nlin0302037/la.tex
1: \documentstyle[preprint,eqsecnum,aps]{revtex}
2: %\def\btt#1{{\tt$\backslash$#1}}
3: %\def\BibTeX{\rm B{\sc ib}\TeX}
4: \begin{document}
5: 
6: %\documentstyle{article}
7: %\begin{document}
8: %\baselineskip 24pt
9: 
10: %\magnification=\magstep1
11: %\overfullrule=0pt
12: %\baselineskip=18pt
13: 
14: 
15: \draft
16: \title{Direct Scattering Processes  \\
17: and Signatures of Chaos in Quantum Waveguides}
18: \author{G. B. Akguc and L. E. Reichl}
19: \address{ Center for Studies in Statistical Mechanics and Complex Systems\\
20: The University of Texas at Austin\\
21: Austin, Texas 78712
22: }
23: \date{\today}
24: \maketitle
25: \begin{abstract}
26: 
27: The effect of direct processes on the statistical
28: properties of deterministic scattering processes in a chaotic 
29: waveguide is examined.
30: The single channel Poisson
31: kernel describes well the distribution of S-matrix eigenphases when
32: evaluated over an
33: energy interval. When direct processes are transformed away, the scattering
34: processes
35: exhibit universal random matrix behavior. The effect of chaos on scattering
36: wavefunctions,eigenphases, and time delays is discussed.
37: 
38: \end{abstract}
39: \pacs{05.45.Mt, 05.60.Gg, 73.23.Ad, 73.50.Bk}
40: \narrowtext
41: \section{Introduction}
42: \label{sec:intro}
43: 
44: \bigskip
45: \bigskip
46: %-----------------------------------------------------------------------
47: 
48: 
49: In 1957, Wigner proposed the use of statistical measures to analyze
50:    complex nuclear scattering data \cite{kn:wigner}. It was soon found
51: that,  in some nuclear scattering data, the  spacing between
52: scattering resonances was distributed in a manner similar to the
53: spacing of energy
54: levels of a Gaussian random Hamiltonian \cite{kn:gary}. In the 1960's,
55: extensive work was done to develop a systematic theory of the statistical
56: properties of
57: random hermitian matrices
58: \cite{kn:brody.etal,kn:mehta} and random unitary matrices \cite{kn:4,kn:hua}.
59: The general criterion for
60: constructing these random matrices is that they minimize information.
61: In 1979, the appearance of random matrix-like behavior in quantum
62: systems was linked to
63: underlying chaos in
64: the classical deterministic dynamics of these  systems
65: \cite{kn:macdonald,kn:boh,kn:casati,kn:berry}.
66:      Since then a large
67: body of work
68: has developed linking the statistical properties of bounded and open
69: quantum systems to
70: underlying deterministic chaos \cite{kn:reichl,kn:guhr}.
71: 
72: Historically there have been two different approaches to describing the
73: statistics of
74: quantum scattering processes in chaotic systems. One approach
75: \cite{kn:weiden} begins
76: directly with the Hamiltonian and uses it to build the scattering
77: matrix. The other
78: approach
79: \cite{kn:mello3,kn:mello2}, begins directly with the scattering matrix. In both
80: cases, random matrices are used to describe scattering processes and the
81: statistical
82: properties of the eigenphases of the scattering matrix and partial delay
83: times can be
84: obtained and
85: compared to experiment.  There has been considerable success in recent
86: years to link
87: the results and predictions of these two approaches \cite{kn:brouwer}.
88: 
89: The connection between the statistical properties of scattering processes and
90: underlying chaos is not straightforward because scattering events may
91: involve either
92: reactive processes or direct processes.
93: Reactive scattering processes are those for which an incident
94: particle becomes engaged
95: with the dynamics in the reaction region, and may be
96: delayed there for a considerable time. Direct processes
97: are those for which the particle passes through the reaction region without
98: becoming significantly
99: engaged in the reaction region. One prediction of the random
100: matrix theory (RMT) of
101: scattering is that, if the scattering process is truly chaotic, the average
102: value of the S-matrix will be zero. However, if direct processes are
103: present this
104: will not happen.
105: 
106: In this paper, we will study the deterministic scattering of a 
107: quantum particle in
108: a two
109: dimensional ballistic waveguide which has a classically chaotic cavity
110: formed by a
111: ripple billiard connected to a single lead at one end (see Figure 1). The
112: ripple
113: billiard is particularly well suited to the use of the reaction 
114: matrix theory approach
115: to scattering \cite{kn:eisenbud}, 
116: because a
117: simple coordinate transformation \cite{kn:luna,kn:akg1} allows us to
118: construct a
119: Hamiltonian matrix and thus an
120: eigenvalue equation for the basis states inside the cavity.
121: We will compare the results of deterministic scattering from the chaotic
122: cavity to some
123: recent predictions of random matrix theory as regards the scattering
124: process. The open
125: geometry of the waveguide in Fig. (1) ensures that direct processes will play
126: an important role in scattering. We show that the contribution of the
127: direct processes
128: to the scattering can be transformed out and the statistical properties of
129: the reactive
130: part of the scattering process can be compared to random matrix 
131: predictions. Our
132: discussion focuses on quantum particles and we will use parameters 
133: appropriate for
134: electrons in waveguides made with GaAs, for which a number of 
135: experiments have been done
136: \cite{kn:marcus1,kn:chang,kn:bird}. However, our results also apply 
137: to electro-magnetic
138: waves in flat microwave cavities, because the eigenmodes in these 
139: cavities satisfy a
140: Schrodinger-like equation \cite{kn:graf,kn:stein,kn:stockmann}.
141: 
142: 
143: We begin in Section (2), by reviewing the  reaction matrix theory of
144: deterministic
145: scattering in the ballistic waveguide and we study some properties of the
146: cavity basis
147: states. In Section (3), we study various statistical properties of the
148: eigenphases of
149: the waveguide scattering matrix (S-matrix). We show that, when the cavity
150: dynamics is
151: chaotic, the deterministic S-matrix eigenphases exhibit level repulsion and
152: their
153: distribution is well described by a Poisson kernel.  In Section (4), we
154: compare the
155: distribution of partial delay times for the deterministic scattering
156: process to the
157: predictions of random matrix theory. 
158: Finally, in Section (5) we give
159: concluding
160: remarks.
161: 
162: %
163: %
164: \section{Reaction matrix description of scattering}
165: %
166: %
167: 
168: We will explore the behavior of a particle of mass, $m$, in a ballistic
169: waveguide as it scatters from the ripple cavity shown in Fig. (1). A 
170: particle with
171: energy,
172: $E$, enters the cavity from the left  along a straight lead which
173: has infinitely hard walls. The particle wave is reflected back to the left
174: by an
175: infinitely hard
176: wall located at $x=L$. The dynamics inside the cavity,
177: $0<x<L$, can range from mixed to fully chaotic as the ripple amplitude is
178: varied.
179: The Schrodinger equation, which describes propagation of a particle wave,
180: $\Psi(x,y,t)$, in the waveguide at time, $t$, is given by
181: %
182: \begin{equation}
183: i{\hbar}{{\partial}{\Psi}(x,y,t)\over {\partial}t}={\hat H}{\Psi}(x,y,t)=
184: {\biggl[}-{{\hbar}^2\over 2m}{\biggl(}{{\partial}^2\over
185: {\partial}x^2}+{{\partial}^2\over
186: {\partial}y^2}{\biggr)}+V(x,y){\biggr]}{\Psi}(x,y,t),
187: \end{equation}
188: %
189: where ${\hat H}$ is the total Hamiltonian of the particle in the waveguide
190: and ${\hbar}$
191: is Planck's constant. The potential,
192: $V(x,y)$, has the following
193: properties: $V(x,y)={\infty}$ for $(L{\leq}x<\infty)$;
194: $V(x,0)=\infty$ for $(-\infty{\leq}x{\leq}L)$;
195: $V(x,y=g(x))=\infty$ for $(0<x<L)$; and
196: $V(x,y=d)=\infty$ for $(-\infty<x<0)$;
197: where $g(x)=d+{\rm a}~{\sin}(5{\pi}x/L)$ gives the contour of the ripple, $d$ is
198: the
199: average width of the cavity, $L$ is the length, and a is the ripple
200: amplitude. In all subsequent sections, we take the particle mass to be the
201: effective mass
202: of an electron in GaAs, $m=0.067m_e$, where $m_e$ is the free electron mass.
203: An energy eigenstate, $|E{\rangle}$, with energy, $E$, satisfies
204: the equation, ${\hat H}|E{\rangle}=E|E{\rangle}$, and evolves as
205: ${\Psi}(x,y,t)={\langle}x,y|E{\rangle}~{\rm e}^{-iEt/{\hbar}}$.
206: 
207: 
208: As shown in reference \cite{kn:akg1} the waveguide energy eigenstates can be
209: expressed in the form
210: %
211: \begin{equation}
212: {\langle}x,y|E{\rangle}={\sum_{j=1}^{\infty}}{\gamma}_j{\phi}_j(x,y)
213: +{\sum_{n=1}^{\infty}}{\Gamma}_n~{\Phi}_n(x,y).
214: \label{energystate}
215: \end{equation}
216: %
217: The states, ${\Phi}_n(x,y)$, are the basis states in the lead ($x{\leq}0$),
218: %
219: \begin{equation}
220: {\Phi}_n(x,y){\equiv}{\langle}x,y|{\Phi}_n{\rangle}=\sqrt{\frac{2}{d}}{\chi}_{n}
221: (x){\sin}{\biggl(}{n
222: \pi y\over d}{\biggr)}
223: \label{leadbasis}
224: \end{equation}
225: %
226: ($n=1,2,...,{\infty}$). These will consist of both propagating and
227: evanescent modes, as
228: we will discuss below. The states,
229: ${\phi}_j(x,y){\equiv}{\langle}x,y|{\phi}_j{\rangle}$ ($j=1,2,...,\infty$),
230: are the
231: basis states in the cavity ($0{\leq}x{\leq}L$). In practice, we truncate
232: the number of
233: cavity basis states to some large but finite number, $M$. The value of $M$
234: is determined
235: by the accuracy desired for the calculation.  The coefficients,
236: ${\Gamma}_n$ and
237: ${\gamma}_j$,  in Eq. (\ref{energystate}) are defined
238:       ${\Gamma}_n={\langle}{\Phi}_n|E{\rangle}$ and
239: ${\gamma}_j={\langle}{\phi}_j|E{\rangle}$.
240: 
241: 
242: A complete set of basis states, ${\phi_j}(x,y)$, inside the ripple cavity has
243: been obtained by solving the Schrodinger equation in the cavity, using Neumann
244: boundary conditions (${\bigr(}{d{\phi}_j\over dx}{\bigl)}_{x=0}=0$)
245: at the entrance
246: ($x=0$),
247: and Dirichlet boundary conditions, (${\phi}_j(x,y)=0$) at the walls. We
248: obtained the eigenfunctions, ${\phi_j}(x,y)$ and the associated eigenvalues,
249:        ${\lambda_j}$, using the transformation technique discussed in Ref.
250: \cite{kn:akg1},
251: Sect. V.
252: 
253: In Fig. (2.a), we show a Poincare surface of section  for
254: a classical particle in a closed ripple cavity with the same shape as 
255: in Fig. 1, and
256: with hard walls. We choose $d=101\AA$, $L=511\AA$, and ${\rm a}=1.0\AA$. We 
257: have plotted
258: Birkhoff coordinates, $p_x/p={\cos}(\alpha)$ versus $x$, each time
259: the particle  hits  the lower wall at point $x$ ($\alpha$ is the 
260: angle between the wall
261: and the momentum).  For these billiard dimensions, the classical 
262: phase space contains a
263: mixture of regular orbits, nonlinear resonances, and chaotic motion. 
264: If we increase the
265: ripple amplitude,
266: a, there is a range of values of
267: a for which the classical motion appears to become totally chaotic.  For the
268: scattering system (with one end of the cavity open) periodic orbits 
269: and nonlinear
270: resonance regions can't be reached classically by a particle that 
271: enters from the
272: left, but quantum mechanically tunneling into these regions is possible as
273: we will show Section IV (see also \cite{kn:her}).
274: In Figs.~(2.b-e), we show Husimi plots of quantum Poincare surfaces of section
275: \cite{kn:li} of  cavity basis states, ${\phi}_j(x,y)$, with eigenvalues
276: $\gamma_{983}=257.1934E_1$, $\gamma_{985}=257.9339E_1$,
277: $\gamma_{989}=258.6655E_1$, and $\gamma_{990}=258.9072E_1$, where
278: $E_1={{\hbar}^2{\pi}^2\over 2md^2}$. The  Husumi plots in Fig.~(2.c) and Fig.~(2.d) show
279: that these bases states reside on nonlinear resonances of the underlying classical phase
280: space. In Section IV, we will show that these bases states give the primary contribution
281: to sharp resonances in the transmission at these cavity parameters.
282: 
283: 
284: 
285: Inside the lead,  we must distinguish between propagating and
286: evanescent modes. The longitudinal component of the eigenstates in the
287: leads, for
288: propagating modes, can be written
289: %
290: \begin{equation}
291: {\Gamma}_n{\chi}_n(x)=\frac{a_n}{\sqrt{k_n}}
292: {\rm e}^{-ik_nx}+\frac{b_n}{\sqrt{k_n}}{\rm e}^{ik_nx},
293: \label{eq:chan1}
294: \end{equation}
295: %
296: where the wavevector, $k_n$ is given by $k_n=\sqrt{{2mE\over
297: {\hbar}^2}-{\bigl(}{n{\pi}\over d}{\bigr)}^2}$.
298: If there are $N$ propagating modes then $n=1,...,N$. Here we use a
299: unit current normalization. A particle propagating in the $n^{th}$ channel
300: has energy in
301: the interval, $n^2E_1{\leq}E{\leq}(n+1)^2E_1$, where
302: $E_1={{\hbar}^2{\pi}^2\over
303: 2md^2}=0.0738$eV.
304: All remaining modes,
305: $n=N+1,...,\infty$,  are evanescent and can be represented in terms of  the
306: states,
307: %
308: \begin{equation}
309: {\Gamma}_n{\chi}_n(x)=\frac{c_n}{\sqrt{{\kappa}_n}}{\rm e}^{-{\kappa}_n|x|},
310: \label{eq:chan3}
311: \end{equation}
312: %
313: where ${\kappa}_n=\sqrt{{\bigl(}{n{\pi}\over d}{\bigr)}^2-{2mE\over
314: {\hbar}^2}}$.
315: In Ref. \cite{kn:akg1},
316: we showed
317: that the evanescent modes for this system do not make a significant
318: contribution to the
319: scattering properties that we are interested in. Therefore we will neglect the
320: evanescent modes in subsequent sections.
321: 
322: 
323: As shown in Ref. \cite{kn:akg1}, The eigenvalue equation,
324: ${\hat H}|E{\rangle}=E|E{\rangle}$, leads to the relation,
325: %
326: \begin{equation}
327: {\gamma}_j={{\hbar}^2\over 2m}{1\over
328: (E-{\lambda}_j)} {\sum_{n=1}^N}~{\phi}^{*}_{j,n}(0){\biggl(}{d{\chi}_n\over
329: dx}{\biggr)}_{x=0}{\Gamma}_{n}.
330: \label{eq:reac1}
331: \end{equation}
332: %
333: Also, continuity of energy eigenstates at $x=0$ gives
334: %
335: \begin{equation}
336: {\Gamma}_n{\chi}_n(0)={\sum_{j=1}^M}{\gamma}_j{\phi}_{j,n}(0)=
337: {\sum_{n'=1}^N}~{R}_{n,n'}~{\biggl(}{d{\chi}_n\over
338: dx}{\biggr)}_{x=0}{\Gamma}_{n'}.
339: \label{eq:reac2}
340: \end{equation}
341: %
342: where
343: %
344: \begin{equation}
345: {R}_{n,n'}={{\hbar}^2\over 2m}{\sum_{j=1}^M}
346: {{\phi}_{j,n'}^*(0){\phi}_{j,n}(0)\over (E-{\lambda}_j)}
347: \label{eq:reac3}
348: \end{equation}
349: %
350: is the $(n,n')^{th}$ matrix element of the reaction matrix  \cite{kn:eisenbud}.
351: The quantity, ${\phi}_{j,n}(0)$ is a measure of the overlap between
352: the  $j^{th}$
353: cavity state, and the $n^{th}$ channel in the lead, evaluated at the interface,
354: %
355: \begin{equation}
356: {\phi}_{j,n}(0)=\sqrt{2\over
357: d}{\int_{0}^d}~dy~{\phi}_{j}(0,y)~{\sin}{\bigl(}{n{\pi}y\over d}{\bigr)}.
358: \label{eq:reac4}
359: \end{equation}
360: %
361: 
362: 
363: Let us now form an $N{\times}1$ column matrix, ${\bar b}$ (${\bar a}$)
364: whose matrix
365: elements consist of the $N$ probability amplitudes ${\{}b_n{\}}$
366: (${\{}a_n{\}}$) of the
367: outgoing (incoming) {\it propagating} modes.   The waveguide scattering matrix
368: (S-matrix), ${\bar S}$, is a $N{\times}N$ matrix which connects the
369: incoming propagating
370: modes to the outgoing propagating modes, ${\bar b}={\bar S}{\cdot}{\bar a}$.
371: 
372: For the case when there are $N$ propagating modes in the lead we can obtain an
373: $N{\times}N$ S-matrix which may be written
374: %
375: \begin{equation}
376: {\bar S}=-{({\bar 1}_N-i{\bar {\rm K}})
377: \over ({\bar 1}_N+i{\bar {\rm K}})},
378: \label{eq:scat1}
379: \end{equation}
380: %
381: where ${\bar 1}_N$ is $N{\times}N$ unit matrix, and the $N{\times}N$
382: matrix, ${\bar {\rm
383: K}}$, has matrix elements,  ${\bar {\rm
384: K}}_{n,n'}=\sqrt{k_n}R_{n,n'}\sqrt{k_{n'}}$ and can be written
385: %
386: \begin{equation}
387: {\bar {\rm K}}={\bar w}^\dagger{\cdot} \frac{1}{E{\bar 1}_M-{\bar
388: H}_{in}}{\cdot}{\bar w}.
389: \label{eq:scat1.a}
390: \end{equation}
391: %
392: In Eq. (\ref{eq:scat1.a}),  ${\bar 1}_M$ is the $M{\times}M$ unit matrix,
393: ${\bar H}_{in}$
394: is an $M{\times}M$ diagonal matrix  formed with the eigenvalues,
395: $\lambda_j$ ($j=1,...M$)
396: in the cavity, and ${\bar w}$ is an $M{\times}N$ coupling matrix,
397: %
398: \begin{equation}
399: {\bar w}\equiv \pmatrix{w_{1,1}& \ldots& w_{1,{N}}\cr
400: \vdots&&\vdots \cr w_{M,1}& \ldots& w_{M,N}\cr },
401: \label{eq:scat2}
402: \end{equation}
403: %
404: where $w_{j,n'}=\phi_{j,n'}(0)\sqrt{k_{n'}}$.   With some algebra, the
405: S-matrix can also
406: be written  in the form
407: %
408: \begin{eqnarray}
409: {\bar S}=-{\biggl(}{\bar 1}_N-2i{\bar w}^\dagger {\cdot}\frac{1}{E{\bar
410: 1}_M-{\bar
411: H}_{in}+i{\bar w}{\cdot}{\bar
412: w}^\dagger}{\cdot}{\bar
413: w}{\biggr)}.
414: \label{eq:scat3}
415: \end{eqnarray}
416: %
417: In Ref. \cite{kn:akg1}, we showed that, if evanescent modes are included, an
418: additional term appears in the denominator in Eq. (\ref{eq:scat3}).
419: 
420: 
421: 
422: 
423: The reaction matrix approach to waveguide scattering
424: provides a very efficient means of computing the statistical properties of the
425: scattering process because the Schrodinger equation only needs to be solved
426: once to
427: obtain the basis states and eigenvalues in the cavity. Using these values,
428: the S-matrix
429: can then be obtained at all other particle energies, $E$. Typically for the
430: ripple cavity in
431: Figure (1), we can obtain the scattering matrix at $10^5$ different values
432: of incident
433: energy in a reasonable amount of time on a Cray machine.
434: 
435: 
436: One of the goals of this paper is to compare the statistical properties of the
437: deterministic scattering process in the ripple cavity to statistical
438: properties of a
439: hypothetical scattering process in which ${\hat H}_{in}$ is replaced by a diagonal
440: matrix, ${\hat H}_{goe}$, composed of the $M$ eigenvalues of an
441: $M{\times}M$
442: Gaussian Orthogonal Ensemble (GOE) Hamiltonian, ${\hat H}_{goe}'$, and the
443: $N$ columns of
444: the coupling matrix, ${\bar w}$, are replaced by $N$ of the $M$ normalized
445: eigenvectors
446: of ${\hat H}_{goe}'$ to yield a coupling matrix, ${\hat w}_{goe}$
447: \cite{kn:brody,kn:mehta}. In this random matrix theory approach, the strength of the
448: coupling between the cavity and the lead is given by the phenomenological
449: parameter,
450: $g$. The parameter, $g$, does not appear in the deterministic scattering process. For
451: deterministic scattering the strength of the coupling is entirely determined by the
452: geometry and the potentials at the interface. The  scattering matrix, obtain from RMT,  can
453: then be written
454: %
455: \begin{eqnarray}
456: {\bar S}_{goe}=-{\biggl(}{\bar 1}_N-2ig{\bar w}_{goe}^\dagger
457: {\cdot}\frac{1}{E{\bar
458: 1}_M-{\bar H}_{goe}+ig{\bar w}_{goe}{\cdot}{\bar
459: w}_{goe}^\dagger}{\cdot}{\bar
460: w}_{goe}{\biggr)}.
461: \label{eq:scatgoe}
462: \end{eqnarray}
463: %
464: It was shown in Ref. (\cite{kn:vaarbar}), using supersymmetry
465: techniques, that for
466: the case when the distribution of energy eigenvalues of ${\hat H}_{goe}$
467: is centered at $E=0$ and
468: $M{\rightarrow}\infty$, the average S-matrix can be written
469: %
470: \begin{equation}
471: {\langle}{\bar S}_{goe}{\rangle}=s{\bar 1}_N
472: ~~~{\rm with}~~~s=\frac{1-g[iE/2+\pi\nu(E)]}
473:                     {1+g[iE/2+\pi\nu(E)]},
474: \label{eq:avsmat2}
475: \end{equation}
476: %
477: where $\nu(E)={\pi}^{-1}\sqrt{1-(E/2)^2}$ is the average density of energy
478: eigenstates. It is useful to introduce the quantity
479: %
480: \begin{equation}
481: \mu={\mu}_r+i{\mu}_i={1-s\over 1+s^*}=g{\nu}{\pi}+ig{E\over 2},
482: \label{eq:avsmat3}
483: \end{equation}
484: %
485: where ${\mu}_r=g{\nu}{\pi}$ and ${\mu}_i=g{E\over 2}$,
486: respectively, are
487: the real and imaginary parts of $\mu$.  The case when
488: $g=1$, corresponds to ideal coupling. In the neighborhood of
489: $E{\approx}0$, the
490: eigenvalues of ${\hat H}_{goe}$ have a constant density, ${1\over 2\pi}$,
491: and the
492: average S-matrix ${\langle}{\bar S}_{goe}{\rangle}=0$. When
493: $g{\neq}1$, the
494: average S-matrix cannot be zero.
495: 
496: 
497: %
498: %
499: %
500: \section{Eigenphases of the Scattering Matrix }
501: %
502: %
503: %
504: 
505: We have analyzed some of the statistical properties of the eigenphases of
506: the S-matrix
507: for the case of  deterministic  scattering from the ripple cavity for the
508: cases when the
509: internal dynamics in the cavity is completely chaotic and when it is
510: near-integrable. In
511: this section, we consider the
512: energy interval  $256E_1{\leq}E{\leq}289E_1$ when 16 channels are present
513: in the lead.
514: The S-matrix is a $16{\times}16$ matrix, and for each value of incident
515: energy it has 16
516: eigenvalues, ${\rm e}^{i{\delta}_{\alpha}}$ (${\alpha}=1,...,16$) and 16
517: eigenfunctions, $|{\delta}_{\alpha}{\rangle}$ (${\bar
518: S}|{\delta}_{\alpha}{\rangle}=
519: {\rm e}^{i{\delta}_{\alpha}}|{\delta}_{\alpha}{\rangle}$). The S-matrix is
520: unitary so
521: the eigenfunctions, $|{\delta}_{\alpha}{\rangle}$, form a complete
522: orthonormal set.
523: We can use the orthonormality of the eigenfunctions to follow each eigenphase,
524: ${\delta}_{\alpha}$, continuously as a function of energy
525: \cite{kn:akg}. The eigenfunctions, for two S-matrices evaluated at nearby
526: energies, will
527: be approximately orthogonal if they do not belong to the same eigenphase.
528: Thus we
529: can plot each of the 16 different eigenphases as a function of the 
530: incident energy.
531: These are
532: shown in Figure (3) where
533: the eigenphases, which are defined mod $2\pi$, are ``unwrapped" and allowed
534: to evolve
535: continuously as a function of energy. In Fig. (3.a), we show the case with
536: ripple
537: amplitude, ${\rm a}=25\AA$, where the classical cavity dynamics is chaotic, and in Fig.
538: (3.b) we show the
539: case ${\rm a}=1\AA$ where the classical cavity dynamics is mixed (see Fig. (2)). The
540: case of mixed dynamics shows many more
541: abrupt changes of phase as a function of energy than the chaotic case. This is due to
542: the fact that the mixed dynamics has many long lived resonances not found in the chaotic
543: case. This was also seen in Ref. \cite{kn:akg,kn:her}. We shall return to 
544: this feature in
545: Section (IV).
546: 
547: Below we first discuss the effect of direct processes on the distribution of
548: eigenphases, and  then we determine the distribution of nearest 
549: neighbor spacings of
550: these eigenphases.
551: 
552: 
553: %
554: %
555: %
556: \subsection{Distribution of Eigenphases }
557: %
558: %
559: %
560: 
561:       When a scattering process has a non-zero average S-matrix, ${\langle}\bar
562: S{\rangle}$, it  indicates that direct processes may play a 
563: significant role in the
564: scattering process. Direct processes are generally scattering events which
565: do not interact significantly with the reaction region (cavity) 
566: \cite{kn:mello4}. When
567: direct processes are present, the distribution of S-matrix elements 
568: that minimizes
569: information about the scattering process is the Poisson kernel. For 
570: the case of an $N$
571: channel process whose dynamics is time reversal invariant, the 
572: Poisson kernel has the
573: form,
574: %
575: \begin{equation}
576: P_N({\bar S})=\frac{1}{{\Omega}} \frac{[{\rm
577: Det}(1-{\langle}S{\rangle}^{*}{\langle}S{\rangle}]^{(N+1)/2}} {|{\rm
578: Det}(1-{\langle}S{\rangle}^{*}{\bar S})|^{(N+1)}},
579: \label{eq:poiskern1}
580: \end{equation}
581: %
582: where ${\Omega}$ is a normalization factor that ensures that the Poisson kernel
583: satisfies the normalization condition, ${\int}d{\bar S}P_N({\bar S})=1$.
584: 
585: The S-matrix  can be diagonalized by a unitary matrix, ${\bar U}$,
586: and,  as mentioned
587: earlier, the eigenvalues
588: of the S-matrix are denoted, ${\rm e}^{i{\delta}_{\alpha}}$,
589: ${\alpha}=1,...,N$.  In
590: terms of the eigenphases, ${\delta}_{\alpha}$, the normalization condition
591: for the
592: Poisson kernel, Eq. (\ref{eq:poiskern1}), can be written
593: %
594: \begin{eqnarray}
595: {\int}d{\bar S}P_N({\bar S})={\int}...{\int}d{\delta}_1...d{\delta}_N~
596: P_N({\delta}_1,...,{\delta}_N)\nonumber\\
597: =\frac{1}{{\Omega}_U}{\int}...{\int}d{\delta}_1...d{\delta}_N~
598: {\times}{\prod_{1{\leq}{\alpha}<{\alpha}'{\leq}N}}|{\rm
599: e}^{i{\delta}_{\alpha}}-{\rm
600: e}^{i{\delta}_{\alpha}'}|\nonumber\\
601: {\times}{\biggl(}{(1-s^*s)^N\over
602: {\prod_{{\alpha}=1}^N}(1-s^*{\rm e}^{i{\delta}_{\alpha}}) (1-s{\rm
603: e}^{-i{\delta}_{{\alpha}'}})}{\biggr)}^{(N+1)/2}=1,~~
604: \label{poiskern2}
605: \end{eqnarray}
606: %
607: where ${\Omega}_U$ is a normalization constant.
608: Note that $P_N({\delta}_1,...,{\delta}_N)$ is the joint probability density
609: to find the
610: angles, ${\delta}_{\alpha}$, in the intervals
611: ${\delta}_{\alpha}{\rightarrow}{\delta}_{\alpha}+d{\delta}_{\alpha}$,
612: (${\alpha}=1,...,N$).
613: 
614: In Ref. \cite{kn:fyo2}, it is shown   that if the following change of angles is
615: introduced
616: %
617: \begin{equation}
618: {\tan}(\frac{{\theta}_{\alpha}}{2})={1\over
619: g{\pi}{\nu}}{\biggl(}{\tan}(\frac{{\delta}_{\alpha}}{2})+g{E\over
620: 2}{\biggr)},
621: \label{eq:avsmat}
622: \end{equation}
623: %
624: and if one assumes ideal coupling, $g=1$, then Eq. (\ref{poiskern2})
625: reduces to
626: %
627: \begin{eqnarray}
628: {\int}d{\bar S}P_N({\bar
629: S})={\int}...{\int}d{\theta}_1...d{\theta}_N~P_N({\theta}_1,...,{\theta}_N)~~~~~~~
630: ~~\nonumber\\
631: =\frac{1}{{\Omega}_U}{\int}...{\int}d{\theta}_1...d{\theta}_N~
632: {\prod_{1{\leq}{\alpha}<{\alpha}'{\leq}N}}|{\rm e}^{i{\theta}_{\alpha}}-{\rm
633: e}^{i{\theta}_{{\alpha}'}}|,
634: \label{poiskern3}
635: \end{eqnarray}
636: %
637: which is just the distribution for the Circular Orthogonal Ensemble (COE)
638:  \cite{kn:4}, \cite{kn:hua}.  Thus, even for scattering processes which
639: include direct
640: processes, it is possible in principle to transform away the direct
641: processes and compare
642: the eigenphase distribution with that of COE (note that similar
643: ideas first appeared in literature in Ref.~\cite{kn:mello2}). It is important to note that
644: the transformation that removes direct processes is not the same as the unfolding process
645: that occurs on the energy spectrum of bounded systems to give a constant average density. 
646: 
647: In our subsequent analysis, the case of scattering with only a single
648: channel will be  useful for analyzing data. For single channel scattering ($N=1$),  the
649: S-matrix reduces to the complex
650: function, $S={\rm e}^{i\delta}$, and  the Poisson kernel, reduces to \cite{kn:mello3}
651: %
652: \begin{equation}
653: P_1{s}(\delta)=\frac{1}{2\pi} \frac{[(1-s^{*}s]}
654: {|(1-s^{*}{\rm e}^{i\delta})|^{2}},
655: \label{eq:poiskern4}
656: \end{equation}
657: %
658: with normalization condition, ${\int_{-\pi}^{\pi}}d{\delta}P_1(\delta)=1$.
659: Under the
660: transformation above, $P_1(\delta){\rightarrow}P_1(\theta)=1/2\pi$,
661: which is the COE prediction.
662: 
663: Having obtained numerical values of the eigenvalues,
664: ${\rm e}^{i{\delta}_{\alpha}}$, as a function of energy,  we can compute
665: an average
666: value for each of the  16 eigenvalues,
667: %
668: \begin{equation}
669: s_{\alpha}={\langle}{\rm e}^{i{\delta}_{\alpha}}{\rangle}={1\over
670: \eta}{\sum_{k=1}^{\eta}}{\rm
671: e}^{i{\delta}_{\alpha}(E_k)},
672: \label{eq:poiskern5}
673: \end{equation}
674: %
675: where ${\eta}$ is the number of energy values used. The apparently continuous
676: eigenphase curves actually consist of about
677: $40,000$ discrete energy points. The approximate orthonormality of S-matrix 
678: eigenvectors for neighboring energies has been used to sort the eigenphases. 
679: Thus, the eigenphases and eigenvectors have an energy interval over which they are
680: correlated, and we have used that fact in our sorting process. On the other hand, this 
681: correlation of the S-matrices at neighboring energies can prevent us from obtaining
682: statistics that can be compared to RMT predictions. Comparison to RMT requires
683: use of independent data points. Therefore, in order to study the statistical properties
684: of the eigenphases, we must choose values of the eigenphases separated in energy a
685: distance greater than the correlation length. For each eigenphase curve we select points
686: which have an energy spacing, 
687: $\Delta E=0.495E_1$. We choose this spacing based on an analysis of the delay time
688: correlation discussed in Section IV. (The delay time auto-correlation function is the second
689: derivative of the eigenphase auto-correlation function.) 
690: 
691:       We have computed a histogram of the number of eigenphases,
692: $N(\delta)={\eta}P_N(\delta)$, versus value of eigenphase,
693: ${\delta}_{\alpha}$, in the 16 channel region, where ${\eta}$ is the number of data
694: points. In order to improve the statistics, we use data from four different ripple
695: amplitudes, 
696: ${\rm a}=25\AA, ~30\AA,~35\AA,45\AA$, all of which lie in the chaotic regime. All
697: eigenphases lie in the energy interval, $256<E/E_1<289$ and have energy spacing, $\Delta
698: E=0.495E_1$. Thus the histogram includes $67 X 16 X 4=4288$ data points.
699: We have found that the distribution of eigenphases, along a given eigenphase curve, is well
700: described by the Poisson kernel for the single channel case.  We proceed as follows. We 
701: compute the average eigenvalue, $s_{\alpha}={\langle}{\rm
702: e}^{i{\delta}_{\alpha}}{\rangle}$, for each eigenphase curve.   
703: We can form a histogram using the 67 data points from a single eigenphase curve. We do
704: this for each of the 64 eigenphase curves, using the same number of bins and bin width
705: for each curve. We then add these 64 histograms together to form a single histogram,
706: which is   shown in Fig. (4.a).  The solid line in Fig.
707: (4.a), is the single channel Poisson kernel, $P_1(\delta)$, but with
708: ${\langle}s{\rangle}={1\over 64}{\sum_{\alpha=1}^{64}}s_{\alpha}$.
709: We can use Eq. (\ref{eq:avsmat}) to
710: transform away the effects of the direct interactions
711: and find the
712: distribution of the transformed angles, ${\theta}_{\alpha}$. If we use Eq.
713: (\ref{eq:avsmat3}) and
714: obtain ${\mu}$ from the numerically calculated values of
715: $s_{\alpha}={\langle}{\rm
716: e}^{i{\delta}_{\alpha}}{\rangle}$, then  the transformed angles,
717: ${\theta}_{\alpha}$,
718: are given by
719: %
720: \begin{equation}
721: {\tan}(\frac{{\theta}_{\alpha}}{2})={1\over
722: {\mu}_r}{\biggl(}{\tan}(\frac{{\delta}_{\alpha}}{2})+{\mu}_i{\biggr)},
723: \label{eq:avsmat4}
724: \end{equation}
725: %
726: with ${\mu}_r$ and ${\mu}_i$ computed numerically from $s_{\alpha}$.
727: In Fig.~(4.b), we show how the histogram in Fig. (4.a) changes if we
728: transform each
729: eigenphase, ${\delta}_{\alpha}$, using Eq. (\ref{eq:avsmat4}). In this case
730: distribution
731: is approximately constant and equal to the area,
732: $\frac{N}{2\pi}$.  Thus, having transformed away the contribution
733: from the direct
734: processes, we obtain the COE eigenphase distribution for this chaotic
735: scattering
736: process, with fairly high confidence level.
737: 
738: For ${\rm a}=1\AA$ the plots of the eigenphase distributions look very similar
739: to the chaotic case shown in Fig. (4), and it appears that the
740: eigenphase distribution is not as sensitive an indicator for
741: underlying chaos as is
742: the nearest
743: neighbor spacing distribution, at least with this type of analysis.
744: 
745: 
746: %
747: %
748: %
749: \subsection{Nearest Neighbor Eigenphase Spacing }
750: %
751: %
752: %
753: 
754: In this section we consider the nearest neighbor spacings between eigenphases,
755: ${\delta}_{\alpha}$, of the
756: scattering matrix
757: for the 16 channel case in the energy interval $256<E/E_1<289$. For any
758: given value of the energy, the S-matrix only has 16 eigenphases.
759: However, we can form a
760: histogram of
761: nearest neighbor eigenphase spacings if we obtain eigenphase spacings for a
762: sequence of different energies in the range,
763: $256<E/E_1<289$.
764: In Fig. (5.a) we show histograms of 1005 (15X67)
765: nearest neighbor spacings for eigenphases
766: computed at energy increments,
767: ${\Delta}E=0.495E_1$ and obtained by averaging over histograms
768:  for each ripple amplitude, ${\rm a}=25\AA,~30\AA,~35\AA,~45\AA$.
769: We have fit the histogram to the Brody distribution,
770: \cite{kn:brody}
771: %
772: \begin{equation}
773: P_B({\sigma})=A{\biggl(}{{\sigma}\over
774: {\langle}{\sigma}{\rangle}}{\biggr)}^{\beta}\exp{\biggl[}-{\xi}
775: {\biggl(}{{\sigma}\over {\langle}{\sigma}{\rangle}}{\biggr)}^{1+\beta}{\biggr]}
776: ~~~{\rm with}~~~
777: \xi={\biggl[}{1\over
778: {\langle}{\sigma}{\rangle}}\Gamma{\biggl(}\frac{2+\beta}{1+\beta}{\biggr)}{\biggr]}^{1+\beta},
779: \end{equation}
780: %
781: ${\langle}{\sigma}{\rangle}$ is the average spacing between nearest neighbor
782: eigenphases, and ${\Gamma}(x)$ is the Gamma function. In Fig. (5.a), the solid line
783: is a fit to the Brody distribution for $\beta=0.635$.
784: In Fig.~(5.b) same calculation is performed after direct processes were
785: transformed away. In this case $\beta=0.865$.
786: Note that the GOE prediction ($\beta=0.95$) for the closed system eigenvalue
787: spacings is fairly close to our value of $\beta$, after the effects of 
788: direct processes are transformed  away. In Ref. \cite{kn:li}, the nearest neighbor
789: energy  eigenvalue spacings for a closed ripple billiard were fit to the Brody
790: distribution with  $\beta=0.806$. In that case, the deviation from GOE predictions was
791: found to be due to  bouncing ball orbits. Our result also contains bouncing ball
792: contributions.  It is  useful to note that in our scattering system there is no long
793: range energy correlation  for 16 channel region (we have explicitly removed energy
794: correlations by taking data points  at large energy increments), in contrast to the case
795: reported in reference \cite{kn:dietz}
796: 
797: 
798: We also obtained a nearest neighbor spacing histogram  for the case with mixed
799: phase space in the 16 channel energy interval.  In
800: Fig.~(6.a) we show the histogram of nearest neighbor eigenphase spacings for
801: ${\rm a}=0.5\AA,~ 1\AA,~ 2\AA,~ 3\AA,~ 4\AA,~ 5\AA$ in the
802: energy interval $256<E/E_1<289$ before direct process are transformed away.
803: We use an energy spacing ${\Delta}E=0.495E_1$ and obtain 1005 data points for 
804: each amplitude. We then average over the histograms  for the six ripple amplitudes.
805: The solid line is a fit to the Brody
806: distribution for ${\beta}=0.2$. In Fig.~(6.b), the same histogram is 
807: shown after
808: direct processes are transformed away. The Brody parameter in this case,
809: ${\beta}=0.116$,
810: which is closer to a Poisson distribution (the Brody
811: distribution
812: becomes a Poisson distribution for $\beta=0$).
813: 
814: The distribution of nearest
815: neighbor eigenphase spacings has been computed for a energy independent
816: scattering matrix in \cite{kn:jung} for a very different physical 
817: system. They also
818: report close agreement with COE predictions for the chaotic region, 
819: although they do
820: not have to deal with direct processes. It is clear that direct 
821: processes can play an
822: important role in causing deviations from random matrix theory predictions for
823: scattering processes.
824: 
825: 
826: 
827: %
828: %
829: %
830: \section{Partial Delay Times}
831: %
832: %
833: %
834: 
835: In this section we compare the partial delay time distribution, computed
836: for the
837: deterministic scattering process, to values obtained  from random
838: matrix theory.  The partial delay times are given by the energy derivative of
839: eigenphases,
840: $\tau_{\alpha}=\hbar{d{\delta}_{\alpha}\over dE}$~\cite{kn:bohm}.
841: The average partial
842: delay time
843: density  for a scattering process governed by the scattering matrix,
844: ${\bar
845: S}_{goe}$, has been computed by \cite{kn:fyo1,kn:fyo2}, using supersymmetry
846: techniques, and is given by,
847: %
848: \begin{equation}
849: {\rho}(\tau)=(1/N)\sum_{\alpha}{\langle}{\delta}(\tau
850: -\tau_{\alpha}){\rangle}_{goe}=
851: \frac{(1/2)^{N/2}}{\Gamma(N/2)}\frac{\exp(-1/2(\tau/{\langle}\tau{\rangle}))}{(\tau/{\langle}\tau{\rangle})^{N/2+2}}
852: \label{eq:td1}
853: \end{equation}
854: %
855: where ${\langle}\tau{\rangle}=1/N$.
856: 
857: Before showing the distribution obtained for the partial delay times, 
858: it is useful to
859: discuss energy correlations contained in the partial  delay time curves.
860: In Fig.~7, we show the auto-correlation function for the partial 
861: delay times obtained
862: in the 16 channel energy regime $256<E/E_1<289$ and averaged over
863: 6 different ripple sizes, ${\rm a}=22\AA,~23\AA,~24\AA,~25\AA,~26\AA,~27\AA$. For each
864: partial delay time curve we obtain a auto-correlation function, and then we  average
865: over all 96 curves. 
866: We also show the GOE  prediction for the 16 channel case as well as the partial delay
867: time auto-correlation function for
868: the near integrable regime (${\rm a}=1\AA$). The GOE prediction is obtained after
869: performing the triple integration given in reference \cite{kn:leh}. The energy scale is
870: adjusted to correspond to the relevant scale for our data.  We also note the partial
871: delay time auto-correlation function is  the second derivative of the eigenphase
872:  auto-correlation function. Therefore the eigenphase auto-correlation function decays more slowly
873: than the partial time delay auto-correlation function.
874: 
875: In Fig.~(8.a), we show a histogram of the scaled partial delay times,
876: $\tau/{\langle}\tau{\rangle}$.
877: We again
878: consider  the energy
879: regime with 16 channels and vary the energy in the
880: interval, $256E_1<E<289E_1$. To obtain enough values  to build
881: good statistics,
882: we use 4 different ripple sizes, ${\rm a}=25\AA,~30\AA,~35\AA,~45\AA$.
883: For these ripple
884: amplitudes, the ripple cavity dynamics is chaotic.   We
885: used 100  energy points per specific ripple size, and therefore an 
886: energy increment
887: of $\Delta E=0.33E_1$. The average, delay time, ${\langle}\tau{\rangle}$, is 
888: obtained numerically
889: for each partial  delay time curve. Then the histograms for 64
890: scaled partial delay times are combined into one histogram by simply 
891: adding values in the corresponding bins.
892:   The solid line in Fig.~(8.a), is a plot of the RMT prediction, 
893: $N(\tau)={\eta}'{\rho}(\tau)$, where ${\eta}'$ is the area under the curve and 
894: ${\langle}\tau{\rangle}=1/N$. The agreement
895: is not good because our data contains the effect of
896: direct scattering processes. In Fig.~(8.b) we show the partial delay time
897: density obtained
898: from the eigenphases, ${\theta}_{\alpha}$, which no longer contain the
899: effect of direct
900: scattering processes. The solid line is a plot of $N(\tau)$ with
901: ${\langle}\tau{\rangle}=1/N$ \cite{kn:fyo1}. The agreement is
902: very good.  Finally in
903: Fig.~(8.c) we plot histogram of 4000  partial delay times obtained from
904:  a $16{\times}16$ ${\bar S}_{goe}$ by using different realizations.
905: Again, the agreement between the data and Eq. (\ref{eq:td1})  is very good.
906:     Thus, after the removal of the effects of direct
907: scattering processes
908: our deterministic scattering from the chaotic ripple cavity behaves very
909: much like the
910: RMT prediction.
911: (It is useful to note that in Ref. \cite{kn:akg1}, we compared the
912: {\it Wigner-Smith} delay
913: time distribution with numerically computed predictions of RMT. The
914: Wigner-Smith delay
915: time is defined,
916: ${\tau}_{ws}={1\over
917: N}{\sum_{\alpha=1}^N}{\tau}_{\alpha}$.)
918: 
919: 
920: In Fig. 9, we show the delay time distributions
921: for the near integrable case, ${\rm a}=1\AA$.  We have used energy increments,
922: ${\Delta}E=0.1E_1$ justified from Fig.~7.
923:   The delay time distribution for the near integrable case deviates significantly from the  random
924: matrix result and the results for chaotic cavity shown in Fig. (8).
925: 
926: Let us now return to the eigenphase curve in Fig. (3.b). We see that 
927: the curves for the near integrable case have a sequence of fairly abrupt large changes 
928: of phase. These are due to resonance structures that cause larger than average delays of 
929: the particle in the cavity. In Fig. (10), we plot the Wigner-Smith delay time (which is 
930: an average over all partial delay times) in the energy interval, 
931: $257.4{\leq}E/E_1{\leq}259.5$. This energy interval contains two of the large phase
932: changes in the eigenphase  curves in Fig. (3.b).
933: We see that each large phase change gives rise to a large peak in the 
934: delay time. The crosses in Fig. (10) give the energies of the cavity basis states, 
935: $\lambda_j$, in that energy interval. There appears to be one cavity state which lies at 
936: each resonance energy. In Fig. (10), we have also plotted the configuration space 
937: distribution of four of the cavity eigenstates, two at resonance and two off resonance.
938: In  Figs. (2.b)-(2.e), we have shown Husimi plots of the quantum Poincare surface of 
939: section for each of these four states. The two states giving rise  to the delay time
940: resonance  peaks lie in the dominant nonlinear resonance structures in the classical
941: phase space.  The quantum particle appears to tunnel into these dynamical resonance
942: structures,  and is delayed there for a considerable length of time.
943: 
944: 
945: 
946: 
947: 
948: 
949: 
950: 
951: 
952: %
953: %
954: \section{Conclusion}
955: %
956: %
957: 
958: 
959: We have analyzed the statistical properties of a scattering
960: process in a waveguide with a cavity which allows a range of 
961: dynamics, including
962: integrable, mixed, or chaotic. In this waveguide, direct processes 
963: also play an important
964: role.  The ``ripple" cavity that we use has the special feature that 
965: it allows us to form
966: a Hamiltonian matrix to describe the dynamics interior to the cavity. 
967: This, in turn,
968: allows us to use the reaction matrix approach to scattering for our 
969: deterministic
970: scattering process.  The reaction matrix approach is one of the most 
971: efficient methods
972: for obtaining the large amounts of data necessary to obtain good statistics.
973: Until now, mesh based models (like the boundary element method, finite 
974: element method, or
975: recursive Green's function method) were the main numerical methods to deal with
976: scattering problems. However, these methods use an energy dependent 
977: boundary condition
978: which makes it a formidable task to obtain solutions for very large 
979: numbers of energy
980: points. The reaction matrix approach allows us to circumvent this problem.
981: It is also useful to note that the reaction matrix approach has been used extensively to study
982: properties of the complex poles of the S-matrix. This is discussed in some detail in 
983: \cite{kn:jung,kkn:rotter1,kkn:rotter2,kkn:stockman2002}.
984: 
985: We have obtained a number of results. We find that, in the near 
986: integrable regime,
987: nonlinear resonances in the classical phase space give rise to large eigenphase
988: excursions and long delay times for quantum particles that can tunnel 
989: into these
990: dynamical structures.
991: 
992: We have focused much of our discussion on the energy regime in
993: which sixteen channels are open in the lead. We have been  
994: able to follow each
995: eigenphase of the S-matrix continuously as a function of energy. We 
996: have examined the
997: statistical properties of the scattering process by gathering data about each
998: eigenphase at  discrete energy intervals in the sixteen channel 
999: regime. This assumes a
1000: kind of ``stationarity" as a function of energy, of the underlying
1001: scattering process.   We have chosen the energy intervals so that our 
1002: data points are
1003: statistically independent.
1004: 
1005: We have shown, for the scattering system considered here,  that the affect of direct
1006: processes on the eigenphase  curves can be
1007: transformed away. We find that, for the case where the cavity 
1008: dynamics is classically
1009: chaotic, a partial time delay density histogram, formed  from all 
1010: sixteen transformed
1011: eigenphase curves, agrees to 96$\%$ confidence level with  a 
1012: Brody distribution
1013: with Brody parameter, ${\beta}=0.87$. 
1014: Similar deviations
1015: from the GOE prediction of ${\beta}=0.95$ have been seen in the 
1016: nearest neighbor energy
1017: eigenvalue spacing distributions of closed ripple billiards \cite{kn:li} and in that
1018: case are  caused by bouncing
1019: ball orbits. We expect the same mechanism is having an affect here. 
1020: 
1021: 
1022: 
1023: 
1024: %
1025: \section {Acknowledgements}
1026: 
1027: The authors wish to thank the Welch Foundation, Grant
1028: No.F-1051; NSF Grant INT-9602971;  and DOE contract No.DE-FG03-94ER14405
1029: for partial support of this work. We also thank the University of
1030: Texas at Austin High Performance Computing Center for use of their computer
1031: facilities.
1032: 
1033: \pagebreak
1034: 
1035: 
1036: \begin{thebibliography}{99}
1037: 
1038: \bibitem{kn:wigner} E.P. Wigner, Can. Math. Congr. Proc. (University of
1039: Toranto Press, Toranto, 1957).
1040: 
1041: \bibitem{kn:gary} J. Garg, J. Rainwater, J.S. Petersen, and W.W. Havens, Jr., Phys. Rev.
1042: {\bf 134} B985 (1964).
1043: 
1044: \bibitem{kn:brody.etal} T.A. Brody, J. Flores, J.B. French, P.A.
1045: Mello, A. Pandey, and
1046: S.S.M. Wong, Rev. Mod. Phys. {\bf 53} 385 (1981).
1047: 
1048: \bibitem{kn:mehta} M.L. Mehta, {\it Random Matrices and the
1049: Statistical Theory of Energy
1050: Levels} 2nd Ed. (Academic Press, New York, 1991).
1051: 
1052: \bibitem{kn:4} F.J. Dyson, J. Math Phys. {\bf 3} 140
1053: (1962); {\bf 3} 157 (1962); {\bf 3} 166 (1962); {\bf 3} 344 (1962).
1054: 
1055: \bibitem{kn:hua} L.K. Hua   {\it Harmonic Analysis of Functions of 
1056: Several Complex
1057: Variables in the Classical Domains} (American Mathematical Society, 
1058: Providence, 1963).
1059: 
1060: \bibitem{kn:macdonald} S.W. MacDonald and A.N. Kaufman, Phys. Rev. Lett.
1061: {\bf 42} 1189
1062: (1979).
1063: 
1064: \bibitem{kn:boh} O. Bohigas, M.J. Giannoni, C. Schmit,
1065: Phys. Rev. Lett. {\bf 52} 1. (1984)
1066: 
1067: \bibitem{kn:casati} G. Casati and F. Valz-Gris, and I. Guarnieri,  Lett. Nuovo
1068: Cimento {\bf 28} 279 (1980).
1069: 
1070: \bibitem{kn:berry} M.V.  Berry and M. Tabor, Proc. Roy. Soc. Lond. {\bf A356}
1071: 375 (1977).
1072: 
1073: 
1074: \bibitem{kn:reichl} L.E. Reichl, {\it The Transition to Chaos in Classical
1075: Conservative Systems: Quantum Manifestations} (Springer-Verlag, New York,
1076: 1992).
1077: 
1078: \bibitem{kn:guhr} T. Guhr, A. Muller-Groeling, and H.A. Weidenmuller,
1079: Phys. Rept. {\bf
1080: 299} 189 (1998).
1081: 
1082: 
1083: \bibitem{kn:weiden} C. Mahaux and H.A. Weidenmuller {\it Shell Model
1084: Approach to Nuclear
1085: Reactions} (North-Holland, Amsterdam, 1969)
1086: 
1087: \bibitem{kn:mello3} G. Lopez, P. A. Mello and T. H. Seligman, 
1088: Z. Phys. A {\bf 302} 351 (1981)
1089: 
1090: \bibitem{kn:mello2} P. A. Mello, P. Pereyra and T.H. Seligman,
1091: Annals of Physics {\bf 161} 254
1092: (1985), W. A. Friedman and P. A. Mello,
1093: Annals of Physics {\bf 161} 276
1094: (1985)
1095: 
1096: \bibitem{kn:brouwer} P.W. Brouwer, Phys. Rev. B {\bf 51} 16878 (1995)
1097: 
1098: \bibitem{kn:eisenbud} E.P. Wigner and L.E. Eisenbud, Phys. Rev. {\bf
1099: 72} 29 (1949).
1100: 
1101: \bibitem{kn:luna} G.A. Luna-Acosta, Kyungsan Na,  L.E. Reichl, and
1102: A. Krokhin, Phys. Rev. E (1996) {\bf 53} 3271 (1996).
1103: 
1104: \bibitem{kn:akg1} Gursoy B. Akguc and L.E. Reichl,
1105: Phys. Rev. E. {\bf 64} 056221 (2001).
1106: 
1107: \bibitem{kn:marcus1}  C.M. Marcus, A.J. Rimberg, R.M. Westervelt, 
1108: P.F. Hopkins,  and
1109: A.C. Gossard,  Phys. Rev. Lett. {\bf 69} 506 (1992).
1110: 
1111: \bibitem{kn:chang} A.M. Chang,  H.U.  Baranger, L.N. Pfeiffer, and K.W. West,
1112:   Phys. Rev. Lett. {\bf 73} 2111 (1994).
1113: 
1114: \bibitem{kn:bird} J.P. Bird, J. Phys.: Condens. Matter {\bf 11} R413 (1999).
1115: 
1116: \bibitem{kn:graf}  H-D. Graf, H.L.  Harney, H. Lengeler,  C.H.  Lewenkopf,
1117:   C. Rangacharyulu, A. Richter,  P. Schardt,  and H.A. Weidenmuller,  Phys. Rev.
1118: Lett. {\bf 69} 1296  (1992).
1119: 
1120: \bibitem{kn:stein} J. Stein,  and H.-J. St{\"o}ckmann,   Phys. Rev. 
1121: Lett. {\bf 68} 2867
1122: (1992).
1123: 
1124: \bibitem{kn:stockmann} H.-J. St{\"o}ckmann,   {\it Quantum Chaos - An 
1125: Introduction}
1126: (Cambridge University Press, Cambridge, UK, (1999)).
1127: 
1128: 
1129: 
1130: \bibitem{kn:her} G.A. Luna-Acosta, J. A.  Mendez-Bermudez, P. Seba, and
1131: K. N. Pichugin, Phys. Rev. E (1996) {\bf 65} 046605 (2002).
1132: 
1133: \bibitem{kn:jung} C. Jung, T. H. Seligman, J. Phys. A. {\bf 28} 1507 (1995)
1134: 
1135: 
1136: \bibitem{kn:vaarbar} J.J.M. Verbaarschot, H.A. Weidenmuller, and M.R.
1137: Zirnbauer,
1138: Phys. Rep. {\bf 129} 367 (1985).
1139: 
1140: \bibitem{kn:akg} Gursoy B. Akguc and L.E. Reichl,
1141: J. Stat. Phys. {\bf 98} 813 (2000).
1142: 
1143: \bibitem{kn:li} Wenjun Li, L.E. Reichl, and Biao Wu, 
1144:  Phys. Rev. E {\bf 65} 056220 (2002)
1145: 
1146: \bibitem{kn:dietz} B. Dietz, M. Lombardi, and T. H. Seligman
1147: J. Phys, A {\bf 29} L95 (1996)
1148: 
1149: \bibitem{kn:brody} T.A. Brody, Lett. Nuov. Cim. {\bf 7} 482 (1973)
1150: 
1151: \bibitem{kn:leh} N. Lehman, D.V. Savin, V.V. Sokolov, H.-J. Sommers,
1152:    Physica D {\bf 86} 572
1153: (1995)
1154: 
1155: \bibitem{kn:mello4} P. A. Mello and H. U. Baranger
1156: Waves Random Media {\bf 9} 105
1157: (1999)
1158: 
1159: \bibitem{kn:bohm} E.P. Wigner, Phys. Rev. {\bf 98} 145
1160: (1955); F.T. Smith, Phys. Rev. {\bf 118} 349 (1960); D. Bohm, {\it
1161: Quantum Theory}
1162: (Prentice-Hall, Inc. Englewood Cliffs, N.J., 1950), p. 260.
1163: 
1164: 
1165: \bibitem{kn:fyo1} Y.V. Fyodorov and H-J Sommers,
1166:        J. Math. Phys. {\bf 38} 1918
1167: (1997).
1168: 
1169: \bibitem{kn:fyo2} Dmitry V. Savin, Yan V. Fyodorov and H-J Sommers,
1170: Phys. Rev. E. {\bf 63} 035202(R)
1171: (2001).
1172: 
1173: 
1174: \bibitem{kn:haake}F. Haake, F. Izrailev, N. Lehman, D. Saher, and
1175: H.-J. Sommers. Z. Phys. B {\bf 88} 359 (1992)
1176: 
1177: \bibitem{kn:jung} C. Jung, M. Muller, and I. Rotter, Phys. Rev. E
1178: {\bf 60} 114 (1999).
1179: 
1180: \bibitem{kkn:rotter1}  I. Rotter, E. Persson, K. Pichugin, and P. 
1181: Seba, Phys. Rev. E
1182: {\bf 62} 450 (2000).
1183: 
1184: \bibitem{kkn:rotter2}  I. Rotter,  Phys. Rev. E {\bf 64} 36213 (2001).
1185: 
1186: \bibitem{kkn:stockman2002} H.-J. Stockmann, E. Persson, Y.-H. Kim, M. 
1187: Barth, U. Kuhl, and
1188: I. Rotter, Phys. Rev. E {\bf 65} 66211 (2002).
1189: 
1190: 
1191: 
1192: \end{thebibliography}
1193: 
1194: \pagebreak
1195: 
1196: 
1197: \newpage
1198: 
1199: 
1200: \begin{figure}
1201: \caption{The geometry of the two dimensional ballistic waveguide
1202: used in our calculations; $a$ is the half-width of the  ripple,
1203: $d$ is the width of the lead and the average width of the cavity. The ripple cavity
1204: extends from $x=0$ to $x=L$}
1205: \label{figure1}
1206: \end{figure}
1207: %
1208: 
1209: 
1210: \begin{figure}
1211: \caption{Surfaces of section for $L=511\AA$, $d=101\AA$, and ${\rm a}=1\AA$.
1212: (a) A Poincare surface of section showing $p_x/p={\cos}(\alpha)$ 
1213: versus $x$, each time
1214: the particle hits the bottom wall. (b) Husimi plot of quantum surface 
1215: of section (QSS)
1216: for cavity eigenstate with ${\lambda}_j=257.1934E_1$. (c) QSS for 
1217: cavity eigenstate with
1218: ${\lambda}_j=257.9339E_1$. (d) QSS for cavity eigenstate with
1219: ${\lambda}_j=258.6655E_1$. (e) QSS for cavity eigenstate with
1220: ${\lambda}_j=258.9072E_1$.}
1221: \label{figure2}
1222: \end{figure}
1223: %
1224: 
1225: \begin{figure}
1226: \caption{Eigenphases, ${\delta}_{\alpha}$ versus $E/E_1$ for the
1227: energy interval,
1228: $256{\leq}E/E_1{\leq}272$: (a)
1229: ${\rm a}=25\AA$,  (b) ${\rm a}=1\AA$.}
1230: \label{figure3}
1231: \end{figure}
1232: %
1233: 
1234: 
1235: 
1236: %
1237: \begin{figure}
1238: \caption{ (a) Histogram of number of eigenphases, $N(\delta)$, versus $\delta$  for the
1239: 16 channel energy interval $256{\leq}E/E_1{\leq}289$, and for four
1240:    different ripple sizes, ${\rm a}=25\AA,~30\AA,~35\AA,~45\AA$. The solid line is a
1241: plot of the single channel Poisson kernel with
1242: ${\langle}S{\rangle}={\sum_{\alpha=1}^{64}}{\langle}{\rm
1243: e}^{i{\delta}_{\alpha}}{\rangle}$ and normalized to the number of eigenphases.   (b)
1244: Histogram of transformed eigenphases, ${\theta}_{\alpha}$, for all 64 eigenphase curves.
1245: For all cases, $d=101\AA$ and  $L=511\AA$. A $\chi^2$ test result is also shown for both plots
1246:  with 17 bins taken into account}
1247: \label{figure4}
1248: \end{figure}
1249: %
1250: 
1251: 
1252: \begin{figure}
1253: \caption{ Histogram of number, $N({\sigma})$, of nearest neighbor scaled eigenphase
1254: spacings,
1255: ${\sigma}$,  for the chaotic regime with
1256: $d=101\AA$, $L=511\AA$. The average spacing, ${\langle}{\sigma}{\rangle}$ is obtained for
1257: each eigenphase curve.  The histograms contain a total of $15{\times}67=1005$ data
1258: points averaged over four different ripple sizes, ${\rm
1259: a}=25\AA,~30\AA,~35\AA,~45\AA$. (We obtain a histogram for each of the four 
1260: values of the ripple amplitude. We then add them and divide by four.)
1261:  (a) Before direct processes are transformed away. The
1262: thin solid line is the Brody distribution with $\beta=0.635$.
1263: (b) After direct processes transformed away. The thin solid line
1264: is the Brody distribution with $\beta=0.865$.A $\chi^2$ test result is also shown for both plots
1265:  with 13 bins in it taken into account
1266: }
1267: \label{figure5}
1268: \end{figure}
1269: %
1270: 
1271: %
1272: \begin{figure}
1273: \caption{  Histogram of number, $N({\sigma})$, of nearest neighbor eigenphase spacings,
1274: ${\sigma}$, 
1275: for the near integrable  regime with
1276: $d=101\AA$, $L=511\AA$. The histograms contain a total of 1005 data
1277: points averaged over
1278: 6 different ripple sizes, ${\rm a}=0.5\AA,~1\AA,~2\AA,~3\AA,~4\AA,~5\AA$.
1279: (a) Before direct processes are transformed away. The thin solid line
1280: is the Brody distribution with $\beta=0.2$.
1281: (b) After direct processes are transformed away. The thin solid line
1282: is the Brody distribution with $\beta=0.116$. A $\chi^2$ test result is also shown for both plots
1283:  with 8 bins taken into account }
1284: \label{figure6}
1285: \end{figure}
1286: %
1287: 
1288: \begin{figure}
1289: \caption{The auto-correlation function of  time delays
1290: in the 16 channel energy interval, $256<E/E_1<289$.
1291: The thin line is  obtained numerically for the chaotic regime, using six
1292: different ripple amplitudes, ${\rm a}=22\AA,~23\AA,~24\AA,~25\AA,~26\AA,~27\AA$, with
1293: direct processes transformed out of the data. An  auto-correlation function is
1294: obtained for each partial time delay curve and the average of those 96 auto-correlation
1295: functions is shown. 
1296:    The dotted-dashed line
1297: is the GOE result for perfect coupling with 16 modes with the average density of
1298: states chosen equal to $1.25$. The thick line represents the 
1299: numerically obtained auto-correlation function in the region of mixed phase space for 
1300: ${\rm a}=0.5\AA,~1\AA,~2\AA,~3\AA,~5\AA$ }
1301: \label{figure7}
1302: \end{figure}
1303: %
1304: 
1305: 
1306: 
1307: \begin{figure}
1308: \caption{Histogram of number of scaled partial delay times, $N(\tau)$, versus
1309: $\tau/{\langle}\tau{\rangle}$, for the 16 channel energy interval,
1310: $256{\leq}E/E_1{\leq}289$, with $d=101\AA$ and $L=511\AA$. Data for
1311: ripple amplitudes
1312: ${\rm a}=25\AA,~ 30\AA,~ 35\AA, ~45\AA$ is included in the histograms. 
1313: Data points are taken at energy intervals, ${\Delta}E=0.33E_1$. 
1314:  (a) Histogram of scaled partial delay times taken from 
1315: eigenphase curves for
1316: ${\delta}_{\alpha}$. A scaling factor, ${\langle}\tau{\rangle}$, is obtained for each
1317: eigenphase curve. (b) Histogram of scaled partial delay times taken from eigenphase
1318: curves for the transformed eigenphases, ${\theta}_{\alpha}$. (c) Histogram of partial delay
1319: times obtained from the $16{\times}16$ S-matrix, ${\bar S}_{goe}$ (includes 4000 data
1320: points). A $\chi^2$ test result is also shown for the  plots (b) and (c)
1321:  with 13 bins taken into account. 
1322: }
1323: \label{figure8}
1324: \end{figure}
1325: %
1326: 
1327: %
1328: \begin{figure}
1329: \caption{Histogram of partial delay times for the 16 channel energy interval,
1330: $256{\leq}E/E_1{\leq}289$, with $d=101\AA$ and $L=511\AA$. Data for
1331: ripple amplitudes, 
1332: ${\rm a}=0.5\AA,1\AA,2\AA,3\AA,5\AA$, is used to construct the histograms.
1333: Data points are taken at energy spacings, ${\Delta}E=0.1E_1$. A total of
1334: $400 X  16$ points is used.
1335:  (a) Histogram of scaled partial delay time curves taken from eigenphase 
1336: curves for ${\delta}_{\alpha}$. (b)
1337: Histogram of scaled partial delay times taken from curves for transformed eigenphases, 
1338: ${\theta}_{\alpha}$.}
1339: \label{figure9}
1340: \end{figure}
1341: %
1342: 
1343: 
1344: %
1345: \begin{figure}
1346: \caption{Plot of Wigner-Smith delay time in the energy the interval
1347: $257.4{\leq}E/E_1{\leq}259.5$  for  $d=101\AA$, $L=511\AA$, ${\rm a}=1\AA$. 
1348: Crosses show values
1349: of cavity basis state energies in this interval. Inserts show the 
1350: spatial distribution
1351: of four cavity basis states, two at resonance and two off resonance. }
1352: \label{figure10}
1353: \end{figure}
1354: %
1355: 
1356: 
1357: 
1358: %
1359: \end{document}
1360: 
1361: 
1362: