1: \documentclass[figures]{epl}
2: \usepackage{graphicx}
3: \usepackage{rcs}
4:
5: \title{Delocalization and Heisenberg's uncertainty relation}
6: \shorttitle{Delocalization and Heisenberg's uncertainty relation}
7: \shortauthor{G.-L. Ingold \etal}
8: \author{%
9: Gert-Ludwig Ingold\thanks{E-mail: \email{Gert.Ingold@physik.uni-augsburg.de}}
10: \and
11: Andr{\'e} Wobst%\thanks{E-mail: \email{Andre.Wobst@physik.uni-augsburg.de}}
12: \and
13: Christian Aulbach%\thanks{E-mail: \email{Christian.Aulbach@physik.uni-augsburg.de}}
14: \and
15: Peter H{\"a}nggi%\thanks{E-mail: \email{Peter.Hanggi@physik.uni-augsburg.de}}
16: }
17: \institute{Institut f{\"u}r Physik, Universit{\"a}t Augsburg,
18: Universit{\"a}tsstra{\ss}e 1, D-86135 Augsburg,\\ Germany}
19: %\date{\RCSRevision\ / \RCSDate}
20: %\RCS $Revision: 1.19 $
21: %\RCS $Date: 2002/04/29 14:48:03 $
22: \pacs{05.60.Gg}{Quantum transport}
23: \pacs{71.23.An}{Theories and models; localized states}
24:
25: \begin{document}
26: \maketitle
27:
28: \begin{abstract}
29: In the one-dimensional Anderson model the eigenstates are localized for
30: arbitrarily small amounts of disorder. In contrast, the Harper model
31: with its quasiperiodic potential shows a transition from extended to
32: localized states. The difference between the two models becomes particularly
33: apparent in phase space where Heisenberg's uncertainty relation imposes a
34: finite resolution. Our analysis points to the relevance of the coupling
35: between momentum eigenstates at weak potential strength for the delocalization
36: of a quantum particle.
37: \end{abstract}
38:
39: \section{Introduction}
40: The delocalization of two interacting quantum particles in a disordered
41: potential has been the subject of intensive research since it was first
42: addressed \cite{shepe94,imry95,weinm95}. More recent work has generalized
43: the Harper or Aubry-Andr{\'e} model \cite{harpe55,aubry80} to study the
44: behaviour of two interacting particles in a quasiperiodic potential
45: \cite{barel96,shepe96,eilme01}. Here, the interaction rather leads
46: to a tendency towards localization. For finite densities, no clear indication
47: of an interaction dependence of the phase transition in the Harper model has
48: been found \cite{schus02}.
49:
50: Already in the absence of interaction, a quantum particle in one dimension
51: exhibits a very different behaviour depending on the potential in which it
52: is moving. For a periodic potential, the eigenstates are extended Bloch
53: waves characterized by a quasimomentum. On the other hand, already a small
54: amount of disorder suffices to localize the particle \cite{ander58}. A
55: different situation can arise when motion on a lattice is considered. Then,
56: for a periodic potential incommensurate with the underlying lattice, a
57: transition from delocalized to localized states occurs as the potential
58: strength reaches a critical value.
59:
60: In order to gain more insight into the localization properties of a quantum
61: particle in one dimension, we compare two lattice models, the Anderson model
62: \cite{ander58} and the Harper or Aubry-Andr\'e model \cite{harpe55,aubry80}.
63: An analysis of the phase space properties of the energy eigenstates will reveal
64: the dependence on the nature of the coupling between momentum eigenstates due
65: to the potential in which the particle is moving.
66:
67: \section{Random and quasiperiodic potential}
68: The Hamiltonian of the Anderson model is given by \cite{ander58}
69: \begin{equation}
70: H=-\sum_n\left(\vert n\rangle\langle n+1\vert + \vert n+1\rangle\langle n\vert
71: \right) + W\sum_n v_n\vert n\rangle\langle n\vert
72: \label{eq:ha}
73: \end{equation}
74: where the Wannier states $\vert n\rangle$ are localized at the sites
75: $n=1,\dots,L$ of a ring with periodic boundary conditions. The first term on
76: the right-hand side describes the kinetic energy which defines the energy
77: scale. The random potential of strength $W$ is expressed by the second term.
78: The distribution of on-site energies is determined by the coefficients $v_n$
79: distributed uniformly on the interval $[-1/2;1/2]$. In the limit $L\to\infty$
80: the eigenstates of the Anderson model are known to localize for any
81: nonvanishing potential strength $W$ \cite{ander58}.
82:
83: The second model of interest, the Harper model, is defined by the Hamiltonian
84: \cite{harpe55,aubry80}
85: \begin{equation}
86: H=\sum_n\left(\vert n\rangle\langle n+1\vert + \vert n+1\rangle\langle n\vert
87: \right) + \lambda\sum_n\cos\left(2\pi\beta n\right)
88: \vert n\rangle\langle n\vert
89: \label{eq:hhx}
90: \end{equation}
91: where the random potential in (\ref{eq:ha}) has been replaced by a
92: quasiperiodic potential if the parameter $\beta$ assumes an irrational number
93: in the limit $L\to\infty$. Then, the nature of the eigenstates
94: depends on the value of the parameter $\lambda$. For $\lambda<2$, all states
95: are extended while they are localized for $\lambda>2$ \cite{aubry80}.
96:
97: For finite size systems, it is convenient to choose $\beta=F_{i-1}/F_i$ where
98: $F_{i-1}$ and $F_i$ are two successive Fibonacci numbers \cite{kohmo83}. In the
99: limit of large systems $\beta$ approaches the inverse of the golden
100: mean, $(\sqrt{5}-1)/2$. With this choice of $\beta$ the system contains $L=F_i$
101: lattice sites with $F_{i-1}$ periods of the potential.
102:
103: The Harper model possesses an interesting duality property \cite{aubry80}
104: which becomes evident by transforming Wannier states $\vert n\rangle$ into
105: new states
106: \begin{equation}
107: \vert k\rangle = L^{-1/2}\sum_n \exp(i2\pi k\beta n)\vert n\rangle
108: = L^{-1/2}\sum_n \exp\left(i2\pi\frac{kF_{i-1}}{L} n\right)\vert n\rangle\,.
109: \label{eq:harpertrafo}
110: \end{equation}
111: These are eigenstates of the momentum operator to eigenvalues
112: $kF_{i-1}\,\mathrm{mod}\,F_i$. Neighbouring values of $k$ therefore do not
113: imply neighbouring momentum eigenvalues.
114:
115: With the transformation (\ref{eq:harpertrafo}) one obtains the dual Hamiltonian
116: \begin{equation}
117: H=\frac{\lambda}{2}\left[\sum_k\left(\vert k\rangle\langle k+1\vert +
118: \vert k+1\rangle\langle k\vert\right) + \frac{4}{\lambda}\sum_k
119: \cos\left(2\pi\beta k\right)\vert k\rangle\langle k\vert\right]\,.
120: \label{eq:hhk}
121: \end{equation}
122: By this transformation real and momentum space are interchanged and
123: the original potential strength $\lambda$ becomes inverted into $4/\lambda$.
124: Comparison of (\ref{eq:hhx}) and (\ref{eq:hhk}) yields the self-dual point
125: $\lambda=2$ which separates the parameter regimes of extended and localized
126: states. In contrast to the nearest neighbour coupling in real space in
127: (\ref{eq:hhx}), the new Hamiltonian does not couple nearest neighbour momenta.
128: The physical reason is that scattering by the incommensurate potential may
129: change the momentum by a large amount. This coupling to quite different
130: momentum values will be of central importance for our reasoning below.
131:
132: \section{Real and momentum space}
133: After this discussion, the question arises why these two archetypical models
134: behave so differently. In particular, what is the physical reason which allows
135: a localization transition at finite value $\lambda=2$ in the Harper model? In
136: order to answer this question, we first take a look at the structure of the
137: wave function $\vert\psi\rangle=\sum_n c_n\vert n\rangle$ expressed in terms
138: of the Wannier states $\vert n\rangle$, which will provide information about
139: the spatial extension of the state. An often used quantity is the inverse
140: participation ratio in real space \cite{thoul74,hashi92,mirli00}
141: \begin{equation}
142: P_x = \sum_n \vert c_n\vert^4\,.
143: \label{eq:iprx}
144: \end{equation}
145: Provided that $\sum_n\vert c_n\vert^2=1$, the inverse of this quantity
146: indicates the number of lattice sites over which the wave function is
147: distributed. A corresponding quantity
148: \begin{equation}
149: P_k = \sum_n \vert d_n\vert^4
150: \label{eq:iprk}
151: \end{equation}
152: can be defined in momentum space, where
153: \begin{equation}
154: d_n=L^{-1/2}\sum_l \exp\left(i2\pi\frac{nl}{L}\right)c_l\,.
155: \end{equation}
156:
157: The two quantities are depicted as full lines (real space) and dashed lines
158: (momentum space) in fig.~\ref{fig:iprxk}a for the Anderson model of length
159: $L=2048$ and fig.~\ref{fig:iprxk}b for the Harper model of length $L=10946$.
160: For the Anderson model, the curves represent an average over 50 disorder
161: realizations with $L/2$ states around the band center each. For the Harper
162: model, an average over all symmetric eigenstates has been taken \cite{thoul83}.
163:
164: For both models one observes a monotonously increasing inverse participation
165: ratio in real space which corresponds to an increasing localization of the
166: eigenfunctions as the potential strength is increased. Correspondingly, the
167: inverse participation ratio in momentum space decreases with increasing
168: potential strength, indicating a delocalization in momentum. The different
169: limiting values of $P_x$ for strong potential reflect the fact that in the
170: Anderson model the eigenfunctions localize at one site while in the Harper
171: model two sites are occupied because we consider here symmetric eigenstates.
172:
173: \begin{figure}
174: \centerline{\includegraphics[width=0.8\textwidth]{iprxk.eps}}
175: \caption{The inverse participation ratio in real space (full line) and
176: momentum space (dashed line) is shown (a) for the Anderson model with $L=2048$
177: and (b) for the Harper model with $L=10946$. The curves represent averages over
178: an ensemble of eigenstates of the respective Hamiltonian as explained in the
179: text.}
180: \label{fig:iprxk}
181: \end{figure}
182:
183: While the overall picture is qualitatively the same for both models, we note
184: an important difference which becomes apparent when the system size is changed.
185: In the Anderson model, the transition from extended to localized states is
186: smooth and notably shifts to lower values of $W$ as $L$ is increased. As a
187: consequence, in the limit of infinite system size all states are localized if
188: finite disorder is present. In contrast, for the Harper model one observes a
189: sharp transition at $\lambda=2$ for sufficiently large system sizes.
190:
191: \section{Phase space approach}
192: The differences between the two models become much more distinct in phase
193: space. This approach has the advantage of providing a consistent description
194: for arbitrary potential strength \cite{weinm99}. The inverse participation
195: ratio in phase space \cite{manfr00,sugit02,wobst02}
196: \begin{equation}
197: P=\int\frac{\mathrm{d}x\mathrm{d}k}{2\pi}[\varrho(x,k)]^2
198: \end{equation}
199: is based on the positive definite phase space density provided by the Husimi
200: function \cite{husim40} or Q function \cite{cahil69}
201: \begin{equation}
202: \varrho(x_0,k_0) = \vert\langle x_0,k_0\vert\psi\rangle\vert^2.
203: \end{equation}
204: Here, the state $\vert\psi\rangle$ is projected onto a minimal uncertainty
205: state centered around position $x_0$ and momentum $k_0$. In order to ensure
206: equal resolution in the two directions of phase space we choose the width
207: of the Gaussian as $\sigma=\Delta x=\sqrt{L/4\pi}=1/2\Delta k$.
208:
209: In fig.~\ref{fig:ipr} we present the inverse participation ratio in phase space
210: scaled with $L^{1/2}$ which is appropriate in the absence of a potential as
211: well as for very strong potentials \cite{wobst02}. For the Anderson model
212: (fig.~\ref{fig:ipr}a), one obtains an increased inverse participation ratio at
213: intermediate potential strengths implying that the eigenstates contract in
214: phase space. As we will demonstrate below, the behaviour to the left of the peak
215: is dominated by a contraction in real space corresponding to the increase of
216: $P_x$ (cf.\ fig.~\ref{fig:iprxk}a) while to the right of the peak the decrease
217: is dominated by the decrease of $P_k$.
218:
219: \begin{figure}
220: \centerline{\includegraphics[width=0.8\textwidth]{ipr.eps}}
221: \caption{The inverse participation ratio in phase space is shown (a) for the
222: Anderson model with $L=2048$ and (b) for the Harper model with $L=10946$.
223: The averages have been taken with respect to the same states as in
224: fig.~\protect\ref{fig:iprxk}.}
225: \label{fig:ipr}
226: \end{figure}
227:
228: The qualitative agreement between the inverse participation ratios in
229: real and momentum space for the Anderson and Harper model suggests that
230: the same should hold true for the inverse participation ratio in phase space.
231: This is even more so since the scenario just described for the Anderson model
232: is consistent with the duality property of the Harper model where an inversion
233: of the potential strength is accompanied by a transformation between real and
234: momentum space. However, the results depicted in fig.~\ref{fig:ipr}b tell a
235: different story.
236:
237: In contrast to the Anderson model, the inverse participation ratio $P$
238: shown in fig.~\ref{fig:ipr}b initially decreases with increasing potential
239: strength up to $\lambda=2$. In this regime, the eigenstates therefore become
240: more and more delocalized in phase space. Then, at $\lambda=2$, the phase space
241: distribution contracts and starts to become delocalized in momentum as
242: $\lambda$ is increased further. Therefore, for almost all values of $\lambda$,
243: the phase space behaviour is dominated by the momentum component except for the
244: transition which is dominated by the real space behaviour.
245:
246: The question now arises, why for weak potential the Anderson and Harper models
247: behave so differently while for strong potentials they behave in the same way.
248: The mechanism at work for weak potential can be considered to be responsible
249: for the localization transition in the Harper model because the jump found in
250: fig.~\ref{fig:ipr}b can only occur if the phase space distribution broadens
251: as $\lambda$ is increased from zero.
252:
253: \section{Uncertainty in phase space}
254: Before addressing this question in detail, we recall the Gaussian smearing
255: arising from the projection onto minimal uncertainty states which is inevitable
256: if a positive definite phase space density is required. As a consequence, the
257: Husimi function only provides limited resolution which, as discussed above, we
258: have chosen to be equal for the spatial and momentum components. The resolution
259: is of the order of $\sqrt{L}$ sites which in the thermodynamic limit becomes
260: small compared to the number $L$ of sites in real as well as momentum space. It
261: should be kept in mind, however, that for large $L$ a phenomenon occurring on a
262: fixed and finite number of sites cannot be resolved. These considerations would
263: still hold, if we chose $\Delta x$ and $1/\Delta k$ to scale according to
264: $L^{\alpha}$ with $0<\alpha<1$. In contrast, an entirely different situation
265: arises, if we keep the resolution fixed in one of the two phase space
266: directions as is the case for the inverse participation ratios in real and
267: momentum space, (\ref{eq:iprx}) and (\ref{eq:iprk}). Then, no effects occurring
268: in the other direction can be resolved even in the thermodynamic limit.
269:
270: \section{Limit of strong potential}
271: We are now in a position to answer the question raised above. It is useful to
272: start by considering the limit of strong potential where the phase space
273: behaviour of the Anderson and the Harper model agree. For infinitely strong
274: potential, the eigenstates of the Anderson model are localized at one site
275: while for the Harper model it is sufficient to restrict the discussion to one
276: of the two sites occupied by a symmetric eigenstate. The kinetic energy may now
277: be considered as a perturbation coupling in lowest order to the two neighbouring
278: sites. The dominant contribution will be due to the neighbouring site closer in
279: energy so that the problem reduces to the solution of a two-level system
280: \cite{wobstxx}.
281:
282: The energy eigenstates at very large but finite potential strength will be
283: delocalized over two sites. While this will reduce the inverse participation
284: ratio $P_x$ in real space, it will not affect the inverse participation ratio
285: $P$ in phase space which can only resolve spatial structures of size $\sqrt{L}$
286: and larger. We recall, that this argument is independent of our particular
287: choice of $\sigma$ since the absolute width of the minimal uncertainty state
288: has to increase with increasing system size even though its relative width
289: decreases to ensure a proper classical limit.
290:
291: The two original eigenstates which are coupled by means of the nearest-neighbour
292: hopping term were localized on one site each and therefore totally delocalized
293: in momentum space. Introducing the coupling, one finds a reduced spreading in
294: momentum in order to ensure the orthogonality of the two states. This can
295: easily be verified by considering the superpositions $(\vert n\rangle\pm\vert
296: n+1\rangle)/\sqrt{2}$ which in momentum space exhibit large scale density
297: oscillations of period $L$. By means of the Fourier transformation the effect
298: of the coupling on short real space distances, too small to be resolved in phase
299: space, is turned into a large scale phenomenon. Therefore, in the momentum
300: component the coupling is easily detected even with the finite resolution of
301: the Husimi function. As a consequence, the behaviour of the inverse
302: participation ratio in phase space is dominated by momentum and one finds an
303: increase of the inverse participation ratio with decreasing potential strength
304: as depicted in fig.~\ref{fig:ipr}.
305:
306: \section{Limit of weak potential}
307: We now apply similar considerations to the regime of weak potential by starting
308: from momentum eigenstates, \textit{i.e.}\ states well localized in momentum
309: analogous to the localized states in real space considered before. In the
310: Anderson model the random potential will lead to a coupling among all momentum
311: eigenstates. However, as before, the coupling between states close in energy,
312: and therefore close in momentum, will be most effective. As a consequence, the
313: role of position and momentum are interchanged and with the same arguments as
314: above, we find an increase of the inverse participation ratio in phase space
315: with increasing potential strength, albeit now due to the behaviour in real
316: space. Within this perturbative treatment, we can readily understand the
317: behaviour of $P$ depicted in fig.~\ref{fig:ipr}a.
318:
319: The situation is quite different for the Harper model where we may consider
320: the dual model (\ref{eq:hhk}) for small $\lambda$. The perturbation is now
321: represented by the first term on the right-hand side of (\ref{eq:hhk}) which
322: couples to well-defined momentum eigenstates. However, as remarked below
323: (\ref{eq:hhk}), due to the scattering by the incommensurate potential these
324: eigenstates in general do not correspond to nearest neighbour momenta. For most
325: of the energy eigenstates, the momentum eigenstates to which the coupling
326: occurs are far away on the scale of the resolution of the Husimi function. The
327: resulting broadening of the momentum distribution leads to a reduction of the
328: inverse participation ratio. In real space, on the other hand, the coupling
329: will lead to short scale oscillations which are not resolved because of the
330: finite resolution $\sigma$ in phase space. Therefore, the influence of the
331: momentum component dominates and the inverse participation ratio decreases with
332: increasing potential strength.
333:
334: Even in the Harper model there exist few particular states which couple to
335: states close in momentum. Then, the inverse participation ratio will initially
336: rise. However, the next order coupling leads to a distant momentum value and
337: the inverse participation ratio in phase space will eventually decrease before
338: reaching $\lambda=2$.
339:
340: It follows from this discussion that, in contrast to the Anderson model, the
341: Harper model for both weak and strong potential is dominated by the momentum
342: properties. It is only around $\lambda=2$ that real space becomes important.
343: From the comparison of the Anderson and the Harper model we conclude, that the
344: form of the coupling between the momentum eigenstates due to a weak potential
345: plays a decisive role for the structure of the eigenstates in phase space
346: and for the appearance of a delocalization-localization transition.
347:
348: As a further example we briefly comment on the Anderson model in two and three
349: dimensions where the inverse participation ratio in phase space behaves very
350: much like in the case of the Harper model (cf.\ fig.~\ref{fig:ipr}b)
351: \cite{wobst02}. In the marginal case of two dimensions, the tendency towards a
352: transition is therefore clearly visible, even though the critical disorder
353: strength vanishes in the thermodynamic limit \cite{wobstxx} and no true phase
354: transition occurs. In contrast, in three dimensions the Anderson transition is
355: recovered \cite{abrah79,lee85}.
356: The main difference between the Anderson model in one dimension on the one hand
357: and in two and three dimensions on the other hand lies again in the coupling
358: of momentum eigenstates by a weak random potential. In the one-dimensional
359: case, eigenstates close in energy are necessarily close in momentum. In
360: contrast, in higher dimensions there may exist even energetically degenerate
361: states far away in momentum, so that they can be resolved in phase space.
362:
363: \section{Conclusions}
364: A crucial aspect of our discussion was the finite resolution available in
365: phase space. This is in strong contrast to the ideal resolution available
366: with the inverse participation ratio in real or momentum space, albeit only
367: in one direction of phase space. As a consequence, there is no possibility
368: to resolve the other direction even in the limit of large system size.
369: Accepting Heisenberg's uncertainty relation and thus the finite resolution
370: in phase space allows one to analyze the structure of the eigenstates in
371: real as well as momentum space and, as was demonstrated above, to obtain
372: valuable information about the model of interest.
373:
374: This is corroborated by a recent observation by Varga \textit{et al.}\
375: \cite{varga02}, that instead of a full-fledged phase space calculation, one can
376: alternatively make use of marginal distributions in real and momentum space.
377: Even though the inverse participation ratios deduced from them resemble those
378: defined in eqs.\ (\ref{eq:iprx}) and (\ref{eq:iprk}) a Gaussian smearing is
379: again crucial.
380:
381: We therefore conclude that for the understanding of the localization properties
382: of a quantum particle, where both position and momentum are relevant, the
383: smearing in phase space called for by the uncertainty relation is not only
384: necessary but also represents an essential ingredient of the physical
385: argumentation.
386:
387: \acknowledgments
388: The authors thank D.~Weinmann, I.~Varga, C.~Schuster and H.~J.~Korsch for
389: interesting discussions. This work was supported by the
390: Son\-der\-for\-schungs\-be\-reich 484 of the Deutsche Forschungsgemeinschaft.
391: The numerical calculations were carried out partly at the Leibniz-Rechenzentrum
392: M{\"u}nchen.
393:
394: \begin{thebibliography}{00}
395:
396: \bibitem{shepe94}
397: \Name{Shepelyansky D. L.}
398: \REVIEW{Phys. Rev. Lett.}{73}{1994}{2607}.
399:
400: \bibitem{imry95}
401: \Name{Imry Y.}
402: \REVIEW{Europhys. Lett.}{30}{1995}{405}.
403:
404: \bibitem{weinm95}
405: \Name{Weinmann D., M{\"u}ller-Groeling A., Pichard J.-L. \and Frahm K.}
406: \REVIEW{Phys. Rev. Lett.}{75}{1995}{1598}.
407:
408: \bibitem{harpe55}
409: \Name{Harper P. G.}
410: \REVIEW{Proc. Phys. Soc. London Sect. A}{68}{1955}{874}.
411: This work had focussed on $\lambda=2$.
412:
413: \bibitem{aubry80}
414: \Name{Aubry S. \and Andr{\'e} G.}
415: \REVIEW{Ann. Israel Phys. Soc.}{3}{1980}{133}.
416:
417: \bibitem{barel96}
418: \Name{Barelli A., Bellisard J., Jacquod P. \and Shepelyansky D. L.}
419: \REVIEW{Phys. Rev. Lett.}{77}{1996}{4752}.
420:
421: \bibitem{shepe96}
422: \Name{Shepelyansky D.L.}
423: \REVIEW{Phys. Rev. B}{54}{1996}{14896}.
424:
425: \bibitem{eilme01}
426: \Name{Eilmes A., R{\"o}mer R. A. \and Schreiber M.}
427: \REVIEW{Eur. Phys. J. B}{23}{2001}{229}.
428:
429: \bibitem{schus02}
430: \Name{Schuster C., R{\"o}mer R. A. \and Schreiber M.}
431: \REVIEW{Phys. Rev. B}{65}{2002}{115114}.
432:
433: \bibitem{ander58}
434: \Name{Anderson P. W.}
435: \REVIEW{Phys. Rev.}{109}{1958}{1492}.
436:
437: \bibitem{kohmo83}
438: \Name{Kohmoto M.}
439: \REVIEW{Phys. Rev. Lett.}{51}{1983}{1198}.
440:
441: \bibitem{thoul74}
442: \Name{Thouless D. J.}
443: \REVIEW{Phys. Rep.}{13}{1974}{93}.
444:
445: \bibitem{mirli00}
446: \Name{Mirlin A. D.}
447: \REVIEW{Phys. Rep.}{326}{2000}{259}.
448:
449: \bibitem{hashi92}
450: \Name{Hashimoto Y., Niizeki K. \and Okabe Y.}
451: \REVIEW{J. Phys. A: Math. Gen.}{25}{1992}{5211}.
452:
453: \bibitem{thoul83}
454: \Name{Thouless D. J.}
455: \REVIEW{Phys. Rev. B}{28}{1983}{4272}.
456:
457: \bibitem{weinm99}
458: \Name{Weinmann D., Kohler S., Ingold G.-L. \and H{\"a}nggi P.}
459: \REVIEW{Ann. Phys. (Leipzig)}{8}{1999}{SI-277}.
460:
461: \bibitem{manfr00}
462: \Name{Manfredi G. \and Feix M. R.}
463: \REVIEW{Phys. Rev. E}{62}{2000}{4665}.
464:
465: \bibitem{sugit02}
466: \Name{Sugita A. \and Aiba H.}
467: \REVIEW{Phys. Rev. E}{65}{2002}{036205}.
468:
469: \bibitem{wobst02}
470: \Name{Wobst A., Ingold G.-L., H{\"a}nggi P. \and Weinmann D.}
471: \REVIEW{Eur. Phys. J. B}{}{2002}{} (in press);
472: arXiv:cond-mat/0110028.
473:
474: \bibitem{husim40}
475: \Name{Husimi K.}
476: \REVIEW{Proc. Phys. Math. Soc. Jpn.}{22}{1940}{264}.
477:
478: \bibitem{cahil69}
479: \Name{Cahill K. E. \and Glauber R. J.}
480: \REVIEW{Phys. Rev.}{177}{1969}{1882}.
481:
482: \bibitem{wobstxx}
483: \Name{Wobst A., Ingold G.-L., H{\"a}nggi P., Weinmann D.}
484: (unpublished).
485:
486: \bibitem{abrah79}
487: \Name{Abrahams E., Anderson P. W., Licciardello D. C. \and Ramakrishnan T. V.}
488: \REVIEW{Phys. Rev. Lett.}{42}{1979}{673}.
489:
490: \bibitem{lee85}
491: \Name{Lee P. A. \and Ramakrishnan T. V.}
492: \REVIEW{Rev. Mod. Phys.}{57}{1985}{287}.
493:
494: \bibitem{varga02}
495: \Name{Varga I. \and Pipek J.}
496: arXiv:cond-mat/0204041.
497:
498: \end{thebibliography}
499:
500: \end{document}
501: