1: \documentclass[11pt]{article}
2: %\documentclass[pra,twocolumn,showpacs,superscriptaddress,floatfix]{revtex4}
3:
4: \def\be{\begin{equation}}\def\ee{\end{equation}}
5: \renewcommand{\theequation}{\arabic{section}.\arabic{equation}}
6:
7: \let\a=\alpha \let\b=\beta \let\g=\gamma \let\d=\delta \let\e=\varepsilon
8: \let\z=\zeta \let\h=\eta \let\th=\theta \let\k=\kappa \let\l=\lambda
9: \let\m=\mu \let\n=\nu \let\x=\xi \let\p=\pi \let\r=\rho
10: \let\s=\sigma \let\t=\tau \let\f=\varphi \let\ph=\varphi\let\c=\chi
11: \let\ch=\chi \let\ps=\psi \let\y=\upsilon \let\o=\omega\let\si=\varsigma
12: \let\G=\Gamma \let\D=\Delta \let\Th=\Theta\let\L=\Lambda \let\X=\Xi
13: \let\P=\Pi \let\Si=\Sigma \let\F=\Phi \let\Ps=\Psi
14: \let\O=\Omega \let\Y=\Upsilon
15:
16: \font\tenmib=cmmib10%
17: \font\sevenmib=cmmib7%
18: \font\fivemib=cmmib5%
19: \textfont5=\tenmib\scriptfont5=\sevenmib\scriptscriptfont5=\fivemib
20: \mathchardef\Bx = "0518 %xi bold greek
21: \def\ie{{\it i.e.\ }}\def\etc{\it e.t.c.}
22:
23: \title{\bf Thermostats, chaos and Onsager reciprocity}
24: \author{\small\textsc{Giovanni Gallavotti}%
25: \\
26: \small Dipartimento di Fisica and INFN\\
27: \small Universit\`a di Roma
28: {\it La Sapienza}\\
29: \small P.~A.~Moro 2, 00185, Roma, Italy\\
30: \small \texttt{giovanni.gallavotti@roma1.infn.it}
31: }
32: \date{\today}
33: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
34: %%%%%%%%%%%%%%% TO INSERT FIGUREE ( for dvips ) %%%%%%%%%%%%%%
35: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
36: \newdimen\xshift \newdimen\xwidth \newdimen\yshift \newdimen\ywidth
37:
38: \def\ins#1#2#3{\vbox to0pt{\kern-#2pt\hbox{\kern#1pt #3}\vss}\nointerlineskip}
39:
40: \def\eqfig#1#2#3#4#5{
41: \par\xwidth=#1pt \xshift=\hsize \advance\xshift
42: by-\xwidth \divide\xshift by 2
43: \yshift=#2pt \divide\yshift by 2
44: %\line%%%in plain tex togliere il commento
45: {\hglue\xshift \vbox to #2pt{\vfil
46: #3 \special{psfile=#4.eps}
47: }\hfill\raise\yshift\hbox{#5}}}
48: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
49:
50: \begin{document}
51: \maketitle
52: \begin{abstract}{\it Finite thermostats are studied in the context of
53: nonequilibrium statistical mechanics. Entropy production rate has
54: been identified with the mechanical quantity expressed by the
55: phase space contraction rate and the currents have been linked to
56: its derivatives with respect to the parameters measuring the
57: forcing intensities. In some instances Green-Kubo formulae, hence
58: Onsager reciprocity, have been related to the fluctuation
59: theorem. However, mainly when dissipation takes place at the
60: boundary (as in gases or liquids in contact with thermostats),
61: phase space contraction may be independent on some of the forcing
62: parameters or, even in absence of forcing, phase space contraction
63: may not vanish: then the relation with the fluctuation theorem
64: does not seem to apply. On the other hand phase space contraction
65: can be altered by changing the metric on phase space: here this
66: ambiguity is discussed and employed to show that the relation
67: between the fluctuation theorem and Green-Kubo formulae can be
68: extended and is, by far, more general.}\end{abstract} \*
69: \vskip3mm
70:
71: I am honored to be given this occasion to thank J\"urgen Fr\"olich and
72: Tom Spencer for the time they spent discussing with me and
73: communicating their insights, projects and ideas. In particular
74: I have drawn inspiration from the one among the two to whom I have
75: been in closer contact, (JF), particularly at the time when he worked at
76: IHES: his strong criticism of any ``mathematical deviations'' from
77: physical problems has been always extremely effective on my work.
78:
79: \section{Thermostats}\label{sec1}\setcounter{equation}{0}
80:
81: A mechanical interpretation of the entropy production rate in
82: nonequilibrium systems interacting with thermostats and possibly
83: subject to external non conservative (``stirring'') forces has emerged
84: from simulations and studies on nonequilibrium statistical mechanics
85: since the early 1980's, \cite{No84,Ho85,EM90,Ga08}. It is interpreted
86: as phase space contraction rate, as defined by the divergence of the
87: equations of motion which we write symbolically $\dot x=f(x)$, \ie
88: $\s(x)=-\sum_i \partial_{x_i} f(x)$.
89:
90: General thermostats acting on a mechanical system, on which also
91: external non conservative forces may act, will be modeled as described
92: in Fig.1 and illustrated in the caption, \cite{Ga08,Ga07b}:
93:
94: \eqfig{302}{85}{\ins{90}{60}{${\bf X}_0,{\bf X}_1,\ldots,{\bf X}_n$}
95: \ins{60}{27}{$\ddot{{\bf X}}_{0i}=-\partial_i U_0({\bf X}_0)-\sum_{j}
96: \partial_i U_j({\bf X}_0,{\bf X}_j)+{\bf E}_i({\bf X}_0)$}
97: \ins{60}{10}{$\ddot{{\bf X}}_{ji}=-\partial_i U_j({\bf X}_j)-
98: \partial_i U_j({\bf X}_0,{\bf X}_j)-\a_j {\bf{\dot X}}_{ji}$}
99: }{fig1}{(fig1)\label{fig1}}
100: \*
101:
102: \noindent{}{Fig.1: \small\ The $1+n$ boxes $C_0,T_1,\ldots,T_n$ contain
103: $N_0,N_1,\ldots, N_n$ particles, of mass $m=1$, whose positions and
104: velocities are denoted ${\bf X}_0,{\bf X}_1,\ldots,{\bf X}_n$, and
105: $\dot{\bf X}_0,\dot{{\bf X}}_1,\ldots,\dot{{\bf X}}_n$
106: respectively. The ${\bf E}$ denote external, non conservative,
107: forces and the multipliers $\a_j$ model the thermostats and are so
108: defined that the kinetic energies $K_j=\frac12 \dot{{\bf X}}_j^2$
109: are exactly constants of motion with values $K_j=\frac32 N_j k_B
110: T_j$, $k_B=$ Boltzmann's constant, $j=1,\ldots,n$. The energies
111: $U_0,U_j,W_{0,j},\,j>0,$ should be imagined as generated by pair
112: potentials $\f_0,\f_j,\f_{0,j}$ short ranged, smooth, or with a
113: singularity like a hard core.\footnote{\small Singularities of
114: different type but care has to be excercised in formulating and by
115: external potentials modeling the containers walls and for simplicity
116: the assumption of smoothness (possibly in presence of a hard core)
117: is made here. For the more general cases, like Lennard-Jones
118: potentials, see \cite{BGGZ05}.}}
119:
120: \* To imply $\dot K_j=0$ in the above model the multiplier $\a_j$ has
121: to be $\a_j=-\frac{(Q_j+\dot U_j)}{3 N_j k_B T_j}$, where
122:
123: \be Q_j{\buildrel def\over=}-\dot{\bf X}_j\cdot\partial_{{{\bf X}}_j}
124: W_{0,i}({\bf X}_{0},{{\bf X}}_j)\label{e1.2}\ee
125: %
126: is naturally interpreted as the {\it heat} ceded per unit time to the
127: thermostat $C_j$. The phase space contraction rate, neglecting for
128: simplicity $O(N_j^{-1})$, is computed from the equation in Fig.1 to be
129:
130: \be\s(X)=\sum_j\frac{Q_j+\dot U_j}{k_B T_j}\label{e1.1}\ee
131: %
132: (each addend should be multiplied by the factor $(1-\frac2{3N_j})$ if
133: $O(N_j^{-1})$ is not neglected).
134:
135: Of course $\s(x)$ depends upon the metric used on phase space and on
136: the density giving the volume element: both are arbitrary and
137: Eq.(\ref{e1.1}) yields the contraction rate for the Euclidean metric
138: and density $1$: \ie for the Liouville volume. Because of such
139: ambiguity $\s(x)$ {\it cannot} have an immediate physical
140: meaning. However its time average, and the fluctuations of its finite
141: time averages over long time intervals, have an intrinsic meaning,
142: independent of the choices of the metric and the density,
143: \cite{Ga07b}, at least if the motions are ``chaotic'', see below.
144:
145: Some interesting concrete examples of the above systems are
146: illustrated in Fig.2 and Fig.3.
147:
148: \eqfig{290}{70}{
149: \ins{24}{63}{${{\bf E}}\ \to$}
150: \ins{70}{50}{periodic boundary (``{\it wire}'')}
151: \ins{70}{32}{$m\ddot{\bf x}={\bf E} -\a \dot{\bf x}$}
152: }{fig2}{(fig.2)}
153: \*
154:
155: \noindent{}{Fig2. \small A modern version of the classical Drude's model for
156: electric conductivity.}
157:
158: \*
159: \noindent{}In Fig.2 $\a=\frac{{\bf E}\cdot\dot{\bf x}}{m\dot{\bf x}^2}$ and this is
160: an electric conduction model of $N$ charged particles ($N=2$ in the
161: figure) in a constant electric field ${\bf E}$ and interacting with a
162: lattice of obstacles (circles in the figure); it is
163: ``autotermostatted'' (because $C_0$ and $T_1$ coincide) in $2$
164: dimensions. This is a model that appeared since the early days (Drude,
165: 1899, \cite{Be64}) in a slightly different form (\ie in dimension $3$
166: and with the thermostatting realized by replacing the $-\a\dot{\bf x}$
167: force with the prescription that after collision with an obstacle
168: velocity is rescaled to $|\dot{\bf x}|=\sqrt{\frac{3}m k_B T}$.
169:
170: The thermostat forces are a model of the effect of the interactions
171: between the particle (electron) and a background lattice (phonons).
172: This model is remarkable because it is the first nonequilibrium
173: problem that has been treated with real mathematical attention and for
174: which the analog of Ohm's law for electric conduction has been proved
175: if $N=1$, \cite{CELS93}.
176:
177: Another example is a model of thermal conduction, Fig.3:
178:
179: \eqfig{360}{60}{
180: \ins{90}{60}{$T_1$}
181: \ins{170}{61}{$C_0$}
182: \ins{250}{61}{$T_2$}
183: }{fig3}{(fig.3)\label{fig3}}
184: \*
185:
186: \noindent{}{Fig3. \small A model for thermal and electric conduction.}
187: \*
188:
189: \noindent{}in which $N_0$ hard disks interact by elastic collisions with each
190: other and with other hard disks ($N_1=N_2$ in number) in the
191: containers labeled by their temperatures $T_1,T_2$: the latter are
192: subject to elastic collisions between themselves and with the disks in
193: the central container $C_0$; the separation reflect elastically the
194: particles when their {\it centers} reach them, thus allowing interactions
195: between the thermostats and the main container particles.
196: Interactions with the thermostats take place only near the separating
197: walls.
198:
199: If one imagines that the upper and lower walls of the central
200: container are identified (realizing a periodic boundary condition) and
201: that a constant field of intensity $E$ acts in the vertical direction
202: then two forces conspire to keep it out of equilibrium, and the
203: parameters ${\bf F}=(T_2-T_1,E)$ characterize their strength: matter and
204: heat currents flow.
205:
206: The case $T_1=T_2, E\ne0$ has been studied in simulations to check that the
207: thermostats are ``efficient'': \ie that the simple interaction, via
208: collisions taking place across the boundary, is sufficient to allow the
209: systems to reach a stationary state, \cite{GG07}.
210:
211: Thermostat models similar to the above have been considered in the
212: literature, \cite{EM90,ECM93,GC95}. A fundamental problem with the
213: model in Fig.1 is that it is not clear which detailed assumptions have
214: to be made on the interactions to insure that almost all initial
215: conditions evolve staying in a bounded region in phase space so that
216: they can be expected to determine a stationary state. This can be
217: called the ``thermostat efficiency problem'' and it is, for
218: nonequilibrium, the analogue of the Hamiltonian stability problem in
219: equilibrium, \cite{Ga00}. The experiment in \cite{GG07} encourages
220: the idea that the assumptions could be very general and fairly simple.
221: In \cite{Ga96} a model like the one in Fig.3 was studied but the
222: confinement difficulty was avoided by requiring that also the total
223: kinetic energy $K_0$ in the central container was constant thanks to
224: an extra thermostatting force $-\a_0 \dot{{\bf X}}_0$ with a properly
225: chosen $\a_0$.
226:
227: The model in Fig.3 {\it without} thermostatting forces to
228: keep $K_j,\,j>0$ constant, hence with a purely Hamiltonian evolution,
229: has been carefully studied in \cite{Ja99} which also gives the
230: clearest account on the so called ``transient fluctuation theorem''
231: improving and extending its earlier formulation in \cite{BK81a}, and
232: obtains implicitly also a transient version of the result on
233: fluctuation patterns, analogous to the one derived earlier for steady
234: states in \cite{Ga99}.
235:
236: In \cite{Ja99} there is also a careful analysis of the model in Fig.3
237: with the aim of obtaining results for stationary states:
238: stationarity is made possible by taking the thermostats infinitely
239: large stressing the (formidable) problems that one should encounter in
240: attempting a rigorous proof.
241:
242: In this paper (and in all my preceding ones) I have chosen to consider
243: only finite thermostats with empirical thermostat forces and studied a
244: few problems by introducing a single assumption, the chaotic
245: hypothesis.
246:
247: \section{Chaos}\label{sec2}\setcounter{equation}{0}
248:
249: Microscopic motions are in all possible empirical senses
250: ``chaotic''. The paradigm of chaotic motions are the hyperbolic
251: transitive systems: these are smooth systems whose evolution can be
252: intuitively described by saying that each phase space point moves
253: being seen by the comoving neighboring points as a hyperbolic fixed point.
254:
255: Another intuitive way to look at such systems is to say that the phase
256: space points can be coded into sequences $\Bx=(\x_i)_{i=-\infty}^{\infty}$
257: of symbols, say the digits $0,1,2,$ $\ldots,q<\infty$, in such a way that
258: the dynamics becomes the trivial shift of the sequence $\Bx$, and all
259: sequences which satisfy $M_{\x_i,\x_{i+1}}\equiv1$ represent one phase
260: space point, $M$ being a ``compatibility matrix'' with elements
261: $M_{ij}=0,1$ which is transitive (\ie $M^s_{ij}>0$ for some
262: $s$). There may be ambiguities, \ie different sequences may represent
263: the same point, but this can happen on a zero volume set of points
264: only, in close analogy with the familiar ambiguity in the
265: representation of number by digits (where $0.9999..$ and $1.0000...$
266: are the same number).
267:
268: It is natural, at least for some \cite{Ru78b,GC95,Ru98}, to imagine
269: that motions of complex systems, like gases or liquids, are chaotic in
270: the simplest sense (which is also the strongest) of being hyperbolic
271: transitive on the attracting sets (also called Anosov systems). The
272: {\it chaotic hypothesis}, proposed in \cite{GC95}, see also
273: \cite{Ga00}, reflects this remark.
274:
275:
276: \*
277:
278: \noindent{}{\bf Chaotic hypothesis} {\it Attracting sets for mechanical systems
279: are smooth surfaces on which motion is smooth, hyperbolic and transitive.} \*
280:
281: \noindent{}This is an hypothesis that has to be considered in the same sense
282: as the ergodic hypothesis for equilibrium statistical mechanics,
283: \cite{Ru73}. Hence {\it it might be at first disturbing}.
284:
285: However disturbing assumptions are common in the literature and,
286: nevertheless, are often fruitful. I just mention the assumption of
287: periodicity with equal period (``monocyclicity'') of the motions of
288: mechanical systems: it was employed in the derivation of the second
289: law from the action principle in Boltzmann, \cite{Bo866}: this
290: assumption was considered also by Clausius, Maxwell, Helmholtz and was
291: the basis of the early works on the mechanical interpretation of the
292: second law, \cite{Cl871,He884a}. At the time there must have been
293: objections to such a bold assumption and someone must have declared,
294: as it was done a little later about its modification into the ergodic
295: hypothesis (and as it is done today about the chaotic hypothesis),
296: that it is ``a strong assumption as the periodicity (or ergodicity)
297: hypothesis raises the question of which systems of practical interest
298: are ``periodic'' (``ergodic''), since almost none of them is actually
299: such'', see \cite{MR007}. Similar statements can be found in the
300: literature, even in good papers.
301:
302: Chaotic systems (in the above sense) admit a statistics (called
303: SRB statistics, \cite{Si68a,Bo70a,BR75}), \ie a probability
304: distribution $\m$ on each attracting set which, by integration, gives
305: the average values of the observables $G(x)$ on trajectories whose
306: initial data $x$ are randomly chosen, near enough to an
307: attracting set, with a distribution with some (arbitrary) density:
308:
309: \be \langle{G}\rangle\,=
310: \lim_{T\to\infty}\frac1T\sum_{j=0}^{T-1} G(S^jx)=\int
311: G(y)\m(dy),\qquad \hbox{with probability $1$}\label{e2.1}\ee
312: %
313: where $x\to Sx$ is a discretized time evolution map, obtained by
314: timing observations on the occurrence of some selected event. Or in
315: the (unphysical, yet customary and interesting) case of observations
316: in continuous time
317:
318: \be \langle{G}\rangle\,=\lim_{T\to\infty}\frac1T \int_0^T G(S_t x)\,dt =\int
319: G(y)\m(dy),\qquad \hbox{with probability $1$}\label{e2.2}\ee
320: %
321: where $x\to S_tx$ denotes the evolution of the initial data $x$ via
322: the equations of motion, \cite{Ge98}.
323:
324: If motion is chaotic (\ie hyperbolic, regular, transitive) the finite
325: time averages
326:
327: \be \g=\langle{G}\rangle_\t= \frac1\t\sum_{j=0}^{\t-1} G(S^jx)\label{e2.3}\ee
328: %
329: satisfy a {\it large deviations law}, \ie fluctuations off the average
330: $\langle{G}\rangle$ as large as $\t$ itself are controlled by a function
331: $\z(\g)$ convex and analytic in a (finite) interval $(\g_1,\g_2)$,
332: maximal at $\langle{G}\rangle$. This means that the probability that $\g\in
333: [a,b]$ satisfies
334:
335: \be P_\t(\g\in [a,b])\simeq \, e^{\t\,\max_{[a,b]}\z(\g)},\qquad \forall
336: a,b\in (\g_1,\g_2)\label{e2.4}\ee
337: %
338: and the interval $(\g_1,\g_2)$ is non trivial if
339: $\langle{G^2}\rangle-\langle{G}\rangle^2>0$,
340: \cite{Si68a,Si77,GBG04}. If $\z(\g)$ is
341: quadratic at its maximum (\ie at $\langle{G}\rangle$) then this implies a
342: central limit theorem for the fluctuations of
343: $\sqrt{\t}\,\langle{G}\rangle_\t$, but Eq.(\ref{e3.4}) is a much stronger
344: property.
345:
346: \*
347: %
348: \noindent{}{\it Remarks:} (1) The hypothesis holds also in equilibrium; if
349: the system admits a dense trajectory in phase space it implies the
350: classical ergodic hypothesis.
351: \\
352: (2) If the observable $G$ has nonzero SRB-average
353: it is convenient to consider instead the observable
354: $\frac{G}{\langle{G}}\rangle$ because it is dimensionless, just as in the case
355: of $\langle{G}\rangle=0$ it is convenient to consider the dimensionless
356: observable $\frac{G}{\sqrt{\langle{G^2}\rangle}}$.
357: %
358: \\ (3) If the dynamics is {\it reversible}, \ie there is a smooth,
359: isometric, map $I$ of phase space such that $I^2=1$ and $IS_t=S_{-t}
360: I$ or in the discrete case $IS=S^{-1}I$, then any {\it time reversal odd}
361: observable $G$, with non zero average and nonzero dispersion
362: $\langle{G^2}\rangle-\langle{G}\rangle^2>0$, is such that the interval of
363: $(\g_1,\g_2)$ of large deviations for $\frac{G}{\langle{G}\rangle}$ is at
364: least $(-1,1)$ provided there is a dense orbit (which also implies
365: existence of only one attracting set).
366: %
367: \\ (4) The systems in the thermostats model of Sec.\ref{sec1} are all
368: reversible with $I$ being the ordinary time reversal, change in sign of
369: velocity with positions unaltered, and the phase space contraction
370: $\s(x)$ is odd under time reversal, see Eq.(\ref{e1.1}). Therefore if
371: $\s_+=\langle{\s}\rangle>0$ it follows that the observable
372:
373: \be p'=\frac1\t\sum_{j=0}^{\t-1}
374: \frac{\s(S^jx)}{\s_+}\label{e2.5}\ee
375: %
376: has domain of large deviations of the form $(-\overline g,\overline g)$
377: and contains $(-1,1)$.
378: %
379: \\
380: (5) Since by Eq.(\ref{e1.1}) $\s$ differs from
381: $\e(x)=\sum_{j>0}\frac{Q_j}{k_B T_j}$ by the time derivative of an
382: observable, it follows that the finite or infinite time averages $ \s$
383: and of $\e$ have, for large $\t$, the same distribution. Therefore the
384: same large deviations function $\z(p)$ controls the fluctuations of
385: $p'$ above and of
386:
387: \be p=\frac1\t\sum_{j=0}^{\t-1} \frac{\e(S^jx)}{\s_+},
388: \qquad\s_+\equiv\langle{\s}\rangle_{SRB}=\langle{\e}\rangle_{SRB},\label{e2.6}\ee
389: %
390: and it has been shown, \cite{GC95,GC95b} and in a mathematical form in
391: \cite{Ga95b}, that under the chaotic hypothesis and reversibility of
392: motions on the attracting set, the function $\z(p)$ has the
393: symmetry property
394:
395: \be \z(-p)=\z(p)-p\s_+, \qquad\hbox{\rm for all}\ p\in(-\overline p,\overline
396: p)\label{e2.7}\ee
397: %
398: and $\overline p\ge1$. This is the {\it fluctuation theorem} of
399: \cite{GC95} (it requires a proof and therefore it should not be
400: confused with several identities, see for instance \cite{CG99}, with
401: which, for reasons that I fail to understand, it has been often
402: identified). The interest of the theorem is that it is {\it
403: universal}, model independent yielding a parameter free relation which
404: deals with a quantity which has the physical meaning of entropy
405: production rate and therefore has an independent macroscopic
406: definition and is accessible to experiments.
407: %
408: \\ (6) Eq.(\ref{e2.7}) is closely related to the theorem in
409: \cite{Ja99}, from which it differs only because it deals with finite
410: thermostats assuming the (strong) chaotic hypothesis, rather than
411: dealing with infinite thermostats and assuming (strong) ergodicity
412: properties. In spite of the latter work several paper have appeared in
413: the literature trying to get rid of the chaotic hypothesis without
414: adding much (if anything) to the lucid discussion in \cite{Ja99} about
415: the necessity of suitable assumptions in order to allow extending a
416: transient fluctuation relation (which is an identity, requiring no
417: assumption, on the full phase space, \cite{CG99}) to a stationary one
418: (which deals with properties that hold on a subset of zero probability
419: with respect to the initial data sampling).
420: %
421: \\
422: (7) The fluctuation theorem has several extensions including a
423: remarkable, parameter free relation that concerns the relative
424: probability of patterns of evolution of an observable and their
425: reversed patterns, \cite{Ga97,Ga00,Ga02}, related to the
426: Onsager--Machlup fluctuations theory, which keeps being rediscovered
427: in various forms and variations in the literature.
428:
429: \section{Onsager reciprocity}\label{sec3}\setcounter{equation}{0}
430:
431: Another consequence of the fluctuation theorem are the Onsager
432: reciprocity and Green-Kubo formulae for the infinitesimal deviations
433: from equilibrium, \cite{Ga96}; the latter can be independently derived
434: (in a simpler way) from the chaotic hypothesis and time reversal
435: symmetry assumed only at equilibrium, \cite{CELS93}, as it will be
436: shown in the concluding comments, or as discussed from a somewhat
437: different viewpoint in \cite{GR97}.
438:
439: Here the aim is to show that {\it the Green-Kubo formulae, hence Onsager's
440: reciprocity, can be regarded as the version at zero forcing of the
441: fluctuation theorem for stationary states}.
442:
443: In the case in which $T_1=T_2=\ldots=T$ and ${\bf E}={\bf0}$ the
444: system is in thermal equilibrium and its state is characterized by a
445: probability distribution $\m_0$ which invariant under the time
446: evolution $x\to S_tx$ generated by the equations in Fig.1. Setting
447: $x=({\bf X}_0,\dot{{\bf X}}_0,{\bf X}_1,\dot{{\bf X}}_1,\ldots,{\bf
448: X}_n,\dot{{\bf X}}_n)$ it is remarkable that the distribution can be
449: explicitly found, \cite{EM90}, as
450:
451: $$\m_0(dx)=\, const\, e^{-\b \big(U_0({\bf X}_0)+\sum_{j>0}
452: \big(U_j({\bf X}_j)+W({\bf X}_0,{\bf X}_j)\big)+K_0(\dot{{\bf X}}_0 )\big)} $$
453: \be\cdot\textstyle
454: (\prod_{j>0}\d(K_j-\frac32 N_j T))(\prod_{j\ge0} d\dot{{\bf X}}_j\, d{\bf
455: X}_j)\label{e3.1}\ee
456: %
457: where $\b=\frac1{k_B T}$ (neglecting $O(N_j^{-1})$ for
458: simplicity). Calling the ``unperturbed'' energy $H_0(x)=K_0(\dot{\bf
459: X}_0 )+\sum_{j\ge0} U_j({\bf X}_j) +\sum_{j>0} W_j({\bf X}_0,{\bf
460: X}_j)$\footnote{\small The kinetic energy of the thermostats is an
461: additive constant and therefore is not explicitly written.} and
462: $\widetilde\d({\bf K}_j(x), T_j)= \d(K_j(\dot{{\bf X}}_j)-\frac32 N_j T_j))$,
463: Eq.(\ref{e3.1}), written more compactly, is
464:
465: \be\m_0(dx)=
466: const\,e^{-\b H_0(x)} \,\prod_{j>0}\widetilde\d(K_j(x),T)\,dx\label{e3.2}\ee
467: %
468: which is a distribution in an ensemble which, for the system in $C_0$,
469: is equivalent to the canonical one (for $N_0,L_0\to\infty,\,
470: N_0/L_0^3=const$ if $L_0$ is the side of the container).
471:
472: We now want to compare the average values of various currents that are
473: swit\-ched on when ${\bf E}$, the external forces, become non zero and the
474: temperatures of the thermostats become different: ${\bf E}\ne{\bf0}$ and
475: $T_j=T+\e_j$. More precisely we look for the relations between
476: infinitesimal forcing {\it actions} and the corresponding {\it
477: currents}, \ie the susceptibility coefficients.
478:
479: The currents are related to the average values of the derivatives of
480: the entropy production with respect to the forces (material currents)
481: or to the temperature inequalities (heat currents). {\it However} the
482: arbitrariness inherent in the phase space contraction generates
483: interesting questions: for instance in the model in Fig.1 the
484: phase space contraction with respect to the Liouville volume {\it is
485: independent of the external forces ${\bf E}$}, see Eqs.(\ref{e1.2}),(\ref{e1.1}),
486: so that $\partial_E\s\equiv0$, while it is {\it obvious} that the external
487: forces generate material currents, being non conservative.
488:
489: On the other hand even in equilibrium a thermostatted system exchanges
490: heat with the thermostats: hence there is a production of entropy
491: which has a zero average but which is not zero and equal to
492: $\sum_j\frac{Q_j}{k_B T}$.
493:
494: It is therefore interesting to see, first, why in equilibrium (\ie
495: when the thermostats have all the same temperature and no external
496: forces act) the SRB-average of $\sum_j\frac{ Q_j}{k_B T}$ vanishes,
497: \cite{Ru96}. This is the case because the latter quantity is the
498: derivative of $\b H_0(x)$. In fact the derivative $\b H_0$ is $\b$
499: times the work done on the system by the forces $-\a \dot{{\bf X}}_j$
500: which equals $\sum_{j>0} \frac{Q_j+\dot U_j}{k_B T}$. This means that
501: $\s(x)-\b \dot H_0(x)\equiv 0$ and therefore
502:
503: \be \sum_{j>0} \frac{Q_j}{k_B T}=\b \dot H_0(x)-\sum_{j>0}
504: \frac{\dot U_j(x)}{k_B T}\label{e3.3}\ee
505: %
506: and the r.h.s. is a time derivative, hence it has $0$ time average.
507:
508: When the system is out of equilibrium (\ie $T_j\not\equiv T$ and ${\bf
509: E}\ne {\bf0}$) the heat currents flowing into the thermostats divided by
510: the temperature are generated by the entropy production rate
511: $j_k(x)=\partial_{T_k} \s(x)$, while the material currents through the
512: system are defined by minus the derivatives with respect to the acting
513: forces of the work per unit time that they do, given by the
514: corresponding derivatives of $\dot H_0$. Thus given arbitrarily $\b$
515: the quantity
516:
517: \be \overline \s(x)=\s(x)-\b \dot H_0(x)\label{e3.4}\ee
518: %
519: generates all currents up to a proportionality factor (here $\b$ is
520: arbitrary). It can be computed as
521:
522: \be \overline\s(x)=\sum_{j>0} \frac{Q_j+\dot U_j}{k_B T_j}-\b\, {\bf
523: E}\cdot \dot {{\bf X}}_0-\b\,\sum_{j>0}(Q_j+\dot U_j)\label{e3.5}\ee
524: %
525: because, by the equations in Fig.1, $\dot H_0={\bf E}\cdot
526: \dot{{\bf X}}_0-\sum_{j>0}\a_j \dot{{\bf X}}_j^2$ and therefore
527: $\dot H_0={\bf E}\cdot
528: \dot{{\bf X}}_0+\sum_{j>0}(Q_j+\dot U_j)$, see Eq.(\ref{e1.2}).
529:
530: Hence, discarding the time derivatives terms involving the $\dot U_j$
531: (parameters independent), the currents (at infinitesimal forcing) can
532: be generated by the function
533:
534: \be\s_0(x)=\sum_{j>0} Q_j(\frac1{k_B T_j}-\frac1{k_B
535: T})-{\bf E}\cdot\dot{{\bf X}}_0\frac1{k_B T}\label{e3.6}\ee
536:
537: The generating function $\s_0$ is odd under time reversal and {\it
538: vanishes at equilibrium} $T_j=T_i,{\bf E}={\bf0}$ if $T$ is chosen $T=T_j$;
539: its derivatives with respect to the forcing parameters $T_j,E_k$
540: generate the heat and material currents and, at the same, time
541: $\s_0(x)$ differs from the phase space contraction by a time
542: derivative.
543:
544: Note that $\overline \s$ is also the phase space contraction of the volume in
545: phase space, {\it provided} the latter is measured by the distribution
546:
547: \be\overline\m(dx)=
548: const\,e^{-\b H_0(x)} \,\prod_{j>0}\widetilde\d(K_j(x),T_j)\,dx\label{e3.7}\ee
549: %
550:
551: In \cite{Ga96a} a reversible system (like the model in Fig.1), has
552: been considered in which the generating function for the currents {\it
553: $\s_0$ vanishes for vanishing ``thermodynamic forces'' ${\bf F}=
554: (T_1-T,\ldots,T_n-T, E_1,\ldots ,E_q)={\bf0}$} and satisfies the
555: fluctuation relation or, better, its extension in
556: \cite[Eq.(14)]{Ga96a}, has been considered.
557:
558: And it has been shown, \cite{Ga96a}, that the products of the currents,
559: generated by the thermodynamic forces, times $\b=\frac1{k_B T}$ , and
560: defined by
561:
562: \be j_m=\partial_{F_r}\overline\s(x)\equiv \partial_{F_r}\s_0(x)\label{e3.8}\ee
563: %
564: are such that their averages $J_m=\langle{j_m}\rangle_{SRB}$ have
565: susceptibilities $L_{mp}=\partial_{F_m} J_p\Big|_{{\bf F}={\bf0}}$ which
566: satisfy
567:
568: \be L_{mp}=
569: \frac12\int_{-\infty}^\infty
570: dt\, \big(\langle j_m(S_t\cdot) j_p(\cdot)\rangle_{SRB}-
571: \langle j_m\rangle_{SRB}\langle j_p\rangle_{SRB}
572: \big)\big|_{{\bf F}={\bf0}}\label{e3.9}\ee
573: %
574:
575: If the parameter $\b$ is properly chosen as mentioned above, \ie
576: $\b=\frac1{k_BT}$ (and only if so chosen), $\s_0$ will vanish when ${\bf
577: F}={\bf0}$. Since $\overline\s$ and $\s_0$ differ by a time derivative they
578: can be interchangeably used in the theory of the SRB distribution and
579: therefore $\s_0$ satisfies the fluctuation theorem (because $\overline\s$
580: does); the assumptions in the derivation in \cite{Ga96a} apply and
581: therefore Eq.(\ref{e3.9}) yields Onsager's reciprocity
582: $L_{mp}=L_{pm}$, and Green--Kubo formula.
583:
584: \section{Work and entropy theorems. Comments}\label{sec4}
585: \setcounter{equation}{0}
586:
587: (1) This extends considerably the results in \cite{Ga96a,Ga96}
588: removing the restriction on the phase space contraction to be the
589: generating function of the currents. The key is that the phase space
590: contraction is only defined up to a time derivative of an observable
591: and the generating function of the currents coincides with the phase
592: space contraction only if the observable is properly chosen. \*
593:
594: \noindent{}(2) it is worth stressing that the extension of the fluctuation
595: theorem needed to derive from it Onsager reciprocity is an
596: important one: in \cite{Ga97} it was further extended to show
597: ({\it conditional reversibility theorem}) that there is a simple
598: relation between the probability that an observable $F(x)$, even
599: or odd under time reversal (for simplicity), follows in a time
600: interval ${-\t,\t}$ a ``pattern'' $F(S_t x)=\f(t)$ or the
601: ``reversed pattern'' $F(S_t x)=\f(-t)$ {\it provided} the entropy
602: production rate is fixed, \cite{Ga97}. A statement that can be
603: colorfully quoted as {\it .. relative probabilities of patterns
604: observed in a time interval of size $\t$ and in presence of an
605: average entropy production $p$ are the same as those of the
606: corresponding anti-patterns in presence of the opposite average
607: entropy production rate}, \cite[p.476]{Ga02}, or also
608: \cite[p.476]{Ga02}, or {\it ... it ``suffices'' to change the sign
609: of the entropy production to reverse the arrow of time}, or also
610: {\it ... a waterfall will go up, as likely as we see it going
611: down, in a world in which for some reason, or by the deed of a
612: Daemon, the entropy creation rate has changed sign during a long
613: enough time}, \cite[p.288]{Ga00}. We can also say that the motion
614: on an attractor is reversible, even in the presence of
615: dissipation, once the dissipation is fixed. Again variations of
616: this property keep being rediscovered, see for instance \cite{GPB08}. \*
617:
618: \noindent{}(3) In the case of systems in contact with a single
619: thermostat but in a stationary nonequilibrium because of the
620: action of external forces the above analysis has also interesting
621: consequences. The phase space contraction can be written as
622: $\s(x)=\sum_{j>0}\frac{ Q_j+\dot U_j}{k_B T}$, as in
623: Eq.(\ref{e1.1}), or by adding to it a time derivative as
624: $\overline \s(x)=\s(x)+\b \dot H_0(x)$ which in this case is
625: simply $\overline\s(x)=\frac{{\bf E}({\bf X}_0)\cdot\dot{{\bf
626: X}}_0}{k_B T}=\frac{\dot W}{k_B T}$. Therefore the fluctuation
627: theorem, as pointed out by Bonetto: see \cite[Eq.(9.10.4)]{Ga00},
628: yields the following ``{\it work theorem}''
629:
630: \be \langle{ e^{-\b\, w\, \t}}\rangle_{SRB}=1,
631: \qquad w{\buildrel def\over=}
632: \frac1\t\int_0^\t \dot W(S_tx)\, dt\label{e4.1}\ee
633: %
634: in the sense that the logarithm of the l.h.s. divided by $\t$ tends to
635: $0$ as $\t\to\infty$. More generally the identity up to a time
636: derivative of $\s$, $\overline\s$, $\sum_{j>0}\frac{Q_j}{k_B T_j}$ and
637: $\s_0 =\sum_{j>0}(\frac{Q_j}{k_B T_j}-\b Q_j)-\b {\bf E}\cdot\dot{\bf
638: X}_0$, see Eqs.(\ref{e3.3})-(\ref{e3.6}), implies that, in the same
639: sense as in Eq.(\ref{e4.1}), the finite time average $P$ of {\it any}
640: of the latter four quantities, denoted $\widetilde\s$, over a time $\t$ will
641: satisfy
642:
643: \be \langle{ e^{-P\t}}\rangle_{SRB}=1,\qquad P{\buildrel def\over=}
644: \frac1\t\int_0^\t \widetilde\s(S_tx)\label{e4.2}\ee
645: %
646: which can be called an ``{\it entropy theorem}'': not only remarkable
647: because it involves quantities that can be measured in experiments,
648: \cite{Ga08}, but also because here $\b$ can be taken {\it arbitrary},
649: so that Eq.(\ref{e4.2}) is an infinite number of relations. Actually
650: if $p=P/\langle{\s_0}\rangle_{SRB}$ the large deviations of $p$ satisfy the
651: fluctuation theorem symmetry Eq.(\ref{e2.7}). Note however that all
652: such relations are special cases of the theorem in
653: \cite{Ga99}.
654:
655: \noindent{}(4) A further alternative method to derive the Green-Kubo relations
656: is in \cite{CELS93a}. It will be illustrated, for completeness, in
657: the simple case of a system interacting with only one thermostat
658: and subject to several nonconservative external forces that will
659: be proportional to parameters ${\bf E}=(E_1,\ldots,E_q)$. Under
660: the chaotic hypothesis the SRB average of the currents $J_m=\int
661: \m_{SRB}(dx) j_m(x)$, with $j_m(x){\buildrel
662: def\over=}\partial_{E_m}\overline\s(x)$ in presence of
663: thermodynamic forces ${\bf E}$, can be computed as the limit
664: $J_m=\lim_{t\to\infty} \m_0(j_m(S_t^{{\bf E}}x))$, if $S_t^{\bf E}$
665: is the map such that $x\to S^{{\bf E}}_tx$ solves the equations of
666: motion in presence of forcing forces with parameters ${\bf E}$,
667: and $\m_0$ is the equilibrium distribution Eq.(\ref{e3.1}),
668: \cite{CELS93a,CELS93}. Therefore
669:
670: $$J_m=\lim_{t\to\infty}\m_0(j_m(S_t^{{\bf E}}x))=
671: \int_0^{+\infty}dt\, \frac{d}{dt} \int \m_0(dx) J_m(S_t^{{\bf E}}x)$$
672: \be=\int_0^{+\infty}dt\, \frac{d}{dt}\int \frac{\m_0(dx)}{\m_0(dS^{{\bf E}}_tx)}
673: \m_0(dS^{{\bf E}}_tx) j_m(S^{{\bf E}}_tx)\label{e4.3} \ee
674: $$=\int_0^{+\infty}dt\, \frac{d}{dt}\int
675: \frac{\m_0(dS^{{\bf E}}_{-t}x)}{\m_0(dx)}\m_0(dx) j_m(x)$$
676: %
677: but by the comment preceding Eq.(\ref{e3.7}) (considered with
678: $T_j\equiv T$)
679:
680: \be \frac{d}{dt} \frac{\m_0(dS^{{\bf E}}_{-t}x)}{\m_0(dx)}=\overline\s(S^{\bf
681: E}_{-t}x)\label{e4.4}\ee
682: %
683: so that the chain of equalities in Eq.(\ref{e4.3}) yields
684:
685: \be J_m=\int_0^\infty dt\, \int \overline\s(S^{{\bf E}}_{-t}x)
686: j_m(x)\m_0(dx)\label{e4.5}\ee
687: %
688: And taking into account that $\overline\s(x)\equiv0$, if ${\bf E}={\bf0}$, and
689: $j_m(x)=\partial_{E_m}\overline\s(x)$
690:
691: $$L_{pm}=\partial_{E_p} J_m|_{{\bf E}={\bf0}}= \int_0^\infty
692: dt\,\big( \int \partial_{E_p}\overline\s( S^{\bf
693: E}_{-t}x)\,\partial_{E_m}\overline\s(x)\, \m_0(dx)\big)
694: \big|_{{\bf E}={\bf0}}$$
695: \be=\frac12 \int_{-\infty}^\infty dt \big(\int
696: \partial_{E_p}\overline\s( S^{\bf
697: E}_{-t}x)\,\partial_{E_m}\overline\s(x)\,\m_0(dx)\big)
698: \big|_{\bf E}= {\bf 0}=L_{mp}\label{e4.6}\ee
699: %
700: by time reversal invariance of the equilibrium distribution $\m_0$,
701: which is the Green-Kubo formula.
702: %\bibliography{0Bib}
703: \bibliographystyle{unsrt}
704: \begin{thebibliography}{10}
705:
706: \bibitem{No84}
707: S.~Nos\'e.
708: \newblock A unified formulation of the constant temperature molecular dynamics
709: methods.
710: \newblock {\em Journal of Chemical Physics}, 81:511--519, 1984.
711:
712: \bibitem{Ho85}
713: W.~Hoover.
714: \newblock Canonical equilibrium phase-space distributions.
715: \newblock {\em Physical Review A}, 31:1695--1697, 1985.
716:
717: \bibitem{EM90}
718: D.~J. Evans and G.~P. Morriss.
719: \newblock {\em Statistical Mechanics of Non{\-}equilibrium Fluids}.
720: \newblock Academic Press, New-York, 1990.
721:
722: \bibitem{Ga08}
723: G.~Gallavotti.
724: \newblock Heat and fluctuations from order to chaos.
725: \newblock {\em European Physics Journal B, EPJB}, 61:1--24, 2008.
726:
727: \bibitem{Ga07b}
728: G.~Gallavotti.
729: \newblock Fluctuation relation, fluctuation theorem, thermostats and entropy
730: creation in non equilibrium statistical physics.
731: \newblock {\em C.R. Physique}, 8:486--494, 2007,
732: (doi:10.1016/j.crhy.2007.04.011).
733:
734: \bibitem{BGGZ05}
735: F.~Bonetto, G.~Gallavotti, A.~Giuliani, and F.~Zamponi.
736: \newblock Chaotic {H}ypothesis, {F}luctuation {T}heorem and {S}ingularities.
737: \newblock {\em Journal of Statistical Physics}, 123:39--54, 2006.
738:
739: \bibitem{Be64}
740: R.~Becker.
741: \newblock {\em Electromagnetic fields and interactions}.
742: \newblock Blaisdell, New-York, 1964.
743:
744: \bibitem{CELS93}
745: N.~I. Chernov, G.~L. Eyink, J.~L. Lebowitz, and {Ya.}~G. Sinai.
746: \newblock Steady state electric conductivity in the periodic {L}orentz gas.
747: \newblock {\em Communications in Mathematical Physics}, 154:569--601, 1993.
748:
749: \bibitem{GG07}
750: P.~Garrido and G.~Gallavotti.
751: \newblock Boundary dissipation in a driven hard disk system.
752: \newblock {\em Journal of Statistical Physics}, 126:1201--1207, 2007.
753:
754: \bibitem{ECM93}
755: D.~J. Evans, E.~G.~D. Cohen, and G.~P. Morriss.
756: \newblock Probability of second law violations in shearing steady flows.
757: \newblock {\em Physical Review Letters}, 71:2401--2404, 1993.
758:
759: \bibitem{GC95}
760: G.~Gallavotti and E.~G.~D. Cohen.
761: \newblock Dynamical ensembles in nonequilibrium statistical mechanics.
762: \newblock {\em Physical Review Letters}, 74:2694--2697, 1995.
763:
764: \bibitem{Ga00}
765: G.~Gallavotti.
766: \newblock {\em Statistical Mechanics. A short treatise}.
767: \newblock Springer Verlag, Berlin, 2000.
768:
769: \bibitem{Ga96}
770: G.~Gallavotti.
771: \newblock Chaotic hypothesis: {O}nsager reciprocity and
772: fluctuation--dissi\-pation theorem.
773: \newblock {\em Journal of Statistical Physics}, 84:899--926, 1996.
774:
775: \bibitem{Ja99}
776: C.~Jarzynski.
777: \newblock Hamiltonian derivation of a detailed fluctuation theorem.
778: \newblock {\em Journal of Statistical Physics}, 98:77--102, 1999.
779:
780: \bibitem{BK81a}
781: G.~N. Bochkov and Yu.~E. Kuzovlev.
782: \newblock Nonlinear fluctuation-dissipation relations and stochastic models in
783: nonequilibrium thermodynamics: I. generalized fluctuation-dissipation
784: theorem.
785: \newblock {\em Physica A}, 106:443--479, 1981.
786:
787: \bibitem{Ga99}
788: G.~Gallavotti.
789: \newblock New methods in nonequilibrium gases and fluids.
790: \newblock {\em Open Systems and Information Dynamics}, 6:101--136, 1999
791: (preprint chao-dyn/9610018).
792:
793: \bibitem{Ru78b}
794: D.~Ruelle.
795: \newblock What are the measures describing turbulence.
796: \newblock {\em Progress in Theoretical Physics Supplement}, 64:339--345, 1978.
797:
798: \bibitem{Ru98}
799: D.~Ruelle.
800: \newblock Natural nonequilibrium states in quantum statistical mechanics.
801: \newblock {\em Journal of Statistical Physics}, 98:57--75, 1998.
802:
803: \bibitem{Ru73}
804: D.~Ruelle.
805: \newblock {\em Ergodic theory}, volume Suppl X of {\em The Boltzmann equation,
806: ed. E.G.D Cohen, W. Thirring, Acta Physica Austriaca}.
807: \newblock Springer, New York, 1973.
808:
809: \bibitem{Bo866}
810: L.~Boltzmann.
811: \newblock {\em {\"U}ber die mechanische {B}edeutung des zweiten {H}auptsatzes
812: der {W\"a}rme\-theorie}, volume 1, p.9 of {\em {W}is\-sen\-schaft\-li\-che
813: {A}bhandlungen, ed. {F}. {H}asen{\"o}hrl}.
814: \newblock Chelsea, New York, 1968.
815:
816: \bibitem{Cl871}
817: R.~Clausius.
818: \newblock Ueber die zur{\"u}ckf{\"u}hrung des zweites hauptsatzes der
819: mechanischen w{\"a}rmetheorie aud allgemeine mechanische prinzipien.
820: \newblock {\em Annalen der Physik}, 142:433--461, 1871.
821:
822: \bibitem{He884a}
823: H.~Helmholtz.
824: \newblock {\em Prinzipien der Statistik monocyklischer Systeme}, volume III of
825: {\em {W}is\-sen\-schaft\-li\-che {A}bhandlungen}.
826: \newblock Barth, Leipzig, 1895.
827:
828: \bibitem{MR007}
829: C.~Mejia-Monasterio and L.Rondoni.
830: \newblock On the fluctuation relation for {N}ose-{H}oover locally thermostated
831: systems.
832: \newblock {\em arXiv: cond-mat}, 0710.3673:1--27, 2007.
833:
834: \bibitem{Si68a}
835: {Ya.}~G. Sinai.
836: \newblock Markov partitions and {$C$}-diffeomorphisms.
837: \newblock {\em Functional Analysis and Applications}, 2(1):64--89, 1968.
838:
839: \bibitem{Bo70a}
840: R.~Bowen.
841: \newblock Markov partitions for axiom {A} diffeomorphisms.
842: \newblock {\em American Journal of Mathematics}, 92:725--747, 1970.
843:
844: \bibitem{BR75}
845: R.~Bowen and D.~Ruelle.
846: \newblock The ergodic theory of axiom {A} flows.
847: \newblock {\em Inventiones Mathematicae}, 29:181--205, 1975.
848:
849: \bibitem{Ge98}
850: G.~Gentile.
851: \newblock A large deviation theorem for {A}nosov flows.
852: \newblock {\em Forum Mathematicum}, 10:89--118, 1998.
853:
854: \bibitem{Si77}
855: {Ya.}~G. Sinai.
856: \newblock {\em Lectures in ergodic theory}.
857: \newblock Lecture notes in Mathematics. Princeton University Press, Princeton,
858: 1977.
859:
860: \bibitem{GBG04}
861: G.~Gallavotti, F.~Bonetto, and G.~Gentile.
862: \newblock {\em Aspects of the ergodic, qualitative and statistical theory of
863: motion}.
864: \newblock Springer Verlag, Berlin, 2004.
865:
866: \bibitem{GC95b}
867: G.~Gallavotti and E.G.D. Cohen.
868: \newblock Dynamical ensembles in stationary states.
869: \newblock {\em Journal of Statistical Physics}, 80:931--970, 1995.
870:
871: \bibitem{Ga95b}
872: G.~Gallavotti.
873: \newblock Reversible {A}nosov diffeomorphisms and large deviations.
874: \newblock {\em Mathematical Physics Electronic Journal (MPEJ)}, 1:1--12, 1995.
875:
876: \bibitem{CG99}
877: E.~G.~D. Cohen and G.~Gallavotti.
878: \newblock Note on two theorems in nonequilibrium statistical mechanics.
879: \newblock {\em Journal of Statistical Physics}, 96:1343--1349, 1999.
880:
881: \bibitem{Ga97}
882: G.~Gallavotti.
883: \newblock Fluctuation patterns and conditional reversibility in nonequilibrium
884: systems.
885: \newblock {\em Annales de l' Institut H. Poincar\'e}, 70:429--443, 1999 and
886: chao-dyn/9703007.
887:
888: \bibitem{Ga02}
889: G.~Gallavotti.
890: \newblock {\em Foundations of Fluid Dynamics}.
891: \newblock (second printing) Sprin\-ger Verlag, Berlin, 2005.
892:
893: \bibitem{GR97}
894: G.~Gallavotti and D.~Ruelle.
895: \newblock {SRB} states and non\-equi\-li\-brium statistical mechanics close to
896: equi\-li\-brium.
897: \newblock {\em Com\-mu\-ni\-ca\-tions in Mathematical Physics}, 190:279--285,
898: 1997.
899:
900: \bibitem{Ru96}
901: D.~Ruelle.
902: \newblock Positivity of entropy production in nonequilibrium statistical
903: mechanics.
904: \newblock {\em Journal of Statistical Physics}, 85:1--25, 1996.
905:
906: \bibitem{Ga96a}
907: G.~Gallavotti.
908: \newblock Extension of {O}nsager's reciprocity to large fields and the chaotic
909: hypothesis.
910: \newblock {\em Physical Review Letters}, 77:4334--4337, 1996.
911:
912: \bibitem{GPB08}
913: A.~Gomez-Marin, J.M.R. Parondo, and C.~Van den Broeck.
914: \newblock The footprints of irreversibility.
915: \newblock {\em European Physics Letters}, 82:5002+4, 2008.
916:
917: \bibitem{CELS93a}
918: N.~I. Chernov, G.~L. Eyink, J.~L. Lebowitz, and {Ya.}~G. Sinai.
919: \newblock Derivation of {O}hm's law in a deterministic mechanical model.
920: \newblock {\em Physical Review Letters}, 70:2209--2212, 1993.
921:
922: \end{thebibliography}
923:
924: \end{document}
925: