1: % jltpcls.tex v0.2
2: % Example JLTP camera-ready manuscript.
3:
4: \documentclass{jltp}
5:
6: \usepackage{graphicx} % uncomment this line to include the graphicx package
7:
8: \title{Ballistic transport in classical and quantum integrable systems}
9:
10: \author{Xenophon Zotos}
11:
12: \address{Institut Romand de Recherche Num\'erique en Physique des
13: Mat\'eriaux,\\ PPH-Ecublens, CH-1015 Lausanne, Switzerland }
14:
15: \runninghead{X. Zotos}{Ballistic transport}
16:
17: \begin{document}
18:
19: \maketitle
20:
21: \begin{abstract}
22: In this essay, we first sketch the development of ideas on the
23: extraordinary dynamics of integrable classical nonlinear
24: and quantum many body Hamiltonians. In particular, we comment on the state
25: of mathematical techniques available for analyzing their thermodynamic and
26: dynamic properties.
27:
28: Then, we discuss the unconventional finite temperature transport
29: of integrable systems using as example the classical Toda chain and
30: the toy model of a quantum particle interacting with a fermionic bath in
31: one dimension; we focus on the singular long time asymptotic of
32: current-current correlations, we introduce the notion of the Drude weight
33: and we emphasize the role played by conservation laws in establishing
34: the ballistic character of transport in these systems.
35:
36: PACS numbers: 71.27.+a, 05.45, 72.10.-d
37: \end{abstract}
38:
39: \section{INTRODUCTION}
40: The extraordinary stability of solitons upon collisions in integrable
41: nonlinear systems was first discovered in a numerical simulation
42: of the Korteweg-de-Vries evolution equation, a system commonly studied
43: in hydrodynamics and plasma physics\cite{zk}. This discovery was soon followed
44: by the development of a beautiful mathematical theory - the Inverse Scattering
45: Method (ISM) - that allows the analytical evaluation of the time evolution of
46: an initial pulse configuration using linear operations\cite{ism}; this is the
47: analogue of the Fourier Transform for linear systems.
48: These seminal works opened the new and still rapidly expanding field
49: of nonlinear physics, with developments ranging from mathematical physics to
50: applications\cite{os}.
51:
52: Here, we should point out that there is a fundamental distinction between
53: {\it integrable systems} (mostly one dimensional) characterized by the presence
54: of ``mathematical solitons" and those with ``topological solitons".
55: On the one hand, the stability of mathematical solitons is guaranteed by a
56: subtle interplay between dispersion and nonlinearity; this interplay is
57: expressed by the existence of a macroscopic number of
58: {\it conservation laws} constraining the dynamical evolution.
59: On the other hand, the stability of topological solitons
60: is enforced by a topological constraint; of course there are examples, like
61: the sine-Gordon field theory, which possess topological solitons but they
62: are also integrable. In this work we will focus on the
63: transport properties of integrable systems.
64:
65: Due to the presence and stability upon scattering of nonlinear excitations,
66: integrable systems are expected to show unconventional finite temperature
67: transport properties, as ideal thermal, charge or spin
68: conductivities, ballistic rather than diffusive transport.
69: Within the traditional framework of linear response theory,
70: the finite temperature dynamic correlations characterize the transport
71: behavior and they are directly linked to experimental observations.
72: However, although integrable models are
73: considered as exactly soluble, meaning that the initial
74: value problem can be exactly analyzed using the ISM, rather little
75: progress has been done in the evaluation of
76: dynamic correlations which remains at best technically very involved.
77:
78: In the quantum domain,
79: parallel to developments on the analysis of classical integrable nonlinear
80: systems, in the early 80's it was realized that the exact solution of a
81: certain class of one dimensional quantum models by the Bethe ansatz (BA) method
82: was equivalent to a quantum version of the ISM\cite{kor}.
83: In this class belong well known prototype systems as the Hubbard or spin-1/2
84: Heisenberg model, commonly used for the description of (quasi) one-dimensional
85: electronic or magnetic materials.
86: The BA method provides the exact eigenfunctions and eigenvalues
87: and by a certain procedure (and assumptions) the exact thermodynamic properties
88: and excitation spectrum. Similarly to their classical counterparts, the
89: quantum systems possess a macroscopic number of conservation laws,
90: characteristic of their integrability.
91: It should therefore come as no surprise the proposition that quantum
92: integrable systems should also exhibit unconventional transport\cite{czp}.
93: The situation however is similar to the one of classical systems;
94: although the exact eigenfunctions and eigenvalues are known, the
95: calculation of finite temperature dynamic correlations is still out of
96: reach for most of the models.
97:
98: In the following, we discuss two simple examples, one classical and one
99: quantum, in order to show the
100: ideal conducting properties of integrable systems.
101: Our strategy is to focus on the {\it long time} asymptotic
102: value of the current correlations in order to demonstrate the ballistic
103: transport instead of attempting to evaluate the full
104: frequency dependence of the conductivity (or mobility).
105: Furthermore, instead of linking the ideal conductivity
106: to the dynamics of soliton excitations, not very transparent
107: for quantum many body systems, we directly relate it to
108: the conservation laws characterizing integrable systems.
109:
110: First, we present a study of the energy current - current correlations
111: (related to the thermal conductivity) for the classical
112: Toda chain.
113: Second, we introduce the notion of the {\it Drude weight} as a criterion of
114: ideal conductivity and evaluate it exactly using the Bethe ansatz
115: method in the context of the mobility for a toy model describing a
116: quantum particle interacting with a fermionic bath.
117:
118: \section{A classical system: the Toda chain}
119:
120: The classical Toda lattice is a prototype model for studying
121: the physics of nonlinear excitations\cite{toda}. It is one of the first
122: models analyzed using the Inverse Scattering method\cite{flaschka},
123: the conservation laws characterizing this system\cite{henon} have been
124: presented and it has even been invoked in attempts to describe nonlinear
125: transport in DNA molecules\cite{dna}.
126: As we mentioned above,
127: although the initial value problem and the thermodynamic properties can
128: be analytically studied, there is no clear picture on the finite
129: temperature dynamic correlations\cite{ss}.
130:
131: A physical quantity of interest in an anharmonic chain is the heat
132: conductivity. In a generic case, it is expected that the
133: energy current correlations decay to zero at long times and the decay is
134: fast enough so that a transport coefficient can be defined.
135: This behavior seems rather difficult to observe in several one dimensional
136: systems\cite{livi,gr}, where the currents decay to zero but often too slowly,
137: leaving the issue of diffusive transport controversial.
138:
139: For an integrable model, ideal conducting behavior is expected
140: with current correlations decaying to a finite value at long times.
141: To quantify the contribution of nonlinear excitations, different ingenious
142: methods have been devised\cite{dna} (soliton counting procedures).
143: Here, we will use the long time asymptotic value of current
144: correlations as a measure of ideal transport, related in integrable systems
145: to the existence of conservation laws.
146:
147: To establish this relation, we will use an inequality proposed
148: by Mazur\cite{mazur}, linking the long time
149: asymptotic of dynamic correlations functions to the presence of
150: conservation laws:
151: \begin{equation}
152: \lim_{T\rightarrow \infty} \frac{1}{T} \int_0^T \langle A(t)A\rangle
153: dt \geq \sum_n
154: \frac{\langle A Q_n\rangle^2}{\langle Q_n^2\rangle} .
155: \label{mazur}
156: \end{equation}
157: Here $\langle \rangle$ denotes thermodynamic average,
158: the sum is over a set of conserved quantities ${Q_n}$, orthogonal
159: to each other $\langle Q_n Q_m\rangle=\langle Q_n^2\rangle\delta_{n,m}$
160: and we suppose that $\langle A\rangle=0$.
161:
162: The classical Toda Hamiltonian for a chain of $L$ sites with periodic
163: boundary conditions is given in reduced units by:
164: \begin{equation}
165: H=\sum_{l=1}^L \frac{p_l^2}{2}+e^{-q_l}
166: \label{toda}
167: \end{equation}
168: where $p_l$ is the momentum of particle $l$, $x_l$ its position and
169: $q_l=x_{l+1}-x_l$.
170:
171: The energy current for a system of interacting particles is given
172: by\cite{mcl}:
173: \begin{equation}
174: j^E=\sum_{l=1}^L p_l h_l+\frac{(p_{l+1}+p_l)}{2}q_l e^{-q_l}
175: \label{je}
176: \end{equation}
177: where $h_l=\frac{p_l^2}{2}+\frac{1}{2}(e^{-q_l}+e^{-q_{l-1}})$.
178:
179: We consider dynamic correlation functions in the fixed
180: temperature-pressure thermodynamic ensemble:
181: \begin{equation}
182: \langle A(t)A\rangle=Z^{-1}\int \prod_{l=1}^L dp_l dq_l A(t)A e^{-\beta(H+PL)}
183: \label{ave}
184: \end{equation}
185: where $Z=\int \prod_{l=1}^L dp_l dq_l e^{-\beta(H+PL)}$,
186: $L=\sum_{l=1}^L q_l$, $P$ is the pressure and $\beta$ the inverse
187: of the temperature.
188:
189: In this thermodynamic ensemble, equal time correlation functions can be
190: calculated analytically. For instance the average distance is
191: given by:
192: \begin{equation}
193: \langle q\rangle=\ln(\beta)-\Psi(\beta P)
194: \label{q}
195: \end{equation}
196: where $\Psi(z)$ is the digamma function.
197:
198: The classical Toda lattice is characterized by a macroscopic number of
199: conservation laws. The first few ones are:
200: \begin{eqnarray}
201: Q_1&=&\sum_{l=1}^L p_l\\
202: Q_2&=&\sum_{l=1}^L \frac{p_l^2}{2}+e^{-q_l}\\
203: Q_3&=&\sum_{l=1}^L \frac{p_l^3}{3}+(p_l+p_{l+1})e^{-q_l}\\
204: Q_4&=&\sum_{l=1}^L \frac{p_l^4}{4}+(p_l^2+p_lp_{l+1}+p_{l+1}^2)e^{-q_l}
205: +\frac{1}{2}e^{-2q_l}+e^{-q_l}e^{-q_{l+1}}\\
206: Q_5&=&\sum_{l=1}^L \frac{p_l^5}{5}+(p_l^3+p_l^2p_{l+1}+p_lp_{l+1}^2+p_{l+1}^3)
207: e^{-q_l}\\
208: &+&(p_l+p_{l+1})e^{-2q_l}+(p_l+2p_{l+1}+p_{l+2})e^{-q_l}e^{-q_{l+1}}...
209: \label{laws}
210: \end{eqnarray}
211: with $Q_1$ the total momentum, of course present in all
212: translationally invariant systems, integrable or not, as also $Q_2$ the total
213: energy. According to the standard Green-Kubo formulation of
214: transport theory\cite{mcl} ``subtracted fluxes" should be used in the
215: dynamic correlation functions determining the transport coefficients.
216: So in the case of energy transport we will study the decay of the
217: ``subtracted" energy current\cite{am}:
218: \begin{equation}
219: \tilde j^E=j^E-\frac{\langle Q_1 j^E\rangle}{\langle Q_1^2\rangle}Q_1
220: \label{jet}
221: \end{equation}
222: We see that the use of a subtracted flux is equivalent to removing the
223: contribution of $Q_1$ in the long time asymptotic bound\cite{mazur}
224: for $\langle j^E(t)j^E\rangle$.
225:
226: We will now calculate a bound on the long time asymptotic value of
227: $\langle j^E(t)j^E\rangle$ by the Mazur inequality eq.(\ref{mazur})
228: using the first $m$ conservation laws.
229: We should note that $Q_3$ has a structure very similar to the energy
230: current, so we expect a large contribution from this term; actually
231: in some quantum models like the spin-1/2 Heisenberg chain or the t-J model,
232: the energy current is identical to a conservation law, directly implying
233: a nondecaying energy current and thus infinite thermal conductivity\cite{znp}.
234: Here, $Q_n's$ with $n=$even do not couple
235: to $\tilde j^E$ so we will consider only $Q_n, n=3,5,7$. Higher
236: $Q_n's$ can of course be included but the calculations become
237: rather cumbersome.
238: Orthogonalizing the conserved quantities which appear in the right hand
239: side of eq.(\ref{mazur}) is equivalent to evaluating the expression:
240: \begin{equation}
241: C^m_{j^Ej^E}=\langle\tilde j^E|Q\rangle\langle Q|Q\rangle^{-1}
242: \langle Q|\tilde j^E\rangle
243: \label{ortho}
244: \end{equation}
245: where $\langle Q|Q\rangle$ is the $m\times m$ overlap matrix of $Q_n'n$ and
246: $\langle Q|\tilde j^E\rangle$ the overlap vector of $\tilde j^E$
247: with the $Q_n's$.
248:
249: \begin{figure}
250: %
251: \centerline{\includegraphics[height=3.0in]{fig1_pw.eps}}
252: %
253: %\framebox[5in]{\rule[1.125in]{0in}{1.125in}}
254: %\makebox[5in]{\rule[1.125in]{0in}{1.125in}}
255: \caption{Lower bounds on the long time asymptotic value of energy current
256: correlations.}
257: \label{fig1}
258: \end{figure}
259:
260: In Fig.1 we show the temperature dependence of
261: $C^m_{j^Ej^E}/\langle\tilde {j^E}^2\rangle$ for $m=1~(n=3)$, $m=2~(n=3,5)$ and
262: $m=3~(n=3,5,7)$.
263: At low T the behavior is linear with slope $\frac{3}{35}$ for $m=2$ and
264: $\frac{4}{63}$ for $m=3$. It is interesting that this value is comparable
265: to the value for the density of solitons\cite{mb}
266: $N_s/N=ln(2)/\pi^2 T$. So, we find that the long time asymptotic value of the
267: subtracted energy current correlations is finite and most interestingly that
268: it increases with temperature. This trend we can interpret as evidence for an
269: increasing contribution of thermally excited nonlinear excitations on the
270: ballistic transport.
271:
272: The idea presented here provides, on the one hand a
273: conceptual understanding of the role played by the conservation laws
274: on the finite temperature dynamic correlations and on the other hand
275: a simple analytical method for evaluating, or at least giving bounds on their
276: long time asymptotic value. A similar analysis can be carried out for
277: quantum integrable systems\cite{znp} although the complexity of the quantum
278: conservation laws renders their wide use rather limited.
279:
280: \section{A quantum system: the ``heavy particle" model}
281:
282: The Drude weight $D$ (or charge stifness) was introduced as a
283: criterion of an ideal conducting or insulating state at zero
284: temperature\cite{kohn} and
285: recently extended as a measure of ideal transport at finite
286: temperatures\cite{czp}. Within linear response theory,
287: it is essentially the prefactor of the low frequency
288: reactive part of the conductivity,
289: $\sigma''=2D/\omega|_{\omega \rightarrow 0}$, a finite $D$
290: implying a freely accelerating system.
291: For a normal, diffusive system the Drude weight is zero at any finite
292: temperature in the thermodynamic limit; according to the standard scenario
293: the weight at zero frequency spreads to a ``Drude peak" with width
294: proportional to the inverse of the collision time.
295: As we will see below, in integrable systems the Drude weight remains
296: finite at all temperatures indicative of ballistic rather than diffusive
297: transport.
298:
299: The Drude weight can be conveniently evaluated\cite{kohn} as the thermal
300: average of curvatures of energy levels $\epsilon_n$ of the system subject
301: to a fictitious flux $\phi$,
302:
303: \begin{equation}
304: D=\frac{1}{2L}\sum_n p_n \frac{\partial^2 \epsilon_n}{\partial \phi^2}|_
305: {\phi \rightarrow 0}
306: \label{drude}
307: \end{equation}
308: where $p_n$ are the Boltzmann weights and the sum is over all eigenstates of
309: the system. It is also equal to the
310: long time asymptotic value of the current-current correlations\cite{znp},
311:
312: \begin{equation}
313: D=\frac{\beta}{2L}\langle j(t)j\rangle|_{t\rightarrow \infty}=
314: \frac{\beta}{2L} \sum_n p_n |\langle n|j|n\rangle|^2
315: \label{ltc}
316: \end{equation}
317: so useful bounds on $D$ can be obtained using the Mazur inequality following
318: the same formulation as in the previous example.
319:
320: For integrable quantum many body models it can evaluated exactly
321: following recent developments in the Bethe ansatz technique,
322: thus providing essential information on the transport
323: properties of these systems without requiring the full calculation of the
324: frequency dependence of the conductivity.
325: This type of analysis, still under discussion as it is technically
326: involved, has been carried out
327: for several one dimensional integrable quantum models as the Hubbard chain,
328: the spin 1/2 Heisenberg and the nonlinear-$\sigma$ model\cite{fk,f,xz}.
329: These calculations show that in most of the cases the Drude weight is
330: finite at all temperatures implying ideal thermal, charge or spin
331: conductivity. Recently, investigation of the finite temperature of these
332: systems was also carried out by a semiclassical approach\cite{sachdev} and
333: within the Luttinger liquid description\cite{gm,ra}.
334:
335: Here, we will demonstrate the idea behind this type of Bethe ansatz analysis
336: by evaluating the Drude weight related to the mobility of a quantum particle
337: interacting with a bath of fermions in a one dimensional system\cite{mcguire}.
338: A similar analysis was carried out for a particle moving on a lattice\cite{czp}
339: but the case discussed below is simpler and shows a qualitatively
340: new behavior.
341:
342: We consider a particle with coordinate $y$ moving on a system of length $L$
343: with periodic boundary conditions and interacting with a set of N
344: fermions described by the coordinates $x_j$ via a $\delta-$function
345: interaction of strength $c$,
346:
347: \begin{equation}
348: H=-\sum_j\frac{\partial^2}{\partial x_j^2}
349: -\frac{\partial^2}{\partial y^2}+2c\sum_j\delta(x_j-y).
350: \label{hham}
351: \end{equation}
352:
353: When the mass of the ``heavy particle" is equal to the mass of the fermions
354: the model is integrable and so we expect a ballistic behavior of the
355: mobility and therefore a finite Drude weight.
356: To evaluate $D$ using eq. (\ref{drude}), we
357: consider the dependence of the energy levels
358: on a flux $\phi$ acting only on the heavy particle.
359: The momenta $k_j$ and the collective coordinate $\Lambda$
360: describing the Bethe ansatz wavefunctions are then given by the following
361: standard equations obtained by applying periodic boundary conditions,
362:
363: \begin{eqnarray}
364: Lk_j&=&2\pi I_j +\theta(k_j-\Lambda),~~~j=1,...,N+1,\\
365: \theta(p)&=&-2\tan^{-1}(2p/c),\\
366: L\sum_{j=1}^{N+1}k_j&=&2\pi\sum_{j=1}^{N+1}I_j+2\pi J+L \phi.
367: \label{pbc}
368: \end{eqnarray}
369: The eigenstates are characterized by the quantum numbers $(I_j,J)$ and their
370: energy is given by:
371:
372: \begin{equation}
373: E=\sum_{j=1}^{N+1} \epsilon(k_j)=\sum_{j=1}^{N+1} k_j^2.
374: \label{energy}
375: \end{equation}
376: These equations can be solved to order $1/L$ as we consider the effect
377: of the one particle on the ensemble of fermions.
378: \begin{equation}
379: k_j=k_j^0+\frac{1}{L}\theta(k_j-\Lambda),~~~k_j^0=\frac{2\pi I_j}{L}
380: \label{k}
381: \end{equation}
382: Thus the total energy can be written as,
383: \begin{eqnarray}
384: &&E=\sum_j\epsilon(k_j^0)+\frac{2}{L} k_j^0 \theta(k_j^0-\Lambda),\\
385: &&\frac{1}{L}\sum_j\theta(k_j^0-\Lambda)=\frac{2\pi J}{L}+\phi.
386: \label{o1}
387: \end{eqnarray}
388: Going to the continuum limit we obtain:
389: \begin{eqnarray}
390: &&E(\rho(k),\Lambda)=\frac{L}{2\pi}\int dk \rho(k) (k^2
391: +\frac{2}{L} k \theta(k-\Lambda)),\\
392: &&P+\phi=\frac{1}{2\pi}\int dk \rho(k) \theta(k-\Lambda),
393: ~~~P=\frac{2\pi J}{L}.
394: \label{cont}
395: \end{eqnarray}
396:
397: Now we can define a correlation energy $\epsilon_c(\Lambda)$
398: assuming that the distribution of the fermion momenta is not affected
399: by the presence of the extra particle and replacing the density
400: $\rho(k)$ by the Fermi-Dirac distribution $f(k)$,
401:
402: \begin{equation}
403: \epsilon_c(\Lambda)=\frac{1}{2\pi}\int dk f(k) 2k \theta(k-\Lambda).
404: \label{ec}
405: \end{equation}
406: Using this formulation and the definition of the Drude weight eq.(\ref{drude})
407: we obtain:
408:
409: \begin{equation}
410: D=\frac{1}{2\pi Z_{\Lambda}}\int d\Lambda g(\Lambda) w(\Lambda)
411: \frac{1}{2}\frac{\partial^2 \epsilon_c(\Lambda)}{\partial \phi^2}
412: \label{d}
413: \end{equation}
414: where,
415: \begin{equation}
416: g(\Lambda)=\frac{\partial P}{\partial\Lambda}=\frac{1}{2\pi}
417: \int dk f(k) \frac{\partial \theta (k-\Lambda)}{\partial \Lambda},
418: \label{g}
419: \end{equation}
420:
421: \begin{equation}
422: Z_{\Lambda}=\frac{1}{2\pi}\int d\Lambda g(\Lambda) w(\Lambda),~~~
423: w(\Lambda)=e^{-\beta\epsilon_c(\Lambda)}.
424: \label{z}
425: \end{equation}
426:
427:
428: \begin{figure}
429: %
430: \centerline{\includegraphics[height=3.0in]{fig2_pw.eps}}
431: %
432: %\framebox[5in]{\rule[1.125in]{0in}{1.125in}}
433: %\makebox[5in]{\rule[1.125in]{0in}{1.125in}}
434: \caption{Drude weight of the ``heavy particle".}
435: \label{fig2}
436: \end{figure}
437:
438: In Fig. 2 we show the normalized Drude weight $D/D_0$ for different values
439: of the interaction $c$ as a function of temperature, with
440: $D_0=D(T=0)=(\pi/2)(\tan^{-1}(2k_F/c)-(2k_F/c)/(1+(2k_F/c)^2))/
441: \tan^{-1}(2k_F/c)^2$ and $k_F=\pi n$.
442: The chemical potential is chosen so that we consider density $n=1$;
443: upon scaling $n\rightarrow nc, \beta\rightarrow \beta/ c^2$
444: the Drude weight $D$ remains the same.
445: We note that the behavior of $D$ is not monotonic, initially decreasing at low
446: temperatures because of the interaction and then tending to the free
447: particle value at high temperatures. This is in contrast to the Drude weight
448: of systems on a lattice (tight binding models) where $D$ goes always to
449: zero as $\beta$ at high temperatures.
450: The difference in behavior can be attributed to the bounded spectrum
451: in lattice models in contrast to the unbounded one for continuous models.
452: Furthermore, numerical calculations on this model show that,
453: the Drude weight vanishes at any finite temperature when the mass of the
454: ``heavy particle" is not equal to that of fermions, as expected for any
455: normal system.
456: In conclusion, the presented analysis demonstrates the basic features of
457: the generic finite temperature ballistic transport behavior of integrable
458: quantum many body systems.
459:
460: \section*{ACKNOWLEDGMENTS}
461: It is my honor and pleasure to dedicate this paper to Professor Peter
462: W\"olfle on the occasion of his 60th birthday.
463: This research was supported by the Swiss National Science Foundation,
464: the University of Fribourg and Neuch\^atel.
465: I would also like to thank P. Prelov\v sek, H. Castella and F. Naef for
466: our collaboration in the development of the ideas here presented.
467:
468: \begin{thebibliography}{9}
469: \bibitem{zk} N. J. Zabusky and M. D. Kruskal, {\it Phys. Rev. Lett.}
470: {\bf 15}, 240 (1965).
471: \bibitem{ism} C. S. Gardner, J. M. Greene, M. D. Kruskal and R. M. Miura,
472: {\it Phys. Rev. Lett.} {\bf 19}, 1095 (1967).
473: \bibitem{os}``Optical solitons: theoretical challenges and
474: industrial perspectives", editors: V.E. Zakharov and S. Wabnitz,
475: {\it Les Houches Workshop}, Springer (1998).
476: \bibitem{kor} ``Quantum Inverse Scattering Method and Correlation
477: Functions", V.E. Korepin, N.M. Bogoliubov and A.G. Izergin,
478: {\it Cambridge Univ. Press} (1993).
479: \bibitem{czp} H. Castella, X. Zotos, P. Prelov\v sek, {\it Phys. Rev. Lett.}
480: {\bf 74}, 972 (1995).
481: \bibitem{toda} M. Toda, ``Theory of Nonlinear Lattices",
482: {\it Springer Series in Solid-State Sciences 20, Springer-Verlag}, (1981).
483: \bibitem{flaschka} H. Flaschka, {\it Phys. Rev.} {\bf B9}, 1924 (1974).
484: \bibitem{henon} M. Henon, {\it Phys. Rev.} {\bf B9}, 1921 (1974).
485: \bibitem{dna} V. Muto, A.C. Scott and P.L. Christiansen, {\it Physica}
486: {\bf D44}, 75 (1990).
487: \bibitem{ss} See for instance: T. Schneider and E. Stoll,
488: {\it Phys. Rev. Lett.}
489: {\bf 45}, 997 (1980); A. Cuccoli et al., {\it Phys. Rev.}
490: {\bf B47}, 7859 (1993); N. Theodorakopoulos and M. Peyrard, {\it Phys. Rev.
491: Lett.} {\bf 83}, 2293 (1999).
492: \bibitem{livi} C. Giardina et al., {\it Phys. Rev. Lett.} {\bf 84},
493: 2144 (1998).
494: \bibitem{gr} See however the system analyzed in,
495: M. Garst and A. Rosch, {\it Eur. Lett.} {\bf 55}, 66 (2001).
496: \bibitem{mazur} P. Mazur, {\it Physica} {\bf 43}, 533 (1969);
497: M. Suzuki, {\it Physica} {\bf 51}, 277 (1971).
498: \bibitem{mcl} J.A. McLennan, ``Introduction to non equilibrium statistical
499: mechanics", {\it Prentice-Hall Advanced Reference Series}, (1989).
500: \bibitem{am} Basically equivalent to the entropic component, N.W. Ashcroft and
501: N.D. Mermin, ``Solid State Physics", {\it Holt, Rinehart and Winston},
502: p. 253 (1976).
503: \bibitem{znp} X. Zotos, F. Naef and P. Prelov\v sek,
504: {\it Phys. Rev.} {\bf B55}, 11029 (1997).
505: \bibitem{mb} F.G. Mertens and H. B\"uttner, {\it Phys. Lett.} {\bf 84A},
506: 335 (1981).
507: \bibitem{kohn} W. Kohn, {\it Phys. Rev.} {\bf A171}, 133 (1964).
508: \bibitem{mcguire} J. B. McGuire, {\it J. Math. Ph.} {\bf 6}, 432 (1965);
509: C. N. Yang, {\it Phys. Rev. Lett.} {\bf 19}, 1312 (1967).
510: \bibitem{fk} S. Fujimoto and N. Kawakami, {\it J. Phys. A.}
511: {\bf 31}, 465 (1998).
512: \bibitem{f} S. Fujimoto, {\it J. Phys. Soc. Jpn.} {\bf 68}, 2810 (1999).
513: \bibitem{xz} X. Zotos, {\it Phys. Rev. Lett.} {\bf 82}, 1764 (1999).
514: \bibitem{sachdev} K. Damle and S. Sachdev,
515: {\it Phys. Rev.} {\bf B57}, 8307 (1998).
516: \bibitem{gm} T. Giamarchi and A. J. Millis,
517: {\it Phys. Rev.} {\bf B46}, 9325 (1992).
518: \bibitem{ra} A. Rosch and N. Andrei, {\it Phys. Rev. Lett.} {\bf 85},
519: 1092 (2000).
520: \end{thebibliography}
521: \end{document}
522: