1: \documentstyle[prl,aps,epsfig]{revtex}
2:
3:
4: \begin{document}
5:
6: \draft
7:
8: \title{Dynamical mechanisms of DC current generation in
9: driven Hamiltonian systems}
10:
11:
12: \author{S. Denisov and S. Flach}
13: \address{Max-Planck-Institut f\"ur Physik komplexer Systeme,
14: N\"othnitzer Str. 38, D-01187 Dresden, Germany}
15:
16: \date{\today}
17:
18: \wideabs{
19:
20: \maketitle
21:
22: \begin{abstract}
23: Recent symmetry considerations (Phys. Rev. Lett. {\bf 84} 2358 (2000))
24: have shown that dc currents may be generated in the stochastic layer
25: of a system describing the motion of a particle in a one-dimensional
26: potential in the presence of an ac time-periodic drive. In this paper
27: we explain the dynamical origin of this current.
28: We show that the dc current is induced
29: by the presence and desymmetrization of ballistic channels
30: inside the stochastic layer. The existence of these
31: channels is due to resonance islands with
32: non-zero winding numbers. The characterization of the
33: flights dynamics inside ballistic channels is described
34: by distribution functions. We obtain these distribution functions
35: numerically and find very good agreement with simulation data.
36: \end{abstract}
37:
38: \pacs{05.45.-a, 05.60.Cd, 05.45.Ac}
39:
40: }
41:
42:
43:
44: Transport in driven systems has received widespread interest
45: for several years because of its potential applicability to
46: nonequilibrium processes \cite{fjaajp97}. One canonical model reduces to the
47: motion of a particle in a one-dimensional space-periodic potential
48: in the presence of friction and a time-dependent stochastic force
49: $\chi(t)$ .
50: If $\chi(t)$ contains correlations, a nonzero current
51: may be realized even in the case of zero average $< \chi(t) > =0$
52: \cite{fjaajp97}.
53: Despite an enormous accumulation of results in this area
54: \cite{pr01} we are still lacking
55: a full understanding of the {\it microscopic} mechanisms of current occurrence.
56: If such an understanding is realizable, it should make use of the
57: true dynamical evolution of the system rather than of properties of
58: equations for probabilities. A first step in such a direction
59: requires to separate the essential time correlations from the pure
60: Gaussian white noise in $\chi(t)$. The simplest way is to assume
61: that $\chi(t)=E(t)+\xi(t)$ where $E(t)=E(t+T)$ is a time-periodic function
62: with zero mean $<E(t)>=0$ and $\xi(t)$ is a Gaussian white noise term.
63:
64: The next step is to skip the $\xi(t)$ term which leaves us with
65: a regular dynamical problem. In \cite{sfoyyz00} such a case was
66: considered and the relevant space-time symmetries of the dynamical problem
67: have been obtained. It was shown that a breaking of those symmetries
68: leads to a nonzero dc current. The mechanism of current occurrence
69: for the dissipative case was identified with a desymmetrization
70: of attractor basins. In contrast the nondissipative (Hamiltonian)
71: case is much less understood \cite{igph00}.
72: A time-dependent Hamiltonian system is usually
73: nonintegrable \cite{Licht}.
74: A strong dc current component was found in the corresponding
75: stochastic layer of such a system \cite{sfoyyz00}.
76: While its presence or absence
77: was clearly connected to the above mentioned absence or presence of
78: symmetries, the dynamical nature of directed transport in the
79: stochastic layer lacks full understanding.
80: The importance of this understanding
81: can be seen from e.g. results in \cite{oysfaaoyz01} where kinetic equations
82: for probability functions have been studied. In particular it was
83: found that the approaching of the Hamiltonian (dissipationless) limit
84: leads to an increase of the dc current value by 2-3 orders of magnitude.
85: Thus the description of the dynamical
86: mechanisms of directed current generation in
87: the stochastic layer of a driven Hamiltonian system will provide with
88: very useful information for dissipative systems as well.
89:
90: Let us consider the canonical example of a
91: particle moving in a spatially periodic nonlinear potential $U(x)=-\cos x$
92: under
93: the influence of
94: time-periodic zero-mean force $E(t)$. The Hamiltonian and the
95: equation of motion are
96: given by:
97: \begin{equation}\label{1}
98: H=\frac{p^2}{2} -\cos x - xE(t)\;\;,\;\;
99: \ddot{x} = -\sin x + E(t)\;\;.
100: \end{equation}
101: Here $p$ and $x$ are canonically conjugated momentum and coordinate
102: and $\ddot{x}\equiv {\rm d}^2 x / {\rm d} t^2$.
103:
104: We restrict our consideration to the choice
105: \begin{equation}
106: E(t)=E_1 \cos (t) + E_2 \cos (2t + \phi)\;\;.
107: \label{e(t)}
108: \end{equation}
109: According to \cite{sfoyyz00} for $E_2 \neq 0$ and $\phi \neq 0,\pi$
110: all possible symmetries which yield zero dc current are broken.
111: Note that the phase space dimension $d$ of (\ref{1}) is $d=3$.
112:
113: In the case of a nonzero field $E(t)$
114: the phase space of
115: (\ref{1})
116: is characterized by the presence of a stochastic layer
117: which originates from the
118: destroyed separatrix of the undriven system \cite{Licht}.
119: For $\phi=0,\pi$ this layer is invariant under the transformation
120: ($p \rightarrow -p$, $t \rightarrow -t$, $x \rightarrow x$). At the same
121: time the average velocity for any trajectory in this layer
122: vanishes, so we find zero dc current.
123: The symmetry will be broken when
124: tuning $\phi$ away from the values $(0,\pi)$.
125: The stochastic layer will deform. Most importantly any trajectory in
126: the layer will then be characterized
127: by a nonzero value of the average velocity.
128: Due to ergodicity inside the layer this value will be unique
129: for all trajectories from the layer.
130: While the fact that it may become nonzero
131: is understandable using the symmetry analysis,
132: its appearance and magnitude is due to dynamical mechanisms of motion inside
133: the stochastic layer.
134: In this paper, we show that the dc current is induced
135: by the presence and desymmetrization of ballistic channels
136: inside the stochastic layer. The existence of these
137: channels is due to resonances. The characterization of the
138: realization of flights inside ballistic channels is described
139: by distribution functions. We obtain these distribution functions
140: numerically and find very good agreement with simulation data.
141:
142: System (\ref{1}) has a mixed phase space, which
143: contains of chaotic areas and regular resonance
144: islands \cite{Zas}. These islands are impermeable for chaotic
145: trajectories and, at a first glance,
146: may be excluded from the consideration of
147: the phase space flow inside the stochastic layer.
148: In reality the phase space topology inside the stochastic layer
149: is very complex
150: precisely at the boundary between chaotic and
151: regular
152: regions \cite{Zas}. Close to resonances the stochastic
153: layer shows up with a hierarchical
154: set of
155: cantori, which form partial barriers for a trajectory from the
156: layer.
157: Due to the presence of these
158: barriers a chaotic trajectory can be trapped
159: for a long time near a regular island resonance. This trapping or
160: sticking effect
161: leads to the appearance of strongly nonergodic
162: episodes during the overall chaotic motion.
163: Regular islands are characterized by a corresponding
164: rational winding number $\omega=\Delta x / T$ which defines the distance
165: $\Delta x$ travelled
166: during one period $T=2\pi$ of the drive.
167: If the winding number
168: $\omega$ is nonzero, the corresponding sticking episode
169: of the chaotic trajectory is a ballistic-like unidirectional
170: flight. For $\omega=0$ the sticking episode corresponds to trapped oscillations.
171:
172: Thus the complicated evolution of a trajectory in the stochastic layer
173: can be subdivided into several parts \cite{Zas}. The first one is a fast
174: diffusion in the bulk of the layer, while the other ones are
175: stickings to the above mentioned regular islands and correspond
176: to propagation in
177: ballistic channels. The switching from the diffusion
178: process into a ballistic flight will be described by some
179: probability distribution. The same will be true for the actual
180: residence or sticking time inside a given channel.
181: We will show for the cases studied that the fast diffusion
182: alone is not capable of explaining the observed dc current.
183: The main point is, that the leading mechanism of current
184: generation in the {\it stochastic layer} is related to the
185: desymmetrization of the strongly {\it non-stochastic} part
186: of the overall stochastic dynamics inside the layer.
187: Note that our kinetic energy choice
188: $p^2/2$ in (\ref{1}) implies that the stochastic layer is bounded
189: in $p$, so we will always expect ballistic channels to appear.
190:
191: Let us study the case of weak driving $E_1=0.252$ and $E_2=0.052$.
192: A Poincare map of the phase space flow for
193: $\phi=0$ is shown in Fig.1(a) \cite{com_poinc}.
194: The main stochastic layer (central location) shows up
195: with zero average velocity due to symmetry arguments (Fig.2).
196: The large hole in the middle of this layer corresponds to
197: regular trapped motion in the well of $U(x)$.
198: Additional resonances are seen above and below
199: the central layer. These thin ballistic-like but yet stochastic channels
200: have no overlap with the central layer.
201: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
202: %%%%%%%%%%%% FIG 1 %%%%%%%%%%%%%%%%%%%%%%%
203: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
204: \begin{figure}[htb]
205: \vspace{10pt}
206: \centerline{
207: \epsfig{file=fig1.eps,width=82mm,height=90mm,angle=270}
208: }
209: \vspace{2pt}
210: \caption{Poincare map for (a) $\phi=0$ and (b) $\phi=\pi/2$.}
211: \label{fig1}
212: \end{figure}
213: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
214:
215: A weak asymmetry $\phi=\pi/5$
216: leads to a slight deformation of the main stochastic layer and to a
217: desymmetrization of the overlap of the chaotic layer with higher-order
218: resonances and to the appearance of a positive current in the system.
219: Note that it still does not overlap with
220: the thin ballistic channels seen in Fig.1(a).
221: Most importantly we observe a nonzero average velocity
222: $<\dot x> \approx 0.05$ (Fig.2).
223:
224: Further increase
225: of the asymmetry, $\phi=\pi/2$,
226: results in an overlapping of the main stochastic
227: layer with the upper ballistic resonance
228: Fig.1(b). Note that at the same time the lower ballistic resonance
229: is not overlapping.
230: The average velocity increases to $<\dot x > \approx 0.2$,
231: which is four times larger than the result for $\phi=\pi/5$.
232: Standard harmonic mixing theories (see \cite{pr01},\cite{igph98}) would predict a
233: dependence $< \dot x > \sim \sin \phi$ and thus only an increase by
234: a factor of 1.7.
235:
236: In the $x(t)$ curves in Fig.2 we observe many ballistic flights.
237: For one of them an inset shows the corresponding Poincare map result,
238: which verifies that these flights correspond to stickings of the
239: chaotic trajectory to the upper balistic resonances.
240: A zooming of the $x(t)$ curves shows {\it self-similarity},
241: i.e. the seemingly random dynamics between observable long flights
242: is actually again composed of shorter flights and seemingly random dynamics
243: etc (cf. insets in Fig.2).
244:
245: In order to quantify our analysis of the symmetry broken dynamics
246: we compute the distribution of travelling times of
247: 'uniform' flights to the left $P_-(t_f)$ and
248: to the right $P_+(t_f)$ separately. Here 'uniform' means no change
249: of direction of motion \cite{com_num}. For each separate flight we note
250: both the time $t_f$ spent in this motion and the distance $x_f$ travelled.
251: The dependence of $x_f$ on $t_f$ is shown in the inset of
252: Fig.3(a) for $\phi=\pi/2$.
253: Similar to the other cases $\phi=0,\pi/5$
254: we observe a simple fork-like structure.
255: This is due to the fact that any considerable distance
256: covering in the stochastic
257: layer is realized through flights while sticking to the boundary
258: of the stochastic layer. The slopes in the inset of
259: Fig.3(a) are given by the corresponding
260: winding numbers of the layer boundaries. Note that the two fork parts
261: merge at values of $t_f \approx 10 T$.
262: In principle other ballistic channels with different
263: winding numbers might be present. Here they were too weak
264: to be detected.
265: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
266: %%%%%%%%%%%% FIG 2 %%%%%%%%%%%%%%%%%%%%%%%
267: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
268: \begin{figure}[htb]
269: \vspace{10pt}
270: \centerline{
271: \epsfig{file=fig2.eps,width=62mm,height=90mm,angle=270}
272: }
273: \vspace{2pt}
274: \caption{$x(t)$ for $\phi=0,\pi/5,\pi/2$ (lower, middle and upper
275: curves respectively). Left upper inset: Poincare map for ballistic
276: flight with $\phi=\pi/2$ as indicated by arrow. Right inset: zoom
277: of $x(t)$ for $\phi=\pi/2$.}
278: \label{fig2}
279: \end{figure}
280: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
281:
282:
283: In Fig.3(a) we show the corresponding distribution functions
284: $P_{\pm}(t_f)$ (again for $\phi=\pi/2$). They are obtained by counting
285: the number of flights with $t_f$ falling into a time window of size $T$.
286: For $t_f < 10 T$ we observe exponential
287: dependence of $P_{\pm}$ on $t_f$. These short flight distributions
288: are in fact independent of the direction of flight.
289: Skipping all longer flights would lead to the prediction
290: of nearly zero average velocity (restricting to flights of length
291: $t_f < 10T$ yields about one percent of the numerically observed
292: current).
293: Thus the desymmetrization will manifest itself for longer
294: flights.
295: For $t_f > 10 T$ a crossover to
296: a power law $P_{\pm} \sim t_f^{\alpha_{\pm}}$ takes place. Here we find
297: a significant desymmetrization for $\phi=\pi/5,\pi/2$. Estimating the
298: exponents \cite{num_errors}
299: we find for $\phi=\pi/5$: $\alpha_-\approx 2.5$,
300: $\alpha_+\approx 2.4$ and for $\phi=\pi/2$: $\alpha_-\approx 3.7$,
301: $\alpha_+\approx 2.3$. It is worthwhile noting that $\alpha < 3$
302: implies unidirectional anomalous diffusion with diverging second moments
303: of $P(t)$. The flights are coined {\it Levy flights} in such a case
304: \cite{Levy}.
305:
306: Following the continuous-time random
307: walk (CTRW) formalism \cite{Klafter} we propose
308: a generalized asymmetrical flight model capable of reproducing
309: the above results.
310: The applicability of the CTRW model follows from the
311: assumption that the presence of a random phase
312: with fast decaying correlations leads
313: to the absence of
314: correlations between consecutive flights, since they are
315: almost always
316: separated by
317: dispersive chaotic motion.
318: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
319: %%%%%%%%%%%% FIG 3 %%%%%%%%%%%%%%%%%%%%%%%
320: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
321: \begin{figure}[htb]
322: \vspace{10pt}
323: \centerline{
324: \epsfig{file=fig3.eps,width=62mm,height=90mm,angle=270}
325: }
326: \vspace{2pt}
327: \caption{$P_+(t_f)$ (solid line) and $P_-(t_f)$ (thick dashed line).
328: Insets: $x_f$ versus $t_f$. For parameters see text.}
329: \label{fig3}
330: \end{figure}
331: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
332:
333:
334: Assume that there
335: exist $N$ different resonances with winding numbers
336: $w_i$, $i=1,...,N$. Every resonance is characterized by a probability
337: distribution function (PDF) of sticking time
338: $S_i(t)$.
339: After finishing a random phase event, the probability of sticking to
340: the $i$th resonance
341: is $\rho_{i}$, $\sum_{i=1}^{N}\rho_{i}=1$.
342: The random phase residing time is characterized by a PDF
343: $S_{r}(t)$.
344: All functions $S_{i}(t)$ and $S_{r}(t)$ must have finite first moments,
345: due to the Kac theorem about finiteness of recurrence times in Hamiltonian
346: systems \cite{Zas}.
347: With these definitions we obtain
348: the following expression for the current:
349: \begin{equation}
350: J=\frac{\sum_{i=1}^{N}\omega_{i}\rho_i\langle t_{i} \rangle}
351: {\sum_{i=1}^{N}\rho_i\langle t_{i} \rangle + \langle t_{r} \rangle}
352: \label{6}
353: \end{equation}
354: where $\langle t_i \rangle = \int t P_i(t) {\rm d}t$.
355:
356: For the above discussed cases of $\phi=\pi/5,\pi/2$ we find
357: only two relevant ballistic channels - one with positive
358: winding number and a second one with negative winding number.
359: In order to properly obtain $S(t)$, we note that our numerically
360: obtained function $P(t)$ consists of a lot of short 'flights' as
361: defined through the numerics \cite{com_num}. These may be either
362: stickings to islands with zero winding number or chaotic motion.
363: We observe that for flight times $t_f > 10 T$ only ballistic flights
364: with nonzero winding number are obtained. So the functions $S_{\pm}(t)$
365: may be easily obtained from $P_{\pm}(t_f)$ by cutting the central part
366: $t < 10 T$ out and properly normalizing. In this case the expression
367: for the average current simplifies to
368: \begin{equation}
369: J = \frac{1}{\kappa(1+f)}(\omega_+ \langle t_+ \rangle
370: + f \omega_-\langle t_- \rangle )
371: \label{jnew}
372: \end{equation}
373: where the two constants $\kappa$ and $f$ can be obtained from
374: the total time of a simulation $T_{tot}$ and the numbers $N_{\pm}$ of ballistic
375: flights, $\kappa = T_{tot} / (N_+ + N_-)$ and $f=N_-/N_+$.
376:
377: For $\phi=\pi/5$ we obtain from the numerical runs
378: $f\approx 0.57$, $\kappa \approx 1900$, $\langle t_+\rangle \approx
379: \langle t_- \rangle \approx 220$ and $\omega_+ =10/6 \approx 1.67$,
380: $\omega_- = -1.5$.
381: In this case of {\it weak desymmetrization} the main source of
382: a nonzero current is the different probability to enter a right or
383: left going flight because $f \neq 1$. At the same time the
384: average flight times in both ballistic channels are nearly identical.
385: With the help of (\ref{jnew}) we find
386: $J \approx 0.056$ which is close to the numerically observed value 0.05.
387:
388: For the case $\phi=\pi/2$ we find
389: $f\approx 0.16$, $\kappa \approx 2600$, $\langle t_+\rangle \approx
390: 400$,
391: $\langle t_- \rangle \approx 150$ and $\omega_+ = 2$,
392: $\omega_-\approx -1.4$. Note that the above discussed overlap with
393: the upper resonance yields a further {\it strong desymmetrization} in the
394: probability to realize a left or right going flight, and in addition
395: the average flight times in both channels significantly differ.
396: Expression (\ref{jnew}) yields $J\approx 0.22$ which is in good agreement
397: with the numerically observed value 0.2.
398:
399: For stronger driving amplitude $E_1=3.26$, $E_2=1.2$ and $\phi=\pi/2$
400: we obtain an average velocity $\langle \dot{x} \rangle \approx 0.85$.
401: The corresponding $x_f(t_f)$ dependence
402: and the PDFs $P_{\pm}(t_f)$ are shown in Fig. 3(b).
403: The $x_f(t_f)$ dependence shows that more than two ballistic channels
404: are involved.
405: The asymmetry of the PDFs at short flight times indicates
406: that a considerable number of left-going flights becomes
407: dominating at short times,
408: in agreement with the tendency of the previous results.
409: This makes the application of the simplified sum rule (\ref{jnew})
410: impossible, instead the original definition (\ref{6}) should be
411: used. Careful analysis of the structure of the stochastic layer
412: shows that relevant resonances become embedded in the bulk of the
413: stochastic layer. While these structures are of rather small size,
414: they are frequently visited. A restriction to short flights $t_f <
415: 10 T$ now yields a considerable nonzero current which is however
416: {\it negative}, i.e. opposite to the total current value. Again
417: the long ballistic flights are necessary in order to properly
418: obtain the observed current value.
419:
420: In summary, we have explained the dynamical
421: mechanisms of current appearance in
422: driven Hamiltonian systems inside the stochastic layer
423: with broken time-reversal symmetry. The
424: key
425: source of such a directed transport is
426: the desymmetrization of flight probabilities in ballistic channels
427: inside the layer.
428: As it follows from our sum rule (\ref{6}) the resulting current
429: depends among other parameters on the average time spent in a
430: ballistic channel. These times will sensitively depend on
431: control parameters of the system if the exponent $\alpha$ becomes
432: less than 3. In such a case small changes may significantly alter
433: the current value as shown above.
434:
435: A recently proposed geometric approach of counting areas
436: and winding numbers is in principle also capable of obtaining
437: the observed mean value for the current \cite{sumrule}.
438: This approach may also require sophisticated studies of the fractal
439: structure of the chaotic layer. It represents
440: a nontrivial complementary result, since it, although not
441: explaining the dynamical mechanisms of current generation,
442: is capable of obtaining the average current value (provided the
443: sums in Eq.(3) of \cite{sumrule} converge fast enough).
444:
445: We thank M. Fistul and A. A. Ovchinnikov for useful discussions.
446:
447:
448:
449: \begin{thebibliography}{99}
450:
451: \bibitem{fjaajp97}
452: P. H\"anggi and R. Bartussek, in: Nonlinear Physics of Complex
453: Systems - Current Status and Future Trends, J. Parisi, S. C. M\"uller
454: and W. Zimmermann (Eds), Lecture Notes in Physics {\rm 476}, p. 294
455: (Springer, Berlin 1997);
456: F. J\"ulicher, A. Ajdari and J. Prost,
457: Rev. Mod. Phys. {\bf 69}, 1269 (1997).
458:
459: \bibitem{pr01}
460: P. Reimann, {\em Brownian Motors: Noisy Transport Far From
461: Equilibrium}, cond-mat/0010237 (2001).
462:
463: \bibitem{sfoyyz00}
464: S. ~Flach,O.~Yevtushenko, and Y.~Zolotaryuk,
465: Phys. Rev. Lett. {\bf 84}, 2358 (2000).
466:
467: \bibitem{igph00}
468: I. Goychuk and P. H\"anggi, in: Stochastic Processes in Physics,
469: J. Freund and T. P\"oschel (Eds) {\rm 557}, p. 7
470: (Springer, Berlin 2000);
471: T. Dittrich, R. Ketzmerick, M.-F. Otto and H. Schanz,
472: Ann. Phys. (Leipzig) {\bf 9}, 755 (2000).
473:
474: \bibitem{Licht}
475: A.~J. Lichtenberg and M.~A. Lieberman,
476: \newblock {\em Regular and chaotic dynamics}.
477: \newblock (Springer-Verlag., 2d. ed., 1992).
478:
479: \bibitem{oysfaaoyz01}
480: O. Yevtushenko, S. Flach, A. A. Ovchinnikov and Y. Zolotaryuk,
481: Europhys. Lett., in press (2001).
482:
483: \bibitem{Zas}
484: G.~M. Zaslavsky,
485: \newblock {\em Physics of chaos in Hamiltonian systems}.
486: \newblock (Imperial College Press, 1998).
487:
488: \bibitem{com_poinc}
489: The starting time of output was set to zero, subsequent points
490: are drawn after multiples of the drive period. The value of $x$
491: is mod $2\pi$, so the true trajectory may actually cover
492: large distances in $x$.
493:
494: \bibitem{igph98}
495: I. Goychuk and P. H\"anggi,
496: Europhys. Lett. {\bf 43}, 503 (1998).
497:
498: \bibitem{com_num}
499: Flights are monitored on a time grid with spacing $T$.
500: A flight end is defined by either a change of the velocity sign
501: or a shift in $x$ in the opposite direction.
502:
503:
504: \bibitem{num_errors}
505: Here and in the following our estimated numbers from fitting numerical
506: data are meant with a relative error of at least 10 percent.
507:
508: \bibitem{Levy}
509: {\it Levy Flights and Related Topics in Physics.},Lect. Not. in Phys.,
510: ed. M. ~F. Shlesinger and G. ~M. Zaslavsky, (Springer, 1994).
511:
512: \bibitem{Klafter}
513: G. Zumofen, J. Klafter, and A. Blumen, Phus. Rev.E, {\bf 47}, 2183
514: (1993);
515: J. Klafter and G. Zumofen, Phys. Rev. E, {\bf 49}, 4873 (1994).
516:
517: \bibitem{sumrule}
518: H. Schanz, M.-F. Otto, R. Ketzmerick and T. Dittrich,
519: nlin.CD/0011038.
520:
521:
522: \end{thebibliography}
523:
524:
525: \end{document}
526: