1: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2: %
3: % Matrix element correlations and fluctuations
4: %
5: % Bruno Eckhardt, Imre Varga, Peter Pollner
6: %
7: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
8: \documentstyle[twoside,fleqn,epsf,espcrc2,draft]{article}%
9: \topmargin -1.0cm
10: \begin{document}
11: \renewcommand{\refname}{\normalsize\bf References}
12: \newcommand{\mquad}{\!\!\!\!\!}
13: \title{
14: Correlations and fluctuations of matrix elements and cross sections
15: }
16: \author{
17: Bruno Eckhardt, Imre Varga and P{\'e}ter Pollner
18: \address{Fachbereich Physik, Philipps--Universit{\"a}t Marburg,\\
19: Renthof 6, D-35032 Marburg an der Lahn, Germany}
20: \thanks{Supported by OTKA T029813, T024136 and F024135, the Alexander
21: von Humboldt foundation,the Hungarian Committee
22: for Technical Development (OMFB) and DAAD-M\"OB}
23: }
24: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
25: %
26: % abstract
27: %
28: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
29: \begin{abstract}
30: \hrule
31: \mbox{}\\[-0.2cm]
32:
33: \noindent{\bf Abstract}\\
34:
35: The fluctuations and correlations of matrix elements of
36: cross sections are investigated in open systems that are chaotic in the
37: classical limit. The form of the correlation functions is discussed
38: within a statistical analysis and tested in calculations for
39: a damped quantum kicked rotator.
40: We briefly comment on the modifications expected for
41: systems with slowly decaying correlations,
42: a typical feature in mixed phase spaces.
43: \\[0.2cm]
44: {\em PACS}: 05.45.+b,03.65.Sq\\[0.1cm]
45: {\em Keywords}: random matrix theory, quantum kicked rotator, cross section
46: correlations \\
47: \hrule
48: \end{abstract}
49:
50: \maketitle
51:
52: \section{Introduction}
53: The photo-dissociation cross section as determined by Fermi's golden
54: rule contains a combination of final density of states
55: and transitional matrix elements. The statistical properties of
56: this cross section in a situation of a chaotic dynamics
57: will thus be determined by the statistical properties of both the
58: density of states and the matrix elements. Moreover, if there is
59: a classical underlying dynamics, both will be connected, in
60: a suitable semiclassical limit, to the properties of the
61: classical dynamics. Our aim here is to present a few
62: numerical and theoretical considerations connected to these
63: observations.
64:
65: We focus on the form of the correlation function
66: of photo-absorption cross sections. Using
67: a random matrix theory approach Alhassid and Fyodorov \cite{AF}
68: found a correlation function that consisted of a Lorentzian and
69: a derivative of a Lorentzian with respect to the width.
70: The first term is familiar from the analysis of Ericsson
71: fluctuations in nuclear physics \cite{USLeshouches,BS}
72: and the second one from
73: the correlation function of the Wigner time delay
74: \cite{Chaos}. The
75: form of the correlation function can now be made plausible
76: if within a statistical model for the cross section the
77: fluctuations in the density of states and in the transition
78: matrix elements are independent, as discussed in section 2.
79:
80: A semiclassical analysis of this correlation function \cite{EFV}
81: shows that the relative weight of the two contributions depends
82: on the ratio of the fluctuations of the observable to the
83: mean. In case of a single initial state, random matrix theory
84: fixes this ratio to universal numbers. However, in situations
85: an incoherent superposition of initial states contributes
86: to the cross section, variations in the relative weight are
87: possible. This is illustrated within a numerical analysis for
88: a damped kicked rotator in section 3.
89:
90: The semiclassical connection also suggests certain modifications
91: in the correlation functions if the classical decay is not
92: purely exponential\cite{other}. In particular, in situations
93: with mixed phase space an algebraic decay is expected\cite{chirikov}.
94: The modifications
95: in the correlation functions include a slower decay
96: and the formation of a cusp at the origin, depending on
97: the exponent of the decay law. In section 4 we propose
98: a model for the form factor and analyze some of the consequences.
99:
100: We conclude with a brief summary in section~5.
101:
102: \section{Matrix element correlations within random matrix theory}
103:
104: The quantity we focus on is the density of states weighted by
105: the matrix elements of the observable,
106: \begin{equation}
107: \rho_A(E)=\sum_{\mu} A_{\mu}\delta_{\eta}(E-E_{\mu}),
108: \label{ro}
109: \end{equation}
110: where the sum runs over the eigenstates of the system having
111: eigenvalues $E_{\mu}$. The photo-absorption cross section is
112: proportional to this expression if the observable $A$ contains
113: the projection onto the initial state and the dipole operator.
114: The expectation value of $\hat A$ can be
115: written as $A_{\mu }=\bar{A}+\delta A_{\mu}$, where
116: $\bar{A}$ is the mean and $\delta A_{\mu}$ is the random fluctuation
117: around its mean. Statistically we assume
118: \begin{equation}
119: \langle\delta A_n\rangle=0, \qquad {\rm and} \qquad
120: \langle\delta A_n\delta A_m\rangle=\sigma^2_A\delta_{n,m}.
121: \label{stat}
122: \end{equation}
123: The function $\delta_{\eta}(x)$ is a Lorentzian function with half
124: width parameter $\eta$, normalized so that as $\eta\to 0$
125: it approaches a Dirac--$\delta$.
126: We take the same value of $\eta$ for all the eigenstates implying a uniform
127: damping. The mean density of states is simply
128: \begin{equation}
129: \bar{\rho}_A = \langle\rho_A(E)\rangle_E =
130: \int_B \frac{dE}{B}\rho_A(E) = \frac{N{\bar A}}{B},
131: \label{meanrho}
132: \end{equation}
133: where $N$ is the number of levels and $B$ is the energy width of the
134: subset of the spectrum, over which the average $\langle\dots\rangle_E$ is
135: calculated.
136:
137: The normalized autocorrelation function of the fluctuations of the density of
138: states, $\delta\rho_A(E)=\rho_A(E)-\bar{\rho}_A$ is defined as
139: \begin{eqnarray}
140: C(\varepsilon )\mquad &=&\mquad \frac
141: {\langle\delta\rho_A(E+\varepsilon )\delta\rho_A(E)\rangle_E}
142: {\bar{\rho}^2_A} \label{cf} \\
143: \mquad &=&\mquad \left (\frac{B}{N{\bar A}}\right )^2
144: \int_B \frac{dE}{B}\rho_A(E+\varepsilon)\rho_A(E) - 1.
145: \nonumber
146: \end{eqnarray}
147: Inserting the definition (\ref{ro}) in (\ref{cf}) we obtain
148: \begin{eqnarray}
149: C(\varepsilon)\mquad &=&\mquad \left(\frac{B}{N{\bar A}}\right)^2
150: \sum_{\mu,\nu}\big[{\bar A}^2+{\bar A}(\delta A_{\mu} + \delta A_{\nu})
151: \nonumber \\
152: & & \qquad \quad
153: + \delta A_{\mu}\delta A_{\nu}\big]g_{\eta}(\varepsilon,E_{\mu},E_{\nu})
154: -1,
155: \label{cf1}
156: \end{eqnarray}
157: where we have inserted the shorthand notation
158: \begin{equation}
159: g_{\eta}(\varepsilon,E_{\mu},E_{\nu})=
160: \langle\delta_{\eta}(E-E_{\mu}+\varepsilon)\delta_{\eta}(E-E_{\nu})\rangle_E.
161: \label{shn}
162: \end{equation}
163: With the assumption that matrix elements and resonances are uncorrelated
164: \cite{physicad}, averaging over the $\delta A$'s eliminates two terms,
165: \begin{eqnarray}
166: C(\varepsilon)&=&
167: \left (\frac{B}{N}\right )^2\sum_{\mu,\nu}
168: g_{\eta}(\varepsilon,E_{\mu},E_{\nu})-1
169: \label{cf2} \\
170: &+&
171: \left (\frac{B}{N{\bar A}}\right )^2\sum_{\mu.\nu}
172: \langle\delta A_{\mu}\delta A_{\nu}\rangle g_{\eta}(\varepsilon,E_{\mu},E_{\nu}).
173: \nonumber
174: \end{eqnarray}
175: In the second term we will utilize (\ref{stat}).
176: Also we will split the first double sum in diagonal ($\mu=\nu$) and
177: non-diagonal ($\mu\neq\nu$) parts. For sufficiently small $\eta$ the
178: diagonal term is the autocorrelation function of Lorentzian
179: that also yields a Lorentzian.
180: \begin{eqnarray}
181: C(\varepsilon)&=&
182: \left (1+\frac{\sigma^2_A}{{\bar A}^2}\right )\frac{B^2}{N}
183: \langle\delta_{\eta}(E-\varepsilon)\delta_{\eta}(E)\rangle_E -1 \nonumber \\
184: &+&
185: \left (\frac{B}{N}\right )^2\sum_{\mu\neq\nu}
186: g_{\eta}(\varepsilon,E_{\mu},E_{\nu}).
187: \label{cf3}
188: \end{eqnarray}
189: After performing the averaging in the first term and in the sum together with
190: the definition in (\ref{shn}), one arrives at
191: \begin{eqnarray}
192: C(\varepsilon)&=&
193: \Delta\frac{\sigma_A^2}{{\bar A}^2}\delta_{2\eta}(\varepsilon)-1 \nonumber \\
194: &+&
195: \Delta\frac{1}{N}\sum_{\mu\neq\nu}
196: \delta_{2\eta}[\varepsilon-(E_{\mu}-E_{\nu})],
197: \label{cf4}
198: \end{eqnarray}
199: where $\Delta =B/N$. In the sum one can recognize the appearance of the
200: two--level correlation function in the limit of $\eta\to 0$. For strongly
201: overlapping resonances, i.e.\ when $\eta\gg\Delta$ the above expression
202: reduces to
203: \begin{equation}
204: C(\varepsilon)\propto
205: \left (\frac{\sigma_A^2}{{\bar A}^2}\frac{\eta}{\eta^2+\varepsilon^2} +
206: \frac{1}{2\pi}\frac{\eta^2-\varepsilon^2}{(\eta^2+\varepsilon^2)^2}
207: \right )
208: \label{cf5}
209: \end{equation}
210: Thus the correlation function is characterized by two terms, a Lorentzian
211: and a derivative of a Lorentzian with respect to the broadening parameter
212: $\eta$. The weight of the first term comes from the fluctuations of the
213: observable $A$. In the case when $\hat A=|i\rangle\langle i|$ is a
214: projection on the basis state $|i\rangle$. This quantity in random
215: matrix theory (RMT) \cite{RMT} is $(\beta +2)/\beta$, where $\beta
216: =1,2,$ and 4 for the different universality classes the system belongs
217: to (orthogonal, unitary, and symplectic, respectively). Hence we
218: recover the correlation function derived in Ref.~\cite{AF} for
219: channels with uniform resonance width.
220:
221: However, if $\bar{A}$ vanishes, the normalization by $\bar{A}$ is not
222: possible and the second term, which comes from the correlation
223: function of the density of states, cf.~(\ref{cf1}) and (\ref{shn}),
224: disappears and the correlation function becomes a pure
225: Lorentzian. These findings are thus in accordance with what has been
226: argued on the basis of semiclassical periodic orbit theory\cite{EFV}.
227:
228: \section{Correlations in the quantum kicked rotator}
229:
230: To illustrate the above calculations we calculate the correlation
231: function of the matrix element weighted density of states for a damped
232: kicked rotator.
233:
234: We consider the statistical properties of observables that are projections
235: onto a subset of the basis states
236: \begin{equation}
237: \hat A=\sum_{n\in I(m)} |n\rangle\langle n|,
238: \label{Adef}
239: \end{equation}
240: where $I(m)$ is a subset of size $m<N$ of the basis set. The fluctuations
241: of the matrix elements $\langle\mu |\hat A|\mu\rangle $ over the eigenstates
242: $\mu$ of the system describe the cross section fluctuations of the
243: excitation of the system from an initial state $I(m)$ to the final
244: state $\mu$ \cite{AF,FA,OA,EFV}. A possible dipole operator has been
245: absorbed into the definition of $|\mu\rangle$.
246:
247: The model system we considered is a quantum kicked rotator with a kicking
248: potential
249: \begin{equation}
250: V(\phi )=k(\cos\phi - \sin 2\phi).
251: \end{equation}
252: It is known \cite{BS,other} that this model belongs to the unitary
253: universality class since the second term in the potential breaks the
254: conjugation symmetry.
255:
256: We have diagonalized the unitary one--step evolution operator $U$ at a
257: value of the classical kicking strength, $K$, where complete ergodicity
258: was expected ($K=7$~\cite{shepel}). The size of the system was fixed
259: to $N=201$. The matrix element weighted density of states was defined as
260: \begin{equation}
261: \rho_A(\phi )=\sum_{\mu}\langle\mu|{\hat A}|\mu\rangle
262: \left (1-e^{i(\phi -\phi_{\mu})-\Gamma/2}\right )^{-1}.
263: \end{equation}
264: The mean value of this observable is simply $\bar A=m/N$ where
265: $1\leq m\leq N$ is the number of states contributing in the projection.
266: \begin{figure}[ht]
267: \vspace{-0.3in}
268: \leavevmode
269: \epsfxsize=7cm
270: \epsfbox{g1.eps}
271: \vspace{-0.5cm}
272: \caption{\label{qkr}
273: The correlation function $C(x)$ as a function of the eigenvalue separation
274: in units of mean level spacing. The data points are obtained for observables
275: of projections over $m$ basis states indicated in the legend. The continuous
276: curves are fits of the form (\protect\ref{sc-cf}) fitting $\alpha$ and
277: $\gamma$ independently. The width of the resonances was uniform throughout
278: the spectrum: $\Gamma=\Delta$.}
279: \vspace{-0.5cm}
280: \end{figure}
281: We have calculated the correlation function (\ref{cf}) applied for our
282: case
283: \begin{equation}
284: C(x)=\langle \delta\rho_A(\phi +x)\delta\rho_A(\phi) \rangle_{\phi},
285: \end{equation}
286: i.e.\ averaging $\langle\dots\rangle_{\phi}$ is done over the eigenphase
287: spectrum extending over $B=2\pi$. The variables $\phi$ and $x$ are measured
288: in units of the mean level spacing $\Delta=B/N$ therefore
289: $\bar\rho_A=2\pi{\bar A}/N$. The arguments of \cite{EFV} and the RMT
290: arguments of the previous section shows, that the correlation function
291: is composed of a Lorentzian and a derivative of a Lorentzian
292: \begin{equation}
293: C(x)\propto\left (\alpha\frac{\gamma}{\gamma^2+x^2}+
294: \frac{1}{2\pi}\frac{\gamma^2-x^2}{(\gamma^2+x^2)^2}\right ).
295: \label{sc-cf}
296: \end{equation}
297: In (\ref{sc-cf}) $\alpha=\sigma^2_A/({\bar A}^2 T_H)$, with the
298: Heisenberg time $T_H=N$. The width $\gamma$ should come out to be the
299: damping $\eta$ of (\ref{cf5}), in appropriate units.
300:
301: This expression can be compared to numerical
302: results on the kicked rotator. In Fig.~\ref{qkr} we plot the
303: correlation function obtained for observables that are projections extending
304: over $m=1,3,10,50$ basis states. The continuous curves in the figure are
305: fits of the form (\ref{sc-cf}) allowing the two parameters $\alpha$ and
306: $\gamma$ to vary. The classical estimate for $\alpha$ is \cite{EFV,other}
307: \begin{equation}
308: \alpha=\frac{\sigma^2_A}{{\bar A}^2 N}=\frac{1-{\bar A}}{{\bar A}N}
309: \propto\frac{1}{m}
310: \label{var}
311: \end{equation}
312: The variance $\sigma^2_A$ follows its classical value for low values of
313: $p={\bar A}=m/N$ \cite{other}. We obtained a fitted value of
314: $\gamma=1.11\,\Gamma$ (instead of $\Gamma$)
315: independent of $m$ and $\alpha=0.284$, 0.135, 0.042, and 0.009 for
316: $m=1$, 3, 10, and 50, respectively. The dependence on $m$ is close to
317: the one predicted by (\ref{var}) at least for $m>1$.
318:
319: \section{Open systems and slowly decaying correlations}
320:
321: The correlations described in the previous sections can be derived also
322: using the following phenomenological procedure\cite{AF}. The two--level correlation
323: function of a closed chaotic system, $C_0(\varepsilon)$, consists of a
324: Dirac--$\delta$ at the origin and a smooth function decreasing to zero for
325: large level separations. This is the Fourier transform of the form factor
326: $K_0(\tau )$. From now on for sake of simplicity we restrict ourselves to
327: the unitary universality class, i.e.\ we write the RMT form factor in its
328: standard form \cite{RMT},
329: $K_0(\tau )=1-b(\tau )$, with
330: \begin{equation}
331: b(\tau )=(1-|\tau |)\Theta (1-|\tau |)
332: \end{equation}
333: where $\Theta(x)$ is the step--function. Time $\tau$ is measured in units
334: of the Heisenberg time $T_H$, therefore energy separation $\varepsilon$,
335: in units of mean level spacing $\Delta$. According to RMT the
336: correlation function is
337: \begin{eqnarray}
338: C_0(\varepsilon )&=&\int_{-\infty}^{\infty}d\tau
339: K_0(\tau )\cos (2\pi\varepsilon\tau) \nonumber \\
340: &=&\delta(\varepsilon )-
341: \left [\frac{\sin(\pi\varepsilon)}{\pi\varepsilon}\right ]^2.
342: \end{eqnarray}
343: Uniform damping, i.e.\ opening up the system, can be introduced by
344: multiplying the form factor with an exponential decay.
345: \begin{equation}
346: K(\tau )=K_0(\tau )e^{-\Gamma|\tau |},
347: \label{ffdamp}
348: \end{equation}
349: where $\Gamma$ measures the relaxation rate in units of the Heisenberg
350: time. In fact $\Gamma=T_c/T_H$, where $T_c$ is the relaxation time. When
351: $\Gamma<1$ ($\Gamma>1$) the relaxation happens over a time scale longer
352: (shorter) than $T_H$, hence produces correlations over energy scales
353: lower than (beyond) mean level spacing. The resulting form factor for
354: different values of $\Gamma=0.0$, 0.1, 1.0 and 5.0 are plotted in
355: Fig.~\ref{ffexp}. The Fourier transform of (\ref{ffdamp}) results in
356: \begin{equation}
357: C(\varepsilon )=C_1(\varepsilon )+C_2(\varepsilon )
358: \label{damp}
359: \end{equation}
360: where
361: \begin{equation}
362: C_1(\varepsilon)=\int_{-\infty}^{\infty}d\tau
363: e^{-\Gamma|\tau |}\cos(2\pi\varepsilon\tau),
364: \end{equation}
365: \begin{equation}
366: C_2(\varepsilon)=-\int_{-1}^1 d\tau (1-|\tau |)e^{-\Gamma|\tau |}
367: \cos(2\pi\varepsilon\tau).
368: \label{damp-2}
369: \end{equation}
370: The first term $C_1(\varepsilon)$ is a Lorentzian from the broadening
371: of the $\delta$-function and the second term, the convolution of the
372: Lorentzian with the two-level cluster function, leads, in the limit of
373: large $\Gamma$, to the derivative of the Lorentzian.
374: The corresponding
375: correlation functions $C(\varepsilon)$ are plotted in Fig.~\ref{cfexp}.
376: \begin{figure}[ht]
377: \vspace{-0.3in}
378: \leavevmode
379: \epsfxsize=7cm
380: \epsfbox{ffexp.eps}
381: \vspace{-0.5cm}
382: \caption{\label{ffexp}
383: The form factor for different uniform damping constants $\Gamma=0.0$, 0.1, 1.0,
384: and 5.0.}
385: \vspace{-0.5cm}
386: \end{figure}
387: \begin{figure}[ht]
388: \leavevmode
389: \epsfxsize=7cm
390: \epsfbox{cfexp.eps}
391: \vspace{-0.5cm}
392: \caption{\label{cfexp}
393: The correlation function, $C(\varepsilon )$ (see Eq.~(\protect\ref{damp})),
394: of an open system for different uniform damping rates $\Gamma=0.0$,
395: 0.1, 1.0, and 5.0. The inset shows $C_2(\varepsilon )$
396: (\protect\ref{damp-2}).}
397: \vspace{-0.5cm}
398: \end{figure}
399:
400: These considerations led us to extend the above approach in another direction.
401: Let us assume that the quantum return probability, the form factor, is
402: damped slower than exponential, i.e.\ in an algebraic fashion. The
403: correlation function in this case will contain signatures of the slowing
404: down of the classical dynamics, a behavior that is expected to be prominent
405: in systems with mixed phase space \cite{HLK}.
406:
407: Such an algebraic damping may be introduced in the phenomenological ansatz
408: \begin{equation}
409: K(\tau )=K_0(\tau )\left (1+c\tau\right )^{-a},
410: \label{alg1}
411: \end{equation}
412: where $c=T_H/T_c$ is the ratio of the decorrelation time $T_c$ compared
413: to the Heisenberg time $T_H$. As in the case of exponential
414: damping, for $c<1$ ($c>1$) the slow decorrelation of classical trajectories
415: due to the presence of the hierarchy of stable islands \cite{HLK,rk} occur
416: on a time scale that is longer (shorter) than the Heisenberg time that
417: produces correlations over energy scales lower than (beyond) mean level
418: spacing.
419: \begin{figure}[ht]
420: \vspace{-0.3in}
421: \leavevmode
422: \epsfxsize=7cm
423: \epsfbox{ffpl4_3.eps}
424: \vspace{-0.5cm}
425: \caption{\label{ffpl4_3}
426: The form factor with algebraic damping ($a=4/3$) for different ratios of
427: the Heisenberg time $T_H$ compared to the decorrelation time $T_c$,
428: $c=T_H/T_c=0.0$, 0.1, 1.0, and 5.0.}
429: \vspace{-0.5cm}
430: \end{figure}
431: \begin{figure}[ht]
432: \leavevmode
433: \epsfxsize=7cm
434: \epsfbox{cfpl4_3.eps}
435: \vspace{-0.5cm}
436: \caption{\label{pl-4_3}
437: The correlation function, $C(\varepsilon )$, of a system containing algebraic
438: decorrelation with exponent $a=4/3$ (see Eq.~(\protect\ref{pow}).
439: The curves stand for different ratios of the Heisenberg time $T_H$
440: compared to the decorrelation time $T_c$, $c=T_H/T_c$, with $c=0.0$,
441: 0.1, 1.0, and 5.0. The inset shows $C_2(\varepsilon )$~(\protect\ref{pow-2}).}
442: \vspace{-0.5cm}
443: \end{figure}
444: By fixing one of the parameters $a$ and $c$ the variation of the other results
445: in changes in different parts of the correlation function. Similarly as
446: in (\ref{damp}) we find
447: \begin{equation}
448: C(\varepsilon )=C_1(\varepsilon )+C_2(\varepsilon )
449: \label{pow}
450: \end{equation}
451: where
452: \begin{eqnarray}
453: C_1(\varepsilon)&=&\int_{-\infty}^{\infty}d\tau\frac{\cos(2\pi\varepsilon\tau)}
454: {(1+c|\tau |)^a}, \\
455: C_2(\varepsilon)&=&-\int_{-1}^1 d\tau\frac{1-|\tau |}{(1+c|\tau |)^a}
456: \cos(2\pi\varepsilon\tau).
457: \label{pow-2}
458: \end{eqnarray}
459: We have plotted the correlation function for the case when $a=4/3$. This
460: type of correlations can be expected to occur in systems with mixed
461: phase space\cite{rk}. In Fig.~\ref{pl-4_3} the curves demonstrate that
462: a low value of $c$ results in only a slight modification of the correlations
463: while large $c$, i.e.\ when $T_c<T_H$ the algebraic decorrelation results
464: in changes both in the $\varepsilon<1$ and $\varepsilon>1$ regimes. This
465: is in contrast to previous expectations \cite{rk} that the dynamics occuring
466: over time scales up to $T_H$ should show up in the correlation function on
467: energy scales beyond mean level spacing only.
468:
469: To show the similarities and differences between correlation functions
470: obtained for exponential and power law damping we plot the value at
471: $\varepsilon=0$, i.e.\ the variance of the fluctuating observable under
472: investigation. In Fig.~\ref{c0fig} we can see that for $T_c>T_H$ the
473: two types of damping functions yield the same correlation time dependence,
474: $\propto (T_H/T_c)^{-1}$. In the case of $T_c<T_H$ the exponential
475: damping produces a $(T_H/T_c)^{-2}$ dependence and the power law version a
476: $(T_H/T_c)^{-a}$ (in the case shown $a=4/3$).
477:
478: \begin{figure}[ht]
479: \vspace{-0.3in}
480: \leavevmode
481: \epsfxsize=7cm
482: \epsfbox{c0.eps}
483: \vspace{-0.5cm}
484: \caption{\label{c0fig}
485: The value of $C(\varepsilon )$ at $\varepsilon=0$ for exponential
486: (continuous) and power law damping (dashed curve). The ratio $T_H/T_c$ is
487: $\Gamma$ for exponential and $c$ for power law damping. The power law curve
488: was obtained for the exponent of $a=4/3$.}
489: \vspace{-0.5cm}
490: \end{figure}
491:
492: An even more striking difference between the fast and slow decorrelations
493: shows up in the low--$\varepsilon$ behavior of the correlation function. As
494: pointed out already by Lai {\it et al.} \cite{Lai} non-hyperbolic systems
495: produce a cusp in the correlation function
496: \begin{equation}
497: C(\varepsilon )\sim C(0)-C_a\varepsilon^{a-1}
498: \label{cusp}
499: \end{equation}
500: where $C_a=b_ac^{-(a-1)}$ is a positive constant and $a$ is the exponent
501: of the classical return probability $P(t)\sim t^{-a}$. This behavior is
502: to be contrasted with the hyperbolic case when the low--$\varepsilon$
503: behavior is expected to be
504: \begin{equation}
505: C(\varepsilon )\sim C(0)-C_2\varepsilon^2,
506: \label{nocusp}
507: \end{equation}
508: where $C_2\sim\Gamma^{-2}$ is a constant depending on the exponential
509: relaxation rate. In Fig.~\ref{cuspfig} we show the cusp at
510: low--$\varepsilon$ in the case of power law decorrelation
511: and the parabolic behavior for exponential damping. The curves represent especially the
512: cases when $T_H/T_c\geq 1$, i.e.\ when the classical correlations appear at
513: time scales below the Heisenberg time. In this case one expects deviations
514: at $\varepsilon\geq 1$. The cusp at low--$\varepsilon$, however, is
515: present for any value of $c$.
516:
517: \begin{figure}[ht]
518: \vspace{-0.3in}
519: \leavevmode
520: \epsfxsize=7cm
521: \epsfbox{cusp.eps}
522: \vspace{-0.5cm}
523: \caption{\label{cuspfig}
524: The correlation function for low--$\varepsilon$ for exponential
525: (continuous) and power law damping (dashed curve) for two different values
526: of the decorrelation time $T_c/T_H=1$, and 5. The power law curve
527: was obtained for the exponent of $a=4/3$.}
528: \vspace{-0.5cm}
529: \end{figure}
530:
531: We would like to emphasize that the apparent contradiction between the
532: behavior of the correlation function for power law decorrelation (\ref{cusp})
533: and that of refs.~\cite{HLK,rk} resides probably in the difference of the
534: classical return probability functions considered. We took the modification
535: of the RMT form factor (\ref{alg1}) while in refs.~\cite{HLK,rk} the
536: classical return probability of $P(t)\sim t^z$ was used.
537:
538: \section{Final remarks}
539: The variations in the correlation functions for the different
540: situations are not very large and perhaps difficult to detect.
541: This seems to apply in particular to cases with an algebraic
542: decay, especially in view of the fact that the exponents are not
543: universal and might be clouded by a distribution of algebraic
544: decay laws. The modifications due to variations in matrix elements
545: and thus a stronger emphasis of the derivative part of
546: the correlation function could perhaps be achieved in
547: incoherent superpositions
548: of cross sections from different initial states. Analysis
549: of experimental data in this direction seems worthwhile.
550:
551:
552: \begin{thebibliography}{xxx}
553:
554: \bibitem{AF}Y. Alhassid and Y. V. Fyodorov,
555: J. Phys. Chem. {\bf 102}, 9577 (1998).
556:
557: \bibitem{USLeshouches} U. Smilansky, Semiclassical Quantization of Chaotic
558: Billiards - A Scattering Approach. in
559: {\em Proc. of the Les Houches Summer School on Mesoscopic Quantum Physics},
560: Elsevier Science Publ. (1995)
561: Ed. E. Akkermans, G. Montambaux and J. L. Pichard.
562:
563: \bibitem{BS}R. Bl\"umel and U. Smilansky,
564: Phys. Rev. Lett. {\bf 69}, 217 (1992).
565:
566: \bibitem{Chaos} B. Eckhardt,
567: % Correlations in quantum time delay,
568: Chaos {\bf 3}, 613 (1993)
569:
570: \bibitem{EFV}B. Eckhardt, S. Fishman and I. Varga, in preparation.
571:
572: \bibitem{other}B. Eckhardt, I. Varga and P. Pollner, in preparation.
573:
574: \bibitem{chirikov} B.V. Chirikov and D.L. Shepelyansky,
575: Phys. Rev. Lett. {\bf 82}, 528 (1999).
576:
577: \bibitem{physicad} B. Eckhardt,
578: %Fluctuations and correlations in matrix elements,
579: Physica {\bf D 109}, 53, (1997)
580:
581: \bibitem{RMT}M.L. Mehta, {\it Random Matrices}
582: (Academic Press, Boston, 1991);
583: T. Guhr, A. M\"uller--Groeling, and H.A. Weidenm\"uller,
584: Phys. Rep. {\bf 299}, 189 (1998).
585:
586: \bibitem{FA}Y.V. Fyodorov and Y. Alhassid,
587: Phys. Rev. {\bf A58}, R3375 (1998).
588:
589: \bibitem{OA}O. Agam,
590: Phys. Rev. {\bf A60}, R2633 (1999); preprint cond-mat/9906361.
591:
592:
593: \bibitem{shepel}G. Casati, G. Maspero, and D.L. Shepelyansky,
594: Phys. Rev. Lett. {\bf 82}, 524 (1999).
595:
596: \bibitem{HLK}B. Huckestein, R. Ketzmerick and C. Lewenkopf, preprint
597: cond-mat/9908090.
598:
599: \bibitem{rk}R. Ketzmerick,
600: Phys. Rev. {\bf B54}, 10841 (1996).
601:
602: \bibitem{Lai}Y.-Ch. Lai, R. Bl\"umel, E. Ott, and C. Grebogi,
603: Phys. Rev. Lett. {\bf 68}, 3491 (1992).
604:
605: \end{thebibliography}
606:
607:
608: \end{document}
609: