1: \documentclass[prl,showpacs,twocolumn]{revtex4}
2:
3: \usepackage{graphicx}
4:
5: \usepackage{amsmath}
6: \usepackage{amsfonts}
7: \usepackage{amssymb}
8: \usepackage{bm}
9:
10: \begin{document}
11:
12: \title{$2\delta$-Kicked Quantum Rotors: Localization and `Critical' Statistics}
13:
14: \author{C.E.~Creffield, G.~Hur and T.S.~Monteiro}
15:
16: \affiliation{Department of Physics and Astronomy, University College London,
17: Gower Street, London WC1E 6BT, United Kingdom}
18:
19: \date{\today}
20:
21: \begin{abstract}
22: The quantum dynamics of atoms subjected to pairs of
23: closely-spaced $\delta$-kicks from optical potentials are shown to be
24: quite different from the well-known
25: paradigm of quantum chaos, the singly-$\delta$-kicked system.
26: We find the unitary matrix has a new oscillating band structure
27: corresponding to a cellular structure of phase-space and observe a spectral signature of
28: a localization-delocalization transition from one cell to several.
29: We find that the eigenstates
30: have localization lengths which scale with a fractional
31: power $L \sim \hbar^{-.75}$ and obtain a regime of near-linear spectral variances
32: which approximate the
33: `critical statistics' relation
34: $\Sigma_2(L) \simeq \chi L \approx \frac{1}{2}(1-\nu) L$,
35: where $\nu \approx 0.75$ is related to the fractal classical phase-space
36: structure. The origin of the $\nu \approx 0.75$ exponent is analyzed.
37: \end{abstract}
38:
39: \pacs{ 05.45.Mt, 05.60.-k, 72.15Rn, 32.80.Pj}
40:
41: \maketitle
42:
43: The $\delta$-kicked quantum rotor (QKR)
44: is one of the most studied paradigms of quantum chaos.
45: Its implementation using cold atoms in
46: optical lattices \cite{Raizen} provided a convincing demonstration of
47: a range of quantum chaos phenomena including
48: Dynamical Localization \cite{Fish}, the quantum suppression
49: of chaotic diffusion.
50: Recently, an experimental study \cite{Jones} of
51: cesium atoms subjected to {\em pairs} of $\delta$-kicks (2$\delta$-KR)
52: showed surprisingly different behavior.
53: The classical phase-space is chaotic but
54: is made up fast diffusing regions which are partly separated by
55: slow-diffusing `trapping regions', where the classical trajectories stick;
56: the classical analysis revealed a regime of anomalous
57: diffusion corresponding to long-lived correlations between kicks.
58:
59: Further details are given in \cite{Web} but we show here that the
60: 2$\delta$-KR has some unexpected {\em quantum} properties.
61: We show that there is a cellular phase-space structure which
62: arises from a novel oscillatory band structure of the corresponding
63: unitary matrix. One consequence is a new type of localization-delocalization
64: transition not seen in the QKR, where states delocalize from single
65: to multiple-cell occupancy; we show it has a clear spectral signature.
66: We have also found scaling behavior of the localization lengths
67: associated with a {\em fractional} exponent, i.e. $L \sim \hbar^{-0.75}$,
68: whereas for the well-studied QKR, $L \sim \hbar^{-1}$.
69: A similar exponent is found for the decay of return probabilities in the
70: trapping regions, $P(t) \sim t^{-0.75}$.
71: We argue that the exponent $0.75$ corresponds closely to
72: the value obtained for the dominant exponent of the
73: golden ratio cantorus \cite{Fish2,Maitra}.
74: We show that the spectral fluctuations (both the nearest-neighbor
75: statistics (NNS) and spectral
76: variances) show important differences with the QKR in regimes
77: where the delocalization of eigenstates is hindered by cantori
78: bordering the cells. We find a regime
79: approximating the form found in `critical statistics': the number variances
80: of the spectra are linear $\Sigma_2(L) \simeq \chi L$ for $L \gg 1$,
81: where $\chi \simeq 1/2(1-\nu) < 1$ and $\nu \simeq 0.75$.
82:
83: The term `critical statistics' arose originally in relation to the
84: Metal Insulator Transition (MIT) in systems with
85: disorder \cite{Shapiro,Chalker}.
86: A new universal form of the distribution of nearest-neighbor
87: eigenvalue spacings, termed `semi-Poisson', $P(s)\sim s \exp{-2s}$
88: was associated with the MIT \cite{Shapiro}.
89: For critical statistics a very interesting connection has been
90: established between
91: the multifractal characteristics of the wavefunctions and those of the
92: spectral fluctuations \cite{Chalker}: the number variances
93: of the spectra are linear $\Sigma_2(L) \simeq \chi L$ for $L \gg 1$.
94: The slope, $\chi \simeq 1/2(1-D_2/D) < 1$, was shown to be related to a fractal
95: dimension $D_2$ obtained from the second moment of the wavefunction and
96: to $D$, the spatial dimension of the system. For integrable
97: dynamics, in contrast, $\Sigma_2(L) = L$ while
98: for a GOE, $\Sigma_2(L) \sim Ln(L)$.
99: There is much current interest in so-called `critical' statistics
100: in non-KAM billiards (typically
101: systems where the dynamics would be integrable were it not for a discontinuity
102: in the potential) \cite{Garcia}, which show multifractal scalings and linear
103: variances related to $D_2$. Below we apply the term `critical statistics'
104: in this broader sense, rather than the MIT critical point.
105: Multifractal behavior has been demonstrated for Cantor spectra
106: \cite{Ketz} where the level density itself is not smooth.
107: However, until now, critical statistics have not been seen -- and were not
108: thought to be relevant to -- KAM systems. These are systems, ubiquitous in
109: many areas of physics, where the transition to chaos
110: as a perturbing parameter is increased is quite gradual.
111:
112: \begin{figure}[b]
113: \includegraphics*[width=0.125\textwidth,angle=-90,clip=true]{Fig1a.eps}
114: \hspace*{5 mm}
115: \includegraphics[width=0.125\textwidth,angle=-90,clip=true]{Fig1b.eps}
116: \caption{Left: Structure of time-evolution matrix $U(T,0)$,
117: for the Quantum Kicked Rotor (QKR),
118: in a basis of momentum states,
119: exemplifying the typical band-structure of a Band Random Matrix (BRM).
120: Right: $U(T,0)$ for our system, the 2$\delta$-KR, showing the new form with
121: oscillating bandwidth. Before delocalization,
122: eigenstates are confined within a single `momentum cell' of dimension $N$.}
123: \label{Fig1}
124: \end{figure}
125:
126: The Hamiltonian of the 2$\delta$-KR is
127: $H(x,p)= \frac{p^2}{2} + K \cos x \sum_n \delta(t-nT) + \delta(t-nT+\epsilon)$;
128: there is a short time interval $\epsilon$ between the kicks in each pair
129: and a much longer time $\simeq T$ between the pairs themselves.
130: In experiments $\epsilon \sim 0.01 - 0.1 \ll T$ and $ \hbar \simeq 2 \to 1/4$
131: in the usual re-scaled units \cite{Jones}.
132:
133: A study of the spectral fluctuations of a time-periodic system involves a
134: study of the eigenstates and eigenvalues
135: of the one-period time-evolution operator $U(T,0)$.
136: For the QKR, the matrix representation, in an angular
137: momentum basis $|\ l \rangle$, has elements
138: $U_{lm}= U_l^{free}. \ U_{lm}^{kick}=\exp {-i l^2T\hbar/2} \ . \
139: J_{l-m}(\frac{K}{\hbar})$.
140: The `kick' terms, $J_{l-m}(\frac{K}{\hbar})$
141: are Bessel functions and
142: give the matrix the banded form illustrated in Fig.\ref{Fig1}(a).
143: Since $J_{l-m}(x) \simeq 0$ if $|l-m| >x$ we define the
144: usual bandwidth $b = K/\hbar$.
145: The resulting statistics are approximated by those of Band Random Matrix Theory
146: (BRMT) \cite{Izraelev} rather than of RMT: i.e.
147: if the dimension of the $U(T,0)$
148: matrix is $N_{tot}$, the statistics are {\em Poissonian} for $N_{tot}\gg b$;
149: the eigenstates of the BRM are exponentially localized in $l$,
150: with a localization length in momentum ($p=l\hbar$) which equals
151: $L_p \sim K^2/\hbar$, so states separated
152: in $p$ by $\gg L_p$ will be largely uncorrelated.
153:
154: For the 2$\delta$-KR, the corresponding matrix elements are:
155: \begin{equation}
156: U_{lm}=\\
157: e^{-i l^2 (T-\epsilon) \hbar/2} \ . \
158: \sum_k J_{l-k}\left(\frac{K}{\hbar}\right)
159: J_{k-m}\left(\frac{K}{\hbar}\right)
160: e^{-i \frac{k^2}{2} \hbar\epsilon} .
161: \end{equation}
162: As the $U(T,0)$ matrix is quite insensitive to $(T-\epsilon)$
163: the quantum dynamics largely depends only on
164: two scaled parameters $K_\epsilon= K\epsilon$
165: and $\hbar_{\epsilon}= \hbar\epsilon$, rather than
166: on $K,\epsilon$ and $\hbar$ independently: the remainder of the matrix is
167: invariant if $K_\epsilon$ and $\hbar_{\epsilon}$
168: are kept constant.
169:
170: An analytical form for the bandwidth of $U(T,0)$ was obtained
171: in \cite{Web}: the bandwidth oscillates sinusoidally between
172: a maximum value $b_{max}= 2K/\hbar$ for angular momenta
173: $l \simeq 2n\pi/(\hbar\epsilon)$
174: and a minimum value $b_{min} \approx 0$ for
175: $l \simeq (2n+1)\pi/(\hbar\epsilon)$.
176: These minima correspond to the trapping momenta
177: $p= l\hbar \simeq (2n+1)\pi/\epsilon$
178: seen in experiment. The corresponding band-structure of $U$ is
179: illustrated in Fig.\ref{Fig1}(b): the
180: band oscillates, and $U$ is approximately partitioned into sub-matrices
181: of dimension $N=\frac{2\pi}{\epsilon \hbar}$ corresponding to
182: separate momentum cells.
183:
184: The key to our work is our ability to vary the {\em transport} between the cells
185: (by opening/closing the classical fractal `gates' between them) separately
186: from the degree of {\em filling} of each individual cell.
187: We begin by introducing a `filling factor' $R$ where
188: \begin{equation}
189: R=\frac{K^2}{N \hbar^2} = \frac{K^2_\epsilon}{2 \pi \hbar_{\epsilon}}
190: \label{eq2}
191: \end{equation}
192: measures the degree of filling of a cell by a typical state in the
193: absence of confinement.
194: Clearly, if there is no transport between cells the states simply
195: fill the cell uniformly
196: and even if $R \gg 1$, the localization lengths $L_p \sim N\hbar$.
197: We begin by defining a {\em localized} limit, where $R \ll 1$ and
198: $L_p \ll N\hbar \sim \frac{2\pi}{\epsilon} $; typical states are
199: insensitive to the boundary conditions of the cell and this limit
200: is Poissonian. At the other extreme, if
201: we allow strong coupling between cells, we move to an opposite limit as an
202: increasing proportion of eigenstates become delocalized over several cells.
203:
204: We now investigate the transport. A classical analysis \cite{Web,Mischa} shows
205: that if we take $K \epsilon \ll 1$ and expand initial momenta
206: of the $j-th$ trajectory about the trapping values
207: $p_j = (2n+1)\pi/\epsilon + \delta p_j$, we can show that much of the trapping
208: region is given by a classical map
209: quite similar to the well-known Standard Map:
210: \begin{eqnarray}
211: p_{j+2} &\simeq& p_{j} - K^2 \frac{\epsilon}{2} \ \sin 2 x_{j} - K \delta p^j \ \epsilon \cos x_{j}\label{eq13} \\
212: x_{j+2} &\simeq& x_{j} + p_{j+2} T .
213: \label{eq13a}
214: \end{eqnarray}
215:
216: Over much of the trapping region, the second term in Eq.\ref{eq13}
217: is dominant, for parameter regimes of interest.
218: Then the kick impulse has a $\pi/2$ phase relative to the full map
219: (the 2-kick map gives a pair of $V'(x) = -\sin x$ type impulses)
220: and a momentum dependent effective kick strength
221: $K'= K \epsilon \delta p_j $. A detailed study of classical
222: phase-space \cite{Mischa} shows that at low $K'$ the resonance structure is locally
223: quite similar to
224: the Standard Map. Hence, in the regime $ K \epsilon \delta p_j \sim 1$
225: we expect the
226: `golden ratio' cantori, which result from the last invariant manifold
227: of a KAM system, to provide the strongest barrier to transport \cite{Ott}.
228: Though we want $K \epsilon$ to be small, if $K \epsilon < 0.1$
229: phase space becomes too regular. Hence here we find that the regime of interest is within the interval $0.1 \leq K_\epsilon \leq 0.7$.
230:
231: \begin{figure}[htb]
232: \includegraphics[height=2.25in]{Fig2a.eps}
233: \includegraphics[height=2.in]{Fig2b.eps}
234: \caption { {\bf(a)} Final ($t \to \infty$) momentum distributions, $N(p)$,
235: (slightly smoothed) for quantum wavepackets of the 2$\delta$-QKR
236: for $K=20$, $\epsilon=0.0175$ and $\hbar=1/8$ and $ 1/30$ respectively.
237: $N(p)$ for both the eigenstates and wavepackets shows a long-range
238: `staircase' form which on average follows the exponential
239: $N(p) \sim \exp{-2(p-{\overline{p}})/L_{exp}}$ where
240: $L_{exp}/2 = 2\pi/\hbar\epsilon$;
241: the $\hbar$-dependence of $L_{exp}$ is determined
242: by the drop in probability, $d$, at each step.
243: {\bf(b)} shows that $Ln(2d)$ plotted against $Ln (\hbar \epsilon)$
244: lies on straight lines of invariant $K_{\epsilon} = K \epsilon$,
245: with constant slope $0.75$. Hence $d \propto (\hbar \epsilon)^{0.75}$
246: and $L_{exp} \propto \hbar^{-0.75}$ -- in contrast to the well-known
247: QKR result $L_{exp} \propto \hbar^{-1}$.
248: The $R=1$ border is shown: for $R>1$ (below the line) eigenstates
249: fill much of a single cell. The {\em delocalization} border is shown
250: at $d \simeq 2$: for $d>2$ (above the line), over $98\%$ of the probability
251: of typical eigenstates is
252: confined to a single cell; for $d<2$, eigenstates begin to occupy
253: multiple cells. Statistics are presented
254: later in Fig.3 for points corresponding to the dotted line. }
255: \label{Fig2}
256: \end{figure}
257:
258: We investigated the corresponding quantum transport
259: by evolving a set of wavepackets $\Phi(p,t)$ in time,
260: (where $\Phi(p,t=0) = \delta(p)$) for a range
261: of $K_\epsilon $ and $\hbar_\epsilon$, until the momentum
262: spreading is arrested by dynamical localization at $t \simeq t_H$.
263: The resulting probability
264: distributions $|\Phi(p,t \gg t_H)|^2= N(p)$ have a characteristic `staircase' structure, shown in Fig.\ref{Fig2}(a).
265: At each step there is a steep drop in probability:
266: \begin{equation}
267: N(p)_+= e^{-{2d}}N(p)_-
268: \end{equation}
269: (where $N(p)_\pm$ represent probabilities before(-) and
270: after(+) the step)
271: concentrated over the trapping region
272: ($\sim 1/6$ of a cell in every case \cite{Web}).
273: The staircase tracks an exponential envelope
274: $N(p) \sim \exp-{2|p|/L_{exp}}$, where
275: $L_{exp}=\frac{\pi}{\epsilon d}$. We average over
276: several steps, to obtain $d$ as a function of $K_\epsilon$ and
277: $\hbar_\epsilon$. In Fig.\ref{Fig2}(b) we show that, quite accurately,
278: $d \propto \hbar_\epsilon^{0.75}/f(K_\epsilon)$ where $f(K_\epsilon)$ is
279: some function of the scaled kick-strength, $K_\epsilon$. We estimate
280: \begin{equation}
281: d \approx \frac{3.5\hbar_\epsilon^{0.75}} {K_\epsilon^3} .
282: \label{eq4}
283: \end{equation}
284: This defines our transport parameter, $d$, and complements the `filling factor'
285: $R$. The inner steps of the staircase have been seen in the
286: momentum distributions of atoms in optical lattices
287: \cite{Christensen,Jones}; so,
288: though existing data is
289: not in the critical regime, in principle the form of $d$ is
290: experimentally verifiable.
291:
292: We are unaware of another KAM system where a single power-law exponent
293: is so dominant. Typically, power-law behavior is associated
294: with mixed phase space behavior, where many
295: competing exponents are found \cite{Ketz}. We note also that the
296: value $0.75$ coincides closely with one of the scaling exponents found
297: in \cite{Fish2} for the golden ratio cantori: these were
298: $\sigma \approx 0.65$ (in the most unstable part of the cantori) and $\sigma \approx 0.76$ (most stable regions).
299: There have been previous studies of
300: transport in a region near golden ratio cantori. These have found
301: $L \sim \hbar^{+0.66}$ \cite{Geisel}
302: but only in a momentum band region `local' to the cantori.
303: We note that the $L \sim \hbar^{+\sigma}$ dependence is associated with
304: the physical process termed `retunnelling' \cite{Maitra} associated
305: with cantori which are classically `open'
306: but for which $\hbar$ is too large to permit free quantum transport.
307: An abrupt
308: change to an $ L \sim \hbar^{-\sigma}$ is observed when the cantori `open'
309: for quantum transport and $L$ is determined instead by localization.
310: It has been argued that the reason all previous studies have
311: found $L \sim \hbar^{+0.66}$
312: \cite{Maitra} is that the retunnelling transport favors
313: the unstable direction. To our knowledge ours is the only example
314: corresponding to the dominant exponent of
315: the golden ratio cantori; we attribute this to the fact
316: that we are always in a dynamical
317: localization regime, and localization will select the most stable parts of the
318: fractal cantori regions, where at low $K$, elliptic fixed points are found.
319:
320: We can now investigate the statistics as a function of the filling
321: factor $R$ and the inter-cell transport parameter $d$.
322: Full details are given in \cite{Web}, but
323: in brief: we considered two types of boundary conditions (BCs).
324: (1) Periodic BCs, i.e. solving the problem on a `torus' in momentum space,
325: a well-known procedure for the QKR \cite{Izraelev}.
326: (2) Open BCs, where we diagonalize
327: $U(T,0)$ with $N_{tot}=10,000$,
328: but $N \approx 1000$; we then assigned the $i-th$ eigenstate
329: to the $n-th$ cell if
330: $(2n+1)\pi/\epsilon \leq \langle p_i \rangle \leq (2n+3)\pi/\epsilon$
331: and calculated statistics for
332: each cell. In both cases we averaged over $\approx 20$ cells to improve
333: significance. For periodic BCs, eigenstates cannot escape from a single cell
334: of width $N\hbar$.
335: For open BCs, however, they can delocalize onto neighboring cells.
336:
337: \begin{figure}[htb]
338: \includegraphics[height=1.5in,width=3.in]{Fig3a.eps}
339: \vspace*{10 mm}
340: \includegraphics[height=1.5in,width=3.in]{Fig3b.eps}
341: \caption {{\bf(a)} shows the $\Sigma_2(L)$ statistics for the
342: $2\delta-KR$. Approximately linear variances, with slope $\chi \approx 0.125$
343: are found in the intermediate regime. Solid line indicates the line
344: $\Sigma_2(L) = 0.125 L + c$ corresponding to the `critical statistics' slope
345: $\chi \approx 1/2(1-\nu)$ where $\nu \simeq 0.75$ is the fractional
346: exponent obtained
347: previously. {\bf(b)} shows corresponding nearest neighbor statistics
348: for $R=0.8$ and $R=1.3$ of an intermediate form, which is compared with the
349: Semi-Poisson and GOE forms.
350: The insets show the effect of the boundary conditions. Before
351: delocalization ($d>2$) the results are insensitive to boundary conditions.
352: At delocalization, if the eigenstates are confined to a single cell
353: (periodic BCs, points indicated by asterisks) the
354: statistics make a transition to GOE; if they can delocalize (open BCs, points
355: indicated by circles),
356: the statistics tend to Poissonian behavior.
357: The arrows indicate the $d=2$ border.}
358: \label{Fig3}
359: \end{figure}
360:
361: In Fig.\ref{Fig3}(a), the $\Sigma_2(L)$ statistics are presented.
362: These represent the
363: variances in the spectral number density,
364: $\Sigma_2(L)= \langle L^2 \rangle - \langle L \rangle^2$, where we consider a
365: stretch of the spectrum with an average $\langle L \rangle$ levels.
366: A fit to the best straight line
367: in the range $L \simeq 5-40$ yields an estimate of the slope $\chi$.
368: In Fig.\ref{Fig3}(b) we show the
369: nearest neighbor $P(S)$ statistics. We quantify the deviation of $P(S)$ from
370: $P_{P}(S)$ and $P_{GOE}(S)$,
371: its Poisson and GOE limits respectively,
372: with a parameter $Q$ \cite{Casati2}:
373: \begin{equation}
374: 1-Q= \frac{ \int_0^{S_0} ( P(S') - P_{GOE}(S') dS')}
375: {\int_0^{S_0} (P_{P}(S') - P_{GOE}(S'))dS'} .
376: \end{equation}
377: Hence $Q=0$ indicates a Poisson distribution, while $Q=1$
378: signals a GOE distribution. We take $S_0=0.3$.
379:
380: The results of Fig.\ref{Fig3} demonstrate that the statistics are
381: not too sensitive
382: to boundary conditions for $d>2$ (see also \cite{Web}). However,
383: for $d \leq 2$, while the states with periodic BCs
384: (effectively restricted to a matrix of dimension $N$)
385: move gradually to the GOE limit, for the open BCs, at delocalization
386: the statistics tend back to the Poisson limit. This initially surprising behavior occurs
387: because, after delocalization, one finds increasing numbers of states for
388: which $\langle p_i \rangle$ assigns them to the $n-th$ cell, but for
389: which much of the state's probability is actually found in neighboring cells
390: \cite{Web}. Hence
391: states within each spectrum of $N$ eigenvalues
392: become progressively uncorrelated. This apparent failure of the
393: procedure of assigning states to a given cell, in fact provides a
394: rather good `marker'
395: for the onset of delocalization, and
396: yields a clear `turning point' in both the NNS
397: and $\Sigma_2(L)$ behavior. A detailed study of the multi-cell regime
398: remains to be undertaken; models of the statistics for chaotic systems
399: with non-uniform rate of exploration of phase-space \cite{Bohigas} may
400: be relevant here.
401:
402: Of further interest here is an intermediate regime, found for both boundary
403: conditions for $R \approx 1$ and $d > 2$. Over a wide range of parameters
404: and different cell sizes \cite{Web}, we find approximately
405: linear variances, for $L \gg 1$ and $L \ll N$, with slope $\chi \approx 0.125$.
406: The inset of Fig.\ref{Fig3}(a) plots the values of $\chi$ calculated along
407: the vertical dotted line
408: of Fig\ref{Fig2}(b) for periodic and open boundary conditions. We note that
409: a study of the decay of return probabilities
410: $P(t)= |\langle \psi(t=0)| \psi(t)| \rangle^2 \sim t^{-0.75}$
411: for wavepackets started in the trapping regions (see \cite{Web}).
412: The value of $\chi \approx 0.125$ corresponds to the value
413: which would be obtained from the MIT relation,
414: $\chi \approx 1/2(1-D_2)$ with, if $D_2 \approx 0.75$; in the MIT,
415: return probabilities with $P(t) \sim t^{-D_2}$ were similarly found.
416: We suggest that this represents a KAM analogue of behavior associated with
417: `critical statistics'.
418:
419: In summary, we have shown that the behavior of the 2$\delta$-kicked system
420: is rather different from the standard QKR. We have identified
421: a spectral signature of the novel localization-delocalization
422: transition that the system exhibits. We have also identified signatures
423: of the fractal phase-space structure of the cell borders.
424: The trapping regions may have applications
425: in atom optics experiments, as a means of manipulating the momentum
426: distribution of the atomic cloud.
427:
428: This work was supported by the EPSRC. We thank Shmuel Fishman
429: and Antonio Garcia-Garcia for helpful comments and advice.
430:
431: \begin{thebibliography}{99}
432: \bibitem{Raizen} F.L. Moore, J.C. Robinson, C.F. Bharucha, B. Sundaram,
433: M.G. Raizen, Phys. Rev. Lett. {\bf 75}, 4598 (1995).
434: \bibitem{Fish} G. Casati, B.V. Chirikov, F.M. Izraelev, J. Ford,
435: in "Lecture notes in Physics", Springer, Berlin, {\bf 93 }, 334 (1979);
436: S. Fishman, D.R. Grempel, R.E. Prange, Phys. Rev. Lett. {\bf 49}, 509 (1982).
437: \bibitem{Jones} P.H. Jones, M.M. Stocklin, G. Hur, T.S. Monteiro,
438: Phys. Rev. Lett. {\bf 93},223002 (2004).
439: \bibitem{Web} C.E. Creffield, S. Fishman, T.S. Monteiro, physics/0510161.
440: \bibitem{Fish2} S. Fishman, D.R. Grempel, R.E. Prange,
441: Phys. Rev. A.{\bf 36}, 289 (1987).
442: \bibitem{Maitra} N.T. Maitra and E. J. Heller,
443: Phys. Rev. E.{\bf 61}, 3620 (2000).
444: \bibitem{Shapiro}B.I. Shklovskii et al,
445: Phys. Rev. B.{\bf 47}, 11487 (1993)
446: \bibitem{Chalker}J.T. Chalker, I.V. Lerner, R.A. Smith,
447: Phys. Rev. Lett. {\bf 77}, 554 (1996);
448: V.E. Kravtsov and K.A. Muttalib, Phys. Rev. Lett. {\bf 79}, 1913 (1997);
449: D. Braun, G. Montambaux and M. Pascaud, Phys. Rev. Lett.{\bf 81}, 1062 (1998)
450: F. Evers and A.D. Mirlin, Phys. Rev. Lett.{\bf 84}, 3690 (2000).
451: \bibitem{Garcia}E. Bogomolny, U. Gerland and C. Schmit,
452: Phys. Rev. E.{\bf 59}, R1315 (1999);
453: A.M. Garcia-Garcia and J.J.M. Verbaarschot,
454: Phys. Rev. E.{\bf 67}, 046104 (2003);
455: E.B. Bogomolny and C. Schmit, Phys. Rev. Lett. {\bf 92}, 244102 (2004).
456: A.M. Garcia-Garcia and J. Wang, Phys. Rev. Lett. {\bf 94}, 244102 (2005).
457: \bibitem{Ketz} R. Ketzmerick, G. Petschel, T. Geisel,
458: Phys. Rev. Lett.{\bf 69}, 695 (1992).
459: \bibitem{Izraelev}M. Feingold et al, Phys. Rev. B {\bf 31}, R6852 (1985);
460: F.M. Izraelev, Phys. Rep. {\bf 196}, 299 (1990).
461: \bibitem{Mischa}M. Stocklin, PhD. thesis in preparation.
462: \bibitem{Ott}E. Ott, `Chaos in dynamical systems',
463: Cambridge University Press (1993).
464: \bibitem{Christensen} K. Vant, G. Ball, H. Ammann, N. Christensen.
465: Phys. Rev. E {\bf 59}, 2846 (1999).
466: \bibitem{Geisel} T. Geisel, G. Radons and J. Rubner,
467: Phys. Rev. Lett.{\bf 57}, 2883 (1986).
468: \bibitem{Casati2} C. Mejia-Monasterio, G. Benenti, G.G. Carlo, and G. Casati,
469: quant-ph/0410246 (2004).
470: \bibitem{Bohigas} O. Bohigas, S. Tomsovic, D. Ullmo,
471: Phys. Rep. {\bf 232}, 43 (1993).
472:
473: \end{thebibliography}
474:
475: \end{document}
476:
477: