1: \documentclass[prl,aps,twocolumn,showpacs]{revtex4}
2:
3: \usepackage{graphicx}
4:
5: \begin{document}
6:
7: \title{Classical and Quantum Fluctuation Theorems for Heat Exchange}
8: \author{Christopher Jarzynski}
9: \affiliation{Theoretical Division, T-13, MS B213,
10: Los Alamos National Laboratory,
11: Los Alamos, NM 87545}
12: \email{chrisj@lanl.gov}
13: \author{Daniel K. W\'ojcik}
14: \affiliation{Center for Nonlinear Science, School of Physics, Georgia
15: Institute of Technology, 837 State Street, Atlanta, GA 30332-0430,
16: USA}
17: \affiliation{Department of Neurophysiology, Nencki Institute of
18: Experimental Biology, 3 Pasteur Str., 02-093 Warsaw, Poland}
19: \email{danek@cns.physics.gatech.edu}
20: \date{\today}
21:
22: \begin{abstract}
23: The statistics of heat exchange between two classical or quantum
24: finite systems initially prepared at different temperatures are
25: shown to obey a fluctuation theorem.
26: \end{abstract}
27:
28: \pacs{05.70.Ln, 05.20.-y \\
29: Keywords: {\bf fluctuation theorem},
30: {\bf irreversible processes}
31: %LAUR-99-5580
32: }
33: \maketitle
34:
35:
36: The {\it fluctuation theorem} (FT) refers to a collection of
37: theoretical predictions~\cite{EvansCM93, EvansS94, GallavottiC95,
38: Kurchan98, LebowitzS99, Maes99, Evans02}, recently confirmed
39: experimentally~\cite{WangSMSE02}, pertaining to a system evolving
40: under non-equilibrium conditions. These results are roughly
41: summarized by the equation
42: \begin{equation}
43: \label{eq:ft}
44: \ln \frac{p(+\Sigma)}{p(-\Sigma)} = \Sigma,
45: \end{equation}
46: where $p(\Sigma)$ denotes the probability that an amount of entropy
47: $\Sigma$ is generated during a specified time interval. Both {\em
48: transient\/} and {\em steady state\/} versions of the FT have been
49: obtained. The definition of ``entropy generated'' ($\Sigma$) depends
50: on the dynamics used to model the evolution of the system under
51: consideration. However, for a variety of physical situations, and a
52: variety of equations of motion (both deterministic and stochastic)
53: used to model them, the FT has been established under reasonable
54: definitions of entropy generation. Moreover, the FT is
55: related~\cite{crooks99} to a set of {\em free energy relations\/} (see
56: e.g.~\cite{Jarzynski97,Crooks98}) which connect equilibrium free
57: energy differences to non-equilibrium work values, and which have
58: recently been confirmed experimentally~\cite{Liphardt02}.
59:
60: The situations modeled in Refs.~\cite{EvansCM93, EvansS94,
61: GallavottiC95, Kurchan98, LebowitzS99, Maes99, crooks99, Evans02,
62: WangSMSE02, Jarzynski97, Crooks98} all involve an externally driven
63: system, in the presence of a heat reservoir. The purpose of this
64: paper is to point out that a similar result can be derived in a
65: different setting. Namely, we will obtain a symmetry relation
66: constraining the statistics of heat exchange between two bodies
67: initially prepared at different temperatures. We will present both
68: classical and quantum derivations, and will use the term {\it exchange
69: fluctuation theorem} (XFT) to refer to these results.
70:
71: In what follows, the XFT (Eq.~(\ref{eq:heatft})) will be stated and
72: derived. A corollary result related to the Second Law of
73: Thermodynamics will then be presented (Eq.~(\ref{eq:bound})).
74:
75: Consider two finite bodies, $A$ and $B$, separately prepared in
76: equilibrium states at temperatures $T_A$ and $T_B$, respectively, then
77: placed in thermal contact with one another for a time $\tau$, and then
78: separated again. Let $Q$ denote the net heat transfer from $A$ to $B$
79: during the interval of contact, i.e. the amount of energy lost by $A$
80: and gained by $B$. Now imagine repeating this experiment many times,
81: always initializing the two bodies at the specified temperatures, and
82: let $p_\tau(Q)$ denote the observed distribution of values of $Q$ over
83: the ensemble of repetitions. Then we claim that this distribution
84: satisfies
85: \begin{equation}
86: \label{eq:heatft}
87: \ln
88: {p_\tau(+Q)\over p_\tau(-Q)}
89: = \Delta\beta \cdot Q,
90: \end{equation}
91: where $\Delta\beta=T_B^{-1}-T_A^{-1}$ is the difference between the
92: inverse temperatures at which the bodies are prepared.
93:
94: In the quantum case we must define $Q$ through an experimental
95: procedure: starting with the two systems initially prepared at
96: different temperatures, we first measure the energy of each system,
97: then we allow them to weakly interact over a time $\tau$, and finally
98: we again measure the energy of each system. We then interpret heat
99: transfer in terms of the changes in these measured energies
100: (Eq.~(\ref{eq:12})). This approach is similar in spirit to that taken
101: by~\cite{kurchan00s, Mukamel03, monnai03}, who considered related
102: problems. For an alternative approach see e.g.~\cite{deroeck03}.
103:
104:
105: Eq.~(\ref{eq:heatft}) clearly resembles the usual FT,
106: Eq.~(\ref{eq:ft}). Indeed, if we invoke macroscopic thermodynamics to
107: argue that $-Q/T_A$ is the entropy change of $A$, and $+Q/T_B$ is that
108: of $B$, then the net entropy generated by the exchange of heat is
109: given by $\Sigma = \Delta\beta \cdot Q$, and Eq.~(\ref{eq:heatft})
110: becomes Eq.~(\ref{eq:ft}). However, this argument works only if the
111: heat transferred is very small in comparison with the internal energy
112: of either body, whereas the validity of Eq.~(\ref{eq:heatft}) does
113: {\it not} require this assumption. Therefore, we will leave
114: Eq.~(\ref{eq:heatft}) as a statistical statement about heat exchange,
115: rather than trying to force it to be a statement about entropy
116: generation {\it per se}.
117:
118: To derive Eq.~(\ref{eq:heatft}) from classical equations of motion,
119: let ${\bf z}_A$ denote the phase space coordinates specifying the
120: microstate of body $A$ (e.g.\ the positions and momenta of all its
121: degrees of freedom); and let $H^A({\bf z}_A)$ be a Hamiltonian whose
122: value defines the internal energy of $A$, as a function of its
123: microstate. Similarly for $H^B({\bf z}_B)$. Let $h^{\rm int}({\bf
124: z}_A,{\bf z}_B)$ denote a small interaction term, turned ``on'' at
125: $t=0$, and ``off'' at $t=\tau$, coupling the two bodies. Let ${\bf
126: y}=({\bf z}_A,{\bf z}_B)$ specify a point in the {\it full} phase
127: space of all participating degrees of freedom. During any realization
128: of the process in which we are interested, the microscopic evolution
129: of the two bodies is described by a trajectory ${\bf y}(t)$, evolving
130: from $t=0$ to $t=\tau$ under Hamilton's equations, as derived from the
131: Hamiltonian
132: \begin{equation}
133: {\cal H}({\bf y}) = H^A({\bf z}_A) + H^B({\bf z}_B)
134: + h^{\rm int}({\bf y}).
135: \end{equation}
136:
137: We now further assume {\it time-reversal invariance}:
138: \begin{equation}
139: \label{eq:tri}
140: H^i({\bf z}_i) = H^i({\bf z}_i^*) ,\quad
141: h^{\rm int}({\bf y}) = h^{\rm int}({\bf y}^*),
142: \end{equation}
143: where $i=A,B$ and the asterisk (*) denotes the time-reversal
144: operation, usually the reversal of momenta: $({\bf q},{\bf p})^* =
145: ({\bf q},-{\bf p})$. This assumption has the crucial consequence that
146: microscopic realizations of the process come in pairs related by
147: time-reversal: for any trajectory ${\bf y}(t)$ which is a solution of
148: Hamilton's equations, its time-reversed twin, $\overline{\bf y}(t) =
149: {\bf y}^*(\tau-t)$, is also a solution. For future reference let
150: ${\bf y}^0$ and ${\bf y}^\tau$ denote the initial and final conditions
151: of the ``forward'' realization~\footnote{ Of course, for any pair of
152: realizations related by time-reversal, the designation of which is
153: the forward realization, and which is the reverse, is arbitrary.},
154: ${\bf y}(t)$; hence the ``reverse'' realization, $\overline{\bf
155: y}(t)$, evolves from $\overline{\bf y}^0={\bf y}^{\tau *}$ to
156: $\overline{\bf y}^\tau={\bf y}^{0*}$, as illustrated in
157: Figure~\ref{fig:twins}.
158: \begin{figure}[htbp]
159: \centering
160: \includegraphics[scale=0.5]{twins.eps}
161: \caption{Twin trajectories $y(t)$ and $\bar{y}(t) = y^*(\tau -t)$
162: related by time reversal.}
163: \label{fig:twins}
164: \end{figure}
165:
166: By our assumption regarding the equilibrium preparation of the two
167: bodies, the probability distribution for sampling initial conditions
168: ${\bf y}^0$ is given by:
169: \begin{equation}
170: \label{eq:heatinitprobdist}
171: P({\bf y}^0) =
172: {1\over Z_AZ_B}
173: e^{-H^A({\bf z}_A^0)/T_A}
174: e^{-H^B({\bf z}_B^0)/T_B},
175: \end{equation}
176: where the $Z$'s are partition functions. Given a trajectory ${\bf
177: y}(t)$ and its time-reversed twin $\overline{\bf y}(t)$, the ratio
178: of probabilities of sampling their respective initial conditions is
179: then:
180: \begin{equation}
181: {P({\bf y}^0)\over P(\overline{\bf y}^0)} =
182: e^{\Delta E_B/T_B} e^{\Delta E_A/T_A},
183: \end{equation}
184: where $ \Delta E_A = H^A({\bf z}_A^\tau) - H^A({\bf z}_A^0) =
185: H^A(\overline{\bf z}_A^0) - H^A({\bf z}_A^0) $, and similarly for
186: $\Delta E_B$. The quantities $\Delta E_A$ and $\Delta E_B$ represent
187: the net change in the internal energies of the two bodies, over the
188: course of the realization described by ${\bf y}(t)$. If we neglect
189: the small amount of work performed in switching on and off the
190: interaction term $h^{\rm int}$, then the net change in the energy of
191: one system is compensated by an opposite change in the energy of the
192: other, i. e. $\Delta E_B \approx -\Delta E_A$, and it is natural to
193: view these changes as representing a quantity of heat transfered from
194: $A$ to $B$: $Q := \Delta E_B \approx -\Delta E_A$. Hence,
195: \begin{equation}
196: \label{eq:heatratio}
197: { P({\bf y}^0)\over P(\overline{\bf y}^0)} =
198: e^{\Delta\beta\cdot \hat{Q}({\bf y}^0)},
199: \end{equation}
200: where the function $\hat Q({\bf y})$ denotes the value of $Q$ during a
201: realization evolving from initial conditions ${\bf y}$. Note that
202: \begin{equation}
203: \hat Q(\overline{\bf y}^0) = -\hat Q({\bf y}^0),
204: \label{eq:heatopp}
205: \end{equation}
206: that is, the heat transfer during the forward realization is opposite
207: to that during the reverse realization.
208:
209: Combining Eqs.~(\ref{eq:heatratio}) and~(\ref{eq:heatopp}) we get:
210: \begin{eqnarray}
211: p_\tau(Q) &=& \int d{\bf y}^0
212: P({\bf y}^0) \delta[Q-\hat Q({\bf y}^0)] \nonumber \\
213: &=& e^{\Delta\beta\cdot Q}
214: \int d\overline{\bf y}^0
215: P(\overline{\bf y}^0)
216: \delta[Q + \hat Q(\overline{\bf y}^0)] \nonumber \\
217: &=&
218: e^{\Delta\beta\cdot Q}
219: p_\tau(-Q),\label{eq:advertised}
220: \end{eqnarray}
221: which is equivalent to Eq.~(\ref{eq:heatft}). Here the change in the
222: variables of integration between the first and second lines is
223: justified by the invariance of the Liouville measure under time
224: evolution ($d{\bf y}^0 = d{\bf y}^\tau$), as well as under time
225: reversal ($d{\bf y}^\tau = d{\bf y}^{\tau*} = d{\overline{\bf y}}^0$).
226:
227:
228: These formal manipulations can be understood intuitively. $p_\tau(Q)$
229: is a sum of contributions from all realizations for which the heat
230: transfer takes on a specified value, $Q$; and $p_\tau(-Q)$ is a sum
231: over those for which the heat transfer is $-Q$. But these two sets of
232: realizations are in one-to-one correspondence; for every trajectory
233: ${\bf y}(t)$ belonging to the former set, its twin $\overline{\bf
234: y}(t)$ belongs to the latter (Eq.~\ref{eq:heatopp}). Moreover, the
235: ratio of initial condition sampling probabilities for such a twinned
236: pair of realizations is $e^{\Delta\beta\cdot Q}$
237: (Eq.~(\ref{eq:heatratio})). Therefore, when we add the sampling
238: probabilities $P({\bf y}^0)$ from the first set of realizations to get
239: $p_\tau(Q)$, and $P(\overline{\bf y}^0)$ from the second set to get
240: $p_\tau(-Q)$, the ratio of the sums is $e^{\Delta\beta\cdot Q}$.
241:
242:
243: The above derivation, based on comparing the sampling probabilities
244: for pairs of twinned trajectories, is similar to that carried out by
245: Evans and Searles~\cite{EvansS94} for the transient FT. Note also
246: that this derivation is valid for arbitrary times $\tau$; there are no
247: hidden assumptions that the temperatures of the two systems remain
248: constant, or even well-defined after $t=0$.
249:
250: The sole approximation that we have made is the neglect of the
251: interaction term, $h^{\rm int}$. In reality, a finite amount of work
252: is required to turn on this interaction initially, $\delta w_{\rm
253: on}$, and then to turn it off finally, $\delta w_{\rm off}$. The
254: resulting balance of energy reads: $\Delta E_A + \Delta E_B = \delta
255: w_{\rm on} + \delta w_{\rm off}$, hence $\delta w = \delta w_{\rm on}
256: + \delta w_{\rm off}$ enters as a correction to the approximation
257: $\Delta E_B \approx -\Delta E_A$ used earlier. The validity of our
258: final result thus requires that the work performed in coupling and
259: later uncoupling the systems ($|\delta w|$) be much smaller than the
260: typical energy change in either system ($|\Delta E_A|$, $|\Delta
261: E_B|$). Whether or not this condition is met depends, of course, on
262: details of the two systems, on the strength of the interaction term
263: ($\delta w \sim h^{\rm int}$), and on the duration $\tau$. It will be
264: interesting to investigate this issue in the context of specific
265: models.
266:
267: We proceed now to the proof of the quantum version of our theorem. We
268: assume that systems $A$ and $B$ have equilibrated to temperatures
269: $T_A, T_B$ before the experiment, and are thus described by density
270: matrices $\rho_i = \exp(- \beta_i H^i)/Z_i$, where $i=A,B$. At time
271: $t = 0^-$ we separate the systems from the reservoirs and measure
272: their energies. As a result, each system $i$ is projected onto a pure
273: state $|n_i\rangle$ with probability $e^{-\beta_i E_{n_i}^i}/Z_i$, and
274: the combined system is described by the product state $|n_A
275: n_B\rangle$. We then allow the systems to interact through a weak
276: coupling term $h^{\mathrm{int}}$. Thus the Hamiltonian takes the form
277: ${\cal H} = H^A\otimes I^B + I^A \otimes H^B + h^{\mathrm{int}}$.
278:
279: Let us now assume, as in the classical case (Eq.~(\ref{eq:tri})), that
280: the system and both its subsystems are time-reversal invariant. In
281: quantum mechanics the time-reversal invariance of a system is
282: expressed by the condition
283: \begin{equation}
284: \label{eq:qtri}
285: \Theta H = H \Theta,
286: \end{equation}
287: where $H$ is the system Hamiltonian, and $\Theta$ is the quantum
288: time-reversal operator~\cite{Merzbacher98,Ballentine98}. This
289: operator reverses linear and angular momentum while keeping position
290: unchanged, and is {\it anti-linear}:
291: \begin{equation}
292: \Theta \Bigl( \alpha_1 |\psi\rangle + \alpha_2 |\phi\rangle \Bigr)
293: = \alpha_1^\dagger \Theta |\psi\rangle +
294: \alpha_2^\dagger \Theta |\phi\rangle,
295: \end{equation}
296: where the dagger denotes complex conjugation. When dealing with such
297: operators, the Dirac bra-ket notation, invented to deal with linear
298: operators, becomes cumbersome: the expression
299: $\langle\phi|\Theta|\psi\rangle$ is ambiguous until we specify whether
300: $\Theta$ is acting to the right or to the left. To avoid this
301: inconvenience we will use the standard product in Hilbert space
302: $(|\phi\rangle , | \psi \rangle)$, rather than the more abbreviated
303: Dirac bra-ket, $\langle \phi | \psi \rangle$, to denote the inner
304: product between two wave functions. From Eq.~(\ref{eq:qtri}) it
305: follows that, for every eigenstate $|n\rangle$ of $H$ there
306: corresponds a time-reversed eigenstate $\Theta |n \rangle$ with the
307: same energy; these two states are either linearly independent, or else
308: identical apart from an overall phase. Moreover, since $\Theta$
309: preserves wave function normalization, it is not just anti-linear but
310: also {\it anti-unitary}: $( \Theta |\phi\rangle , \Theta |\psi\rangle
311: ) = ( |\psi\rangle , |\phi\rangle )$. We will make use of these
312: properties in the analysis below.
313:
314: Having turned on the interaction term at $t=0$, we allow the systems
315: to evolve for a time $\tau$. The combined system then reaches a state
316: $|\Psi\rangle$, obtained from the initial state $|n_A n_B\rangle$ by
317: evolution under Schr\" odinger's equation. We now separate the two
318: systems -- that is, we turn off the interaction term -- and once again
319: measure their energies. The state $|\Psi\rangle$ is thus projected
320: onto a product state $|m_A m_B\rangle$. As before, we make no
321: assumptions regarding $\tau$, in particular the systems have not
322: necessarily equilibrated.
323:
324: Letting $P_\tau(|n\rangle \rightarrow | m \rangle)$ denote the
325: probability of observing a transition from $| n \rangle \equiv | n_A
326: n_B \rangle$ to $| m \rangle \equiv | m_A m_B \rangle$, we have
327: \[
328: P_\tau(|n\rangle \rightarrow | m \rangle) = |( | m \rangle ,
329: U_\tau | n \rangle) |^2 \frac{ e^{-\beta_A E_{n_A}^A -\beta_B
330: E_{n_B}^B }}{Z_A Z_B} ,
331: \]
332: where $U_\tau = e^{- i \tau {\cal H}}$ is the quantum evolution
333: operator, and $\hbar = 1$. The second factor on the right is the
334: probability for sampling the initial state $| n \rangle$; the first
335: factor is the transition probability from $|n\rangle$ to $| m\rangle$.
336: Similarly, the probability of observing the time-reversed transition
337: from $ \Theta | m \rangle $ to $ \Theta | n\rangle $ is
338: \[
339: \!\! P_\tau(\Theta |m \rangle \rightarrow \Theta |n\rangle) = |(
340: \Theta |n \rangle , U_\tau \Theta| m \rangle) |^2 \frac{e^{-\beta_A
341: E_{m_A}^A -\beta_B E_{m_B}^B }}{Z_A Z_B}.
342: \]
343: But, since $\Theta$ is anti-unitary, and $U_\tau \Theta = \Theta
344: U_{-\tau}$~\footnote{ This follows from Eq.\ref{eq:qtri} and the
345: anti-linearity of $\Theta$.}, we have
346: \begin{eqnarray*}
347: ( \Theta |n \rangle , U_\tau \Theta| m \rangle) & = &
348: ( \Theta |n \rangle , \Theta U_{-\tau}| m \rangle)
349: = ( U_{-\tau}| m \rangle, |n\rangle) \\
350: & = & ( | m \rangle, U_\tau|n\rangle),
351: \end{eqnarray*}
352: therefore
353: \begin{equation}
354: \frac{
355: P_\tau(|n\rangle \rightarrow |m\rangle )
356: }{
357: P_\tau(\Theta |m\rangle \rightarrow \Theta |n\rangle )
358: } = e^{-\beta_A (E_{n_A}^A -E_{m_A}^A)} e^{-\beta_B (E_{n_B}^B -
359: E_{m_B}^B )}.
360: \end{equation}
361: Since we assumed that the interaction is weak, we expect the energy of
362: the total system to be almost preserved:
363: \begin{equation}
364: E_n^A + E_n^B \approx E_m^A + E_m^B.
365: \end{equation}
366: It follows that the energy changes in the two systems are
367: approximately equal
368: \begin{equation}
369: \label{eq:12}
370: Q_{n \rightarrow m} := E_m^B - E_n^B \approx E_n^A - E_m^A.
371: \end{equation}
372: We interpret $Q$ as the heat exchange between the systems $A$ and $B$.
373: Thus,
374: \begin{equation}
375: \frac{
376: P_\tau(|n\rangle \rightarrow |m\rangle )
377: }{
378: P_\tau(\Theta |m\rangle \rightarrow \Theta |n\rangle )
379: } \approx e^{\Delta \beta \cdot Q_{n \rightarrow m}}.
380: \end{equation}
381: Since every eigenstate has a corresponding time-reversed twin, the net
382: probability of the heat transfer $Q$ in time $\tau$ is
383: \begin{eqnarray}
384: p_\tau(Q) & = & \sum_{n,m} P_\tau(|n\rangle \rightarrow |m\rangle)
385: \delta(Q - Q_{n \rightarrow m}) \nonumber \\
386: & = & e^{\Delta \beta \cdot Q} \sum_{\Theta n,\Theta m}
387: P_\tau(\Theta |m\rangle \rightarrow \Theta |n\rangle ) \delta(Q +
388: Q_{\Theta m
389: \rightarrow \Theta n}) \nonumber \\
390: & = & e^{\Delta \beta \cdot Q} p_\tau(-Q). \label{eq:13.5}
391: \end{eqnarray}
392: This result is true for the quantities as we have defined them. We
393: can rewrite Eq.~(\ref{eq:13.5}) in the form of Eq.~(\ref{eq:heatft})
394: if we further assume a sufficiently dense spectrum, so that
395: $p_\tau(Q)$ can be replaced by a locally smooth function.
396:
397:
398:
399: At the level of macroscopic thermodynamics (and in the absence of
400: external work), the passage of heat from a colder to a hotter body
401: constitutes a violation of the Second Law. From
402: Eq.~(\ref{eq:heatft}), we can derive an upper bound on the probability
403: of observing such a ``violation'', of at least some finite magnitude,
404: as follows. Assume that $T_A>T_B$, i.e.\ $\Delta\beta>0$. The
405: probability that the heat transfer from $A$ to $B$ will fall below a
406: specified value $q$ is given by $\int_{-\infty}^q p_\tau(Q) dQ$.
407: Using Eq.~(\ref{eq:heatft}) to replace $p_\tau(Q)$ by
408: $p_\tau(-Q)\exp(\Delta\beta\cdot Q)$, and then invoking the inequality
409: chain
410: \[
411: \int_{-\infty}^q p_\tau(-Q)
412: e^{\Delta\beta\cdot Q} dQ
413: \le
414: e^{\Delta\beta\cdot q}
415: \int_{-\infty}^q p_\tau(-Q) dQ
416: \le
417: e^{\Delta\beta\cdot q},
418: \]
419: we get
420: \begin{equation}
421: \label{eq:bound}
422: \int_{-\infty}^q p_\tau(Q)\, dQ
423: \le e^{\Delta\beta\cdot q}.
424: \end{equation}
425: Choosing $q<0$, this result tells us that the probability of observing
426: a net heat transfer in the ``wrong'' direction ($Q<0$), from $B$
427: (cold) to $A$ (hot), of at least some magnitude $\vert q\vert$, dies
428: exponentially (or faster) with that magnitude. Eq.~(\ref{eq:heatft})
429: also implies that the average of $\exp(-\Delta\beta\cdot Q)$, over the
430: ensemble of realizations for any time $\tau$, is unity:
431: \begin{equation}
432: \overline{e^{-\Delta\beta\cdot Q}}
433: \equiv \int dQ p_\tau(Q) e^{-\Delta\beta\cdot Q} = 1.
434: \end{equation}
435:
436: In conclusion, a result analogous to the FT for entropy generation
437: (Eq.~(\ref{eq:ft})), and valid for arbitrary times $\tau$, has been
438: derived for the statistics of heat exchange between finite classical
439: or quantum systems separately prepared in equilibrium
440: (Eqs.~(\ref{eq:heatft})). In our derivation we invoke statistical
441: mechanics to describe the initial preparation of the systems, then
442: treat their evolution during the interval of contact dynamically. We
443: also assume a negligible energy of interaction between the two
444: systems, and a time-reversal invariant Hamiltonian. In the quantum
445: case, an additional source of randomness arises from the fact that the
446: initial quantum state of the system does not uniquely determine the
447: outcome of the final energy measurements. Nevertheless, this does not
448: spoil our result. We finally mention that a similar theorem can be
449: derived for particle exchange between two reservoirs, driven by a
450: difference in initial chemical potentials (unpublished).
451:
452: It is a pleasure to thank J.R. Anglin, H. van Beijeren, and D.J.
453: Thouless for stimulating and useful conversations. This research was
454: supported by the Department of Energy, under contract W-7405-ENG-36,
455: and by the Polish-American Maria Sk\l odowska-Curie Joint Fund II,
456: under project PAA / DOE-98-343. DW thanks the Center for Nonlinear
457: Science, School of Physics, Georgia Institute of Technology for
458: support as a Joseph Ford Fellow.
459:
460: \begin{thebibliography}{18}
461: \expandafter\ifx\csname natexlab\endcsname\relax\def\natexlab#1{#1}\fi
462: \expandafter\ifx\csname bibnamefont\endcsname\relax
463: \def\bibnamefont#1{#1}\fi
464: \expandafter\ifx\csname bibfnamefont\endcsname\relax
465: \def\bibfnamefont#1{#1}\fi
466: \expandafter\ifx\csname citenamefont\endcsname\relax
467: \def\citenamefont#1{#1}\fi
468: \expandafter\ifx\csname url\endcsname\relax
469: \def\url#1{\texttt{#1}}\fi
470: \expandafter\ifx\csname urlprefix\endcsname\relax\def\urlprefix{URL }\fi
471: \providecommand{\bibinfo}[2]{#2}
472: \providecommand{\eprint}[2][]{\url{#2}}
473:
474: \bibitem[{\citenamefont{Evans et~al.}(1993)\citenamefont{Evans, Cohen, and
475: Morriss}}]{EvansCM93}
476: \bibinfo{author}{\bibfnamefont{D.~J.} \bibnamefont{Evans}},
477: \bibinfo{author}{\bibfnamefont{E.~G.~D.} \bibnamefont{Cohen}},
478: \bibnamefont{and} \bibinfo{author}{\bibfnamefont{G.~P.}
479: \bibnamefont{Morriss}}, \bibinfo{journal}{Phys. Rev. Lett.}
480: \textbf{\bibinfo{volume}{71}}, \bibinfo{pages}{2401} (\bibinfo{year}{1993}).
481:
482: \bibitem[{\citenamefont{Evans and Searles}(1994)}]{EvansS94}
483: \bibinfo{author}{\bibfnamefont{D.~J.} \bibnamefont{Evans}} \bibnamefont{and}
484: \bibinfo{author}{\bibfnamefont{D.~J.} \bibnamefont{Searles}},
485: \bibinfo{journal}{Phys. Rev. E} \textbf{\bibinfo{volume}{50}},
486: \bibinfo{pages}{1645} (\bibinfo{year}{1994}).
487:
488: \bibitem[{\citenamefont{Gallavotti and Cohen}(1995)}]{GallavottiC95}
489: \bibinfo{author}{\bibfnamefont{G.}~\bibnamefont{Gallavotti}} \bibnamefont{and}
490: \bibinfo{author}{\bibfnamefont{E.~G.~D.} \bibnamefont{Cohen}},
491: \bibinfo{journal}{J. Stat. Phys.} \textbf{\bibinfo{volume}{80}},
492: \bibinfo{pages}{931} (\bibinfo{year}{1995}).
493:
494: \bibitem[{\citenamefont{Kurchan}(1998)}]{Kurchan98}
495: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Kurchan}}, \bibinfo{journal}{J.
496: Phys. A-Math. Gen.} \textbf{\bibinfo{volume}{31}}, \bibinfo{pages}{3719}
497: (\bibinfo{year}{1998}).
498:
499: \bibitem[{\citenamefont{Lebowitz and Spohn}(1999)}]{LebowitzS99}
500: \bibinfo{author}{\bibfnamefont{J.~L.} \bibnamefont{Lebowitz}} \bibnamefont{and}
501: \bibinfo{author}{\bibfnamefont{H.}~\bibnamefont{Spohn}}, \bibinfo{journal}{J.
502: Stat. Phys.} \textbf{\bibinfo{volume}{95}}, \bibinfo{pages}{333}
503: (\bibinfo{year}{1999}).
504:
505: \bibitem[{\citenamefont{Maes}(1999)}]{Maes99}
506: \bibinfo{author}{\bibfnamefont{C.}~\bibnamefont{Maes}}, \bibinfo{journal}{J.
507: Stat. Phys.} \textbf{\bibinfo{volume}{95}}, \bibinfo{pages}{367}
508: (\bibinfo{year}{1999}).
509:
510: \bibitem[{\citenamefont{Evans and Searles}(2002)}]{Evans02}
511: \bibinfo{author}{\bibfnamefont{D.~J.} \bibnamefont{Evans}} \bibnamefont{and}
512: \bibinfo{author}{\bibfnamefont{D.}~\bibnamefont{Searles}},
513: \bibinfo{journal}{Adv. in Phys.} \textbf{\bibinfo{volume}{51}},
514: \bibinfo{pages}{1529} (\bibinfo{year}{2002}).
515:
516: \bibitem[{\citenamefont{Wang et~al.}(2002)\citenamefont{Wang, Sevick, Mittag,
517: Searles, and Evans}}]{WangSMSE02}
518: \bibinfo{author}{\bibfnamefont{G.~M.} \bibnamefont{Wang}},
519: \bibinfo{author}{\bibfnamefont{E.~M.} \bibnamefont{Sevick}},
520: \bibinfo{author}{\bibfnamefont{E.}~\bibnamefont{Mittag}},
521: \bibinfo{author}{\bibfnamefont{D.~J.} \bibnamefont{Searles}},
522: \bibnamefont{and} \bibinfo{author}{\bibfnamefont{D.~J.} \bibnamefont{Evans}},
523: \bibinfo{journal}{Phys. Rev. Let.} \textbf{\bibinfo{volume}{89}},
524: \bibinfo{pages}{050601} (\bibinfo{year}{2002}).
525:
526: \bibitem[{\citenamefont{Crooks}(1999)}]{crooks99}
527: \bibinfo{author}{\bibfnamefont{G.~E.} \bibnamefont{Crooks}},
528: \bibinfo{journal}{Phys. Rev. E} \textbf{\bibinfo{volume}{60}},
529: \bibinfo{pages}{2721} (\bibinfo{year}{1999}).
530:
531: \bibitem[{\citenamefont{Jarzynski}(1997)}]{Jarzynski97}
532: \bibinfo{author}{\bibfnamefont{C.}~\bibnamefont{Jarzynski}},
533: \bibinfo{journal}{Phys. Rev. Lett.} \textbf{\bibinfo{volume}{78}},
534: \bibinfo{pages}{2690} (\bibinfo{year}{1997}).
535:
536: \bibitem[{\citenamefont{Crooks}(1998)}]{Crooks98}
537: \bibinfo{author}{\bibfnamefont{G.~E.} \bibnamefont{Crooks}},
538: \bibinfo{journal}{J. Stat. Phys.} \textbf{\bibinfo{volume}{90}},
539: \bibinfo{pages}{1481} (\bibinfo{year}{1998}).
540:
541: \bibitem[{\citenamefont{Liphardt et~al.}(2002)\citenamefont{Liphardt, Dumont,
542: Smith, Tinoco, and Bustamante}}]{Liphardt02}
543: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Liphardt}},
544: \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Dumont}},
545: \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Smith}},
546: \bibinfo{author}{\bibfnamefont{I.}~\bibnamefont{Tinoco}}, \bibnamefont{and}
547: \bibinfo{author}{\bibfnamefont{C.}~\bibnamefont{Bustamante}},
548: \bibinfo{journal}{Science} \textbf{\bibinfo{volume}{296}},
549: \bibinfo{pages}{1832} (\bibinfo{year}{2002}).
550:
551: \bibitem[{\citenamefont{Kurchan}(2000)}]{kurchan00s}
552: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Kurchan}}
553: (\bibinfo{year}{2000}), \bibinfo{note}{cond-mat/0007360}.
554:
555: \bibitem[{\citenamefont{Mukamel}(2003)}]{Mukamel03}
556: \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Mukamel}},
557: \bibinfo{journal}{Phys. Rev. Lett.} \textbf{\bibinfo{volume}{90}},
558: \bibinfo{pages}{170604} (\bibinfo{year}{2003}).
559:
560: \bibitem[{\citenamefont{Monnai and Tasaki}(2003)}]{monnai03}
561: \bibinfo{author}{\bibfnamefont{T.}~\bibnamefont{Monnai}} \bibnamefont{and}
562: \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Tasaki}}
563: (\bibinfo{year}{2003}), \bibinfo{note}{cond-mat/0308337}.
564:
565: \bibitem[{\citenamefont{De~Roeck and Maes}(2003)}]{deroeck03}
566: \bibinfo{author}{\bibfnamefont{W.}~\bibnamefont{De~Roeck}} \bibnamefont{and}
567: \bibinfo{author}{\bibfnamefont{C.}~\bibnamefont{Maes}}
568: (\bibinfo{year}{2003}), \bibinfo{note}{cond-mat/0309498}.
569:
570: \bibitem[{\citenamefont{Merzbacher}(1998)}]{Merzbacher98}
571: \bibinfo{author}{\bibfnamefont{E.}~\bibnamefont{Merzbacher}},
572: \emph{\bibinfo{title}{Quantum Mechanics}} (\bibinfo{publisher}{J. Wiley \&
573: Sons}, \bibinfo{address}{New York}, \bibinfo{year}{2001}),
574: \bibinfo{edition}{3rd} ed.
575:
576: \bibitem[{\citenamefont{Ballentine}(1998)}]{Ballentine98}
577: \bibinfo{author}{\bibfnamefont{L.}~\bibnamefont{Ballentine}},
578: \emph{\bibinfo{title}{Quantum Mechanics: A Modern Development}}
579: (\bibinfo{publisher}{World Scientific}, \bibinfo{address}{Singapore},
580: \bibinfo{year}{1998}), \bibinfo{edition}{2nd} ed.
581:
582: \end{thebibliography}
583:
584: \end{document}
585: