations exist between kHz QPOs and low-frequency phenomena which may link neutron stars and BHCs. It is useful to first examine a correlation between two low-frequency phenomena. At low $`\dot{M}`$, BHCs and atoll sources (van der Klis 1994a and references therein), the millisecond pulsar SAX J1808.4-3658 (§ 3; Wijnands & van der Klis 1998c), and perhaps even Z sources have very similar power spectra (Fig. 14; Wijnands & van der Klis 1999a), with a broad noise component that shows a break at low frequency and often a QPO-like feature above the break. Break and QPO frequency both vary in excellent correlation (Fig. 15, left), and similarly in neutron stars and BHCs. This suggests that (with the possible exception of the Z sources, which are slightly off the main relation) these two phenomena are the same in neutron stars and black holes. This would exclude spin-orbit beat-frequency models and any other models requiring a material surface, an event horizon, a magnetic field, or their absence, and would essentially imply the phenomena are generated in the accretion disk around any low-magnetic field compact object.
The good correlations between kHz QPOs and low-frequency phenomena in Z and atoll sources suggest that kHz QPOs might also fit in with schemes linking neutron stars and BHCs (§6.1). However, no twin kHz QPOs have been reported from BHCs. Psaltis et al. (1999a) pointed out that many Z and atoll sources, the peculiar source Cir X-1 and a few low luminosity neutron stars and BHCs sometimes show two QPO or broad noise phenomena whose centroid frequencies, when plotted vs. each other, seem to line up (Fig. 15, right). This suggests that perhaps “kHz QPOs” do occur in BHCs, but as features at frequencies below 50 Hz. These features have very low Q, and although the data are suggestive they are not conclusive. The implication would be that the lower kHz QPO peak (whose frequency is the one that lines up with those seen in the BHCs) is not unique to neutron stars, but a feature of disk accretion not related to neutron star spin. The coincidence of kHz QPO frequencies with burst oscillation frequencies (§4, §5.1) would then require some other explanation. Orbital motion in the disk would remain an attractive interpretation for some of the observed frequencies. Stella et al. (1999b) showed that for particular choices of neutron star and black hole angular momenta their relativistic precession model can fit these data. The phenomenology is quite complex; in particular, no way has been found yet to combine the Wijnands & van der Klis (1999a) work with the Psaltis et al. (1999a) results in a way that works across all source types. New low-frequency phenomena are still being discovered as well (di Salvo et al. 1999, Jonker et al. 1999c).
## 7 Kilohertz QPO models
The possibility to derive conclusions of a fundamental nature has led to a relatively large number of models for kHz QPOs. Most, but not all of these involve orbital motion around the neutron star. It is beyond the scope of the present work to provide an in-depth discussion of each model. Instead, I point out some of the main issues, and provide pointers to the literature.
In early works, the magnetospheric beat-frequency model was implied when beat-frequency models were mentioned (e.g., van der Klis et al. 1996b, Strohmayer et al. 1996c), but this model has not recently been applied much to kHz QPOs; see however Cui et al. (1998a). Most prominent recently have been the sonic point beat-frequency model of Miller et al. (1996, 1998a) and the relativistic precession model of Stella and Vietri (1998, 1999), but also the the photon bubble model (Klein et al. 1996b) and the disk transition layer models (Titarchuk et al. 1998, 1999) have been strongly argued for. Additional disk models have been proposed as well (§7.4). Neutron star oscillations have been considered (Strohmayer et al. 1996c, Bildsten et al. 1998, Bildsten & Cumming 1998), but probably can not produce the required combination of high frequencies and rapid changes in frequency.
Most models have evolved in response to new observational results. The sonic point model was modified to accommodate the observed deviations from a pure beat-frequency model (§5.5; Lamb & Miller 1999, Miller 1999c), the relativistic precession model initially explained the lower kHz peak as a spin/orbit beat frequency (Stella & Vietri 1998) and only later by apsidal motion (Stella & Vietri 1999), the photon bubble model is based on numerical simulations which in time have become more similar to what is observed (R Klein, priv. comm.), and also the disk transition layer models have experienced considerable evolution (e.g., Osherovich & Titarchuk 1999).
The relativistic precession model makes the strongest predictions with respect to observable quantities and hence allows the most direct tests, but the near-commensurability of kHz QPO and burst oscillation frequencies (§5.5) is unexplained in that model. The sonic-point model provides specific mechanisms to modulate the X-rays and to make the frequency vary with mass-accretion rate. Other models usually discuss at least one of these issues only generically (usually in terms of self-luminous and/or obscuring blobs, and arbitrary preferred radii in the accretion disk).
### 7.1 The sonic point beat-frequency model
Beat-frequency models involve orbital motion at some preferred radius in the disk (§5.1). A beat-frequency model which uses the magnetospheric radius $`r_M`$ was proposed by Alpar & Shaham (1985) to explain the HBO in Z sources (§6.3; see also Lamb et al. 1985). In the Z sources HBO and kHz QPOs have been seen simultaneously, so at least one additional model is required.
Miller et al. (1996, 1998a) suggest to continue to use the magnetospheric model for the HBO (see also Psaltis et al. 1999b), and for the kHz QPOs propose the sonic-point beat-frequency model. In this model the preferred radius is the sonic radius $`r_{sonic}`$, where the radial inflow velocity becomes supersonic. This radius tends to be near $`r_{ISCO}`$5.1) but radiative stresses change its location, as required by the observation that the kHz QPO frequencies vary. Comparing the HBO and kHz QPO frequencies, clearly $`r_{sonic}r_M`$, so part of the accreting matter must remain in near-Keplerian orbits well within $`r_M`$.
At $`r_{sonic}`$ orbiting clumps form whose matter gradually accretes onto the neutron star following a fixed spiral-shaped trajectory in the frame corotating with their orbital motion (Fig. 16). At the “footpoint”of a clump’s spiral flow the matter hits the surface and emission is enhanced. The footpoint travels around the neutron star at the clump’s orbital angular velocity, so the observer sees a hot spot move around the surface with the Keplerian frequency at $`r_{sonic}`$. This produces the upper kHz peak at $`\nu _2`$. The high Q of the QPO implies that all clumps are near one precise radius and live for several 0.01 to 0.1 s, and allows for relatively little fluctuations in the spiral flow. The beat frequency at $`\nu _1`$ occurs because a beam of X-rays generated by accretion onto the magnetic poles sweeps around at the neutron star spin frequency $`\nu _s`$ and hence irradiates the clumps at $`r_{sonic}`$ once per beat period, which modulates, at the beat frequency, the rate at which the clumps provide matter to their spiral flows and consequently the emission from the footpoints.
So, the model predicts $`\mathrm{\Delta }\nu =\nu _2\nu _1`$ to be constant at $`\nu _s`$, contrary to observations (§5.5). However, if the clumps’ orbits gradually spiral down, then the observed beat frequency will be higher than the actual beat frequency at which beam and clumps interact, because then during the clumps’ lifetime the travel time of matter from clump to surface gradually diminishes. This puts the lower kHz peak closer to the upper one, and thus decrease $`\mathrm{\Delta }\nu `$, more so when at higher $`L_x`$ due to stronger radiation drag the spiralling-down is faster, as observed (Lamb & Miller 1999, Miller 1999). As the exact way in which this affects the relation between the frequencies is hard to predict, this makes testing the model more difficult. A remaining test is that a number of specific additional frequencies is predicted to arise from the beat-frequency interaction (Miller et al. 1998a), which are different from additional frequencies in, e.g., the relativistic precession model.
### 7.2 The relativistic precession model
Inclined eccentric free-particle orbits around a spinning neutron star show both nodal precession (a wobble of the orbital plane) due to relativistic frame dragging (Lense & Thirring 1918), and relativistic periastron precession similar to Mercury’s (Einstein 1915). The relativistic precession model (Stella & Vietri 1998, 1999) identifies $`\nu _2`$ with the frequency of an orbit in the disk and $`\nu _1`$ and the frequency $`\nu _{LF}`$ of one of the observed low-frequency (10–100 Hz) QPO peaks (§6.3) with, respectively, periastron precession and nodal precession of this orbit.
To lowest order, the relativistic nodal precession is $`\nu _{nod}=8\pi ^2I\nu _2^2\nu _s/c^2M`$ and the relativistic periastron precession causes the kHz peak separation to vary as $`\mathrm{\Delta }\nu =\nu _2(16GM/rc^2)^{1/2}`$, where $`I`$ is the star’s moment of inertia and $`r`$ the orbital radius (Stella et al. 1998, 1999; see also Marković & Lamb 1998). Stellar oblateness affects both precession rates and must be corrected for (Morsink & Stella 1999, Stella et al. 1999b). For acceptable neutron star parameters, there is an approximate match (e.g., Fig. 17) with the observed $`\nu _1`$, $`\nu _2`$ and $`\nu _{LF}`$ relations if $`\nu _{LF}`$ is twice (or perhaps sometimes four times) the nodal precession frequency, which could in principle arise from a warped disk geometry (Morsink & Stella 1999). In this model the neutron star spin frequencies do not cluster in the 250–350 Hz range (§5.6) and $`\mathrm{\Delta }\nu `$ and $`\nu _{burst}`$ are not expected to be equal as in beat-frequency interpretations. A clear prediction is that $`\mathrm{\Delta }\nu `$ should decrease not only when $`\nu _2`$ increases (as observed) but also when it sufficiently decreases (Fig. 17).
For a precise match between model and observations, additional free parameters are required. Stella & Vietri (1999) propose that the orbital eccentricity $`e`$ systematically varies with orbital frequency. A critical discussion of the degree to which the precession model and the beat-frequency model can each fit the data can be found in Psaltis et al. (1999b). Vietri & Stella (1998) and Armitage et al. (1999) have performed calculations relevant to the problem of sustaining the tilted orbits required for Lense-Thirring precession in a viscous disk, where the Bardeen-Petterson effect (1975) drives the matter to the orbital plane. Karas (1999a,b) calculated frequencies and light curves produced by clumps orbiting the neutron star in orbits similar to the ones discussed here. Miller (1999d) calculated the effects of radiation forces on Lense-Thirring precession. Kalogera & Psaltis (1999) explored how the Lense-Thirring precession interpretation constrains neutron star structure.
Relatively high neutron star masses (1.8–2M), relatively stiff equations of state, and neutron star spin frequencies in the 300–900 Hz range follow from this model. Because it requires no neutron star, but only a relativistic accretion disk to work, the model can also be applied to black holes. Stella et al. (1999b) propose it explains the frequency correlations discussed in §6.3 (Fig. 17).
The idea that the three most prominent frequencies observed are in fact the three main general-relativistic frequencies characterizing a free-particle orbit is fascinating. However, questions remain. How precessing and eccentric orbits can survive in a disk is not a priori clear (see Psaltis & Norman 1999 for a possible way to obtain these frequencies from a disk). How the flux is modulated at the predicted frequencies, why the basic, orbital, frequency $`\nu _2`$ varies with luminosity, and how the burst oscillations fit in are all open questions. With respect to the burst oscillations the model requires other explanations than neutron star spin (§4) and these are being explored (Stella 1999, Kluźniak 1999, Psaltis & Norman 1999).
### 7.3 Photon bubble model
A model based on numerical radiation hydrodynamics was proposed by Klein et al. (1996b) for the kHz QPOs in Sco X-1. In this model accretion takes place by way of a magnetic funnel within which accretion is super-Eddington, so that photon bubbles form which rise up by buoyancy and burst at the top in quasi-periodic sequence. In some of the simulations one or two strong QPO peaks are found, whose frequencies increase with accretion rate, as observed (R. Klein, priv. comm.). The model stands out by not requiring rotational phenomena to explain the QPOs, and does not naturally produce beat frequencies. In recent work, attention with respect to this model has shifted to the classical accreting pulsars for which it was originally conceived (Klein et al. 1996a; §6.2).
### 7.4 Disk mode models
In several models the observed frequencies are identified with oscillation modes of an accretion disk. From an empirical point of view, these models fall into two classes. Some can be seen as “implementations” of a beat-frequency (e.g. Alpar et al 1997) or a precession (Psaltis & Norman 1999) model that they provide ways in accretion disk physics to produce signals at the frequencies occurring in those models. Others produce new frequencies. In all disk models one of the two kHz QPOs is a Keplerian orbital frequency at some radius in the disk.
The disk transition layer models (Titarchuk & Muslimov 1997, Titarchuk et al. 1998,1999, Osherovich & Titarchuk 1999a,b, Titarchuk & Osherovich 1999) have evolved into a description (the “two-oscillator model”) where $`\nu _1`$ is identified with the Keplerian frequency at the outer edge of a viscous transition layer between Keplerian disk and neutron-star surface. Oscillations in this layer occur at two low frequencies, producing the noise break and a low frequency QPO (§6.3). Additionally, blobs described as being thrown out of this layer into a magnetosphere oscillate both radially and perpendicular to the disk, producing two harmonics of another low-frequency QPO (in the Z sources this is the HBO) as well as the upper kHz peak. So, altogether this description provides six frequencies which can all fit observed frequencies. Further work on disk oscillations was performed by Lai (1998, 1999), Lai et al. (1999) and Ghosh (1998; see also Moderski & Czerny 1999).
## 8 Final remark
RXTE has opened up a window that allows us to see down to the very bottoms of the potential wells of some neutron stars and perhaps to near to the horizons of some black holes. Three millisecond phenomena have been found, whose interpretation relies explicitly on our description of strong-field gravity and neutron-star structure. In order to take full advantage of this, it will be necessary to observe the new phenomena with larger instruments (in the 10 m<sup>2</sup> class). This will allow to follow the motion of clumps of matter orbiting in strong gravity and of hot spots corotating on neutron star surfaces and thereby to to map out curved spacetime near accreting compact objects and measure parameters such as the compactness of neutron stars and the spin of black holes.
Acknowledgements: It is a pleasure to acknowledge the help of many colleagues who either made data available before publication, sent originals of figures, read versions of the manuscript or provided insightful discussion: Didier Barret, Lars Bildsten, Deepto Chakrabarty, Wei Cui, Eric Ford, Peter Jonker, Richard Klein, Fred Lamb, Phil Kaaret, Craig Markwardt, Mariano Méndez, Cole Miller, Mike Nowak, Dimitrios Psaltis, Luigi Stella, Tod Strohmayer, Rudy Wijnands and Will Zhang. This work was supported in part by the Netherlands Organization for Scientific Research (NWO) and the Netherlands Research School for Astronomy (NOVA).
## Literature cited |
no-problem/0001/astro-ph0001229.html | ar5iv | text | # Ages, Distances, and the Initial Mass Functions of Stellar Clusters
## 1. Introduction
Rather than attempt to present an extremely abbreviated synopsis of all of the material covered by the panel, we give here a summary of a few of the presentations. For each main topic, I (JRS) will give a brief outline of the current paradigm and/or any well-known controversies. Individual presentations by members of the panel will follow, with the author of the contribution indicated by his initials at the end of the section title.
## 2. The Distances to Open Clusters
The classical open cluster distance scale is based on determining an astrometric distance to the Hyades, either via trigonometric parallax or convergent point or some similar analysis, and then using main-sequence fitting techniques to determine the distances of other clusters relative to the Hyades (thus requiring a small correction for the higher-than-solar metallicity of the Hyades). This method obviously requires an accurate knowledge of the cluster’s reddening and metallicity. It further requires good photometry over the color range where the cluster members lie on or near the ZAMS. Finally, large number statistics are useful in order that the single-star sequence be well-defined. The method assumes that the photometry for each cluster is accurately calibrated, that the metallicities are accurately known and the effects of metallicity correctly modeled, and that the corrections for interstellar extinction are accurately derived from the data.
For the nearer, richer open clusters, where photometry of large numbers of stars is available (often in several different photometric systems), the main sequence fitting distances are typically accurate to of order 5% in the distance (0.1 mag in the distance modulus) unless there is some significant error in the assumptions of the method. Using the available data, the commonly quoted distances to the Pleiades, Praesepe and Coma clusters from main sequence fitting are of order 130 pc, 158 pc, and 80 pc, respectively. Standard compilations of open cluster distances include Becker & Fenkart (1971), the Lynga catalog (Lynga 1987), and J.-C. Mermilliod’s on-line “Database for Stars in Open Clusters” (BDA) at http://cdsweb.u-strasbg.fr/online/mermio.html .
### 2.1. Comparison of Hipparcos Open Cluster Distances to MS Fitting Distances (DMT)
The high precision and large number of stars in the Hipparcos catalog made it possible for the first time to derive trigonometric distances to open clusters other than the Hyades (Robichon et al. 1999, van Leeuwen 1999), as well as an extremely precise (average) distance to the Hyades cluster itself ($`mM=3.34\pm 0.01`$, Perryman et al. 1998). In several cases, particularly for the Pleiades and Coma clusters, the derived distances (118 pc and 90 pc, respectively for Pleiades and Coma) are significantly different from distances inferred from the technique of main-sequence fitting (MSF). If the Hipparcos parallax to the Pleiades were correct, then the main sequence of that cluster would be as much as 0.3 mag fainter than expected from stellar evolution theory.
Prompted by this discrepancy, Pinsonneault et al. (1998) re-examined the MSF distances to several open clusters with distances from Hipparcos, and in particular discussed how the luminosity of the main sequence depends on metallicity, helium abundance, and interstellar extinction. They concluded that none of these effects could produce the anomalous main sequence luminosity, and suggested that there could in some cases be systematic errors in the Hipparcos parallaxes on the order of 1 milliarcsecond (mas), about 10 times larger than the systematic errors in the parallaxes on a global scale.
Soderblom et al. (1998) examined the Hipparcos data base, selecting young main-sequence stars with parallax errors small enough that systematic errors at the 1 mas level would be unimportant. If nature were able to produce subluminous stars in the Pleiades, they reasoned, then there should be similar objects in the field. They were unable to find any, further suggesting that the Pleiades distance from Hipparcos was incorrect.
In the design of the Hipparcos experiment, it was known that a potential problem in determining the distances to open clusters is the possibility of correlated errors on small angular scales (e.g., Lindegren 1988). Furthermore, the typical density of stars in open clusters is significantly higher than the average across the sky. (Both these effects were included in the Hipparcos analysis of the cluster distances.) Narayanan & Gould (1999a) proposed a new method of examining the Hipparcos parallaxes in open clusters. Applying this to the Pleiades and Hyades (Narayanan & Gould 1999b), they showed that the Hipparcos parallaxes toward these open clusters are spatially correlated over angular scales of $`2^{}3^{}`$, with an amplitude of up to 2 mas. This correlation is stronger than expected based on the analysis of the Hipparcos catalog. Using a distance method based on the Hipparcos proper motions which should not be biased by these spatial correlations, Narayanan & Gould derived a distance to the Pleiades essentially equal to the main-sequence fitting distance but significantly different from that derived from the Hipparcos parallaxes. For the Hyades, by chance, the structure of the spatial correlations is such that the errors average out and the distance derived from the Hipparcos parallaxes agrees with that derived from main-sequence fitting.
The issue is not settled, however, because those most closely connected to the Hipparcos parallax measurements maintain that their distances are correct, and that there is an astrophysical explanation for the discrepant Pleiades and Coma cluster distances (see the talk in these proceedings by M. Grenon). Fortunately, there is reason to believe that a definitive resolution can now be expected from new observations because NASA has approved construction of a new astrometric mission called FAME, which should obtain parallaxes for many more stars and to significantly higher accuracy than Hipparcos – and provide the answers within the next decade based on an expected 2004 launch (see http://aa.usno.navy.mil/fame/).
## 3. Open Cluster Ages
The traditional open cluster age scale is based on fitting photometry of stars near the upper main sequence turnoff of a given cluster to theoretical evolutionary isochrones. The most widely referenced compendium of open cluster ages derived in that manner is probably Mermilliod (1981). In that paper, the ages of the Alpha Persei, Pleiades, and Hyades open clusters are listed as 51 Myr, 78 Myr, and 660 Myr, respectively. While other age indicators confirm the relative ordering of cluster ages given by Mermilliod, the absolute age scale has been a subject of some controversy. Most particularly, by adjusting the amount of “convective core overshoot” for stars near the MS turnoff, it is possible to derive cluster ages up to a factor of two older than those given in Mermilliod (1981): see, for example, Mazzei & Pigatto (1989) or Meynet et al. (1993).
An alternate way to estimate cluster ages is by determining the location of the pre-main sequence (PMS) turn-on point or the displacement of the PMS locus above the ZAMS. Because of fairly obvious realities, this method has mostly been applied to very young clusters. However, a few – perhaps foolhardy – authors have attempted to apply the method to “oldish” clusters like the Pleiades or Alpha Persei (Stauffer 1984; Stauffer et al. 1989). With the development of more accurate theoretical evolutionary models that extend to lower masses, it is now becoming more feasible to extend this type of age-dating to “oldish” open clusters, and two discussions of this topic are included below.
A new way to estimate the age of open clusters has recently been proposed by Basri, Marcy, & Graham (1996=BMG), Bildsten et al. (1997), Ushomirsky et al. (1998), Ventura et al. (1998) and others. The idea behind this method is that for stars near the substellar mass limit, the age at which stars become hot enough in their cores to burn lithium is a sensitive function of mass. Furthermore, it is argued that the physics required to predict the location of the “lithium depletion boundary” (LDB) as a function of age is very well understood and not subject to significant uncertainty (Bildsten et al. 1997). In the mass range of interest, stars are fully convective, and the core lithium abundance will be directly reflected in the surface lithium abundance - and the latter can be determined by use of the 6708Å Li I doublet.
BMG and Rebolo et al. (1996) were the first to successfully apply this test by detecting lithium in three substellar objects in the Pleiades. However, it was later determined (Basri & Martín 1999a) that the exact location of the lithium boundary (and hence the Pleiades age) was uncertain because the brightest of the three objects (PPL15) is a nearly equal mass binary. This problem was resolved by Stauffer, Schultz, & Kirkpatrick (1998=SSK), who obtained spectra of an additional 10 faint Pleiades members, five of which still retain their lithium and by Martín et al. (1998) who obtained spectra of an additional two Pleiades members, one of which had detected lithium. By providing measurements of a large number of stars near the lithium boundary, SSK claimed to determine the absolute magnitude of the lithium depletion boundary to an accuracy of about 0.1 mag, corresponding to an age uncertainty of about 8 Myr. The age derived for the Pleiades by SSK was 125 Myr.
More recently, Basri & Martín (1999b) and Stauffer et al. (1999) have determined an LDB age for the Alpha Persei cluster (based, respectively, on one and five members with detected lithium), and Barrado y Navascués, Stauffer & Patten (1999) have derived an LDB age for the IC2391 open cluster. The ages for all three clusters derived in this manner are systematically older (by of order 50%) than the ages quoted in Mermilliod (1981). If these LDB ages are as accurate and precise as believed, it should be possible to define a new open cluster age scale to replace Mermilliod (1981). Two of the panelists however urge caution at this point.
### 3.1. Limitations of the Lithium Depletion Boundary Method (RDJ)
Whilst the Lithium Depletion Boundary Method is potentially a very precise way of determining the age of a cluster, it is worth considering in some detail to what extent the age error budget is influenced by various sources of uncertainty. These can be placed into two categories; random and systematic.
My baseline assumption is that the age is found by locating the LDB in a colour-magnitude diagram (CMD), using the colour of the LDB and an empirical bolometric correction (BC) to find $`L_{\mathrm{bol}}`$ at the LDB and then using a model for Li depletion in cool stars to translate this into an age (see Fig. 1a). Random errors can be ascribed to uncertain placement of the LDB in the CMD, uncertain photometric calibrations and uncertainties in the distance and reddening of the cluster in question. Systematics arise from the chosen BC-colour relation, whether one assumes the LDB defines the point at which say 90% or 99% of Li has been depleted and the choice of evolutionary model which defines the $`L_{\mathrm{bol}}`$-age relation at the LDB. The first two of these systematics will alter the ages of all clusters in qualitatively the same way, whereas the latter systematic may cause individual clusters to be older or younger, as the $`L_{\mathrm{bol}}`$-age relations cross for differing evolution models. There may also be additional systematics due to common assumptions made by all the current generation of models.
I have investigated these uncertainties, focussing on clusters with an LDB defined in the $`I`$ vs $`RI`$ CMD, and using the theoretical models of Burrows et al. (1997), Chabrier & Baraffe (1997) and D’Antona & Mazzitelli (1997). Note that model colours and magnitudes from the former two sets of models have not been used as their uncertainties are unquantified. I find that for clusters with data quality similar to that available for the Pleiades, random errors in the age rise from $`\pm 9\%`$ at $`25`$ Myr, to $`\pm 16\%`$ at 200 Myr. These errors are dominated by uncertain placement of the LDB in the CMD ($`\pm 0.15`$ mag in $`I`$ and $`\pm 0.05`$ in $`RI`$) along with distance modulus errors and uncertainties in the $`I`$ photometric calibration (both around $`\pm 0.1`$ mag). Obtaining Li data for more points around the LDB, distinguishing binaries from single stars and tightening up the photometric calibration could reduce these errors considerably. The systematic errors are more constant, ranging between $`7\%`$ and $`11\%`$ over a similar age range. The systematics are due mainly to the choice of model and whether there is any age dependence in the BC-colour relation. The influence of whether Li is depleted by 90% or 99% at the LDB (say for Li 6708Å EWs $`<0.3`$Å) only alters ages by $`\pm 3`$%.
In Table 1, I present the results of my analysis for the Pleiades, Alpha Per and IC 2391 clusters using Li spectra, $`I`$ and $`RI`$ data from the literature and averaging over the three evolution models. These ages are similar to those previously estimated (Stauffer et al. 1998, 1999; Barrado y Navascués et al. 1999), but the estimated errors are larger by a more than a factor of two. This does not alter the conclusion that LDB ages are older than turn-off ages with no core overshoot, but certainly obscures whether core overshoot might be mass dependent (see §3.2).
I also have evidence that there may be additional systematics in the LDB method that are not adequately reflected by simply choosing a variety of models. If I assume the Pleiades age is given by its LDB age for a given set of evolutionary tracks, then I can force an isochrone in the CMD to match the available cluster photometry by choosing a particular form of the $`RI`$-$`T_{\mathrm{eff}}`$ relation. The only other parameters involved here are cluster distance and reddening and the empirical BCs – none of which are uncertain enough to affect my conclusions. If I further assume that this $`RI`$-$`T_{\mathrm{eff}}`$ relation applies to younger stars, in IC 2391 for example, then the isochrone defined by their LDB age should coincide with the observed cluster CMD. If it does not then there is either something wrong with the evolution models or the assumption that a single colour-$`T_{\mathrm{eff}}`$ relation holds over a range of gravity is flawed. Figure 1b shows the result of doing this, assuming a Pleiades distance modulus of 5.6. There is clearly a discrepancy. Using the Hipparcos distance modulus of 5.3 makes the Pleiades LDB age older by $`25`$ Myr and merely increases the discrepancy. I conclude that there may yet be a systematic problem in the evolutionary models (all of them) that could be solved if IC 2391 were younger than suggested by its LDB age. An age-dependent colour-$`T_{\mathrm{eff}}`$ relation seems less likely as it is not predicted in currently published models and spectroscopic indices in the $`R`$ and $`I`$ bands give reliable estimates of $`RI`$ colours in all these clusters. This suggests that any gravity dependence of spectral features is heavily subordinate to the dominant $`T_{\mathrm{eff}}`$ dependence.
### 3.2. Age Estimates from Isochrone Fitting (DMT)
There are new efforts underway to compare the MSF and Hipparcos distances for more clusters. For example, Pinsonneault et al. (2000) have expanded their MSF technique, which employs the Yale YREC isochrones, to work over a wider range of luminosity on the main sequence than previously, by recomputing color-effective temperature transformations to reproduce the morphology of the Pleiades main sequence. For clusters younger than the age of the Hyades, the technique can be used to estimate the age of the cluster, since the lowest-mass stars are still descending towards the main sequence. This technique is illustrated in Figure 2, which shows a color-magnitude diagram for Alpha Per compared to isochrones of various ages. An age of $`60\pm 6`$ Myr results. This method also generates a metallicity estimate for each cluster by comparing the distance derived using $`BV`$ photometry to that obtained in $`VI`$; since the luminosity depends on metallicity differently in the two colors, one derives distances which disagree unless the metallicity is correct.
Pinsonneault et al. (2000) derive distances and photometric metallicities for Alpha Per, the Pleiades, NGC 2516, NGC 6475, and NGC 6633; for other clusters they adopt the high-resolution spectroscopic abundances of Boesgaard & Friel (1992) and the distances computed in Pinsonneault et al. (1998). There are some systems (the Pleiades and Coma Ber) which still have large differences between the MSF and Hipparcos distances, while there are others (NGC 6475 and Alpha Per) which may possibly be in conflict depending on the treatment of the Hipparcos errors. There is no obvious pattern to the deviations with age or metal abundance, and furthermore the differences are not consistent with a simple scale shift to systematically shorter or longer distances.
In most cases, the ages derived in Pinsonneault et al. (2000) are somewhat higher than had been found from the main-sequence turnoff (as in Mermilliod 1981), but are still rather younger than the ages inferred from the LDB method. The two ages are compared in Figure 3 for three open clusters (from youngest to oldest being Alpha Per, IC 2391, and the Pleiades). The upper panel compares the ages directly, with the dashed line indicating equality. The ratio of the two ages is shown in the lower panel; there is probably an age-dependent offset in the two scales.
## 4. Cluster and Star-Forming Region Mass Functions
It is only within the past few years that open cluster mass functions that extend to low masses ($``$0.1-0.2 $``$$`_{\mathrm{}}`$) and that are based on proper motions have started to become available. This situation has resulted from a lack of the appropriate raw materials (deep, 1st and 2nd epoch imaging data) and a lack of sufficiently good and automated measuring machines (cf. Hambly 1998). Much of the literature for open clusters in fact has not been directed at determining a mass function, but instead has concentrated on the more directly determined luminosity function. At least up until 1980, and possibly up until much later, the most cited primary reference for open cluster luminosity functions was the work by van den Bergh & Sher (1960). Those authors estimated luminosity functions for a large number of open clusters using photographic plates and estimated cluster-member data derived from counting the number of stars in annuli centered on the apparent cluster center. van den Bergh and Sher’s conclusion from these data was that most open clusters had falling or flat luminosity functions faintward of M<sub>B</sub> $``$ 6, and hence that open clusters are quite deficient in low mass stars compared to the field. This result was later used as one piece of evidence in favor of bimodal star formation – in particular in favor of the idea that low mass stars are formed in loose associations (like Taurus-Auriga) and high mass stars are formed in dense clusters. However, the observational result in fact was almost certainly wrong because the outer annuli used to derive the “field” star density was chosen at too small a radius (which preferentially affects the derived number of faint cluster members due to mass segregation).
The situation in star-forming regions was not much better until quite recently, again due to a combination of lack of the right equipment (mostly sensitive, low-noise IR detectors) and lack of the right apparatus to utilize those observations (in this case, theoretical evolutionary tracks and ways to compare them to observations for very young ages). Much progress has been made in the past few years on both fronts, with the results that it is possible to begin to have confidence in the mass functions being derived for regions like rho Oph and IC 348. It is particularly likely that the derived mass functions will be close to reality when spectra have been obtained for all or most of the low mass objects so that the estimated effective temperatures are determined better and non-members can be excluded (see especially recent papers by K. Luhman on this topic).
### 4.1. Open Cluster Mass Functions (RDJ)
The problem of determining cluster mass functions (MFs) can be broken down into three parts: surveying for cluster members, converting the observed luminosity function into a MF and then correcting this MF for any mass-dependent spatial dispersion of stars (mass segregation).
A survey for cluster members might ideally use proper motions, photometry, radial velocities, spectral types, the presence of Li, X-ray emission or some subset of these to form an opinion on individual membership. It is clearly of vital importance to understand both the completeness (to include members) and specificity (to exclude non-members) of any membership tests. These need not be known on an individual basis, but must at least be known statistically as a function of luminosity (e.g., Hambly et al. 1999).
The conversion from luminosity functions to a MF is not trivial, relying on the age and distance of the clusters (with age becoming more important in younger clusters), a choice of stellar evolution models and a knowledge of the fraction of stars that are unresolved binary systems (with consequently larger luminosities).
The final step of correcting for the effects of mass segregation is often forgotten. For instance, the Pleiades should be almost dynamically relaxed and equipartition will have ensured that lower mass stars have a larger core radius ($`r_cM^{1/2}`$). This effect has been detected by Pinfield et al. (1998) and could have a large effect on the low mass end of the mass function. Unfortunately, the relaxation time also becomes longer at lower masses and this combined with the relatively low number of brown dwarfs uncovered in the Pleiades makes normalization of the brown dwarf MF in the Pleiades uncertain by a factor of $`2`$ relative to the higher mass stars (Hodgkin & Jameson 2000). Finally, there is of course the possibility of preferential ejection of lower mass cluster members that stray beyond the tidal radius. There is no alternative to a theoretical correction for this effect if one wishes to estimate the initial MF from a present-day MF. Clearly these dynamical effects have considerably less influence in younger clusters and associations.
The state of the art MF for open clusters is that derived in the Pleiades (Bouvier et al. 1998, Hambly et al. 1999, Hodgkin & Jameson 2000). This now extends from the A stars down to brown dwarfs at $`0.04M_{}`$. Whilst there is still disagreement on the exact representation of the MF, a number of clear points have emerged. (i) The MF cannot be represented as a single power law, but a log-normal MF, similar to that proposed by Miller & Scalo (1979) certainly does a reasonable job down to 0.08$`M_{}`$. (ii) A power law fit to the brown dwarf MF still has uncertainties of 50% or so in the power law index because of the uncertainties in membership, small number statistics and uncertain dynamical effects. However, if the cluster MF is similar to the field MF, it is clear that brown dwarfs ($`0.04<M<0.08M_{}`$) contribute negligibly to the Galactic disk mass. (iii) There is no evidence for a steep turnover or truncation of the MF down to 0.04$`M_{}`$.
### 4.2. Comments on the Mass Function of the Pleiades (ELM)
My collaborators and I have obtained new IR photometry and optical spectroscopy of all of the very low mass and brown dwarf candidate members of the Pleiades proposed in Bouvier et al. (1998). We have used those new data (Martín et al. 2000) to attempt to determine which of these stars are real members of the cluster, and which are instead field star contaminants. The criteria used include spectral type and luminosity class, presence and strength of H$`\alpha `$ emission, and location in color-magnitude diagrams. In addition, we consider other published information such as the presence of lithium in absorption and radial velocity.
Table 2 provides a summary of the membership information for these stars. The bottom line is that the success rate for the Bouvier et al. candidate list is in the range 63$`\%`$ to 78$`\%`$, depending on the status of the YES? stars (for which the results are currently not definitive). Bouvier et al. had predicted a probable success rate of 75$`\%`$ based on a simple field star luminosity function and survey volume argument. The new data are consistent with the original estimate, with a possibility of somewhat more field star contamination than expected. Therefore, the Bouvier et al. estimate of the Pleiades IMF down to 0.04 M is supported by the new membership study, indicating a slightly rising mass function below the hydrogen burning mass limit. This mass function is slightly less steep than the mass function derived by Martín et al. (1997), based on the IAC Pleiades surveys which concentrated on fields somewhat closer to the cluster center than the CFHT fields. This could be a hint that the brown dwarfs are more concentrated to the cluster center than the VLM stars.
## References
Barrado y Navascués, D., Stauffer, J. R., & Patten, B. M. 1999, ApJ, 522, L53
Basri, G., Marcy, G., & Graham, J. 1996, ApJ, 458, 600 (BMG)
Basri, G., & Martín, E. L. 1999a, AJ, 118, 2460
Basri, G., & Martín, E. L. 1999b, ApJ, 510, 266
Becker, W., & Fenkart, R. 1971, A&AS, 4, 241.
Bildsten, L., Brown, E., Matzner, C., & Ushomirsky, G. 1997, ApJ, 482, 442
Boesgaard, A. M., & Friel, E. D. 1990, ApJ, 351, 467
Bouvier, J., et al. 1998, A&A, 336, 490
Burrows, A., et al. 1997, ApJ, 491, 856
Chabrier, G., & Baraffe, I. 1997, A&A, 327, 1039
D’Antona, F., & Mazzitelli, I. 1997, MSAI, 68, 807
Hambly, N. C. 1998, in ASP Conf. Ser. 134, Brown Dwarfs and Extrasolar Planets Proceedings, ed. R. Rebolo, E. L. Martín, M. R. Zapatero-Osorio (San Francisco: ASP), 11
Hambly, N. C., Hodgkin, S. T., Crossburn, M. R., & Jameson, R. F. 1999, MNRAS, 303, 835
Hodgkin, S. T., & Jameson, R. F. 2000, in Stellar clusters and associations: Convection, rotation and dynamos, ed. R. Pallavicini, ASP Conference series, in press
Lindegren, L. 1988, in Scientific Aspects of the Input Catalog Preparation II, January 1988, ed. J. Torra, & C. Turon (Noordwijk: ESA), 179
Lynga, G. 1987, Catalog of Open Cluster Data (5th Edition, Lund: Lund Observatory)
Martín, E. L., Zapatero Osorio, M. R., Rebolo, R. 1997, in ASP Conf. Ser. 134, Brown Dwarfs and Extrasolar Planets Proceedings, ed. R. Rebolo, E. L. Martín, M. R. Zapatero-Osorio (San Francisco: ASP), 507
Martin, E. L.; Basri, G.; Gallegos, J. E.; Rebolo, R.; Zapatero-Osorio, M. R.; Bejar, V. J. S. 1998, ApJ, 499, 61
Mazzei, P., & Pigatto, L. 1988, A&A, 193, 148
Mermilliod, J.-C. 1981, A&A, 97, 235
Meynet, G., Mermilliod, J.-C., & Maeder, A. 1993, A&AS, 98, 477
Miller, G. E., & Scalo, J. M. 1979, ApJS, 41, 513
Narayanan, V. K., & Gould, A. 1999a, ApJ, 515, 256
Narayanan, V. K., & Gould, A. 1999b, ApJ, 523, 328
Perryman, M. A. C., et al. 1998, A&A, 331, 81
Pinfield, D., Jameson, R. F., & Hodgkin, S. T. 1998, MNRAS, 299, 955
Pinsonneault, M. H., Stauffer, J., Soderblom, D. R., King, J. R., & Hanson, R. B. 1998, ApJ, 504, 170
Pinsonneault, M. H., Terndrup, D. M., & Yuan, Y. 2000, in Stellar Clusters and Associations: Convection, Rotation, and Dynamos, ed. R. Pallavicini, ASP Conf. Ser., in press
Rebolo, R., Martín, E. L., Basri, G., Marcy, G. & Zapatero-Osorio, M. R. 1996, ApJ, 469, L53
Robichon, N., Arenou, F., Mermilliod, J.-C., & Turon, C. 1999, A&A, 345, 471
Soderblom, D. R., King, J. R., Hanson, R. B., Jones, B. F., Fischer, D., Stauffer, J. R., & Pinsonneault, M. H. 1998, ApJ, 504, 192
Stauffer, J. 1984, ApJ, 280, 189
Stauffer, J., et al. 1989, ApJ, 346, 160
Stauffer, J. R., Schultz, G., & Kirkpatrick, J. D. 1998, ApJ, 499, L199 (SSK)
Stauffer, J. R. et al. 1999, ApJ, 527, 219
Ushomirsky, G., Matzner, C., Brown, E., Bildsten, L., Hilliard, V., & Schroeder, P. 1998, ApJ, 497, 253
van den Bergh, S., & Sher, D. 1960, Pub. DDO, 2, 203
van Leeuwen, F. 1999, A&A, 341, L71
Ventura, P., Zeppieri, A., Mazzitelli, I., & D’Antona, F. 1998, A&A, 334, 953 |
no-problem/0001/astro-ph0001227.html | ar5iv | text | # Irradiation of the secondary star in X-ray Nova Scorpii 1994 (=GRO J1655–40)
## 1 Introduction
The soft X-ray transients, a subclass of the low-mass X-ray binaries distinguished by their X-ray outbursts, have proved to be an ideal hunting ground for stellar-mass black hole candidates (Tanaka & Shibazaki 1996). The system Nova Sco 1994 (=GRO J1655–40) is particularly interesting, since as well as being a source of superluminal jets (Zhang et al., 1994; Harmon et al., 1995), its optical brightness and partial eclipse features mean that it is one of the few systems that has yielded a reliable estimate for the mass of the collapsed star.
Nova Sco 1994 was discovered on July 27 1994 with BATSE on board the Compton Gamma Ray Observatory (Zhang et al., 1994). It has been studied extensively during the past few years in X-rays and at optical and radio wavelengths (Bailyn et al., 1995a and b, Zhang et al., 1995, van der Hooft et al. 1998). Strong evidence that the compact object in Nova Sco 1994 is a black hole was presented by Bailyn et al. (1995b) who initially established a spectroscopic period of $`2.601\pm 0.027`$ days, classified the secondary as an F2–F6$`\mathrm{iv}`$ type star and suggested a mass function $`f(M)`$=3.16$`\pm `$0.15 M. An improved value of $`f(M)`$=3.24$`\pm `$0.09 M was presented by Orosz & Bailyn (1997) using both quiescent and outburst data, derived from a radial velocity semi-amplitude of 228.2$`\pm `$2.2 km s<sup>-1</sup>. Shahbaz et al. (1999) using only quiescent data, determined the true radial velocity semi-amplitude $`K_2`$=215.5$`\pm `$2.4 km s<sup>-1</sup>which gives a revised value for the mass function of $`f(M)`$ = 2.73$`\pm `$0.09 M. They also measured the rotational broadening of the secondary star which then gives the binary mass ratio $`q0.39`$ (=M<sub>2</sub>/M<sub>1</sub>, where M<sub>1</sub> and M<sub>2</sub> are the masses of the compact object and secondary star respectively).
The effect of heating of the secondary is to shift the ‘effective centre’ of the secondary, weighted by the strength of the absorption lines, from the centre of mass of the star. One expects that this results in a significant distortion of the radial velocity curve and renders a sinusoidal fit clearly inadequate, leading to a spuriously high radial velocity semi-amplitude. In order to quantify this effect we have determined the radial velocity variations of the secondary star in Nova Sco 1994, when it was in outburst and compared our results with others obtained using data taken when the source was in different X-ray states.
## 2 Observations and Data reduction
### 2.1 Spectroscopy
Intermediate resolution optical spectra of Nova Sco 1994 were obtained on 1996 June 20–24 with the 1.54-m Danish Telescope at the European Southern Observatory (ESO) in Chile using the Danish Faint Object Spectrograph and Camera (DFOSC). We used grating #8 which gave a dispersion of 1.26 Å per pixel and a wavelength coverage from 5865–8336Å. The Loral 2048$`\times `$2048 CCD was used, binned by a factor two in the spatial direction in order to reduce the readout noise, but not binned in the dispersion direction. The seeing during the observations was poor and variable (see section 2.2) so we used a slit width of $`2\stackrel{}{.}5`$ on the first night and then $`2\stackrel{}{.}0`$ for the other nights. This resulted in spectral resolutions of 7.6 Å and 5.5 Å for the first and other nights respectively. Wavelength calibration was performed using a Cu-Ar arc. A total of 47 spectra were taken each having exposure times of 1800s (see Table 1 for details).
The data reduction and analysis was performed using the Starlink figaro package, the pamela routines of K. Horne and the molly package of T. R. Marsh. Removal of the individual bias signal was achieved through subtraction of a median bias frame. Small scale pixel-to-pixel sensitivity variations were removed with a flat-field frame prepared from observations of a tungsten lamp. One-dimensional spectra were extracted using the optimal-extraction algorithm of Horne (1986), and calibration of the wavelength scale was achieved using 5th order polynomial fits which gave an rms scatter of 0.03 Å. The stability of the final calibration was verified with the OH sky line at 6300.3Å whose position was accurate to within 0.1 Å.
### 2.2 Photometry
Using the same setup as for the spectroscopy, we also obtained limited Bessell $`r`$-band images of Nova Sco 1994 every night. The data were debiased using a median bias frame, but not flat-fielded, as none were taken. These images were used to estimate the seeing each night (see Table 1). We applied aperture photometry to Nova Sco 1994 and several nearby comparison stars within the field of view. Johnson $`V`$\- and $`R`$-band magnitudes of these comparison stars were made available to us by J. Orosz. We determined the relative magnitude of Nova Sco 1994 with respect to three stars having a range of colours \[($`VR`$)=0.49, 0.77 and 1.23\]. Assuming that the colour correction between the two filter systems is small ($`<`$0.05 mags; similar to the accuracy of our photometry) and that Nova Sco 1994 has a colour in the same range as the comparison stars used, we estimate $`R`$15.05 for Nova Sco 1994.
## 3 The spectra of Nova Sco 1994
In Figure 1 we show the variance-weighted average and also the nightly averages of the Nova Sco 1994 spectra. A strong H$`\alpha `$ emission line (mean equivalent width of 7.5$`\pm `$0.06Å) and a much weaker He$`\mathrm{i}`$ 6678Å (equivalent width of 0.4$`\pm `$0.03Å) can be seen. In Table 2 we list the H$`\alpha `$ equivalent width for the nightly averages. The Fe$`\mathrm{i}`$ absorption blend at 6485, 6496, 6499, 6502Å is also visible. These features are used to determine the radial velocity of the secondary star (see section 4). The 6613Å diffuse interstellar band is also present.
The emission lines in Nova Sco 1994 are double-peaked, which is presumably a consequence of the system being at high inclination. We can compare the observed peak-to-peak half separation of the H$`\alpha `$ emission line (which arises from the accretion disc) with the projected velocity of the outer disc edge. In a binary system with a mass ratio $`>`$ 0.25 it is generally assumed that the accretion disc cannot grow larger than the tidal truncation radius, $`r_d`$ (Paczynski 1977; Whitehurst 1988; Osaki, Hirose & Ichikawa 1993), which is approximately given by $`r_d=0.60a/(1+q)`$ for 0.03 $`<q<`$ 1, where $`a`$ is the binary separation (Warner 1995). Given the system parameters ($`P_{orb}`$=2.62168 days; $`q0.39`$; $`i69`$ degrees; $`M_1`$6.7 $`M_{}`$ see Shahbaz et al., 1999) the minimum value for the projected velocity of the accretion disc rim is $``$394 km s<sup>-1</sup>. The observed peak-to-peak half separation of the H$`\alpha `$ emission line (see Figure 1) in late June 1996 is 385$`\pm 8`$ km s<sup>-1</sup>(measured by fitting the profile with a double Gaussian), which implies that the accretion disc is close to its maximum possible size. Soria et al. (1998) estimate the H$`\alpha `$ half peak-to-peak separation to be $`<`$350/2 km s<sup>-1</sup>and $`<`$550/2 km s<sup>-1</sup>for their August/September 1994 and June 1996 H$`\alpha `$ observations respectively, velocities much lower than expected, suggesting that the H$`\alpha `$ emission line arises from non-Keplerian regions/flows in the accretion disc.
## 4 The radial velocity of the secondary star
The radial velocities of the F-type secondary star in Nova Sco 1994 were measured from the spectra by the method of cross-correlation (Tonry & Davis 1979) with a template star. Prior to cross-correlation the spectra were interpolated onto a logarithmic wavelength scale (pixel size 55 km s<sup>-1</sup>) using a $`\mathrm{sin}x/x`$ interpolation scheme to minimize data smoothing (Stover et al. 1980), and then normalised. The template star spectrum (HR2906; F6$`\mathrm{v}`$) was then artificially broadened by 90 km s<sup>-1</sup>(Shahbaz et al., 1999) to account for the rotational velocity of the secondary star. Note that the orbital smearing of the Nova Sco 1994 spectra through the 1800s exposure is at most only 10 km s<sup>-1</sup>, much less than the resolution of the data. Only regions of the spectrum devoid of emission lines (6400-6520Å) were used in the cross-correlation. The radial velocity of the template star (derived using the position of the H$`\alpha `$ absorption line to be $``$7 km s<sup>-1</sup>) was then added to the radial velocities of Nova Sco 1994.
Using the orbital ephemeris given by van der Hooft et al (1998) we phase-folded and binned the heliocentric radial velocities (see Figure 2). From figure 2, it can be seen that the radial velocity measurement at phase 0.2 does not fit the general pattern of the sinusoidal modulation present in the data. This data point was the total of three radial velocity measurements taken on the second night (21st June 1996). Although the seeing and quality of the spectra taken during this night were not as good as the others, no obvious reason could be found as to why these spectra gave much lower radial velocities than expected. A sine wave fit to all the data points does not give an adequate fit ($`\chi _\nu ^2`$=6.9). However, removing the discrepant data point and then performing a sine wave fit yields a $`\chi _\nu ^2`$ of 1.5, a semi-amplitude $`K_2=279\pm 10`$ km s<sup>-1</sup>, systemic velocity $`\gamma =155\pm 7`$ and a phase shift of $`0.043\pm 0.005\varphi `$ (1-$`\sigma `$ errors are given). We also fitted the radial velocity curve with an eccentric orbit, but found the fit to be less than 50 percent significant.
## 5 The effect of irradiation on the secondary star’s radial velocity
Three absorption line radial velocity curves have been obtained for Nova Sco 1994, using the same absorption features of the F6$`\mathrm{iv}`$ secondary star and the standard method of cross-correlation. However, in each case the system was observed to be in a different X-ray state. A sinusoidal fit to the outburst data taken in April/May 1995 of Orosz & Bailyn (1998) gives a radial velocity semi-amplitude of $`K_{obs}`$=230$`\pm 2`$ km s<sup>-1</sup>. During this period BATSE did not detect the source, so we can only put an upper limit of 2.4$`\times `$10<sup>36</sup> erg s<sup>-1</sup>($`<`$0.03 photons erg cm<sup>-2</sup> s<sup>-1</sup> in the BATSE 20-350 keV energy range) to the X-ray luminosity of the source. This upper limit alone does not allow us to state unequivocally that the source was not active at X-ray energies, but optical observations suggest that the source was not in quiescence ($`V`$=16.5; Orosz & Bailyn 1998). In section 4 we determined $`K_2`$=279$`\pm 10`$ km s<sup>-1</sup>from data taken in June 1996 when RXTE ASM (2-12 keV) observations give an X-ray luminosity of $`L_x`$=6.8$`\times `$10<sup>37</sup> erg s<sup>-1</sup>, and the $`R`$-band brightness was $``$ 1 mag brighter than its quiescent value. The BATSE (20–350 keV) count rate was at least a factor of 4 higher than in April/May 1995. Shahbaz et al. (1999) determined the true radial velocity of the secondary star ($`K_2`$=215.5$`\pm 2.4`$ km s<sup>-1</sup>) in 1998 May/June, when the source was finally in optical quiescence. The only X-ray quiescent observations were obtained during March 1996 using ASCA (1-10 keV; Robinson et al., 1997) which gave $`L_x`$=2$`\times `$10<sup>32</sup> erg s<sup>-1</sup>.
In Figure 3 we show the observed radial velocity amplitudes relative to the quiescent value as a function of the observed X-ray luminosity at the time of the measurements. We have converted the X-ray luminosities, which were observed with different instruments, into a common energy range (0.4–10 keV) using a hydrogen column density of $`N_h=0.89\times 10^{22}`$ cm<sup>-2</sup> and a photon power-law model with indices 2.8 and 1.5 for the X-ray high and quiescent states respectively (see Table 3; Zhang et al., 1997; Robinson et al., 1997; Hameury et al., 1997). This energy range is where we expect the total radiated power for X-ray transients in both outburst and quiescence to lie (Chen, Shrader & Livio 1997). Note that there is a correlation between X-ray luminosity and the observed radial velocity semi-amplitude; the higher the X-ray luminosity the larger the observed radial velocity semi-amplitude, exactly as expected. We can use our model to estimate the X-ray luminosity at the time when Orosz & Bailyn (1997) took their radial velocity measurements. We find $`L_x5\times 10^{35}`$ erg s<sup>-1</sup>which is consistent with the BATSE upper limit.
## 6 Irradiation of the secondary star
It has been known for some time, especially in studies of dwarf novae and polars, that substantial heating of the secondary star shifts the effective centre of the secondary, weighted by the strength of the absorption lines, from the centre of mass of the star. This results in a significant distortion of the radial velocity curve leading to a spuriously high semi-amplitude and a radial velocity curve that may be eccentric. Davey & Smith (1992) describe a procedure for detecting the effects of irradiation on the radial velocity curve of the secondary star, whereby one tests the significance of an eccentricity in the orbital solution. However, it should be noted that, although our data does not allow this eccentricity test, due to the poor orbital phase coverage, we can use the spuriously high radial velocity semi-amplitude to show that X-ray heating is present.
In order to investigate the effects of X-ray heating on the secondary star’s radial velocity curve we used the model described by Phillips, Shahbaz & Podsiadlowski (1999). The model uses a crude treatment for X-ray heating, since no satisfactory robust model exists for the effects of external heating in stars. However, it serves to illustrate the extreme effects of X-ray heating. It should be noted that the first order model of X-ray heating by Brett & Smith (1993), which does not include energy transport effects, does show that the whole temperature structure of the outer layers of the secondary is upset by external heating. Figure 3 shows the effects of different amounts of X-ray luminosity on the secondary star’s radial velocity amplitude. $`K_{\mathrm{obs}}`$ is computed by fitting the predicted curve with an eccentric orbit. The regions on the secondary star that are heated do not contribute to the absorption line flux. The maximum possible change that irradiation can have on $`K_{obs}`$, based purely on geometry, is 15 per cent. However, from our data presented in this paper, we observe $`\mathrm{\Delta }K_2/K_2`$=0.30$`\pm `$0.05, which when compared with maximum possible value based on geometry, is significant at the 3-$`\sigma `$ level.
In Figure 4 we show how much of the secondary star’s surface needs to be heated in order to produce the observed radial velocity amplitude. We find that based purely on geometry 35 per cent of the secondary star’s surface is directly heated by X-rays produced at the compact object. (This fraction only depends on the shape and size of the secondary star, which in turn is determind by the $`q`$. Using the extreme values for $`q`$ (Shahhaz et al., 1999), we find that this fraction changes by less than 1 per cent.) However, in order to produce the observed large radial velocity semi-amplitude, 60$`{}_{7}{}^{}{}_{}{}^{+9}`$ per cent of the secondary star needs to be heated. The 1-$`\sigma `$ uncertanties quoted here were estinated using the 1-$`\sigma `$ uncertanties in $`\mathrm{\Delta }K_2/K_2`$. This result may seem surprising at first, since one expects only the regions of the secondary star facing the compact object to be irradiated and yet our result implies that some of the regions not directly seen by the compact object are also affected by irradiation. However, one should note that effects such as X-ray scattering and irradiation-induced flows on the surface of the secondary star (Phillips & Podsiadlowski 1999) can increase the fraction of the secondary star that responds to the X-ray source. Note that the regions on the secondary star that are shadowed by the accretion disc will be indirectly heated by such mechanisms. Therefore $`K_{\mathrm{obs}}`$ can be larger than that expected from heating the inner face of the secondary star alone.
## 7 Discussion
The existence of circulation in rotating stars was first proved in 1924 by von Zeipel (von Zeipel 1924). He demonstrated that for a rotating homogeneous star, the radiative transport equation and equation of conservation of energy cannot be fulfilled simultaneously. This results in the formation of meridional motions. In order to maintain a stationary state as assumed, one has to demand that these meridional motions contribute to the energy transport. In the case of an irradiated rotating star, the situation is still more complicated, since the radiation will induce additional circulation currents.
Evidence for the existence of significant irradiation-driven circulation is provided by several sources. For example, the analysis of the optical light curve of HZ Herculis has shown this to be heated by its accompanying X-ray source HER X-1. Although the main features of the optical light variation are well understood (HZ Her is bright when the X-ray source is in front of it, its brightness is reduced during the occultation of the X-ray source by the secondary), the minimum at phase 0.0 is sharper than expected and indicates some additional source of optical radiation at this phase. Strittmatter et al. (1973) tried to explain this via the illumination of the disk by HZ Her. Other attempts were made by Pringle (1973) and Bahcall, Joss & Avni (1974). However, the most successful explanation was due to Kippenhahn & Thomas (1979). They estimated the energy transported from the X-ray illuminated part of the stellar surface to the shadowed side, and demonstrated that the minimum at phase 0.0 could be reasonably well accounted for (X-ray heating without horizontal transport leads to a flat minimum at phase 0.0).
In addition, Schandl et al. (1997) found circulation to be necessary in order to accurately model the optical light curve of CAL 87, an eclipsing supersoft X-ray source. They calculated the light curve based on the assumption that an accreting, steadily burning white dwarf irradiates the accretion disk and the secondary star, as suggested by van den Heuvel et al. (1992). A simple description of energy transport on the secondary surface was used and then integrated over the whole surface, while conserving the total luminosity. They found that significant energy transport of the irradiated flux to non-illuminated parts on the secondary surface is required to simulate the observed lightcurve, particularly around the primary eclipse, when the shadowed hemisphere of the secondary is in view.
Recent models for irradiation-induced flows in binary stars have been computed by Martin & Davey (1995). They considered circulation in gently-heated secondary stars (where the incident flux is less than the intrinsic flux). Their 2-dimensional calculations included the effects of the Coriolis force and showed upwelling of hot material being carried preferentially towards the direction of rotation of the star. They also concluded that all secondary stars should show asymmetric heating, because of the presence of Coriolis forces. Phillips (1999) has recently extended the study of circulation to 3-dimensions. As well as including the effects of X-ray irradiation i.e. the anisotropic heating of the irradiated surface, and the effects of surface radiation stress, he also considers the large-scale effects of the rotation of the system and includes an approximate treatment of the Coriolis force. His results suggest a realistic analysis of the Coriolis force is essential for a full description of stellar circulation.
In order to study the extent of irradiation of the secondary star one requires good quality spectrophotometric studies throughout an X-ray outburst, during which the level of X-ray irradiation and induced heating changes. This will allow the surface intensity distribution across the secondary star to be mapped (see Rutten & Dhillon 1994 and Davey & Smith 1996), from which effects such as irradiation-induced circulation or star-spots can be investigated.
## Acknowledgements
We would like to thank the referee, Prof. Robert Smith for useful comments. |
no-problem/0001/astro-ph0001431.html | ar5iv | text | # A GRB Tool Shed
## Introduction
We are implementing a suite of software tools to aid Gamma-Ray Burst (GRB) researchers in working with the GRB data. The major features of the tool shed — a SHell for Expeditions using Datamining — are a web-based data query facility, web-based data visualization capability, and a web-based interface to data mining software tools. The tool shed maintains a database of users allowing each user to store their own work at the tool shed site. Each user’s data will not be visible to other users of the system.
Our GRB tool shed is populated with a standard set of preprocessed GRB data such as the basic table data, flux/fluence data, and duration 4BCAT . These data are stored in tabular form as rows (burst instances) and columns (attributes). Each burst has the same attributes as all other bursts, with a provision for indicating “missing” attributes. Users may augment this database by uploading their own table or performing SQL database queries for data selection and calculations.
The data mining tools can be given any of the queried data and produce either rules for classification of bursts or an identification of classes of bursts by identifying which bursts belong to which class. Note that these “identified” classes are not necessarily classes based on physical properties of bursts; they may be due to instrumental bias, or even statistical fluctuations from the small numbers of instances (c.f. HHP+99 ). This step can lead to defining new data in new tables for further exploration. An example of applying these tools is given in HHR+99 .
### Data Storage
Once a registered user has logged in, the data can be manipulated in a variety of ways. The user may simply query the existing data, selecting a subset of the bursts and/or attributes for later processing. They may augment the data by uploading their own attributes for existing bursts. Or they may augment the data by uploading information for new bursts.
As an example database query, consider selecting all bursts from the 4B catalog with high relative measurement errors on the channel 1 fluence measurement. To determine what might be considered “high”, it may be necessary to see all of the bursts sorted by relative measurement error, easily done with this SQL query:
> SELECT burstnum, channel1FluenceError FROM 4BFluxTable
> WHERE channel1Fluence \<\> 0
> ORDER BY (channel1FluenceError/channel1Fluence);
Now, after viewing the results of this query, one may decide that a “quality” threshold of 1.0 standard deviations would be required on the fluence data. The quality data can be extracted using this query:
> SELECT burstnum, channel1Fluence, channel1FluenceError
> FROM 4BFluxTable WHERE channel1Fluence \<\> 0
> AND (channel1FluenceError/channel1Fluence) \> 1.0;
For those users unfamiliar with SQL, the web application will provide point-and-click, fill-in-the-box forms for generating a database query. These web pages should provide the user with most of the features they would be interested in. There will also be a blank form for users who may wish to enter the SQL query directly. Either way, the names of the tables and attributes will be shown on the web browser so the user need not memorize them.
### Data Visualization
At any time the user may decide to invoke visualization tools to help “see” the data in their scratch area or in the system tables. The web application will send these requests to ION (IDL On the Net) to produce graphical views of the data. The display of these graphs will be done on the user’s web browser window.
### Data Mining Tools
Several data mining tools will be available as part of the web application. Initially, there will be the classic classification tool, C4.5 C45 , a tool developed by a member of our group, ESX RGHH99 , and at least one Neural Network package. These software packages run on the web server and the web application will guide the user, requesting information needed by the specific data mining tool being invoked.
The output from the data mining tools differ from one tool to the next. Our web application will be able to capture the output and transform it to an internal “rule” format. Once captured, the user may view the rules, or go even one step further by applying the rules to a database in their scratch area. The application of the rules to the data is a significant feedback component of the tool shed.
### Online Help
There will be extensive online help to guide the user through the data mining process. The help system will be written in a hyper-linked book format complete with a table of contents, an index, and a search engine. There will also be context-sensitive help in the sense of hyper-links from web forms to relevant pages of the help system. The help system will address how to use the tool. The GRB data will be minimally documented, with pointers (hyper-links) to existing help in understanding the meaning of the data attributes.
A tutorial will be provided that brings the user through a data mining session. This tutorial will provide (scientific) rationale for selecting options along with way. Although the tool shed need will not be specific to GRB data, the tutorial will be.
## GRB Tool Shed System Platform
The GRB tool shed will initially run on a pentium-class computer system running RedHat Linux. To maximize portability, the Java language will be used. Since this tool is web-based, all of the Java code will run as Java Servlets. The Apache web server will be used along with its companion Java Servlet Apache-JServ. The GRB data will be stored in a PostgreSQL database with access to the Java code provided via JDBC. All of these software packages are available on many platforms, including Unix, Windows, and Mac, and they are all freely available. The flow of information through these packages is shown in Figure 1.
This whole process is initiated by aiming a web browser at the appropriate URL (http://grb.mankato.msus.edu/) where the Apache web server is configured to start a Java Servlet via Apach JServ. The Java Servlet will then make requests to the PostgreSQL database via the JDBC package that comes with PostgreSQL. The major development effort for this application is in the creation of the Java Servlet code, which we call “the web application.”
## GRB Tool Shed Data Flow
The Java Servlet application is a complex set of Java code with many data structures and interfaces. There are two major data formats that will be used: Standard User Interface Format (SUIF) and Standard Internal Classifier Format (SICF). The SUIF will be used when presenting data to the user on the client machine (web browser). And SICF will be used when presenting data to any of the supported classifiers (data mining tools). The SICF format is proposed as a tool-independent representation format for holding all information necessary to conduct a classification/data mining run.
We expect to provide a user interface on the web browser that looks very much like a spreadsheet program. This familiar view will allow the user to inspect and possibly update data easily.
Note that a “donut” object indicates a data format conversion is needed at that point. Some of these data format conversions may require additional information, which would require the use of web pages to interact with the user. And the wide arrows indicate data flowing across the internet from server to client (browser).
The control of when and where the data flows is completely independent of this diagram. Imagine a controller sitting above this page directing the data to flow along the various paths of the diagram. This controller is a web-based menu system under the direction of the user.
## Web Pages (Control Flow)
The design of this perspective of the GRB Tool Shed is least developed at this time. It is clear there needs to be a login menu screen. Once that information has been verified, the user will be placed in a main menu/dispatcher web page. This page will allow the user to click on various functions on a menu bar either along the side, the top, or the bottom. The menu selections/actions might be:
| $``$ graph/visualize data | $``$ show data in spreadsheet form |
| --- | --- |
| $``$ initiate upload/download of data | $``$ work with the data (query database) |
| $``$ initiate classification | $``$ view documentation |
Each of these items constitutes a large implementation effort. For example, the “work with the data” selection initiates a sequence of web pages/forms where the user is required to make selection criteria for the various rows and columns of the data. The “initiate classification” selection will cause the user to be prompted for which classifier tool to invoke, then guide the user through selecting parameter values that are specific to that tool.
## Conclusions
With the creation of this software comes a powerful research tool capable of automating many aspects of manipulating GRB data. Our goal is to go beyond the tool creation and build up the collection of attributes about the GRBs, emerging as a significant repository of GRB information with a built-in efficient methodology.
Our implementation strategy is to incrementally develop components. That way, as the development progresses, the web application will contain some usable software and data. |
no-problem/0001/cond-mat0001456.html | ar5iv | text | # Antiferromagnetic Order of the Ru and Gd in Superconducting RuSr2GdCu2O8
\[
## Abstract
Neutron diffraction has been used to study the magnetic order in RuSr<sub>2</sub>GdCu<sub>2</sub>O<sub>8</sub>. The Ru moments order antiferromagnetically at $`T_N=136(2)K`$, coincident with the previously reported onset of ferromagnetism. Neighboring spins are antiparallel in all three directions, with a low T moment of 1.18(6) $`\mu _B`$ along the c axis. Our measurements put an upper limit of $``$0.1 $`\mu _B`$ to any net zero-field moment, with fields exceeding $`0.4T`$ needed to induce a measurable magnetization. The Gd ions order independently at $`T_N=2.50(2)K`$ with the same spin configuration.
. \]
The ruthenate class of materials has been the focus of considerable work recently because of their interesting magnetic and superconducting properties. SrRuO<sub>3</sub>, for example, is a 4d band ferromagnet that orders at 165 K, while Sr<sub>2</sub>RuO<sub>4</sub> is an exotic p-wave superconductor ($`T_C=1.5K`$). Of particular interest here is the recent report of ferromagnetic ordering of the Ru at 133 K in RuSr<sub>2</sub>GdCu<sub>2</sub>O<sub>8</sub>, while bulk superconductivity is established at lower temperatures as observed in susceptibility and specific heat. In these hybrid ruthenate-cuprate systems both the Cu-O and Ru-O planes form very similar square-planar arrays, and the coexistence of superconductivity and long range magnetic order at high temperatures is intriguing. Previous “magnetic-superconductors” such as the Chevrel phases ($`RMo_6S_8`$, $`R`$ = rare earth ion), borocarbides ($`RNi_2B_2C`$), and cuprates ($`RBa_2Cu_3O_7`$ and related materials) show rare earth ordering at low temperature ($`10K`$ ), and almost all are antiferromagnets that do not couple strongly to the superconductivity. The rare occurrence of ferromagnetism, as found in $`ErRh_4B_4`$, $`HoMo_6S_8`$, and $`HoMo_6Se_8`$, revealed the strongly competitive nature of these two cooperative phenomena in the form of long wavelength oscillatory magnetic states at low temperature ($`1K`$) and a ferromagnetic lock-in transition that quenches the superconductivity. It would then be quite interesting if RuSr<sub>2</sub>GdCu<sub>2</sub>O<sub>8</sub> were a ferromagnetic superconductor with such a high magnetic ordering temperature, as this would suggest a superconducting order parameter of the Fulde-Ferrell-Larkin-Ovchinnikov type that could exhibit $`\pi `$-phase behavior. Our diffraction results, however, demonstrate that the magnetic order of the Ru is predominantly antiferromagnetic, instead making this by far the highest known antiferromagnetic ordering to coexist with superconductivity. An upper limit of $`0.1\mu _B`$ is obtained for the ferromagnetic component, consistent with recent magnetization data, and the system is then similar to $`ErNi_2B_2C`$, where a net magnetization develops below 2.3 K. The Gd moments also order magnetically, but at low temperatures in a manner analogous to previous magnetic-superconductor systems.
A polycrystalline sample of RuSr<sub>2</sub>GdCu<sub>2</sub>O<sub>8</sub> was prepared by the solid state reaction technique, using the <sup>160</sup>Gd isotope to avoid the huge nuclear absorption cross section of natural Gd. The single-phase sample weighed $`1.5`$ g, has an onset superconducting temperature of 35 K and bulk superconducting T<sub>C</sub> of 21 K, and is the identical sample used in a previous susceptibility and neutron crystallographic study. All the present neutron data were collected at NIST. BT-2 was employed at a neutron wavelength of 2.359 Å, with a pyrolytic graphite filter to suppress higher-order wavelengths. Polarized neutron measurements were carried out with a Heusler monochromator and analyzer. A <sup>3</sup>He refrigerator was employed for the lowest temperature measurements, and a vertical field 7T superconducting magnet for the field measurements. Small angle neutron scattering data were also collected with a wavelength of 5Å on the NG-1 spectrometer from 6K to 300K. Statistical uncertainties quoted in this article represent one standard deviation.
Fig. 1 shows a portion of the diffraction pattern obtained at a temperature of 16 K. The peak at 23.5 is the weak {002} nuclear Bragg peak; for comparison, the strong {103}+{110} Bragg peak at 51.5 has 4809 counts/min. The peaks at 25.8 and 30.8 can be indexed as the $`\{\frac{1}{2},\frac{1}{2},\frac{1}{2}\}`$ and $`\{\frac{1}{2},\frac{1}{2},\frac{3}{2}\}`$ reflections. At a temperature of 150 K we see that these peaks have completely disappeared, indicating that they are magnetic and originate from the magnetic ordering of the Ru. There is no change in the nuclear Bragg intensity, where a ferromagnetic component would appear, as clearly indicated by the difference scattering shown at the bottom of the figure. The temperature dependence of the integrated intensity for the $`\{\frac{1}{2},\frac{1}{2},\frac{1}{2}\}`$ peak is shown in Fig. 2. The solid curve is a simple mean field fit to estimate a Néel temperature of 136(2) K. This is in excellent agreement with the reported Ru magnetic ordering temperature.
The data in Fig. 1 show that the Ru moments order antiferromagnetically, with nearest neighbor spins in all three crystallographic directions aligned antiparallel. The magnetic scattering for a collinear structure is
$$I_M=C\left|F_M\right|^2\frac{m_{hkl}A_{hkl}}{\mathrm{sin}(\theta )\mathrm{sin}(2\theta )}1\left(\widehat{\tau }\widehat{M}\right)^2$$
(1)
where $`C`$ is an instrumental constant, $`m_{hkl}`$ is the multiplicity of the powder peak with Miller indices $`hkl`$ for the reciprocal lattice vector $`\tau `$, $`A_{hkl}`$ is the absorption factor, $`\widehat{M}`$ is a unit vector in the direction of the moment, and the brackets indicate a powder/domain average. The magnetic structure factor is given by
$$F_M=\underset{j=1}{\overset{N}{}}\mu _j^zf_j(\tau )e^{i\tau 𝐫_j}e^{W_j}$$
(2)
where $`\mu _j^z`$ is the ordered moment and $`f_j(hkl)`$ is the magnetic form factor for the $`j^{th}`$ ion at position $`𝐫_j`$ in the unit cell, W<sub>j</sub> is the Debye-Waller factor, and the sum is over all atoms in the unit cell. The magnetic intensities can be put on an absolute scale by comparison with the nuclear intensities. The intensity for the $`\{\frac{1}{2},\frac{1}{2},\frac{3}{2}\}`$ peak in Fig. 1 is clearly reduced in intensity compared to the $`\{\frac{1}{2},\frac{1}{2},\frac{1}{2}\}`$ peak, and the orientation factor in Eq. (1) then suggests that the direction of the moment is along the tetragonal c axis. The observed intensity ratio of the $`\{\frac{1}{2},\frac{1}{2},\frac{1}{2}\}`$ to $`\{\frac{1}{2},\frac{1}{2},\frac{3}{2}\}`$ peaks is 2.49(40), which is indeed in good agreement with the calculated value of 2.21. Of course, with only two observable magnetic peaks this moment direction assignment is tentative rather than definitive. The ordered Ru moment at low temperatures is then 1.18(6) $`\mu _B`$, which agrees nicely with the moment of 1.05(5) $`\mu _B`$ obtained from susceptibility.
Any ferromagnetic contribution to the scattering will occur at the same positions as the nuclear Bragg peaks, and the data in Fig. 1 indicate no magnetic contribution to the {002} peak within experimental error. We measured the intensities of the ten lowest-angle nuclear Bragg reflections above and below the Ru magnetic ordering temperature, and these data provide an upper limit of $``$0.1 $`\mu _B`$ to any ferromagnetic component. This result was substantiated by polarized beam measurements, although the error limit was comparable to that obtained with unpolarized neutrons. We also monitored the flipping ratio of the intensity transmitted through the sample to determine if there were any depolarization of the beam as might be expected for a ferromagnet. No change with temperature was observed. Finally, we measured the small angle scattering on NG-1 in an attempt to see any critical scattering associated with ferromagnetic correlations that might develop, but no scattering was observed. Therefore the neutron data so far do not reveal the ferromagnetic component associated with the Ru ordering.
The field dependence of the magnetic scattering of the $`\{\frac{1}{2},\frac{1}{2},\frac{1}{2}\}`$ and $`\{002\}`$ peaks, corresponding to the antiferromagnetic order and induced ferromagnetic moment, respectively, is shown in Fig. 3. A temperature of 80 K was chosen for these measurements since this is well below the Ru ordering temperature so that the sublattice magnetization is near its saturated value, but it is high enough in temperature that the Gd paramagnetic moment should not dominate the net magnetization except at the highest fields. No significant change in either intensity is observed up to $`0.4T`$. With further increase of field the intensity of the antiferromagnetic reflection begins to decrease, while the induced magnetization increases. At the highest field of $`7T`$ there is no significant antiferromagnetic intensity remaining (as indicated by full angular scans), while the induced magnetization corresponds to a net moment of 1.4(1) $`\mu _B`$ perpendicular to c. The calculated induced Gd paramagnetic moment is shown by the dashed curve, and for fields above $`0.4T`$ the data systematically lie above the curve indicating a Ru contribution. However, the value is $`0.2`$ $`\mu _B`$, which suggests the Ru moments are rotating into another antiferromagnetic (spin-flop) structure, rather than becoming fully aligned with the field. Returning to zero field, all the peaks recover their zero-field intensities, indicating that the effect of the field is reversible, and also that no preferred orientation of the loose powder particles occurred when the field was applied. The Gd anisotropy is very small since it is an S-state ion. The low spin-flop field and the lack of any field-induced preferred orientation then suggests that the Ru crystalline anisotropy is also relatively weak, indicating that a Heisenberg Hamiltonian is appropriate to describe the Ru spin system.
We now turn to measurements of the Gd magnetic order. Fig. 4 shows the magnetic diffraction pattern, obtained by subtracting the data at 5 K from the data at 1.6 K, below T<sub>N</sub>. We see that the peak positions and relative intensities are identical to that for the Ru, so that nearest-neighbor Gd spins are also coupled antiferromagnetically along all three crystallographic directions, with the moment direction along the tetragonal c axis. The calculated and observed intensities then agree to within the statistical uncertainties.
The temperature dependence of the sublattice magnetization for the Gd is shown in Fig. 5. We obtain the expected 7 $`\mu _B`$ moment within experimental uncertainties at low T. Near T<sub>N</sub> a small correction has been applied to the observed intensity to account for critical scattering, and the solid curve is a fit to a modified power law, with a fitted ordering temperature of 2.50(2) K. The sharpness of the ordering might at first be surprising since the Ru and Gd magnetic structures are identical, and hence one might expect them to be strongly coupled, smearing the Gd order parameter. However, the Gd ions sit at the body-centered position of the simple tetragonal Ru lattice. The antiferromagnetic magnetic structure for the Ru then results in a cancellation of the average interaction between the Gd and Ru (Fig. 4), rendering the two spin systems fully frustrated (neglecting quantum fluctuations) with respect to each other and thus behaving independently to a good approximation. It is noteworthy that this frustration is relieved by a (zero-field) ferromagnetic component on the Ru sublattice, so that the sharpness of the Gd order parameter is another indication that the Ru magnetic structure can have only a modest net moment.
The Ru antiferromagnetic moment we observe is in good agreement with the moment obtained from susceptibility, accounting for essentially the full ordered moment. Our experimental upper limit of $`0.1\mu _B`$ on the ferromagnetic component is consistent with the spontaneous moment derived from low-field data, but is inconsistent with band structure calculations predicting full ferromagnetic spin polarization of the Ru subsystem. The crystallographic data indicate a rotation of the RuO<sub>6</sub> octahedra about the c-axis, with a small rotation around an axis perpendicular to c. It is the latter rotation that would be needed for conventional mechanisms such as antisymmetric exchange or single-ion anisotropy to produce a canting and net Ru moment as observed in a variety of measurements. In any case, the antiferromagnetism is dominant, and the coexistence of superconductivity with the high ordering temperature and consequent large exchange interactions of the Ru make this an especially interesting system for further investigations.
Acknowledgments. We would like to thank R.W. Erwin and Q. Huang for their assistance, and J. D. Jorgensen for communicating their results prior to publication. |
no-problem/0001/astro-ph0001171.html | ar5iv | text | # 2 µm Spectroscopy within 0.┴{′′}3 of Sgr A∗ 1footnote 11footnote 1Data presented herein were obtained at the W.M. Keck Observatory, which is operated as a scientific partnership among the California Institute of Technology, the University of California and the National Aeronautics and Space Administration. The Observatory was made possible by the generous financial support of the W.M. Keck Foundation.
## 1 Introduction
Recent high-resolution near-infrared imaging reveals a tight cluster of at least a dozen stellar sources projected within 0$`\stackrel{}{\mathrm{.}}`$5 of the putative massive black hole in the Galactic Center (Genzel et al. 1997; Ghez et al. 1998). Genzel et al. (1997) suggest that this cluster contains early-type stars with initial masses $``$15 to 20 M. While their low-resolution spectra (R $``$ 35) and photometry can be fit by early-type stars, they can also be fit by much lower mass K giants of a few M. The implications for star formation near a massive black hole are heavily dependent on whether the stars are of early or late type. In the former case, the stars are a few Myr old and probably formed very near to the central black hole. In the latter case, they are on the order of a Gyr old and represent a central concentration of the general old population seen throughout the Galactic Center (Alexander 1999). If stars have formed near the central black hole, then it is important to know the physical properties of the gas there. The intensity of recombination line emission can be used to constrain the gas density and ionizing environment near the center, and the line width might be used to probe the central mass to smaller size scales than the stellar velocity dispersions.
With these issues in mind, we obtained high resolution (R $``$ 22,400) and moderate resolution (R $``$ 2,700) long-slit spectra of the central few arcseconds of the Galaxy in the 2.0$``$2.4 µm region (K-band). We present spectra of the combined light from the central stellar sources (hereafter the “S sources”; Genzel et al. 1997), nearby stars, and ionized gas.
## 2 Observations and Data Reduction
The observations were obtained with NIRSPEC, the facility near-infrared spectrometer, on the Keck II telescope (McLean et al. 1998, 1999). A log of observations is given in Table 1. The plate scale was measured by comparing the locations of the spectra of IRS7, IRS16NW, IRS33E, and IRS33W to the positions in Eckart & Genzel (1997), giving 0$`\stackrel{}{\mathrm{.}}`$14$`\times `$0$`\stackrel{}{\mathrm{.}}`$20 per pixel in the spectral$`\times `$spatial directions of the high-resolution mode, and 0$`\stackrel{}{\mathrm{.}}`$20$`\times `$0$`\stackrel{}{\mathrm{.}}`$14 per pixel in the spectral$`\times `$spatial directions of the low-resolution mode (note that the axes are flipped with respect to the camera in the two modes). The slit viewing camera (SCAM) was used to obtain images simultaneously with the spectra. We measured a plate scale for the SCAM of 0$`\stackrel{}{\mathrm{.}}`$18 per pixel by comparing the locations of IRS16NW, IRS16NE, IRS33E, and IRS33W to those given by Eckart & Genzel (1997). From SCAM images, we estimate seeing (FWHM) of 0$`\stackrel{}{\mathrm{.}}`$5 on 28 April 1999 and 0$`\stackrel{}{\mathrm{.}}`$3 on 3 June 1999. We chose to use the 3-pixel-wide slit (0$`\stackrel{}{\mathrm{.}}`$43) in high-resolution mode and the 2-pixel-wide slit (0$`\stackrel{}{\mathrm{.}}`$39) in low-resolution mode in order to match the seeing.
Quintuplet Star #3, which is featureless in this spectral region (see Figure 1 in Figer et al. 1998), was observed as a telluric standard (nomenclature from Moneti, Glass, & Moorwood 1994). Arc lamps containing Ar, Ne, Kr, and Xe, were observed to set the wavelength scale. A field relatively devoid of stars (RA 17<sup>h</sup> 44<sup>m</sup> 49$`\stackrel{\mathrm{s}}{\mathrm{.}}`$8, DEC $``$28<sup>°</sup> 54 6$`\stackrel{}{\mathrm{.}}`$8 , J2000) was observed to provide a dark current plus bias plus background image. A quartz tungsten halogen lamp was observed to provide a “flat” image.
All data reduction was accomplished using IRAF routines<sup>2</sup><sup>2</sup>2IRAF is distributed by the National Optical Astronomy Observatories, which are operated by the Association of Universities for Research in Astronomy, Inc., under cooperative agreement with the National Science Foundation.. Bad pixel removal, flat-fielding, and coadding of object-sky frame pairs were performed to produce the final spectral images. The spectra were extracted and then divided by a similarly extracted spectrum of the telluric standard corrected for its apparent spectral energy distribution (Figer et al. 1998).
## 3 Analysis and Results
We have compared the flux in the spectrum to that expected from the stars in the slit (Wizinowich et al. 1999), and the detected flux in the low-resolution spectrum confirms our pointing. The total background-subtracted flux from the combined low-resolution spectrum is K = 12.8, in good agreement with the sum of the previously reported fluxes for the sources, reported as K = 12.6 (Genzel et al. 1997).
Figure 1 shows the low resolution spectra for the S sources, IRS7, and IRS16NW. The spectra demonstrate the very deep CO absorption in the spectrum of a cool star (IRS7) and the total lack of any similar feature in that of a hot star (IRS16NW). The spectra for the S sources were extracted from synthetic apertures centered 1$`\stackrel{}{\mathrm{.}}`$38 south of IRS16NW and 0$`\stackrel{}{\mathrm{.}}`$75 wide to give W$`_{{}_{}{}^{12}\mathrm{CO}(20)}`$ $``$ 6$`\pm `$2 Å. Figure 2 shows that the CO bandhead strength gradually increases in apertures to the south of Sgr A, and to the north of IRS16NW. For instance, W$`_{{}_{}{}^{12}\mathrm{CO}(20)}`$ $``$ 11$`\pm `$2 Å for an aperture centered 0$`\stackrel{}{\mathrm{.}}`$78 to the south of Sgr A. A similar measurement for $`\iota `$ Cep (K0III), whose spectrum was taken from the Kleinmann & Hall atlas (1986), gives W$`_{{}_{}{}^{12}\mathrm{CO}(20)}`$ = 11.5 Å; note that later red giants would have deeper CO absorption. It appears, then, that the off-source spectrum is similar to that of a K giant. The total on-source continuum flux level (with background) is about 1.8 times the off-source flux level. So, if the collective spectrum for the S sources is truly featureless, and the spectrum of the background mimics that of a K giant, then we would expect to measure an equivalent width of 6 Å in the combined spectrum, in good agreement with our measurements. We find that no more than 30% of the light in the composite spectrum can come from K giants or later types. Note that this limit is conservatively determined by using the upper error limit of W$`_{{}_{}{}^{12}\mathrm{CO}(20)}`$ = 8 Å, for the combined light spectrum.
The spectrum of the light from all sources in the central 0$`\stackrel{}{\mathrm{.}}`$39 $`\times `$ 0$`\stackrel{}{\mathrm{.}}`$85 (EW $`\times `$ NS) suggests at least some population of stars earlier than K0, and we find that the nearest blue supergiant, IRS16NW, does not contribute significantly to the light falling in the aperture. The peak of the S sources is 1$`\stackrel{}{\mathrm{.}}`$43 south of IRS16NW, and the point spread function (measured from IRS7) suggests that the light from IRS16NW should contribute $`<`$ 1% of the total flux in the synthetic aperture.
We are also interested in detecting possible emission-line flux from the stellar sources, in order to constrain their nature, and from any gas associated with the black hole, in order to constrain the physical properties of material near the black hole. The high resolution data are particularly useful for this purpose (see Table 1 and Figure 3). We find three distinct components to the Br-$`\gamma `$ emission: 1) a faint “zero-velocity” component which is distributed throughout the low-resolution slit region, and is not easily identified in Figure 3, 2) a very bright, high-velocity, component associated with the “mini-spiral,” and 3) a near-zero-velocity component near the position of Sgr A and extending 1$`\stackrel{}{\mathrm{.}}`$5 to the north and 7″ to the south. The first component can also be seen in the H<sub>2</sub> (2.122 µm) and He I (2.058 µm) lines, suggesting that H I and He I gas are distributed in projection over the whole region, and are being ionized by the ambient radiation field. The third component varies considerably along the north-south direction in peak location, 0 $``$ V<sub>LSR</sub> $``$ +55 km s<sup>-1</sup>, and line width, 25 km s<sup>-1</sup> $``$ V<sub>FWHM</sub> $``$ 110 km s<sup>-1</sup>, gradually increasing from south to north. At the position of Sgr A, the emission is centered at V<sub>LSR</sub> = +15 km s<sup>-1</sup>, and V<sub>FWHM</sub> = 80 km s<sup>-1</sup> at the peak emission. Within 2″ of Sgr A, the line center has a linear gradient of $`+`$18 km s<sup>-1</sup> arcsec<sup>-1</sup> (south-to-north), and the line width has a peak value of +90 km s<sup>-1</sup> at $``$0$`\stackrel{}{\mathrm{.}}`$6 south of Sgr A. The only distinguishing characteristic of the emission at the location of Sgr A with respect to surrounding regions is the local maximum in the line width which is 50% greater than the value $``$ 1″ in either direction. The otherwise nondescript appearance of the line emission near Sgr A suggests that it has little to do with the black hole. Forbidden iron-line emission can also be seen throughout the spectroscopic field. There appears to be a broad \[Fe III\] emission line near 2.2178 µm with V<sub>FWHM</sub> $``$ 1,000 km s<sup>-1</sup>, peaking near the center of the spectroscopic field shown in Figure 2; a similar broad feature is not seen in the Br-$`\gamma `$ line. There is also a velocity component of the \[Fe III\] emission lines at 2.1451 µm, 2.2178 µm, 2.2420 µm, and 2.3479 µm which is narrow and follows the pattern of the mini-cavity, the spatial distribution of which agrees well that described in Lutz, Krabbe, & Genzel (1993).
## 4 Discussion
The S sources span a range of brightness of 14.0 $``$ K $``$ 16.0, implying a range of absolute magnitudes of $``$3.2 $``$ M<sub>K</sub> $``$ $``$1.1, assuming d=8000 pc (Reid 1993) and A<sub>K</sub> $``$ 2.7. The absolute magnitudes match those of O9V to B1V stars having 25 M $`>`$ M<sub>initial</sub> $`>`$ 10 M, and 6 Myr $`<`$ $`\tau _{\mathrm{age}}`$ $`<`$ 20 Myr (assuming the Geneva models, twice mass loss rate, twice solar metallicity, and $`\tau _{\mathrm{age}}`$ $``$Myr old; Meynet et al. 1994). The absolute magnitudes can also be fit by red giants of type K0III to K3III with ages $``$ 1 Gyr. The early-type stars are most easily distinguished from the red giants by the presence of CO absorption in the K-band spectra of the latter. For instance, the deepest CO bandhead in the K-band (2.2935 µm) for an early K giant has W$`_{{}_{}{}^{12}\mathrm{CO}(20)}`$ $``$ 11 Å.
The data suggest that the majority of the S sources in our synthetic aperture are hot stars formed within 0.1 pc of the black hole. This is in agreement with Genzel et al. (1997) and Eckart, Ott, & Genzel (1999). It is unlikely that the observed stars are actually red giants minus their outer envelopes, such as might be produced via stellar collisions (Lacy, Townes, & Hollenbach 1982; Bailey & Davies 1999). Even a particularly luminous red giant (L $``$ 10<sup>4</sup> L) with its envelope stripped to an extent consistent with the lack of strong Br-$`\gamma `$ absorption would have K $`>`$ 16, far too faint to be a likely candidate for the S sources. It is also unlikely that they have formed outside of the center and have been transported inward via dynamical friction. Using Figure 1 and equation 1 in Morris (1993), we find that it would take longer than the lifetime of an O- or B-star to transport them into the center if they were formed more than 0.1 pc from the center.
The requirements for star formation so near the supermassive black hole are extreme. Consider a protostellar clump of sufficient density to form an O-star near the black hole. Such a clump would need to be very dense to be bound against tidal disruption:
$$\rho _{\mathrm{clump}}3.53\frac{M}{R_{\mathrm{GC}}^3},$$
(1)
where R<sub>GC</sub> is the distance between the clump and the Galactic Center, and M is the enclosed mass within the orbit of the clump. Let’s assume that the clump is as far away from the center as possible while still allowing for dynamical friction to operate as described above, i.e., R<sub>GC</sub> $``$ 0.1 pc. With M = 2.6(10<sup>6</sup>M, we find, n<sub>clump</sub> $``$ 4(10<sup>11</sup>) cm<sup>-3</sup>. Since gas near the GC is presently at least 5 orders of magnitude less dense than this, one must appeal to a much denser environment during the formation epoch, or to an event which was exceptionally strongly compressive, or to both.
As discussed by Morris, Ghez & Becklin (1999), the formation of stars this close to the supermassive black hole would inevitably be accompanied by the violent release of accretion energy with a total luminosity near the Eddington limit of the black hole, since the black hole would then be immersed in a relatively dense medium. Indeed, this outpouring of energy may be required to compress the gas to densities sufficient to overcome the tidal forces. Given the challenge of forming stars in this tidally extreme environment, other possibilities might be considered.
For example, have the masses of these stars been built up by stellar coalescence or by continuous accretion, making them much older than we infer? The calculations of Lee (1996) suggest that this is unlikely, although Bonnell, Bate, & Zinnecker (1998) argue that all stars with M<sub>initial</sub> $`>`$ 10 M form by coalescence in very young dense clusters; it remains to be seen if the steady-state conditions in the central parsec can mimic those in the formation epoch of a young cluster. Or are these “stars” really more exotic objects such as compact stars with atmospheres acquired by passage through a dense medium or by collisions with red giant stars (Morris 1993)? Perhaps they are stars powered by dark matter annihilation in their interiors (Salati & Silk 1989), in which case they would be much longer-lived than a star of comparable mass.
In any of these cases, the explanation for the presence of these stars would be exceedingly interesting, and continued investigation well worthwhile.
## 5 Conclusions
We find that about half a dozen of the stars projected within a few thousand AU of Sgr A have little, if any, CO absorption in their K-band spectra, indicating that the stars are hot. Coupled with their brightnesses, we suggest that the stars are OBV types, and therefore $`<`$ 20 Myr old. Given the lifetimes of such stars, it is improbable that they formed beyond 0.1 pc of the Galactic Center, forcing us to consider the possibility that gas clumps having n $``$ 10<sup>11</sup> cm<sup>-3</sup> can exist within a few thousand AU of a supermassive black hole.
It is a pleasure to acknowledge the hard work of past and present members of the NIRSPEC instrument team at UCLA: Maryanne Angliongto, Oddvar Bendiksen, George Brims, Leah Buchholz, John Canfield, Kim Chin, Jonah Hare, Fred Lacayanga, Samuel B. Larson, Tim Liu, Nick Magnone, Gunnar Skulason, Michael Spencer, Jason Weiss and Woon Wong. In addition, we thank the Keck Director Fred Chaffee, CARA instrument specialist Thomas A. Bida, and all the CARA staff involved in the commissioning and integration of NIRSPEC. We especially thank our Observing Assistants Joel Aycock, Gary Puniwai, Charles Sorenson, Ron Quick and Wayne Wack for their support. Finally, we thank Diane Gilmore of STScI for assisting in preparing the figures. |
no-problem/0001/nucl-ex0001004.html | ar5iv | text | # Emission time scale of light particles in the system Xe+Sn at 50 AMeV. A probe for dynamical emission ?
## 1 Introduction
The microscopic description of hot nuclear matter is fundamental to the understanding of energetic heavy ion collisions. Many of the theoretical models developed so far suffer from a shortage of observables to directly test the properties of nuclear matter early after the reaction. An example is the time scale of the reaction. By the very nature, experimenters observe only the signals reaching the detectors at infinite time. Subsequently only asymptotic properties of particles leaving the reaction zone are monitored. This shortcoming is partly bypassed in studies using interferometry of light particles -. It has been shown that the two-body correlation function is not only sensitive to emission time and source size, but also to charge and slope parameter of the source which affect the velocity distribution of the particles and thus the relative distance between them.
It is quite obvious that a 4$`\pi `$ detector array can be superior to a finite angle hodoscope, as used in many previous investigations, provided the granularity (angular resolution), energy threshold, energy resolution and the particle identification are of sufficient quality. A good 4$`\pi `$ setup allows minimally biased event selection avoiding many possible distortions.
At GANIL, the system Xe on Sn has been extensively investigated at 50 AMeV using the 4$`\pi `$ multidetector INDRA , , . Here experimental data are further analyzed by means of correlation functions aiming at the correlation properties of light particles . They allow to perform event by event the following tasks : (i) determine an ’experimental’ impact parameter, (ii) select and characterize the emission sources, and (iii) build the correlation functions. Instead of imposing a spatially fixed particle correlator, we now can use a correlator, which fully exploits the event topology. This correlator continuously adapts to the kinematical configuration of every single collision. So each emission process can be characterized individually.
This study focuses on two-proton (p-p), two-deuteron (d-d) and proton-deuteron (p-d) correlation functions for the system $`{}_{}{}^{129}Xe+^{Nat}Sn`$ at 50 AMeV which aims at providing time scales for the emission of the light charged particles. This should shed light on the type of processes, in particular if the reaction is dominated by preequilibrium emission occurring at the early stage of the reaction (typically with times between 0 and 100 fm/c) or by thermal emission from the projectile and target like fragments (typically with times of a few hundreds fm/c). With the knowledge on the chronology pattern, the emission time of deuterons might give a hint to the mechanism for the production of this lowly bound particle in hot nuclear matter. For this purpose the design of the INDRA detector is well-suited . A large range of excitation energies (up to 12 AMeV) is covered and the light charged particle multiplicity which can be measured is adequate for our selection criteria.
With a conventional reaction picture in mind, and supported by recent studies showing that a sizeable fraction of fragments are emitted in the mid-rapidity region , only thermalized particles are expected in the forward hemisphere of the momentum space of the projectile-like source. Such an assumption has for example important consequences for the estimation of the excitation energy and the slope parameter of the projectile source , . Conversely semiclassical calculations of heavy ion reactions in this energy domain have shown that the projectile-like and the prompt emissions from the interacting zone present a large overlap in their rapidity distributions . We have constructed our correlation functions with particles selected in this forward region to find out if the thermalized component is really the single contribution.
The extraction of the emission time was performed with a quantum model whose interesting feature is to take into account the Coulomb effect of the source charge by analytically solving the three-body problem , .
## 2 INDRA setup and light charged particle resolutions
The experiment was performed at the GANIL facility where the INDRA detector has been installed with a target of 350 $`\mu g/cm^2`$. The beam intensity was limited to 0.4 nA to avoid a saturation of the data acquisition software.
INDRA has been designed to maximize the detection efficiency of charged particles at intermediate energy. It reaches a total detection efficiency of 90%. The fine granularity chosen is such that double counts stay below 5%. INDRA consists of an array of 336 modules reparted on 17 rings centered along the beam axis. Each module is a telescope composed of an ionization chamber (ChIo) filled with $`C_3F_8`$ gas followed by a Cesium Iodide (CsI) scintillator. For forward angles below 45 degrees, the resolution is further improved by insertion of a 300 $`\mu m`$ silicon (Si) wafer between the ChIo and the CsI. With $`\mathrm{\Delta }EE`$ methods in the telescope, the charge identification goes up to Z=54. Isotopic resolution (PID) is obtained for Z=1,2 (and up to Z=5 for ring 2 to 8) by pulse shape analysis of the CsI light output. Lowest energy threshold for the identification of protons and deuterons is 6 MeV using the matrix \[CsI(fast) + Si\] versus \[CsI(slow)\].
In particular p-p correlation functions require the resolution of very small relative momenta, less than 20 MeV/c. A minimum relative momentum of 10 MeV/c can be reached for forward angles under $`20^o`$. For the determination of the particle coordinates, the angle from the target to the middle of the detector has been used instead of a random distribution over the spatial extension. The energy resolution of light particles is between 100 keV and 200 keV depending on the module. With exception of INDRA’s first ring ($`\mathrm{\Delta }\mathrm{\Theta }=2^o3^o`$) which consists of plastic phoswich detectors (NE102 and NE115) for standing higher particle rates in this region, the light isotope separation could be performed on the overall domain.
Fig. 1 shows the isotope resolution summed up for rings 2 to 9 ($`3^o<\theta <45^o`$). The insert shows it separately for rings 2, 5 and 9. The projectile-like fragments at small angles come out at larger energy. To avoid saturation there, the photomultiplier gains steadily increase from ring 2 to ring 17 by about an overall factor of 10. Subsequently ring 2 has less PID resolution, seen in the insert to Fig. 1. However the angular resolution of the first rings is superior and therefore they contribute important information to the correlation function at small relative momentum.
The p-p correlation function of Fig. 2 shows the data from the forward hemisphere of the projectile source (FHPS, see section 3) without impact parameter selection. It can be continuously constructed from 10 MeV/c up to 250 MeV/c relative momenta due to the 4$`\pi `$ coverage and the good angular resolution (yet the forward source selection slightly increases the minimal relative momentum). The structure in the correlation function at 20 MeV/c is due to the attractive s-wave p-p interaction . There is a Coulomb suppression at very small relative momentum and possibly in the range between 50-75 MeV/c. The normalization has been applied to the data points between 100 and 120 MeV/c which is above any remaining two-body effects and below any kinematical effects at higher momentum. A normalization at lower relative momentum would have introduced a misinterpretation of the correlation effect.
## 3 Event sorting and source selection
We take in our analysis events in which the total longitudinal momentum of detected ejectiles is above 80% of the initial momentum. We refrain from further cuts, for instance the totally detected charge $`Z_{tot}`$, to conserve a representative impact parameter distribution. For most of the events the target like ejectile is lost due to the velocity thresholds. This missing fragment has been kinematically reconstructed and taken into account in the calculation of the momentum tensor and thrust variable. This event class gives an unbiased starting point for the analysis of light particle correlations.
The calculation of the impact parameter is based on the total transverse energy of the light charged particles (Z$``$2) whose experimental detection is quite independent of the reaction mechanism. Furthermore a recent analysis of the correlation between the total multiplicity versus the transverse energy ($`E_T`$) for that system has demonstrated the validity of $`E_T`$ .
In order to observe impact parameter dependent properties while maintaining sufficient statistics, we have defined three $`E_T`$ bins for which the correlation function is constructed. The “peripheral”, “intermediate" and “central” events have a $`E_T`$ range between 0-280 MeV, 280-420 MeV and above 420 MeV corresponding to reduced impact parameters in \[1-0.65\], \[0.65-0.35\] and smaller than 0.35 respectively.
INDRA allows to build event by event a momentum tensor defined in the center of mass by :
$$\textcolor[rgb]{0,0,0}{Q}_{\textcolor[rgb]{0,0,0}{i}\textcolor[rgb]{0,0,0}{j}}\textcolor[rgb]{0,0,0}{=}\underset{\textcolor[rgb]{0,0,0}{k}\textcolor[rgb]{0,0,0}{=}\textcolor[rgb]{0,0,0}{1}}{\overset{\textcolor[rgb]{0,0,0}{M}}{\textcolor[rgb]{0,0,0}{}}}\frac{\textcolor[rgb]{0,0,0}{1}}{\textcolor[rgb]{0,0,0}{p}}\textcolor[rgb]{0,0,0}{p}_\textcolor[rgb]{0,0,0}{i}\textcolor[rgb]{0,0,0}{(}\textcolor[rgb]{0,0,0}{k}\textcolor[rgb]{0,0,0}{)}\textcolor[rgb]{0,0,0}{p}_\textcolor[rgb]{0,0,0}{j}\textcolor[rgb]{0,0,0}{(}\textcolor[rgb]{0,0,0}{k}\textcolor[rgb]{0,0,0}{)}$$
where M is the multiplicity of fragments with a charge Z greater than 2, p is the momentum of the k’th particle in M and $`p_i`$, $`p_j`$ two of the Cartesian momentum components. The eigenvectors of this tensor establish a reference frame. The main axis (eigenvector of the largest eigenvalue) gives the average direction of nuclear matter emission. The eigenvectors associated to the two largest eigenvalues define a reaction plane. The FHPS selections and the calculations have been performed with regards to this new reference system. Fig. 3 shows the transverse versus the parallel velocity of the protons when $`E_T`$ is smaller than 40 MeV. Since this selection implies only very peripheral events, a clearer separation of the sources is exhibited.
The next task consists of recognizing the fragments emitted either from the target-like or from the projectile-like source. The thrust variable defined by
$$\textcolor[rgb]{0,0,0}{T}\textcolor[rgb]{0,0,0}{=}\underset{\textcolor[rgb]{0,0,0}{c}_\textcolor[rgb]{0,0,0}{1}\textcolor[rgb]{0,0,0}{,}\textcolor[rgb]{0,0,0}{c}_\textcolor[rgb]{0,0,0}{2}}{\textcolor[rgb]{0,0,0}{\mathrm{max}}}\frac{\textcolor[rgb]{0,0,0}{}\textcolor[rgb]{0,0,0}{}_{\textcolor[rgb]{0,0,0}{i}\textcolor[rgb]{0,0,0}{}\textcolor[rgb]{0,0,0}{c}_\textcolor[rgb]{0,0,0}{1}}\stackrel{\textcolor[rgb]{0,0,0}{}}{\textcolor[rgb]{0,0,0}{P}_\textcolor[rgb]{0,0,0}{i}}\textcolor[rgb]{0,0,0}{}\textcolor[rgb]{0,0,0}{+}\textcolor[rgb]{0,0,0}{}\textcolor[rgb]{0,0,0}{}_{\textcolor[rgb]{0,0,0}{j}\textcolor[rgb]{0,0,0}{}\textcolor[rgb]{0,0,0}{c}_\textcolor[rgb]{0,0,0}{2}}\stackrel{\textcolor[rgb]{0,0,0}{}}{\textcolor[rgb]{0,0,0}{P}_\textcolor[rgb]{0,0,0}{j}}\textcolor[rgb]{0,0,0}{}}{\textcolor[rgb]{0,0,0}{}_{\textcolor[rgb]{0,0,0}{k}\textcolor[rgb]{0,0,0}{=}\textcolor[rgb]{0,0,0}{1}}^\textcolor[rgb]{0,0,0}{M}\textcolor[rgb]{0,0,0}{|}\stackrel{\textcolor[rgb]{0,0,0}{}}{\textcolor[rgb]{0,0,0}{P}_\textcolor[rgb]{0,0,0}{k}}\textcolor[rgb]{0,0,0}{}}$$
divides up the fragments in two groups $`c_1`$ and $`c_2`$ corresponding to the two emitting sources. The velocity of each of them is determined by a kinematic reconstruction within these two ensembles . As an example, the average velocity of the projectile-like source is drawn with a vertical bold line in Fig. 3.
To disentangle the projectile-like source emission from mid-rapidity contributions, we have taken particles with a parallel velocity larger than the projectile source velocity. This region (FHPS) is on the right of the bold line in Fig. 3. In the following, this sample of particles is used as the base for the extraction of the slope parameter in the energy spectra as well as for the construction of the correlation functions. We remind the reader here that our correlator works dynamically within the reference frame, a fact being imperative in order to optimize the FHPS selection.
## 4 Source parameters : disentangle size and time
### 4.1 Size
The shape and the height of the correlation function is given by the strength of the interactions which themselves depend on the average distance between the two particles detected in coincidence. This distance depends on the average emission time and on the spatial source extension. The double parametrization can be circumvented only for two extreme cases: at high energy the emission time is set to zero while at low energy the emission time is very long and consequently the source size is negligible . In our energy domain, both parameters are relevant. Using the complete detection by INDRA of all charged products we alternatively can determine the source size directly.
For this estimation, ejectiles are grouped according to ($`ZZ_L`$) and ($`Z>Z_L`$) where $`Z_L`$ is an adjusting parameter. In the first group only the particles faster than the projectile-like source velocity are included (particles from the FHPS) and the sum of their charges is multiplied by 2 taking into account the isotropical projectile-like emission. This sum is $`Z_P`$ in which the additional mid-rapidity contribution mainly composed of light charges is suppressed. The second group contains only the particles faster than the center of mass velocity, the sum of their charges is $`Z_F`$. This separate treatment of the heaviest particles (second group) takes into account an asymmetry of emission in the projectile-like reference frame (for example only one big remnant, or two fission fragments). The total charge of the source is $`Z_{Total}=Z_P+Z_F`$. To test the quality of this procedure, the calculation of $`Z_{Total}`$ has been performed for different values of the parameter $`Z_L`$(2, 4, 8, 10). Fig. 4 shows that $`Z_{Total}`$ only varies by less than 6% with $`Z_L`$. For later calculation of $`Z_{Total}`$ we chose $`Z_L=4`$.
For each of the three impact parameter bins we have also defined $`Z_{Max}`$ as the largest fragment being faster than the center of mass. $`Z_{Max}`$ decreases with the centrality as expected in the geometrical, simple picture of the collision also shown in Fig. 4. $`Z_{Total}`$ unexpectedly remains constant.
To estimate the source size, we make two assumptions about the projectile-like source. First, the $`A_{Total}/Z_{Total}`$ ratio is fixed to the one from the valley of nuclear stability. Secondly, with this value of $`A_{Total}`$ we determine the source radius r by assuming a normal nuclear density $`\rho _0`$ and by simply applying $`r=r_0.A_{Total}^{1/3}`$ with $`r_0=1.2`$ fm. The central ’single source’ events which have a radial flow of 2 AMeV have a very small cross section. This implies that a density different from $`\rho _0`$ does not make sense for our calculation even in the central impact parameter bin. The extracted source sizes are given in Table 1. The errors are derivated from the $`Z_{Total}`$ distribution widths.
### 4.2 Slope parameter
We have extracted the slope parameter from the experimental energy spectra using the formula for surface emission. As for the source size estimation, we have selected only the protons and the deuterons located in the FHPS region. Their kinetic energy is given in the projectile-like source reference frame. As example for the peripheral collisions, Fig. 5 shows the energy spectra restricted to ring 2, 4 and 6 where the double slope is the most clearly visible. The low energy one originates from the projectile-like thermal emission, the other at higher energy, presumably from a preequilibrium emission. We observe that the importance of high energy particles decreases with the radial angle as if this emission were concentrated along the beam axis, and as expected in a Fermi-jet picture . The slope parameters averaged over all the rings included in FHPS are given in Table 1. For the quantum calculation code presented below we have used a weighted average value of the double slopes.
## 5 Description of the quantum model
The extraction of the emission time has been performed by using the three-body quantum model developed by R. Lednicky , . This code calculates the quantum statistics for identical particles and the final state interaction by taking into account the nuclear and the Coulomb potentials. The Coulomb repulsion on the particle pair due to the emitting projectile-like source is also included . The three-body problem is analytically solved by making an adiabatic assumption : the relative motion between the two particles has to be much slower than their velocity in the source reference frame. Due to a sizeable angle between neighboring detectors, the small relative momentum region can only be populated by pairs of particles with almost equal velocity. Thus the adiabatic assumption is fulfilled in the region of the signal.
The introduction of the emitter Coulomb effect in the quantum calculation is a new feature brought by this model. We feel that this type of description is required to correctly reproduce the experimental data because the presence of the remnant source charge is intrinsic to the model. In so far it may surpass models where the Coulomb influence of the emitter is only treated as a correction . Limits of this model are certainly related to the static source description which does not take into account the correlation between momentum and position, nor the dynamical emission.
The particle pairs are generated by a static surface emitter describing the projectile-like source with size and slope parameter (see Table 1). The choice of a surface emission instead of a volume one is justified on grounds of the dominance of binary processes , meaning a big fragment at the speed of the projectile remains in the exit channel. Combined with the fact that no radial flow is observed (see 4.1), the picture of a surface emission of particles seems to be the most appropriate. The time distribution for particle emission follows an exponential decay law with average emission times to be determined from comparison with the data. The energy distribution is taken from the experimental slope parameter. The distributions have been filtered according to the geometrical granularity, including also double counts. In addition, energy and particle identification thresholds as well as the source selection criteria (FHPS) have been folded in.
## 6 Analysis of the light particle correlation functions
The experimental correlation function is defined by the yield of true coincidences as a function of the two particle relative momentum divided by the so-called false coincidences obtained by the technique of event mixing . It consists of taking two particles from two different events which assures a full decorrelation and has the advantage to use the same sample of events for true and false coincidences. Possible distortions coming from our FHPS selection are then largely eliminated. In general, event mixing introduces an additional term in the relative momentum due to different source velocities. This influence which otherwise would disturb the correlation function is reduced by the impact parameter selection.
### 6.1 The Proton-Proton correlation function
The p-p correlation functions for the three intervals in impact parameter are represented in Fig. 6. The statistics of 2.3 millions of reconstructed events give a reasonable correlation function above 10 MeV/c relative momentum. In all three event classes, the resonance at 20 MeV/c is clearly visible and well described by a time value of $`\tau `$=80 fm/c with a surface emission model. A time variation of 25% changes the height of the resonance by a factor of two. This demonstrates the high sensitivity of the correlation function on the emission time parameter. The fact that the time estimation strongly depends on the emission description has also been investigated : a simulation using volume break-up, closer to the scenario of a preequilibrium emission, leads to shorter times in the range of 25 fm/c.
The undershoot in the shape of the calculated correlation function visible at 50 MeV/c is induced by the Coulomb repulsion between the two particles. The additional boost due to the repelling charge of the third body (emitting source) shifts the Coulomb suppression, which is usually located at small relative momentum, to higher values. In a pure two body calculation this undershoot almost disappears.
The experimental data also show this behavior at 50 MeV/c for the central events contributing to a very good agreement with the calculation. On the contrary the experimental correlation function for the peripheral events does not show this anticorrelation feature. It is not fully reproduced by the calculation based still on a surface emission. This might indicate the presence of a preequilibrium component, for which no coherent Coulomb influence of a source is expected and for which a volume break-up simulation would be more appropriate. A recent QMD calculation for the system Xe+Sn at 50 AMeV predicts the compression to be maximal at 50 fm/c after the beginning of the collision compatible with the total time of spatial overlap ($``$ 40 fm/c). The end of fragment interaction occurs in this model at 120 fm/c considered as the end of the reaction. Taking this time scale into account, our measured times in between 25 and 80 fm/c are compatible with a preequilibrium emission. It is confirmed by the disappearance of the Coulomb undershoot at 50 MeV/c in the case of the peripheral collisions. The appearance of the double slope in the energy spectra supports this interpretation.
One must ask here, how the emission time alters for a given error in source size. To test this we have reduced the charge from Z=46 to Z=36, simply assuming all the Z=1 particles do not belong to the projectile-like emission. The result for the emission time is then 100 fm/c instead of 80 fm/c which still stays short enough to be compatible with our conclusion saying, these light particles characterize a prompt process of pre-equilibrium emission which covers a large domain of rapidity. By lack of statistics resulting from our source and impact selections, the method of simultaneous determination of the source size and the emission time from parallel and transverse correlation functions could not be tested here.
### 6.2 The Deuteron-Deuteron correlation function
The d-d correlation function has been constructed on the same base of events as p-p. The normalization has been applied to the data points between 150 and 200 MeV/c. Again the three impact parameter classes have been separately analyzed and the results are shown on Fig. 7. We immediately observe the anticorrelation effect in the d-d correlation function for small relative momenta. Despite the fact that data do not go below 30 MeV/c the fit of the quantum model gives the following results : for peripheral reactions the emission time is as least 200 fm/c, for semi-central reactions it is 100 fm/c and for central it is 25 fm/c. This behavior can be interpreted as an increasing contribution of out-of-equilibrium emission.
Yet the creation of barely bound particles is not the prefered mechanism of hot nuclear matter to dissipate energy. A second scenario could be imagined, assuming the deuteron creation happens only at a certain low density . Then the extracted emission time would give a direct hint on when this state of the nuclear matter is reached during the reaction process. However the double slope of the deuteron energy spectra seems to favor the out-of-equilibrium emission. In the picture of the coalescence model , the deuteron formation is directly connected to the proton creation. Consequently it is not surprising to find also two components in the deuteron energy spectra.
In conclusion the process of deuteron production remains an open question. Still we tend to favor the preequilibrium emission over the other explanations. Unfortunately, the present sample of data is insufficient to disentangle more.
### 6.3 The Proton-Deuteron correlation function and the emission chronology
The correlation function of non-identical particles can give model independent information about their mean order of emission simply making velocity selections , , . We have applied this method to the p-d correlation function for particles emitted in the FHPS region. The principle is to compare two functions. The first, ( $`1+R^+`$), is constructed with pairs where the proton is faster than the deuteron in the projectile-like source reference frame. The second function, ( $`1+R^{}`$), corresponds to the reverse situation . When the first emitted particle is slower than the second, the average distance will be reduced and the Coulomb suppression effect enhanced, and vice versa. The comparison of the two functions gives the mean order of emission as it is shown in Fig. 8 for the peripheral collisions. The Coulomb suppression is more pronounced in $`1+R^+`$, which the ratio clearly demonstrates. This means that the deuteron is on average emitted earlier than the proton, namely $`\tau _{deuteron}<\tau _{proton}`$. The same time sequence is observed for the two other impact parameter selections. It is important to note that this result is only validated between 0-120 MeV/c relative momentum where the anticorrelation effect leaves a measurable signal.
The chronology of emission in p-d spectra of peripheral collision can be considerated as surprising since we just learned from previous paragraphs that the mean emission time is 80 fm/c for p-p and 200 fm/c for d-d. This apparent contradiction can be resolved by postulating that the protons which contribute to p-p are not identical with those contributing to p-d. Indeed the proton energy spectra show a fast and a slow component (Fig. 5). Furthermore, from the energy slopes of Fig. 9, we infer that the protons of p-p in the relative momentum range of 0-50 MeV/c have a higher mean kinetic energy (12.9 MeV) than the one of p-d in the same range (9.9 MeV). Subsequently it is clear that the p-p correlation function is more influenced by the preequilibrium protons. The extracted times between 25 and 80 fm/c must be seen as an upper limit reflecting the mixing of a fast and a slow component. In the p-d correlation function the situation is different : the coincidences of ”early” protons with the deuterons are shifted to higher relative momentum value since the speed differs more than for thermal proton-deuteron pairs. Moreover, the functions $`1+R^+`$ and $`1+R^{}`$ are built in the projectile-like source frame. So protons from the interacting zone which feed both correlation functions are per definition faster. Consequently they put only little weight to the emission time order determination. The p-d correlation function informs rather about the thermal protons
Taking the deuteron emission time extracted from d-d, $`\tau =`$ 310, 110 and 25 fm/c for the peripheral, intermediate and central events respectively, we have deduced the corresponding time for the protons (Fig. 10) and obtained 400 fm/c for central collisions. Unfortunately no satisfying agreement can be found between the quantum model and the p-d data for peripheral and intermediate reactions, possibly due to the different proton contributions, one acting at low the other at high relative momentum. In addition, the description of the source without dynamical features might prevent a better agreement. Therefore the extracted parameters do not retain the meaning of a physical time.
Although both components of protons simultaneously play a role in the correlation function, we can deduce a chronology pattern of the light particle emission. The fast protons from the interacting zone come first, then the deuterons and still later the protons thermally emitted by the projectile-like source. It is possible to better separate the two proton components by making a cut on the parallel velocity as Fig. 11 shows. The clear enhancement of the p-p resonance at 20 MeV/c and the disappearance of the Coulomb suppression at 50 MeV/c reveal the enlarged part of fast protons not feeling the charge of an emitter. The importance of differentiating between short-lived and long-lived emission components and subsequent space-momentum correlations has also been discussed by using source imaging methods . Furthermore, making the plausible hypothesis that fastest particles are emitted earlier than the slowest ones because the available energy is greater at the beginning of the reaction , we can extrapolate from the slope parameter pattern an equivalent chronology :
$$\textcolor[rgb]{0,0,0}{T}_{\textcolor[rgb]{0,0,0}{P}\textcolor[rgb]{0,0,0}{r}\textcolor[rgb]{0,0,0}{o}\textcolor[rgb]{0,0,0}{t}\textcolor[rgb]{0,0,0}{o}\textcolor[rgb]{0,0,0}{n}\textcolor[rgb]{0,0,0}{\mathrm{\hspace{0.17em}2}}}\textcolor[rgb]{0,0,0}{>}\textcolor[rgb]{0,0,0}{T}_{\textcolor[rgb]{0,0,0}{D}\textcolor[rgb]{0,0,0}{e}\textcolor[rgb]{0,0,0}{u}\textcolor[rgb]{0,0,0}{t}\textcolor[rgb]{0,0,0}{e}\textcolor[rgb]{0,0,0}{r}\textcolor[rgb]{0,0,0}{o}\textcolor[rgb]{0,0,0}{n}\textcolor[rgb]{0,0,0}{\mathrm{\hspace{0.17em}2}}}\textcolor[rgb]{0,0,0}{,}\textcolor[rgb]{0,0,0}{T}_{\textcolor[rgb]{0,0,0}{D}\textcolor[rgb]{0,0,0}{e}\textcolor[rgb]{0,0,0}{u}\textcolor[rgb]{0,0,0}{t}\textcolor[rgb]{0,0,0}{e}\textcolor[rgb]{0,0,0}{r}\textcolor[rgb]{0,0,0}{o}\textcolor[rgb]{0,0,0}{n}\textcolor[rgb]{0,0,0}{\mathrm{\hspace{0.17em}1}}}\textcolor[rgb]{0,0,0}{>}\textcolor[rgb]{0,0,0}{T}_{\textcolor[rgb]{0,0,0}{P}\textcolor[rgb]{0,0,0}{r}\textcolor[rgb]{0,0,0}{o}\textcolor[rgb]{0,0,0}{t}\textcolor[rgb]{0,0,0}{o}\textcolor[rgb]{0,0,0}{n}\textcolor[rgb]{0,0,0}{\mathrm{\hspace{0.17em}1}}}$$
## 7 Conclusion
We have taken the large data set for collisions of Xe on Sn at 50 AMeV which INDRA has accumulated at GANIL to examine correlations of protons and deuterons. Such studies have many attractive aspects in view of the complete detection of all collision residues by a 4$`\pi `$ detector. Foremost to name is the unique possibility to well determine the emitting source in particular for symmetric systems at intermediate energy. In addition each event can be individually characterized in its own frame of reference, due to full charge, angular and energy coverage of INDRA.
The time scale as well as the chronology of emission of light projectile-like particles could be determined from two particles correlation functions interpreted by a full three body quantum code. Total charge measurements as a function of impact parameter indicate possible out-of-equilibrium emission of protons from the forward hemisphere of the projectile source. The observation that the slope parameter of the energy spectra exhibits two components points to a similar conclusion. Strong confirmation of these findings stems from the very short emission time extracted from p-p correlation functions. While this process is expected to dominate in central collisions our study unveils that it also contributes to the forward zone of binary peripheral collisions. We explained in the text how both proton components are not equally shared in p-p and p-d correlation functions. So the whole emission chronology pattern remains self-consistent. It is also in good agreement with the measurement of slope parameters. The short time scale in p-p reveals the presence of fast hot protons from an out-of-equilibrium process. Protons emitted later than the deuterons correspond to the really equilibrated production from the projectile-like source. The light particle emission chronology including the deuteron formation via the $`NNNdN`$ process has been calculated in the Boltzmann-Uehling-Uhlenbeck (BUU) approach . The theoretical results although for lighter systems are in excellent agreement with the present experimental study.
We would have liked to look into the hydrogenic correlation function in more detail by finer selecting impact parameter intervals. Furthermore, the inclusion of tritons could have given valuable additional information. This task cannot yet be performed on the same footing as with protons and deuterons due to a serious shortage of data statistics.
We therefore suggest to perform a high statistics experiment especially dedicated to light particle correlations. INDRA parameters, optimized toward this goal, could contribute important and still better information on the dynamics of light particle emission.
## 8 Acknowledgments
The authors whish to thank H.Orth for his careful reading of the manuscript, R.Lednicky for discussions and providing his code. D.G acknowledges the support of the ALADIN group at GSI.
| | | T <sub>Proton</sub> (MeV) | | T <sub>Deuteron</sub>(MeV) | |
| --- | --- | --- | --- | --- | --- |
| Centrality | Radius (fm) | T<sub>1</sub> | T<sub>2</sub> | T<sub>1</sub> | T<sub>2</sub> |
| Peripheral | 5.9$`\pm `$0.3 | 3.7$`\pm `$0.1 | 9.5$`\pm `$0.1 | 4.1$`\pm `$0.1 | 8.9$`\pm `$0.1 |
| Intermediate | 5.9$`\pm `$0.4 | 4.8$`\pm `$0.1 | 11.3$`\pm `$0.1 | 6.0$`\pm `$0.1 | 10.7$`\pm `$0.1 |
| Central | 5.9$`\pm `$0.5 | 5.2$`\pm `$0.1 | 12.8$`\pm `$0.1 | 11.3$`\pm `$0.1 | \*) |
Table 1 : Parameters of the projectile-like source. The radius has been calculated from the total reconstructed charge of the source, by assuming a A/Z ratio in the valley of nuclear stability and a normal nuclear density. The slope parameter has been extracted from the energy spectra in the source reference frame.
\*) Only one slope could be extracted.
Fig. 1 : Light isotopes resolution (Z=1) of the 8 first rings and in the insert for the rings 2, 5 and 9 separately.
Fig. 2 : Two-proton correlation function for the FHPS selection.
Fig. 3 : Invariant velocity plot of the protons in the center of mass for a total light particles transverse energy smaller than 40 MeV which corresponds to a normalized impact parameter larger than 0.9. The average value of the reconstructed forward source is represented by a vertical line at $`V_{//}=0.155c`$. This line position is in accordance to the middle of the Coulomb circle. The FHPS is defined by the particles in each event which are faster than the reconstructed forward source velocity.
Fig. 4 : The reconstructed total charge of the projectile-like $`Z_{Total}`$ for three centrality bins and for different calculations. The meaning of the parameter $`Z_L`$ is given in the text. The $`Z_{Total}`$ dependance on $`Z_L`$is less than 10% which places confidence in this estimation. The largest fragment $`Z_{Max}`$ (open stars) gets smaller with centrality as expected in a geometrical picture. This is not the case of $`Z_{Total}`$ which remains constant.
Fig. 5 : The energy spectra of the protons (top) and deuterons (bottom) for the ring 2, 4 and 6 in the case of the peripheral collisions. The shapes clearly exhibit two components, better separated in the case of the protons. For comparison reasons the relative scale is the same for all panels.
Fig. 6 : The experimental p-p correlation functions (black circles) of the projectile-like source for the three impact parameter intervals. Each case has been calculated with different emission times using the source parameters of Table 1. A time of 80 fm/c was found to be the best for all impact parameter bins. The resonance of the calculated function decreases with the centrality because the slope parameter of the emitter increases.
Fig. 7 : The experimental d-d correlation functions of the projectile-like source for the three impact parameter bins (black circles). Each case has been calculated with different emission times using the source parameters of Table 1. There is only a weak dependence on the long emission time parameter for the peripheral reactions.
Fig. 8 : The experimental proton-deuteron correlation functions of the projectile-like source for the peripheral collisions grouped into two velocity bins (left and middle panel). $`1+R^+`$ contains all the pairs of particles where the proton was faster than the deuteron. $`1+R^{}`$ is the reverse situation. The ratio of both functions (right panel) which is smaller than unity indicates that the deuteron is on the average emitted earlier than the proton.
Fig.9 : The kinetic energy of the protons in the projectile-like source frame which contribute to p-p (open circle) or to p-d (black circle) with a relative momentum selection smaller than 50 MeV/c for peripheral collisions. For p-p the average energy and the slope parameter are higher than for p-d.
Fig.10 : The experimental proton-deuteron correlation functions of the projectile-like source for the three impact parameter selections (black circles). The calculation (lines) fails to reproduce the data for the peripheral and the intermediate collisions which may reveal the double contribution of fast and slow protons in the interference pattern.
Fig. 11 : The experimental proton-proton correlation functions for all impact parameters with increasing selection on their longitudinal velocity in the projectile-like reference frame. The higher resonance indicates a faster emission time. |
no-problem/0001/astro-ph0001442.html | ar5iv | text | # RR Lyrae – Theory vs Observation
## 1. Introduction
The recent work of Kovács & Jurcsik (Kovács & Jurcsik 1996 \[KJ96\], 1997 \[KJ97\], Jurcsik 1998 \[J98\]) proposes an almost purely empirical method of extracting the absolute magnitudes, colors and effective temperatures ($`T_{\mathrm{ef}\mathrm{f}}`$) and metallicities directly from the observed periods and light curves of RR Lyrae stars. The only theoretical input appears in the transformation from M<sub>V</sub> and V – K to L and $`T_{\mathrm{ef}\mathrm{f}}`$ via Kurucz’s static model atmospheres. The very empirical nature and potential usefulness of the approach has attracted a great deal of attention from observers.
The most recent work (J98) compiles and analyzes a sizeable set of observational data of RR Lyrae cluster variables and RR Lyrae field stars. This study thus includes RR Lyrae stars with metallicities ranging from Z=0.00001 to almost solar Z=0.020. The end products of her analysis that we are most concerned about here are Log$`L`$-Log$`T_{\mathrm{ef}\mathrm{f}}`$ and Log$`L`$-Log$`P`$ plots, and the byproducts which are relations between luminosity L, mass M and metallicity Z.
The J98 Log$`L`$-Log$`T_{\mathrm{ef}\mathrm{f}}`$ and Log$`L`$-Log$`P`$ data are shown in Figure 1 as small dots, circles for RRab and triangles for RRc stars. The figures indicate well defined slopes for the fundamental and overtone blue edges and red edges, all four of which have approximately the same values, although there may be a slight broadening of the instability strip with luminosity. Hereafter we call this slope $`\mathrm{\Xi }`$ ($`=\mathrm{\Delta }`$ Log$`L`$/ $`\mathrm{\Delta }`$ Log$`T_{\mathrm{ef}\mathrm{f}}`$).
J98 pointed out that no model calculations can explain the strong dependence of the temperature on the luminosity. In §2 we first confirm that indeed the slope $`\mathrm{\Xi }`$ of this empirical Log$`L`$ vs. Log$`T_{\mathrm{ef}\mathrm{f}}`$ diagram is in irreconcilable disagreement with that of radiative models. Next, in §3 we show that the inclusion of turbulent convection in the models only shifts the Log$`L`$ vs. Log$`T_{\mathrm{ef}\mathrm{f}}`$ line, but does not change the slope. The discrepancy is therefore very basic because it is already inherent in models that are as fundamental as purely radiative ones. In §4 we reexamine the uncertainties of both radiative and turbulent model calculations and conclude that the theoretical $`\mathrm{\Xi }`$ is very robust and cannot be changed very much without introducing new physics. In §5 we discuss the theory that goes into the empirical relations of Jurcsik, in particular the influence of the use of the static Kurucz color – temperature transformation. We conclude in §6.
## 2. Radiative Models of RR Lyrae Stars
In the following we examine how well RR Lyrae models agree with the J98 data. We first examine radiative models, and thus limit ourselves to the vicinity of the blue edges.
Figure 1 shows a Log$`L`$-Log$`T_{\mathrm{ef}\mathrm{f}}`$ plot on the left. On these we superpose the linear fundamental and first overtone blue edges of radiative RR Lyrae models. We recall here that among the linear edges of the instability strip only the overtone linear blue edge and the fundamental linear red edge coincide with the observable edges of the instability strip. Nonlinear dynamical effects shift the observable fundamental blue edge and the overtone red edge to the red and to the blue, respectively, compared to their corresponding linear edges, by up to several 100 K (e.g.cf. Buchler 2000). Strictly speaking, for radiative models, it is only the linear overtone blue edge that is relevant, but the linear fundamental blue edge can give a rough indication of the slope $`\mathrm{\Xi }`$. Furthermore, the red edges are determined by convection, and radiative red edges are not relevant and are therefore not shown in Fig. 1.
The solid lines have been computed for models with M=0.65, X=0.75 and Z=0.001. The dotted lines represent models with M and Z from Jurcsik’s {L, M, Z} relations, the dashed lines those with M from Jurcsik, but with Z=0.001. Finally, the large open circles are models from Tuggle and Iben (1972) for M=0.6 and Z=0.001 (with the old Los Alamos opacities). They have essentially the same $`\mathrm{\Xi }`$ as ours which are computed with the OPAL opacities (Iglesias & Rogers 1996) merged with the low temperature Alexander-Ferguson (1994) opacities.
Again, because we are computing radiative models we do not expect the location of the blue edge to be in perfect agreement with the edge of the instability strip, but we note a very large discrepancy in the slope $`\mathrm{\Xi }`$.
Figure 1 also displays a Log$`L`$ – Log$`P`$ plot on the right. Only the slope of the constant mass models is in almost acceptable agreement with the J98 data, but those calculated with the J98 $`\{M,L,Z\}`$ are in strong disagreement.
In Figures 2 we display again a Log$`L`$ – Log$`T_{\mathrm{ef}\mathrm{f}}`$ plot which shows the effect of composition on the location of the linear blue edges. The (radiative) models were computed with M=0.65. Here the solid lines have X=0.75, Z=0.001, the dotted lines have X=0.75, Z=0.004 and the dashed lines have X=0.70, Z=0.001.
The location of the blue edge displays very little sensitivity to either metallicity Z or helium content Y, within a reasonable range of values. Adjusting the helium content or metallicity does not provide a resolution of the slope discrepancy. We describe further tests in §4.
We conclude this section by noting that purely radiative RR Lyrae models are in severe disagreement with Jurcsik’s Log$`L`$-Log$`T_{\mathrm{ef}\mathrm{f}}`$ data.
## 3. Convective Models of RR Lyrae Stars
The addition of time-dependent turbulent convection in the models has led to some spectacular successes compared to radiative models, in particular, in finally predicting double mode behavior both in RR Lyrae (Feuchtinger 1998) and in Cepheids (Kolláth, Beaulieu, Buchler & Yecko 1998). One would hope therefore that the inclusion of turbulent convection might also remove the discrepancy between observations and theory described in the previous section.
Yet, in that respect all of our modelling efforts with turbulent convection have proved in vain, despite the flexibility afforded by the 8 free order unity ($`\alpha `$) parameters that the turbulent convective equations contain (e.g., Yecko, Kolláth & Buchler 1997, Buchler& Kolláth 2000). We find that turbulent convection can shift the blue edges, but cannot produce the differential effect with respect to luminosity that is required to give the right slope $`\mathrm{\Xi }`$ in the Log$`L`$ vs. Log$`T_{\mathrm{ef}\mathrm{f}}`$ plots. In Figure 3 we show the results for three different combinations of these parameters.
We note that different codes and slightly different recipes for convection give essentially the same theoretical slopes. For example, even though no linear models are computed, the nonlinear hydrodynamical models of Bono et al. (1997) indicate a similarly steeper slope than that of J98.
## 4. Discussion of the Modeling Assumptions
In this section we discuss the effects of several of the approximations that are inherent in the numerical modeling to see how robust the theoretical Log$`L`$-Log$`T_{\mathrm{ef}\mathrm{f}}`$ slope is.
### 4.1. Equation of State
Our code uses as equation of state a simple iteration of the Saha equations (e.g. Stellingwerf 1982). This equation of state is very similar to the OPAL (Rogers et al. 1996) and the MHD (Däppen et al. 1988) equations of state and we do not believe that the tiny differences can be responsible for the discrepancy between the models and the J98 data.
### 4.2. Radiative Transport
Our radiative code uses a standard equilibrium diffusion approximation for the radiative transport, i.e.,
$$L=(4\pi R^2)^2\frac{c}{3\kappa }\frac{}{m}aT^4$$
(1)
in which one uses for the opacity $`\kappa `$ the Rosseland mean (e.g., Mihalas & Mihalas 1984). Eq. 1 has several shortcomings in the region above the photosphere. First, it implies an Eddington factor of $`f_E=1/3`$, which is not correct in the optically thin outer region where $`f_E`$ approaches 0.4–0.5 (Feuchtinger & Dorfi 1994, Fig. 3). Second, in this regime, the opacities should be higher than those given by the Rosseland mean (Alexander & Ferguson 1994). Finally, for the computation of the periods and growth rates it would be more appropriate to linearize the radiation hydrodynamics equations rather than the equilibrium diffusion equation. The magnitude of the errors introduced by the first effect can easily be estimated.
One can approximate the effect of the Eddington factor in Eq. 1. by replacing $`3\kappa `$ by $`\kappa /f_E`$. The Eddington factor also appears in the radiation pressure which becomes $`p_{\mathrm{rad}}3f_Ep_{\mathrm{rad}}`$. We can disregard sphericity effects in the momentum equation contained in the term $`(p_{\mathrm{rad}}f_Ee_{\mathrm{rad}})/r`$.
To simulate the effect of an increasing $`f_E`$ we assume that for $`T<`$ $`T_{\mathrm{ef}\mathrm{f}}`$, $`f_E`$ increases smoothly from 1/3 to 1/2 with decreasing temperature. The results confirm our hunch that an increase of $`f_E`$ in the very outer region has very little influence on the growth rates and almost none on the periods. More importantly it has no differential sensitivity to luminosity.
Fig. 2 : Log$`L`$-Log$`T_{\mathrm{ef}\mathrm{f}}`$ plot: RRab (filled circles) and RRc (triangles) stars of Jurcsik; Radiative blue edges for M=0.65; solid line: X=0.75 Z=0.001, dotted line: X=0.75 Z=0.004, dashed line: X=0.70 Z=0.001.
Fig. 3 : Log$`L`$-Log$`T_{\mathrm{ef}\mathrm{f}}`$ plot: RRab (filled circles) and RRc (triangles) stars: Convective models: Solid, dotted and dashed lines are for three different combinations of the $`\alpha `$ parameters in the convective model equations, chosen to give a reasonable width for the instability strip.
We have artificially, and somewhat arbitrarily, increased the opacity from the vicinity of $`T_{\mathrm{ef}\mathrm{f}}`$ outward. We have not found an appreciable differential effect with $`T_{\mathrm{ef}\mathrm{f}}`$, and have therefore not pursued this avenue with more detailed modelling.
Finally, the insensitivity of the slope to $`f_E`$ makes us believe that the linearization of the radiation hydrodynamics equations instead of the common equilibrium diffusion equation (requiring a serious coding effort) would also make little difference.
We conclude that the discrepancy between the models and the observations is not due to an inadequate treatment of the radiation transport.
### 4.3. Composition
Could compositional make-up be more important than expected? We have computed a number of RR Lyrae models with various combinations of Y and Z, but find that the effect on $`\mathrm{\Xi }`$ is negligible. Next we have artificially increased the abundances of the easily ionizable light elements such as Mg and Na in the OPAL opacities. Again this has such a negligible effect on $`\mathrm{\Xi }`$ that the inclusion of this data would unnecessarily clutter Figure 2.
An inhomogeneous composition is unlikely to exist in RR Lyrae envelopes because of convection, but even if existed our tests with various changes in composition make us doubt that it would resolve the discrepancy.
### 4.4. Rotation
Could rotation be responsible for the discrepancy? In order to estimate the magnitude of the effect of rotation we have included a spherical pseudo-centrifugal acceleration $`\omega ^2r`$ in the equilibrium model and in the computation of the periods and growth rates. We find that a rather short rotation period, of order of a few days, would be necessary to have an impact. Furthermore the disagreement is worsened by rotation because the slope steepens.
We conclude that moderate rotation rates cannot be the cause of the discrepancy.
### 4.5. Evolutionary Effects
The blueward moving evolutionary paths turn around at some $`T_{\mathrm{ef}\mathrm{f}}`$ and then move upward (e.g., Dorman 1992, Lee, Demarque & Zinn 1990). If this happened inside the linear instability strip, then the leftmost blue edge, that of the overtone mode, would be defined by the topology of these paths, rather than by the pulsational stability of the models. However, the fact that the Jurcsik blue and red edges of the fundamental and the overtone are all essentially parallel eliminates this possibility as an explanation for the shallow observed slope. Moreover, for some of the clusters, full photometry of the horizontal branch exists, and there is no indication of any gap next to the RR Lyrae region. Clearly there are stars next to the empirical red edge for which the amplitude (if not zero) is less than the observational limit.
### 4.6. Conclusion
The theoretical $`\mathrm{\Xi }`$ slope is very robust with respect to the uncertainties or approximations inherent in our code, but is in disagreement with the Jurcsik empirical relations. We have therefore been led to reexamine the latter for their robustness, especially since they also make use of some theory, viz. color to temperature transformation.
## 5. Discussion of the Empirical Relations
Because of the strong disagreement between Jurcsik’s processed observational data and the predictions from something as basic as radiative models, it is worth while to reexamine some of the most uncertain points in the reduction of the observational data.
The observational material consists of RRab periods and lightcurves. KJ96, JK97 and J98 deduce the color, the visual magnitude M<sub>v</sub> and the metallicity Z (\[Fe/H\] in the form of linear fully empirical relations
$`\left(\begin{array}{c}M_v\\ VK\\ [Fe/H]\end{array}\right)`$ $``$ $`c_0+A\left(\begin{array}{c}P\\ A_1\\ \varphi _{31}\\ \varphi _{41}\end{array}\right)`$ (2)
The connection with Log$`T_{\mathrm{ef}\mathrm{f}}`$ and with Log$`L`$ was then made through static envelope models of Kurucz (1993, cf. J98 for details). This is the only theoretical input into the otherwise empirical relations.
$$\left(\begin{array}{c}\mathrm{L}ogL\\ ŁT_{eff}\end{array}\right)=c_1+C\left(\begin{array}{c}M_v\\ VK\\ [Fe/H]\end{array}\right)=c_2+CA\left(\begin{array}{c}P\\ A_1\\ \varphi _{31}\\ \varphi _{41}\end{array}\right)$$
(3)
### 5.1. Color – $`T_{\mathrm{ef}\mathrm{f}}`$ Transformation
J98 uses the static envelope calculations of Kurucz’s ATLAS code (for details cf. J98), to make the transformation from average color to average $`T_{\mathrm{ef}\mathrm{f}}`$. The $`T_{\mathrm{ef}\mathrm{f}}`$ however varies considerably during one pulsation cycle, compared to the width of the instability strip, and the question arises whether the correct mean color and mean temperature are still related through the static Kurucz relations. To check the validity of this assumption we have computed the spectra of two full amplitude RR Lyrae models.
We investigate the theoretical color variation for two RR Lyrae models with different luminosities (L = 40 $`L_{}`$ and L = 52 $`L_{}`$, both for M = 0.65 $`M_{}`$, $`T_{\mathrm{ef}\mathrm{f}}`$= 6500 K, X = 0.76 and Z=0.001). Nonlinear full amplitude models are taken from Feuchtinger (1999) and the color variations in V and I are computed as outlined in Dorfi & Feuchtinger (1999). The results can be inferred from Fig. 4.2 which displays the color difference V – I as a function of the effective temperature (left panel) and the corresponding color-magnitude diagram (right panel), both over one oscillation cycle. Full and dotted lines refer to the L = 40 $`L_{}`$ and L = 52 $`L_{}`$ models, respectively. The differential effect with L is seen to be small. Dynamical atmospheric effects thus cannot account for the discrepancy.
There remains the question of the consistency of the static Kurucz atmospheres with our frequency dependent calculation. For that purpose we have checked the color – temperature transformation against the Kurucz tables at selected points on the pulsation cycle. The agreement is quite satisfactory, and any discrepancies are irrelevant for our purpose here.
### 5.2. RR Lyrae Period - Radius Relation
The radii of RR Lyrae stars can be obtained independently either from a Baade-Wesselink approach or from the J98 empirical L and $`T_{\mathrm{ef}\mathrm{f}}`$. In Fig. 5, on the left, we have plotted as open circles the Baade-Wesselink radii of the 15 RRab stars of Jones et al. (1992) (we have omitted RR Leo). Superposed as crosses are the radii for the same stars derived from the J98 L and $`T_{\mathrm{ef}\mathrm{f}}`$. The agreement is remarkably good, but worsens for the longer periods. The solid and dashed lines represent linear regression lines for the BW and for the J98 period – radius relations.
We have also derived the radii of the additional J98 stars that were not in the Jones sample and show them as small dots in the righthand figure. The thin lines represent constant mass and constant luminosity radiative models, and the thick lines radiative models that obey the J98 L, M, Z relations Since the radiative models do not provide a red edge, the upward extent of the thin or thick lines is not significant. The latter models are seen to give better agreement with the swarm of the J98 empirical radii.
We conclude that the agreement between the empirical J98 and Jones’ Baade-Wesselink radii is remarkably good, and so is the agreement that can be achieved with theoretical models. A consideration of the period – luminosity relations therefore does not give us any clues as to the origin of the slope discrepancy.
## 6. Conclusions
We have shown that the Jurcsik relations lead to RR Lyrae radii that are in good agreement with both Baade-Wesselink radii and with theoretical radii. On the other hand, there exists a strong discrepancy between the slope of the theoretical Log$`L`$– Log$`T_{\mathrm{ef}\mathrm{f}}`$ relation and the slope of the empirical Jurcsik relation.
We have reviewed the physical and numerical uncertainties that enter the theoretical calculations. The discrepancy exists already at the level of purely radiative modelling. We have further shown that our model equations for turbulent convection do not alter the slope of the Log$`L`$–Log$`T_{\mathrm{ef}\mathrm{f}}`$ relation despite the large number (8) of adjustable parameters ($`\alpha `$’s) that we have at our disposal. Consequently, apart from the uncertainties inherent in a 1D recipe for turbulent convection, it appears that the discrepancy is not caused by a deficiency of the theoretical models. We note that the same slope discrepancy is implicit in older and more recent independent calculations (Tuggle & Iben 1972, Bono et al. 1997).
The derivation of the empirical relations makes use of a static Kurucz color - $`T_{\mathrm{ef}\mathrm{f}}`$ transformation. By computing the behavior of color versus temperature over the pulsation cycle we have shown that the use of static envelopes is in fact a very good approximation, and therefore cannot be the culprit.
Finally, the shape of the evolutionary tracks, through their potential avoidance of certain regions of the instability strip cannot be responsible either.
At this time the origin of this disturbing discrepancy constitutes an unsolved puzzle. It remains to be seen whether an improved treatment of turbulent convection or more complete observations will resolve the difficulty.
## 7. Acknowledgements
The authors would like to thank Johanna Jurcsik for kindly providing us with copies of her data sets. They also acknowledge fruitful discussions with Johanna Jurcsik and Géza Kovács. This work has been supported by NSF (AST 95-28338 and AST 98-19608) and the Hungarian OTKA (T-026031). |
no-problem/0001/nlin0001061.html | ar5iv | text | # Wavelength Scaling and Square/Stripe and Grain Mobility Transitions in Vertically Oscillated Granular Layers
## 1 Introduction
Granular materials are collections of discrete solids for which even the simplest realizations — ensembles of identical solid spheres interacting only via contact forces — exhibit a wealth of surprising behaviors . Our work on granular media has focused on a dynamic phenomenon, the formation of subharmonic standing waves in vertically oscillated granular layers . These strongly nonlinear waves form patterns of stripes, squares, hexagons and more complex patterns , as well as localized structures called oscillons , as a function of three dimensionless control parameters: the acceleration $`\mathrm{\Gamma }`$, the frequency $`f^{}=f\sqrt{H/g}`$, and the layer thickness $`N=H/D`$. This paper concerns the square or stripe patterns that arise at the primary instability as $`\mathrm{\Gamma }`$ is increased; examples of these patterns are shown in Fig. 1. Despite being composed of discrete grains and having typical wavelengths $`\lambda `$ of only 20-30 $`D`$, the appearance of granular patterns is similar to that of patterns in fluid systems , the closest example being standing surface waves in a vertically oscillated liquid layer (the Faraday instability) .
Pattern formation in granular media can perhaps be described by continuum equations analogous to those used in fluid system , but the continuum equations remain largely untested by experiment. Aspects of granular patterns have been described using phenomenological models , but making the connection to real systems requires an understanding of granular media at a more microscopic level. Recently, molecular dynamics simulations have begun to yield new details concerning the behavior of granular waves . In particular, a simulation developed by Bizon et al. quantitatively reproduces the granular patterns observed in three dimensions and thus allows calculation of quantities not readily accessible by experiment.
In this paper, we will first describe our experimental system and then discuss some generic features of the flat-layer/wave transition. Next, we will present results for the scaling of the square/stripe transition frequency and the wavelength with particle size and layer depth. Last, we will use results from local measurements of the collision pressure and from dispersion relations to demonstrate and characterize the grain mobility transition.
## 2 Experiment
Our experimental apparatus is similar to that used in Refs. and is described in detail in Ref. . A layer of granular material is placed in the bottom of an evacuated container mounted on the armature of an electro-mechanical shaker and sinusoidally oscillated in the vertical direction (see Fig. 2(a)). Our granular layers consist of spherical non-cohesive and non-magnetic particles of bronze, 316 stainless steel, titanium, or lead with as-poured densities of 5.1, 2.6, 4.7 and 6.7 g/cm<sup>3</sup> respectively, and with $`0.05<D<0.8`$ mm and a poly-dispersity of $`\pm 10\%`$. Data is obtained for $`D=0.17`$ mm bronze particles except as noted. Three different containers are used (see Table 1); each has a polished aluminum base to prevent static charge accumulation and clear plastic side walls and top for lighting and visualization respectively. An accelerometer mounted on the underside of the container measures the acceleration. The patterns are visualized using as a light source a ring of strobed light-emitting diodes, which encircle the container and illuminate the layer from the side (Fig. 2(b)). Images are acquired by a digital CCD camera mounted on axis.
The momentum transfer generated during layer-container contact is measured by an acceleration compensated pressure sensor with a resonant frequency of 300 kHz (PCB 112A22). This sensor has a circular sensing area 5.5 mm in diameter; the sensor size can be compared with the wavelength of the pattern, which varies from 2 to 40 mm for the present measurements, and the particle size, typically 0.17 mm. The sensor is flush mounted ($`\pm 0.02`$ mm) 13 mm from one container sidewall and 36 mm from the other in the bottom of cell S2. To avoid high sampling frequencies, a circuit consisting of a peak detector, a level shifter, and a sample-and-hold is used to record the maximum pressure from each collision of the layer with the container.
## 3 Wave Onset
Here we investigate the transition from a flat layer to standing waves in the regime where the layer free-flight time $`t_{flt}`$ is less than one period of oscillation of the container $`T`$. The layer state is characterized by local collision pressure measurements . Related work is reported in Ref. and in Refs. , where in the latter, global force and optical measurements characterize the flat layer state. For $`\mathrm{\Gamma }<1`$, the layer is always in contact with the plate and there is no significant relative grain motion. For $`\mathrm{\Gamma }>1`$, we identify three distinct stages of layer motion during each cycle: free-flight – layer not in contact with container; impact – layer and container collide; contact – layer and container in contact and moving with the same vertical velocity. Impact imparts relative kinetic energy to the grains which, if large enough, enables the layer to change its configuration.
We first describe some generic features of the layer-container pressure in the vicinity of the flat-layer/waves transition. A time series for the local pressure exerted by the granular layer on the container below the onset of the waves is plotted in Fig. 3(a). The pressure is expressed in non-dimensional form, $`P^{}=P/\rho gH`$, where $`P`$ is the pressure and $`\rho `$ is the as-poured layer density. $`P^{}=0`$ indicates the layer is in free-flight, the sharp jump in $`P^{}`$ occurs at impact, while contact is shown by $`P^{}a(t)/g+1`$ — the pressure calculated using the measured plate acceleration and assuming a solid layer attached to the plate (dashed curve). Several features of $`P^{}`$ should be noted. As the inset in Fig. 3(a) emphasizes, the collision duration is short, less than $`0.02T`$, which indicates the layer is relatively compact at the time of collision. Furthermore, immediately following the collision, $`P^{}=a(t)/g+1`$, which shows the collision is strongly inelastic (if the layer bounced, $`P^{}`$ would be reduced). Finally, $`P^{}`$ and $`a(t)/g+1`$ go to zero simultaneously, which indicates the entire layer is moving upward with the container velocity when free-flight begins.
To contrast the layer behavior above and below wave onset, a time series of $`P^{}`$ in the wave regime is presented in Fig. 3(b). Here the layer motion consists of just two stages, a free-flight followed by an extended collision. There is no longer a contact stage as indicated by the observation that $`P^{}`$ is always larger than $`a(t)/g+1`$, and that $`a(t)/g+1`$ goes to zero before $`P^{}`$ goes to zero. From these observations, we infer that the layer is expanded at impact, that at no time during the collision is the entire layer totally at rest with respect to the container, and that the collision continues for a short time after $`a(t)`$ becomes less than $`g`$.
At the onset of waves, the maximum collision pressure per cycle, $`P_{max}^{}`$, decreases rapidly, as Fig. 4(a) illustrates. This decrease is associated with an increase in layer dilation (see Fig. 3). With increasing $`\mathrm{\Gamma }`$, the decrease in $`P_{max}^{}`$ occurs at $`\mathrm{\Gamma }=2.7`$. However, the transition is hysteretic and the layer reverts to the flat state at $`\mathrm{\Gamma }=2.2`$, where $`P_{max}^{}`$ increases abruptly (see Fig. 4(a)). For the flat layer, $`P_{max}^{}`$ increases with increasing $`\mathrm{\Gamma }`$, while in the wave state, $`P_{max}^{}`$ decreases with increasing $`\mathrm{\Gamma }`$ . Additional measurements using steel and titanium particles show a somewhat different behavior . Before the transition to waves ($`\mathrm{\Gamma }1.9`$), $`P^{}`$ in these larger restitution coefficient materials drops from 80 to approximately 40 but then continues to increase until the onset of waves at which point $`P^{}`$ drops to 20 as for the bronze layers.
Figure 4(b) compares the layer free-flight time measured from the experiment with the time calculated from the completely inelastic layer model . When the layer is flat, the measured and calculated flight times are nearly identical. However, when waves arise, the measured flight time is smaller than the model predicts (see inset in Fig. 4(b)). This discrepancy arises because the velocity of a portion of the layer is less than the container velocity when $`a(t)`$ becomes less than $`g`$. These slower grains reduce the effective layer take-off velocity, which decreases $`t_{flt}`$. For increasing $`\mathrm{\Gamma }`$, there is a sudden decrease in $`t_{flt}`$ when waves arise. However, for decreasing $`\mathrm{\Gamma }`$, and unlike $`P_{max}^{}`$, $`t_{flt}`$ continuously decreases until it equals the model value at which point the layer reverts to the flat state. The equality of measured and model values of $`t_{flt}`$ at this transition implies that for waves to exist, a portion of the layer must be dilated when the bottom of the layer leaves the container. This necessary condition appears to be due to the strongly dissipative nature of the granular layer, which rapidly removes kinetic energy when layers remain in contact with the container.
The qualitative features of the transition from a flat layer to waves, shown in Figs. 3 and 4, are the same for the entire range of $`H`$, $`D`$, and $`f`$ we examine. However, other features of the transition are non-trivial functions of $`H`$ and $`f`$. For instance, the $`\mathrm{\Gamma }`$ value at which the transition from a flat layer to waves occurs, $`\mathrm{\Gamma }_c`$, is plotted in Fig. 5(a) as a function of $`f^{}`$. For increasing $`\mathrm{\Gamma }`$, $`\mathrm{\Gamma }_c`$ decreases monotonically with $`f^{}`$, while for decreasing $`\mathrm{\Gamma }`$, $`\mathrm{\Gamma }_c`$ is nearly constant except for a slight local increase near $`f^{}0.4`$ . The observation that the existence of waves depends on the layer never reaching the contact stage gives the following necessary condition for wave existence
$$Tt_{flt}=t_{coll}=h/v_{coll},$$
(1)
where $`h`$ is the layer dilation (actual layer thickness minus $`H`$), $`t_{coll}`$ is the collision duration, and $`v_{coll}`$ is the collision velocity. At the flat-layer/waves transition for decreasing $`\mathrm{\Gamma }`$, $`\mathrm{\Gamma }_c`$ is nearly constant which implies $`hf^2`$ since $`Tt_{flt}`$ and $`v_{coll}`$ are both proportional to $`f^1`$. For the data in Fig. 5(a), $`\mathrm{\Gamma }_c2.2`$ which gives $`h=0.25v_{coll}t_{flt}`$. Similarly, applying Eq. (1) in the flat layer regime where $`\mathrm{\Gamma }_c`$ is a function of $`f`$, yields $`hf^{1.8}`$ — a slower decrease with increasing $`f`$ than for waves.
$`\mathrm{\Gamma }_c`$ is also a function of $`N`$, as Fig. 5(b) shows. $`\mathrm{\Gamma }_c`$ is nearly independent of $`N`$, while $`\mathrm{\Gamma }_c`$ is roughly proportional to $`N`$. The transition as measured by the rapid drop in $`P_{max}`$ is no longer subcritical for $`N2`$ — the same $`N`$ value below which coherent wave patterns disappear. Interpreting these results in light of Eq. (1), the increase in $`\mathrm{\Gamma }_c`$ with increasing $`N`$ is due to the greater dissipation in the thicker layer which reduces $`h`$. The disappearance of waves for $`N2`$ is likely a result of large grain velocity fluctuations that destroy coherent layer motion. This interpretation is also supported by experiments with different materials : materials with a restitution coefficient larger than that of bronze, such as stainless steel and titanium, require deeper layers than does bronze to generate waves. Lead, which has a smaller restitution coefficient than bronze, requires a layer of only $`N=1.1`$ for waves to exist. Additionally, in layers deep enough to form waves, we find that higher (lower) restitution materials have higher (lower) values of $`\mathrm{\Gamma }_c`$ for equal $`N`$.
In Ref. , Bizon et al. note that waves in their simulations form when the time for peak growth is greater than that for peak decay. From this observation, they calculate $`\mathrm{\Gamma }_c2.47`$ independent of either $`f`$, $`D`$, or $`H`$. The data presented above is in reasonable agreement with their prediction. Their observation does not address the internal layer state and thus can not predict the value of $`\mathrm{\Gamma }_c`$ and its dependence on $`f`$ and $`N`$. With results from simulations, it will be possible to measure $`h`$ and thus check the validity of the criteria for wave onset proposed in Eq. 1.
## 4 Square/Stripe Transition
The wave patterns at the primary bifurcation are squares for low $`f^{}`$ and stripes for high $`f^{}`$, while for intermediate $`f^{}`$, both patterns are observed simultaneously (see Fig. 1). To quantify the square/stripe transition, we divide the pattern images into $`3\lambda \times 3\lambda `$ regions, calculate the spatial power spectra, radially average the power within one full width at half maximum of the dominant wavenumber, and subtract the mean to obtain $`I(\theta )`$. The autocorrelation function $`C(\pi /2)=I(\theta )I(\theta +\pi /2)_\theta /I^2(\theta )_\theta `$ is then used to characterize the pattern. For perfect squares $`C(\pi /2)=1`$, whereas for an image of perfect stripes $`C(\pi /2)=0`$. Differences between the measured and ideal values (i.e., $`C(\pi /2)<0`$) are due to the finite angular width of the spectral peak. Figure 6 is a plot of $`C(\pi /2)`$ versus $`f^{}`$, which shows the transition from squares to stripes occurs in a narrow frequency range about the transition frequency $`f_{ss}^{}=0.31`$, with $`C(\pi /2)`$ nearly constant above and below $`f_{ss}^{}`$. The value of $`f_{ss}^{}`$ is not sensitive to details of the method: plotting $`I(\theta _{max}\pm \pi /2)`$, where $`\theta _{max}`$ is the azimuthal location of the maximum value of $`I(\theta )`$, as well as varying the sub-image size, yield the same result.
Figure 7(a) is a plot of $`f_{ss}^{}`$ as a function of $`\mathrm{\Gamma }`$ for fixed $`N`$. There is little change in $`f_{ss}^{}`$ as $`\mathrm{\Gamma }`$ is increased from near wave onset to close to the transition to hexagons . Figure 7(b) plots $`f_{ss}^{}`$ against $`N`$. Although $`N`$ varies by more than an order of magnitude, the non-dimensionalized square/stripe transition frequency remains nearly fixed. Included in Fig. 7(b) are additional measurements that indicate $`f_{ss}^{}`$ also does not depend strongly on $`D`$, material, or container size or shape.
Bizon et al. have proposed that the square/stripe transition occurs when the distance that the layer falls from its maximum height to the container is equal to the layer depth: for fall distances larger (smaller) than $`H`$, squares (stripes) are the preferred pattern . Using a simple model that considers the layer as a totally inelastic object with no internal modes , they calculate $`f_{ss}^{}=1/\sqrt{8}0.35`$. The data presented here gives an average value of $`f_{ss}^{}0.33`$ over a large range of experimental parameters. The preceding measurements as well as the criterion of Bizon et al. suggest that the square/stripe transition is primarily a function of the layer depth $`H`$. Layers with differing $`N`$ and $`D`$ but equal $`H`$ make the transition from squares to stripes at the same $`f`$ (except for layers with $`NO(1)`$ where the pattern weakens due to large relative particle velocities). This result implies that the layer acts as a continuum, i.e., the size of the particles does not affect the bulk behavior although the discrete nature does. In contrast, the dependence of the critical acceleration amplitude for waves on $`N`$ and $`f`$, where scaling with $`H`$ fails (Fig. 5(b)), suggests that the physics determining subcriticality and square/stripe pattern selection are unrelated.
## 5 Wavelength Scaling
The dispersion relation for surface waves in inviscid fluids in the absence of surface tension and in the shallow water limit ($`\lambda >H`$) is $`\lambda /H=f^1\sqrt{g/H}`$. When plotted in this form, which uses $`H`$ and $`\sqrt{H/g}`$ as the characteristic length and time scales respectively, the dispersion relations for different depth layers lie on a single curve. In granular layers, however, there are two natural length scales – layer depth and grain diameter. We show in this section that below a critical frequency, wavelength scaling is governed by $`H`$ as is the case for fluids, while above this frequency, scaling with $`H`$ fails. In the next section (Sec. 6), we will show that the scaling breakdown is associated with a transition in the horizontal grain mobility.
Figure 8(a) presents results for the wavelength as a function of frequency for layers of bronze particles with $`N=5`$, $`\mathrm{\Gamma }=3.0`$, and a range of particle sizes ($`0.07<D<0.8`$ mm). The wavelength decreases with increasing $`f`$ and increases with increasing $`D`$. When $`H`$ is used to non-dimensionalize the wavelength $`\lambda ^{}=\lambda /H`$ and the frequency $`f^{}=f\sqrt{H/g}`$, the data collapse onto a single curve (Fig. 8(b)). A well-defined kink in the dispersion curve occurs near $`f^{}=0.4`$, above which $`\lambda ^{}`$ decreases more slowly with increasing $`f^{}`$. A similar slowing has been observed in other experiments but only for one and two disjoint values of $`D`$.
Because $`N`$ is constant for the data presented in Fig. 8, it is unclear whether $`H`$ or $`D`$ is the length scale governing the collapse since $`H=ND`$. To test whether the data collapse obtained in Fig. 8(b) is due to scaling with $`H`$, we plot in Fig. 9 the wavelength versus frequency with fixed $`D=0.17`$ mm but with $`N`$ varying from 2 to 31. $`\lambda `$ increases with increasing $`H`$ and again the unscaled data collapse when $`\lambda ^{}`$ is plotted as a function of $`f^{}`$. Similar results were obtained by Bizon et al. but only for two values of $`N`$ ($`N=2.7,5.4`$) .
The rescaled data in Fig. 9(b) fit the functional form $`\lambda ^{}=a+bf_{}^{}{}_{}{}^{c}`$ with the best fit given by
$$\lambda ^{}=1.0+1.1f_{}^{}{}_{}{}^{1.32\pm 0.03}.$$
(2)
The residuals from the fit to the rescaled dispersion data in Fig. 9(b) (see inset) for $`f^{}0.4`$ exhibit a marked decrease in the quality of the data collapse, which is a function of $`H`$ and which is characterized by a slowing in the decrease of $`\lambda ^{}`$ with increasing $`f^{}`$. This slowing is more pronounced in small $`N`$ layers. We will return to this observation in the next section.
Three points should be noted concerning the dispersion relation data. First, although the granular data collapse below the kink in the dispersion curves when scaled with $`H`$ as for fluid surface waves, the dispersion relations for these two media are different. Second, the exponent in Eq. 2 is not equal to $`2`$ as was found in our earlier work on glass beads in air , by Metcalf et al. for glass beads in an evacuated container, and by Clément et al. in a two-dimensional layer of aluminum balls . Possible explanations for the larger value of exponent found in the other studies include the influence of air viscosity, static charging in non-conducting grains, small aspect ratios, and drag associated with the side walls. Finally, the fit predicts a minimum wavelength equal to $`H`$ for large $`f`$. The $`\lambda ^{}=1`$ limit is clearly unattainable in small $`N`$ layers due to the finite size of the layer’s constituent grains; thus, scaling with $`H`$ will fail for small $`N`$. However, as we will show next, breakdown of scaling normally occurs before this limit is reached due to a reduction in the grain mobility.
## 6 Grain Mobility Transition
As discussed in the previous two sections, the layer depth is the characteristic length scale which determines the scaling properties of the square/stripe transition and the wavelength. However, as mentioned in Secs. 1 and 5, the grain diameter $`D`$ is also expected to influence the layer response. Consider the initial velocity required to raise a grain a distance $`D`$ in the presence of gravity, $`v_D=\sqrt{2gD}`$. If the grain is embedded in a plane of identical particles, horizontal motion will only be possible if $`v_D>\sqrt{2gD}`$; for lower velocities, the grain will be stuck. Assuming further that the fluctuation velocity of a typical grain with respect to its neighbors is proportional to the peak container velocity $`2\pi Af`$, a relevant dimensionless parameter characterizing grain mobility is $`\stackrel{~}{\mathrm{v}}=2\pi Af/\sqrt{Dg}`$. Section 6.1 presents evidence of a qualitative change in the layer response at a particular value of $`\stackrel{~}{\mathrm{v}}`$ which we call the grain mobility transition, $`\stackrel{~}{\mathrm{v}}_{gm}2.5`$, for large $`\mathrm{\Gamma }`$ or small $`N`$. Section 6.2 then shows that this change appears related to the loss of horizontal grain mobility for $`\stackrel{~}{\mathrm{v}}<\stackrel{~}{\mathrm{v}}_{gm}`$.
### 6.1 Changes in Layer Response
The value of $`f^{}`$ below which data collapse occurs for constant $`D`$ and varying $`N`$ is marked by a kink in the dispersion curves (see Fig. 9(b)) and is an increasing function of $`N`$. The data in Fig. 8(b) for constant $`N`$ and varying $`D`$ show a similar kink which is, in contrast to the data for constant $`D`$, independent of $`H`$ for fixed $`N`$. To characterize the location of the kink, Fig. 10 presents a plot of $`\stackrel{~}{\mathrm{v}}`$ at the kink (e.g. $`\stackrel{~}{\mathrm{v}}_{gm}`$) as a function of $`N`$ for $`\mathrm{\Gamma }=3.0`$ and $`\mathrm{\Gamma }=2.5`$ . For $`\mathrm{\Gamma }=3.0`$, $`\stackrel{~}{\mathrm{v}}_{gm}2.5`$ and shows no systematic dependence on $`N`$. This is nearly the same as $`\stackrel{~}{\mathrm{v}}_{gm}=2.6`$, the value found for varying $`D`$, constant $`N=5`$ and $`\mathrm{\Gamma }=3.0`$ (see Fig. 8). For $`\mathrm{\Gamma }=2.5`$, $`\stackrel{~}{\mathrm{v}}_{gm}`$ is slightly larger and is a slowly (slower than $`\sqrt{N}`$) increasing function of $`N`$.
To summarize, for $`\mathrm{v}>\mathrm{v}_{gm}`$, $`\lambda ^{}`$ is independent of $`H`$ and scales with $`f^{}`$, while for $`\mathrm{v}<\mathrm{v}_{gm}`$, $`\lambda ^{}`$ scaling with $`H`$ fails. This breakdown appears to be related to a transition in the layer response when $`\stackrel{~}{\mathrm{v}}`$ is small. Observations at $`\stackrel{~}{\mathrm{v}}0.5`$ using relatively large grains show that there is little if any horizontal particle motion at the surface. In contrast, for $`\stackrel{~}{\mathrm{v}}>\stackrel{~}{\mathrm{v}}_{gm}`$, grains slosh back and forth – in thinner layers ($`N4`$) the sloshing is so vigorous that the layer depth goes to zero in the pattern minima. To illustrate the difference between these two wave regimes, Fig. 11 presents a plot of $`\lambda ^{}`$ versus $`\stackrel{~}{\mathrm{v}}`$ for $`N=5`$ and $`\mathrm{\Gamma }=2.5`$ and $`\mathrm{\Gamma }=3.0`$. The wavelength is approximately 10 percent larger for $`\mathrm{\Gamma }=2.5`$ than for $`\mathrm{\Gamma }=3.0`$ when $`\stackrel{~}{\mathrm{v}}>\stackrel{~}{\mathrm{v}}_{gm}2.6`$, but for $`\stackrel{~}{\mathrm{v}}<\stackrel{~}{\mathrm{v}}_{gm}`$, $`\lambda ^{}`$ for $`\mathrm{\Gamma }=2.5`$ jumps to 1.5 times $`\lambda ^{}`$ for $`\mathrm{\Gamma }=3.0`$. If $`\lambda ^{}`$ is instead plotted versus $`f^{}`$, the change in behavior is even more evident: for $`f^{}`$ below the transition $`\lambda ^{}`$ increases with increasing $`\mathrm{\Gamma }`$, while for $`f^{}`$ above, $`\lambda ^{}`$ decreases with increasing $`\mathrm{\Gamma }`$.
In addition to the breakdown in $`\lambda `$ scaling with $`H`$, a qualitative change in the collision pressure is also observed for $`\stackrel{~}{\mathrm{v}}3`$. The $`\stackrel{~}{\mathrm{v}}`$ dependence of the $`\mathrm{\Gamma }`$ averaged collision pressure $`P_{max}^{}_\mathrm{\Gamma }=_{\mathrm{\Gamma }_{min}}^{\mathrm{\Gamma }_{max}}P^{}(\mathrm{\Gamma })𝑑\mathrm{\Gamma }`$ is plotted in Fig. 12 for both the flat-layer and wave regimes. For the flat layer, $`P_{max}^{}_\mathrm{\Gamma }`$ increases continuously with increasing $`\mathrm{v}`$. However, in the wave regime $`P_{max}^{}_\mathrm{\Gamma }`$ is nearly constant for $`\stackrel{~}{\mathrm{v}}>3`$ but decreases in a similar fashion to the flat layer for $`\stackrel{~}{\mathrm{v}}<3`$ . The latter result suggests that the flat layer and wave states are similar for $`\stackrel{~}{\mathrm{v}}<3`$ as would be expected below the grain mobility transition.
Finally, as Sec. 4 describes, the frequency of the square stripe transition is given by $`f_{ss}=0.33\sqrt{g/H}`$ and shows no systematic dependence on the grain size. However, in thin layers when $`\mathrm{\Gamma }\sqrt{N}<2\pi \stackrel{~}{\mathrm{v}}_{gm}f_{ss}^{}`$, the frequency associated with the grain mobility transition $`f_{gm}`$ is less than the frequency of the square/stripe transition. In this case and for $`\stackrel{~}{\mathrm{v}}\stackrel{~}{\mathrm{v}}_{gm}`$, the resulting wave patterns are noticably more tenuous and have significantly shorter spatial correlation lengths (on the order of $`\lambda `$) than do the corresponding patterns for $`f_{ss}>f_{gm}`$.
### 6.2 Loss of Horizontal Grain Motion
In this section, the breakdown in $`\lambda `$ scaling with $`H`$ and the decrease in the collision pressure are identified with a transition in the horizontal grain mobility. We present data suggesting that for $`\stackrel{~}{\mathrm{v}}>\stackrel{~}{\mathrm{v}}_{gm}`$, the local layer height is changed by lateral grain motion, while for $`\stackrel{~}{\mathrm{v}}<\stackrel{~}{\mathrm{v}}_{gm}`$, grains are essentially immobile . In Ref. , Mujica and Melo similarly propose that waves at low container velocities result from bending of the layer and not from horizontal grain transfer. Also, in our direct visual observations we observe significant horizontal particle motion for waves at large $`\stackrel{~}{\mathrm{v}}`$ but no noticeable horizontal grain motion for waves at small $`\stackrel{~}{\mathrm{v}}`$. In general for waves at both high and low $`\stackrel{~}{\mathrm{v}}`$, patterns persist when the container is rapidly brought to rest. When the stationary container is lightly and repeatedly tapped, the patterns at low $`\stackrel{~}{\mathrm{v}}`$ slowly disappear without any apparent grain motion on the layer surface, whereas the high $`\stackrel{~}{\mathrm{v}}`$ patterns exhibit significant grain rearrangement.
To quantitatively investigate the proposed grain mobility transition, fluctuations in the local collision pressure are studied. In the presence of waves and for $`\stackrel{~}{\mathrm{v}}>\stackrel{~}{\mathrm{v}}_{gm}`$, relatively large fluctuations in the collision pressure associated with changes in the local layer height due to grain motion are expected. Conversely, for $`\stackrel{~}{\mathrm{v}}<\stackrel{~}{\mathrm{v}}_{gm}`$ fluctuations should be smaller because the layer height is essentially constant. Figure 13(a) plots the relative standard deviation of the maximum pressure $`\sigma _{P_{max}}=\sqrt{(\mathrm{\Delta }P_{max})^2}/P_{max}`$ versus $`\stackrel{~}{\mathrm{v}}`$ for both the flat-layer and wave regimes where $`\sqrt{(\mathrm{\Delta }P_{max})^2}`$ and $`P_{max}`$ are the standard deviation and mean of $`P_{max}`$ respectively. For the flat-layer, $`\sigma _{P_{max}}`$ is small and nearly independent of $`\stackrel{~}{\mathrm{v}}`$. For the wave regime, $`\sigma _{P_{max}}`$ shows three distinct behaviors — $`\stackrel{~}{\mathrm{v}}<3`$: $`\sigma _{P_{max}}`$ is independent of $`\stackrel{~}{\mathrm{v}}`$ and equal to the flat-layer value; $`3<\stackrel{~}{\mathrm{v}}<7`$: $`\sigma _{P_{max}}`$ increases with increasing $`\stackrel{~}{\mathrm{v}}`$ and is larger than the flat-layer value; $`\stackrel{~}{\mathrm{v}}>7`$: $`\sigma _{P_{max}}`$ decreases with increasing $`\stackrel{~}{\mathrm{v}}`$.
As a further check that the equality of $`\sigma _{P_{max}}`$ for the flat-layer and waves for $`\stackrel{~}{\mathrm{v}}<3`$ is due to a loss of horizontal grain motion, Fig. 13(b) presents measurements of the autocorrelation of $`P_{max}`$ at a delay of 4T, $`C_{P_{max}}(4T)`$ . $`C_{P_{max}}(4T)`$ is sensitive to periodic variations in $`P_{max}`$ even when the intrinsic pressure noise is larger than the fluctuations associated with grain motion. As is true for $`\sigma _{P_{max}}`$, $`C_{P_{max}}(4T)`$ for the flat-layer is independent of $`\stackrel{~}{\mathrm{v}}`$. For waves, the dependence of $`C_{P_{max}}(4T)`$ on $`\stackrel{~}{\mathrm{v}}`$ can also be divided into three regimes — $`\stackrel{~}{\mathrm{v}}<3`$ : $`C_{P_{max}}(4T)`$ is constant and equal to the flat-layer value; $`3<\stackrel{~}{\mathrm{v}}<7`$ : $`C_{P_{max}}(4T)`$ is constant and larger than the flat-layer value; $`\stackrel{~}{\mathrm{v}}>7`$ : $`C_{P_{max}}(4T)`$ decreases with increasing $`\stackrel{~}{\mathrm{v}}`$. The transition in $`C_{P_{max}}(4T)`$ to the flat-layer value is sharp and occurs at the same $`\stackrel{~}{\mathrm{v}}`$ where $`\sigma _{P_{max}}`$ for waves reaches its minimum value. Also, note that $`\sigma _{P_{max}}`$ peaks at the velocity ($`\stackrel{~}{\mathrm{v}}=7`$) at which $`C_{P_{max}}(4T)`$ begins to decrease for increasing $`\stackrel{~}{\mathrm{v}}`$.
Our interpretation of the three regimes for $`\sigma _{P_{max}}`$ and $`C_{P_{max}}`$ shown in Fig. 13 is as follows. For $`\stackrel{~}{\mathrm{v}}<3`$ horizontal grain mobility is strongly reduced and the layer depth is everywhere equal. Waves in this regime are likely bending waves since mass transfer, dilational waves , or any other mode giving rise to a periodic variation in pressure would produce positive correlations in the pressure fluctuations. For $`\stackrel{~}{\mathrm{v}}>3`$ waves are mass transfer waves. We speculate that the decrease in $`\sigma _{P_{max}}`$ and $`C_{P_{max}}`$ above $`\stackrel{~}{\mathrm{v}}=7`$ results from the disordered patterns found in this regime which possibly result from an instability associated with the more rapid growth of the wave amplitude in comparison to $`\lambda `$ as $`f`$ decreases .
Having made the case for a grain mobility transition in terms of $`\sigma _{P_{max}}`$ and $`C_{P_{max}}`$, we now examine the dependence of the transition on $`\mathrm{v}`$, $`N`$, and $`D`$. Figure 14 presents $`\sigma _{P_{max}}`$ for varying $`\mathrm{\Gamma }`$ and nine distinct $`f`$ values ranging from $`0.05\sqrt{g/D}`$ to $`0.33\sqrt{g/D}`$, as a function of $`\stackrel{~}{\mathrm{v}}`$. Most importantly, Fig. 14 shows that variations in $`\mathrm{\Gamma }`$ at constant $`f`$ produce the same changes in $`\sigma _{P_{max}}`$ as do variations in $`f`$ at constant $`\mathrm{\Gamma }`$ (see Fig. 13(b)). This finding strengthens our proposition that $`\stackrel{~}{\mathrm{v}}=\mathrm{\Gamma }/(2\pi f\sqrt{D/g})`$ is the correct dimensionless parameter specifying the grain mobility. The data also show that $`\stackrel{~}{\mathrm{v}}_{gm}3`$, which is in agreement with the value obtained from the pressure fluctuation data for $`\mathrm{\Gamma }=3.0`$.
Figure 15(a) examines the dependence of grain mobility on layer depth by plotting $`C_{P_{max}}(4T)`$ versus $`N`$ for $`\stackrel{~}{\mathrm{v}}=3.4`$ and $`\stackrel{~}{\mathrm{v}}=4.2`$. By choosing $`\stackrel{~}{\mathrm{v}}`$ slightly above $`\stackrel{~}{\mathrm{v}}_{gm}3`$ (for $`N=9`$), $`C_{P_{max}}(4T)`$ is sensitive to the grain mobility transition. Below $`N=14`$, the layer response for the different $`\stackrel{~}{\mathrm{v}}`$ is similar. For $`N<5`$, $`C_{P_{max}}(4T)`$ decreases to near $`0`$ at $`N2`$, while for $`5<N<14`$, $`C_{P_{max}}(4T)`$ decreases slightly with increasing $`N`$. The dependence of $`C_{P_{max}}(4T)`$ on $`N`$ at small $`N`$ appears similar to that of $`C_{P_{max}}(4T)`$ on $`\stackrel{~}{\mathrm{v}}`$ at large $`\stackrel{~}{\mathrm{v}}`$, suggesting that the two limits of large $`\stackrel{~}{\mathrm{v}}`$ and small $`N`$ are related. For $`N>14`$, $`C_{P_{max}}(4T)`$ continues to decreases slowly with increasing $`N`$ for $`\stackrel{~}{\mathrm{v}}=4.2`$, but, for $`\stackrel{~}{\mathrm{v}}=3.4`$ and $`N14`$, $`C_{P_{max}}(4T)`$ drops to approximately the flat layer value. Possibly, the reduction in grain mobility with increasing $`N`$ is due to a decrease in grain velocity associated with an increase in the grain collision frequency. A more mundane explanation is as $`N`$ increases, the layer mass becomes significant with respect to the container mass. This reduces the change in layer velocity at impact and subsequently decreases the effective driving.
Finally, Fig. 15(b) looks at the effect of varying $`D`$ on the grain mobility by plotting $`C_{P_{max}}(4T)`$ versus $`\stackrel{~}{\mathrm{v}}`$ for $`D=0.17`$, 0.33, 0.46, and 0.66 (pressure data for constant $`\stackrel{~}{\mathrm{v}}`$ and varying $`N`$ was not collected). The dependence of $`C_{P_{max}}(4T)`$ on $`\stackrel{~}{\mathrm{v}}`$ and the value of $`\stackrel{~}{\mathrm{v}}_{gm}3`$ is in accord with the data for constant $`D`$ and $`N`$ and varying $`f`$ shown in Fig. 13 and is consistent with our proposal that $`\stackrel{~}{\mathrm{v}}_{gm}`$ is independent of $`D`$. In addition to bronze, Fig. 15(b) also includes data for layers of lead and steel particles.
In summary, $`\stackrel{~}{\mathrm{v}}_{gm}`$ values obtained from local pressure measurements and from identifying the breakdown in wavelength scaling with $`H`$ are in good qualitative agreement. Both show the same weak dependence on $`N`$ for small $`\mathrm{\Gamma }`$ and are independent of $`D`$. Quantitatively, the value of $`\stackrel{~}{\mathrm{v}}_{gm}`$ for $`N<12`$ obtained from the dispersion data ($`2.6`$) is somewhat smaller that the value obtained from the local pressure fluctuations ($`3`$). A possible explanation for this discrepancy is the layer “freezes” from the bottom up as $`\stackrel{~}{\mathrm{v}}`$ is reduced. For $`2.6<\stackrel{~}{\mathrm{v}}<3.0`$ the upper portion of the layer may be mobile but the associated mass variations are not seen in the pressure fluctuations because the variation in $`P`$ is small relative to the inherent pressure noise associated with a flat layer and/or stress chains in the frozen portion of the layer wash out the pressure fluctuations by distributing them over an area comparable to or larger than $`\lambda `$.
## 7 Conclusion
Two length scales, the layer depth $`H`$ and the particle diameter $`D`$, control the scaling properties of granular waves and granular wave patterns. As a guide, Table 2 provides a summary of the relevant dimensionless parameters and their associated transition values and figures. The horizontal mobility of grains appears to be determined by the container velocity — above a critical value $`\mathrm{v}_{gm}3\sqrt{Dg}`$, grains are mobile while below $`\mathrm{v}_{gm}`$, the layer depth remains uniform throughout the oscillation cycle. The physical mechanism responsible for the transition remains to be understood. The sharpness of the grain mobility transition stands in contrast to the gradual change in properties of gases as the length scale of disturbances approaches the mean free path. Additionally, it is surprising that finite grain size effects are made manifest via $`\stackrel{~}{\mathrm{v}}`$ rather than $`H/D`$ or $`\lambda /D`$ as mentioned in Sec. 5.
For $`\mathrm{v}>\mathrm{v}_{gm}`$, the layer behaves as a continuum with $`\lambda `$ and $`f`$ scaling with $`H`$ and $`\sqrt{H/g}`$, respectively. With this scaling, dispersion curves for different layer depth and particle size collapse onto the curve given by Eq. 2. The square/stripe transition for most parameter values occurs within the continuum regime ($`\mathrm{v}>\mathrm{v}_{gm}`$), depends simply on the layer depth $`H`$ ($`f_{ss}^{}=f_{ss}\sqrt{H/g}0.33`$), and shows no systematic dependence on $`\mathrm{\Gamma }`$, $`N`$, $`D`$, or particle composition (for lead, bronze, and steel). We do not yet understand why $`f_{ss}^{}=0.33`$, but speculate it is perhaps related to the dynamic angle of repose. For $`\mathrm{v}<\mathrm{v}_{gm}`$, scaling with $`H`$ fails. Within this regime, local pressure measurements support the idea that waves result from a bending of the grain layer (see Mujica et al. Ref. ), although there is a large discrepancy between our value of $`\stackrel{~}{\mathrm{v}}_{gm}3`$ and that found by Mujica et al.$`\stackrel{~}{\mathrm{v}}_{gm}0.6`$. Possibly, our measurements indicate where horizontal grain mobility stops whereas the value reported by Mujica et al. represents the cessation of relative vertical motion as well. Another interesting question is the nature of stripe patterns above the grain mobility transition. Does wave scaling with $`H`$ exist in this regime? Casual observations show that stripes above the transition have shorter spatial and temporal correlations than do stripes in the sloshing regime.
Finally, the flat-layer/waves transition occurs when the layer remains dilated after making contact with the plate. For decreasing $`\mathrm{\Gamma }`$, this transition occurs at a nearly constant value of $`\mathrm{\Gamma }=2.2`$, which is independent of $`f`$ and $`N`$. Associated with the onset of waves is a sudden decrease in the collision pressure. Possibly the least understood aspect of the flat-layer/waves transition is the causative connection between the dilation and the pattern.
There remain numerous unanswered questions concerning wave patterns in vertically oscillated granular layers. Increased computational power, better simulation techniques, improved measurement capabilities, and application of Navier-Stokes-like continuum equations promise significant new insights into this system in particular and into the dynamics of granular media in general.
## Acknowledgments
The authors thank Chris Bizon and Mark Shattuck for helpful comments, Dan Goldman for experimental assistance, and Francisco Melo for illuminating discussions. This research was supported by the Engineering Research Program of the Office of Basic Energy Sciences of the U.S. Department of Energy and by the U.S. National Science Foundation Division of International Programs (Chile). |
no-problem/0001/astro-ph0001244.html | ar5iv | text | # New Models for X-Ray Synchrotron Radiation from the Remnant of Supernova 1006 AD
## X-ray Synchrotron Emission from SNRs
Young supernova remnants are in a period of transition to the Sedov-Taylor phase, sweeping up several times their original ejected mass as they expand into a circumstellar medium. The shocked gas at temperatures of 10<sup>7</sup> K emits a thermal spectrum (bremsstrahlung and lines) through interactions of electrons with ions. However, the synchrotron emission, which produces the radio spectrum, persists through the X-ray regime (and in some case dominates the thermal X-rays koyama95 ).
In order to study the energetics of the accelerated electrons and protons in the shock, as well as to obtain accurate abundances from thermal models, it is imperative to understand this synchrotron spectrum. SNR observations show that the synchrotron emission drops significantly below a powerlaw at X-ray frequencies, implying a curved synchrotron spectrum. To accurately describe the synchrotron spectrum we have developed the following models:
* SRCUT A homogeneous population of electrons, with an exponentially cut off power-law energy distribution, radiating in a constant magnetic field, produces the sharpest physically plausible cutoff in the emitted spectrum. Given a radio flux and spectral index, the models depend on a single parameter: the frequency at which the spectrum has dropped by a factor of 10 below the power-law extrapolation. If a remnant emits primarily thermal X-rays, SRCUT will give the maximum nonthermal-electron energies allowed, so as not to exceed observed X-ray fluxes. SRCUT fits to 14 galactic SNRs give electron upper limits of 100 TeV or less, well below the 1000 TeV “knee” in the cosmic-ray (ion) spectrum.keohane99
* SRESC In some remnants the energy of the most energetic electrons will be limited because the shock cannot effectively scatter particles above a certain gyroradius. The SRESC model describes Sedov expansion into a uniform magnetic field, appropriate for Type Ia and late core-collapse SNR. Details are discussed in Reynolds (1998).reynolds98
These models are currently available and will be distributed in the next release of XSPEC. The XSPEC model formats should also be compatible with CXC, the Chandra software. More sophisticated models, as described in Reynolds (1998)reynolds98 , will be released in the future.
## The Escape Model
The electron scattering required for acceleration to high enough energies to produce synchrotron radiation may become less efficient above some energy. Or, in the limiting case, electrons above some energy may freely escape the SNR. In the escape model we assume that magnetohydrodynamic scattering waves are much weaker above some wavelength $`\lambda _{max}`$. Since electrons with gyroradius $`r_g`$ scatter resonantly with waves of wavelengths $`\lambda =2\pi r_g`$, electrons will escape once their energy reaches $`E_{max}`$ given by:
$$E_{max}=\lambda _{max}B_1/4$$
(1)
where $`B_1`$ is the upstream magnetic field strength.
Reynolds (1998) produced a detailed model for SNR emission with electron energies limited by escape, which includes correct accounting for variation of shock-acceleration efficiency and post-shock radiative and adiabatic losses, assuming Sedov dynamics.
The model has three parameters: 1) the radio flux measurement at 1 GHz, 2) $`\alpha `$, the radio spectral index (flux density $`\nu ^\alpha `$), 3) a characteristic rolloff frequency, the frequency at which the spectrum has dropped by approximately 10 below a straight powerlaw. The spectral index and 1 GHz flux for SNR are fixed by observations. For Galactic SNRs, they can be found at Green’s website.green98
## SN 1006 AD
SN 1006 AD is an example of a young remnant where the synchrotron emission dominates in the X-ray regimekoyama95 . Early observations, with low signal to noise, could be adequately fit by curved synchrotron models reynolds96 . Observations from RXTE are a 10-fold improvement in sensitivity and provide a more sensitive test.
ASCA and RXTE observations of SN 1006 AD can be adequately fit with two components: the escape model and a thermal plane-parallel nonequilibrium-ionization shock model (VSHOCK) with variable abundances (Borkowski et al., in preparation). The model fit depends relatively weakly on the flux at 1 GHz. For SN 1006 AD the flux was fixed at the observed value of 19 Jygreen98 . The column density was fixed at $`5\times 10^{20}`$ cm<sup>-2</sup>, a value consistent with optical and ROSAT PSPC observations. The best fit, shown in Figure 1, has a $`\alpha `$=0.58 and the rolloff frequency of 1.7$`\times 10^{17}`$ Hz with a $`\chi ^2`$ of 665 for 376 degrees of freedom. The VSHOCK obtains abundances (relative to solar) of O 0.27, Ne 0.16, Mg 1.56, Si 6.43 and S 8.16, compared in Figure 2 to predictions for Type Ia supernova by Nomoto et al.(1984).nomoto84 From the synchrotron formula, $`\nu _mE^2B`$, and equation 1, assuming a magnetic field of 3$`\mu `$G, we calculate a $`\lambda _{max}`$ of 1.2$`\times 10^{17}`$ cm.
X-ray emission from SN1006 AD is well fit by synchrotron models with electron energies limited by escape. Comparing the escape model to future models that are limited by radiation losses or SNR age will give us a better understanding of synchrotron emission at X-ray energies. Our future work involves applying these models to SNRs and making them widely available to the community by distribution through XSPEC.
Thanks to G. Allen for sharing RXTE X-ray observations of SN1006 AD and J. Keohane for research notes and advice. This research is supported by NASA grant NAG5-7153 and NGT5-65 through the Graduate Student Researchers Program. |
no-problem/0001/cond-mat0001018.html | ar5iv | text | # 1 Introduction
## 1 Introduction
In the last 40 years history of surface growth science, computer simulations were particularly useful and important for better understanding of modes and mechanisms of growth. Great number of simulations is based on so-called solid-on-solid (SOS) model approximation, when each particle seats on the top of the others and film surface may be described by a single-valued function of its height $`h(𝐫)`$ in planar substrate coordinate $`𝐫`$ (see \[1-3\] for review). In this paper, we present a short review of our recent works \[4-6\] as well as current investigation and oncoming ones . In all these papers we concentrate on films growth on lattice with simple cubic symmetry.
Section 2 is devoted to presentation of the model. In Section 3 we present results which prove that a considerable speed-up of computations may be achieved at cost of some assumed simplification of growth rules and particle behaviors in computer simulation. Then influence of model control parameters on surface morphology and its dynamics is presented in Section 3 and summarized in Section 4.
## 2 Model
Presented here results of simulations are based on simple SOS model of epitaxial thin films growth \[4-8\] based on Wolf-Villain model , in which after randomly chosen place of a particles initial contact with the surface, they relax to the site offering the strongest bonding (i.e. they maximize the number of particle-particle lateral bonds (PPLB)). The place of particle settling is probed among site of initial deposition plus their four nearest-neighbors. Then each particle repeats relaxation $`L_{dif}`$ times and so $`L_{dif}`$ may be considered as a range of diffusion.
This simple and deterministic rules of growth are generalized in this paper by accounting on:
* strengths of horizontal bonds ($`J_x`$, $`J_y`$) and vertical bonds ($`S_x`$, $`S_y`$) of PPLB in particle level layer and one layer below, respectively,
* diffusion barrier ($`V_x`$, $`V_y`$), which restricts migration and stops it for infinitely large values of $`V`$,
* temperature of substrate $`T`$.
The temperature $`T`$ entries calculation by the Boltzmann factor $`\mathrm{exp}(E/kT)`$ for probability $`p(E)`$ of picking a site of energy $`E`$, which replaces the deterministic rule of fixing the particle at site of minimum $`E`$. Desorption process is neglected becouse typical desorption energies for metals are of the order of several eV. With such modification we may grow anisotropic surfaces for model control parameters ($`S_xS_y`$, $`J_xJ_y`$, and/or $`V_xV_y`$) . The diffusion barrier $`V`$ reduces probability of movement by factor $`\mathrm{exp}(V_x/kT)`$ or $`\mathrm{exp}(V_y/kT)`$ for considered direction $`x`$ or $`y`$ of migration. $`V`$ values are always positive, while negative $`J`$ and $`S`$ correspond to the assumed tendency to maximize PPLB. Additionally, the presence of $`S`$ bonds to particles below in particle energy may play role of the Ehrlich-Schwoebel barrier near the film edge . The $`S`$ term makes it difficult to reach the step edge and thus limiting horizontal growth of a cluster of atoms .
## 3 Results of Simulations
The simulation were carried out on $`L\times L`$ large square lattice with periodic boundary conditions to minimize influence of boundary effects. The particles are represented as unit cubes which can occupy only discrete position in (2+1)-D space.
### 3.1 Diffusion Range
Firstly, we evaluate speed-up of computation when Monte Carlo simulations based on Arrhenius dynamics are reduced to oversimplified random deposition (RD) model through intermediate stages. If $`t_d`$ is computing time to deposite a particle at the first contact site somewhere on the surface, then we need at least $`t_{RD}=Nt_d`$ time, where $`N`$ is total number of the deposited particles, to complete simulation. The RD model correspond to $`L_{dif}=0`$. Increasing $`L_{dif}=1`$ results in changing model to local diffusion with single step relaxation and increase time of computation by the additional time $`t_r`$ necessary for relaxation procedure: $`t_{SSR}=Nt_d+Nt_r`$. Case $`L_{dif}>1`$ corresponds to the local diffusion model when simulation time may be evaluated as $`t_{LD}=Nt_d+NL_{dif}t_r`$. Finally, for full-diffusion model ($`L_{dif}\mathrm{}`$) we have $`t_{FD}=Nt_d+(1+2+\mathrm{}+(N1)+N)t_r`$. Note, that almost deterministic rules of growth speed-up computation on a ratio $`k=t_{FD}/t_{LD}(N+1)/2L_{dif}\theta L^2/2L_{dif}`$, where the total number of particles $`N=\theta L^2`$ is given by lattice lateral size $`L`$ and film thickness/substrate coverage $`\theta `$. And — as we shall show later — for such simplification we may loose some significant film properties and/or misinterpret particles dynamics.
The range of diffusion $`L_{dif}`$ influences surface roughness measured by standard surface width $`\sigma `$, defined as the root-mean-square of surface heights. We found that increasing of diffusion range make surface smoother, however some saturation level $`\sigma _{sat}(L_{dif}\mathrm{})>0`$ is observed. The saturation level depends on strong of attractive forces among particles which is governed by $`J`$ . The results are collected in Tab. 1. Note, that effective diffusion range $`L_{eff}`$ may be also controlled by substrate temperature $`T`$ as we show later in Section 3.4.
### 3.2 Growth Dynamics
The roughness $`\sigma `$ for film surfaces growing by deposition often obeys Family-Vicsek scaling law: $`\sigma L^\alpha f(\theta /L^z)`$, where scaling function is given by $`f(x)x^\beta `$ for $`x1`$ and $`f(x)1`$ for $`x1`$ . For substrate low coverage ($`\theta <1`$ ML) also third regime is observed with growth exponent $`\beta =\beta _{RD}=1/2`$, similar to simple RD . Deviation from scaling hypotheses may be also observed when particles are allowed to climb on the top of other ones which results an unstable growth when surface width growth infinitely with increasing of coverage $`\theta `$ . This may be considered as infinite value of roughness exponent ($`\alpha \mathrm{}`$).
When also anisotropy in growth is included (e.g. for Ag/Ag(110) or Cu/Cu(110)), then we observe power law for roughness $`\sigma \theta ^\beta `$ for small $`\theta `$ (model parameters for Cu and Ag are taken after ). Before reaching $`\theta =10^4`$ ML we still do not observe roughness saturation which may indicate that particles jumps to the higher levels may be not essential for presence of unstable growth ($`\alpha \mathrm{}`$) as it was earlier suggested in . In both cases, either stable or unstable growth, surface roughness $`\sigma `$ increases initially as $`\theta ^\beta `$. The growth exponent $`\beta `$ may also depend on substrate crystallographic orientation as was shown by Elliot et al. for Ag(111) and Ag(100) . This can be explained only within models that allow for anisotropy.
Family-Vicsek law fails for inhomogeneous multilayerd $`a/b/a`$-like sandwiches as a result of effective changes of growth rules — or more precisely — strength of interaction during deposition .
### 3.3 Surface Morphology Anisotropy
As we mentioned in Section 2, the anisotropy may originate from either anisotropy in bonds ($`S_xS_y`$, $`J_xJ_y`$) or anisotropy in barriers $`V_xV_y`$ . Then anisotropy may be quantitatively measured from the difference in the height-height correlation function $`G(𝐬)h(𝐫+𝐬)h(𝐫)h(𝐫)^2`$ in two $`x`$ and $`y`$ directions, where $`\mathrm{}`$ denotes spatial average over all sites $`𝐫`$. Correlation $`G(1,0)`$ are positive for terrace-like, smooth surfaces (in $`x`$\- direction) while its negative values correspond to spiky and rough structures. Thus $`G(1,0)G(0,1)`$ normalized to $`\sigma ^2=G(0,0)`$ may be used as a quantitative measure of anisotropy $`\epsilon [G(1,0)G(0,1)]/G(0,0)`$ and may be competitive to a direct pictures of the surface morphology obtained from simulations . We found that anisotropy parameter $`\epsilon `$ depends on the ratio of $`|J_xJ_y|/|J_x+J_y|`$ and not on the simple difference $`|J_xJ_y|`$ only. As it may be expected, the anisotropy vanishes for very large diffusion barriers $`V_x=V_y\mathrm{}`$, when RD results are reproduced. When the first monolayer is completed, the anisotropy parameter $`\epsilon `$ seems to become insensitive to increasing coverage $`\theta `$ .
### 3.4 Influence of Substrate Temperature on Film Properties
We expect that increasing temperature should make the surface smoother since energetic barriers are easier to overcome and also surface bonds may be broken more easily. The effect of temperature increase is presented in Tab. 2. Model parameters are taken after where simulations of Pt on Pt(111) behaviors were studied.
Dependence of effective diffusion range with substrate temperature may also be easily predicted; for low temperatures particles freeze at the place of deposition and migration is stopped ($`T0L_{eff}=0`$). Then we reproduce RD results with Poisson distribution of film heights. On contrary, hot substrate make particles mobile ($`T\mathrm{}L_{eff}\mathrm{}`$) and particles stop only if diffusion range is reached in the number of subsequent jumps . The quantity introduced to describe particle mobility is number $`M`$ of particles which are still unsettled on the surface when the diffusion range is reached, that is the number $`M`$ of particles which changed their position during the last step ($`L_{dif}`$) any iteration cycle. $`M`$ is increasing function of substrate temperature $`T`$, as expected (see Tab. 3). Note, that the increase in $`L_{dif}`$ plays similar role as an increase in substrate temperature $`T`$ for stochastic deposition processes.
Finally, we check how temperature influences surface morphology measured by $`\epsilon `$ and applied to Cu/Cu(110) and Ag/Ag(110). For absolute zero temperature ($`T0`$) we recover RD results ($`\epsilon 0`$). With increasing temperature the anisotropy coefficient reaches maximum and then drops, so that, we again have RD results $`\epsilon =0`$ for $`T\mathrm{}`$. In intermediate temperature surface morphology evolves from isolated small islands, through strongly anisotropy long string of atoms along dimmer rows to larger and more isotropic islands also across dimmer rows .
## 4 Summary
The described model proposed in \[4-8\] allows for qualitatively correct description of thin solid films grown by molecular beam epitaxy where dominant physical processes responsible for film growth is surface diffusion. We show that simplification of rules of growth — when next particle incoming only when position of a previous one was fixed — speeds up computation drastically particularly for large lattices and large coverage. This is justified for not very large flux of the incoming particles. Actually, bigger $`L_{dif}`$ also mimics smaller flux of the deposited particles. The anisotropy in the model parameters allows to consider not only sc(100) symmetry surfaces, but also particles deposited on bcc(110) surfaces .
From simulation we were able to reproduce data on thermal, time and structural evolution of surface morphology. Usually the surface roughness measured by standard deviation of film height increases with time (or coverage) accord to power law . Such behavior is characteristic for early stages of growth of self-affine surfaces (see for review).
We believe that it is important to study the fractal dimension for the (2+1)-D surfaces for better understanding of the growth phenomena. We also intend to investigate more throughly the anisotropic growth .
## Acknowledgments
I am very grateful to A.Z.Maksymowicz for fruitful discussions and scientific guidance. Paper was partially supported by Polish Committee for Scientific Research (KBN) under grant 8 T11F 02616. The calculations were carried out in ACK-CYFRONET- AGH. The machine time is financed by KBN with grants KBN/C3840/AGH/008/1997, KBN/S2000/AGH/069/1998 and KBN/HP-K460-XP/AGH/069/1998. |
no-problem/0001/astro-ph0001030.html | ar5iv | text | # A Hard Medium Survey with ASCA.III.: a Type 2 AGN revealed from X-ray spectroscopy Based on observations performed at the European Southern Observatory, Paranal, Chile
## 1 Introduction
Absorbed (Type 2) AGN have been proposed as the major contributors to the Cosmic X-ray Background (CXB) above 2 keV (see among others Setti and Woltjer, 1989; Madau, Ghisellini and Fabian, 1994; Comastri et al., 1995). Indeed, recent results from ASCA and Beppo-SAX observations (Akiyama et al., 1998a; Bassani et al.,1999; Risaliti, Maiolino and Salvati, 1999; Fiore et al., 1999) seem to favor this hypothesis but deeper investigations are still needed to confirm and refine this scenario and to test competing models. For example it is not clear if high luminosity Type 2 AGN exist or not (see e.g. Halpern, Turner and George, 1999; Akiyama et al., 1998a) or if a significant fraction of the hard CXB deriving from star-burst and star-forming galaxies at moderate redshift can be excluded or not (see e.g. Moran, Lehnert and Helfand, 1999).
At the Osservatorio Astronomico di Brera, with the specific aim of extending to faint fluxes the census of the X-ray sources shining in the hard X-ray sky, we have initiated a few years ago the ASCA Hard Serendipitous Survey (HSS: see Della Ceca at al., 1999a for a progress update): a systematic search for sources in the $`210`$ keV band, using data from the GIS2 instrument on board the ASCA satellite (Tanaka, Inoue and Holt, 1994). This effort has lead to a “pilot” sample of 60 sources (Cagnoni, Della Ceca and Maccacaro, 1998) whose spectral properties have been presented and discussed by Della Ceca at al., (1999b). The ASCA HSS has been extended and it now covers $`71`$ deg<sup>2</sup> of sky. The resulting source sample consists of 189 sources detected with a signal-to-noise ratio $`4.0`$ (Della Ceca at al., 1999a, 2000), a more restrictive criterion than that used in Cagnoni, Della Ceca and Maccacaro, 1998 where a signal-to-noise ratio $`3.5`$ was used.
In this paper we report the serendipitous discovery of a low luminosity Type 2 AGN: AXJ2254+1146. Its absorbed Type 2 nature has been revealed by the X-ray data alone, thanks to the good spectral resolution of the ASCA GIS ($`\mathrm{\Delta }E/E8\%`$ at $`67`$ keV). We are convinced that the new generation of X-ray telescopes (e.g. Chandra, XMM) will make AXJ2254+1146 only the first one of several objects whose nature will be discovered by means of the X-ray data alone.
The paper is organized as follows. In section 2 we present the X-ray data and discuss the X-ray spectral properties of AXJ2254+1146. In section 3 we present its optical identification (UGC 12237), while in section 4 we discuss the overall Spectral Energy Distribution (SED) of the source and compare it with the SED of Seyfert 2 and non-AGN spiral galaxies. An optical spectrum of UGC 12237 is shown and discussed in section 5. Finally, summary and conclusions are presented in section 6. We adopt $`H_0=50`$ km s<sup>-1</sup> Mpc<sup>-1</sup> and $`q_0=0`$ throughout.
## 2 X-ray Data and Analysis
### 2.1 X-ray Imaging
AXJ2254+1146 was discovered in the ASCA GIS field 74076000 <sup>1</sup><sup>1</sup>1The ASCA satellite also has two solid-state imaging spectrometer (SIS) on the focal plane. Unfortunately this observation was operated in the “2 CCD mode” configuration and AXJ2254+1146 is out of the investigated sky region. So we do not have SIS data on this source. pointed at the radio loud quasar PKS2251+113 ($`\alpha _{2000}=`$ $`22^h54^m10.5^s`$ and $`\delta _{2000}=+11^o36^{}38\mathrm{"}`$; z=0.323). The field is one of the many used to extend the survey of Cagnoni, Della Ceca and Maccacaro 1998; data preparation and analysis have been discussed in detail therein and in Della Ceca et al., 1999b. This source attracted our attention because its hardness ratios (HR1 = $`0.41\pm 0.12`$; HR2 = $`0.47\pm 0.06`$) <sup>2</sup><sup>2</sup>2 HR1 and HR2 are defined in the following way:
$$HR1=\frac{MS}{M+S}HR2=\frac{HM}{H+M}$$
where S, M and H are the corrected net counts in the 0.7-2.0, 2.0-4.0 and 4.0-10.0 keV energy band respectively (cf. Della Ceca et al., 1999b). can not be explained by a single absorbed power law model, implying a more complex and structured X-ray spectrum (cf. Della Ceca et al., 1999b).
Panel (a) of Figure 1 shows the central ($`32^{}\times 32^{}`$) smoothed <sup>3</sup><sup>3</sup>3The smoothing function is a two-dimensional Gaussian filter with $`\sigma =1^{}`$, which is comparable to the “core” of the (XRT+GIS2) point-spread function (PSF). GIS2 hard (2-10 keV) image with overlaid a contour plot indicating the X-ray emission intensity. Three sources are clearly visible in the 2-10 keV band image: the target of the observation, (AXJ2254+1136 $``$ PKS2251+113), plus two serendipitous sources, AXJ2254+1141 and AXJ2254+1146. AXJ2254+1141 is identified with MS2252.2+1126, a Type 1 AGN at z=0.0281 (Stocke et al., 1991, Maccacaro et al, 1994).
Panel (b) shows a “hard-over-soft” image obtained dividing the smoothed “hard” (2-10 keV) image by the smoothed “soft” (0.7 - 2.0 keV) image. The same contours of panel (a) have been overlaid on the image for ease of comparison. AXJ2254+1146 stands out as a very hard source compared to the other ones.
In Table 1 the X-ray position, the optical position and the offset between the two (X-ray - optical) are listed for the three sources previously mentioned. The measured offsets are consistent with the positional uncertainties of the ASCA GIS images <sup>4</sup><sup>4</sup>4The positional uncertainty of the ASCA detected sources has been investigated by Gotthelf (1996) using 48 point-like sources. He reached the conclusion that, in the case of the combination XRT+SIS, the 90% confidence level error circle is of the order of $`40\mathrm{"}`$ radius, while in the case of the XRT+GIS additional instrumental and calibration uncertainties can produce larger errors. From the analysis of a sample of 168 point-like ASCA targets within the data set used for the ASCA HSS we have estimated a 90% confidence level error circle of $`2`$ arcmin radius. .
### 2.2 X-ray Spectrum
To maximize statistics and signal-to-noise ratio, total counts (source + background) were extracted from a circular region of 2 arcmin radius around the source centroid in the GIS2 and GIS3 images. Background counts were taken from two circular uncontaminated regions of 4.75 arcmin radius close to the source. Source and background data were extracted in the “Pulse Invariant” (PI) energy channels, which have been corrected for spatial and temporal variations of the detector gain. The Ancillary Response File (ARF) was created with version 2.72 of the FTOOLS task ASCAARF at the location of the source in the detector. In order to improve the statistics, we have produced a combined GIS spectrum (adding the GIS2 and GIS3 data) and the corresponding background and response matrix files, following the recipe given in the ASCA Data Reduction Guide (rev 2.0; see section 8.9.2 and 8.9.3 and reference therein) and using the FTOOLS 4.2 software package (supplied by HEASARC at the Goddard Space Flight Center). The total ($``$ GIS2 + GIS3) net counts in the source extraction region and in the energy interval used (0.7 – 10 keV) are $`290\pm 20`$, overimposed over a total $`110`$ background counts. In order to use the $`\chi ^2`$ statistics in the spectral fitting procedure the combined GIS spectrum was rebinned to give at least 15 total counts per energy bin. Spectral analysis has been performed using XSPEC 10.0. All the models discussed below have been filtered by the Galactic absorption column density along the line of sight ($`N_{H_{Gal}}=5.2\times 10^{20}`$ cm<sup>-2</sup>; Dickey and Lockman, 1990). All the errors reported in the spectral analysis are at the 68% confidence level for 1 interesting parameter ($`\mathrm{\Delta }\chi ^2=1.0`$), unless explicitly quoted.
Single-component models do not provide an adequate description of the spectrum of AXJ2254+1146. A single temperature Raymond-Smith model or a single power-law model are both rejected at a high confidence level ($`99.999\%`$).
We have first tried a single absorbed power law model. The best fit values for the power law photon index ($`\mathrm{\Gamma }`$) and for the absorbing column density ($`N_H`$) are $`2.18_{1.68}^{2.80}`$ and $`1.67_{1.32}^{2.15}\times 10^{23}`$ cm<sup>-2</sup>, respectively. Although this fit is formally acceptable ($`\chi ^2=29.6`$ for 23 degree of freedom) the residuals (reported in figure 2) show an excess below $``$ 2 keV and also around $``$ 6.5 keV; a more complex spectrum is thus needed to describe the overall spectral properties of AXJ2254+1146.
We have then tried a spectral model consisting of a narrow (unresolved) Gaussian line at $``$ 6.4 keV, plus the so called “leaky-absorber” continua; the latter being composed by an absorbed plus a non-absorbed power law model having the same photon index. The absorbed power law model represents the “first order” AGN spectrum transmitted through an absorbing cold medium (torus ?) while the non-absorbed power law model represents the primary AGN spectrum scattered into our line of sight by a warm, highly ionized gas located outside the absorbing medium. The scattered fraction is given by the ratio between the normalization of the non-absorbed and absorbed power law components.
The results are reported in Table 2, while in Figure 3 we show the unfolded and folded spectrum and the residuals to the best fit model. The best fit values are $`\mathrm{\Gamma }`$ = $`2.51_{2.17}^{2.76}`$ and $`N_H`$ = $`1.85_{1.47}^{2.24}\times 10^{23}`$ cm<sup>-2</sup>. The scattering fraction implied from the fit is $`0.7_{0.1}^{1.4}\%`$; the position and the observed equivalent width (EW<sub>obs.</sub> hereafter) of the line are $`6.43_{6.33}^{6.52}`$ keV and $`0.6_{0.36}^{0.84}`$ keV, respectively. Note that EW<sub>obs.</sub> has been computed against the observed total (i.e. absorbed plus non-absorbed) source continuum. As evident from the residuals shown in figure 2 and from the $`\chi _\nu ^2`$ in Table 2 this model represent a good description of the spectrum of AXJ2254+1146. We have also tried a fit with the width of the line as a free parameter; the other best fit values are unaffected from this new fit, while, at the spectral resolution of the GIS2 instrument, the line is unresolved with a $`1\sigma `$ upper limit on the width of 0.6 keV. The best fit intrinsic photon index ($`\mathrm{\Gamma }`$ = $`2.51_{2.17}^{2.76}`$) is steeper than the mean observed photon index in Seyfert 1 galaxies ($`\mathrm{\Gamma }=1.7\pm 0.1`$, Nandra et al., 1997). We have then tried to fit the data using the “leaky absorber” continua plus the narrow Gaussian line model but fixing $`\mathrm{\Gamma }=1.7`$; the absorbing column density, the scattered fraction, the EW<sub>obs.</sub> and the position of the line are consistent, within the errors, with that reported in table 2.
Seyfert 1 and Seyfert 2 galaxies are usually characterized by an emission line at 6.4 keV produced by the K<sub>α</sub> emission of Iron species less ionized than Fe XVI, while in some Seyfert 2 object emission lines from He-like (6.7 keV) and H-like Iron (6.96 keV) have also been observed (see Turner et al., 1997a,b and reference therein). The last two are emitted from highly ionized gas and are supposed to derive from the warm scattering medium located outside the absorbing material, i.e. the same medium which is supposed to scatter the primary AGN spectrum into our line of sight. As shown in Matt et al., 1996 the EW of the He-like and H-like Iron lines, if computed with respect to the scattered component (EW<sub>scatt.</sub> hereafter), can be as high as a few keV; these lines should emerge in Compton thick system ($`N_H>few\times 10^{24}`$ cm<sup>-2</sup>) while they should be strongly diluted if the primary radiation is directly visible (as in the case discussed here above 4 keV). If we take an implausible upper limit of EW<sub>scatt.</sub> = 10 keV for the He-like or the H-like Iron line, we should observe, in the case of AXJ2254+1146, a line with EW<sub>obs</sub> $``$ 0.07 keV; this is about a factor 9 less than that observed. We can then rule out an origin of the observed line from He-like and H-like iron species, leaving the fluorescent line at 6.4 keV produced by low ionization Iron the more plausible possibility. The measured position of the line ($`6.43`$ keV) is consistent, within its 95% confidence range of ($`6.206.65`$ keV), with the low ionization Iron $`K_\alpha `$ line emitted at the redshift of the object (z = 0.029, see section 5).
The EW<sub>obs.</sub> is significantly larger than that observed in Seyfert 1 galaxies (EW = $`230\pm 60`$ eV, Nandra et al., 1997) and is similar to that measured in others Seyfert 2 galaxies (Bassani et al., 1999). The measured EW ($`0.6`$ keV) is about a factor 4.5 greater than that expected from transmission through a spherical uniform shell of neutral material (solar abundances) having the measured intrinsic $`N_H2\times 10^{23}`$ cm<sup>-2</sup> (Leahy and Creighton, 1993) and about a factor 7.5 greater than that expected to be produced from transmission through an absorbing torus with a geometrical configuration like that assumed by Ghisellini, Hardt and Matt, 1994 and having $`N_H2\times 10^{23}`$ cm<sup>-2</sup>. This could suggest a combined origin either from the accretion disk and from the absorbing torus. The scattering fraction ($`1\%`$) is consistent with that observed in other well studied Seyfert 2 galaxies (see Turner et al., 1997a,b and reference therein).
The Iron line equivalent width along with the scattering fraction and the measured absorbing column density clearly allow us to classify AXJ2254+1146 as a Type 2 object directly from X-ray spectroscopy.
We note that two data points at about 3 and 3.5 keV (see figure 3) seem to be in excess with respect to the best fit model; these line energies are in very good agreement with the $`K_\alpha `$ neutral emission lines of Ar (2.96 keV) and Ca (3.7 keV) respectively. Both lines are expected to be present in a Compton reflected spectrum (see e.g the theoretical reflected X-ray spectrum presented in Reynolds et al., 1994).
The observed (de-absorbed from the Galactic $`N_H`$ value) flux implied from the best fit model (see Table 2) in the 0.5$``$2.0 keV, 0.5$``$4.5 keV and 2$``$10 keV energy band are $`3.86\times 10^{14}`$ erg cm<sup>-2</sup> s<sup>-1</sup>, $`2.46\times 10^{13}`$ erg cm<sup>-2</sup> s<sup>-1</sup> and $`1.14\times 10^{12}`$ erg cm<sup>-2</sup> s<sup>-1</sup>, respectively. The intrinsic luminosity (i.e. the luminosity emitted from the nucleus) in the 0.5$``$2.0 keV and in the 2$``$10 keV energy band are $`2.35\times 10^{43}`$ erg s<sup>-1</sup> and $`1.30\times 10^{43}`$ erg s<sup>-1</sup>, respectively.
## 3 Source Identification
Figure 4 shows the Palomar Observatory Sky Survey II (POSS II) image centered on the X-ray position of AXJ2254+1146 with its 90% error circle of 2.0 arcmin radius (see footnote 4). We have also marked with a cross the X-ray position of AXJ2254+1146 if we apply the offset we are measuring for the target of the ASCA observation (see table 1).
The fairly bright ($`m_{B_o}=14.26`$) spiral (Sbc) galaxy present in the error circle is UGC 12237. Using the surface density of similarly bright galaxies of $`<1`$ deg<sup>-2</sup> (Maddox et al., 1990) we estimated that the probability for a galaxy brighter than m = 15 to fall by chance in an X-ray error circle of 2 arcmin radius is $`<3.5\times 10^3`$. This, combined with the other data discussed in this paper (i.e. SED and the optical spectral properties), clearly shows that UGC 12237 is the most probable optical counterpart of AXJ2254+1146.
## 4 Spectral Energy Distribution
UGC 12237 has been detected in the NVSS (Condon et al., 1998) as a 1.4 GHz compact radio source and also in the IRAS survey at 25, 60 and 100 $`\mu m`$ (but not at 12 $`\mu m`$). There are no pointed ROSAT observations containing UGC 12237 nor is the galaxy detected in the ROSAT All Sky Survey. The only available X-ray measurement is an upper limit of $`2.5\times 10^{13}`$ erg cm<sup>-2</sup> s<sup>-1</sup> in the 0.5-4.5 keV energy band from the Einstein satellite (Burstein et al., 1997). This value is consistent with the 0.5 - 4.5 keV flux measured here. Photometric data on this galaxy are reported in Table 3.
In Figure 4 we show the Spectral Energy Distribution (SED) of UGC12237: the filled dots represent the photometric data from radio to optical wavelengths while the tick solid line represents the $`110`$ keV best fit spectral model reported in Table 2. Also shown in Figure 5 is the mean SED of a sample of Seyfert 2 and a sample of non-AGN Spiral galaxies as determined by Schmitt et al., 1997. The SEDs have been normalized at $``$ 2 keV.
It is clear from Figure 4 that the mean SED of non-AGN Spiral galaxies is at variance with the SED of AXJ2254+1146, while the SED of Seyfert 2 galaxies seem to be more consistent with the data. The largest discrepancy between the Seyfert 2 SED and the SED of AXJ2254+1146 is in the optical domain, where only the total magnitude (nucleus + host galaxy) is available; this discrepancy is expected since the SEDs reported in Schmitt et al., 1997 were based on the optical nuclear flux. In the far infrared regime Schmitt et al., 1997 use the IRAS fluxes as we have done for AXJ2254+1146.
## 5 Optical Spectroscopy
While the paper was in the refereeing process an optical spectrum was obtained with VLT/FORS in long slit mode as a backup program during a period of poor seeing. The exposure time was 5 minutes long and the slit ($`1.6\mathrm{}`$ wide) was set perpendicular to the galaxy major axis. The Grism 150I was used providing a resolution $`\mathrm{\Delta }\lambda 9\AA `$ with little sensitivity below $`4000\AA `$. The extracted spectrum (Fig.6) shows H<sub>β</sub>, \[OIII\], \[OI\], H<sub>α</sub>, \[NII\] and \[SII\] narrow emission lines as well as absorption lines originating in the host galaxy, at redshift $`z_{\mathrm{opt}}=0.029`$. The weak H<sub>β</sub> line is likely partially obscured by blending with the stellar H<sub>β</sub> absorption line. We estimate a flux ratio$`[\mathrm{O}\mathrm{III}]\lambda 5007/\mathrm{H}_\beta >3`$. At the above stated spectral resolution, we also estimate a FWHM for all the observed emission lines less than $`800`$ km/s. Together with the presence of a weak \[OI\]$`\lambda 6300`$ line, these line features and ratios lend support to the identification of AXJ2254+1146 as a Seyfert 2, as opposed to a LINER or starburst galaxy.
## 6 Summary and Conclusion
In this paper we have reported the X-ray discovery of a Type 2 AGN - AXJ2254+1146 - which is part of the ASCA Hard Serendipitous Survey (Della Ceca et al., 1999a, 2000); this object must be added to the scanty number of hard X-ray selected type 2 AGN reported in literature (see e.g. Akiyama et al., 1998b and reference therein).
The X-ray spectrum is best described by a model consisting of an unresolved Gaussian line at $`6.43\pm 0.1`$ keV plus the so called “leaky-absorber” continua. The best fit values of the absorbing column density ($`N_H`$ = $`1.85_{1.47}^{2.24}\times 10^{23}`$ cm<sup>-2</sup>), of the line equivalent width ($`0.6_{0.36}^{0.84}`$ keV) and of the scattering fraction ($`0.7_{0.1}^{1.4}\%`$) clearly reveal its Type 2 nature.
A fairly bright ($`m_{B^o}=14.26`$) spiral galaxy (UGC 12237) is present inside the ASCA error circle: this is the most probable optical counterpart of AXJ2254+1146. Subsequent optical spectroscopy has confirmed that UGC 12237 is a Seyfert 2 at z=0.029.
Which are the most important implications of this finding? As with many discovery of this kind we are probably seeing the “tip of the iceberg”. As soon as the new generation of X-ray telescopes (e.g. Chandra, XMM) is fully operative, we predict that it will become common to investigate and classify X-ray selected objects by using X-ray data alone (see also Rosati et al., 1995 for a classification based on the source extent). AXJ2254+1146 could be indeed one of the brightest examples of a population of very hard X-ray sources that is showing up at faint fluxes (cf. Della Ceca et al., 1999a,b). The current optical identification status of the ASCA HSS prevents us from speculating any further and discussing constraints on the synthesis models for the CXB. We only note here that in the ASCA HSS sample there are about 8 more sources with broad band spectral properties (as derived from the hardness ratios) similar to those of AXJ2254+1146. One of these 8 sources is NGC 6552, a well known and studied Seyfert 2 galaxy (Reynolds et al., 1994). Finally, we recall that Fabian and Iwasawa (1999) have recently pointed out that if absorbed AGNs are the major contributors to the CXB above 2 keV then about 85% of the accretion power in the Universe is absorbed. In AXJ2254+1146 about 99% (65%) of the intrinsic flux in the 0.5-2.0 (2 -10) keV energy range is absorbed !
###### Acknowledgements.
We are grateful to the referee, M. Cappi, and to A. Wolter for the careful reading of the manuscript and for useful comments. V.B. acknowledges financial support from the Osservatorio Astronomico di Brera. This work received partial financial support from the Italian Ministry for University and Research (MURST) under grant Cofin98-02-32. This research has made use of the NASA/IPAC extragalactic database (NED), which is operated by the Jet Propulsion Laboratory, Caltech, under contract with the National Aeronautics and Space Administration. We thank all the members of the ASCA team who operate the satellite and maintain the software data analysis and the archive. |
no-problem/0001/hep-ph0001230.html | ar5iv | text | # Bremsstrahlung from an Equilibrating Quark-Gluon Plasma *footnote **footnote *Supported by BMBF, GSI Darmstadt, DFG, and Humboldt foundation
## I Introduction
In ultrarelativistic heavy ion collisions photons are emitted during the entire life time of the fireball. Since their mean free path is large compared to the size of the fireball , energetic photons escape from it without any interaction. Therefore photons with large transverse momenta probe the early, hot stage of the collision and may serve as a direct signature for the quark-gluon plasma (QGP) formation at RHIC and LHC . In order to use photons as a probe for the QGP, we have to predict the thermal spectrum from the QGP as well as from the hadronic phase. For this purpose we have to convolute the photon production rates from both phases with the space-time evolution of the fireball. Here we want to focus on a QGP phase possibly created at RHIC and LHC energies. Recent calculations of the photon production rate from the QGP, taking into account bremsstrahlung, resulted in a significantly larger rate compared to previous investigations . These rates have been implemented in a hydrodynamical calculation assuming a local thermal and chemical equilibrium . However, at RHIC and LHC the QGP phase is not expected to be in chemical equilibrium, i.e., the phase space of quarks and gluons will be undersaturated probably . Using non-equilibrium photon rates without bremsstrahlung and generalizing the hydrodynamical calculations to chemical non-equilibrium, the photon spectrum from a chemically non-equilibrated QGP has been estimated . Due to the undersaturation in particular of the quark component a suppression of the photon yield has been observed . In order to make up-to-date predictions for the thermal photon spectrum of a QGP at RHIC and LHC, the both competing effects of the enhanced photon rate considering bremsstrahlung and the reduction of the photon yield due to the chemical non-equilibrium should be combined. This is the aim of the present investigation.
In the next section we present an estimate of the non-equilibrium photon rates taking into account bremsstrahlung. In Sec.III we summarize briefly our hydrodynamical description, before we will discuss our results in Sec.IV and draw our conclusions in Sec.V.
## II PHOTON PRODUCTION FROM QUARK-GLUON PLASMA
The production rate of energetic real photons in an equilibrated QGP has been calculated using the Hard Thermal Loop (HTL) resummation technique . Here only the lowest order contributions, namely quark-antiquark annihilation and Compton scattering (corresponding to a one-loop polarization tensor containing a HTL resummed quark propagator in the case of a soft quark momentum), has been taken into account. An estimate for these contributions in chemical non-equilibrium has been given by computing these rates from the tree level scattering diagrams and using off-equilibrium distribution functions for the external quarks and gluons . For this purpose the equilibrium distributions have been multiplied by space-time dependent fugacity factors $`\lambda _{g,q}`$ describing the deviation from equilibrium according to Ref.. Also the infrared cutoff in these rates given by the effective in-medium quark mass has been generalized to non-equilibrium. Neglecting Pauli blocking and Bose enhancement effects, which is justified since the average momenta of the partons is $`3T`$ in the baryon free QGP, the following result for the rates due to Compton scattering and quark-antiquark annihilation has been obtained ,
$$E\frac{\mathrm{d}N_\gamma }{\mathrm{d}^4x\mathrm{d}^3p}|_{\mathrm{𝖼𝗈𝗆}.}^{\mathrm{𝖼𝖾𝗊}.}=\frac{2\alpha \alpha _s}{\pi ^4}\lambda _q\lambda _gT^2\left(\underset{f}{}e_f^2\right)e^{E/T}\left[\mathrm{ln}\left(\frac{4ET}{\kappa _c^2}\right)+\frac{1}{2}C\right],$$
(1)
and
$$E\frac{\mathrm{d}N_\gamma }{\mathrm{d}^4x\mathrm{d}^3p}|_{\mathrm{𝖺𝗇𝗇}.}^{\mathrm{𝖼𝖾𝗊}.}=\frac{2\alpha \alpha _s}{\pi ^4}\lambda _q\lambda _{\overline{q}}T^2\left(\underset{f}{}e_f^2\right)e^{E/T}\left[\mathrm{ln}\left(\frac{4ET}{\kappa _c^2}\right)1C\right].$$
(2)
Here $`C=`$ 0.577216…, and $`\kappa _c=2m_q^2`$, where the thermal mass of the quarks in the non-equilibrated medium is given as
$$m_q^2=\frac{4\pi \alpha _s}{9}\left(\lambda _g+\frac{\lambda _q}{2}\right)T^2.$$
(3)
Alternatively non-equilibrium rates can be calculated by generalizing the HTL technique to quasistatic chemical non-equilibrium situations . In the case of the photon production rate this more consistent, but also more elaborate, method leads to quantitatively very similar results as the simplification used above .
Recently Aurenche et al. have shown that there are additional contributions to the thermal photon rate at the same order $`\alpha _s`$ coming from two-loop diagrams within the HTL method. The corresponding physical processes are bremsstrahlung and quark-antiquark annihilation, where the quark (or antiquark) scatters of a parton from the QGP. The latter process becomes important especially for energetic photons as the rate is proportional to $`ET`$, whereas the rates from bremsstrahlung, annihilation without rescattering, and Compton scattering are proportional to $`T^2`$. Although the Compton and annihilation contributions are enhanced in the weak coupling limit by a logarithmic factor $`\mathrm{ln}(1/\alpha _s)`$ compared to the bremsstrahlung processes, the latter ones dominate for realistic values of the coupling constant by a factor of 5 or more . Of course, the extrapolation of these perturbatively calculated rates to realistic values of the coupling constant, for which we will use $`\alpha _s=0.3`$ in the following, is questionable. However, in view of the lack of other methods for calculating dynamical quantities such as production rates in thermal field theory so far, we assume that these results can be used as a rough estimate at least in the case of high energy photons ($`ET`$), in which we are interested.
For estimating the bremsstrahlung contributions to the non-equilibrium rate we adopt the same approximations as in (1) and (2). This means that we start from the scattering diagrams corresponding to these processes and ascribe non-equilibrium distributions to the external partons in the entrance channels. Restricting ourselves to $`t`$-channel diagrams, which dominate because the exchanged gluon is soft , and considering the different spin, color, and flavor ($`N_f=2`$) factors, we find for the bremsstrahlung process
$$E\frac{\mathrm{d}N_\gamma }{\mathrm{d}^4x\mathrm{d}^3p}|_{\mathrm{𝖻𝗋𝖾𝗆}.}^{\mathrm{𝖼𝖾𝗊}.}=\frac{2N_cC_F}{\pi ^5}\alpha \alpha _s\left(\frac{4}{7}\lambda _q^2+\frac{3}{7}\lambda _g\lambda _q\right)\left(\underset{f}{}e_f^2\right)T^2e^{E/T}(J_TJ_L)\mathrm{ln}(2),$$
(4)
and for the annihilation with scattering process
$$E\frac{\mathrm{d}N_\gamma }{\mathrm{d}^4x\mathrm{d}^3p}|_{\mathrm{𝖺𝗐𝗌}}^{\mathrm{𝖼𝖾𝗊}.}=\frac{2N_cC_F}{3\pi ^5}\alpha \alpha _s\left(\frac{2}{5}\lambda _q^3+\frac{3}{5}\lambda _g\lambda _q^2\right)\left(\underset{f}{}e_f^2\right)ETe^{E/T}(J_TJ_L).$$
(5)
Here the constants $`J_T4.45`$ and $`J_L4.26`$ for two flavors are the same as in equilibrium, since their dependence on the fugacities can be neglected. For they are functions of $`m_q/m_g`$ only and the square of the effective quark mass $`m_q^2`$, Eq.(3), and of the effective gluon mass $`m_g^2`$ are proportional to the gluon fugacity, if the much smaller quark fugacity is neglected. It should be noted that in Eq.(5) the combination of the different parton fugacities appears in cubic power, instead of quadratically as for the other processes, due to the fact that this particular process involves three particles in the entrance channel.
## III HYDRODYNAMIC EXPANSION AND CHEMICAL EQUILIBRATION
To evaluate the thermal photon spectrum one has to convolute these emission rates with the space time history of the expanding fireball out of chemical equilibrium. We do not want to repeat here the calculations of the chemical evolution of the QGP by means of rate equations and its implementation in a hydrodynamical code, but refer the reader to Ref.. In this hydrodynamical model, which we are adopting here, the transverse expansion of the QGP has been taken into account.
The essential input needed are the initial conditions for the fugacities and temperature at the time at which local thermal equilibrium is achieved. In order to take into account the uncertainties in these initial conditions we consider various possibilities predicted by two microscopic models for ultrarelativistic heavy ion collisions, namely SSPC and HIJING . The initial conditions we are using here are tabulated in Tab.I. The ones denoted by (I) are the original HIJING predictions, while the set (II) is obtained by multiplying the original initial fugacities by a factor of 4 and by decreasing the initial temperature somewhat, in order to take into account possible uncertainties in the model such as the neglect of soft parton production from the color field .
In addition to the different initial conditions, we also need a nuclear profile function for the fireball to solve the hydrodynamical equations numerically as matter distributions with sharp edges are difficult to use in numerical simulations . Another aim of our investigation is to study the dependence of the photon spectra on different choices of the profile function. For this purpose we consider in the following two profile functions. The first one is a Fermi-like profile function
$$T_A(r)=\frac{1}{e^{\left(rR_T\right)/\delta }+1},$$
(6)
where $`r`$ is the transverse coordinate, $`R_T`$ is the transverse radius of the nucleus and $`\delta `$ is the surface thickness, which has also been used in Ref.. The second one is the wounded-nucleon profile given by
$$T_A(r)=\frac{3}{2}\sqrt{1\frac{r^2}{R_T^2}}.$$
(7)
Since the initial conditions shown in Tab.I describe averages over the transverse cross section, $`\pi R_T^2`$, of the colliding nuclei in a central collision, one needs to modify them for different nuclear profiles. This implies that the system will have higher initial energy density and fugacities using the wounded-nucleon profiles, but also the preexistence of a density gradient over the whole transverse area resulting in a faster onset of the transverse expansion throughout the plasma volume .
## IV RESULTS
In Fig.1 we present the thermal photon production from the quark phase as a function of its transverse momentum at RHIC energies with SSPC initial conditions and the Fermi-like profile function both in the equilibrium and equilibrating scenario, where we assumed the same initial energy densities in both cases given in Tab.I. The initial temperature in the equilibrium case corresponding to $`ϵ_i=61.4`$ GeV/fm<sup>3</sup> is $`T_i^{eq}=0.429`$ GeV. Therefore there is an interplay in the photon production between the enhanced temperature and the reduced fugacities in the non-equilibrium. As a result the total photon yield is suppressed at $`p_T=1`$ GeV by a factor of five but almost identical at $`p_T=5`$ GeV in the equilibrating compared to the equilibrated scenario. Keeping, however, the initial temperature ($`T_i=0.668`$ GeV) instead of the initial energy density fixed leads to a strong reduction of the non-equilibrium compared to the equilibrium rate by one (low $`p_T`$) to three (high $`p_T`$) orders of magnitude. The strong suppression at high $`p_T`$ is due to the small fugacities in the early hot stage, from which the energetic photons are emitted mainly. Keeping, however, the initial energy density fixed this effect is counterbalanced by the higher initial temperature in the non-equilibrated case.
The contribution from different physical processes discussed in the preceding section can easily be identified from Fig.1 itself. As expected from the equilibrium static rates, the photon spectrum from the equilibrated QGP (see upper panel of Fig.1) is dominated by the annihilation with scattering contribution. In the case of a chemically equilibrating plasma (lower panel of Fig.1) the most dominant contribution comes from the usual bremsstrahlung processes, even though the aws static equilibrium rate was higher by almost an order of magnitude. The reason is obvious as the rate for the aws processes for an equilibrating plasma involves the parton fugacities to cubic power (Eq.(5)) whereas the other processes are only quadratic in the fugacities (Eqs.(1,2,4)), and the fugacities at RHIC are small for the entire life time of the plasma. At higher $`p_T`$ the Compton and annihilation one-loop contributions even exceed clearly the aws contribution as energetic photons have their origin in the early hot stage of the plasma, where the fugacities are very small. It is also worthwhile to note that the SSPC model predicts a gluon dominated plasma implying larger contributions from processes involving gluons in the entrance channel.
The total thermal photon yield at RHIC dominates over the prompt photon contributions for $`p_T4.5`$ GeV. However, if only the Compton and annihilation contributions are considered the prompt photons overshadow the thermal photon yield already beyond $`p_T3`$ GeV. Since the life time of the plasma at RHIC is small, the photon yield is not affected by flow .
The corresponding results for LHC energies are shown in Fig.2. Here the non-equilibrium rate is suppressed compared to the equilibrium at the same initial energy density ($`T_i^{eq}=0.695`$ GeV) by about a factor of three for all photon momenta and by one to two orders of magnitude in the case of the same initial temperature. The photon yields from different processes as well as the total yield are much larger than at RHIC as the life time of the plasma, likely to be created at LHC energy, is expected to be much larger and the initial temperature and fugacities are higher. The upper panel shows the equilibrated scenario, whereas the lower panel applies to the equilibrating one. Now the usual bremsstrahlung and the aws contribution originating from two-loop calculations are similar in the non-equilibrium case, whereas the one-loop contribution is smaller at all $`p_T`$. The reduced suppression of the aws contribution compared to RHIC can be understood in the following way. The rate depends linearly on the temperature and cubicly on the fugacities. In addition the aws is also proportional to the energy of the photon. Now with the passage of time the system expands, the temperature falls and the chemical reactions pushes the system towards equilibrium causing the fugacities to increase, though at later time they decrease significantly caused by transverse expansion. Hence, there is a competition between the space-time evolution of the temperature and the fugacities though they are not exactly counterbalanced. Rather, the interplay of these two quantities along with the photon energy dependence causes a reduced suppression of the aws contribution compared to RHIC.
The prompt photon productions due to lowest order QCD (Born) and inclusive photons (background photons fragmented off high-$`p_T`$ quark jets) will remain buried under the thermal photon yield for all $`p_T`$-values considered here.
Comparing the photon spectra obtained with the wounded-nucleon profile we observe that the photon yield following from the Fermi-like profile is a little bit higher. This enhancement, which is always less than a factor of two, is caused by the fact that a Fermi-like profile associated with a slower cooling implies a higher temperature.
In the following we present our results with HIJING initial conditions. If we use the original prediction, $`i.e.`$, HIJING-I, then the life time of the QGP phase is very small (less than 2 fm/c) for RHIC and 7.5 fm/c for LHC . Also the matter will be very dilute due to the very small initial values of the fugacities at RHIC. Furthermore, for LHC only the fluid beyond $`4`$ fm from the centre participates in flow. Now for HIJING-II initial condition, the life time of the plasma increases substantially both for RHIC ($``$ 4 fm/c) and LHC ($``$ 12 fm/c). Also the initial fugacities differ by one order of magnitude for RHIC, and are reasonably higher for LHC. For LHC the entire fluid will participate in the transverse flow since the life time is large and initially the system will approach equilibrium but then be driven away from it as soon as the large transverse velocity gradient develops. Fig.3 exhibits the photon production from the equilibrating plasma with HIJING-I initial conditions and a Fermi-like nuclear profile both at RHIC (upper panel) and LHC (lower panel) energies. Because of the low values of the fugacities during the entire life time of the plasma , the contribution of the different processes as well as the total yield are strongly suppressed compared to the case using SSPC initial conditions. In particular there is a huge suppression of the aws contribution, depending cubicly on the fugacities, compared to the other processes at RHIC. Also at LHC the strong suppression of the aws contribution at high-$`p_T`$ is due to the very small initial fugacities as high-$`p_T`$ photons are mostly emitted from the initial stages.
Finally the results for an equilibrating plasma with HIJING-II initial conditions are given in Fig.4. Since the initial fugacities are an order of magnitude higher for RHIC (upper panel), the contributions from each process, particularly from the aws, are enhanced substantially compared to HIJING-I. Since the initial fugacities are higher compared to SSPC but the initial temperature is smaller, there is a counterbalance of these two effects, and the total photon production for LHC (lower panel) is almost the same as that using SSPC initial conditions (lower panel of Fig.2).
## V SUMMARY AND CONCLUSION
We have considered the photon production from a chemically non-equilibrated QGP at RHIC and LHC energies. As a new aspect we have included bremsstrahlung processes. The photon production rate due to these processes has been calculated recently using the HTL resummation technique by Aurenche et al. in the case of a fully equilibrated QGP. Instead of repeating this calculation in the non-equilibrium plasma, we estimated the non-equilibrium photon production rate in the following way: we simply assigned fugacity factors, describing the deviation of the parton densities from chemical equilibrium, to the partons in the entrance channels of the matrix elements corresponding to the different processes (annihilation, Compton scattering, bremsstrahlung, annihilation with scattering). In the case of the first two processes this approach has to be shown to give quantitatively very similar results as the more elaborate HTL method extended to chemical non-equilibrium . Moreover, regarding the uncertainties, such as the validity of perturbation theory and the initial conditions in the computation of the photon spectrum, this simplification is justified.
The photon spectra have been calculated from these rates by using a hydrodynamical calculation, describing the space-time evolution of the QGP phase of the fireball, where we have taken into account the transverse expansion of the fireball. The initial conditions for the temperature and the fugacities have been taken from microscopic models (SSPC, HIJING).
We found that the photon yield is reduced by a factor one to five depending on the collision energy and photon momentum compared to the fully equilibrated plasma, assuming the same initial energy density taken from the SSPC model. This moderate suppression is the result of the interplay between the small fugacities and the increased temperature in the non-equilibrium compared to the equilibrium scenario. In general the suppression is more pronounced for energetic photons coming from the highly dilute early stage.
Assuming the same initial temperature the photon yield from the non-equilibrium QGP is reduced by one to three orders of magnitude. The reason for this large difference is the reduction of the initial temperature for the equilibrated scenario assuming a fixed initial energy density. In ultrarelativistic heavy ion collisions assuming free flow the initial energy density is determined from the measured particle multiplicity, which is related to the collision energy. Therefore the assumption of a fixed initial energy density instead of temperature appears to be more physical.
Although the total photon yield does not differ much in the both scenarios, using the same initial energy density, its composition by the individual processes is completely different. Whereas the contribution from the annihilation with scattering process dominates the photon production over the entire momentum range in equilibrium, in a chemically non-equilibrated plasma it is suppressed at RHIC energies compared to the bremsstrahlung contribution, which is now dominating, and even to the one-loop (annihilation, Compton scattering) contributions, which is the smallest in equilibrium. The reason for this behavior is that the annihilation with scattering process depends cubicly on the fugacities, which are very small at RHIC, whereas the other processes only quadratically. At LHC energies, on the other hand, where the fugacities are significantly larger, the annihilation with scattering and the bremsstrahlung contributions are of the same order and exceed the one-loop contributions clearly.
We have also investigated the dependence of our results on different initial conditions and the choice of the profile of the fireball. The photon yield can vary by more than an order of magnitude, depending on the choice of the initial conditions, while the dependence on the profile function is weak.
Although in this investigation all important effects (bremsstrahlung, chemical non-equilibrium, transverse expansion), known so far, are incorporated, there are still uncertainties and open questions. For example, the applicability of perturbative methods and the role of higher order contributions is unknown. Also a consistent treatment of the chemical equilibration, the space-time evolution of the fireball, and the non-equilibrium photon rates within the framework of the kinetic theory would be desirable.
ACKNOWLEDGMENTS
The authors are grateful to H. Zaraket for helpful discussions concerning the extension of the bremsstrahlung rate to non-equilibrium. |
no-problem/0001/cond-mat0001445.html | ar5iv | text | # Clustering data by inhomogeneous chaotic map lattices
## Ackowledgements
The authors thank G. Nardulli, for valuable discussions, and P. Blonda, G. Pasquariello, G. Satalino (IESI-CNR) for providing the LANDSAT data set.
FIG. Captions
FIG. 1: Scaling function for the mutual information for $`D=2`$ (solid line), $`D=3`$ (dashed line) and $`D=4`$ (dotted line).
FIG. 2: Artificial data set consisting of two dense regions of $`400`$ and $`1900`$ points, in a dilute background of $`200`$ points.
FIG. 3: Frequency distribution of (a) distances between neighboring points in Fig. 2 and (b) mutual information of neighboring points.
FIG. 4: Size of the three biggest clusters obtained by our algorithm, on the data-set in Fig.2, as a function of the threshold $`\theta `$.
FIG. 5(color): First two principal components of the LANDSAT data-set (see the text). Comparing the variances along the six principal axis, it turns out that also the third and fourth principal components are relevant to this data-set.
FIG. 6: Hierarchical structure of the LANDSAT data-set as it has been found by our algorithm; the $`\theta `$ values at which the clusters split can be read on the axis at the bottom. This results have been obtained using $`k=20`$, however values in the range $`1050`$ give similar results. |
no-problem/0001/cond-mat0001029.html | ar5iv | text | # Doppler shift on local density of states and local impurity scattering in the vortex state
## I Introduction
The theoretical predictions of Volovik and Kopnin and Volovik that the low temperature specific heat in a $`d`$-wave superconductor would vary as the square root of the applied magnetic field $`(\sqrt{H})`$ was verified in several experiments. To understand the experimental data it is assumed that the extended quasiparticle states outside the vortex cores can be treated semiclassically taking account only of the Doppler shift due to the circulating supercurrents. To obtain the specific heat, which is a bulk property of the system, it is necessary to carry out a spacial average over the vortex unit cell of the local density of states (LDOS) which varies as a function of distance from the vortex core center. The low temperature limit of the specific heat depends on the zero frequency limit of this averaged density of quasiparticle states. This quantity is not only affected by the magnetic field H but can also depend on impurity content, particularly in the resonant scattering limit. In this limit the theory of a $`d`$-wave superconductor without magnetic field predicts a finite, constant value for its zero frequency limit which is proportional to the self consistent effective scattering rate and this leads to the well known universal limit for the transport.
When the magnetic field is oriented in the copper oxide plane rather than perpendicular to it new anisotropy effects in the specific heat are predicted. In addition transport properties are also affected by the Doppler shift of a vortex state. The thermal conductivity has been particularly widely studied.
The semiclassical Doppler shift of the electron due to the circulating supercurrents outside the vortex core should also affect the frequency dependence of the LDOS which can be measured in scanning tunneling microscope (STM) experiments. In view of the above successes in our understanding of the thermal and transport properties of the copper oxides it is of interest to understand the signature of the Doppler shift on the LDOS and the effect of impurities on it. Very recently Franz and Tešanović considered the LDOS problem and compared an average over vortex winding angles of the semi classical LDOS with equivalent results obtained from complete solutions of the BdG equations for a single vortex. Their main conclusion was that the semiclassical approach does indeed provide a good approximation for this quantity.
For no magnetic field $`(H=0)`$ there is a single van Hove singularity in the density of states at energy $`\omega `$ equal to the gap amplitude which does not shift with position $`𝐫`$ (homogeneous case). For finite $`H`$ new van Hove ridges are predicted. The actual number of such ridges and their topology depends on the winding angel $`\beta `$ and simple formulas can be obtained from which their contour in $`(\omega ,𝐫)`$-space can be traced. The effect of impurities on the topology of these ridges is considered as is the effect of orthorhombicity. Impurities are treated in both Born and unitary limit, and the influence of the circulating supercurrents on the local effective impurity scattering is also considered. This topic was recently discussed by Barash and Svidzinskii. Orthorhombicity, present in YBa<sub>2</sub>Cu<sub>3</sub>O<sub>6.95</sub> (YBCO) because of the existence of chains, is treated within a simple effective mass model for the in-plane electronic dispersion curves and a possible subdominant $`s`$-wave component is added to the dominant $`d`$-wave gap.
The tetragonal case is the subject of section II. Impurities are included. Generalization to the orthorhombic case is given in section III. Section IV deals with effective local impurity scattering rates. A short conclusion is found in section V.
## II Tetragonal case
In a semiclassical approximation the local Green’s function at position r is given in terms of the Doppler shift $`𝐯_F(𝐤)𝐪_s`$ by
$$G(𝐤,\omega _n;𝐫)=\frac{\left[i\omega _n𝐯_F(𝐤)𝐪_s\right]\tau _0+\epsilon _𝐤\tau _3+\mathrm{\Delta }_𝐤\tau _1}{\left[\omega _n+i𝐯_F(𝐤)𝐪_s\right]^2+\epsilon _𝐤^2+\mathrm{\Delta }_𝐤^2}$$
(1)
where $`\tau _{1,2,3}`$ are the $`2\times 2`$ Pauli matrices, $`i\omega _n`$ the Matsubara frequencies $`i(2n+1)\pi T,n=0,\pm 1,\pm 2,\mathrm{}`$ and $`T`$ is the temperature. The electronic dispersion in momentum space k is $`\epsilon _𝐤`$ and the gap in the two-dimensional copper oxide Brillouin zone is $`\mathrm{\Delta }_𝐤`$. The electronic Fermi velocity is $`𝐯_F(𝐤)`$ and $`𝐪_s`$ is the momentum associated with the superfluid flow about the vortex. In the magnetic field region $`H_{c1}<|𝐇|H_{c2}`$ with $`H_{c1,2}`$ the lower and upper critical field respectively, and H the external magnetic field taken to be perpendicular to the CuO<sub>2</sub>-planes, the intervortex distance $`R=\frac{1}{a}\sqrt{\frac{\varphi _0}{\pi |𝐇|}}`$ where $`a`$ is a geometrical factor of order one which is associated with the vortex arrangement, and $`\varphi _0`$ is the fundamental quantum of flux. In what follows, the position r will be measured relative to the center of the vortex core. We use polar coordinates $`(r,\beta )`$ with $`\beta `$ the vortex winding angle. The dimensionless variable $`\rho =r/R`$ takes on the value 1 at the boundary of the vortex unit cell. It is expected that $`R\xi _0`$ ($`\xi _0`$ is the coherence length) and we will be interested only in values of $`r>\xi _0`$.
Assuming a circular velocity field for the supercurrents around a single vortex, $`𝐯_s=\frac{\mathrm{}\widehat{\beta }}{2mr}`$ with $`\widehat{\beta }`$ a unit vector along the current, the Doppler shift
$$𝐯_F(𝐤)𝐪_s=\frac{E_H}{\rho }\mathrm{sin}(\varphi \beta ),$$
(2)
where $`E_H`$ is the magnetic energy scale that enters our problem and is equal to $`\frac{a}{2}v_F\sqrt{\frac{\pi H}{\varphi _0}}\nu \mathrm{\Delta }_0`$ where the last identity measures $`E_H`$ in units of the gap amplitude $`\mathrm{\Delta }_0`$. In Eq. (2) $`\varphi `$ is the polar angle associated with momentum k on the Fermi surface and the $`d`$-wave gap $`\mathrm{\Delta }_𝐤=\mathrm{\Delta }_0\mathrm{cos}2\varphi \mathrm{\Delta }_d(\varphi )`$ in the same model.
The local quasiparticle density of states at relative position $`\rho `$ and energy $`\omega `$, $`N_L(\rho ,\omega )`$, follows directly from the analytic continuation of $`G`$ defined in Eq. (1) to real frequencies $`i\omega _n\omega +i0^+`$ with
$$\frac{N_L(\rho ,\omega )}{N_0}=\underset{0}{\overset{2\pi }{}}\frac{d\varphi }{2\pi }\mathrm{}\mathrm{e}\left\{\frac{\left|\omega 𝐯_F(𝐤)𝐪_s\right|}{\sqrt{\left[\omega 𝐯_F(𝐤)𝐪_s\right]^2\mathrm{\Delta }_d^2(\varphi )}}\right\},$$
(3)
where $`N_0`$ is the normal state density of states. Note that the Doppler shift of Eq. (2) which enters (3) depends on the magnetic field $`|𝐇|=H`$ through $`E_H\sqrt{H}`$ and on the position of the STM tip away from the vortex core given by $`\rho `$ and $`\beta `$. We will be interested only in the region where r is outside the vortex core of size $`\xi _0`$. This is mandated by the fact that our approach does not treat in detail the core interior but instead includes only the supercurrents outside the core through the superfluid velocity field $`𝐯_s=\frac{\mathrm{}\widehat{\beta }}{2mr}`$ which itself is an approximation. Nevertheless, it is the aim of this work to understand how the presence of $`𝐯_s`$ modifies the LDOS in the region between the vortex cores.
The integral in Eq. (3) is easily evaluated. We present a first set of results which include some impurity scattering in Born approximation. This smooths out the logarithmic singularity at the gap amplitude predicted for pure $`d`$-wave with $`H=0`$. To include static impurity scattering it is simply necessary to exchange the frequency $`\omega `$ and the gap $`\mathrm{\Delta }_d(\varphi )`$ in Eq. (3) by their renormalized values denoted by a tilde, i.e.: $`\stackrel{~}{\omega }`$ and $`\stackrel{~}{\mathrm{\Delta }}(\varphi )`$. The renormalized frequency
$$\stackrel{~}{\omega }(\rho ,\beta ,\omega )=\omega +i\pi t^+\mathrm{\Omega }(\rho ,\beta ,\omega )$$
(4)
in Born approximation with $`t^+`$ the impurity scattering rate in the normal state, and
$$\stackrel{~}{\omega }(\rho ,\beta ,\omega )=\omega +i\pi \mathrm{\Gamma }^+\frac{\mathrm{\Omega }(\rho ,\beta ,\omega )}{\mathrm{\Omega }^2(\rho ,\beta ,\omega )+D^2(\rho ,\beta ,\omega )}$$
(5)
in the unitary limit with $`\mathrm{\Gamma }^+`$ replacing $`t^+`$ of Eq. (4). The renormalized gap $`\stackrel{~}{\mathrm{\Delta }}(\varphi )=\mathrm{\Delta }_0\mathrm{cos}2\varphi +i\stackrel{~}{\mathrm{\Delta }}_s(\rho ,\beta ,\omega )`$ in a $`(d+is)`$-wave symmetry with
$$\stackrel{~}{\mathrm{\Delta }}_s(\rho ,\beta ,\omega )=i\pi t^+D(\rho ,\beta ,\omega )$$
(6)
for Born scattering and
$$\stackrel{~}{\mathrm{\Delta }}_s(\rho ,\beta ,\omega )=i\pi \mathrm{\Gamma }^+\frac{D(\rho ,\beta ,\omega )}{\mathrm{\Omega }^2(\rho ,\beta ,\omega )+D^2(\rho ,\beta ,\omega )}$$
(7)
in the unitary limit. Here, the two functions $`\mathrm{\Omega }(\rho ,\beta ,\omega )`$ and $`D(\rho ,\beta ,\omega )`$ are
$$\mathrm{\Omega }(\rho ,\beta ,\omega )=\frac{\stackrel{~}{\omega }(\rho ,\beta ,\omega )𝐯_F(𝐩)𝐪_s}{\sqrt{\left[\stackrel{~}{\omega }(\rho ,\beta ,\omega )𝐯_F(𝐩)𝐪_s\right]^2\stackrel{~}{\mathrm{\Delta }}_s^2(\rho ,\beta ,\omega )\mathrm{\Delta }_d^2(\varphi )}}_\varphi $$
(8)
and
$$D(\rho ,\beta ,\omega )=\frac{\stackrel{~}{\mathrm{\Delta }}_s(\rho ,\beta ,\omega )}{\sqrt{\left[\stackrel{~}{\omega }(\rho ,\beta ,\omega )𝐯_F(𝐩)𝐪_s\right]^2\stackrel{~}{\mathrm{\Delta }}_s^2(\rho ,\beta ,\omega )\mathrm{\Delta }_d^2(\varphi )}}_\varphi ,$$
(9)
where $`\mathrm{}_\varphi `$ indicates an angular average over the angles $`\varphi `$ around the circular Fermi surface in the two-dimensional copper oxide Brillouin zone.
In the top frame of Fig. 1 we show our results for a $`d`$-wave superconductor (tetragonal symmetry) with a gap amplitude $`\mathrm{\Delta }_0=24`$meV. We take the magnetic field $`𝐇c`$-axis, i.e. H is perpendicular to the copper oxygen planes and the vortex winding angle $`\beta =0^{}`$. The vertical axis is labeled by $`N_L(\rho ,\omega )`$ which is given by
$$N_L(\rho ,\omega )=\mathrm{}\mathrm{e}\left\{\mathrm{\Omega }(\rho ,\beta ,\omega )\right\}_{\beta =const},$$
(10)
and features the LDOS as a function of $`\rho `$ and $`\omega `$. Estimates of the magnetic energy scale $`E_H`$ in YBCO have been considered by Vekhter et al. This quantity is not very well known but best estimates give between
$`E_H=2030[\mathrm{K}]\sqrt{H[\mathrm{Tesla}]}.`$
Therefore the magnetic energy scale has been set at 10% of the gap amplitude for illustrative purposes, i.e. $`\nu =0.1`$ in our notation and the static impurity content in Born approximation is $`t^+=0.1`$meV, enough to smear out the van Hove singularity but not enough to make them disappear. (This value of $`t^+`$ corresponds to a scattering rate of about $`6`$K. Recent experiments by Hosseini et al. seem to indicate an even smaller scattering rate.) Three prominent sets of ridges are seen for $`\rho `$ in the range 0.1 (at the vortex core, i.e. at $`r=\xi _0`$) and 1.0 (at the vortex cell boundary). The case with no magnetic field is shown as the bottom frame of Fig. 1 for comparison. Only one prominent van Hove ridge is seen in this second case and it is centered at $`\omega =\mathrm{\Delta }_0=24`$meV and, of course, there is no $`\rho `$ dependence since there is no vortex core and thus no spatial inhomogeneities in the problem. A main ridge around $`\omega =24`$meV is still seen in the top frame although as $`\rho `$ is decreased and the vortex core is approached at $`\rho =0.1`$ it has shifted in energy significantly. The other two ridges are due entirely to the Doppler shift provided by the supercurrents and can be understood as follows. In the pure case (no impurity scattering) the ridges in $`N_L(\rho ,\omega )`$ correspond to extrema in the zeros of the denominator of Eq. (3), i.e. they correspond to the extrema of the equation
$$\omega =𝐯_F(𝐤)𝐪_s\pm \mathrm{\Delta }(\varphi )=\mathrm{\Delta }_0\left[\frac{\nu }{\rho }\mathrm{sin}(\varphi \beta )\pm \mathrm{cos}2\varphi \right],$$
(11)
i.e. to solutions of equation
$$\frac{d\omega }{d\varphi }=0=\frac{\nu }{\rho }\mathrm{cos}(\varphi \beta )2\mathrm{sin}2\varphi ,$$
(12)
which have been written for a general value of vortex winding angle $`\beta `$. There are three solutions of this equation for the case considered in the top frame of Fig. 1 $`(\beta =0)`$. The critical values of frequency $`\omega _{ci}`$ with $`i=1,2,3`$ are shown as a function of distance from the vortex core $`\rho `$ in the top frame of Fig. 2. For $`\omega 0`$ the solutions occur at the critical angles $`\varphi _{c1,2}=\pm \pi /2`$ and $`\varphi _{c3}=\mathrm{sin}^1(\nu /4\rho )`$ respectively and the corresponding values of $`\omega _{ci}`$ are
$$\omega _{c1,2}=\mathrm{\Delta }_0(1\pm \nu /\rho ),$$
(13)
which are valid without restriction; in the limit $`\nu /\rho 0`$ we find
$$\omega _{c3}=\mathrm{\Delta }_0\left[1+\frac{1}{2}\left(\frac{\nu }{2\rho }\right)^2\right].$$
(14)
Numerical solutions can always be obtained and are needed except for a few simple cases. For $`\nu =0`$, no magnetic field, all $`\omega _{ci}=\mathrm{\Delta }_0`$ as expected. At the vortex cell boundary $`\rho =1`$, i.e. $`r=R`$, the three critical frequencies are in units of $`\mathrm{\Delta }_0`$ 1.1, 0.9, and 1.0013 respectively, and at the vortex core which corresponds to $`\rho =0.1`$ for $`\nu =0.1`$ these values are 2, 0, and 1.125. These expectations are bourn out in the top frame of Fig. 1. We note from Eq. (13) that the spacing of the two additional structures away from $`\omega =\mathrm{\Delta }_0`$ goes in the antinodal direction $`(\beta =0^{})`$ like
$`\omega _{c1,2}\sqrt{H},`$
i.e. is linear in $`\nu `$; the shift of the main van Hove ridge from $`\mathrm{\Delta }_0`$ due to the Doppler shift goes instead like
$`\omega _{c3}H,`$
i.e. is linear in the magnetic field in the limit $`\nu /\rho 0`$. Other directions, $`\beta 0^{}`$, will develop other $`H`$-dependencies of the spacing of the van Hove ridges which will have to be evaluated numerically in most cases.
Also note the $`\omega 0`$ limit of the LDOS. It goes to zero linearly in the $`H=0`$ case but, in contrast, it takes on a finite value when vortices are present. Its value depends strongly on position $`\rho `$ in the vortex unit cell and increases strongly with decreasing $`\rho `$. It is the average of the LDOS at $`\omega =0`$ over the vortex unit cell, i.e. the average over $`\rho `$ and winding angle $`\beta `$ that determines the low temperature bulk specific heat and this varies with the value of H. Franz and Tešanović discuss in detail how this region of small $`\omega `$ can be strongly modified by an in-plane angular dependence to the tunneling matrix element that comes into the relationship between STM measurements and the LDOS itself. These are only the same when the tunneling element is constant.
In the bottom frame of Fig. 2 we show again the same data as shown in the top frame of Fig. 1 but in a different representation which allows to see the ridges a little more clearly. What is presented here is $`N_L(\rho ,\omega )`$ vs. $`\omega `$ for fixed values of $`\rho `$, namely $`\rho =0.2`$ (solid line), $`\rho =0.3`$ (dashed line), $`\rho =0.6`$ (dotted line), and $`\rho =1.0`$ (dash-dotted line). The vortex core radius equal to the coherence length $`\xi _0`$ corresponds to $`\rho =0.1`$ and the inter vortex distance is $`\rho =1.0`$. Finally, a thin solid line indicates $`N_L(\omega )`$ for no magnetic field, $`H=\nu =0`$.
Furthermore, the data shows clearly the main van Hove singularity around, but not precisely at $`\omega \mathrm{\Delta }_0`$, with some variation with $`\rho `$ as detailed in the top frame of Fig. 2 (solid line, $`\omega _{c3}`$). This peak exists even when $`H=0`$ (light solid line) when there are no supercurrents and it is the usual $`d`$-wave gap peak. Note that here we have included some small amount of impurity scattering in Born approximation, namely $`t^+=0.1`$meV. The other secondary peaks above and below $`\omega 24`$meV are due to the supercurrents and are the semiclassical signature of the resulting Doppler shifts. (Similar ridges with somewhat different geometry appear in earlier work which makes use of the semiclassical Eilenberger approach. This approach should be better inside the core, a region we do not treat properly. Other related work in the $`s`$-wave case is the very recent paper by Eschrig et al.) As the STM tip is moved more towards the vortex core, the secondary peak at low energy becomes fairly prominent and also the $`\omega =0`$ limit becomes finite as noted before. The topology of the surface $`N_L(\rho ,\omega )`$ vs. $`\rho `$ and $`\omega `$ is affected by several external variables such as the magnitude of the magnetic field, the vortex winding angle, the type of impurities involved, i.e. Born or unitary scattering, and its strength as we now detail.
The ridge structure in the LDOS changes considerably as the position of the STM tip is changed. This is illustrated in the top frame of Fig. 3 where we show our result for a winding angle $`\beta =45^{}`$, i.e. in the nodal direction. There are striking differences with the top frame of Fig. 1. Only two ridges are now seen in contrast to three in the $`\beta =0^{}`$ case (the antinodal direction), and the main gap peak around $`\omega =24`$meV at the vortex cell boundary $`\rho =1`$ (at $`r=R`$) moves downward in energy as $`\rho `$ is reduced toward the vortex core. The $`\omega =0`$ value of $`N_L(\rho ,\omega )`$ near $`\rho =0.1`$, i.e. $`r=\xi _0`$, is also much smaller than in the $`\beta =0^{}`$ case.
In the bottom frame if Fig. 3 we show our solutions for the critical values of $`\omega `$ ($`\omega _{ci},i=1,2`$) which locate the ridges in the top frame. The critical angles $`\varphi _{ci},i=1,2`$ at which the extrema responsible for the ridges occur are also presented as functions of the distance from the vortex center $`\rho `$ in this frame. It is obvious that geometry strongly affects the topology of the $`N_L(\rho ,\omega )`$ surface.
Impurity scattering can also affect things in an important way. In Fig. 4 we show results for the LDOS $`N_L(\rho ,\omega )`$ as a function of the distance to the vortex core $`\rho `$ and energy $`\omega `$ for a magnetic energy $`\nu =0.1`$, $`𝐇c`$-axis, and a winding angle $`\beta =0^{}`$ (antinodal direction). This figure is to be compared directly with the top frame of Fig. 1 as the only difference between the two figures is that in Fig. 3 the unitary impurity limit is used, Eqs. (5) and (7), with $`\mathrm{\Gamma }^+=0.1`$meV instead of the Born limit, Eqs. (4) and (6), with $`t^+=0.1`$meV. The structure of the ridges is now more pronounced and the zero frequency limit of the LDOS near the vortex cell boundary is now larger.
## III Orthorhombicity
YBCO is orthorhombic because of the existence of chains along the $`b`$-axis in one of the three copper oxide two-dimensional planes per unit cell. This implies that the gap is in principle a mixture of $`s`$\- and $`d`$-wave symmetry which do belong to the same irreducible representation of the crystal lattice group, and therefore has the form
$$\mathrm{\Delta }(\varphi )=\mathrm{\Delta }_0\left(\mathrm{cos}2\varphi +s\right)$$
(15)
on the Fermi surface. Here $`s`$ is the subdominant $`s`$-wave component. A simple way to include band anisotropy is to introduce anisotropic effective masses with $`m_b<m_a`$ in an infinite band model. This gives an ellipsoidal Fermi surface which can always be mapped onto a circular one by scaling of momentum variables. At the same time the gap is transformed to
$$\mathrm{\Delta }(\varphi )=\mathrm{\Delta }_0\left(s+\frac{\alpha +\mathrm{cos}2\varphi }{1+\alpha \mathrm{cos}2\varphi }\right).$$
(16)
There are no additional changes and the calculations proceed as before with Eqs. (11) and (12). For the direction $`\beta =0^{}`$, they are replaced by:
$$\omega =\mathrm{\Delta }_0\left[\frac{\nu }{\rho }\frac{1+\alpha }{\sqrt{1\alpha }}\mathrm{sin}\varphi \pm \left(\frac{\alpha +\mathrm{cos}2\varphi }{1+\alpha \mathrm{cos}2\varphi }+s\right)\right],$$
(17)
and
$$\frac{d\omega }{d\varphi }=0=\frac{\nu }{\rho }\frac{1+\alpha }{\sqrt{1\alpha }}\mathrm{cos}\varphi \frac{2(1\alpha ^2)\mathrm{sin}2\varphi }{(1+\alpha \mathrm{cos}2\varphi )^2},$$
(18)
with $`\alpha `$ the effective mass anisotropy parameter defined as $`\alpha =(m_bm_a)/(m_a+m_b)`$. Again, for a general winding angle numerical solutions are needed although analytic ones are still possible in the limit $`\nu /\rho 0`$ in some cases. If we restrict ourselves to solutions of Eq. (16) which have at least some positive values for the critical frequency $`\omega _c`$ in the range $`\nu \rho 1`$ then we get for $`\beta =0^{}`$, and the critical angles $`\varphi _c=\pm \pi /2`$ the following three solutions of Eq. (17)
$$\frac{\omega _c}{\mathrm{\Delta }_0}=\{\begin{array}{cc}\frac{\nu }{\rho }\frac{1+\alpha }{\sqrt{1\alpha }}\pm (1s),\hfill & \varphi _c=\frac{\pi }{2}\hfill \\ \frac{\nu }{\rho }\frac{1+\alpha }{\sqrt{1\alpha }}+(1s),\hfill & \varphi _c=\frac{\pi }{2}\hfill \end{array}$$
(19)
without restrictions on the value of $`\nu /\rho `$, and another solution
$$\frac{\omega _c}{\mathrm{\Delta }_0}=1+s+\frac{1}{2}\left(\frac{\nu }{2\rho }\right)^2\frac{(1+\alpha )^3}{(1\alpha )^2},$$
(20)
which is valid only for $`\nu /\rho 0`$ and corresponds in that case to $`\varphi _c`$ near zero radians.
First we note the well known limit of $`\nu =0`$ (no magnetic field) in which case the van Hove singularity in the density of states is split by the existence of the subdominant $`s`$-wave component of the gap and there are two singularities in the density of states positioned at $`\omega _c=\mathrm{\Delta }_0(1\pm s)`$. This holds whatever the value of the effective mass anisotropy $`\alpha `$ might be. This mass anisotropy does not lead to a corresponding splitting of the van Hove singularities in the density of states. In a previous paper Schürrer et al. have considered the specific case of optimally doped YBCO and have suggested that reasonable model parameters are $`\alpha =0.4`$ to account for a factor of a little more than 2 in the effective conductivity between $`b`$\- and $`a`$-direction. Further consideration of finite temperature penetration depth data suggests $`s=0.25`$. This gives agreement between the model and the measured penetration depth data at low but finite temperatures $`T`$.
In the top frame of Fig. 5 we show results for $`N_L(\rho ,\omega )`$ for a case with $`\alpha =0.4`$, $`s=0.25`$, and $`\beta =0^{}`$, i.e. in the antinodal direction. The top frame is a three dimensional plot of $`N_L(\rho ,\omega )`$ vs. $`\rho `$ and $`\omega `$ and is to be compared with the top frame of Fig. 1 (tetragonal case). We note striking differences in the ridge structure. Now, the two ridges at $`\rho =1`$ and about $`\omega =\mathrm{\Delta }_0(1\pm s)`$ come together as the vortex core is approached and they eventually cross each other before diverging in the opposite direction. The contours for the critical frequencies, $`\omega _c`$, are shown in the bottom frame of Fig. 5. The two ridges that cross are the solutions of Eq. (19) for $`\varphi _c=\pi /2`$ and of Eq. (20). The solution of Eq. (19) for $`\varphi _c=\pi /2`$ and the plus sign gives the higher energy ridge shown in the top frame of Fig. 5 (and dash-dotted line in the bottom frame of this figure).
As in the tetragonal case (Fig. 4) the ridge topology in the LDOS changes considerably as the position of the STM tip is changed. We illustrate this in the top frame of Fig. 6 where we present results for a winding angle $`\beta =45^{}`$. Now all four ridges are seen in the entire range $`\nu \rho 1`$ and two cross overs can be observed. The contours of the critical frequencies $`\omega _c`$, are shown in the bottom frame of Fig. 6. A comparison of the top frame of this figure with the top frame of Fig. 4 (tetragonal case, $`\beta =45^{}`$) makes the splitting of the van Hove singularity as a result of the subdominant $`s`$-wave component of the gap particularly transparent.
## IV Local impurity scattering rate
Finally, it is of interest to discuss briefly the local impurity scattering rates which are modified by the presence of supercurrents. Similar work has already appeared in a paper by Barash and Svidzinskii. In Born approximation $`\mathrm{}\mathrm{m}\stackrel{~}{\omega }(\rho ,\beta ,\omega )`$ given by Eq. (4) is given by
$$\mathrm{}\mathrm{m}\stackrel{~}{\omega }(\rho ,\beta ,\omega )=\pi t^+\mathrm{}\mathrm{e}\mathrm{\Omega }(\rho ,\beta ,\omega ),$$
(21)
where $`\mathrm{}\mathrm{e}\mathrm{\Omega }(\rho ,\beta ,\omega )`$ is just the LDOS of Eq. (3) modified by the impurity scattering and plotted already in the top frame of Fig. 1 for the antinodal direction and in Fig. 3 for the nodal direction. It is clear then that the Doppler shift accounting for the supercurrents significantly affects the local scattering rates in the same way as they modify the LDOS. The changes due to the supercurrents are very different when unitary scattering is considered. In this case, we find from Eq. (5),
$$\mathrm{}\mathrm{m}\stackrel{~}{\omega }(\rho ,\beta ,\omega )=\pi \mathrm{\Gamma }^+\mathrm{}\mathrm{e}\left\{\frac{\mathrm{\Omega }(\rho ,\beta ,\omega )}{\mathrm{\Omega }^2(\rho ,\beta ,\omega )+D^2(\rho ,\beta ,\omega )}\right\}.$$
(22)
For a fixed winding angle $`\beta `$ defining $`\gamma (\rho ,\omega )=\mathrm{}\mathrm{m}\stackrel{~}{\omega }(\rho ,\beta ,\omega )`$ we have for $`\nu =0.1`$, $`𝐇c`$-axis, and $`\beta =0^{}`$ the results shown in the top frame of Fig. 7 where we give a three dimensional plot of $`\gamma (\rho ,\omega )`$ as a function of distance $`\rho `$ from the vortex core and of energy $`\omega `$ for a system of tetragonal symmetry. This quantity is strikingly different from its Born limit counterpart which just corresponds to the $`N_L(\rho ,\omega )`$ as plotted in the top frame of Fig. 1 scaled by a constant factor of $`\pi t^+`$.
The $`H=0`$ case in the unitary limit is quite similar to the $`H0`$ case at $`\rho =1`$. The only difference is that for $`H=0`$ there is only one shallow dip at $`\omega =\mathrm{\Delta }_0=24`$meV which does not change position as one moves from $`\rho =1`$ to $`\rho =\nu `$. It is quite important to notice that in the limit $`\omega 0`$ $`\gamma (\rho ,\omega )`$ decreases by almost one order of magnitude as one approaches the vortex core. This means that the supercurrents surrounding the vortex core will also influence substantially the value of the universal limit for the transport.
The equivalent results for a system with orthorhombic symmetry with an effective mass anisotropy parameter $`\alpha =0.4`$ and an $`s`$-wave component $`s=0.25`$ to the gap are shown in the bottom frame of Fig. 7 to contrast the results for the system with tetragonal symmetry. The variation of $`\gamma (\rho ,\omega )`$ in the limit $`\omega 0`$ is less dramatic in this particular case.
It is obvious from these results that the supercurrents modify significantly the local effective impurity scattering which depends strongly on $`\rho `$. This has already been emphasized by Barash and Svidzinskii in another context.
## V Conclusion
In this work we have described the effect of supercurrents around the vortex core on the LDOS and on the local impurity scattering. We have included in the work only the effect of the Doppler shift in a semiclassical approximation on the extended quasiparticles. The work was motivated by the considerable success such an approach has had recently in describing the effect of vortex cores on the thermodynamics and some transport properties of $`d`$-wave superconductors for magnetic fields in the range $`H_{c1}<HH_{c2}`$, between the lower and upper critical fields. The calculations predict definite modifications of the expected van Hove singularity at the gap amplitude $`\mathrm{\Delta }_0`$ in a $`d`$-wave superconductor. There are additional ridges which depend on the geometry of the STM arrangement, i.e. its distance from the vortex core and direction of the STM with respect to nodal or antinodal direction. They also depend on orthorhombicity and are modified by impurity scattering. So far we are not aware of experiments which confirm these predictions. We hope our work will stimulate more experiments. Of course more sophisticated approaches to the structure of a vortex core in a $`d`$-wave superconductor are possible. We have already mentioned the work of Franz and Tešanović. They compare results for the LDOS obtained from a semiclassical approach with results from solutions for a single vortex within a BdG self consistent approach. For the angular averages over the vortex winding angle they consider, they do find good agreement between the two sets of results. An important additional aspect of these author’s work is that they investigate the possibility that the tunneling matrix element may have angular dependence and that consequently STM does not directly measure the LDOS. In the particular model considered, the perpendicular tunneling is modulated by a $`\mathrm{cos}^22\varphi `$ in-plane dependence which strongly affects the low energy part of the STM spectrum because there can be no tunneling right on the diagonal of the Brillouin zone. This modulation, however, will have less effect away from the $`\omega 0`$ region and is not important for the ridges. Here our own emphasis has been on the topology of the secondary van Hove ridges introduced in the LDOS by supercurrents. We have derived simple formulas for the position of these ridges in configuration space, for their dependence on $`H`$ which varies with position about the vortex, and have examined the effect of impurities both in Born and in unitary scattering limit on the topology of the ridges. We have also considered the modifications that are introduced when orthorhombicity is introduced within an anisotropic effective mass model.
## Acknowledgment
Research supported in part by NSERC (Natural Sciences and Engineering Research Council of Canada) and CIAR (Canadian Institute for Advanced Research). We thank E.J. Nicol for pointing out the relevance of Ref. 26 to our work. |
no-problem/0001/cond-mat0001173.html | ar5iv | text | # Yang–Lee Theory for a Nonequilibrium Phase Transition
\[
## Abstract
To analyze phase transitions in a nonequilibrium system we study its grand canonical partition function as a function of complex fugacity. Real and positive roots of the partition function mark phase transitions. This behavior, first found by Yang and Lee under general conditions for equilibrium systems, can also be applied to nonequilibrium phase transitions. We consider a one-dimensional diffusion model with periodic boundary conditions. Depending on the diffusion rates, we find real and positive roots and can distinguish two regions of analyticity, which can identified with two different phases. In a region of the parameter space both of these phases coexist. The condensation point can be computed with high accuracy.
\]
The investigation of nonequilibrium systems is a growing field in statistical mechanics and currently attracts much attention. In this context simple models such as driven diffusive systems play a paradigmatic role similar to the Ising model in equilibrium statistical mechanics. These systems establish a simple framework in which many phenomena can extensively be studied. Moreover, driven diffusive systems can easily be mapped to other nonequilibrium models, e.g. of polymer dynamics, interface growth and traffic flow .
A hallmark of many nonequilibrium systems is the absence of detailed balance and the support of stationary states with non-vanishing currents. Hence these systems build a larger class than respective equilibrium systems and phase transitions may appear under less restrict conditions. For example, it is known that spontaneous symmetry breaking and a first-order phase transition may occur in one-dimensional nonequilibrium systems with short-range interactions .
In thermal equilibrium the probability measures can in principle be expressed through an appropriate ensemble. For driven systems an equally powerful concept is missing. In Ref. a grand canonical partition function for nonequilibrium systems has been introduced for the first time. For the definition of this function one uses the matrix-product representation (see below) of the stationary state. Such a representation is not known for every system. But for several models it is known and, for instance, its existence is guaranteed for open reaction-diffusion models . In this Letter we utilize the partition function and show that more concepts from equilibrium physics may be applied to nonequilibrium systems. We develop a Yang–Lee theory giving us a very powerful method to analyze phase transitions.
We start from the grand canonical partition function $`Z(x)`$ and study its behavior as a function of complex fugacity $`x`$. Although only real values of the fugacity are of physical interest, the analytical behavior of thermodynamic functions can be revealed completely only by allowing the fugacity to be complex. This was first found by Yang and Lee for equilibrium systems : For finite systems the roots of the grand canonical partition function $`Z(x)`$ are in general complex or negative if real. But in the thermodynamic limit roots may approach the positive real axis. This marks a phase transition; in equilibrium systems the pressure $`p=k_\mathrm{B}Tlim_V\mathrm{}((1/V)\mathrm{log}Z)`$ is non-analytical, the density $`\rho =(/\mathrm{log}x)(p/k_\mathrm{B}T)`$ is discontinuous, and one can distinguish different phases.
We show here that the above behavior of the roots of the partition function can also be found for nonequilibrium systems. By similar reasoning, one can clearly define phases and determine their properties and furthermore transfer other concepts known from equilibrium statistical mechanics to nonequilibrium systems. A new method to compute the phase transition point with great accuracy is established, as well.
To present our results let us consider an one-dimensional stochastic diffusion model with $`L`$ sites and periodic boundary conditions. Each site $`k`$ may be occupied by one particle of type 1 or 2, or may be vacant (denoted by 0). Starting from random initial conditions with fixed particle densities, $`\rho _1`$ and $`\rho _2`$, the particles can rearrange themselves. The only allowed processes are local interchanges conserving the number of particles: $`(\alpha )_k(\beta )_{k+1}(\beta )_k(\alpha )_{k+1}`$ with $`\alpha ,\beta \{0,1,2\}`$. In an infinitesimal time interval $`d\tau `$ these diffusion processes occur with probability $`g_{\alpha ,\beta }d\tau `$. Here we consider the rates:
$$g_{1,2}=q,g_{2,1}=1,g_{1,0}=g_{0,2}=1$$
(1)
All other $`g_{\alpha ,\beta }`$ are zero. The stationary state of this model has been studied recently by Monte-Carlo simulations and mean-field approximations . For a snapshot of a Monte-Carlo simulation see Fig. 1. For the parameters chosen one observes one macroscopic droplet of particles which is surrounded by vacancies with a finite density of particles. This is a first sign of a first order phase transition. Increasing $`q`$ the droplet shrinks and disappears for $`q`$ larger than some $`q_0(\rho )`$; decreasing $`q`$ it grows until, for $`q=1`$, all particles are bound in the droplet. For $`q>1`$ the position of droplet fluctuates and is fixed for $`q<1`$, i.e. translational invariance is spontaneously broken. Similar one- and higher-dimensional models have been studied in .
The probability, $`𝒫(𝜷,t)`$, for configurations $`𝜷=(\beta _1,\beta _2,\mathrm{},\beta _L)`$ at time $`t`$ follows a Master Equation . The stationary probability distribution $`𝒫_{\mathrm{st}}(𝜷)`$ can be expressed as the trace over algebra elements $`D_\beta `$ :
$$𝒫_{\mathrm{st}}(𝜷)=\frac{1}{Z}\mathrm{tr}(D_{\beta _1}D_{\beta _2}\mathrm{}D_{\beta _L})$$
(2)
The normalization factor, $`Z`$, is defined as
$$Z=\underset{\beta _1,\beta _2,\mathrm{},\beta _L}{}\mathrm{tr}(D_{\beta _1}D_{\beta _2}\mathrm{}D_{\beta _L})=\mathrm{tr}(C^L)$$
(3)
with $`C=D_0+D_1+D_2`$. In the following we recognize that $`Z`$ plays the role of a grand canonical partition function. For Eq. (2) to hold, the $`D_\beta `$ have to fulfill the quadratic algebra
$$qD_1D_2D_2D_1=x_2D_1+x_1D_2$$
(5)
$$D_1D_0=x_1D_0$$
(6)
$$D_0D_2=x_2D_0$$
(7)
The two free parameters $`x_1`$ and $`x_2`$ reflect the freedom to choose the two densities of particles. A representation of this algebra is known
$$D_0=𝒢_0,D_1=x_1𝒢_1,D_2=x_2𝒢_2$$
(8)
with the matrices $`𝒢`$ given by
$`(𝒢_0)_{ij}=\delta _{1i}\delta _{1j}`$
$`𝒢_1=\left(\begin{array}{ccccc}a_1\hfill & t_1\hfill & 0\hfill & 0\hfill & \mathrm{}\hfill \\ 0\hfill & a_2\hfill & t_2\hfill & 0\hfill & \\ 0\hfill & 0\hfill & a_3\hfill & t_3\hfill & \\ 0\hfill & 0\hfill & 0\hfill & a_4\hfill & \mathrm{}\hfill \\ \mathrm{}\hfill & & & & \mathrm{}\hfill \end{array}\right),𝒢_2=\left(\begin{array}{ccccc}a_1\hfill & 0\hfill & 0\hfill & 0\hfill & \mathrm{}\hfill \\ s_1\hfill & a_2\hfill & 0\hfill & 0\hfill & \\ 0\hfill & s_2\hfill & a_3\hfill & 0\hfill & \\ 0\hfill & 0\hfill & s_3\hfill & a_4\hfill & \\ \mathrm{}\hfill & & & \mathrm{}\hfill & \mathrm{}\hfill \end{array}\right)`$
where we have introduced the notations
$`a_k=r\left(2\{k1\}_r\{k2\}_r\right)`$
$`s_kt_k=\{k\}_r\left(3r1+(2r1)^2\{k1\}_r\right)`$
$`\{k\}_r={\displaystyle \frac{r^k1}{r1}},r={\displaystyle \frac{1}{q}}`$
In the following we restrict ourselves to equal densities of particles, $`\rho _1=\rho _2=\rho `$, and have $`x_1=x_2=x`$. Then
$$Z=Z(x)=\underset{\beta _1,\beta _2,\mathrm{},\beta _L}{}x^{N(𝜷)}\mathrm{tr}(𝒢_{\beta _1}𝒢_{\beta _2}\mathrm{}𝒢_{\beta _L})$$
(9)
takes the form of a grand canonical partition function where $`N(𝜷)`$ counts the number of particles in a configuration $`𝜷`$ and therefore $`x`$ can be interpreted as a fugacity. The density of particles 1 is given by
$$\rho (x)=\frac{1}{Z}\mathrm{tr}(D_1C^{L1})=\frac{1}{2}\frac{}{\mathrm{log}x}P(x)$$
(10)
For the last expression in Eq. (10) we introduced, analogously to equilibrium physics, the “pressure”
$$P(x)=\frac{1}{L}\mathrm{log}Z(x)$$
(11)
Note, that in this context $`P`$ is not the physical pressure of the particles. The singularities and the analytical behavior of $`\rho (x)`$ and $`P(x)`$ are determined by the roots of $`Z(x)`$. To perform a numerical calculation of $`\mathrm{tr}(C^L)`$ we change the ensemble slightly and require that at least one vacancy is present. In this case we have to compute $`\mathrm{tr}(𝒢_0C^{L1})=(C^{L1})_{11}`$ which is much simpler and one has to handle $`(L/2)\times (L/2)`$ matrices only. Accordingly,
$$Z(x)\underset{j=1}{\overset{L1}{}}(xx_j)$$
(12)
is a polynomial of degree $`L1`$ in $`x`$ with roots $`x_j`$.
To study the analytical behavior of $`P`$ and $`\rho `$ we first determine the roots of $`Z(x)`$ in the complex plane. Since the partition function is a real polynomial with positive coefficients the roots come in complex conjugated pairs and real roots are negative. The roots of $`Z`$ for two different choices $`q>1`$ are shown in Fig. 2. For small $`q`$ the roots are lying on an elliptic curve. For $`q`$ larger than a critical value, $`q_{\mathrm{crit}}2.15`$, the curve is a hyperbola.
Let us treat the elliptic before the hyperbolic case. At the end we give some remarks on the case $`q<1`$.
The elliptic case ($`1<q<q_{\mathrm{crit}}`$). — In this case, the imaginary part of the root closest to the positive real axis (e.g. root No. 1 for $`q=2.0`$ in Fig. 2) vanishes for large $`L`$ with $`L^1`$ while the real part of this root stays finite and reaches a positive value, $`\stackrel{~}{x}(q)`$, in the limit $`L\mathrm{}`$.
To investigate further, let us denote the curve from the first to the last root (see Fig. 2) by $`c(\tau )`$ with $`\tau [0,l]`$ where $`l`$ is the arclength of $`c`$. The roots are not equally spaced and their density varies with the position on the ellipse. Furthermore, with increasing system size the density of roots along the curve grows linearly in leading order and we can denote the number of roots in a line element $`d\tau `$ by $`g(\tau )Ld\tau `$. The function $`g(\tau )`$ has a non-vanishing large $`L`$ limit and is shown in Fig. 3. Since in this limit the curve $`c(\tau )`$ is as well defined we find the pressure (up to an additive constant)
$$P(x)=𝑑\tau g(\tau )\mathrm{log}(xc(\tau ))$$
(13)
It is a thermodynamic quantity in the sense that its limit $`L\mathrm{}`$ exists. If $`q<q_{\mathrm{crit}}`$ the curve $`c`$ closes to give an ellipse with a positive intercept of the real axis: $`c(0)=\stackrel{~}{x}>0`$ and $`g(0)>0`$ in the limit $`L\mathrm{}`$. One has to distinguish two regions of analyticity, the inner and the outer of the ellipse in the complex plane. In both regions all physical quantities are continuous and differentiable. For a plot of $`P(x)`$ and $`\rho (x)`$ see Fig. 4.
In the outer region of the ellipse ($`x>\stackrel{~}{x}`$) the asymptotic behavior of $`P`$ is $`P(x)=\mathrm{log}x`$ and we have $`\rho (x)=1/2`$ in this phase.
For $`x`$ in the inner of the ellipse we find a $`\stackrel{~}{\rho }`$ with $`\rho (x)<\stackrel{~}{\rho }<1/2`$. Crossing the ellipse at $`\stackrel{~}{x}`$ the pressure $`P(x)`$ is not differentiable and $`\rho (x)`$ has a finite discontinuity; the transition is of first order. The jump in the density is related to the density of roots $`g(0)`$ at the real axis
$$1/2\stackrel{~}{\rho }=\pi \stackrel{~}{x}g(0)$$
(14)
The value of $`\stackrel{~}{x}`$ can easily be obtained by extrapolating the real part of the first root for $`L\mathrm{}`$. Using Eq. (10) and extrapolating one gets $`\stackrel{~}{\rho }`$.
If the density $`\rho `$ is fixed in the interval $`\stackrel{~}{\rho }<\rho <1/2`$ one finds the two phases coexistent in the stationary state; the dense one with $`\rho =1/2`$ (i.e. without vacancies) and the other with $`\rho =\stackrel{~}{\rho }`$ (see again Fig. 1). The dense phase acts like a bottle-neck and accordingly the current is $`(q1)\rho (1\rho )=(q1)/4`$.
Increasing $`q`$ to $`q_{\mathrm{crit}}`$, the density of particles above that condensation starts, $`\stackrel{~}{\rho }`$, reaches $`1/2`$ and the density of roots on the positive real axis $`g(0)`$ vanishes according to Eq. (14). The phases in the inner and outer region become similar and one finds a second order phase transition at $`q=q_{\mathrm{crit}}`$.
Next we want to fix the density of particles and determine the value $`q_0(\rho )`$ of $`q`$ below which two phases coexist. Therefore, the values of $`\stackrel{~}{\rho }`$ are plotted versus $`q`$ in Fig. 5. In the same plot we give the mean-field approximation for the phase transition $`q_0^{\mathrm{MF}}(\rho )=(1+6\rho )/(1+2\rho )`$ as it has been calculated in . In the latter paper the case $`\rho =0.2`$ was studied by Monte-Carlo simulations as well and the phase transition was found at $`q_0^{\mathrm{MC}}=1.62\pm 0.05`$. With the above method we find the more precise value $`q_0(0.2)=1.617\pm 0.001`$.
The hyperbolic case ($`q>q_{\mathrm{crit}}`$). — Also shown in Fig. 2 are the roots for $`q=2.5`$. The curve $`c(\tau )`$ is a hyperbola. The distribution of roots $`g(\tau )`$ (see Fig. 3) is similar to the elliptic case. The real and imaginary part of the first and last roots, and hence the arclength, $`l`$, diverges with $`L`$. We find only the phase with $`0<\rho (x)<1/2`$ continuous and monotonically increasing with $`x`$.
The case $`q<1`$. — In this case the roots lie nearly equally spaced on a circle. But the diameter of the circle shrinks exponentially with $`L`$ and in the large $`L`$ limit all roots are located at $`x=0`$. In the limit $`L\mathrm{}`$ we find $`\rho (x)=1/2`$ for $`x>0`$. Details on the $`L`$-dependence and geometric properties of $`c(\tau )`$ for all above cases will be given elsewhere . In physically interesting cases the density of particles is fixed in the interval $`0<\rho <1/2`$ and again one finds again coexistent phases.
In summary, we have demonstrated that by allowing for complex values of the fugacity in the grand canonical partition function one can analyze the phases and phase transitions in nonequilibrium models. In the model presented, a first order phase transition has been investigated. In particular, the transition point can be computed with great accuracy.
Due to the general nature of the method it can also be applied numerically or analytically to other models where the density of particles controls a phase transition, e.g. jamming transitions in traffic related models. Furthermore, the method is not constrained to investigate roots in the fugacity plane. One can study complex roots of some other parameter (e.g. the input rate in the totally asymmetric exclusion process ) as well .
I want to thank M.R. Evans, C. Godrèche, T. Heinzel, D. Mukamel and V. Rittenberg for discussions. |
no-problem/0001/astro-ph0001402.html | ar5iv | text | # CHANDRA X-RAY OBSERVATIONS OF THE HYDRA A CLUSTER: AN INTERACTION BETWEEN THE RADIO SOURCE AND THE X-RAY-EMITTING GAS
## 1. Introduction
The Hydra A radio galaxy is associated with a relatively poor cluster of galaxies at redshift $`z=0.052`$. The cluster harbors an atmosphere of X-ray emitting gas of luminosity $`L_x(0.54.5)=2.2\times 10^{44}\mathrm{ergs}\mathrm{s}^1`$, and a mean gas temperature of $`4`$ keV based on $`Einstein`$ MPC observations (David et al. 1990) and ASCA observations (Ikebe et al. 1997). A cooling flow is present with an accretion rate of $`\dot{M}250\mathrm{M}_{}\mathrm{yr}^1`$ (assuming $`\mathrm{H}_0=50\mathrm{km}\mathrm{s}^1\mathrm{Mpc}^1`$) within a radius of $`170`$ kpc (White et al. 1997; Peres et al. 1998). The hot atmosphere is centered on a brightest cluster elliptical galaxy (BCG) that hosts a large (80 arcsec or 84 kpc in projection), unusually powerful ($`P=1.6\times 10^{26}\mathrm{W}\mathrm{Hz}^1`$ at 178 MHz) Fanaroff-Riley type 1 (FR 1) radio source 3C 218 (Ekers & Simkin 1983; Taylor et al. 1990; Taylor 1996). The twin jet-lobe radio source emerges from a disk of young stars (McNamara 1995; Hansen et al. 1995; Melnick et al. 1997) and nebular emission (Ekers & Simkin 1983; Hansen et al. 1995; Baum et al. 1988) several kpc in size that is in rotation about the nucleus. H I is seen in absorption toward the nucleus (Dwarakanath, Owen, & van Gorkom 1995) and has been mapped in absorption against the parsec scale nuclear radio source with the VLBA (Taylor 1996).
Hydra A was observed by the Chandra X-ray Observatory during its orbital verification and activation phase. Hydra A’s several interesting properties provide a first opportunity to investigate a cooling flow and potential interactions between
<sup>1</sup>Harvard-Smithsonian Center for Astrophysics, 60 Garden St. Cambridge, MA 02138
<sup>2</sup>Department of Engineering Physics, University of Wollongong, Wollongong NSW 2522, Australia
<sup>3</sup>Massachusetts Institute of Technology, Center for Space Research, Cambridge, MA 02139
<sup>4</sup>Astronomy Department, University of Virginia, Charlottesville, VA
the radio source and X-ray-emitting gas using Chandra’s unprecedentedly high spatial resolution. We present an analysis of Hydra A based on the preliminary telescope calibration in this Letter, and report the discovery of an interaction between the radio source and the X-ray-emitting gas. We assume $`\mathrm{H}_0=70\mathrm{km}\mathrm{s}^1\mathrm{Mpc}^1`$, $`\mathrm{\Omega }_\mathrm{M}=0.3`$, $`\mathrm{\Omega }_\mathrm{\Lambda }=0.7`$, a luminosity distance of 240 Mpc, and 1 arcsec = 1.05 kpc.
## 2. Data Analysis
The calibration observations of Hydra A were performed on 1999 October 30. A total integration time of 40 ksec was obtained, 20 ksec centered at the aim point of the S3 back-illuminated ACIS chip (OBSID 576), and 20 ksec at the ACIS I aim point on the I3 front-illuminated device (OBSID 575). X-ray events with energies below 300 eV and above 10 keV were not considered in our analysis, and flight grades 0, 2, 3, 4, 6 were retained. The particle background was generally stable throughout the S3 observations. Only 300s of the 20 ksec exposure experienced a $`30\%`$ increase in particle background. Therefore no S3 data were rejected on this basis. The spectral analysis presented here is limited to a $`0.57`$ keV bandpass for the S3 data.
### 2.1. X-ray Morphology
A Chandra image of the X-ray emission from the central $`118\times 118`$ arcsec of the cluster is shown in Fig. 1. The image is a summed, 40 ksec exposure obtained with the ACIS-I3, front illuminated device, and the ACIS-S3, back-illuminated device. The top panel shows the unsmoothed image, after event filtering, centered on the X-ray point source. The X-ray emission shows a great deal of structure on scales ranging from less than a few kpc to tens of kpc that have not been seen in earlier X-ray imagery. An X-ray point source ($`RA=091805.77`$ , $`Dec=120542.53`$, J2000), shown inset to the top panel of Fig. 1, coincides with the central radio core and BCG nucleus. The lower panel of Fig. 1 shows the wavelet-smoothed X-ray image. The emission is concentrated in a central triangular region, 10–15 arcsec in size, and in fainter fingers of emission extending northeast and southwest of the center. The emission spectrum of this material is consistent with thermal emission from $`3`$ keV cluster gas. Two depressions in the X-ray emission, 20–30 arcsec in diameter, are seen $`20`$ arcsec to the north and south of the cluster center. The surface brightness within these cavities is a factor of $`1.5`$ lower than the mean surface brightness of the emission at similar radii from the center. This decline in surface brightness is consistent with these regions being devoid of gas at the ambient temperature and density.
Fig. 1. Top Panel: 40 ksec integration X-ray image of the central region of the Hydra A cluster centered on the nuclear X-ray point source (shown inset). Bottom Panel: Wavelet smoothed and reconstructed image of the same region superposed on the 6 cm VLA Radio image of Hydra A (contours). The scale bar is 20 arcsec in length.
### 2.2. Radio Morphology
A 6 cm VLA image obtained in the A-array, obtained by G. Taylor, is shown superposed on the wavelet-reconstructed X-ray image in Fig. 1. The cavities in the X-ray emission are filled by emission from the radio lobes, and the X-ray cavities and radio lobes are remarkably similar in shape. Similar cavities have been seen in $`ROSAT`$ imagery of the Perseus cluster (Böhringer et al. 1992) and in Cygnus A (Carilli et al. 1994). Heinz et al. (1998) proposed a model to explain such cavities as a shell of shocked gas displaced by the relativistic material from the radio jet. This model would imply that the X-ray emission surrounding the radio source should be considerably hotter than the material away from the radio source. We evaluate this interpretation of the radio cavities, and we explore the physical state of the cluster gas in §2.4.
### 2.3. Optical Morphology: The Central Disk
Spatial correlations between the X-ray, optical, and radio emission are present in the inner 20 kpc of the BCG. The $`U`$-band contours are superposed on a grayscale X-ray image in Fig. 2. The X-ray point source is centered on a gaseous disk of nebular emission and young blue stars in rotation about the nucleus (Ekers & Simkin 1983; Baum et al. 1988; McNamara 1995; Hansen et al. 1995; Melnick et al. 1997). The disk of star formation is about 10 kpc by 7 kpc in size, and is roughly $`0.7`$ magnitudes bluer than the $`UI`$ color of the galaxy’s halo (McNamara 1995). The nebular emission (Baum et al. 1988; Hansen et al. 1995) extends over a similar region.
The brightest X-ray emission, apart from the point source, appears in a flattened structure coincident with the disk, and in an irregular structure several arcsec east of the disk. The X-ray point source is within a half arcsec of the radio core (Fig. 1) and the BCG’s nucleus. This positional correspondence is within Chandra’s absolute pointing error. The surface brightness of the light gray region in Fig. 2 is $`30\%`$ fainter than the dark gray structure just east of the disk. This fainter emission forms a wedge-shaped structure that widens toward two companion galaxies projected onto the BCG (see also Fig. 1). The asymmetry in the X-ray structure may be caused, in part, by an irregular potential well associated with a merger between the BCG and the companion galaxies (Ekers & Simkin 1983).
Fig. 2. Detail of the X-ray emission, grayscale smoothed with a $`FWHM=1.75`$ pix gaussian, superposed on the $`U`$-band contours of the BCG (McNamara 1995). The central disk of gas and young stars are the elliptical contours centered on the X-ray point source. The scale bar is 10 arcsec in length.
The radio jets (Fig. 1) emerge from the optical disk at roughly a $`20\mathrm{deg}`$ angle from the disk’s minor optical axis and rotation axis, and flare into lobes where the gas pressure reaches $`6\times 10^{10}\mathrm{erg}\mathrm{cm}^3`$. The jets show no obvious signs of interacting with the optical disk emission or the X-ray emission in the central flattened structure.
### 2.4. Physical State of the X-ray Emitting Gas
We computed the radial distribution of density, temperature, and pressure of the gas in the central 84 kpc. These parameters were computed by fitting a single temperature XSPEC model to the X-ray emission in annuli centered on a central X-ray point source. The fluxes were corrected for cluster emission at large radii projected onto the inner regions (deprojected), and the abundance and foreground absorption were free parameters in the models. The abundances were found to be $`0.4`$ solar, and the foreground absorption $`2\times 10^{20}\mathrm{cm}^2`$. No evidence for excess absorption from cold gas within the cluster is found, with the exception of a large column toward the nuclear point source (§4). In addition, we constructed hardness ratio profiles, $`\kappa (R)`$, by taking the ratios of the X-ray surface brightnesses, $`I`$, in several passbands. These profiles are plotted in Fig. 3. The hardness ratios are defined as $`\kappa _1=I(0.51.5)/I(1.52.5)`$ (open symbols) and $`\kappa _2=I(0.52.5)/I(2.56.0)`$ (filled symbols), where the figures in brackets are the passbands in keV. The spectrum is harder as $`\kappa (R)`$ decreases. The $`n_e`$, $`kT`$, and $`P`$ profiles exclude the central point source, while the hard (absorbed) point source is included in the hardness ratio plot.
The temperature increases from $`3`$ keV in the central 10 kpc to $`4`$ keV at 80 kpc. The rms density within the central 10 kpc is $`n_e=0.06\mathrm{cm}^3`$. The density declines with radius as $`r^1`$ to a radius of 70 kpc, and the gas pressure declines by a factor of $`6`$ from the center to 70 kpc. The central point source is considerably harder than the surrounding emission within a 10 kpc radius. Beyond the central source, $`\kappa `$ decreases (hardens) $`10\%30\%`$ between the center and 80 kpc. The broad-band $`\kappa _2`$ shows the larger decline. There is no significant difference between hardness ratio profiles including or excluding the flux from the cavities.
We note that in the 0.5 – 7 keV Chandra band, the count rate for gas at 80 keV is about 2/3 of that for gas with the same emission measure at 5 keV. If gas in or near the cavities has been compressed in a shock, its emission in the Chandra band would be larger than from the same amount of gas at ambient conditions. Then the regions within the radio lobes would be brighter than the surrounding emission, which is not observed.
## 3. Discussion
### 3.1. Origin of the X-ray Cavities
There is no indication that the gas surrounding the radio source is hotter than the ambient cluster gas. This behavior is inconsistent with strongly shock heated gas, as is suggested by the Heinz et al. (1998) model, but is consistent with the cool central material being displaced as the radio source expands subsonically. There is no evidence that the radio source is heating the gas outside the cavities (Nulsen et al. in preparation).
Fig. 3. Run of electron density (upper left), temperature (upper right), pressure (lower left), and hardness ratio (lower right). In this convention, hardness increases as $`\kappa `$ decreases.
We now calculate the minimum energy required by the radio source to displace the gas using the southern cavity for illustration. The center of the southern cavity is approximately $`26`$ arcsec from the X-ray nuclear point source. The electron density in the annulus centered on the southern cavity is $`n_\mathrm{e}=0.027\mathrm{cm}^3`$ and the temperature is $`kT=3.4\mathrm{keV}`$, giving a pressure $`p=2.8\times 10^{10}\mathrm{erg}\mathrm{cm}^3`$. Based on the deprojection, undisturbed gas in a volume the size of this cavity would have contributed about 2000 counts. This is consistent with our estimate for the count deficit in the cavity. We therefore assume that the ambient ICM has been displaced from this region by the radio source. Since there is no spectral hardening at the edges of the cavities, we assume that the pressure within them is similar to the ambient gas pressure.
The minimum energy required to push the gas out of a sphere of radius 15 kpc is then $`pV1.2\times 10^{59}\mathrm{erg}`$. Since there are no signs of shocks, we assume that gas motions are subsonic and can be ignored, so that the total energy required to inflate the cavity is a modest multiple of this.
The cavity would take $`2\times 10^7`$ yr to expand at approximately the sound speed. The cavity is also buoyant, so it must be refilled on a similar timescale, $`2R\sqrt{r/GM(R)}`$, where $`r`$ is the radius of the cavity and $`R`$ is the distance to the cluster center. From the deprojection, a rough estimate of the total gravitating mass within 30 kpc of the cluster center is $`M(R)=3\times 10^{12}\mathrm{M}_{}`$, giving a refilling time of about $`6\times 10^7\mathrm{yr}`$. Thus, the minimum mechanical power from the southern radio jet required to maintain the cavity is $`6\times 10^{43}\mathrm{erg}\mathrm{s}^1`$, which is comparable to the total radio power from Hydra A (Ekers & Simkin 1983). Either timescale implies an efficiency of conversion of kinetic energy to radio power of order unity, and we need not postulate the existence of kinetic luminosity substantially in excess of the luminous radio power (e.g. Heinz et al. 1998).
Our interpretation requires the radio lobes and surrounding gas to be nearly in pressure equilibrium. However, the gas pressure appears to be more than an order of magnitude larger than the minimum energy pressure of the radio lobes (Taylor et al. 1990). If the lobes are in pressure balance with the cluster gas, and if equipartition between the magnetic field and relativistic particles is to be maintained, then one or more of the following conditions is implied. There is a significant additional contribution to the particle energy density from low energy electrons or from protons (e.g. Böhringer et al. 1993), the radio filling factior is significantly less than one, or the radio lobes are at significantly larger radii in the cluster atmosphere than their projected positions.
### 3.2. The Cooling Flow
Gas with the observed central density $`n_e(r<10\mathrm{kpc})=0.06\mathrm{cm}^3`$, temperature $`kT=3`$ keV, and abundance of $`0.4`$ solar has a radiative cooling time of $`6\times 10^8`$ yr. If such cool material has been present in the cluster for $`>`$1 Gyr, the gas should cool to low temperatures and flow to the center of the cluster. The Chandra data allow us to obtain high quality spectra of several independent regions within the central 80 kpc of the cluster. We have fitted an absorbed thermal (mekal) plus constant pressure cooling flow (mkcflow) model to a series of 8 circular apertures ranging in equal steps from $`10`$ arcsec to $`80`$ arcsec in radius. The central $`1.5`$ arcsec was excluded. Assuming spherical symmetry, we attribute the differences in mass deposition rate, $`\dot{M}`$, for successive apertures to the annulus they define, apply a standard geometric deprojection to convert these to mass deposition rates per unit volume, and calculate total $`\dot{M}`$’s for the spheres. The total mass deposition rate within a 74 kpc sphere is $`34\pm 5\mathrm{M}_{}\mathrm{yr}^1`$, and $`\dot{M}(R)R^{1.1}`$.
### 3.3. Cooling Rate vs Star Formation Rate
The most controversial issue concerning cooling flows is the fate of the cooling material. We now ask whether the accretion rates derived from the new X-ray data are consistent with the star formation rate. Star formation appears to be confined to the disk (§2.3), whose young stellar population mass estimate ranges from $`10^{7.79}\mathrm{M}_{}`$ (McNamara 1995; Hansen et al. 1995). Estimates of the average star formation rates vary from $`<1\mathrm{M}_{}\mathrm{yr}^1`$ for continuous star formation for $``$Gyr duration, to $`15\mathrm{M}_{}\mathrm{yr}^1`$ for a younger, $`10^8`$ yr old burst population. The degree to which star formation may be consuming the cooling gas depends on factors including the relative ages of the cooling flow and the stellar population, and the stellar initial mass function. In the central 10 kpc, a region somewhat larger than the central optical disk, the mass deposition rate is $`4\pm 2\mathrm{M}_{}\mathrm{yr}^1`$. Thus, the observed rates of star formation can account entirely for the mass cooling within this region regardless of the age of the cooling flow. However, the disk star formation would provide a sink for the cooling gas within the entire 74 kpc region for $`<3\times 10^6`$ yr, which is much smaller than the minimum time $`1`$ Gyr required to establish a cooling flow. Most of the cooling material remains unaccounted for, and there is no evidence that the radio source is heating the gas and reducing the cooling rates.
## 4. The Central Point Source
We extracted the spectrum of the central point source within a 1 arcsec radius aperture. The background was taken from the annular region between $`r=12.5`$ arcsec centered on the point source. The net point source spectrum was found by subtracting the background spectrum, after normalizing by the relative areas of the central and annular regions. The net spectrum appears to be a power law absorbed by a $`N_\mathrm{H}=4.5(2.29.9)\times 10^{22}\mathrm{cm}^2`$ ($`90\%`$ errors). The point source flux, uncorrected for absorption, is $`f(0.66.0)=1.0\times 10^{13}\mathrm{ergs}\mathrm{s}^1\mathrm{cm}^2`$, which corresponds to a luminosity of $`6.9\times 10^{41}\mathrm{ergs}\mathrm{s}^1`$. The absorbing column toward the point source is more than two orders of magnitude larger than the Galactic foreground column, and is confined to a spatial extent of $`<1.5`$ arcsec.
This column density is in reasonably good agreement with the column of neutral hydrogen absorption seen toward the VLBI radio source in the nucleus of the galaxy (Taylor 1996). The radio absorption map confines the spatial extent of this high column density material to a region of $`<24`$ pc in the nucleus of the galaxy. Assuming the X-ray and radio flux is absorbed by the same material, the combined observations restrict the X-ray-emitting region of the point source to $`<24`$ pc.
## Acknowledgments
PEJN gratefully acknowledges the hospitality of the Harvard-Smithsonian Center for Astrophysics. BRM acknowledges grant NAS8-39073.
## References
Baum, S.A., Heckman, T., Bridle, A., van Breugel, W., & Miley, G., 1988, ApJS, 68, 643
Böhringer, H., Voges, W., Fabian, A.C., Edge, A.C., & Neumann, D.M. 1993, MNRAS, 264, L25
Carilli, C.L., Perley, R.A., & Harris, D.E. 1994, MNRAS, 270, 173
David, L.P., Arnaud, K.A., Forman, W., & Jones, C. 1990, ApJ, 356, 32
Dwarakanath, K.S., Owen, F.N., & van Gorkom, J.H., 1995, ApJ, 442, L1
Ekers, R.D., & Simkin, S.M. 1983, ApJ, 265, 85
Hansen, L., Jørgenson, H.E., & Nørgaard-Nielson, H.U. 1995, AA, 297, 13
Heinz, S., Reynolds, C.S., & Begelman, M.C. 1998, ApJ, 501, 126
Ikebe, Y., Makishima, K., Ezawa, H., Fukazawa, Y., Hirayama, M., Honda, H., Ishisaki, Y., Kikuchi, K., Kubo, H., Murakami, T., Ohashi, T., Takahashi, T., & Yamashita, K. 1997, ApJ, 481, 660
McNamara, B.R. 1995, ApJ, 443, 77
Melnick, J., Gopal-Krishna, & Terlevich, R. 1997, AA, 318, 337
Peres, C. B., Fabian, A. C., Edge, A. C., Allen, S. W., Johnstone, R. M., & White, D. A. 1998, MNRAS, 298, 416
Taylor, G.B. 1996, ApJ, 470, 394
Taylor, G.B., Perley, R.A., Inoue, M., Kato, T., Tabara, H., & Aizu, K., 1990, ApJ, 360, 41
White, D. A., Jones, C., & Forman, W. 1997, MNRAS, 292, 419 |
no-problem/0001/cond-mat0001233.html | ar5iv | text | # Non-Gaussian velocity distributions in excited granular matter in the absence of clustering
## Abstract
The velocity distribution of spheres rolling on a slightly tilted rectangular two dimensional surface is obtained by high speed imaging. The particles are excited by periodic forcing of one of the side walls. Our data suggests that strongly non-Gaussian velocity distributions can occur in dilute granular materials even in the absence of significant density correlations or clustering. When the surface on which the particles roll is tilted further to introduce stronger gravitation, the collision frequency with the driving wall increases and the velocity component distributions approach Gaussian distributions of different widths.
Recent simulations and experiments have shown that clusters are formed due to inelastic collisions in excited granular matter. The formation of clusters was shown to lead to non-Gaussian velocity distributions. Here we report that even in the absence of cluster formation or significant density correlations, dissipation leads to strongly non-Gaussian velocity distributions.
The statistical description of dissipative systems such as granular flows are of fundamental interest. Because thermal energies are irrelevant, the concept of “granular temperature” has been introduced based on the velocity fluctuations of particles. A dissipative kinetic theory has been developed to describe rapid granular flows. In this theory, velocity distributions are assumed to be locally Gaussian. Deviations from Maxwell-Boltzmann are explained by averaging over local Gaussian velocity distributions with different widths which are inversely related to the local clustering. Given the success of this local Gaussian distribution assumption, we might conclude that the velocity distributions can be considered to be approximately Gaussian for most practical situations where clustering does not occur.
To test this idea, we study the velocity distributions of steel spheres rolling inside a two-dimensional box for which energy is supplied to the particles by oscillating one of the side walls. A similar system has been used previously to demonstrate the formation of clusters at high densities due to inelastic collisions. However, at low densities corresponding to longer mean free paths and lower collision rates, the particles appear to be randomly distributed and no clustering is observed. We use high speed and high resolution imaging to obtain accurate velocity distributions. Because one of the sides of the rectangular geometry is used to input energy and the particle-particle and particle-wall interactions are dissipative, the velocity distribution is asymmetrical. An obvious consequence is that the velocity fluctuations in the direction parallel to the driving wall motion is greater than in the perpendicular direction. We find that the velocity distribution is strongly non-Gaussian even in the direction perpendicular to the motion of the oscillating wall. By varying the frequency $`f`$ and the angle of inclination $`\theta `$ of the surface, we show that non-Gaussian velocity distributions can occur when the collision rate due to dissipative interactions is greater than the particle-driving wall collision rate.
The experimental setup consists of one hundred 0.32 cm diameter steel (or brass) spheres rolling on an optically smooth glass surface (29.5 cm $`\times 24.7`$ cm) which is enclosed by steel side walls. The setup is similar to that in Ref. with a few enhancements. The coefficient of rolling friction of the spheres on the glass plate is less than $`2\times 10^3`$. In contrast, the coefficient of sliding friction is $`0.2`$. The particles are visualized using a SR-1000 Kodak digital camera capable of taking 1000 frames per second (fps). The camera and lights are placed such that the centers of the spheres appear bright. The best pixel resolution of $`512\times 480`$ can be obtained at 250 fps, which was used for most of the data presented here. A typical image is shown in Fig. 1a. The particles correspond to $`4`$ pixels in diameter and a centroid technique was used to find the position of the particle to within 0.01 cm. The particles with the highest velocities were determined by using images separated by 4 ms, and the velocity of the slowest particles were determined by using images separated by at least 12 ms. The positions of the slow moving particles do not change appreciably in comparison to the pixel size if a very high frame rate is used, resulting in substantial roundoff errors. We note that a combination of both high resolution and high frame rate is required to obtain accurate velocities.
The driving wall is connected to a linear solenoid which is controlled by a computer, and the resulting motion of the piston is measured with a displacement sensor. The displacement of the piston as a function of time, $`d(t)`$, for three typical driving frequencies is plotted in Fig. 1b. A square waveform of fixed amplitude (4 volt) and various frequencies is used to obtain the displacements.
By averaging over at least $`10^5`$ images, we obtain velocity distributions of the particles rolling inside the rectangular box. The distributions of the velocity components, $`P(v_x)`$ and $`P(v_y)`$, are plotted in Fig. 2a and 2b respectively. The directions $`x`$ and $`y`$ and the origin are indicated in Fig. 1a. The surface is slightly tilted ($`\theta =0.1^{}`$) to continously excite the particles. Particles that are less than 2 cm from the side walls are excluded to avoid direct effects of the boundaries. We checked that if a narrow region in $`y`$ is used, the form of the velocity distributions is independent of $`y`$. The plots in Fig. 2 correspond to various driving frequencies. Note that the form of $`P(v_x)`$ and $`P(v_y)`$ does not depend strongly on the driving frequency $`f`$. The distribution $`P(v_x)`$ is symmetrical about $`v_x=0`$, but is strongly non-Gaussian as can be seen from the poor fit to a Gaussian which is also shown in Fig. 2a.
The distribution $`P(v_y)`$ shown in Fig. 2b is non-Gaussian and asymmetrical. The reason for the asymmetry is not difficult to understand. The energy is supplied to the system by the driving wall. Therefore, particles move with higher velocities in the negative-$`y`$ direction (indicated by the bump at $`v_y35`$ cm/s). The second more subtle reason for the asymmetry is that the particles suffer an inelastic collision either with the wall opposite to the driving wall or with other particles before moving in the positive $`y`$ direction. Therefore, particles on average move with lower velocities in the $`+y`$ direction than in the $`y`$ direction. Consequently, the particles also spend a longer time interval traveling in the $`+y`$ direction than in the $`y`$ direction. Thus, asymmetrical $`v_y`$ distributions are obtained by averaging over time. Such effects were also observed in the distributions obtained numerically by Luding et al.. The form of $`P(v_x)`$ is relatively independent of the energy input, and therefore its non-Gaussian behavior is surprising.
In Fig. 3, we plot $`P(n)`$, the distributions of the number of particles $`n`$, in a 3 cm $`\times `$ 3 cm area averaged over the box and over all images. For comparison, the distribution of randomly distributed particles is also shown (dashed curve). Our data for $`P(n)`$ is similar to the randomly distributed case with a small deviation at higher $`n`$. Because of the dilute nature of our experiments, no effects of clustering are observed as reported in Ref. .
The effect of gravity and the collision rate with the driving wall is studied by lifting the surface on which the particles roll by the tilt angle $`\theta `$. The $`v_x`$ and $`v_y`$ distributions are plotted in Fig. 4 corresponding to four values of $`\theta `$. The density distribution $`P(n)`$ is no longer uniform in $`y`$ as in the $`\theta =0.1^{}`$ case, but decreases strongly with distance from the driving wall. (At the highest angle $`\theta =6.7^{}`$, the particles do not reach the top wall.) The velocity distributions correspond to particles in a 2 cm wide region at a distance $`y=4`$ cm from the driving wall to avoid the effects of the increase in potential energy and the decrease in the average $`v_y`$ at higher $`\theta `$. The distributions $`P(v_x)`$ and $`P(v_y)`$ grow broader as $`\theta `$ increases. A Gaussian fit to the data corresponding to $`\theta =6.7^{}`$ is shown in Fig. 4. The data for $`P(v_x)`$ is reasonably described by a Gaussian especially at low $`v_x`$. However, $`P(v_y)`$ deviates systematically from Gaussian because of the asymmetry. Furthermore, the width of $`P(v_y)`$ is large compared to the width of $`P(v_x)`$ because energy is mostly supplied to the $`x`$ component in an inelastic collision consistent with earlier observations of vertically vibrated granular matter. We further note that $`P(v_x)`$ averaged over all $`y`$ is non-Gaussian for $`\theta =6.7^{}`$ because of the slight increase in the widths of the Gaussians as a function of $`y`$. This might also be the reason for deviations from Gaussian observed in velocity distributions of vertically vibrated monolayer layer of particles.
As $`\theta `$ increases, the acceleration in the $`y`$ direction increases. Therefore the particles fall back to the driving wall faster as $`\theta `$ is increased. This results in higher collision rates of particles with the driving wall leading to higher average velocities. Thus as the collision rate with the driving wall increases, the distributions become broader and more Gaussian. Furthermore, the total number of inelastic collisions the particles undergo before returning to the driving wall decreases because fewer particles hit the top wall at higher $`\theta `$.
We note that there is no discrepancy with the observation of the broadening of the velocity distributions with an increase in $`\theta `$ and the data plotted in Fig. 2. It might appear that changing the frequency of the driving piston should also increase the collision rate with the bottom wall and hence have the same effect. However, the distributions plotted in Fig. 2 do not depend strongly on the driving frequency. The lack of the strong dependence of $`P(v_x)`$ and $`P(v_y)`$ on $`f`$ in Fig. 2 can be explained by the fact that fewer particles are struck by the forward moving wall at higher $`f`$. Thus the net energy input (and collision rate with the bottom wall) appears to not depend strongly on $`f`$.
Next we discuss the effect of rolling on the distributions. In a previous study, it was noted that rolling appears to lead to a lower effective coefficient of restitution $`r`$ during collisions. Some theoretical work has been done studying rolling and collisions. In Fig. 5a, we plot the $`r`$ as a function of scattering angle $`\varphi `$ of a rolling steel sphere on a glass surface colliding with steel boundary walls. The coefficient of restitution is calculated as the square root of the ratio of the final to the initial kinetic energy. The data is distributed over a range of $`r`$ which becomes smaller for higher $`\varphi `$. The large fluctuations are not related to measurement noise nor can they be accounted for by variations in the initial velocity (see Fig. 5b.)
The mean value of $`r`$ depends on $`\varphi `$ and is lower for higher $`\varphi `$ (indicated by the dashed line in Fig. 5a.) The mean value of $`r`$ during a wall collision averaged over all $`\varphi `$ was $`0.5\pm 0.02`$. In contrast the average loss of energy during a particle-particle collision was lower and the mean $`r`$ was found to be 0.65 by averaging over 20 collisions. In comparison, the coefficient of restitution of the steel sphere bouncing off a steel block after being dropped is 0.93. The larger losses in kinetic energy occur mostly because the particles slide as their angular velocity changes to match their new translational velocity. The change in angular velocity was observed to occur within a centimeter even at the highest velocities observed. It may be possible to approximately model the complications due to rolling by simply assuming a lower value of $`r`$. However, an event-driven simulation using only a loss of kinetic energy in the normal direction does not produce non-Gaussian distributions in $`v_x`$, but yields asymmetric distributions for $`v_y`$. The subtle effects introduced by rolling may help us understand the effect of higher dissipation on velocity distributions.
Limited experiments were also performed with brass spheres. The distributions obtained were identical to those corresponding to steel spheres under the same conditions. Thus the velocity distributions obtained are not very sensitive to the details of the surface properties of the spheres.
The distribution of the speed of the particles is a combination of the non-Gaussian or approximately Gaussian velocity distributions of different widths (such as in Figs. 2 and 4) and deviates from the Maxwell-Boltzmann form. Subtle effects on the velocity distributions such as the scaling of the temperature (the second moment of the distributions) in the near elastic limit has been studied theoretically by Kumaran, Grossman et al., and experimentally by Warr et al.. However, the deviations observed in our experiments are stronger as the loss of energy during a collision is significant compared to the energy of the particle. Therefore scaling around the Maxwell-Boltzmann distributions as discussed in Refs. for $`r1`$ cannot be applied to our data.
In conclusion, our experiments on excited dissipative granular matter rolling and bouncing inside a rectangular box have yielded velocity distributions that are strongly non-Gaussian in the absence of significant clustering and density correlations. Our data suggests that caution has to be used in assuming Gaussian velocity distribution and Maxwell-Boltzmann distributions even for rapid granular flows where position correlations are negligible.
We thank J. Tobochnik and H. Gould for many useful discussions, and J. Norton for technical support. This work was supported by the donors of the Petroleum Research Fund, and Grant No. DMR-9983659 from the National Science Foundation. A. K. thanks the Alfred P. Sloan Foundation for its support. |
no-problem/0001/astro-ph0001208.html | ar5iv | text | # Structure of the solar core: Effect of asymmetry of peak profiles
## 1 Introduction
Accurate determination of the frequencies of low-degree oscillation of the Sun is essential in inferring the structure of the solar core, owing to the small effect of this part of the Sun on the oscillation frequencies (e.g. Turck-Chièze, Brun, & Garcia 1999). Thus particular care is required in interpreting the data obtained. Important aspects, which have received considerable attention recently, are the effect on the frequencies of the variations of solar activity during the solar cycle (Dziembowski & Goode 1997; Dziembowski et al. 1997), as well as the different results obtained from intensity and velocity measurements (Toutain et al. 1997). These different observations have led to a better understanding of how the data must be analyzed, emphasizing, for example, the need to use the same period of time for low- and high-degree modes to avoid differences in the near-surface effects, and ideally considering data sets of one or a maximum of two years around the minimum of activity.
There is now a number of experiments designed specifically to investigate the properties of the solar core; one of these is the Global Oscillations at Low Frequencies (GOLF) instrument (e.g. Gabriel et al. 1997) on board the Solar and Heliospheric Observatory (SoHO) (e.g. Domingo, Fleck & Poland 1995). A comparison of different data series still shows differences in the inferences in the solar core substantially exceeding the error bars (e.g. Figure 6 of Turck-Chièze et al. 1997). In order to make progress on this point, we propose here a differential study, using GOLF data taken near the solar minimum, to show the influence of three ingredients of the investigation: (a) the asymmetry of the frequency peaks due to the localized nature of the source of the oscillation and its interaction with the noise; (b) the role of the methods used for inverting the data; and (c) the influence of the solar model used (with different prescriptions for the atmosphere) to perform the inversion. The purpose of the paper is to give quantitative estimates of these effects and a better estimate of the uncertainty they may introduce in our determination of the structure of the solar core.
It has been demonstrated that in general the peaks in solar oscillation power spectra are not symmetric (e.g. Duvall et al. 1993; Nigam & Kosovichev 1998; Toutain et al. 1998; Chaplin & Appourchaux 1999). This is believed to be a consequence of the localized nature of the source which drives the oscillations (e.g. Gabriel 1993; Abrams & Kumar 1996; Roxburgh & Vorontsov 1997; Nigam et al. 1998; Rosenthal 1998). Despite the evidence for asymmetry, most analyses of observed solar power spectra involve the fitting of symmetric Lorentzian profiles to the peaks in power. This leads to a systematic error in the inferred frequencies.
While there is no longer any doubt that the peaks are asymmetric and that there is a frequency shift, what is still not clear is whether the shift in the frequencies changes results obtained by inverting the frequencies. Tests by Christensen-Dalsgaard et al. (1998) and Rabello-Soares et al. (1999b) suggest that the frequency shift is a smooth function of frequency, which may possibly be removed while inverting the frequencies. Also, using frequencies determined from $`m`$-averaged spectra obtained the Global Oscillation Network Group (GONG), Basu & Antia (2000) showed that frequencies obtained by using asymmetric fits to the peaks do not significantly change results obtained by inversion. However, Toutain et al. (1998) found that for low-degree data obtained by the Michelson Doppler Imager (MDI) on board SoHO for 679 days of observation, the shift in the frequencies does change the result substantially in the solar core.
In this paper, we use solar data obtained by GOLF to check whether there is a significant change in the inferred solar structure when one shifts from using frequencies from Lorentzian fitted peaks to those from asymmetric profiles. The details of the fitting procedure and the results of the fits have been described by Thiery et al. (1999). We use only the $`l=0`$, 1 and 2 modes from the GOLF data. Since it is essential to have intermediate- and high-degree modes to do a reliable inversion, we use the $`l=`$ 3-250 data obtained from MDI (Schou et al. 1998). We note that only symmetrically fitted frequencies are available for these observations. This introduces an unfortunate inconsistency in our analysis which, however, would most likely increase the error in the inferred sound speed (e.g. Rabello-Soares et al. 1999b).
The rest of the paper is organized as follows: we describe the inversion techniques and solar models used in this work in §2, the results of the inversions are discussed in §3 and our conclusions are stated in §4.
## 2 The inversion technique
Inversions to determine solar structure from solar oscillation frequencies proceed through the linearization of the equation for linear adiabatic oscillations around a known solar model. When the oscillation equation is linearized — under the assumption of hydrostatic equilibrium — the fractional change in the frequency can be related to the fractional changes in the squared sound speed ($`c^2`$) and density ($`\rho `$). Thus,
$$\frac{\delta \omega _i}{\omega _i}=K_{c^2,\rho }^i(r)\frac{\delta c^2}{c^2}(r)dr+K_{\rho ,c^2}^i(r)\frac{\delta \rho }{\rho }(r)dr+\frac{F_{\mathrm{surf}}(\omega _i)}{E_i}$$
(1)
(cf. Dziembowski et al. 1990). Here $`\delta \omega _i`$ is the difference in the frequency $`\omega _i`$ of the $`i`$th mode between the solar data and a reference model. The kernels $`K_{c^2,\rho }^i`$ and $`K_{\rho ,c^2}^i`$ are known functions of the reference model which relate the changes in frequency to the changes in $`c^2`$ and $`\rho `$, respectively; and $`E_i`$ is the inertia of the mode, normalized by the photospheric amplitude of the displacement. The term $`F_{\mathrm{surf}}`$ results from the near-surface differences between the Sun and models because of the difficulty in modeling the outer layers.
### 2.1 The inversion methods
We have used two different methods to invert the frequencies: the Regularised Least Squares (RLS) and the Subtractive Optimally Localized Averages (SOLA) methods.
For RLS inversions, the sound-speed and density differences between the Sun and the reference model are described by a set of basic functions in radius, $`r`$ (in this case splines). The surface term is described as a spline in frequency. The spline coefficients are found by minimizing the difference between the left-hand side of Eq. (1) and the right-hand side expanded in splines, subject to the condition that the resulting $`\delta c^2/c^2`$ is smooth.
The number of knots in radius $`r`$ (a total of 120) and the number of knots in frequency to describe the surface term (25) are determined by the fact that for a proper inversion we need enough knots to ensure that the residuals of the fit are randomly distributed as a function of frequency and lower turning point of the modes for a proper inversion; on the other hand, the condition number of the system of equations (which increases with an increase in the numbers of knots) should be as small as possible to ensure that the system is sufficiently well-conditioned to allow a stable numerical solution (see e.g. Basu & Thompson 1996). The knots were distributed according to the density of turning points of the set of modes along the radius. The trade-off parameter which controls the error in the solution and its smoothness was determined by plotting the so-called L-curve which gives the Tikhonov smoothing term (here the norm of the first derivative of the solution) as a function of $`\chi ^2`$ to find a compromise between a good fit of the data (small $`\chi ^2`$) and a rather smooth physically acceptable solution (Gonczi et al. 1998).
The principle of the SOLA inversion technique (Pijpers & Thompson 1992) is to form linear combinations of Eq. (1) with weights $`d_i(r_0)`$ chosen so as to obtain an average of $`\delta c^2/c^2`$ localized near $`r=r_0`$ while suppressing the contributions from $`\delta \rho /\rho `$, and the near-surface errors, when inverting for $`\delta c^2/c^2`$. In addition, the statistical errors in the combination must be constrained. The result of the inversion is then an average of $`\delta c^2/c^2`$, with a weight determined by the averaging kernel, defined by
$$𝒦(r_0,r)=d_i(r_0)K_{c^2,\rho }^i(r)$$
(2)
and normalized so that $`𝒦(r_0,r)dr=1`$. Details of the implementation were provided by Basu et al. (1996) and a procedure to find the parameters required in the inversion was discussed by Rabello-Soares, Basu & Christensen-Dalsgaard (1999a).
### 2.2 The reference solar models
We have used two reference models for this work. The first model (hereafter referred to as the “Saclay/Nice model”) is an updated calculation of the standard model of Brun, Turck-Chièze & Morel (1998) based on the CESAM code (Morel 1997). Nuclear reaction rates of Adelberger et al. (1998) were used with screening effects from Dzitko et al. (1995). The most recent OPAL opacity tables (Iglesias & Rogers 1996) and the OPAL equation of state (Rogers, Swenson & Iglesias 1996) have been introduced in constructing the model. This model converged at the solar age with the observed abundances of thirteen elements from Grevesse & Noels (1993) and microscopic diffusion of each of these elements was computed using diffusion coefficients from Michaud & Proffitt (1993). A reconstructed atmosphere was deduced from the ATLAS9 atmosphere code of Kurucz (1991). The computation included a pre-main-sequence evolution phase and the model has an age of 4.6 Gyr, including this phase. A more detailed description of the model was given by Brun et al. (1998). Some results of comparisons of this model with others and preliminary comparisons with the Sun were made by Turck-Chièze et al. (1998).
The second model is Model S of Christensen-Dalsgaard et al. (1996). This model is used because many helioseismological results in literature are based on this reference model. This is a standard solar model constructed with the Livermore (OPAL) equation of state (Rogers et al. 1996). For temperatures higher than $`10^4`$ K, an early version of the OPAL opacities was used (Rogers & Iglesias 1992); at lower temperatures, opacities from the tables of Kurucz (1991) were taken. The model incorporates the diffusion of helium and heavy elements below the convection zone. The surface heavy element ratio is $`Z/X=0.0245`$ (Grevesse & Noels 1993). The model has an age of 4.6 Gyr without pre-main-sequence. The model was described in detail by Christensen-Dalsgaard et al. (1996).
## 3 Results
The differences in frequencies obtained by the fitting Lorentzian profiles and those obtained by fitting an asymmetric profile are shown in Fig. 1. Note that the differences are systematic, not random, and hence one could expect that the result of inverting the two sets of frequencies will show systematic differences too.
Figure 2 shows the sound-speed difference between the Saclay/Nice model and the Sun obtained using GOLF data combined with MDI data for $`l3`$. Both sets of GOLF frequencies, i.e., the one obtained by fitting a Lorentzian profile and the one obtained by fitting the asymmetric profile of Thiery et al. (1999), are shown.
The results of the SOLA and RLS inversions agree within errors in most of the Sun. The height of the bump at the base of the convection zone is higher in the case of RLS than SOLA and is a result of slightly different resolutions in the two inversions. For the same error magnification, RLS generally has a better resolution than SOLA, though with the drawback that the RLS averaging kernels have some structure far away from the target radius. There is some difference in the convection zone, which is most probably a reflection of differences in error correlation.
For a given inversion method, the results for the outer layers are almost identical regardless of the data set used. This is expected since the same set of high-degree modes is used in both cases. There are changes in the core when the symmetric data set is replaced by the asymmetric set; however, we can see that they are within the errors, and hence not significant. In Fig. 3 we show some of the averaging kernels in the region of the core. The figure also shows the difference between the averaging kernels obtained for the inversion of the two data sets. We see that there is very little difference between them.
We get similar conclusions using Model S (cf. Fig. 4): the introduction of the asymmetry does not affect significantly the inversion in the core. The results for the outer layers are identical, while the change in the core is very small when data sets are changed. The two models, however, have different sound-speed profiles which warrant some comments. The sound-speed differences between the two models can be attributed completely to the physical inputs of the models. The differences are mainly due to the following three effects: (1) a reestimate of the solar age (about 4.55 Gyr without including the pre-main-sequence) in the Saclay/Nice model which lowers the relative sound-speed difference in the very inner core and slightly increases the peak in the sound-speed difference relative to the Sun below the base of the convection zone, (2) a reestimate of the nuclear reaction rates and the screening effect which has similar consequences (see also the discussion of Turck-Chièze et al. 1998), and (3) the upgrade to the most recent OPAL opacity tables which has little effect in the core but dominates the increase in the sound-speed difference below the convection zone.
Our main result is that frequencies obtained using the asymmetric fits do not change inversion results. This is consistent with the estimates by Christensen-Dalsgaard et al. (1998) of the functional form of frequency shifts caused by asymmetry, based on artificial data, which indicated that such shifts would largely be eliminated together with the surface term in $`F_{\mathrm{surf}}`$. This conclusion was confirmed by the inverse analyses carried out by Rabello-Soares et al. (1999b) of such artificial data which showed that asymmetric fits had a very modest effect on the results of structure inversion. However, the results are strikingly different from those obtained by Toutain et al. (1998) using only MDI data.
The question then arises as to why the MDI results of Toutain et al. are so discrepant. In Fig. 5 we show the inversion of the symmetric MDI set (Schou et al. 1998), as well as the inversion of the low-degree MDI data obtained by fitting an asymmetric profile (Toutain et al. 1998) combined with the $`l3`$ data of the MDI data set; in both cases the reference model was Model S. We see that while the symmetric MDI data give results quite similar to those based on GOLF, the asymmetric MDI data do indeed give quite different results.
The MDI asymmetric set has some modes with much lower errors than their adjacent modes. These modes thus get very large weights in the inversion process and, since there are very few low-degree modes anyway, they can indubitably influence the inversion result. This does indeed seem to be the case. Furthermore, the modes $`l=2,n=6`$ and $`l=2,n=7`$ are suspect because of the fact that they have extremely large residual ($`>10\sigma `$) in the RLS inversions. Removal of these modes reduces the difference in the results between the symmetric and asymmetric MDI sets (cf. Fig. 6). Therefore we find no evidence even from the thus corrected MDI data that frequencies obtained with asymmetric profiles fitted to the peaks in the power spectrum cause significantly different results, compared with frequencies obtained with a Lorentzian fit.
It may be noted here that the $`l=1`$ modes of the GOLF and MDI sets show a fairly large difference at high frequency. However, the errors on the modes are also very high and the difference does not seem to cause substantial difference in the inversion results. There is, however, some remaining difference between the GOLF inversion results and the MDI inversion results. To check whether that is merely an artifact of having different numbers of modes in the different data sets, we have inverted a common set of modes from each data set. The set has $`l=0`$ and $`l=1`$ modes of $`n=13`$ to $`n=25`$, and $`l=2`$ modes with $`n=13`$ to $`n=23`$. The higher-degree modes are from the MDI set as before. The inferred sound-speed differences in the inner parts of the Sun are shown in Fig. 7. Note that the results are actually quite similar for both reference models.
## 4 Conclusions
We find very little evidence that the shift in frequencies between symmetric and asymmetric fits to solar oscillation power spectra changes our inferences concerning the sound speed in the solar core. The changes we see are well within one standard deviation. Larger differences occur as a result of addition of modes to the set, or when using different inversion methods. It should be kept in mind that this conclusion is based on combining symmetrically or asymmetrically fitted low-degree data with higher-degree frequencies obtained from symmetric fits. This introduces an inconsistency in the results based on the asymmetrical fits, which could be significant in view of the fact that the low-degree modes form a very small fraction of the total mode set. On the other hand, an inconsistency of this kind, by introducing a degree-dependent systematic error in the frequencies, would appear likely if anything to increase the effect on the inversions. We note also that Rabello-Soares et al. (1999b) found little effect of asymmetry in inversions of artificial data including asymmetry. Similarly, the analysis by Basu & Antia (2000), including both low- and intermediate-degree asymmetrically fitted modes, indicates that the results obtained in this work will not change; however, it should be kept in mind that Basu & Antia obtained the frequencies from $`m`$-averaged spectra and errors in averaging may affect the results. Thus it is obvious that our study must be repeated when asymmetric fits to peaks of intermediate- and high-degree modes are also available. Of course, when this analysis will be generalized, careful attention will be needed to other sources of distortion, such as those resulting from the effect of the solar cycle on the outer layers which will be considerable for data taken between 1998 and 2003.
This study confirms that there is a dip in the sound-speed difference at about 0.18 $`R_{}`$, which appears independent of the model and inversion procedure used. Even deeper into the core, the difference observed appears more dependent on the details of a few modes of low order which we can determine with high accuracy, and on the use of symmetric or asymmetric profiles in fitting the modes, as well as on the surface treatment in the inversion. In addition, the errors in the inversion results caused by data errors are larger. Knowledge about this region of the Sun, crucial for the neutrino predictions, will be improved by extended data series and coherent treatment of asymmetric fitting of all the modes.
This work utilizes data from GOLF and the Solar Oscillations Investigation / Michelson Doppler Imager (SOI/MDI) on the Solar and Heliospheric Observatory (SoHO). SoHO is a project of international cooperation between ESA and NASA. The work was supported in part by the Danish National Research Foundation through its establishment of the Theoretical Astrophysics Center. |
no-problem/0001/astro-ph0001008.html | ar5iv | text | # 1 Introduction
## 1 Introduction
Gamma-Ray Bursts (GRBs) are bright, transient events in the gamma-ray sky, unpredictable in time and location, with a typical duration of $``$ seconds (for a review, see Fishman 1995; 1999; Fishman & Meegan 1995; Kouveliotou 1995a; Meegan 1998). The brightest bursts have gamma-ray fluences of order 10<sup>-4</sup> erg cm<sup>-2</sup>, strong enough to lead to detectable disturbances of the Earth’s upper atmosphere (Fishman & Inan 1988). Most of the energy of the bursts is released in the 0.1-1 MeV range. Spectra generally display featureless smooth continua (for a review, see Teegarden 1998).
The first GRB was recorded with the Vela satellites on July 2, 1967 (Klebesadel, Strong, & Olson 1973; Strong, Klebesadel, & Olson 1974; for a historical review see Bonnell & Klebesadel 1996). The precise localization of a GRB on the sky has been an unsolved challenge for ground-based astronomy for 30 years, although more than 2000 bursts were detected by numerous space-based experiments during this timespan. This unsatisfactory observational situation led to an enormous flood of publications (see Hurley 1998a), and culminated in more than 100 theories about the nature of the bursters, ranging from solar-wind models to topological defects at cosmological distances (in the early years of GRB research there were in fact more theories than bursts; see Nemiroff 1994; Ruderman 1975). The main reason for the missing identification of the burster population was that prior to 1997 no small, arcmin-sized GRB error boxes were available for rapid follow-up observations.
In the mid 1980’s it was recognized that there is a subclass among the GRBs which are distinguished from the majority of the bursts (Laros et al. 1986, 1987). These bursts have a soft, thermal Bremsstrahlung spectrum ($`kT`$ about 30 keV), short durations, and come from certain, well-localized regions in the sky. These Soft Gamma-Ray Repeaters (SGR) represent a population of objects in the Galaxy and the Large Magellanic Cloud which occasionally emit soft GRBs. This class of objects, currently consisting of four confirmed sources (SGR 0525–66, 1627–41, 1806–20, 1900+14), is not considered here, although great progress was also achieved in SGR research in recent years (for reviews, see Hartmann 1995; Kouveliotou 1995b; Hurley 1999a).
## 2 Steps toward the GRB distance scale
### 2.1 The period 1991 - 1996
The 1990’s have seen two observational breakthroughs in GRB research. The first came with the Burst and Transient Source Experiment (BATSE) on the Compton Gamma-Ray Observatory (CGRO; Fishman 1981; Fishman et al. 1993, 1994). It characterizes the period from 1991 to 1996. During this period several GRB experiments aboard various satellites were carried out, but BATSE was most successful.
In operation since 1991, BATSE detects about 1 burst/day, corresponding to a full sky rate of about 800 a year (Meegan et al. 1992). Based on BATSE observations it became obvious in the early 1990’s that the bursters must be located either deep in a Galactic halo or at cosmological distances (Meegan et al. 1992; for reviews, see Blaes 1994; Dermer & Weiler 1995; Fishman & Meegan 1995; Harding 1994; Hartmann 1994, 1996; Hurley 1994; Lamb 1997; Mészáros 1997). Compared to the observational situation in the 1970’s and 1980’s (for a review, see Higdon & Lingenfelter 1990; Houston & Wolfendale 1983; Puget 1981) this evidence represented substantial progress. However, it also meant that the distance scale of the bursters was still unknown by orders of magnitude, lying either at the kpc level or in the Gpc range (for a detailed study see, e.g., Mao & Paczyński 1992a,b). Correspondingly uncertain was the GRB luminosity distribution. The arguments for and against GRBs inside or outside the Galaxy led to a lively debate, which was eventually presented to the general public (see the dedicated publications of ’The Great Debate’: Fishman 1995; Lamb 1995; Nemiroff 1995; Paczyński 1995; Rees 1995; Trimble 1995).
The most promising way to distinguish between the Galactic halo model and the cosmological model of GRBs seemed to be either to check with high precision the angular isotropy of the bursts on the sky (cf. Briggs et al. 1996; Hakkila et al. 1994; Lamb 1997; Tegmark et al. 1996b), or to search for GRBs and their accompanying optical and X-ray flashes from a potential Galactic halo population of bursters in M 31 (cf. Klose 1995a; Lamb 1995; Li & Liang 1992). Evidence for the latter was not found, but the former method worked very well and based on BATSE gamma-ray data alone over the years it became more and more difficult to postulate a Galactic halo origin of all bursts. This included constraints on burst repetition (cf. Meegan et al. 1995; Tegmark et al. 1996a).
There have been several other proposals to determine the distance scale of the bursters (cf. Paczyński 1991a,b). At gamma-ray energies these investigations involved attempts to detect a cosmological signature in GRB durations and spectral hardness (cf. Brainerd 1994; Fenimore & Bloom 1995; Horack, Mallozzi, & Koshut 1996; Lee & Petrosian 1997; Norris et al. 1994, 1995), studies of luminosity function effects on the observed brightness distribution of the bursts (cf. Fenimore et al. 1993; Horváth, Mészáros, & Mészáros 1996; Mao & Paczyński 1992a; Mészáros & Mészáros 1995, 1996; Mészáros et al. 1996), analyses of the spectral energy distribution of the bursts in combination with pair production opacities in GRB photon fields (cf. Baring & Harding 1998), and theoretical estimates of the energy reservoir of potential Galactic neutron stars as GRB sources (cf. Hartmann & Narayan 1996). In the X-ray band a Galactic origin of the burster population would have been revealed by scattering and absorption of GRB X-rays by the Galactic interstellar medium (Klose 1994, 1995b; Owens, Schaefer, & Sembay 1995). However, this required a dedicated X-ray satellite, which was not available at that time. At optical/infrared/radio wavelengths one way was to perform rapid follow-up searches for transients following or accompanying GRBs, the occurrence of which was predicted in theoretical models (cf. Mészáros, Rees, & Papathanassiou 1994). Another way was to look for a potential GRB source population in GRB error boxes. Both strategies were confronted with the problem that BATSE 1$`\sigma `$ error boxes are several degrees in radius (cf. Briggs et al. 1999). Although this is a very small localization size for a gamma-ray telescope, it is a giant error box for follow-up observations in the radio, optical, or near-infrared bands. Searches for optical transients in GRB error boxes on wide-field photographic plates fortuitously taken at the right time and area of the sky have been one method to address the problem (cf. Greiner et al. 1992; Schaefer 1990; for a review, see Hudec 1993). In general, this method was not very efficient since the chance of finding suitable photographic plates is very small (cf. Greiner et al. 1996), and variable stars could mimic optical transients (cf. Greiner & Motch 1995; Hudec & Wenzel 1996; Klose 1995c; Pedersen 1994). Indeed, these searches did not lead to the unambiguous detection of a GRB source. A more promising way was to set-up automatic wide-field cameras, which could be triggered on-line by GRB detections on board the CGRO. Based on dedicated international distribution networks for GRB data (cf. Barthelmy et al. 1998; Kippen et al. 1998a; McNamara, Harrison, & Williams 1995), response times on the order of seconds were achieved, however no optical transients were discovered with this method either (cf. Akerlof et al. 1995; Lee et al. 1997; Park et al. 1997; Vanderspek, Krimm, & Ricker 1995).
Another technique to observe a burster was based on GRB error boxes provided by the Interplanetary Network (IPN) of satellites operating in the solar system (cf. Laros et al. 1998). For most of the 1990’s the IPN consisted only of Ulysses-spacecraft (cf. Hurley et al. 1995) and CGRO. The IPN makes use of the light travel time delays between different satellites to localize a burst on the sky (cf. Cline et al. 1980). Compared to BATSE-only error boxes, if triggered, the IPN can provide GRB error boxes with considerable smaller sizes (for a review, see Lund 1995), but not in real-time. International campaigns of follow-up observations of the most promising half-dozen well-localized IPN bursts only provided upper limits to the flux density of any transient source in the X-ray, optical, or radio band (cf. Frail et al. 1994; Galama et al. 1997a; Greiner et al. 1997a; Hurley et al. 1994; Laros et al. 1997; Palmer et al. 1995; Pedersen 1994; Schaefer et al. 1994; for a review, see McNamara, Harrison, & Williams 1995). Furthermore, deep surveys of arcmin-sized IPN error boxes have been performed, mostly months or years after the corresponding burst, with the aim of finding evidence for a statistical excess of a certain class of astronomical objects in GRB error boxes. No unambiguous GRB sources were found in the optical (cf. Gorosabel et al. 1995; Schaefer et al. 1997; Vrba, Hartmann, & Jennings 1995; Webber et al. 1995), although in the case of quasars and active galactic nuclei the situation is still somewhat controversial (cf. Burenin et al. 1998; Fruchter et al. 1999b; Hurley et al. 1999a; Luginbuhl et al. 1995, 1996; Schartel, Andernach, & Greiner 1997; Vrba et al. 1994, 1999a). Similar surveys in the radio (cf. Palmer et al. 1995; Schaefer et al. 1989), the X-ray band (cf. Boër et al. 1993, 1997; Greiner et al. 1991; Hurley et al. 1999b; Pizzichini et al. 1986), and the near-infrared (Blaes et al. 1997; Klose, Eislöffel, & Richter 1996; Klose et al. 1998; Larson & McLean 1997; Larson, McLean, & Becklin 1996) also failed to identify the burster population. The deepest optical survey (Vrba, Hartmann, & Jennings 1995) and the near-infrared surveys revealed, however, that there is no lack of potential GRB host galaxies in IPN error boxes. The missing item was the direct observational proof of an extragalactic origin of the bursts. This second observational breakthrough was initiated by the Italian-Dutch gamma-ray/X-ray satellite BeppoSAX launched in 1996 (Boella et al. 1997; Costa et al. 1998; Frontera 1998), and later it included successful GRB localizations with RXTE (see Bradt & Smith 1999; Smith et al. 1999). It characterizes the period from 1997 to 1999. The observational situation before that time is summarized by e.g. Greiner (1995), Lamb (1997), McNamara, Harrison, & Williams (1995), Mészáros (1997), and Vrba (1996).
### 2.2 The period 1997 - 1999
Contrary to BATSE, BeppoSAX can provide an arcmin-sized GRB error box within hours after the GRB trigger, though only about ten times a year because of its small field of view. The first burst detected by BeppoSAX in 1997 was GRB 970111 with an error box of 10 arcmin radius provided to observatories via electronic mail only hours after the occurrence of the burst (Costa et al. 1997a). Although no GRB afterglow was detected either in the optical (cf. Galama et al. 1997b) or in the radio band (Frail et al. 1997a), this burst made it immediately clear that an observational breakthrough was near. It came with GRB 970228 (Costa et al. 1997b,c) when for the first time the optical afterglow of a burst was discovered (Groot et al. 1997; van Paradijs et al. 1997), meaning that the burst was localized with high angular precision. Since then much has been learned about GRBs based on multi-wavelengths observations of afterglows. The establishment of a sophisticated electronic network for GRB messages, the GRB coordinated network GCN (Barthelmy et al. 1998, see appendix), was another milestone on this route. Various aspects of these exciting, new discoveries have been summarized by numerous authors from the theoretical as well as from the observational point of view (e.g., Blinnikov 1999; Castro-Tirado 1999; Frontera 1998; Ghisellini 1999; Greiner 1998; Hartmann 1999; Hurley 1998b; Lamb 1999; McNamara & Harrison 1998; Mészáros 1999a,c; Mészáros, Rees, & Wijers 1999; Paczyński 1999; Piran 1999a,b; Rees 1998, 1999; Stern 1999; Vietri 1999).
## 3 Recent observations of GRB afterglows
The occurrence of broad-band afterglows following GRBs was expected on theoretical grounds (cf. Rees 1998, and references therein). Compared to the duration of the bursts, afterglows in the long-wavelengths bands can be long-lived, making the precise localization of the bursters possible and extending GRB research into international multi-wavelength observing campaigns. Greiner’s WWW-page at the Astrophysical Institute Potsdam (see appendix) provides informations about ongoing observational activities in GRB follow-up studies.
According to the currently most accepted theoretical GRB model, the afterglows are due to external shocks when a relativistically expanding fireball (a $`\gamma `$-ray fireball, see Cavallo & Rees 1978) released by a compact source sweeps up matter from the “interstellar” medium surrounding the burster (for an introduction into this subject and/or a review, see Mészáros 1997, 1999a; Piran 1997, 1999a,b; Rees 1999). This medium could be, for example, the ordinary interstellar medium in a spiral galaxy, or the stellar wind environment from the GRB progenitor (cf. Chevalier & Li 1999; Halpern et al. 1999). The afterglow emission process is most likely synchrotron radiation (see, e.g., Sari, Piran, & Narayan 1998; Wijers, Rees, & Mészáros 1997, and references therein).
At the date of submission of this review, long-lasting optical afterglows have been observed from about a dozen GRBs, and about ten radio afterglows were discovered (see Greiner’s WWW-page \[appendix\]; see also Table 2 in Wheeler 1999). Evidence for a short-lived afterglow in the gamma-ray band has been reported too (Giblin et al. 1999). About half of all GRB afterglows seen in the X-ray band showed no detectable optical emission, whereas almost all GRBs detected by BeppoSAX exhibited an X-ray afterglow (Costa 1999).
To date, there are eleven reported spectroscopic redshift measurements of GRB afterglows and/or host galaxies (mid December 1999; Table 1), confirming the cosmological distance scale of GRBs, which was considered soon after their discovery (e.g., Usov & Chibisov 1975; van den Bergh 1983; Paczyński 1987). The data seem to indicate that the redshift distribution of the bursters peaks around 1, with a long tail towards higher redshifts (for model fits see, e.g., Schmidt 1999). There is also one burst that came from the local universe ($`z`$=0.0085; GRB 980425). The question if all GRBs are of extragalactic nature is not yet completely solved, however (cf. Cline, Matthey, & Otwinowski 1999; Tavani 1998).
Usually the optical transient following a GRB has an $`R`$-band magnitude of about 18…22 when it is detected some hours after the burst, provided that no strong extinction occurs in the GRB host galaxy or in our Galaxy. This can make the optical transient detectable to 1-m class telescopes. For example, the optical afterglow of GRB 970508 was discovered with the Kitt Peak 0.9-m reflector (Bond 1997), the afterglow of GRB 980329 with the Tautenburg 1.34-m (Klose, Meusinger, & Lehmann 1998; Reichart et al. 1999), and the afterglow of GRB 990308 with the Venezuelan 1-m Schmidt telescope (Schaefer et al. 1999). The burst 980329 is the only GRB afterglow detected in the submm-band (Smith et al. 1999). Presumably, there was something special about this burst, which is also indicated by the color of its afterglow (for a discussion see, e.g., Draine 1999; Fruchter 1999; in ’t Zand et al. 1998; Palazzi et al. 1998). Detailed color measurements of afterglows are published for, e.g., GRB 980703 (Vreeswijk et al. 1999) and 990510 (Israel et al. 1999).
The time-dependent flux density of an afterglow follows a power-law decay, $`F_\nu (t)t^\beta `$, in accordance with fireball models (cf. Sari, Piran, & Narayan 1998; Wijers, Rees, & Mészáros 1997). For example, a decay constant in the optical of $`\beta =1.2`$ leads to a dimming of the optical transient by 3 photometric magnitudes between day 1 and day 10 after the occurrence of the burst. Rapid follow-up observations are therefore crucial for the detection of GRB afterglows. Exceptions to this rule do occur, however. The afterglow flux of GRB 970508 showed a fall and rise during the first 2 days, before it entered a power-law decline (cf. Castro-Tirado et al. 1997; Fruchter et al. 1999b; Pedersen et al. 1998; for a possible theoretical explanation, see Panaitescu, Mészáros, & Rees 1998; Pugliese, Falcke, & Biermann 1999).
In its early phase a GRB afterglow can outshine its host galaxy, which becomes visible weeks or months after the burst when the light curve of the afterglow seems to flatten. This was first seen for GRB 970508 (Zharikov & Sokolov 1999). This was also the first burst where a radio counterpart of the afterglow was detected with the Very Large Array (VLA) which suggested extreme relativistic expansion velocities of the radio source (Frail et al. 1997b; Waxman, Kulkarni, & Frail 1998) based on the radio scintillation technique (Goodman 1997). Meanwhile, it has been proven that observations in the radio band can localize GRB afterglows in BeppoSAX error boxes when optical identifications are still missing (Taylor et al. 1998a,b, in case of GRB 980329; Frail et al. 1999a, in case of GRB 991208) and even when they do not exist at all (Frail 1999, in case of GRB 970828; Frail et al. 1999b, in case of GRB 981226).
## 4 Afterglows and GRB energetics
Perhaps the most impressive property of the bursts is the huge amount of electromagnetic energy released (Table 1). This is especially apparent in the case of GRB 990123, where an optical flash was detected for the first time when the burst was still in progress in the gamma-ray band (Akerlof et al. 1999). This phenomenon was predicted from theory (Mészáros 1999a, and references therein; Sari & Piran 1999, and references therein; for a phenomenological approach, see Ford & Band 1996). The optical flash peaked about 45 sec after the onset of the burst at a mean $`V`$-band magnitude of about 9 on a frame with a 5 sec exposure (Akerlof et al. 1999). The (minimum) redshift of the burster was found to be 1.60 (Andersen et al. 1999; Kulkarni et al. 1999), so that the optical flash translates into an ultraviolet rest-frame luminosity of about $`3\times \mathrm{\hspace{0.17em}10}^{16}L_{}`$ (Kulkarni et al. 1999). If the burster had been at a distance corresponding to that of M 81, its $`V`$-band magnitude would have reached –9 (a cosmological $`K`$-correction neglected). If it had been located at the Galactic center, but assuming no extinction, it would have peaked at a $`V`$-band magnitude of –22, comparable to the apparent magnitude of the Sun. Assuming isotropic emission, the gamma-ray energy emitted by this burst amounts to $`3.4\times \mathrm{\hspace{0.17em}10}^{54}`$ erg (Table 1), about twice the rest mass energy of the Sun and an order of magnitude higher than the previous record holder, GRB 971214 (Kulkarni et al. 1998a). The term ’hypernova’ (Paczyński 1998a,b) has become a popular word to describe such energetic events.
Two basic ideas are being discussed on how to reduce the required energy budget; gravitational lensing and non-isotropic emission. This discussion peaked when GRB 990123 was detected and its redshift measured (see the archive of the GCN circulars; appendix). Direct observational support for lensing has never been found in GRB data (cf. Marani et al. 1999) but evidence for beaming seems to be present. Beaming reduces the energy requirement by a factor of order 10–100, depending on the assumed or calculated solid angle of emission, but it increases by the same factor the required GRB event rate.
It was recently predicted that beaming of the relativistic outflow should lead to a break in the afterglow light curve (Mészáros & Rees 1999b; Panaitescu, Mészáros, & Rees 1998; Piran 1999b; Rhoads 1997, 1999; Sari, Piran, & Halpern 1999) and to a non-zero and time-dependent polarization in the optical/near-infrared (Ghisellini & Lazzati 1999; Gruzinov 1999; Sari 1999). This is expected to occur when the decreasing bulk Lorentz-factor of the radiating shock front which runs into the ambient interstellar medium of the burster allows the observer to see the edge of the jet (Piran \[1999b\] notes that ’jet’ is not the appropriate description of the phenomenon, since this is a transient collimated outflow). The smaller the opening angle of the jet, the earlier the observer should see a break, a steepening in the afterglow light curve. Within this theoretical context it would be interesting to learn if a time-delayed signal from the counterjet is expected to be seen too.
Observational evidence for beaming was first found in the optical afterglow of GRB 990123, where the light curve steepened two days after the burst (Castro-Tirado et al. 1999; Kulkarni et al. 1999). The measured steepening is in agreement with the picture that at this time the observer began to see the edge of a jet. Recently, a steepening in the light curve has also been observed for another burst, GRB 990510 (Beuermann et al. 1999; Harrison et al. 1999; Stanek et al. 1999). In both cases a beaming factor in the order of 200…300 has been deduced from the observations (Harrison et al. 1999; Mészáros 1999b; Sari, Piran, & Halpern 1999). This seems to bring very energetic bursts to the energy output of less energetic ones (Table 1), for which no observational evidence for strong beaming has been found, like GRB 970228 and 970508 (Sari, Piran, & Halpern 1999). More afterglow observations are required, however, to check this hypothesis.
An exciting consequence of beaming is that there could exist GRBs which develop an X-ray, optical, or radio afterglow, but have no detected gamma-ray burst (for a discussion see, e.g., Mészáros, Rees, & Wijers 1998, 1999; Perna & Loeb 1998b; Rhoads 1997). Archived X-ray data have been searched for such events, but no strong evidence for them was found (Greiner et al. 1999a,b; Grindlay 1999). However, possible evidence for a relation of optical transients detected on photographic plates to underlying blue galaxies was reported by Hudec et al. (1996). Results of wide-field CCD surveys to search for this phenomenon have not yet been published.
There are numerous possible causes for a break in the light curve of an afterglow (for a discussion see, e.g., Kulkarni et al. 1999; Wei & Lu 1999). An independent observational test for beaming is therefore desirable. It could be provided by polarimetric observations. According to recent theoretical studies, optical afterglows could be linearly polarized up to the 10% level if the radiation comes from a collimated outflow moving with relativistic velocity toward the observer and the observer is not directed exactly at the center of the jet (Ghisellini & Lazzati 1999; Gruzinov 1999; Sari 1999; see also Hughes, Aller, & Aller 1985). However, microlensing is also considered as a possible option for producing time-dependent linearly polarized afterglows (Loeb & Perna 1998a). If the degree of linear polarization of an afterglow can be as high as predicted, even if no beaming occurs (Gruzinov & Waxman 1999), this could allow observers to detect an afterglow in a GRB error box at high Galactic latitude without the need of second-epoch data. This holds particularly in the near-infrared bands (Klose, Stecklum, & Fischer 1999).
Linear polarization was first detected at the 2% level in the afterglow of GRB 990510 about 1…2 days after the burst based on observations with the ESO Very Large Telescope (Covino et al. 1999; Wijers et al. 1999). Unfortunately, only three data points could be obtained, and within the measurement errors no time-dependency of the degree of linear polarization was found. However, such a time-dependence appears to be present in the afterglow of GRB 990712 (Rol et al. 1999).
## 5 What is the nature of the bursters?
Among the most exciting discoveries about the nature of the GRBs is the bimodality of their duration distribution (Kouveliotou et al. 1993; Mazets et al. 1981; McBreen et al. 1994). Recently, evidence for an intermediate class of bursts was also reported (Horváth 1998; Mukherjee et al. 1998). Short burst durations range between 0.01 to 2 seconds, long bursts last from about 2 to a few hundred seconds. Since BeppoSAX is only sensitive to long bursts, it is currently unknown whether short bursts do also produce detectable GRB afterglows. It is also unknown what the origin of this bimodality is although various studies have been performed to gain insights into this issue (cf. Balázs, Mészáros, & Horváth 1998; Balázs et al. 1999; Dezalay et al. 1996; Katz & Canel 1996; Mao, Narayan, & Piran 1994; Mészáros, Bagoly, & Vavrek 1999; Mitrofanov 1998).
One working hypothesis is that the bimodal duration distribution reflects different GRB engines. According to the most accepted picture of the bursters today, GRBs can be made either by merger events that include compact stars and/or stellar-mass black holes, or by the gravitational collapse of single stars (cf. Fryer & Woosley 1998; Fryer, Woosley, & Hartmann 1999; Fryer et al. 1999; in ’t Zand 1998; Ruffert & Janka 1998, 1999; Woosley, MacFadyen, & Heger 1999; for a discussion see also, e.g., Mészáros 1999a; Rees 1999). In all cases it is believed that the relic of the explosion is a stellar-mass black hole. In other words, within this picture every detected GRB (excluding SGR bursts) represents either a black hole formation event in the universe, or at least a signal from a pre-existing stellar-mass black hole.
Although the central engine of a GRB is hidden from observation, both classes of burst models make certain predictions that can be tested by observations, mainly based on the ages of the objects involved (cf. Bloom, Sigurdsson, & Pols 1999; Fryer, Woosley, & Hartmann 1999). As noted by Paczyński (1998b), massive stars will be close to their birthplaces when they explode and hence they could be embedded in a dust-rich environment. Several pieces of evidence seem to favor this GRB model: First, within the context of the fireball models, multi-wavelength observations of afterglows can be used to determine the gas density of the “interstellar” medium surrounding the burster (cf. Vreeswijk et al. 1999; Wijers & Galama 1999). Model fits led to the conclusion that the density of the external medium into which the fireball of GRB 980329 expands (see section 3) is $``$1000 cm<sup>-3</sup> (Lamb et al. 1999). In other words, this burst presumably occurred in an interstellar gas cloud in a remote galaxy. (In contrary cases, e.g. GRB 970508, an ambient gas density of about 1 cm<sup>-3</sup> or less was deduced from the afterglow data \[Waxman 1997; Wijers & Galama 1999\].) Second, only about 50% of all GRBs whose X-ray afterglows were detected have been discovered in optical bands. This could be due to extinction by dust in the GRB host galaxies. An example of an optically obscured burst is GRB 970828. The burst developed a bright X-ray afterglow detected by RXTE (Marshall et al. 1997; Remillard et al. 1997), ASCA (Murakami et al. 1997), and ROSAT (Greiner et al. 1997b), but it was not seen in the optical down to an $`R`$-band magnitude of 24 (Groot et al. 1998), although follow-up observations started only hours after the burst.<sup>2</sup><sup>2</sup>2The only near-infrared follow-up observations of this event were reported by Klose, Eislöffel, & Stecklum (1997), but they were performed 13 days after the burst. A red object was detected by these authors in the 30<sup>′′</sup> ASCA X-ray error circle (Murakami et al. 1997) which was later found to be also located in the 10<sup>′′</sup> ROSAT HRI error circle (Greiner et al. 1997b). However, second epoch observations five months later showed this object again (Klose 1998, unpublished). So, it is not the GRB afterglow, despite it having an underlying faint galaxy (Pedersen 1997, private communication). Third, the GRB hosts detected so far are actively star-forming galaxies (Fruchter et al. 1999a). Their star-formation rates range between about 1 to 10 $`M_{}`$ yr<sup>-1</sup> (cf. Bloom et al. 1998b,1999b; Djorgovski et al. 1998; Kulkarni et al. 1998a; Odewahn et al. 1998), although this is not unusual for galaxies at these redshifts. There is presently no evidence for GRBs occurring in elliptical galaxies.
A number of future tests of the GRB environment have been proposed in the literature. For example, the gaseous component of the GRB environment could manifest itself by certain spectral features in the X-ray afterglow (cf. Böttcher et al. 1999; Lazzati, Campana, & Ghisellini 1999; Mészáros & Rees 1999a; Piro et al. 1999b) or in the optical afterglow (cf. Perna & Loeb 1998a), whereas a test of the dusty component could make use of the scattering properties of the dust grains in the soft X-ray band (Klose 1998).
## 6 Evidence for a GRB - Supernova association
The idea of a GRB-Supernova (SN) association was first considered by Colgate (1968, 1974; see also the review by Chupp 1976) but it was not supported by observations at that time. About 30 years later it turned out that a GRB-SN link does indeed exist.
GRB 980425 was detected by BeppoSAX and an anomalous SN (1998bw) was found in the 8 arcmin error circle of the X-ray afterglow (Galama et al. 1998; Kulkarni et al. 1998b). The SN spectrum indicated ejection speeds approaching 60 000 km s<sup>-1</sup> (Kulkarni et al. 1998b). SN 1998bw is classified as a peculiar Type Ic supernova. Its host galaxy is at a redshift of 0.0085 (Tinney et al. 1998), corresponding to a distance of 38 Mpc ($`H_0`$ = 65 km s<sup>-1</sup> Mpc<sup>-1</sup>). The gravitational collapse of a massive stellar iron core that formed a black hole can explain the observations (Iwamoto et al. 1998; Woosley, Eastman, & Schmidt 1999; Woosley, MacFadyen, & Heger 1999). The radio-to-X-ray light curves of this event is not in conflict with the fireball model (Iwamoto 1999; for the optical light curve see McKenzie & Schaefer 1999). Since GRB 980425 was so close to our Galaxy, but not exceptionally bright, one has to conclude that BATSE does only see the ’tip of the iceberg’ of all GRBs occurring in the universe.
Although up to December 1999 no other direct GRB-SN association has been found, there is in fact increasing observational evidence that at least a subclass of all bursts is associated with SNe. The strongest evidence comes from the recent discovery that the light curves of GRB 970228, 980326, and possibly 990712, show a bump at late times which cannot be explained within the context of simple afterglow models, but can be fitted as a redshifted SN 1998bw light curve which adds to the flux of the GRB afterglow (Bloom et al. 1999a; Dar 1999; Galama et al. 1999; Hjorth et al. 1999; Reichart 1999). If that interpretation is correct, then GRB 970228, the first burst with a detected optical afterglow, represents one of the most distant SNe ever seen ($`z`$=0.695; Djorgovski et al. 1999). Moreover, if it turns out that SN light curves do appear in GRB afterglows, this could represent a new powerful method to measure the cosmological parameters. A check of this potential application might be an interesting long-term project for 8-m class optical telescopes.
It has recently been proposed that SN 1997cy (Germany et al. 1999) and SN 1999E (Kulkarni & Frail 1999; Thorsett & Hogg 1999), which seem to resemble SN 1998bw, could be related to the BATSE bursts 970514 and 980910a, respectively, which were not seen by BeppoSAX (for a discussion, see Wheeler 1999). It has also been suggested that there is observational evidence for a relativistic jet from SN 1987A (Cen 1999). Finally, the detection of a redshifted iron K<sub>α</sub> emission line in the X-ray afterglows of GRB 970508 (Piro et al. 1999a,b) and possibly GRB 970828 (Yoshida et al. 1999) has been reported, which could indicate an SN event before the corresponding GRB (Lazzati, Campana, & Ghisellini 1999; Vietri & Stella 1998). Naturally, because of these findings the question arises if all GRBs are physically related to SNe. Several authors have therefore compared the BATSE GRB catalog (Meegan et al. 1998) with SN catalogs (Bloom et al. 1998a; Hudec, Hudcova, & Hroch 1999; Kippen et al. 1998b; Klose 1999; Norris, Bonnell, & Watanabe 1999; Wang & Wheeler 1998). However, no convincing statistical evidence for an excess of any subclass of SNe in GRB error boxes has been found in that way. The reason for this is that BATSE-only error boxes are very large in radius and that the SN database is very inhomogeneous. Therefore, these studies cannot exclude the existence of a general association of GRBs to SNe. In any case, GRB research gives a strong impulse to SN research (Wheeler 1999; Woosley, MacFadyen, & Heger 1999). If GRBs do indeed reflect certain SN explosions, BATSE detects more SNe per year than all current SN search campaigns combined.
Based on these findings, is it possible that we have already seen a few SN-GRBs in our Galactic neighborhood but not recognized them as such? A search for potential GRB afterglows in the CBAT SN catalog has not been successful (Klose 1999), although such a study is confronted with the lack of detailed observational data about most known SNe. Another way to tackle the question under consideration is to look for bright, nearby galaxies in arcmin-sized IPN error boxes. A working hypothesis can be that a SN-GRB association was not recognized in these cases because these GRB error boxes could not be determined in a timely manner, i.e., any SN light curve was already below the detection threshold when these error boxes were imaged in the optical. A visual inspection of published optical images of IPN error boxes of the 1970’s (cf. Schaefer et al. 1998; Vrba, Hartmann, & Jennings 1995) reveals one potential candidate for such a case, GRB 781104b.<sup>3</sup><sup>3</sup>3Schaefer (1999) made a similar suggestion at the 5th Huntsville Symposium on GRBs, Huntsville, AL, October 1999. The only 14 arcmin<sup>2</sup> large error box of this burst contains a relatively bright galaxy ($`B`$=15, Simbad data base). This galaxy, MCG 04-47-011, is at a redshift of $`z`$ = 0.0024 (Simbad data base), corresponding to a distance of about 11 Mpc ($`H_0`$ = 65 km s<sup>-1</sup> Mpc<sup>-1</sup>). The latter estimate is uncertain however, because the redshift is so small. Similar to the duration of GRB 980425 (Galama et al. 1998a; Pian et al. 1999), the duration of GRB 781104b was about 30 seconds (Mazets et al. 1981; Teegarden & Cline 1980). Although no SN is known in MCG 04-47-011, another way to check whether GRB 781104b was a nearby SN burst could be to search in the radio band for its potential GRB remnant.
A number of authors has recently addressed the question of how the relics of GRB explosions might look and where they are in our local universe (e.g., Hansen 1999). At a GRB rate in the order of 1 per 10<sup>6±1</sup> yr per Milky-Way like galaxy, GRB remnants in the local universe could possibly be detectable. Woods & Loeb (1999) proposed to search for radio emission from GRB remnants in the Virgo cluster of galaxies ($`z`$=0.0038; Simbad data base). This would require surveying on the order of 1000 galaxies down to a flux level below 100 $`\mu `$Jy. Efremov, Elmegreen, & Hodge (1998) as well as Loeb & Perna (1998b) brought attention to HI supershells (cf. Lee & Irwin 1997; Walter et al. 1998), whose energy content is comparable to the electromagnetic energy release of GRBs. Potential X-ray selected GRB remnant candidates in M 101 have also been proposed (Wang 1999). First calculations of the emission spectrum of GRB remnants have been published and phenomenological differences to supernova remnants (SNRs) have been outlined (Perna, Raymond, & Loeb 1999). Since there are some similarities to SNRs, however, in the optical the search strategy for GRB remnant candidates could follow those for HII-regions/SNRs (cf. Elmegreen & Salzer 1999; Gordon et al. 1993; Matonick & Fesen 1997). The close-by, face-on spirals M 33 and NGC 300 could be potential targets for such an investigation.
## 7 The future of GRB research
BeppoSAX currently detects about 10 X-ray afterglows of GRBs per year. The HETE-2 satellite (Vanderspek et al. 1999), scheduled to be launched early 2000, will increase this rate by a factor of 3 to 5, and the Swift GRB mission (Gehrels 1999), scheduled to fly in 2003…2005, will provide about 100 or more very small GRB error boxes per year. This, combined with the next generation high-energy satellites (XMM, AXAF-Chandra, ASTRO-E, Integral), will guarantee a large increase in our knowledge about GRBs in the coming decade (see also Hurley 1999b).
For optical telescopes there are three observing strategies: 1) Imaging of GRB error boxes when the burster is still active in the gamma-ray band. GRB 990123 has already demonstrated that this is feasible with current CCD cameras. Great progress will be achieved when automatic 1-m class telescopes will become available, like the Flagstaff 1.3-m (Vrba et al. 1999b) and the Super-Lotis 0.6-m (Park et al. 1999). Prism spectroscopy is another future option. 2) Investigation of GRB afterglows. In most cases this will require ’Target of Opportunity’ observations scheduled within hours or days after a burst. 3) Study of GRB host galaxies. This can be performed even months after a burst. It requires the largest telescopes.
Some open questions for the coming decade are obvious: Are all GRBs related to SNe? What is the physical difference between the short and the long bursts? How many distinct populations of bursters are there? Do GRBs occur in elliptical galaxies? Do short bursts develop afterglows? Does any subclass of GRBs represent a cosmological standard candle? Can we measure GRB redshifts based on observations in the gamma-ray band alone? What are the most distant GRBs? Where are the GRB remnants in the local universe, in our Galaxy, and in M 31? Finally, do GRBs affect the evolution of life?
GRBs are the most energetic electromagnetic phenomena in the universe. Although for an individual GRB this holds only for a very small timespan, it can make GRBs a giant observational tool to investigate the high-$`z`$ universe (cf. Lamb & Reichart 1999). This includes the cosmological parameters (cf. Holz, Miller, & Quashnock 1999; Horack et al. 1996; Marani et al. 1999), the cosmic radiation background (Mannheim, Hartmann, & Funk 1996), stellar evolution, the cosmic star formation rate (Hartmann & Band 1998; Jorgensen et al. 1995; Krumholz, Thorsett, & Harrison 1998; Mao & Mo 1998; Totani 1999; Wijers et al. 1998;), large-scale structure (Hartmann & Blumenthal 1989; Lamb & Quashnock 1993), etc. GRBs could also prove to be a useful tool to investigate the interstellar medium in our Galaxy (and in intergalactic space) through which they are observed. Thus, there are good reasons to believe that the future of GRB research is bright, in all bands of the electromagnetic spectrum.
Appendix: selected WWW-pages about GRBs
* GRB coordinated network GCN: http://gcn.gsfc.nasa.gov/gcn/
* GRB missions: Holger Pedersen’s WWW-page at
http://www.astro.ku.dk/holger/ddirGAMMA/dF/OFSAT.html
* SGRs: Robert Duncan’s WWW-page at the University of Texas Austin,
http://solomon.as.utexas.edu/duncan/
* IPN: Kevin Hurley’s WWW-page at Berkeley,
http://ssl.berkeley.edu/ipn3/interpla.html
* GRB bibliography: Kevin Hurley’s WWW-page at Berkeley,
http://ssl.berkeley.edu/ipn3/bibliogr.html
* reports about GRB follow-up observations: Jochen Greiner’s WWW-page at the AIP, http://www.aip.de/jcg/grbgen.html
* BATSE: http://gammaray.msfc.nasa.gov/batse/
* BeppoSAX: http://www.sdc.asi.it/
* HETE-2 satellite: http://space.mit.edu/HETE/
* Swift mission: http://swift.gsfc.nasa.gov/
Acknowledgements
I am deeply indebted to Bringfried Stecklum (Thüringer Landessternwarte Tautenburg) and Olaf Fischer (Universitäts-Sternwarte Jena), who share my interest in GRBs and their afterglows. I am grateful to Jochen Greiner (Astrophysical Institute Potsdam), Dieter Hartmann (Clemson University, Clemson, SC), Attila Mészáros (Astronomical Institute, Charles University Prague), Bringfried Stecklum, and Fred Vrba (U.S. Naval Observatory, Flagstaff, AZ) for many valuable comments on the manuscript. I thank Josef Solf for his interest in and support of GRB research during his tenure as director of the Thüringer Landessternwarte Tautenburg in the years 1994-99. I greatly appreciate assistance during observing campaigns at the Calar Alto Observatory, Spain, from Kurt Birkle, Ulrich Thiele, and Markus Feldt, Max-Planck-Institut für Astronomie, Heidelberg. This research has made use of the Simbad database, operated at CDS, Strasbourg, France.
References
Akerlof, C. et al., 1995, Astroph. Sp. Sci. 231, 255
Akerlof, C. et al., 1999, Nature 398, 400
Andersen, M. I. et al., 1999, Science 283, 2075
Balázs, L. G., Mészáros, A., & Horváth, I., 1998, A&A 339, 1
Balázs, L. G. et al., 1999, A&A Suppl. Ser. 138, 417
Baring, M. G. & Harding, A. K., 1998, Adv. Space Res. 22 (7), 1115
Barthelmy, S. D. et al., 1998, in: C. A. Meegan, R. D. Preece, & T. M. Koshut (eds.), 4th Huntsville Symposium on Gamma-Ray Bursts (AIP: New York), AIP Conf. Proc. 428, 99
Beuermann, K. et al., 1999, A&A 352, L26
Blaes, O. M., 1994, ApJ Suppl. Ser. 92, 643
Blaes, O. et al., 1997, ApJ 479, 868
Blinnikov, S., 1999, preprint astro-ph/9911138
Bloom, J. S., Sigurdsson, S., & Pols, O. R., 1999, MNRAS 305, 763
Bloom, J. S. et al., 1998a, ApJ 506, L105
Bloom, J. S. et al., 1998b, ApJ 507, L25
Bloom, J. S. et al., 1999a, Nature 401, 453
Bloom, J. S. et al., 1999b, ApJ 518, L1
Boella, G. et al., 1997, A&A Suppl. Ser. 122, 299
Boër, M. et al., 1993, A&A 277, 503
Boër, M. et al., 1997, ApJ 481, L39
Bond, H. E., 1997, IAU Circ. 6654
Bonnell, J. T. & Klebesadel, R. W., 1996, in: C. Kouveliotou, M. F. Briggs, & G. J. Fishman (eds.), 3rd Huntsville Symposium on Gamma-Ray Bursts (AIP: New York), AIP Conf. Proc. 384, 977
Böttcher, M. et al., 1999, A&A 343, 111
Bradt, H. V. & Smith, D. A., 1999, A&A Suppl. Ser. 138, 423
Brainerd, J. J., 1994, ApJ 428, L1
Briggs, M. et al., 1996, ApJ 459, 40
Briggs, M. et al., 1999, ApJ Suppl. Ser. 122, 503
Burenin, R. A. et al., 1998, Astron. Lett. 24, 427; astro-ph/9804274
Castro-Tirado, A. J., 1999, Astroph. Sp. Sci. 263, 15; astro-ph/9903187
Castro-Tirado, A. J. et al., 1997, Science 279, 1011
Castro-Tirado, A. J. et al., 1999, Science 283, 2069
Cavallo, G. & Rees, M. J., 1978, MNRAS 183, 359
Cen, R., 1999, ApJ 524, L51
Chevalier, R. A. & Li, Z.-Y., 1999, preprint astro-ph/9908272
Chupp, E. L., 1976, Gamma Ray Astronomy (Reidel: Dordrecht, Boston)
Cline, D. B., Matthey, C., & Otwinowski, S., 1999, preprint astro-ph/9905346
Cline, T. L. et al., 1980, ApJ 237, L1
Colgate, S. A., 1968, Canadian J. Phys. 46, S476
Colgate, S. A., 1974, ApJ 187, 333
Costa, E., 1999, talk given at the 5th Huntsville Symposium on Gamma-Ray Bursts, AIP Conf. Proc., to be published
Costa, E. et al., 1997a, IAU Circ. 6533
Costa, E. et al., 1997b, IAU Circ. 6572
Costa, E. et al., 1997c, IAU Circ. 6576
Costa, E. et al., 1998, Adv. Space Res. 22 (7), 1129
Covino, S. et al., 1999, A&A 348, L1
Dar, A., 1999, GCN 346
Dermer, C. D. & Weiler, T. J., 1995, Astroph. Sp. Sci. 231, 377
Dezalay, J. P. et al., 1996, ApJ 471, L27
Djorgovski, S. G. et al., 1998, ApJ 508, L17
Djorgovski, S. G. et al., 1999, GCN 289
Draine, B. T., 1999, ApJ, submitted, preprint astro-ph/9907232
Efremov, Y. N., Elmegreen, B. G., & Hodge, P. W., 1998, ApJ 501, L163
Elmegreen, D. M. & Salzer, J. J., 1999, ApJ Suppl. Ser. 117, 764
Fenimore, E. E. & Bloom, J. S., 1995, ApJ 453, 25
Fenimore, E. E. et al., 1993, Nature 366, 40
Fishman, G. J., 1981, Astroph. Sp. Sci. 75, 125
Fishman, G. J., 1995, PASP 107, 1145
Fishman, G. J., 1999, A&A Suppl. Ser. 138, 395
Fishman, G. J. & Inan, U. S., 1988, Nature 331, 418
Fishman, G. J. & Meegan, C. A., 1995, Annu. Rev. Astron. Astrophys. 33, 415
Fishman, G. J. et al., 1993, A&A Suppl. Ser. 97, 17
Fishman, G. J. et al., 1994, ApJ Suppl. Ser. 92, 229
Ford, L. A. & Band, D. L., 1996, ApJ 473, 1013
Frail, D. A., 1999, talk given at the 5th Huntsville Symposium on Gamma-Ray Bursts, AIP Conf. Proc., to be published
Frail, D. A. et al., 1994, ApJ 437, L43
Frail, D. A. et al., 1997a, ApJ 483, L91
Frail, D. A. et al., 1997b, Nature 389, 261
Frail, D. A. et al., 1999a, GCN 451
Frail, D. A. et al., 1999b, ApJ 525, L81
Frontera, F., 1998, in: M. S. Potgieter, B. C. Raubenheimer, & D. J. van der Wait (eds.), 25th International Cosmic Ray Conference, Vol. 8, (World Scientific: Singapore), p. 307; astro-ph/9802157
Fruchter, A., 1999, ApJ 512, L1
Fruchter, A. et al., 1999a, ApJ 516, 683
Fruchter, A. et al., 1999b, preprint astro-ph/9903236
Fryer, C. L. & Woosley, S. E., 1998, ApJ 502, L9
Fryer, C. L., Woosley, S. E., & Hartmann, D. H., 1999, ApJ 526, 152
Fryer, C. L. et al., 1999, ApJ 520, 650
Galama, T. J. et al., 1997a, A&A 321, 229
Galama, T. J. et al., 1997b, ApJ 486, L5
Galama, T. J. et al., 1998, Nature 395, 670
Galama, T. J. et al., 1999, preprint astro-ph/9907264
Gehrels, N. 1999, talk given at the 5th Huntsville Symposium on Gamma-Ray Bursts, AIP Conf. Proc., to be published
Germany, L. M. et al., 1999, preprint astro-ph/9906096
Ghisellini, G., 1999, preprint astro-ph/9907376
Ghisellini, G. & Lazzati, D., 1999, MNRAS 309, L7
Giblin, T. W. et al., 1999, ApJ 524, L47
Goodman, J., 1997, New Astron. 2, 449
Gordon, S. M. et al., 1993, ApJ 418, 743
Gorosabel, J. et al., 1995, Astroph. Sp. Sci. 231, 297
Greiner, J., 1995, Astroph. Sp. Sci. 231, 263
Greiner, J., 1998, in: Proc. Multifrequency behaviour of high energy cosmic sources, Vulcano, Mem. Societa Astron. Italiana, in press; preprint astro-ph/9802222
Greiner, J. & Motch, C., 1995, A&A 294, 177
Greiner, J. et al., 1991, in: Proc. 22nd International Cosmic Ray Conference (Reprint Ltd: Dublin), ISBN 1 85500 995 1, Vol. 1, 53
Greiner, J. et al., 1992, in: W. Paciesas & G. Fishman (eds.), Gamma-Ray Bursts (AIP: New York), AIP Conf. Proc. 265, 327
Greiner, J. et al., 1996, in: C. Kouveliotou, M. F. Briggs, & G. J. Fishman (eds.), 3rd Huntsville Symposium on Gamma-Ray Bursts (AIP: New York), AIP Conf. Proc. 384, 622
Greiner, J. et al., 1997a, A&A 325, 640
Greiner, J. et al., 1997b, IAU Circ. 6757
Greiner, J. et al., 1999a, A&A Suppl. Ser. 138, 441
Greiner, J. et al., 1999b, preprint astro-ph/9910300
Grindlay, J. E. 1999, ApJ 510, 710
Groot, P. J. et al., 1997, IAU Circ. 6584
Groot, P. J. et al., 1998, ApJ 493, L27
Gruzinov, A., 1999, ApJ 525, L29
Gruzinov, A. & Waxman, E., 1999, ApJ 511, 852
Hakkila, J. et al., 1994, ApJ 422, 659
Halpern, J. P. et al., 1999, ApJ 517, L105
Hansen, B. M., 1999, ApJ 512, L117
Harding, A. K., 1994, ApJ Suppl. Ser. 90, 863
Harrison, F. A. et al. 1999, ApJ 523, L121
Hartmann, D. H., 1994, Science 263, 47
Hartmann, D. H., 1995, A&A Rev. 6, 225
Hartmann, D. H., 1996, A&A Suppl. Ser. 120, 31
Hartmann, D. H., 1999, Perspectives of the National Acad. of Sciences (PNAS) 96, 4752
Hartmann, D. H. & Band, D. L., 1998, in: C. A. Meegan, R. D. Preece, & T. M. Koshut (eds.), 4th Huntsville Symposium on Gamma-Ray Bursts (AIP: New York), AIP Conf. Proc. 428, 615
Hartmann, D. H. & Blumenthal, G. R., 1989, ApJ 342, 521
Hartmann, D. H. & Narayan, R., 1996, ApJ 464, 226
Higdon, J. C. & Lingenfelter, R. E., 1990, Annu. Rev. Astron. Astrophys. 28, 401
Hjorth, J. et al., 1999, GCN 403
Holz, D. E., Miller, M. C., & Quashnock, J. M., 1999, ApJ 510, 54
Horack, J. M., Mallozzi, R. S., & Koshut, T. M., 1996, ApJ 466, 21
Horack, J. M. et al., 1996, ApJ 472, 25
Horváth, I., 1998, ApJ 508, 757
Horváth, I., Mészáros, P., & Mészáros, A., 1996, ApJ 470, 56
Houston, B. P. & Wolfendale, A. W., 1983, Vistas Astron. 26, 107
Hudec, R., 1993, A&A Suppl. Ser. 97, 49
Hudec, R. & Wenzel, W., 1996, A&A Suppl. Ser. 120, 707
Hudec, R., Hudcova, V., & Hroch, F. 1999, 138, 475
Hudec, R. et al., 1996, in: H. R. Miller, J. R. Webb, & J. C. Noble (eds.), Blazar Continuum Variability (ASP: San Francisco), ASP Conf. Ser. 110, 129
Hughes, P. A., Aller, H. D., & Aller, M. F., 1985, ApJ 298, 301
Hurley, K., 1994, ApJ Suppl. Ser. 90, 857
Hurley, K., 1998a, in: C. A. Meegan, R. D. Preece, & T. M. Koshut (eds.), 4th Huntsville Symposium on Gamma-Ray Bursts (AIP: New York), AIP Conf. Proc. 428, 87
Hurley, K., 1998b, in: Proc. 3rd Integral Workshop, preprint astro-ph/9812052
Hurley, K., 1999a, preprint astro-ph/9912061
Hurley, K., 1999b, A&A Suppl. Ser. 138, 553
Hurley, K. et al., 1994, ApJ Suppl. Ser. 92, 655
Hurley, K. et al., 1995, Astroph. Sp. Sci. 231, 227
Hurley, K. et al., 1999a, ApJ 515, 497
Hurley, K. et al., 1999b, ApJ 524, 92
in ’t Zand, J. J. M., 1998, ApJ 503, L53
in ’t Zand, J. J. M. et al., 1998, ApJ 505, L119
Israel, G. L. et al., 1999, A&A 348, L5
Iwamoto, K., 1999, ApJ 512, L47
Iwamoto, K. et al., 1998, Nature 395, 672
Jorgensen, H. et al., 1995, Astroph. Sp. Sci. 231, 389
Katz, J. I. & Canel, L. M., 1996, ApJ 471, 915
Kippen, R. M. et al., 1998a, in: C. A. Meegan, R. D. Preece, & T. M. Koshut (eds.), 4th Huntsville Symposium on Gamma-Ray Bursts (AIP: New York), AIP Conf. Proc. 428, 119
Kippen, R. M. et al., 1998b, ApJ 506, L27
Klebesadel, R. W., Strong, I. B., & Olson, R. A., 1973, ApJ 182, L85
Klose, S., 1994, ApJ 423, L23
Klose, S., 1995a, A&A 303, 75
Klose, S., 1995b, ApJ 446, 261
Klose, S., 1995c, ApJ 446, 357
Klose, S., 1998, ApJ 507, 300
Klose, S., 1999, in: Proc. 19th Texas Symposium on Relativistic Astrophysics, Paris, France, submitted
Klose, S., Eislöffel, J., & Richter, S., 1996, ApJ 470, L93
Klose, S., Eislöffel, J., & Stecklum, B., 1997, IAU Circ. 6756
Klose, S., Meusinger, H., & Lehmann, H., 1998, IAU Circ. 6864
Klose, S., Stecklum, B., & Fischer, O., 1999, Proc. 5th Huntsville Symposium on Gamma-Ray Bursts, AIP Conf. Proc., submitted
Klose, S. et al., 1998, in: C. A. Meegan, R. D. Preece, & T. M. Koshut (eds.), 4th Huntsville Symposium on Gamma-Ray Bursts (AIP: New York), AIP Conf. Proc. 428, 635
Kouveliotou, C., 1995a, in: H. Böhringer, G. E. Morfill, & J. E. Trümper (eds.), Proc. 17th Texas Symposium on Relativistic Astrophysics (The New York Academy of Sciences: New York), Ann. N.Y. Acad. Sci. 759, 411
Kouveliotou, C., 1995b, Astroph. Sp. Sci. 231, 49
Kouveliotou, C. et al., 1993, ApJ 413, L101
Krumholz, M., Thorsett, S. E., & Harrison, F. A., 1998, ApJ 506, L81
Kulkarni, S. R. & Frail, D. A., 1999, GCN 198
Kulkarni, S. R. et al., 1998a, Nature 393, 35
Kulkarni, S. R. et al., 1998b, Nature 395, 663
Kulkarni, S. R. et al., 1999, Nature 398, 389
Lamb, D. Q., 1995, PASP 107, 1152
Lamb, D. Q., 1997, in: Gravitation, R. E. Schielicke (ed.), Rev. Mod. Astron. 10, 101
Lamb, D. Q., 1999, A&A Suppl. Ser. 138, 607
Lamb, D. Q. & Quashnock, J. M., 1993, 415, L1
Lamb, D. Q. & Reichart, D., 1999, preprint astro-ph/9909002
Lamb, D. Q. et al., 1999, A&A Suppl. Ser. 138, 479
Laros, J. G. et al., 1986, Nature 322, 152
Laros, J. G. et al., 1987, ApJ 320, L111
Laros, J. G. et al., 1997, ApJ Suppl. Ser. 110, 157
Laros, J. G. et al., 1998, ApJ Suppl. Ser. 118, 391
Larson, S. B. & McLean, I. S., 1997, ApJ 491, 93
Larson, S. B., McLean, I. S., & Becklin, E. E., 1996, ApJ, 460, L95
Lazzati, D., Campana, S., & Ghisellini, G., 1999, MNRAS 304, L31
Lee, B. et al., 1997, ApJ 482, L125
Lee, S.-W. & Irwin, J. A., 1997, ApJ 490, 247
Lee, T. T. & Petrosian, V., 1997, ApJ 474, 37
Li, H. & Liang, E. P., 1992, ApJ 400, L59
Loeb, A. & Perna, R., 1998a, ApJ 495, 597
Loeb, A. & Perna, R., 1998b, ApJ 503, L35
Luginbuhl, C. et al., 1995, Astroph. Sp. Sci. 231, 289
Luginbuhl, C. et al., 1996, IAU Circ. 6526
Lund, N., 1995, Astroph. Sp. Sci. 231, 217
Mannheim, K., Hartmann, D. H., & Funk, B., 1996, ApJ 467, 532
Mao, S. & Mo, H.-J., 1998, A&A 339, L1
Mao, S., Narayan, R., & Piran, T., 1994, ApJ 420, 171
Mao, S. & Paczyński, B, 1992a, ApJ 388, L45
Mao, S. & Paczyński, B, 1992b, ApJ 389, L13
Marani, G. F. et al., 1999, ApJ 512, L13
Marshall, F. E. et al., 1997, IAU Circ. 6727
Matonick, D. M. & Fesen, R. A., 1997, ApJ Suppl. Ser. 112, 49
Mazets, E. P. et al., 1981, Astroph. Sp. Sci. 80, 3
McBreen, B. et al., 1994, MNRAS 271, 662
McKenzie, E. H. & Schaefer, B. E., 1999, PASP 111, 964
McNamara, B. J. & Harrison, T. E., 1998, Nature 396, 233
McNamara, B. J., Harrison, T. E., & Williams, C. L., 1995, ApJ 452, L25
Meegan, C. A., 1998, Adv. Space Res. 22 (7), 1065
Meegan, C. A. et al., 1992, Nature 355, 143
Meegan, C. A. et al., 1995, ApJ 446, L15
Mészáros, A., Bagoly, Z., & Vavrek, R., 1999, preprint astro-ph/9912037
Mészáros, A. & Mészáros, P., 1996, ApJ 466, 29
Mészáros, A. et al., 1996, J. Korean Astron. Soc. 29, S43
Mészáros, P., 1997, in: Gravitation, R. E. Schielicke (ed.), Rev. Mod. Astron. 10, 127
Mészáros, P., 1999a, preprint astro-ph/9904038
Mészáros, P., 1999b, Nature 398, 368
Mészáros, P., 1999c, A&A Suppl. Ser. 138, 533
Mészáros, P. & Mészáros, A., 1995, ApJ 449, 9
Mészáros, P. & Rees, M. J., 1999a, ApJ 502, L105
Mészáros, P. & Rees, M. J., 1999b, MNRAS 306, L39
Mészáros, P., Rees, M. J., & Papathanassiou, H., 1994, ApJ 432, 181
Mészáros, P., Rees, M. J., & Wijers, R. A. M. J., 1998, ApJ 499, 301
Mészáros, P., Rees, M. J., & Wijers, R. A. M. J., 1999, New Astron. 4, 303
Mitrofanov, I. G., 1998, Adv. Space Res. 22 (7), 1077
Mukherjee, S. et al., 1998, ApJ 508, 314
Murakami, T. et al., 1997, IAU Circ. 6732
Nemiroff, R. J., 1994, Comm. Astrophys. 17, 189
Nemiroff, R. J., 1995, PASP 107, 1131
Norris, J. P., Bonnell, J. T., & Watanabe, K., 1999, ApJ 518, 901
Norris, J. P. et al., 1994, ApJ 424, 540
Norris, J. P. et al., 1995, ApJ 439, 542
Odewahn, S. C. et al., 1998, ApJ 509, L5
Owens, A., Schaefer, B. E., & Sembay, S., 1995, ApJ 447, 279
Paczyński, B., 1987, ApJ 317, L51
Paczyński, B., 1991a, Acta Astron. 41, 157
Paczyński, B., 1991b, Acta Astron. 41, 257
Paczyński, B., 1995, PASP 107, 1167
Paczyński, B., 1998a, in: C. A. Meegan, R. D. Preece, & T. M. Koshut (eds.), 4th Huntsville Symposium on Gamma-Ray Bursts (AIP: New York), AIP Conf. Proc. 428, 783
Paczyński, B., 1998b, ApJ 494, L45
Paczyński, B., 1999, in: M. Livio, K. Sahu, & N. Panagia (eds.), The Largest Explosions Since the Big Bang: Supernovae and Gamma-Ray Bursts (Cambridge University Press: Cambridge), preprint astro-ph/9909048
Palazzi, E. et al., 1998, A&A 336, L95
Palmer, D. M. et al., 1995, Astroph. Sp. Sci. 231, 315
Panaitescu, A., Mészáros, P., & Rees, M. J., 1998, ApJ 503, 314
Park, H. S. et al., 1997, ApJ 490, 99
Park, H. S. et al., 1999, Proc. 5th Huntsville Symposium on Gamma-Ray Bursts, AIP Conf. Proc., to appear
Pedersen, H., 1994, A&A 291, L17
Pedersen, H. et al., 1998, ApJ 496, 311
Perna, R. & Loeb, A., 1998a, ApJ 501, 467
Perna, R. & Loeb, A., 1998b, ApJ 509, L85
Perna, R., Raymond, J., & Loeb, A., 1999, preprint astro-ph/9904181
Pian, E. et al., 1999, preprint astro-ph/9903113
Piran, T., 1997, in: J. N. Bahcall & J. P. Ostriker (eds.), Unsolved Problems in Astrophysics (Princeton University Press: Princeton), p. 343
Piran, T., 1999a, Physics Reports 314, 575; astro-ph/9810256
Piran, T., 1999b, preprint astro-ph/9907392
Piro, L. et al., 1999a, preprint astro-ph/9906363
Piro, L. et al., 1999b, ApJ 514, L73
Pizzichini, G. et al., 1986, ApJ 301, 641
Puget, J.-L., 1981, Astroph. Sp. Sci. 75, 109
Pugliese, G., Falcke, H., & Biermann, P. L., 1999, A&A 344, L37
Rees, M. J., 1995, PASP 107, 1176
Rees, M. J., 1998, in: A. V. Olinto, J. A. Frieman, & D. N. Schramm (eds.), Proc. 18th Texas Symposium on Relativistic Astrophysics (World Scientific: Singapore), p. 34
Rees, M. J., 1999, A&A Suppl. Ser. 138, 491
Reichart, D., 1999, ApJ 521, L111
Reichart, D. et al., 1999, ApJ 517, 692
Remillard, R. et al., 1997, IAU Circ. 6726
Rhoads, J. E., 1997, ApJ 487, L1
Rhoads, J. E., 1999, A&A Suppl. Ser. 138, 539
Rol, E. et al., 1999, Proc. 5th Huntsville Symposium on Gamma-Ray Bursts, AIP Conf. Proc., to appear
Ruderman, M., 1975, Ann. N.Y. Acad. Sci. 262, 164
Ruffert, M. & Janka, H.-Th., 1998, A&A 338, 535
Ruffert, M. & Janka, H.-Th., 1999, A&A 344, 573
Sari, R., 1999, ApJ 524, L43
Sari, R. & Piran, T., 1999, ApJ 517, L109
Sari, R., Piran, T., & Halpern, J. P., 1999, ApJ 519, L17
Sari, R., Piran, T., & Narayan, R., 1998, ApJ 497, L17
Schaefer, B. E., 1990, ApJ 364, 590
Schaefer, B. E., 1999, Proc. 5th Huntsville Symposium on Gamma-Ray Bursts, AIP Conf. Proc., to appear
Schaefer, B. E. et al., 1989, ApJ 340, 455
Schaefer, B. E. et al., 1994, ApJ 422, L71
Schaefer, B. E. et al., 1997, ApJ 489, 693
Schaefer, B. E. et al., 1998, ApJ Suppl. Ser. 118, 353
Schaefer, B. E. et al., 1999, ApJ 524, L103
Schartel, N., Andernach, H., & Greiner, J., 1997, A&A 323, 659
Schmidt, M., 1999, ApJ 523, L117
Smith, D. A. et al., 1999, ApJ 526, 683
Smith, I. A. et al., 1999, A&A 347, 92
Stanek, K. Z. et al., 1999, ApJ 522, L39
Stern, B., 1999, in: J. Poutanen & R. Svensson (eds), High Energy Processes in Accreting Black Holes (ASP: San Francisco), ASP Conf. Ser. 161, 277
Strong, I. B., Klebesadel, R. W., & Olson, R. A., 1974, ApJ 188, L1
Tavani, M., 1998, ApJ 497, L21
Taylor, G. B. et al., 1998a, GCN 40
Taylor, G. B. et al., 1998b, ApJ 502, L115
Teegarden, B. J., 1998, Adv. Space Res. 22 (7), 1083
Teegarden, B. J. & Cline, T. L., 1980, ApJ 236, L67
Tegmark, M. et al., 1996a, ApJ 466, 757
Tegmark, M. et al., 1996b, ApJ 468, 214
Thorsett, S. E. & Hogg, D. W., 1999, GCN 197
Tinney, C. et al., 1998, IAU Circ. 6896
Totani, T. 1999, ApJ 511, 41
Trimble, V., 1995, PASP 107, 1133
Usov, V. V. & Chibisov, G. V., 1975, Sov. Astron. 19, 115
van den Bergh, S., 1983, Astroph. Sp. Sci. 97, 385
van Paradijs, J. et al., 1997, Nature 386, 686
Vanderspek, R., Krimm, H. A., & Ricker, G. R., 1995, Astroph. Sp. Sci. 231, 259
Vanderspek, R. et al., 1999, A&A Suppl. Ser. 138, 565
Vietri, M., 1999, preprint astro-ph/9911523
Vietri, M. & Stella, L., 1998, ApJ 507, L45
Vrba, F. J., 1996, in: C. Kouveliotou, M. F. Briggs, & G. J. Fishman (eds.), 3rd Huntsville Symposium on Gamma-Ray Bursts (AIP: New York), AIP Conf. Proc. 384, 565
Vrba, F. J., Hartmann, D. H., & Jennings, M. C., 1995, ApJ 446, 115
Vrba, F. J. et al., 1994, ApJ 424, 68
Vrba, F. J. et al., 1999a, ApJ 511, 298
Vrba, F. J. et al., 1999b, Proc. 5th Huntsville Symposium on Gamma-Ray Bursts, AIP Conf. Proc., to appear
Vreeswijk, P. M. et al., 1999, ApJ 523, 171
Wang, D. Q., 1999, ApJ 517, L27
Wang, L. & Wheeler, J. C., 1998, ApJ 504, L87
Walter, F. et al., 1998, ApJ 502, L143
Waxman, E., 1997, ApJ 489, L33
Waxman, E., Kulkarni, S. R., & Frail, D. A., 1998, ApJ 497, 288
Webber, W. R. et al., 1995, AJ 110, 733
Wei, D. M. & Lu, T., 1999, preprint astro-ph/9908273
Wheeler, C., 1999, in: M. Livio, K. Sahu, & N. Panagia (eds.), The Largest Explosions Since the Big Bang: Supernovae and Gamma-Ray Bursts (Cambridge University Press: Cambridge), preprint astro-ph/9909096
Wijers, R. A. M. J. & Galama, T. J., 1999, ApJ 523, 177
Wijers, R. A. M. J., Rees, M. J., & Mészáros, P., 1997, MNRAS 288, L51
Wijers, R. A. M. J. et al., 1998, MNRAS 294, L13
Wijers, R. A. M. J. et al., 1999, ApJ 523, L33
Woods, E. & Loeb, A., 1999, preprint astro-ph/9907110
Woosley, S. E., Eastman, R. G., & Schmidt, B. P., 1999, ApJ 516, 788
Woosley, S. E., MacFadyen, A. I., & Heger, A., 1999, in: M. Livio, K. Sahu, & N. Panagia (eds.), The Largest Explosions Since the Big Bang: Supernovae and Gamma-Ray Bursts (Cambridge University Press: Cambridge), preprint astro-ph/9909034
Yoshida, A. et al., 1999, A&A Suppl. Ser. 138, 433
Zharikov, S. V. & Sokolov, V. V., 1999, A&A Suppl. Ser. 138, 485 |
no-problem/0001/nucl-th0001029.html | ar5iv | text | # Sensitivities of the Proton-Nucleus Elastic Scattering Observables of 6He and 8He at Intermediate Energies
## I INTRODUCTION
The advent of radioactive accelerator beams during the past decade has enhanced the variety of nuclear reactions available for study. In the present research we concentrate on the neutron rich isotopes of helium, <sup>6</sup>He and <sup>8</sup>He, which have been produced as secondary beams at intermediate energies . The simplified shell model structure of these isotopes is thought to be a core <sup>4</sup>He, surrounded by loosely bound valence neutrons located in the $`p`$ shell, denoted often as the nuclear halo. A significant amount of research has been done on constructing models that reproduce data from experimental reactions involving these isotopes (Refs. , for example). To use radioactive beams effectively in nuclear studies the present uncertainty in the ground state wave functions of <sup>6</sup>He and <sup>8</sup>He must be reduced. Once the wave functions are known with better precision, the radioactive beam experiments may produce significant implications for neutron stars, shell model calculations, the two body nuclear force, and the three body nuclear force.
One way to ascertain the physical structure of the exotic helium nuclei would be to use elastic scattering data. We want to address the feasibility of this method by developing proton-<sup>6</sup>He and proton-<sup>8</sup>He first order optical potentials at intermediate energies (60 MeV-100 MeV per nucleon) using different structure models as inputs to this optical potential model. A fair amount of earlier work examines this sensitivity , , but there is not full agreement in the literature on the strength of this sensitivity of structure to elastic scattering data. For example Ref. found that at these energies proton-nucleus elastic scattering data was not an effectual tool in determining structure. Korsheninnikov et al. also did a detailed study on the sensitivities of proton elastic scattering not only of the helium isotopes but also the lithium isotopes (<sup>9</sup>Li and <sup>11</sup>Li), using an eikonal approach. They concluded that elastic scattering “is not a very promising tool” to determine structure of the valence neutrons. Their belief is that the size of the core plays a more important role in determining the differential cross section than the lower density valance neutrons. More recently, and in contrast, Karataglidas et al. have performed calculations on the same exotic helium reactions at intermediate energy range using a few different variations of a structure calculation in a g-matrix elastic optical potential calculation. They concluded for <sup>6</sup>He that the data available was insufficient for the elastic scattering calculations to discern the existence of a halo. For <sup>8</sup>He, they ascertained that there was enough data to conclude that it is not a halo nucleus from comparing differences in the elastic differential cross section calculations.
In the literature there is not general agreement to the question of sensitivity of the elastic proton-nucleus differential cross section at intermediate energies to the structure calculation of the target nuclei <sup>6</sup>He and <sup>8</sup>He. Most research concludes that the sensitivity is not there to determine structure, but some authors have used elastic proton-nucleus scattering to put constraints on the details of their physical structure, specifically the halo. In this work, we will systematically examine this sensitivity using four independent structure models and conclude whether elastic scattering is a tool that should be used to ascertain the structure of <sup>6</sup>He and <sup>8</sup>He.
In section II, we will briefly summarize our full-folding optical potential calculation technique which we use to describe elastic proton-<sup>6</sup>He and proton-<sup>8</sup>He scattering and we outline the four different structure models used to describe the helium isotopes. Our results are in section III and our conclusions are in section IV.
## II FULL-FOLDING OPTICAL POTENTIAL
A standard microscopic approach to the elastic scattering of a strongly interacting projectile from a target of $`A`$ particles is given by the formulation of an optical potential in ‘$`\tau \rho `$’ form where $`\tau `$ contains information about the nucleon-nucleon interaction and $`\rho `$ is a nuclear structure calculation (ground state density) of the target. The development of this optical potential begins with the separation of the Lippmann-Schwinger equation for the transition amplitude
$$T=V+VG_0(E)T$$
(1)
into two parts, namely an integral equation for $`T`$:
$$T=U+UG_0(E)PT,$$
(2)
where $`U`$ is the optical potential operator given through a second integral equation
$$U=V+VG_0(E)QU.$$
(3)
In the above equations, the potential operator $`V`$ represents the external many-body interaction. The potential operator $`V=_{i=1}^Av_{0i}`$ consists of the two-body potential $`v_{0i}`$ acting between the projectile and the nucleons in the target nucleus. The operators $`P`$ and $`Q`$ are projection operators, $`P+Q=1`$, and $`P`$ is defined such that Eq. (2) is solvable. In this case, $`P`$ is conventionally taken to project onto the elastic channel. For more details see Refs. .
The evaluation of the full-folding optical potential requires a fully off-shell nuclear density matrix, which in its most general form is given as
$$\rho (𝐫^{},𝐫)=\mathrm{\Phi }_{A}^{}{}_{}{}^{}(𝐫^{})\mathrm{\Phi }_A(𝐫).$$
(4)
Here, $`\mathrm{\Phi }_A(𝐫)`$ is the wave function describing the nuclear ground state in position space. We choose four models for this work to describe the <sup>6</sup>He and <sup>8</sup>He ground state. The structure models vary in rigor, quality, and applicability in describing these exotic helium nuclei. First, a general description of each model will be given in part A, to be followed by comparisons of all four models in part B.
### A Descriptions of the four off-shell densities
Our first model, proposed by Sherr , will be referred to as the ‘boot-strap’ model (BS). This model was created to describe the root mean squared (rms) radii of a variety of exotic nuclei using a simple description of the nucleus. It represents the valence neutrons by using a Woods-Saxon potential which is fit to the two neutron binding energy. It then follows a sequential step procedure to build the exotic nuclei. Explicitly for helium, the model starts with the well known core <sup>4</sup>He, and then builds <sup>6</sup>He by calculating the wave function for the valence neutrons. Likewise, to construct <sup>8</sup>He, <sup>6</sup>He is considered the core and the 2 neutron wave function generated from a Woods-Saxon potential is calculated. Wave functions for the <sup>4</sup>He core and valence neutrons were calculated in r space (relative to the center of mass of the whole nucleus), then Fourier transformed to momentum space
$$\rho ^{}(𝐩^{},𝐩)=\frac{1}{8\pi ^3}d^3𝐫^{}e^{i𝐫^{}𝐩^{}}d^3𝐫e^{i𝐫𝐩}\rho (𝐫^{},𝐫),$$
(5)
where they were used in construction of the off-shell density via Eq. (4). The validity of this model is questionable due to its extreme simplicity. It is thought that the size of the <sup>4</sup>He core in the exotic isotopes is larger than the bare <sup>4</sup>He radius which this model assumes . The mode of construction of <sup>8</sup>He is also in contradiction with most other structure calculations for <sup>8</sup>He. To first order, <sup>6</sup>He could be approximated as a <sup>4</sup>He + 2n; however, it is apparent that <sup>8</sup>He is closer to <sup>4</sup>He+4n than (<sup>4</sup>He +2n) +2n as the BS model would suggest .
The second model is a relativistic point coupling model within the framework of a chiral effective field theory by Rusnak and Furnstahl . A Lagrangian is constructed: an expansion in powers of the scalar,vector, isovector-vector, tensor, and isovector-tensor densities, and their derivatives. The theory contains all the symmetries of QCD and is able to calculate low energy features, such as the structure of nuclei ground states adequately. For this paper, we used what Ref. refers to as the ‘FZ4’ scheme. Here, the vector meson and $`\rho `$ meson masses are fixed, while the coefficients of the densities to fourth order are varied to produce a low chi-square to experimental observables. Most varieties of the chiral effective theory reproduce the bulk nuclear observables of spherically symmetric nuclei. Questions of applicability to exotic helium nuclei can of course be raised while using this model, for it was not developed for the non-spherical, non-bulk nuclei <sup>6</sup>He and <sup>8</sup>He. The numerical procedure to create the off-shell densities (Eq. 4) used for the optical potential is given in Ref. .
The third nuclear structure model used to describe the densities <sup>6</sup>He and <sup>8</sup>He, a Dirac-Hartree model (DH), has been used extensively by two of the authors (Ch. Elster and S. P. Weppner) to describe the structure of doubly magic spherical nuclei with success . This is the oldest structure model of the four models discussed. The FZ4 model (detailed above) has the same structure wave function, so the method used to create the momentum off-shell density of Eq. (4) is the same for both models and is detailed in Refs. . Applicability is a concern for this model also. When developed, it was fit to the bulk properties of <sup>16</sup>O, <sup>40</sup>Ca, <sup>48</sup>Ca, <sup>90</sup>Zr, and <sup>208</sup>Pb, all doubly magic nuclei. Furthermore, this model, as well as FZ4, falls under the mean field ansatz, which is rather tenable when describing nuclei with only 6 or 8 nucleons. It was also developed well before the general structure of <sup>6</sup>He and <sup>8</sup>He were apparent, thereby making it a candidate to test whether the elastic observables can detect this non-applicability.
The last nuclear model to be discussed was developed explicitly for exotic nuclei. The COSMA model (cluster orbit shell model approximation) , an approximation to the three body problem, has been used extensively in the literature to describe elastic scattering with the exotic helium nuclei . The COSMA model is a combination of nucleon clustering and the standard shell model, which obeys the Pauli exclusion principle by using Slater determinants to produce a fully antisymmetrized wave function in r-space. They are then translated as single particle wave functions relative to the center of mass of the entire nucleus. These wave functions are Fourier transformed into momentum space using the method of Eq. (5), thus making it possible to create a fully off-shell density in momentum space.
There have been more rigorous models developed for <sup>6</sup>He and <sup>8</sup>He (Refs. among others) which we have not used. In all cases, these models treat the three body problem (core + nucleon + nucleon) or (core +di-neutron +di-neutron) with fewer approximations than the models presented here. Our goal is to present a sensitivity test, rather than produce the most fundamental calculations possible. To that end, we have used four highly varying models, with different characteristics, which are easy to calculate. As an aside, in a comparison between a realistic three body approach (treating the antisymmetry correctly) and the reducible two body approximation, the authors of Ref. note little difference at intermediate and long ranges. They do find small differences in the short range behavior of the wave functions, but these are of little consequence when describing intermediate energy elastic scattering. Therefore we would expect our results using these simpler models to differ little from results using the more rigorous realistic models.
### B Nuclear structure of the four models
In Fig. 1, we have plotted the proton densities in r-space of all four structure models described above. The general characterization is that the core two protons and two neutrons are more tightly bound in the two models that are designed for exotic nuclei, these being the BS model (solid line) and COSMA model (long dashed line). These models have a core close to that of the lone <sup>4</sup>He nucleus ($`1.6`$ fm). In contrast, the FZ4 model (short dashed line) and the DH model (dashed-dotted line) have a core which ranges from 12$`\%`$ to 20$`\%`$ larger than their exotic counterparts.
The total neutron densities in r-space have also been plotted for all four structure models for <sup>6</sup>He and <sup>8</sup>He in Fig. 2. Comparing the exotic nuclei structure models first, the BS model (solid line) and COSMA model (long dashed line) have a tight two-neutron core because their densities are higher in the 0 fm to 2 fm range. At about 3 fm, differences begin to emerge between these two exotic models. All models do have an extended neutron wave function of varying degrees, with the COSMA model having the most unique shape.
In Table I, we list the four models and the characteristics they describe. All position measurements are relative to the center of mass of the <sup>6</sup>He or <sup>8</sup>He system. Calculated in Table I are the root mean squared matter radius ($`r_{rms}`$) of the whole nucleus, the rms radius of the two neutron-two proton core wave function($`r_c`$), the standard deviation of the core ($`\mathrm{\Delta }r_c`$), the valence neutron matter radius ($`r_v`$), the standard deviation of the valence neutron wave function ($`\mathrm{\Delta }r_v`$), the separation energy of the valence neutrons, the separation distance of one standard deviation of the core and valence wave functions, and a statement on whether the model has a halo structure as defined by this work. The definition of the standard deviation is
$$\mathrm{\Delta }r=\sqrt{<r^2><r>^2},$$
(6)
whereas the separation distance is defined as
$$S_d=r_v\mathrm{\Delta }r_v\mathrm{\Delta }r_cr_c.$$
(7)
Simply if $`S_d>0`$ then we define this nucleus as having a halo because there is a well defined separation between the core and halo centers. The only discernible halo nucleus is the COSMA model of <sup>6</sup>He. It contains a tightly bounded core wave function with adequate spacing between core and valence nucleons. The COSMA model also has a significantly different asymptotic wave function shape. The other three models do not define a halo for the <sup>6</sup>He nucleus, as there is too much significant overlap between the core and valence wave functions. No model produces a definitive halo for <sup>8</sup>He, although COSMA comes closest.
In summary it is concluded that the most disparate structure is the COSMA model. One would expect that because this model is used often to describe exotic nuclei, it should also best describe elastic scattering if the observables are sensitive to the nuclear structure calculation. In the next section we will use these four models as input into our optical potential to describe elastic scattering at the intermediate energy range.
## III RESULTS AND DISCUSSION
The elastic scattering of protons from <sup>6</sup>He and <sup>8</sup>He at incident proton energies from 66 MeV to 100 MeV are calculated. At this energy range, the proton-nucleus elastic scattering data are scarce and exist only at forward angles. We will focus on the reactions where experimental data for the elastic differential cross-sections exists, but we will also comment on how our conclusions would change if the experimental database were enlarged. Other observables which are calculated in this work, for which no data exists, are the spin rotation function ($`Q`$) and the analyzing power ($`A_y`$). We will also comment on sensitivity to these observables.
The full-folding optical potential used for these results is calculated as outlined in Refs. , and we use the four model densities as described in Sec. II. We will refer to the model of Ref. as ‘BS’. The Dirac-Hartree model of Refs. will be labeled ‘DH’. The chiral point coupling model of Ref. will be labeled ‘FZ4’. The cluster model of Refs. will be referred to as ‘COSMA’.
The full-folding optical potential also requires a model of the NN interaction. In this work, we use the Nijmegen I interaction . We have also calculated some optical potentials using the CD Bonn NN potential . This potential has the same tight constraints of the Nijmegen for its on shell values to agree with np and pp data, but the off shell amplitudes are different. The elastic scattering calculations using the CD Bonn potential show very little difference with those that use the Nijmegen potential. The conclusions drawn in this work are therefore independent of the choice of which modern NN potential was used.
### A Elastic scattering results: effects of structure
The scattering observables for elastic proton scattering from <sup>6</sup>He at 71 MeV are displayed in Fig. 3. There are four calculations on the figure (using the same legend as Figs. 1-2). The solid line represents the elastic differential cross-section calculated from a full-folding optical potential using the BS model as the structure calculation, and the short-dashed line represents a calculation of the observables from a full-folding optical potential using the FZ4 model as the structure calculation. The DH version of the calculation is represented by the dot-dashed line while the optical potential using the COSMA structure model was used in the calculation of the long-dashed line. All models use the Nijmegen I nucleon-nucleon interaction. The experimental data for this reaction were given in Ref. . As also seen in Fig. 1, the only calculation which does not adequately describe the limited experimental data for this reaction is the BS model. Referring to Table I, the only significant difference between this model and other models is the extremely small core which mimics the size of a lone <sup>4</sup>He nucleus. The other three models agree favorably with the limited data, yet when one looks at their features there are large differences in their binding energies, rms radii, and presence of a discernible halo. Therefore, it is impossible to draw any conclusions about the structure of <sup>6</sup>He valance neutrons from this reaction. We may draw some inferences on the appropriate size of the core from looking at the results produced using the BS model, but the shape and existence of a halo cannot be determined from this reaction.
In Fig. 4, we calculate elastic scattering from <sup>8</sup>He at a projectile energy of 66 MeV. The legend represents the same calculations as in Fig. 3. The experimental data are from Ref. . All four models adequately represent the data. The one model with the most significant differences in shape is again the BS model (and only at higher angles). This model has the tightest core, and a loose valence wave function, as it had for <sup>6</sup>He. There is nothing that can be learned about the valance structure of <sup>8</sup>He from studying this reaction’s differential cross section for the data which exist. Polarization measurements (specifically $`Q`$) may be used if polarized experiments are done at large angles to high accuracy ($`>`$60<sup>o</sup>).
We move to a slightly higher projectile energy in Fig. 5, where protons are scattered from <sup>8</sup>He at a projectile energy of 72.5 MeV. Again, the calculations have the same legend as given in Figs. 3 and Fig. 4. The data for this calculation are the most extensive in this energy range, they approach the 65<sup>o</sup> center of mass angle. Unfortunately the structure of <sup>8</sup>He still cannot be determined from this experiment. The sensitivities due to the structure calculation are not strong enough, given the experimental error, to ascertain the structure of <sup>8</sup>He. According to Table I, the COSMA model has a more defined valence ring than the others, but the differential cross section experimental data are unable to differentiate the validity of any of these disparate models unless experimenters were able to measure the differential cross-section with a margin of error of less than 20% at angles above 65<sup>o</sup>.
In Fig. 6, we explore a higher energy reaction where no experimental data exist. Here, we calculated protons scattering from <sup>8</sup>He at 100 MeV. Once more, the calculations use the same models and line definitions as in Figs. 3-5. Again, the three structure calculations give similar results (DH,FZ4, COSMA). The BS model runs high through the whole calculation, similar in significance to the previous figures. Polarization experiments also tell us little below 70<sup>o</sup> scattering center of mass angle. Even at 100 MeV the elastic reaction is insensitive to the structure of the valence neutrons in <sup>8</sup>He.
Recently, in Ref. , a similar study was done using these elastic reactions and pion production. The authors concluded that neither <sup>6</sup>He nor <sup>8</sup>He were halo nuclei. Their inclusion of a halo in their structure calculations always lowered the differential cross section at angles greater than 15<sup>o</sup>. In this present work, we did not find such a simple relationship between non-halo and halo nuclei. In fact, the most extreme halo model of <sup>6</sup>He (dashed line - COSMA) was not at either extreme of the differential cross section calculation of Fig. 5. We therefore conclude that the differential cross section is only slightly sensitive to the existence of a halo ($`S_d`$), the spread of the halo ($`\mathrm{\Delta }r_v`$), the radius of the core nucleons ($`r_c`$), and the binding energy of the core nucleons ($`\mathrm{\Delta }r_c`$), all of which are coupled to each other in a complex fashion. The structure parameter that seems to have the most influence at this energy is the radius of the core <sup>4</sup>He particle. There seems to be almost complete insensitivity to the valence neutron wave functions. To reiterate, the COSMA model has a very distinct asymptotic shape for the valence neutrons, and this uniqueness does not transfer into the differential cross section as exhibited by the similarity in the calculations.
### B Medium effects
So far in this work, we have used the impulse approximation, setting the medium field to zero. In previous work two of the authors (Ch. Elster and S. P. Weppner) showed that at 65 MeV, if a medium field was used (as outlined briefly in Section II, and in more detail in Refs. ) then there was a systematically better fit with elastic scattering observables across a wide range of stable spin-0 nuclei.
For two of the structure calculations, the DH and the FZ4, we added a mean field consistently. If we used a DH structure model then we used the same DH model to simulate our mean field; likewise, this was also done using the FZ4 structure model. Overall the effects of adding this mean field to the <sup>6</sup>He and <sup>8</sup>He calculations of elastic scattering observables were smaller than seen previously for other nuclei.
In Fig. 7, we compare two calculations of <sup>8</sup>He elastically scattering off a proton at 66 MeV. Both calculations use the DH structure calculation and the Nijmegen I interaction. The difference is that the solid line sets the mean field to zero, while the dashed line includes it. For comparison, in Fig. 8, the same calculation is done using the FZ4 structure calculation and mean field using the same FZ4 model. Both calculations give the same results: when the medium effect is added, it systematically lowers the differential cross section slightly. In general, the effect is smaller than for larger spin-0 nuclei previously studied . Since these are smaller nuclei, and less tightly bound, this conclusion seems reasonable. However, it is important to note that this small change did not lead to a better description of the experimental data, in contrast to earlier work with other nuclei where there was a systematic improvement.
At higher energies, these trends continue, although their effects are smaller. We plot in Fig. 9 the elastic observables of <sup>8</sup>He at an energy of 100 MeV colliding with a proton. As in Fig. 7, the solid line represents the DH calculation without mean field effects, while the dashed line includes the effects. Both calculations use the Nijmegen I potential. These medium effects are barely discernible at this higher energy. This trend has been seen before in earlier work with other nuclei . For completeness, in Fig. 10 we have calculated the same reaction as Fig. 9 except we now use the FZ4 structure calculation and mean field (dashed line) for <sup>8</sup>He. The same conclusions are reached. By using two different models, we conclude that this mean field procedure leads to results that are model independent and smaller than doubly magic nuclei at the same energies.
## IV SUMMARY and CONCLUSION
We have presented sensitivity tests for elastic scattering observables of protons bombarded with <sup>6</sup>He and <sup>8</sup>He. Here, we found that elastic scattering is a weak tool for determining the structure of these isotopes. These conclusions were drawn by using four different nuclear structure models that had different spatial characteristics in the calculation of the proton-nucleus optical potential. All calculations using these structure models were in good agreement with the data that exist. In fact, the models not designed for exotic nuclei (Dirac-Hartree and chiral models) did as well as, and sometimes slightly better than their made-for-exotic-nuclei counter parts (COSMA and a simple ‘boot-strap’ model). We agree with the results of earlier work of Korsheninnikov et al. . They believe that the size of the core plays a more important role in determining the differential cross section than the lower density valance neutrons. The only potential area for significant nuclear structure sensitivity with elastic scattering is with the large angle ($`>`$ 70<sup>o</sup>) spin observables. Since the radioactive beams are secondary beams, to produce enough polarized statistics to measure these reactions with any accuracy is beyond experimental and theoretical capabilities at the present time. It is, therefore, possible to conclude that one should look beyond intermediate elastic reactions when trying to determine the structure of the neutron rich helium isotopes. Higher energy elastic scattering ($`>`$ 500 MeV/nucleon) has had some success in determining structure, although they warn against using an optical model approach, as used here . Inelastic hadron reactions (momentum distributions following fragmentation , transfer reactions , Coulomb breakup , excitation , and charged pion photo production ) and an interesting concept using electron scattering offer hope as tools to determine conclusively the structure of <sup>6</sup>He and <sup>8</sup>He.
###### Acknowledgements.
This work was performed in part under the auspices of the National Science Foundation under grant No. PHY-9804307 with Eckerd College, the auspices of Eckerd College under the Hughes Foundation, and the auspices of the Department of Energy under grant No. DE-FG02-93ER40756 with Ohio University. We thank the National Partnership for Advanced Computational Infrastructure (NPACI) under grant No. ECK200 and the Ohio Supercomputer Center (OSC) under grant No. PHS206 for the use of their facilities. We would also like to thank Richard Furnstahl for the use of his chiral structure code. |
no-problem/0001/astro-ph0001234.html | ar5iv | text | # Symbiotic Miras can do it
## 1. Introduction
Symbiotic Miras are interacting binary systems composed of a late AGB star and a hot white dwarf. Even though the orbital periods of none of these systems have been measured, they are thought to be of the order of 20-100 yr (cf. Mikołajewska 1997). Then interaction between the two stars does not occur via Roche lobe overflow, but it is estimated that the white dwarf accretes mass from the Mira wind at a rate of 10<sup>-8</sup> M yr<sup>-1</sup> or slightly more. The accreted mass amounts to about 1$`\%`$ of the Mira wind; the rest distributes around the binary system in a slowly expanding circumbinary nebula.
Symbiotic Miras are a key case for the study of the effects on the AGB mass loss caused by interaction in detached binaries. The Mira wind is partially ionised by the white dwarf radiation, and thus it is more easily observable than in single Miras. In addition, accretion of the Mira wind onto the white dwarf causes nova–like eruptions and the consequent production of fast winds which can last for decades (slow-novae). These winds interact and shape the bulk of the Mira wind which is not accreted. The velocities and mass loss rate of the fast winds from the erupting hot component of symbiotic stars are not very different from those produced by the central stars of planetary nebulae (PNe). Thus, in symbiotic Miras not only we can observe the effects of the binary interactions on the geometry of the Mira wind, but also a “simulation” of its post–AGB evolution (PN) under the action of a fast wind.
## 2. The nebulae around symbiotic stars
Our knowledge about the nebular environment of symbiotic stars has improved significantly in recent years. Basic information about the morphology of the innermost regions of these interacting binaries was obtained by means of radio observations (see the list of references in Corradi et al. 1999a), ground based long slit spectroscopic imaging (Solf 1983, 1984), and HST imagery (Paresce & Hack 1994; this paper). In addition, ground–based CCD imaging using specific narrow band filters allowed us to discover extended ionised nebulae with sizes typical of planetary nebulae (up to more than one parsec). In spite of the many similarities, it should be stressed that the nebulae around symbiotic stars are not genuine PNe, since the gas is donated by a star which is still on the AGB phase (pre-PN), while ionisation is provided by the hot companion which has already lost its own PN.
To date, 8 optically extended nebulae are known around symbiotic Miras (for a complete list of references of individual studies, see Corradi et al. 1999a). Three of them have a bipolar shape (R Aqr, BI Cru, and He 2-104), one is an inclined ring (He 2-147), three have a more irregular but markedly aspherical shape (HM Sge, V1016 Cyg and RX Pup), and one is barely resolved (H 1-36).
One of the most important things that we want to point out here, is that the bipolar symbiotic nebulae have morphological and kinematical properties (including high polar velocities of 200-300 km s<sup>-1</sup>) which are strikingly similar to those of some bipolar PNe (e.g. Hb 5 and NGC 6537, Corradi & Schwarz 1993). This suggests a link between the two classes. The fact that among symbiotic Miras half of the nebulae are bipolar/ring types, and all are markedly aspherical, while among PNe only about 15$`\%`$ is bipolar, and some 20$`\%`$ is spherical, indicates that interaction in detached binaries favours the formation of bipolar and aspherical PNe. This is confirmed by theoretical modelling (Mastrodemos & Morris 1999).
Moreover, several symbiotic Miras possess multiple nebulae which are ascribed to the shaping action of fast winds set up during recurrent outbursts. This also offers a natural explanation for the origin of quadrupolar PNe (Manchado, Stanghellini & Guerrero 1996), whose formation is difficult to understand in terms of interacting-winds evolution from single stars.
Some of the properties of the extended nebulae around symbiotic Miras are illustrated by the two cases below.
### 2.1. He 2-104, the Southern Crab
He 2-104 is a symbiotic star containing a Mira with a pulsational period of 400 days. Note that the presence of a Mira in this system is only detected by the modulation of the IR luminosity, while the optical spectrum does not show any sign of the cool stars, and appears as a typical spectrum of a PN except for a core with peculiarly high (for PNe) densities.
We recently obtained an HST image of the multiple-bipolar nebula around He 2-104 in the light of the \[Nii\]658.3 nm line. We present in Fig. 1 the image as it appeared in an HST News Release on August 1999 (see http://oposite.stsci.edu/ pubinfo/pr/1999/32/). This image nicely resolves the inner bipolar nebula, which is a small scale reproduction of the larger crab-like one. The inner nebula is strikingly similar to the prototypical bipolar PN MyCn 18 (Sahai & Trauger 1998), suggesting a similar formation process. At the distance of 800 pc preferred by Schwarz et al. (1989), the inner and outer bipolar nebulae of He 2-104 would have a kinematical age of 200 and 900 yr, respectively.
### 2.2. He 2-147
According to Munari (1997), He 2–147 is a template case of a symbiotic nova whose PN–like spectrum (which is typical of the outburst phase of these systems) has “retreated” to the blue part of the spectrum leaving the red region dominated by strong TiO bands and continuum from the Mira (which is instead characteristic of the quiescent phase of symbiotic novae). The nebula around He 2–147 (Fig. 2) is a ring expanding with a velocity of $``$100 km s<sup>-1</sup> which is inclined at $``$ 55 to the line of sight (Corradi et al. 1999b). The kinematical age of the ring is of about 300 yr, and its size is 0.05 pc. Such a ring nebula can be considered an “extreme bipolar”, in which material in the polar directions and at intermediate latitudes (if any) is exceedingly tenuous, or has already vanished into the surrounding space.
#### Acknowledgments.
The work of RLMC is supported by a grant of the Spanish DGES PB97-1435-C02-01.
## References
Corradi, R.L.M., Schwarz, H.E.: 1993, A&A 268, 714
Corradi, R.L.M., Brandi, E., Ferrer, O., Schwarz , H.E. 1999a, A&A, 343, 841
Corradi, R.L.M., Ferrer, O., Schwarz , H.E., Brandi, E., Garcia, L. 1999b, A&A, 348, 978
Manchado, A., Stanghellini, L., Guerrero, M. 1996 ApJL 466, L95
Mastrodemos, N., Morris, M. 1999 ApJ in the press
Mikołajewska, J. (Editor) 1997, Physical Processes in Symbiotic Binaries and Related Systems, Copernicus Foundation for Polish Astronomy, Warsaw
Munari, U.: 1997, in Physical processes in Symbiotic binaries and related systems ed. J. Mikolajewska, Copernicus Foundation for Polish Astronomy, Warsaw, p. 37.
Paresce, F., Hack, W.J. 1994, A&A 287, 154
Sahai, R., Trauger J.T. 1998 AJ 116, 1357
Schwarz, H.E., Aspin, C., Lutz, J.H. 1989 ApJ 344, L29
Solf, J.: 1983, ApJ 266, L113
Solf, J.: 1984, A&A 139, 296 |
no-problem/0001/hep-ph0001186.html | ar5iv | text | # Models of Neutrino Mass and Mixing
## Introduction
There are numerous studies of the neutrino mass matrix for explaining the disappearance and appearance of $`\nu _e`$ and $`\nu _\mu `$ in various experiments1 . In this talk, I will only address the theoretical issue of how neutrinos obtain mass and the accompanying consequences beyond neutrino oscillations. The starting point of any such discussion is the 1979 observation by Weinberg2 that given the particle content of the Minimal Standard Model at low energies, neutrinos acquire mass only through the following unique effective dimension-5 operator:
$$\mathrm{\Lambda }^1(\nu \varphi ^0e\varphi ^+)^2.$$
(1)
A nonzero Majorana mass for $`\nu `$ is obtained as $`\varphi ^0`$ acquires a vacuum expectation value in electroweak symmetry breaking. Models of neutrino mass differ only in how this operator is realized3 .
## Examples of the Minimalist Approach
(1) Canonical Seesaw4
Add 3 heavy singlet right-handed neutrinos to the Minimal Standard Model: 1 $`\nu _R`$ for each $`\nu _L`$. Then the Weinberg operator is realized because each heavy $`\nu _R`$ is linked to $`\nu _L\varphi ^0`$ with a Yukawa coupling $`f`$; and since $`\nu _R`$ is allowed to have a large Majorana mass $`M_R`$, the famous seesaw realtionship $`m_\nu =m_D^2/M_R`$ is obtained, where $`m_D=f\varphi ^0`$. This mechanism dominates the literature and is usually implied when a particular pattern of neutrino mass and mixing is proposed.
(2) Minimal Seesaw5
Add just 1 $`\nu _R`$. Then only 1 linear combination of $`\nu _e,\nu _\mu ,\nu _\tau `$ gets a seesaw mass. The other 2 neutrino masses are zero at tree level, but since there is in general no more symmetry to protect their masslessness, they must become massive through radiative corrections. As it turns out, this happens in two loops through double $`W`$ exchange and the result is doubly suppressed by the charged-lepton masses. Hence it is not a realistic representation of the present data for neutrino oscillations.
(3) Next-to-Minimal Seesaw6
Add 1 $`\nu _R`$ and 1 extra Higgs doublet. Then 1 neutrino gets a seesaw mass. Another gets a one-loop mass through its coupling to $`\varphi _2^0`$, where $`\varphi _2^0=0`$. This second mass is proportional to the coupling of the term $`(\overline{\varphi }_2^0\varphi _1^0)^2`$ times $`\varphi _1^0^2`$ divided by $`M_R`$. The third neutrino gets a two-loop mass as in (2). This scheme is able to fit the present data.
(4) Purely Radiative Mechanism7
Add 1 extra Higgs doublet $`\mathrm{\Phi }_2`$ and 1 charged singlet $`\chi ^+`$. Then the coexistence of the terms $`(\nu _il_j\nu _jl_i)\chi ^+`$ and $`(\varphi _1^+\varphi _2^0\varphi _2^+\varphi _1^0)\chi ^{}`$ allows the following radiative mass matrix to be obtained:
$$_\nu =\left[\begin{array}{ccc}0& f_{\mu e}(m_\mu ^2m_e^2)& f_{\tau e}(m_\tau ^2m_e^2)\\ f_{\mu e}(m_\mu ^2m_e^2)& 0& f_{\tau \mu }(m_\tau ^2m_\mu ^2)\\ f_{\tau e}(m_\tau ^2m_e^2)& f_{\tau \mu }(m_\tau ^2m_\mu ^2)& 0\end{array}\right].$$
(2)
This model has been revived in recent years and may be used to fit the neutrino-oscillation data.
(5) Heavy Higgs Triplet8
Add 1 heavy Higgs triplet $`(\xi ^{++},\xi ^+,\xi ^0)`$. Then the coexistence of the terms $`\nu _i\nu _j\xi ^0`$ and $`\varphi ^0\varphi ^0\overline{\xi }^0`$ allows a tree-level neutrino mass given by
$$m_\nu =\frac{2f\mu \varphi ^0^2}{m_\xi ^2}=2f\xi ^0.$$
(3)
This shows the interesting result that $`\xi `$ has a very small vacuum expectation value inversely proportional to the square of its mass. Note also that the effective operator of Eq. (1) should now be written as
$$(\nu _i\varphi ^0l_i\varphi ^+)(\nu _j\varphi ^0l_j\varphi ^+)=\nu _i\nu _j(\varphi ^0\varphi ^0)(\nu _il_j+l_i\nu _j)(\varphi ^0\varphi ^+)+l_il_j(\varphi ^+\varphi ^+),$$
(4)
which shows clearly the role of $`\xi `$.
(6) Radiative Splitting of Neutrino Mass Degeneracy9
Add 1 Higgs triplet as in (5). Assume further that
$$_\nu =\left[\begin{array}{ccc}0& m_0& 0\\ m_0& 0& 0\\ 0& 0& m_3\end{array}\right]$$
(5)
at tree level. Then the mass eigenstates corresponding to the mass eigenvalues $`\pm m_0`$ are radiatively corrected to have slightly different masses at one-loop level, resulting in the following successful connection between atmospheric and solar neutrino vacuum oscillations:
$$\frac{(\mathrm{\Delta }m^2)_{sol}(\mathrm{\Delta }m^2)_{atm}}{m_\nu ^4(\mathrm{sin}^22\theta )_{atm}}=2I^2,$$
(6)
where $`I=(3G_Fm_\tau ^2/16\pi ^2\sqrt{2})\mathrm{ln}(m_\xi ^2/m_W^2)`$, and
$$\left(\begin{array}{c}\nu _1\\ \nu _2\\ \nu _3\end{array}\right)=\left(\begin{array}{ccc}1/\sqrt{2}& c/\sqrt{2}& s/\sqrt{2}\\ 1/\sqrt{2}& c/\sqrt{2}& s/\sqrt{2}\\ 0& s& c\end{array}\right)\left(\begin{array}{c}\nu _e\\ \nu _\mu \\ \nu _\tau \end{array}\right)$$
(7)
has been assumed.
## Some Generic Consequences
(A) Once neutrinos have mass and mix with one another, the radiative decay $`\nu _2\nu _1\gamma `$ happens in all models, but is usually harmless as long as $`m_\nu <`$ few eV, in which case it will have an extremely long lifetime, many many orders of magnitude greater than the age of the Universe.
(B) The analogous radiative decay $`\mu e\gamma `$ also happens in all models, but is only a constraint for some models where $`m_\nu `$ is radiative in origin, such as in (3).
(C) Neutrinoless double $`\beta `$ decay occurs10 , but is sensitive only to the $`\nu _e\nu _e`$ entry of $`_\nu `$, which may be assumed to be zero as in (6).
(D) Leptogenesis is possible in the 2 simplest models of neutrino mass, i.e. (1) and (5). In the canonical seesaw scenario, $`\nu _R`$ may decay into both $`l^{}\varphi ^+`$ and $`l^+\varphi ^{}`$11 . In the Higgs triplet scenario, $`\xi ^{++}`$ may decay into both $`l^+l^+`$ and $`\varphi ^+\varphi ^+`$8 . The lepton asymmetry thus generated may be converted into the present observed baryon asymmetry of the Universe through the electroweak sphalerons12 .
## Examples of the Holistic Approach
(7) Grand unification usually requires new particles at high energies and lepton-number conservation to be violated at some scale. Hence it is ideal for the consideration of neutrino mass. There is a vast literature on this subject and I will not discuss anything more in this talk other than the simple observations that $`SO(10)`$ contains $`\nu _R`$ and that $`E_6`$ contains both $`\nu _R`$ and $`\nu _S`$, where the latter may be regarded as a sterile neutrino which has a natural reason to be light13 .
(8) $`R`$ parity nonconserving supersymmetry is another very fruitful approach which has received a lot of attention in the past 2 years or so. If only $`B`$ is assumed to be conserved but not $`L`$, then the superpotential also contains the terms
$$\mu _iL_iH_2+\lambda _{ijk}L_iL_je_k^c+\lambda _{ijk}^{}L_iQ_jd_k^c,$$
(8)
which violates $`R(1)^{3B+L+2J}`$. As a result, a radiative neutrino mass $`m_\nu \lambda ^2(Am_b^2)/16\pi ^2m_{\stackrel{~}{b}}^2`$ may be obtained14 . Furthermore, from the mixing of $`\nu _i`$ with the neutralino mass matrix through the bilinear term $`L_iH_2`$ and the induced vacuum expectation value of $`\stackrel{~}{\nu }_i`$, a tree-level mass $`m_\nu (\mu _i/\mu \stackrel{~}{\nu }_i/h_1^0)^2m_{eff}`$ is also obtained15 .
## More Side Effects
(E) New particles at the 100 GeV mass scale exist in some radiative models. They can be searched for in future accelerators.
(F) Lepton-flavor changing processes at tree level provide another mechanism for matter-induced neutrino oscillations.
(G) Lepton-number violating interactions at the TeV mass scale may erase any preexisting $`B`$ or $`L`$ asymmetry of the Universe16 . In $`R`$ parity nonconserving supersymmetry, $`\lambda ^{}>10^4`$ is required for realistic $`m_\nu `$, but $`\lambda ^{}<10^7`$ is needed to avoid erasure17 .
## Leptogenesis from R Parity Violation
As remarked already earlier in (G), whereas lepton-number violating trilinear couplings in Eq. (8) are able to generate neutrino masses radiatively, they also wash out any preexisting $`B`$ or $`L`$ asymmetry during the electroweak phase transition. On the other hand, successful leptogenesis may still be possible as shown recently18 .
Assume the lightest and 2nd lightest supersymmetric particles to be
$$\stackrel{~}{W}_3^{}=\stackrel{~}{W}_3ϵ\stackrel{~}{B},\stackrel{~}{B}^{}=\stackrel{~}{B}+ϵ\stackrel{~}{W}_3,$$
(9)
respectively, where $`\stackrel{~}{W}_3`$ and $`\stackrel{~}{B}`$ are the $`SU(2)`$ and $`U(1)`$ neutral gauginos, and $`ϵ`$ is a very small number. Note that $`\stackrel{~}{B}`$ couples to $`\overline{\tau }_L^c\stackrel{~}{\tau }_L^c`$ but $`\stackrel{~}{W}_3`$ does not, because $`\tau _L^c`$ is trivial under $`SU(2)`$. Assume $`\stackrel{~}{\tau }_Lh^{}`$ mixing to be negligible but $`\stackrel{~}{\tau }_L^ch^+`$ mixing to be significant and denoted by $`\xi `$. Obviously, $`\stackrel{~}{\tau }`$ may be repalced by $`\stackrel{~}{\mu }`$ or $`\stackrel{~}{e}`$ in this discussion.
Given the above assumptions, $`\stackrel{~}{B}^{}`$ decays into $`\tau ^{}h^\pm `$ through $`\xi `$, whereas $`\stackrel{~}{W}_3^{}`$ decays (also into $`\tau ^{}h^\pm `$) are further suppressed by $`ϵ`$. This allows $`\stackrel{~}{W}_3^{}`$ decay to be slow enough to be out of equilibrium with the expansion of the Universe at a temperature $``$ 2 TeV, and yet have a large enough asymmetry $`(\tau ^{}h^+\tau ^+h^{})`$ in its decay to obtain $`n_B/n_\gamma 10^{10}`$. See Figure 1.
This unique scenario requires $`\stackrel{~}{W}_3^{}`$ to be lighter than $`\stackrel{~}{B}^{}`$ and that both be a few TeV in mass so that the electroweak sphalerons are still very effective in converting the $`L`$ asymmetry into a $`B`$ asymmetry. It also requires very small mixing bewteen $`\stackrel{~}{\tau }_L`$ with $`h^{}`$, which is consistent with the smallness of the neutrino mass required in the phenomenology of neutrino oscillations. On the other hand, the mixing of $`\stackrel{~}{\tau }_L^c`$ with $`h^+`$, i.e. $`\xi `$, should be of order $`10^3`$ which is too large to be consistent with the usual terms of soft supersymmetry breaking. For successful leptogenesis, the nonholomorphic term $`H_2^{}H_1\stackrel{~}{\tau }_L^c`$ is required.
## CONCLUSION
Models of neutrino mass and mixing invariably lead to other possible physical consequences which are important for our overall understanding of the Universe, as well as other possible experimentally verifiable predictions.
## ACKNOWLEDGEMENTS
I thank D. Cline and the other organizers of $`\mu \mu 99`$ for a great meeting. This work was supported in part by the U. S. Department of Energy under Grant No. DE-FG03-94ER40837. |
no-problem/0001/hep-ph0001242.html | ar5iv | text | # 1 Introduction
## 1 Introduction
The typical temperature scales of hadron matter in heavy-ion collisions at CERN-SPS energies, extracted from hadron abundances and hadron transverse momentum spectra, are in the order of $`TT_c𝒪(m_\pi )`$, where $`T_c`$ stands for the expected deconfinement and/or chiral symmetry restoration temperature. It is the chiral symmetry restoration which is thought to cause a drastic reshaping of the low-lying hadron spectrum: In a low-temperature expansion the vector and axial-vector parts of the current-current correlator, which determines the emissivity of matter, are mixed and cause a dilepton spectrum which resembles a simple $`q\overline{q}`$ rate at given temperature for invariant masses $`M>1`$ GeV (cf. fig. 2.9 in ); at lower invariant masses the various interaction processes in hot matter cause a disappearance of pronounced structures emanating from the lowest vector mesons, and also here the resulting spectrum can be approximated by the $`q\overline{q}`$ rate (cf. fig. 4.6 in ). As a result, the descriptions of a dense hadron system and a quark-gluon system become dual to each other.
Here we employ this duality and analyze the dilepton spectra obtained in sulfur and lead beam reactions at the CERN-SPS. Since we use duality (i) in the full invariant mass range considered and (ii) for the full time evolution of matter we phrase our approach as a test of the extended duality hypothesis.
## 2 The model
Since the experiments have no space-time resolution we parameterize the dilepton spectra by
$$\frac{dN}{p_1dp_1p_2dp_2dy_1dy_2d\varphi _1d\varphi _2}=\frac{5\alpha ^2}{72\pi ^5}N_{\mathrm{eff}}\mathrm{exp}\left\{\frac{M_{}\mathrm{cosh}(YY_{cms})}{T_{\mathrm{eff}}}\right\},$$
(1)
where $`p_{1,2}`$, $`y_{1,2}`$ and $`\varphi _{1,2}`$ denote the transverse momenta, rapidities and azimuthal angles of the individual leptons 1 and 2, which must be appropriately combined to construct the pair mass $`M`$, the pair transverse momentum $`Q_{}`$ and transverse mass $`M_{}`$. This expression stems from the $`q\overline{q}`$-annihilation dilepton yield of a source localized at midrapidity $`Y_{cms}`$. The two parameters $`T_{\mathrm{eff}}`$ and $`N_{\mathrm{eff}}`$ are to be adjusted to the experiment. In a more detailed description one has to introduce the space-time dependence of the source parameters such as volume, temperature, chemical potentials and matter velocity (cf. ) and to integrate over the evolution up to freeze-out. In this way the two parameters $`T_{\mathrm{eff}}`$ and $`N_{\mathrm{eff}}`$ are mapped on a much larger parameter space, which however is constrained by hadronic observables and allows a detailed explanation of $`T_{\mathrm{eff}}`$ and $`N_{\mathrm{eff}}`$. According to our experience the detailed space-time models do not improve the description of the data discussed below (cf. vs. ). As matter of fact we mention that the dilepton rate, in particular the transverse momentum spectrum, depends sensitively on the flow of matter when using the correct Lorentz invariant expression . Fortunately, the transverse flow is constrained by hadron data , and for realistic scenarios the flow is masked by background dilepton contributions.
## 3 Analysis of dilepton spectra
### 3.1 Lead beam data
A comparison of the model (1) with lead beam data is displayed in fig. 1. With a uniquely fixed set of $`T_{\mathrm{eff}}=`$ 170 MeV and $`N_{\mathrm{eff}}=3.3\times 10^4`$ fm<sup>4</sup> a fairly well description of the data is accomplished. We mention that the use of the hadronic decay cocktail (dashed curves in the upper panels in fig. 1) and the normalization of $`\frac{dN_{ch}}{d\eta }=250`$ are essential for describing the CERES $`e^+e^{}`$ data in central reactions Pb(158 AGeV) + Au. For a description of the NA50 $`\mu ^+\mu ^{}`$ data in central reactions Pb(158 AGeV) + Pb, the Drell-Yan contribution (dashed curves in the lower panels in fig. 1) and the correlated semileptonic decays of open charm mesons (dot-dashed curves in the lower panels in fig. 1) are needed. The latter ones are generated with PYTHIA (D-’ structure functions, charm mass 1.5 GeV, default fragmentation ”hybrid”), where the Drell-Yan K factor of 1.2 is adjusted to the data and the open charm K factor of 4 to the compilation of identified hadronic charm channels in . Confidence in these important K factors is gained by a comparison with $`\mu ^+\mu ^{}`$ data in the reaction p(450 GeV) + W . Note that we here exploit the approximate NA50 acceptance according to for the lead beam data. To translate the cross sections delivered by PYTHIA into rates we use a thickness function of 31 mb<sup>-1</sup> for central collisions Pb + Pb. Also $`J/\psi `$ and $`\psi ^{}`$ contributions à la are included.
Fig. 1. Comparison of our model with the preliminary CERES data (upper panels, $`Q_{}`$ spectrum for $`M=0.25\mathrm{}0.68`$ GeV) and NA50 data (lower panels, $`Q_{}`$ spectrum for $`M=1.5\mathrm{}2.5`$ GeV). The solid curves are the thermal yields, and the uppermost curves depict the sum of all contributions. For further details see text.
### 3.2 Sulfur beam data
Let us now turn to the older sulfur beam data (cf. for a recent survey). Since a much larger rapidity interval is covered (see fig. 2) we smear the source distribution (1) by a Gaussian function with a width of 0.8. For $`Y_{cms}`$ we choose 2.45 as suggested by an analysis of the hadronic rapidity distribution (cf. fig. 3a). For the CERES $`e^+e^{}`$ data in S(200 AGeV) + Au reactions, $`\frac{dN_{ch}}{d\eta }=125`$ and the published hadronic cocktail is used. The NA38 $`\mu ^+\mu ^{}`$ data in the Drell-Yan region $`M>4.2`$ GeV can be used to pin down the intrinsic transverse momentum distribution of partons. A value of $`k_{}^2(0.8\mathrm{}1)^2`$ GeV<sup>2</sup> is found (see fig. 3b), which we use in all PYTHIA simulations.
Fig. 2. Coverage of the rapidity $`Y_{lab}`$ and transverse mass $`M_{}`$ of the various dilepton experiments. (a) left panel: lead beam; (b) right panel: sulfur beam.
Fig. 3. (a) left panel: Rapidity distribution of negatively charged hadrons (from ) and a Gaussian fit centered at $`y=2.45`$; (b) right panel: Transverse momentum distribution of dileptons in the Drell-Yan region in the NA38 experiment .
Fig. 4. Comparison of our model calculations with the data with separately adjusted normalization factors. Meaning of the curves: (a) upper left panel: lower solid curve with $`\rho `$ hump: cocktail, curve above cocktail: thermal yield; (b) upper right panel: thermal yield, open charm contribution and Drell-Yan (from left to right at larger $`M`$); the uppermost curves in (a, b) display the sum of all contributions; lower panels: solid lines: thermal yield.
While we can nicely reproduce the Drell-Yan background for the HELIOS-3 experiment, our PYTHIA simulations deliver another open charm contribution than the one used in previous analyzes . Since the accurate knowledge of the background contributions is necessary prerequisite, we use therefore for our analysis the difference $`\mu ^+\mu ^{}`$ spectra of S(200 AGeV) + W and p(200 GeV) + W reactions , thus hoping to get rid of the background since these spectra are appropriately normalized.
The comparison of our calculations with the data of CERES and NA38 and HELIOS-3 are displayed in fig. 4 for $`T_{\mathrm{eff}}=160`$ MeV. One observes a fairly well reproduction of the spectral shapes. The available transverse momentum spectrum of NA38 is also nicely reproduced in shape (see fig. 5a). We leave here the normalization $`N_{\mathrm{eff}}`$ as free parameter to achieve an optimum description of the data. It turns out that the found normalization factors for the different data sets are quite different. We therefore conclude that a unique description of the sulfur beam data within statistical errors is not possible even though the target nuclei have very similar masses.
We mention that adjusting the normalization to the CERES data the published upper bounds of the direct photon yields are partially below our model calculations when adopting the model described in section 2 for photons too (see fig. 5b).
Fig. 5. (a) left panel: Comparison of our model calculations with the transverse momentum spectrum of dileptons from NA38 in the intermediate-mass region. (b) right panel: A comparison of the photon spectrum with the experimental upper bounds when adjusting the source strength to the CERES data ; dashed curve: Gaussian smearing of the source, solid curve: point like source.
## 4 Summary
An attempt is reported to explain the dilepton data in recent CERN-SPS experiments with heavy-ion beams. In doing so we employ the duality hypothesis, i.e. we assume that due to strong in-medium effects the hadron spectrum is drastically changed and the resulting dilepton emissivity looks as the one of $`q\overline{q}`$ annihilation. The applicability of this hypothesis relies on the chiral symmetry restoration expected at temperatures achieved in the given beam energy range. While we find a good overall reproduction of the shapes of the experimental spectra, only the lead beam data can be explained with a unique and reasonable normalization.
For next future an analysis of the $`E_{}`$ dependence of the NA50 data is envisaged. |
no-problem/0001/astro-ph0001328.html | ar5iv | text | # Oxygen Absorption in Cooling Flows
## 1. Introduction
The inhomogeneous cooling flow scenario (e.g., Fabian 1994) is often invoked to interpret the X-ray observations of massive elliptical galaxies, groups, and clusters. The key prediction of this scenario is the existence of large quantities of gas that have cooled out of the hot phase and dropped out of the flow. The only evidence for large amounts of mass drop-out arises from the excess soft X-ray absorption from cold gas found for many cooling flows especially from spectral analysis of Einstein and ASCA data (e.g., White et al 1991; Fabian et al 1994; Buote & Fabian 1998; Buote 1999, 2000a).
This interpretation is highly controversial because for systems with low Galactic columns no excess absorption from cold gas is ever found with the ROSAT PSPC which should be more sensitive because of its softer bandpass, 0.1-2.4 keV (e.g., David et al 1994; Jones et al 1997; Briel & Henry 1996). Furthermore, the large intrinsic columns of cold H indicated by the Einstein and ASCA data are in embarrassing disagreement with H i and CO observations (e.g., Bregman, Hogg, & Roberts 1992; O’Dea et al 1994).
We have found new evidence for absorption in the PSPC data during our investigation of the radial metallicity profiles of several of the brightest, nearby cooling flows (Buote 2000b, hereafter PAPER2). In this Letter we present absorption profiles for a subset of these cooling flows obtained from analysis of the deprojected PSPC spectra and briefly compare these results to single-aperture ASCA spectra. We focus on the galaxy, NGC 1399, the group, NGC 5044, and the cluster, A1795 because these systems have (1) low Galactic $`N_\mathrm{H}`$ which facilitates analysis of any intrinsic absorption, (2) the most significant excess columns measured from two-component spectral models with ASCA, and (3) the most clearly significant differences in $`N_\mathrm{H}`$ obtained from PSPC data when energies below $`0.5`$ keV are included/excluded from analysis. The temperature and metallicity profiles, as well as results for a few other systems not satisfying all the above conditions, will be given in PAPER2 and Buote (2000c, hereafter PAPER3).
## 2. Spatially Resolved ROSAT Spectra
We obtained archival ROSAT PSPC observations of NGC 1399, NGC 5044, and A1795 and reduced the data as described in PAPER2. Here we mention that particular attention was given to excluding any significant fluctuations in the light curves since contamination from solar emission would preferentially appear in the low-energy channels. We also developed our own software to properly scale the background spectra to the source positions.
For each object we extracted spectra in concentric circular annuli located at the X-ray centroid such that for each annulus the width was $`1\mathrm{}`$ and the background-subtracted counts was larger than some value chosen to minimize uncertainties on the spectral parameters for each system while maintaining as many annuli as possible. Data with energies $`0.2`$ keV were excluded to insure that the PSF was $`<1\mathrm{}`$ FWHM. For our on-axis sources $`99\%`$ of the PSF at 0.2 keV is contained within $`R=1\mathrm{}`$.
We deproject the data following McLaughlin (1999) who elaborates on the original paper by Fabian et al (1981). The spectral fitting is performed with XSPEC v10.0 using the MEKAL code to model the (single-phase) hot plasma emission. We used the photoelectric absorption cross sections of Balucińska-Church & McCammon (1992). Although Arabadjis & Bregman (1999) point out that the He cross section at 0.15 keV is in error by 13%, since we analyze $`E>0.2`$ keV we find that our fits do not change when using the Morrison & McCammon (1983) cross sections which have the correct He value. Further details of our deprojection analysis are discussed in PAPER2.
### 2.1. Intrinsic Absorption from Cold Gas
No excess absorption from cold gas is found in cooling flows with the PSPC when energies down to $`0.1`$-0.2 keV are included, whereas significant absorption (similar to that obtained by ASCA) is found when the PSPC spectrum is restricted to energies above $`0.5`$ keV (Allen & Fabian 1997; Buote 1999). Presently, the only viable explanation of why large excess absorbing columns are not inferred when energies down to $`0.1`$-0.2 keV are included is that the standard foreground screen model systematically underestimates the true column intrinsic to the cooling flow (Allen & Fabian 1997; Sarazin, Wise, & Markevitch 1998). Allen & Fabian show in their Figure 9 that a simulated PSPC cluster spectrum modified by both a foreground column of $`N_\mathrm{H}=10^{20}`$ cm<sup>-2</sup> plus an intrinsic column of $`N_\mathrm{H}=10^{21}`$ cm<sup>-2</sup> with covering fraction 0.5 will always yield a value similar to the foreground column if fitted only with a standard foreground model. However, such two-component models usually represent the projection of extended gas (with no intrinsic absorption) with centrally concentrated, intrinsically absorbed gas, and thus the suitability of these models can be tested via deprojection.
In Figure 2.1 we display the PSPC spectrum for NGC 1399 within $`R=1\mathrm{}`$ (2D) and show the model representing the projection of gas from $`r>1\mathrm{}`$ (3D) into the $`R=1\mathrm{}`$ aperture modified by a foreground Galactic absorber with solar abundances. The deprojection separates the foreground component (i.e. the projected model in Fig 2.1) from the intrinsic gas and shows that the intrinsic component cannot be absorbed very differently from the outer gas since their spectral shapes below $`0.5`$ keV are so similar; e.g., for an intrinsic absorber with covering factor of 0.5 we obtain a best-fit $`\mathrm{\Delta }N_\mathrm{H}=0`$ and $`\mathrm{\Delta }N_\mathrm{H}<N_\mathrm{H}^{\mathrm{Gal}}`$ at $`>90\%`$ confidence. Since we obtain analogous results for NGC 5044 and A1795, we conclude that the deprojected PSPC spectra for each system rule out absorption by large amounts of intrinsic cold gas (especially H and He) as a viable explanation of the different excess absorbing columns obtained from analyses in different X-ray bandpasses.
### 2.2. Oxygen Edge
A solution to this problem becomes apparent upon examination of how $`N_\mathrm{H}`$varies with the lower energy limit in the centers of cooling flows. For example, in the central $`1\mathrm{}`$ of NGC 1399 we find that $`N_\mathrm{H}N_\mathrm{H}^{\mathrm{Gal}}`$ when $`E_{\mathrm{min}}0.2`$-0.3 keV, though when $`E_{\mathrm{min}}0.4`$ keV $`N_\mathrm{H}`$increases to $`2N_\mathrm{H}^{\mathrm{Gal}}`$. A dramatic change occurs for $`E_{\mathrm{min}}0.5`$ keV where $`N_\mathrm{H}`$increases to several times $`N_\mathrm{H}^{\mathrm{Gal}}`$, although for larger $`E_{\mathrm{min}}`$ $`N_\mathrm{H}`$remains nearly constant. In Figure 1 we plot $`N_\mathrm{H}(R)`$ for $`E_{\mathrm{min}}=0.2`$, 0.5 keV for NGC 1399, 5044, and A1795. The profiles radically differ for each $`E_{\mathrm{min}}`$: $`N_\mathrm{H}(R)N_\mathrm{H}^{\mathrm{Gal}}`$ when $`E_{\mathrm{min}}=0.2`$ keV, but $`N_\mathrm{H}(R)`$ increases significantly at small $`R`$ when $`E_{\mathrm{min}}=0.5`$ keV such that $`N_\mathrm{H}>N_\mathrm{H}^{\mathrm{Gal}}`$ at $`>90\%`$ confidence within $`R=1\mathrm{}`$. (The uncertainties are computed from 100 Monte Carlo realizations of the best-fitting models in each 2D annulus. Our simple procedure for defining confidence limits (see PAPER2) occasionally results in the best-fitting value lying outside the $`1\sigma `$ errors.)
Since $`N_\mathrm{H}(E_{\mathrm{min}})const`$ for $`E_{\mathrm{min}}>0.5`$ keV the portion of the spectrum responsible for the absorption must be $`0.4`$-0.7 keV considering the PSPC resolution. The most important absorption (and emission) features over these energies are due to oxygen. In Figure 1 we also show the deprojection results for $`E_{\mathrm{min}}=0.2`$ keV including an edge at 0.532 keV (rest frame) corresponding to cold oxygen (O i). In this case $`N_\mathrm{H}(R)`$ is very similar to the previous fits with $`E_{\mathrm{min}}=0.2`$ keV and no edge (i.e., similar to Galactic), and the edge optical depth profiles, $`\tau (R)`$, closely resemble $`N_\mathrm{H}(R)`$ when $`E_{\mathrm{min}}=0.5`$ keV.
The large optical depths at small radii and the improvements in the fits when adding the edge are highly significant. In the $`R=1\mathrm{}`$ bins the 95% confidence lower limits on $`\tau `$ are 0.97, 0.59, and 0.21 respectively for NGC 1399, 5044, and A1795. Furthermore, 99 out of 100 Monte Carlo simulations predict $`\tau `$ larger than 0.31, 0.10, and 0.17 respectively; i.e., $`\tau >0`$ is highly significant for these systems in the inner radial bin. The fits are improved substantially when the edge is introduced in the central bin of each system – see Table 1 where we list the values of $`\chi ^2`$, degrees of freedom (dof), and null hypothesis probability $`(P)`$. (Note that for NGC 1399 the edge improves the fits significantly more than by letting the oxygen abundance in the hot gas go to zero: i.e., $`\chi ^2`$ improves only to 95.9 as opposed to 83.7 for the edge.) We emphasize that for the large values of $`\tau 1`$ obtained in the central regions the absorption from the edge affects a large energy range comparable to the resolution of the PSPC over $`0.5`$-0.7 keV; e.g., the O i edge absorbs 25% of the flux at 0.8 keV for $`\tau =1`$.
For most radii the constraints on the edge energy are not very precise which is why we fixed the edge energy in our analysis. If we allow the edge energy to be a free parameter in the central radial bin we obtain $`0.51_{0.05}^{+0.05}`$ keV, $`0.51_{0.05}^{+0.09}`$ keV, and $`0.56_{0.05}^{+0.05}`$ keV (90% confidence) for the edge energies of NGC 1399, 5044, and A1795. These constraints are consistent with the lower ionization states of oxygen but not edges from the highest states O vi-viii. Due to the limited resolution we can add additional edges to share the $`\tau `$ obtained for the O i edges, although even with a two-edge model a significant $`\tau `$ cannot be obtained for edge energies above $`0.65`$ keV corresponding to $``$O vi.
(We mention that the models for NGC 5044 with $`E_{\mathrm{min}}=0.2`$ keV not only imply $`N_\mathrm{H}<N_\mathrm{H}^{\mathrm{Gal}}`$ for small $`R`$ but also predict very large metallicities that are inconsistent with the ASCA data. If $`N_\mathrm{H}`$ is fixed to $`N_\mathrm{H}^{\mathrm{Gal}}`$ then consistent metallicities are obtained as are $`\tau (R)`$ and $`\chi ^2`$ improvements very similar to the free $`N_\mathrm{H}`$case – see Figure 1 and Table 1; see also PAPER2 and PAPER3.)
## 3. Single-Aperture ASCA Spectra
The intrinsic oxygen absorption indicated by the PSPC data is most significant in the central $`1\mathrm{}`$ which is much smaller than the width of the ASCA PSF. In addition, since the ASCA SIS is limited to $`E>0.5`$ keV, and the efficiency near 0.5 keV is limited due to instrumental oxygen absorption, it cannot be expected that ASCA can distinguish an oxygen edge from an absorber with solar abundances. Nevertheless, we briefly summarize the results obtained when adding an oxygen edge to the ASCA data.
Previously we (Buote 1999) have fitted two-temperature models to the accumulated ASCA SIS and GIS data within $`R5\mathrm{}`$ of NGC 1399 and 5044 and obtained $`N_\mathrm{H}^\mathrm{c}=49_9^{+6},25_6^{+5}\times 10^{20}`$ cm<sup>-2</sup> respectively for the cooler temperature components. Examination of Figure 1 shows that these columns are $`2`$ times the values in the central arcminute and are consistent with the total columns within $`R5\mathrm{}`$ obtained from the PSPC for $`E_{\mathrm{min}}=0.5`$ keV. If instead we add an O i edge to the cooler ASCA model components (while keeping $`N_\mathrm{H}`$ fixed to Galactic on each component) we obtain (1) fits of comparable quality (slightly better) to the original models and (2) optical depths that agree with the cumulative values obtained from the PSPC.
For A1795 we have re-analyzed the ASCA data within $`R4\mathrm{}`$ and obtained results very similar to Fabian et al (1994) for two-temperature models. As above, we obtain fits with an oxygen edge supplying the excess absorption that are as good as those obtained with an excess absorber having solar abundances. Unlike the ellipticals and groups, for the cluster A1795 the cooler temperature component contributes only $`1/7`$ to the emission measure, and thus its fitted parameters are not nearly as well constrained as for NGC 1399 and 5044; i.e., although the best-fitting $`\tau =2.5`$, the 90% lower limit is 0.6 which is comparable to the total optical depth inferred from the PSPC (Fig 1).
Hence, the oxygen edge provides as good or better description of the excess absorption inferred from multitemperature models of ASCA data within the central few arcminutes of NGC 1399, 5044, and A1795 as a cold absorber with solar abundances, and yields optical depths that are consistent with those obtained with the PSPC data.
## 4. Discussion
Using spatially resolved, deprojected X-ray spectra from the ROSAT PSPC we have detected strong oxygen absorption intrinsic to the central $`1\mathrm{}`$ of the elliptical galaxy, NGC 1399, the group, NGC 5044, and the cluster, A1795 which are amongst the largest nearby cooling flows in their respective classes and have low Galactic columns (e.g., Fabian et al 1994; Buote & Fabian 1998; Buote 1999, 2000a). Modeling the oxygen absorption with an edge (rest frame $`E=0.532`$ keV) produces the necessary absorption in both the PSPC and ASCA data for $`E0.5`$ keV without violating the PSPC constraints over $`0.20.3`$ keV for which no significant excess absorption is indicated. This reconciles many reported discrepancies between absorbing columns inferred from ROSAT with those obtained from ASCA and other instruments with bandpasses above $`0.5`$ keV. Moreover, there is no need for large absorbing columns of cold H which are known to be very inconsistent with the negligible atomic and molecular H measured in cooling flows (e.g., Bregman et al 1992; O’Dea et al 1994).
Since no excess absorption is detected over $`0.20.3`$ keV, and since the edge energies allow for ionized states of oxygen, the most reasonable model for the absorber is warm, collisionally ionized gas with $`T=10^{56}`$ K. At these temperatures the majority of the absorption arises from oxygen but with a sizeable contribution from ionized carbon and nitrogen (see PAPER3).
Arnaud & Mushotzky (1998) have reported the detection of an O i edge in the Perseus cluster using BBXRT data. Their cooling flow model (which has stronger oxygen lines than two-temperature models) requires an edge energy that is consistent only with cold oxygen, and thus warm ionized gas could not be responsible for the absorption. However, it was not shown that multiple edges from different ionization states are inconsistent with the data, and thus we expect that a multi-edge model would fit as well as a single O i edge since the BBXRT has lower energy resolution than ASCA. (We do not find any evidence for excess oxygen absorption in Perseus from the PSPC data probably because of the large Galactic column, but $`\tau 2`$ for the O i edge obtained by Arnaud & Mushotzky is similar to what we found for A1795 with ASCA.)
The warm ionized gas implied for NGC 1399, 5044, and A1795 (and possibly Perseus) could be the much sought-after mass that has dropped out of the cooling flows. In PAPER3 we show that the absorber masses indicated by the oxygen edge optical depths are consistent with the mass expected to have been deposited by the cooling flow over the lifetime of the system, while the emission expected from the warm gas does not violate published optical and UV constraints. Understanding the details of how warm ionized gas is maintained and supported in such large quantities is a serious theoretical challenge.
Since detection of the warm ionized gas at other wavelengths is difficult, e.g., background QSOs have not been found within $`R=1\mathrm{}`$ of strong cooling flows (Miller, Bregman, & Knezek 2000), confirmation will probably have to be realized in the X-ray band. Fortunately, the CCDs of the Chandra and XMM missions have the combined spatial and spectral resolution and include energies down to $`0.1`$ keV to verify our prediction of warm ionized gas in these and other cooling flows. The $`10`$ eV resolution of the ASTRO-E XRS will place strong constraints on the allowed ionization states of oxygen. Hence, data from these new missions will elucidate the properties of the warm ionized gas and thus the role of cooling flows in the formation and evolution of ellipticals, groups, and clusters.
I thank M. Bolte and the referee for comments on the manuscript. Support for this work was provided by NASA through Chandra Fellowship grant PF8-10001 awarded by the Chandra Science Center, which is operated by the Smithsonian Astrophysical Observatory for NASA under contract NAS8-39073. |
no-problem/0001/astro-ph0001184.html | ar5iv | text | # 1 Introduction
## 1 Introduction
One of the most significant properties of galaxy clusters is the relation between their dynamical state and the underlying cosmology. In an open universe, clustering effectively freezes at high redshifts and clusters today should appear more relaxed with weak or no indications of substructure. Instead, in a critical density model, such systems continue to form even today and are expected to be dynamically active. The percentage and morphologies of disordered objects in a cluster sample could lead to crucial constraints on $`\mathrm{\Omega }_{}`$ and $`\mathrm{\Lambda }`$, especially if combined with N-body/gas-dynamic numerical simulations spanning different dark matter (DM) scenarios (; ; ).
A large number of relevant analyses have been devoted to this study and an accordingly varying and large number of optical and X–ray cluster compilations have been utilised to this aim (; ; and references therein). In the present work, we use a sample of 22 galaxy clusters (APM and ROSAT) in a complementary fashion with the aim to address the following two questions:
* Is substructure in the X–ray also corroborated by the optical observations and in what percentage?
* What is the percentage of systems depicting strong indications of subclumping and what does it imply for the existing cosmology?
## 2 Data & Methodology
The present dataset follows from a double cross-correlation between rich ACO clusters ($`\mathrm{R}`$1,2,3) with the APM cluster catalogue and the X–ray (0.1 - 2.4) keV ROSAT pointed observations archive, finally resulting in 27 common entries. Due to problematic regions of the APM catalogue, low signal to noise X–ray observations and contamination by known foreground or background objects, we exclude 5 clusters reducing our cluster sample to 22 systems. The redshift range of our sample is $`0.04z0.13`$ with $`z0.074`$ and median $`0.069`$. For the needs of our analysis we transform cluster redshifts to distances using the luminosity-distance relation for a critical density model, $`q_{}=0.5`$ and $`H_{}=100h`$ km $`s^1`$ Mpc<sup>-1</sup>.
In order to construct a common comparison base, we create a continuous density field for both optical and X–ray data by using a Gaussian Kernel and a variable smoothing length according to each cluster redshift. However, so as to take into account the reduction of the number of cluster members as a function of distance (due to the APM magnitude limit), and thus the corresponding increase of discreteness effects, we have investigated, using Monte-Carlo cluster simulations, the necessary size of the smoothing window in order to minimise such effects and optimize the performance of our procedure (cf. ). We then compute the optical and X–ray cluster shape parameters utilising the method of moments of inertia. The eigenvalues and the eigenvectors of the inertia tensor can provide us with the cluster ellipticities and major axis orientations (position angles) respectively. We also define the centroid shift as the vectorial difference between the weighted cluster center-of-mass and the highest cluster density peak (cf. , ). These shape parameters are estimated using all cells that have densities above three thresholds. These are defined as the average density of all cells that fall within a chosen radius. The three radii used are $`r_\rho =0.3,0.45`$ and $`0.6h^1`$ Mpc, whereas the maximum searching radius for all subsequent calculations is $`0.8h^1`$Mpc. We finally utilise a friend-of-friends algorithm to investigate possible substructure by joining all cells having common boundaries and fall above each density threshold. We therefore create and register all subgroups as a function of density threshold and rank substructure according to different criteria (cf. ).
## 3 Quantifying substructure results
Looking at the cluster shape parameters and visually inspecting the isodensity contour maps (see Figure 1 for a subsample of 4 objects), we do observe a remarkable 1-to-1 correspondence in $`80\%`$ of our sample regarding the gross structural features (prime and secondary components, elongations, irregular activity, collision vestiges, unimodality). The majority of the optical and X–ray images are very well aligned with $`\delta \theta 20^{}`$ and relative correlation coefficient of order of $`0.9`$, which is also highly significant. Furthermore, the ellipticities and the centroid shifts between optical and X–ray data do correlate well with coefficient of order $`0.7`$ in both cases. On the other hand, we have found important intrinsic correlations between ellipticities and centroid variations in the optical and the X–ray configurations separately, with relative coefficients ranging from 0.6 to 0.8 respectively. Cross-correlating the optical and X–ray substructure measures, we discover that they also correlate nicely. Probably the most interesting correlation is that between optical centroid shifts and X–ray ellipticities, with a value exceeding 0.8. This indicates that we can deduce the shape of the DM gravitational potential from optical cluster data.
Since random density fluctuations as well as background contamination may introduce spurious substructure, we quantify the significance of our substructure measures, as revealed by the center of mass shift in the optical, using Monte Carlo cluster simulations with the same number of galaxies, ellipticity and estimated background as that of each cluster in our sample. This significance is estimated by measuring the deviation of the true cluster center-of-mass shift, from the corresponding simulated value in units of the estimated $`\sigma `$ from 100 Monte-Carlo simulations of each cluster. We also compare this significance measure to the results of the subgroup statistics algorithm. The two measures are significantly correlated with a value of $`>0.7`$.
Finally we classify our clusters according to their morphological parameters using a scheme which is very close to the one developed by . Results indicate that our findings are in very good agreement with those of both on a quantitative and qualitative basis. Note also, that 11 out of our 22 clusters have been examined for substructure signals elsewhere in the literature. We have checked that our computations on the cluster shape parameters do accord with those of the other studies. (; ; )
From our prime substructure analysis we confirm that at least 9 out of 22 systems display strong substructure indications visible in both parts of the spectrum. We also find that 4 clusters ($`20\%`$) show clear disparities between the optical and X–ray maps, with apparent substructure in the optical not corroborated by the X–ray data. The rest of our sample exhibits no or insignificant substructure indications. We finally observe that our present study is compatible with that of (their Figure 2) regarding the cluster substructure frequency, setting a rather frail lower limit on the density parameter ($`\mathrm{\Omega }_{}\mathrm{\hspace{0.17em}0.5}`$).
In the near future we plan to apply the methodology of this work to the large optical APM sample of galaxy clusters ($`>900`$ entries), in order to investigate in more detail the issue of cluster substructure.
Acknowledgements. V. Kolokotronis and S. Basilakos wish to acknowledge financial support from the Greek State Fellowship Foundation. |
no-problem/0001/cond-mat0001454.html | ar5iv | text | # Quasiparticle resonant states as a probe of short-range electronic structure and Andreév coherence
## Abstract
The recently observed properties of quasiparticle resonant states near impurities on the surface of superconducting Bi<sub>2</sub>Sr<sub>2</sub>CaCu<sub>2</sub>O<sub>8+δ</sub> demonstrate that in-plane Andreév processes are either absent or phase-incoherent. Analysis of the spectral and spatial details of the electronic structure near a Zn impurity also suggest an effective magnetic component of the impurity potential. Further experiments are proposed to clarify whether the effective moments of nearby impurities are correlated.
Over the past few years several authors have emphasized the wealth of information available from local probes of impurity properties in correlated electron systems, and particularly in superconductors whose homogeneous order parameters (OP) are anisotropic in momentum. A parallel improvement in scanning tunneling spectroscopy (STS) has allowed this vision to become a reality through direct observation of the local density of states (LDOS), first in niobium, which has a momentum-independent OP, and this year in the high-temperature superconductor Bi<sub>2</sub>Sr<sub>2</sub>CaCu<sub>2</sub>O<sub>8+δ</sub> (BSCCO), which has an anisotropic OP. The electronic structure of the BSCCO surface is much more complex than that of the niobium surface; there are local moments in the copper-oxygen planes and, under certain conditions, a pseudogap state. Recent work has emphasized the role of the pseudogap state in determining the properties of clean surfaces of high-temperature superconductors at temperatures near T<sub>c</sub>, and STS and photoemission have directly demonstrated its existence on the surface of BSCCO. The pseudogap state is characterized by a single-particle gap of $`d_{x^2y^2}`$ symmetry, but the Andreév processes one would expect in a superconducting state are (according to differences in mechanism) either absent or phase-incoherent.
Here the STS measurements near impurities will be shown to unambiguously demonstrate essentially complete suppression of Andreév processes in the vicinity of the impurity, even at very low temperatures. This low-temperature indication of a pseudogap implies its importance to the nature of the superconducting state and the operation of devices (such as Josephson junctions) with such materials. Indications of an effective magnetic component can also be seen in Ref. . Thus STS provides a direct probe of both the local Andreév (superconducting) coherence and the local magnetic properties on the BSCCO surface.
A brief review of the experimental results from Refs. is in order. Theoretical predictions which have been confirmed include the presence of quasiparticle resonances near nonmagnetic impurities in anisotropic OP superconductors, as well as the suppression of the gap feature near the impurity and the asymmetry of the resonance peak due to the energy dependence of the quasiparticle density of states. The disagreements with previous theory, however, are striking. The most noticeable one is that the resonant state has only been detected on the hole side of the spectrum, both on the impurity site and everywhere else around the impurity. Whereas previous theories are consistent with an LDOS measured at the impurity which is entirely hole-like, these same theories unambiguously predict the LDOS at nearest-neighbor states will be almost entirely electron-like. Indeed these theories predict the spatially-integrated LDOS (or DOS) will be nearly particle-hole symmetric even though the LDOS at any particular site is not. A second unexpected element in the data is the presence of a second, much smaller, spectral peak on the hole side in Ref. . Whereas previous theories are consistent with two resonances at a single impurity, the spatially-integrated amplitude of each resonance should be approximately the same, unlike what is seen in Ref. .
The above two issues are disagreements between the experimental and theoretical DOS, but there are also two significant disagreements in the LDOS. The resonance has a large amplitude at the impurity site, whereas calculations indicate that the largest amplitude in the LDOS should occur at the nearest-neighbor sites. Finally, the gap feature is seen on the impurity site, where it does not appear in calculations.
The calculations here incorporate the pseudogap state into the evaluation of the differential conductance ($`dI/dV`$). After evaluating the LDOS for several impurity potential models, allowing for spatial extent and magnetic character in the potential, the impurities of Refs are found to be highly localized and the four discrepancies above can be reconciled. Finally, in the case of Ref. , a magnetic component to the effective Zn impurity potential is evident.
The calculations of the LDOS are based on the Hamiltonian
$$H=\underset{ij,\sigma }{}\left[t_{ij}c_{i\sigma }^{}c_{j\sigma }+\mathrm{\Delta }_{ij}c_i^{}c_j^{}+\mathrm{\Delta }_{ij}^{}c_jc_i\right]+\underset{i}{}\left[(V_{0i}+V_{Si})c_i^{}c_i+(V_{0i}V_{Si})c_i^{}c_i\right],$$
(1)
which includes a site-dependent potential which can be magnetic ($`V_S`$), nonmagnetic ($`V_0`$), or a combination of both. $`i`$ and $`j`$ label sites and $`\sigma `$ labels spin. The homogeneous electronic structure, expressed as hopping matrix elements ($`t_{ij}`$) for the first five nearest neighbors, is taken from a single-band parametrization of photoemission data. Large variations in hopping matrix elements ($`>50`$ meV) produce results very much at odds with experiment, whereas smaller variations are mimicked by the site-dependent potential. Hence such changes will be ignored here. Only on-site and nearest-neighbor order parameters $`\mathrm{\Delta }_{ij}`$ are nonzero, and the maximum OP on the Fermi surface, $`\mathrm{\Delta }_{max}=40`$ meV.
The electronic structure of the inhomogeneous system (including the impurity) is determined by direct numerical solution in real space of the Gor’kov equation (in Nambu form) for the inhomogeneous Green’s function, $`𝐆=𝐠+\mathrm{𝐠𝐕𝐆}=(𝐈\mathrm{𝐠𝐕})^1𝐠`$, within a real-space region around the impurity beyond which the potential is negligible. The $`\mathrm{\Delta }_{ij}`$’s are found self-consistently in this process, for they determine the off-diagonal components of the potential $`𝐕`$. Spectra outside this real-space region are constructed according to the generalized $`𝐓`$-matrix equation: $`𝐆=𝐠+\mathrm{𝐠𝐕}\left[𝐈\mathrm{𝐆𝐕}\right]^1𝐠`$. Once $`𝐆`$ has been calculated throughout the region near the impurity, the LDOS and DOS are obtained from its imaginary part and the lattice Wannier functions. Then
$$\frac{dI(𝐱;V)}{dV}=𝑑\omega \underset{\sigma }{}\frac{1}{\pi }\left(\frac{n_{STM}(\omega )}{\omega }\right)|\varphi _\sigma (𝐱;i)|^2\mathrm{Im}G_\sigma (i,i;\omega ),$$
(2)
where $`\varphi _\sigma (𝐱;i)`$ is the overlap of the Wannier function at site $`i`$ and spin $`\sigma `$ with the STM tip at $`𝐱`$, and $`n_{STM}(\omega )`$ is the occupation function of the STM tip. Resonances correspond to new peaks in the differential DOS (the difference between the inhomogeneous and homogeneous DOS); their energies are shown in Fig. 1(ab) for magnetic and nonmagnetic single-site impurities.
If Andreév processes are suppressed, either by reduction in their amplitude or phase coherence, a resonance’s DOS will become more electron-like or more hole-like. Reduction of the amplitude of the homogeneous anomalous Green’s function $`f(i,j;\omega )`$, due perhaps to a local antiferromagnetic (AF) order, decreases the mean-field coupling between electron and hole excitations. Note that this is very different from the fully electron or hole-like character of the LDOS at the impurity, which originates from a vanishing $`f(i,i;\omega )`$ in the $`d_{x^2y^2}`$ state. Figure 1(cd) shows the DOS of a resonance for three systems with a 40 meV $`d_{x^2y^2}`$ gap: a fully superconducting gap (solid line), a gap with a 25 meV superconducting component (dashed line), and a pseudogap with no superconducting component (dot-dashed line). As the superconducting component is reduced, the electron-hole symmetry diminishes. The nonmagnetic potentials of Fig. 1(c) are chosen ($`1.375`$ eV, $`1.000`$ eV, and $`0.833`$ eV, respectively) so the resonance peak is at $`1.5`$ mV (the same as Ref. ). The magnetic potentials in Fig. 1(d) are the same as those in Fig. 1(c).
The reduction of the electron-like peak from phase decoherence is similar to the effect of amplitude suppression of $`f(i,i;\omega )`$. For a resonance at $`\omega `$, the peak at $`\omega `$ comes from terms with products of pairs of anomalous Green’s functions. For a phase incoherent pseudogap state the expectation value of these pairs, and thus the amplitude of the electron-like peak, is diminished. The effect of this is shown in Fig. 1(ef) as a dashed line corresponding to partial (half) and a dot-dashed line corresponding to no phase coherence. The nonmagnetic potential is $`1.375`$ eV in Fig. 1(e) and the magnetic potential is $`1.375`$ eV in Fig. 1(f). Note that for the purely magnetic impurities even in the absence of Andreév coherence there is a peak on each side of zero energy.
For Refs. there is no apparent electron-like component of the resonance in the DOS, thus local Andreév coherence is absent. In measurements of the LDOS near metal islands on BSCCO, however, both hole-like and electron-like peaks are apparent. This may indicate that the metal plays an important role in maintaining phase coherence at the surface, or that the metal overlayer is less disruptive to superconductivity than impurities in the plane. The presence of the in-plane Andreév processes, indicated by the electron-like peak, is essential to the operation of Josephson junctions.
The on-site LDOS of Ref. is shown in Fig. 2(a). The second (hole-like) peak is not as clearly evident in the results of Refs. , and thus may be peculiar either to the Zn impurity or to the impurity site in the BSCCO unit cell. Additional resonances around impurities can originate from additional orbital states around spatially-extended potentials or from spin-splitting near magnetic potentials. Note that an effective magnetic potential could also originate from a nonmagnetic impurity potential placed in a spin-polarized host electronic structure.
Figure 2(abc) shows the best fit of $`dI/dV`$ to the data of Ref. for phase incoherent Andreév processes and i a single-site nonmagnetic potential (1.375 eV), ii a nonmagnetic potential with onsite (0.360 eV) and nearest-neighbor (0.150 eV) values, and iii a mixed nonmagnetic and magnetic potential ($`V_0=0.825`$ eV, $`V_S=0.550`$ eV). Also shown is the best fit using iv a pseudogap with no superconducting component and a mixed potential ($`V_0=0.543`$ eV, $`V_S=0.290`$ eV). The three panels show $`dI/dV`$ (a) at the impurity site, (b) at the nearest-neighbor site along the gap nodes, and (c) along the gap maxima. Measurements of Ref. for (a) and (b) are shown; (c) is not available.
The large size of the resonance on-site and the simultaneous presence of the gap feature occur because of junction normalization (equal resistance at $`200`$ mV) and the finite width of the $`|\varphi _\sigma |^2`$ (modeled as Gaussians of range $`0.8`$Å, $`3.8`$Å, $`0.8`$Å, and $`1.0`$Å for i-iv). The very small LDOS in this energy range at the impurity site causes the tip to approach closer to the surface and (1) enlarge the apparent size of the resonance on-site, and (2) pick up the gap features from the nearest-neighbor sites. The relative size of the on-site resonance to the gap features is largely determined by the overlap of the nearest-neighbor Wannier functions with the tip when the tip is over the impurity site.
Judging from the comparison with experiment, iii and iv appear most in agreement. i does not have a second resonance, and whereas ii does show one in the proper location, its relative magnitude is incorrect. The smaller amplitude of the second resonance is obtained for iii and iv because the overlaps with the STM tip are spin-dependent ($`|\varphi _{}(𝐱)|^2/|\varphi _{}(𝐱)|^240`$). If the second peak were absent, either the impurity would lack magnetic character, or it would occur in less magnetic regions of the BSCCO unit cell. The remaining disagreement is in the amplitude of the resonance in (B), where iv is best, but still too small.
Figure 3 shows the amplitude of the resonance as a function of distance from the impurity along the gap maxima (a) and the gap nodes (b). The squares are the data from Ref. , whereas the solid line corresponds to iii, the dotted line to ii, and the dot-dashed line to iv. The plot for i looks identical to that of iii. The agreement of iii and iv are quite good along the maxima direction. The absence of a well-defined maximum at the nearest-neighbor in (a), which was pointed out in Ref. , is due to the normalization procedure. An inset in Fig. 3 shows the difference between the junction normalized (solid) and unnormalized (dashed) $`dI/dV`$. The main discrepancy is with the amplitude of the signal along the node directions (b).
Figure 4 shows the $`dI/dV`$ of the resonance for iv; iii is similar. The differences between the Figs. 2-4 and the measurements of Ref. may be due to errors in the homogeneous electronic structure of BSCCO used in the calculation, particularly the low-energy electronic structure which dominates the longer-range LDOS. These errors may be due to inaccuracies in the model for the electronic structure measured by photoemission, or they may be due to the neglect of other collective effects on the surface. Another likely source of error is that the electronic structure model of the host is not spin dependent.
One of the possible mechanisms of a pseudogap is local AF order, such as occurs in a stripe. The magnetic component apparent in the impurity potential suggests this origin as well. If two nearby local moments are aligned parallel, then the resonances associated with them will hybridize and split, whereas if they are antiparallel the resonances will be degenerate. A careful examination of the $`dI/dV`$ for two Zn atoms near each other on the surface may clarify whether there is local AF order.
The LDOS reported in Refs. are best explained by the presence of a pseudogap state on the surface of BSCCO. The relative height of the electron-like and hole-like resonances in the DOS depends directly on the amplitude of local Andreév processes, and thus shows the degree of local superconducting coherence. This is of great practical interest, for the presence of these processes is essential to forming a proper Josephson junction across an interface. The information obtained about the superconducting state at the surface of BSCCO indicates the clear promise of future STM measurements near defects in other correlated electron systems.
I would like to thank J. C. Davis for discussions and for providing the data of Ref. . This work has been supported in part by ONR through contract No. N00014-99-1-0313. |
no-problem/0001/hep-ph0001135.html | ar5iv | text | # Resummation of Nonalternating Divergent Perturbative Expansions
## Abstract
A method for the resummation of nonalternating divergent perturbation series is described. The procedure constitutes a generalization of the Borel-Padé method. Of crucial importance is a special integration contour in the complex plane. Nonperturbative imaginary contributions can be inferred from the purely real perturbative coefficients. A connection is drawn from the quantum field theoretic problem of resummation to divergent perturbative expansions in other areas of physics.
In view of the probable divergence of quantum field theory in higher order , the resummation of the perturbation series is necessary for obtaining finite answers to physical problems. While the divergent expansions probably constitute asymptotic series , it is unclear whether unique answers can be inferred from perturbation theory . Significant problems in the resummation are caused by infrared (IR) renormalons. These are contributions corresponding to nonalternating divergent perturbation series. The IR renormalons are responsible for the Borel-nonsummability of a number of field theories including quantum chromodynamics (QCD) and quantum electrodynamics (QED) .
Here I advocate a modification of the resummation method proposed in for nonalternating divergent perturbation series. The method starts with a given input series,
$$f(g)\underset{n=0}{\overset{\mathrm{}}{}}c_ng^n,c_n>0,g>0,$$
(1)
where $`g`$ is the coupling parameter and the perturbative coefficients $`c_n`$ are expected to diverge as follows ,
$$c_nK\frac{n!n^\gamma }{S^n},n\mathrm{},$$
(2)
with $`K`$, $`\gamma `$ and $`S`$ being constant. The Borel transform $`f_\mathrm{B}`$ of the perturbation series (1)
$$f_\mathrm{B}(g)=\underset{n=0}{\overset{\mathrm{}}{}}\frac{c_n}{n!}g^n$$
(3)
has a finite radius of convergence about the origin. For the evaluation of the Borel integral, $`f_\mathrm{B}(g)`$ has to be continued analytically beyond the radius of convergence. Strictly speaking, this analytic continuation has to be done on the branch cut in view of the nonalternating character of the series (1). This requirement can be relaxed slightly by performing the analytic continuation into regions where $`g`$ acquires at least an infinitesimal imaginary part $`gg\pm \mathrm{i}ϵ`$. In this case, the analytic continuation can be achieved by evaluating Padé approximants . The first $`n+1`$ terms of the Borel transformed series (3) can be used to construct a diagonal or off-diagonal Padé approximant (for the notation see )
$$𝒫_n(z)=\left[[[n/2]]/[[(n+1)/2]]\right]_{f_\mathrm{B}}\left(z\right),$$
(4)
where $`[[x]]`$ denotes the integral part of $`x`$. The resummation is accomplished by constructing the sequence of transforms $`\{𝒯f_n(g)\}_{n=0}^{\mathrm{}}`$ where
$`𝒯f_n(g)`$ $`=`$ $`{\displaystyle _{C_j}}dt\mathrm{exp}(t)𝒫_n(gt),`$ (5)
and the integration contour $`C_j`$ ($`j=1,0,+1`$) is as shown in Fig. 1 (for $`j=1`$ and $`j=+1`$). The result obtained along $`C_1`$ is the complex conjugate of the result along $`C_{+1}`$. The arithmetic mean of the results of the integrations along $`C_1`$ and $`C_{+1}`$ is associated with $`C_0`$. Therefore, the result along $`C_0`$ is real rather than complex. The limit of the sequence $`\{𝒯f_n(g)\}_{n=0}^{\mathrm{}}`$ (provided it exists),
$$\underset{n\mathrm{}}{lim}𝒯f_n(g)=f(g),$$
(6)
is a plausible complete nonperturbative result inferred from the perturbative expansion (1). Which of the contours $`C_j`$ ($`j=1,0,+1`$) is chosen, has to be decided on the basis of additional considerations which do not follow from perturbation theory alone.
The zeros of the denominator polynomial of the Padé approximant \[see Eq. (4)\] correspond to the poles of the integrand in Eq. (5). Denote by $`t`$ the integration variable for the evaluation of the generalized Borel integral in Eq. (5), then the poles lie at $`t=z_i`$ (where the index $`i`$ numbers the poles) along the positive real axis ($`\mathrm{Im}z_i=0`$) and in the complex plane ($`\mathrm{Im}z_i0`$). The poles lying on the positive real axis are treated as half-poles encircled in the mathematically positive sense for $`C_1`$ and as half-poles encircled in the mathematically negative sense for $`C_{+1}`$. The contour $`C_1`$ encircles all poles at $`t=z_i`$ in the lower right quadrant of the complex plane ($`\mathrm{Re}z_i>0`$, $`\mathrm{Im}z_i<0`$) in the positive sense (see Fig. 1). The contribution of these poles should be added to the final result. The contour $`C_{+1}`$ is understood to encircle all poles in the upper right quadrant of the complex plane in the mathematically negative sense. In general, the integrations along $`C_1`$ and $`C_{+1}`$ lead to a nonvanishing imaginary part in the final result for $`f(g)`$ \[see Eq. (6)\], although all the perturbative coefficients $`c_n`$ are by assumption real and positive \[see Eq. (1)\]. It might be interesting to note that, as with any complex integration, it is permissible to deform the integration contours shown in Fig. 1 in accord with the Cauchy Theorem as long as all pole contributions are properly taken into account.
This paper represents a continuation of previous work on the subject . The resummation method defined in Eqs. (1)–(6) differs from in the combination of Borel and Padé techniques and, if compared to the remarkable investigations in on the resummation of QCD perturbation series, in the integration contour used for the evaluation of the generalized Borel integral. It is argued here that, when the Borel transform (3) is analytically continued with Padé approximants (4), the contribution of poles lying off the positive real axis has to be taken into account in order to obtain consistent results in the resummation (see Fig. 1). In it is argued that the Borel integral should be evaluated by principal value. It could appear that the $`C_0`$ contour corresponds to the principal-value prescription. However, this is not necessarily the case, if there are poles present which lie off the positive real axis (i.e., at $`t=z_i`$ with $`\mathrm{Re}z_i>0`$, $`\mathrm{Im}z_i0`$). The contribution of these poles not only modifies the imaginary, but also the real part of the final nonperturbative result. Of course, when there are no poles lying off the positive real axis, as it is the case for the problems discussed in , then the principal-value prescription used in is equivalent to the $`C_0`$ contour. Because the result obtained along $`C_0`$ is real, this contour should be used whenever the existence of an imaginary part is discouraged by physical reasons.
It is important to mention that the method presented here is not the only prescription currently available for the resummation of divergent perturbative expansions in quantum field theory. For example, the $`\delta `$ transformation (see Eq. (4) in ) is a very useful method for the resummation of divergent perturbation series. The $`\delta `$ transformation has a number of appealing mathematical properties, including rapid and numerically stable convergence, and it has been shown to yield consistent results in many cases, including applications from quantum field theory and from other areas of physics . Because the $`\delta `$ transformation fulfills an accuracy-through-order relation (see Eq. (9) in ), it can be used to predict perturbative coefficients. The $`\delta `$ transformation is primarily useful for alternating series. It fails, in general, in the resummation of the nonalternating series discussed here. The $`\delta `$ transformation and the resummation method introduced here complement each other.
Three applications of the resummation method defined in Eqs. (1)–(6) are considered below: (i) the QED effective action in the presence of a constant background electric field, (ii) a mathematical model series which simulates the expected large-order behavior of perturbative coefficients in quantum field theory, (iii) the perturbation series for the energy shift of an atomic level in a constant background electric field (including the auto-ionization width). The nonperturbative imaginary contributions obtained along $`C_1`$ and $`C_{+1}`$ find a natural physical interpretation in all cases considered.
The QED effective action, or vacuum-to-vacuum amplitude, in the presence of a constant background electric field has been treated nonperturbatively in , and the result is proportional to the integral
$$S(g_\mathrm{E})=\underset{0\mathrm{i}ϵ}{\overset{\mathrm{}\mathrm{i}ϵ}{}}\frac{\mathrm{d}s}{s^2}\left\{\mathrm{cot}s\frac{1}{s}+\frac{s}{3}\right\}\mathrm{exp}\left[\frac{1}{\sqrt{g_\mathrm{E}}}s\right],$$
(7)
where $`g_\mathrm{E}`$ is a coupling parameter proportional to the square of the electric field strength, $`g_\mathrm{E}=e^2E^2/m_\mathrm{e}^4`$. Here, $`m_\mathrm{e}`$ is the electron mass, and $`e`$ is the elementary charge. The natural unit system ($`\mathrm{}=c=1`$) is used. The imaginary part of $`S(g_\mathrm{E})`$ is proportional to the electron-positron pair production amplitude per space-time interval \[there is of course also a muon-antimuon pair-production amplitude, obtained by the imaginary part of (7) under the replacement $`m_\mathrm{e}m_\mu `$, which is not discussed here\]. $`S(g_\mathrm{E})`$ has the following asymptotic expansion in the coupling parameter,
$$S(g_\mathrm{E})16\left[\underset{n=0}{\overset{\mathrm{}}{}}\frac{4^n|_{2n+4}|}{(2n+4)(2n+3)(2n+2)}g_\mathrm{E}^{n+1}\right]$$
(8)
where $`_{2n+4}`$ is a Bernoulli number. In view of the asymptotics
$$\frac{4^n|_{2n+4}|}{n^3}\frac{\mathrm{\Gamma }(2n+2)}{\pi ^{2n+4}}\left[1+\mathrm{O}\left(\frac{1}{n}\right)\right],n\mathrm{},$$
(9)
the perturbative coefficients, which are nonalternating in sign, diverge factorially in absolute magnitude. The asymptotic series (8) for $`S(g_\mathrm{E})`$ is taken as the input series for the resummation process \[Eq. (1)\], and a sequence of transforms $`𝒯S_n(g_\mathrm{E})`$ is evaluated using the prescription (5). The results have to be compared to the exact nonperturbative expression (7). This is done in Table I for $`g_\mathrm{E}=0.05`$. The partial sums of the asymptotic series (8) are listed in the second column.
Numerical results from perturbation theory are normally obtained by (optimal) truncation of the perturbation series. For the example considered, (i) the partial sums do not account for the imaginary part and (ii) due to the divergence of the perturbative expansion, no improvement in the final result could be obtained by adding more than the first 7 perturbative terms. It requires a valid resummation procedure to go beyond the accuracy obtainable by optimal truncation of the perturbation series. The transforms $`𝒯S_n(g_\mathrm{E})`$ displayed in the third column of Table I apparently converge to the full nonperturbative result given in Eq. (7), and the nonperturbative imaginary part, which corresponds to the pair production amplitude, is reproduced although the input series (8) has purely real perturbative coefficients.
Two specific mathematical model series are considered next. The series
$`𝒩(g){\displaystyle \underset{n=0}{\overset{\mathrm{}}{}}}n!g^n`$
has been used as a paradigmatic example for nonalternating divergent series in the literature . This series can be resummed by the method (5). Moreover, this resummation is even exact for all transformation orders $`n2`$. This can be seen as follows. The Borel transform
$`𝒩_\mathrm{B}(g)={\displaystyle \underset{n=0}{\overset{\mathrm{}}{}}}g^n=1/(1g)`$
is a geometric series. The summation of geometric series inside and outside of the circle of convergence by Padé approximants is exact in all transformation orders $`n2`$. So, for all $`n2`$ the transforms $`𝒯𝒩_n(g)`$ fulfill the equality $`𝒯𝒩_n(g)=1/g\mathrm{exp}(1/g)\mathrm{\Gamma }(0,1/g)=𝒩(g)`$, where $`\mathrm{\Gamma }(0,x)`$ is the incomplete Gamma function (see ), and the choice of the contour ($`C_1`$ or $`C_{+1}`$) determines on which side of the branch cut the incomplete Gamma function is evaluated.
The asymptotic series,
$$(g)\underset{n=0}{\overset{\mathrm{}}{}}\frac{\mathrm{\Gamma }(n+\gamma )}{\mathrm{\Gamma }(n)}n!g^n,$$
(10)
constitutes a more interesting application of the resummation method than $`𝒩(g)`$. On account of the asymptotics,
$$\frac{\mathrm{\Gamma }(n+\gamma )}{\mathrm{\Gamma }(n)}n^\gamma \left(1+\mathrm{O}\left(\frac{1}{n}\right)\right),n\mathrm{},$$
(11)
the series $`(g)`$ serves as a model for the expected large-order behavior of perturbative coefficients in quantum field theory \[see Eq. (2)\]. The analytic summation of (10) leads to
$$(g)=\mathrm{\Gamma }(\gamma )\left(g\frac{}{g}\right){}_{2}{}^{}F_{0}^{}(1,\gamma ;g),$$
(12)
where the hypergeometric $`{}_{2}{}^{}F_{0}^{}`$ function has a branch cut along the positive real axis (see ). The imaginary part of (12) for $`g>0`$ as a function of $`g`$ and $`\gamma `$ is $`\mathrm{Im}(g)=\pi (1g\gamma )g^{\gamma 1}\mathrm{exp}(1/g)`$, where the integration is assumed to have been performed along the contour $`C_{+1}`$. For $`C_1`$, the sign of the imaginary part is reversed. The numerical example considered here is $`\gamma =2.3`$, $`g=0.1`$. In the Table II, numerical results are displayed for the $`n`$th partial sums of the asymptotic series (10) and the transforms $`𝒯_n(g)`$ calculated according to Eq. (5) in the range $`n=2,\mathrm{},12`$. While the partial sums eventually diverge, the transforms $`𝒯_n(g)`$ exhibit apparent convergence to about 6 significant figures in ($`n=12`$)th transformation order, and the transforms reproduce the imaginary part although the coefficients of the series (10) are all real rather than complex. The integration is performed along the contour $`C_{+1}`$. The exact result in the last row of Table II is obtained from Eq. (12). For the evaluation of the transforms $`𝒯_n(g)`$ it is crucial to use the contour $`C_{+1}`$ rather than a contour infinitesimally above the real axis. For example, in order to obtain consistent numerical results, it is necessary to take into account the poles at $`t=9.99\pm \mathrm{i}\mathrm{\hspace{0.17em}0.578}`$ in ($`n=11`$)th transformation order, encountered in the evaluation of the transform $`𝒯_{11}(g)`$ according to Eq. (5), and the pole at $`t=9.99\pm \mathrm{i}\mathrm{\hspace{0.17em}0.495}`$ in ($`n=12`$)th order for the evaluation of $`𝒯_{12}(g)`$. These poles approximately correspond to the triple pole at $`t=1/(0.1)=10`$ which would be expected in the case $`\gamma =2`$.
When an atom is brought into an electric field, the levels become unstable against auto-ionization, i.e. the energy levels $``$ acquire a width $`\mathrm{\Gamma }`$ ($`\mathrm{Re}\mathrm{i}\mathrm{\Gamma }/2`$ where $`\mathrm{\Gamma }`$ is the width). Perturbation theory cannot account for the width. The coefficients are real, not complex . An established method for the determination of the width is by numerical diagonalization of the Hamiltonian matrix . It is argued here that the full complex energy eigenvalue, including the width, can also be inferred from the divergent perturbation series by the resummation method defined in Eqs. (1)–(6), where the appropriate integration contour is $`C_{+1}`$. Perturbative coefficients for the energy shift in arbitrarily high order can be inferred from the Eqs. (9,13–15,28–33,59–67,73) in .
The symmetry of the problem suggests the introduction of the parabolic quantum numbers $`n_1`$, $`n_2`$ and $`m`$ (the principal quantum number is $`n=n_1+n_2+m+1`$). Here, calculations are performed for the ground state with parabolic quantum numbers $`n_1=0`$, $`n_2=0`$, $`m=0`$ and two L shell states, both of which are coherent superpositions of the 2S and 2P states. One of the L shell states investigated here has the parabolic quantum numbers $`n_1=1`$, $`n_2=0`$, $`m=0`$, and the other L shell state has the quantum numbers $`n_1=0`$, $`n_2=1`$, $`m=0`$. The Stark effect is interesting because, depending on the atomic state, the perturbation series are either completely nonalternating in sign (e.g., for the ground state), or they constitute nonalternating divergent series with a subleading divergent alternating component (e.g., for $`n_1=0`$, $`n_2=1`$, $`m=0`$), or the series are alternating with a subleading divergent nonalternating component (e.g., for $`n_1=1`$, $`n_2=0`$, $`m=0`$). The perturbation series for the Stark effect do not strictly fulfill the assumptions of Eq. (1), and the successful resummation of these series might indicate that the method introduced here is in fact more generally applicable. The large-order asymptotics of the perturbative coefficients for the Stark effect are given in Eqs. (4,5) in . In quantum field theory, the alternating and nonalternating components correspond to ultraviolet (UV) and IR renormalons. Using the first 20 coefficients of the perturbation series for the energy and evaluating the first 20 transforms according to Eq. (5), estimates for the real part of the energy (Stark energy shift) and the imaginary part of the energy (decay width of the state) may be obtained. The apparent convergence of the first 20 transforms for the real part of the energy extends to 6–8 significant figures, whereas the convergence of the imaginary part is much slower (2–3 significant figures). In all cases considered, both the real and the imaginary part of the energy obtained by resummation compare favorably with values for the decay width obtained by numerical diagonalization of the Hamiltonian matrix . Here we concentrate on the decay width, the full calculation will be described in detail elsewhere. The atomic unit system is used in the sequel, as is customary for this type of calculation . In the atomic unit system, the unit of energy is $`\alpha ^2m_\mathrm{e}c^2=27.211\mathrm{eV}`$ where $`\alpha `$ is the fine structure constant, and the unit for the electric field is the field strength felt by an electron at a distance of one Bohr radius $`a_{\mathrm{Bohr}}`$ to a nucleus of elementary charge, which is $`1/(4\pi ϵ_0)(e/a_{\mathrm{Bohr}}^2)=5.142\times 10^{11}\mathrm{V}/\mathrm{m}`$ (here, $`ϵ_0`$ is the permittivity of the vacuum).
Evaluations have been performed for all atomic levels and field strengths of Table III in . Three examples are presented here. For the ground state, at an electric field strength of $`E=0.1`$ in atomic units, the imaginary part of the first 20 transforms calculated according to Eq. (5) exhibits apparent convergence to $`\mathrm{\Gamma }=1.46(5)\times 10^2`$ which has to be compared to $`\mathrm{\Gamma }=1.45\times 10^2`$ obtained from numerical diagonalization of the Hamiltonian matrix . For the L shell state with quantum numbers $`n_1=0`$, $`n_2=1`$, $`m=0`$, at a field strength of $`E=0.004`$, the first 20 transforms exhibit apparent convergence to an imaginary part of $`\mathrm{\Gamma }=4.46(5)\times 10^6`$ which compares favorably to $`\mathrm{\Gamma }=4.45\times 10^6`$ from . The most interesting case is the state $`n_1=1`$, $`n_2=0`$, $`m=0`$, for which the nonalternating component of the perturbation series is subleading. At $`E=0.006`$, resummation of the complete perturbation series (including the leading alternating part) leads to a decay width of $`\mathrm{\Gamma }=6.08(5)\times 10^5`$, which is again consistent with the result of $`\mathrm{\Gamma }=6.09\times 10^5`$ from . The contour $`C_{+1}`$ is crucial, due to poles lying off the real axis.
With the help of Carleman’s Theorem it is possible to formulate a criterion which guarantees that there is a one-to-one correspondence between a function and its associated asymptotic series (see for example , Theorems XII.17 and XII.18 and the definition on p. 43 in , p. 410 in , or the comprehensive and elucidating review ). Let $`f(z)`$ be a function which is analytic in the interior and continuous on a sectorial region $`𝒮=\{z||\mathrm{arg}(z)|k\pi /2+ϵ,0<|z|<R\}`$ of the complex plane for some $`ϵ>0`$. Let the function $`f`$ have an asymptotic expansion $`f(z)_{n=0}^{\mathrm{}}c_nz^n`$ (for $`z0`$). The function $`f`$ obeys a strong asymptotic condition (of order $`k`$) if there are suitable positive constants $`C`$ and $`\sigma `$ such that $`|f(z)_{n=0}^mc_nz^n|C\sigma ^{m+1}[k(m+1)]!|z|^{m+1}`$ holds for all $`m`$ and for all $`z𝒮`$. The validity of such a condition implies that the function $`f(z)`$ is uniquely determined by its asymptotic series (see Theorem XII.19 of ). Typically, series which entail nonperturbative (imaginary) contributions do not fulfill the Carleman condition. The resulting ambiguity is reflected in the three integration contours in Fig. 1, only one of which gives the physically correct result.
It has not escaped our attention that specialized variants of the method introduced here can be constructed in those cases where additional information about the perturbative coefficients (large-order asymptotics, location of poles in the Borel plane, etc.) is available.
Finite and consistent answers in quantum field theory are obtained after regularization, renormalization and resummation. Using a resummation method, as shown in the Tables I and II, it is possible to go beyond the accuracy obtainable by optimal truncation of the perturbation series. The purpose of resummation is to eventually reconstruct the full nonperturbative result from the divergent perturbation series (see also ). I have examined two physical examples, the QED effective action in a constant background electric field \[Eq. (8)\] and the Stark energy shift. The perturbation series for the Stark effect contains nonalternating and alternating divergent contributions, which correspond in their mathematical structure to IR and UV renormalons in quantum field theory, respectively. It has been shown in each case that complete nonperturbative results, including the pair-production amplitude for electron-positron pairs and the atomic decay width, can be inferred from the divergent nonalternating perturbation series by the resummation method defined in Eqs. (1)–(6). A mathematical model series (10), which simulates the expected large-order growth of perturbative coefficients in quantum field theory \[see Eq. (2)\], can also be resummed by the proposed method (see Table II). In all cases considered, the full nonperturbative result involves an imaginary part, whereas the perturbative coefficients are real. The advocated method of resummation makes use of the Padé approximation applied to the Borel transform of the divergent perturbation series. Advantage is taken of the special integration contours $`C_j`$ ($`j=1,0,1`$) shown in Fig. 1. The author acknowledges helpful discussions with G. Soff, P. J. Mohr and E. J. Weniger. |
no-problem/0001/astro-ph0001175.html | ar5iv | text | # Ammonia observations of the nearby molecular cloud MBM 12
## 1 Introduction
MBM 12 (Magnani, Blitz, & Mundy 1985) is the closest known molecular cloud. Its distance has been estimated to be $`65`$ pc (Hobbs, Blitz, & Magnani 1986). This means that the distance to MBM 12 is less than half of that to the closest star-formation complexes (e.g., 140 pc to Taurus). Therefore, determining the presence of star-forming processes in MBM 12 would allow us to get deep into the smallest possible physical scales of these processes. For instance, for young low-mass stars a critical scale is $`100`$ AU, which is the estimated size of protoplanetary discs, and the distance at which jets are being collimated \[Rodríguez 1989\]. A size of 100 AU at 65 pc will extend $`1\stackrel{}{.}5`$, which is a resolution attainable at present, for instance by millimetre interferometers already operating. However, similar studies in more distant star-forming regions would require sub-arcsecond resolution, for which we will have to wait until the next generation of interferometers is available.
Although no ongoing star-formation has been detected so far within this high-latitude cloud, the presence of T Tauri stars \[Pound 1996\] outside, but close to the bulk of the CO emission (Pound, Bania, & Wilson 1990; Zimmerman & Ungerechts 1990; Moriarty-Schieven, Andersson, & Wanner 1997), suggests that, at least in the near past, this cloud has been a star-forming site. However, it might be possible that these T Tauri stars are run-away objects from other star-forming regions. The detection of embedded young stellar objects would then be the only way to confirm whether MBM 12 is actually a star-forming region.
Reach et al. \[Reach et al. 1995\] have determined the presence of dense gas ($`10^4`$ cm<sup>-3</sup>, traced by CS emission) in some areas of MBM 12, forming clumps in near virial equilibrium (i.e., dynamically bound, not just density fluctuations in a turbulent medium). Therefore, these are possible pre-stellar cores (Ward-Thompson et al. 1994) that merit further study.
In this paper, we present single-dish NH<sub>3</sub> observations aimed to search for evidence of on-going star formation in MBM 12. NH<sub>3</sub> observations can be used to identify the presence of current star formation processes. The ratio between different NH<sub>3</sub> transitions give us temperature information (Ho & Townes 1983). Local heating effects traced by ammonia have been successfully used to locate young stellar objects (e.g., Gómez et al. 1994; Girart et al. 1997). Moreover, local peaks of linewidth could indicate local turbulence enhancements, which are also very likely associated with star-forming processes.
We also used radio-continuum observations from the Very Large Array (VLA) of NRAO<sup>1</sup><sup>1</sup>1The National Radio Astronomy Observatory is a facility of the National Science Foundation operated under cooperative agreement by Associated Universities, Inc. archive, to look for possible embedded young stellar objects.
## 2 Observations
Observations of the NH<sub>3</sub>(1,1) and (2,2) inversion lines (rest frequencies $`\nu _{11}=23.6944955`$ GHz, and $`\nu _{22}=23.7226333`$ GHz) were taken with the 37m antenna at Haystack Observatory<sup>2</sup><sup>2</sup>2Radio Astronomy at MIT Haystack Observatory is supported by the US National Science Foundation. on 1998 March 17-24, and 1998 May 13-15. The beam size of the telescope at this frequency is $`1\stackrel{}{.}44`$ (which corresponds to $`0.03`$ pc on MBM 12). We used a dual maser, cryogenically cooled receiver, which was tuned at a frequency midway between those of the observed lines, and with $`V_{\mathrm{LSR}}=5.0`$ km s<sup>-1</sup>. The autocorrelation spectrometer was configured to measure a bandwidth of 53.3 MHz with 1024 lags, which gave a spectral resolution of 62.85 kHz (0.8 km s<sup>-1</sup>). This system setup allowed us to obtain spectra that included both NH<sub>3</sub> lines. Two circular polarizations were also obtained simultaneously. The rms pointing error was estimated to be $`11^{\prime \prime }`$, with observations of Saturn. The reference position of our NH<sub>3</sub> maps is $`\alpha (1950)=02^h53^m08\stackrel{s}{.}9`$, $`\delta (1950)=+19^{}14^{}15^{\prime \prime }`$. The spectra have all been corrected for antenna gain and atmospheric attenuation. Typical system temperatures are $`100`$ K. The final spectra have been smoothed up to a velocity resolution of 1 km s<sup>-1</sup>, yielding an rms noise level of $`0.03`$ K. Data reduction was performed with the program CLASS of IRAM and Observatoire de Grenoble.
We also used archive VLA data towards this region. These data correspond to B-array, 3.6 cm continuum observations, originally taken by D. Helfand and J. Halpern on 1990 Aug 2, to search for a radio counterpart of the X-ray pulsar H0253+193 (see Patterson & Halpern 1990; Zuckerman et al. 1992). An effective bandwidth of 100 MHz, and two circular polarizations were observed. The phase centre was located at $`\alpha (1950)=02^h53^m20\stackrel{s}{.}5`$, $`\delta (1950)=+19^{}14^{}38^{\prime \prime }`$. The source 1328+307 was used as the primary flux calibrator, with an assumed flux density of 5.19 Jy. The phase calibrator was 0235+164, for which a flux density of $`3.44\pm 0.06`$ was derived. Calibration and further image processing was performed with the program AIPS of NRAO. The resulting synthesized beam was $`0\stackrel{}{.}84\times 0\stackrel{}{.}72`$, and the rms noise level $`1.0\times 10^2`$ mJy beam<sup>-1</sup>.
## 3 Results
Figure 1 shows the emission of the NH<sub>3</sub>(1,1) line detected at different velocities. Figure 2 shows the integrated intensity map. Several clumps of ammonia emission are evident in this latter figure, the most intense one centred around position offset $`(1\stackrel{}{.}44,0\stackrel{}{.}00)`$. The observed clumps tend to align forming two clear filamentary structures, one extending south to north between position offsets $`(1\stackrel{}{.}44,2\stackrel{}{.}88)`$ and $`(1\stackrel{}{.}44,4\stackrel{}{.}32)`$, and the other one from southeast to northwest between $`(2\stackrel{}{.}88,1\stackrel{}{.}44)`$ and $`(7\stackrel{}{.}20,4\stackrel{}{.}32)`$. The morphology seen with our NH<sub>3</sub> maps is similar to the one in CO (Pound et al. 1990; Zimmerman & Ungerechts 1990; Moriarty-Schieven et al. 1997), and specially to the CS map \[Reach et al. 1995\]. Figure 3 shows an averaged spectrum within a box of $`3^{}\times 3^{}`$ around $`(1\stackrel{}{.}44,0\stackrel{}{.}00)`$, which includes the NH<sub>3</sub> emission from the most intense clump. Apart from the central, main hyperfine component, the satellite lines of NH<sub>3</sub>(1,1) are also evident. No NH<sub>3</sub>(2,2) line emission was detected in our spectra. This emission is not present even in the average within the $`3^{}\times 3^{}`$ box around $`(1\stackrel{}{.}44,0\stackrel{}{.}00)`$, with an upper limit<sup>3</sup><sup>3</sup>3Upper limits and errors are given in this paper for a 99% confidence level. of $`0.020`$ K.
No radio continuum emission is detected within the bounds of the ammonia emission, with an upper limit to the flux density of $`2.3\times 10^2`$ mJy. Only an unresolved source is present within the primary beam of the VLA, with a flux density of $`0.114\pm 0.026`$ mJy, at position $`\alpha (1950)=02^h53^m26\stackrel{s}{.}38`$, $`\delta (1950)=+19^{}14^{}44\stackrel{}{.}4`$ (we name this object as VLA B0253+192). Note that the phase centre of the VLA observations is located at position offset (2$`\stackrel{}{.}`$74,0$`\stackrel{}{.}`$38) in our NH<sub>3</sub> maps, and the primary beam at 3.6 cm is $`5\stackrel{}{.}4`$ (FWHM). Therefore, this primary beam includes the whole central clump. The position of VLA B0253+192 is indicated in Figures 1 and 2.
## 4 Discussion and Conclusions
The main goal of our NH<sub>3</sub> observations was to identify the presence of on-going star formation processes, by measuring local temperature and turbulence enhancements.
In our data, the lack of NH<sub>3</sub>(2,2) emission indicates a low temperature throughout the mapped area. In particular, from the averaged data for the most intense clump, around $`(1\stackrel{}{.}44,0\stackrel{}{.}00)`$ (see Fig. 3), and using the values for the main and inner satellite components $`T_A(1,1;m)=0.117`$ K $`\pm 0.022`$ K, $`T_A(1,1;is)=0.041`$ K $`\pm 0.016`$ K, $`T_A(2,2;m)<0.020`$ K, we derive an optical depth of $`\tau (1,1;m)=0.69`$ and an upper limit to the rotational temperature of $`12`$ K. This temperature is similar to those found in other high-latitude clouds, including MBM 7 \[Turner 1995\], which is close to MBM 12. It is also expected for this rotational temperature to be close to the value of the local kinetical temperature \[Turner 1995\]. No local linewidth enhancement is seen in our data. Typical linewidths seem to be unresolved with our 1 km s<sup>-1</sup> velocity resolution.
From these results we conclude that no indication of current star formation is evident in our NH<sub>3</sub> data. This suggests that T Tauri stars seen close to MBM 12 \[Pound et al. 1990\] were formed in a near past, but star-forming processes are no longer going on, at least in the area we mapped. Alternatively, those T Tauri stars may be runaway objects from nearby star-formation sites.
However, Reach et al. \[Reach et al. 1995\] estimated that MBM 12 is a potential star-forming cloud, given the similar mass derived from the CS integrated intensity and the virial mass. In our data, for the main clump around $`(1\stackrel{}{.}44,0\stackrel{}{.}00)`$, and assuming a $`T_k=T_{\mathrm{rot}}=12`$ K, we estimate a mass of $`0.12\left(\frac{X_{\mathrm{NH}_3}}{10^8}\right)^1`$ M, where $`X_{\mathrm{NH}_3}`$ is the molecular abundance of NH<sub>3</sub> with respect to H<sub>2</sub> . The total mass for the clumps mapped in our data is $`0.65\left(\frac{X_{\mathrm{NH}_3}}{10^8}\right)^1`$ M. These values are consistent with those obtained by Reach et al. \[Reach et al. 1995\]. Although we see no sign of current star formation processes, the presence of self-gravitating clumps suggests the possibility of these processes to occur in the future. Therefore, given its proximity to the Sun, this region is an ideal laboratory to study precollapsing physical conditions, preceding new star formation (e.g., Ward-Thompson et al. 1994).
On the other hand, the radio-continuum source that lie within our sampled area does not seem to be a young stellar object associated to MBM 12, since it is outside the bounds of the NH<sub>3</sub> emission. Association with NH<sub>3</sub> emission is a known characteristic of young stellar objects \[Anglada et al. 1989\]. The source VLA B0253+192 is likely to be an extragalactic object.
## acknowledgments
We thank John Ball, Kevin Dudevoir, and Phil Shute for their help during the observations at Haystack Observatory. JFG and JMT are supported in part by DGICYT grant PB95-0066 and by Junta de Andalucía (Spain). SP is supported by DGICYT grant PB96-0610. |
no-problem/0001/cond-mat0001157.html | ar5iv | text | # Photonic excess noise and wave localization
## Abstract
This is a theory for the effect of localization on the super-Poissonian noise of radiation propagating through an absorbing disordered waveguide. Localization suppresses both the mean photon current $`\overline{I}`$ and the noise power $`P`$, but the Fano factor $`P/\overline{I}`$ is found to remain unaffected. For strong absorption the Fano factor has the universal value $`1+\frac{3}{2}f`$ (with $`f`$ the Bose-Einstein function), regardless of whether the waveguide is long or short compared to the localization length.
PACS numbers: 42.25.Dd, 42.50.Ar
The coherent radiation from a laser has Poisson statistics . Its noise power $`P_{\mathrm{Poisson}}`$ equals the mean current $`\overline{I}`$ (in units of photons per second). Elastic scattering has no effect on the noise, because the radiation remains in a coherent state. The coherent state is degraded by absorption, resulting in an excess noise $`PP_{\mathrm{Poisson}}>0`$ . The Fano factor $`P/P_{\mathrm{Poisson}}`$ deviates from unity by an amount proportional to the Bose-Einstein function $`f`$. It is a small effect ($`f10^2`$ at room temperature for infrared frequencies), but of interest because of its fundamental origin: The excess noise is required to preserve the canonical commutation relations of the electromagnetic field in an absorbing dielectric .
The interference of multiply scattered waves may lead to localization . Localization suppresses both the mean current and the fluctuations — on top of the suppression due to absorption. Localization is readily observed in a waveguide geometry , where it sets in once the length $`L`$ of the waveguide becomes longer than the localization length $`\xi Nl`$ (with $`l`$ the mean free path and $`N`$ the number of propagating modes). Typically, $`\xi `$ is much larger than the absorption length $`\xi _\mathrm{a}`$, so that localization and absorption coexist. The interplay of absorption and localization has been studied previously for the mean current . Here we go beyond these studies to include the current fluctuations.
It is instructive to contrast the super-Poissonian photonic noise with the sub-Poissonian electronic analogue. In the case of electrical conduction through a disordered wire, the (zero-temperature) noise power is smaller than the Poisson value as a result of Fermi-Dirac statistics. The reduction is a factor $`1/3`$ in the absence of localization . The effect of localization is to restore Poisson statistics, so that the Fano factor increases from $`1/3`$ to 1 when $`L`$ becomes larger than $`\xi `$. What we will show in this paper is that the photonic excess noise responds entirely differently to localization: Although localization suppresses $`P`$ and $`\overline{I}`$, the Fano factor remains unaffected, equal to the value $`1+\frac{3}{2}f`$ obtained in the absence of localization .
Let us begin our analysis with a more precise formulation of the problem. The noise power
$$P=_{\mathrm{}}^{\mathrm{}}𝑑t\overline{\delta I(0)\delta I(t)}$$
(1)
quantifies the size of the time-dependent fluctuations of the photon current $`I(t)=\overline{I}+\delta I(t)`$. (The bar $`\overline{\mathrm{}}`$ indicates an average over many measurements on the same system.) For a Poisson process, the power $`P_{\mathrm{Poisson}}=\overline{I}`$ equals the mean current and the Fano factor $`=P/P_{\mathrm{Poisson}}`$ equals unity. We consider monochromatic radiation (frequency $`\omega _0`$) incident in a single mode (labeled $`m_0`$) on a waveguide containing a disordered medium (at temperature $`T`$). (See Fig. 1.) The incident radiation has Fano factor $`_{\mathrm{in}}`$. We wish to know how the Fano factor changes as the radiation propagates through the waveguide.
Starting point of our investigation is a formula that relates the Fano factor to the scattering matrix of the medium ,
$``$ $`=`$ $`1+[t^{}t]_{m_0m_0}(_{\mathrm{in}}1)`$ (3)
$`+2f(\omega _0,T){\displaystyle \frac{[t^{}(𝟙𝕣𝕣^{}𝕥𝕥^{})𝕥]_{𝕞_\mathrm{𝟘}𝕞_\mathrm{𝟘}}}{[t^{}t]_{m_0m_0}}}.`$
(We have assumed detection with quantum efficiency 1 in a narrow frequency interval around $`\omega _0`$.) The function $`f(\omega ,T)=[\mathrm{exp}(\mathrm{}\omega /kT)1]^1`$ is the Bose-Einstein function. The transmission matrix $`t`$ and the reflection matrix $`r`$ are $`N\times N`$ matrices, with $`N`$ the number of propagating modes at frequency $`\omega _0`$. The term proportional to $`f`$ in Eq. (3) is the excess noise. For a unitary scattering matrix, $`rr^{}+tt^{}`$ equals the unit matrix $`𝟙`$, hence the excess noise vanishes.
In what follows we will assume that the incident radiation is in a coherent state, so that $`_{\mathrm{in}}=1`$ and the deviation of $``$ from unity is entirely due to the excess noise. Since the Bose-Einstein function at room temperature is negligibly small at optical frequencies, one would need to use the coherent radiation from an infrared or microwave laser. Alternatively, one could use a non-coherent source and extract the excess noise contribution by subtracting the noise at low temperature from that at room temperature.
The absorbing disordered waveguide is characterized by four length scales: the wavelength $`\lambda `$, the mean free path for scattering $`l`$, the absorption length $`\xi _\mathrm{a}`$, and the localization length $`\xi =(N+1)l`$. We assume the ordering of length scales $`\lambda l\xi _\mathrm{a}\xi `$, which is the usual situation . We ask for the average $``$ of the Fano factor, averaged over an ensemble of waveguides with different realizations of the disorder. For $`L\xi _\mathrm{a}`$ we may neglect the matrix $`tt^{}`$ with respect to $`𝟙`$ in Eq. (3), so that the expression for the Fano factor (with $`_{\mathrm{in}}=1`$) takes the form
$$=1+2f(1C_1),C_p\frac{[t^{}(rr^{})^pt]_{m_0m_0}}{[t^{}t]_{m_0m_0}}.$$
(4)
In the absence of localization, for $`L\xi `$, one can simplify the calculation of $``$ by averaging separately the numerator and denominator in the coefficient $`C_1`$, since the sample-to-sample fluctuations are small. This diffusive regime was studied in Refs. . Such a simplification is no longer possible in the localized regime and we should proceed differently.
We follow the general approach of Ref. , by considering the change in $``$ upon attaching a short segment of length $`\delta L`$ to one end of the waveguide. Transmission and reflection matrices are changed to leading order in $`\delta L`$ according to
$$tt_{\delta L}(1+rr_{\delta L})t,rr_{\delta L}^{}+t_{\delta L}(1+rr_{\delta L})rt_{\delta L}^\mathrm{T},$$
(5)
where the superscript $`\mathrm{T}`$ indicates the transpose of a matrix. (Because of reciprocity the transmission matrix from left to right equals the transpose of the transmission matrix from right to left.) The transmission and reflection matrices $`t_{\delta L}`$, $`r_{\delta L}`$, $`r_{\delta L}^{}`$ of the short segment have zero mean and variances
$`[r_{\delta L}]_{kl}^{}[r_{\delta L}]_{mn}^{}`$ $`=`$ $`[r_{\delta L}^{}]_{kl}^{}[r_{\delta L}^{}]_{mn}^{}`$ (7)
$`=`$ $`(\delta _{km}\delta _{ln}+\delta _{kn}\delta _{lm})\delta L/\xi ,`$ (8)
$`[t_{\delta L}]_{kl}^{}[t_{\delta L}]_{mn}^{}`$ $`=`$ $`N^1\delta _{km}\delta _{ln}(1\delta L/l\delta L/l_\mathrm{a}),`$ (9)
where $`l_\mathrm{a}=2\xi _\mathrm{a}^2/l`$ is the ballistic absorption length. All covariances vanish. Substituting Eq. (5) into Eq. (4) and averaging we find the evolution equation
$`\xi {\displaystyle \frac{dC_1}{dL}}`$ $`=`$ $`2C_1\rho _1+\rho _2`$ (13)
$`{\displaystyle \frac{\xi l}{\xi _\mathrm{a}^2}}C_1+1+2C_2C_1C_1^2`$
$`4\mathrm{Re}[t^{}t]_{m_0m_0}^2[t^{}rt^{}]_{m_0m_0}[t^\mathrm{T}r^{}rr^{}t]_{m_0m_0}`$
$`+2(1+C_1)[t^{}t]_{m_0m_0}^2\left|[t^{}rt^{}]_{m_0m_0}\right|^2,`$
where we have defined $`\rho _p=\mathrm{tr}(1rr^{})^p`$.
For $`L\xi _\mathrm{a}`$ we may replace the average of the product $`C_1\rho _1`$ by the product of averages $`C_1\rho _1`$, because statistical correlations with traces that involve reflection matrices only are of relative order $`\xi _\mathrm{a}/\xi `$ — which we have assumed to be $`1`$. The moments of the reflection matrix are given for $`L\xi _\mathrm{a}`$ by
$$\rho _p=\frac{\mathrm{\Gamma }(p1/2)}{\sqrt{\pi }\mathrm{\Gamma }(p)}\frac{\xi }{\xi _\mathrm{a}},$$
(14)
hence they are $`1`$ and also $`\xi l/\xi _\mathrm{a}^2`$. We may therefore neglect the terms in the second, third, and fourth line of Eq. (13). What remains is the differential equation
$$\xi \frac{dC_1}{dL}=2C_1\rho _1+\rho _2,$$
(15)
which for $`L\xi _\mathrm{a}`$ has the solution
$$C_1=\frac{\rho _2}{2\rho _1}=\frac{1}{4}.$$
(16)
We conclude that the average Fano factor $`=1+2f(1C_1)1+\frac{3}{2}f`$ for $`L\xi _\mathrm{a}`$, regardless of whether $`L`$ is small or large compared to $`\xi `$.
To support this analytical calculation we have carried out numerical simulations. The absorbing disordered waveguide is modeled by a two-dimensional square lattice (lattice constant $`a`$). The dielectric constant $`\epsilon `$ has a real part that fluctuates from site to site and a non-fluctuating imaginary part. The multiple scattering of a scalar wave $`\mathrm{\Psi }`$ is described by discretizing the Helmholtz equation $`[^2+(\omega _0/c)^2\epsilon ]\mathrm{\Psi }=0`$ and computing the transmission and reflection matrices using the recursive Green function technique . The mean free path $`l=20a`$ and the absorption length $`\xi _\mathrm{a}=135a`$ are determined from the average transmission probability $`N^1\mathrm{tr}tt^{}=l/\xi _\mathrm{a}\mathrm{sinh}(L/\xi _\mathrm{a})`$ in the diffusive regime . Averages were performed over the $`N/2`$ modes $`m_0`$ near normal incidence and over some $`10^210^3`$ realizations of the disorder. Results are shown in Figs. 2 and 3.
The length dependence of the average Fano factor is plotted in Fig. 2, for $`N=50`$ and $`L`$ ranging from $`0`$ to $`2\xi `$. Clearly, localization has no effect. The limiting value of $`f^11`$ resulting from this simulation is slightly smaller than the value $`3/2`$ predicted by the analytical theory for $`N1`$. The $`N`$-dependence of $``$ in the localized regime is shown in Fig. 3. A line through the data points extrapolates to the theoretical expectation $`f^113/2`$ for $`N\mathrm{}`$. Fig. 3 also shows the variance of the Fano factor. The variance extrapolates to $`0`$ for $`N\mathrm{}`$, indicating that $`=P/\overline{I}`$ becomes self-averaging for large $`N`$. This is in contrast to $`P`$ and $`\overline{I}`$ themselves, which fluctuate strongly in the localized regime.
In conclusion, we have demonstrated that localization of radiation in an absorbing disordered waveguide has no effect on the ratio of the excess noise and the mean current. In the limit of a large number of propagating modes, this ratio is self-averaging and takes on the universal value of $`3/2`$ times the Bose-Einstein function. Observation of this photonic analogue of the universal $`1/3`$ reduction of electronic shot noise presents an experimental challenge.
This work was supported by the Dutch Science Foundation NWO/FOM. |
no-problem/0001/nlin0001044.html | ar5iv | text | # Detection of Nonlinear Coupling and its Application to Cardiorespiratory Interaction
## Abstract
We present here a modification of the Lagrangian measures technique, which allows a reliable detection of interdependency among simultaneous measurements of different variables. This method is applied to a simulated multivariate time series and to a bivariate cardiorespiratory signal. By using this methodology, it is possible to reveal a nonlinear interaction among cardiac and respiration rhythms in pathological conditions.
The study of natural complex phenomena has urged for the reformulation of signal processing for nonlinear systems. Moreover, with the advent of chaos theory and the ”suspicion” that many natural phenomena may behave in a chaotic, although deterministic, way, new analytical methods have been developed. Many of these complex behavior arise as the interaction of the variables involved in the process under consideration. This is especially true in the case of physiological monitoring, where the signals are correlated with one another by feedback mechanisms. Very recently , cardiorespiratory interaction has attracted the interest of the nonlinear community as a subject area where application of new methodologies can be tested.
Cardiorespiratory interaction is one of these examples where traditional tools have been used in order to gain information about its underlying dynamics. In this case, an interdependency between heart rate (HR) and respiration (R) rhythms, in physiological conditions, is almost present and is known as Respiratory Sinus Arrhithmia (see and references therein). This phenomenon is especially evident in the usual analysis by means of linear tools, like power spectrum. Accordingly, the power spectrum of the heart rate contains a peak centered at the respiratory frequency . However, there may exist situations where this peak will not be present. This fact is interpreted as a blockade in the interaction between both systems.
In this Letter, we use a recently developed technique , originally aimed to detect hidden frequencies in time series, with the purpose to demonstrate the interdependence among cardiac and respiratory signals, even in pathological conditions, where traditional tools, like power spectrum, show no signs of interactions. As we will explain below, a simple modification of our original approach will allow us to deal with this rather noisy and short time series. For illustration, we will show the technique in a well known system, the Lorenz system, in order to understand the basic steps.
From the point of view of nonlinear dynamics , a time series is considered as the ”output” or observable of a dynamical system, which can be described by the $`p`$-first order differential equation. In the case of chaotic behavior, the system must be necessarily multivariate (that is, $`p>2`$). Embedding techniques provide a way to reconstruct geometric (attractor structure) and statistical information of the original system by constructing an equivalent representation of it, using time-delayed coordinates. Given a time series of $`m`$ values, an $`n`$-dimensional vector can be constructed:
$`\overline{x_i}=(x_i,x_{i+\tau },\mathrm{},x_{i+(n1).\tau })`$
where $`\tau `$ is the time delay. In this way, from a single set of observations, multivariate vectors in the reconstructed $`n`$-dimensional space are used to trace the orbit of the system.
One of the many ways to describe the dynamics is by studying the long-term behavior of the system. When performing a statistical description of dynamical systems, a central role is played by the natural (or physical) probability measure, which describes where the orbit has been, and whose integral over a volume of state space counts the number of points within that volume . By performing a partition in the reconstructed phase space of dimension $`n`$, this probability density can be estimated as,
$$\widehat{\mu }(𝐱_i)=\frac{n(𝐱_i)}{_jn(𝐱_j)}$$
(1)
with $`n(𝐱_i)`$ equal to the number of points in partition element $`i`$ around the point $`𝐱_i`$. Using this approach, we have recently proposed a new technique to detect hidden frequencies in chaotic time series . Here we use an extension of this technique which gives a reliable detection of intrinsic frequencies in noisy and short time series.
Basically, the procedure is the following: for a given time series, a fixed, relatively high embedding dimension is chosen (high enough in order to unfold the geometrical structure of the attractor), and a reconstruction for a given range of time lags $`\tau `$ (the number of lags should be several times the correlation length of the time series) is then performed. For each $`\tau `$ the density of points the trajectory encounters as it evolves is calculated. In this case, the density along the reconstructed trajectory is estimated as
$$\widehat{\mu }(𝐱_i,\tau )=\frac{n(𝐱_i,\tau )}{_jn(𝐱_j,\tau )}$$
(2)
that is, every $`𝐱_i`$ is ordered consecutively along the trajectory. Roughly speaking, this is an estimate of the probability measure the system trajectory encounters as it evolves. In order to estimate the density given by Eq. 1 we have used small spheres around each point $`𝐱_i`$, and count the number of other points $`𝐱_j`$ inside this volume. Typically, we have used a radius of 5% to 20% of the total extent of the attractor. In this way, a new time series may be constructed with the density data, which now gives information of the different regions in the reconstructed phase space that the system visits. This information reveals recurring motion in the phase space, which is ultimately transferred to the observable time series as periodicity information. A periodogram is performed over this density time series, $`\widehat{P}`$, for each $`\tau `$ (see reference for numerical implementation) as
$$\widehat{P}(f_k,\tau )=\frac{1}{N^2}\left[\underset{j=0}{\overset{N1}{}}\widehat{\mu }(𝐱_i,\tau )e^{\frac{i2\pi jk}{N}}\right]^2$$
(3)
with $`N`$ representing the number of data points in the density time series.
In order to plot all the periodograms in a single graph we have used a gray-scale map, and 3 contour curves were superimposed in each plot in order to clarify visual inspection.
A note about the use of the periodogram is in order here. The periodogram is based in the direct Fourier transformation of the signal . If the signal comes from a deterministic system, no further modification is needed in order to the correct interpretation of the periodogram in terms of the Fourier transform. However, in the case of more noise-like signals, it is best to introduce statistical analysis, because for each frequency $`f_k`$, $`P(f_k)`$ is in fact a random variable, which can introduce fluctuations in the estimation process. This is usually accomplished by using averaging , with several realizations of the estimate. In our case, we have used plain periodograms because in plotting them side by side, several realizations of the periodograms is in fact equivalent to the process of averaging. True frequencies must remain constant in the embedding process, so any kind of fluctuations in the estimate can be readily detected.
The Lorenz system is a model proposed to explain the convective dynamics in the atmosphere (known as the Rayleigh-Benard convection), with the following variables:
$`\dot{x}`$ $`=`$ $`sx+sy`$ (4)
$`\dot{y}`$ $`=`$ $`y+rxxz`$ (5)
$`\dot{z}`$ $`=`$ $`bz+xy`$ (6)
and the standard parameters $`s`$ = 10.0, $`r`$ = 28.0 and $`b`$ = 2.66, that yields a chaotic regime. The $`x`$ coordinate is proportional to the velocity of the circulating flux, while the $`z`$ coordinate represents the distortion of the temperature with respect to a linear profile between the upper and lower temperature. Figures 1(a) and 1(b) (and first and third stripes in Figure 1(c)) shows the power spectra of the $`x`$ and $`z`$ coordinates, which look totally different, forcing the preliminary conclusion that there is no dynamical connection between both phenomena. Moreover, the power spectrum of the $`x`$ coordinate shows no characteristic frequency at all. Its power is mainly distributed in the lower band; in contrast, the power spectrum of the $`z`$ coordinate shows a very definite frequency around harmonic number 54. However, both time series belong to the same system. It must be remarked that more involved linear methods, like cross-correlation and coherence gives no further insight in this problem. The above fact reveals the main disadvantage in applying spectral methods for analyzing time series coming from chaotic systems. We have used the $`x`$ variable as our experimental time series with 2048 data points. An embedding dimension of 3 has been used, and a 10% radius of the attractor was employed in order to estimate the density. The main panel of Figure 1(c) shows the power spectrum of the density time series after applying our method to the $`x`$ coordinate, for each $`\tau `$ of the embedding process. As the $`\tau `$ parameter varies, the strong frequency which appears in the $`z`$ coordinate is remarkably recovered, as well as some of the frequencies in the lower band. Inversely, it is also possible to use the $`z`$ coordinate as the experimental one. Figure 2 shows the same as Figure 1 this time using the $`z`$ coordinate. Although the power spectrum of this coordinate shows a single sharp peak, perhaps superimposed to a continuous background, our method allows to predict the existence of lower band frequencies in its dynamics. This is evident in the power spectrum of the $`x`$ or $`y`$ coordinates.
Two additional points are noteworthy. Some spurious frequencies may appear in this procedure, for example due to an effect of the finite size of time series. However, these are easily detected because they do not remain constant along the $`\tau `$ scan. By varying the $`\tau `$ parameter in the reconstruction, a nonlinear transformation is performed, and intrinsic properties of the system must remain constant along this procedure, a fact guaranteed by the embedding theorems In fact, we are seeking that $`\widehat{P}(f_i,\tau )=\widehat{P}(f_i)`$, independently of the particular $`\tau `$ used. The second point is that intrinsic frequencies (as is the case of harmonic number 54) may disappear for some values of the $`\tau `$ parameter. This is due to the embedding itself, which is only guaranteed to give faithful results in the case of infinitely long data sets and without noise. Some values of the time lag $`\tau `$ may introduce false information, or even hide important frequencies. By performing a scan along a series of $`\tau `$ we can be assured of extracting information of the system itself.
Now, we use our approach in the multivariate time series from heart rate and respiration rhythms. The experiment is fully described in reference . This is a multivariate time series recorded from a patient in the sleep laboratory of the Beth Israel Hospital in Boston. The HR-record is the heart rate, the R-record is the chest volume (respiration force) and the BOC-record is the blood oxygen concentration (measured by ear oximetry). 2048 data points were chosen from the total record, corresponding to the periods of the sleep apnea episodes of the patient. Figure 3 shows the results of applying our analysis to the respiration force time series. We have used an embedding dimension of 4 for the reconstruction process and a radius of the ball of 15% of the reconstructed attractor.
Classical power spectrum analysis of the HR-record shows a strong peak at harmonic number 22 (which corresponds to a frequency of 0.021 Hz). However, the R-record shows several peaks around the harmonic number 260 (0.254 Hz), and no traces of the main frequency of the HR-record. Therefore, it could be concluded that both variables are (linearly) independent. In the original analysis , based on standard spectral analysis there is no mention at all of any kind of coupling among both variables in this pathological case, i.e. sleep apnea, contrary to the case of the RSA in the physiological case.
By applying our analysis to the R-record (Figure 3(c), main panel), it appears that the main frequencies in the reconstructed system are those related with the HR record, and some traces of the original R time series can be observed (around $`\tau =25`$). This fact, as in the previous case of the Lorenz system, could be considered as a justification of a nonlinear relation between both variables, and, moreover, suggests that they are related with the same dynamics. However, in this particular case, the flow of information is in the oposite direction as in the case of the RSA.
As in the case of the Lorenz system, we can apply the procedure to the other variable, that is, the HR record (Figure 4). Although some traces of the R record are recovered (specially around $`\tau `$ = 90) almost all of the power is concentrated in the characteristic frequency of the HR. This would imply a very weak coupling from the heart rate to the respiration variable. In the case of the Lorenz system, the equations shows that both variables used, $`x`$ and $`z`$ are in fact coupled to each other. That is why it is possible to recover the whole spectrum by using whatever of both variables. One can conclude from the above argument that the coupling among HR and R variables could have a preferential flow direction from the HR to the R, in this pathological condition.
Experience in numerical models shows that hidden frequencies, as detected by our approach, are caused fundamentally by chaotic systems, when one of the variables ”fails” in projecting its periodic content over the measured time series. In a linear system, or a nonlinear system without a chaotic behavior, this does not happen. Considering the cardiorrespiratory interaction, the modulation of HR by respiration is clearly seen in the power spectrum of the HR signal under physiological conditions. However, in the pathological condition we are considering, this coupling is not revealed by linear analysis, and moreover, it appears to be in the opposite direction. It could be argued that the underlying dynamics has changed from a linear/nonlinear state to a chaotic behavior. This could be a reason for the dissapearance of the RSA phenomena and the appearance of an inverse coupling. This fact could also be in accordance with the so-called ”dynamical diseases” where an abrupt change in the dynamics of the interacting variables is evidenced.
It is very worthy to mention that we have used the data set as it is. No further filter or smoothing routines have been applied to the data, beyond those of the recording process and to the conversion from the ECG to Heart rate data . We think that it is very important, from the point of view of nonlinear dynamics analysis, to minimize the use of any kind of ”treatment” of the experimental record, which may alter the underlying dynamics, especially in this short and very noisy data sets.
Data requirements for applying this method are almost the same as in any classical spectral analysis. Stationarity of the data is the most important prerequisite, because of the probability estimate that is being calculated. The method is fairly robust against noise contamination, mainly because of the density estimation step, as we have shown with other methods . Other than these considerations, the algorithm is of wide use, and could be applied to any kind of experimental data, from biological to geophysical and so on. We are currently applying this analysis to physiological data derived from biological rhythmicity recording in mammals . It should be noted that even more information could be gathered from the reconstructed density time series by applying nonlinear analytical methods . However, the use of power spectra is still the most common tool in time series analysis because of its simplicity and ease of interpretation , and here we propose to extend its well known capabilities with the aids of nonlinear methods. The most important benefits derived from our methodology might be observed in the case of several variables, where one or more of them are not available experimentally. In this way, more information could be achieved that the one currently available from classical spectral analysis techniques.
The authors are members of CONICET Argentina. Studies in D.A.G. s laboratory are funded by grants from CONICET, UNQ, ANPCyT and Fundación Antorchas. |
no-problem/0001/hep-ph0001201.html | ar5iv | text | # Top Polarization at a 𝜇⁺𝜇⁻ Collider1footnote 11footnote 1Talk presented by GM at the 5th International Conference on Physics Potential and Development of 𝜇⁺𝜇⁻ Colliders, San Francisco, CA, December 15–17, 1999.
## Abstract
The top quark pairs produced at a polarized muon collider are in a (nearly) pure spin configuration. This result holds for all center-of-mass energies, and is insensitive to the next-to-leading order QCD radiative corrections. The decay products of a polarized top quark show strong angular correlations. We describe an interesting interference effect between the left-handed and longitudinally polarized $`W`$ bosons in top quark decay. This effect is easily observable in the angular distribution of the charged lepton with respect to the beam axis.
With a mass of $`173.8\pm 5.2`$ GeVPDG1999 , the top quark is by far the heaviest fermion in the Standard Model (SM). Within the SM, the top quark decays very rapidly to a $`W`$ boson and a $`b`$ quark. Since the $`t`$ decay width ($`\mathrm{\Gamma }_t1.5`$ GeV) is much greater than the spin decorrelation scale ($`\mathrm{\Lambda }_{\mathrm{QCD}}^2/m_t230`$ keV), a polarized $`t`$ quark decays before QCD can randomize its spinBigi . Thus, the decay products of polarized $`t`$ quarks exhibit strong angular correlations.
Because the focus of this talk is on a muon collider environment, we should begin with a few words about the Higgs boson. Although much has been said about the usefulness of a muon collider as a “Higgs factory” for low-mass Higgs bosons, from the point of view of studying $`t\overline{t}`$ pair production, the Higgs boson does not play an important role. A light SM Higgs boson shifts the total $`t\overline{t}`$ production cross section at threshold by only a few percentTopThresh . This effect, which is comparable in size to the (incompletely known) 2-loop QCD corrections, becomes even smaller as the machine energy is increased beyond the threshold region. The other case to consider is a Higgs which is heavy enough for the on-shell decay $`ht\overline{t}`$ to occur, allowing for resonant $`t\overline{t}`$ production. Assuming SM couplings, a 400 GeV Higgs boson contributes only a fraction of a percent to the total cross section through this channel. Consequently, as far as top quark pair production is concerned, the same analysis applies equally to $`e^+e^{}`$ and $`\mu ^+\mu ^{}`$ colliders.
The production of top quark pairs at a muon collider proceeds through the $`s`$-channel via an off-shell photon or $`Z`$ boson. For the purposes of our initial discussion, we will assume that the $`\mu ^{}`$ beam is 100% polarized in a left-handed helicity state and that the $`\mu ^+`$ beam is 100% polarized in a right-handed helicity state. Details concerning the more general case with arbitrary polarizations of the two beams may be found in Refs. ParkeShadmi ; mumu97 ; DrellTalk . In the situation at hand, the matrix element for the production of a $`t\overline{t}`$ pair whose spins are labelled by $`\lambda `$ and $`\overline{\lambda }`$ takes the form
$$\frac{e^2}{s}\left[\overline{\text{v}}(\overline{\mu })\gamma ^\mu \mathrm{P}_\text{𝕃}\text{u}(\mu )\right]\left[\overline{\text{u}}(t,\lambda )\gamma ^\mu \{f_{\text{𝕃𝕃}}\mathrm{P}_\text{𝕃}+f_{\text{𝕃ℝ}}\mathrm{P}_{\text{}}\}\text{v}(\overline{t},\overline{\lambda })\right],$$
(1)
where we have defined the chirality projection operators $`\mathrm{P}_\text{𝕃}\frac{1}{2}(1\gamma _5)`$ and $`\mathrm{P}_{\text{}}\frac{1}{2}(1+\gamma _5)`$. The effective couplings $`f_{\text{𝕃𝕃}}`$ and $`f_{\text{𝕃ℝ}}`$ incorporate the different center-of-mass energy dependence of the photon and $`Z`$ boson propagators: consequently they are weakly dependent on $`\sqrt{s}`$. The ratio $`f_{\text{𝕃ℝ}}/f_{\text{𝕃𝕃}}`$ ranges from 0.343 near threshold to 0.365 in the ultra-relativistic limit. Note that since the muon is effectively massless at the center-of-mass energies we are considering, the helicity of the muons is linked to their chirality in the usual fashion. On the other hand, both chiralities contribute to the spin eigenstates of the $`t`$ quark, even if we choose to work in the helicity basis. Nevertheless, Parke and Shadmi ParkeShadmi have constructed a spin basis, called the off-diagonal basis, in which more than 99% of the $`t\overline{t}`$ pairs are produced in the UD spin state at $`\sqrt{s}=400\mathrm{GeV}`$.
In order to motivate the off-diagonal basis, let us imagine that we can turn off the $`f_{\text{𝕃ℝ}}`$ coupling. Then, it is easy to see what happens in the two energy extremes (Fig. 1). At ultrarelativistic energies, the left-handed chirality of the produced $`t\overline{t}`$ pair translates into left-handed helicity for the $`t`$ and right-handed helicity for the $`\overline{t}`$: we obtain 100% $`t_L\overline{t}_R`$. Near threshold, however, where there is no orbital angular momentum involved, the $`t`$ and $`\overline{t}`$ spins must be parallel to the beam axis, since the initial $`\mu ^{}`$ and $`\mu ^+`$ spins add up to one unit along that direction. At intermediate energies, it is not surprising to learn that in the lab frame the direction of the $`t`$ and $`\overline{t}`$ spins is not entirely along either the beam axis or the $`t\overline{t}`$ production axis. Instead, when viewed in the $`t`$ rest frame, the $`t`$ spin is parallel to the $`\mu ^+`$ momentum. In the $`\overline{t}`$ rest frame, the $`\overline{t}`$ spin is antiparallel to the $`\mu ^{}`$ momentum. These associations ($`t`$ with $`\mu ^+`$ and $`\overline{t}`$ with $`\mu ^{}`$) may be understood by Fierzing the first term of Eq. (1). Thus, in this basis, the $`t\overline{t}`$ spin state is 100% UD in the limit $`f_{\text{𝕃ℝ}}0`$.
Of course, we cannot set $`f_{\text{𝕃ℝ}}=0`$ in the real world. However, because $`f_{\text{𝕃ℝ}}/f_{\text{𝕃𝕃}}`$ is not very large, it is possible to tweak the spin axis choice a bit and retain the overwhelming dominance of the UD spin configuration. The result is the off-diagonal basis of Parke and ShadmiParkeShadmi . In the off-diagonal basis, the UU+DD spin state vanishes identically for $`\mu _L^{}\mu _R^+`$ collisions. The DU spin state is suppressed by $`\beta ^4(f_{\text{𝕃ℝ}}/f_{\text{𝕃𝕃}})^2`$. Since the speed of the produced top quarks is only $`\beta 0.5`$ at $`\sqrt{s}=400\mathrm{GeV}`$, this is a significant suppression, and $`\mu _L^{}\mu _R^+t_U\overline{t}_D`$ accounts for 99.88% of the tree-level cross section. Fig. 2 shows the contributions to the $`t\overline{t}`$ cross section as a function of the $`t\overline{t}`$ production angle in the lab frame, decomposed into the various $`t\overline{t}`$ spin states for a 400 GeV collider.
All of the preceding discussion has been at tree level. Because the resulting spin state is so pure, we might worry that the next-to-leading-order QCD corrections might spoil this result, especially since the total (unpolarized) cross section grows from 0.63 pb to 0.80 pb at $`\sqrt{s}=400\mathrm{GeV}`$, an enhancement of more than 25%. The spin dependence of the NLO QCD corrections has been studied by Kodaira, Nasuno and Parke OneLoopQCD , who find that the off-diagonal basis defined at tree level is still the appropriate basis, and that the purity of the $`t_U\overline{t}_D`$ state produced in $`\mu _L^{}\mu _R^+`$ collisions is still 99.85%. The physics behind this result is easily understood. The large enhancement in the total cross section is mostly attributable to the production of soft gluons. However, soft gluons are incapable of causing spin flips of the top quark. In the soft gluon limit, matrix elements with an extra gluon factorize into matrix elements without the extra gluon times an eikonal factor SoftGluons . This factorization occurs independently and with the same eikonal factor for each spin configuration. Thus, the spin composition of the produced $`t\overline{t}`$ pairs is altered only slightly by the inclusion of NLO QCD effects.
The decay products of a polarized top quark are strongly correlated with the top quark spin axis. These correlations are most simply viewed in the $`t`$ rest frame, where the dependence upon the angle $`\chi _i^t`$ between the $`i`$th decay product and the $`t`$ spin axis is of the form
$$\frac{1}{\mathrm{\Gamma }_T}\frac{d\mathrm{\Gamma }}{d(\mathrm{cos}\chi _i^t)}=\frac{1}{2}(1+\alpha _i\mathrm{cos}\chi _i^t)$$
(2)
for a spin up top quarkDecayCorrel (see Fig. 4). The charged lepton or $`d`$-type quark from the decaying $`W`$ boson displays maximal correlation with the top quark spin. The values of the $`\alpha _i`$’s for the other decay products depend on the top and $`W`$ masses (see Ref. xidep for a convenient tabulation).
The spin states of the $`W`$ bosons coming from a spin up top quark decay display interesting interference effects. The simplest description utilizes the $`W`$ helicity measured in the $`t`$ rest frame. Then, in the decay of a spin up top quark, the $`W`$’s are produced with left-handed or longitudinal helicity only. The ratio of left-handed to longitudinal $`W`$’s is $`2m_W^2:m_t^2`$ ($`=30\%:70\%`$), and is reflected in the decay angular distribution
$$\frac{1}{\mathrm{\Gamma }_T}\frac{d\mathrm{\Gamma }}{d(\mathrm{cos}\chi _{\mathrm{}}^W)}=\frac{3}{4}\frac{m_W^2(1\mathrm{cos}\chi _{\mathrm{}}^W)^2+m_t^2\mathrm{sin}^2\chi _{\mathrm{}}^W}{2m_W^2+m_t^2}.$$
(3)
In Eq. (3), we have denoted $`W`$ rest frame angle between the $`b`$ quark and charged lepton by $`\pi \chi _{\mathrm{}}^W`$.
In Fig. 4 we present a contour plot of the lepton emission angle in the $`W`$ rest frame versus the $`W`$ emission angle in the $`t`$ rest frame. From this plot, we see that the longitudinal $`W`$’s are emitted mainly parallel to the top quark spin, while the left-handed $`W`$’s are emitted mainly antiparallel to the top quark spin. Near $`\mathrm{cos}\chi _W^t=0`$, interference between the left-handed and longitudinal $`W`$’s dominates.
The interference between the left-handed and longitudinal $`W`$ bosons from $`t`$ quark decay is the explanation for an apparent mystery contained in Fig. 4. Recall that the $`W`$ bosons are only moderately correlated with the top quark spin: $`\alpha _W=0.40`$. However, one of the $`W`$ decay products (the charged lepton or $`d`$-type quark) is maximally correlated with the top quark spin: $`\alpha _{\mathrm{}}=1`$! In Fig. 5, we show the angular distribution of the charged leptons coming from left-handed and longitudinal $`W`$ decays separately,<sup>2</sup><sup>2</sup>2The $`\alpha _i`$’s of Eq. (2) for the left-handed and longitudinal $`W`$ bosons are given by
$$\alpha _{\mathrm{left}}=\frac{\xi ^3+8\xi ^24\xi +1}{(\xi 1)^3}+\frac{6\xi ^3\mathrm{ln}\xi }{(\xi 1)^4}$$
and
$$\alpha _{\mathrm{long}}=\frac{(\xi +1)(\xi ^28\xi +1)}{(\xi 1)^3}+\frac{12\xi ^2\mathrm{ln}\xi }{(\xi 1)^4}$$
where $`\xi (m_t/m_W)^2`$. For $`m_t=173.8\mathrm{GeV}`$ and $`m_W=80.4\mathrm{GeV}`$ these expressions yield $`\alpha _{\mathrm{left}}=0.041`$ and $`\alpha _{\mathrm{long}}=0.55`$. for the left-handed as well as their quantum mechanical sum, including interference effects. We see that at $`\mathrm{cos}\chi _{\mathrm{}}^t=1`$ there is total destructive interference. For $`\mathrm{cos}\chi _{\mathrm{}}^t>0`$, the interference is, on average, constructive. Thus, the interference term allows the charged lepton or $`d`$-type quark to be more correlated with the top quark spin than the parent $`W`$.
This interference effect also leaves a visible imprint in the lab frame. In Fig. 6 we show the distribution of the angle between the charged lepton and the direction of the $`\mu ^{}`$ beam, again broken down into the contributions from left-handed and longitudinal $`W`$ bosons. The effect is particularly striking for $`\mu _L^{}\mu _R^+`$ collisions, as the shape of the quantum mechanical sum of the two contributions is completely dominated by presence of the interference term. The dotted line in Fig. 6 shows the top quark production angle. Even though the $`t`$ quark is predominantly produced in the forward direction in $`\mu _L^{}\mu _R^+`$ collisions, the charged lepton from its decay tends to move backward in the lab frame. This is a consequence of the interference between the left-handed and longitudinal $`W`$ bosons. |
no-problem/0001/hep-ex0001039.html | ar5iv | text | # Physics Beyond the Standard Model 11footnote 1presented at the 1999 UK Phenomenology Workshop, Durham; to appear in Journal of Physics G.
## 1 The Standard Model Higgs
The search for the SM Higgs at LEP2 is well understood. The reach depends on the centre of mass energy which can be achieved, and the integrated luminosity obtained. For 150 pb<sup>-1</sup> per experiment at $`\sqrt{s}=200`$ GeV, Higgs masses $`m_H<109`$ GeV will be ruled out. The LEP energy may be pushed a few GeV higher, but to go much beyond this will require the use of the Tevatron. Much interest in the Tevatron’s potential for Higgs searches was sparked by work carried out during the Run II SUSY/Higgs workshop in 1998.
The Higgs cross section at the Tevatron is large (1 pb for $`m_H100`$ GeV) but since the $`ggHb\overline{b}`$ process dominates, there is a huge QCD background. For Higgs masses below 130–140 GeV, the best potential seems to come from the processes where a $`W`$ or $`Z`$ is produced in association with the Higgs:
* $`WH\mathrm{}\nu b\overline{b}`$: Backgrounds from $`Wb\overline{b}`$, $`WZ`$, $`t\overline{t}`$, single top. A factor of 1.3 improvement in signal to background has been demonstrated in this channel by using a neural network compared to standard cuts. It is also possible that additional gains may be had if the angular distributions ($`WH`$, spin zero, vs. $`Wb\overline{b}`$, spin one) can be exploited. $`WHq\overline{q}^{}b\overline{b}`$ is overwhelmed by the QCD background.
* $`ZH\mathrm{}\mathrm{}b\overline{b}`$: backgrounds from $`Zb\overline{b}`$, $`ZZ`$, $`t\overline{t}`$.
* $`ZH\nu \nu b\overline{b}`$: backgrounds from QCD, $`Zb\overline{b}`$, $`ZZ`$, $`t\overline{t}`$. This requires a relatively soft missing $`E_T`$ trigger (35 GeV) but is powerful because of the large $`Z\nu \nu `$ branching ratio.
For Higgs masses above 130–140 GeV:
* $`ggHWW^{}`$: backgrounds from Drell-Yan, $`WW`$, $`WZ`$, $`ZZ`$, $`t\overline{t}`$, $`Wt`$. The initial signal to background ratio is 1:140, so many, rather finely tuned, selections are required, culminating in angular cuts to separate the signal from the “irreducible” $`WW`$ background.
Combining all these search channels, and assuming a 10% systematic error on the backgrounds, one finds the sensitivity shown in Fig. 1. With 2 fb<sup>-1</sup> of data, only a modest extension of the LEP reach is possible. With 15 fb<sup>-1</sup>, on the other hand, it should be possible to either exclude, or observe at the 3–5$`\sigma `$ level, a SM Higgs up to $`m_H180`$ GeV.
This is an exciting prospect. Is it credible? In my view, yes: it is an exercise similar in scale to the top discovery, with a similar number of backgrounds, and requiring a similar level of detector understanding. It will be harder — the irredicible signal to background ratio is worse — but it has caught the imagination of the experimenters. One problem with the studies so far (in my opinion) is the $`b\overline{b}`$ mass resolution. Can the assumed resolution really be attained in a high luminosity environment? The resolution not only affects the mass bins over which a Higgs signal would be smeared, but a detailed understanding of the shape of the $`b\overline{b}`$ mass spectrum will be needed to separate a putative Higgs signal from the nearby mass peak due to $`Zb\overline{b}`$ decays (see Fig. 1). For this reason, calibration samples of $`Zb\overline{b}`$ events will be very important. Tens of thousands of events (after cuts) are expected for 2 fb<sup>-1</sup>, using a displaced vertex trigger, in both CDF and DØ.
## 2 Minimal Supersymmetry
Even the minimal spectrum of supersymmetry (the SM particles plus two Higgs doublets and their superpartners) can have many faces. Firstly, whether or not $`R`$ parity is conserved determines whether the lightest supersymmetric particle (LSP) will or will not be stable. Secondly, how supersymmetry is broken determines the mass heirarchy of states and hence their decays. The typical benchmark, the supergravity-inspired or “minimal SUGRA” model, is determined by five parameters $`m_0`$, $`m_{1/2}`$, $`A_0`$, $`\mathrm{tan}\beta `$ and $`\mathrm{sign}\mu `$. In this picture, radiative electroweak symmetry breaking occurs naturally from the large top mass. The lightest neutralino is the LSP; the two lightest neutralinos, the lighest chargino and the lightest Higgs $`h`$ are all “light”, while the other charginos and neutralinos, squarks, gluinos and the other Higgs states are all “heavy” (above a few hundred GeV, perhaps). In gauge-mediated models, by contrast, the LSP is a gravitino, and there are signatures with photons and/or slow moving particles which may decay within or outside the detector. Recent anomaly-mediated models suggest that the lightest chargino and neutralino may be almost degenerate.
At LEP2, the increased centre of mass energy will allow the present SUSY limits to be raised by 5–10 GeV (my guess) with full use of data at and above 200 GeV.
At the Tevatron, the highest cross sections are for the pair production of colored particles like squarks and gluinos. As long as $`R`$-parity is conserved, the signature will be jets plus missing transverse energy ($`E\text{/}_T`$ ). The Run I DØ analysis required three jets (one with $`E_T>115`$ GeV) and $`E\text{/}_T\text{ }>75`$ GeV; it excludes squark masses below about 250 GeV and gluino masses below about 200 GeV. This search may be combined with complementary channels where one or more leptons are required, together with jets and $`E\text{/}_T`$ . With 2 fb<sup>-1</sup> of data the reach will approximately be doubled, to gluino masses of order 400 GeV. For masses much larger than this, falling parton distributions kill the production rate very quickly. To extend the reach in parameter space, chargino/neutralino production will become increasingly important. Present searches in the “golden” trilepton mode do not really constrain models, but with 2 fb<sup>-1</sup> of data, chargino masses up to 180 GeV should be probed (150 GeV if $`\mathrm{tan}\beta `$ is large). This places an emphasis on low-$`p_T`$ lepton triggering, since one or more of the leptons tends to be soft. It would also help a lot if $`\tau `$ modes could be included.
In many models, the stop and sbottom squarks are significantly lighter than the others. It is therefore interesting to search for them separately. The decay channels involve $`b`$ or $`c`$ jets, and for stop there is also the possibility of $`t\stackrel{~}{t}`$ decays or vice versa (depending on the masses). Present CDF limits explore $`m_{\stackrel{~}{t}}<120`$ GeV and $`m_{\stackrel{~}{b}}<145`$ GeV; with 2 fb<sup>-1</sup>, the sensitivity should improve to about 200 GeV.
Much interest in gauge-mediated supersymmetry was sparked a few years ago by the observation of a single $`ee\gamma \gamma E\text{/}_T\text{ }`$ event at CDF. This final state is consistent with selectron production in a gauge mediated scenario. All we can now say is that searches for the expected related signatures have all proved negative: $`\gamma \gamma E\text{/}_T\text{ }`$ and $`\gamma +\mathrm{jets}+E\text{/}_T\text{ }`$ at the Tevatron, and one or more photons plus missing energy at LEP.
This type of signature is just one example of what may be expected in gauge mediated models. Depending on the details of the model, the next-to-lightest superpartner or NLSP (which then decays into the gravitino LSP) may be a neutralino, a stau, or effectively more than one slepton state (if they are almost degenerate in mass). Prompt decays of the NLSP will then give final states containing photons$`+E\text{/}_T\text{ }`$, taus, and multileptons, respectively. Searching for such signatures is relatively straightforward. A more challenging possibility is that the NLSP has a finite decay length — since this is determined by the SUSY breaking interactions it is totally unknown. For neutralino NLSP’s, decay lengths $`c\tau `$ more than a few metres will give standard $`E\text{/}_T`$ signatures since the NLSP will escape the detector. Decay lengths of a few centimetres to a metre result in photons which do not point back to the primary vertex: searches at LEP for such photons have excluded neutralino NLSP masses less than 85 GeV, for $`c\tau <1`$ m. In Run II at the Tevatron, DØ will be able to make use of new preshower detectors upstream of the electromagnetic calorimeter to detect non-pointing photons. The resolution in the distance of closest approach to the vertex will be 2.2 cm along the beamline and 1.4 cm radially.
Charged NLSP’s with long decay lengths appear as massive, slow-moving particles which exit the detector (so called “cannonballs”). LEP limits exclude stau NLSP masses less than 76 GeV, or slepton co-NLSP masses less than 85 GeV. In Run II at the Tevatron, $`dE/dx`$ and timing information will be available (using time of flight counters in CDF, and the muon system in DØ). CDF expect sensitivity up to stau masses of 180 GeV using the time of flight system. Short decay lengths $`c\tau <1`$ cm will give reconstructable impact parameters in the vertex detectors; 1 cm $`<c\tau <1`$ m is harder, especially to trigger. (With enough data, a combination of impact parameter and cannonball searches may be sensitive enough to exclude this intermediate region as well.) A general problem with understanding delayed decays is that event generators are not widely available, and the interface to the detector simulation is non-trivial.
Recently there has been considerable interest in anomaly-mediated SUSY models. The phenomenology of such models contains a light chargino which is almost degenerate with the LSP (a neutralino). Such small mass differences may result in delayed chargino decays, with cannonball type signatures, or decays inside the detector $`\stackrel{~}{\chi }^\pm \stackrel{~}{\chi }^0+\mathrm{soft}\pi `$: very challenging indeed.
There is also considerable interest in scenarios with large extra dimensions. Gravitons may propagate into the higher dimensional space. Searches have been carried out for $`e^+e^{}\gamma +`$nothing, $`p\overline{p}\gamma +`$nothing, and $`p\overline{p}`$jet$`+`$nothing. Indirect effects have been searched for in $`e^+e^{}\gamma \gamma ,\mu ^+\mu ^{},\tau ^+\tau ^{}`$.
Turning now to SUSY with $`R`$-parity violating decays, the usual assumption is that the $`R`$ violating couplings are small compared with the $`R`$ conserving ones. The production processes and subsequent decay chain then proceed as in minimal SUGRA, but the final LSP’s decay via baryon or lepton number violating interactions. Hence there is no $`E\text{/}_T`$ from the escaping LSP, but there are usually multi-leptons from cascade decays and $`E\text{/}_T`$ from neutrinos. Both at LEP and the Tevatron, it is found (perhaps surprisingly) that the excluded regions end up very comparable to those for mSUGRA with $`R`$ conserved. $`R`$-violating couplings could also enter in the production of SUSY particles. At HERA, “leptoquark” searches have been interpreted in terms of $`ep\stackrel{~}{q}`$, while at LEP, limits have been placed on processes like $`e^+e^{}\stackrel{~}{\nu }\tau ^+\tau ^{}`$.
Both at LEP2 and the Tevatron, we therefore have a basic menu of supersymmetry searches which is well-defined and we should have no trouble in exploring:
* minimal SUGRA;
* gauge mediated supersymmetry with prompt photon signatures;
* some subset of $`R`$ violation.
Our concerns are what we may have forgotten, especially at the Tevatron where triggering is a crucial issue. Can we cover:
* slow moving massive particles;
* gauge mediated supersymmetry with detached photons or taus;
* anomaly-mediated supersymmetry (e.g. $`\stackrel{~}{\chi }^\pm \stackrel{~}{\chi }^0+\mathrm{soft}\pi `$);
* general extra dimension signatures, etc?
A critical look at the proposed trigger lists is probably called for.
## 3 Supersymmetric Higgs
At LEP2, two complementary processes allow the neutral Higgs states of supersymmetric models to be explored: $`e^+e^{}(h/H)Z`$ and $`e^+e^{}(h/H)A`$. Combining these channels and the four experiments, one can exclude (Summer 1999) $`m_h<81`$ GeV and $`m_A<81`$ GeV. Moreover, $`0.9<\mathrm{tan}\beta <1.6`$ ($`0.6<\mathrm{tan}\beta <2.6`$) is excluded for maximal (and for minimal) stop mixing. The $`\mathrm{tan}\beta `$ exclusion is very sensitive to the top mass, and it should be noted that there is no excluded range in $`\mathrm{tan}\beta `$ if $`m_t=180`$ GeV. General scans of minimal SUSY parameter space find some points with strange decay patterns that can evade the limits, but these are rather few. Invisible Higgs decays have been included in the searches.
In Run II of the Tevatron, rather stringent limits will be set on the SUSY Higgs sector, given sufficient luminosity, since the whole allowed mass range for the lightest Higgs $`h`$ ($`m_h<130`$ GeV) is covered. One may also exploit the enhanced couplings between the Higgs states and $`b`$ quarks at large $`\mathrm{tan}\beta `$ to search for $`p\overline{p}b\overline{b}(h/A)4b`$. Using present data and requiring three $`b`$-tagged jets, CDF have been able to explore the region of $`\mathrm{tan}\beta 50`$ and above. With 10 fb<sup>-1</sup> the sensitivity will extend to $`\mathrm{tan}\beta =30`$ for $`m_A`$ up to 150 GeV.
Supersymmetry also predicts the existence of charged Higgs states. Searches for pair production at LEP have excluded masses $`m_{H^\pm }<77`$ GeV. In minimal SUSY models, it is expected that $`m_{H^\pm }>m_W`$, and LEP is not really sensitive to this region. Searches at the Tevatron have concentrated on the production of $`H^\pm `$ in top decays in competition with the standard $`tW`$ mode. This has been carried out both as an “appearance” experiment (looking for $`H^\pm \tau `$) and as a “disappearance” experiment (looking for fewer than expected $`te\mathrm{and}\mu `$). Present limits are sensitive only for $`\mathrm{tan}\beta <1`$ and $`>40`$, but with 2 fb<sup>-1</sup> the sensitivity will extend to $`\mathrm{tan}\beta 2`$ and 20.
## 4 Non-supersymmetric Electroweak Symmetry Breaking
There is no fully worked-out scenario for electroweak symmetry breaking through a new strong interaction, but this does not mean that it is not possible. Indeed, many schemes have been outlined in some detail, and a straw-man technicolor model is now implemented in pythia. In general, dynamical symmetry breaking schemes like technicolor and topcolor predict:
* new particles in the mass range 100 GeV – 1 TeV,
* with strong couplings and large cross sections,
* decaying to vector bosons and (preferentially third generation?) fermions.
Recently L3 has reported searches for technicolor resonances at LEP2, and several technicolor and topcolor searches have been made on present Fermilab data. All are negative so far. With Run II, the reach of the Tevatron will be greatly extended.
Besides all of the above scenarios, one should always be looking (at LEP, HERA and the Tevatron) for anything unexpected:
* leptoquarks
* fourth generation fermions, or isosinglet fermions
* $`W^{}`$ and $`Z^{}`$
* contact interactions or compositeness, etc.
## 5 Conclusions
There are plenty of opportunities for us to find “something new” before the LHC starts operation. The standard scenarios have been explored in some detail, but different supersymmetry breaking schemes can produce radically different signatures. Theoretical fashion moves fast, and event generators are not always available. We need to keep an open mind in our searches. This is easy to say once one has the data in hand; the challenge right now is to ensure that we do our best to trigger on all possible interesting things, especially in the hadron collider environment.
## Acknowledgments
I would like to note my appreciation for the hospitality of St. John’s College, Durham, and thank M. Whalley, J. Forshaw and E.W.N. Glover for all their organisational work.
## References |
no-problem/0001/astro-ph0001391.html | ar5iv | text | # ON THE ORIGIN OF QUASI-PERIODIC OSCILLATIONS AND BROAD-BAND NOISE IN ACCRETING NEUTRON STARS AND BLACK HOLES
## 1 INTRODUCTION
Weakly-magnetic accreting neutron stars and black holes are known to exhibit a variety of phenomena on timescales that span a very wide range, from days to milliseconds. Soon after their discovery, it became clear that these objects show large brightness fluctuations as well as flickering or shot noise (see, e.g., Oda et al. 1971). It was, however, the launch of X-ray timing satellites (such as EXOSAT, GINGA, and especially RXTE) that led to the discovery of quasi-periodic oscillations in their X-ray brightness, with frequencies $`10^110^3`$ Hz (see van der Klis 1998 for a recent review).
Quasi-periodic oscillations (QPOs) are a common feature of the power-density spectra of accreting compact objects and come in many shapes and flavors. In weakly-magnetic accreting neutron stars, at least four distinct classes of QPOs have been identified so far: the $`615`$ Hz normal branch oscillation (NBO; Middleditch, & Priedhorsky 1986), the $`1070`$ Hz horizontal branch oscillation (HBO; van der Klis et al. 1985), and the $`1001200`$ Hz lower- and upper-kHz QPOs that typically occur in a pair (van der Klis et al. 1996). In black-hole systems, the QPO phenomenology has not been developed to the same extent so far, mostly because the majority of black-hole binaries are transient systems and, therefore, not always observable. In most canonical spectral states, however, low-frequency ($`10^110`$ Hz) QPOs and broader peaked noise components (at $`1100`$ Hz) with varying centroid frequencies have been detected (see van der Klis 1995 for a review). Moreover, in microquasars, QPOs with even higher (up to $`300`$ Hz) and possibly stationary frequencies are often observed (see, e.g., Morgan, Remillard, & Greiner 1997).
The physical mechanism that produces any (or all) of the rapid variability components in accreting compact objects is still the subject of active research. The detection of nearly coherent oscillations during Type I X-ray bursts in four low-luminosity neutron stars with frequencies that are constant in the tails of bursts that are separated by $``$20 months provides an important clue towards understanding the origin of the kHz QPOs (see Strohmayer et al. 1996). These oscillations are thought to occur at the stellar spin frequencies, as the latter are coherent and stable to the degree originally inferred observationally for the burst oscillations. This property motivated the beat-frequency interpretation of the kHz QPOs (Strohmayer et al. 1996; Miller, Lamb, & Psaltis 1998), in which the upper kHz QPO occurs at the Keplerian orbital frequency in a characteristic radius in the inner disk, whereas the lower kHz QPO occurs at the beat of the orbital frequency with the stellar spin (note here that beat-frequency models were originally developed for the HBO; see Alpar & Shaham 1985; Lamb et al. 1985).
Subsequent detailed analyses of the properties of the burst oscillations and the kHz QPOs have challenged the beat-frequency interpretation (see, however, Miller 2000). The frequencies and amplitudes of the burst oscillations were observed to vary non-monotonically during the bursts (Miller 1999; Strohmayer 1999). The peak separation of the kHz QPOs was shown to decrease with increasing kHz QPO frequencies (van der Klis et al. 1997; Méndez et al. 1998; Psaltis et al. 1998) and be statistically different than the burst oscillation frequencies (Méndez & van der Klis 1999). Moreover, the frequencies of QPOs (and noise components) in both neutron-star and black-hole systems follow similar tight correlations, suggesting that all these variability components cannot directly be related to the presence of a compact object with a hard surface or large-scale magnetic field (Wijnands & van der Klis 1999; Psaltis, Belloni, & van der Klis 1999; see also van der Klis 1994a, 1994b).
The frequencies of the HBO and kHz QPOs in neutron stars, as well as of the corresponding QPOs and peaked noise components in black holes, were recently shown to be nearly equal to the orbital, periastron-precession, and twice the nodal-precession frequencies of tilted test-particle orbits at the inner accretion disks (Stella, Vietri, & Morsink 1999; see also Stella & Vietri 1997, 1998, 1999). These frequencies depend entirely on the gravitational field around the compact object and not on the accretion flow properties and can, therefore, account naturally for the tight correlations between the observed QPO frequencies and their common occurrence in both types of accreting compact objects.
Modeling theoretically the time-dependence of inner accretion disks and their emission is challenging and is hampered mostly by our lack of understanding of the viscous mechanism for angular momentum transport. In the case of modeling high-frequency QPOs, the problem becomes even more difficult because radiation forces and general relativistic effects are non negligible (see, e.g., Miller et al. 1998). Studies of the inner disk normal modes have been very fruitful in revealing the potential of trapping low-frequency oscillations very close to the compact objects (see, e.g., Kato 1989; Nowak & Wagoner 1991; Perez et al. 1997). Some of the resulting diskoseismic modes have frequencies comparable to the orbital, periastron-precession, and nodal-precession frequencies of tilted test-particle orbits at the inner accretion disks. Exciting or trapping these oscillations at a characteristic radius in the inner disk is therefore a prime candidate for producing the modulations required in the QPO interpretation proposed by Stella & Vietri (1998, 1999).
In this paper we explore an alternative mechanism for producing high-amplitude variability predominantly at these characteristic general relativistic frequencies in the inner accretion disks. We are motivated by the similarity between the observed power-density spectra of accreting compact objects and the response function to external perturbations of driven mechanical and electrical systems (such as a pendulum in a viscous medium or an RLC circuit; see Bendat & Piersol 1971, §2.4). We do not require modes with only particular frequencies to be excited or trapped at a characteristic radius, but rather study the response of the accretion disk inside this radius to a broad spectrum of perturbations imposed outside of it. The broad-band noise component of the power-density spectrum of the compact object is then dictated by the broad-band frequency response of the inner disk. Resonances occur at the normal modes of the accretion disk with a frequency response that is less suppressed than in the nearby frequencies and, therefore, appear as the QPO peaks in the power-density spectra. In this picture, both the break frequency of the broad-band noise component and the QPO frequencies are associated with properties in the same narrow annulus in the inner accretion disk and therefore the correlations between them can be naturally accounted for.
## 2 ASSUMPTIONS
The observed QPOs are typically narrow, with fractional widths as low as $`\delta \nu /\nu 10^2`$ (see, e.g., van der Klis et al. 1997). Because all characteristic frequencies in an accretion flow have a strong dependence on radius and height above the equatorial plane, such small fractional widths severely constrain the location of the physical mechanism that determines the QPO frequencies, i.e., it should be associated with a narrow annulus in a geometrically thin accretion disk, so that $`\delta r/r\delta \nu /\nu 10^2`$ and $`h/r(\delta \nu /\nu )^{1/2}10^1`$. This narrow annulus represents a transition in the accretion disk properties but its exact nature is not important for the current study. Hereafter, for simplicity, we will refer to the flow outside the transition radius as the outer accretion disk and to the flow inside as the inner accretion disk, even though the transition radius can be as small as a few Schwarzchild radii.
We describe the structure of the accretion disk using the continuity equation for the density $`\rho `$,
$$\frac{\rho }{t}+\frac{}{rr}(r\rho u_r)+\frac{}{r\varphi }(\rho u_\varphi )+\frac{}{z}(\rho u_z)=0,$$
(1)
and Euler’s equations for the three components of the velocity field, $`(u_r,u_\varphi ,u_z)`$, in a cylindrical reference frame,
$`{\displaystyle \frac{u_r}{t}}+u_r{\displaystyle \frac{u_r}{r}}+u_\varphi {\displaystyle \frac{u_r}{r\varphi }}+u_z{\displaystyle \frac{u_r}{z}}{\displaystyle \frac{u_\varphi ^2}{r}}={\displaystyle \frac{\psi }{r}}{\displaystyle \frac{1}{\rho }}{\displaystyle \frac{P}{r}}+N_r,`$ (2)
$`{\displaystyle \frac{u_\varphi }{t}}+u_r{\displaystyle \frac{u_\varphi }{r}}+u_\varphi {\displaystyle \frac{u_\varphi }{r\varphi }}+u_z{\displaystyle \frac{u_\varphi }{z}}+{\displaystyle \frac{u_ru_\varphi }{r}}={\displaystyle \frac{1}{\rho }}{\displaystyle \frac{P}{r\varphi }}+N_\varphi ,`$ (3)
$`{\displaystyle \frac{u_z}{t}}+u_r{\displaystyle \frac{u_z}{r}}+u_\varphi {\displaystyle \frac{u_z}{r\varphi }}+u_z{\displaystyle \frac{u_z}{z}}={\displaystyle \frac{\psi }{z}}{\displaystyle \frac{1}{\rho }}{\displaystyle \frac{P}{z}}+N_z,`$ (4)
where $`P=\rho c_\mathrm{s}^2`$, $`c_\mathrm{s}`$ is the local speed of sound, and its value can be determined by solving a local energy equation. The structure and response of the accretion disk depends on the gravitational potential $`\psi `$, the viscous torque described by the vector $`(N_r,N_\varphi ,N_z)`$, which is the projection of the viscous tensor on the coordinate vectors, and the local heating and cooling of the accretion flow. We describe now our assumptions regarding these properties of the central object and the accreting gas.
We perform our analysis neglecting any explicit special or general relativistic effects. We specify the gravitational potential of the central object through the relations
$`\mathrm{\Omega }^2`$ $`=`$ $`{\displaystyle \frac{1}{r}}\left({\displaystyle \frac{\psi }{r}}\right)_{z=0},`$ (5)
$`\kappa ^2`$ $`=`$ $`2\mathrm{\Omega }\left(2\mathrm{\Omega }+r{\displaystyle \frac{d\mathrm{\Omega }}{dr}}\right),`$ (6)
$`\mathrm{\Omega }_{}^2`$ $`=`$ $`{\displaystyle \frac{1}{z}}\left({\displaystyle \frac{\psi }{z}}\right)_r,`$ (7)
where $`\mathrm{\Omega }`$ is the Keplerian orbital frequency, $`\kappa `$ is the epicyclic frequency, and $`\mathrm{\Omega }_{}`$ is the frequency of vertical oscillations in an annulus in the accretion disk. In Newtonian dynamics, $`\mathrm{\Omega }_{}=\kappa =\mathrm{\Omega }`$ for a non-rotating central object. However, we allow here for the possibility that the three frequencies do not satisfy these relations, taking thus into account, in an approximate way, the general-relativistic effects of periastron precession and frame dragging (see, e.g., Misner, Thorne, & Wheeler 1973). This is equivalent with assuming a pseudo-potential for the gravitational field for which the three characteristic frequencies have the same value as the exact general relativistic frequencies.
We study the response to perturbations of an annulus in a geometrically thin accretion disk that is localized in the radial and vertical directions. We therefore assume that the $`r`$ and $`z`$ components of the vector $`\stackrel{}{N}`$ are negligible and set
$$N_\varphi u_\varphi \mathrm{\Omega }_\mathrm{d},$$
(8)
where $`\mathrm{\Omega }_\mathrm{d}`$ is defined in this way to be the inverse of the radial drift timescale. Equation (8) allows us to avoid specifying any particular recipe for the largely unknown mechanism that transports angular momentum in the disk.
Finally, for simplicity, we assume that the perturbations do not alter the local temperature, and hence the sound speed, in the accretion flow. Strictly speaking, this assumption is justified for fluctuations on timescales much shorter than the thermal timescale but breaks down for slower phenomena. However, our results do not depend strongly on this approximation, as it will become obvious later. Indeed, previous studies of isothermal or adiabatic diskoseismic modes produced qualitatively similar results (see, e.g., Nowak & Wagoner 1991).
## 3 THE RESPONSE OF AN ACCRETION DISK TO A SPECTRUM OF PERTURBATIONS
As discussed in §2, we assume that the properties in the accretion disk exhibit a transition over an annulus of width $`\delta r`$ at a characteristic radius $`r`$. Defining radial averages over the annulus width of all physical quantities as
$$f=\frac{1}{r\delta r}_r^{r+\delta r}rf𝑑r,$$
(9)
and dropping for clarity the angle brackets, the continuity equation becomes
$$\frac{\rho }{t}+\frac{}{r\varphi }(\rho u_\varphi )+\frac{}{z}(\rho u_z)=\frac{1}{r\delta r}_r^{r+\delta r}\frac{}{r}(r\rho u_r)𝑑r=\frac{1}{r\delta r}\left(\left[r\rho u_r\right]_{r+\delta r}\left[r\rho u_r\right]_r\right)$$
(10)
The first term in the parenthesis in the right-hand side of equation (10) corresponds to the incoming mass flux from the outer radius of the annulus, whereas the second term corresponds to the outgoing mass flux from the inner radius. We set $`[r\rho u_r]_{r+\delta r}[r\rho u_r]_rr^2(\rho _{\mathrm{in}}\rho )\mathrm{\Omega }_\mathrm{d}`$, where $`\rho _{\mathrm{in}}`$ describes the density at the outer edge of the annulus. The continuity equation then becomes
$$\frac{\rho }{t}+\frac{r}{\delta r}\mathrm{\Omega }_\mathrm{d}\rho +\frac{}{r\varphi }(\rho u_\varphi )+\frac{}{z}(\rho u_z)=\frac{r}{\delta r}\rho _{\mathrm{in}}\mathrm{\Omega }_\mathrm{d}.$$
(11)
We assume that the steady-state disk is geometrically thin and in vertical hydrostatic equilibrium, with $`u_\varphi >>u_r,u_z`$. We also assume that, in steady state, $`u_\varphi =\mathrm{\Omega }r`$. We introduce small perturbations to all physical quantities (denoted by the superscript ‘1’) and linearize equations (2)–(4) and (11) keeping only terms up to first order in all small quantities. The resulting linearized equations are
$`\left({\displaystyle \frac{}{t}}+\mathrm{\Omega }{\displaystyle \frac{}{\varphi }}+{\displaystyle \frac{r}{\delta r}}\mathrm{\Omega }_\mathrm{d}\right)\rho ^1+{\displaystyle \frac{\rho }{r}}{\displaystyle \frac{u_\varphi ^1}{\varphi }}+{\displaystyle \frac{}{z}}(\rho u_z^1)`$ $`=`$ $`{\displaystyle \frac{r}{\delta r}}\rho _{\mathrm{in}}^1\mathrm{\Omega }_\mathrm{d},`$ (12)
$`\left({\displaystyle \frac{}{t}}+\mathrm{\Omega }{\displaystyle \frac{}{\varphi }}\right)u_r^12\mathrm{\Omega }u_\varphi ^1`$ $`=`$ $`{\displaystyle \frac{c_\mathrm{s}^2}{r}}\left({\displaystyle \frac{\delta r}{r}}\right)\left({\displaystyle \frac{\rho _{\mathrm{in}}^1\rho ^1}{\rho }}\right),`$ (13)
$`\left({\displaystyle \frac{}{t}}+\mathrm{\Omega }{\displaystyle \frac{}{\varphi }}+\mathrm{\Omega }_\mathrm{d}\right)u_\varphi ^1+{\displaystyle \frac{\kappa ^2}{2\mathrm{\Omega }}}u_r^1`$ $`=`$ $`{\displaystyle \frac{c_\mathrm{s}^2}{r}}{\displaystyle \frac{}{\varphi }}\left({\displaystyle \frac{\rho ^1}{\rho }}\right),`$ (14)
$`\left({\displaystyle \frac{}{t}}+\mathrm{\Omega }{\displaystyle \frac{}{\varphi }}\right)u_z^1`$ $`=`$ $`c_\mathrm{s}^2{\displaystyle \frac{}{z}}\left({\displaystyle \frac{\rho ^1}{\rho }}\right).`$ (15)
Because the above system of equations is linear, the solution for the perturbed density $`\rho ^1`$ can be written as the convolution
$$\rho ^1(t,\varphi ,z)=_{\mathrm{}}^{\mathrm{}}h(\tau ;\varphi ,z)\rho _{\mathrm{in}}^1(t\tau ,\varphi ,z)𝑑t,$$
(16)
where the function $`h(\tau )`$ is called the ‘weighting function’ of the linear system. For the linear system to be physically realizable, it is necessary that it responds only to past inputs (Bendat & Piersol 1971, §2.3), and hence that
$$h(\tau <0)=0.$$
(17)
Instead of solving for the weighting function $`h(\tau )`$, we define the response function $`A(\omega )`$ as
$$A(\omega )_{\omega =0}^{\mathrm{}}h(\tau )\mathrm{exp}(i\omega \tau )𝑑\tau $$
(18)
such that the Fourier transform $`\stackrel{~}{\rho }^1(\omega )`$ of the solution is related to the Fourier transform $`\stackrel{~}{\rho }_{\mathrm{in}}^1(\omega )`$ of the driving term by
$$\stackrel{~}{\rho }^1(\omega )=A(\omega )\stackrel{~}{\rho }_{\mathrm{in}}^1(\omega ).$$
(19)
Note that in the definition (18) the lower limit of the integral is at $`\omega =0`$, because of condition (17).
We can obtain the complete solution of the linear system of equations by expanding the temporal and $`\varphi `$-dependence of the perturber in modes of the form
$$\rho _{\mathrm{in}}^1(t,\varphi ,z)=\underset{m}{}_\omega \rho _{\mathrm{in},m}(\omega ;z)e^{i(\omega tm\varphi )}.$$
(20)
Because $`|A(\omega ,m)|=|A(\omega ,m)|`$ (see also Bendat & Pierson 1971), we will consider only the response of the system to perturbations with positive frequencies. Moreover, we will consider only positive values for $`m`$, as negative values correspond to perturbations that rotate in the retrograde sense with respect to the local Keplerian velocity and are expected to have weak intrinsic amplitudes. Setting $`\xi \rho ^1/\rho `$ and $`\xi _{\mathrm{in},m}\rho _{\mathrm{in},\mathrm{m}}/\rho `$ we obtain
$`i(\omega m\mathrm{\Omega })\xi +{\displaystyle \frac{r}{\delta r}}\mathrm{\Omega }_\mathrm{d}\xi +i{\displaystyle \frac{m}{r}}u_\varphi ^1{\displaystyle \frac{\mathrm{\Omega }_{}^2}{c_\mathrm{s}^2}}zu_z^1+{\displaystyle \frac{u_z^1}{z}}`$ $`=`$ $`{\displaystyle \frac{r}{\delta r}}\mathrm{\Omega }_\mathrm{d}\xi _{\mathrm{in},m}`$ (21)
$`i(\omega m\mathrm{\Omega })u_r^12\mathrm{\Omega }u_\varphi ^1`$ $`=`$ $`{\displaystyle \frac{c_\mathrm{s}^2}{r}}\left({\displaystyle \frac{\delta r}{r}}\right)\left(\xi _{\mathrm{in},\mathrm{m}}\xi \right)`$ (22)
$`i(\omega m\mathrm{\Omega })u_\varphi ^1+\mathrm{\Omega }_\mathrm{d}u_\varphi ^1+{\displaystyle \frac{\kappa ^2}{2\mathrm{\Omega }}}u_r^1`$ $`=`$ $`{\displaystyle \frac{1}{r}}mc_\mathrm{s}^2i\xi `$ (23)
$`i(\omega m\mathrm{\Omega })u_z^1`$ $`=`$ $`c_\mathrm{s}^2{\displaystyle \frac{\xi }{z}}.`$ (24)
Combining the above into a single equation for $`\xi `$ gives
$`{\displaystyle \frac{r}{\delta r}}\mathrm{\Omega }_\mathrm{d}\left({\displaystyle \frac{\xi }{\xi _{\mathrm{in},m}}}\right)^1`$ $`=`$ $`{\displaystyle \frac{i}{\omega \omega _m}}\left(\mathrm{\Omega }_{}^2z{\displaystyle \frac{\xi }{\xi z}}c_\mathrm{s}^2{\displaystyle \frac{^2\xi }{\xi z^2}}\right)+\left[i(\omega \omega _m)+{\displaystyle \frac{r}{\delta r}}\mathrm{\Omega }_\mathrm{d}\right]`$ (25)
$`+{\displaystyle \frac{m^2\mathrm{\Omega }_\mathrm{c}^2(\omega \omega _m)}{i(\omega \omega _p)(\omega \omega _{p+})+\mathrm{\Omega }_\mathrm{d}(\omega \omega _m)}},`$
where $`\omega _mm\mathrm{\Omega }`$, $`\omega _{p\pm }m\mathrm{\Omega }\pm \kappa `$, $`\mathrm{\Omega }_\mathrm{c}c_\mathrm{s}/r`$, and we have neglected for simplicity the radial pressure forces..
In principle, equation (25) must be solved for the functional dependence of $`\xi `$ on $`z`$ by imposing proper boundary conditions at the equatorial plane and at infinity. In previous studies of trapped oscillatory modes in accretion disks (see, e.g., Okazaki, Kato, & Fukue 1987), it was shown that the $`z`$dependence of $`\xi `$ can be written in terms of Hermite polynomials, each order of which represent a different vertical mode. Driven by this result, we will simply study here the response of the disk to modes for which
$$\frac{\xi }{z}=n^2\frac{\xi }{z};n=0,1,$$
(26)
where $`n`$ represents the vertical mode number. This is equivalent with retaining only the Hermite polynomials of the lowest two orders in the expansion of the $`z`$dependence of $`\xi `$. Equation (25) then becomes
$`{\displaystyle \frac{r}{\delta r}}\mathrm{\Omega }_\mathrm{d}\left({\displaystyle \frac{\xi }{\xi _{\mathrm{in},m}}}\right)^1`$ $`=`$ $`{\displaystyle \frac{i(\omega \omega _{n+})(\omega \omega _n)+(r/\delta r)\mathrm{\Omega }_\mathrm{d}(\omega \omega _m)}{\omega \omega _m}}`$ (27)
$`+{\displaystyle \frac{m^2\mathrm{\Omega }_\mathrm{c}^2(\omega \omega _m)}{i(\omega \omega _{p+})(\omega \omega _p)+\mathrm{\Omega }_\mathrm{d}(\omega \omega _m)}},`$
where $`\omega _{n\pm }m\mathrm{\Omega }\pm n\mathrm{\Omega }_{}`$. The response function of the transition annulus in the accretion disk to external perturbations of a given mode is then simply (see, eq.)
$$A_{\mathrm{mn}}(\omega )=\frac{\xi }{\xi _{\mathrm{in},m}}.$$
(28)
In the above analysis, we have made the explicit assumption that all perturbations are isothermal. Therefore, by construction, the radiation emerging from the transition annulus in the accretion disk will be unperturbed and therefore time-independent. However, the mass flux to the inner disk and possibly the loading of the hot Comptonizing medium responsible for the hard X-ray spectral component will be modulated with a response function of similar qualitative characteristics as equation (28). Modulation of the scattering optical depth in the Comptonizing medium can produce the strong photon-energy dependence of the observed modulation amplitudes (see, e.g., Lee & Miller 1998; Miller, Lamb, & Psaltis 1998). We, therefore, make the working hypothesis that the function (28) describes also the response of the radiation emerging from the inner accretion flow to perturbations produced outside the transition radius.
### 3.1 Mode Analysis
In this section we study in more detail the response function (28) derived in §3 for the three lowest-order modes of perturbation, $`(m=0,n=0)`$, $`(m=1,n=0)`$, and $`(m=1,n=1)`$. The response of the inner accretion disk described by these three modes is shown in Figure 1, for typical values of the characteristic frequencies.
#### 3.1.1 Mode $`(m=0,n=0)`$
This mode describes time-dependent perturbations of the gas density in the accretion disk that are uniform in both the azimuthal and vertical directions. The magnitude of its response function is
$$A_{00}=\frac{1}{\left[1+(\omega /\omega _\mathrm{b})^2\right]^{1/2}},$$
(29)
where $`\omega _\mathrm{b}(r/\delta r)\mathrm{\Omega }_\mathrm{d}`$. It describes a low band-pass filter with a response that cuts off at $`\omega _\mathrm{b}`$ and decreases as $`A_{00}\omega ^1`$ at higher frequencies.
The characteristic properties of $`A_{00}`$ can be easily understood given the fact that $`\omega _\mathrm{b}^1`$ is the radial drift timescale in the annulus with radial width $`\delta r`$. Any density perturbations with characteristic timescales longer than $`\omega _\mathrm{b}^1`$ propagate inwards unaffected, whereas perturbations at increasingly shorter timescales are dumped with an increasingly higher efficiency. Note that there is no characteristic frequency in the disk that corresponds to the $`(m=0,n=0)`$ mode and hence its response function shows no resonances.
#### 3.1.2 Mode $`(m=1,n=0)`$
This is an one-armed mode that is uniform in the vertical direction. As it is also evident in Figure 1, the magnitude of its response is negligible for most frequencies, besides the frequency ranges near $`\omega \mathrm{\Omega }`$, $`\omega \mathrm{\Omega }\kappa `$, and $`\omega \mathrm{\Omega }+\kappa `$. This is caused by the differential rotation of gas in the accretion disk that can strongly dump any azimuthal density perturbations that are driven at frequencies other than one of the local characteristic frequencies.
Although the complete response is given by the analytic form (28), we can obtain some simpler expressions by expanding relation (28) around the resonant frequencies. For $`\omega \mathrm{\Omega }`$,
$$A_{10}\left[\frac{\omega _\mathrm{b}^2}{\omega _\mathrm{b}^2+(1+\mathrm{\Omega }_\mathrm{c}^2/\kappa ^2)^2(\omega \mathrm{\Omega })^2}\right]^{1/2}.$$
(30)
Therefore, the resonance occurs at the orbital frequency
$$\omega _3\mathrm{\Omega }$$
(31)
and the FWHM of the response around $`\omega _3`$ is
$$\delta \omega _32\sqrt{3}\omega _\mathrm{b}\left(1+\frac{\mathrm{\Omega }_\mathrm{c}^2}{\kappa ^2}\right)^1.$$
(32)
The response function around the other two resonances is more complicated and can be simplified analytically in the limit $`\kappa /\omega _\mathrm{b}1`$, which is usually true for the cases of interest here. The resonances occur at the frequencies
$$\omega _{2\pm }\mathrm{\Omega }\pm \kappa \left(1+\frac{\mathrm{\Omega }_\mathrm{c}^2}{2\kappa ^2}\right)$$
(33)
and the FWHM of the response around these frequencies is
$$\delta \omega _{2\pm }\sqrt{3}\omega _\mathrm{b}\left(\frac{\mathrm{\Omega }_\mathrm{c}^2}{\kappa ^2}\right).$$
(34)
If, moreover, $`(\delta r/r)(\kappa /\mathrm{\Omega }_\mathrm{c})^2<<1`$ then the response function around the resonances can be put in the form
$$A_{10}\left\{1+\left[\frac{\kappa }{\omega _\mathrm{b}}+\left(\frac{\mathrm{\Omega }_\mathrm{c}}{2\omega _\mathrm{b}}\right)\frac{\mathrm{\Omega }_c}{\omega (\mathrm{\Omega }\pm \kappa )}\right]^2\right\}^1,\text{for}\omega \mathrm{\Omega }\pm \kappa .$$
(35)
Note that the frequencies $`\omega _{2\pm }`$ derived here are consistent with the ones obtained simply by using the dispersion relation of the inertial-acoustic waves derived by Kato (1989; eq. \[3.13\] for $`m=1`$ and $`n=0`$)
$$(\omega \mathrm{\Omega })^2\kappa ^2=k_r^2c_\mathrm{s}^2$$
(36)
and setting their radial wavenumber equal to $`k_r1/r`$. Note, also, that the amplitude of the response at $`\omega _{2\pm }`$ is comparable to unity only in the limit $`(\delta r/r)(\kappa /\mathrm{\Omega }_\mathrm{c})^2<<1`$ and because we have neglected radial pressure forces, which damp radial oscillations.
#### 3.1.3 Mode $`(m=1,n=1)`$
This mode describes a precessing, one-armed azimuthal density perturbation with a vertical tilt. As it is also evident in Figure 1, the differential rotation of tilted orbits in the disk strongly dumps all such perturbations, unless $`\omega \mathrm{\Omega }\pm \mathrm{\Omega }_{}`$ (the upper resonance of this mode is not plotted in the figure).
The response function around the precession frequency is
$$A_{11}\left\{\left[1+\left(\frac{\delta r}{r}\right)\frac{\mathrm{\Omega }_\mathrm{c}^2\mathrm{\Omega }_{}^2}{\mathrm{\Omega }_\mathrm{d}^2\mathrm{\Omega }_{}^2+(\kappa ^2\mathrm{\Omega }_{}^2)^2}\right]^2+\left(\frac{\omega \omega _{1\pm }}{\omega _\mathrm{b}/2}\right)^2\right\}^1,\text{for}\omega \mathrm{\Omega }\pm \mathrm{\Omega }_{},$$
(37)
where
$$\omega _{1\pm }\mathrm{\Omega }\pm \mathrm{\Omega }_{}\left[1+\frac{\mathrm{\Omega }_\mathrm{c}^2(\mathrm{\Omega }_{}^2\kappa ^2)}{2\mathrm{\Omega }_\mathrm{d}^2\mathrm{\Omega }_{}^2+2(\kappa ^2\mathrm{\Omega }_{}^2)^2}\right]$$
(38)
Typically, the response $`A_{11}`$ peaks at $`\omega \omega _{1\pm }`$ with a FWHM of
$$\delta \omega _{1\pm }\sqrt{3}\omega _\mathrm{b}\left[1+\left(\frac{\delta r}{r}\right)\frac{\mathrm{\Omega }_\mathrm{c}^2\mathrm{\Omega }_{}^2}{\mathrm{\Omega }_\mathrm{d}^2\mathrm{\Omega }_{}^2+(\kappa ^2\mathrm{\Omega }_{}^2)^2}\right].$$
(39)
How pronounced the peak at $`\omega _1`$ is depends on the relative magnitude of the radial drift timescale and the resonance frequency (Fig. 2). As $`\omega _1`$ decreases, the resonance peak becomes broader, the low-frequency response for this mode becomes comparable to unity, and the overall response shows a break at a frequency comparable to $`\omega _1`$. For the response function to show a narrow and pronounced resonance peak, two conditions need to be satisfied, i.e., $`(\delta \omega _1/\omega _1)1`$ and $`lim_{\omega 0}A_{11}1`$, both of which lead to the single requirement $`\omega _\mathrm{b}\omega _1`$. Note also that the amplitude of the response at $`\omega _{1\pm }`$ is unity only because we neglected $`N_z`$ and hence the damping of this mode caused by the differential precession of nearby fluid elements..
## 4 APPLICATION TO GALACTIC NEUTRON STARS AND BLACK HOLES
In this section we study the applicability of the disk modes analyzed in §3 in modeling the X-ray variability properties of galactic neutron stars and black holes. We first address the basic assumptions in our analysis and then compare directly the predicted mode frequencies to the observed QPO and noise frequencies in a variety of sources.
The presence of a sharp transition in the properties of a geometrically thin accretion disk is a generic requirement of any model that attributes any of the observed QPO and noise frequencies to characteristic frequencies in the disk. The physical mechanism that produces this sharp transition is not specified in our analysis. Indeed, the predicted mode frequencies depend mostly on the gravitational field around the compact object and only weakly on the hydrodynamic properties of the flow itself. As a result, different mechanisms may give rise to the sharp transition in different sources and yet produce QPOs and noise with the same characteristic frequencies. This is an important property of the model given the fact that similar QPOs and noise components have been identified in sources with widely different masses, accretion rates, and magnetic field strengths, but with centroid frequencies that follow very similar correlations (see, e.g., Psaltis et al. 1999a). Sharp transitions in accretion disk properties have been discovered in a variety of situations, such as the transition around the sonic point caused by radiation drag (Miller et al. 1998), the heating front during a dwarf nova instability (see, e.g., Menou, Hameury, & Stehle 1999), or the transition between a thin accretion disk and an advection-dominated accretion flow (see, e.g., Kato & Nakamura 1998). Any or all of these mechanisms might be responsible for the transition that causes the observed X-ray variability in different accreting compact objects.
In the physical picture presented here, the various disk modes need not be excited or indefinitely trapped only in a localized region of the accretion flow. In fact, some of the basic features of the model depend on the slow leakage of power from the different modes excited outside the transition radius, which is modeled by the parameter $`\mathrm{\Omega }_\mathrm{d}`$ in §2. This leakage is what generates the broad-band response plotted in Figure 1 and hence the noise component of the observed power-density spectra in galactic sources. Moreover, if the leakage of power is slower than the mode period, it will not affect the presence of the resonance peaks but will determine only their widths. This is a general argument regarding the observational signatures of disk modes, applicable also to the trapped but slowly leaking modes reported elsewhere (e.g., Okazaki et al. 1987; Kato 1989, 1990; Nowak & Wagoner 1991; Perez et al. 1997, etc.). It is, however, important that significant power is generated close to the transition radius, because the characteristic timescales at larger radii are too short for strong perturbations to be generated at frequencies comparable to the higher-frequency resonance peaks.
Although the small fractional QPO widths imply that their frequencies are determined in a characteristic radius in the accretion disk, their hard X-ray spectra (see, e.g., Berger et al. 1996) suggest that the emission from the hot component of the accretion flow is also modulated at the QPO frequency. If the hard power-law components of the X-ray spectra are produced by thermal Comptonization of soft photons by hot electrons, then a small-amplitude oscillation of the density in the Comptonizing medium can account for the observed QPO energy spectra (Lee & Miller 1998; Miller et al. 1998). This has been our motivation for calculating the frequency response of the density in the inner accretion disk to a broad spectrum of external perturbations.
We have performed our analysis neglecting any explicit special or general relativistic effects. However, we have allowed for the possibility that the azimuthal and vertical frequencies, $`\mathrm{\Omega }`$ and $`\mathrm{\Omega }_{}`$, are not equal and that the periastron precession frequency is not zero. Following Stella et al. (1999), we will use the weak-field and slow-rotation limits of the general relativistic expressions for these frequencies and compare the predicted frequencies to the observed QPOs (note here that our expressions are not strictly valid for a rapidly rotating neutron star, the external spacetime of which is not described by the Kerr metric). Given a transition radius $`r`$, as well as a mass $`M`$ and specific angular momentum per unit mass $`a_{}`$ of the central compact object, we obtain for the corresponding frequencies for a prograde disk, as measured by a static observed at infinity (see Bardeen et al. 1972; Perez et al. 1997; Stella et al. 1999),
$`\mathrm{\Omega }^2`$ $`=`$ $`{\displaystyle \frac{M}{r^3[1+a_{}(M/r)^{3/2}]}},`$
$`\kappa ^2`$ $`=`$ $`\mathrm{\Omega }^2(16M/r),`$ (40)
and
$$\mathrm{\Omega }_{}^2=\mathrm{\Omega }^2[14a_{}(M/r)^{3/2}],$$
(41)
where we have set $`G=c=1`$, $`G`$ is the gravitational constant and $`c`$ is the speed of light. Finally, we estimate the hydrodynamic timescales related to the accretion flow using the standard $`\alpha `$-prescription for the viscosity (Shakura & Sunyaev 1973; see also Frank, King, & Raine 1992), as
$$\mathrm{\Omega }_\mathrm{V}\alpha (h/r)^2\mathrm{\Omega },$$
(42)
and
$$\mathrm{\Omega }_\mathrm{c}(h/r)\mathrm{\Omega }_{},$$
(43)
where $`\mathrm{\Omega }_\mathrm{V}^1`$ is the viscous timescale, $`h`$ is the scale height of the disk, and $`\alpha `$ is the viscosity parameter, and set for simplicity the radial drift timescale at the transition radius equal to the viscous timescale, i.e., $`\mathrm{\Omega }_\mathrm{d}=\mathrm{\Omega }_\mathrm{V}`$.
Following Stella et al. (1999) we identify (a) the upper kHz QPO with the resonance at $`\omega _3`$ (i.e., the Keplerian frequency) of the $`(m=1,n=0)`$ mode, (b) the lower kHz QPO, or peaked noise component identified as such, with the resonance at $`\omega _2`$ of the $`(m=1,n=0)`$ mode, and (c) the HBO, or low-frequency QPO identified as such, with twice the frequency of the resonance at $`\omega _1`$ of the $`(m=1,n=1)`$ mode; the only reason for choosing twice the frequency $`\omega _1`$ is to account for the relatively high observed HBO frequencies in neutron star systems, which are otherwise inconsistent with the properties of any stable star in general relativity (Stella et al. 1999; see also Psaltis et al. 1999b; Kalogera & Psaltis 2000). The latter choice can be motivated by the presence of strong subharmonics of the QPOs in both black-hole and neutron-star systems<sup>1</sup><sup>1</sup>1We treat here the ‘intermediate’ branch of the HBO discussed by Psaltis et al. (1999a; see also Ford & van der Klis 1998) as evidence for subharmonic structure. and justified by the two-fold symmetry of the $`(m=1,n=1)`$ mode.
Figure 3 shows the observed correlation between the lower and upper kHz QPOs in a number of neutron-star systems (after Psaltis et al. 1998) and compares it to the predicted relation between the resonant frequencies $`\omega _2`$ and $`\omega _3`$. The predicted relation has a very weak dependence on the compact object spin and we therefore set $`a_{}=0`$ for the curves plotted. On the other hand, we allow for different values of the compact object mass and the scale height of the inner accretion disk. We also assume that the maximum possible Keplerian frequency is that of the innermost stable circular orbit and this results in an upper bound on the upper kHz QPO frequency (vertical lines in Fig. 3; see also Miller et al. 1998). We find that the observed correlation can be accounted for, in the present picture, if the neutron stars are relatively massive ($`1.92.2M_{}`$), in agreement with Stella et al. (1999). Moreover, small discrepancies between the general relativistic frequencies and the observed QPO frequencies can be easily accommodated in the present model (for reasonable values of the disk scale height), because of the corrections introduced to the mode frequencies by the properties of the accretion flow.
Note here that in the absence of any broadening mechanisms, the lower kHz QPO can be narrower than the upper kHz QPO (compare equations and ). This is consistent with the observational fact that, when the two kHz QPOs have different fractional widths, the lower kHz QPO is systematically narrower than the upper kHz QPO (see, e.g., Méndez et al. 1998; also van der Klis, private communication). Accounting for this property in a beat-frequency model of the kHz QPOs, in which the lower kHz QPO is produced by the beat of the upper kHz QPO with a coherent oscillation, requires the upper kHz QPO peak to be additionally broadened by an external mechanism.
Figure 4 shows the observed correlation between the HBO and lower kHz QPO (or the peaked noise component identified as such) in a number of neutron-star and black-hole systems (after Psaltis et al. 1999a) and compares it to the predicted relation between the resonant frequencies $`2\omega _1`$ and $`\omega _2`$. The predicted relation has a weak dependence on the mass of the compact object (see also Stella et al. 1999) and we therefore set it equal to $`2.2M_{}`$, allowing for different values of the compact object spin and disk scale height. We again find that the observed correlation can be accounted for in the present picture, if the spins of the compact objects are in the range $`a_{}0.10.4`$. Note here that, when the hydrodynamic corrections are non-negligible and hence when $`\omega _1\mathrm{\Omega }\mathrm{\Omega }_{}`$, the resonance peaks at $`2\omega _1`$ are broad and insignificant and therefore will not appear as prominent QPO peaks (cf. Fig 2).
In the physical mechanism we discuss in this paper, the shape of the broad-band noise of neutron-star and black-hole systems is the result of multiplying the intrinsic power spectrum of density perturbations at the outer disk with the square of the frequency response of the $`(m=0,n=0)`$ mode. If the power spectrum of the density perturbations is relatively flat over a range of frequencies around $`\omega _\mathrm{b}`$, then the resulting power spectrum will be flat at low frequencies, have a cut-off at
$$\omega _\mathrm{b}10\left(\frac{\alpha }{0.1}\right)\left(\frac{\delta r/r}{0.1}\right)^1\left(\frac{h/r}{0.1}\right)^2\left(\frac{\mathrm{\Omega }}{1\text{kHz}}\right)\text{Hz},$$
(44)
and drop above it as $`\omega ^2`$. Assuming that the radial drift timescale in the transition radius is equal to the viscous timescale, the relation between the break and HBO frequencies depends mostly on the fractional width $`\delta r/r`$ and scale height $`h/r`$ of the transition annulus. Modeling, therefore, the correlation discovered by Wijnands & van der Klis (1999) between these two variability components requires a detailed study of the accretion flow properties, which is beyond the scope of the current paper. According to the observed correlation $`\omega _\mathrm{b}\omega _1`$, and since in our model $`\omega _1\mathrm{\Omega }^2`$, this would require
$$\left(\frac{\delta r}{r}\right)\left(\frac{h}{r}\right)^2\mathrm{\Omega }^1.$$
(45)
Note though that the radial drift timescale at the transition radius cannot be, by construction, strictly equal to the viscous timescale (even though we assumed it to be for simplicity) and, therefore, the above correlation is not a strict requirement for our model. In any case, however, equation (44) shows that the overall properties of the $`(m=0,n=0)`$ mode are consistent with the shapes and break frequencies of the noise components observed in both neutron-star and black-hole systems (see, e.g., Wijnands & van der Klis 1999).
We do not attempt here to compare in detail the predictions of our model to specific sources and in particular to neutron stars for which the quality of the data is best, because of a number of systematic effects that introduce uncertainties significantly larger than those of the frequency measurements. First, in calculating the QPO frequencies we assumed a Kerr spacetime around the compact object, which is inconsistent with the high neutron-star spin frequencies required in this model (see discussion below). Second, the hydrodynamic corrections to the QPO frequencies depend strongly on the local temperature and the largely unknown viscosity law and are specific to the response of the particular physical quantity (i.e., the density) we chose to calculate (see, e.g., the discussion in Bendat & Pierson 1971, §2.4). Third, the centroid frequencies of relatively wide QPO peaks depend not only on the shape of the response function but on the power spectrum of the perturber, as well. Finally, for relatively weak QPO peaks, the uncertainties in the measurement of their centroid frequencies is often dominated by uncertainties in the subtraction of the noise continuum.
Bearing in mind all these caveats, neglecting the hydrodynamic corrections, and for a Kerr spacetime and transition radii larger than the radius of the innermost stable circular orbit, the resonance frequencies in our model follow the relations (Stella et al. 1999)
$$\omega _2\omega _3^{5/3}$$
(46)
and
$$\omega _1\omega _3^2.$$
(47)
The predicted power-law dependences are consistent with the observed correlations empirically inferred in detailed statistical studies of a number of neutron-star sources (Fig. 2b in Psaltis et al. 1998; Fig. 8 in Psaltis et al. 1999b). The clear exception appears to be the correlation between the HBO and upper kHz QPO frequencies in Sco X-1, for which the data favor a significantly flatter relation. It seems, however, likely (see Fig. 2b in Psaltis et al. 1999a) that the apparent flat correlation in this source is an artifact of treating as similar data points that follow two distinct branches (i.e., the ‘HBO’ and ‘intermediate’ branches) of the correlations discussed in Psaltis et al. (1999a). Whether Sco X-1 exhibits a behavior similar to that of 4U 1728$``$34 (see Ford & van der Klis 1998) or not cannot be resolved with the currently available data.
## 5 DISCUSSION
We have studied a mechanism for explaining the broad-band X-ray variability properties of accreting compact objects, in which a sharp transition in the accretion disk properties acts as a low-band pass filter with multiple resonances of the density perturbations produced outside of it. Similarly to the model proposed by Stella et al. (1999), all QPO frequencies (which we identify with the various resonances) are determined mostly by the properties of the metric exterior to the compact object. As a result, the mechanism we studied here is applicable to both accreting neutron stars and black holes and can account naturally for the observed correlations between QPO frequencies in systems with widely different properties. Moreover, the same physical mechanism is responsible for the broad-band noise spectra of such systems and is roughly consistent with their observed shapes and break frequencies.
Besides the particular modes we have identified with the observed QPOs in §4 and their harmonics, a number of additional resonances exist that could, in principle, produce detectable QPOs at different frequencies. First, for the lowest order modes, resonances occur also at $`\omega _{2+}`$ and $`\omega _{1+}`$, which have not been detected as QPOs in any galactic source. However, both these frequencies are significantly larger than the Keplerian frequency at the transition radius, which corresponds to the fastest timescale there. It is therefore plausible that the power spectrum of the perturber shows a sharp cut-off at $`\omega _3`$ and hence the power at a resonance with higher frequency is reduced below the detection limits. Moreover, modes of order ($`m`$,0) have resonances at frequencies $`\omega m\mathrm{\Omega }\pm \kappa `$ with small corrections similar to the ones of equation (33) and modes of order $`(0,n)`$ have resonances at frequencies $`\omega n\mathrm{\Omega }_{}`$ with small corrections similar to the ones of equation (38). The amplitudes of the potentially detectable QPOs at these frequencies depend on the strength of the density perturbations at these modes generated in the outer accretion disks, which is unknown a priori. Their predicted frequencies, however, are particular to the mechanism studied here (see discussion in Miller 2000) and their detection will provide additional support to the hypothesis that QPOs occur at fundamental dynamical frequencies in the accretion disks. It is important to stress, however, that all the additional resonances occur at frequencies comparable to or higher than the local Keplerian frequency $`\mathrm{\Omega }`$. Indeed the only low-frequency QPOs in the current picture occur at frequencies comparable to $`\omega _1`$ and its harmonics.
In the model discussed here, the peak separation of the kHz QPOs is not related to the spin frequency of the neutron star. Instead, it is nearly equal to the epicyclic frequency of a Keplerian orbit close to the compact object. This property of the model solves a number of problems related to the beat-frequency interpretation of kHz QPOs, such as their variable peak separation (see, e.g., van der Klis et al. 1997; Méndez et al. 1998; see, however, Miller 2000), and of the nodal-precession interpretation of the HBO, such as the unphysically large required moments of inertia (see, e.g., Stella & Vietri 1998; Psaltis et al. 1999b; Kalogera & Psaltis 2000). Moreover, it does not require all neutron stars that show kHz QPOs to be spinning at very similar frequencies (Stella et al. 1999), a result that has important implications for their spin-up and the possibility of detecting gravitational waves from such systems (see, e.g., Bildsten 1998; Andersson, Kokkotas, & Stergioulas 1999).
The correlation between the HBO and upper kHz QPO frequencies in neutron-star systems depends on their angular momenta and hence allows us to estimate their spin frequencies, assuming a value for the stellar moment of inertia $`I`$ (see, e.g., Psaltis et al. 1999b). Adopting $`I10^{45}M`$ gr cm<sup>3</sup> (see, e.g., Cook, Shapiro, & Teukolsky 1994; also Kalogera & Psaltis 2000), we find that, in all sources showing HBO and kHz QPOs, the neutron stars must be spinning at frequencies $`500`$ Hz. Therefore, the burst oscillations, which are observed at frequencies $`300350`$ Hz, cannot occur at the stellar spin frequencies. Indeed, it would be too coincidental for the stellar spin frequency, as inferred from the burst oscillations, to be as similar to the peak separation of the kHz QPOs as is inferred observationally for some sources (see, e.g., Méndez & van der Klis 1999).
The similarity between the frequencies of the burst oscillations and the peak separation of kHz QPOs in the four sources in which all these phenomena have been observed (see, van der Klis 1998) makes tempting the identification of the burst oscillations with a disk mode that is excited only when the neutron stars show thermonuclear flashes. In fact, this disk mode must occur at a frequency comparable to the maximum epicyclic frequency of a Keplerian orbit near the compact object and may be related to the trapping of modes studied by Kato (1990) and Nowak & Wagoner (1991). (See also the discussion of $`g`$modes in Titarchuk, Lapidus, & Muslimov 1998). This would eliminate the problems that arise from the fact that the frequencies of the burst oscillations are not constant but vary in time (see, e.g., Strohmayer et al. 1996) and that their amplitudes are non-negligible even in the tails of X-ray bursts (see, e.g., Strohmayer et al. 1998). Detection of burst oscillations in the millisecond X-ray pulsar SAX J1808.4$``$3658 (Wijnands & van der Klis 1998; Chakrabarty & Morgan 1998) or of coherent pulsations in any source showing kHz QPOs will be crucial in assessing these possibilities.
We thank Paolo Coppi, Jean-Pierre Lasota, Wlodek Kluźniak, Rashid Sunyaev, Luigi Stella, and especially Shoji Kato for many useful discussions on the dynamics of accretion disks. We are grateful to the organizers of the ITP workshop “Black Holes in Astrophysics”, during which this work was initiated and to the participants of the 1999 Aspen Summer Workshop on X-ray Astronomy for many useful discussions. We thank Deepto Chakrabarty, Fred Lamb, Cole Miller, Luigi Stella, and Michiel van der Klis for reading critically the manuscript. We also thank an anonymous referee for comments and suggestions that improved the paper. D. P. acknowledges the support of a postdoctoral fellowship of the Smithsonian Institution. This research was supported in part by the National Science Foundation under Grant No. PHY94-07194. |
no-problem/0001/astro-ph0001007.html | ar5iv | text | # Wide Field CCD Surface Photometry of the Giant Elliptical Galaxy NGC 4472 in the Virgo Cluster
## 1 INTRODUCTION
NGC 4472 (M49) is a giant elliptical galaxy in the Virgo cluster, located at 4 south of NGC 4486 (M87) at the center of the cluster. NGC 4472 is the brightest member of the Virgo cluster, some 0.2 mag brighter than the cD galaxy M87. NGC 4472 is an outstanding example of giant elliptical galaxies showing a bimodality in the color distribution of globular clusters (Geisler, Lee & Kim 1996; Lee, Kim & Geisler 1998). The bimodal color distribution of the globular clusters in NGC 4472 has shown that there are two kinds of cluster populations in this galaxy: a metal-poor population with a mean metallicity of \[Fe/H\] $`=1.3`$ dex and a more spatially concentrated metal-rich population with a mean metallicity of \[Fe/H\] $`=0.1`$ dex. Interestingly it is found that the metal-rich globular clusters show some properties in common with the galaxy halo stars in their spatial distribution and color profiles, while the metal-poor globular clusters do not show such behavior. This result indicates that there may exist some connection between the metal-rich globular clusters and halo stars in NGC 4472 (Lee et al. 1998).
NGC 4472 is an ideal target to investigate the spatial distribution of stellar light as well as the properties of globular clusters in giant elliptical galaxies, because it is relatively nearby and is the brightest galaxy in the Virgo cluster. Information on the spatial distribution of galaxy light is very useful for understanding the structure and evolution of galaxies, and it provides important constraints for modeling galaxy formation.
To date there have been many surface photometry studies of NGC 4472, as summarized in Table 1. However, there is a large discrepancy in the photometry of the outer region of NGC 4472 among them, the details of which will be shown later. This large difference leads to different conclusions about the properties of NGC 4472. For example, Mihalas & Binney (1981) showed that the surface brightness profile of NGC 4472 given by King (1978) is beautifully fit by a King model with a concentration parameter (c = 2.35), while the surface brightness profile of this galaxy published later by Caon et al. (1994) is much flatter than that of King’s in the outer region. McLaughlin (1999) pointed out this significant difference, and he adopted the data given by Caon et al. for the comparison of the halo stellar light and the globular clusters in NGC 4472. He described that the surface brightness profile of NGC 4472 is similar to the surface number density profile of globular clusters in NGC 4472, which is contrary to the case of M87, as seen in his Fig. 3. If the surface photometry given by King were used instead, this conclusion no longer remains valid. To resolve this discrepancy requires good wide field surface photometry of NGC 4472. Until now, the combined requirements of accurate photometry over a wide field were difficult to meet in a single study, given the photometric limitations of photographic plates and the small size of CCDs.
In this paper we present wide field surface photometry of NGC 4472 based on deep CCD images taken with Washington $`CT_1`$ filters. This paper is organized as follows: In Section 2 observations and data reduction are described. Section 3 presents the results and Section 4 compares the results of this study with those of previous studies. Section 5 discusses the surface brightness profile and the color gradient. Finally the primary results are summarized in Section 6.
## 2 OBSERVATIONS AND DATA REDUCTION
Washington $`CT_1`$ CCD images of NGC 4472 were obtained at the prime focus of KPNO 4m telescope on the photometric night of 1993 February 26, with the primary purpose of studying the globular clusters in NGC 4472. We used Washington $`CT_1`$ filter system which is very efficient for measuring the metallicity of extragalactic globular clusters. The effective central wavelengths and bandpasses of $`C`$ and $`T_1`$ filters are $`\lambda _c=3910\AA `$, $`\mathrm{\Delta }\lambda =1100\AA `$, and $`\lambda _c=6330\AA `$, $`\mathrm{\Delta }\lambda =800\AA `$, respectively (Canterna 1976). The size of the field is $`16^{}.4\times 16^{}.4`$. The pixel scale of the CCD is 0.48 arcsec. We took $`60`$s and $`3\times 1000`$s $`T_1`$ exposures (short and long exposures, respectively, hereafter), and $`5\times 1000`$s C exposures. The seeing was 1.25 arcsec. The images of the central region with $`r<8^{\prime \prime }`$ of NGC 4472 in the $`C`$ images were saturated so that we could not derive the colors for the central region. The details of the observations and transformation of the photometry to the standard system were described in Geisler et al. (1996).
Fig.1 displays a greyscale map of the short $`T_1`$ exposure image of NGC 4472, overlayed with isophotes. Fig. 1 shows that the inner region of NGC 4472 is less elliptical than the outer region. Surface photometry of NGC 4472 was obtained using the ellipse fitting software ELLIPSE in STSDAS/IRAF<sup>1</sup><sup>1</sup>1IRAF is distributed by the National Optical Astronomical Observatories, which is operated by the Association of Universities for Research in Astronomy, Inc., under cooperative agreement with the National Science Foundation., and polygonal aperture photometry software POLYPHOT in APPHOT/IRAF, independently. Comparison of the resulting surface photometries showed an excellent agreement within the errors between the two methods. We adopt the results obtained using ELLIPSE for the final analysis.
We have determined the sky background brightness from the clear region at the north-east corner of the image (the distance along the minor axis from the center of NGC 4472 is $`10^{}.5`$), obtaining $`20.80\pm 0.02`$ mag arcsec<sup>-2</sup> for $`T_1`$ and $`22.23\pm 0.02`$ mag arcsec<sup>-2</sup> for $`C`$, respectively. These results correspond to $`V=21.22`$ mag arcsec<sup>-2</sup> and $`B=22.09`$ mag arcsec<sup>-2</sup>, respectively. The night sky brightness at the zenith measured at the Kitt Peak National Observatory under new moon is known to be $`V=21.91`$ and $`B=22.91`$ mag arcsec<sup>-2</sup> (Pilachowski et al. 1989). Considering that the lunar phase when our data were obtained is 5 days from new moon (leading to a difference in the night sky brightness of $`\mathrm{\Delta }B=0.7`$ and $`\mathrm{\Delta }V=0.2`$ mag arcsec<sup>-2</sup> (Elias 1994)) and that the air mass of our target is 1.1–1.4, our sky estimtes are approximately consistent with the night sky brightness. This result shows that our sky measurement is reasonable for the surface photometry of the galaxy. We have estimated the contribution from the galaxy light at the radius for the sky measurement as follows. Using the de Vacouleurs law, we have fit the surface brightness profile for the inner region of the galaxy where the surface photometry is affected little by the uncertainty in the sky value, and calculated the expected sky brightness at the position of the sky region, obtaining $`T_126.50\pm 0.33`$ mag arcsec<sup>-2</sup>. This value is only 0.5 % of the sky brightness. Therefore any contribution from the galaxy light at the sky position is estimated to be negligible. The flat-fielding is accurate with an error smaller than one percent. So that a sky level measured near one corner of the chip can be applied to the whole area without introducing any significant error for the surface photometry of NGC 4472.
Since we have short and long $`T_1`$ exposure images, we can check the accuracy of our surface photometry by comparison. Fig. 2 displays the comparison of the $`T_1`$ surface brightness between the short and long exposure images. Fig. 2 shows good agreement between the two results: the differences in the surface brightness of the two exposures are on average smaller than 0.03 mag arcsec<sup>-2</sup> over the region with $`r<420^{\prime \prime }`$, but get as large as $`0.1`$ mag arcsec<sup>-2</sup> beyond $`r=420^{\prime \prime }`$. Final data of the surface photometry were prepared by combining the long exposure data for $`r>3^{}`$ and the average of the short and long exposure data for $`r<3^{}`$, after matching the zero points in the brightness of the two. Table 2 lists the final surface photometry of NGC 4472 including the surface brightness, color, ellipticity and position angle as a function of the major axis. Preliminary results of this study were used for the comparison of halo light and globular clusters in NGC 4472 in Lee et al. (1998).
## 3 RESULTS
### 3.1 Surface Brightness Profiles
Radial surface brightness profiles of NGC 4472 are displayed in Fig. 3(a). The $`T_1`$ surface brightness covers a full 8.7 magnitude range. The shapes of the $`C`$ and $`T_1`$ profiles are very similar in general. The shapes of the surface brightness profiles of NGC 4472 are typical for giant elliptical galaxies: flattening in the core region and falling off smoothly with increasing radius.
### 3.2 Surface Color Profiles
Fig. 3(b) displays the radial surface $`(CT_1)`$ color profile of NGC 4472. Here the surface color means the differential color per square-arcsecond. The surface colors for the central region with $`r<8^{\prime \prime }`$ were not obtained, because of the saturation in the $`C`$ image. Fig. 3(b) shows that there exists clearly a radial color gradient in NGC 4472. The surface color gets bluer as the radius increases until $`r3^{}`$. This feature is often seen in other giant elliptical galaxies (Cohen 1986; Kormendy & Djorgovski 1989; Peletier et al. 1990). However, the surface color gets slowly redder with increasing radius for the region with $`3^{}<r<7^{}.3`$, and gets much redder with increasing radius beyond $`r=7^{}.3`$. The outer region suffers more from the photometric errors than the inner region as shown by the error bars in the figure, but the reversal of the color gradient at $`r3^{}`$ appears to be real.
The mean color gradient for $`r<3^{}`$ is measured to be $`\mathrm{\Delta }\mu (CT_1)=0.08`$ mag arcsec<sup>-2</sup> for $`\mathrm{\Delta }\mathrm{log}r=1`$, from the weighted linear least square fitting. We have transformed the $`(CT_1)`$ color gradient into the metallicity gradient using the relation given by Geisler & Forte (1990): \[Fe/H\] = $`2.35(CT_1)4.39`$, assuming the color gradient is entirely due to the metallicity gradient \[Peletier et al. 1990\]. The resulting metallicity gradient of $`\mathrm{\Delta }`$ \[Fe/H\] $`=0.2`$ dex. This value is similar to a mean value of the metallicity gradients derived from the color gradients for giant elliptical galaxies \[Peletier et al. 1990\]. The color profile for $`r<3^{}`$ appears to consist of two linear components breaking at $`30^{\prime \prime }`$, the slopes of which are derived to be $`0.046`$ and $`0.111`$ mag arcsec<sup>-2</sup>, respectively. The color gradient for the outer region with $`3^{}<r<9^{}`$ is measured to be d $`\mathrm{\Delta }\mu (CT_1)=+0.199`$ for $`\mathrm{\Delta }\mathrm{log}r=1`$. If the fit is limited to $`r<7^{}.3`$, the slope will be slightly reduced to 0.181.
### 3.3 Ellipticities and Position Angles
Radial profiles of ellipticity and position angle of NGC 4472 are presented in Figs. 4(a) and 4(b), respectively. Fig. 4 shows several features as follows. First, both $`C`$ and $`T_1`$ profiles of ellipticity and position angle are almost the same over the entire range of radius measured in this study. This is again confirmed in the $`C`$ and $`T_1`$ isophotal contour map of the fitted ellipses in Fig. 5. Fig. 5 shows that there is little difference in the shape and orientation of the $`C`$ and $`T_1`$ isophotal contours of the fitted ellipses. This result shows clearly that the shapes of the isophotes and isochromes of NGC 4472 are almost identical, which is consistent with previous findings based on the data of much smaller area than ours (Cohen 1986; Peletier et al. 1990). Secondly, the ellipticity decreases slightly from 0.1 at $`r=1^{\prime \prime }`$ to 0.07 at $`r4^{\prime \prime }`$, and increases with increasing radius to $`0.18`$ at $`r=20^{\prime \prime }`$. For the region at $`20^{\prime \prime }<r<130^{\prime \prime }`$ the ellipticity remains at $`0.2`$ with small fluctuations. Beyond this region the ellipticity continues to increase to 0.22 at the outer limit. Thus the inner region of NGC 4472 is more circular than the outer region. Thirdly, the position angle decreases rapidly from $`172`$ degrees at $`r=1^{\prime \prime }`$ to 158 degrees at $`r5^{\prime \prime }`$, then decreases slowly to 153 degrees with increasing radius at the outer region, showing that the isophotes of NGC 4472 rotate almost 20 degrees with increasing radius.
We have compared the radial profiles of the ellipticity and position angle for the central region at $`r<15^{\prime \prime }`$ with those based on the HST observations by van den Bosch et al. (1994). The comparison, included in Figure 4, shows that the two results are in good agreement. Rapid changes of the ellipticity and position angle in the inner $`r<2^{\prime \prime }`$ are probably caused by the presence of dust which is visible at $`0^{\prime \prime }.3<r<1^{\prime \prime }.5`$ at position angle 140 \[Jaffe et al. 1994\].
### 3.4 Total Magnitude, Total Color and the Size of NGC 4472
We have calculated the total magnitude and color of NGC 4472 by integrating the radial surface brightness profiles out to the limiting radius, obtaining $`T_1`$(total) = $`7.76\pm 0.02`$ mag and $`(CT_1)`$(total) = $`1.87\pm 0.03`$ mag, respectively. We have transformed these results into $`B`$ magnitude and $`(BV)`$ color using the relations given by Geisler (1996). The transformed magnitude and color ($`B_T=9.26`$ mag, $`(BV)_T=0.92`$) are in excellent agreement with those in de Vaucouleurs et al. (1991), $`B_T=9.25`$ mag, and $`(BV)_T=0.95`$. However, our magnitude is $`0.4`$ mag fainter than that given by Caon et al. (1994). Lee et al. (1998) used the total magnitude of NGC 4472 given by de Vaucouleurs et al. (1991) which agrees well with ours, for calculating the specific frequency of the globular clusters in NGC 4472. Therefore the results on the specific frequency of the globular clusters in NGC 4472 given by Lee et al. (1998) remain valid.
The integrated $`(CT_1)`$ color (within given radius) gets bluer with increasing radius out to $`r3^{}`$. Beyond this radius, the integrated color remains at $`(CT_1)1.87`$. The surface color gets slowly redder outward for $`r>3^{}`$, but the surface brightness is so low in the outer region that the integrated color change little with increasing radius.
We have also derived a standard radius of NGC 4472, which is defined as the radius where the surface brightness in $`B`$ band is $`25`$ mag arcsec<sup>-2</sup>: $`r_{25}313^{\prime \prime }`$, which corresponds to a linear size of 51.7 kpc for the adopted distance of 17.4 Mpc \[Lee, Kim & Geisler 1998\]. This value is a little smaller than the result of Caon et al. (1990), $`350^{\prime \prime }`$. In Table 3 we have summarized the photometric parameters of NGC 4472 derived in this study.
## 4 Comparison with Previous Studies
We have compared our results with those given by previous studies which are listed in Table 1. While previous photographic studies cover a wide field of NGC 4472, previous CCD studies are limited to a small field except for the study by Cohen (1986) which covers up to $`r400^{\prime \prime }`$. Previous surface photometry of NGC 4472 was presented in various photometric systems. For the comparison we transformed approximately the results of surface photometry in different filters into the $`BVR`$ system, using the conversion relations given in the literature which are listed in Table 4.
### 4.1 Surface Brightness Profiles
Fig. 6 displays the comparison of the surface brightness profiles between this study and others, plotting the differences between our value and theirs. The surface brightness profiles of this study and others agree roughly for the inner region at $`r<100^{\prime \prime }`$, except for the $`B`$ profile given by Boroson & Thompson (1987) which is about 0.4 mag brighter than the others. In particular, the profiles of this study and Peletier et al. (1990) show an excellent agreement for $`r<100^{\prime \prime }`$. However, the differences among these results become significant for the outer region at $`r>100^{\prime \prime }`$. For $`r>100^{\prime \prime }`$ the surface brightness profile given by Caon et al. becomes fainter than ours, while those of the others become brighter than ours. Note the two photographic results by King (1978) and Caon et al. (1994) show opposite trends to each other for the outer region, and that our photometry stays between the two. The large differences in the surface brightness profiles for the outer region are probably due to the difficulty in estimating sky values in images, as well as the larger photometric errors from the diminishing galaxy light.
### 4.2 Surface Color Profiles
Fig. 7 displays the comparison of the surface color profiles of NGC 4472 obtained in this study and others. It is seen that for the inner region at $`r<100^{\prime \prime }`$ $`(UR)`$, $`(CT_1)`$, $`(VI)`$, $`(BR)`$, and $`(vg)`$ colors show obviously negative radial gradients, while $`(ri)`$ and $`(gr)`$ show positive radial gradients. For the very outer region, $`(gr)`$ and $`(gi)`$ colors change much more significantly than does $`(CT_1)`$. This appears to be due to large errors in the $`(gr)`$ and $`(gi)`$ colors, as seen in Fig. 6.
## 5 Discussion
### 5.1 The Shape of the Surface Brightness Profiles
In Fig. 8 we have fit the surface brightness profiles of NGC 4472 using King models and de Vaucouleurs law. Fig. 8 displays the $`T_1`$ surface brightness profile of NGC 4472 obtained in this study and the surface brightness of the central region at $`r<15^{\prime \prime }`$ based on the HST observation by Ferrarese et al. (1994). We have transformed the $`V`$ surface photometry of the central region into the $`T_1`$ magnitude using the conversion relation given by Geisler (1996). Here the surface brightness profile is plotted against the geometric mean of the major and minor radii ($`r=\sqrt{(r_{\mathrm{major}}r_{\mathrm{minor}})}`$). Fig. 8 shows that both photometry sets are in excellent agreement.
In Fig. 8(a) it is found that the entire surface brightness profile cannot be fit by any single King model, but it can be fit approximately by two King models. The outer region at $`r>10^{\prime \prime }`$ is approximately fit by a King model with a concentration parameter $`c=(\mathrm{log}r_t/r_c)`$ = 2.35 and $`r_c=5^{\prime \prime }`$ (= 420 pc), while the inner region at $`r<7^{\prime \prime }`$ is fit well by a King model with $`c=2.50`$ and $`r_c=4^{\prime \prime }`$ (= 340 pc) (where $`r_t`$ and $`r_c`$ represent, respectively, tidal radius and core radius). Ferrarese et al. (1994) also pointed out that the surface brightness profile of the central region ($`r<15^{\prime \prime }`$) could not be fit by any single King model.
The surface brightness profile of the outer region $`r>7^{\prime \prime }`$ of NGC 4472 is also fit approximately by a deVaucouleurs law, as shown in Figs. 8(a) and 8(b), while that of the inner region is far from being fit by a deVaucouleurs law. The solid line in Fig. 8 represents a fit to the data for $`7^{\prime \prime }<r<260^{\prime \prime }`$, $`\mu (T_1)=2.52(\pm 0.07)r^{1/4}+13.09`$ with $`\sigma =0.08`$. The corresponding effective radius is derived to be $`r_e=120\pm 2`$ arcsec, (= 10.1 kpc).
The evidence found in this study clearly indicates that there is a distinct component in the central region of NGC 4472: the presence of two components in the surface brightness profile, the reversal of the ellipticity profile, and the rapid change in the profile of the position angle in the inner region at $`r<4^{\prime \prime }`$. Interestingly it is known that NGC 4472 has a kinematically decoupled core and shows enhanced Mg<sub>2</sub> in the inner $`5^{\prime \prime }`$ (Davis & Birkinshaw 1988; Ferrarese et al. 1989; Davies, Sadler & Peletier 1993). Therefore it is suspected that the central component may be related with the kinematically decoupled core, although van den Bosch et al. (1994) pointed out that the isophotal profiles show no evidence for a photometrically distinct nucleus.
### 5.2 Color Gradient
It is found that there is clearly a negative color gradient for $`r<3^{}`$ (= 15 kpc) in NGC 4472 in the sense that the color gets bluer with increasing radius: $`\mathrm{\Delta }\mu (CT_1)=0.08`$ mag arcsec<sup>-2</sup> for $`\mathrm{\Delta }\mathrm{log}r=1`$. In general, color gradients in giant elliptical galaxies are interpreted as evidence for a metallicity gradient, while color gradients in dwarf galaxies may be due to age effects (Vader et al. 1988; Peletier et al. 1990). The color gradient in the inner region of NGC 4472 corresponds to a metallicity gradient of $`\mathrm{\Delta }`$\[Fe/H\] $`=0.2`$ dex. This result is similar to that derived from the line-strength gradient for $`3^{\prime \prime }<r<50^{\prime \prime }`$ of NGC 4472 by Davies et al. (1993). This value for NGC 4472 is similar to the mean value known for giant elliptical galaxies, $`\mathrm{\Delta }`$\[Fe/H\]$`=0.2\pm 0.1`$ dex (Kormendy & Djorgovski 1989; Peletier et al. 1990; Davies, Sadler & Peletier 1993).
The metallicity gradient in NGC 4472 may result from dissipational collapse. However, the metallicity gradients predicted by conventional models of galaxy formation based on dissipational collapse (Larson 1975; Carlberg 1984) are much steeper than the value observed in NGC 4472. Considering that the metallicity gradients can be diluted by a factor of 2 over three merger events \[White 1980\], Davies et al. (1993) pointed out that the shallow metallicity gradients support the hypothesis that giant elliptical galaxies form by stellar mergers, and that the line-strength (metallicity) gradient may be due to their progenitors which formed predominantly by dissipational collapse. Some of the observed properties of the globular clusters in NGC 4472 also support the merger hypothesis for the formation of NGC 4472, while some do not. This point was discussed in detail by Lee et al. (1998).
On the other hand, it is difficult to understand in terms of galaxy formation that the color profile of NGC 4472 shows a positive gradient in the outer region at $`r>3^{}`$. This trend was also shown by the $`(gr)`$ color profile given by Cohen (1986), but this $`(gr)`$ color profile showed a positive color gradient even for the inner region which is contrary to others, as seen in Fig. 7. Other deep photometry of the outer region of NGC 4472 and other giant elliptical galaxies are needed to investigate this feature further.
## 6 SUMMARY AND CONCLUSIONS
We have presented surface photometry in Washington $`C`$ and $`T_1`$ filters for a wide field centered on the brightest elliptical galaxy in Virgo, NGC 4472. Our data cover a wider field than those of any previous CCD surface photometry, and our photometry goes deeper than any previous photometry. Primary results obtained in this study are summarized as follows:
(1) Surface brightness profiles for the outer region of NGC 4472 obtained in this study lie between those found in the wide field photographic studies of King (1978) and Caon et al. (1994). The surface brightness profiles of NGC 4472 are not fit well by a single King model, but they can be fit approximately by two King models: one for the inner region and the other for the outer region. Surface brightness profiles for the outer region can be fit approximately also by a deVaucouleurs law.
(2) There is obviously a negative color gradient for the region at $`r<3^{}`$ of NGC 4472 in the sense that the colors get bluer with increasing radius.. The slope of the color gradient for this region is derived to be $`\mathrm{\Delta }\mu (CT_1)=0.08`$ mag arcsec<sup>-2</sup> for $`\mathrm{\Delta }\mathrm{log}r=1`$, which corresponds to a metallicity gradient of $`\mathrm{\Delta }`$\[Fe/H\] $`=0.2`$ dex. However, the surface color appears to get redder slowly with increasing radius for the region with $`r>3^{}`$.
(3) A comparison of the structural parameters of NGC 4472 in $`C`$ and $`T_1`$ images has shown that there is little difference in the ellipse shapes between isochromes and isophotes.
(4) Photometric and structural parameters of NGC 4472 have been determined, which are listed in Table 3.
## Acknowledgments
This research is supported by the Ministry of Education, Basic Science Research Institute grant No.BSRI-98-5411 (to M.G.L.). |
no-problem/0001/hep-th0001140.html | ar5iv | text | # I Introduction
## I Introduction
Nonsingular spacetimes and those containing black holes are qualitatively quite different. Nevertheless, it is possible to find sequences of spacetimes that, while remaining nonsingular, come arbitrarily close to having horizons . In a previous paper we studied a class of such solutions that are associated with self-gravitating monopoles in a spontaneously broken Yang-Mills theory. The emphasis there was on the detailed behavior of the fields as one approaches the critical solution in which a horizon first appeared. In this paper, we concentrate instead on the geometrical aspects of the spacetimes associated with these objects near criticality, and on using these to gain insights into the properties of true black holes.
As in Ref. , we restrict ourselves to spherically symmetric spacetimes and write the metric in the form
$$ds^2=Bdt^2Adr^2r^2(d\theta ^2+\mathrm{sin}^2\theta d\varphi ^2).$$
(1)
A horizon corresponds to a zero of $`1/A`$; the horizon is extremal if $`d(1/A)/dr`$ also vanishes. We work in the context of an SU(2) gauge theory with gauge coupling $`e`$ and a triplet Higgs field $`\varphi `$ whose vacuum expectation value $`v`$ breaks the symmetry down to U(1). In flat spacetime this theory possesses a finite energy monopole solution with magnetic charge $`Q_M=4\pi /e`$ and mass $`Mv/e`$. It has a core region, of radius $`1/ev`$, with nontrivial Higgs and massive vector boson fields. Beyond this core is a Coulomb region in which all massive fields approach their vacuum values exponentially rapidly, leaving only the Coulomb magnetic field. The effects of adding gravitational interactions depend on the value of $`v`$. For $`v`$ much less than the Planck mass $`M_{\mathrm{Pl}}`$, one finds that $`1/A`$ is equal to unity at the origin, decreases to a minimum at a radius of order $`1/ev`$, and then increases again with $`A(\mathrm{})=1`$. As $`v`$ is increased, this minimum becomes deeper, until an extremal horizon develops at a critical value $`v_{\mathrm{cr}}`$ of the order $`M_{\mathrm{Pl}}`$. As we describe in more detail in Sec. II, two distinct types of critical behavior are possible, depending on the ratio of the Higgs and gauge boson masses. For lower values of this ratio, one finds “Coulomb-type” critical solutions, in which the horizon occurs in the Coulomb region of the monopole at $`r_0=\sqrt{4\pi G/e^2}`$. Outside the horizon, the metric is that of an extremal Reissner-Nordstrom black hole, with
$$B(r)=\frac{1}{A}=1\frac{2MG}{r}+\frac{Q^2G}{4\pi r^2},$$
(2)
while the massive fields take on their vacuum values. As the Higgs self-coupling increases, there is a transition to “core-type” critical solutions that have a horizon inside the monopole core and nontrivial matter fields (or hair) outside the horizon.
In both types of critical monopole solutions the fields remain nonsingular at $`r=0`$. However, it is also possible to have solutions with singularities at $`r=0`$ that can be viewed as self-gravitating monopoles with Schwarzschild black holes at their center. As long as the mass of the central black hole is not too great, the variation of these solutions with $`v`$ is quite similar to that of the nonsingular monopoles, and one finds the same two types of critical behavior .
After reviewing these solutions, we discuss in Sec. III how near-critical monopoles might appear to an outside observer. One would expect the measurements made by such an observer to vary continuously with the parameters of the monopole and to show no discontinuity at the critical solution. The external observer could probe the monopole with either particles or waves. In the case of the particle, we find that the time needed for the particle to emerge from the interior (as measured by a static external observer using “Schwarzschild time”) diverges as the critical solution is approached. When a wave is sent in, there is a reflected wave due to the gravitational field just outside the horizon and a transmitted wave that passes through the interior and then emerges with a time delay. As before, the time delay diverges as $`vv_{\mathrm{cr}}`$, while the reflected wave becomes indistinguishable from that due to a black hole. Using either type of probe, an observer whose lifetime is finite cannot distinguish between a true black hole and a nonsingular, subcritical solution that is sufficiently close to being critical. We discuss the implications for our understanding of black hole entropy.
We also find that the near-critical Coulomb-type solutions display what Horowitz and Ross have termed naked-black-hole behavior, even though there is no black hole at all. This is characterized by the fact that a freely-falling observer passing through the minimum of $`1/A`$ (we shall refer to the location of this minimum as the quasi-horizon) feels a tidal force that diverges as the critical solution is approached. For core-type solutions, on the other hand, no such behavior is observed.
In Sec. IV, we consider the effect of having additional matter fall into a near-critical solution, addressing in particular the question of whether this process could produce an extremal black hole. Extremal black holes are especially interesting from the standpoint of black hole thermodynamics because they have vanishing Hawking temperature. The analogies between black hole dynamics and thermodynamics thus suggest that they should be rather difficult, if not impossible, to create. Indeed, one of the formulations of the third law of black hole dynamics asserts the impossibility (under certain technical assumptions) of making a nonextremal black hole extremal. One could also envision producing an extremal black hole starting from a nonsingular spacetime. Boulware showed that this can be done by the collapse of a charged shell of matter. However, this mechanism relies critically on the shell being infinitely thin; shells of finite thickness and density do not collapse to an extremal configuration.
It is easy to understand the difficulty of making an extremal black hole if one recalls that the extremal Reissner-Nordstrom black hole is characterized by having a mass and a charge that (in appropriately rescaled Planck units) are equal. Forming such an object by the collapse of a shell with equal charge and mass densities involves a delicate balance between electromagnetic and gravitational forces. One could instead try to achieve extremality by adding matter to a pre-existing nonextremal Reissner-Nordstrom black hole (i.e. one with greater mass than charge). However, because the added matter would have to have more charge than mass, the Coulomb repulsion between the black hole and the infalling matter would tend to overcome their gravitational attraction.
The situation is rather different in our case, because the nonsingular monopole solutions are overcharged; i.e., their long range fields are those of a Reissner-Nordstrom solution with greater charge than mass.In the pure Reissner-Nordstrom case, this leads to a naked singularity. The singularity is avoided here by the same mechanism that makes the mass of the flat-space monopole finite: the orientation of the massive gauge fields in the monopole core is such that their magnetic dipole energy just cancels the singular Coulomb energy at the origin. Allowing uncharged matter to fall into these objects increases their mass and should bring them closer to criticality. If the amount of matter entering is just sufficient to create a zero of $`1/A`$, one would expect an extremal solution to result. We will present numerical arguments that support this expectation. Finally, we make some concluding remarks in Sec. V.
## II Review of previous results
As in Ref. , we consider an SU(2) gauge theory that is spontaneously broken to U(1) by the vacuum expectation value $`v`$ of a triplet Higgs field $`\varphi `$. This theory has magnetic monopole solutions that can be described by the spherically symmetric ansatz
$`\varphi ^a`$ $`=`$ $`v\widehat{r}^ah(r)`$ (3)
$`A_{ia}`$ $`=`$ $`ϵ_{iak}\widehat{r}^k{\displaystyle \frac{1u(r)}{er}}.`$ (4)
Finiteness of the energy requires that $`u(\mathrm{})=0`$ and $`h(\mathrm{})=1`$. If the solutions are also required to be nonsingular at $`r=0`$, then $`u(0)=1`$ and $`h(0)=0`$.
In a spacetime with a metric of the form of Eq. (1), the static field equations for these matter fields can be derived from a $`(1+1)`$-dimensional action of the following form
$$S_{\mathrm{matter}}=4\pi 𝑑t𝑑rr^2\sqrt{AB}\left[\frac{K(u,h)}{A}+U(u,h)\right]$$
(5)
where $`U(u,h)`$, involves the fields but not their derivatives, while
$$K=\frac{1}{e^2r^2}\left(\frac{du}{dr}\right)^2+\frac{v^2}{2}\left(\frac{dh}{dr}\right)^2.$$
(6)
The Euler-Lagrange equations for the matter fields that follow from Eq. (5) must be supplemented by the gravitational field equations. For static, spherically symmetric field configurations these reduce to
$$G_{\widehat{t}\widehat{t}}=\frac{1}{2r^2}\frac{d}{dr}\left[r\left(\frac{1}{A}1\right)\right]=4\pi G\left(\frac{K}{A}+U\right)$$
(7)
and
$$G_{\widehat{t}\widehat{t}}+G_{\widehat{r}\widehat{r}}=\left(\frac{2}{rA}\right)\frac{1}{\sqrt{AB}}\frac{d\sqrt{AB}}{dr}=\frac{16\pi GK}{A}.$$
(8)
Here carets indicate orthonormal components.
Equation (8) can be immediately integrated to obtain
$$B(r)=\frac{1}{A(r)}\mathrm{exp}\left[16\pi G_r^{\mathrm{}}𝑑r^{}r^{}K\right].$$
(9)
Using this to eliminate $`B`$, one is left with two second order and one first order equation for the functions $`u`$, $`h`$, and $`A`$. These equations must be solved numerically. Up to a rescaling of distances, the solutions of these equations depend only on the two dimensionless parameters $`a=8\pi Gv^2`$ and $`b=(m_H/2m_W)^2`$.
For small values of $`b`$ (roughly $`b25`$ for a quartic Higgs) one finds Coulomb-type solutions in which the minimum of $`1/A`$ is located outside the monopole core. This minimum decreases<sup>§</sup><sup>§</sup>§This behavior is modified slightly for very small $`b`$. For a detailed description, see . with increasing $`a`$, until the critical solution is reached at $`a_{\mathrm{cr}}=8\pi Gv_{\mathrm{cr}}^2`$. In the critical solution, the matter fields $`u`$ and $`h`$ reach their asymptotic values $`u=0`$ and $`h=1`$ at the horizon and are then constant for all $`r>r_0`$; because both fields fall as fractional powers of $`r_0r`$, the derivatives $`du/dr`$ and $`dh/dr`$ both diverge as $`r`$ approaches $`r_0`$ from below. (This nonanalytic behavior is possible because an extremal horizon is a singular point of the matter field equations.)
The metric of the critical solution is identical to the extremal Reissner-Nordstrom metric outside the horizon, but differs from it for $`r<r_0`$. The metric function $`1/A`$ varies relatively smoothly, falling monotonically from unity at the origin to a zero at the horizon. Just inside the horizon $`1/Ak(r_0r)^2`$, with $`k`$ being larger than for the corresponding Reissner-Nordstrom solution. The behavior of $`B`$ contrasts sharply with this. Equation (9) shows that the product $`AB`$ (which is identically equal to unity in both the Schwarzschild and Reissner-Nordstrom solutions) is given by an integral of the functional $`K(u,h)`$. The singularities in the derivatives of $`u`$ and $`h`$ at the horizon are strong enough to cause this integral to diverge, so that the ratio
$$c\frac{\sqrt{AB}|_{\mathrm{outside}r_0}}{\sqrt{AB}|_{\mathrm{inside}r_0}}$$
(10)
is infinite in the critical limit. If we adopt the conventional normalization $`B(\mathrm{})=1`$, then $`B`$ vanishes identically inside the horizon. If we instead set $`B(0)=1`$, then $`B`$ is finite and varying inside the horizon and infinite for $`r>r_0`$; depending on the value of $`b`$, the minimum of $`B`$ may be at $`r=0`$ or at some finite radius, but in neither case does $`B`$ have a zero. For near-critical solutions where the minimum value $`(1/A)_{\mathrm{min}}ϵ`$ is small but nonzero, we find that the ratio $`c`$ varies as $`ϵ^q`$, where $`q`$ ranges from about 0.7 to unity.
A rather different type of critical solution is found for larger $`b`$. For these core-type solutions the horizon occurs at a radius $`r_{}<r_0`$, with the values $`u_{}`$ and $`h_{}`$ of the matter fields at this point being different than their asymptotic values. Although the solutions are still nonanalytic at the horizon, this nonanalyticity occurs only in subdominant terms. Thus, $`1/A`$ again vanishes as $`(rr_0)^2`$ as one approaches the horizon, but the coefficient is now the same inside and outside the horizon.For intermediate values of $`b`$, subcritical monopoles exhibit both core-type and Coulomb-type quasi-horizons. However, as one approaches criticality for a given value of $`b`$, only one quasi-horizon actually becomes a horizon. The other quasi-horizon, though interesting, is essentially irrelevant for our purposes. The radial derivatives of both matter fields are finite at the horizon, so $`K`$ remains finite and there is no sharp change in $`AB`$ at the horizon. Because $`AB`$ remains finite and nonzero, $`B`$ has a zero at the horizon that coincides with the zero of $`1/A`$.
These solutions can be generalized to include a black hole in the center of the monopole. Instead of requiring that the fields be nonsingular at $`r=0`$, one instead requires that there be a zero of $`1/A`$ at a nonzero radius $`r_\mathrm{H}`$. At this zero, the equations for the matter fields become constraint equations relating the fields and their first derivative; solving these constraints yields enough boundary conditions to determine a solution.
If $`r_\mathrm{H}`$ is not too large, the effect of increasing $`a`$ is similar to what it is in the absence of a central black hole. For larger values of $`r_H`$, see the discussion in . There is an outer minimum of $`1/A`$ that moves downward, finally reaching zero and becoming an extremal horizon at some critical value $`a_{\mathrm{cr}}(r_\mathrm{H})`$. For small values of $`b`$ the solutions are Coulomb-type, while for large $`b`$ one finds core-type critical solutions.
Rather than increasing $`a`$ with $`r_\mathrm{H}`$ fixed, one can instead increase $`r_\mathrm{H}`$ with $`a`$ held fixed; this is much more analogous to the process of actually dropping matter into a near-critical solution that we will consider in Sec. IV. For initial values of $`a`$ that are sufficiently close to $`a_{\mathrm{cr}}(r_\mathrm{H}=0)`$, this gives a family of solutions with a critical limit. In Figs. 1 and 2 we illustrate this with Coulomb-type solutions with $`b=1.0`$ and core-type solution with $`b=10^6`$.
## III Probing the almost black hole
For any $`a<a_{\mathrm{cr}}`$, the self-gravitating monopole solution is a nonsingular spacetime with a Penrose diagram of the same form as that of Minkowski spacetime (Fig. 3a). The critical solution, on the other hand, can be extended beyond the original coordinate patch to yield a spacetime with the Penrose diagram shown in Fig. 3b. This diagram is quite similar to that of an extremal Reissner-Nordstrom black hole, but differs from it by not having a singularity<sup>\**</sup><sup>\**</sup>\**Although there is no singularity at the origin for a critical monopole black hole, there are singularities at the extremal horizon resulting from nonanalytic behavior of the monopole fields. These singularities are relatively mild in the core-type case, but are more dramatic in the Coulomb-type case. at $`r=0`$. The difference between the two diagrams is striking and seems to indicate a discontinuity at $`a=a_{\mathrm{cr}}`$, in contradiction with the usual expectation that physical quantities should vary continuously with the parameters of a theory.
However, this discontinuity can been seen as an artifact of the conformal transformation that produces the Penrose diagram from an infinite spacetime. This can be illustrated by considering the points A and B that lie on the curves of constant $`r`$ that are shown in Figs. 3a and 3b. These have been chosen so that it is possible for an object to start at A, move in to $`r=0`$ at C, and then travel out again to B. The total proper time along this worldline (or the total affine parameter, if the worldline is lightlike) is finite. This should be compared with the proper time along the worldline of constant $`r`$. This is finite for the subcritical case, whereas in the critical case the proper times along the segments AD and DB are both infinite, corresponding to the fact that an observer in the exterior region containing A cannot detect objects behind the horizon PD. In order to obtain consistency with our physical expectations of continuity, we should require that the proper time along the world of constant $`r`$ should diverge as $`aa_{\mathrm{cr}}`$. More generally, the time required for an external observer to receive information from a probe of the interior regions solution should diverge in the critical limit.
### A Particle and Wave Probes
To see how this works out, we consider probing the interior region (i.e., the region $`r<r_{}`$, where $`r_{}`$ is the quasi-horizon) of a near-critical solution by sending in either a particle or a wave. In both cases, we assume that the probe interacts only gravitationally, and has no direct interaction with the monopole fields.
To begin, imagine releasing a massive particle from an initial radius $`r_1r_{}`$ that is large enough that we may approximate $`B(r_1)1`$. The rotational and time-translation symmetries of the metric allow us to take the motion to lie in the $`\theta =\pi /2`$ plane and guarantee the conservation of the angular momentum per unit mass
$$J=r^2\frac{d\varphi }{d\tau }$$
(11)
and the energy per unit mass
$$E=B(r)\frac{dt}{d\tau }.$$
(12)
These, together with Eq. (1), imply that
$$\frac{dr}{d\tau }=\frac{1}{\sqrt{AB}}\left[E^2B\left(\frac{J^2}{r^2}+1\right)\right]^{1/2}$$
(13)
If $`J=0`$, the particle falls radially in, pass through the origin, and emerge on the other side of the monopole. If instead $`J0`$, the particle turns around after reaching a minimum radius $`r_{\mathrm{min}}(J)`$ and return to $`r_1`$ with its trajectory advanced by an angle
$$\mathrm{\Delta }\varphi =2_{r_{\mathrm{min}}}^{r_1}𝑑r\frac{d\varphi /d\tau }{dr/d\tau }=2_{r_{\mathrm{min}}}^{r_1}𝑑r\frac{J}{r^2}\sqrt{AB}\left[1\frac{B}{E^2}\left(\frac{J^2}{r^2}+1\right)\right]^{1/2}.$$
(14)
We are interested in the proper time measured by an observer who remains at $`r=r_1`$. Assuming that the observer is not moving at relativistic speeds, this is approximately the same as the Schwarzschild coordinate time $`t`$, which over the course of the entire trajectory increases by an amount
$$\mathrm{\Delta }t=2_{r_{\mathrm{min}}}^{r_1}𝑑r\frac{dt/d\tau }{dr/d\tau }=2_{r_{\mathrm{min}}}^{r_1}𝑑r\frac{A}{\sqrt{AB}}\left[1\frac{B}{E^2}\left(\frac{J^2}{r^2}+1\right)\right]^{1/2}.$$
(15)
There are two potential sources of divergences in this integral as $`ϵ=(1/A)_{\mathrm{min}}0`$ and the critical solution is approached. In both types of critical solutions there is a contribution from $`rr_{}`$ associated with the growth of $`A(r_{})`$. In the Coulomb-type solutions the near-vanishing of $`\sqrt{AB}`$ \[see Eq. (9)\] gives a second contribution from the entire region $`r<r_{}`$. Let us examine these in more detail.
For core-type solutions, in the region $`rr_{}`$ we can write
$$Ak_1\left[\left(\frac{rr_{}}{r_{}}\right)^2+ϵ\right]^1$$
(16)
with $`k_1`$ of order unity, while $`\sqrt{AB}`$ is roughly constant and independent of $`ϵ`$. Because $`B(r_{})`$ is small, the $`J`$-dependent term in $`\mathrm{\Delta }t`$ can be neglected for any $`J`$ such that the particle could have reached $`r_{}`$. By a similar argument, we see that any particle that reaches $`r_{}`$ goes through the peak of $`A`$ before turning around. Hence,
$$\mathrm{\Delta }t\frac{2k_1\pi r_{}}{\sqrt{AB}|_{r=r_{}}}ϵ^{1/2}+\mathrm{}$$
(17)
where the ellipses represent subdominant terms.
For Coulomb solutions the dominant effect is due to the fact that $`\sqrt{AB}ϵ^q`$ is almost vanishing throughout the interior region. Because our numerical solutions show that $`q`$ ranges between 0.7 and unity, the divergence due to this effect is greater than that from the region near $`r_{}`$. Furthermore, the near-vanishing of $`B`$ in the interior implies that any particle that enters the interior almost reaches the origin, so that $`r_{\mathrm{min}}0`$. Thus,
$$\mathrm{\Delta }tk_2r_{}ϵ^q+\mathrm{}$$
(18)
where $`k_2`$ is of order unity.
Rather than sending in a particle, one can also probe the near-black hole by sending in a wave packet. As an example, let us consider a free massive scalar field $`\varphi `$, whose field equation in curved spacetime takes the form
$$0=\frac{1}{\sqrt{g}}_\mu \left[\sqrt{g}g^{\mu \nu }^\nu \varphi \right]+m^2\varphi .$$
(19)
This can be put into a more tractable form by writing
$$\psi =r\varphi $$
(20)
and defining a new coordinate $`y(r)`$ satisfying
$$\frac{dr}{dy}=\frac{\sqrt{AB}}{A}.$$
(21)
Equation (19) then takes the form of a one-dimensional wave equation
$$0=\frac{^2\psi }{t^2}\frac{^2\psi }{y^2}+\left[U(r)+m^2B\right]\psi $$
(22)
with a scattering potential
$$U(r)=\frac{1}{2r}\frac{d}{dr}\left[\frac{AB}{A^2}\right]+\frac{J(J+1)B}{r^2}.$$
(23)
When a wave packet incident from large $`r`$ reaches the region near the quasi-horizon, a portion is reflected by the scattering potential, while a portion is transmitted and emerge with some time delay. If a near-critical solution is to appear essentially indistinguishable from a black hole to an outside observer, two conditions must hold. First, the reflection coefficient as a function of wave number must approach that of the black hole as $`ϵ0`$. Second, the time delay in the emergence of the transmitted wave should diverge in the critical limit.
To see how the first of these conditions comes about, let us use Eq. (8) to rewrite the scattering potential as
$$U(r)=\frac{AB}{rA}\left[\frac{8\pi GK}{A}\frac{d}{dr}\left(\frac{1}{A}\right)+\frac{J(J+1)}{r}\right].$$
(24)
For both core- and Coulomb-type solutions the quantity $`K/A`$ remains finite in the critical limit, while the second term in the brackets is zero at the quasi-horizon. Since $`AB1`$, it is clear that $`U(r_{})`$ vanishes at least as fast as $`ϵ`$ as the critical limit is approached. Hence, in the limit the scattering potential splits into two parts, one inside and one outside the horizon. The outer potential is equal either to that of an extremal Reissner-Nordstrom black hole (in the Coulomb case) or that of a black hole with hair (in the core case). Because the variation of the outer potential with $`ϵ`$ is smooth in both cases, our conditions on the reflection coefficients are satisfied if we can ignore reflection from the inner part of the potential.
This can be understood by noting that the natural distance variable in which to discuss the motion of the waves is $`y`$. By integrating Eq. (21) inward from some reference point $`r_1r_{}`$, we obtain
$$y(r)=y(r_1)_r^{r_1}𝑑r\frac{A}{\sqrt{AB}}.$$
(25)
The behavior of this integral as the critical limit is approached is very similar to that of the integral in the expression for $`\mathrm{\Delta }t`$, Eq. (15). For either type of solution, the region near the quasi-horizon gives a contribution that diverges at least as fast as $`ϵ^{1/2}`$. There is a corresponding growth in both the effective distance from the inner portion of the potential to any external point and in the time delay of the corresponding reflected wave. As $`ϵ`$ is increased, an external observer at fixed $`r`$ first sees the reflections from the inner and outer parts of the potential split into two distinct reflected waves, and then finds that the time delay of the second reflected wave (from the inner potential) grows without bound.
The portion of the wave that is transmitted through the region near the quasi-horizon either continues through the origin and then outward or reflects off a central centripetal barrier, according to whether or not $`J`$ vanishes. In either case, the time delay accumulated by this wave before it returns to the quasi-horizon grows in essentially the same manner as the travel time for a massive particle traversing the same path: as $`ϵ^{1/2}`$ for a core-solution and as $`ϵ^q`$ for a Coulomb-type solution.
### B Information and Entropy
Thus, regardless of the type of probe used, an external observer at fixed $`r_{\mathrm{obs}}`$ must wait for at least a time $`\mathrm{\Delta }tO(ϵ^{1/2})`$ before the probe emerges from the region inside the quasi-horizon. To leading order, this time delay is independent of the energy or angular momentum of the probe, and is instead determined solely by the spacetime geometry. Hence, to an observer with a finite lifetime $`T`$, the interior region of any near-critical configuration with $`ϵT^2`$ is inaccessible.<sup>††</sup><sup>††</sup>††Note that once $`ϵ`$ is less than $`T^2`$, the the boundary of the inaccessible region depends only very weakly on $`T`$ and $`r_{\mathrm{obs}}`$, and is essentially indistinguishable from the quasi-horizon. He would most naturally describe any larger system containing this configuration by a density matrix $`\rho `$ obtained by tracing over the degrees of freedom inside the quasi-horizon. From this density matrix one can derive an entropy $`S_{\mathrm{interior}}=\mathrm{Tr}\rho \mathrm{ln}\rho `$ that can be associated with the interior of the quasi-black hole.
One could, of course, proceed in this manner to define an entropy for any arbitrary region in space, just as one can choose to make the information in any subsystem inaccessible by putting the subsystem behind a locked door. The crucial difference here is that the inaccessibility is due to the intrinsic properties of the spacetime, and that the boundary of the inaccessible region is defined by the system itself rather than by some arbitrary external choice. It is thus reasonable to define $`S_{\mathrm{interior}}`$ as the entropy of the quasi-black hole.
A precise calculation of this entropy is clearly infeasible. Among other problems, such a calculation would require a correct implementation of an ultraviolet cutoff, which presumably would require a detailed understanding of how to perform the calculation in the context of a consistent theory of quantum gravity. As an initial effort, one can take the ultraviolet cutoff as the Planck mass $`M_{\mathrm{Pl}}`$ and ask for an order of magnitude calculation. Such a calculation was done by Srednicki , who showed that tracing over the degrees of freedom of a scalar field inside a region of flat spacetime with surface area $`A`$ led to an entropy $`S=\kappa M^2A`$ where $`M`$ is the ultraviolet cutoff and $`\kappa `$ is a numerical constant. Furthermore, although the precise calculations depend on the details of the theory, Srednicki gave general arguments suggesting that such an entropy should always be proportional to the surface area. This leads us to expect that $`S_{\mathrm{interior}}M_{\mathrm{Pl}}^2A`$.
This result is, of course, consistent with the possibility that in the critical limit $`S_{\mathrm{interior}}`$ goes precisely to the standard black hole result $`S_{\mathrm{BH}}=M_{\mathrm{Pl}}^2A/4`$. However, in contrast with the usual black hole case, our spacetime configurations are topologically trivial. The “interior” region enclosed by the quasi-horizon is nonsingular and static. Furthermore, this region can be unambiguously defined, so that it is conceptually clear what is meant by tracing over the interior degrees of freedom, even though it may not yet be possible to implement this calculation in complete detail. We find it quite striking that by this approach one can arrive so nearly at the standard entropy result.
### C Curvature and Naked-Black-Hole Behavior
In our discussion above of the trajectory of a particle probe, we focussed on the coordinate time that elapses over the course of the particle’s passage through the monopole. However, it also of interest to consider the elapsed proper time, which can be found by integrating $`d\tau /dr`$ \[see Eq. (13)\]. For core-type solutions this gives a finite nonzero result with no unusual behavior as the critical limit is approached. The situation with Coulomb-type solutions is, on the other hand, quite striking. The sharp decrease in $`AB`$ at the quasi-horizon leads to a corresponding decrease in $`d\tau /dr`$, so that the proper time elapsed while the probe is within the quasi-horizon is<sup>‡‡</sup><sup>‡‡</sup>‡‡The drop in $`AB`$ also has consequences for the shape of the trajectory through the interior. By combining Eqs. (11) and (13), one finds that in the critical limit all probes follow a straight line passing through the origin, regardless of the incident angle with which they hit the horizon.
$$\mathrm{\Delta }\tau \frac{2r_{}}{E}\sqrt{AB}|_{r=0}ϵ^q.$$
(26)
In the critical limit $`AB`$ vanishes identically for $`r<r_{}`$, and $`\mathrm{\Delta }\tau =0`$.
This vanishing of $`\mathrm{\Delta }\tau `$ is related to another interesting property of these solutions. It is well known that the Riemann tensor is nonsingular at a black hole horizon. It therefore does not seem surprising that in the most familiar black hole solutions, the Schwarzschild and Reissner-Nordstrom, a particle suffers no unusual effects as it crosses the horizon. However, Horowitz and Ross showed that this is not always the case. Because of the acceleration of a particle as it approaches the horizon, the components of the Riemann tensor in a coordinate frame that is freely falling with the particle can be quite different from the components measured in a static frame. With a metric of the form of Eq. (1), the components $`R_{t^{}kt^{}k}`$ (where $`k`$ denotes a transverse spatial direction and $`t^{}`$ the time in the boosted frame) are given by
$$R_{t^{}kt^{}k}=\frac{1}{2r}\frac{d}{dr}\left[\frac{E^2}{AB}\frac{1}{A}\right],$$
(27)
where $`E`$ is the energy per unit mass of the infalling particle.
The fact that this curvature component is never large (with $`E`$ of order unity) for the Schwarzschild and Reissner-Nordstrom black holes is a consequence of the fact that $`AB`$ is constant in both cases. This is not true in general. Horowitz and Ross found several examples of dilaton black holes for which $`R_{t^{}kt^{}k}`$, and thus the tidal forces felt by an infalling particle, could be made arbitrarily large near the horizon by taking the solution to be sufficiently close to extremality. They introduced the term naked black hole to indicate the fact that this (almost) singular behavior occurs outside the horizon. Subsequently , they showed that in these examples $`R_{t^{}kt^{}k}`$ was inversely proportional to the square of the proper time remaining before the particle reached the singularity at $`r=0`$.
Applying their results to our solutions, we find that near-critical Coulomb-type solutions display naked black hole behavior, even though they are not black holes at all. This can be seen by noting that Eq. (8) implies that
$$\frac{d}{dr}\left(\frac{1}{AB}\right)=\frac{16\pi GrK}{AB}.$$
(28)
Inside the quasi-horizon $`ABϵ^{2q}`$, while the radial derivatives of $`u`$ and $`h`$, and hence $`K`$, are of order unity. Inserting this result into Eq. (27), gives
$$R_{t^{}kt^{}k}ϵ^{2q}.$$
(29)
Note that this is proportional to $`(\mathrm{\Delta }\tau )^2`$, giving a relationship between tidal forces and proper time reminiscent of the examples described in Ref. .
## IV Collapse to an Extremal Black Hole
To gain an insight into the third law of thermodynamics, as noted in Sec. I, it is of interest to determine whether systems with either initially nonsingular spacetimes or initially non-extremal black holes can evolve into systems with an extremal horizon. It appears that under reasonable conditions of finiteness and causality this cannot be done by adding charge to an undercharged object . The discussion in the previous two sections, however, suggests an alternative. Recall that our subcritical monopoles are overcharged; i.e., they have a charge larger than their mass. In the normal Reissner-Nordstrom case, such a system would exhibit a naked singularity. However, in the monopole the Coulomb core is screened by the massive particles in such a way that no gravitational singularity exists. This suggests a scenario by which an extremal black hole is dynamically formed from a monopole by dropping in uncharged matter.
Let us add to our theory a chargeless matter field that is coupled to the monopole fields only through gravity. We then allow a spherical shell (of small, but finite thickness) of this matter fall into the monopole. If the mass of the shell is sufficiently small, we do not expect a horizon to be formed. On the other hand, if the shell contains enough matter, the system should collapse to form a black hole. It seems plausible that threshold case between these two regimes should produce an extremal horizon.
In the case of a Coulomb-type solution, one might run into difficulties because of the naked-black-hole behavior it exhibits. This concern, however, should not be an issue for the case of a core-type solution. Here the fields of the critical solution are much better behaved; what nonanalyticities exist at the horizon are mild, and become increasingly so as one increases $`b`$. Moreover, nothing unusual happens to the near-critical core-solutions as one parametrically approaches criticality. Adding a small Schwarzschild black hole at the center of the self-gravitating monopole should not significantly change the scenario. The infall of a spherical shell of appropriate mass should still turn the quasi-horizon of a near-critical solution into an extremal horizon. The black hole would essentially play a spectator role, as its presence is largely irrelevant to the dynamics of the system.
We have carried out numerical simulations to test these ideas. We begin with a core-type monopole solution that, for numerical convenience, has a small Schwarzschild black hole with horizon radius $`r_H`$ at its center. The parameters are chosen to be such that the solution is near criticality, so that only a small amount of additional matter is needed for the quasi-horizon at $`r=r_{}`$ to collapse to a true horizon. We add to the theory a massive scalar field $`\chi (r,t)`$ that is coupled only to gravity. We then send a spherically symmetric Gaussian pulse of $`\chi `$ field in toward the monopole and watch the system evolve. To simplify the computation, we freeze the matter field variables $`u`$ and $`h`$ at their initial values; because the fields for near-critical core-type monopoles are not very sensitive to the metric (cf. Fig 2), this approximation should cause little error.
When the pulse amplitude is larger than some threshold value, the pulse falls into the monopole until the metric function $`1/A`$ develops a simple zero near the location of the quasi-horizon of the initial monopole configuration. This newly formed horizon is non-extremal. More interesting is the situation where the pulse amplitude is at, or just below, this threshold value. Figure IV shows a sequence of snapshots illustrating this scenario at four different points during the pulse’s motion inward.<sup>\**</sup><sup>\**</sup>\**Note that we have chosen a normalization of time such that $`AB1`$ as $`rr_H`$, the horizon of the internal black hole; this corresponds to using a time variable appropriate to an observer in the interior of the monopole. If we had used the more conventional normalization with $`AB`$ equal to unity at spatial infinity, the time coordinate would be that appropriate to an external observer and would grow rapidly as the pulse approached the quasi-horizon. (The distortion of the pulse is due to the backreaction of the monopole metric.)
As required by Birkhoff’s theorem, $`1/A`$ is is undisturbed ahead of the pulse, but undergoes a shift as the pulse passes. Thus, in the region ahead of the pulse $`1/A`$ is the same as it was in the original configuration, while behind the pulse it has the form corresponding to a static monopole with a central black hole whose horizon radius exceeds $`r_H`$ by an amount determined by the energy-momentum of the infalling pulse. These two are joined by a kink at the pulse position. As the pulse passes through the quasi-horizon, $`1/A`$ reaches its minimum.
Similarly, the plot of $`\sqrt{AB}`$ appears as nearly a step function centered at the pulse position, with the inner and outer regions corresponding to the initial and final configurations. The jump in $`\sqrt{AB}`$ at the pulse position varies with time. It reaches its maximum when the pulse is passing through the quasi-horizon, and then decreases. (The resulting variation in the value of $`\sqrt{AB}`$ at large $`r`$ may appear to violate causality, but is actually a just a consequence of the gauge choice implicit in our choice of coordinates.)
As the pulse continues past the quasi-horizon into the monopole core, the metric function $`1/A`$ remains static outside the pulse. However, $`\sqrt{AB}_{\mathrm{outside}}`$ decreases from its maximum value as the pulse continues inward. Eventually, the pulse bounces off the central black hole (with a small amount of its energy being absorbed) and the process reverses itself. The pulse passes by the quasi-horizon and retreats to infinity. As it does so, one sees the metric variables restore themselves to almost the original values they had before the insertion of the pulse.
As the initial pulse amplitude is increased towards its threshold value, the minimum value achieved by $`1/A`$ approaches zero, while the maximum value of $`\sqrt{AB}`$ appears to grow without bound. The time $`t_{\mathrm{qh}}`$ at which the pulse passes through the quasi-horizon does not vary appreciably.<sup>\*†</sup><sup>\*†</sup>\*†This would not be the case if we had fixed the normalization of $`AB`$ at spatial infinity; $`t_{\mathrm{qh}}`$ would then diverge as the threshold amplitude was approached.
A numerical simulation will not, of course, be able to produce a precisely extremal horizon. In our simulations, we have been able to adjust the pulse amplitude to make the minimum $`1/A`$ be as small as $`2.3\times 10^5`$. In analyzing the behavior of the solutions as the pulse amplitude is varied, we see no indication of any singularity as the threshold is approached. We therefore expect that a pulse precisely at threshold would produce a nonsingular extremal horizon. In this case of critical collapse, the subsequent evolution of the system would be quite similar to the subcritical case, with the pulse continuing inward, bouncing off the central black hole, and then retreating outward. However, in this case the spacetime into which it moves is causally distinct from the one where the pulse had originated; i.e., it is a new sector of the Penrose diagram.
As a final comment, in all this analysis, dropping in pressureless dust should give analogous results. One can invoke Birkhoff’s law so that the metric behind the (radially thick) dust shell must be represented by a static metric. But one should expect the same sorts of naked singularity behavior since this results from the interaction of metric variables.
## V Concluding Remarks
In this paper we have used near-critical self-gravitating monopoles as tools for studying the transition from a nonsingular spacetime to one with a horizon. By analyzing the properties of trajectories that pass through the quasi-horizon and then emerge again, we have seen that the observations made by an external observer vary continuously and show no evidence of discontinuity when the critical limit is reached. This analysis also shows how the many causally distinct regions of the extremal black hole spacetime naturally emerge from the simple Penrose diagram of the subcritical monopole.
A somewhat unexpected result from this analysis is that for the Coulomb-type solutions the proper time required to traverse the interior region vanishes in the critical limit. This is closely associated with the fact that near-critical Coulomb solutions display naked black hole behavior; these are the first examples of configurations without horizons that do so. However, the absence of this behavior in core-type solutions shows that this is not a universal property of near-critical solutions.
Our analysis also sheds light on some aspects of black holes themselves. We have seen that the region bounded by the quasi-horizon becomes effectively inaccessible to outside observers when the solution is sufficiently close to criticality. The interior degrees of freedom thus become unmeasurable. Tracing over them then leads a naturally defined entropy that can be attributed to this configuration. An order of magnitude estimate of this entropy agrees with the Hawking-Bekenstein formula for the entropy of a black hole; it seems plausible that in the critical limit the agreement would be precise. These nonsingular near-critical solutions thus provide a concrete and unambiguous framework for implementing the old idea that black hole entropy might be understood in terms of the degrees of freedom hidden behind the horizon.
Finally, we have argued that an extremal black hole can be produced by allowing additional matter to fall into a near-critical monopole. We have illustrated this by numerical simulations. Starting with an initially nonsingular monopole, this leads to a zero-temperature black hole where there had previously been no horizon at all. Alternatively, one can start with a small black hole at the center of the monopole. In this latter case, a configuration with a nonzero Hawking temperature evolves into one with $`T=0`$. The existence of these possibilities gives additional clues for, and constraints on, a more precise formulation of the third law of black hole thermodynamics.
###### Acknowledgements.
We wish to thank Gary Horowitz and Krishna Rajagopal for helpful conversations. This work was supported in part by the U.S. Department of Energy. |
no-problem/0001/cond-mat0001134.html | ar5iv | text | # First and second order transition in frustrated XY systems
## I INTRODUCTION
Phase transitions of frustrated spin systems have been extensively studied during the last decade (for reviews see ). In particular the nature of the phase transition of the stacked triangular antiferromagnet (STA) with $`XY`$ spins interacting via nearest–neighbor bonds has been extensively investigated . At high temperatures the symmetry group of this system is $`O(2)Z_2`$ whereas at low temperatures this symmetry is completely broken. The Ising symmetry $`Z_2`$ has its origin in the non collinearity of the spins in the ground state which can be classified as chirality plus or minus. For non frustrated systems the symmetry group in the high temperature region is simply $`O(2)`$ and this difference in symmetry between frustrated and non frustrated spin systems should lead to a different critical behavior. Bailin and Garel , using the renormalization group (RG), proved that there is no stable fixed point close to space dimension $`d=4`$ and they concluded that the transition is of first order. Extending the RG technique to a $`N`$ component spin system, that is using $`4ϵ`$ expansion to first order in $`ϵ`$ (two-loops) and expanding also in $`1/N`$, Kawamura suggested a new universality class linked to the chirality for the transition with the exponents given by Monte Carlo simulations . With the same technique in $`d=4ϵ`$ and in three dimensions to three-loops, it was shown later that the transition for $`N=2`$ must be of first order . Further Monte Carlo studies have confirmed that the exponents for STA–system (see table V) are different from the ones of the standard $`O(N)`$ universality class (given in table VI). Simulations seem to favor the concept of a new chiral universality class or tricritical behavior. However, Plumer and Mailhot used different exchange constants for spins along the c–axis and spins in the triangular planes, so that the hexagonal STA–system is quasi one-dimensional. They concluded that the transition is weakly first order.
There are two principal groups of magnetic materials which can be modeled by our system. The first group are Hexagonal perovskites ABX<sub>3</sub>, which are quasi one dimensional systems which however order at low temperatures and have a planar anisotropy so that the spins are in the $`XY`$–plane. The most studied examples are CsMnBr<sub>3</sub> , RbMnBr<sub>3</sub> , CsVBr<sub>3</sub> , CsVCl<sub>3</sub> , and CsCuCl<sub>3</sub> . For a review see ref. . The results of the first four compounds are compatible with second order transitions with exponents more or less in agreement with the MC simulations: for example $`\nu =0.50(1)`$ in MC and between 0.54(3) and 0.57(3) experimentally (for details see Tables I). However, the specific heat measurement of CsCuCl<sub>3</sub> indicates a cross over to first order in zero magnetic field .
The second group are helimagnetic systems. Since the angle between the spins can be different from 120 for the STA–structure without changing the critical behavior, helimagnetic rare earths (see ref. , for Ho , for Dy and for Tb ) could also be analyzed by the STA–model. For helimagnets the critical behavior is quite varied (see the review for Ho and Dy ref. ). Essentially three types of results exist: one in favor of a $`O(4)`$ class , another in favor of a new universality class , and a third class which favor first order transition . See Tables II-IV for details. With these results a definite answer cannot be given about the order of transition. In section V we will come back to this point.
In order to check the results of the renormalization group studies we have studied the STAR and the Stiefel model . The first is derived from the STA model by imposing the constraint that in each triangle the sum of the spins is zero at all temperatures. The modes removed are irrelevant for the RG and the two models STA and STAR should be in the same universality class, provided such a class exists. As is explained in the next section the Stiefel model we use for the simulation is connected to the STAR model. Each cell of three spins plus constraint is equivalent to a system of dihedral, i.e. an ensemble of two perpendicular vectors. Neighboring pairs of vectors interact ferromagnetically, but only vectors of the same kind. Since these two systems have the same number of degrees of freedom, they should belong to the same universality class. We think that the two models are closer to the RG studies than the original stacked triangle antiferromagnet. We can therefore check the predictions of the RG and gain an understanding of the difficulties one has with the results of Monte Carlo simulations and measurements in the critical region.
In section II we present the two models. The simulations and the details of the finite size scaling analysis for a first order transition are explained in section III. The results are shown in section IV. Discussion and conclusions are in section V.
## II Model and simulation
### A The STAR model
Starting from the stacked triangular antiferromagnet (STA) we take the simplest Hamiltonian with one antiferromagnetic interaction constant $`J>0`$
$$H=J\underset{(ij)}{}𝐒_i.𝐒_j,$$
(1)
where $`𝐒_i`$ are two component classical vectors of unit length. The sum runs over all nearest neighbor pairs, that is the spin $`𝐒_i`$ has six nearest neighbor spins in the same $`XY`$–plane and two in $`Z`$–direction in adjacent planes. The ground state is characterized by a planar spin configuration with three spins on each triangle forming a $`120^{}`$ structure with either positive or negative chirality (see Fig. 1). The ground state degeneracy is thus twofold like the Ising symmetry, in addition to the continuous degeneracy due to global rotations.
In the RG theory one uses the concept of local rigidity which means that the sum of three spins $`𝐒_1`$, $`𝐒_2`$, $`𝐒_3`$ on the corners of a triangle is set to zero
$$𝐒_1+𝐒_2+𝐒_3=\mathrm{𝟎}.$$
(2)
In this theory the local fluctuations violating this constraint become modes with a gap. Thus they do not contribute to the critical behavior and can be neglected . The constraint in (2) used for an ordinary collinear antiferromagnet with two spins instead of three eliminates also one degree of freedom so that only one is left which means that the critical behavior of antiferromagnet is the same as that of a ferromagnet. In our case we are left with two degrees of freedom. We choose one spin direction and then have one more choice for the chirality, that is the direction of the second spin could be chosen clockwise or counter–clockwise with respect to the first spin.
In order to impose local rigidity for the MC simulation, we first partition the lattice into interacting triangles which do not have common corners. This can be done as follows. In each $`XY`$–plane one selects in a row one ”supertriangle”. Then one finds two nearest supertriangles which do not share a common corner in the row (they are separated by two head-up and three head-down triangles). This process is repeated. Then one takes the next row until all rows in the $`XY`$–planes are filled with triangles which do not share corners, see Fig. 1. All spins are then located on the supertriangles and each spin belongs to only one supertriangle. Local rigidity means that the three spins in each supertriangle form a $`120^{}`$ structure. Only in the ground state all the supertriangle have the same orientation. At finite temperature local rigidity means that there are no local fluctuations within a supertriangle, but fluctuations between supertriangles are allowed.
The MC updating procedure for the state of the supertriangles is made as follows. At a supertriangle, we take a new random orientation for one of its three spins; next we choose a second spin so as to form a $`\pm 120^{}`$ angle with the first spin. For the orientation of the third no freedom is left. The interaction energy between the spins of this supertriangle with the spin of the neighboring supertriangles is calculated in the usual way and we follow the standard Metropolis algorithm to update one supertriangle after the other.
We consider $`LLL_z`$ systems , where $`LL`$ is the size of the planes, and $`L_z=2L/3`$ the number of planes. $`L`$ must be a multiple of three so that no frustration occurs because of periodic boundary conditions in the $`XY`$–planes. Simulations have been done for systems sizes with $`L=12,\mathrm{\hspace{0.17em}18},\mathrm{\hspace{0.17em}24},\mathrm{\hspace{0.17em}30},\mathrm{\hspace{0.17em}36}`$.
The order parameter $`M`$ used in the calculation is
$$M=\frac{1}{N}\underset{s=1}{\overset{3}{}}|M_s|,$$
(3)
where $`M_s`$ $`(s=1,2,3)`$ is the s-th sublattice magnetization and $`N=L^2L_z`$ is the total number of the lattice sites. This definition corresponds to the one for the ordinary antiferromagnet with only two sublattices. Instead of alternating signs for the collinear case the non collinear staggered magnetization is obtained by making a rotation of $`+120^0`$ ($`120^0`$) for the second (third) magnetization before summing over the three sublattice magnetizations.
The chirality $`\kappa `$ is defined in the usual way
$`\kappa _i`$ $`=`$ $`{\displaystyle \frac{2}{3\sqrt{3}}}\left[𝐒_i^1\times 𝐒_i^2+𝐒_i^2\times 𝐒_i^3+𝐒_i^3\times 𝐒_i^1\right],`$ (4)
$`\kappa `$ $`=`$ $`{\displaystyle \frac{1}{N^{}}}\left|{\displaystyle \underset{i}{}}\kappa _i\right|,`$ (5)
where the summation is over all supertriangles and $`N^{}=N/3`$ is their number. The chirality $`\kappa _i`$ of one triangle is parallel to the $`Z`$-axis and equal to $`\pm 1`$.
### B The Stiefel model
The Stiefel model can be derived from the STAR model . We give the main points. In each elementary cell an orthonormal basis
$$𝐞_a(i);a=1,2$$
(6)
is defined, where $`i`$ is the superlattice index. Each spin located in the cell can be represented in this basis
$$𝐒_i=\underset{a}{}s_a(i)𝐞_a(i).$$
(7)
If we put this expression into the Hamiltonian (1) we obtain a new Hamiltonian with interactions between the orthogonal vectors $`𝐞_a(x)`$:
$$H=J\underset{ij}{}\left[𝐞_1(i)𝐞_1(j)+𝐞_2(i)𝐞_2(j)\right].$$
(8)
The interaction can be chosen negative (or ferromagnetic) and the sum $`_{ij}`$ is for simplicity over nearest neighbor pairs of a simple cubic lattice instead of a hexagonal lattice since the new spins $`𝐞_a`$ (see Fig. 2 and 3) are no longer frustrated. The chirality $`\kappa `$ for the Stiefel model is defined as
$$\kappa =\frac{1}{N}\left|_i𝐞_1(i)\times 𝐞_2(i)\right|.$$
(9)
The Hamiltonian (8) is similar to the one of the Ashkin-Teller model . Indeed we can give this Hamiltonian a form which is close to it. The interaction energy of two nearest neighbors (ij) with opposite chirality (9) is zero and with the same chirality it is $`\mathrm{\hspace{0.17em}2}𝐞_1(i)𝐞_1(j)`$ (see Fig. 3). Therefore the Hamiltonian can be written as
$$H=J\underset{i,j}{}(1+\sigma _i\sigma _j)𝐒_i𝐒_j$$
(10)
where $`\sigma =\pm 1`$ is an Ising spin representing the chirality and $`𝐒_i`$ is an $`XY`$–spin. In Hamiltonian of Ashkin and Teller only Ising spins appear.
Despite the fact that the Stiefel model is extensively studied, especially in two dimensions, no clear picture emerged. The problem is to know whether there are two transitions, an Ising and a Kosterlitz–Thouless transition, or only one transition of a new type . In three dimensions it has been shown that there is only one transition . Here the problem is to determine the order of the transition.
The procedure of MC procedure is as follows. At each site one takes a new random orientation for the first vector and chooses for the second vector a perpendicular direction (we have two choices: $`\pm 90^0`$, the Ising degrees of freedom). We have two degrees of freedom the same number as for the STAR model. The interaction energy between this dihedral and its neighbors is calculated. If it is lower than the energy of the old state, the new state is accepted. Otherwise, it is accepted only with a probability according to the standard Metropolis algorithm. It is possible to use a cluster MC algorithm , but in the case of a strong first order transition there is no reduction of the critical slowing down . Periodic boundary conditions are used.
Systems with $`L=12,\mathrm{\hspace{0.17em}15},\mathrm{\hspace{0.17em}18},\mathrm{\hspace{0.17em}21},\mathrm{\hspace{0.17em}24}`$ have been simulated. To compare with the size $`L`$ of the STA or the STAR model, we must multiply $`L`$ by $`\sqrt{3}`$. One supertriangle or triangle contains three spins and is represented by one site in the Stiefel model. So we obtain equivalent sizes from 20 to 40.
The order parameter $`M`$ for this model is
$$M=\frac{1}{2N}\underset{s=1}{\overset{2}{}}\left|M_s\right|,$$
(11)
where $`M_s`$ $`(s=1,2)`$ is the magnetization for the vectors $`𝐞_\alpha `$ over all sites and $`N=L^3`$ is the total number of sites.
### C Definitions, histogram methods and finite-size scaling
We use in this work the histogram MC technique developed by Ferrenberg and Swendsen . The histogram for the energy $`P_T(E)`$ is very useful for identifing a first order transition. Also the data obtained by simulation at $`T_0`$ can be used to obtain thermodynamic quantities at temperature $`T`$ close to $`T_0`$. Since the energy spectrum of a Heisenberg spin system is continuous, the data list obtained from a simulation is basically a histogram with one entry per energy value. In order to use the histogram method efficiently, we divided the energy range $`E<0`$ by 10 000 bins. We have verified that we obtain the same results, with our precision, for 30 000 bins.
The critical slowing down in a first order transition is greater than in a second order transition because of energy barriers, and thus the time of the simulation, to go from one state to another grows exponentially with the size of the lattice. For this reason we restricted our simulations to systems not too large to have good enough statistics. In each simulation, at least 2 millions (3 millions for the greater sizes) measurements were made after enough Metropolis updating (500 000) were carried out to reach equilibration.
For each temperature T we calculate the following quantities
$`C`$ $`=`$ $`{\displaystyle \frac{(<E^2><E>^2)}{Nk_BT^2}},`$ (12)
$`\chi `$ $`=`$ $`{\displaystyle \frac{N(<M^2><M>^2)}{k_BT}},`$ (13)
$`\chi _\kappa `$ $`=`$ $`{\displaystyle \frac{N(<\kappa ^2><\kappa >^2)}{k_BT}},`$ (14)
$`V`$ $`=`$ $`1{\displaystyle \frac{<E^4>}{3<E^2>^2}},`$ (15)
where $`M`$ is the order parameter, $`C`$ the specific heat per site, $`\chi `$ the magnetic susceptibility per site, $`V`$ the fourth order energy cumulant, $`<\mathrm{}>`$ means the thermal average.
The finite size scaling (FSS) for a first order transition has been extensively studied . A first order transition should be identified by the following properties:
* $`P_T(E)`$ has a double peak.
* The maximum of the specific heat $`C`$ and the susceptibilities $`\chi `$ and $`\chi _\kappa `$ are proportional to the volume $`L^d`$.
* The minimum of the fourth order energy cumulant $`V`$ varies as:
$`V`$ $`=`$ $`V^{}+bL^d,`$ (16)
where $`V^{}`$ is different from 2/3.
* The temperatures $`T(L)`$ at which the quantities $`C`$, $`\chi `$ or $`\chi _\kappa `$ have a maximum should vary as:
$`T(L)=T_c+aL^d.`$ (17)
All this conditions will be verified for our systems.
## III Results
In Fig. 4 the specific heat $`C`$ of the STAR model is plotted as function of the temperature for various sizes (we note that the maximum is 30 times as large as the usual value of STA which is a sign of a first order transition). The value of the maximum as function of the volume is shown in Fig. 5 for $`C`$, $`\chi `$ and $`\chi _\kappa `$. We note that the maxima vary like the volume except for the smaller sizes where further corrections are important.
In Fig. 6 the same quantities are shown for the Stiefel model. In all cases the maxima vary for the greater sizes proportional to the volume as they should for a first order transition.
In Fig. 7 and 8 we have plotted $`V`$ as function of $`T`$ for different sizes $`L`$ for the STAR and the Stiefel model respectively. We can see that $`V`$ does not tend to 2/3 (for a second order transition) but to a value $`V^{}<2/3`$. This value is calculated by fiting the minimum with (16). As result we yield for the STAR model
$`V^{}=0.652(2)`$ (18)
and for the Stiefel model
$`V^{}=0.625(3).`$ (19)
The Fig. 9 and 10 show the energy distribution for different sizes at different temperatures for the STAR and Stiefel model. The double peaks observed, even for a very small $`L`$, indicate a strong first order transition. With increasing sizes, these two peaks are separated by a region of zero probability, indicating a discontinuity of the energy at the transition. We estimate the correlation length $`\xi _0`$ by $`1/3`$ of the first size where the two peaks are well separated by a region of zero probability. This is our estimate of the distance needed for two phases to coexist. This method yields the correct answer in the case of Potts models. We obtain $`\xi _012`$ for the STAR model and $`\xi _09`$ for the Stiefel model (see Fig. 9 and 10).
To obtain the critical temperature we can use (17). The results are
$`T_c=2.2990(5)`$ (20)
for the STAR model and
$`T_c=2.4428(4)`$ (21)
and for the Stiefel model. We have sizable corrections for the small systems. Comparing the last results with those of Kunz and Zumbach for the case $`V_{2,2}`$ similar to ours we agree with their result $`T_c`$=2.445.
Our results show clearly the first order transition for the STAR and Stiefel model. So we confirm the indication given in for the Stiefel model where in high temperature region $`\nu `$ was determined not too far away from 1/3 which is the value for a first order transition.
## IV Discussion
We have shown that the STAR and the Stiefel model have first order transitions. These models are equivalent to the STA in the RG theory, because the constraint of local rigidity is not relevant in the transition region. Consequently the phase transition for the original triangular antiferromagnet STA must also be of first order, and this result holds generally for all systems with a breakdown of symmetry of type $`O(2)Z_2`$. How can we reconcile our results with those of the MC of the STA model and the experimental results?
For the MC simulation of the original frustrated spin system the second order transition is an effect of the finite system size according to Zumbach’s analysis of ”almost second order phase transitions” . The main point of this analysis is that the stable fixed point $`F_c`$, known to exist only for the number of components $`N>N_c`$, moves into the unphysical complex plane when $`N<N_c`$. In our case $`N=2`$, the estimation for $`N_c`$ is $`N_c=3.91`$ and second order could never occur. Nevertheless this complex fixed point has a large basin of attraction and mimics a behavior of a real fixed point. Only if the system is very large, i.e. if $`L\xi _0`$, where $`\xi _0`$ is the largest correlation length, the transition will appear of first order. The phenomenon of a crossover between a second order to a first order transition is not so uncommon. An extreme case is the Potts model in two dimensions with $`q=5`$ components, where the transition is known to be of first order . The MC gives always a second order transition with critical exponents of an instable fixed point due to the enormous correlation length $`\xi _0`$.
If the Hamiltonian is a sum of two terms, one interaction is of the Heisenberg symmetry and the other favors Ising symmetry: $`H=H_{Heisenberg}+H_{Ising}`$ and if $`H_{Ising}H_{Heisenberg}`$, one has a crossover between a region of Heisenberg type to Ising behavior close to $`T_c`$. If the system size is too small we will only see the region controlled by the Heisenberg fixed point and therefore obtain the exponents of the Heisenberg magnet. In a sense we have the same situation if we replace the Heisenberg fixed point by the Zumbach fixed point. However, there is an essential difference: for the fixed point in the complex plane, one has to modify the scaling relation
$$\gamma /\nu =2\eta +c$$
(22)
by a constant $`c`$ different from zero . If the fixed point is real $`c`$ must be zero.
We can use this relation as a criterion for real or complex fixed points. In three dimensions $`\eta `$ is usually small, that is $`0.03`$ (see table VI), but it must be positive and therefore $`\gamma /\nu 2`$ for a genuine second order phase transition. If the ratio $`\gamma /\nu >2`$ the fixed point must be complex. The correction $`c`$ in the scaling law (22) will depend on the distance of the fixed point from real space. Therefore one expects that $`c`$ will be greater for the $`XY(N=2)`$ case than for the Heisenberg $`(N=3)`$ case. For frustrated Heisenberg systems it will be difficult to find out whether $`\eta _{eff}=\eta c`$ is negative.
With the relation $`\gamma /\nu =2\eta `$ we obtain $`\eta =0.06`$ using the results of MC simulation and $`\eta =0.16`$ from experimental values of Ho . Thus the fixed point is in the complex plan and the second order transition observed in the XY systems has only an “almost second order” character.
The effect of imposing local rigidity obviously forces the system to stay away from the region of influence of the complex fixed point $`F_c`$ and thus permits to ”see” the true first order behavior. Introducing larger coupling constants for inter–plane interactions than for intra–plane interactions seems to have a similar effect as the local rigidity constraint.
A further remark concerning MC simulations and the fixed point $`F_c`$ in the complex plane; Zumbach has shown that the finite size scaling in this case does not hold. The FSS ansatz for the free energy should be replaced by
$$f=L^dg[L/\xi ,c_2\text{ln}(L)],$$
(23)
where $`g`$ is a function of $`L/\xi `$, but also of ln$`L`$. The constant $`c_2`$ is proportional to the constant $`c`$ of (22). This constant is small and therefore the correction to FSS. Indeed if we take for the true value of $`\eta `$ the value of the ferromagnetic case $`\eta 0.03`$ (see table VI) the value of $`c`$ will be at most equal to $`c0.03+0.060.1`$ (see discussion above) and if we compare with $`2\eta `$ in (22) this gives an error of 5 %. However, small but not negligible corrections to the standard FSS could explain the differences in MC simulations obtained with different methods (see table V) and also some of the differences in the experimental values (tables I-IV).
We will discuss now the experiments in the light of the concepts used. In order to see the first order region the temperature resolution is the limiting factor not the finite size. However, they are linked through $`t_0\xi _0^{1/\nu }`$ with $`\nu 0.5`$ found by MC and $`\xi _0`$ depending on the materials studied. The temperature distance $`t_0=(TT_c)/T_c`$ could be too small to be measurable.
The experiments on CsMnBr<sub>3</sub> , RbMnBr<sub>3</sub> and CsVBr<sub>3</sub> (I) give exponents compatible with those of MC on STA and a second order transition (see table I and V). We can interpret this result by the fact that the systems are under the influence of a complex fixed point and $`t_0`$ is too small to observe a first order transition.
The case CsCuCl<sub>3</sub> (table I) is different since the authors observe a crossover from a second order region with exponents compatible with MC on STA for $`10^3<t<5.10^2`$ to a region of first order transition for $`5.10^5<t<5.10^3`$. For $`t<t_010^3`$ one seems to observe the true first order region.
Helimagnetic rare earth metals are more complicated as already discussed in the introduction (see also tables II-IV). The results compatible with those of the MC on STA for Ho (table II), Dy (table III) and Tb (table IV) can be interpreted as before: the systems are under the influence of $`F_c`$. The first order transition for Ho and Dy is due to the fact that the measurements were done in the first order region near the critical temperature. The values of the exponent $`\beta 0.39`$ in the case of Ho and Dy (see table II-III) are not compatible with those found by MC ($`\beta 0.25`$). This fact can be explained by the presence of a second length scale in the critical fluctuations near $`T_c`$ related to random strain fields which are localized at or near the sample surface . Thus the critical exponent $`\beta `$ measured is of this second length. However the result of $`\beta `$ for Tb (table IV) shows values compatible with MC but it has been proved that this second length is present also in Tb . Further measurements to determine $`\beta `$ should help in the interpretation.
We have tried to give a general picture of the phase transition of frustrated XY spins where the breakdown of symmetry is of type $`O(2)Z_2`$. We have shown that this transition is first order but usually not seen because of the presence of a fixed point in the complex plane. One way to observe that the behavior is really driven by such a fixed point is the existence of a negative value for $`\eta `$ in the Monte Carlo simulations and experimental systems. The method used here is to impose local rigidity. This constraint does not change the behavior of the system but permits the system to stay outside the region of influence of the complex fixed point. So we can rely on the renormalization group study and the true behavior is first order. There is no ”new chiral universality class” in strict sense for our system ($`N=2`$). Another possibility discussed in the literature is that the transition is influenced by the presence of topological defaults which are not visible in the continuum formulation of the RG (for the presence of topological defaults in Stiefel model see ).
From our experience with the frustrated $`XY`$–model we conclude that the true first order transition for the frustrated Heisenberg model cannot be reached in MC simulations. The experimental situation should be similar. However, due to presence of the anisotropies we never reach the Heisenberg first order region but will have a crossover to the Ising or $`XY`$ region .
## V Acknowledgments
This work was supported by the Alexander von Humboldt Foundation. One of the authors (D.L.) is grateful to Professors B. Delamotte, H.T. Diep and A. Dobry for discussions, and for A.I. Sokolov for the reference to the proof of $`\eta 0`$. |
no-problem/0001/hep-ph0001170.html | ar5iv | text | # FINAL STATE INTERACTION IN HEAVY HADRON DECAY
## 1 Does the FSI phase diminish with the initial hadron mass ?
For a limited number of decay modes, we can extract the FSI phase directly from experiment. It is large for the $`D`$ decays and small for the $`B`$ decays:
$`\mathrm{\Delta }\delta `$ $`=`$ $`80^{}\pm 7^{};DK^{}\pi ^+/\overline{K}^0\pi ^+,`$ (1)
$`\mathrm{\Delta }\delta `$ $`<`$ $`\{\begin{array}{cc}11^{};\hfill & BD^+\pi ^{}/D^0\pi ^{}\hfill \\ 16^{};\hfill & BD^+\rho ^{}/D^0\rho ^{}\hfill \\ 29^{};\hfill & BD^+\pi ^{}/D^0\pi ^{}.\hfill \end{array}`$ (5)
Though nothing is known for $`B\pi \pi ,K\pi ,K\overline{K}`$, we might conjecture that the FSI phases diminish as the initial mass increases. Meanwhile the relative FSI phase is known for the $`ggg`$ and ”$`\gamma `$” decay amplitudes of the $`J/\psi `$ decays: $`\mathrm{\Delta }\delta 90^{}`$ and $`75^{}`$ for $`J/\psi 0^{}0^{}`$ and $`0^{}1^{}`$, respectively. Therefore, if the FSI phases diminish with the increasing initial mass, it should start diminishing somewhere between $`m_{J/\psi }`$ and $`m_B`$ or at higher energies.
Can we argue that short-distance (SD) QCD should dominate in the FSI of $`B`$ meson decays ? I am skeptical. Take the spectator decay $`B^+(\overline{b}u)K^0(\overline{s}d)\pi ^+(\overline{d}u)`$. $`K^0`$ is formed with energetic $`\overline{s}`$ and $`d`$ flying away colinearly, while $`\overline{d}`$ picks up the soft spectator $`u`$ to form $`\pi ^+`$. The gluons exchanged between $`\overline{d}`$ and $`\overline{s}`$ (or $`d`$) are hard. But the gluons between the soft spectator $`u`$ and $`\overline{s}`$ (or $`d`$) are not so hard. The CM energy of the $`u`$ and $`\overline{s}`$ (or $`d`$) grows only with $`\sqrt{m_b}`$, not linearly in $`m_b`$. By simple kinematics, $`m_{\overline{d}u}(\mathrm{\Lambda }_{\mathrm{QCD}}m_b)^{1/2}`$1.2 GeV for $`E_u\mathrm{\Lambda }_{\mathrm{QCD}}`$, which is in the middle of the resonance region of the $`\overline{s}d`$ channel. Then long-distance (LD) interactions cannot be ignored between $`K^0`$ and $`\pi ^+`$. One may stick to the quark picture by invoking the local quark-hadron duality. While it may be appropriate for inclusive decays, the duality is a dynamical hypothesis yet to be tested in two-body B decays.
## 2 Why is the FSI complicated in the $`B`$ decay ?
The FSI phases of $`K\pi \pi `$ and the hyperon decay are equal to the strong interaction phase shifts by the phase theorem, since the final states are the eigenstates of S matrix. In contrast, the FSI phases of the $`B`$ decay depend on both strong and weak interactions. Take the decay $`BK\pi `$. Since $`K\pi `$ is made of a large number of eigenchannels $`a`$ of S matrix, $`|K\pi ^{out}=_ac_a|a^{out}`$, the decay amplitude through the operator $`𝒪`$<sub>i</sub> acquires an eigenphase $`\delta _a`$ from each eigenchannel as
$$M^{(i)}(BK\pi )=\underset{a}{}c_aM_a^{(i)}e^{i\delta _a},(c_a^{}=c_a;M_a^{(i)}=M_a^{(i)}).$$
(6)
The net phase of $`M^{(i)}`$ depends on $`𝒪_i`$ even for the same isospin eigenstate of $`K\pi `$. The FSI phase has nothing to do with the phase shift of elastic $`K\pi `$ partial-wave amplitude, $`\eta e^{i\delta _{K\pi }}\mathrm{sin}\delta _{K\pi }(_a|c_a|^2e^{2i\delta _a}1)/2i`$ at the $`B`$ mass.
## 3 Meson-meson interaction at $`E_{\mathrm{cm}}`$ 5 GeV
To understand the FSI, we need to know how two mesons interact at $`E_{\mathrm{cm}}=m_B`$. The elastic meson-meson scattering amplitude can be written as
$$T(s,t)=i\sigma _{\mathrm{tot}}se^{bt}+\underset{i}{}\gamma _i(t)s^{\alpha _i(t)}+\mathrm{}.$$
(7)
The dominant term is the Pomeron exchange, which is purely imaginary and sharply peaked in the forward direction. With a bit of theory, we find $`\sigma _{\mathrm{tot}}=1520`$mb and $`b2.8\mathrm{GeV}^2`$ for $`K\pi `$. These values are much the same for other meson scatterings. The non-Pomeron terms are down by $`\sqrt{s}`$ at $`t=0`$.
The partial-wave projection of $`T(s,t)`$ gives $`a_0(m_B)0.17i`$. It leads to a large inelasticity:
$$\sigma _{\mathrm{inel}}/\sigma _{\mathrm{tot}}=1\mathrm{Im}a_00.83.$$
(8)
A large portion of inelasticity presumably goes into multibody channels.
What does this imply in the two-body $`B`$ decays? For the factorization-allowed processes, the intermediate states would be similar to those which appear in elastic meson scattering. Among two-body intermediate states, those connected to the final state by the Pomeron exchange are the most important. For the factorization-forbidden processes in which a quark-pair annihilation must occur, the intermediate states of the Pomeron-exchange type are forbidden. Since the composition of the intermediate states are completely different, the factorization-allowed and forbidden amplitudes generally have quite different FSI phases even for the same isospin state of, say, $`K\pi `$.
## 4 Two-body intermediate approximation
In computing the FSI phases, one makes the two-body approximation for the intermediate states. It is a dubious approximation when inelasticity of scattering is high. Getting a number even in this approximation is a daunting task, if one tries to do it right.
One starts with the unitarity sum of the absorptive part of decay amplitude for $`Bn`$ (two-body) and keep only the two-body intermediate states:
$$\mathrm{Im}M_n=\underset{n^{}}{}a_0(s)_{nn^{}}M_n^{}^{},(s=m_B^2;n^{}=\text{two-body})$$
(9)
where $`a_0(s)_{nn^{}}(=a_0(s)_{n^{}n})`$ is the $`J=0`$ amplitude for $`nn^{}`$. Let us consider only the factorization-allowed processes. The intermediate states connected to $`n`$ by Pomeron exchange dominate in the sum. Therefor all $`a_0(s)_{nn^{}}`$ are purely imaginary. Then we can easily show that $`M_n`$ are also all purely imaginary. This is a special case where the FSI phases do not depend on $`𝒪`$<sub>i</sub>.
Phase differences of $`M_n`$ arise from the non-Pomeron exchanges, $`\rho `$, $`a_2,\mathrm{}`$. Computing the non-Pomeron contributions in Eq.(9) is nearly impossible: First of all, the individual terms in the unitarity sum are not real. Only the sum total is. One has to know the phases of $`M_n^{}`$. This is circular. To cut the circle, one might approximate $`M_n^{}`$ with the factorization amplitudes. That is the Born approximation valid only for small phases. After obtaining $`\mathrm{Im}M_n`$, one has to compute $`\mathrm{Re}M_n`$ by a dispersion integral of $`\mathrm{Im}M_n`$ over all energies. Even if one could compute for all two-body intermediate states, one has no idea as to how much the multibody states modify the answer.
## 5 Random approximation
Computing the FSI phases is difficult because there are so many intermediate states that we know very little about. Can we take advantage of the presence of many states? A statistical approach or a random approximation comes to our mind. Such an attempt was actually made with a limited success.
Start with the unitarity sum, this time including all intermediate states:
$$\mathrm{Im}M_n=\underset{n^{}}{}a_0(s)_{nn^{}}M_n^{}^{},(s=m_B^2;n^{}=\mathrm{all}).$$
(10)
Each term in the sum is a product of the pure strong-interaction amplitude and the weak decay amplitude. Let us postulate that the sign of their product is random from one intermediate state to another when there are very many states. This is a dynamical hypothesis. It is not true if SD dominates. Under this postulate one can compute the statistically likely values of the FSI phases as their standard deviations from zero. The likely value $`\overline{\delta }`$ of the FSI phase depends on the “favoredness” parameter $`\rho `$. It parametrizes how much favored or disfavored the decay modes are for a given decay operator: $`\rho <1`$ for the favored processes such as the factorization-allowed and color-favored modes, while $`\rho <1`$ for the unfavored processes such as factorization-forbidden and/or color-suppressed modes. The relation between $`\overline{\delta }_n`$ and $`\rho `$ is given by
$$\mathrm{tan}^2\overline{\delta }_n=\frac{\tau ^2(\rho ^2\tau ^2)}{1\rho ^2\tau ^2},(\tau <\rho <\frac{1}{\tau })$$
(11)
where $`\tau ^2`$ is a number determined by elasticity of the final-state rescattering ($`\tau ^20.20`$ for $`K\pi `$). The likely phase $`|\overline{\delta }_n|`$ of $`M_n`$ is plotted in Fig.1. It is small for the favored modes but can be large for unfavored modes. For the “typical processes” of $`\rho =1`$, the value is $`20^{}`$. It means that the FSI phases of the typical processes are most likely between $`20^{}`$ and $`20^{}`$.
This trend of $`\overline{\delta }`$ can be understood as follows: For the favored modes, the direct decay dominates over the indirect decay going through on-shell intermediate states. Since the latter is the origin of the FSI phase, the FSI phase is small. For the disfavored modes, the indirect decays dominate since the direct decay is suppressed. The random approximation quantifies this feature in terms of the favoredness parameter $`\rho `$, given elasticity of FSI. A similar qualitative reasoning was made by Rosner in the quark picture.
The limitation of the random approximation is that we cannot predict values of individual phases, but only their likely values.
## 6 Factorization and enhancement by FSI
### 6.1 Enhancement by FSI
The FSI not only generates a strong phase but also changes a magnitude of amplitude. Fermi argued in potential scattering that the FSI should modify the amplitude by the wave function at origin $`\phi (0,k)`$ of final two particles:
$$M\phi (0,k)M.$$
(12)
For slow final particles, the attractive force enhances the decay by pulling them close together while the repulsive force suppresses by pushing them away. Since $`\phi (0,k)`$ is equal to the Jost function $`1/f(k)^{}`$, it also proves the phase theorem.
This theory can be extended to relativistic rescattering. In relativistic particle physics, we use the phase-amplitude dispersion relation generically referred to as the Omnès-Mushkelishvili representation :
$$M(s+iϵ)=P(s)\mathrm{exp}\left(\frac{1}{\pi }_{s_{\mathrm{th}}}^{\mathrm{}}\frac{\delta (s^{})}{s^{}siϵ}𝑑s^{}\right),$$
(13)
where $`s`$ is the $`B`$ mass squared taken as a dispersion variable, and $`P(s)`$ is a polynomial arising from zeros of $`M(s)`$. $`P(s)`$ is related to the factorization amplitude since it contains no FSI. Therefore the FSI enhancement factor of the factorization amplitude is given by<sup>1</sup><sup>1</sup>1 For $`s^{}>(m_B^{}+m_\pi )^2`$, $`\delta (s^{})`$ contains the initial-state interaction or dissociation, part of which enters the form factor of the factorization amplitude. It is understood that the initial-state interaction contributing only to the form factor has been removed from $`\delta (s^{})`$ of Eq.(14).
$$E=\mathrm{exp}\left(\frac{𝒫}{\pi }_{s_{\mathrm{th}}}^{\mathrm{}}\frac{\delta (s^{})}{s^{}m_B^2}𝑑s^{}\right).$$
(14)
This form of enhancement was actually proposed and studied for elastic rescattering at low energies in the late 1950’s. The enhancement $`E`$ includes the FSI of both SD and LD. While the low energy portion of the phase integral is purely LD for an obvious reason, the high energy portion is not entirely SD.
### 6.2 Numerical exercise
I first illustrate the enhancement factor with the decay $`K\pi \pi `$. Since the FSI phase is equal to the scattering phase shift below the inelastic threshold, we can evaluate the low-energy portion of the phase integral up to $`1`$ GeV with the $`\pi \pi `$ phase shifts. The result is $`E=2.2`$ and $`0.8`$ for $`I=0`$ and $`I=2`$ channels, respectively, which amounts to the relative enhancement of $`M_{\mathrm{\Delta }I=1/2}/M_{\mathrm{\Delta }I=3/2}=2.9`$. This is a LD enhancement independent of $`𝒪`$<sub>i</sub>. Above the inelastic threshold, $`\delta (s^{})`$ is not longer equal to the phase shift and depends on decay operators. Therefore the integral above 1 GeV provides an operator-dependent enhancement similar to the SD enhancement of RG, but not identical to it.
Let us turn to the $`B`$ decay.
In the two-body intermediate approximation, the FSI phase approaches $`\pm 90^{}`$ at high energies. Such an amplitude potentially acquires a dangerously large enhancement or suppression since the phase integral can be large. If the factorization gives a right ballpark value for a decay amplitude, $`\delta (s^{})`$ must not stay too large at high energies. In the random approximation, the sign of $`\delta (s^{})`$ may fluctuate with $`s^{}`$. In this case, the effect of enhancement and suppression would be much smaller.
Here I present a numerical estimate for $`BK\pi (\pi \pi )`$ with one model FSI phase motivated by experiment. Knowing that the phases are small for $`D\overline{K}\pi `$ but possibly small for $`BK\pi `$, I choose the FSI phase as shown in Fig.2: $`\delta (s^{})`$ rises to large values ($`90^{}`$) around 2 GeV and falls linearly to zero at $`E_{\mathrm{max}}=O(m_B)`$.<sup>2</sup><sup>2</sup>2To be precise, this $`\delta (s^{})`$ includes part of the initial-state interaction above $`s^{}=(m_B^{}+m_\pi )^2`$. See the footnote in the preceding page.
This $`\delta (s^{})`$ suppresses the decay amplitude since the support of the phase integral is mostly below $`m_B`$ where $`1/(s^{}m_B^2)`$ is negative. The suppression factor is given by the lower curve in the right of Fig.2. It can easily a 10$``$40% correction to the factorization as $`E_{\mathrm{max}}`$ varies.
We learn two lessons from this exercise. Unlike the phase, the magnitude of amplitude receives the FSI effect from all energies. Even if the FSI phase happens to be zero at $`m_B`$, the FSI can generate a substantial correction to the factorization amplitude. Next, the low-energy (LD) contribution is power-suppressed in $`1/m_B^2`$, but not necessarily like $`\mathrm{\Lambda }_{\mathrm{QCD}}^2/m_B^2`$. It may be more like $`m_D^2/m_B^2`$ or $`m_{J/\psi }^2/m_B^2`$ ($`=1530\%`$).
## 7 Conclusion
If LD physics is important in two-body $`B`$ decays, computing individual FSI phases is nearly an impossible proposition. What we can predict at best is their statistically likely values.
The phase and the magnitude of amplitude are related by analyticity. Multiplying a large FSI phase on a factorization amplitude is very likely inconsistent with analyticity. If an amplitude is significantly enhanced or suppressed relative to its factorization value, it indicates that the FSI phase is large and LD physics is important. When we attempt to extract the CP phases from experiment, we should be apprehensive about the level of accuracy of factorization amplitudes. We must not use calculated values of FSI phases in analysis.
## Acknowledgment
This work was supported in part by the Director, Office of Science, Office of High Energy and Nuclear Physics, Division of High Energy Physics, of the U.S. Department of Energy under contract DE-AC03-76SF00098 and in part by National Science Foundation under grant PHY-95-14797.
## References |
no-problem/0001/cond-mat0001267.html | ar5iv | text | # Collective Modes and Electronic Raman Scattering in the Cuprates
## 1 INTRODUCTION
In spite of the considerable efforts to explain the experimental Raman spectra of cuprate superconductors, the $`A_{1g}`$ superconducting response is not yet completely understood. It has been shown that the theoretical description of the $`A_{1g}`$ Raman response was very sensitive to small changes in the Raman vertex harmonic representations, yielding peak positions varying between $`\mathrm{\Delta }`$ and 2$`\mathrm{\Delta }`$ . However, the data show peaks consistently slightly above $`\mathrm{\Delta }`$ for both YBCO and BSCCO.
In this paper we present calculations suggesting that the $`A_{1g}`$ peak position is largely controlled by a collective spin fluctuation (SF) mode near 41 meV, consistent with inelastic neutron scattering (INS) observations . We show that the $`A_{1g}`$ response is strongly modified by the SF term and is not sensitive to small changes in the Raman vertex. The experimental peak position is well reproduced by our model whereas the $`B_{1g}`$ and $`B_{2g}`$ response remain essentially unaffected by the SF mode.
## 2 MODEL CALCULATION
The CuO<sub>2</sub> bilayer is modeled by a tight binding band structure with a nearest ($`t`$) and a next nearest neighbor hopping ($`t^{}`$) parameter and an inter-plane hopping given by
$$t_{}(𝐤)=2t_{}\mathrm{cos}(k_z)[\mathrm{cos}(k_x)\mathrm{cos}(k_y)]^2.$$
(1)
$`k_z`$ can be 0 or $`\pi `$, for bonding or anti-bonding bands of the bilayer, respectively.
The spin susceptibility ($`\chi _s`$) is modeled by extending the weak coupling form of a $`d_{x^2y^2}`$ superconductor to include antiferromagnetic spin fluctuations by an RPA form with an effective interaction $`\overline{U}`$; i.e. $`\chi _s=\chi _0/(1\overline{U}\chi _0)`$ where $`\chi _0`$ is the simple bubble in the d-wave state. This form of the spin susceptibility is motivated by the fact that it contains a strong magnetic resonance peak at $`𝐪=𝐐=(\pi ,\pi ,\pi )`$ which was proposed to explain the INS resonance at energies near 41 meV in YBCO and BSCCO .
The Raman response function in the superconducting state is evaluated using Nambu Green’s functions. The spin fluctuations contribute to the Raman response via a 2-magnon process as shown in Fig. 1 where a schematic representation of the Feynman diagrams of the SF and the bubble contribution is plotted. For the electronic propagators we have used the bare BCS Green’s functions and a d-wave superconducting gap $`\mathrm{\Delta }_𝐤=\mathrm{\Delta }_0[\mathrm{cos}(k_x)\mathrm{cos}(k_y)]/2`$.
The total Raman response is calculated in the gauge invariant form which results from taking into account the long wavelength fluctuations of the order parameter . The total Raman susceptibility is thus given by
$$\chi _{tot}(𝐪=\mathrm{𝟎},i\mathrm{\Omega })=\chi _{\gamma \gamma }(\mathrm{𝟎},i\mathrm{\Omega })\frac{\chi _{\gamma 1}^2(\mathrm{𝟎},i\mathrm{\Omega })}{\chi _{11}(\mathrm{𝟎},i\mathrm{\Omega })}$$
(2)
where $`\chi _{ab}(𝐪=\mathrm{𝟎},i\mathrm{\Omega })`$ is determined according to Fig. 1. The analytical continuation to the real axis is performed using Padé approximants.
We have used several different forms for the Raman vertex $`\gamma `$ which possess the correct transformation properties required by symmetry. Our calculations show that the SF term yields vanishingly small corrections to the response in the $`B_{1g}`$ and $`B_{2g}`$ channels, but contributes substantially to the $`A_{1g}`$ channel. The shape of the total response in the $`A_{1g}`$ geometry is mainly dependent on the value of the effective interaction $`\overline{U}`$. Variations of $`\overline{U}`$ change the relative magnitude of the two diagrams summed in Fig. 1, changing the position of the peak in $`A_{1g}`$ geometry. Importantly, we find that the $`A_{1g}`$ response shows little dependence on the form used for the vertex: $`\mathrm{cos}(kx)+\mathrm{cos}(ky),\mathrm{cos}(kx)\mathrm{cos}(ky)`$, or the vertex calculated in an effective mass approximation. These results can be explained by symmetry reasons given that the SF propagator is strongly peaked for $`𝐐`$ momentum transfers.
## 3 COMPARISON WITH DATA
We compare the calculated Raman response with the experimental spectra of an optimally doped Bi-2212 sample in Fig. 2. Adding the SF contribution leads to a shift of the peak position from near $`\mathrm{\Delta }_0`$ for $`\overline{U}=0`$ to higher frequencies, allowing a better agreement with the experimental relative positions of the peaks in $`A_{1g}`$ and $`B_{1g}`$ geometries. For the fit we have adjusted $`t`$ to achieve a good agreement with the $`B_{1g}`$ channel, obtaining $`t=130`$ meV, and then adjusted $`\overline{U}`$ to match both the $`A_{1g}`$ peak position as well as the peak in the SF propagator to be consistent with the INS peak at 41 meV.
From this work we conclude that including the SF contribution in the Raman response solves the previously unexplained sensitivity of the $`A_{1g}`$ response to small changes in the Raman vertex. Whereas the SF (two-magnon) contribution controls the $`A_{1g}`$ peak, the $`B_{1g}`$ and $`B_{2g}`$ scattering geometries are essentially unaffected and determined by the bare bubble alone.
We would like to thank R. Hackl for numerous discussions. One of the authors (F.V.) would like to thank the Gottlieb Daimler and Karl Benz Foundation for financial support. |
no-problem/0001/cond-mat0001266.html | ar5iv | text | # Electron states in aone-dimensional random binary alloy
## I Introduction
Electron states in random systems have become an active research topic since the generality of localization phenomena in one dimension (1D) . Although it is well established that almost any nonzero disorder causes exponential localization of all eigenstates in 1D systems, regardless their energy (see, e. g., Ref. and references therein), there exist several exceptions. It is nowadays well known that extended states may arise in random systems where disorder exhibits short-range or long-range correlations . Spatial correlation means that random variables are not independent within a given correlation length. Suppression of localization by correlations was further put forward for the explanation of high conductivity of doped polyaniline as well as transport properties of random semiconductor superlattices .
In this work, we report further progress along the lines in the preceding paragraph. We turn ourselves to one of the pioneering works in the field, namely the work of Wu and Phillips on polyaniline (see Ref. and references therein). These authors showed that polyaniline can be mapped onto a random dimer model that has a set of extended states, originated by a resonance of a single dimer defect (two neighbor sites with the same energy). Electron states whose energy is close to this resonance turn out to be extended (in the sense that their localization length is larger than the system size) when dimers are placed at random in the 1D system. In this work we show that dimers are not needed to observe extended states in 1D random systems with short-range correlated disorder. To this end, we built up a simple model of semiconductor binary alloy —like ternary III-V compounds—. In these alloys (say Al<sub>x</sub>Ga<sub>1-x</sub>As), the cation sublattice is occupied by the same atoms (say As) while anions (say Al and Ga) are randomly distributed over the other sublattice. We model these alloys by considering a 1D random binary alloy with two species, referred to as A and B atoms hereafter. In order to mimic the disorder present in the anion sublattice, we further assume that the site energy of A atoms is randomly distributed from site to site while that of B atoms is the same over the entire cation sublattice. As a major point, we demonstrate the occurrence of extended states in the vicinity of the site energy of B atoms in spite of the fact that the system is purely 1D and random. Thus, we conclude that dimers (or larger defects like trimers or n-mers) are not required to observe extended states in 1D systems with short-range correlated disorder.
## II Model
We consider a 1D binary system where A (B) atoms are placed at odd (even) positions of the otherwise regular lattice, whose corresponding site energies are $`ϵ_{2n1}`$ ($`ϵ_{2n}`$) with $`n=1,2,\mathrm{}N`$, $`N`$ being the number of unit cells of the alloy. The Schrödinger equation for stationary eigenstates $`\psi _n(E)`$ is
$$(Eϵ_n)\psi _n+\psi _{n+1}+\psi _{n1}=0,n=1,2,\mathrm{},𝒩,$$
(1)
where the $`E`$ is the corresponding eigenenergy, $`ϵ_n`$ is the site energy and $`𝒩2N`$ is the number of atoms in the system. According to our model, site energy at even positions is the same and we can set $`ϵ_{2n}=0`$ without loss of generality. The source of disorder in this model arise from the stochastic fluctuations of site energy at even positions. We assume that $`\{ϵ_{2n1}\}_{n=1}^N`$ is a set of uncorrelated random Gaussian variables with mean value $`v`$ and variance $`\sigma ^2`$. Hereafter $`\sigma `$ will be referred to as degree of disorder. The joint distribution function of a realization of disorder is represented by the direct product of single Gaussians. Thus
$$ϵ_{2n}=v,ϵ_{2n}^2=v^2+\sigma ^2.$$
(2)
The Schrödinger equation (1) can be written via the $`2\times 2`$ promotion-matrix $`P_n`$ as follows
$$\left(\begin{array}{c}\psi _n\hfill \\ \psi _{n+1}\hfill \end{array}\right)=\left(\begin{array}{cc}0& 1\\ 1& E+ϵ_n\end{array}\right)\left(\begin{array}{c}\psi _{n1}\hfill \\ \psi _n\hfill \end{array}\right)P_n\left(\begin{array}{c}\psi _{n1}\hfill \\ \psi _n\hfill \end{array}\right).$$
(3)
By iterating this equation we can relate $`(\psi _n,\psi _{n+1})`$ and $`(\psi _0,\psi _1)`$ with $`\psi _00`$:
$$\left(\begin{array}{c}\psi _n\hfill \\ \psi _{n+1}\hfill \end{array}\right)=\underset{k=n}{\overset{1}{}}P_k\left(\begin{array}{c}\psi _0\hfill \\ \psi _1\hfill \end{array}\right)M_n\left(\begin{array}{c}\psi _0\hfill \\ \psi _1\hfill \end{array}\right),$$
(4)
where $`M_n`$ is referred to as the transfer-matrix. We find most convenient to deal with the promotion-matrix of the (diatomic) unit cell instead of that corresponding to a single atom (3), namely $`T_nP_{2n}P_{2n1}`$. For real $`E`$ and $`ϵ_n`$, the promotion-matrix $`T_n`$ can be regarded as an element of the $`SO(1,2)`$ group, isomorphic to $`SL(2,R)`$. It can be cast in the following form via the Pauli matrices $`\sigma _\mu `$
$$T_n=\left[\frac{E}{2}(Eϵ_{2n1})1\right]_2\frac{E}{2}(Eϵ_{2n1})\sigma _3+\frac{ϵ_{2n1}}{2}\sigma _1+i\left(E\frac{ϵ_{2n1}}{2}\right)\sigma _2,$$
(5)
where $`_m`$ denotes the $`m\times m`$ unit matrix. It is easy to demonstrate the following useful property $`T_n^1=\sigma _2T_n^{}\sigma _2`$. The transfer-matrix of the entire system ($`N`$ unit cells) is obtained as $`M_N=_{n=N}^1T_n`$. Oseledec’s theorem states that the following limiting matrix $`\mathrm{\Gamma }`$ exists
$$\mathrm{\Gamma }=\underset{N\mathrm{}}{lim}\left(M_N^{}M_N\right)^{1/2N},$$
(6)
with eigenvalues $`e^\gamma `$. The Lyapunov exponent $`\gamma `$ is nothing but the inverse of the localization length $`\lambda ^1`$, where $`\lambda `$ is given in units of the length of the unit cell.
## III Existence of extended states
In order to find the localization length one should calculate the matrix $`M_N^{}M_N`$ for large $`N`$. We will perform this task following the technique developed in Ref. . By using the formula for the decomposition of the product of two spin-1/2 states into the direct sum of scalar and spin-1 states, we have
$$(T_j)_\alpha ^{}^\alpha (T_j^1)_\beta ^\beta ^{}=\frac{1}{2}(\delta )_\beta ^\alpha (\delta )_\alpha ^{}^\beta ^{}+\frac{1}{2}(\sigma ^\mu )_\alpha ^{}^\beta ^{}\mathrm{\Lambda }_j^{\mu \nu }(\sigma ^\nu )_\beta ^\alpha ,$$
(7)
where
$$\mathrm{\Lambda }_j^{\mu \nu }=\frac{1}{2}\mathrm{Tr}\left(T_j\sigma ^\mu T_j^1\sigma ^\nu \right)$$
(8)
is the spin-1 part. Multiplying the expression (7) by the left and right by $`\sigma _2`$ we have
$$(T_j)_\alpha ^{}^\alpha (T_j^+)_\beta ^\beta ^{}=\frac{1}{2}(\sigma _2)_\beta ^\alpha (\sigma _2)_\alpha ^{}^\beta ^{}+\frac{1}{2}(\sigma ^\mu \sigma _2)_\alpha ^{}^\beta ^{}\mathrm{\Lambda }_j^{\mu \nu }(\sigma ^\nu \sigma _2)_\beta ^\alpha .$$
(9)
Now we should take into account the disorder and calculate the average of $`\mathrm{\Gamma }`$ by random distribution of $`ϵ_{2n1}`$ at odd sites
$$\mathrm{\Gamma }=\frac{1}{2}\sigma _2\sigma _2+\frac{1}{2}(\sigma ^\mu \sigma _2)(\sigma ^\nu \sigma _2)\left(\underset{j=1}{\overset{N}{}}\mathrm{\Lambda }_j\right)^{\mu \nu },$$
(10)
where $`\mathrm{\Lambda }_j`$ is defined by (8). According to Oseledec’s theorem , the Lyapunov exponent and, correspondingly, the localization length will be given by
$$\lambda ^1=\mathrm{log}[\xi (E)],$$
(11)
where $`\xi (E)`$ is the closest to unity eigenvalue of $`\mathrm{\Lambda }_j`$, whose elements are
$`\mathrm{\Lambda }_j^{11}`$ $`=`$ $`{\displaystyle \frac{1}{2}}\left(2+E^4+4Ev2E^3v\sigma ^2v^2+(\sigma ^2+v^24)E^2\right)`$ (12)
$`\mathrm{\Lambda }_j^{12}`$ $`=`$ $`{\displaystyle \frac{i}{2}}\left(E^42E^3v+\sigma ^2+v^2+(\sigma ^2+v^22)E^2\right)`$ (13)
$`\mathrm{\Lambda }_j^{13}`$ $`=`$ $`E^3v+2E^2v(\sigma ^2+v^22)E`$ (14)
$`\mathrm{\Lambda }_j^{21}`$ $`=`$ $`{\displaystyle \frac{i}{2}}\left(E^4+4Ev2E^3v\sigma ^2v^2+(\sigma ^2+v^22)E^2\right)`$ (15)
$`\mathrm{\Lambda }_j^{22}`$ $`=`$ $`{\displaystyle \frac{1}{2}}\left(2+E^42E^3v+\sigma ^2+v^2+(\sigma ^2+v^2)E^2\right)`$ (16)
$`\mathrm{\Lambda }_j^{23}`$ $`=`$ $`i\left(E^3+v2E^2v+(\sigma ^2+v^2)E\right)`$ (17)
$`\mathrm{\Lambda }_j^{31}`$ $`=`$ $`2E+E^3+vE^2v`$ (18)
$`\mathrm{\Lambda }_j^{32}`$ $`=`$ $`i\left(E^3vE^2v\right)`$ (19)
$`\mathrm{\Lambda }_j^{33}`$ $`=`$ $`12E^2+2Ev.`$ (20)
Delocalized states have an infinite localization length and, therefore, at some particular energy $`E`$, the matrix $`\mathrm{\Lambda }_j`$ should have an eigenvalue equal to one. Hence we obtain the following condition for obtaining delocalized states
$$det\left[_3\mathrm{\Lambda }_j\right]=2\sigma ^2E^2=0.$$
(21)
As we see, there is a delocalized state at $`E=0`$. One can calculate the localization length (11) and expand it around $`E=0`$. For $`v>0`$ one gets
$$\lambda ^1=\{\begin{array}{cc}\frac{\sigma ^2}{4v}E+𝒪(E^2),\hfill & E<0,\hfill \\ 2v^{1/2}E^{1/2}+𝒪(E^{3/2}),\hfill & E>0.\hfill \end{array}$$
(22)
Notice that the localization length is asymmetric around the energy of the extended state since it scales as $`E`$ at the left and $`E^{1/2}`$ at the right. The situation is just the opposite for $`v<0`$. Remarkably, the prefactor at the left depends on the degree of disorder of the alloy but becomes independent of disorder at the right.
When the degree of disorder vanishes ($`\sigma =0`$), the alloy is simply a diatomic periodic chain with site energies $`0`$ and $`v`$ in each unit cell and, consequently, there are two allowed bands. The lower band ranges from $`v/2\left[(v/2)^2+4\right]^{1/2}`$ up to $`0`$ while the upper band ranges from $`v`$ up to $`v/2+\left[(v/2)^2+4\right]^{1/2}`$ for $`v>0`$. Obviously all eigenstates are Bloch functions and spread over the entire chain. Localization occurs as soon as a small degree of disorder is introduced in the system. But, according to our previous results, the eigenstate with $`E=0`$ remains extended. This is clearly seen in Fig. 1, where the inverse of the localization length obtained from (11) is plotted against $`E`$ for $`v=1`$ and different degrees of disorder $`\sigma `$. In all cases the inverse of the localization length is nonzero except at $`E=0`$, where $`\lambda ^1=0`$. This suggests the occurrence of a delocalized states at $`E=0`$.
## IV Numerical results
To confirm the above analytical results we have also numerically diagonalized the Schrödinger equation (1). We will mainly focus our attention on the normalized density of states $`\rho (E)`$ and on the degree of localization (inverse participation ratio, IPR) for the states at energy $`E`$. They are defined respectively as follows
$$\rho (E)=\frac{1}{𝒩}\underset{k}{}\left(\frac{1}{R}\right)\theta \left[\frac{R}{2}|EE_k|\right],$$
(24)
$$L(E)=\frac{1}{𝒩\rho (E)}\underset{k}{}\left(\frac{1}{R}\right)\theta \left[\frac{R}{2}|EE_k|\right]\left(\underset{n=1}{\overset{𝒩}{}}a_{kn}^4\right),$$
(25)
where the angular brackets indicate an average over an ensemble of disordered linear chains and the $`a_{kn}`$ is the eigenvector of (1) corresponding to the eigenvalue $`E_k`$ with $`k=1,2,\mathrm{},𝒩`$. Here $`R`$ is the spectral resolution and $`\theta `$ is the Heaviside step function. The IPR behaves like $`1/𝒩`$ for delocalized states spreading uniformly over the entire system on increasing $`𝒩`$. In particular, the IPR can be exactly computed for the eigenstates of the periodic lattices. In doing so we obtain the expected behavior for $`𝒩\mathrm{}`$. On the contrary, localized states exhibit much higher values. In the extreme case, when the eigestate is localized at a single site, the IPR becomes unity.
We have fixed $`v=1`$ and studied several values of the degree of disorder $`\sigma `$, ranging from $`0.1`$ up to $`1.0`$. The highest value of the degree of disorder considered in the present work means that the typical fluctuations of the site energy is of the order of the nearest neighbor coupling. The maximum number of atoms in the chain was $`𝒩=1000`$ ($`500`$ unit cells) although larger systems were also studied to check that our main results are independent of the size. The results comprise the average over $`100`$ realizations of the disorder for each given pair of parameters $`v`$ and $`\sigma `$. The spectral resolution was $`R=4\times 10^3`$.
Let us comment the results we have obtained numerically. Figure 2 shows the DOS for two different values of the degree of disorder ($`\sigma =0.1`$ and $`\sigma =1.0`$) when the system size is $`𝒩=1000`$ and $`v=1`$. The DOS presents the usual U-shape within the bands when the degree of disorder is small. The singularities at the edge of the allowed bands are smeared out on increasing the degree of disorder except at $`E=0`$, where the divergence remains even for the largest degree of disorder ($`\sigma =1.0`$). This result provides further evidence that the states at $`E=0`$ are delocalized.
The degree of localization (IPR) presents an overall increase when the degree of disorder increases, meaning that the larger the degree of disorder, the smaller the localization length. This is clearly observed in Fig. 3, where we show the IPR as a function of the energy for the same parameters of Fig. 1. However, the increase of the IPR strongly depends on the energy, being more pronounced close to the center of both allowed bands. Interestingly, the IPR at $`E=0`$ becomes independent on the degree of disorder although depends on the system size, as expected. This peculiarity manifest the delocalized character of the states at this spectral region. Finally, notice the good correspondence between Fig. 1 (analytical results) and Fig. 3 (numerical results).
## V Conclusions
In this paper we have considered electron dynamics in a one-dimensional model of binary alloy where disorder lies in one of the two sublattices. Although the system is purely one-dimensional and random, we have demonstrated analytically the existence of delocalized states close to a resonant energy, which match that of the atoms of the other sublattice. Numerical results from the evaluation of the DOS and IPR (degree of localization) strongly suggest that there exist many states close to the resonant energy that remain unscattered in finite systems.
###### Acknowledgements.
Work at Madrid is supported by CAM under Project 07N/0034/98. Work at Yerevan is supported by INTAS grant. |
no-problem/0001/astro-ph0001092.html | ar5iv | text | # Hubble Space Telescope imaging and ground-based spectroscopy of old nova shells – I. FH Ser, V533 Her, BT Mon, DK Lac, V476 Cyg
## 1 Introduction
Every classical nova outburst should result in the ejection of $`10^5`$ to $`10^4M_{}`$ of material at velocities of the order of hundreds to thousands of kilometres per second (Bode & Evans 1989). Study of these expanding shells of nova ejecta has importance for a range of astrophysical areas including the physics of thermonuclear reactions, mixing mechanisms, nebular shaping, radiation-driven winds, clumping mechanisms, astrophysical chemistry and the formation of dust. Novae provide a real-time laboratory in which these processes can be investigated. The work presented in this series of papers builds on our ground-based imaging survey of the ejecta of old novae using 4-m class telescopes (Slavin, O’Brien & Dunlop 1995, Gill & O’Brien 1998) in two ways. Firstly we present high-resolution imaging obtained with the Wide Field Planetary Camera 2 (WFPC2) on the Hubble Space Telescope (HST). This is then complemented by kinematical information obtained from spatially-resolved spectroscopy using the William Herschel and Anglo-Australian Telescopes (WHT & AAT respectively).
This first paper presents the HST images and medium-resolution long-slit WHT spectra of the shells of FH Ser and V533 Her in the region of H$`\alpha `$. We also present non-detections in the HST imaging of BT Mon, DK Lac and V476 Cyg. §2 describes the observations and §3 presents the results. The data are interpreted and discussed in §4. The corresponding observations of HR Del, RR Pic, T Aur, V1500 Cyg and V842 Cen will be presented in papers to follow.
## 2 Observations
Hubble Space Telescope (HST) observations of the novae were made during 1997/98 – see Table 1. In each case the nova was positioned on the Planetary Camera chip of WFPC2. The pixel scale is 46 mas with a complex point spread function distributing 60% of the flux from a point source over the 3$`\times `$3 pixel region centred on the source.
The standard pipeline data processing was used (Leitherer 1995) and the images combined eliminating cosmic rays and bad pixels using standard tasks in STSDAS/IRAF. Narrow band photometry was performed using the prescription described by Dudziak and Walsh (1997). For the images where there was no immediately obvious detection of an ejected shell we applied a 7$`\times `$7 sliding-box median-filter to the images to reduce the effects of noise and contamination from low level cosmic rays. This process enhanced the slower varying underlying background which is more consistent with an extended nova shell and thus allowed visual detection of shells to be investigated.
Long-slit spectra of the shells of FH Ser and V533 Her were taken on the 4.2m WHT in La Palma, Canary Islands using the ISIS spectrometer on 1996 Aug 3 – see Table 1. This instrument has the advantage of having both a blue and red arm which allows simultaneous observations of two wavelength ranges with a suitable dichroic. Using the TEK-1 and TEK-2 chips with the R1200B and R1200R gratings this instrument has a spectral range of 420Å and a dispersion of 0.41Å/pixel in both arms. Observations were taken with central wavelengths of $`\lambda \lambda `$ 4950 and 6600Å for the blue and red arms respectively to allow primarily the lines of H$`\alpha `$ and H$`\beta `$ ($`\lambda \lambda `$ 6562.8 and 4861.3Å respectively) to be examined as well as \[O III\] ($`\lambda \lambda `$ 4959 and 5007Å), \[N II\] ($`\lambda \lambda `$ 6549 and 6583Å) and He I ($`\lambda \lambda `$ 4921, 5015 and 6677Å). The dispersion along the slit was 0.36 arcsec per pixel. Slits of width 1 arcsec were used at position angles chosen to investigate features revealed in previous imaging. The spectra were reduced using standard techniques.
## 3 Results
### 3.1 FH Ser
FH Ser (Nova Ser 1970) erupted in 1970 and was discovered by Honda (1970). It reached a peak visual magnitude of 4.4 on Feb 18 (Burkhead et al. 1970, Borra & Andersen 1970). The lightcurve shows a significant DQ-Her-type dip (e.g. Fig. 2 of Rosino et al. 1986) and is characterized by a $`t_3`$ time of 62 days (Duerbeck 1987). Recently it has been classifed as a ‘slow’ nova belonging to the Fe II spectroscopic class (Della Valle & Livio 1998, Williams 1992). Duerbeck (1992) produced the first image of the shell from which he measured a radius of 2.7 arcsec and described an equatorial band of enhanced emission. He derived an orbital inclination of 58 and a distance of 850 pc. Later images were taken by Slavin at al. (1995) in 1993 September. Their images were deeper and in several narrow wavebands covering approximately $`\lambda \lambda `$ 6530 to 6608Å. They also resolved the equatorial enhancement and derived an inclination angle of $`58\pm 14^{}`$. They suggested that the equatorial band is due to H$`\alpha `$ emission from a density enhancement.
Della Valle et al. (1997) imaged this nova shell in 1996 March with a H$`\alpha `$+\[N II\] filter. They showed the shell to be elliptical with outer diameters of 7.6 and 9 arcsec with a peak-to-peak minor axis of 5.4 arcsec. No peak-to-peak major axis was given. They also showed that the shell had an equatorial ring. They derived an inclination angle of 60 and a distance of 870$`\pm `$90 pc.
#### 3.1.1 Imaging
The FH Ser remnant was imaged with HST on 1997 May 11 with two 1200s integrations, Fig. 1. The maximum pixel value attributable to the shell is 10.9 counts (a 7-$`\sigma `$ detection corresponding to a surface brightness of 9.1$`\times 10^{15}`$ erg cm<sup>-2</sup> s<sup>-1</sup> arcsec<sup>-2</sup>). The shell takes the form of a very clumpy limb-brightened prolate ellipsoid. There is a clear equatorial ring which is brighter in the western half compared to the eastern. If we assume this equatorial enhancement is an inclined circular ring then the derived inclination angle is 62$`\pm `$1. The apparent major axis is 7.0$`\pm `$0.3 arcsec and the minor axis 5.8$`\pm `$0.1 arcsec (measured peak to peak). When this ellipsoid is de-projected the true axial ratio of the shell is found to be 1.26$`\pm `$0.08.
#### 3.1.2 Spectroscopy
The long-slit spectra for both slit positions for the region of 6520–6610Å are shown in Fig. 2. Both slits show velocity ellipses of the shell in the lines of H$`\alpha `$ ($`\lambda `$ 6562.8Å) and \[N II\] ($`\lambda `$ 6549Å and $`\lambda `$ 6583Å). The only other detection of emission was from H$`\beta `$ which has the same appearance as that of the H$`\alpha `$ ring. The two slit positions have been displayed using the same high and low levels to allow relative comparison of fluxes (it should be noted that the run was not photometric but a first order comparison is possible).
In slit position 1 the velocity ellipse in H$`\alpha `$ of the shell can clearly be seen between the two fainter \[N II\] ellipses. The H$`\alpha `$ ellipse shows clumpiness on small scales super-imposed on a well defined velocity ellipse. The \[N II\] ellipses are both barely detected apart from a feature above (blue-shifted) and below (red-shifted) the central continuum source corresponding to the equatorial ring seen in the HST image. For the $`\lambda `$ 6549Å line these features appear at $`\lambda `$$``$6538Å and 6555Å (within the H$`\alpha `$ ellipse), and for the $`\lambda `$ 6583Å line they appear at $`\lambda `$$``$6573Å (co-incident with the red-shifted H$`\alpha `$ ellipse) and 6591Å. The fact that the equatorial ring feature above the stellar continuum (the west) is blue-shifted and the feature below is red-shifted indicates that the shell is inclined so that the western part of the shell is tilted away from the observer.
In slit position 2 the velocity ellipses are quite different to those in slit 1. The northern and southern halves look the same and the brightness of the \[N II\] $`\lambda `$ 6583Å ellipse is comparable to that from the H$`\alpha `$ ellipse. The H$`\alpha `$ ellipse seems to be of fairly uniform intensity whereas the \[N II\] ellipse is brighter near the extremities of the shell (top and bottom of the slit). By inspection of the image in Fig. 1 it can be seen that the edges of the shell correspond with the region where the equatorial ring is fully within the slit. The brightest region in the \[N II\] $`\lambda `$ 6583Å ellipse is brighter than any region in the H$`\alpha `$ ellipse.
### 3.2 V533 Her
V533 Her (Nova Herculis 1963) was discovered by Dahlgren (1963) and extensively studied by Chincarini & Rosino (1964). It declined with a $`t_3`$ time of 44d (see Duerbeck 1981 using AAVSO light curves). However combining the V light curve presented by Chincarini (1964) which shows that it declined to 6th magnitude sometime between March 4 and 8 with the post-discovery results (see Haddock et al. 1963) that it actually reached magnitude 3 on Jan 30 shows it may be more accurate to quote a $`t_3`$ time of around 35d. It has been classified as a ‘slow’ nova belonging to the Fe II spectroscopic class (Della Valle & Livio 1997). The nebulosity around V533 Her was first detected by Cohen & Rosenthal (1983) by comparison with a point spread function. The first clear image was provided by Slavin at al. (1995) who detected a smooth circular shell, radius $``$4.5 arcsec, with a thick equatorial band. This image is reproduced at the left of Figure 3.
#### 3.2.1 Imaging
V533 Her was imaged with HST for a total of 2600s with 2 cosmic ray split observations on 1997 September 3. The median smoothed image of V533 Her is shown at the right of Figure 3. An approximately circular shell has been detected with radius $``$5 arcsec, consistent with unifirm expansion since the Slavin et al. epoch. The brightest knot at the bottom of the median-smoothed image of the shell has a maximum value of 1.8 counts per pixel corresponding to a 7$`\sigma `$ detection at a surface brightness of 1.3$`\times 10^{15}`$ erg cm<sup>-2</sup> s<sup>-1</sup> arcsec<sup>-2</sup>. The signal to noise in the HST image is however not sufficient to provide any more information than was already apparent from the earlier ground-based image.
#### 3.2.2 Spectroscopy
WHT ISIS spectra were obtained on 1996 Aug 3. Two slit positions were used; slit 1 at PA 160 along the equatorial ring and slit 2 at PA 70 orthogonal to the equatorial ring. Both used a slit of width 1 arcsec. The two positions are indicated on the WHT image in Figure 3 reproduced from Slavin et al. (1995) but rotated to correspond to the orientation of the HST image shown alongside.
The spectra in the region of H$`\alpha `$ are shown in Figure 4 (the H$`\beta `$ range is not reproduced here as, although the shell is just visible, the signal to noise is very poor). From these spectra we see that the radius of the shell at the first slit position is $`4.5\pm 0.5`$ arcsec whilst at the second slit position this increases to $`5.5\pm 0.5`$ arcsec. Hence there is some evidence that the shell is elliptical with an apparent axial ratio of $`1.2\pm 0.2`$. The major axis is orthogonal to the equatorial band which is consistent with a prolate ellipsoidal shell as observed in other novae e.g. FH Ser. We are unable to estimate an inclination and hence a true axial ratio as the equatorial band is not clearly detected in the HST image. The maximum line of sight velocity (obtained from the points where the slits cross at the centre of the nebula) is estimated to be $`850\pm 150`$ km s<sup>-1</sup>. Expansion parallax, taking into account the estimated errors in the shell diameter and expansion velocity, leads to a distance estimate of $`1250\pm 300`$ pc.
### 3.3 BT Mon
BT Mon reached maximum in 1939 September. The $`t_3`$ time is disputed in the literature due to the exact date of maximum light being missed. Payne-Gaposchkin (1964) quotes a $`t_3`$ of 36 days calculated from spectral data whereas Schaefer & Patterson (1983) quote a $`t_3`$ of 190 days from the light curve after maximum. The shell has been detected around this nova in spectroscopic observations by Marsh et al. (1983). The only image of the shell is that by Gill & O’Brien (1998). They claimed detection of a ring of emission of diameter $``$7 arcsec after deconvolution of the image with various techniques.
BT Mon was imaged with HST in two observations for a total of 2500s. The median-smoothed image of BT Mon is shown in Fig. 5. No shell is detected putting a 3$`\sigma `$ upper limit on the surface brightness of the shell at 6$`\times 10^{16}`$ erg cm<sup>-2</sup> s<sup>-1</sup> arcsec<sup>-2</sup>.
### 3.4 V476 Cyg
V476 Cyg (Nova Cygni 1920) reached maximum on 1920 August 24. The measured $`t_3`$ time for its decline was 16.5d (Duerbeck 1987) classifying it as a very fast nova. A diffuse shell of radius $`5`$ arcsec was detected by Slavin et al. (1995). With their integration time of 900s it was barely detected above the 3 sigma level.
V476 Cyg was imaged with the HST for a total of 2600s (cosmic ray split by a factor 2). The median-smoothed image of V476 Cyg is shown at the top left of Fig. 5 where no nova shell is detected. This provides a 3$`\sigma `$ upper limit to the shell surface brightness of 5$`\times `$10<sup>-16</sup> erg cm<sup>-2</sup> s<sup>-1</sup> arcsec<sup>-2</sup>.
### 3.5 DK Lac
DK Lac (Nova Lacertae 1950) was discovered during its rise in 1950 January. The nova declined at a moderately fast rate with a $`t_3`$ of 32d, accompanied by large variations of brightness. Slavin et al. (1995) detected the shell around DK Lac, measuring is radius as 2.0 to 2.5 arcsec, although the shell is poorly distinguished from the central source.
The median-smoothed HST image of DK Lac (total integration time 2500s) is shown at the bottom left of Fig. 5. No shell is detected providing a 3$`\sigma `$ upper limit on the shell surface brightness of 6$`\times 10^{16}`$ erg cm<sup>-2</sup> s<sup>-1</sup> arcsec<sup>-2</sup>.
## 4 Further Interpretation of the FH Ser Results
### 4.1 Observations
The \[N II\] features seen in the spectrum from the first slit position (S1) directly above and below the central source and the \[N II\] features at the extremities of the shell in the spectrum from the second slit position (S2) are associated with the equatorial ring clearly seen in the image. We would expect H$`\alpha `$ to mirror this behaviour if the ring results from an enhancement in density. However it can clearly be seen in S2 that H$`\alpha `$ is not greatly enhanced in the ring at all. The \[N II\] ring features in S1 are approximately 40 times greater than the typical \[N II\] flux in the rest of the shell; although the H$`\alpha `$ ellipse is rather clumpy in S1, there is no obvious comparative increase in brightness in the corresponding regions for the ring. Therefore the equatorial ring is dominated by enhanced emission in \[N II\].
This picture is supported by the images in Slavin et al. (1995). In their figure 3c they show an image of the FH Ser remnant taken with a filter central wavelength $`\lambda `$ 6560Å, FWHM 17Å. From examining the long slit spectra shown in this paper we can see that this will include almost all light in H$`\alpha `$ apart from the parts of the shell moving directly towards or away from us. There should be no \[N II\] contamination. This image shows only a smooth ring of emission resembling a simple limb-brightened shell. There is no evidence in this image for an equatorial enhancement. Slavin et al. also show an image with filter central wavelength $`\lambda `$ 6537Å, FWHM $`>`$ 17Å (their figure 3a). From our spectra we can see that this corresponds primarily to blue-shifted \[N II\] $`\lambda `$ 6549Å. This image shows a half-crescent of emission with very little emission in the rest of the shell which again implies the equatorial enhancement is an \[N II\] feature. This crescent of emission is to the west of the central source which agrees with our suggestion that the eastern pole is tilted towards us.
### 4.2 Synthetic Images and Spectra
Using the code described in Gill & O’Brien (1999) it is possible to produce synthetic images and spectra for simple models of the FH Ser nova shell. The code calculates emission as a function of density and therefore separate models needed to be produced for the H$`\alpha `$ and \[N II\] emission lines. The lines can be scaled relative to one another by arbitrary factors and then combined, taking into account the rest wavelengths of the lines and the Doppler shifting of each element, to produce a spectral cube for a given wavelength range.
To generate the synthetic spectra the spectral cube was then smoothed in the spatial directions with the equivalent of a 0.5 arcsec FWHM Gaussian to simulate the seeing of the ground-based observations. A ‘slit’ was then positioned on the cube at positions and widths to correspond with the observations. Smoothing must be performed before the slit is positioned onto the cube to allow light to be scattered into and out of the slit. The possibility of correction for a systemic velocity for any model was incorporated to allow a best fit to the data.
Synthetic images were generated by convolving the unsmoothed spectral cubes with the system throughput of the HST with the F656N filter (Leitherer 1995). This was approximated by a profile starting with zero throughput at $`\lambda `$ 6545Å, rising linearly to peak at $`\lambda `$ 6554Å, remaining constant to $`\lambda `$ 6576Å, then falling linearly back to zero at $`\lambda `$ 6582Å. The images were then generated by collapsing the filtered spectral cube in the spectral direction.
We have generated models corresponding to a thin prolate ellipsoidal shell with ellipticity 1.3 and inclination 62. For the \[N II\] components an equatorial enhancement by a factor of 40 in intensity was also added in a thin ring. The light from H$`\alpha `$ was increased by a factor of 8.6 relative to the typical (i.e. not the enhanced value appropriate to the ring) \[N II\] $`\lambda `$ 6583Å line (to best fit the observations), and the \[N II\] $`\lambda `$ 6549Å scaled by a factor 0.34 relative to the \[N II\] $`\lambda `$ 6583Å line. Velocities were set proportional to the distance from the centre of the ellipsoid and the maximum line of sight velocity was set to be 530 km/s. For the model spectra to best fit the observations it was found that a systemic velocity of $``$45 km/s was required. The resulting synthetic spectra for positions S1 and S2 can be seen in Fig. 7 and the synthetic image in Fig. 6. The model agrees closely with the observations shown in Fig. 2. This also allows us to derive a distance to FH Ser of 950$`\pm `$50 pc which takes full account of the asphericity of the shell.
If we examine the observed image in Fig. 1 we see that the equatorial ring is brighter in one half than the other, behaviour replicated in our synthetic image in Fig. 6. Our model shows that this is because the system throughput of the HST with the F656N filter lets through the red-shifted $`\lambda `$ 6549Å light and the blue-shifted $`\lambda `$ 6583Å light. Therefore the ring seen in the image isn’t actually a whole ring from one line but two halves of the ring each from a different line. The ratio of the lines in this doublet is always $`[6549]/[6583]=0.34`$, so the half of the ring visible in the image which is produced by the $`\lambda `$ 6583Å line is 3 times brighter than the $`\lambda `$ 6549Å half.
### 4.3 The origin of the \[N II\] ring
For the equatorial ring to be enhanced in \[N II\] relative to the rest of the shell requires either a density, ionization or metallicity effect.
A simple density increase would produce more \[N II\] (providing that the level did not start to become collisionally de-excited) but also more H$`\alpha `$ and is therefore presumably not the solution.
An increase in the amount of nitrogen in the equatorial plane of the FH Ser shell could explain the \[N II\] enhancement. There is the possibility of a metallicity variation across the surface of the white dwarf before eruption due to rotation of the white dwarf although it is difficult to see how a thin enhancement of nitrogen could be produced rather than a more wide-spread variation around the shell.
The final possibility is an ionization effect which would require either an increase or decrease in ionization in the equatorial region. A good candidate for providing such an effect is the accretion disc around the central source. An explanation could be that the main shell is very highly ionized, doubly ionized or above, by the central parts of the accretion disk. The equatorial region is then shadowed from this hot central UV source, causing the relative abundance of nitrogen in the singly ionized state to be increased. This would then produce an increase in \[N II\] emission in a thin equatorial region with no increase in H$`\alpha `$. This proposal needs testing by further observation and modelling. A similar suggestion was made by Petitjean et al. (1990) for the shell of DQ Her.
## 5 Conclusions
HST WFPC2 images and WHT spectroscopy of FH Ser have allowed us to derive a consistent model for its shell as an ellipsoid with ellipticity 1.3$`\pm `$0.1, and inclination 62$`\pm `$4. The equatorial expansion rate is found to be 490$`\pm `$20 km s<sup>-1</sup> and the systemic velocity $`45`$ km s<sup>-1</sup>. This gives a distance to FH Ser of 950$`\pm `$50 pc. These values agree within the errors with the results presented in previous papers (Della Valle et al. 1997, Slavin et al. 1995, Duerbeck 1992). The equatorial ring seen in the HST images and previous ground-based images is found to be due to a 40$`\times `$ increase in emission from \[N II\] rather than H$`\alpha `$+\[N II\]. It is suggested that the origin of this ring lies either in abundance gradients in the shell or in aspherical illumination of the shell by the central photoionizing source. This latter suggestion will be developed in a paper to follow.
Similar imaging and spectroscopy of the nova V533 Her reveal a shell of radius $``$5 arcsec with an axial ratio of $`1.2\pm 0.2`$. The expansion velocity of this shell is $`850\pm 150`$ km s<sup>-1</sup> and the distance is estimated to be $`1250\pm 300`$ pc. The quality of the data is however not sufficient for us to examine the structure of the shell in as much detail as was possible for the shell of FH Ser.
Our HST images of BT Mon, V476 Cyg and DK Lac provided only upper limits on the surface brightness of their shells. This may seem surprising given that these shells had been previously detected from the ground (Gill & O’Brien 1998, Slavin et al. 1995). However it should be noted that the ground-based images were taken using the large aperture 4.2m WHT and 3.9m Anglo-Australian Telescope. Although the improved spatial resolution of HST helps overcome its reduced collecting area when the nebula is clumpy and these clumps are unresolved from the ground this does not apply when the emission is more smoothly distributed. These nova shells were not particularly well-detected in the ground-based survey (BT Mon requiring image deconvolution) so it is perhaps not too surprising that we did not detect them here.
Other novae to be investigated in the following papers of this series include HR Del, RR Pic, T Aur, V842 Cen and V1500 Cyg.
## 6 Acknowledgements
We would like to thank Dr M. Della Valle for valuable comments at the refereeing stage. CDG would like to thank PPARC for his studentship, and Nial Tanvir and Rachel Johnson for their help with the reduction of the HST images. This work is based on observations 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 No. NAS5-26555) and the William Herschel Telescope (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). |
no-problem/0001/astro-ph0001246.html | ar5iv | text | # References
The Chandrasekhar limit for quark stars
Shibaji Banerjee<sup>1</sup><sup>1</sup>1email: phys@bosemain.boseinst.ernet.in, Sanjay K. Ghosh<sup>2</sup><sup>2</sup>2email: phys@bosemain.boseinst.ernet.in and Sibaji Raha<sup>3</sup><sup>3</sup>3email: sibaji@bosemain.boseinst.ernet.in
Department of Physics, Bose Institute,
93/1, A. P. C. Road, Calcutta 700 009, INDIA
## Abstract
The Chandrasekhar limit for quark stars is evaluated from simple energy balance relations, as proposed by Landau for white dwarfs or neutron stars. It has been found that the limit for quark stars depends on, in addition to the fundamental constants, the Bag constant.
Compact objects like the white dwarfs and neutron stars have been a topic of interest for several decades. These objects are produced as the end product of the stellar evolution, i.e. , when the nuclear fuel of the normal stars have been consumed. White dwarfs are supported against gravitational collapse by the degeneracy pressure of electrons whereas for neutron stars, this pressure comes mainly from the degenerate neutrons. It is well known that both these classes of collapsed stars have a maximum mass ( $`1.5M_{}`$ ), beyond which they collapse to black holes . Although the Chandrasekhar limit refers strictly to white dwarfs, the limiting mass for neutron stars is also loosely called the Chandrasekhar limit, primarily because the limits in the two cases turn out to be the same .
The underlying quark structure of the nucleons suggests a possible hadron - quark phase transition at high density and/or temperature. This, in turn, implies the possible existence of another kind of compact objects, namely quark stars . The suggestion of Witten that the strange quark matter may be the true ground state of the strongly interacting matter started a new era in this subject. In this circumstance, quark stars, if they are formed, would preferably convert to strange stars, comprising u, d and s quarks, under weak interaction. Several authors ( for example, ) have used different models to understand the properties of strange stars. For a review, see .
The unique feature of strange stars, which separates them from other compact objects, is that these need not be the direct product of stellar evolution. These are most likely to be produced due to the hadron quark phase transition inside the neutron stars as the density is high enough to favor such a transition. For such quark stars, the maximum mass would indeed be almost the same as that for neutron stars. On the other hand, it is also conceivable that if a large amount of quark matter exists in the universe as a relic of the cosmological quark-hadron phase transition , it could clump under gravitational interaction and even form invisible quark galaxies . The ‘stars’ of such a galaxy would be strange stars which do not evolve from neutron stars and thus are not governed by the Chandrasekhar limit for neutron stars. It is therefore most pertient to ask if there exists, just like the case of ordinary compact stars, an upper limit on the strange stars beyond which they would be gravitationally unstable. Starting with the seminal work of Witten , most authors have concentrated on solving the Tolman-Oppenheimer-Volkov (TOV) equation (see, for example, ) numerically for the quark matter equation of state. While the results show that indeed there exists a limiting mass for quark stars (which is very close to that for neutron stars), there is no a priori argument to prove that such a limit should exist or that it should depend mostly on fundamental constants, as is the case for the ordinary compact stars . Our aim in this letter is to show analytically, from first principles , that such a limit exists for compact quark stars and that it is indeed determined mostly by universal constants.
We start with the most general and simple picture of energy balance, proposed by Landau . The strange star is composed of a 3 flavour system of massless quarks, confined in a large bag characterized by a constant energy density $`B`$. As in the case of white dwarfs and neutron stars, the equilibrium should occur at a minimum of the total energy per fermion $`e`$, where $`ee_F+e_G`$, in which $`e_F`$ is the fermi energy and $`e_G`$ is the gravitational energy per fermion. The crucial difference between the cases of quark stars and the ordinary compact stars lies in the estimation of the (Newtonian) gravitational energy, considered to be a macroscopic quantity. For ordinary compact stars, this mass is due almost entirely to the baryons. For quark stars, however, the total mass is the total ( thermodynamic as well as the confining ) energy in the star. For the purpose of estimating the gravitational energy per fermion, one can then define an effective quark mass which incorporates both these contributions. A suitable prescription for such a definition was formulated quite some time ago .
The number density of fermions is related to the chemical potential as
$$n=\frac{g}{6\pi ^2}\mu ^3$$
which dictates that
$$\mu =(\frac{9\pi }{2g})^{\frac{1}{3}}\frac{N^{\frac{1}{3}}}{R}$$
(1)
In the above relations, $`n`$ is the number density, $`N`$ the total number of fermions in a star of radius $`R`$, $`g`$ the statistical degeneracy factor and $`\mu `$ is the chemical potential.
The fermion energy density is given by
$$\epsilon _F=\frac{g}{8\pi ^2}\mu ^4$$
(2)
and hence the fermi energy per particle of the quarks becomes
$$e_F=\frac{\epsilon _F}{n}=\frac{3}{4}(\frac{9\pi }{2g})^{\frac{1}{3}}\frac{N^{\frac{1}{3}}}{R}$$
(3)
The mass $`M`$ of the star can be written in terms of $`N`$ and $`B`$ (the bag constant), if the density $`\rho (r)`$ in the star is assumed to be roughly constant throughout the volume of the star. Hence using eq.(3),
$$M=_0^R4\pi r^2\rho (r)𝑑r=\frac{4}{3}\pi R^3B+e_FN=\frac{3}{4}(\frac{9\pi }{2g})^{\frac{1}{3}}\frac{N^{\frac{4}{3}}}{R}+\frac{4}{3}\pi BR^3$$
(4)
Extremising the mass $`M`$ (eq. 4) with respect to $`R`$ gives,
$$(\frac{9\pi }{2g})^{\frac{1}{3}}\frac{N^{\frac{4}{3}}}{R^4}=\frac{16}{3}\pi B\epsilon _F=3B$$
Substituting eq.(S0.Ex2) in the expression for $`M`$ (eq. 4),
$$M=4BV=\frac{16}{3}\pi BR^3$$
(5)
Note that this is very similar to the condition obtained for hadronic bags . The task at hand then is to find the $`R`$ for which the total energy per fermion would be maximum.
The gravitational energy per fermion $`e_G`$ is
$$e_G=\frac{GMm_{eff}}{R}$$
(6)
where $`m_{eff}`$ is the effective quark mass inside the star. Assuming that the effective quark mass contributes to the total star mass $`M`$, one can write for a strange star with $`N`$ quarks,
$$M=Nm_{eff}m_{eff}=\frac{4B}{n}$$
(7)
As mentioned above, the effect of confinement in a quark matter system was shown to be incorporable in the effective quark mass, which, the quarks being fermions, coincides with the quark chemical potential . As a result, one gets, in the limit of vanishing quark density ,
$$\mu =\frac{B}{n}$$
. This, together with the eq.(7), gives
$$m_{eff}=4\mu $$
(8)
where all the energy ( thermodynamic and confining ) is included in the effective gravitational mass of the quarks inside the strange star.
Using equations (6), (7) and (8) we get
$$e_G=\frac{64}{3}(\frac{9\pi }{2g})^{\frac{1}{3}}G\pi BRN^{\frac{1}{3}}$$
(9)
Minimising the total energy $`e=e_F+e_G`$ with respect to $`N`$, we get the expression for maximum value of $`R`$ as
$$R_{max}=\frac{3}{16}\frac{1}{\sqrt{\pi GB}}$$
(10)
Finally, the maximum mass of the strange star is computed by substituting the value of $`R_{max}`$ (from equation 10) in equation ( 5).
$$M_{max}=\frac{16}{3}\pi BR_{max}^{}{}_{}{}^{3}$$
(11)
The chemical potential $`\mu `$ can be evaluated in terms of $`B`$ using equations (2) and (S0.Ex2). Substituting this in eq.(7) gives the value of $`N_{max}`$. The values of $`R_{max}`$, $`M_{max}`$ and $`N_{max}`$ are tabulated below for various values of the Bag constant $`B`$.
| $`B^{1/4}(MeV)`$ | $`R_{max}(Km)`$ | $`\frac{M_{max}}{M_{}}`$ | $`N_{max}`$ |
| --- | --- | --- | --- |
| 145 | 12.11 | 1.54 | 1.55 $`\times `$ 10<sup>57</sup> |
| 200 | 6.36 | 0.81 | 5.90 $`\times `$ 10<sup>56</sup> |
| 245 | 4.24 | 0.54 | 3.21 $`\times `$ 10<sup>56</sup> |
We thus find that even for quark stars, there does exist a limiting mass, the so-called Chandrasekhar limit, which is mostly determined by the universal constants ( $`G`$ as well as $`\mathrm{}`$ and $`c`$, which do not occur explicitly due to our use of the naturalised units ) and the Bag energy $`B`$. Although treated as a parameter here, the Bag energy too has, roughly, the status of another universal constant, being the difference between the non-perturbative and the perturbative vacua of Quantum Chromodynamics. The physical radius $`R_{max}`$, corresponding to the maximum mass as well as the maximum mass itself, are independent of the number of flavours, as seen from equations (10, 11). Although $`N_{max}`$ depends on the statistical degeneracy factor $`g`$ ( or equivalently, the number of flavours), the dependence is extremely weak, as can be readily checked from equation (1). In fact, we have verified that there is almost no difference in $`N_{max}`$ between the cases with $`g`$ = 2 and 3. This, in turn, implies that the assumption of massless quarks ( even for $`s`$ quarks ) does not affect these results.
Admittedly, our purpose here has been mostly illustrative. Nevertheless, these limits agree very well with those found in the numerical solutions of the TOV equation (see, for example, ). Even under the simplifying assumption of a constant density profile, necessary for an analytical solution, the scaling behavior ( $`R_{max}B^{1/2},M_{max}B^{1/2}`$ , obtained from the numerical solutions ) hold true. This proves that the simple picture presented here adequately incorporates the essential physics.
To conclude, we have shown that there exists a limiting mass (the Chandrasekhar limit) for compact quark stars, beyond which they would be gravitationally unstable. As with white dwarfs and neutron stars, this mass depends mostly on universal constants.
The work of SB and SKG were supported in part by Council of Scientific and Industrial Research, Govt. of India. |
no-problem/0001/astro-ph0001110.html | ar5iv | text | # BeppoSAX observations of the nearby low-mass X-ray binary and fast transient SAX J1819.3–2525
## 1 Introduction
The X-ray source SAX J1819.3–2525 was discovered in February, 1999, independently with the Wide Field Cameras (WFCs) on BeppoSAX (In ’t Zand et al. 1999a) and with the Proportional Counter Array (PCA) on RossiXTE (ergo, its alternative designation XTE J1819–254, Markwardt et al. 1999a). The WFC detection involved an hour-long flare with a peak of 80 mCrab on Feb. 20. The first PCA detection occurred during regular scans of the Galactic Center field, on Feb. 18.
For some time there was confusion over the identification of the optical counterpart. The variable star GM Sgr is within the 2′-radius WFC error box of the X-ray source as noted by In ’t Zand et al. (1999a). However, it was later discovered that the actual counterpart is another, previously unknown, variable star only $``$1′ north of GM Sgr: V4641 Sgr (Williams 1999, Samus et al. 1999).
V4641 Sgr was found to show a large outburst in visual magnitude $`m_\mathrm{V}`$ from about 11 on Sep. 14.8 up to 8.8 on Sep. 15.4 (Stubbings 1999). X-ray observations with the All-Sky Monitor on RossiXTE also showed a bright outburst of SAX J1819.3–2525 up to a peak intensity of 12.2 Crab on Sep. 15.7 (2-12 keV, Smith et al. 1999). The giant X-ray outburst was preceded by a 4.5 Crab precursor 0.8 d earlier. Between both peaks the flux decreased to a quiescent level. Wijnands & Van der Klis (2000) report that during the tail of the giant X-ray outburst, strong variability was observed of factor-of-four on seconds time scales to factor-of-500 on minutes time scales. The giant outburst was also seen at radio wavelengths (Hjellming et al. 1999a, 1999b). The radio source was resolved and had the appearance of a twin jets structure extending 0.25″ each jet. The giant and precursor outburst were, furthermore, seen in 20-100 keV (McCollough et al. 1999).
Within 2 d after the giant outburst, V4641 Sgr settled down to a quiescent brightness of $`m_\mathrm{V}=13.5`$. Combined with the failure of any X-ray detection since the giant outburst (C.B. Markwardt, priv. comm.), this shows that the flaring activity faded considerably, although the Balmer lines continued to show additional activity up to at least Sep. 30 (Wagner 1999). If one assumes that the emission is due to a main sequence A star (Garcia et al. 1999), one obtains with B–V=–0.2 and E(B–V)=0.24 (Wagner 1999) a distance of 0.4 kpc. We will adhere to this distance in the present paper but acknowledge its uncertainty. From the magnitudes alone, we estimate the uncertainty to be about 50%. With a 0.4 kpc distance, the giant outburst would have a peak X-ray luminosity of $`5\times 10^{36}`$ erg s<sup>-1</sup> in 2-10 keV. This luminosity, and the spectrum discussed in the present paper, clearly suggest that there is a compact object at play that is either a neutron star or black hole. It also identifies the object as a low-mass X-ray binary (LMXB), and the short distance implies it is possibly the nearest LMXB known so far (c.f., Van Paradijs & White 1995). This opens up interesting perspectives for SAX J1819.3–2525 as a case study of the quiescent emission from transient low-mass X-ray binaries. We note that recently Rutledge et al. (1999) introduced diagnostics for the nature of the compact object from the quiescent emission.
In this paper, we present the observations of SAX J1819.3–2525 with the WFCs and Narrow Field Instruments (NFI) on BeppoSAX in the spring and fall of 1999. This includes the longest semi-continuous observation of the system so far. SAX J1819.3–2525 was observed in two states: a flaring state with peak fluxes that are one to two orders of magnitude less than that of the giant outburst, and a ’calm’ state with an average flux that is about three orders of magnitude smaller (though still considerably higher than that of the quiescent emission).
## 2 Observations
The BeppoSAX platform carries two sets of astrophysical X-ray and $`\gamma `$-ray devices in space (Boella et al. 1997a). One pertains to two identical Wide Field Cameras (WFCs, Jager et al. 1997) that view the sky with $`40\times 40`$ square degrees field-of view in opposite directions with 5′ spatial resolution in the 2 to 26 keV bandpass. The other set includes the Narrow Field Instruments (NFI) that are co-aligned in a direction that is perpendicular to that of both WFCs. The NFI include 2 imaging instruments that are active below 10 keV, the Low-Energy and the Medium-Energy Concentrator Spectrometer (LECS and MECS respectively, see Parmar et al. 1997 and Boella et al. 1997b respectively), and two non-imaging instruments that have bandpasses of $`12`$ to 300 keV (the Phoswich Detector System or PDS, Frontera et al. 1997) and 4 to 120 keV (the High-Pressure Gas Scintillation Proportional Counter or HP-GSPC, Manzo et al. 1997). The MECS has a photon energy resolution of 8% (full width at half maximum) at 6 keV.
Since mid 1996, the WFCs carry out a long-term program of monitoring observations in the field around the Galactic center. The program consists of campaigns during the spring and fall of each year. Each campaign lasts about two months and typically comprises weekly observations. SAX J1819.3–2525 was detected twice during these campaigns so far, in hourly exposures above a $``$50 mCrab detection limit, on Feb. 20 and Sep. 10, 1999. The first WFC-detection triggered a target-of-opportunity observation (TOO) with the NFI on March 13.22-14.02, 1999 (this is 186 d before the giant outburst). SAX J1819.3–2525 was strongly detected in three NFI (the HP-GSPC was not turned on), and the average intensity was found to be about 12 mCrab (2-10 keV). The NFI net exposure times on SAX J1819.3–2525 are 10.2 ks for LECS, 27.7 ks for MECS and 14.7 ks for PDS. The LECS and MECS images show a single bright source, the position as determined from the MECS image is R.A. 18<sup>h</sup>19<sup>m</sup>22$`\stackrel{}{.}`$2, Decl. = –25°24′03″ (Eq. 2000.0, error radius 0$`\stackrel{}{.}`$8) which is 1$`\stackrel{}{.}`$0 from that determined with the WFC (In ’t Zand et al. 1999a) and 0$`\stackrel{}{.}`$4 from the optical counterpart V4641 Sgr.
## 3 Light curves
The background-subtracted light curves of the two flares as measured with the WFCs are presented in Figs. 1 and 2. Besides these two occasions, the source was never detected in either short 1-hr or long 24-hr WFC exposures with typical upper limits of 50 and 12 mCrab respectively. The light curves of both detections are characterized by sporadic flaring activity up to 0.2 ct cm<sup>-2</sup>s<sup>-1</sup>, which corresponds to 0.1 Crab units, with a larger and longer flare on Sep. 10 with a peak of 0.3 Crab units. The large flare was above the detection limit for about 6 hr and occurred in the middle of a 34 hr long observation. It started on Sep. 10.0, which is 5.4 d before the giant outburst. Its duration is comparable to that of the giant flare.
The flaring activity on a time scale of hours is in line with the behavior as measured with the Proportional Counter Array (PCA) on RossiXTE (Markwardt et al. 1999b). The PCA covers SAX J1819.3–2525 bi-weekly since Feb. 5, 1999, for individual snapshot exposures of about 60 s with a sensitivity of about 1 mCrab. SAX J1819.3–2525 was first seen on Feb. 18 and subsequently showed an erratic variable behavior until immediately after the giant outburst, with fluxes ranging from 0.5 to 30 mCrab (Markwardt et al. 1999b). Since the giant outburst, no emission was detected anymore (C.B. Markwardt, priv. comm.).
The WFC data do not reveal any type I X-ray burst from SAX J1819.3–2525 in about 0.8 Ms of source coverage for the year 1999 when it was seen to be active with the PCA (Markwardt et al. 1999b), or for $``$3 Ms over all WFC observations since 1996.
Fig. 3 shows the evolution of the background-corrected photon count rates in various bandpasses of the three NFI, in 400 s time resolution. For an explanation of the method of data extraction, we refer to In ’t Zand et al. (1999b). The light curves show slow variability on a time scale of a few hours with an amplitude of about 50%. The average flux level is 12 mCrab (2-10 keV) which, relative to the flares, can be regarded as calm emission. However, we note that this cannot be regarded as quiescent emission because the PCA has seen flux levels from this source at least one order of magnitude smaller (Markwardt et al. 1999b).
A power density spectrum of the MECS photon count rate, generated with a timing resolution of 0.02 s and averaged over 256 s time intervals, reveals no measurable narrow features nor broad-band noise. The upper limit on the variability, integrated between 0.01 and 10 Hz with power laws with indices of –0.5, –1.0 and –2.0, are 18, 8 and 3.5% fractional rms respectively (90% confidence). The upper limit on the pulsed amplitude is 1.3% fractional rms (95% confidence) in 0 to 25 Hz.
## 4 Spectrum of calm emission
We first extracted a spectrum for the average emission over the whole NFI observation. The spectral channels were rebinned so as to sample the spectral full-width at half-maximum resolution by three bins and to accumulate at least 20 photons per bin. The bandpasses were limited to 0.4-4.0 keV (LECS), 1.8-10.5 keV (MECS) and 15-120 keV (PDS) to avoid photon energies where either the spectral calibration of the instruments is not complete or no flux was measured above the statistical noise. We tried to model the spectrum with various descriptions. In all models, an allowance was made to leave free - within reasonable limits - the relative normalization of the spectra from LECS and PDS to that of the MECS spectrum, to accommodate cross-calibration uncertainties in this respect. Publicly available instrument response functions and software were used (version November 1998).
The continuum could best be fitted ($`\chi _\mathrm{r}^2=1.30`$ for 96 dof) with a Comptonization model (Titarchuk 1994) plus black body radiation, see Table 1. Next to that there is a strong emission line at 7 keV. A fit with a single narrow line results in a centroid energy of 6.85$`\pm 0.02`$ keV. We identify this as K$`\alpha `$ fluorescence in strongly ionized iron. The centroid energy is between the expected Fe-K lines for helium-like (6.68 keV) and hydrogen-like Fe (6.96 keV). We included in the model narrow lines at these fixed energies. The best-fit parameter values are given in Table 1 and a graph of the spectrum and the model fit in Fig. 4. The fits with other continuum models, in combination with a black body and two narrow line components results in fit qualities of $`\chi _\mathrm{r}^2=3.19`$ for 98 dof (thermal bremsstrahlung), $`\chi _\mathrm{r}^2=1.83`$ for 96 dof (broken power law) and $`\chi _\mathrm{r}^2=1.59`$ for 96 dof (power law with high-energy cut off).
Terada et al. (1999) reported about the 1996 transient AX J1842.8–0423 which exhibited an Fe-K line at 6.80 keV with a large equivalent width of 4 keV. The 0.5-10 keV continuum plus line spectrum was successfully fitted with a thin hot thermal plasma emission model of temperature 5.1 keV. We fitted such a model according to the MEKAL code implementation (Mewe et al. 1995). The fit was reasonable, provided two additional components were included. With a power law and black body as additional components, $`\chi _\mathrm{r}^2=1.73`$ (98 dof), which is a worse fit than the Comptonization model in Table 1. The resulting plasma temperature is $`9.1\pm 0.5`$ keV. The fitted contribution of the thin plasma to the flux is of order 10%. The emission measure of the thermal plasma is $`(1.33\pm 0.08)\times 10^{56}(d/0.4\mathrm{kpc})^2`$ cm<sup>-3</sup>.
E(B–V)=+0.24 (Wagner 1999) implies $`N_\mathrm{H}=0.13\times 10^{22}`$ cm<sup>-2</sup> according to the relationship defined by Predehl & Schmitt (1995). If we assume that the uncertainty in E(B–V) is 0.10, where most of the uncertainty comes from the uncertainty in the calibration of the relationship used by Wagner (1999) between the equivalent width of the 578.0 nm interstellar absorption line to E(B–V) (see Herbig 1975), then the error in $`N_\mathrm{H}`$ is 0.05$`\times 10^{22}`$ cm<sup>-2</sup>. If we fix $`N_\mathrm{H}`$ to $`0.08\times 10^{22}`$ cm<sup>-2</sup> and leave free the remaining parameters of the Comptonized model in Table 1, $`\chi _\mathrm{r}^2`$ is 1.31 (97 dof). We conclude that $`N_\mathrm{H}`$ as determined from the X-ray spectrum is consistent with E(B–V)=+0.24$`\pm 0.10`$.
To determine whether the variability as illustrated in Fig. 3 is accompanied by strong spectral changes, we extracted a spectrum for times when the source was relatively faint and one for times when the source was relatively bright. These times are indicated by hatched areas in Fig. 3. Subsequently, we employed the same Comptonized model as for the whole observation, leaving free only the normalizations of the different contributions. The resulting values for $`\chi _\mathrm{r}^2`$ are 1.36 for the bright data (104 dof) and 1.22 (103 dof) for the faint data. The 44% difference in the 0.4-10 keV flux between the faint and bright data is due to approximately equal changes in blackbody and Comptonized components (i.e., 25 and 19% respectively). The flux of the emission lines scales with the integral flux: the combined equivalent widths of both lines is identical in both cases at 0.26 keV. This spectral behavior is illustrated in Fig. 5 which zooms in on the MECS part of the spectrum (including the emission lines) for the two extremes. In Fig. 6 the ratio between both spectra is presented. This is consistent with a constant ratio of $`1.44\pm 0.02`$ throughout the spectrum ($`\chi _\mathrm{r}^2=1.18`$ for 65 dof). The apparent bump between 4 and 5 keV is statistically not significant.
## 5 Spectrum of flare
We extracted a 2-26 keV spectrum of the complete Sep. 10 flare which amount to an elapsed time of 5.5 hr and an exposure time of 2.3 hr, and attempted to fit the same Comptonization model as applies to the NFI data when the source is $`20`$ times fainter. Leaving free only the normalizations of the different spectral components, the fit was unsatisfactory with $`\chi _\mathrm{r}^2=2.59`$ for 24 dof. However, if $`N_\mathrm{H}`$ was allowed to vary, the fit became dramatically better with $`\chi _\mathrm{r}^2=0.89`$ (23 dof) and $`N_\mathrm{H}=(2.5\pm 0.4)\times 10^{22}`$ cm<sup>-2</sup>. If instead the black body temperature was allowed to vary, an even better improvement occurred with $`\chi _\mathrm{r}^2=0.64`$ (23 dof) and k$`T_{\mathrm{bb}}=2.4\pm 0.2`$ keV. If the Fe-K emission lines at 6.68 and 6.96 keV are replaced for a single one whose energy is left free, the fit improves further to $`\chi _\mathrm{r}^2=0.50`$ for 23 dof. The line energy then is $`6.39\pm 0.18`$ keV. This suggests a shift of the ionization balance to a lower degree (Fe I-XVII). We determined an average flux over the 3 hr duration of the peak of $`5\times 10^9`$ erg cm<sup>-2</sup>s<sup>-1</sup>. The equivalent width of the emission line is $`1.0\pm 0.3`$ keV. The flare spectrum is presented in Fig. 7.
## 6 Discussion
We have analyzed the spectrum at basically two flux levels which correspond to 2-10 keV luminosities of $`5\times 10^{33}(d/0.4\mathrm{kpc})^2`$ erg s<sup>-1</sup> and $`1\times 10^{35}(d/0.4\mathrm{kpc})^2`$ erg s<sup>-1</sup>, where $`d`$ is the distance to SAX J1819.3–2525. These compare to a maximum luminosity during the giant outburst of $`5\times 10^{36}(d/0.4\mathrm{kpc})^2`$ erg s<sup>-1</sup> (Smith et al. 1999) and a minimum of $`4\times 10^{32}(d/0.4\mathrm{kpc})^2`$ erg s<sup>-1</sup> (see below). These numbers may be multiplied by a factor of 2 to obtain a rough extrapolation of the X-ray luminosity to the 0.4-120 keV range. They show that, though bright, SAX J1819.3–2525 in fact has a peak luminosity that is at least an order of magnitude below the Eddington limit for a mass of the compact object that is equal to or larger than that of a canonical 1.4 M neutron star. However, one should bear in mind that the sampling of the ASM light curve is sparse (about ten 90 s observations per day) and we do not know what the source did between ASM measurements.
The eye-catching feature of the BeppoSAX data is the strong Fe-K emission line complex. The line energies during the calm phase point to a large ionization degree of the iron. Two origins for the line can be considered: thermal emission from a thin hot plasma, where iron atoms are collisionally excited, or fluorescent emission from a plasma illuminated by the continuum X-radiation. The thin hot plasma model gave an unsatisfactory fit to the NFI data. Also, the fact that no difference was seen in equivalent width between the low- and high-flux intervals of the NFI data seems to be at odds with a thin thermal plasma origin of the emission line. Therefore, we believe it is more probable that we are dealing with fluorescent emission in a photo-ionized medium. The ionization degree of the medium is fairly high, the ionization parameter $`\xi =L/nr^2`$ (Kallman & McCray 1982), where $`r`$ is the size of the fluorescent material and $`n`$ the atom density, is of order 10<sup>3</sup> to 10<sup>4</sup> which implies that the medium must be close to the source of the continuum radiation.
The 2 to 26 keV spectrum of the 2nd WFC-detected flare (Fig. 7) shows an ionization balance of iron that is at much lower degrees than that of the calm emission half a year earlier. Apparently, $`\xi `$ is at least two orders of magnitude smaller. This implies that $`n`$ and/or $`r`$ increased substantially. Therefore, the cloud of fluorescent material must have been, during the flare, denser and/or farther away from the source of the fluorescing radiation (probably near to the compact object). The same probably applies to the giant outburst 5 days later because the iron line was seen at an energy of 6.5 keV then (Markwardt et al. 1999b).
We do not detect K-edge absorption of iron at the appropriate energies of 8.8 (Fe XXV) and 9.3 keV (Fe XXVI). The 3$`\sigma `$ upper limit on the relative depth of an edge at 9.0 keV is 0.2 which does not appear to be very constraining (e.g., Makishima 1986).
The equivalent width of the Fe-K line complex is relatively large. In the WFC-detected flare in September, it was 3 to 4 times as large as during the March NFI observation. Markwardt et al. (1999b) also observed the Fe-K complex during the tail of the giant flare, at a photon energy that is close to that measured with WFC six days earlier and with an equivalent width of about 500 eV. Model calculations of the structure of the accretion disk coronae and data analyses of Fe-K lines in other X-ray binaries by Vrtilek et al. (1993) suggest that large equivalent widths in Fe-K lines may be due to high (edge-on) inclination angles. This appears to be in line with the high inclination angle inferred from the radio jets.
Hjellming et al. (1999b) reported a double-sided jet structure from SAX J1819.3–2525 in a VLA image taken on Sep. 16.02 UT, with sizes of roughly 0.25″ in both directions. A subsequent VLA image taken on Sep. 17.93 only shows one of the two jets at the same position. If $`v`$ is the intrinsic velocity of the jets, $`i`$ the angle between the jets and the line of sight, and $`c`$ the velocity of light, then the angular proper motion $`\mu _\pm `$ of the approaching (–) and receding jet (+) are given by:
$`\mu _\pm `$ $`=`$ $`(v/d)\mathrm{sin}i/[1\pm (v/c)\mathrm{cos}i]`$
(e.g., Hjellming & Rupen 1995). We can only make a rough estimate of $`\mu _+`$ and $`\mu _{}`$. The uncertainty results from the facts that no moving radio blobs were seen, like in GRO J1655–40 (c.f., Hjellming & Rupen 1995), and that no difference was measured between $`\mu _+`$ and $`\mu _{}`$ (this is partly due to insufficient accuracy of the position of the optical counterpart). If we assume that the difference between $`\mu _+`$ and $`\mu _{}`$ is less than 10%, we find that $`(v/c)\mathrm{cos}i<0.048`$. The value for $`\mu _+\mu _{}`$ is determined by the time of jet ejection. Depending on whether one takes this to be at the time of the measured optical peak (Stubbings 1999), X-ray peak or X-ray pre-peak (see Smith et al. 1999), the travel time of the jets on Sep. 16.02 was between 0.3 and 1.1 d. This results in an allowed range for $`\mu _+\mu _{}`$ of 224 to 806 mas d<sup>-1</sup>. This implies $`(v/c)>0.53`$. The above constraint on $`(v/d)\mathrm{cos}i`$ then results in $`i>84.8`$°. Additionally, we can obtain a constraint on the distance from $`(v/c)<1`$ and $`v\mathrm{sin}i=\mu d`$. For $`\mu =224`$ mas d<sup>-1</sup>, $`d<0.8`$ kpc. Conversely, a distance of 0.4 kpc implies a jet ejection on Sep. 15.4 for $`v=c`$ or Sep. 14.9 for $`v=0.53c`$
The angle $`i`$ is probably close to the inclination angle of the binary orbit, since one may expect the jets to be ejected close to perpendicular to the binary’s orbital plane. The high value of $`i`$ then suggests that at times the compact object should be eclipsed by the companion star. Our NFI data involve the longest continuous observation ever performed of SAX J1819.3–2525 while it was above the detection threshold (all RossiXTE TOOs lasted shorter than 3 hr). During 19 hr no eclipse was seen. Since the earth obscured the view to SAX J1819.3–2525 each BeppoSAX orbit, there is some uncertainty in the non-detection of an eclipse. With this reservation, the non-detection suggests that either the binary has an orbital period in excess of 19 hr, that the jets are misaligned from the orbital rotation axis by more than $``$10° (like seems to be the case for the micro quasar GRO J1655–40, see Van der Hooft et al. 1997), or that the companion star is not an A-type main sequence star.
The X-ray flux of SAX J1819.3–2525 behaved quite erratic. Whenever SAX J1819.3–2525 was detected, it exhibited at least 50% variability on hourly time scales. The variability is not (quasi) periodic. The scan observations with the PCA show that the source turned on probably on Feb. 18 and disappeared immediately after the giant outburst on Sep. 15. During the seven months in between, the snapshot measurements revealed fluxes varying at least between an upper limit of 0.4 and detections of up to 12$`\times 10^{10}`$ erg cm<sup>-2</sup>s<sup>-1</sup> (2-60 keV; Markwardt et al. 1999b, C.B. Markwardt priv. comm.). The two WFC-detected flares show that the flux sometimes reached flux levels one order of magnitude higher. Where does this variability come from and why did it turn off immediately after the giant outburst?
The mass transfer from the companion star should be mediated either through an (irregular) wind from the companion star or an accretion disk. Which one applies to SAX J1819.3–2525 is unclear, given the observations thus far. The high variability is perhaps most easily explained by an irregular wind from the companion star. Such wind accretion is very common among the high-mass X-ray binaries (see review by White et al. 1995). We note that SAX J1819.3–2525 shows similarities to CI Cam. CI Cam was fast (e-folding decay time 0.6 d, Harmon et al. 1998), exhibited a bright Fe-K line (EW up to 597 eV, Orr et al. 1998), was a radio jet source (Hjellming & Mioduszewski 1998), and was detectable in the 20-70 keV band (Harmon et al. 1998). The optical counterpart of CI Cam is a symbiotic B star with an irregular wind. The alternative, in the accretion disk interpretation, is that disk or thermal instabilities are continuously important. The apparent 0.3 to 0.9 day head start in the giant outburst of the optical to the X-ray emission suggests that a disk instability may be at work during this particular flare which moves from the outside to the inside of the disk, like has been seen in dwarf novae (e.g., Meyer & Meyer-Hofmeister 1994) and in the LMXB black-hole transient GRO J1655–40 (Orosz et al. 1997). During the 5 days before the giant outburst V4641 Sgr is continuously 2 mag brighter than immediately after the giant outburst when there is hardly any X-ray emission (Kato et al. 1999). This can be explained as optical emission from a disk which disappeared with the giant outburst. The alternative in the wind accretion interpretation is that the 2 mag come from the companion star. This too is not unreasonable, increased levels of mass loss are likely to go hand in hand with such brightening. Finally, there is the observation that the largest flare forced the system to go in quiescence. Again, this can be explained either way. In the accretion disk interpretation, the giant outburst drained the disk in such a manner that no accretion occurs anymore. The implied time scale of this drainage is about 10 hr. This can only be explained if the drained mass is small though enough to induce a strong flare.
The short distance enables one to set a rather strict limit on the flux for quiescent emission. The PCA monitoring program reveals that the flux went below an upper limit of 0.4$`\times 10^{10}`$ erg cm<sup>-2</sup>s<sup>-1</sup> (2-60 keV) or 1 mCrab (C.B. Markwardt, priv. comm.). Assuming the same spectrum holds as during our NFI observation (when the flux was approximately 25 times higher), the 0.4-120 keV luminosity upper limit is $`<8\times 10^{32}(d/0.4\mathrm{kpc})^2`$ erg s<sup>-1</sup>. Unfortunately, this flux limit is too high to say something sensible about the nature of the compact object (see, e.g., Rutledge et al. 1999).
We reiterate the point made by Wijnands & Van der Klis (2000) that this transient may be an example of a separate class of fast and faint X-ray transients whose existence has been proposed by Heise et al. (1999) and is characterized by peak luminosities that are two orders of magnitude less than the typical bright soft X-ray transient and by outburst durations of order one week instead of months. They estimate a rate of 18 such transients per year within 20° from the Galactic center. Such transients are easily over-seen by surveying instruments with small duty cycles if they are at the usual few-kpc distances. It is only because of its brightness due to its nearness that SAX J1819.3–2525 was easily noticed. At the distance to the Galactic center, the source would only have had a peak flux of 30 mCrab.
It is not clear whether the compact object is a neutron star or a black hole candidate. No X-ray characteristics that unambiguously identify the nature, such as type I X-ray bursts or pulsations, are present. Optical Doppler measurements during quiescence will give the best opportunity to obtain constraints on the mass of the compact object.
###### Acknowledgements.
We thank Frits Paerels for useful discussions. This research has made use of linearized and cleaned event files from the ASI-BeppoSAX SDC on-line database and archive system. BeppoSAX is a joint Italian and Dutch program. |
no-problem/0001/hep-ph0001127.html | ar5iv | text | # Constraints on the Gluon Density from Lepton Pair Production
## 1 INTRODUCTION
The production of lepton pairs in hadron collisions $`h_1h_2\gamma ^{}X;\gamma ^{}l\overline{l}`$ proceeds through an intermediate virtual photon via $`q\overline{q}\gamma ^{}`$, and the subsequent leptonic decay of the virtual photon. Traditionally, interest in this Drell-Yan process has concentrated on lepton pairs with large mass $`Q`$ which justifies the application of perturbative QCD and allows for the extraction of the antiquark density in hadrons .
Prompt photon production $`h_1h_2\gamma X`$ can be calculated in perturbative QCD if the transverse momentum $`Q_T`$ of the photon is sufficiently large. Because the quark-gluon Compton subprocess is dominant, $`gq\gamma X`$, this reaction provides essential information on the gluon density in the proton at large $`x`$ . Unfortunately, the analysis suffers from fragmentation, isolation, and intrinsic transverse momentum uncertainties. Alternatively, the gluon density can be constrained from the production of jets with large transverse momentum at hadron colliders , but the information from different experiments and colliders is ambiguous.
In this paper we demonstrate that, like prompt photon production, lepton pair production is dominated by quark-gluon scattering in the region $`Q_T>Q/2`$. This realization means that new independent constraints on the gluon density may be derived from Drell-Yan data in kinematical regimes that are accessible in collider and fixed target experiments but without the theoretical and experimental uncertainties present in the prompt photon case.
In Sec. 2, we review the relationship between virtual and real photon production in hadron collisions in next-to-leading order QCD. In Sec. 3 we present our numerical results, and Sec. 4 is a summary.
## 2 NEXT-TO-LEADING ORDER QCD FORMALISM
In leading order (LO) QCD, two partonic subprocesses contribute to the production of virtual and real photons with non-zero transverse momentum: $`q\overline{q}\gamma ^{()}g`$ and $`qg\gamma ^{()}q`$. The cross section for lepton pair production is related to the cross section for virtual photon production through the leptonic branching ratio of the virtual photon $`\alpha /(3\pi Q^2)`$. The virtual photon cross section reduces to the real photon cross section in the limit $`Q^20`$.
The next-to-leading order (NLO) QCD corrections arise from virtual one-loop diagrams interfering with the LO diagrams and from real emission diagrams. At this order $`23`$ partonic processes with incident gluon pairs $`(gg)`$, quark pairs $`(qq)`$, and non-factorizable quark-antiquark $`(q\overline{q}_2)`$ processes contribute also. Singular contributions are regulated in $`n`$=4-2$`ϵ`$ dimensions and removed through $`\overline{\mathrm{MS}}`$ renormalization, factorization, or cancellation between virtual and real contributions. An important difference between virtual and real photon production arises when a quark emits a collinear photon. Whereas the collinear emission of a real photon leads to a $`1/ϵ`$ singularity that has to be factored into a fragmentation function, the collinear emission of a virtual photon yields a finite logarithmic contribution since it is regulated naturally by the photon virtuality $`Q`$. In the limit $`Q^20`$ the NLO virtual photon cross section reduces to the real photon cross section if this logarithm is replaced by a $`1/ϵ`$ pole. A more detailed discussion can be found in .
The situation is completely analogous to hard photoproduction where the photon participates in the scattering in the initial state instead of the final state. For real photons, one encounters an initial-state singularity that is factored into a photon structure function. For virtual photons, this singularity is replaced by a logarithmic dependence on the photon virtuality $`Q`$ .
A remark is in order concerning the interval in $`Q_T`$ in which our analysis is appropriate. In general, in two-scale situations, a series of logarithmic contributions will arise with terms of the type $`\alpha _s^n\mathrm{ln}^n(Q/Q_T)`$. Thus, if either $`Q_T>>Q`$ or $`Q_T<<Q`$, resummations of this series must be considered. For practical reasons, such as event rate, we do not venture into the domain $`Q_T>>Q`$, and our fixed-order calculation should be adequate. On the other hand, the cross section is large in the region $`Q_T<<Q`$. In previous papers , we compared our cross sections with available fixed-target and collider data on massive lepton-pair production, and we were able to establish that fixed-order perturbative calculations, without resummation, should be reliable for $`Q_T>Q/2`$. At smaller values of $`Q_T`$, non-perturbative and matching complications introduce some level of phenomenological ambiguity. For the goal we have in mind, viz., contraints on the gluon density, it would appear best to restrict attention to the region $`Q_TQ/2`$, but below $`Q_T>>Q`$.
## 3 PREDICTED CROSS SECTIONS
In this section we present numerical results for the production of lepton pairs in $`p\overline{p}`$ collisions at the Tevatron with center-of mass energy $`\sqrt{S}=1.8`$ and 2.0 TeV. We analyze the invariant cross section $`Ed^3\sigma /dp^3`$ averaged over the rapidity interval -1.0 $`<y<`$ 1.0. We integrate the cross section over various intervals of pair-mass $`Q`$ and plot it as a function of the transverse momentum $`Q_T`$. Our predictions are based on a NLO QCD calculation and are evaluated in the $`\overline{\mathrm{MS}}`$ renormalization scheme. The renormalization and factorization scales are set to $`\mu =\mu _f=\sqrt{Q^2+Q_T^2}`$. If not stated otherwise, we use the CTEQ4M parton distributions and the corresponding value of $`\mathrm{\Lambda }`$ in the two-loop expression of $`\alpha _s`$ with four flavors (five if $`\mu >m_b`$). The Drell-Yan factor $`\alpha /(3\pi Q^2)`$ for the decay of the virtual photon into a lepton pair is included in all numerical results.
In Fig. 1 we display the NLO QCD cross section for lepton pair
production at the Tevatron at $`\sqrt{S}=1.8`$ TeV as a function of $`Q_T`$ for four regions of $`Q`$. The regions of $`Q`$ have been chosen to avoid resonances, i.e. between $`2`$ GeV and the $`J/\psi `$ resonance, between the $`J/\psi `$ and the $`\mathrm{{\rm Y}}`$ resonances, above the $`\mathrm{{\rm Y}}`$’s, and a high mass region. The cross section falls both with the mass of the lepton pair $`Q`$ and, more steeply, with its transverse momentum $`Q_T`$. No data are available yet from the CDF and D0 experiments. However, prompt photon production data exist to $`Q_T100`$ GeV, where the cross section is about $`10^3`$ pb/GeV<sup>2</sup>. It should be possible to analyze Run I data for lepton pair production to at least $`Q_T30`$ GeV where one can probe the parton densities in the proton up to $`x_T=2Q_T/\sqrt{S}0.03`$. The UA1 collaboration measured the transverse momentum distribution of lepton pairs at $`\sqrt{S}=630`$ GeV up to $`x_T=0.13`$ , and their data agree well with our theoretical results .
The fractional contributions from the $`qg`$ and $`q\overline{q}`$ subprocesses through NLO are shown in Fig. 2. It is evident
that the $`qg`$ subprocess is the most important subprocess as long as $`Q_T>Q/2`$. The dominance of the $`qg`$ subprocess diminishes somewhat with $`Q`$, dropping from over 80 % for the lowest values of $`Q`$ to about 70 % at its maximum for $`Q`$ 30 GeV. In addition, for very large $`Q_T`$, the significant luminosity associated with the valence dominated $`\overline{q}`$ density in $`p\overline{p}`$ reactions begins to raise the fraction of the cross section attributed to the $`q\overline{q}`$ subprocesses. Subprocesses other than those initiated by the $`q\overline{q}`$ and $`qg`$ initial channels are of negligible import.
We update the Tevatron center-of-mass energy to Run II conditions ($`\sqrt{S}=2.0`$ TeV) and use the latest global fit by the CTEQ collaboration (5M). Figure 3
demonstrates that the larger center-of-mass energy increases the invariant cross section for the production of lepton pairs with mass 5 GeV $`<Q<`$ 6 GeV by 5 % at low $`Q_T1`$ GeV and 20 % at high $`Q_T100`$ GeV. In addition, the expected luminosity for Run II of 2 fb<sup>-1</sup> should make the cross section accessible to $`Q_T100`$ GeV or $`x_T0.1`$. This extension would constrain the gluon density in the same regions as prompt photon production in Run I.
Next we present a study of the sensitivity of collider and fixed target experiments to the gluon density in the proton. The full uncertainty in the gluon density is not known. Here we estimate this uncertainty from the variation of different recent parametrizations. We choose the latest global fit by the CTEQ collaboration (5M) as our point of reference and compare results to those based on their preceding analysis (4M) and on a fit with a higher gluon density (5HJ) intended to describe the CDF and D0 jet data at large transverse momentum. We also compare to results based on global fits by MRST , who provide three different sets with a central, higher, and lower gluon density, and to GRV98 <sup>1</sup><sup>1</sup>1In this set a purely perturbative generation of heavy flavors (charm and bottom) is assumed. Since we are working in a massless approach, we resort to the GRV92 parametrization for the charm contribution and assume the bottom contribution to be negligible..
In Fig. 4 we plot the cross section for lepton pairs with mass between the
$`J/\psi `$ and $`\mathrm{{\rm Y}}`$ resonances at Run II of the Tevatron in the region between $`Q_T=10`$ and 30 GeV ($`x_T=0.01\mathrm{}0.03`$). For the CTEQ parametrizations we find that the cross section increases from 4M to 5M by 2.5 % ($`Q_T=30`$ GeV) to 5 % ($`Q_T=10`$ GeV) and from 5M to 5HJ by 1 % in the whole $`Q_T`$-range. The largest differences from CTEQ5M are obtained with GRV98 at low $`Q_T`$ (minus 10 %) and with MRST(g$``$) at large $`Q_T`$ (minus 7%).
The theoretical uncertainty in the cross section can be estimated by varying the renormalization and factorization scale $`\mu =\mu _f`$ around the central value $`\sqrt{Q^2+Q_T^2}`$. Figure 5
shows this variation for $`p\overline{p}\gamma ^{}X`$ at $`\sqrt{S}=2.0`$ TeV in the region between the $`J/\psi `$ and $`\mathrm{{\rm Y}}`$ resonances. In the interval $`0.5<\mu /\sqrt{Q^2+Q_T^2}<2`$ the dependence of the cross section on the scale $`\mu =\mu _f`$ drops from $`\pm 15\%`$ (LO) to the small value $`\pm 2.5\%`$ (NLO). The $`K`$-factor ratio (NLO/LO) is approximately 2, as one might expect naively.
A similar analysis for Fermilab’s fixed target experiment E772 is shown in Fig. 6.
In this experiment, a deuterium target is bombarded with a proton beam of momentum $`p_{\mathrm{lab}}=`$ 800 GeV, i.e. $`\sqrt{S}=38.8`$ GeV. The cross section is averaged over the scaled longitudinal momentum interval 0.1 $`<x_F<`$ 0.3. In fixed target experiments one probes substantially larger regions of $`x_T`$ than in collider experiments. Therefore one expects greater sensitivity to the gluon distribution in the proton. We find that use of CTEQ5HJ increases the cross section by 7 % (26 %) w.r.t. CTEQ5M at $`Q_T=3`$ GeV ($`Q_T=6`$ GeV) and by 134 % at $`Q_T=10`$ GeV. With MRST(g$``$) the cross section drops relative to the CTEQ5M-based values by 17 %, 40 %, and 59 % for these three choices of $`Q_T`$.
Figure 7
shows the variation of the fixed target cross section on the renormalization and factorization scale $`\mu =\mu _f`$. In the interval $`0.5<\mu /\sqrt{Q^2+Q_T^2}<2`$ the dependence decreases from $`\pm 49\%`$ (LO) to $`\pm 37\%`$ (NLO). An optimal scale choice might be $`\mu =\mu _f=\sqrt{Q^2+Q_T^2}/4`$, where the points of Minimal Sensitivity (maximum of NLO) and of Fastest Apparent Convergence (LO=NLO) nearly coincide. At $`\mu =\mu _f=\sqrt{Q^2+Q_T^2}`$, the $`K`$-factor ratio is 2.6. The NLO cross section turns negative at the lowest scale shown $`\mu =\mu _f=\sqrt{Q^2+Q_T^2}/81`$ GeV, a value too low to guarantee perturbative stability.
## 4 SUMMARY
The production of Drell-Yan pairs with low mass and large transverse momentum is dominated by gluon initiated subprocesses. In contrast to prompt photon production, uncertainties from fragmentation, isolation, and intrinsic transverse momentum are absent. The hadroproduction of low mass lepton pairs is therefore an advantageous source of information on the parametrization and size of the gluon density. The increase in luminosity of Run II increases the accessible region of $`x_T`$ from 0.03 to 0.1. The theoretical uncertainty has been estimated from the scale dependence of the cross sections and found to be very small for collider experiments.
## Acknowledgment
It is a pleasure to thank L. E. Gordon for his collaboration. |
no-problem/0001/nucl-th0001030.html | ar5iv | text | # Effects of the secondary decays on the isotopic thermometers
## Abstract
The sharp microcanonical multifragmentation model from \[Al. H. Raduta and Ad. R. Raduta, Phys. Rev. C 55, 1344 (1997); Phys. Rev. C, in press\] is employed for evaluating the nuclear caloric curve predictions of nine isotopic thermometers for three representative nuclei. Evaluations are performed for both primary decay and asymptotic stages. Effects of the secondary decays on the primary decay caloric curves are evidenced and discussed. In both cases a dispersive character of the isotopic caloric curves with increasing the source excitation energy is observed. A procedure of calibrating the isotopic thermometers on the microcanonical predictions for both primary decay and asymptotic stages is proposed. The resulting set of calibrating parameters for each thermometer is independent on the source size, on its excitation energy and, in the case of the primary decay, on the freeze-out radius assumption.
Nuclear caloric curve is presently one of the most studied subjects in nuclear multifragmentation research. The reason lay in the resemblance of the nucleon-nucleon interaction with the van der Waals forces which motivates the expectation of a liquid-gas phase transition in nuclear matter. While statistical multifragmentation models (SMM and MMMC ) predicted caloric curves with transition-like regions since 1985, the first experimental evaluation of the nuclear caloric curve was performed ten years later by the ALADIN collaboration . A wide plateau interrupting the monotonical increase of the caloric curve at 5 MeV temperature was evidenced with that occasion. This was interpreted as a sign of liquid-gas phase transition. Experiments of INDRA and EOS performed on other reactions followed showing sightly different results. Two years latter, papers of the ALADIN collaboration announced a reevaluation of the neutron kinetic energies and of the energies of the spectator parts resulting in a modification of the ALADIN caloric curve from 1995. Reasons for such discrepancies lay in both the different employed reactions leading to different sequences of equilibrated sources ($`E^{}(A)`$) and, as shown above, in the degree of precision of the experimental measurements. The question is to what extent the experimentally measured caloric curves relate to the real one. Apart from the nonequilibrium parts of the excitation energies of the nuclear sources which have to be properly removed, one has to use a temperature formula which accounts for the finite size effects which manifest in nuclear systems.
In this respect, is widely used the Albergo isotopic formula , expressed as a double ratio of isotopic yields:
$$T_{12/34}=\mathrm{\Delta }B/\mathrm{ln}\{s\left[Y(A_1,Z_1)/Y(A_2,Z_2)\right]/\left[Y(A_3,Z_3)/Y(A_4,Z_4)\right]\},$$
(1)
where $`\mathrm{\Delta }B=\left[B(A_1,Z_1)B(A_2,Z_2)\right]\left[B(A_3,Z_3)B(A_4,Z_4)\right]`$, $`B(A_i,Z_i)`$ and $`Y(A_i,Z_i)`$ are respectively the binding energy and the yield corresponding to the isotope $`(A_i,Z_i)`$, $`s`$ is a statistical factor deduced from the ground state spin and masses of the isotopes and, finally, the isotope pairs (1,2) and (3,4) must differ by the same number of neutrons and/or protons. Being deduced on grandcanonical grounds, formula (1) does not account for finite size effects. Pochodzalla et al. tried to remediate this drawback by including a calibrating factor $`f_T=1.2`$ chosen as to average the results of the QSM, GEMINI and MMMC models. Obviously, the efficiency of this method depends on the compatibility of the employed model with the physical phenomenon. Due to the finite dimensions of the systems currently involved, the microcanonical ensemble is the best solution. Moreover, one has to take into account the process of secondary decays from the primary decay excited fragments which modify the yields entering formula (1).
The calibration of seven isotopic thermometers on the prediction of the sharp microcanonical model from Ref. was described in . There, the Albergo thermometers were only microcanonically corrected. In order to allow a direct application of the thermometers on the experimental yields one has to apply a second correction for eliminating the effects of the secondary decays. This is now possible with the new version of the sharp microcanonical multifragmentation model . In comparison with its initial formulation, in Ref. the microcanonical model is refined and improved by considering the experimental energy levels for fragments with $`A6`$ (instead of the level density Fermi gas formula) and by including the secondary decay stage. Using this version, a very good agreement was obtained in Ref. between the predicted uncorrected HeLi caloric curve and the experimental (uncorrected) HeLi caloric curve of ALADIN from 1997 .
The present paper intends to use the new version of the sharp microcanonical model in order to analyze the effects of the secondary decays on the primary decay isotopic caloric curve and to perform a microcanonical calibration of nine isotopic thermometers for both primary decay and asymptotic stages. Also, a study is made concerning the sensitivity of the resulting parametrization to the freeze-out radius assumption.
Isotopic caloric curve evaluations are presented in Fig. 1 for nine isotopic thermometers: <sup>6,7</sup>Li/<sup>3,4</sup>He, <sup>7,8</sup>Li/<sup>3,4</sup>He, <sup>8,9</sup>Li/<sup>3,4</sup>He, <sup>12,13</sup>C/<sup>3,4</sup>He, <sup>13,14</sup>C/<sup>3,4</sup>He, <sup>11,12</sup>B/<sup>3,4</sup>He, <sup>12,13</sup>B/<sup>3,4</sup>He, <sup>1,2</sup>H/<sup>3,4</sup>He, <sup>2,3</sup>H/<sup>3,4</sup>He together with the microcanonical caloric curve, in the case of three representative source nuclei (70,32), (130,54) and (190,79) and a freeze-out radius parameter of 2.5 $`A^{1/3}`$ fm. Calculations have been performed for both primary decay stage (left column) and asymptotic stage (right column). In all cases the isotopic curves show a dispersive character growing with the increase of the source excitation energy. An interesting effect is the strong diminish of this dispersion after the secondary decays (see Fig. 1, right column). When analyzing the asymptotic curve this effect may induce the exaggerate impression that the uncorrected Albergo isotopic thermometers give quite close results and consequently, the finite size effects affecting formula (1) are small. In fact, secondary decays can bring some isotopic caloric curves quite close to the microcanonical ones in spite of the great deviation between the above mentioned curves appearing in the primary decay stage (e.g. the case of the <sup>13,14</sup>C/<sup>3,4</sup>He curve).
In order to calibrate formula (1) on the microcanonical predictions for both primary decay and asymptotic stages we apply the method described in Ref. . First, the calibrating factors $`f_T=T_{micro}/T_{iso}`$ (where $`T_{micro}`$ denotes the microcanonical temperature and $`T_{iso}`$ the corresponding isotopic temperature) corresponding to each temperature point from Fig. 1 are represented in Fig. 2. In all cases, the $`f_T(E_{ex})`$ curves are almost straight so one can use linear expressions for fitting them:
$$f_T(E_{ex})=aE_{ex}+b,$$
(2)
where $`a`$ and $`b`$ are calibrating parameters. A set of calibrating parameters $`a`$, $`b`$ was derived for each considered thermometer for both the primary decay and the asymptotic situations. The values of these parameters are given in Table I for the case of the primary decay and in Table II for the asymptotic case.
In order to check the efficiency of the obtained parametrization, the isotopic caloric curves corresponding to both primary decay and asymptotic situations (Fig. 1) are adjusted using the obtained calibration parameters according to the relation:
$$T_{iso}^{corr}(E_{ex})=T_{iso}(E_{ex})f_T(E_{ex}).$$
(3)
The result is the overlapping between all the isotopic caloric curves and the microcanonical one for both the primary decay stage and the asymptotic stage presented in Fig. 3. In conclusion, $`f_T(E_{ex})`$ is perfectly applicable for calibrating formula (1) correctly. It is remarkable that the obtained parameters do not depend on the source size or its excitation energy. This makes them quite applicable on experimental data where the source size is strongly dependent on its excitation energy. While the parametrization corresponding to the primary decay case (Table I) needs a further correction as to erase the effects of the secondary decays, the parameterization corresponding to the asymptotic situation (Table II) is directly applicable on the experimental isotopic yields.
The modifications brought by the secondary decays to the primary decay isotopic caloric curve are clearly evidenced in Fig. 4. There it is represented the ratio between the primary decay caloric curves and the asymptotic caloric curve for all cases except <sup>13,14</sup>C/<sup>3,4</sup>He. For the clarity of the figure, this ratio can be very well approached by the ratio of the factors $`f_T(E_{ex})`$ corresponding to the asymptotic situation and those corresponding the the primary decay situation, $`f_T^{asymp}(E_{ex})/f_T^{prim}(E_{ex})`$. Except for one situation (<sup>6,7</sup>Li/<sup>3,4</sup>He), the curves after the secondary decays are lower than those corresponding to the primary decay on the entire considered domain of excitation energy. Also, the deviation of the asymptotic curves from the primary decay curves increases in five cases (<sup>7,8</sup>Li/<sup>3,4</sup>He, <sup>8,9</sup>Li/<sup>3,4</sup>He, <sup>12,13</sup>C/<sup>3,4</sup>He, <sup>11,12</sup>B/<sup>3,4</sup>He, <sup>12,13</sup>B/<sup>3,4</sup>He) and decreases in other three cases (<sup>6,7</sup>C/<sup>3,4</sup>He, <sup>1,2</sup>H/<sup>3,4</sup>He, <sup>2,3</sup>H/<sup>3,4</sup>He).
Finally, a study is made concerning the influence of the freeze-out radius assumption on the resulting parametrization. To this aim, the isotopic caloric curves corresponding to the primary decay for the (70,32) source for the freeze-out radii 2.25 $`A^{1/3}`$ fm and 2.75 $`A^{1/3}`$ fm are plotted in the left column of Fig. 5. The overall aspect is maintained except the position of the plateau-like region. In the left column of Fig. 5 the caloric curves corresponding to each considered freeze-out radius, adjusted according to equation (3) using the parameters listed in Table I are represented. Remarkably, the corrected curves overlap over the microcanonical ones. This means that the parametrization corresponding to the primary decay (Table I) is also independent to the freeze-out radius assumption. A similar study concerning the asymptotic isotopic caloric curves using the parameters from Table II showed the same tendency of overlapping between the isotopic curves but a small global deviation from the microcanonical curve was present for the $`2.25A^{1/3}`$ case.
Summarizing, isotopic caloric curves have been evaluated for nine isotopic thermometers in both primary decay and asymptotic stages. In both cases a dispersive character of the isotopic caloric curve monotonically increasing with the increase of the excitation energy is evidenced. In the asymptotic stage this dispersion is strongly diminished. A set of microcanonical calibrating parameters was deduced for each of the considered thermometers for both primary decay and asymptotic situations. Remarkably, these parameters are independent on the dimension of the source, on its excitation energy and, in the primary decay stage, on the freeze-out radius assumption. Finally, a study was made concerning the modifications brought by the secondary decays to the primary decay isotopic caloric curves. |
no-problem/0001/physics0001031.html | ar5iv | text | # Application of Conditioning to the Gaussian-with-Boundary Problem in the Unified Approach to Confidence Intervals
## I Introduction
Roe and Woodroofe have made an interesting suggestion for modifying the “unified approach” to classical confidence intervals which Feldman and I advocated in Ref. . They invoke the use of “conditioning”, namely replacing frequentist coverage probabilities with conditional probabilities, still calculated in a frequentist manner, but conditioned on knowledge gained from the result of the particular experiment at hand.
Roe and Woodroofe (RW) illustrate their suggestion using one of the two prototype problems, that of Poisson processes with background. Suppose, for example that an experiment observes 3 events (signal plus background). Then the experimenters know that, in that particular experiment, there were 3 or fewer background events. RW therefore calculate the frequentist coverage using an ensemble of experiments with 3 or fewer background events, rather than the larger unrestricted ensemble which we used. Thus, the RW ensemble changes from experiment to experiment. Conditioning on an equality has a long history in classical statistics. (Ref. contains key references.) However, conditioning on an inequality, as RW do when the number of events is greater than zero, is perhaps less well founded, and it is interesting to explore the consequences.
In this paper, we attempt to apply RW-like conditioning to the other prototype problem, that of Gaussian errors with a bounded physical region. The result is similar to the Poisson problem analyzed by RW, but difficulties which were apparently masked by the discrete nature of the Poisson problem now arise. In particular, the lower endpoints of confidence intervals are moved significantly in an undesirable direction.
## II The Unified Approach to the Gaussian-with-Boundary Problem
As in Ref. , we consider an observable $`x`$ which is the measured value of parameter $`\mu `$ in an experiment with a Gaussian resolution function with known fixed rms deviation $`\sigma `$, set here to unity. I.e.,
$$P(x|\mu )=\frac{1}{\sqrt{2\pi }}\mathrm{exp}((x\mu )^2/2).$$
(1)
We consider the interesting case where only non-negative values for $`\mu `$ are physically allowed (for example, if $`\mu `$ is a mass).
The confidence-belt construction in Ref. proceeded as follows. For a particular $`x`$, we let $`\mu _{\mathrm{best}}`$ be the physically allowed value of $`\mu `$ for which $`P(x|\mu )`$ is maximum. Then $`\mu _{\mathrm{best}}=\mathrm{max}(0,x)`$, and
$$P(x|\mu _{\mathrm{best}})=\{\begin{array}{cc}1/\sqrt{2\pi },\hfill & x0\hfill \\ \mathrm{exp}(x^2/2)/\sqrt{2\pi },\hfill & x<0\text{.}\hfill \end{array}$$
(2)
We then compute the likelihood ratio $`R`$,
$$R(x)=\frac{P(x|\mu )}{P(x|\mu _{\mathrm{best}})}=\{\begin{array}{cc}\mathrm{exp}((x\mu )^2/2),\hfill & x0\hfill \\ \mathrm{exp}(x\mu \mu ^2/2),\hfill & x<0\text{.}\hfill \end{array}$$
(3)
During our Neyman construction of confidence intervals, $`R`$ determines the order in which values of $`x`$ are added to the acceptance region at a particular value of $`\mu `$. In practice, this means that for a given value of $`\mu `$, one finds the interval $`[x_1,x_2]`$ such that $`R(x_1)=R(x_2)`$ and
$$_{x_1}^{x_2}P(x|\mu )𝑑x=\alpha ,$$
(4)
where $`\alpha `$ is the confidence level (C.L.). We solve for $`x_1`$ and $`x_2`$ numerically to the desired precision, for each $`\mu `$ in a fine grid. With the acceptance regions all constructed, we then read off the confidence intervals $`[\mu _1,\mu _2]`$ as in Ref. .
## III Invoking Conditioning in the Gaussian-with-Boundary Problem
In order to formulate the conditioning, we find it helpful to think of the measured value $`x`$ as being the sum of two parts, the true mean $`\mu _t`$ and the random “noise” which we call $`\epsilon `$:
$$x=\mu _t+\epsilon .$$
(5)
We are considering the case where it is known on physical grounds that $`\mu _t0`$. Thus, if an experimenter obtains the value $`x_0`$ in an particular experiment, then he or she knows that, in that particular experiment,
$$\epsilon x_0.$$
(6)
For example, if the experimenter measures $`\mu `$ and obtains $`x_0=2`$, then the experimenter knows that $`\epsilon 2`$ in that particular experiment. This information is analogous to the information in the Poisson problem above in which one knows that in the particular experiment, the number of background events is 3 or fewer. We thus use it the manner analogous to that of RW: our particular experimenter will consider the ensemble of experiments with $`\epsilon x_0`$ when constructing the confidence belt relevant to his or her experiment.
We let $`P(x|\mu ,\epsilon x_0)`$ be the (normalized) conditional probability for obtaining $`x`$, given that $`\epsilon x_0`$. In notation similar to that of RW, this can be denoted as $`q_\mu ^{x_0}(x)`$:
$$q_\mu ^{x_0}(x)P(x|\mu ,\epsilon x_0)=\{\begin{array}{cc}\frac{2}{\sqrt{2\pi }}\mathrm{exp}((x\mu )^2/2)/(\mathrm{erf}(x_0/\sqrt{2})+1),\hfill & x\mu +x_0\hfill \\ 0,\hfill & x>\mu +x_0\text{.}\hfill \end{array}$$
(7)
Given $`x_0`$, at each $`x`$ we find $`\mu _{\mathrm{best}}`$, that value of $`\mu `$ which maximizes $`P(x|\mu ,\epsilon x_0)`$:
$$\mu _{\mathrm{best}}=\{\begin{array}{cc}x,\hfill & x_00\text{ and }x0\hfill \\ xx_0,\hfill & x_0<0\text{ and }xx_0\hfill \\ 0,\hfill & \text{otherwise}\hfill \end{array}$$
(8)
In the notation of Ref. , $`P(x|\mu _{\mathrm{best}},\epsilon x_0)`$ is then
$$\underset{\mu ^{}}{\mathrm{max}}q_\mu ^{}^{x_0}(x)=\frac{2}{\sqrt{2\pi }(\mathrm{erf}(x_0/\sqrt{2})+1)}\times \{\begin{array}{cc}1,\hfill & x_00\text{ and }x0\hfill \\ \mathrm{exp}(x_0^2)/2,\hfill & x_0<0\text{ and }xx_0\hfill \\ \mathrm{exp}(x^2)/2,\hfill & \text{otherwise}\hfill \end{array}$$
(9)
Then the ratio $`R`$ of Eqn. 3 is replaced by
$$\stackrel{~}{R}^{x_0}(\mu ,x)=\frac{q_\mu ^{x_0}(x)}{\mathrm{max}_\mu ^{}q_\mu ^{}^{x_0}(x)},$$
(10)
which vanishes if $`x>\mu +x_0`$, and otherwise is given by
$$\stackrel{~}{R}^{x_0}(\mu ,x)=\{\begin{array}{cc}\mathrm{exp}((x\mu )^2/2),\hfill & x_00\text{ and }x0\hfill \\ \mathrm{exp}(((x\mu )^2+x_0^2)/2),\hfill & x_0<0\text{ and }xx_0\hfill \\ \mathrm{exp}(x\mu \mu ^2/2),\hfill & \text{otherwise}\hfill \end{array}$$
(11)
Figures 1 through 3 show graphs of $`q_\mu ^{x_0}(x)`$, $`\mathrm{max}_\mu ^{}q_\mu ^{}^{x_0}(x)`$, and $`\stackrel{~}{R}^{x_0}(\mu ,x)`$, for three values of $`\mu `$, for each of three values of $`x_0`$.
We let $`\stackrel{~}{c}_{x_0}(\mu )`$ be the value of $`c`$ for which
$$_{x:\stackrel{~}{R}^{x_0}(\mu ,x)<c}q_\mu ^{x_0}(x)𝑑x=\alpha .$$
(12)
The modified confidence interval consists of those $`\mu `$ for which
$$\stackrel{~}{R}^{x_0}(\mu ,x_0)\stackrel{~}{c}_{x_0}(\mu ).$$
(13)
Note that this entire construction depends on the value of $`x_0`$ obtained by the particular experiment. An experiment obtaining a different value of $`x_0`$ will have a different function in Eqn. 7, and hence a different confidence belt construction. Figure 4 shows examples of such constructions for six values of $`x_0`$. The vertical axis gives the endpoints of the confidence intervals. Each different confidence belt construction is used only for an experiment obtaining the value $`x_0`$ which was used to construct the belt. The interval $`[\mu _1,\mu _2]`$ at $`x=x_0`$ is read off for that experiment; the rest of that plot is not used.
Finally, we can form the graph shown in Fig. 5 by taking the modified confidence interval for each $`x_0`$, and plotting them all on one plot. These are tabulated in Table I, which includes for comparison the unconditioned intervals from Table X of Ref. .
Fig. 6 shows the modified intervals plotted together with the unified intervals of Ref. . The modified upper curve is shifted upward for negative $`x`$, which results in a less stringent upper limit when $`\epsilon `$ is known to be negative; this feature is considered desirable by some. The lower curve, however, is also shifted upward: for all $`x_0>0`$, the interval is two-sided. We find this to be a highly undesirable side-effect.
It is interesting to consider what happens if one applies Fig. 5 to an unconditioned ensemble. The result can be seen by drawing a horizontal line at any $`\mu `$ in Fig. 5 and integrating $`P(x|\mu )`$ (Eqn.1) along that line between the belts. For small $`\mu `$, there is significant undercoverage, while for $`\mu `$ near 1.0, there is significant overcoverage. The undercoverage was surprising, since the conditioned intervals always cover within the relevant subset of the ensemble. However, conditioning on an inequality means that these subsets are not disjoint.
The undesirable raising of the lower curve is present in the Poisson case, as can be seen in Figure 1 of Ref. . However, there the discreteness of the Poisson problem apparently prevents the curve from being shifted so dramatically, and the two-sided intervals do not extend to such low values of the measured $`n`$.
## IV Conclusion
In this paper, we apply conditioning in the style Roe and Woodroofe to the Gaussian-with-boundary problem. We find that the transition from one-sided intervals to two-sided intervals undesirably moves to the origin. This reflects a general feature of confidence interval construction: when moving one of the two curves, the other curve moves also. In the Poisson-with-background problem, the undesirable movement was not large, but in the Gaussian-with-boundary problem, the effect is quite substantial.
###### Acknowledgements.
I thank Gary Feldman, Byron Roe, and Michael Woodroofe for comments on the paper. This work was supported by the U.S. Department of Energy. |
no-problem/0001/astro-ph0001043.html | ar5iv | text | # Photometric Properties of 47 Clusters of Galaxies: I. The Butcher-Oemler Effect
## 1 INTRODUCTION
The existence of a linear locus in the color-magnitude diagram is a main characteristic of the early-type galaxy population in clusters. The effect in which the brightest galaxies are also the most red ones was studied in a systematic way by various authors. Visvanathan & Sandage (1977) obtained spectral energy distributions (SEDs) over a wide range of wavelengths for Virgo cluster galaxies and were one of the first to establish not only the existence, but also the dependence of the slope of the color-magnitude relation with wavelength.
The slope of the color-magnitude relation can be explained as an age effect, implying that the most luminous, massive galaxies, are older, or as a metallicity effect, consistent with the most luminous galaxies being more metal rich. Hubble Space Telescope (HST) observations have shown that the color-magnitude relation continues to be very well defined for early-type galaxies in high redshift clusters up to $`z1`$ (Standford et al. 1998, Kodama et al. 1998, Ellis et al. 1997), as well as in the field (Kodama et al. 1999). These observations suggest that at high redshift the early-type population already consists of old, passively evolving systems. Ellis et al. (1997), Standford et al. (1998) estimated that these galaxies were formed at $`z>2`$, and therefore the color-magnitude relation can not be explained as a purely age effect. Also, Terlevich et al. (1999) used spectral absorption line indices, together with broad band photometry, to investigate how mean age and metal abundance correlate with galaxy luminosity and find that the color-magnitude relation in Coma is driven primarily by a luminosity-metallicity correlation.
However, recent observations by Worthey (1996) indicate that star formation is still occurring in $`2/3`$ of low redshift ellipticals and that many of these galaxies have ages less than half a Hubble time. The scatter in age of the early-type galaxy population can still be consistent with a well defined color-magnitude relation if galaxies assembling more recently are on average more metal-rich than older galaxies of same luminosity (Ferreras et al. 1998), which is consistent with Worthey’s observations that show a trend for younger large galaxies to be more metal rich.
The luminosity-metallicity relation can be created as a result of the different efficiency of supernova galactic winds to eject gas from galaxies with different masses (Larson 1974). Massive galaxies are able to retain a high fraction of their gas, becoming more enriched than less massive galaxies which tend to lose their gas more easily. Another possible explanation for the origin of the luminosity-metallicity relation is the difference in merging dynamics between systems of different luminosities (Bekki and Shioya 1997). If the more luminous elliptical galaxies are formed by galaxy merging with more rapid star formation, less gas is tidally stripped from these systems during merging, and as a result a greater amount of the gas is enriched.
The universality of the color-magnitude relation (Visvanathan & Sandage 1977, Bower et al. 1992) makes it a powerful tool to study characteristics of the cluster galaxy population and its evolution. Butcher and Oemler (1978) observed the central regions of two rich, high redshift clusters of galaxies (CL 3C 295 at z=0.46 and CL 0024+1654 at z=0.39) and found an excess of blue galaxies in comparison to the typical early-type population which is found in the central region of local clusters (Dressler 1980). In 1984, Butcher & Oemler published a study of 33 clusters in the redshift range between z=0.003 and 0.54 and confirmed that the fraction of blue galaxies in the inner region of clusters, where $`30\%`$ of their galaxy population is found, increases with redshift. Butcher and Oemler’s results were interpreted as detection of evolution in the population of galaxies, and many works regarding the nature of the blue galaxies followed.
Larson et al. (1980) suggested an evolutionary connection between S0 and spiral galaxies as the physical origin of the enhanced star formation in moderate redshift galaxies. The authors argue that S0s might be disk systems that lost their gas envelopes during the cluster collapse and consumed their remaining gas by star formation. This idea can explain the population of blue galaxies observed by Butcher and Oemler as spiral galaxies seen just before running out of gas, and the disappearance of this population in more evolved, low redshift rich clusters (Dressler et al. 1997, Couch et al. 1998). This idea is also in agreement with observations showing that the blue population lies preferentially in the outer edges of the cluster (Butcher and Oemler 1984, Rakos et al. 1997). Also, early-type galaxies in dense clusters are less metal-enriched than their counterparts in lower density environments (Rose et al. 1994), suggesting the truncation of star-formation in high density regions. The idea that the Butcher-Oemler effect is an evolutionary phenomenon is also reinforced by observations showing that star formation similar to that seen in distant clusters is still ongoing, although at a reduced level, in local clusters (Caldwell and Rose 1997).
HST images allowed the determination of the morphology of galaxies in high redshift clusters (Dressler et al. 1994, Couch et al. 1994, 1998). These data suggest that the blue Butcher-Oemler galaxies are predominantly normal late-type (i.e. small B/D ratios) spirals but also that dynamical interactions and mergers between galaxies may be an important process responsible for the star formation enhancement in those galaxies. Oemler et al. (1997) presented HST obsevations of four rich clusters at $`z0.4`$ confirming that most of the blue Butcher-Oemler galaxies have colors, luminosities and spatial distributions similar to the normal galaxies observed at low redshift. The authors show however that $`30\%`$ of these “normal” galaxies show abnormal patterns of star formation such as rings. Oemler et al. (1997) also find a large fraction of m/i galaxies, but argue that these objects can not account for most of the blue Buther-Oemler galaxies. m/i galaxies do not follow the spatial distribution of the the blue Buther-Oemler objects, instead they are concentrated at the dense regions of the cluster following the distribution observed for E/S0 galaxies. Furthermore, Oemler et al. (1997) show that the m/i galaxies are too few to account for the Buther-Oemler galaxies. Rakos et al. (1996) used narrowband photometry to study Butcher-Oemler galaxies and also found that the blue population consists of spiral-like systems with the addition of a small fraction of bright, blue interacting and merger systems. Rakos and Schombert (1995) find that the fraction of blue galaxies increases from $`20\%`$ at $`z=0.4`$ to $`80\%`$ at $`z=0.9`$, suggesting that the evolution in clusters is even stronger than previously indicated by Butcher and Oemler.
Although the Butcher-Oemler effect has been debated in many studies, and many evolutionary theories have been proposed, there are some suggestions of strong selection biases in Butcher and Oemler’s original sample of clusters. Newberry et al. (1988) measured velocity dispersions and surface density of galaxies in clusters and found that there is a marked difference between local clusters and intermediate redshift ones. More recently, Andreon and Ettori (1999) measured x-ray surface brightness profiles, size, and luminosity of the Butcher-Oemler sample of clusters and conclude that this sample is not formed from the same kind of objects observed over a range of look-back times. The selection effects are not well understood, and might be mimicking evolutionary effects. Smail et al. (1998) used a well defined sample of 10 clusters in the redshift range $`z=0.220.28`$ with high x-ray luminosity, and found that these massive clusters contain only a small fraction of blue galaxies. The Butcher-Oemler effect is not observed in this sample.
Also, in a visually selected sample of clusters, the presence of false structures, identified as clusters due to projection effects, is expected. The color of the galaxy population at such regions will tend to be bluer than observed in clusters, and the inclusion of these false clusters in the study of the Butcher-Oemler effect will tend to increase the number of structures with high fraction of blue galaxies. It is therefore important to be very careful in interpreting results from samples which might be biased, and the need for a well defined, statistically significant sample of clusters to study evolutionary effects is evident.
Garilli et al. (1995, 1996) observed clusters in the redshift range $`0.05z0.25`$ and used the color-magnitude relation to study their galaxy populations. The authors estimated the fraction of blue galaxies and did not find signs of evolution, but noticed the presence of a rather large number of red galaxies, with both $`gr`$ and $`ri`$ colors at least $`0.3^m`$ redder than the early-type sequence in the color-magnitude diagrams. These objects accounted for $`7\%`$ of the total galaxy population. They also noticed that about $`1/3`$ of these red galaxies had $`19.5<m_r<22`$ and $`ri>1`$ which are typical of field galaxies at $`z>0.7`$, but their $`gr1.2`$ colors were not red enough for these galaxies to be at such a high redshift. The existence of such a population of galaxies would be very important for the models of stellar population in elliptical galaxies, and for this reason we used the same methodology as Garilli et al. (1996) to search for a red galaxy population in our sample.
The slope, intercept, and scatter of the color-magnitude relation can also be used to study the evolution of galaxy population in clusters. Many authors (Rakos and Schombert 1995, Ellis et al. 1997, Gladders et al. 1998, Kodama et al.1998, Standford et al. 1998, Pahre 1998, 1999) have found that the observations are consistent with models in which most early-type galaxies in rich clusters are old, passively evolving systems. Bower et al. (1998) proposed a model on which star formation occurs over an extended period of time in most galaxies with star formation being truncated randomly. This type of star formation allows both for the small scatter of the color-magnitude relation which is observed up to $`z1`$, and for the presence of the blue galaxy population at intermediate redshift clusters (Butcher-Oemler effect).
Bower et al. (1998) have also shown how is it possible to use the color-magnitude relation as a constraint on the formation of rich clusters. Mergers will tend to reduce the slope and increase the scatter of the color-magnitude relation, therefore, the ratio between the scatter and the slope of the relation can be used to study the degree of merging between pre-existing stellar systems. The authors analyze the cases of random and hierarchical mergers and show that the first case would tend to very rapidly destroy the relation and that the second case allows for the color-magnitude relation to persist through a larger number of mergers.
In this work we present the color-magnitude relation for 48 clusters, and then use them to study the Butcher-Oemler effect, and to search for a population of red galaxies in the sample. In order to study the Butcher-Oemler effect, we need to know the redshifts of the clusters. We use the 21 clusters for which there are spectroscopic redshift measurements to construct an empirical relation that allows the estimation of photometric redshifts for the remainder. The paper is organized as follows. In §2 we describe the observations and data reduction. §3 gives a description of the galaxy catalog construction. The results are presented in §4, and the summary and conclusions are shown in §5. We will examine the evolution of the color-magnitude relation and investigate the merger history of the galaxy population in a forthcoming paper.
## 2 OBSERVATIONS AND DATA REDUCTION
### 2.1 Observations
We obtained CCD images of 48 clusters in four runs between March 1997 and November 1998. All data were taken with the Tek 2K-3 detector at the Cassegrain focus of the 0.9m telescope at the Cerro Tololo Interamerican Observatory (CTIO). The CCD has 2048 x 2048 pixels, each pixel covering $`0.396^{\prime \prime }`$, corresponding to a field of $``$ $`13.5^{}`$ x $`13.5^{}`$, or $`0.56.7`$ Mpc at $`z=0.030.38`$ ($`H_0=67`$ Km s<sup>-1</sup> Mpc<sup>-1</sup>). We observed 44 Abell clusters and 4 cluster candidates detected in POSS-II (Second Palomar Sky Survey) (Gal et al. 1999) photographic plates. The images were taken through the g, r and i filters of the Thuan & Gunn (1976) photometric system, with effective wavelengths and widths of 5118 Å and 900 Å for g, 6798 Å and 1000 Å for r, and 8100 Å and 1500 Å for i. The exposure times varied between 15-60 minutes in the g band, 10-20 minutes in the r band, and 10-15 minutes in the i band.
Details of the observations are shown in Table 1. In column 1 the ACO (Abell, Corwin & Olowin 1989) cluster number or the candidate identification is given; column 2 shows the observation date; columns 3, 4 and 5 list the exposure times in the g, r and i bands; the seeing is given in columns 6, 7 and 8, and the limiting magnitudes (see section 3.2) are listed in columns 9, 10 and 11 for each band.
### 2.2 Data Reduction
The raw images were corrected for the usual instrumental effects of “bias”, “dome flat”, “sky flat”, and “illumination”. It was also necessary to apply a “shutter” correction for short exposure images ($`T_{exp}<20sec`$) such as “dome flats” and standard stars. In this case, the time for the shutter to open and to close is important compared to the total exposure time and therefore the borders of the CCD collect photons for a shorter time than its central part.
### 2.3 Photometric Calibration
The photometric calibration of the magnitude scale was determined by the observation of $`5`$ standard stars from the Thuan & Gunn (1976) list in each night. The color equation, which establishes the relation between instrumental and calibrated magnitudes, is given by eq.(1).
$$g_{cal}=g_{inst}+A+B\mathrm{sec}(z)+C(gr)_{inst}$$
(1)
where A is the zero point of the magnitude scale; B is the extinction coefficient; and C is the color term coefficient. Similar color equations were determined for the r and i bands, with the color terms established from the $`gr`$ and $`gi`$ colors respectively.
When possible, the color equations were determined for each night in the three photometric bands. Because the number of standard stars with i calibration is very limited, there were nights when we did not observe enough standard stars in this band to allow us to determine this calibration. However, we found that the color equation coefficients were always in close agreement for the nights within a specific run. We therefore used the stars observed in each run to derive mean g, r and i color equations (i.e. A, B and C coefficients) for each run. The color equations derived for each run showed rms errors of at most 0.025<sup>m</sup> in g, 0.017<sup>m</sup> in r, and 0.009<sup>m</sup> in i. These errors could justify the use of the same color equations for each entire run, but in order to avoid small systematic errors we decided to use the zero points of the magnitude scale estimated individually for each night. The zero points determined using the mean extinction and color term coefficients previously calculated for each run are listed in Table 2, and have errors that are consistent with the rms values of the color equations. In Figure 1 we show the comparison between magnitudes of standard stars published by Thuan & Gunn (1976) and our final calibrated measurements.
#### 2.3.1 External Errors
The external photometric errors were estimated by comparison between the measurements of clusters that were observed in different nights and runs. Abell cluster 2700 was observed in three different nights, and two other clusters (A248, and A324) were observed twice. Figure 2 shows the residuals of the magnitudes, in each band, for Abell 2700. The residuals are calculated from the galaxies brighter than the limiting magnitude (see §3.2) of the CCD images, and we found mean values of 0.042 in g, 0.017 in r, and 0.048 in i, which are of the same order as the rms values of the color equation.
## 3 GALAXY CATALOG CONSTRUCTION
### 3.1 Detection and Classification
The detection and classification of objects in the CCD images were performed using a modified version of the FOCAS (Faint Object Classification and Analysis System, Valdes 1982) package. The first step in the construction of those catalogs is the detection of objects. We used a detection limit of two times the standard deviation of the background sky, which corresponds to $``$7% of the local sky value. We used also a minimal detection area of 25 pixels, which is $`2`$ times greater than the average seeing disc.
FOCAS calculates about 30 photometric attributes for each detected object, including the total and aperture magnitudes. The total magnitude is determined with the use of an algorithm which extends the detection isophote to lower surface brightness limits. When the area of the extended isophote become two times bigger than the area of the initial detection isophote the total magnitude is computed. The aperture magnitude is calculated from the flux inside a central circular region of radius $`5.15^{\prime \prime }`$.
Another important step in the construction of the CCD catalogs is the determination of the point spread function (PSF) which is used in the classification procedure. Stellar, non-saturated, objects with magnitudes between $`16^m`$ and $`19^m`$ are selected to construct the PSF. A visual inspection on the objects selected by FOCAS as stars allowed us to exclude asymmetric objects, and the final PSF for each image was determined with typically 35 stars. The classification of the objects was performed by FOCAS using the determined PSF (Valdes 1982). To avoid the inclusion of clearly misclassified objects we carried out a visual inspection before the construction of the catalog (the data are available upon request).
### 3.2 Limiting Magnitude
The histogram of total magnitudes of all the detected objects (stars and galaxies) was used to estimate the limiting magnitude for each CCD image. This limiting magnitude was determined by the last bin of magnitude before the counts start to drop significantly (Picard 1991). The values are listed in Table 1 for each cluster, in the three photometric bands.
## 4 RESULTS
### 4.1 Color-Magnitude Diagrams
In Figure 3 we present the $`ri\times r`$ and $`gr\times r`$ color-magnitude relations for all the clusters in our sample. The linear fit shown by a solid line was, in general, determined using the galaxies in the magnitude range between $`M^{}1`$ and $`M^{}+2`$, where $`M^{}=20.91`$ ($`h=0.75`$) (Lin et al. 1996). Objects with $`r<16.0`$ can be saturated in our images and were excluded from our analysis, and for this reason the limiting magnitude to fit the color-magnitude relation of the lower redshift clusters was extended to $`M^{}+3`$ (Abell 119, 168, 509, 1134, 2103) and $`M^{}+4`$ (Abell 189, 261, 1260, P861C1). Also, because the average limiting magnitude of our sample is $`r22.0`$, we excluded fainter objects from our fitting procedure. The data was binned in intervals of 0.5 mag, except for the first bin for which we impose that there should be at least 5 galaxies. We then used ROSTAT (Beers et al. 1990) to determine the median color and magnitude in each bin, as well as its bi-weight dispersion. The fitting procedure was done with the GaussFit program (Jefferys et al. 1988), which allows the usage of several methods for solving least squares and robust estimation problems. After testing many of these procedures, we decided to use the “Orthogonal Regression M-Estimates” (orm), considering the color dispersion in each magnitude bin, to fit the color-magnitude relation.
### 4.2 Photometric Redshifts
In order to study the Butcher and Oemler effect we need to know the redshift of the clusters. The 21 clusters with spectroscopic measured redshift were used in the construction of an empirical relation between redshift and mean $`gri`$ magnitudes, that was then applied to estimate photometric redshifts for the remainder. The idea of estimating cluster redshifts using magnitudes and colors is based on the fact that in the central region of clusters the population is dominated by early-type galaxies (Dressler 1980, 1997). If we assume that the same SED is typical of all clusters, the apparent magnitude and consequentially the colors of these galaxies can be used to estimate redshifts because of the strong 4000 Åbreak feature typical of this population. We used the thirty brightest galaxies, after a statistical background correction, to estimate mean magnitudes representative of each cluster and fit a first order polynomial relation $`z(g,r,i)`$.
The background correction is very important if one is trying to study the cluster galaxy population, and it is also a difficult matter due to the field-to-field fluctuations on the number counts. Usually the best approach in estimating the back and foreground contamination is the use of an annulus around the cluster. Unfortunately the CCD images presented in this work sample only the central region of clusters and therefore it is not feasible to determine corrections for each cluster individually. We used the mean counts in 5 control fields to estimate the background magnitude distribution. The mean number of interlopers subtracted from each field is 12 with an rms variation of $`\pm 4.6`$ among the 5 different control regions.
The comparison between spectroscopic and photometric redshifts is shown in Figure 4 and the dispersion of $`z_{spect}z_{phot}`$ is $`\sigma _z=0.03`$. This residual is of the same order as found by other works such as Connolly et al. (1995), Brunner et al. (1997), and Yee et al. (1999). Connolly et al. (1995) were the first to use an empirical approach to estimate photometric redshifts and using photographic data found a dispersion of $`\sigma _z=0.047`$ when comparing spectroscopic and photometric redshifts for galaxies out to $`z0.5`$. Brunner et al. (1997), using CCD observations in 4 photometric bands, were able to obtain a relation with dispersion of $`\sigma _z=0.023`$ for clusters in the redshift range between $`0.0`$ and $`0.4`$. Recently, Yee et al. (1999) used the C-M relation to determine redshifts with a mean dispersion of $`\sigma _z=0.028`$ for clusters at $`0.1<z<0.7`$.
We also estimated photometric redshifts using only the early-type population of the clusters. The selection of cluster early-type galaxies can be done with an efficiency better than $`90\%`$, based on two-color and morphological information (Pahre 1998, 1999). The dispersion between spectroscopic and photometric redshifts obtained from the early-type galaxies brighter then $`21^m`$ is $`\sigma _z=0.04`$. Comparing this result with the previous one, it is clear that using all the galaxies in the central region of the clusters is consistent with using elliptical galaxies only. This is consistent with the central region of rich, regular clusters being populated mainly by early-type galaxies (morphology-density relation, Dressler 1980, 1997).
Cluster coordinates and redshifts are presented in Table 3. The ACO cluster number or candidate identification is given in column 1; R.A. and Dec. coordinates for 1950 are listed in columns 2 and 3; the Abell richness class is given in column 4; The redshift is given in column 5, and the reference for this redshift is listed in column 6. Column 7 gives the fraction of blue galaxies calculated as described in 4.3.
### 4.3 Butcher-Oemler Effect
Butcher & Oemler (1984) (hereafter BO84) searched for signs of evolution in the galaxy population of clusters studying a sample of 33 clusters of galaxies with redshifts between 0.003 (Virgo) and 0.54. The galaxies brighter than $`M_V=20`$, inside the central region containing 30% of the cluster population ($`R_{30}`$) were selected, and those with rest-frame $`BV`$ colors at least 0.2 mag bluer than the ridge of the early-type galaxies in the color-magnitude diagrams were classified as “blue”. The authors defined the fraction of blue galaxies ($`f_B`$) as the ratio between the number of blue and total number of galaxies, and found that $`f_B`$ is approximately constant until $`z0.1`$ and starts to increase linearly for higher redshift clusters.
Our CCD images are $``$ $`13.5^{}`$ x $`13.5^{}`$ and sample only the central part of the clusters. For this reason it was not possible to determine $`R_{30}`$ for each cluster individually and we decided to adopt a fixed physical size of radius $`0.7`$ Mpc, which corresponds to the mean $`R_{30}`$ used by BO84. We then selected the galaxies inside this region which, after k(z) correction, were found to be in the magnitude range between $`M^{}1`$ and $`M^{}+3`$, where $`M^{}=20.91`$ (Lin et al. 1996) assuming $`h=0.75`$. The galaxies with $`gr`$ colors at least 0.2 mag bluer than the locus of the early-type galaxies in the color-magnitude diagrams (Figure 3) were identified as blue, because a change of 0.2 mag in the $`BV`$ color of an elliptical galaxy corresponds to approximately the same change in its $`gr`$ color (Jorgensen 1995).
The background correction is very important in this study, and it tends to decrease $`f_B`$ since the field galaxies are in general bluer than the early-type population typically found in the central region of clusters (Dressler 1980). BO84 estimated background corrections in different ways for different clusters, but their main idea was to use the population of galaxies around the clusters. Our fields are too small to allow such corrections, and therefore we used 5 control fields observed during the first run to estimate the background correction. Since the number of background galaxies contaminating each cluster depends on its redshift, and the number of background galaxies which would be identified as blue depends also on each cluster C-M relation, the background correction was applied individually. Usually, the number of galaxies r between $`M^{}1`$ and $`M^{}+3`$ (where $`M^{}=20.91`$ assuming $`h=0.75`$, Lin et al. 1996) detected in a cluster image is 94, and the background counts are 38, with a variation of $`\pm 8`$ galaxies among the 5 different control regions. The number of blue galaxies in on average 23, and there are usually $`9\pm 2`$ background galaxies that would be selected as blue.
Several corrections need to be applied to guarantee that the same physical region, and the same interval in the luminosity function (LF) of the clusters, are being used to compute the fraction of blue galaxies. Clusters at low redshift suffer from two different problems: 1) our CCD fields are not big enough to observe a region of radius $`0.7`$ Mpc; and 2) the brightest objects (around $`M^{}1`$) are saturated. For clusters at higher redshifts, the fainter objects (around $`M^{}+3`$) are not detected. To estimate the effect of measuring $`f_B`$ at different regions in the cluster we used 5 clusters in the redshift range $`z=0.160.18`$ for which the angular sizes are small, and for which the entire magnitude interval from $`M^{}1`$ to $`M^{}+3`$ can be observed. The mean fraction of blue galaxies computed at different radius is shown in Figure 5, and the errorbars are $`rms/\sqrt{5}`$. Fitting a linear relation to the points we find:
$$f_B=0.103\times (R/0.7)+0.185(\sigma _{f_B}=0.093)$$
(2)
where $`R`$ is the radius in Mpc.
The idea that the fraction of blue galaxies increases in the outer parts of the cluster is in agreement with previous results. This behavior was already noticed by BO84 in their original work, and Rakos et al. (1997) also observe this trend in the detailed study of Abell 2317. This is also consistent with the work by Dressler and Gunn (1992), Dressler et al. (1994) and Oemler et al. (1997), which shows that early-type galaxies cluster more strongly than those with signs of recent or ongoing star-formation.
To estimate the effect of losing objects off the bright or faint ends of the chosen magnitude range, we use the sample of clusters at moderate redshift ($`0.085<z<0.175`$), for which we are able to observe the entire interval of $`M^{}1`$ to $`M^{}+3`$ and therefore to calculate $`f_B`$ at various limiting magnitudes. The sample was subdivided in two ($`0.085<z<0.130`$ and $`0.130<z<0.175`$) in order to check if clusters at different redshifts show a different distribution of luminosities of its blue galaxy population. Figure 6a shows the result of missing the cluster brightest objects ($`f_B`$ versus $`\mathrm{\Delta }_{mag}`$, where $`\mathrm{\Delta }_{mag}=(M^{}+3)(M^{}\mathrm{?})`$), and Figure 6b indicates the effect of losing the faintest ones ($`f_B`$ versus $`\mathrm{\Delta }_{mag}`$, where $`\mathrm{\Delta }_{mag}=(M^{}+\mathrm{?})(M^{}1)`$). Two main results can be draw from these figures: (1) the fraction of blue galaxies inside a fixed radius of $`0.7`$ Mpc is larger for clusters at higher redshift (open circles) than for the lower redshift ones (open squares) (Buther-Oemler effect); (2) $`f_B`$ increases as brighter objects are excluded from the sample (Fig 6a) and decreases if the faint objects are missing (Fig 6b), indicating that the blue galaxy population is faint. The results are the same for the two subsamples, and the slope of the $`f_B`$ versus $`\mathrm{\Delta }_{mag}`$ relations does not show significant change with redshift. We therefore use the entire sample of clusters (between $`0.085<z<0.175`$) to estimate the effect of measuring $`f_B`$ at different intervals of magnitude. The following two equations indicate the linear fits for losses from the bright and faint ends, respectively, obtained from the total mean points (solid dots).
$$f_B=0.034\times \mathrm{\Delta }_{mag}+0.362(\sigma _{f_B}=0.064)$$
(3)
$$f_B=+0.055\times \mathrm{\Delta }_{mag}+0.016(\sigma _{f_B}=0.050)$$
(4)
The point at $`\mathrm{\Delta }_{mag}=2.0`$ in Figure 6b, corresponding to the magnitude interval between $`M^{}1`$ and $`M^{}+1`$, was excluded from the fitting procedure. This point was ignored because when the faint objects are excluded the number of cluster galaxies drops dramatically, significantly increasing the errors in $`f_B`$. This effect is not as important when the brightest objects are missing, since the number of faint objects is always much larger than the number of bright objects. Although BO84 were limited to much brighter objects, they find the same tendency observed in our sample. The blue galaxies are also the faintest ones. Rakos et al. (1997) found an opposite trend in A2317 ($`z=0.211`$). The authors calculated $`f_B`$ at four magnitude bins and found that $`f_B`$ is highest in the brighter bin, drops in the two intermediate bins, and rises again in the fainter one. The fact that the blue galaxies in more evolved, low redshift clusters are faint, and that a greater number of bright blue galaxies is found in higher redshift or less evolved clusters seems to support the idea that BO galaxies are late-type, low surface brightness objects, which fade after a burst of star formation (RS95).
The CCD images presented here are deeper than the photographic plates used by BO84, therefore allowing us to include fainter galaxies in our $`f_B`$ estimate. On the other hand, BO84 were able to measure the brightest cluster objects that are usually saturated in our images. As it can be seen from Figures 6a and 6b the differences in the magnitude interval used to determine $`f_B`$ strongly affects its measurements. In order to compare our results with BO84 we need to “correct” our fraction of blue galaxies as if we had used the same limiting magnitudes. If we assume that the typical $`gr`$ color of an elliptical galaxy at $`z=0`$ is 0.44 (Small 1996), the $`M_V=20`$ limiting magnitude used by BO84 corresponds approximately to $`M_r=20.19`$ (Jorgensen et al. 1995). This is the same as limiting our sample at $`M^{}+0.72`$, which results in a reduction of 0.13 (Figure 6b) from our $`f_B`$ values. BO84 give no limiting magnitude for brighter objects, so we apply no correction for this effect. However, we do not expect this to be a major problem since it can be seen from Figures 6a and 6b (and equations 3 and 4) that the effect of missing bright objects is weaker than that of missing faint objects. Comparing the two clusters that we have in common with BO84, we find that their $`f_B`$ measurements are $`0.13`$ (for Abell 1689) and $`0.16`$ (for Abell 370) lower than ours, indicating that our correction of 0.13 is reasonable.
The final fractions of blue galaxies are shown in column 7 of Table 3 and in Figure 7a. The quoted $`f_B`$ errors are $`(N)^{1/2}`$ of the number of galaxies after background correction. Abell 1993 shows as extremely high fraction of blue galaxies ($`0.606\pm 0.141`$) that can be due to a wrong redshift estimation and/or to the common problem of projection effects existent in catalogs of clusters constructed from visual inspection of images. This is a not very well studied cluster and the last reference is from Abell et al. (1989) and for this reason we decide not to include it in Figure 7 and in the following discussion. Figure 7b shows the mean $`f_B`$ values for intervals of 0.04 in redshift for the clusters with spectroscopic redshifts (filled circles), and for the clusters with estimated photometric redshifts (open circles). Both photometric and spectroscopic redshift samples show the same behavior in the $`f_B\times z`$ plot. In Figure 7c we plot the mean $`f_B`$ values for the clusters in our sample (filled circles), for BO84’s clusters (open squares), and for 10 x-ray clusters observed by Smail et al. (1998) (crosses). The solid and dashed lines shown in figure 7 represent respectively the relation originally found by BO84, and the one estimated by Rakos and Schombert (1995, hereafter RS95) in their study of 17 clusters. The dotted line represent a linear fit to the binned mean points from our sample.
In order to check the adherence of the points to BO84, RS95 and to our estimated $`f_B\times z`$ relation we plot in Figure 7d the accumulated product $`f_B\times \mathrm{\Delta }z`$ for our data and for the three relations. The data seems more consistent with our relation and also with RS95 relation which is very similar to ours except for an offset due to the negative fraction of blue galaxies that we find at lower redshifts. The data shows an stronger increase in the fraction of blue galaxies with redshift than estimated by B084. The dot-dash line indicates the behavior assuming no evolution, or a constant $`f_B`$ of 0.09. This is a mean value estimated from the clusters with spectroscopic redshiftin our sample, and it is clear that $`f_B`$ is not constant: there is an absence of blue galaxies at local clusters, and an excess of blue objects at higher redshifts.
Andreon and Ettori (1999) used x-ray measurements of BO84 sample to argue that these clusters are not part of the same class of objects, and that therefore the increase in the fraction of blue galaxies with redshift might be simply a selection effect. Smail et al. (1998) studied the Butcher-Oemler effect in clusters at $`0.220.28`$ with similar x-ray luminosity, which should result in a sample consisting of the same kind of objects, and did not find the evolutionary trend shown in BO84, RS95 and in our data. Their observations, binned in redshift, are indicated by two crosses in Figure 7c and lie well below previous observations. It is interesting to note that if we concentrate our attention only in the redshift range between $`0.08`$ and $`0.2`$ (see Figure 7a), where we observed 10 times more clusters than BO84 and where most of our sample lies, we find objects with a wide range in $`f_B`$. These observations, as well as Smail et al. (1998) measurements, might be indicating that the Butcher-Oemler effect is at least partially due to selection effects that are still not well understood. A complete, statistically significant sample of clusters, representative of all kinds of clusters at a certain redshift, should be able to resolve this question. Unfortunately, such a complete sample is very hard to observe, and any sample will always be biased toward richer, more luminous systems, but a very large sample should contain most kinds of clusters at a given redshift. With this purpose we are now analyzing a sample of $`500`$ clusters that were observed in the same way described in this work. This large CCD sample of Abell clusters was primarily obtained to calibrate the POSS-II (Second Palomar Sky Survey) photographic plates and is now also being used to study the population of galaxies in clusters and its evolution.
### 4.4 Anomalously Red Galaxies?
In each cluster we selected as red all the galaxies with $`gr`$ and $`ri`$ colors $`0.3^m`$ above the lines in Figure 3, and brighter than $`m_r=22^m`$. The upper panels in Figure 8 show the histograms of $`m_r`$, $`gr`$, and $`ri`$ of the red galaxies. The filled histograms represent objects with $`ri1`$. As it can be seen from upper panel, most of those objects are faint having $`m_r`$ between $`20.5^m`$ and $`22^m`$ and have $`gr`$ colors varying from $`0.8`$ up to $`2.4`$. Objects with $`gr1.8`$ are red enough to be in agreement with the observation of a high redshift ($`z0.7`$) population of galaxies with $`ri1`$. Such red colors are due to the heavy k-corrections for elliptical galaxies at $`z>0.4`$ (see Figure 2 of Gal et al. 1999 for $`(gr)\times (ri)`$ tracks for diferent types of galaxies as a function of redshift). At $`z0.4`$ the 4000 Åbreak feature typical of elliptical galaxies enters the $`g`$ band causing a dramatic drop in its blue luminosity. The objects with $`gr1.2`$ seem to correspond to the anomalously red galaxies found by Garilli.
Most of the stars with $`ri1`$ have $`gr1.2`$ (Kennefick 1996), so stellar contamination could be responsible for at least part of this red population. In order to study the seeing effect in this population, the sample was divided in two: one with better seeing ($`FWHM<1.44^{\prime \prime }`$), and other with worse seeing ($`FWHM>1.44^{\prime \prime }`$) in r band. The bad seeing observations shown in the central panels of Figure 8 present mainly one peak around $`1.2^m`$ on the $`gr`$ distribution. But the $`gr`$ distribution from the better seeing images (lower panels in Figure 8) shows also the presence of redder objects, which indicates that at least part of those anomalously red objects are in fact stars that were misclassified as galaxies. This effect should be even stronger for the Garilli et at. (1996) data, which had an average poor seeing of $`2^{\prime \prime }`$. Furthermore, only $`42\%`$ of the “anomalously” red galaxies classified as galaxies in $`r`$ were also classified as galaxies in $`g`$ and $`i`$, while $`71\%`$ of the rest of the $`r`$ galaxies were classified as galaxies in all filters.
## 5 SUMMARY AND CONCLUSIONS
We present a photometric study of 48 clusters of galaxies. Color-magnitude diagrams are shown for the entire sample, and used to study the Butcher-Oemler effect and the population of red galaxies. We also present photometric redshifts for the clusters for which there are not spectroscopic measurements. The photometric redshift is estimated from the colors and magnitudes of cluster galaxies and has an accuracy of $`0.03`$. Our main conclusions are summarized as follows:
1. The blue, Butcher-Oemler galaxies, lie preferentially in the outer edges of the cluster, and are in general fainter than the cluster early-type population.
2. We find that the increase in the fraction of blue galaxies with redshift seems more consistent with the relation estimated by RS95 than with the original one by BO84. We predict a linear $`f_B\times z`$ relation consistent with RS95 estimation for $`z>0.08`$ but in which the fraction of blue galaxies continues to decrease for lower redshifts. This result is consistent with the observation of a smaller number of blue galaxies in the cluster region than would be observed in a blank field region of the same angular area. However, there is no redshift evolution of $`f_B`$ in the range between $`0.08`$ and $`0.2`$, where most of our clusters lie. Also, a large range of $`f_B`$ values is seen between $`0.2`$ and $`0.3`$ when Smail et al. (1998) x-ray clusters are added to the $`f_B\times z`$ plot. We are currently analyzing a sample of $`500`$ clusters which might help in understanding the selection effects and may clarify the results.
3. Studying the population of red galaxies in our images we found that some of these objects have colors that are normal for background galaxies, and that the colors of the supposably anomalous red objects are typical of stars. The fact that these objects are near the limiting magnitude of our sample, and that they are found in greater number at the bad seeing images, lead us to the conclusion that at least part of those objects are stars that were misclassified as galaxies.
We would like to thank D. Wittman, J.A. Tyson, H.V. Capelato, R.R. Gal, and S.C. Odewhan for very helpful comments and suggestions which helped to improve the paper. We also thank the anonymous referee for the detailed revision and usefull recommendations provided for this work. |
no-problem/0001/math0001078.html | ar5iv | text | # 1 Introduction
## 1 Introduction
The Rogers-Ramanujan identities , ,
$$1+\underset{n=1}{\overset{\mathrm{}}{}}\frac{q^{n^2}}{(1q)(1q^2)\mathrm{}(1q^n)}=\underset{n=1}{\overset{\mathrm{}}{}}\frac{1}{(1q^{5n1})(1q^{5n4})}$$
$$1+\underset{n=1}{\overset{\mathrm{}}{}}\frac{q^{n(n+1)}}{(1q)(1q^2)\mathrm{}(1q^n)}=\underset{n=1}{\overset{\mathrm{}}{}}\frac{1}{(1q^{5n2})(1q^{5n3})}$$
are among the most interesting partition identities in number theory and combinatorics. Combinatorial aspects of these identities are discussed by Andrews ,. Lepowsky and Wilson , , connect the Rogers-Ramanujan identities with affine Lie algebras and conformal field theory. Feigen and Frenkel interpret them as a character formula for the Virasoro algebra. The author applies the product form to computational group theory, obtaining a simple formula for the $`n\mathrm{}`$ limit of the probability that an element of $`GL(n,q)`$ is semisimple. Stembridge proves them by adapting a method of Macdonald for calculating partial fraction expansions of symmetric formal power series. Garsia and Milne offer a lengthy bijective proof, giving birth to the “involution principle” in combinatorics. Garrett, Ismail, and Stanton usefully recast Roger’s original proof using orthogonal polynomials. Relations with statistical mechanics appear in , , and . The paper gives an account of their appearance in physics.
One of the main results of this note is a first probabilistic proof of the Rogers-Ramanujan identities. In fact the $`i=1`$ and $`i=k`$ cases of Theorem 1 will be proved. Theorem 1 is due to Andrews . Bressoud then connected it with an earlier combinatorial result of Gordon .
###### Theorem 1
() Let $`x_n`$ denote $`(1x)\mathrm{}(1x^n)`$. For $`1ik,k2`$,
$$\underset{n_1,\mathrm{},n_{k1}0}{}\frac{x^{N_1^2+\mathrm{}+N_{k1}^2+N_i+\mathrm{}+N_{k1}}}{(x)_{n_1}\mathrm{}(x)_{n_{k1}}}=\underset{\genfrac{}{}{0pt}{}{r=1}{r0,\pm i(mod2k+1)}}{\overset{\mathrm{}}{}}\frac{1}{1x^r}$$
where $`N_j=n_j+\mathrm{}n_{k1}`$.
There are other elementary proofs of the Rogers-Ramanujan identities available (e.g. ), but the one offered here is possibly the most natural and gives insight into Bailey’s Lemma.
The basic object of study in this paper is a certain one-parameter family of probability measures $`M_u`$ on the set of all partitions of all natural numbers, studied in the prior article . Section 2 recalls these measures and gives their group theoretic motivation. (For now we remark that they arise in the study of conjugacy classes of finite classical groups. As conjugacy classes of compact Lie groups are essentially eigenvalues up to the action of the Weyl group, their probabilistic study can be regarded as philosophically analogous to work of Dyson , who described the eigenvalues of random matrices of compact Lie groups in terms of Brownian motion).
Section 3 shows how to construct these measures using a non-upward-moving Markov chain on the integers. The Markov chain is diagonalizable with eigenvalues $`1,\frac{u}{q},\frac{u^2}{q^4},\mathrm{}`$, and a basis of eigenvectors is given. Analogous computations are done for a measure related to the uniform measure on partitions, which by work of Fristedt has a Markov chain approach. It would be interesting to make a connection with the articles , in which a fascinating continuous space Markov chain arises in the asymptotic probability theory of the symmetric group.
With these preliminaries in place, Section 4 gives a proof of the Rogers-Ramanujan identities. The idea of the proof is simple. We compute in two ways the $`L\mathrm{}`$ probability that the Markov chain started at $`L`$ is absorbed at the point $`0`$ after $`k`$ steps. (Since the Markov chain is absorbed at 0 with probability 1 and the measure $`M_u`$ corresponds to the $`L\mathrm{}`$ limit, the time to absorption really is the most natural quantity one could consider). The sum side of the Rogers-Ramanujan identities follows from the definition of the probability measures. For the product side, the fact that the transition matrix is explicitly diagonalizable gives a sum expression. One then applies the Jacobi triple product identity (which as explained on page 21 of follows easily from the $`q`$-binomial theorem) and the proof is complete.
Section 4 continues by discussing Bailey’s Lemma, which is the only non-trivial step in many of the simplest proofs of the Rogers-Ramanujan identities. The most useful case of Bailey’s Lemma follows immediately once a basis of eigenvectors of the transition matrix has been found. The only non-trivial step in our proof of the Rogers-Ramanujan identities is finding a basis of eigenvectors; however with Mathematica this was easy. By contrast it is unclear how one would guess at Bailey’s Lemma.
Section 5 reviews the theory of quivers and shows that the Markov chain method extends to quivers. Although we have not invested serious effort into finding analogs of Bailey’s Lemma for quivers other than the one point quiver (which corresponds to conjugacy classes of the finite general linear groups), it is not hard to see that the resulting Bailey Lemmas differ from those of and .
It is tempting to speculate that there is a relationship between conjugacy classes of the finite general linear groups and modular forms. Aside from this paper, there are two good reasons to suspect this. One reason is that the conjugacy classes are related to Hall-Littlewood polynomials, which in turn are related to vertex operators . Other evidence is work of Bloch and Okounkov , who relate a version of the uniform measure on partitions to quasi-modular forms.
## 2 Measures on Partitions and Group Theory
We begin by reviewing some standard notation about partitions, as on pages 2-5 of Macdonald . Let $`\lambda `$ be a partition of some non-negative integer $`|\lambda |`$ into parts $`\lambda _1\lambda _2\mathrm{}`$. Let $`m_i(\lambda )`$ be the number of parts of $`\lambda `$ of size $`i`$, and let $`\lambda ^{}`$ be the partition dual to $`\lambda `$ in the sense that $`\lambda _i^{}=m_i(\lambda )+m_{i+1}(\lambda )+\mathrm{}`$. Let $`n(\lambda )`$ be the quantity $`_{i1}(i1)\lambda _i`$. It is also useful to define the diagram associated to $`\lambda `$ as the set of points $`(i,j)Z^2`$ such that $`1j\lambda _i`$. We use the convention that the row index $`i`$ increases as one goes downward and the column index $`j`$ increases as one goes across. So the diagram of the partition $`(5441)`$ is:
$$\begin{array}{ccccc}.& .& .& .& .\\ .& .& .& .& \\ .& .& .& .& \\ .& & & & \end{array}$$
The rest of this section follows the paper . Let $`q`$ be the size of a finite field. To begin we recall a way of defining a one parameter family of probability measures $`M_u(\lambda )`$ on the set of all partitions of all natural numbers. If one simply wants a formula, then all of the following definitions are equivalent. In the third expression, $`P_\lambda `$ denotes a Hall Littlewood polynomial as in .
$`M_u(\lambda )`$ $`=`$ $`\left[{\displaystyle \underset{r=1}{\overset{\mathrm{}}{}}}(1{\displaystyle \frac{u}{q^r}})\right]{\displaystyle \frac{u^{|\lambda |}}{q^{_i(\lambda _i^{})^2}_i(\frac{1}{q})_{m_i(\lambda )}}}`$
$`=`$ $`\left[{\displaystyle \underset{r=1}{\overset{\mathrm{}}{}}}(1{\displaystyle \frac{u}{q^r}})\right]{\displaystyle \frac{u^{|\lambda |}}{q^{2[_{h<i}hm_h(\lambda )m_i(\lambda )+\frac{1}{2}_i(i1)m_i(\lambda )^2]}_i|GL(m_i(\lambda ),q)|}}`$
$`=`$ $`\left[{\displaystyle \underset{r=1}{\overset{\mathrm{}}{}}}(1{\displaystyle \frac{u}{q^r}})\right]{\displaystyle \frac{u^{|\lambda |}P_\lambda (\frac{1}{q},\frac{1}{q^2},\mathrm{};0,\frac{1}{q})}{q^{n(\lambda )}}}`$
The only fact from this section which is required for the proof of the Rogers-Ramanujan identities is the fact that $`M_u`$ defines a probability measure for $`0<u<1`$. This can be seen using an identity from either or else following Stong , who uses the fact that there are $`q^{n^2n}`$ unipotent elements in $`GL(n,q)`$. The first proof of Theorem 2 will use the fact that $`M_u`$ is a probability measure without further comment. As there has been interest in simplifying the proofs of the Rogers-Ramanujan identities as much as possible, a second completely elementary proof of Theorem 2 will be given. From this second proof it will follow that $`M_u(\lambda )`$ is a probability measure.
Although not logically necessary for this paper, we mention that for $`0<u<1`$ and $`q`$ a prime power, the measures $`M_u`$ have a group theoretic description, referring the reader to the survey for further discussion. Recall that the conjugacy classes of $`GL(n,q)`$ are parameterized by rational canonical form. Each such matrix corresponds to the following combinatorial data. To every monic non-constant irreducible polynomial $`\varphi `$ over $`F_q`$, associate a partition (perhaps the trivial partition) $`\lambda _\varphi `$ of some non-negative integer $`|\lambda _\varphi |`$. The only restrictions necessary for this data to represent a conjugacy class are that $`|\lambda _z|=0`$ and $`_\varphi |\lambda _\varphi |deg(\varphi )=n.`$ To be explicit, a representative of the conjugacy class corresponding to the data $`\lambda _\varphi `$ may be given as follows. Define the companion matrix $`C(\varphi )`$ of a polynomial $`\varphi (z)=z^{deg(\varphi )}+\alpha _{deg(\varphi )1}z^{deg(\varphi )1}+\mathrm{}+\alpha _1z+\alpha _0`$ to be:
$$\left(\begin{array}{ccccc}0& 1& 0& \mathrm{}& 0\\ 0& 0& 1& \mathrm{}& 0\\ \mathrm{}& \mathrm{}& \mathrm{}& \mathrm{}& \mathrm{}\\ 0& 0& 0& \mathrm{}& 1\\ \alpha _0& \alpha _1& \mathrm{}& \mathrm{}& \alpha _{deg(\varphi )1}\end{array}\right)$$
Let $`\varphi _1,\mathrm{},\varphi _k`$ be the polynomials such that $`|\lambda _{\varphi _i}|>0`$. Denote the parts of $`\lambda _{\varphi _i}`$ by $`\lambda _{\varphi _i,1}\lambda _{\varphi _i,2}\mathrm{}`$. Then a matrix corresponding to the above conjugacy class data is:
$$\left(\begin{array}{cccc}C(\varphi _i^{\lambda _{\varphi _i,1}})& 0& 0& \\ 0& C(\varphi _i^{\lambda _{\varphi _i,2}})& 0& \\ 0& 0& \mathrm{}& \end{array}\right)$$
Now consider the following procedure for putting a measure on the set of all partitions of all natural numbers. Fix $`u`$ such that $`0<u<1`$. Pick a non-negative integer such that the chance of choosing $`n`$ is equal to $`(1u)u^n`$. Then pick $`\alpha `$ uniformly in $`GL(n,q)`$ and take $`\lambda `$ to be the paritition corresponding to the polynomial $`z1`$ in the rational canonical form of $`\alpha `$. If $`n=0`$ take $`\lambda `$ to be the trivial partition. The random partition so defined obeys $`M_u`$ measure. (The polynomial $`z1`$ is considered without loss of generality. Partitions corresponding to other irreducible polynomials are probabilistically independent, and one just replaces $`q`$ by raising it to the degree of the polynomial). In the limit $`u1`$, one is simply studying random elements in a fixed $`GL(n,q)`$ with $`n\mathrm{}`$. The substitutions $`uu`$ and $`qq`$ correspond to the finite unitary groups. The idea of auxilliary randomization of the dimension $`n`$ is analogous to the idea of canonical ensembles in statistical mechanics.
## 3 Markov chains
### 3.1 Group theoretical measures
The first result of this paper describes the measure $`M_u`$ in terms of Markov chains. Two proofs will be given. The first proof is given in the interest of clarity and assumes that $`M_u`$ is a probability measure. The second proof is more elementary.
It is convenient to set $`\lambda _0^{}`$ (the height of an imaginary zeroth column) equal to $`\mathrm{}`$. For the entirety of this subsection, $`(x)_n`$ will denote $`(1x)(1x/q)\mathrm{}(1x/q^{n1})`$. Thus $`(x)_0=1`$ and $`(x)_n=0`$ for $`n<0`$. For convenience of notation, let $`P(a)`$ be the $`M_u`$ probability that $`\lambda _1=a`$. $`Prob(E)`$ will denote the probability of an event $`E`$ under the measure $`M_u`$.
###### Theorem 2
Starting with $`\lambda _0^{}=\mathrm{}`$, define in succession $`\lambda _1^{},\lambda _2^{},\mathrm{}`$ according to the rule that if $`\lambda _i^{}=a`$, then $`\lambda _{i+1}^{}=b`$ with probability
$$K(a,b)=\frac{u^b(\frac{1}{q})_a(\frac{u}{q})_a}{q^{b^2}(\frac{1}{q})_{ab}(\frac{1}{q})_b(\frac{u}{q})_b}.$$
Then the resulting partition is distributed according to $`M_u`$.
Proof: The $`M_u`$ probability of choosing a partition with $`\lambda _i^{}=r_i`$ for all $`i`$ is
$$Prob.(\lambda _0^{}=\mathrm{})\frac{Prob.(\lambda _0^{}=\mathrm{},\lambda _1^{}=r_1)}{Prob.(\lambda _0^{}=\mathrm{})}\underset{i=1}{\overset{\mathrm{}}{}}\frac{Prob.(\lambda _0^{}=\mathrm{},\lambda _1^{}=r_1,\mathrm{},\lambda _{i+1}^{}=r_{i+1})}{Prob.(\lambda _0^{}=\mathrm{},\lambda _1^{}=r_1,\mathrm{},\lambda _i^{}=r_i)}.$$
Thus it is enough to prove that
$$\frac{Prob.(\lambda _0^{}=\mathrm{},\lambda _1^{}=r_1,\mathrm{},\lambda _{i1}^{}=r_{i1},\lambda _i^{}=a,\lambda _{i+1}^{}=b)}{Prob.(\lambda _0^{}=\mathrm{},\lambda _1^{}=r_1,\mathrm{},\lambda _{i1}^{}=r_{i1},\lambda _i^{}=a)}=\frac{u^b(\frac{1}{q})_a(\frac{u}{q})_a}{q^{b^2}(\frac{1}{q})_{ab}(\frac{1}{q})_b(\frac{u}{q})_b},$$
for all $`i,a,b,r_1,\mathrm{},r_{i1}0`$.
For the case $`i=0`$, the equation
$$P(a)=\frac{u^a(\frac{u}{q})_{\mathrm{}}}{q^{a^2}(\frac{1}{q})_a(\frac{u}{q})_a}$$
is given a probabilistic proof in . For an elementary proof of this identity, see the second proof of this theorem. For $`i>0`$ one calculates that
$$\underset{\genfrac{}{}{0pt}{}{\lambda :\lambda _1^{}=r_1,\mathrm{},\lambda _{i1}^{}=r_{i1}}{\lambda _i^{}=a}}{}M_u(\lambda )=\frac{u^{r_1+\mathrm{}+r_{i1}}}{q^{r_1^2+\mathrm{}+r_{i1}^2}(\frac{1}{q})_{r_1r_2}\mathrm{}(\frac{1}{q})_{r_{i2}r_{i1}}(\frac{1}{q})_{r_{i1}a}}P(a).$$
Similarly, observe that
$$\underset{\genfrac{}{}{0pt}{}{\lambda :\lambda _1^{}=r_1,\mathrm{},\lambda _{i1}^{}=r_{i1}}{\lambda _i^{}=a,\lambda _{i+1}^{}=b}}{}M_u(\lambda )=\frac{u^{r_1+\mathrm{}+r_{i1}+a}}{q^{r_1^2+\mathrm{}+r_{i1}^2+a^2}(\frac{1}{q})_{r_1r_2}\mathrm{}(\frac{1}{q})_{r_{i2}r_{i1}}(\frac{1}{q})_{r_{i1}a}(\frac{1}{q})_{ab}}P(b).$$
Thus the ratio of these two expressions is
$$\frac{u^b(\frac{1}{q})_a(\frac{u}{q})_a}{q^{b^2}(\frac{1}{q})_{ab}(\frac{1}{q})_b(\frac{u}{q})_b},$$
as desired. Note that the transition probabilities automatically sum to 1 because
$$\underset{ba}{}\frac{_{\genfrac{}{}{0pt}{}{\lambda :\lambda _1^{}=r_1,\mathrm{},\lambda _{i1}^{}=r_{i1}}{\lambda _i^{}=a,\lambda _{i+1}^{}=b}}M_u(\lambda )}{_{\genfrac{}{}{0pt}{}{\lambda :\lambda _1^{}=r_1,\mathrm{},\lambda _{i1}^{}=r_{i1}}{\lambda _i^{}=a}}M_u(\lambda )}=1.$$
$`\mathrm{}`$
Proof: (Second Proof) This proof needs only that $`M_u`$ is a measure; it will emerge that $`M_u`$ is a probability measure. For this proof $`P(a)`$ denotes the $`M_u`$ mass that $`\lambda _1^{}=a`$.
One calculates that
$$\underset{\genfrac{}{}{0pt}{}{\lambda :\lambda _1^{}=r_1,\mathrm{},\lambda _{i1}^{}=r_{i1}}{\lambda _i^{}=a}}{}M_u(\lambda )=\frac{u^{r_1+\mathrm{}+r_{i1}}}{q^{r_1^2+\mathrm{}+r_{i1}^2}(\frac{1}{q})_{r_1r_2}\mathrm{}(\frac{1}{q})_{r_{i2}r_{i1}}(\frac{1}{q})_{r_{i1}a}}P(a).$$
Similarly, observe that
$$\underset{\genfrac{}{}{0pt}{}{\lambda :\lambda _1^{}=r_1,\mathrm{},\lambda _{i1}^{}=r_{i1}}{\lambda _i^{}=a,\lambda _{i+1}^{}=b}}{}M_u(\lambda )=\frac{u^{r_1+\mathrm{}+r_{i1}+a}}{q^{r_1^2+\mathrm{}+r_{i1}^2+a^2}(\frac{1}{q})_{r_1r_2}\mathrm{}(\frac{1}{q})_{r_{i2}r_{i1}}(\frac{1}{q})_{r_{i1}a}(\frac{1}{q})_{ab}}P(b).$$
Thus the ratio of these two expressions is
$$\frac{P(b)u^a}{P(a)q^{a^2}(\frac{1}{q})_{ab}}.$$
Since $`M_u`$ is a measure, it follows that
$$\underset{ba}{}\frac{P(b)u^a}{P(a)q^{a^2}(\frac{1}{q})_{ab}}=1.$$
From this recursion and the fact that $`P(0)=(\frac{u}{q})_{\mathrm{}}`$, one solves for $`P(a)`$ inductively, finding that
$$P(a)=\frac{u^a(\frac{u}{q})_{\mathrm{}}}{q^{a^2}(\frac{1}{q})_a(\frac{u}{q})_a}.$$
Identity 2.2.8 on page 20 of now implies that $`_aP(a)=1`$, so that $`M_u`$ is a probability measure. $`\mathrm{}`$
The algorithm of Theorem 2 is nice in that it can be implemented on a computer. An analogous Markov chain approach to the conjugacy classes of the finite symplectic and orthogonal groups has recently been worked out in .
Theorem 3 explicitly diagonalizes the transition matrix $`K`$, which is fundamental for understanding the Markov chain. Note that if the current distribution of the Markov chain is given by a row vector, the distribution at the next step is obtained by multiplying the row vector on the right by $`K`$.
###### Theorem 3
1. Let $`C`$ be the diagonal matrix with $`(i,i)`$ entry $`(\frac{1}{q})_i(\frac{u}{q})_i`$. Let $`M`$ be the matrix $`\left(\frac{u^j}{q^{j^2}(\frac{1}{q})_{ij}}\right)`$. Then $`K=CMC^1`$, which reduces the problem of diagonalizing $`K`$ to that of diagonalizing $`M`$.
2. Let $`A`$ be the matrix $`\left(\frac{1}{(\frac{1}{q})_{ij}(\frac{u}{q})_{i+j}}\right)`$. Then the columns of $`A`$ are eigenvectors of $`M`$ for right multiplication, the $`j`$th column having eigenvalue $`\frac{u^j}{q^{j^2}}`$.
3. The inverse matrix $`A^1`$ is $`\left(\frac{(1u/q^{2i})(1)^{ij}(\frac{u}{q})_{i+j1}}{q^{\left(\genfrac{}{}{0pt}{}{ij}{2}\right)}(\frac{1}{q})_{ij}}\right)`$.
Proof: The first part is obvious. The second part is a special case of Lemma 1 of . The third part is a lemma of . $`\mathrm{}`$
The point of the proof of Theorem 3 is that once one knows (either from Mathematica or by implementing algorithms from linear algebra) what the eigenvectors are, it is a simple matter to verify the computation.
The following corollary will be useful for the proof of the Rogers-Ramanujan identities.
###### Corollary 1
Let $`E`$ be the diagonal matrix with $`(i,i)`$ entry $`\frac{u^i}{q^{i^2}}`$. Then $`K^r=CAE^rA^1C^1`$. More explicitly,
$$K^r(L,j)=\frac{(\frac{1}{q})_L(\frac{u}{q})_L}{(\frac{1}{q})_j(\frac{u}{q})_j}\underset{n=0}{\overset{\mathrm{}}{}}\frac{u^{rn}(1u/q^{2n})(1)^{nj}(\frac{u}{q})_{n+j1}}{q^{rn^2}(\frac{1}{q})_{Ln}(\frac{u}{q})_{L+n}q^{\left(\genfrac{}{}{0pt}{}{nj}{2}\right)}(\frac{1}{q})_{nj}}.$$
Proof: This is immediate from Theorem 3. $`\mathrm{}`$
### 3.2 Mixture of uniform measures
For this subsection $`q<1`$. The measure assigns probability $`q^{|\lambda |}_{i=1}^{\mathrm{}}(1q^i)`$ to the partition $`\lambda `$. Conditioning this measure to live on partitions of a given size gives the uniform measure, an observation exploited by Fristedt . As is clear from , this measure is very natural from the viewpoint of representation theory.
Fristedt (loc. cit.) proved that this measure has a Markov chain description. His chain affects row lengths rather than column lengths (though the algorithm would work on columns too as the measure is invariant under transposing diagrams). Nevertheless, we adhere to his notation. We use the notation that $`(x)_n=(1x)\mathrm{}(1x^n)`$.
###### Theorem 4
Starting with $`\lambda _0=\mathrm{}`$, define in succession $`\lambda _1,\lambda _2,\mathrm{}`$ according to the rule that if $`\lambda _i=a`$, then $`\lambda _{i+1}=b`$ with probability
$$K(a,b)=\frac{q^b(q)_a}{(q)_b}.$$
Then the resulting partition is distributed according to the measure of this subsection.
Theorem 5 diagonalizes this Markov chain, giving a basis of eigenvectors. The proof is analogous to that of Theorem 3, the second part being proved by induction.
###### Theorem 5
1. Let $`C`$ be the diagonal matrix with $`(i,i)`$ entry equal to $`\frac{(q)_i}{q^i}`$. Let $`M`$ be the matrix with $`(i,j)`$ entry $`q^i`$ if $`ij`$ and $`0`$ otherwise. Then $`K=CMC^1`$.
2. Let $`D`$ be the diagonal matrix with $`(i,i)`$ entry equal to $`q^i`$. Let $`A`$ be the matrix $`\left(\frac{(1)^{ij}}{q^{\left(\genfrac{}{}{0pt}{}{ij}{2}\right)}(\frac{1}{q})_{ij}}\right)`$, so that its inverse is $`\left(\frac{1}{(\frac{1}{q})_{ij}}\right)`$ by part b of Theorem 3. Then the eigenvectors of $`M`$ are the columns of $`A`$, the $`j`$th column having eigenvalue $`q^j`$.
As a corollary, one obtains a simple expression for the chance that under the measure of this subsection, the $`r`$th row has size $`j`$.
###### Corollary 2
$$K^r(L,j)=\frac{q^jq^{L(r1)}(q)_L(\frac{1}{q})_{Lj+r1}}{(q)_j(\frac{1}{q})_{Lj}(\frac{1}{q})_{r1}}.$$
Letting $`L\mathrm{}`$, the chance that the $`r`$th row has size $`j`$ becomes
$$\frac{(q)_{\mathrm{}}q^{rj}}{(q)_j(q)_{r1}}.$$
Proof: To obtain the first expression, one multiplies out $`K^r=CAD^rA^1C^1`$ and uses the $`q`$-binomial theorem
$$\underset{m=0}{\overset{\mathrm{}}{}}y^mq^{(m^2+m)/2}\frac{(q)_n}{(q)_m(q)_{mn}}=(1+yq)(1+yq^2)\mathrm{}(1+yq^n).$$
$`\mathrm{}`$
The second part of the corollary can be proved directly without recourse to Markov chain theory; one simply attaches to an $`rj`$ square two partitions: one with at most $`r1`$ rows and another with at most $`j`$ columns.
## 4 Rogers-Ramanujan Identities and Bailey’s Lemma
The first result of this section proves the following identity of Andrews, which contains the Rogers-Ramanujan identities. In this section $`(x)_n`$ denotes $`(1x)(1x/q)\mathrm{}(1x/q^{n1})`$.
###### Theorem 6
For $`k2`$,
$$\underset{n_1,\mathrm{},n_{k1}0}{}\frac{1}{q^{N_1^2+\mathrm{}+N_{k1}^2}(1/q)_{n_1}\mathrm{}(1/q)_{n_{k1}}}=\underset{\genfrac{}{}{0pt}{}{r=1}{r0,\pm k(mod2k+1)}}{\overset{\mathrm{}}{}}\frac{1}{1(1/q)^r}$$
$$\underset{n_1,\mathrm{},n_{k1}0}{}\frac{1}{q^{N_1^2+\mathrm{}+N_{k1}^2+N_1+\mathrm{}+N_{k1}}(1/q)_{n_1}\mathrm{}(1/q)_{n_{k1}}}=\underset{\genfrac{}{}{0pt}{}{r=1}{r0,\pm 1(mod2k+1)}}{\overset{\mathrm{}}{}}\frac{1}{1(1/q)^r}$$
where $`N_j=n_j+\mathrm{}n_{k1}`$.
Proof: For the first identity, we compute in two ways
$$\underset{\lambda :\lambda ^{}<k}{}M_1(\lambda ).$$
One obtains the sum side by using the first definition of $`M_1`$ in Section 2. For the product side, let $`u=1`$, $`j=0`$, $`r=k`$, and $`L\mathrm{}`$ in Corollary 1. The rest is now a standard argument.
$`1+{\displaystyle \underset{n=1}{\overset{\mathrm{}}{}}}{\displaystyle \frac{(1+1/q^n)(1)^n}{q^{rn^2}q^{\left(\genfrac{}{}{0pt}{}{n}{2}\right)}}}`$ $`=`$ $`1+{\displaystyle \underset{n=1}{\overset{\mathrm{}}{}}}(1)^n({\displaystyle \frac{1}{q}})^{(k+1/2)n^2n/2}+({\displaystyle \frac{1}{q}})^{(k+1/2)n^2+n/2}`$
$`=`$ $`{\displaystyle \underset{n=\mathrm{}}{\overset{\mathrm{}}{}}}(1)^n({\displaystyle \frac{1}{q}})^{(k+1/2)n^2}({\displaystyle \frac{1}{q}})^{n/2}.`$
Now invoke Jacobi’s triple product identity (e.g. )
$$\underset{n=\mathrm{}}{\overset{\mathrm{}}{}}(1)^nv^nw^{n^2}=\underset{n=1}{\overset{\mathrm{}}{}}(1vw^{2n1})(1w^{2n1}/v)(1w^{2n}).$$
The proof of the second identity is the same except that one takes $`u=\frac{1}{q}`$ instead of $`u=1`$. $`\mathrm{}`$
Next we discuss the most important case of Bailey’s Lemma, which was alluded to in and stated explicitly in . A pair of sequences $`\{\alpha _L\}`$ and $`\{\beta _L\}`$ are called a Bailey pair if
$$\beta _L=\underset{r=0}{\overset{L}{}}\frac{\alpha _r}{(1/q)_{Lr}(u/q)_{L+r}}.$$
Bailey’s Lemma states that if $`\alpha _L^{}=\frac{u^L}{q^{L^2}}\alpha _L`$ and $`\beta _L^{}=_{r=0}^L\frac{u^r}{q^{r^2}(1/q)_{Lr}}\beta _r`$, then $`\{\alpha _L^{}\}`$ and $`\{\beta _L^{}\}`$ are a Bailey pair.
From our viewpoint, this case of Bailey’s Lemma is clear. Namely let $`A,D,M`$ be as in Theorem 3 (recall that $`M=ADA^1`$). Viewing $`\alpha =\stackrel{}{\alpha _L}`$ and $`\beta =\stackrel{}{\beta _L}`$ as column vectors, the notion of a Bailey pair means that $`\beta =A\alpha `$. This case of Bailey’s Lemma follows because
$$\beta ^{}=M\beta =ADA^1\beta =AD\alpha =\alpha ^{}.$$
It would be interesting to obtain all of Bailey’s Lemma (and Theorem 1) by probabilistic arguments, and also to understand the Bailey lattice probabilistically. The reader may enjoy the survey by Bressoud .
## 5 Quivers
This section uses the notion of a quiver, as surveyed in Kac , to which the reader is referred for more detail. The basic set-up is as follows. Let $`\mathrm{\Gamma }`$ be a connected graph with $`n`$ vertices labelled as $`1,\mathrm{},n`$ (where we allow loops). Let $`N,Z`$ denote the natural numbers and integers respectively. Let $`f_{ij}`$ be the number of edges between $`i,j`$. Associated to $`\mathrm{\Gamma }`$ is a natural bilinear form on $`Z^n`$ and a root system $`\mathrm{\Delta }Z^n`$. Choose an arbitrary orientation of $`\mathrm{\Gamma }`$ so that $`\mathrm{\Gamma }`$ is a quiver. For a given dimension $`\alpha N^n\{0\}`$, let $`A_\mathrm{\Gamma }(\alpha ,q)`$ be the number of classes of absolutely indecomposable representations of $`\mathrm{\Gamma }`$ over the algebraic closure of a field of $`q`$ elements. It is proved in that $`A_\mathrm{\Gamma }(\alpha ,q)`$ is a polynomial in $`q`$ with integer coefficients, and that this polynomial is independent of the orientation of the graph. Kac (loc. cit.) formulated many conjectures about this polynomial. One such conjecture, which is still open, is that the constant term in $`A_\mathrm{\Gamma }(\alpha ,q)`$ is the multiplicity of $`\alpha `$ in the root system.
In recent work, Hua has given a completely combinatorial reformulation of this conjecture. To explain, let $`(1/q)_n`$ denote $`(1\frac{1}{q})\mathrm{}(1\frac{1}{q^n})`$, and for any two partitions $`\lambda ,\mu `$ define $`<\lambda ,\mu >=_{i1}\lambda _i^{}\mu _i^{}`$. Let $`t_j^\alpha `$ be the coefficient of $`q^i`$ in the polynomial $`A_\mathrm{\Gamma }(\alpha ,q)`$. Let $`b_\lambda =_{i1}(\frac{1}{q})_{m_i(\lambda )}`$, where $`m_i`$ is the number of parts of $`\lambda `$ of size $`i`$. Let $`\lambda (1),\mathrm{},\lambda (n)`$ be an $`n`$-tuple of partitions. Set $`U_\alpha =U_1^{\alpha _1}\mathrm{}U_n^{\alpha _n}`$. Hua’s result, which reduces Kac’s conjecture to a combinatorial assertion, is that
$$\underset{\lambda (1),\mathrm{},\lambda (n)}{}\frac{_{1ijn}q^{f_{ij}<\lambda (i),\lambda (j)>}U_1^{|\lambda (1)|}\mathrm{}U_n^{|\lambda (n)|}}{_{1in}q^{<\lambda (i),\lambda (i)>}b(\lambda (i))}=\underset{\alpha \mathrm{\Delta }^+}{}\underset{j=0}{\overset{deg(A_{\mathrm{\Gamma }(\alpha ,q)})}{}}\underset{i=1}{\overset{\mathrm{}}{}}(\frac{1}{1U_\alpha q^{ji}})^{t_j(\alpha )}.$$
A few points are in order. First, the right hand side of this equation is different from the statements in , due to a minor slip there. Second, observe that the expression converges in the ring of formal power series in the variables $`U_1,\mathrm{},U_n`$. This leads one to define a “probability” measure on $`n`$-tuples of partitions $`M_{\mathrm{\Gamma },\stackrel{}{U}}`$ by assigning mass
$$\underset{\alpha \mathrm{\Delta }^+}{}\underset{j=0}{\overset{deg(A_{\mathrm{\Gamma }(\alpha ,q)})}{}}\underset{i=1}{\overset{\mathrm{}}{}}(1U_\alpha q^{ji})^{t_j(\alpha )}\frac{_{1ijn}q^{f_{ij}<\lambda (i),\lambda (j)>}}{_{1in}q^{<\lambda (i),\lambda (i)>}b_{\lambda (i)}}$$
to the $`n`$-tuple $`\lambda (1),\mathrm{},\lambda (n)`$. For quivers of finite type this is a true probability measure for values of $`U_1,\mathrm{},U_n`$ sufficiently small, but in general we abuse notation by using terms from probability theory when $`U_1,\mathrm{},U_n`$ are variables.
Note that when the graph consists of a single point, this measure is simply the measure $`M_u`$ from Section 2. Theorem 7 shows that the structure of a Markov chain is still present. As the idea of the proof is the same as the second step of Theorem 2, the algebra is omitted. Let $`P(a_1,\mathrm{},a_n)`$ denote the $`M_{\mathrm{\Gamma },\stackrel{}{U}}`$ probability that $`\lambda (1),\mathrm{},\lambda (n)`$ have $`a_1,\mathrm{},a_n`$ parts respectively.
###### Theorem 7
Let $`\lambda (1)_1^{},\mathrm{},\lambda (n)_1^{}`$ be distributed as $`P(a_1,\mathrm{},a_n)`$. Define $`(\lambda (1)_2^{},\mathrm{},\lambda (n)_2^{})`$ then $`(\lambda (1)_3^{},\mathrm{},\lambda (n)_3^{})`$, etc. successively according to the rule that if $`(\lambda (1)_i^{},\mathrm{},\lambda (n)_i^{})`$ is equal to $`(a_1,\mathrm{},a_n)`$, then $`(\lambda (1)_{i+1}^{},\mathrm{},\lambda (n)_{i+1}^{})`$ is equal to $`(b_1,\mathrm{},b_n)`$ with probability
$$K(\stackrel{}{a},\stackrel{}{b})=\underset{1ijn}{}q^{f_{ij}a_ia_j}\underset{i=1}{\overset{n}{}}\frac{U_i^{a_i}}{q^{a_i^2}(\frac{1}{q})_{a_ib_i}}\frac{P(b_1,\mathrm{},b_n)}{P(a_1,\mathrm{},a_n)}.$$
The resulting $`n`$-tuple of partitions is distributed according to $`M_{\mathrm{\Gamma },\stackrel{}{U}}`$.
As a final remark, observe that letting $`C`$ be diagonal with entries $`\frac{1}{P(a_1,\mathrm{},a_n)}`$, one obtains a factorization $`K=CMC^1`$, where $`M`$ is defined by
$$M(\stackrel{}{a},\stackrel{}{b})=\underset{1ijn}{}q^{f_{ij}a_ia_j}\underset{i=1}{\overset{n}{}}\frac{U_i^{a_i}}{q^{a_i^2}(\frac{1}{q})_{a_ib_i}}.$$
A very natural problem is to investigate the eigenvector matrix $`E`$ of $`M`$, and $`E^1`$, in order to obtain new Bailey Lemmas.
## 6 Acknowledgments
This research was supported by an NSF Postdoctoral Fellowship. The author thanks Persi Diaconis for discussions and a referee for correcting historical mistakes in the bibliography.
Current address: Stanford University Math Department, Building 380, MC 2125, Stanford CA 94305, USA.
Current email: fulman@math.stanford.edu |
no-problem/0001/cond-mat0001115.html | ar5iv | text | # Potential Energy in a Three-Dimensional Vibrated Granular Medium Measured by NMR Imaging
## Abstract
Fast NMR imaging was used to measure the density profile of a three-dimensional granular medium fluidized by vertical vibrations of the container. For container acceleration much larger than gravity, the rise in center of mass of the granular medium is found to scale as $`v_0^\alpha /N^\beta `$ with $`\alpha =1.0\pm 0.2`$ and $`\beta =0.5\pm 0.1`$, where $`v_0`$ is the vibration velocity and $`N`$ is the number of grains in the container. This value for $`\alpha `$ is significantly less than found previously for experiments and simulations in one dimension ($`\alpha =2`$) and two dimensions ($`\alpha =\text{1.3-1.5}`$).
One technologically important way to fluidize a granular medium like sand is to vibrate the walls containing the system. Basic issues include the scaling of the kinetic and potential energy of the grains with parameters such as the frequency and amplitude of the wall vibration, and the number of layers of grains. These scaling relations are not yet fully understood for multidimensional systems, despite progress in simulations, experiments, and theory.
We consider a granular system in gravity excited by periodic vertical motion of the container $`z(t)=z_0cos(\omega t)`$. When the dimensionless acceleration $`\mathrm{\Gamma }=z_0\omega ^2/g`$ is larger than one, the granular system enters a state of internal motion determined in part by $`\mathrm{\Gamma }`$, the dimensionless frequency $`\overline{\omega }=\omega (d/g)^{1/2}`$, and the aspect ratio of the system. Here $`g`$ is the acceleration of gravity and $`d`$ is the grain diameter, so $`(d/g)^{1/2}`$ is the time for a grain to fall through its radius. For $`\overline{\omega }1`$ the container vibration induces large-scale structures in the granular medium such as density and surface waves. Surface structures are also observed for larger $`\overline{\omega }1`$ in very low aspect-ratio systems that permit small-wavenumber symmetry breaking. Conversely, for $`\overline{\omega }1`$ the container vibration is fast compared with the grain-scale motion. In this limit, the granular medium can assume a fluidized state analogous to a gas or liquid, and possibly amenable to theoretical methods developed for such systems. The major distinguishing characteristic of a granular fluid is the continuous energy loss to inelastic collisions, which if sufficiently great can destroy the statistical uniformity of the system via inelastic collapse.
In this paper we report an experimental study of a three-dimensional granular system in the fluidized regime $`\overline{\omega }4`$, $`\mathrm{\Gamma }14`$. Past numerical and experimental work has concentrated on one- and two-dimensional systems, due in part to limited computational resources and the lack of experimental methods suitable for examining the interiors of dense, flowing three-dimensional media. Recently, nuclear magnetic resonance (NMR) has emerged as a powerful noninvasive method for studying three-dimensional granular media. In the experiments reported here, NMR imaging was used to measure the density distribution during continuous vibration of the container. Incoherent grain motion limited image acquisition time to approximately 1 ms, too short to permit ordinary two-dimensional imaging. Therefore one-dimensional images were acquired showing the density projected onto the vertical axis $`\rho (z)`$. We find that $`\rho (z)`$ deviates strongly from the exponential form $`e^{Cz}`$ expected for a dilute isothermal gas in gravity. We focus here on the scaling of the gravitational potential energy, as measured by the center-of-mass height, with $`\mathrm{\Gamma }`$ and the number of grains $`N`$. In the present three-dimensional experiments the scaling is found to deviate even more strongly from ideal-gas predictions than was found in earlier two-dimensional studies.
The granular medium used for these experiments was composed of mustard seeds, of mean diameter 0.18 cm and mass 4.0 mg. Under magnification, the seeds appeared roughly spherical with typical eccentricity $`\pm 15\%`$. Seeds are used because their oily centers give strong, long-lived NMR signals. The seeds were held in a cylindrical container formed by a vertical glass tube of inside diameter 0.8 cm with a flat teflon bottom wall. The container was sufficiently tall to prevent collisions between the seeds and the top wall. It was determined visually that $`18\pm 2`$ seeds were required to complete a monolayer in this container. The experiments were carried out under ambient atmospheric pressure.
A vertical fiberglass tube was used to support the container at the center of the NMR probe, which was mounted into a vertical-bore superconducting solenoid. The static field was set to $`B_0=1.00\text{ T}`$ to reduce susceptibility contrast effects in these highly inhomogeneous samples. The lower end of the fiberglass tube was mounted to a loudspeaker, driven by a function generator and power amplifier, to provide vertical vibrations of the container. A flexible support near the top of the tube permitted the tube and container to move vertically while avoiding contact with the stationary NMR probe.
The waveform and amplitude of vertical motion were continuously measured by a micromachined accelerometer (Analog Devices ADXL50) mounted to the fiberglass tube, close to the NMR probe. This accelerometer was initially calibrated against a more accurate unit, and its output was digitized and fitted. Data were taken at two values of the vibration frequency $`\omega /2\pi =`$ 50 Hz and 40 Hz. Acceleration values were set to within $`\pm 0.05g`$.
One-dimensional NMR images were obtained using a spin-echo sequence. Gradient pulses of fixed amplitude 61.5 mT/m were used to encode vertical position information and an echo time TE = 1.0 ms was selected. Thus, the entire imaging sequence was complete within 1.5 ms. These parameters were chosen to achieve adequate spatial resolution $`\delta z=800\text{ }\mu `$m within a time much smaller than the vibration period while holding signal loss due to incoherent grain motion to an acceptably small level. This loss, measured by the echo size, was greater for smaller samples and larger vibration amplitudes. For a mid-size sample with $`N=50`$ the loss was approximately 10% at the highest amplitude used.
The NMR acquisition was triggered synchronously with the sample vibration. For each value of $`N`$ and $`\mathrm{\Gamma }`$, data for 100 different trigger delays spread uniformly over the vibration period were averaged together. Thus, the data presented here represent the sample density averaged over the vibration cycle. To remove effects of rf-field inhomogeneity the data were normalized to data obtained for a stationary water-filled tube.
Figure 1 shows the vertical density profile as a function of $`\mathrm{\Gamma }`$ for one of the $`(N,\omega )`$ points measured. With no vibration ($`\mathrm{\Gamma }=0`$) the profile is deeply corrugated, reflecting the layering of grains in the stationary chamber. As the vibration amplitude is increased, the layer structure is gradually smoothed out. This smearing is due to relative motion of the grains and not simply the vertical motion of the chamber, as the latter has an amplitude $`z_0=(155\text{ }\mu \text{m})\mathrm{\Gamma }`$ that is typically less than the grain diameter.
Even at the highest vibration amplitude, it is clear from Fig. 1 that the density profile deviates strongly from the exponential form predicted for an isothermal ideal gas. Although the density tail at the upper (free) surface might reasonably be fit to an exponential, the density levels off and then decreases as the lower surface is approached. This agrees qualitatively with features observed in experiments and simulations for two-dimensional systems.
To quantify the gravitational potential energy of the system, we have computed the center-of-mass height $`h_{cm}=(z\rho (z)𝑑z)/(\rho (z)𝑑z)`$. For each $`N`$, the limits of integration were kept fixed as $`\mathrm{\Gamma }`$ was varied to avoid introducing a systematic bias. Then the rise in center-of-mass height $`\mathrm{\Delta }h_{cm}`$ was computed by subtracting the $`\mathrm{\Gamma }=0`$ value of $`h_{cm}`$ for the same $`N`$. Figure 2 shows $`\mathrm{\Delta }h_{cm}(\mathrm{\Gamma },N)`$ for 50 Hz vibration frequency. The time-average height of the container is independent of vibration amplitude, so positive $`\mathrm{\Delta }h_{cm}`$ indicates an upwards displacement of the grains relative to the average container position.
Small systematic dips can be noted in Fig. 2 for specific values of $`N`$, for example $`N=32`$ and $`N=100`$. These do not reflect special states of the fluidized $`(\mathrm{\Gamma }1)`$ state for these $`N`$ values, but rather variations in the static $`(\mathrm{\Gamma }=0)`$ packing of the particles which affect the subtraction used to compute $`\mathrm{\Delta }h_{cm}`$. These dips could be removed from Figs. 2 and 3 by assuming a power-law form for $`\mathrm{\Delta }h_{cm}`$ and fitting to determine the $`\mathrm{\Gamma }=0`$ limit of the fluidized state. We prefer to display the data after subtracting the measured $`\mathrm{\Gamma }=0`$ $`h_{cm}`$ to avoid prejudging the functional form of $`\mathrm{\Delta }h_{cm}(\mathrm{\Gamma },N)`$.
To ascertain that functional form, we have plotted the $`\mathrm{\Delta }h_{cm}`$ data scaled by various powers of the drive amplitude $`z_0`$ and frequency $`\omega `$. Figure 3 shows the best data collapse obtained in this way, which is found to be $`\mathrm{\Delta }h_{cm}/v_0`$. Here $`v_0=z_0\omega `$ is the velocity amplitude of the vibration. In this plot, the data for low acceleration $`\mathrm{\Gamma }6`$ do not collapse while the data for high acceleration $`\mathrm{\Gamma }>6`$ all collapse onto a single function of $`N`$. The collapse is significantly better for $`\mathrm{\Delta }h_{cm}/v_0`$ than it is for $`\mathrm{\Delta }h_{cm}/v_0^{0.5}`$ or $`\mathrm{\Delta }h_{cm}/v_0^{1.5}`$, from which we conclude that $`\mathrm{\Delta }h_{cm}`$ scales as $`v_0^\alpha `$ with $`\alpha =1.0\pm 0.2`$. As shown by the dashed line on Fig. 3, the variation of $`\mathrm{\Delta }h_{cm}`$ with $`N`$ at high $`\mathrm{\Gamma }`$ is approximately $`N^{1/2}`$. From plots like this for a range of exponents we ascertain that $`\mathrm{\Delta }h_{cm}/v_0N^\beta `$ with $`\beta =0.5\pm 0.1`$.
A data set at fixed vibration frequency as shown in Fig. 3 can be used to determine the scaling with $`z_0`$ but not with $`\omega `$. When the data at vibration frequency 40 Hz are added to the scaling plot it appears that velocity $`z_0\omega `$ is a better scaling variable than is the amplitude $`z_0`$ or the acceleration $`z_0\omega ^2`$. However, due to the limited variation of $`\omega `$ this conclusion is not firmly established by our experiments.
The simplest model for a vibro-fluidized granular medium is the elastic hard-sphere gas. In this model the mean grain velocity is proportional to the container vibration velocity, and the potential energy is proportional to the kinetic energy hence $`\mathrm{\Delta }h_{cm}v_0^2`$. This velocity scaling has been observed in experiments and simulations of one-dimensional granular systems (columns of nearly elastic beads). Two-dimensional experiments and simulations found $`\mathrm{\Delta }h_{cm}v_0^\alpha `$ with significantly smaller $`\alpha =\text{1.3-1.5}`$, a result that is only partially understood. From a phenomenological viewpoint, our result $`\alpha 1`$ in three dimensions appears to continue the trend with increasing dimensionality away from elastic kinetic theory predictions.
For a less superficial understanding we must consider the reasons for deviations from ideal-gas theory that have been proposed. The steady state of the granular system is determined by a balance between work done on the grains by collisions with the vibrating container bottom, and energy lost due to inelastic collisions between the grains. (Other loss mechanisms discussed in the literature but not considered here include collisions with the side walls and viscous damping due to interstitial gas.)
We first discuss the power $`P_i`$ that is fed into the translational degrees of freedom by collisions with the vibrating bottom wall. An important dimensionless parameter is the ratio of the mean random grain velocity $`v_g`$ to the characteristic wall vibration velocity $`v_0`$. For nearly elastic systems (restitution coefficient close to one) with a small number of particles per unit wall area, it is possible to have $`v_g/v_01`$. Conversely, less elastic systems (including our experimental system) generally have $`v_g/v_01`$ in the uniformly fluidized state at large $`\overline{\omega }`$.
Collisions between the grains and the vibrating bottom wall always transfer upward momentum to the grains, but only do positive work on the grains when the wall is moving upward. When $`v_g/v_01`$ the grains are unable to catch up with the wall on its downward stroke. In this case, equating the momentum transfer per unit time with the sample weight $`N`$ gives the rigorous result $`P_iv_0N`$ . The same result obtains for an asymmetric sawtooth vibration waveform (infinitely fast downward stroke) for any value of $`v_g/v_0`$. Thus the symmetry of the drive waveform is only significant when $`v_g/v_01`$ . For a symmetric waveform, it has been suggested and verified for two-dimensional simulations that $`P_iv_0Nf(v_g/v_0)`$, where $`f(0)=1`$ and $`f(x)`$ is a decreasing function of $`x`$ . McNamara and Luding fit their simulation data for symmetric drive waveforms to $`f(x)=e^{Cx}`$. We suggest their data may also be approximately represented by $`f(x)=1/(1+x)`$. For large $`v_g/v_0`$, this gives $`P_iv_0^2N/v_g`$, a result that has been derived directly from kinetic theory for symmetric drive waveforms in the nearly elastic limit.
For a dilute, nearly elastic, isothermal system kinetic theory predicts that the density decays exponentially with height and the power lost to collisions $`P_cN^2v_g`$ . Here isothermal denotes uniform granular temperature $`Tv_g^2`$. In this limit potential energy is strictly proportional to kinetic energy, $`\mathrm{\Delta }h_{cm}T`$. Setting $`P_c=P_i`$, for $`v_g/v_01`$ or a sawtooth waveform the granular temperature scales as $`Tv_0^2/N^2`$. For symmetric waveforms with $`v_g/v_01`$ the scaling is $`Tv_0^2/N`$ . The latter corresponds well to experiments and simulations for one-dimensional systems, but not for two-dimensional systems or the three-dimensional experiment reported here.
Huntley has suggested that correlations between grain motions at high density reduce the collision frequency below the kinetic-theory result for an ideal gas. The power lost to collisions is found to scale as $`P_cN^{3/2}v_g^2`$. Combining this with the $`v_g/v_01`$ result for $`P_i`$, Huntley found $`Tv_0^{4/3}/N^{1/3}`$. This is much closer to the experimental and simulation results for two-dimensional systems than is the unmodified kinetic theory prediction. If Huntley’s expression for $`P_c`$ is set equal to $`P_i`$ for the $`v_g/v_01`$ condition appropriate to the present experiments, we obtain $`Tv_0/N^{1/2}`$. This agrees well with the scaling seen in our experiments, suggesting that the degree of inelasticity may be the controlling factor for scaling exponents rather than the system dimension.
Recently, McNamara and Luding carried out a series of two-dimensional simulations in which $`P_i`$, $`P_c`$, $`T`$, and $`\mathrm{\Delta }h_{cm}`$ were measured as the relevant dimensionless parameters were varied. As expected, they found $`\mathrm{\Delta }h_{cm}Tv_0^2`$ in the limit of very low density that occurs at high drive velocity $`v_0`$. For lower $`v_0`$, they found a crossover to $`Tv_0^{3/2}`$ which they traced to a reduction of $`P_c`$ below its kinetic-theory value when $`\mathrm{\Delta }h_{cm}`$ is small. As the detailed mechanism for this reduction is unknown, it is not clear how these results can be applied to the present three-dimensional experiments. McNamara and Luding also found that $`\mathrm{\Delta }h_{cm}T`$ is not well obeyed at high densities.
The trend found in Ref. of reduced power loss to collisions at high density agrees with Huntley’s suggestion, although the functional form may be different. As the power input $`P_i`$ appears reasonbly well understood, we also interpret our experiments as indicating reduced power loss to collisions $`P_c`$ for dense granular systems in three dimensions, as compared with ideal-gas kinetic theory predictions.
We thank N. Menon and R. A. Guyer for useful conversations. This work was supported by NSF grant DMR 9501171. |
no-problem/0001/nucl-ex0001002.html | ar5iv | text | # 1 Introduction
## 1 Introduction
Interesting experimental results on nuclear multifragmentation at low and intermediate energies have recently been obtained. The radial spherical symmetrical flow of fragments, or, in other words, the proportionality of the kinetic energy of a fragment to its mass was discovered in central collisions of gold projectile . The existence of the flow contradicts the assumption of a stochastic character of the nuclear multifragmentation process. The radial flow of spectator fragments in the rest frame of a fragmenting nucleus has been found at study of projectile gold nuclei interactions with photoemulsion nuclei at energy of 10.7 GeV/nucleon . According to estimations of Ref. , the energy of the radial motion ranges between 30 % and 50 % of accessible energy. At the same time, the another of the remarkable experimental result obtained by the ALADIN collaboration – the constancy of isotope temperature over the wide range of the excitation energy, is treated as an evidence of first order phase transition (liquid-gas) and statistical nature of the nuclear multifragmentation. The third experimental result reported by the INDRA collaboration – the independence of isotope composition of the fragments on the mass of fragmenting nuclei at the same excitation energy, - is interpreted as a sign of spinoidal instability of residual nuclei. There is a direct evidence of a perplexing situation.
We believe that the present state of the activity may be settled by studying the fragmentation of light enough systems (in the investigations mentioned heavy systems were analyzed). In fact, the famous statistical multifragmentation model (SMM) was in use to describe both heavy nuclei and oxygen residuals fragmentation . No less famous the quantum molecular dynamic model , in particular, the quantum antisymmetrized molecular dynamics model which incorporates effects of the mean field, are generally applied to light nucleus induced interactions. Hence, there is an intersection region where two approaches can be applied. The purpose of this work is to present the experimental data on multifragmentation of residuals created in 0.9 GeV/nucleon krypton initiated interactions in emulsion.
We start our analysis with the most interesting dependence of multiplicity of intermediate mass fragments (IMF) on the mass of fragmenting system whose measure is the so-called ”bound” charge:
$$Z_{bound}=\underset{F}{}Z_F,(Z_F2)$$
or
$$Z_{b3}=\underset{F}{}Z_F,(Z_F3),$$
where $`Z_F`$ is a charge of a fragment. IMFs have charges $`3Z_F30`$.
Next, we consider intrinsic characteristics of the fragmenting systems such as, the average charge of the largest fragment, asymmetry in the fragment system, etc. Our consideration is summarized in a brief summary.
## 2 Experimental material
Stacks of NIKFI BR-2 emulsion pellicles were exposed to a 1 A GeV $`{}_{}{}^{84}Kr`$ beam at the SIS/GSI and a 10.7 A GeV $`{}_{}{}^{197}Au`$ beam at the BNL/AGS. The sensitivity of the emulsion was not worse than 30 grains per 100 $`\mu `$k for singly charged particles with minimal ionization.
To carry out the analysis, the events induced by krypton nuclei with energy in the interval 0.8 – 0.95 GeV/nucleon were taken. The mean collision energy for this sample was thus reduced to about 0.9 GeV per projectile nucleon.
All the interactions were found by along-the-track-”fast-slow” scanning with a velocity excluding any discrimination in the event selection. A slow scanning (in backward direction) was made to find the events with unbiased projectile track. After excluding the events of electromagnetic dissociation and purely elastic scattering, a total of 677 krypton-emulsion interactions and 1057 gold-emulsion interactions heve been obtained.
Experimentally, the spectator fragments with $`Z=2`$ were classified by the visual inspection of tracks. The ionization of such tracks is constant over the whole length and equals $`g/g_0=4`$, where $`g_0`$ is the minimal ionization of singly charged track. Charge assignment for multiply charged tracks were provided by delta-electron density measurements on the length not less than 10 mm. The calibration was made up on known primary tracks and tracks of the double charged fragments. The accuracy of the measured charges was around three charge units for $`Z>40`$ and one charge unit for$`Z<20`$.
The relativistic particles emitted at $`\theta <\theta _0`$ ($`\theta `$ is the emission angle) were considered as singly charged fragments. $`\theta _0`$ is determined as
$$\mathrm{sin}\theta _0=0.2/P_0$$
where $`P_0`$ is the projectile nucleus momentum per nucleon in GeV/c.
In each investigated event, the polar $`\theta `$ and azimuthal $`\phi `$ angles of all charged particles were measured.
The transverse momentum of a spectator fragment was defined as
$$P_F=2Z_FP_0\mathrm{sin}\theta $$
(1)
The ratio $`A/Z`$ for fragments was assumed to equal two. The mean relative accuracy in the transverse momentum of fragments does not exceed 7%.
It should be noted that at high energies in contrast to low and intermediate ones two clear-out distinguishable regions corresponding to the projectile and target fragmentation are observed. The probability of compound nucleus creation is assumed to be small. Thus, the question how to select the fragments of the projectile and target nuclei is simply settled. Projectile fragments are regarded to have the velocities equal to that of projectile nucleus. Having this in mind, the relation (1) was suggested. Clearly, it is invalid for deep inelastic collisions where the fragments loss significant parts of their longitudinal momentum. The lack of necessary experimental information and specificity of the photoemulsion experiments prevent this circumstance from being taken into consideration.
The assumption of equality between the number of protons $`P`$ and neutrons $`N`$ in the fragments brings main uncertainty to fragment momenta. For heavy nuclei $`(N>P)`$, the relation (1) underestimates the transverse momentum. For double charged fragments among which $`{}_{}{}^{3}He`$ isotopes are presented, the relation (1) overestimates, on average, the transverse momentum. As shown in , 10% $`{}_{}{}^{3}He`$ admixture among all the fragments with $`Z=2`$ gives less than 1% growth in the dispersion of the transverse momentum (the dispersion changes from 162 to 164 MeV/c). This result cannot, of course, have any effect on the conclusions of the present paper.
Identification of the target fragments in photoemulsion experiments requires a special track measurement technique that is not used at our studies. Thus our data concern the fragments of projectile nuclei, which were identified by the commonly accepted emulsion analysis procedure. At photoemulsion studies dedicated to high energies ($`E>1`$ A GeV), the projectile fragments are usually called as spectator fragments. We will follow this tradition, sometimes omitting the assignment ”spectator”.
## 3 IMF multiplicity dependence on nuclear residual mass
It is obvious that the multiplicity of intermediate mass fragments rises with increasing the excitation energy at a fixed nuclear residual mass. However, at high excitation energies, the production of light fragments gets dominant, and $`<N_{IMF}>`$ must decrease. It is just the dependence that has been observed by the ALADIN- group for multifragmentation of gold residuals created in the interactions of the gold projectile with various targets at an energy of 600 MeV/nucleon. The ”bound” charge $`Z_{bound}`$ which includes the charges of $`\alpha `$ -particles was taken as a measure of nuclear residual mass. Since $`\alpha `$-particles can be produced at the pre-equilibrium decay stage, another value, $`Z_{b3}`$ , was suggested to be used. The most remarkable result was that $`<N_{IMF}>`$ as a function of $`Z_{bound}`$ or $`Z_{b3}`$ had proved independence of the target mass. As the dependence under discussion is essentially determined by a relation between the excitation energy and the nuclear residuals mass, one can conclude that the gold residuals of the same masses from various reactions have nearly the same excitation energies. Our results of Fig.1 allows one to make the statement more precise.
Fig.1 displays the $`IMF`$ multiplicity as functions of $`Z_{bound}`$ and $`Z_{b3}`$ (closed circles) in the interactions of krypton nuclei with photoemulsion nuclei at energy of 900 MeV/nucleon in a comparison with the analogous data on 10.7 GeV/nucleon gold nuclei interactions (open circles) . As seen, the data points for $`Kr`$ and $`Au`$ nuclear residual fragmentation are close to each other for $`Z_{bound}22`$ and $`Z_{b3}16`$. Although the error bars for $`Kr`$-points are quite large, we can say that at the average $`IMF`$ multiplicity for $`Kr`$ projectile is larger than that for gold one at the same value of $`Z_{bound}`$. At the same time, the $`<N_{IMF}>`$ as functions of $`Z_{b3}`$ for two projectiles are practically in a coincidence at $`Z_{b3}16`$. This indicates that nuclear residuals of the same masses formed in interactions of different systems at high energies have approximately the same excitation energies if the initial nuclei have lost more than one half of their nucleons at the fast stage of the collisions.
Focus attention on fast growth of the our $`IMF`$ multiplicity in the region of $`Z_{b3}30`$ with decreasing $`Z_{b3}`$. The rise seems to be related to a threshold character of the nuclear multifragmentation. Clearly, at small excitation energy the process of evaporation of nucleons and light nuclei is dominated. At large excitation energy, the multifragment decay channel opens. It was not clear whether the probability of the last process evolves smoothly with excitation energy increase or whether it is of threshold character. It is difficult to note a change in the evolution of $`IMF`$ multiplicity at large $`Z_{b3}`$ for gold residuals fragmentation due to large error bars. The statistics of the data is not rich enough to give conclusive results. The data of the ALADIN collaboration at $`Z_{b3}70`$ have the required statistics but, seemingly, suffer from methodical uncertainties. We believe that a study of $`IMF`$ multiplicity at large $`Z_{b3}`$ intended to look for the threshold character of the nuclear multifragmentation is of great interest.
## 4 Dependencies of intrinsic characteristics of decaying system on $`Z_{bound}`$ and $`Z_{b3}`$
A quite unexpected result was obtained at an analysis of the heaviest fragments. Fig.2 shows the mean charge of the heaviest fragment in an event as a function of $`Z_{bound}`$ or $`Z_{b3}`$ (symbols are the same as in Fig.1). One can see a clear change in the dependence of $`<Z_{max}>`$ on $`Z_{b3}`$ at $`Z_{b3}17`$. The analogous change, but not so pronounced, has been found in the gold residuals fragmentation. This effect is slightly shaded when $`Z_{bound}`$ is used. Thus we can conclude that nuclear residuals fragment in different manners when more or less one half of nucleons from the primary nucleus are ejected. The data presented in Fig.3, where the average asymmetry coefficient as a function of $`Z_{bound}`$ or $`Z_{b3}`$ is plotted, confirms the above conclusion.
The value of $`A_{12}`$ for each event is determined as
$$A_{12}=\frac{Z_1Z_2}{Z_1+Z_2},$$
where $`Z_1`$, $`Z_2`$, etc - the charges of fragments ordered such that $`Z_1Z_2Z_3`$…, $`Z_1Z_{max}`$. As seen, there is not any peculiarity in the dependence of $`<A_{12}>`$ on $`Z_{bound}`$. $`<A_{12}>`$ as a function of $`Z_{b3}`$ remains practically constant at $`Z_{b3}17`$ and then increases sharply with $`Z_{b3}`$ – the events with $`Z_{b3}>17`$ have strong asymmetry. A similar behavior is also observed at the fragmentation of heavier systems.
Summing up, we can conclude that there exist at least two regimes of fragmentation.
## 5 Energy of fragments as functions of $`Z_{bound}`$ or $`Z_{b3}`$
According to the statistical model of nuclear multifragmentation, the kinetic energy of the fragments in the rest frame of fragmenting nucleus is determined by the charge of the residual. Hence, the decrease of the fragment energies can be expected with decreasing $`Z_{bound}`$ . Going from the laboratory system to a rest frame of a fragmenting nucleus, we have used the Gallilean transformation described in Ref. . The average kinetic energy of a fragment is connected with its transverse momentum, assuming the isotropic decay, by
$$<E>=(3/2)<P_F^2>/4Z_Fm_N,$$
Here, $`P_F`$ \- the transverse momentum of the fragment, $`Z_F`$ \- its charge and $`m_N`$ \- the mass of the nucleon.
Fig. 4 presents the mean kinetic energies of the fragments in the events with the number of multiply charged fragments $`(Z_F2)`$ larger or equal three. One can see that the kinetic energy of $`Kr`$ fragments has no tendency to be decreased with decreasing the mass of residuals in the region $`Z_{bound}25`$. Moreover, in the region they are practically permanent. A similar behavior is observed at the fragmentation of gold residuals. The results provide an evidence for the radial flow of the fragments, with the energy of the flow depending on the mass of the initial nucleus (according to our data).
## 6 Summary
The experimental data on the multifragmentation of the krypton residual nuclei formed in the interactions with photoemulsion nuclei at energy 0.9 GeV/nucleon are presented.
The mechanism of the nuclear residual fragmentation is shown to be practically independent on the mass of projectile nucleus if $`Z_{bound}`$ does not exceed a one half of the charge of the initial nucleus.
The evidences are obtained that the multifragmentation is of a threshold character and that there is a radial flow of the fragments that depends upon the mass of the initial nucleus.
The experimental regularities manifest themselves more brightly when the value $`Z_{b3}`$ is used as a measure of the residual mass.
Further experimental and theoretical studies of the multifragmentation of intermediate mass nucleus are of interest.
The authors are thankful to the EMU-01 collaboration for providing us with the experimental material of the gold interactions. The financial support from the Russian Foundation of Fundamental Research (grant No. 99-02-17757), and the Foundation of Fundamental Research of Academy of Sciences of Uzbekistan, are cordially and gratefully acknowledged.
Figure captions
Fig.1: The dependencies of the $`IMF`$ multiplicity on $`Z_{bound}`$ (a) and $`Z_{b3}`$ (b). Closed circles - data for $`Kr`$ interactions at E=900 MeV/nucleon, open circles - data for $`Au`$ interactions at E= 10.7 GeV/nucleon.
Fig.2: The average charge of the largest fragment in the events as a function of ”bound” charge. Symbols are the same as in Fig.1.
Fig.3: The average coefficient of asymmetry in the measured events as a function of ”bound” charge. Symbols are the same as in Fig.1.
Fig.4: The estimated average kinetic energy of the fragments in the rest frame of the fragmenting nucleus. Symbols are the same as in Fig.1. a) - energies of double charged fragments; b) \- energies of the fragments with $`Z_F=35`$; c) - energies of $`IMF`$s. |
no-problem/0001/astro-ph0001381.html | ar5iv | text | # Unsupervised Induction and Gamma-Ray Burst Classification
## Introduction
Induction-based learningls95 attempts to extract interesting patterns from data. These patterns form concept classes with each class containing data instances. When the induction is unsupervised, the learning model has no a priori class knowledge. Rather, the learning algorithm uses one or more statistical or symbolic (machine learning) evaluation functions to cluster instances into concept classes.
Mukherjee et al.mfbmfr98 performed a statistical cluster analysis on a data set containing 797 gamma-ray bursts taken from the BATSE 3B catalogm96 . Assuming all attributes to be distributed logNormally, and utilizing the logarithmic values for T90 duration, total fluence, and hardness ratio HR321 their results showed the instances formed three classes. Class I contained long/bright/intermediate bursts, class II consisted of short/faint/hard bursts and class III was represented by intermediate/intermediate/soft bursts. Table 1 shows the mean and standard deviation values for the three classes. Table 2 offers a best defining rule for each class, as determined by ESXrghh99 . The rule for class I bursts indicates that 82.72% of the bursts in this class have a log T90 value between .70 and 2.66 and a log Fluence between -5.77 and -3.11. The rule also shows that we can be at least 97% confident that a burst with these characteristics is a class I burst. Table 2 shows that the class III rule does not cover its instances as well as the rules for classes I and II.
In this paper we use ESXrghh99 , a machine learning model and product of Information Acumen Corporation, to perform unsupervised learning on these same data for the purpose of comparative analysis. We chose ESX for this research since ESX explains its behavior has been shown to perform well in several real-world environments rghh99 .
## Method
The machine learning component of ESX is an induction-based sequential learning model that creates a concept hierarchyglf89 from a set of input instances. ESX uses knowledge contained in its concept hierarchy to generate a set of production rules to help define and explain what has been discovered. Supervised as well as unsupervised learning is supported.
ESX accepts data in the form of instances represented in attribute-value format. When learning is unsupervised, ESX takes one of two possible actions for each newly presented instance: (1) Place the new instance into an existing cluster, or (2) create a new conceptual cluster containing the instance as its only member.
In addition, ESX allows the user to set a learning parameter so as to encourage or discourage the creation of new clusters. For a given domain, a best value for this parameter can be determined experimentally.
## Results
For our first experiment, we set the ESX learning parameter so as to restrict the formation of new classes. As a result, ESX clustered the data into the two classes found by previous standard techniques cd . Table 3 shows a representative rule for each class. Notice that both clusters are well-defined.
For our second experiment, we allowed ESX to form a best set of three or more clusters. The results of this experiment showed the formation of four clusters. One of the four clusters contained a majority of intermediate bursts (class 1); a second cluster consisted of mostly short bursts (class 2). The remaining two clusters (classes 3 and 4) were subsets of the Mukherjee class I bursts. The class mean and standard deviation values for each of the six burst attributes are shown in Table 4.
Table 5 offers representative rules for each of the four clusters. Figures 1 and 2 as well as Table 4 indicate that class 3 contains mostly long/soft bursts and class 4 contains long/bright bursts. The following rule represents a covering rule for the cluster formed by combining the class 3 and class 4 bursts.
| $`1.19<=\mathrm{log}\mathrm{T90}<=2.66`$ |
| --- |
| :rule accuracy 90.26% |
| :rule coverage 92.68% |
## Conclusions
We used ESX to cluster data about 797 gamma ray bursts. When restricted to forming a small number of classes, two classes were determined. However, when allowed to form more than two classes, four classes were created. Two of the clusters were similar to the class II and class III bursts determined by Mukherjee et al.mfbmfr98 . Taken together, the two remaining clusters represent the class I Mukherjee et al. bursts. ESX differentiated the class I bursts by brightness and hardness. The separation of long bursts into two classes may be due in part to the fact that ESX makes no a priori assumptions about data distribution.
We hypothesize that systematic effects may cause some class I bursts to take on class III characteristicsHak . Systematic biases may explain why class I bursts have been separated into two groups by ESX. Our future work will focus on testing these hypotheses with the help of additional induction-based techniques. |
no-problem/0001/quant-ph0001040.html | ar5iv | text | # References
FIRST SINGLE BUBBLE SONOLUMINESCENCE IN DUBNA.
V.B. Belyaev, M.B. Miller<sup>∗∗</sup>, A.V. Sermyagin<sup>∗∗</sup>
Osaka University, Mihogaoko 10–1, Ibaraki Osaka 567, Japan; (on leave of Joint
Institute for Nuclear Research, Dubna, Russia); belyaev$`\mathrm{@}`$miho.rcnp.osaka–u.ac.jp;
<sup>∗∗</sup>Institute in Physical–Technical Problems, 141980 POB 39, Dubna, Moscow region,
Russia; sermyagin$`\mathrm{@}`$vxjinr.jinr.ru
At the Institute in Physical–Technical Problems experiments on sonoluminescence was started by our group at the beginning of this year. Our study was focused at properties of the single bubble mode of sonoluminescence. First experiments have already taken us, as it seems, to claiming–attention results. In particular, we managed to manipulate the process, varying boundary conditions on resonator outer surface. Some kind of interaction of a few spaced bubbles was observed, enhancing light emission by all of them simultaneously. Now we try to specify this effect (“few bubble sonoluminescence” – FBSL) and use it to affect conditions within the bubble, and study it further. Besides, conditions to induce SBSL in the spherical resonator with very low acoustical quality factor Q were discovered. This result, somehow unexpected, also attracted our attention. In such conditions stable SBSL can easy be supported during long-time runs. The system turns out to be practically non-sensitive to changes in external environment. The above and some other results are presented in the report. Besides, experiments in a stage of preparation are discussed, – a study of mechanism of “Argon rectification” within the bubble under SBSL, measurement of statistical moments of light flash intensity distribution, experiments on Einstein–Podolsky–Rosen correlations, as well as experiments on nuclear aspects in sonoluminescence. The last direction of studies is related, in particular, to a possibility of “molecular–nuclear synthesis” reactions in water under certain conditions, much discussed recently .
This work was supported in part by the Russian Foundation for Basic Research, Grant No. 98–02–16884. |
no-problem/0001/physics0001012.html | ar5iv | text | # I Introduction
## I Introduction
The analysis about motion has never ceased since the old Greece times, but from Zeno Paradox to Einstein’s relativity, only CCM is discussed, and its uniqueness is taken for granted ever since, but as to whether or not CCM is the only possible and objective motion, whether CCM is the real motion or apparent motion, no one has given a definite answer up to now, in fact, people have been indulging in the study of the motion law, but omitted the study of the motion itself.
On the other hand, we have entered into the microscopic world for nearly one century, but our understanding about it is still in confusion, the orthodox view renunciates CCM in microscopic world, but permits no existence of objective motion mode for the microscopic particles, while the opponents still recourse to CCM to lessen the pain of losing realism, no other objective motion modes have been presented for the microscopic particles till now, thus the above problem is more urgent than ever.
In this paper, we will mainly address the above problem, after given a deep logical and physical analysis about motion, we demonstrate that the natural motion in continuous space-time or CSTM is not CCM, but one kind of essentially discontinuous motion, we call it quantum discontinuous motion or quantum motion, since we show that the wave function in quantum mechanics is the very mathematical complex describing it, and Schrödinger equation of the wave function is also its simplest nonrelativistic motion equation; Furthermore, since the combination of quantum mechanics and general relativity will result in the discreteness of space-time, namely the real space-time will be essentially discrete, we further study the motion in discrete space-time, or DSTM, and demonstrate that it will naturally result in the collapse process of the wave function, and finally bring about the appearance of CCM in macroscopic world.
The plan of this paper is as follows: In Sect. 2 we first give a general analysis about CSTM, the motion state of particle is physically defined, its general form and description are also given based on the mathematical analysis in the Appendix. In Sect. 3 we work out the simplest evolution law of CSTM, which turns out to be Schrödinger equation in quantum mechanics. In Sect. 4 we give a strict physical definition of CSTM, and further discuss the constant $`\mathrm{}`$ involved in its law. In Sect. 5 we point out that space-time is essentially discrete due to the ubiquitous existence of gravity, and give a simple demonstration. In Sect. 6 we further give a general analysis about DSTM, and the general form of motion state in such space-time is given. In Sect. 7 the evolution law of DSTM is worked out, and we demonstrate that it will naturally result in the collapse process of the wave function. In Sect. 8 we further show that CCM and its evolution law can be consistently derived from the evolution law of DSTM. At last, conclusions are given.
## II General analysis about motion in continuous space-time (CSTM)
In this section, we will give a deep logical and physical analysis about CSTM.
### A The motion state of particle
First, we should define the motion state of particle, there are two alternatives, one is the instant state of particle, the other is the infinitesimal interval state of particle, it has been generally accepted that the motion state of particle should be the infinitesimal interval state of particle, not the instant state of particle, while people usually omit their essential difference, here we will present some of them.
(1).The instant state of particle contains only one point in space, its potential in mathematics is zero, while the infinitesimal interval state of particle contains infinite innumerable points in space, its potential in mathematics is $`\zeta _1`$.
(2).The instant state of particle contains no motion, but only the existence of particle, while the infinitesimal interval state of particle may contain abundant motion elements, since it contains infinite innumerable points in space.
(3).The instant state of particle possesses no physical meaning, since we can not access it through physical measurement, while the infinitesimal interval state of particle possesses real physical meaning, since we can measure it by means of the following infinite process:$`\mathrm{\Delta }tdt`$.
(4).We can only find and confirm the law for the infinitesimal interval state of particle, while as to the instant state of particle, even if its law exists, we can not find it, let along confirm it.
In fact, in physics there exist only the description quantities defined during infinitesimal time interval, this fact can be seen from the familiar differential quantities such as dt and dx, whereas the quantities defined at instants come only from mathematics, people always mix up these two kinds of quantities, this is a huge obstacle for the development of physics. Thus we can only discuss the motion state and relevant quantities defined during infinitesimal time interval, as well as their differential laws, if we study the point set corresponding to real physical motion.
For simplicity, in the following we say the motion state of particle at one instant, but it still denotes the infinitesimal interval state of particle, not the instant state of particle.
### B The general form of the motion state of particle
Secondly, we will give the general form of the motion state of particle, according to the analysis about point set(see Appendix), the natural assumption in logic is that the motion state of particle in infinitesimal time interval is a general dense point set in space, since we have no a priori reason to assume a special form, its proper description is the measure density $`\rho (x,t)`$ and measure density fluid $`j(x,t)`$.
Certainly, at some instant t the motion state of particle may assume some kind of special form, such as the continuous point set described by dx or $`\rho (x,t)=\delta (xx(t))`$, but whether or not this kind of special form can exist for other instants should be determined by the motion law, not our prejudices.
## III The evolution of motion in continuous space-time (CSTM)
In the following, we will give the main clues for finding the possible evolution equations of CSTM, and show that Schrödinger equation in quantum mechanics is just its simplest nonrelativistic evolution equations. Here we mainly analyze one-dimension motion, but the results can be easily extended to three-dimension situation.
### A The first motion principle
First, we should find the first motion principle similar to the first Newton principle, this means that we need to find the simplest solution of the motion equation, in which we can get the invariant quantity during free motion, it is evident that the simplest solution of the motion equation is:
$$\frac{\rho (x,t)}{t}=0$$
(1)
$$\frac{j(x,t)}{x}=0$$
(2)
$$\frac{j(x,t)}{t}=0$$
(3)
$$\frac{\rho (x,t)}{x}=0$$
(4)
using the relation $`j(x,t)`$=$`\rho (x,t)v`$ we can further get the solution, namely $`\rho (x,t)`$=1,$`j(x,t)`$=$`v`$=$`p`$/m, where $`m`$ is the mass of the particle, $`p`$ is defined as the momentum of particle. Now, we get the first motion principle, namely during the free motion of particle, the momentum of the particle is invariant, but it can be easily seen that, contrary to classical continuous motion, for the free particle with one constant momentum, its position will not be limited in the infinitesimal space dx, but spread throughout the whole space with the same position measure density.
Similar to the quantity position, the natural assumption in logic is also that the momentum (motion) state of particle in infinitesimal time interval is still a general dense point set in momentum space, thus we can also define the momentum measure density $`f(p,t)`$, and the momentum measure fluid density $`J(p,t)`$, their meanings are similar to those of position.
### B Two kinds of description bases
Now, we have two description quantities, one is position, the other is momentum, and position descriptions $`\rho (x,t)`$ and $`j(x,t)`$ provide a complete local description of the motion state, we may call it local description basis, similarly momentum descriptions $`f(p,t)`$ and $`J(p,t)`$ provide a complete nonlocal description of the motion state, since for the particle with any constant momentum, its position will spread throughout the whole space with the same position measure density, we may call it nonlocal description basis.
Furthermore, at any instant the motion state of particle is unique, thus there should exist a one-to-one relation between these two kinds of description bases, namely there should exist a one-to-one relation between position description $`(\rho ,j)`$ and momentum description $`(f,J)`$, and this relation is irrelevant to the concrete motion state, in the following we will mainly discuss how to find this one-to-one relation, and our analysis will also show that this relation essentially determines the evolution of motion.
### C One-to-one relation
First, it is evident that there exists no direct one-to-one relation between the measure density functions $`\rho (x,t)`$ and $`f(p,t)`$, since even for the above simplest situation, we have $`\rho (x,t)=1`$ and $`f(p,t)=\delta ^2(pp_0)`$<sup>*</sup><sup>*</sup>* This result can be directly obtained when considering the general normalization relation $`_\mathrm{\Omega }\rho (x,t)𝑑x=_\mathrm{\Omega }f(p,t)𝑑p`$., and there is no one-to-one relation between them.
Then in order to obtain the one-to-one relation, we have to create new properties on the basis of the above position description $`(\rho ,j)`$ and momentum description $`(f,J)`$, this needs a little mathematical trick, here we only give the main clues and the detailed mathematical demonstrations are omitted, first, we disregard the time variable $`t`$ and let $`t=0`$, as to the above free evolution state with one momentum, we have $`(\rho ,j)=(1,p_0/m)`$ and $`(f,J)=(\delta ^2(pp_0),0)`$, thus we need to create a new position state function $`\psi (x,0)`$ using $`1`$ and $`p_0/m`$, a new momentum state function $`\phi (p,0)`$ using $`\delta ^2(pp_0)`$ and $`0`$, and find the one-to-one relation between these two state functions, this means there exists a one-to-one transformation between the state functions $`\psi (x,0)`$ and $`\phi (p,0)`$, we generally write it as follows:
$$\psi (x,0)=_{\mathrm{}}^+\mathrm{}\phi (p,0)T(p,x)𝑑p$$
(5)
where $`T(p,x)`$ is the transformation function and generally continuous and finite for finite $`p`$ and $`x`$, since the function $`\phi (p,0)`$ will contain some form of the basic element $`\delta ^2(pp_0)`$, normally we may expand it as $`\phi (p,0)=_{i=1}^{\mathrm{}}a_i\delta ^i(pp_0)`$, while the function $`\psi (x,0)`$ will contain the momentum $`p_0`$, and be generally continuous and finite for finite $`x`$, then it is evident that the function $`\phi (p,0)`$ can only contain the term $`\delta (pp_0)`$, because the other terms will result in infiniteness.
On the other hand, since the result $`\phi (p,0)=\delta (pp_0)`$ implies that there exists the simple relation f(p,0)=$`\phi (p,0)^{}\phi (p,0)`$Evidently, another simple relation f(p,0)=$`\phi (p,0)^2`$ permit no existence of one-to-one relation, and owing to the equality between the position description and momentum description, we also have the similar relation $`\rho (x,0)`$=$`\psi (x,0)^{}\psi (x,0)`$, thus we may let $`\psi (x,0)=e^{iG(p_0,x)}`$ and have $`T(p,x)=e^{iG(p,x)}`$, then considering the symmetry between the properties position and momentumThis symmetry essentially stems from the equivalence between these two kinds of descriptions, the direct implication is for $`\rho (x,0)=\delta ^2(xx_0)`$ we also have $`f(p,0)=1`$., we have the general extension $`G(p,x)=_{i=1}^{\mathrm{}}b_i(px)^i`$, furthermore, this kind of symmetry also results in the symmetry between the transformation $`T(p,x)`$ and its reverse transformation $`T^1(p,x)`$, where $`T^1(p,x)`$ satisfies the relation $`\phi (p,0)=_{\mathrm{}}^+\mathrm{}\psi (x,0)T^1(p,x)𝑑p`$, thus we can only have the term $`px`$ in the function $`G(p,x)`$, and the resulting symmetry relation between these two transformations will be $`T^1(p,x)=T^{}(p,x)=e^{ipx}`$, we let $`b_1=1/\mathrm{}`$, where $`\mathrm{}`$ is a constant quantity(for simplicity we let $`\mathrm{}=1`$ in the following discussions), then we get the basic one-to-one relation, it is $`\psi (x,0)=_{\mathrm{}}^+\mathrm{}\phi (p,0)e^{ipx}𝑑p`$, where $`\psi (x,0)=e^{ip_0x}`$ and $`\phi (p,0)=\delta (pp_0)`$, it mainly results from the essential symmetry involved in CSTM itself.
In order to further find how the time variable $`t`$ is included in the functions $`\psi (x,t)`$ and $`\phi (p,t)`$, we may consider the superposition of two single momentum states, namely $`\psi (x,t)=\frac{1}{\sqrt{2}}[e^{ip_1xic_1(t)}+e^{ip_2xic_2(t)}]`$, then the position measure density is $`\rho (x,t)=[1+\mathrm{cos}(\mathrm{}c(t)\mathrm{}px)]/2`$, where $`\mathrm{}c(t)=c_2(t)c_1(t)`$ and $`\mathrm{}p=p_2p_1`$, now we let $`\mathrm{}p0`$, then we have $`\rho (x,t)1`$ and $`\mathrm{}c(t)0`$, especially using the measure conservation relation we can get $`dc(t)/dt=dpp/m`$, namely $`dc(t)=d(p^2/m)t`$ or $`dc(t)=dEt`$, where $`E=p^2/m`$, is defined as the energy of the particle in the nonrelativistic domain, thus as to any single momentum state we have the time-included formula $`\psi (x,t)=e^{ipxiEt}`$.
In fact, there may exist other complex forms for the state functions $`\psi (x,t)`$ and $`\phi (p,t)`$, for example, they are not the above simple number functions but multidimensional vector functions such as $`\psi (x,t)=(\psi _1(x,t),\psi _2(x,t),\mathrm{},\psi _n(x,t))`$ and $`\phi (p,t)=(\phi _1(p,t),\phi _2(p,t),\mathrm{},\phi _n(p,t))`$, but the above one-to-one relation still exists for every component function, and these vector functions still satisfy the above modulo square relations, namely $`\rho (x,t)=_{i=1}^n\psi _i(x,t)^{}\psi _i(x,t)`$ and $`f(p,t)=_{i=1}^n\phi _i(p,t)^{}\phi _i(p,t)`$, these complex forms will correspond to the particles with more complex structure, say, involving more inner properties of the particle such as charge and spin etc.
At last, since the one-to-one relation between the position description and momentum description is irrelevant to the concrete motion state, the above one-to-one relation for the free motion state with one momentum should hold true for any motion state, and the states satisfying the one-to-one relation will be the possible motion states. Furthermore, it is evident that this one-to-one relation will directly result in the famous Heisenberg uncertainty relation $`\mathrm{}x\mathrm{}p\mathrm{}/2`$.
### D The evolution law of motion in continuous space-time (CSTM)
Now, we will work out the evolution law of CSTM.
First, as to the free motion state with one momentum, namely the single momentum state $`\psi (x,t)=e^{ipxiEt}`$, using the above definition of energy $`E=\frac{p^2}{m}`$ and including the constant quantity $`\mathrm{}`$ we can easily find its nonrelativistic evolution law, which is
$$i\mathrm{}\frac{\psi (x,t)}{t}=\frac{\mathrm{}^2}{2m}\frac{^2\psi (x,t)}{x^2}$$
(6)
then owing to the linearity of this equation, this evolution equation also applies to the linear superposition of the single momentum states, namely all possible free notion states, or we can say, it is the free evolution law of CSTM.
Secondly, we will consider the evolution law of CTSM under outside potential, when the potential $`U(x,t)`$ is a constant $`U`$, the evolution equation will be
$$i\mathrm{}\frac{\psi (x,t)}{t}=\frac{\mathrm{}^2}{2m}\frac{^2\psi (x,t)}{x^2}+U\psi (x,t)$$
(7)
then when the potential $`U(x,t)`$ is related to x and t, the above form will still hole true, namely
$$i\mathrm{}\frac{\psi (x,t)}{t}=\frac{\mathrm{}^2}{2m}\frac{^2\psi (x,t)}{x^2}+U(x,t)\psi (x,t)$$
(8)
for three-dimension situation the equation will be
$$i\mathrm{}\frac{\psi (𝐱,t)}{t}=\frac{\mathrm{}^2}{2m}^2\psi (𝐱,t)+U(𝐱,t)\psi (𝐱,t)$$
(9)
this is just the Schrödinger equation in quantum mechanics<sup>§</sup><sup>§</sup>§In fact, Schrödinger equation is also the simplest evolution equation of dense point set, this fact manifests the identity of physical reality and mathematical reality to some extent..
At last, the above analysis also shows that the state function $`\psi (x,t)`$ provides a complete description of CSTM, since CSTM is completely described by the measure density $`\rho (x,t)`$ and measure fluid density $`j(x,t)`$, and according to the above evolution equation the state function $`\psi (x,t)`$ can be expressed by these two functions, namely $`\psi (x,t)=\rho ^{1/2}e^{iS(x,t)/\mathrm{}}`$, where $`S(x,t)=m_{\mathrm{}}^xj(x^{^{}},t)/\rho (x^{^{}},t)𝑑x^{^{}}+C(t)`$When in three-dimension space, the formula for S(x,y,z,t) will be $`S(x,y,z,t)=m_{\mathrm{}}^xj_x(x^{^{}},y,z,t)/\rho (x^{^{}},y,z,t)𝑑x^{^{}}+C(t)=m_{\mathrm{}}^yj_y(x,y^{^{}},z,t)/\rho (x,y^{^{}},z,t)𝑑y^{^{}}+C(t)=m_{\mathrm{}}^zj_z(x,y,z^{^{}},t)/\rho (x,y,z^{^{}},t)𝑑z^{^{}}+C(t)`$, since in general there exists the relation $`\times \{𝐣(x,y,z,t)/\rho (x,y,z,t)\}=0`$., and these two functions can also be expressed by the state function, namely $`\rho (x,t)=|\psi (x,t)|^2`$ and $`j(x,t)=[\psi ^{}\psi /t\psi \psi ^{}/t]/2i`$, then there exists a one-to-one relation between $`(\rho (x,t),j(x,t))`$ and $`\psi (x,t)`$ when omitting the absolute phase, thus the state function $`\psi (x,t)`$ also provides a complete description of CSTM.
On the other hand, we can see that the absolute phase of the wave function $`\psi (x,t)`$, which may depend on time, is useless for describing CSTM, since according to the above analysis it disappears in the measure density $`\rho (x,t)`$ and measure fluid density $`j(x,t)`$, which completely describe CSTM, thus it is natural that the absolute phase of the wave function possesses no physical meaning.
## IV Further discussions about motion in continuous space-time (CSTM)
### A The definition of motion in continuous space-time (CSTM)
Now we can give the physical definition of CSTM in three-dimension space, the definition for other abstract spaces or many-particle situation can be easily extended.
(1). The motion of particle in space is described by dense point set in four-dimension space and time.
(2). The motion state of particle in space is described by the position measure density $`\rho (𝐱,t)`$ and position measure fluid density $`𝐣(𝐱,t)`$ of the corresponding dense point set.
(3). The evolution of motion corresponds to the evolution of the dense point set, and the simplest evolution equation is Schrödinger equation in quantum mechanics.
Compared with classical continuous motion, we may call CSTM quantum discontinuous motion, or quantum motion, the commonness of these two kinds of motion is that they are both the motion of particle, namely the moving object exists only in one position in space at one instant, their difference lies in the moving behavior, namely the behavior of the particle during infinitesimal time interval \[t,t+dt\], for classical motion, the particle is limited in a certain local space interval \[V,V+dV\], while for quantum motion, the particle moves throughout the whole space with a certain position measure density $`\rho (𝐱,t)`$.
In fact, all physical states of CSTM are defined during infinitesimal time interval in the meaning of measure, not at one instant, for example, the single momentum state $`\psi _p(x,t)=e^{ipxiEt}`$, especially even the single position state $`\psi (x,t)=\delta (xx_0)`$ is still defined during infinitesimal time interval.
### B Some discussions about the constant $`\mathrm{}`$
First, from the above analysis about CSTM, we can understand why the constant $`\mathrm{}`$ with dimension $`Js`$ should appear in the evolution equation of CSTM, or Schrödinger equation, the existence of $`\mathrm{}`$ essentially results from the equivalence between the nonlocal momentum description and local position description of CSTM, it is this equivalence that results in the one-to-one relation between these two kinds of descriptions, which requires the existence of a certain constant $`\mathrm{}`$ with dimension $`Js`$ to cancel out the dimension of the physical quantities $`px`$ and $`Et`$ in the relation, at the same time, the existence of $`\mathrm{}`$ also indicates some kind of balance between the dispersion of the position distribution and that of momentum distribution limited by the one-to-one relation ( there is no such limitation for CCM and $`\mathrm{}=0`$ ), or we can say, the existence of $`\mathrm{}`$ essentially indicates some kind of balance between the nonlocality and locality of motion in continuous space-time.
Secondly, even though the appearance of $`\mathrm{}`$ in the evolution equation of CSTM is inevitable, its value can not be determined by CSTM itself, we only know that $`\mathrm{}`$ possesses a finite nonzero value, certainly, just like the other physical constants such as c and G, its value can be determined by the experience, but there may exist some deeper reasons for the special value of $`\mathrm{}`$ in our universe, although motion can not determine this value alone, the solution may have to resort to other subtle realities in this world, for example, gravity (G), space-time(c), or even the existence of our mankind.
## V The real space-time is discrete
The combination of quantum mechanics and general relativity strongly implied space-time is essentially discrete, and the minimum space-time unit will be Planck size $`T_p`$ and $`L_p`$, thus owing to the ubiquitous existence of gravity, the real space-time will be essentially discrete with the minimum size $`T_p`$ and $`L_p`$.
Here we will give a simple operational demonstration about the discreteness of space-time, consider a measurement of the length between points A and B, at point A place a clock with mass $`m`$ and size $`a`$ to register time, at point B place a reflection mirror, when t=0 a photon signal is sent from A to B, at point B it is reflected by the mirror and returns to point A, then the clock registers the return time, for classical situation the measured length will be $`L=\frac{1}{2}ct`$, but when considering quantum mechanics and general relativity, the existence of the clock introduces two kinds of uncertainties to the measured length, the uncertainty resulting from quantum mechanics is: $`\delta L_{QM}(\frac{\mathrm{}L}{mc})^{1/2}`$, the uncertainty resulting from general relativity is: $`\delta L_{GR}\frac{1}{2}\frac{2Gm}{c^2}ln\frac{a+L}{a}\frac{Gm}{c^2}`$, then the total uncertainties is:$`\delta L=\delta L_{QM}+\delta L_{GR}(LL_{p}^{}{}_{}{}^{2})^{1/3}`$, where $`L_p=(\frac{G\mathrm{}}{c^3})^{1/2}`$, is Planck length, thus we conclude that the minimum measurable length is Planck length $`L_p`$, in a similar way, we can also work out the minimum measurable time, it is just Planck time $`T_p`$.
## VI General analysis about motion in discrete space-time (DSTM)
In the discrete space-time, there exist absolute minimum sizes $`T_p`$ and $`L_p`$, namely the minimum distinguishable size of time and position of the particle is respectively $`T_p`$ and $`L_p`$, thus in physics the existence of the particle is no longer in one position at one instant as in the continuous space-time, but limited in a space interval $`L_p`$ during a finite time interval $`T_p`$, it can be seen that this state corresponds to the instant state of particle in continuous space-time, we define it as the instant state of particle in discrete space-time, this state evidently contains no motion, but only the existence of particle.
Furthermore, during the finite time interval $`T_p`$ the particle can only be limited in a space interval $`L_p`$, since if it can move throughout at least two different local regions with separation size larger than $`L_p`$ during the time interval $`T_p`$, then there essentially exists a smaller distinguishable finite time interval than $`T_p`$, which evidently contradicts the fact that $`T_p`$ is the minimum time unit, thus the discreteness of space-time essentially results in the existence of local position state of the particle, in which the particle stays in a local region with size $`L_p`$ for a time interval $`T_p`$, we may call such general local position state Planck cell state.
Similar to the analysis of the motion state in continuous space-time, in discrete space-time, the natural assumption in logic is that the motion state of particle in finite time interval, which is much longer than $`T_p`$ but still small enough, is a general discrete point set or cell set in space, since we have no a priori reason to assume a special form, its proper description is still the measure density $`\rho (x,t)`$ and measure density fluid $`j(x,t)`$, but in the meaning of time average.
Certainly, we can also get the motion state of particle in discrete space-time from that in continuous space-time, since in continuous space-time the particle, which instant state is the particle being in one position at one instant, moves throughout the whole space during infinitesimal time interval, while in discrete space-time the instant state of particle turns to be the particle being in a space interval $`L_p`$ during a finite time interval $`T_p`$, the motion state of particle in discrete space-time will naturally be that during a finite time interval much larger than $`T_p`$ the particle moves throughout the whole space with the position measure density $`\rho (x,t)`$ in the meaning of time average.
Now the visual physical picture of DSTM will be that during a finite time interval $`T_p`$ the particle will stay in a local region with size $`L_p`$, then it will still stay there or ”jump” to another local region, which may be very far from the original region, while during a time interval much larger than $`T_p`$ the particle will move throughout the whole space with a certain average position measure density $`\rho (x,t)`$.
As we can see, on the one hand, the particle undergoing DSTM stays in a local region during infinitesimal time interval, this may generate the display of CCM, on the other hand, during a finite time interval much larger than $`T_p`$ the particle will continually jumps from one local region to another local region, and move throughout the whole space with a certain position measure density $`\rho (x,t)`$, this may generate the display of CSTM, then DSTM is evidently some kind of unification of CCM and CSTM, in fact, owing to the ubiquitous existence of quantum and gravity, space-time is essentially discrete, and DSTM will be the only real motion in Nature, while CSTM and CCM are the apparent motion modes, they are only two ideal approximators of DSTM in microscopic and macroscopic world, thus DSTM is just the lost reality unifying quantum motion (CSTM) and classical motion (CCM), and it will undoubtedly provide an uniform realistic picture for microscopic world and macroscopic world, the following analysis will confirm this conclusion more convincingly.
## VII The evolution of motion in discrete space-time (DSTM)
### A A general discussion
Since CSTM is some kind of time average of DSTM, the evolution of DSTM will follow the evolution law of CSTM in the meaning of time average, on the other hand, the particle undergoing DSTM does stay in a local region for a finite nonzero time interval, and jump from this local region to another local region stochastically, thus the position measure density $`\rho (x,t)`$ of the particle will be essentially changed in a stochastic way due to the finite nonzero stay time in different stochastic regionFor CSTM, the stay time of the particle in any position is zero, so its position measure density $`\rho (x,t)`$ is not influenced by the stochastic jump., and the corresponding wave function will be also stochastically changed, then this kind of stochastic jump inevitably introduce the stochastic element to the evolution, so the evolution law of DSTM will be the combination of the deterministic linear evolution and stochastic nonlinear evolution, in the following we will work out this law.
### B Two rules
At first, since CSTM is some kind of average of DSTM during a finite time interval much larger than $`T_p`$, thus the position measure density $`\rho (x,t)`$ of the particle undergoing CSTM will be also the average of the position distribution of the particle undergoing DSTM during this time interval, then it is natural that the position of the particle undergoing DSTM will satisfy the position measure density $`\rho (x,t)`$, namely for DSTM the stochastic stay position of the particle satisfies the distribution
$$P(x,t)=|\psi (x,t)|^2$$
(10)
this is the first useful rule for finding the evolution law of DSTM.
Secondly, according to the definition of the position measure density $`\rho (x,t)`$, the finite nonzero stay time of the particle in a local region evidently implies that the position measure density $`\rho (x,t)`$ in that region will be increased after this finite nonzero stay time interval, and the increase will be larger when the stay time is longer. We consider the general situation that the particle undergoing DSTM stays in a local region $`L_p`$ for a time interval $`T`$, in the first rank approximation the increase of the position measure density $`\rho (x,t)`$ in this region can be written as follows after normalization:
$$\rho (x,t+T)=\frac{1}{A(T)}(\rho (x,t)+T/T_m)$$
(11)
where $`A(T)`$ is the normalization factor, $`T_m`$ is a certain time size to be determined, which may be relevant to the concrete motion state of the particle, this will be the second useful rule.
We first work out the normalization factor $`A(t)`$, considering the following two conditions:(1)when $`T=0`$,$`\rho (x,t+T)=\rho (x,t)`$, and $`A(0)=1`$;(2)when $`T\mathrm{}`$,$`\rho (x,t+T)1`$, and $`A(\mathrm{})T/T_m`$, we can get $`A(T)=1+T/T_m`$, then the above formula will be:
$$\rho (x,t+T)=\frac{\rho (x,t)+T/T_m}{1+T/T_m}$$
(12)
or it can be written as follows:
$$\mathrm{\Delta }\rho (x,t)=\frac{T}{T_m+T}(1\rho )$$
(13)
in general, when $`T>T_p`$, namely when the particle undergoing DSTM stays in a local region $`L_p`$ for a time interval longer than $`T_p`$, we can divide the whole time interval $`T`$ into many Planck cell $`T_p`$, and the above formula is still valid for every cell, thus in the following discussions we let $`T=T_p`$ for simplicity.
In order to further find the formula of $`T_m`$, we need to study the limitation on the jump of the particle undergoing DSTM, since the item $`T_m`$ just denotes this kind of limitation, this can be seen from the following two extreme situations:(1)when $`T_m\mathrm{}`$, we have $`\mathrm{\Delta }\rho 0`$, this denotes that the position measure density will be not influenced by the jump, and the particle can jump freely;(2)$`T_m0`$, we have $`\rho 1`$, this denotes that the position measure density will turn to be one in the region where the particle stays, and the position measure density in other regions will turn to be zero, so the particle can not jump at all. In fact, from the physical analysis about the jump we can see that the limitation results from the principle of energy conservation, according to which during a finite nonzero time interval $`\mathrm{\Delta }t`$ the possible change of energy $`\mathrm{\Delta }E_j`$ resulting from jump will be limited by the uncertainty relation $`\mathrm{\Delta }E_j\mathrm{}/\mathrm{\Delta }t`$, now we consider two situations, first, if the total difference of energy $`\mathrm{\Delta }E`$ between the original stay region and other regions satisfies the condition $`\mathrm{\Delta }E\mathrm{\Delta }E_j`$, then the particle can hardly jump from its original region to other regions, namely after the stay time $`\mathrm{\Delta }t`$ the position measure density $`\rho (x,t)`$ in the original region will be greatly increased, especially when $`\mathrm{\Delta }E\mathrm{}`$, we have $`\rho (x,t)1`$, and $`T_m0`$<sup>\**</sup><sup>\**</sup>\**In fact, in this situation the wave function has collapsed into this local region in order to satisfy the requirement of energy conservation, and this also indicates that in order to satisfy the principle of energy conservation DSTM will naturally result in the collapse of the wave function.; Secondly, if the total difference of energy $`\mathrm{\Delta }E`$ between the original stay region and other regions satisfies the condition $`\mathrm{\Delta }E\mathrm{\Delta }E_j`$, then the particle can jump more easily from its original region to other regions, namely after the stay time $`\mathrm{\Delta }t`$ the position measure density $`\rho (x,t)`$ will be only changed slightly, especially when $`\mathrm{\Delta }E0`$, we have $`\mathrm{\Delta }\rho (x,t)0`$, and $`T_m\mathrm{}`$. Then we can see that $`T_m`$ is inversely proportional to $`\mathrm{\Delta }E`$, considering the dimension requirement their relation will be $`T_m=\mathrm{}/k\mathrm{\Delta }E`$, where $`k`$ is a dimensionless constant.
Now, the change of the position measure density after stay time $`T_p`$ can be formulated in a more complete way:
$$\rho (x,t+T)=\frac{\rho (x,t)+k\mathrm{\Delta }E/E_p}{1+k\mathrm{\Delta }E/E_p}$$
(14)
or it can be written as follows:
$$\mathrm{\Delta }\rho (x,t)=\frac{\mathrm{\Delta }E}{kE_p+\mathrm{\Delta }E}(1\rho )$$
(15)
where $`E_p=\mathrm{}/T_p`$ is Planck energy, thus we get the second useful rule for finding the evolution law of DSTM.
### C The evolution law of motion in discrete space-time (DSTM)
Now, according to the above two rules, we can give the evolution equation of DSTM.
For simplicity but lose no generality, we consider a one-dimension initial wave function $`\psi (x,0)`$, according to the above analysis, the concrete evolution equation of DSTM will be essentially one kind of revised stochastic evolution equation based on Schrödinger equation, here we assume the form of stochastic differential equation ( SDE ), it can be written as follows:
$$d\psi (x,t)=\frac{1}{i\mathrm{}}H_Q\psi (x,t)dt+\frac{1}{2}[\frac{\delta _{xx_N}}{\rho (x,t)}1]\frac{\mathrm{\Delta }E(x_N,\overline{x_N})}{kE_p+\mathrm{\Delta }E(x_N,\overline{x_N})}\psi (x,t)dt$$
(16)
where the first term in right side represents the evolution element resulting from CTSM, the average behavior of DSTM, $`H_Q`$ is the corresponding Hamiltonian, the second term in right side represents the evolution element resulting from the stochastic jump resulting from DSTM itself, $`\delta _{xx_N}`$ is the discrete $`\delta `$-function, $`k`$ is a dimensionless constant, $`\rho (x,t)=|\psi (x,t)|^2`$, is the position measure density, $`\mathrm{\Delta }E(x_N,\overline{x_N})`$ is the total difference of energy of the particle between the cell containing $`x_N`$ and all other cells $`\overline{x_N}`$, $`x_N`$ is a stochastic position variable, whose distribution is $`P(x_N,t)=\rho (x_N,t)`$.
In physics, this stochastic differential equation is essentially a discrete evolution equation, all the quantities are defined relative to the Planck cells $`T_p`$ and $`L_p`$, and the equation should be also solved in a discrete way.
### D Some further discussions
Now we will give some physical analyses about the above evolution equation of DSTM, first, the linear item in the equation will result in the spreading process of the wave function as for the evolution of CSTM, while the nonlinear stochastic item in the equation will result in the localizing process of the particle or collapse process of the wave function, this can also be seen qualitatively, since according to the nonlinear stochastic item, in the region where the position measure density is larger the stay time of the particle will be longer, moreover, the longer stay time of the particle in one region will further increase the position measure density in that region much more, thus this process is evidently one kind of positive feedback process, the particle will finally stay in a local region, and the wave function of particle will also collapse to that region, so the evolution of DSTM will be some kind of combination of the spreading process and localizing process.
Secondly, the strength of the spreading process and localizing process is mainly determined by the energy difference between different branches of the wave function, if the energy difference is so small, then the evolution of DSTM will be mainly dominated by the spreading process, or we can say, the display of DSTM will be more like that of quantum motion (CSTM), this is just what happens in microscopic world; while if the energy difference is so large, then the evolution of DSTM will be mainly dominated by the localizing process, or we can say, the display of DSTM will be more like that of classical motion (CCM), this is just what happens in macroscopic world, and the boundary of these two worlds can also be estimated, the following example indicates that the energy difference in the boundary may assume $`\mathrm{\Delta }E\sqrt{\mathrm{}E_p}7Mev`$, the corresponding collapse time will be in the level of seconds.
Thirdly, if the particle finally stay in a local region during the evolution of DSTM, the localizing probability of the particle, or the collapse probability the wave function in one local region is just the initial position measure density of the particle in that region, namely the probability satisfies the Born rule in quantum mechanics, since the stochastic evolution of DSTM satisfies the Martingale condition, this can be seen from the following fact, namely during every jump the position measure density $`\rho `$ satisfies the equation $`P(\rho )=\rho P(\rho +\frac{\mathrm{\Delta }E}{kE_p+\mathrm{\Delta }E}(1\rho ))+(1\rho )P(\rho \frac{\mathrm{\Delta }E}{kE_p+\mathrm{\Delta }E}\rho )`$, where $`P(\rho )`$ is the probability of $`\rho `$ turning into one in one local region, namely the probability of the particle localizing in a local region, moreover, the solution of this equation is $`P(\rho )=\rho `$, this just means that the localizing probability of the particle in one region is just the initial position measure density of the particle in that region.
Fourthly, the collapse process resulting from the evolution of DSTM has no tails, since the evolution is essentially discrete, the wave function is just the description of the motion of particle, and its existence is only in the meaning of time average, while the particle, the real object, always exists in one local position, thus in the last stage of the collapse process, when the particle stays in one of the branches long enough it will de facto collapse into that branch owing to the limitation of energy conservation, and the wave function, the apparent ”object”, will also completely disappears in other branches<sup>††</sup><sup>††</sup>††If the wave function is taken as some kind of essential existence, and its evolution is essentially continuous, then the tails problem will be inevitable..
At last, the existence of DSTM will help to tackle the well-known time problem involved in formulating a complete theory of quantum gravity, since as to DSTM, the local position state of particle will be the only proper state, and the only real physical existence, during a finite time interval $`T_p`$ the particle can only be limited in a local space interval $`L_p`$, namely there does not exist any essential superposition of different positions at all, the superposition of the wave function is only in the meaning of time average, thus the essential inconsistency of the superposition of different space-time in the theory of quantum gravity, which results from the existence of the essential superposition of the wave function, will naturally disappear, and the real physical picture based on DSTM will be that at any instant ( during a finite time interval $`T_p`$ ) the structure of space-time determined by the existence of the particle ( in a local space interval $`L_p`$ ) is definite or ”classical”, while during a finite time interval much larger than $`T_p`$ but still small enough it will be stochastically disturbed by the stochastic jump of the particle undergoing DSTM, this kind of stochastic disturbance will be the real quantum nature of the space-time and matter.
### E One simple example
In this section, as one example we will analyze the DSTM evolution of a simple two-state system, and quantificationally show that the evolution of DSTM will indeed result in the collapse process of the wave function.
We suppose the initial wave function of the particle is $`\psi (x,0)=\alpha (0)^{1/2}\psi _1(x)+\beta (0)^{1/2}\psi _2(x)`$, which is a superposition of two static states with different energy levels $`E_1`$ and $`E_2`$, these two static states are located in separate regions $`R_1`$ and $`R_2`$ with the same size.
Since the energy of the particle inside the region of each static state is the same, we can consider the spreading space of both static states as a whole local region, and only study the stochastic jump between these two regions resulting from the evolution of DSTM, namely we directly consider the difference of the energy $`\mathrm{\Delta }E=E_2E_1`$ between these two states, through some mathematical calculations we can work out the density matrix of the two-state system, it is:
$$\rho _{11}(t)=\alpha (0)$$
(17)
$$\rho _{22}(t)=\beta (0)$$
(18)
$$\rho _{12}(t)=[1(\frac{\mathrm{\Delta }E}{kE_p+\mathrm{\Delta }E})^2]^{t/2T_p}\sqrt{\alpha (0)\beta (0)}(1\frac{(\mathrm{\Delta }E)^2}{2k^2\mathrm{}E_p}t)\sqrt{\alpha (0)\beta (0)}$$
(19)
$$\rho _{21}(t)=[1(\frac{\mathrm{\Delta }E}{kE_p+\mathrm{\Delta }E})^2]^{t/2T_p}\sqrt{\alpha (0)\beta (0)}(1\frac{(\mathrm{\Delta }E)^2}{2k^2\mathrm{}E_p}t)\sqrt{\alpha (0)\beta (0)}$$
(20)
It is evident that these results confirm the above qualitative analysis definitely, namely, the evolution of DSTM indeed results in the collapse of the wave function describing DSTM, and the distribution of the collapse results satisfies the Born rule in quantum mechanics, besides, we also get the concrete collapse time for two-state system, it is $`\tau _c2k^2\frac{\mathrm{}E_p}{(\mathrm{\Delta }E)^2}`$<sup>‡‡</sup><sup>‡‡</sup>‡‡This result has also been obtained by Hughston and Fivel from different point of views, and discussed by Adler et al..
## VIII The appearance of classical motion in macroscopic world
The above analysis has indicated that, when the energy difference between different branches of the wave function is large enough, say for the macroscopic situation<sup>\**</sup><sup>\**</sup>\**The largeness of the energy difference for macroscopic object results mainly from the environmental influences such as thermal energy fluctuations., the linear spreading of the wave function will be greatly suppressed, and the evolution of the wave function will be dominated by the localizing process, in fact, the motion state of the particle will be only local position state in appearance, and the evolution of this state will be only still or continuously move in space, this is just the display of CCM in macroscopic world.
Furthermore, we will show that the evolution law of CCM can also be derived, in fact, some people have strictly given the demonstration based on revised quantum dynamics , here we simply use the Enrenfest theorem, namely $`\frac{d<x>}{dt}=<p>`$ and $`\frac{d<p>}{dt}=<\frac{U}{x}>`$, as we have demonstrated, for macroscopic object its wave function will no longer spread, thus the average items in the theorem will represent the effective description quantities for the classical motion of the macroscopic object, and the classical motion law is also naturally derived in such a way, the result is $`\frac{dx}{dt}=p`$, the definition of the momentum, and $`\frac{dp}{dt}=\frac{U}{x}`$, the motion equation.
## IX Conclusions
In this paper, we strictly demonstrate the logical inevitability of the existent form and evolution law of CSTM, the existence of discrete space-time in Nature and resulting real existence of DSTM and its evolution law, this not only explains the appearance of classical motion in macroscopic world, as well as quantum motion in microscopic world consistently and objectively, but also presents a clear logical connection between quantum motion and classical motion, and unveils the unified realistic picture of microscopic and macroscopic world.
Acknowledgments
Thanks for helpful discussions with X.Y.Huang ( Peking University ), A.Jadczyk ( University of Wroclaw ), P.Pearle ( Hamilton College ), F.Selleri ( University di Bari ), Y.Shi ( University of Cambridge ), A.Shimony, A.Suarez ( Center for Quantum Philosophy ), L.A.Wu ( Institute Of Physics, Academia Sinica ), Dr S.X.Yu ( Institute Of Theoretical Physics, Academia Sinica ), H.D.Zeh.
Appendix:Mathematical Analysis About Motion In Continuous Space-time
First, we will give three general presuppositions about the relation between physical motion and mathematical point set, they are basic conceptions and correspondence rules before we discuss the physical motion of particles in continuous space-time.
(1). Time and space in which the particle moves are both continuous point set.
(2). The moving particle is represented by one point in time and space.
(3). The motion of particle is represented by the point set in time and space.
The first presupposition defines the continuity of space-time, the second one defines the existent form of particle in time and space, the last one relates the physical motion of particle with the mathematical point set.
For simplicity but lose no generality, in the following we will mainly analyze the point set in two-dimension space-time, which corresponds to one-dimension motion in continuous space-time.
### A Point set and its law—a general discussion
As we know, the point set theory has been deeply studied since the beginning of this century, nowadays we can grasp it more easily, according to this theory, we know that the general point set is dense point set, whose basic property is the measure of the point set, while the continuous point set is one kind of special dense point set, its basic property is the length of the point set.
Naturally, as to the point set in two-dimension space-time, the general situation is the dense point set in this two-dimension space-time, while the continuous curve is one kind of extremely special dense point set, surely it is a wonder that so many points bind together to form one continuous curve by order, in fact, the probability for its natural formation is zero.
Now, we will generally analyze the law of the point set, as we know, the law about the points in point set, which can be called point law, is the most familiar law, and it is widely taken as the only rational law, for example, as to the continuous curve in two-dimension space-time there may exist a certain expressible analytical formula for the points in this special point set<sup>\*†</sup><sup>\*†</sup>\*†People cherish this kind of point laws owing to their infrequent existence, but perhaps Nature detests and rejects them, since the probability of creating them is zero., but as to the general dense point set in two-dimension space-time the point law possesses no mathematical meaning, since the dense point set is discontinuous everywhere, even if the difference of time is very small, or infinitesimal, the difference of space can be very large, then infinitesimal error in time will result in finite error in space, thus even if the point law exists we can not formulate it in mathematics, and owing to finite error in time determination and calculation, we can not prove it either, let alone predict the evolution of the point set using it, in one word, there does not exist point law for dense point set in mathematics.
### B Deep analysis about dense point set
Now, we will further study the differential description of point set in detail.
First, in order to find the differential description of the special dense point set—continuous curve, we may measure the rise or fall size of the space $`\mathrm{\Delta }x`$ corresponding to any finite time interval $`\mathrm{\Delta }t`$ near each instant $`t_j`$, then at any instant $`t_j`$ we can get the approximate information about the continuous curve through the quantities $`\mathrm{\Delta }t`$ and $`\mathrm{\Delta }x`$ at that instant, and when the time interval $`\mathrm{\Delta }t`$ turns smaller, we will get more accurate information about the curve. In theory, we can get the complete information through this infinite process, that is to say, we can build up the basic description quantities for the special dense point set—continuous curve, which are the differential quantities dt and dx, then given the initial condition the relation between dt and dx at all instants will completely describe the continuous curve.
Then, we will analyze the differential description of the general dense point set, as to this kind of point set, we still need to study the concrete situation of the point set corresponding to finite time interval near every instant. Now, when time is during the interval $`\mathrm{\Delta }t`$ near instant $`t_j`$, the points in space are no longer limited in the local space interval $`\mathrm{\Delta }x`$, they distribute throughout the whole space instead, so we should study this new point set, which is also dense point set, for simplicity but lose no generality, we consider finite space such as x$``$, first, we may divide the whole space in small equal interval $`\mathrm{\Delta }x`$, the dividing points are denoted as $`x_i`$, then we can define and calculate the measure of the local dense point set in the space interval $`\mathrm{\Delta }x`$ near each $`x_i`$, which can be written as $`M_{\mathrm{\Delta }x,\mathrm{\Delta }t}`$($`x_i`$,$`t_j`$), since the measure sum of all local dense point sets in time interval $`\mathrm{\Delta }t`$ just equals to the length of the continuous time interval $`\mathrm{\Delta }t`$, we have:
$$\underset{i}{}M_{\mathrm{\Delta }x,\mathrm{\Delta }t}\text{(}x_i\text{,}t_j\text{)= }\mathrm{\Delta }t$$
(21)
On the other hand, since the measure $`M_{\mathrm{\Delta }x,\mathrm{\Delta }t}`$($`x_i`$,$`t_j`$) will also turn to be zero when the intervals $`\mathrm{\Delta }x`$ and $`\mathrm{\Delta }t`$ turn to be zero, it is not an useful quantity, and we have to create a new quantity on the basis of this measure. Through further analysis, we find that a new quantity $`\rho _{\mathrm{\Delta }x,\mathrm{\Delta }t}(x_i,t_j)=M_{\mathrm{\Delta }x,\mathrm{\Delta }t}(x_i,t_j)/(\mathrm{\Delta }x\mathrm{\Delta }t)`$, which can be called average measure density, will be an useful one, it generally does not turn to be zero when $`\mathrm{\Delta }x`$ and $`\mathrm{\Delta }t`$ turn to be zero, especially if the limit $`lim_{\mathrm{\Delta }x0,\mathrm{\Delta }t0}\rho _{\mathrm{\Delta }x,\mathrm{\Delta }t}(x_i,t_j)`$ exists, it will no longer relate to the observation sizes $`\mathrm{\Delta }x`$ and $`\mathrm{\Delta }t`$, so it can accurately describe the whole dense point set, as well as all local dense point sets near every instant, now we let:
$$\rho (x,t)=lim_{\mathrm{\Delta }x0,\mathrm{\Delta }t0}\rho _{\mathrm{\Delta }x,\mathrm{\Delta }t}(x,t)$$
(22)
then we can get:
$$_\mathrm{\Omega }\rho (x,t)𝑑x=1$$
(23)
this is just the normalization formula, where $`\rho `$(x,t) is called position measure density, $`\mathrm{\Omega }`$ denotes the whole integral space.
Now, we will analyze the new quantity $`\rho `$(x,t) in detail, first, the position measure density $`\rho `$(x,t) is not a point quantity, it is defined during infinitesimal interval, this fact is very important, since it means that if the measure density $`\rho (x,t)`$ exists, then it will be continuous relative to both t and x, that is to say, contrary to the position function x(t), there does not exist the discontinuous situation for the measure density function $`\rho `$(x,t), furthermore, this fact also results in that the continuous function $`\rho `$(x,t) is the last useful quantity for describing the dense point set; Secondly, the essential meaning of the position measure density $`\rho `$(x,t) lies in that it represents the dense degree of the points in the point set in two-dimension space and time, and the points are denser where the position measure density $`\rho `$(x,t) is larger.
### C The evolution of dense point set
Now, we will further discuss the evolution law for dense point set.
Just like the continuous position function $`x(t)`$, although the continuous position measure density function $`\rho `$(x,t) completely describes the dense point set, it is one kind of static description about the point set, and it can not be used for prediction itself, so in order to predict the evolution of the dense point set we must create some kind of quantity describing its change, enlightened by the theory of fluid mechanics we can define the fluid density for the position measure density $`\rho `$(x,t) as follows:
$$\frac{\rho (x,t)}{t}+\frac{j(x,t)}{x}=0$$
(24)
we call this new quantity $`j(x,t)`$ position measure fluid density, this equation measure conservation equation, it is evident that this quantity just describes the change of the measure density of dense point set, thus the general evolution equations of dense point set can be written as in the following:
$$\frac{\rho (x,t)}{t}+\frac{j(x,t)}{x}=0$$
(25)
$$\frac{j(x,t)}{t}+\mathrm{}=0$$
(26) |
no-problem/0001/cond-mat0001090.html | ar5iv | text | # Analytical Studies of Strategies for Utilization of Cache Memory in Computers
## Abstract
We analyze quantitatively several strategies for better utilization of the cache or the fast access memory in computers. We define a performance factor $`\alpha `$ that denotes the fraction of the cache area utilized when the main memory is accessed at random. We calculate $`\alpha `$ exactly for different competing strategies, including the hash-rehash and the skewed-associative strategies which were earlier analyzed via simulations.
The memory of a computer is organized in pages. A single page consists of several words. As the computer performs its computations, it reads from and writes into memory. When a word of main memory is accessed, the entire page containing the word is fetched and stored in a more accessible part of the computer’s hardware called the cache, which admits fast access. Subsequent accesses are likely to be for words in this page and hence the average memory access time is considerably reduced.
Since fast memory is expensive, the cache usually holds far fewer pages than the main memory. When a program requires a certain page that is not already in the cache, the computer has to fetch it from the main memory. But where in the cache should this new page be placed? The placements in the cache of these incoming pages must be organized so that (i) very little time is spent in locating the page where the word is stored in the cache and (ii) the cache area is maximally utilized, i.e., ideally pages should not be sent back to the main memory, when there is space left in the cache.
To fix our notations, we assume that the main memory has $`M=2^m`$ pages and the cache has $`N=2^n`$ pages, where $`mn`$. We use $`m`$-bit and $`n`$-bit 0-1 strings as addresses for pages in the main memory and the cache respectively. The page of the main memory corresponding to the address $`a\{0,1\}^m`$ will be denoted by $`P_a`$. Similarly $`Q_b`$ will denote the page in the cache corresponding to the address $`b\{0,1\}^n`$. When an access to page $`P_a`$ is made in the main memory, it has to be brought to a certain page $`Q_b`$ in the cache. So the question is what is the best strategy for choosing $`Q_b`$ (i.e., cache organization) such that the cache area is maximally utilized.
There are two extreme strategies for cache organization. Strategy $`1`$, usually called direct mapping strategy, assigns a fixed location $`Q_b`$ in the cache for each page $`P_a`$ of the main memory where $`b`$ is the first $`n`$ bits of $`a`$. That is, each time $`P_a`$ is fetched into the cache, it will be placed in page $`Q_b`$; if there is already a page of the main memory residing at $`Q_b`$, then that page will be sent back to the main memory (the main memory will be updated) and $`P_a`$ will replace it in the cache. In this strategy when an access to a page $`P_a`$ is made, we know exactly where to find it in the cache, namely $`Q_b`$. Thus it performs well on point (i) as no time is wasted in searching. However it performs poorly on point (ii) as a page can be sent out even when most of the cache is unused.
The second strategy, Strategy 2, usually called associative mapping strategy, allows $`P_a`$ to reside anywhere in the cache; if all the pages in the cache are already occupied, one of them, chosen according to some rule (e.g. least recently used), will be sent back to the main memory to make space for the $`P_a`$. In this strategy, when a page $`P_a`$ is accessed, determining if it is already present in the cache can be expensive, both from the point of view of time taken and the hardware needed to implement the search. It has, however, its advantage in the utilization of cache area, since no page in the cache is sent out unless the cache is full.
Thus Strategy 1, though preferable from the point of view of design, is likely to be inferior to Strategy 2 in utilization of the memory available in the cache. In order to improve the performance of caches, several other strategies, which try to to combine the advantages of Strategies 1 and 2, have been proposed. In this letter, we will primarily be concerned with three such strategies $`A`$, $`B`$ and $`C`$ mentioned below. These strategies perform considerably better than Strategy 2 on point (i), and are found to be better than Strategy 1 on point (ii). Our goal in this Letter would be to compare quantitatively the performances of these various strategies.
Strategy A: This is known as hash-rehash Strategy . In this strategy, each page of the main memory is allowed to reside in two locations $`Q_{b_1}`$ and $`Q_{b_2}`$ in the cache that are determined as follows: $`b_1`$ is the string of the first $`n`$ bits of $`a`$ and $`b_2`$ is the string of the next $`n`$ bits of $`a`$. When a page $`P_a`$ from the main memory is brought to the cache, it is first put in $`Q_{b_1}`$ provided $`Q_{b_1}`$ is empty; if $`Q_{b_1}`$ is occupied we place $`P_a`$ in $`Q_{b_2}`$. (If there is some page $`P_a^{}`$ residing at $`Q_{b_2}`$, then it is replaced by $`P_a`$ and is sent back to the main memory.)
Strategy B: This is known as two-way skewed-associative strategy . In this case, we divide the cache into two banks $`Q`$ and $`Q^{}`$, each with capacity $`2^{n1}`$. The pages of the two banks are denoted by $`Q_b`$ and $`Q_b^{}`$ respectively where $`b\{0,1\}^{n1}`$. With each page $`P_a`$ are associated two pages $`Q_{b_1}`$ and $`Q_{b_2}^{}`$, one from each bank. When a page $`P_a`$ is brought to the cache, we first try to store $`P_a`$ in the first bank at location $`Q_{b_1}`$, and if that fails, we store it in the second bank at location $`Q_{b_2}^{}`$.
Strategy C: This is known as two-way set-associative strategy . Here the cache is again divided into two banks. The only difference is that we associate pages $`Q_{b_1}`$ and $`Q_{b_1}^{}`$ with page $`P_a`$. That is, if $`Q_{b_1}`$ is already taken, we try to store the page in the second bank but at location $`Q_{b_1}^{}`$ (and not $`Q_{b_2}^{}`$ as in Strategy $`B`$).
Clearly, Strategies $`A`$, $`B`$ and $`C`$ are variants of Strategy $`1`$. Recently, based on data obtained from simulations, Seznec and Bodin have strongly advocated the use of skewed-associative caches (Strategy B). However no attempt appears to have been made to compare the performance of this strategy with older strategies within a theoretical framework. In this letter, we take the first steps in this direction. In particular, we calculate analytically a quantity called performance factor $`\alpha `$ (defined below) for various strategies mentioned above.
The cache has place to store $`N=2^n`$ pages. If $`N`$ pages are accessed at random, a perfect strategy would accomodate them all in the cache, without sending any page back to the main memory. But for most practical strategies some pages will be sent back to the main memory. We define the performance factor $`\alpha `$ of a strategy to be the expected fraction of the randomly accessed $`N`$ pages (in the $`N\mathrm{}`$ limit) that get accomodated in the cache. A perfect strategy has $`\alpha =1`$. For all other strategies, $`\alpha 1`$. We show below that $`\alpha _1=1e^1=0.63212\mathrm{}`$ for Strategy 1 but it increases considerably to $`\alpha _A=(e^21)/(e^2+1)=0.76159\mathrm{}`$, $`\alpha _B=1\frac{1}{2e^2}(1+e^{1e^2})=0.77167\mathrm{}`$ and $`\alpha _C=12e^2=0.72932\mathrm{}`$ for variants $`A`$, $`B`$ and $`C`$ respectively.
It turns out that $`\alpha _1`$ and $`\alpha _C`$ can be computed quite easily using elementary probability arguments. Let us first compute $`\alpha _1`$ for Strategy 1 in its original form. Suppose a page $`P_a`$ of the main memory is chosen at random. The first $`n`$ bits of $`a`$ will be uniformly distributed in $`\{0,1\}^n`$. Thus, each access corresponds to the choice of one of the $`N=2^n`$ pages of the cache, and we want to know how many distinct pages of the cache are expected to be chosen after $`N`$ random pages of the main memory are accessed. In other words, we have $`N`$ bins and $`N`$ balls and each ball is thrown at random into one of the bins; if the bin is already occupied then the new ball replaces the existing ball. (Clearly, it makes no difference to the calculations if we think that the old ball stays and it is the new ball that is discarded.) What is the expected number of bins that will be occupied after all $`N`$ balls have been tried? The probability that any fixed bin remains empty after $`k`$ balls have been tried is $`(11/N)^k`$. Therefore, the expected fraction $`x_1(k)`$ of the occupied bins after $`k`$ trials is given by,
$`x_1(k)=1\left(1{\displaystyle \frac{1}{N}}\right)^k.`$ (1)
The performance factor is then given by,
$`\alpha _1=\underset{n\mathrm{}}{lim}x_1(n)=1{\displaystyle \frac{1}{e}}=0.63212\mathrm{}.`$ (2)
Similar arguments can be used for Strategy C also. Here there are two banks and each bank has $`N^{}=N/2`$ bins. For every bin in the first bank, there is an associated bin in the second bank. A bin from the first bank is chosen at random and a ball is thrown into that bin. If the bin was already occupied then the ball is thrown into the associated bin in the second bank. If the second bin was also occupied, then the ball is rejected. Let $`x_1(k)`$ and $`x_2(k)`$ denote the fraction of occupied bins in the first and the second bank respectively after $`k`$ trials. Then clearly $`x_1(k)=1(11/N^{})^k`$ as given by Eq. (1). We also note that the probability that a bin in the second bank remains unoccupied given that its partner in the first bank is occupied is simply given by $`x_1(k)x_2(k)=\frac{k}{N^{}}(11/N^{})^{k1}`$. This is because out of $`k`$ trials, only one (which can be $`1`$-st, $`2`$-nd,$`\mathrm{}`$, or the $`k`$-th trial) should be succesful in choosing the given bin in the first bank and the others should be unsuccessful. So the performance factor $`\alpha _C`$ is given by,
$`\alpha _C=\underset{N\mathrm{}}{lim}{\displaystyle \frac{x_1(N)+x_2(N)}{N}}=12e^2=0.72932\mathrm{}.`$ (3)
It turns out that such elementary arguments, however, do not give us the expressions for the fraction of occupied bins for Strategies A and $`B`$ and one has to unfortuantely carry out more detailed computations for those two cases which we outline below.
We first consider Strategy $`A`$. In this case there are two fixed locations $`Q_{b_1}`$ and $`Q_{b_2}`$ in the cache corresponding to a page $`P_a`$ in the main memory. If $`a`$ is an $`m`$ bit string then $`b_1`$ consists of the first $`n`$ bits of $`a`$ and $`b_2`$ the next $`n`$ bits of $`a`$ (assuming $`m2n`$). Suppose that an address $`a`$ is accessed randomly by the computer. As $`a`$ varies uniformly over $`\{0,1\}^m`$, the two corresponding strings $`b_1`$ and $`b_2`$ also vary uniformly over $`\{0,1\}^n`$ and their distributions are independent. Thus, in the language of balls and bins, we have the following process. There are $`N`$ balls and $`N`$ bins. The balls are placed one after another using the following strategy. For a given ball, a bin is chosen at random and the ball is attempted to be placed there. If the bin is empty the ball occupies the bin; if the bin is occupied, another bin is picked at random (note $`b_2`$ is independent of $`b_1`$ unlike in Strategy C). If this second bin is empty, the ball occupies it. However, if this bin is also occupied, then the ball is discarded.
We define $`P_A(r,k)`$ to be the probability that $`r`$ bins are occupied after “time” $`k`$, i.e., after $`k`$ trials. Note that in this case each ball is tried at the most twice. The evolution equation for $`P_A(r,k)`$ is then given by,
$`P_A(r,k+1)`$ $`=`$ $`\left({\displaystyle \frac{r}{N}}\right)^2P_A(r,k)+\left[1\left({\displaystyle \frac{r1}{N}}\right)^2\right]`$ (4)
$`\times `$ $`P_A(r1,k).`$ (5)
with the “boundary” condition, $`P_A(0,k)=\delta _{k,0}`$ and the “initial” condition, $`P_A(r,0)=\delta _{r,0}`$. This equation can be written in a compact matrix form, $`[P_A(r,k+1)]=\widehat{W}[P_A(r,k)]`$ where $`\widehat{W}`$ is the $`(n+1)\times (n+1)`$ evolution matrix whose elements can be easily read off Eq. 5. This equation can be solved using the standard techniques of statistical physics which we outline below. The general solution can be written as
$`P_A(r,k)={\displaystyle \underset{\lambda }{}}a_\lambda Q_\lambda (r)\lambda ^k`$ (6)
where $`[Q_\lambda (r)]`$ is the right eigenvector (with eigenvalue $`\lambda `$) of the matrix $`\widehat{W}`$ and $`a_\lambda `$’s are constants to be determined from the initial condition. The eigenvalues of $`\widehat{W}`$ are simply, $`\lambda =l^2/N^2`$, where $`l=0,1,2,\mathrm{},N`$. The $`l`$-th right eigenvector satisfies the equation,
$`{\displaystyle \frac{r^2}{N^2}}Q_l(r)+\left[1{\displaystyle \frac{(r1)^2}{N^2}}\right]Q_l(r1)={\displaystyle \frac{l^2}{N^2}}Q_l(r),`$ (7)
for all $`r1`$. The generating function $`\stackrel{~}{Q}_l(z)=_0^NQ_l(r)z^r`$ satisfies the differential equation,
$`z^2(1z){\displaystyle \frac{d^2\stackrel{~}{Q}_l}{dz^2}}+z(1z){\displaystyle \frac{d\stackrel{~}{Q}_l}{dz}}(l^2N^2z)\stackrel{~}{Q}_l=0.`$ (8)
With a little algebra, it is not difficult to show that the well behaved solution of this differential equation is given by, $`\stackrel{~}{Q}_l(z)=z^l(1z)P_{Nl1}^{1,2l}(2z1)`$ for $`l<N`$ and $`\stackrel{~}{Q}_N(z)=z^N`$ where $`P_n^{a,b}(x)`$’s are Jacobi polynomials defined as,
$`P_n^{a,b}={\displaystyle \frac{1}{2^n}}{\displaystyle \underset{m=0}{\overset{n}{}}}\left({\displaystyle \genfrac{}{}{0pt}{}{n+a}{m}}\right)\left({\displaystyle \genfrac{}{}{0pt}{}{n+b}{nm}}\right)\left(x1\right)^{nm}\left(x+1\right)^m.`$ (9)
The final solution for the generating function, $`\stackrel{~}{P}_A(z,k)=_0^NP_A(r,k)z^r`$ is given by,
$`\stackrel{~}{P}_A(z,k)=z^N+{\displaystyle \underset{l=0}{\overset{N1}{}}}a_lz^l(1z)P_{Nl1}^{1,2l}(2z1)\left({\displaystyle \frac{l}{N}}\right)^{2k}`$ (10)
where the coefficients $`a_l`$’s are determined from the initial condition $`\stackrel{~}{P}_A(z,0)=1`$. After some amount of algebra, we get,
$`a_l=(1)^{l+N1}{\displaystyle \frac{n(2\delta _{m,0})}{nm}}`$ (11)
for $`l=0,1,2,\mathrm{}N1`$. Thus the expected fraction of occupied bins after time $`k`$, $`x_A(k)=\frac{1}{N}\frac{\stackrel{~}{P}_A(z,k)}{z}|_{z=1}`$ is given by,
$`x_A(k)=1+2{\displaystyle \underset{l=1}{\overset{N1}{}}}(1)^{l+N}\left({\displaystyle \frac{l}{N}}\right)^{2k}.`$ (12)
Putting $`k=N`$ in Eq. 12 and taking $`N\mathrm{}`$ limit, we get the performance factor,
$`\alpha _A={\displaystyle \frac{e^21}{e^2+1}}=0.76159\mathrm{}.`$ (13)
Next, we consider Strategy B. In this case, the cache area is divided equally into two banks. Corresponding to every page $`P_a`$ in the main memory, there are two fixed locations $`Q_{b_1}`$ in the first compartment and $`Q_{b_2}`$ in the second. Again $`b_1`$ and $`b_2`$ are uniformly and independently distributed over $`\{0,1\}^{n1}`$. The operation on the cache then corresponds to the following problem with balls and bins. There are $`N=2^n`$ bins, half of which, $`N^{}=2^{n1}`$, belong to the first compartment and the remaining $`N^{}`$ belong to the second. There are $`N`$ balls and each is placed in the bins according to the following rules. One of the first $`N^{}`$ bins is chosen at random. If the bin is occupied one of the $`N^{}`$ bins in the second half is chosen. If even this bin is occupied the ball is discarded.
Let $`P_B(r_1,r_2,k)`$ denote the probability that after “time’ $`k`$, the first compartment has $`r_1`$ occupied bins and the second $`r_2`$ occupied bins. Now the second compartment is tried only if the first compartment fails to accomodate a given ball. If there are $`r_1`$ occupied balls in the first compartment at time $`k`$, clearly the scond compartment has been tried only $`(kr_1)`$ times, out of which $`r_2`$ attempts have been successful. Noting that for each compartment separately, the local strategy is exactly as in Strategy 1 defined earlier, it is easy to see that,
$`P_B(r_1,r_2,k)=P_1(r_1,k)P_1(r_2,kr_1)`$ (14)
where $`P_1(r,k)`$ is the probability of having $`r`$ occupied balls in the first compartment after $`k`$ trials and is the same as in Strategy 1. $`P_1(r,k)`$ can be computed exactly by following the same steps as used for Strategy $`A`$. Using this exact result in Eq. 14 and after some algebra, we finally find the expected fraction of occupied bins (taking into account both the banks) after $`k`$ trials,
$`x_B(k)=1{\displaystyle \frac{1}{2}}\left(1{\displaystyle \frac{1}{N^{}}}\right)^kS(N^{},k)`$ (15)
where $`S(N^{},k)`$ is the sum,
$`S(N^{},k)={\displaystyle \underset{l=0}{\overset{N^{}}{}}}(1)^{N^{}l}\left({\displaystyle \genfrac{}{}{0pt}{}{N^{}}{l}}\right){\displaystyle \frac{N_{}^{}{}_{}{}^{l}}{(N^{}1)^N^{}}}\left({\displaystyle \frac{m}{N^{}}}\right)^k.`$ (16)
Putting $`k=N`$ and taking the limit $`N\mathrm{}`$, we finally get the performance factor $`\alpha _B`$ for Strategy B,
$`\alpha _B=1{\displaystyle \frac{1}{2e^2}}\left(1+e^{1e^2}\right)=0.77167\mathrm{}.`$ (17)
The method illustrated above can not only calculate the final performance factor but it also gives exactly the full probability distribution of the number of occupied bins at any arbitrary “time” $`k`$ and for any cache size $`N`$. It turns out that there is an easier method (to be described below) to compute just the performance factor and is valid only in the $`N\mathrm{}`$ limit. It not only correctly reproduces the exact results for $`\alpha `$’s as found above but can further be used to compute the performance factors for more general and complicated strategies.
Consider, for example, a generalized version of Strategy $`A`$ where instead of $`2`$ fixed locations per page in the cache, one now has $`p`$ fixed locations. In the “ball and bin” language, each ball is tried at most $`p`$ times before being discarded. Let $`r(k)`$ be the random variable that denotes the number of occupied bins after discrete “time” $`k`$. Clearly, the increment $`\mathrm{\Delta }r(k)=r(k+1)r(k)`$ is also a random variable that takes the value $`0`$ with probability $`(r/N)^p`$ (when all $`p`$ trials for a given ball are unsuccessful) and $`1`$ with probability $`1(r/N)^p`$. Taking expectations, we get $`r(k+1)r(k)=1(r/N)^p`$. We then define $`x=r/N`$ and $`t=k/N`$ and note that they become continuous in the $`N\mathrm{}`$ limit and $`x`$ evolves as
$`{\displaystyle \frac{dx}{dt}}=1x^p.`$ (18)
Using the method of bounded differences , it can be shown that for each $`t[0,1]`$ and $`ϵ>0`$
$`\mathrm{Pr}[|x(t)x(t)|ϵ]2\mathrm{exp}(2ϵ^2N/p).`$ (19)
From this, it follows that $`|x^px^p|O(N^{\frac{1}{2}})`$. Thus, one can neglect fluctuations in the $`N\mathrm{}`$ limit. Using this in Eq. 18, one gets a closed equation for $`x`$($`t`$) which can then be integrated to give the performance factor, $`\alpha _p=x`$($`t=1`$). For general $`p`$, we get $`\alpha _p`$ as solution of the equation,
$`{\displaystyle _0^{\alpha _p}}{\displaystyle \frac{dy}{1y^p}}=1.`$ (20)
For example, for $`p=1`$ and $`p=2`$, we reproduce respectively, $`\alpha _1=1e^1`$ and $`\alpha _A=(e^21)/(e^2+1)`$ from Eq. 20. For large $`p`$, we find from Eq. 20,
$`\alpha _p=1{\displaystyle \frac{\mathrm{log}2}{p}}+O\left({\displaystyle \frac{1}{p^2}}\right).`$ (21)
Thus as $`p`$ increases, the deviation from perfect behaviour $`(1\alpha _p)`$ decays as a power law $`1/p`$.
In a similar fashion, one can consider a generalized version of Strategy $`B`$ where instead of 2 equal compartments in the cache, one now considers $`p`$ equal compartments. In this case, once again the performance factor for general $`p`$ can be easily computed using the continuous time method. Apart from reproducing the result for $`p=2`$, we find that for large $`p`$,
$`\alpha _B(p)=1{\displaystyle \frac{1}{(e1)p}}+O\left({\displaystyle \frac{1}{p^2}}\right).`$ (22)
Another generalization of Strategy $`B`$ would be to consider $`2`$ banks only but of unequal sizes $`N`$ and $`\gamma N`$ respectively. Then a similar calculation shows that the performance factor, which is now a function of $`\gamma `$, is given by,
$`\alpha _B(\gamma )=1{\displaystyle \frac{e^{(1+\gamma )}}{1+\gamma }}{\displaystyle \frac{\gamma }{e(1+\gamma )}}e^{\mathrm{exp}[(1+\gamma )/\gamma ]}.`$ (23)
This function has a single maximum at $`\gamma =0.68932\mathrm{}`$ where $`\alpha _B(max)=0.775862\mathrm{}`$, clearly better than $`\alpha _B(1)=0.77167\mathrm{}`$, the result for two equal banks.
In summary, we have studied analytically the performance of several strategies for cache utilization. We have derived exact values for cache utilization factors for these strategies under the assumption that pages in the main memory are accessed at random.
Cache utilization is just one of several issues in the design of caches. Indeed, it has been reported that two-way set associative caches require smaller execution times than hash-rehash caches, although, our analysis indicates that the latter utilize the cache better.
To get performance values that will be useful to designers of cache, one should take into account the amount of time spent in decoding addresses, updating tables and the cost of hardware. Also, instead of assuming that the pages of the main memory are accessed at random, a better understanding of the access patterns in typical applications might be needed.
We are grateful to Abhiram Ranade for bringing the question of analysis of skewed associative caches to our attention; in particular, the idea of using unequal banks in this strategy is due to him. We also thank D. Dhar for useful discussions. |
no-problem/0001/astro-ph0001363.html | ar5iv | text | # Vorticity Budget of Weak Thermal Convection in Keplerian disks
## 1. Introduction
Turbulence is usually driven by waves or instabilities in fluids, taping accessible free energy into fluctuating velocity fields. For astronomical objects, accessible free energy can come from body forces (e.g., gravity or Lorentz force) or from large-scale fluid motions (e.g., rotation or shear).
In Keplerian disks, thermal convection is an important ingredient in the thermal-viscous instability model which explains the semi-periodic changes of light curves of dwarf novae (Meyer & Meyer-Hofmeister (1981)) and soft X-ray transients (Mineshige & Wheeler (1989)). However, it has been argued that convective instability cannot access free thermal energy (i.e., higher entropy near the mid-plane) to transport energy vertically outward if it can access free rotational energy (i.e., higher angular velocities at small radii) to transport angular momentum radially outward. Recently a more general argument has been established to include the turbulence which is not thermally driven and whose azimuthal pressure perturbation is small. The equations of averaged turbulent kinetic energy due to local hydrodynamic mixing can be expressed as (Balbus & Hawley (1998))
$`{\displaystyle \frac{}{t}}{\displaystyle \frac{\rho v_r^2}{2}}+{\displaystyle \frac{1}{2}}\rho v_r^2𝐯`$ (1)
$`=2\mathrm{\Omega }\rho v_rv_\theta v_r{\displaystyle \frac{\delta P}{r}}\mathrm{losses},`$
$`{\displaystyle \frac{}{t}}{\displaystyle \frac{\rho v_\theta ^2}{2}}+{\displaystyle \frac{1}{2}}\rho v_\theta ^2𝐯`$ (2)
$`={\displaystyle \frac{\kappa ^2}{2\mathrm{\Omega }}}\rho v_rv_\theta v_\theta {\displaystyle \frac{\delta P}{r\theta }}\mathrm{losses},`$
where “losses” represent the energy sink due to viscosity, $`v_i`$ is the turbulent velocity field, $`\delta P`$ is the pressure perturbation, $`\rho `$ is the mass density, $`\mathrm{\Omega }`$ is the Keplerian angular velocity, and the epicyclic frequency $`\kappa ^2=(2\mathrm{\Omega }/r)d(r^2\mathrm{\Omega })/dr`$. The notation $`<>`$ in equations (1) and (2) denotes the averaging carried out over the ensemble of turbulent cells, indicating the time/space correlation between two fluctuating quantities. In the case of accretion disks, the time average is taken over several eddy turnover time (but still smaller than the viscous time scale), and the space average is performed over $`2\pi `$ in the azimuthal direction, over the whole disk vertical scale height<sup>1</sup><sup>1</sup>1Averaging over the disk height is not performed when one would like to investigate the detailed vertical structures (e.g., see Stone et al. (1996); Miller & Stone (2000).), and over several eddy sizes in the radial direction (but still much smaller than the disk radius).
As indicated by equation (2), the epicyclic term is an energy sink of $`v_\theta ^2`$ if $`v_rv_\theta >0`$. Therefore any local hydrodynamic instabilities cannot grow when azimuthal pressure perturbation is small. If a disk, however, is heated by other sources, the energy of weak convection can be maintained and then transport angular momentum inward. When thermal convection is weak, the associated negative Reynolds stress is a result of conservation of angular momentum of turbulent elements during the process of local mixing in Rayleigh-stable disks such as Keplerian flows (Balbus (2000)).
Besides considering the epicyclic effect which damps fluctuating velocity fields, the same damping effect should apply to fluctuating vorticity fields which characterize strong turbulence. It is also worth investigating how fluctuating vorticity fields interact with background vorticity when convection is maintained by other heating sources such as MHD turbulence driven by the Balbus-Hawley instability (Balbus & Hawley (1998)), in order to determine the typical linear modes for convection in nonlinear regime (Gu, Vishniac, & Cannizzo (2000), hereafter GVC). As high levels of fluctuating vorticities are observed in non-rotating shear flows at high Reynolds numbers, we must be wondering why the generation of fluctuating vorticities in Rayleigh-stable disks cannot be achieved by the usual vortex dynamics such as vortex stretching or vortex tilting, as noted and numerically simulated by Hawley, Balbus, & Winters (1999).
Perhaps one of the most difficult tasks in dealing with equations of turbulence is to determine the signs of correlations between fluctuating quantities. Positive, negative, or no correlations are usually related to the properties of background flows such as stratification, rotation, background shear, or body forces. Positive Reynolds stresses in non-rotating shear flows, for example, are always related to the mean flow moving in $`+x`$ direction with negative gradient in $`+y`$ direction. This situation is reversed in Keplerian flows for weak convection (Ryu & Goodman (1992); Kley et al. (1993); Kumer et al. (1995); Cabot (1996); Stone & Balbus (1996); Balbus (2000)). Positive kinetic helicity occurring in magnetized disks shown in the numerical simulation by Brandenburg & Donner (1997) probably means that the Coriolis force is of less importance in Keplerian disks where MHD turbulence is driven by the Balbus-Hawley instability. Similar to these examples, we would expect that background shear and rotation can determine the signs of correlations in the equations of fluctuating vorticities.
Beginning with the perturbed vorticity equations in a non-rotating flow, we determine the signs of turbulent correlation based on positive Reynolds stresses without worrying about nonlinear shear instabilities in §2. In §3, we apply the analysis presented in §2 to Keplerian flows, showing that mean-square vorticity perturbations of weak convection cannot grow with positive Reynolds stresses. Positive kinetic helicity driven by the Balbus-Hawley instability in a stratified Keplerian disk is analyzed through vortex equations in §4. In the last section, we comment on the possibility of outward angular momentum by strong convection in accretion disks based on the recent development of theories concerning about azimuthal pressure perturbations and directions of energy cascade.
## 2. momentum conservation and vorticity equations in non-rotating shear flows
The vorticity equation reads
$$\frac{D𝐰}{Dt}=\left(𝐰\right)𝐮\left(𝐮\right)𝐰+\frac{\rho \times P}{\rho ^2}+\times \nu ^2𝐮,$$
(3)
where $`𝐰=\times 𝐮`$, $`D/Dt=_t+(𝐮)`$, $`P`$ is the pressure, $`\rho `$ is the mass density, and $`\nu `$ is the kinematic viscosity. Consider a 3-D turbulent flow with a background shear $`dV/dx`$ without rotation, where $`V(x)`$ is the background flow in $`+y`$ direction. Let the vertical scale height be much smaller than the scale heights in $`x`$ and $`y`$ directions. If the eddy growth rate is smaller than the shearing rate (i.e., weak convection), linearizing the vorticity equation ($`𝐮=V\widehat{\mathbf{ȷ}}+𝐯`$, and $`𝝎=\times 𝐯`$) gives the equations of mean-square vorticity (enstrophy) perturbation:
$`{\displaystyle \frac{D}{Dt}}{\displaystyle \frac{\omega _x^2}{2}}\omega _x{\displaystyle \frac{v_x}{z}}{\displaystyle \frac{dV}{dx}}`$ (4)
$`{\displaystyle \frac{1}{\rho ^2}}{\displaystyle \frac{\rho }{z}}\omega _x{\displaystyle \frac{\delta P}{y}}+{\displaystyle \frac{1}{\rho ^2}}{\displaystyle \frac{P}{z}}\omega _x{\displaystyle \frac{\delta \rho }{y}}\mathrm{losses},`$
$`{\displaystyle \frac{D}{Dt}}{\displaystyle \frac{\omega _y^2}{2}}\omega _x\omega _y{\displaystyle \frac{dV}{dx}}+\omega _y{\displaystyle \frac{v_y}{z}}{\displaystyle \frac{dV}{dx}}`$ (5)
$`+{\displaystyle \frac{1}{\rho ^2}}{\displaystyle \frac{\rho }{z}}\omega _y{\displaystyle \frac{\delta P}{x}}{\displaystyle \frac{1}{\rho ^2}}{\displaystyle \frac{P}{z}}\omega _y{\displaystyle \frac{\delta \rho }{x}}\mathrm{losses},`$
$`{\displaystyle \frac{D}{Dt}}{\displaystyle \frac{\omega _z^2}{2}}v_x\omega _z{\displaystyle \frac{d^2V}{dx^2}}`$ (6)
$`\omega _z\left({\displaystyle \frac{v_x}{x}}+{\displaystyle \frac{v_y}{y}}\right){\displaystyle \frac{dV}{dx}}\mathrm{losses},`$
where the terms “losses” represent the energy sink due to viscosity, $`v_i`$ and $`\omega _i`$ are fluctuating velocity and vorticity respectively, and $`D/Dt`$ denotes $`_t+V_y`$. We have ignored the term $`\omega _i\omega _j_jv_i`$ which represents stochastic stretching of vortex by turbulent shear. This effect is smaller than stretching by background shear as long as background shear rate is larger than turbulent growth rate. Similar to the role played by $`\delta P/r`$ in equation (1), the terms associated with $`\delta P/x`$ and $`\delta \rho /x`$ are energy sources of $`\omega _y^2`$ in equation (5). Since these terms come from the baroclinic term in equation (3), they appear as a result of vertical stratification.
The production or annihilation of turbulent vorticity relies on pressure perturbations, density perturbations, and the signs of correlations between fluctuating vorticities and fluctuating velocity shears as shown in the above equations. However, the term associated with $`_y`$ is usually small for weak convection as a result of strong shear. Assuming that all perturbation quantities are proportional to $`\mathrm{exp}(ik_xx+ik_yy+ik_zz+i\omega t)`$, we have the linear perturbation equations for adiabatic convection under Boussinesq approximation:
$`i\overline{\omega }v_x+ik_x\mathrm{\Psi }`$ $`=`$ $`0,`$ (7)
$`i\overline{\omega }v_y+{\displaystyle \frac{dV}{dx}}v_x+ik_y\mathrm{\Psi }`$ $`=`$ $`0,`$ (8)
$`i\overline{\omega }v_z+ik_z\mathrm{\Psi }+\delta g_z`$ $`=`$ $`0,`$ (9)
$`i\overline{\omega }\delta v_z_z\mathrm{ln}\left({\displaystyle \frac{P^{1/\mathrm{\Gamma }}}{\rho }}\right)`$ $`=`$ $`0,`$ (10)
$`k_xv_x+k_yv_y+k_zv_z`$ $`=`$ $`0,`$ (11)
with the shearing constraint (Shu (1974))<sup>2</sup><sup>2</sup>2This constraint is called the “weak-shear limit” in Shu (1974). In fact, we are dealing with convection with the growth rates smaller than the shearing rate in this paper. In order to avoid confusion, we do not use the term “weak-shear” here.
$$k_y<k_x\frac{i\overline{\omega }}{|dV/dx|},$$
(12)
where $`\overline{\omega }`$ is the frequency measured by a local observer comoving with the mean flow, $`\delta \delta \rho /\rho `$ is the local fractional density perturbation, $`\mathrm{\Psi }\delta P/\rho `$ is the pressure perturbation divided by the density, $`v_x`$, $`v_y`$, and $`v_z`$ are velocity perturbations, $`k_x`$, $`k_y`$ and $`k_z`$ are the wavenumbers in $`x`$, $`y`$ and $`z`$ directions. The validity of the plane-wave approximation for $`k_x`$ is assured by the shearing constraint for thermal convection in the presence of a shear flow. The plane-wave approximation for $`k_z`$ in the presence of a vertical gradient is not a bad assumption for weak convection since first of all, the dispersion relation (19) inferred from the above Boussinesq equations in the case of accretion disks is the same as the one derived by Ruden, Papaloizou, & Lin (1988), who consider the vertical wavelength of axisymmetrical modes for convection as a function of $`z`$ owing to vertical stratification. Second of all, as we shall see in the following text, the equations (7)–(12) serve only as an example to illustrate how the phenomenological interpretation works for explaining the directions of angular momentum transport done by Balbus (2000), and how convective eddies approximately look like in a shear flow, all of which are not very sensitive to the detailed vertical structures of disks. We are going to adopt the same phenomenological strategy to evaluate the signs of correlations in the equations of perturbed enstrophy, and consequently to investigate if turbulent vorticities can be generated by vortex stretching/tilting in the regime where the terms associated with $`_y`$ are small.
By virtue of equations (7) through (12) in the weak-convection limit (i.e., the convective growth rate $`i\overline{\omega }<|dV/dx|`$), one can show that the terms associated with $`_y`$ (or say $`k_\theta `$) in the above Boussinesq equations and in the equations of perturbed enstrophy are smaller than the terms associated with $`dV/dx`$, owing to the shearing constraint. Without the pressure perturbation in the $`y`$ direction, equation (8) states that a turbulent element can conserve momentum in the $`y`$ direction when it moves.
In the case $`dV/dx<0`$, conservation of momentum of turbulent elements in the $`y`$ direction described by equation (8) implies that $`v_xv_y>0`$, i.e., as observed in the comoving frame, positive (negative) values of $`v_x`$ should occur more frequently than negative (positive) ones when $`v_y`$ is positive (negative), or vice versa. By virtue of equation (8), $`v_y^2`$ can grow when $`v_xv_y>0`$ (Balbus & Hawley (1998)). The major concern for a successful growth of turbulent vorticity is that the first term of the right hand side in equation (4) must be positive because the other terms associated with $`_y`$ are small. Since it is the equation (8) which determines the sign of $`v_xv_y`$ and the growth of turbulence, we would expect that the same equation plays the same role in equation (4). By virtue of equation (8) again, we have
$$\frac{1}{2}\frac{\gamma }{z}\frac{v_y^2}{z}+\gamma \left(\frac{v_y}{z}\right)^2=\frac{dV}{dx}\frac{v_y}{z}\frac{v_x}{z},$$
(13)
where the growth rate $`\gamma i\overline{\omega }>0`$. The first term on the left hand is positive since turbulent velocity fields usually increase with convective growth rate. As a result, the correlation $`_zv_y_zv_x`$ is positive. The first term on the right hand side of equation (4) is $`_zv_y_zv_xdV/dx`$ which is therefore positive, leading to successful growth of $`\omega _x^2`$. In fact, this positive sign can be understood in another way. The equation of vorticity $`D\omega _x/Dt_zv_x(dV/dx)`$ suggests that $`\omega _x`$ tends to be negative (positive) when $`_zv_x>0`$ ($`<0`$) and $`dV/dx<0`$ ($`>0`$). Hence $`\omega _x`$ and $`_zv_x`$ are negatively correlated in a non-rotating shear flow. In other words, $`\omega _x^2`$ can grow from the interaction between background ($`dV/dx`$) and turbulent ($`_zv_x`$) shear.
Although the signs of correlations associated with background shear in equation (5) are not important because there are energy source terms due to pressure/density perturbations, we can possibly determine the sign of $`\omega _x\omega _y`$ phenomenologically. In the case of $`dV/dx<0`$, if we consider a turbulent element moving toward $`+x`$ direction with positive sense of $`\omega _y`$, the element will tend to move toward $`+y`$ direction more frequently so that the negative sense of $`\omega _x`$ is created as seen in the comoving frame. This effect of vortex rotation is therefore described by the correlation $`\omega _x\omega _y<0`$. The signs is reversed when $`dV/dx>0`$. After all, the term $`\omega _x\omega _ydV/dx`$ is an energy source of $`\omega _y^2`$.
The term $`v_x\omega _z`$ in equation (6) describes the vorticity transport in the flows which have nonuniform background vorticity (i.e., $`d^2V/dx^20`$). Assume that the angular momentum of vertical vortex tubes are nearly conserved; i.e., the loss term in equation (6) is small. If the gradient of background vorticity $`d^2V/dx^2`$ is positive (negative), the local turbulent mixing will transport the vorticity toward $`x`$ ($`+x`$) direction, or vice versa. This means that the term $`v_x\omega _zd^2V/dx^2>0`$. However, the production provided by vorticity transport is supposed to be small since $`v_x`$ and the correlation length across the mean shear flows are severely shorten by strong shear. This point can be also understood in terms of the fact that the term $`v_x\omega _z`$ is zero in a shear flow without a gradient of a background state in the $`x`$ direction. Introducing a gradient of background state gives rise to a factor about one over the scale height along the $`x`$ direction which is much smaller than one over the vertical scale height associated with other terms in equation (6).
The second term in equation (6) describes vortex generation due to the interaction between background vorticity $`dV/dx`$ and variation of cross sections of vortices. The equation of motion $`D\omega _z/Dt(_xv_x+_yv_y)(dV/dx)`$ suggests that $`\omega _z`$ tends to be negative (positive) when $`_xv_x+_yv_y<0`$ ($`>0`$) and $`dV/dx<0`$ ($`>0`$). This means that the second term in equation (6) is the energy source of $`\omega _z^2`$. For a nearly incompressible disturbance, $`_xv_x+_yv_y<0`$ means vortex stretching (i.e., $`_zv_z>0`$), and $`_xv_x+_yv_y>0`$ means vortex squeezing (i.e., $`_zv_z<0`$) in the vertical direction.
After all, the turbulent vorticities $`\omega _x`$ and $`\omega _y`$ can grow simultaneously via vortex tilting by background shear ($`dV/dx`$) and by turbulent shear ($`_jv_i`$). The turbulent vorticity $`\omega _z`$ can grow via vortex stretching/squeezing. The different vortex dynamics associated with $`\omega _z`$ comes from the fact that $`\omega _z`$ does not couple with $`\omega _x`$ and $`\omega _y`$ as shown in equations (4), (5), and (6). We note that vortex stretching/tilting is a 3-D effect and cannot exist in 2-D fluids ($`\omega _x\omega _y0`$) which are usually employed to model vortex generation in larger scales $`>>`$ vertical scale height $`h`$ (i.e., the shallow water approximation). Large-scale vortex production in 2-D usually relies on the baroclinic term in the first place, and subsequently vortexes can evolve by mutual interactions such as merging into large ones which have long lifetimes (Adams & Watkins (1995); Godon & Livio (1999)). In this paper, we do not consider this evolution.
Before we apply the same approach to Keplerian flows, we need to emphasize that the above analysis just serves as preliminary calculations for Keplerian flows. The turbulent fields driven by thermal convection in a planar shear flow are more complicated than those in a Keplerian flow. In the former case, thermal convection excites a non-linear shear instability which overwhelms the convective motion (Stone & Balbus (1996)), and therefore the above analysis for thermal convection in a planar shear flow is quite questionable; for example, the term $`\omega _i\omega _j_jv_i`$ which we ignored becomes most important (Tennekes & Lumley (1972)). In the case of Keplerian flows, however, numerical simulations thus far have shown that strong epicyclic motion suppresses non-linear shear instabilities (Hawley, Balbus, & Winters (1999), however see Richard & Zahn (1999), and Klahr (2000)). Hence the above analysis should be reasonable for weak convection in a Keplerian flow. We shall see in the next section that the result implied from the equations of mean-square vorticity perturbations are indeed consistent with the result inferred from the equations of mean-square velocity fluctuations.
## 3. angular momentum conservation and vorticity equations in a Keplerian flow
In Keplerian flows, the vorticity equations in cylindrical coordinates become
$`{\displaystyle \frac{D}{Dt}}{\displaystyle \frac{\omega _r^2}{2}}\omega _r{\displaystyle \frac{v_r}{z}}{\displaystyle \frac{\kappa ^2}{2\mathrm{\Omega }}}`$ (14)
$`{\displaystyle \frac{1}{\rho ^2}}{\displaystyle \frac{\rho }{z}}\omega _r{\displaystyle \frac{\delta P}{r\theta }}\mathrm{losses},`$
$`{\displaystyle \frac{D}{Dt}}{\displaystyle \frac{\omega _\theta ^2}{2}}\omega _r\omega _\theta r{\displaystyle \frac{d\mathrm{\Omega }}{dr}}`$ (15)
$`+\omega _\theta {\displaystyle \frac{v_\theta }{z}}{\displaystyle \frac{\kappa ^2}{2\mathrm{\Omega }}}+{\displaystyle \frac{1}{\rho ^2}}{\displaystyle \frac{\rho }{z}}\omega _\theta {\displaystyle \frac{\delta P}{r}}\mathrm{losses},`$
$`{\displaystyle \frac{D}{Dt}}{\displaystyle \frac{\omega _z^2}{2}}v_r\omega _z{\displaystyle \frac{d}{dr}}\left({\displaystyle \frac{\kappa ^2}{2\mathrm{\Omega }}}\right)`$ (16)
$`\omega _z\left({\displaystyle \frac{v_r}{r}}+{\displaystyle \frac{v_\theta }{r\theta }}\right){\displaystyle \frac{\kappa ^2}{2\mathrm{\Omega }}}\mathrm{losses},`$
where $`D/Dt`$ denotes $`_t+\mathrm{\Omega }_\theta `$. In a Keplerian disk threaded with sub-thermal magnetic fields, the “loss” terms in above equations include the turbulent damping due to radial mixing driven by the Balbus-Hawley instability. We do not show the terms associated with density perturbations explicitly because unlike $`\delta P/P\delta \rho /\rho `$ in non-rotating shear flows owing to equation (7), $`\delta P/P\delta \rho /\rho `$ in rotating shear flows due to the Coriolis force (see equation (18) below). Once again $`\omega _z`$ does not couple with $`\omega _r`$ and $`\omega _\theta `$.
Although perturbation equations (7) through (12) should be modified by adding the Coriolis force in a rotating flow, the Coriolis term $`2\mathrm{\Omega }`$ is comparable to the shear term $`rd\mathrm{\Omega }/dr`$. Therefore the terms associated with $`_\theta `$ are small compared to the first term on the right hand side in equation (14), analogous to equation (4). In a Keplerian flow, neglecting azimuthal pressure perturbations means that turbulent elements conserve angular momenta before they mix with background fluid. This leads to a transport of angular momentum down to the angular momentum gradient, giving rise to negative Reynolds stresses in a Keplerian disk (Balbus (2000)). This picture is described by the azimuthal perturbed equation of motion which is similar to equation (8) except dV/dx is replaced by $`\kappa ^2/2\mathrm{\Omega }`$; i.e.,
$$\gamma v_\theta =\frac{\kappa ^2}{2\mathrm{\Omega }}v_r=\left(r\frac{d\mathrm{\Omega }}{r}+2\mathrm{\Omega }\right)v_r,$$
(17)
where we have dropped the term $`ik_\theta \mathrm{\Psi }`$. The above equation indicates that $`v_\theta ^2`$ cannot grow when $`v_rv_\theta >0`$ (Balbus & Hawley (1998)), which is different from the situation in a non-rotating shear flow owing to the appearance of the Coriolis force.
In fact, negative Reynolds stresses can be realized as follows. Consider a turbulent element is moving with positive $`v_r`$. While shear ($`rd\mathrm{\Omega }/dr`$) tries to move the element to the $`+\widehat{\theta }`$, as observed in a corotating frame, the Coriolis force ($`2\mathrm{\Omega }`$) tried to move the element to the $`\widehat{\theta }`$. Since $`2\mathrm{\Omega }>r|d\mathrm{\Omega }/dr|`$ in Keplerian disks (i.e., $`\kappa ^2>0`$), the Coriolis force is the winner. Therefore the turbulent element tends to have $`v_\theta `$, leading to $`v_rv_\theta <0`$. Without rotation (Coriolis effect), $`v_xv_y>0`$ in a non-rotating shear flow with negative $`dV/dx`$. The contribution to $`v_rv_\theta `$ from the radial deviation of turbulent azimuthal motion due to the Coriolis force is small since for weak convection in a Keplerian disk, equation (7) becomes
$$ik_r\mathrm{\Psi }2\mathrm{\Omega }v_\theta ,$$
(18)
where the term $`i\overline{\omega }v_r`$ is small and has been ignored. This means that the strong radial gradient of pressure perturbation is roughly balanced by the Coriolis force, resulting in negligible contribution to $`v_rv_\theta `$. In other words, $`v_rv_\theta `$ is mostly determined by equation (17). We note that in contrast to equation (7), the extra term due to the Coriolis force in equation (18) indicates that the radial gradient of pressure perturbation should be large enough to make convection grow in a rotating flow. Namely, the Coriolis force is a stabilized factor to convection. Owing to equations (17) and (18), the linear theory without radiative and turbulent damping indicates that $`\gamma ^2`$ is smaller than the Brunt-Väisälä frequency $`N^2`$ (GVC; Ruden et al. 1988):
$$\gamma ^2=\frac{N^2A\mathrm{\Omega }^2}{A+1},$$
(19)
where $`A(k_z/k_r)^2`$. It has been known in terrestrial experiments that the critical Rayleigh numbers of rotating Rayleigh-Benard convection are increased above their non-rotating values (e.g., Zhong et al. (1993)).
The above analysis means that for weak convection, the signs of correlations in Keplerian flows in equations (14) (15), and (16) are reversed compared with those in the non-rotating flows with $`dV/dx<0`$ when we replace $`xy`$ with $`r\theta `$. In other words, we have $`v_rv_\theta <0`$, $`\omega _r\omega _\theta >0`$, $`\omega _r_zv_r>0`$, and $`\omega _z(_rv_r+_\theta v_\theta /r)<0`$ when $`\kappa ^2>0`$. The right hand side of equation (14) becomes positive and $`\omega _r^2`$ can grow. Although the first term on the right hand side of equation (15) turns out to be negative, $`\omega _\theta ^2`$ can be maintained by the radial gradient of perturbed pressure owing to strong epicyclic effect.
If we assume that weak convection transports angular momentum down to the angular velocity gradient, then the signs of correlations in equation (14) and equation (15) are the same as those in the non-rotating flows with $`dV/dx<0`$ when we replace $`xy`$ with $`r\theta `$. This means that the first term on the right hand of equation (14) is negative because $`\kappa ^2>0`$. Hence $`\omega _r^2`$ cannot grow and weak convection dies away.
If the disks are Rayleigh-unstable (i.e., $`\kappa ^2<0`$), the signs of all terms in equations (14), (15), and (16) remain unchanged compared to those with positive $`\kappa ^2`$ except that the first term on the right hand side in equation (15) becomes positive because $`\omega _r\omega _\theta <0`$. In other words, a Rayleigh-unstable disk behaves as a non-rotating shear flow in the sense that $`v_rv_\theta >0`$. We note that our analysis for Rayleigh-unstable disks is oversimplified because a non-linear shear instability should be excited.
## 4. positive kinetic helicity as a result of the Balbus-Hawley instability
A weakly magnetized Keplerian disk is linearly unstable to the Balbus-Hawley instability. In nonlinear regime, this instability can drive MHD turbulence which in turn leads to a dynamo process unless the magnetic Reynolds numbers are low (Balbus & Hawley (1998)). Brandenburg & Donner (1997) shows that this dynamo process observed in simulations can be imitated by an $`\alpha \mathrm{\Omega }`$ dynamo in a vertically stratified disk, with a negative $`\alpha _{\theta \theta }`$ in the upper disk plane (where $`\alpha _{\theta \theta }`$ is the $`\theta `$-$`\theta `$ component of kinetic helicity tensor $`\alpha `$). A negative $`\alpha _{\theta \theta }`$ means that the kinetic helicity $`v_z\omega _z`$ is positive for nearly isotropic turbulence. This contradicts with the usual notion that the kinetic helicity of convection driven by the Coriolis force is usually negative. Brandenburg (1998) suggests that the right-handed helical turbulence results from the combined effect of the Balbus-Hawley instability (i.e., $`\delta B_r\delta B_\theta <0`$) and magnetic buoyancy.
Although we have concentrated on convection-like eddies and have not taken into account the effect of sub-thermal magnetic fields in this paper, equations (6) and (16) could shed light on preferred directions of helical turbulence if we realize that a Coriolis-dominated disk characterized by equation (16) becomes a shear-dominated disk symbolized by equation (6) due to the Balbus-Hawley instability. As explained in the preceding section, $`\omega _z(_rv_r+_\theta v_\theta /r)<0`$ in a nonmagnetized Keplerian disk. Assume that a turbulent element expands (contracts) as it goes upward (downward) in the upper disk plane. The above negative correlation means that a turbulent element tends to have a negative (positive) $`\omega _z`$ due to expansion (contraction). Consequently, $`v_z\omega _z<0`$ (see the vortex plot associated with positive $`\kappa ^2`$ in Figure 1). This result is consistent with the picture in which the Coriolis force is the winner over shear so that turbulent eddies are mostly left-handed; i.e., as observed in a corotating frame, left-handed helical turbulent motion results from a positive vorticity ($`\kappa ^2>0`$) of the mean flow. However, introducing the Balbus-Hawley instability changes the sign of Reynolds stresses from negative to positive in Keplerian disks because angular momentum of a turbulent element is changed by fluctuating magnetic torques (Balbus (2000)), leading to the condition for the instability changed from the Rayleigh criterion $`\kappa ^2<0`$ to the Chandrasekhar-Velikov criterion $`d\mathrm{\Omega }^2/dr<0`$ (Velikhov (1959); Chandrasekhar (1960); Fricke (1969)) in a Keplerian flow. Equivalently speaking, the Velikov-Chandrasekhar criterion is a manifestation of conservation of angular velocity of turbulent elements due to the fact that the fluctuating magnetic torque counteracts the torque of the Coriolis force in a weak field limit (Fricke (1969)). Without the torque by the Coriolis force, azimuthal dynamics of turbulence in Keplerian disks becomes that in a non-rotating shear flow. Furthermore, if the contribution to $`v_rv_\theta `$ due to the radial deviation of azimuthal motion is small (like the situation described by equation (18); radial deviation of azimuthal motion is usually small owing to shear), then the signs of correlation quantities in a weakly magnetized disk should behave as the one described by negative vorticity of the mean flow (such as the correlation quantities in equations (4), (5), and (6)), giving rise to a positive kinetic helicity (see the vortex picture associated with negative $`dV/dx`$ in Figure 1). The only physical difference of vortex dynamics between a non-rotating shear flow and a weakly magnetized disk is that in the later case, a positive kinetic helicity coupled with a positive vorticity of the mean flow becomes an energy sink of $`\omega _z^2`$ as shown in equation (16). This is analogous to the result that a positive Reynolds stress is an energy sink of $`v_\theta ^2`$ in Keplerian disks (Hawley, Balbus, & Winters (1999)).
Fig. 1. The expansion of a rising turbulent element described by equation (16) or by equation (6) when $`\kappa ^2>0`$ or $`dV/dx>0`$ (left), and when $`\kappa ^2<0`$ or $`dV/dx<0`$ (right). As observed in a corotating frame, $`\omega _z(_rv_r+_\theta v_\theta /r)<0`$ and $`v_z\omega _z<0`$ when $`\kappa ^2>0`$. These signs are reversed when $`\kappa ^2<0`$. A negative (positive) $`\omega _z`$, as a result of a positive (negative) vorticity of the mean flow, can give rise to a negative (positive) $`v_rv_\theta `$ when radial deviation of azimuthal motion is small. As argued in the text, the picture on the right could also apply to the $`\omega _z`$ produced by the Balbus-Hawley instability which generates positive Reynolds stresses in Keplerian disks. In the magnetic case, however, the vortex motion illustrated in the figure becomes an energy sink of $`\omega _z^2`$. $`k_\theta <k_r`$ is a result of shear.
Although our analysis is related to convection-like eddies (i.e., $`k_z1/h`$), the picture we present here for estimating the sign of kinetic helicity does not contradict the result that the Parker instability, a magnetohydrodynamic instability with $`k_z1/h`$, is suppressed by MHD turbulence (Vishniac & Diamond (1992)). The main point presented here depends solely on the sign of Reynolds stresses which is assumed to be largely caused by the azimuthal deviation of radial motion. Therefore, the picture that we sketch here applies to any turbulent eddies which expand (contract) as they rise (descent). We note that rising turbulent eddies can expand statistically even without any types of buoyancy. As a matter of fact, the vertical structures of large-scale magnetic fields generated by the Balbus-Hawley instability in vertically stratified accretion disks are stable against magnetic buoyancy (Stone et al. (1996); Miller & Stone (2000)). The detailed information, such as the magnitude of this kinetic helicity and the expansion of turbulent eddies in a stably stratified disk, should depend strongly on the detailed structure of the flow, such as azimuthal perturbations (i.e., anisotropic turbulence) and correct vertical dynamics.
## 5. Discussion and Outlook: comments on angular momentum transport by strong convection
In this paper, we demonstrate that weak convection cannot be self-sustained by vortex stretching and tilting in a Keplerian flow, suggested by the equations of enstrophy fluctuations. This is because epicyclic term is an energy sink of $`\omega _r^2`$ when $`v_rv_\theta >0`$. The situation is reversed when weak convection has a negative Reynolds stress, leading to a successful growth of thermal convection. Our determination of the signs of correlation quantities in the equations of enstrophy is based on the point of view in which the signs of Reynolds stresses are determined by conservation of angular momentum for weak convection (Balbus (2000)). Although our determination of signs of correlation quantities is not a rigorous proof but quite suggestive, the results are consistent with those suggested by the equations of velocity fluctuations. Full three-dimensional numerical simulations are needed to verify our thought.
Positive kinetic helicity in a weakly magnetized, vertically stratified Keplerian disk implied from our analysis agrees with the simulation result by Brandenburg & Donner (1997). However, the physical picture shown in Figure 1 is totally different from the plot sketched in Figure 7 in their paper: in a corotating frame of a Keplerian disk, the direction of Coriolis effect ($`2\mathrm{\Omega }`$) should be opposite to the direction of shear effect ($`d\mathrm{\Omega }/dr`$), resulting in a positive vorticity ($`\kappa ^2>0`$). In our analysis for turbulence, $`w_z(_rv_r+_\theta v_\theta /r)<0`$ ($`>0`$) when vorticity is positive (negative), giving rise to negative (positive) $`v_z\omega _z`$ as long as $`v_z(_rv_r+_\theta v_\theta /r)`$ is positive. Since the Balbus-Hawley instability generates a positive $`v_rv_\theta `$ which resembles the hydrodynamic turbulence with a negative vorticity, MHD turbulence driven by the Balbus-Hawley instability should display right-handed helical motion as observed in numerical simulations. However, if rising eddies contract statistically, our analysis will give rise to the conventional sign of kinetic helicities in a rotating fluid influenced mostly by the Coriolis force. The expansion of rising turbulent elements in a stably stratified disk could be just a result of a special profile of underlying background states, or might result from the possible scenario that turbulent mixing, an effect definitely not included in usual laminar analyses, alters the condition of flux freezing in some degree. On the contrary, the simple explanation by Brandenburg (1998) relying on magnetic buoyancy is not consistent with the simulations for vertically stratified accretion disks stirred by the Balbus-Hawley instability.
GVC studied convective energy transport in Keplerian disks stirred by the Balbus-Hawley instability. The typical modes of convection are determined from linear perturbation theory and nonlinear saturation which is caused by the secondary Kelvin-Helmholtz instability (i.e., interaction between convective shear and background vorticity) and by MHD turbulent damping. From equations (14), (15), and (16), we have seen how convective vorticity/shear possibly interacts with background vorticity/shear and how this interaction becomes an energy sink of turbulent vorticity. This means that the quasi-linear approach for the mixing-length theory by GVC based upon the secondary Kelvin-Helmholtz instability is qualitatively correct.
However, the issue of angular momentum transport by weak convection is probably not of great importance in reality. According to the thermal-viscous instability model, weak convection occurs in Keplerian disks when the Shakura-Sunyaev viscosity $`\alpha _{SS}`$ (Shakura & Sunyaev (1973)) is large $`0.1`$, as a result of strong radiative losses (GVC; Cabot (1996)). Furthermore, based on GVC, large $`\alpha _{SS}`$ means strong MHD turbulent mixing driven by the Balbus-Hawley instability, which suppresses weak convection by smoothing away momentum and entropy anisotropies associated with convective bubbles mostly in the radial direction (see equation (19) below): $`A<<1`$ when $`N^2<<\mathrm{\Omega }`$). Consequently, only the convection with larger Rossby numbers $`Ro1`$ can survive in a magnetized Keplerian disk. Weak convection can occur at smaller radii in protostellar disks (e.g., see D’Alessio et al. (1998)). In such a cold disk, a significant ‘dead zone’ implies that the Balbus-Hawley instability may not function properly, but accretion might occur near the surfaces of disks where cosmic rays (Gammie (1996)), irradiation (Glassgold et al. (1997)), or rigorous heating beyond the bottom of photosphere as manifested by protostellar winds and flares might increase the ionization level beyond the critical values (Gammie and Menou (1998)). However, the heat due to layer accretion should reduce the temperature gradient of the ‘dead zone’ in the vertical direction, possibly leading to the suppression of weak convection as a result of enhanced radiative losses in the radial direction (GVC). Therefore, the question becomes, does strong convection transport angular momentum inward or outward? Low Reynolds number simulation by Cabot & Pollack (1992) found strong convection transports angular momentum outward.
Azimuthal pressure perturbations have drawn large attention in the literature recently, as suggested by the equations of mean-square velocity perturbation and by, in this paper, the equations of enstrophy. Axisymmetry/nonaxisymmetry of convection is actually an intimate issue of the damping of convection due to MHD turbulent mixing (Klahr, Henning, & Kley (1999)). When MHD turbulent mixing is large (i.e., large $`\alpha _{SS}`$), modes with large $`k_r`$ are all suppressed owing to strong radial mixing. Similarly, modes with large $`k_r`$ should also succumb to strong radiative losses in radial direction. Modes with small $`k_r`$ can survive in a shearing environment only if $`k_\theta `$ is also very small, suggesting nearly axisymmetrical convection. As mentioned in the preceding paragraph, weak convection occurs when $`\alpha _{SS}`$ is large. Nearly axisymmetrical convection is therefore a manifestation of weak convection which struggles with turbulent and radiative damping in a differentially rotating flow. In other words, nonaxisymmetry of strong ($`Ro1`$) convection means that azimuthal pressure perturbations could be important to violation of conservation of angular momentum. If this happens, the equations of velocity and vorticity fluctuations could allow strong convection to transport angular momentum radially outward. Without a doubt, the quasi-linear analysis based on equations of velocity and vorticity fluctuations is very suggestive.
The other point of view regarding the nature of positive/negative viscosity is the direction of turbulent energy cascade. Negative eddy viscosity manifests the process in which turbulent fields do not only extract energy from the mean flow, but the extracted energy is also passed up to larger scales and finally goes back to the mean flow (Starr (1968)). The connection between negative viscosity and inverse energy cascade in two-dimensional flows has been studied in fluid society (Kraichnan (1976); Pouquet (1978); Chechkin et al. (1999)). Cabot 1996 and Klahr et al. suggested that negative viscosity is a result of an inverse energy cascade of nearly axisymmetrical convection and that positive viscosity results from nonaxisymmetrical convection in which 3D hydrodynamical turbulence is resumed. In fact, the collapse of 3D hydrodynamical turbulence to 2D is not necessarily related only to axisymmetrical patterns, but is a natural result of turbulent anisotropy due to strong body forces, which has been manifested by ample examples in geophysics. Hossain 1994 found that when a strong rigid rotation ($`Ro<1`$) is turned on, the turbulent velocity fields perpendicular to the direction of rotation (i.e. $`xy`$ plane) are strongly correlated along the direction of rotation (i.e. $`z`$ direction). Therefore 3D turbulence (direct cascade) reduces to an approximate 2D state (inverse cascade). In the case of accretion disks, epicyclic effect should enforce turbulent fields of weak convection to have an approximate 2D state, giving rise to negative viscosity due to an inverse energy cascade. If convection is strong, however, the collapse to 2D breaks down and this probably results in positive/zero eddy viscosity due to a direct energy cascade.
The numerical simulation by Klahr (2000) found that the gradient of fluctuating angular momentum generated by weak convection could flatten background gradient of angular momentum. Consequently, a secondary instability occurs at a fast rate of some fraction of $`\mathrm{\Omega }`$. This induced instability overwhelms weak convection and drives angular momentum outward. In terms of the quasi-linear analysis based on velocity and vorticity fluctuations, outward angular momentum transport by this secondary instability is a result of strong azimuthal pressure perturbations and stochastic stretching of vortex by turbulent shear. By virtue of nonlinear phenomena of turbulent cascade, this fast secondary instability diminishes epicyclic effect and thereby forces fluid back to an approximate 3-D hydrodynamical state.
If thermal convection generates positive eddy viscosity in Keplerian disks, it does not necessarily mean that it can be self-sustained. It has been a concern that a self-sustained convection would violate the second law of thermodynamics: in the case of a system in which the eddies are thermally driven, the heat dissipated from the mean flow cannot again be used to drive the eddies (Starr (1968)). If thermal convection can be self-maintained in Keplerian disks, this could mean that convection is not totally thermally driven, in the sense that it can access the available rotational energy at smaller radii via secondary or nonlinear instabilities (Kumer et al. (1995); Richard & Zahn (1999)). However, the reverse of the above statement is not necessarily true. Klahr (2000) shows that circumstellar disks with outward transport of angular momentum by convection cool continuously throughout the simulation.
Direction of angular momentum transport by convection is also an important issue in advection-dominated accretion flows (ADAFs). When $`\alpha _{SS}`$ (not due to convection) is small, strong convection ($`Ro1`$) occurs when convection transports angular momentum inward (or outward less efficiently than it transports energy), and accretion is suppressed in ADAFs (Narayan, Igumenshchev, & Abramowicz (2000); Quataert & Gruzinov (2000)). However, as noted in this section, strong convection is unlikely to be nearly axisymmetrical and is doubtful to be able to collapse to a 2D state especially in a thick disk where the direction of strong entropy gradient has a component along angular momentum gradient. Numerical observations of a transition of direction of angular momentum (if it exists), a transition of direction of energy cascade, and a transition of pattern formation for convection in both Keplerian accretion disks and ADAFs is worthwhile when $`Ro`$ and $`Ra`$ are the control parameters, where $`Ra`$ should be the “turbulent” Rayleigh number characterized by radiative/advection losses and MHD turbulent damping.
While the perturbed baroclinic terms in the equations of perturbed enstrophy act as the energy source of fluctuating vorticities, we do not consider the terms associated with the background baroclinic term in this paper, due to the thin-disk approximation. That is, $`\mathrm{\Omega }`$ does not vary with $`z`$ dramatically in thin disks, resulting in a barotropic and Keplerian disk (e.g., see Frank, King, & Raine (1992)). The background baroclinic effect is usually of less importance on local faster turbulence in disks; for instance, the Biermann battery owing to the same mechanism is usually ignored when a mean-field dynamo is able to operate via mean kinetic helicities driven by local turbulent events such as thermal convection in stars or supernovae/stellar winds in galactic disks. In spite of their appearance in thick disks which behave as fast rotating stars, it is also quite doubtful that the background baroclinic term can have important influence on the onset of convective instability in these cases. As long as the convective growth rate is not too small, the effect of large scale circulation due to the baroclinic term to smooth away entropy inhomogeneity associated with small-scale mixing such as thermal convection in this case is usually negligible, compared with the same damping mechanism achieved by MHD turbulence. On the other hand, as we mention in the end of §2, a large-scale vortex can be initiated from the background baroclinic effect. Large-scale eddies could be also formed via merging convective vortexes (Zhong et al. (1993); Godon & Livio (1999)) or through helicity fluctuations of convection (Branover et al. (1999)). The nonlinear interaction between large-scale vortexes and convective vorticities is worth investigating.
Finally we note that our analysis in this paper is applied to local turbulent mixing in accretion disks, such as thermal convection and the Balbus-Hawley instability. Turbulent transport mediated by global waves, such as internal, Rossby, or spiral shock waves, is not subject to our analysis. Moreover, convective turbulence might behave differently in the environment of MHD turbulence since ideal invariants during turbulent cascades are different between hydrodynamical and MHD turbulence (e.g., see Biskamp (1997)).
We are deeply indebted to Ethan T. Vishniac for many useful discussions, and for providing the information about Klahr 2000. We are also grateful to the anonymous referee for insightful comments and suggestions. This work was completed in the High Energy Physics Laboratory at The University of Texas at Austin, and we would like to thank Roy Schwitters for his generous hospitality. |
no-problem/0001/cond-mat0001221.html | ar5iv | text | # Orbital Structure and Magnetic Ordering in Layered Manganites: Universal Correlation and Its Mechanism
\[
## Abstract
Correlation between orbital structure and magnetic ordering in bilayered manganites is examined. A level separation between the $`3d_{3z^2r^2}`$ and $`3d_{x^2y^2}`$ orbitals in a Mn ion is calculated in the ionic model for a large number of the compounds. It is found that the relative stability of the orbitals dominates the magnetic transition temperatures as well as the magnetic structures. A mechanism of the correlation between orbital and magnetism is investigated based on the theoretical model with the two $`e_g`$ orbitals under strong electron correlation.
\]
Since the discovery of the colossal magnetoresistance (CMR), studies of manganites with cubic perovskite structure have been renewed theoretically and experimentally. Competition and cooperation between spin, charge and orbital degrees of freedom as well as lattice cause the dramatic changes of transport and magnetic properties. Manganites with bilayered structure $`A_{22x}B_{1+2x}`$Mn<sub>2</sub>O<sub>7</sub>, where $`A`$ and $`B`$ are trivalent and divalent cations, respectively, are another class of CMR materials . Since an extremely large MR is observed near the transition from paramagnetic (PM) insulator to ferromagnetic (FM) metal, it has been considered that several concepts proposed in the cubic compounds are applicable to the bilayered ones.
In cubic manganites, one of the key factors dominating the magnetic orderings is the tolerance factor ; a bending of a Mn-O-Mn bond decreases the hopping integral of carriers. As a result, the ferromagnetic transition temperature $`T_c`$ decreases in the double exchange (DE) scenario. However, in bilayered manganites, the Mn-O-Mn bond angle is almost unchanged with changing cations and carrier concentration, as shown later, in spite of a wide variety of the magnetic structures. Various key factors dominating the magnetic ordering, which are not included in the DE model, were experimentally suggested, e.g. the antiferromagnetic (AFM) superexchange (SE) interaction , the local lattice distortion , the charge and orbital degrees of freedom and their orderings and so on. However, systematics in their correlations for a variety of compounds and their mechanisms still remain to be clarified.
In this letter, we study the correlation between magnetic ordering and orbital structure in bilayered manganites. The two $`e_g`$ orbitals, i.e. the $`3d_{3z^2r^2}`$ and $`3d_{x^2y^2}`$ orbitals in a Mn<sup>3+</sup> ion split in the crystalline field of the bilayered structure and one of them is occupied by an electron. It is known that the occupied orbital controls the anisotropy of the magnetic interaction as well as its strength. The level separation between the orbitals is calculated in the ionic model for a large number of the compounds. We find a universal correlation between the relative stability of the orbitals and the magnetic transition temperatures as well as the magnetic structures. A mechanism of the correlation is investigated based on the theoretical model with the $`e_g`$ orbitals under strong electron correlation.
We first show that neither the tolerance factor nor the bond length governs $`T_C`$ and the N$`\stackrel{´}{\mathrm{e}}`$el temperature $`T_N`$ for the A-type AFM ordering . The tolerance factor in the bilayered crystals is defined by $`t=(d_{\mathrm{O}(1)\mathrm{A}(1)}+d_{\mathrm{O}(2)\mathrm{A}(2)})/(2\sqrt{2}d_{\mathrm{Mn}\mathrm{O}(3)})`$ where $`d_{\mathrm{A}\mathrm{B}}`$ is a bond length between A and B ions. The position of each ion is shown in the inset of Fig. 1(a). Being based on the structural data obtained by the neutron and x-ray diffraction experiments , we evaluate $`t`$ and the bond length between nearest neighboring (NN) Mn ions in the $`ab`$ plane $`d_{\mathrm{Mn}\mathrm{Mn}}^{ab}`$ for a variety of compounds. $`T_C`$ and $`T_N`$ are plotted as functions of $`t`$ and $`d_{\mathrm{Mn}\mathrm{Mn}}^{ab}`$ in Figs. 1 (a) and (b), respectively. Almost all $`t`$’s are located in a narrow region where $`T_C`$’s and $`T_N`$’s are distributed randomly. In addition, $`T_C`$ is not correlated with $`d_{\mathrm{Mn}\mathrm{Mn}}^{ab}`$, either. Although $`T_N`$ increases with increasing $`d_{\mathrm{Mn}\mathrm{Mn}}^{ab}`$, this correlation is opposite to that predicted by the DE scenario where, with increasing the bond length, the hopping integral decreases and the FM interaction in the $`ab`$ plane decreases. We conclude that the DE model, which includes the change of the hopping integral caused by the change of the bond angle/length, can not explain $`T_{C(N)}`$. We also examined correlations between $`T_{C(N)}`$ and a number of other quantities: the torelance factor evaluated by the ionic radius, a Mn-O(3)-Mn bond angle, a Mn-O(1)-Mn bond length, Mn-O bond lengths, a lattice spacing between NN bilayers, lattice constants and the valence-bond sum for a Mn ion. However, there are not clear correlations between these parameters and $`T_{C(N)}`$.
Let us focus on the correlation between $`T_{C(N)}`$ and a relative stability of the $`e_g`$ orbitals. We employ the ionic model to examine the electronic energy-level structures. This model may be justified by the following considerations : (1) the manganites at $`x=0`$ are classified as charge-transfer type insulators in which the ionic model provides a good starting point . (2) The ionic property is predominant between bilayers. (3) The energy-level structure given by band-structure calculations shows the same tendency with those by the ionic model . The energy levels of the $`e_g`$ orbitals split due to the electrostatic potential and one of the orbitals is occupied by an electron in a Mn<sup>3+</sup> ion.
By using a large number of the structural data , we calculate the Madelung potential for a hole in the $`3d_{3z^2r^2}`$ and $`3d_{x^2y^2}`$ orbitals at site $`j`$ defined by
$$V_{3z^2r^2}=\frac{1}{2}\left\{V(\stackrel{}{r}_j+r_d\widehat{z})+V(\stackrel{}{r}_jr_d\widehat{z})\right\},$$
(1)
and
$$V_{x^2y^2}=V(\stackrel{}{r}_j+r_d\widehat{x}),$$
(2)
respectively . Here, $`V(\stackrel{}{r}_j)`$ is given by
$$V(\stackrel{}{r}_j)=\underset{ij}{}\frac{Z_ie^2}{|\stackrel{}{r}_j\stackrel{}{r}_i|},$$
(3)
with a point charge $`Z_ie`$ at site $`i`$ and the position $`\stackrel{}{r}_i`$ of the site. $`r_d`$(=0.42Å) is the radius of a Mn $`3d`$ orbital where its radial charge density becomes maximum and $`\widehat{z}(\widehat{x})`$ is the unit vector in the $`z(x)`$ axis. The Ewald method is used for the lattice summation. $`Z_i`$’s for Mn and O ions and a cation at $`A`$ site are chosen to be $`3+x`$, $`2`$ and $`(82x)/3`$, respectively. The difference of the potentials
$$\mathrm{\Delta }V=V_{3z^2r^2}V_{x^2y^2},$$
(4)
represents the relative stability of the orbitals; with increasing $`\mathrm{\Delta }V`$, the energy level of the $`3d_{3z^2r^2}`$ orbital for an electron relatively decreases.
$`T_N`$ and $`T_C`$ are plotted as functions of $`\mathrm{\Delta }V`$ in Fig. 2 where the structural data at room temperature are used. Broad shades are drawn by considering experimental errors. It is clearly shown that both $`T_C`$ and $`T_N`$ are correlated with $`\mathrm{\Delta }V`$; $`T_N`$ increases with decreasing $`\mathrm{\Delta }V`$ and there is an optimal value of $`\mathrm{\Delta }V`$($`0.08`$ eV) for $`T_C`$. We estimate the strength of the correlation between $`T_N`$ and $`\mathrm{\Delta }V`$ by using the correlation coefficient: $`r=\frac{1}{N}_l`$ $`(T_{Nl}\overline{T_N})(\mathrm{\Delta }V_l\overline{\mathrm{\Delta }V})/(\sigma _{T_N}\sigma _{\mathrm{\Delta }V})`$ where $`l`$ indicates a sample and $`N`$ is the number of samples. $`\overline{\mathrm{\Delta }V}`$ ($`\overline{T_N}`$) and $`\sigma _{\mathrm{\Delta }V}`$ ($`\sigma _{T_N}`$) are the mean value and the standard deviation of $`\mathrm{\Delta }V`$ ($`T_N`$), respectively. We obtain $`r=0.89\pm 0.11`$ for single crystal samples and $`r=0.15\pm 0.04`$ for all samples including polycrystals. One might think that the $`T_{C(N)}`$ v.s. $`\mathrm{\Delta }V`$ curve in Fig. 2 just reflects the relation between $`T_{C(N)}`$ and $`x`$ in La<sub>2-2x</sub>Sr<sub>1+2x</sub>Mn<sub>2</sub>O<sub>7</sub> (LSMO). However, when we pay attention to $`T_{C(N)}`$’s for samples with the same $`x`$ (e.g., the samples C and K-R), we notice that the correlation remains. The correlation between $`T_N`$ and $`\mathrm{\Delta }V`$ explains that between $`T_N`$ and $`d_{\mathrm{Mn}\mathrm{Mn}}^{ab}`$ shown in Fig. 1(b), since $`\mathrm{\Delta }V`$ is a decreasing function of $`d_{\mathrm{Mn}\mathrm{Mn}}^{ab}`$ in the region, 3.84$`\mathrm{\AA }<d_{\mathrm{Mn}\mathrm{Mn}}^{ab}<`$3.88$`\mathrm{\AA }`$.
In Fig. 3, we present the magnetic phase diagram at $`T=0`$ as a function of $`\mathrm{\Delta }V`$ and $`x`$. The structural data at room temperature are used. Symbols connected by dotted lines correspond to a series of LSMO with $`x=0.30.5`$ . In addition, $`\mathrm{\Delta }V`$’s calculated by using the data below $`T_{C(N)}`$ in LSMO are also plotted. Bold arrows indicate the change of the Madelung potential with changing temperature from $`T>T_{C(N)}`$ to $`T<T_{C(N)}`$:
$$\delta (\mathrm{\Delta }V)\mathrm{\Delta }V(T<T_{C(N)})\mathrm{\Delta }V(T>T_{C(N)}).$$
(5)
We find that the magnetic structures are governed by $`\mathrm{\Delta }V`$ and $`x`$; the FM (A-type AFM) phase is located in the region with smaller (larger) $`x`$ and moderate (smaller) $`\mathrm{\Delta }V`$. Let us focus on $`\delta (\mathrm{\Delta }V)`$ in LSMO. $`\delta (\mathrm{\Delta }V)`$’s are negative at $`x=0.3`$ and 0.35. The absolute value of $`\delta (\mathrm{\Delta }V)`$ gradually decreases with increasing $`x`$ and $`\delta (\mathrm{\Delta }V)`$ becomes a small positive value at $`x=0.4`$. Below $`T_C`$, $`\mathrm{\Delta }V`$ seems to approach to the optimal value of $`\mathrm{\Delta }V0.08`$ where $`T_C`$ becomes maximum as seen in Fig. 2. On the other hand, $`\delta (\mathrm{\Delta }V)`$’s are negative at $`x`$=0.45 and 0.48 where the A-type AFM structure appears. The orbital structure and its stability in the FM phase have been studied by measuring the striction in Ref. . The difference of the Mn-O bond lengths between PM and FM states was reported in Ref. . These experimental results are consistent with the present results of $`\delta (\mathrm{\Delta }V)`$ in Fig. 3.
Now we theoretically investigate a mechanism of the correlation between magnetic ordering and orbital structure. We start with the following Hamiltonian :
$$=_t+_J+_H+_{AF}+_z.$$
(6)
Instead of the bilayered structure, the simple tetragonal lattice consisting of Mn ions is considered. In this model, the magnetic structure with FM and AFM alignments perpendicular and parallel to the $`c`$ axis, respectively, corresponds to the A-type AFM structure . In each Mn ion, the two $`e_g`$ orbitals are introduced and the $`t_{2g}`$ electrons are treated as a localized spin with $`S=3/2`$. The first two terms in Eq. (6) correspond to the so-called $`t`$\- and $`J`$-terms in the $`tJ`$ model, respectively, with the two $`e_g`$ orbitals under strong electron correlation. The third and fourth terms describe the Hund coupling between $`e_g`$ and $`t_{2g}`$ spins and the AFM SE interaction between $`t_{2g}`$ spins, respectively. The splitting of the energy levels between $`3d_{3z^2r^2}`$ and $`3d_{x^2y^2}`$ orbitals is represented by the last term: $`_z=\mathrm{\Delta }_iT_{iz}`$ where the pseudospin operator is given by $`\stackrel{}{T}_i=\frac{1}{2}_{\gamma \gamma ^{}\sigma }d_{i\gamma \sigma }^{}\stackrel{}{\sigma }_{\gamma \gamma ^{}}d_{i\gamma ^{}\sigma }`$ with $`d_{i\gamma \sigma }`$ being the annihilation operator of an $`e_g`$ electron at site $`i`$ with spin $`\sigma `$ and orbital $`\gamma `$. The $`+()`$ eigenstate of $`T_{iz}`$ corresponds to the state where the $`3d_{3z^2r^2}`$ ($`3d_{x^2y^2}`$) orbital is occupied by an electron. The anisotropies of the hopping integral and the SE interactions due to the layered structure are considered. The explicit expression and derivation of the Hamiltonian are presented in Refs. and .
The calculated magnetic phase diagram at $`T=0`$ is presented in Fig. 4 where the mean field approximation is adopted. We note that the phase diagram derived in this approximation explains that in the cubic manganites . The characteristic features shown in Fig. 3 are well reproduced by the present theory; the A-type AFM phase appears in the region with higher $`x`$ and smaller $`\mathrm{\Delta }`$ than that of the FM one. The range of the horizontal axis in Fig. 4 is larger than that in Fig. 3 by about $`0.25`$. This discrepancy may be attributed to the neglect of the orbital fluctuation . However, the characteristics of the phase diagram are insensitive to the parameters in the model. In the FM (A-type AFM) phase, the orbitals are uniformly aligned with $`0<\theta <0.72\pi `$ ($`0.72\pi <\theta <\pi `$) where $`\theta `$ describes the orbital state as $`|\theta =`$$`\mathrm{cos}(\frac{\theta }{2})|3d_{3z^2r^2}`$$`\mathrm{sin}(\frac{\theta }{2})|3d_{x^2y^2}`$. The present results suggest that a dimensionality of the FM interaction is controlled by the orbital structure; in the A-type AFM phase, the FM ordering in the $`ab`$ plane is caused by the DE interaction, while the AFM in the $`c`$ direction is by the AFM SE. When the $`3d_{x^2y^2}`$ orbital is stabilized, the DE interaction in the $`ab`$ plane ($`c`$ direction) becomes strong (weak) and the A-type AFM phase appears . A mixing of the orbitals is essential in the FM phase where the FM interaction overcomes the AFM SE one in the three directions. We note that, in Fig. 4, the FM phase appears not around $`\mathrm{\Delta }=0`$ but in a region of $`\mathrm{\Delta }>0`$, since the anisotropy in the hopping integral due to the layered structure stabilizes the $`3d_{x^2y^2}`$ orbital more than $`3d_{3z^2r^2}`$. It is worth to mention the change of the orbital structure associated with the magnetic ordering: By utilizing the mean filed approximation at finite temperature, we compare the orbital structures above and below the magnetic transition temperatures. It is found that (1) there is an optimal mixing of the orbitals for the FM state and the orbital structure tends to approach to this structure below $`T_C`$ and (2) the $`3d_{x^2y^2}`$ orbital structure is stabilized below $`T_N`$. The theoretical results are consistent with $`\delta (\mathrm{\Delta }V)`$’s shown in Fig. 3 by considering that the change of $`\mathrm{\Delta }V`$ associated with the magnetic ordering is caused by that of the orbital structure.
In summary, we examine correlation between magnetic ordering and orbital structure in bilayered manganites. A relative stability of the $`e_g`$ orbitals is investigated by calculating the Madelung potentials in a large number of the bilayered compounds. We find that the A-type AFM structure and the $`3d_{x^2y^2}`$ orbital one are stabilized cooperatively and there is an optimal mixing between the $`3d_{3z^2r^2}`$ and $`3d_{x^2y^2}`$ orbitals for the FM ordering. A theory with the two $`e_g`$ orbitals under strong electron correlation explains a mechanism of the universal correlation between orbital and magnetism.
The authors would like to thank Y. Moritomo, T. Akimoto, Y. Tokura, T. Kimura, Y. Endoh, K. Hirota, M. Kubota and G. Khaliullin for their valuable discussions. This work was supported by CREST, NEDO and Grant-in-Aid for Scientific Research Priority Area from the Ministry of Education, Science and Culture of Japan. S. O. acknowledges the financial support of JSPS. Part of the numerical calculation was performed in the HITACS-3800/380 superconputing facilities in IMR, Tohoku Univ. |
no-problem/0001/hep-lat0001003.html | ar5iv | text | # A Chiral Phase Transition using a Fermion Cluster Algorithm
## I Introduction
The past few years has been a time of remarkable progress in our understanding of lattice fermions and their ability to reproduce the chiral properties of the continuum theory . It is now possible to formulate vector-like gauge theories on the lattice with all its chiral symmetry intact . Our understanding of non-perturbative formulations of chiral gauge theories is also advancing at a remarkable rate as is reflected by many recent contributions . On the other hand, our ability to perform reliable (numerical) calculations in many theories involving strongly interacting fermions and especially in the chiral limit has remained primitive. In particular algorithms based on hybrid molecular dynamics, that are popular for gauge theories like QCD, suffer either from sign problems or critical slowing down in the chiral limit. Unless better numerical methods are found, it is likely that we will be forced to perform most of our calculations far from the chiral limit and hence not gain from the progress in our understanding of chiral symmetry.
Numerical methods for fermionic systems are especially difficult due to the Pauli-exclusion principle. The Boltzmann weight of the statistical mechanics problem describing such systems can often become negative thereby invalidating most Monte-Carlo methods. This is usually referred to as the sign problem. In other cases one is often forced to work with bosonic variables with non-local Boltzmann weights which arise through the fermion determinant. Such expressions are not easy to handle and there is no known way to overcome problems like critical slowing.
Recently a new numerical method to perform fermionic path integrals has emerged and is based on a rather novel form of bosonization. Instead of the Grassmann variables, the method starts by representing fermions through their occupation numbers. Such representations, which have been studied in the past , were not considered useful due to sign problems. Over the past year it has become clear that at least in a class of models it is possible to design cluster algorithms which automatically solve the fermion sign problem and provide an alternative way to compute fermionic path integrals. In a sense the algorithm suggests a new way to bosonize the model in terms of dynamics of clusters. The new method works in both relativistic and non-relativistic models. Since cluster algorithms are known to beat critical slowing down in a variety of bosonic models, the new algorithms have the ability to work efficiently even in the presence of long correlation lengths. In particular, they encounter no difficulties in the limit where the fermions become massless.
The present article illustrates the main features of the new approach through an example of a four Fermi model involving staggered fermions. This model undergoes a $`Z_2`$ chiral transition from a high temperature chirally symmetric phase to the broken phase at low temperatures and was studied originally in the chiral limit in . Here the earlier work is extended to include a fermion mass. This extension turns out to be quite useful because on a large but finite lattice all the interesting properties of the chiral condensate, as an order parameter of the transition, emerge only when a tiny mass can be added to the system. In the next section the partition function of the model is written in the occupation number basis and the origin of the fermion sign problem is reviewed. In section 3 the partition function is rewritten in terms of connected fermion world-line configurations that arise due to the introduction of local “bond” variables. In the new variables the solution to the sign problem emerges naturally and leads to a very elegant cluster algorithm. In section 4, the algorithm and some numerical results are discussed. Section 5 contains conclusions along with a description of on-going work and points to new directions for the future.
## II The Model
The model considered in this article involves interacting staggered fermions hopping on a 3-d cubic spatial lattice with $`V=L^3`$ sites $`x`$ ($`L`$ even) and with anti-periodic spatial boundary conditions. The dynamics of the fermions is described through the Hamilton operator
$$H=\underset{x,i}{}h_{x,i}+m\underset{x}{}p_x$$
(1)
where the term
$$h_{x,i}=\frac{1}{2}\eta _{x,i}(\mathrm{\Psi }_x^+\mathrm{\Psi }_{x+\widehat{i}}+\mathrm{\Psi }_{x+\widehat{i}}^+\mathrm{\Psi }_x)+(\mathrm{\Psi }_x^+\mathrm{\Psi }_x\frac{1}{2})(\mathrm{\Psi }_{x+\widehat{i}}^+\mathrm{\Psi }_{x+\widehat{i}}\frac{1}{2}),$$
(2)
couples the fermion operators at the lattice sites $`x`$ and $`x+\widehat{i}`$, where $`\widehat{i}`$ is a unit-vector in the $`i`$-direction and the mass term
$$p_x=(1)^{x_1+x_2+x_3}(\mathrm{\Psi }_x^+\mathrm{\Psi }_x1/2),$$
(3)
is a single site operator. The fermion creation and annihilation operators $`\mathrm{\Psi }_x^+`$ and $`\mathrm{\Psi }_x`$ used in the above equations satisfy the standard anti-commutation relations
$$\{\mathrm{\Psi }_x^+,\mathrm{\Psi }_y^+\}=\{\mathrm{\Psi }_x,\mathrm{\Psi }_y\}=0,\{\mathrm{\Psi }_x^+,\mathrm{\Psi }_y\}=\delta _{xy}.$$
(4)
Further $`\eta _{x,1}=1`$, $`\eta _{x,2}=(1)^{x_1}`$ and $`\eta _{x,3}=(1)^{x_1+x_2}`$ are the standard staggered fermion sign factors. In the chiral limit ($`m=0`$), the above Hamiltonian was discussed in .
In addition to the usual $`U(1)`$ fermion number symmetry, there are a number of discrete symmetries of the staggered fermion Hamiltonian as discussed in . For example, at $`m=0`$ the Hamiltonian is invariant under shifts of $`\mathrm{\Psi }_x^+`$ and $`\mathrm{\Psi }_x`$ by one lattice spacing in the $`x_3`$ direction. The mass term breaks this $`𝐙_2`$ symmetry which is known to be a subgroup of the well known chiral symmetry of relativistic massless fermions. Here this symmetry is broken spontaneously at low temperatures while thermal fluctuations restore it at high temperatures. The associated second order phase transition belongs to the universality class of the 3d ising model as discussed in . The extension discussed here, allows one to add a tiny mass and hence study the transition using the chiral condensate.
In order to construct the path integral representation of the partition function the Hamilton operator is first decomposed into seven terms $`H=H_1+H_2+\mathrm{}+H_6+H_7`$ with
$$H_i=\underset{\stackrel{x=(x_1,x_2,x_3)}{x_ieven}}{}h_{x,i}\text{and}H_{i+3}=\underset{\stackrel{x=(x_1,x_2,x_3)}{x_iodd}}{}h_{x,i}\text{for }i=1,2,3;H_7=m\underset{x}{}p_x.$$
(5)
This decomposition is exactly the same as the one used in for the $`m=0`$ case, where $`H_7`$ was absent. The various steps in the construction of the partition function from here on is exactly the same as in . Leaving the details to that paper only the essential points are sketched here. The Suzuki-Trotter formula leads to the expression
$$Z_f=\text{Tr}[\mathrm{exp}(\beta H)]=\underset{M\mathrm{}}{lim}\text{Tr}[\mathrm{e}^{ϵH_1}\mathrm{e}^{ϵH_7/6}\mathrm{e}^{ϵH_2}\mathrm{e}^{ϵH_7/6}\mathrm{}\mathrm{e}^{ϵH_6}\mathrm{e}^{ϵH_7/6}]^M,$$
(6)
for the fermionic partition function at inverse temperature $`\beta `$, where $`ϵ=\beta /M`$ is the lattice spacing in the Euclidean time direction. Spreading $`H_7`$ symmetrically with each $`H_i,i=1,\mathrm{..6}`$ is not necessary except that it adds some symmetry to the algorithm. Further, in the arguments given below, $`M`$ is assumed to be finite and that the true partition function is obtained from an extrapolation of the data obtained from a series of simulations at larger and larger $`M`$. Fortunately it is also possible to formulate the cluster algorithm in the time continuum limit as demonstrated in .
It is well known that in the occupation number basis a fermionic partition function can be written as a sum over fermion world-line configurations. Since the fermionic creation operators anti-commute, the information that a subset of the lattice sites are occupied defines the state of the system only up to a sign. A complete definition of the occupation number basis states requires an ordering of the entire lattice which then specifies the order in which the fermions on the lattice are created. Taking this ordering into account it is possible to write the partition function as a sum over configurations of fermion occupation numbers $`n(x,t)=0,1`$ on a $`(3+1)`$-d space-time lattice of points $`(x,t)`$ where $`t=ϵn/12,n=0,1,2,\mathrm{}(12M1)`$ labels the time slices. The trace imposes the periodicity constraint $`n(x,0)=n(x,\beta )`$. A typical configuration in one spatial dimension is shown in figure 1, where the shaded regions depict either a two site interaction due to $`\mathrm{exp}(h_{x,i})`$ or a single site interaction due to $`\mathrm{exp}(mp_x)`$.
The Boltzmann weight of each configuration $`n`$ is a product of transfer matrix elements associated with the configuration of fermions on each shaded region. These elements are shown in figure 2.
Representing the magnitude of the Boltzmann weight by $`W[n]`$ and the sign, which takes into account the various anti-commutation relations, by $`\text{Sign}[n]`$ the partition function takes the form
$$Z_f=\underset{n}{}\text{Sign}[n]W[n].$$
(7)
Clearly $`\text{Sign}[n]`$ is the product of the sign factors $`\mathrm{\Sigma }`$ that appear in the transfer matrix elements associated to fermion hops represented through figures 2(e) and 2(f). When the fermion hops from $`x`$ to $`x+\widehat{i}`$ or vice versa due to the action of $`\mathrm{exp}(h_{x,i})`$ it picks up a product of local sign factors due to terms like $`\eta _{x,i}`$ as well as a string of non-local signs that arise due to anti-commutation relations involved when the fermion has to cross other fermions on the ordered lattice while reaching its destination. The evaluation of this non-local part of $`\mathrm{\Sigma }`$ is rather tedious. Separating the two parts, one can define $`\text{Sign}[n]=\text{Sign}_\mathrm{f}[n]\text{Sign}_\mathrm{b}[n]`$, where $`\text{Sign}_f[n]`$ comes from the product of non-local parts of $`\mathrm{\Sigma }`$ and is purely fermionic, where as $`\text{Sign}_\mathrm{b}[n]`$ is the one that comes from the local parts and may arise even in bosonic models. Fortunately $`\text{Sign}_\mathrm{f}[n]`$ has a topological meaning. The occupied lattice sites define fermion world-lines which are closed around the Euclidean time direction. However, during their Euclidean time evolution fermions can interchange their positions, and the fermion world-lines thus define a permutation of particles. The Pauli exclusion principle dictates that the $`\text{Sign}_f[n]`$ is just the sign of that permutation. Further, $`\text{Sign}_\mathrm{b}[n]`$ receives an extra minus-sign for every fermion that hops across a spatial boundary due to anti-periodic boundary conditions,
## III Bond Variables and Meron Clusters
When the partition function is written as a sum over fermion world-line configurations, the Boltzmann weight of each configuration is not guaranteed to be positive definite. This leads to the well known fermion sign problem. It was discovered recently that sometimes it is possible to solve this problem when the Boltzmann weight of each fermion world line configuration is written as a sum of Boltzmann weights describing new type of configurations, referred to as “connected” fermion world line configurations, obtained by the introduction of bond variables that establish connections with in the lattice. An example of this new type of configuration is shown in figure 3. In a class of models, including the one considered in this article, it is possible to regroup these new type of configurations such that the sum of the weights of each group is positive or zero, thus canceling all negative weights. The regrouping is accomplished by identifying all possible connected fermion world-line configurations obtained by flipping clusters of connected lattice sites. By flipping a lattice site one means that if the site is filled (with a fermion) it is emptied and vice versa. The result of a cluster flip is also shown in Figure 3.
In order to calculate the Boltzmann weight of every new type of configurations each interaction plaquette of in the old configuration given in figure 2 is rewritten as a sum of new configurations that involve bonds in addition to the fermions. The sign factors $`\mathrm{\Sigma }`$ are ignored at this step and will be considered during the regrouping step. Further, the bonds are introduced so that they obey the property of a flip symmetry. This means that the magnitude of the Boltzmann weight of connected fermions world line configurations do not change when connected sites are flipped. This property will be useful to identify configurations with equal weight but opposite signs during the regrouping step. For the two site interactions the weights of the new bond variables is shown at the top of figure 4. For the single site interaction due to the mass terms one has to distinguish between even and the odd sites since the weights differ on these two sites. However, now by introducing two types of bonds the property of flip symmetry can be maintained as shown in the bottom part of figure 4. The difference between the two bonds is that when the sites connected by the “dashed” bond are flipped, an extra sign change has to be taken into account. These signs must be multiplied to the $`\text{Sign}[n]`$ and the overall sign should be taken into account at the regrouping stage.
The above procedure gives a new expression for the path integral in terms of connected fermion world-line configurations.
$$Z_f=\underset{𝒞}{}\text{Sign}[𝒞]W[𝒞]$$
(8)
Each configuration $`𝒞`$ now defines a set of clusters in addition to the fermion occupation numbers as shown in figure II. For the present model the clusters form closed loops. The weight $`W[𝒞]`$ of each configuration can be read off using the rules of figure II. The $`\text{Sign}[𝒞]`$ is the product of $`\text{Sign}[n]`$ for the old configuration obtained by ignoring the bond and the extra negative sign factors that may arise from filled dashed-bonds on even sites and empty dashed-bonds on odd sites. A great simplification occurs when the flip symmetry imposed during the bond introduction is used. If a configuration $`𝒞`$ has $`N_𝒞`$ clusters, then all the $`2^{N_𝒞}`$ configurations, obtained by flipping the clusters independently, have the same weight $`W[𝒞]`$. This degeneracy can be used to regroup the configurations and write
$$Z_f=\underset{𝒞}{}\overline{\text{Sign}[𝒞]}W[𝒞]$$
(9)
where $`\overline{\text{Sign}[𝒞]}`$ is the average of $`\text{Sign}[𝒞]`$ over these $`2^{N_𝒞}`$ configurations. The usefulness of eq. (9) can be appreciated when $`\text{Sign}[𝒞]`$ obeys the following two properties:
* $`\text{Sign}[𝒞]=_{i=1}^{N_𝒞}\text{Sign}[c_i]`$, where $`\text{Sign}[c_i]`$ is the sign of the $`i`$th cluster configuration labeled $`c_i`$.
* For every configuration the individual clusters can be flipped to a reference configuration $`c_i^{\mathrm{ref}}`$ such that $`\text{Sign}[c_i^{\mathrm{ref}}]=1`$.
It is then easy to show that $`\overline{\text{Sign}[𝒞]}=0`$, if there is at least a single cluster whose flip changes the sign of the configuration, i.e., there is a cluster $`c_i`$ such that $`\text{Sign}[c_i^{\mathrm{non}\mathrm{ref}}]=1`$. Such clusters are referred to as merons. A configuration with no merons will then have $`\overline{\text{Sign}[𝒞]}=1`$. There are models in which $`\text{Sign}[𝒞]`$ has the above two properties and the present model falls in this class. Although the first property of $`\text{Sign}[𝒞]`$ is difficult to establish, it is easy to show that every cluster in the present model can always be flipped such that all even sites are empty and odd sites are filled in the present model. This leads to the staggered configuration shown on the right in figure II.
The new expression for the partition function derived in eq. (9) shows that fermion dynamics in the present model is equivalent to a statistical mechanics of clusters. Each cluster has two degrees of freedom related to the fermions. The Pauli-exclusion principle is encoded in the fact that meron clusters contribute zero weight to the partition function. In order to determine if a cluster is a meron or not, one has to understand its topology. If $`n_\mathrm{h}`$ is the number of hops of the cluster to the neighboring lattice site on the same time slice, $`n_\eta `$ is the number of local signs that the cluster encounters during the hops, $`n_w`$ is the temporal winding and $`n_d`$ is the number of dashed bonds in the cluster, then the cluster is a meron if $`(n_w+n_\mathrm{h}/2+n_d+n_\eta )`$ is an even integer. Cluster algorithms based on the concept of merons was originally invented in and are referred to as the meron cluster algorithms.
## IV Algorithm and Results
It is easy to construct cluster algorithms that generate clusters with weight $`W[𝒞]`$ obtained using the rules given in figure II. This makes it possible to simulate the modified model obtained by ignoring the sign factors. The partition function of the resulting model, denoted by $`Z_b`$, can be expressed as a sum over contributions from various meron sectors, i.e., $`Z_b=_NZ_N`$ where $`N`$ denotes the number of merons in a cluster configuration. On the other hand the actual partition function $`Z_f=Z_0`$. Since a typical configuration of $`Z_b`$ is filled with merons, it is exponentially difficult to find zero meron sectors and the cluster algorithm is still inefficient for simulating the dynamics of $`Z_f`$. A solution to the problem of avoiding configurations with meron clusters is necessary. Fortunately, using a local Metropolis accept-reject step, it is possible to introduce a re-weighting factor that suppresses higher meron sectors. For example the Metropolis step can modify the weight of cluster configurations to $`W[𝒞]/p^N`$, where $`p>1`$ and $`N`$ represents the number of merons in the configuration. This modifies $`Z_b`$ but leaves $`Z_f`$ unchanged! This Metropolis step then increases the efficiency of the algorithm by an exponential factor.
A variety of quantities can be calculated with the new algorithm. The first step, however, is to find expressions for these quantities in terms of cluster variables. Operators that are diagonal in the occupation number basis are easy to calculate, although more general operators can also be evaluated. For some quantities it is possible to find analytic expressions for the average over all the cluster flips, which are referred to as improved estimators. The chiral condensate in the present problem, which is also the order parameter for the chiral phase transition, is one such quantity. In the operator language the condensate is defined as
$$\overline{\psi }\psi =\frac{1}{V}\frac{\mathrm{Tr}[\mathrm{e}^{\beta H}𝒪]}{\mathrm{Tr}[\mathrm{e}^{\beta H}]}$$
(10)
where $`𝒪=_xp_x`$. The improved estimator on the other hand is given by
$$\overline{\psi }\psi =\frac{1}{2}\frac{p\mathrm{Size}(c_{\mathrm{meron}})\delta _{N,1}}{\delta _{N,0}V\beta },$$
(11)
where $`N`$ is the number of merons in the cluster configuration generated by the algorithm and $`\mathrm{Size}(c_{\mathrm{meron}})`$ is the size of the meron cluster. The extra re-weighting factor $`p`$ takes into account the modifications due to the Metropolis step. It is easy to show that on a finite lattice in the $`m=0`$ limit there are no single meron sectors and hence the chiral condensate simply vanishes. On the other hand, as seen in figure IV, an extremely tiny mass $`m=0.001`$ shows convincingly the effects of spontaneous symmetry breaking. The critical value of $`\beta _c=0.948`$ obtained through finite size scaling analysis in appears consistent with the present results. In the chiral susceptibility was measured in the chiral limit its behavior near the transition was shown to be consistent with the universality class of the 3-d ising model. A more detailed analysis of the data from the present study will be presented elsewhere.
## V Conclusions
The fermion cluster algorithm constructed in this article is an example of a novel way to compute fermionic path integrals. The sign problems that arise in the new method are solved using the concept of a meron which makes it easy to match all the configurations with negative Boltzmann weight with configurations with an equal but positive Boltzmann weight. Some interesting features emerge with the new algorithm. For the first time it is possible to treat fermions and bosons on an equal footing in simulations. Further it is possible to approach chiral limits more easily as discussed here.
A number of applications of the new techniques are presently being studied. For example it appears possible to find models with continuous chiral symmetry in which the sign problem can be completely solved. Chiral symmetry breaking in such a system will lead to the existence of massless Goldstone particles in the chiral limit. Apart from clarifying issues of universality the presence of the low mass particles can help in studying resonance physics that become difficult with conventional algorithms. Applications to non-relativistic many fermion problems is another topic where the present techniques are directly applicable. It is also likely that solutions to Hubbard type models will emerge with these techniques.
Adding gauge fields introduces new negative signs. It will be exciting to find solutions to this class of sign problems using cluster methods. There is evidence from strong coupling limits that a regrouping of configurations that differ in both gauge and fermion content can help solve such problems. This question is presently under investigation.
## Acknowledgments
I would like to thank R. Brower, J. Cox, K. Holland, J. Osborne and U.-J. Wiese for collaboration and insightful discussions. The numerical results presented here were obtained by J. Osborne. The work was supported in part by a grant from the U.S. Department of Energy, Office of Research (DE-FG02-96ER40945). |
no-problem/0001/astro-ph0001434.html | ar5iv | text | # Broad-band colour evolution of star clusters
## 1. Introduction
The basic properties of distant star clusters, like ages, metallicities, or masses, can be derived only from integrated spectral properties. The obtention of spectra with reasonable S/N ratio is, in general, proibitive for galaxies located outside the Local Group (but for some few expections, e.g. NGC 7252; Schweizer & Seitzer 1998). It follows that broad-band magnitudes and colours are the primary source of information about masses and ages of distant star clusters. Up to recently, this was true even for the relatively nearby Magellanic Cloud (MC) clusters. We point out that from a total of about 2400 LMC clusters, only about 300 have ages derived directly from the features in the colour–magnitude diagram (CMD), whereas 600 have them estimated from the colours (Bica et al. 1996; Girardi et al. 1995). This latter situation is however rapidly changing, as massive and high-quality photometric data become available for the MCs.
In most papers in the field, the ages of distant star clusters are estimated by a simple comparison of the observed colours with those predicted by a set of population synthesis models. As simple and easy-to-apply this approach is, it is not demonstratedly the best one. In fact, for the nearby LMC clusters, empirical or semi-empirical methods have so far been preferred, and we are not aware of recent works in which the LMC clusters are age-dated from a direct comparison between model and observed colours. Some of the reasons for this will be mentioned in the following.
## 2. Theoretical grounds
The HR diagram of Fig. 1 shows a sequence of isochrones at equally-spaced intervals of the logarithm of the age, $`\mathrm{log}t`$. From this figure it is evident that both $`\mathrm{log}L`$ and $`\mathrm{log}T_{\mathrm{eff}}`$ of the turn-off stars fade at an almost constant rate with $`\mathrm{log}t`$. This is a simple consequence of the mass–luminosity relation of main sequence (MS) stars being a power-law. Core-helium burning (CHeB) stars follow the same linear trend in $`\mathrm{log}L`$, but only up to ages $`1`$ Gyr. These trends determine, to a large extent, the main features of the evolution of integrated magnitudes: if the initial mass function is also a power law (e.g. the Salpeter one), one expects an almost constant rate of fading of magnitudes with $`\mathrm{log}t`$.
And in fact, this is the case for the blue–visual pass-bands, like $`UBV`$, which depend mostly on the behaviour of stars in the stages of MS termination and CHeB. The roughly linear fading of $`M_U`$, $`M_B`$, $`M_V`$, and $`M_R`$ with $`\mathrm{log}t`$ (with slope $`2`$ mag/dex) is shown in Fig. 2. The same behaviour holds for the near-ultraviolet evolution (e.g. the 1550 pass-band), but only up to an age of $`10^8`$ yr; after that, turn-off stars are too cold to contribute to the UV spectra.
The situation becomes different in the red and near-infrared. Evolved stars such as red supergiants (RSG), asymtotic giant branch (AGB) and first-ascent red giants (RGB) determine the evolution at these wavelengths. Since the number of these stars is a strong (and non-monotonic) function of $`\mathrm{log}t`$, the resulting magnitude and colour evolution is also complicated. This can be seen in Fig. 2: in $`I`$ and more clearly in $`JHK`$, the evolution is marked by the RSGs at $`>10^7`$ yr, and by the presence of AGB stars at $`>10^8`$ yr. Between $`\mathrm{log}t8`$ and $`9.3`$, the details of the colour evolution are essentially determined by AGB stars. At $`10^{9.2}`$ yr, a transient red phase appears due to the increased number of AGB stars which temporarily follows the onset of the RGB (Girardi & Bertelli 1998).
## 3. Trends with metallicity
The above picture is valid for near-solar metallicities. Going to lower metallicities, some of the trends are just obvious: for the same $`\mathrm{log}t`$, colours get bluer. However, this effect is much larger at ages $`2`$ Gyr, and for the red and near-infrared pass-bands. This is shown in Fig. 3, with the colour evolution at different metallicities. For a large age interval, $`7.3<\mathrm{log}t<9.2`$, $`UB`$ and $`BV`$ colours differ by less than 0.15 mag for metallicities in the interval $`0.004<Z<0.019`$. Moreover, the $`UB(t)`$ and $`BV(t)`$ relations are virtually monotonic. It follows that $`UB`$ and $`BV`$ are good age indicators, provided that $`BV<0.6`$ and $`UB<0.2`$.
The metallicity dependence of colours gets larger at ages $`\mathrm{log}t>9.3`$, when subgiant and RGB stars become important in comparison with MS and CHeB ones. This is the range of the age–metallicity degeneracy: ages can be estimated only if independent information on metallicity is provided, and vice-versa.
At low metallicities, the RSG red phase is practically missing, and AGB and RGB phases develop at higher temperatures. This result in small changes in the behaviour of $`UB`$ and $`BV`$ colours, which are, fortunately, of low entity for most of the age interval, i.e. $`7.3<\mathrm{log}t<9.2`$. In the $`IJHK`$ pass-bands the changes become remarkable, since these stars are the main contributors to the red and near-infrared flux. It comes out that colours like $`VI`$ and $`VK`$ are very dependent on the way AGB stars are considered in the models (see Girardi & Bertelli 1998). Moreover, the colour-age relation is not monotonic for $`VI`$ and $`VK`$. Thus, when taken separately from other colours, they are not good age indicators.
A remark is worth here. It is often assumed that all models of population synthesis predict equally well the evolution of broad-band colours, since they are, in general, based on the same sets of evolutionary tracks and synthetic spectra. This is not exactly true. One important source of difference between models is on the way different groups consider the termally pulsing AGB phase. In some cases, TP-AGB stars are even neglected; this is a bad assumption since up to 50 % of the bolometric luminosities, at certain ages, come from these stars (cf. Frogel et al. 1990). In other cases, TP-AGB stars are included according to some empirical prescription, as e.g. that derived from the Frogel et al. (1990) observations of AGB stars in MC clusters. Although apparently better, this latter approach does not include any metallicity dependence in the TP-AGB evolution, whereas every single aspect of this evolution (mass loss, dredge-up, hot-bottom burning, lifetimes) is expected to be strongly dependent on metallicity<sup>1</sup><sup>1</sup>1These dependences, even when not explicitly included in the equations of TP-AGB evolution, come naturally from the changes in the typical temperatures of AGB envelopes with metallicity.. The best approach, probably, is to use TP-AGB models whose parameters have been calibrated in order to reproduce the properties of AGB stars in the MCs, like those by Van der Hoek & Groenewegen (1997) and Marigo et al. (1999). The latter ones are adopted here.
For near-solar metallicities, TP-AGB stars affect mainly the red and near-infrared passbands (e.g. $`IJHK`$) and colours like $`VI`$ and $`VK`$. At low metallicities, TP-AGB stars slightly affect also the visual colours, like $`BV`$ and $`VR`$ (see Girardi & Bertelli 1998). These are the colours and metallicity ranges in which different population synthesis models may differ the most. We point out that the metallicity dependence of $`VK`$ is almost absent in the $`\mathrm{log}t>8`$ models of Fig. 3, just because the underlying TP-AGB models include a considerable dependence on this parameter, which compensates for the changes in the mean AGB temperature. Different trends would be found in other models.
## 4. The problem of age determination
A variety of effects, other than age and metallicity, determine the integrated colours of star clusters. A very subtle and ubiquitous one is that of “stochastic fluctuations” in the integrated colours due to the limited number of stars. The effects are sizeable already in $`BV`$ colours, even for the populous LMC clusters which have $`M_V<6`$. For $`M_V=6`$, the standard deviation of $`BV`$ varies from $`\sigma (BV)0.25`$ at $`\mathrm{log}t=7.3`$, to $`\sigma (BV)0.03`$ at $`\mathrm{log}t=9.0`$ (Girardi et al. 1995). It scales approximately as $`\sigma (BV)0.7^{M_V}`$.<sup>2</sup><sup>2</sup>2Notice that, in Eq. 3 of Girardi et al. (1995), the exponent is presented with the wrong sign. The correct equation should read $`\mathrm{}\times 0.7^{(M_V+6)}`$. Thus, only for intermediate-age clusters, and for young ones with $`M_V<9`$, is this intrinsic dispersion lower than that caused by cluster-to-cluster metallicity variations of $`\sigma ([\mathrm{Fe}/\mathrm{H}])0.3`$ dex.
In contrast, more dramatic stochastic fluctuations are found in the red and near-infrared pass-bands, due to the sampling of a small number of luminous red stars per cluster (e.g. the AGB ones; see Santos & Frogel 1998; and Lançon, this meeting). Also in this respect, the blue-visual pass-bands do a better job than red and near-infrared passbands, since they sample the light coming from stars in well-populated evolutionary stages.
To summarize, colours such as $`BV`$ and $`UB`$ present several potential advantages for the age determination of young clusters, as the lower sensitivity to metallicity, stochastic flutuations, and their monothonic behaviour with age. In fact, the most classic way to determine the ages of MC clusters from colours, namely the Elson & Fall (1985; EF) one, makes use of the $`BV`$ versus $`UB`$ two-colour plane. Clusters draw a fairly regular $`S`$-shaped curve in this plane, so that their position along this curve simply indicate their $`\mathrm{log}t`$ values. The EF method has been slightly modified by Girardi et al. (1995) in order to take into account the fact that the 3 most important factors of dispersion of colours at fixed age – namely the metallicity, reddening, and stochastic dispersions, – act in almost the same direction in the two-colour plane. However, it has been demonstrated that the age determinations with the EF method become unreliable for clusters older than say $`4`$ Gyr, due to the age–metallicity degeneracy above mentioned.
The EF method has the advantage of being empirical, since the age sequence of clusters in the two-colour plane is calibrated using LMC clusters with ages directly determined from the CMD features. In this way, uncertainties in the integrated colours of models (always present and of order $`0.050.1`$ mag in $`UBV`$) do not affect age estimates. Determinations of $`\mathrm{log}t`$ obtained this way present accuracies of $`0.15`$ dex, at least for the most populous LMC clusters (Girardi et al. 1995). We remark that, using a simple comparison between the LMC cluster colours and population synthesis models, the typical accuracies would be much worse, i.e. $`\sigma (\mathrm{log}t)0.3`$ dex. The validity of the EF method has also been confirmed for SMC clusters by Grebel et al. (1999).
Considering these aspects, it is surprising that empirically-calibrated methods are rarely used (and exceptions are e.g. Bresolin et al. 1996; and Larsen & Richtler 1999) for the age-dating of distant clusters. Moreover, very often are the ages of distant clusters derived from $`VI`$ or $`VK`$, which, from a theoretical point of view, are not the most apropriated colours. To be more specific, the age information these colours contain is poorer than that contained in a single measure of $`BV`$ and/or $`UB`$, because of the effects of non-monoticity with $`\mathrm{log}t`$, metallicity dependence, and/or stochastic dispersion. For $`VK`$, the importance of these effects can be confirmed by simply looking at the way $`VK`$ colours change among clusters of similar age in the LMC (Santos & Frogel 1998; Girardi & Bertelli 1998). More compelling reasons for avoiding the red pass-bands in age determinations, can be derived from population synthesis models. One finds that the predicted $`VI`$ and $`VK`$ colour evolution can differ appreciably from author to author, as a result of the different ways in which the several models include the TP-AGB evolutionary phase. This uncertainty is not present in the $`UBV`$ pass-bands.
However, it is also true that most modern instrumentation, and the strong absorption met in several interesting galaxies, favour the observations in pass-bands like $`VRIK`$, instead of the $`UBV`$ necessary for easy age determinations. Measuring clusters in red pass-bands is, in many cases, a necessity.
The $`BVRI`$ pass-bands are, probably, those in which the best data for distant clusters are available. To improve upon the age estimates obtained from these data, it would be adviseable to test, empirically, the age and metallicity dependences of the $`VR`$ and $`VI`$ colours. We notice that this can be done, to some extent, using the present-day photometric databases for MC clusters (e.g. OGLE, MACHO, and the MC Photometric Survey; see the contributions in Chu et al. 1999). Only after these behaviours are well documented, will we be able to check the reliability of such age determinations. Also, this kind of work may constitute an important test to models of spectral evolution of stellar populations, regarding, especially, the way they include (or do not) the latest evolutionary phases for different metallicities.
### Acknowledgments.
We acknowledge support by the European Community under TMR grant ERBFMRX-CT96-0086, and by the Italian MURST. Many thanks are due to G. Bertelli, E. Bica, A. Bressan, C. Chiosi, P. Marigo and A. Weiss for the many useful discussions and remarks.
## References
Bertelli G., Bressan A., Chiosi C., Fagotto F, Nasi E., 1994, A&AS, 106, 275
Bica E., Clariá J.J., Dottori H., Santos Jr. J.F.C., Piatti A., 1996, ApJS 102, 57
Bresolin F., Kennicut Jr. J.C., Stetson P.B., 1996, AJ 112, 1009
Chu Y.-H., Suntzeff N., Hesser J., Bohlender D. (eds.), 1999, IAU Symp. 190, New views of the Magellanic Clouds.
Elson R.A.W., Fall S.M., 1985, ApJ 299, 211 (EF)
Frogel J.A., Mould J., Blanco V.M., 1990, ApJ, 352, 96
Girardi L., Bertelli G., 1998, MNRAS 300, 533
Girardi L., Bressan A., Bertelli G., Chiosi C., 2000, A&AS, in press
Girardi L., Chiosi C., Bertelli G., Bressan A., 1995, A&A, 298, 87
Grebel E.K., Zaritsky D., Harris J., Thompson I., 1999, in IAU Symp. 190, 405
Larsen S.S., Richtler T., 1999, A&A 345, 59
Marigo P., Girardi L., Bressan A., 1999, A&A, 344, 123
Santos Jr. J.F.C., Frogel J.A., 1997, ApJ 479, 764
Schweizer F., Seitzer P., 1998, AJ 116, 2206
van den Hoek L.B., Groenewegen M.A.T, 1997, A&AS 123, 305 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.