1: %\documentclass[prl,twocolumn,aps,superscriptaddress,showpacs]{revtex4}

2: \documentclass[pre,twocolumn,aps,showpacs]{revtex4}

3: %\documentclass[prl,preprint,aps,showpacs]{revtex4}

4: %\documentclass[pre,preprint,showpacs]{revtex4}

5: \usepackage{graphicx}

6: \usepackage{amsmath}

7: \usepackage{amsfonts}

8: %\usepackage[left]{lineno}

9: %\linenumbers

10: %\usepackage{showlabels}

11: \bibliographystyle{apsrev}

12: \def\btt#1{{\tt$\backslash$#1}}

13: \renewcommand{\textfraction}{0}

14: \def\be{\begin{equation}}

15: \def\ba{\begin{eqnarray}}

16: \def\a{\alpha}

17: \def\b{\beta}

18: \def\g{\gamma}

19: \def\G{\Gamma}

20: \def\d{\delta}

21: \def\D{\Delta}

22: \def\e{\epsilon}

23: \def\z{\zeta}

24: \def\h{\eta}

25: \def\th{\theta}

26: \def\Th{\Theta}

27: \def\k{\kappa}

28: \def\l{\lambda}

29: \def\La{\Lambda}

30: \def\m{\mu}

31: \def\n{\nu}

32: \def\x{\xi}

33: \def\p{\pi}

34: \def\r{\rho}

35: \def\s{\sigma}

36: \def\t{\tau}

37: \def\f{\varphi}

38: \def\c{\chi}

39: \def\ps{\psi}

40: \def\o{\omega}

41: \def\vo{\varpi}

42: \def\O{\Omega}

43: \def\i{\int}

44: \def\lg{\langle \g \rangle}

45: \def\bA{{\mathbf A}}

46: \def\bB{{\mathbf B}}

47: \def\bx{{\mathbf x}}

48: \def\bv{{\mathbf v}}

49: \def\bD{{\mathbf D}}

50: \def\bM{{\mathbf M}}

51: \def\bb{{\mathbf b}}

52: \def\bff{{\mathbf f}}

53: \def\bG{{\mathbf G}}

54: \def\bV{{\mathbf V}}

55: \def\bq{{\mathbf q}}

56: \def\bp{{\mathbf p}}

57: \def\bz{{\mathbf z}}

58: \def\bZ{{\mathbf Z}}

59: \def\bp{{\mathbf p}}

60: \def\bq{{\mathbf q}}

61: \def\cP{{\mathcal P}}

62: \def\cG{{\mathcal G}}

63: \def\cT{{\mathcal T_>}}

64: \def\cH{{\mathcal H}}

65: \def\br{\mbox{\boldmath{$\r$}}}

66: \def\tr{\text{tr}}

67: %\def\bfi{\mbox{\boldmath{$\f$}}}

68: %\def\bg{\mbox{\boldmath{$\g$}}}

69: %\def\bps{\mbox{\boldmath{$\ps$}}}

70: %\def\bz{\mbox{\boldmath{$\z$}}}

71: \def\Tr{\mbox{Tr}}

72: \def\tc{\mbox{TC}}

73: %\def\bLa{\mbox{\boldmath{$\La$}}}

74: \def\ee#1{\label{#1}\end{equation}}

75: \def\ea#1{\label{#1}\end{eqnarray}}

76: %\usepackage[left]{lineno}

77: %\linenumbers

78: \begin{document}

79: %\draft

80: %\preprint{HEP/123-qed}

81: \title{Statistics of work performed on a forced quantum  

82: oscillator}

83: \author{Peter Talkner, P. Sekhar Burada, and Peter H\"anggi}

84: \affiliation{Institute of Physics, University of Augsburg,

85: D-86135 Augsburg, Germany}

86: \date{\today}

87: \begin{abstract}

88: Various aspects of the statistics of work performed by an external

89: classical force on a quantum

90: mechanical system are elucidated for a driven harmonic

91: oscillator. In this special case two parameters are introduced that 

92: are sufficient to completely characterize the force protocol. 

93: Explicit results for the

94: characteristic function of work and the respective probability

95: distribution are provided and discussed for three different types of

96: initial states of the oscillator: microcanonical, canonical and coherent

97: states.  Depending on the choice of the initial state the probability

98: distributions of the performed work may grossly differ. This result in

99: particular holds also true for

100: identical force protocols. General fluctuation and work theorems holding for

101: microcanonical and canonical initial states are confirmed.

102: \end{abstract}

103: \pacs{05.30.-d,05.70.Ln,05.40.-a}

104: \maketitle

105: \section{Introduction}

106: During the last decade various fluctuation and work theorems

107: \cite{fwt,ECM} have been formulated and discussed. They inter

108: alia characterize

109: the full nonlinear response of a system under the action of a time

110: dependent force \cite{J,J07}.

111: These theorems have been derived and experimentally

112: confirmed primarily for classical systems \cite{ex,BLR,BHSB}. 

113: Quantum mechanical generalizations

114: were proposed recently \cite{qm,M,RM,EM,TLH,TH,TMH,DL}. 

115: 

116: Conceptual problems though arise 

117: in the context of quantum mechanics 

118: if one tries to generalize those classical

119: relations that require for example the specification of a system's trajectory

120: extending over  

121: some interval of time, or the simultaneous measurement of

122: noncommuting observables.

123: For example, the measurement of work performed by an external force on

124: an otherwise isolated system may be accomplished in the framework of

125: classical physics in principle in two different ways. The first method

126: is based on two measurements of the energy, one at the beginning and

127: the second at the end of the considered process. This method becomes

128: unreliable in practice if the system is large and the work performed on the

129: system is negligibly small compared to the total energy of the system.

130: Such a situation typically arises if the system of interest, on which the

131: force exclusively acts, interacts with its environment. In order to

132: retain an isolated system, the large system made of the open system and

133: its environment must be considered. Again, the work performed on the

134: system results as the difference of the energies of the total system,

135: which both may be very large. 

136: 

137: For classical systems, this

138: unfortunate situation can be circumvented by a second method,  

139: by monitoring the state of

140: the relevant small system during the time when the force is acting. 

141: Having this information at hand one can  

142: determine the work by 

143: integrating the power supplied to the system at each instant of time. 

144: The respective power can be inferred from the registered state of

145: the system and the known force protocol. In a quantum

146: system a continuous measurement of even a single observable would strongly

147: influence and possibly manifestly distort the system's dynamics.

148: Apparently, only the first method of two energy measurements is

149: feasible, at least in principle, in the quantum context.  

150: 

151: An alternative method based on a continuous monitoring 

152: has recently

153: been suggested by Esposito and Mukamel \cite{EM} for open quantum

154: systems described

155: by Markovian quantum master equations. 

156: There the dynamics of the density matrix is 

157: mapped onto a classical rate process for which known fluctuation

158: theorems can be applied \cite{Seif}. This provides an interesting

159: formal approach but its physical meaning has remained unclear

160: \cite{EM}. 

161: Moreover,

162: this approach is restricted to open systems that only weakly

163: interact with their respective environments. 

164: 

165: In the present paper the distribution of work is discussed for the

166: exactly solvable system of a driven harmonic oscillator \cite{DL,H}. In

167: this case, 

168: the distribution of work is discrete. We provide formal expressions

169: for this distribution and its corresponding characteristic function

170: which are valid for all initial states of the system as well as for

171: all possible kinds of force protocols. In particular, we determine the

172: characteristic functions and distributions of the work for

173: microcanonical, canonical and coherent initial states which lead to

174: qualitatively different work distributions.

175: 

176: The paper is organized as follows. In Sect.~\ref{II} we review the general form

177: of the characteristic function of work performed on a system in terms

178: of a correlation function of the exponentiated Hamiltonians at the

179: initial and final time of the force protocol. We prove that this

180: particular expression indeed always represents a characteristic

181: function, i.e. the Fourier transform of a probability density.

182: Sect.~\ref{III} presents various fluctuation and work theorems for

183: canonical and microcanonical initial states. In Sect.~\ref{IV} general

184: expressions for the characteristic function and the corresponding

185: probability distribution of work are derived for a driven harmonic

186: oscillator. Moreover, the expressions for the first four cumulants are

187: derived. The dependence of the work distribution on the force protocol 

188: for microcanonical,

189: canonical and coherent initial states as well as its dependence on the

190: specific parameters of these initial states are investigated.

191:  

192:  

193: %{\it main aspects of the paper, short description of content }

194:            

195: 

196: \section{Characteristic function of work} 

197: \label{II}

198: The response of a quantum system on a perturbation by a classical,

199: external force can be characterized by the change of energies

200: contained in the total system. The energy as an observable coincides

201: with its Hamiltonian $H(t)$  of

202: the total system. It includes

203: the external force and therefore depends on time. We will consider the

204: dynamics of the system only within a finite window of time $[t_0,t_f]$

205: during which the force is acting in a prescribed way, resulting in a

206: protocol of Hamiltonians which is denoted by $\{H(t)\}_{t_f,t_0}$.

207: Apart from the action of the external force the system is assumed to

208: be closed. Its dynamics is consequently governed

209: by a unitary time evolution $U_{t,t_0}$,

210: which is the solution of the Sch\"odinger equation

211: \be

212: \begin{split}

213: i \hbar \partial U_{t,t_0} / \partial t &= H(t) U_{t,t_0}, \\

214: U(t,t_0)&=1. 

215: \end{split}

216: \ee{SE}

217: As explained in the introduction, the work $w$ is measured as the

218: difference of the energies of the system at the final and initial times

219: $t_{f}$ and $t_{0}$. In a single measurement  the work is given by

220: the difference of two eigenvalues $e_{n}(t_{f})$ and $e_{m}(t_{0})$ of the

221: Hamiltonians $H(t)$ at the respective times $t_{f}$ and $t_{0}$,

222: i.e. by $w=e_{n}(t_{f})-e_{m}(t_{0})$. The inherent randomness of the

223: outcome of a quantum measurement in general leads to a  measured work

224: that is random.   

225: A complete description of the statistical properties of the work

226: performed on the system is provided by the characteristic function

227: $G_{t_{0},t_{f}}(u)$ 

228: which presents the Fourier transform of the probability density of the

229: work $p_{t_{f},t_{0}}(w)$, i.e.

230: \be

231: G_{t_{0},t_{f}}(u) = \int dw\: e^{iuw} p_{t_{f},t_{0}}(w).

232: \ee{Gp}

233: It can be expressed as a quantum correlation function of the

234: exponentiated Hamiltonian at the initial and the final time

235: \cite{TMH}, i.e.  

236: \be

237: \begin{split}

238: G_{t_{0},t_{f}}(u) &= \langle e^{iu H(t_{f})} e^{-iu H(t_{0})} \rangle

239: \\ 

240: & \equiv \Tr e^{i u H_{H}(t_{f})} e^{-iu H(t_{0})} \bar{\r}(t_{0}),

241: \end{split}

242: \ee{GC}

243: where 

244: \be

245: H_{H}(t_{f}) = U^{+}_{t_{f},t_{0}} H(t_{f}) U_{t_{f},t_{0}}

246: \ee{HH}

247: denotes the Hamiltonian in the Heisenberg picture. The density

248: matrix $\bar{\r}(t_{0})$ from the initial density matrix $\r(t_{0})$

249: as a result of the

250: measurement of the Hamiltonian $H(t_{0})$. It is given by

251: \be

252: \bar{\r}(t_{0}) = \sum_{n} P_{n}(t_{0}) \r(t_{0}) P_{n}(t_{0}),    

253: \ee{br}

254: where the operators $P_{n}(t_{0})$ denote the eigenprojection operators of the

255: Hamiltonian at time $t_{0}$, which present a partition of the unity

256: \be

257: \sum_{k} P_{n}(t_{0}) = 1.

258: \ee{P1}  

259: Before we apply the general expression (\ref{GC}) 

260: to a particular system and investigate its dependence on the

261: initial state $\r(t_{0})$, we discuss three general properties of the

262: correlation expression (\ref{GC}) which guarantee that the resulting

263: function $G_{t_{f},t_{0}}(u)$ indeed always presents a proper characteristic

264: function of a classical random variable $w$. This is the consequence

265: of the three following properties: \\ 

266: (i) $G_{t_f,t_{0}}(u)$ is a continuous function of $u$.\\

267: (ii)  $G_{t_f,t_0}(u)$ is a positive definite function of $u$, i.e. 

268: for all integer numbers $n$, 

269: all real sequences $u_1,u_2,\ldots u_n$, and all complex numbers

270: $z_{i}$, $i=1,2 \ldots n$ 

271: \be

272: \sum_{i,i'}^n  G_{t_f,t_0}(u_i-u_{i'})

273: z^*_{i} z_{i'} \geq 0 

274: \ee{pd}

275: holds. Here, the asterisk $z_{i}^*$ denotes the complex conjugate of

276: $z_{i}$.\\

277: (iii) $G_{t_f,t_0}(0) = 1$\\

278: According to a theorem by Bochner \cite{Boch} the properties (i-iii)

279: are necessary and sufficient conditions in order that the function

280: $G_{t_f,t_0}(u)$ 

281: is the Fourier transform of the probability measure of a random

282: variable. In short, the first condition insures that, strictly speaking,

283: the function $G_{t_{f},t_{0}}(u)$ is the Fourier transform of a

284: measure, the second condition assures that this measure is positive

285: and the third condition that it is normalized.

286: Hence the correlation expression eq. (\ref{GC}) always

287: defines a proper characteristic function. For a

288: proof of the properties (i-iii) we refer the reader to the appendix \ref{AA}.

289: 

290: \section{Canonical and microcanonical initial states}

291: \label{III}

292: In  experiments an external force is often applied on a system that

293: initially is found in a 

294: thermodynamic equilibrium state. Depending on whether the system was

295: in weak contact with a heat bath or was totally isolated from its

296: environment, the 

297: initial state of the system is described either by a canonical or a 

298: microcanonical density matrix. For both situations fluctuation and

299: work theorems are known. We will shortly review these relations.

300: \subsection{Work and fluctuation theorems for canonical initial

301:   states} 

302: 

303: If the initial density matrix is canonical, i.e. 

304: if 

305: \be

306: \r(t_0) = Z^{-1}(t_0) \exp \{ -\b H(t_0) \},

307: \ee{cs} 

308: where 

309: \be

310: Z(t_0) = \Tr \exp \{ - \b H(t_0) \} = e^{-\b F(t_{0})}

311: \ee{Z} 

312: denotes the partition function and $F(t_{0})$ the free energy, 

313: then $[H(t_0),\r(t_0)]=0$

314: and the first measurement of

315: the energy leaves the density matrix unchanged, such that

316: $\bar{\r}(t_0) = \r(t_0)$. With eq. (\ref{GC}) this leads to the

317: characteristic function of work for a canonical initial state which was

318: derived in Ref.~\cite{TLH}.

319: In this case, $G_{t_f,t_0}(u)$ can be

320: continued to an analytic function of $u$ for all $0\leq \Im u \leq

321: \b$ \cite{TH}. For the particular value $u=i\b$ the characteristic function

322: yields the mean value of the exponentiated work, $\langle \exp\{-\b w\}

323: \rangle$ and the 

324: correlation function expression (\ref{GC}) simplifies to the ratio of

325: the partition functions at the times $t_f$ and $t_0$, resulting in the

326: Jarzynski work theorem

327: \be

328: \langle e^{-\b w} \rangle    = Z(t_f)/Z(t_0) = \exp\left \{-\b

329:   (F(t_{f})-F(t_{0})) \right \},

330: \ee{J}

331: where $Z(t_f) = \tr \exp \{-\b H(t_f) \} = \exp \{- \b F(t_{f}) \}$.

332: 

333: Within the domain of analyticity $\mathcal{S} = \{u | 0 \leq \Im u

334: \leq \b\}$ the characteristic functions for the original and the time reversed

335: protocol are related to each other by the following formula, cf. \cite{TH} 

336: \be

337: G_{t_f,t_0}(u) = \frac{Z(t_f)}{Z(t_0)} G_{t_0,t_f}(-u+i\b),

338: \ee{GTC} 

339: where $G_{t_0,t_f}(u)$ refers to processes under the time reversed

340: protocol $\{H(t)\}_{t_o,t_f}$ starting from the canonical

341: state $Z^{-1}(t_f) \exp \{ -\b H(t_f)\}$. An inverse Fourier

342: transform leads to the Tasaki-Crooks fluctuation theorem, which

343: relates the probability

344: densities  of work $p_{t_f,t_0}(w)$ for a given protocol to

345: the density of

346: the work $p_{t_0,t_f}(w)$ for the time reversed protocol. This theorem

347: explicitly reads \cite{TH}

348: \be

349: \frac{p_{t_f,t_0}(w)}{p_{t_0,t_f}(-w)} = \frac{Z(t_f)}{Z(t_0)} e^{\b

350:   w} = e^{-\b(F(t_{f})-F(t_{0})-w)}.

351: \ee{TC}  

352: \subsection{Fluctuation theorems for microcanonical initial states}

353: A system in a microcanonical state is described by the density matrix

354: \be

355: \r(t_{0}) = \vo_{E}^{-1}(t_{0}) \d (H(t_{0})-E),

356: \ee{rmc}

357: where 

358: \be

359: \vo_{E}(t_{0}) = \Tr\: \d (H(t_{0})-E) = \exp\left \{S(E,t_{0})/k_{B}

360: \right \}

361: \ee{oE}

362: denotes the density of states as a function of the energy $E$ of the

363: system. The density of states can be expressed in terms of the

364: entropy of the system $S_{E}(t_{0})$ provided the spectrum of the

365: system Hamiltonian is sufficiently dense such that the density of

366: states becomes a smooth function on a coarsened energy scale. 

367: The microcanonical density

368: matrix commutes with the Hamiltonian $H(t_{0})$. Consequently,

369: $\bar{\r}(t_{0})$ and $\r(t_{0})$ coincide. 

370: 

371: The microcanonical quantum Crooks theorem

372: assumes the form \cite{TMH}

373: \be

374: \begin{split}

375: \frac{p_{t_{f},t_{0}}(E,w)}{p_{t_{0},t_{f}}(E+w,-w)} &=

376: \frac{\vo_{E+w}(t_{f})}{\vo_{E}(t_{0})}\\

377: & =

378: \exp \left \{\frac{S(E+w,t_{f})-S(E,t_{0})}{k_{B}} \right \}.

379: \end{split}

380: \ee{mcC}

381: Analogous to the canonical case it relates the probability density

382: $p_{t_{0},t_{0}}(E,w)$ of work $w$,

383: for a system starting in a microcanonical state with energy $E$, to the

384: respective quantity for the time reversed process starting

385: at energy $E+w$. This quantum theorem is formally identical to the

386: respective classical theorem \cite{CBK}.     

387: 

388: From the microcanonical Crooks theorem the probability density

389: relating to the time reversed process can be eliminated to yield the

390: so-called entropy-from-work theorem \cite{TMH}, reading:

391: \be

392: \vo_{E_{f}}(t_{f}) = \i dw\: \vo_{E_f-w}(t_{0})

393: p_{t_{f},t_{0}}(E_{f}-w,w) \;.

394: \ee{efw}  

395: This theorem allows one to determine the unknown density of states of a system

396: with Hamiltonian $H(t_{f})$ from the known density of states of a reference

397: system $H(t_{0})$ by means of the statistics of the work that is

398: performed on the system in a process that leads from the reference

399: system to the final system with unknown density of states. In the case

400: of systems with a sufficiently smooth density of states

401: the respective entropy can be determined. For further details

402: see in Ref.~\cite{TMH}.   

403: \section{Driven harmonic oscillator}

404: \label{IV}

405: To illustrate these concepts we consider 

406: an example which allows the analytical construction of the

407: probability of work. Specifically we consider a harmonic oscillator 

408: on which a time

409: dependent force acts during a finite interval of time.  

410: Its time evolution is governed 

411: by the Hamiltonian

412: \be

413: H(t) = \hbar \o a^+ a + f^*(t) a + f(t) a^+,

414: \ee{hoH}

415: where $\o$ denotes the angular frequency, and $a^+$ and $a$ creation and

416: annihilation operators, respectively, which obey the usual commutation

417: relation, i.e. $[a,a^+]=1$. The complex driving force $f(t)$ allows

418: for a coupling

419: to position and/or momentum of the oscillator.  

420: We assume that $f(t)$ vanishes for times $t\leq t_0=0$. 

421: It is our aim to study the influence of the initial state

422: $\r(t_{0})$ on the statistics of work performed on the oscillator. 

423: The measurement of

424: $H(t_{0}) = \hbar \o a^{+} a$

425: at time $t_{0}=0$ then yields the result $\hbar \o n$ with probability 

426: \be

427: p_{n} = \langle n|\r(t_{0})|n\rangle.

428: \ee{pn}

429: Accordingly, the oscillator is found in the state

430: \be

431: \bar{\r}(t_{0}) = \sum_{n} p_{n} |n\rangle \langle n|

432: \ee{bro}

433: immediately after this measurement. Putting this density matrix in the

434: general expression for the characteristic function, eq.~(\ref{GC}) one

435: obtains

436: \be

437: G_{t_f,t_{0}}(u) = \sum_n p_n e^{-iu\hbar \o

438:   n} \langle n |

439: e^{iu H_H(t_f)} |n \rangle\;.

440: \ee{G0} 

441: For the driven harmonic oscillator the diagonal matrix element of the

442: exponentiated Hamiltonian $H_{H}(t_{f})$ can be determined \cite{H}. For

443: details see the Appendix~\ref{B}. With the expression (\ref{nxy}) for

444: the matrix element $\langle n| \exp \left \{ iH_{H}(t_{f})\right \} |n

445: \rangle$ we

446: find

447: \begin{widetext}

448: \be

449: \begin{split}

450: G_{t_f,t_{0}}(u) =& \:  e^{iu |f(t_f)|^2/(\hbar \o)} 

451: \exp \left \{ \left ( e^{iu\hbar \o}-1

452:   \right )|z|^2 \right \}   \sum_{n=0}^\infty\sum_{k=0}^n p_n

453: \binom{n}{k} \frac{|z|^{2(n-k)}}{(n-k)!} e^{-iu\hbar \o (n-k)} \left

454:     (e^{iu\hbar \o}-1 \right )^{2(n-k)} \\

455: &=\:  e^{iu |f(t_f)|^2/(\hbar \o)} 

456: \exp \left \{ \left ( e^{iu\hbar \o}-1

457:   \right )|z|^2 \right \} \sum_{n=0}^{\infty}p_{n}L_{n}\!\left (4 |z|^{2}

458: \sin^{2}\frac{\hbar \o u}{2}\right )\;,

459: \end{split}

460: \ee{Gho} 

461: \end{widetext}

462: where $|f(t_f)|^2/(\hbar \o)$ denotes a uniform shift of the 

463: spectrum of the harmonic

464: oscillator due to the presence of the external force, cf.

465: eq. (\ref{L}), and 

466: \be

467: z = \frac{1}{\hbar \o} \i_0^{t_f} ds \dot{f}(s) \exp \{i \o s\} 

468: \ee{zf}

469: is a dimensionless functional of the driving force $f(t)$, cf. eq. (\ref{z}). 

470: This dimensionless quantity vanishes in particular for all-quasi

471: static forcings,

472: i.e. if the force changes only very slowly with 

473: $f(t)= g(t/t_{f})$ for $t_{f} \to \infty$,  

474: where $g(\t)$ is a continuously differentiable

475: function for $\t \in [0,1]$. We hence call $z(t)$ the {\it rapidity

476: parameter} of the force protocol.  

477: Finally, $L_{n}(x) =\sum_{k=0}^{n} \binom{n}{k} (-x)^{k}/k!$ denotes

478: the Laguerre polynomial of order $n$ \cite{RG}.

479: 

480: 

481: Introducing the cumulant generating function $K(u)= \ln G(u)$ one

482: obtains 

483: the cumulants of work $k_{n}$  

484: as the $n$th derivatives of $K(u)$  with respect to $u$ taken at $u=0$

485: \cite{vK}, i.e. $k_{n} = (-i)^{n} d^{n }K(0)/d u^{n}$. The 

486: first four cumulants become:

487: \begin{align}

488: k_{1} &= \langle w \rangle \nonumber\\ &= \frac{|f(t_f)|^2}{\hbar \o} +\hbar \o

489: |z|^2, \label{aw}\\

490: k_{2} &=\langle w^{2} \rangle -\langle w \rangle^{2}\nonumber\\

491:  &= 2 (\hbar \omega)^2 |z|^2

492: \left ( \langle a^+ a \rangle_0 +\frac{1}{2} \right ),\label{ww}\\

493: k_{3} &=\langle w^{3} \rangle- 3 \langle w^{2} \rangle \langle w

494: \rangle +2 \langle w

495: \rangle^{3}\nonumber\\ 

496: &= \left (\hbar \o \right )^{3} |z|^{2},\\

497: k_{4}&=  \langle w^{4} \rangle -4 \langle w^{3} \rangle\langle w

498: \rangle -3 \langle w^{2} \rangle^{2} \nonumber \\

499: & \quad +12 \langle w^{2} \rangle \langle

500: w \rangle^{2} - 6 \langle w \rangle^{4}\nonumber\\ 

501: &= \left (\hbar \o \right )^{4}

502: |z|^{2} \left \{ 1+4\langle a^{+}a \rangle_{0} + 6 \left [\langle

503:   a^{+}a(a^{+}a-1)\rangle_{0} \right . \right .\nonumber \\

504: &\left . \left .\quad - 2\langle a^{+}a \rangle_{0} \right ]|z|^{2}\right \}.   

505: \label{cu}

506: \end{align}

507: The odd cumulants of the work are independent of the

508: initial preparation. The even cumulants depend on the

509: factorial moments of the number operator $a^{+}a$ with respect to

510: the initial state $\bar{\r}(t_{0})$ such as $\langle

511: a^{+}a \rangle_{0} = \sum_{n} n p_{n}$ and $\langle

512: a^{+}a(a^{+}a -1) \rangle_{0} = \sum_{n} n(n-1) p_{n}$, where $p_{n}$

513: is defined in eq.~(\ref{pn}).  Moreover, all

514: cumulants apart from the first one 

515: vanish for forcings with $z=0$. This holds true in particular for all quasi-static

516: force characteristics. The underlying work probability density then

517: shrinks to a delta function at $w=|f(t_{f})|^{2}/(\hbar \o)$. 

518: 

519: In general, the work probability density follows from the characteristic

520: function by means of an inverse Fourier transformation. Rather than

521: the characteristic function itself we first consider the function

522: $\mathcal{G}(u)\equiv \exp \left \{ -iu|f(t_{f})|^{2}/(\hbar \o)\right\}

523: G_{t_{f},t_{0}}(u)$. Upon expanding $\exp\left \{|z|^{2} \exp\left

524:   \{iu\hbar \o\right \} \right \}$

525: into a series of powers of $|z|^{2}$ we obtain for $\mathcal{G}(u)$ a

526: Laurent series in the variable $\exp\left\{iu\hbar \o\right\}$. The

527: inverse Fourier transformation is given by a series of delta

528: functions $\d(w-\hbar \o r)$, with $r \in \mathbb{Z}$, with

529: weights

530: \be

531: \begin{split}

532: q_r =& e^{-|z|^2 } \sum_{m,n=0}^\infty \sum_{k=0}^n

533: \sum_{l=0}^{2k}(-1)^{2k-l} p_n  \\

534: &\quad\times \frac{|z|^{2(k+m)}}{m!\: k!}

535: \binom{n}{k} \binom{2k}{l} \d_{l+m,k+r}\\

536: =& e^{-|z|^2}\sum_{n=0}^\infty \sum_{k=0}^n \sum_{l=0}^{\text{min}

537:   (k+r,2k)}

538: (-1)^{2k-l}  p_n  \\

539: &\quad \times \frac{|z|^{2(2k+r-l)}}{(k+r-l)!\:k!} \binom{n}{k}

540: \binom{2k}{l}\;. 

541: \end{split}

542: \ee{q}    

543: The factor  $\exp \left \{ -iu|f(t_{f})|^{2}/(\hbar \o)\right\}$, by

544: which $\mathcal{G}(u)$ has to be multiplied to yield

545: $G_{t_{f},t_{0}}(u)$, gives

546: rise to a constant shift such that the probability density of work

547: performed on a harmonic oscillator assumes the result 

548: \be

549: p_{t_f,0}(w) = \sum_r q_{r} \:\d\!\left (w-(\hbar \o r

550: +\frac{|f(t_{f})|^{2}}{\hbar \o})\right ) \;.

551: \ee{pt0}

552: In the next Section we will investigate the influence of the initial

553: state on the statistics of the work.

554: \subsection{Distributions of work for different initial states}

555: As particular examples of initial states we will discuss

556: microcanonical, canonical and coherent states.\\

557: \subsubsection{Microcanonical initial state}

558: For a microcanonical initial state with energy $\hbar \o n_{0}$ the

559: density matrix becomes

560: \be

561: \r(t_{0}) = \bar{\r}(t_{0}) = |n_{0}\rangle \langle n_{0} |.

562: \ee{mco}

563: The characteristic function then reads

564: \be

565: \begin{split}

566: G^{\text{mc}}_{t_{f},t_{0}}(n_{0},u) &= e^{iu|f(t_{f})|^{2}/(\hbar

567:   \o)} \exp\left 

568:   \{ \left( e^{iu\hbar \o} -1\right)|z|^{2}\right \} \\

569: &\quad \times L_{n_{0}}\!\left ( 4|z|^{2} \sin^{2} \frac{\hbar \o u}{2}\right )

570: \end{split}

571: \ee{mccf0}   

572: and, accordingly, the probability $q^{\text{mc}}_{r}(n_{0})$ to find a change of

573: energy by $w=\hbar \o r +|f(t_{f})|^{2}/(\hbar \o)$ emerges as 

574: \be

575: \begin{split}

576: q^{\text{mc}}_{r}(n_{0})& = e^{-|z|^{2}} \sum_{k=0}^{n_{0}} \sum_{l=0}^{\min

577:   (k+r,2k)}\frac{(-1)^{2k-l}}{(k+r-l)!\: k!}\\

578: &\quad \times \binom{n}{k} \binom{2k}{l}

579: |z|^{2(2k+r-l)}.

580: \end{split}

581: \ee{mcq}

582: As expected from the behavior of the moments, all probabilities

583: $q^{\text{mc}}_{r}(n_{0})$ with $r\neq 0$ vanish for quasi-static forcing, i.e. if

584: $z \to 0$. The dependence of $q^{\text{mc}}_{r}(n_{0})$ for $n_{0}=0$ and $3$ 

585: as well as for the eight lowest values of

586: $r$ on the parameter $z$ is displayed in Fig.~\ref{fz}. With

587: increasing values of the rapidity parameter $z$ the distribution is

588: becoming broader.

589: \begin{figure}

590: \includegraphics[width=8cm]{Fig1-a.eps}\hfill 

591: \includegraphics[width=8cm]{Fig1-b.eps} 

592: \caption{The probabilities $q^{\text{mc}}_{r}(n_{0})$ for two

593: microcanonical initial states with $n_{0}=0$ (a) and $n_{0}=3$ (b) are 

594: depicted for $r=-3

595: \ldots 4$, as functions of the rapidity parameter $z$ in

596: eq.~(\ref{z}).  In both cases the

597: distribution collapses at $r=0$ for adiabatic forcing corresponding to

598: $|z|=0$ and broadens with increasing $|z|$. Obviously, when starting in

599: the ground state the oscillator cannot deliver work whence the

600: probability for negative $r$ strictly vanishes. ``Stimulated

601: emission'' becomes possible from an excited state at finite driving

602: rapidity $z$ leading to nonzero probabilities

603: $q^{\text{mc}}_{r}(n_{0})$ at negative values of $r$ in panel (b).}      

604: \label{fz}

605: \end{figure}

606: For the fixed value of $|z|=2$ the distribution

607: $q^{\text{mc}}_{r}(n_{0})$ is compared for the three initial states  with

608: $n_{0}=0$ , $10$ and $30$ in

609: Fig.~\ref{fn}. 

610: \begin{figure}

611: \includegraphics[width=8cm]{Fig2.eps}

612: \caption{(Color online) The probabilities $q^{\text{mc}}_{r}(n_{0})$ for a

613: microcanonical initial state with $n_{0}=0$ (circles) and $n_{0}=10$

614: (diamonds) and $n_{0}=30$ (crosses) are

615: compared for a fixed rapidity parameter $|z|=2$ and $r=-22 \ldots

616: 30$. The lines serve as a guide for the eye.}

617: \label{fn}

618: \end{figure}

619: With increasing value of $n_{0}$ the distributions

620: become broader. They develop a slightly asymmetric shape with 

621: higher peaks at negative values of $r$ compared to those at positive

622: $r$ values. Between these dominant peaks the probability still

623: displays pronounced variations.\\

624: For a harmonic oscillator, the microcanonical Crooks theorem reduces

625: to the relation $q^{\text{mc}}_{r}(n) = q^{\text{mc}}_{-r}(n+r)$. One can show that

626: this symmetry is fulfilled by the probabilities $q^{\text{mc}}_{r}(n)$

627: given by eq.~(\ref{mcq}).

628: \subsubsection{Canonical initial state}

629: For a canonical density matrix 

630: \be

631: \r(t_{0}) = (1-e^{-\b \hbar \o }) e^{-\b \hbar \o a^{+} a}

632: \ee{cdm}

633:  the initial states are

634: distributed according to $p_{n} = e^{-\b

635:   \hbar \o n}/(1- e^{-\b \hbar \o})$. This allows one to perform the

636: sum over $n$ in the expression for the characteristic function

637: (\ref{Gho}) in

638: terms of the generating function of the Laguerre polynomials, cf. \cite{RG}

639: yielding the expression

640: \begin{widetext}

641: \be

642: G^{\text{c}}_{t_{f},t_{0}}(\b,u) = \exp \left \{ \frac{iu

643:     |f(t_{f})|^{2}}{\hbar \o} +\left(e^{iu\hbar

644:       \o}-1 \right )|z|^{2} - 4|z|^{2} \frac{\sin^{2}(\hbar \o u/2)}{e^{\b \hbar

645:       \o}-1}   \right \}. 

646: \ee{Gca}

647: \end{widetext}

648: Putting $u=i\b$ one finds that the two terms in the

649: exponent which are proportional to $|z|^{2}$ cancel each other, such

650: that one obtains 

651: \be

652: \langle e^{\b w} \rangle = G^{\text{c}}_{t_{f},t_{0}}(\b,i\b) =

653: \exp\left \{-\b

654:   |f(t_{f}|^{2}/(\hbar \o)\right \}. 

655: \ee{Jho}

656:  The free energy difference of two oscillators with Hamiltonians

657:  $H(t_{0}) =\hbar \o a^{+}a$ and  $H(t_{f}) =\hbar \o a^{+}a

658:  +f^{*}(t_{f}) a +f(t) a^{+})$ each one staying in a canonical state at the

659:   temperature $\b$ is given by $\D F = F(t_{f}) - F(t_{0}) =

660:  |f(t_{f})|^{2}/(\hbar \o)$ in accordance with Jarzynski's work theorem.

661: 

662: The probability $q^{\text{c}}_{r}(\tilde{\b})$ to find the work $w= \hbar \o r

663: +|f(t_{f})|^{2}/(\hbar \o)$ if the system starts in a canonical state becomes

664: \be

665: \begin{split}

666: q^{\text{c}}_{r}(\tilde{\b})& = e^{-|z|^{2}} \left (1 - e^{-\tilde{\b}

667:   } \right )\sum_{n=0}^{\infty} \sum_{k=0}^{n} \sum_{l=0}^{\min

668:   (k+r,2k)} 

669:   (-1)^{l} e^{-\tilde{\b}  n} \\

670: &\quad \times \frac{|z|^{2(2k+r-l)}}{(k+r-l)!

671:     k!} \binom{n}{k}

672: \binom{2k}{l}

673: \end{split}

674: \ee{qc}

675: where $\tilde{\b} = \b \hbar \o$ denotes the inverse dimensionless

676: temperature of the initial state.

677: The expression for $q^{c}_{r}(\tilde{\b})$ can be further simplified to read

678: \be

679: q^{\text{c}}_{r}(\tilde{\b})= e^{-|z|^{2} \coth

680:   (\tilde{\b}/2)} e^{\tilde{\b}r/2} I_{r}\left (\frac{|z|^{2}}{\sinh

681:     \tilde{\b}/2} \right )

682: %\sum_{m=0}^{\infty}  

683: %    \frac{|z|^{2m}}{m!}  I_{|m-r|}\left ( \frac{2|z|^{2}}{e^{\tilde{\b}

684: %       }-1}\right )

685: \ee{qc2}

686: where $I_{\n}(x)$ denotes the modified Bessel function of first kind of

687: order $\n$ \cite{RG}. For details of the derivation see Appendix~\ref{C}.

688: 

689: Note that the following detailed balance like symmetry relation exists,

690: \be

691: q^{\text{c}}_{-r}(\tilde{\b}) = e^{-\tilde{\b}r} q^{\text{c}}_{r}(\tilde{\b}),

692: \ee{qrmr}

693: relating the occurence of 

694: positive and negative work.   

695: In Fig.~\ref{fcz} the $z$ dependence of $q^{c}_{r}(\tilde{\b})$ for

696: $\tilde{\b}=\ln(4/3)$ is compared for a few small values of $r$. One finds

697: that due to the average over the canonical initial distribution 

698: the multipeaked structure of the microcanonical distribution as a

699: function of the rapidity parameter $|z|$ disappears and only

700: a single peak remains for each value of $r$.

701: The temperature dependence of the work distribution is illustrated

702: in Fig.~\ref{fcb}.

703: \begin{figure}

704: \includegraphics[width=8cm]{Fig3.eps}

705: \caption{The probabilities $q^{\text{c}}_{r}(\tilde{\b}=\ln(4/3))$ for a

706: canonical initial state are displayed for $r=-3

707: \ldots 4$ as a function of the parameter $z$. For the sake of

708: comparability the dimensionless

709: inverse temperature is chosen such that the average energy in the

710: initial state coincides with the energy $3 \hbar \omega$ 

711: of the microcanonical state in Fig.~\ref{fz} (b).   }

712: \label{fcz}

713: \end{figure}

714: \begin{figure}

715: \includegraphics[width=8cm]{Fig4.eps}

716: \caption{(Color online) The probabilities $q^{\text{c}}_{r}(\tilde{\b})$ for a

717: canonical initial state are displayed as functions of  $r$ 

718: for  $|z|=2$ and different values of the dimensionless inverse

719: temperature $\tilde{\b} = 0.5$ (boxes), $1$ (circles) and $5$

720: (crosses). The lines serve as a guide for the eye.}  

721: \label{fcb}

722: \end{figure}

723: Finally, we verify the validity of the Tasaki-Crooks theorem (\ref{TC})

724: for a driven oscillator. For this purpose we

725: consider the probability density $p_{t_{0},t_{f}}(-w)$ for the time

726: reversed protocol. Since the absolute values of the rapidity parameters

727: coincide for the original and the time reversed protocol the

728: probability density of work for the reversed protocols becomes

729: \be

730: p_{t_{0},t_{f}}(-w) = \sum_{r=-\infty}^{\infty} q_{r} \d\left (-w-(\hbar \o r

731: -\frac{|f(t_{f})|^{2}}{\hbar \o}) \right)\; ,

732: \ee{p0f}

733: where we took into account the overall shift of the spectrum by the

734: reversed protocol as

735: $-|f(t_{f})|/(\hbar \o)$.

736: Multiplying both sides of eq.~(\ref{p0f}) with $\exp \left \{ -\b( \D

737:   F -w) \right \}=\exp \left \{ -\b( |f(t_{f})|^{2}/(\hbar \o) -w)

738: \right \} $  one obtains 

739: \be

740: \begin{split}

741: e^{  -\b( \D

742:   F -w) } p_{t_{0},t_{f}}(-w) &= \sum_{r} e^{ -\b

743:   \left (

744:   |f(t_{f})|^{2}/(\hbar \o)  -w \right )}

745: q^{\text{c}}_{r}(\tilde{\b}) \\

746: &\quad \times \d\left (-w-(\hbar \o r 

747: -\frac{|f(t_{f})|^{2}}{\hbar \o}) \right)\\

748: &=\sum_{r} e^{\tilde{\b}r} q^{c}_{r}(\b)\\

749: &\quad \times \d \left (w +(\hbar

750:   \o r -\frac{|f(t_{f})|^{2}}{\hbar \o}) \right )\\

751: &= p_{t_{f},t_{0}}(w),

752: \end{split}

753: \ee{TCho}

754: in accordance with the Tasaki-Crooks theorem (\ref{TC}).     

755:  

756: 

757: 

758: \subsubsection{Coherent initial state}

759: An oscillator prepared in a coherent state $|\a \rangle$ is described

760: by the density matrix

761: \be

762: \r(t_{0}) = |\alpha \rangle \langle \alpha |

763: \ee{rcs}

764: where 

765: \be

766: |\a \rangle = e^{\a a^{+} + \a a} |0 \rangle

767: \ee{acs}

768: and $|0 \rangle $ is the normalized ground state of the oscillator

769: satisfying $a |0 \rangle =0$.

770: Note that the coherent state density matrix does {\it not} commute with the

771: Hamiltonian $H(t_{0})$. The measurement of $H(t_{0})$ modifies the

772: coherent state (\ref{rcs})  by

773: projecting it onto the eigenstates $|n \rangle = (a^{+})^{n}

774: /\sqrt{n!} |0 \rangle$ of this Hamiltonian leading to 

775: \be

776: \bar{\r}(t_{0}) = e^{-|\a|^{2}} \sum_{n} \frac{|\a|^{2n}}{n!} |n

777: \rangle \langle n |

778: \ee{brcs}

779: This implies a Poissonian distribution of the respective  

780: energy eigenvalues $\hbar \o n$ 

781: \be

782: p^{\text{cs}}_{n} = \frac{|\a|^{2n}}{n!} e^{-|\a|^{2}}, 

783: \ee{pcs}

784: which yields for the characteristic function of work (\ref{Gho}) a

785: closed expression of the form

786: \be

787: \begin{split}

788: G^{\text{cs}}_{t_{f},t_{0}}(\a,u) &= \exp\left

789:   \{\frac{iu|f(t_{f})|^{2}}{\hbar \o} +|z|^{2} \left(e^{i\hbar \o u}-1

790:   \right ) \right \}\\ 

791: & \quad \times J_{0}\left (4\left |\a z \sin 

792:   \frac{\hbar \o u}{2} \right | \right ) 

793: \end{split}

794: \ee{Gcs}  

795: where $J_{0}(x)$ is the Bessel function of order zero, cf. Ref.~\cite{RG}.

796: For the probability $q^{\text{cs}}_{r}(\a)$ of work one obtains with

797: eq.~(\ref{q})

798: \begin{widetext}

799: \be

800: q^{\text{cs}}_{r}(\a) = \;e^{-|z|^{2}} \sum_{m=0}^{\infty} 

801:   \frac{|z|^{2m} \:|\a z|^{2|m-r|}}{m! \left( |m-r|!\right)^{2}}

802:   {}_{1}F_{2}

803:   \left ( |m-r|+\frac{1}{2};|m-r|+1,2|m-r|+1;-4|\a z|^{2}\right )

804: \ee{qcs}

805: \end{widetext}

806: where $_{1}F_{2}(a;b,x;x)$ denotes a generalized hypergeometric

807: function \cite{RG}. For details of the derivation see the Appendix~\ref{D}. 

808:  

809: The dependence of  the probabilities $q^{\text{cs}}_{r}(\a)$ on the

810: rapidity parameter $|z|$ is illustrated in Fig.~\ref{fcsz} for $r$

811: values ranging from $-10$ to $20$. 

812: \begin{figure}

813: \includegraphics[width=8cm]{Fig5.eps}

814: \caption{The probabilities $q^{\text{cs}}_{r}(\a)$ for a

815: coherent state with parameter $|\a|=3$ are displayed 

816: for $r = -10 \ldots 20$ as functions of  $z$.}  

817: \label{fcsz}

818: \end{figure}

819: \begin{figure}

820: \includegraphics[width=8cm]{Fig7.eps}

821: \hfill

822: \includegraphics[width=8cm]{Fig6.eps}

823: \caption{(Color online) The distribution of work performed on an

824:   oscillator which 

825:   initially is prepared in a coherent state $|\a \rangle$ for

826:   different values of $\a$ in panel (a) and of the rapidity parameter

827:   $z$ in panel (b).

828: In panel (a) the rapidity parameter has the value $|z|=2$. In panel

829: (b) the coherent state parameter has the value  

830:  $|\a|^{2} =1$. }

831: \label{f2}

832: \end{figure}

833: Increasing values of $z$ lead to a broadening of the distribution

834: and also to a shift towards larger values of $r$, see also panel (b)

835: of Fig.~\ref{f2}.

836: This is in accordance with eq. (\ref{aw}) and (\ref{ww})

837: for the first two cumulants of the work, which both increase with

838: $|z|^{2}$.

839: %\begin{figure}

840: %\includegraphics[width=8cm]{Fig7.eps}

841: %\caption{The distribution of work performed on a harmonic oscillator

842: %  which is

843: %  initially prepared in a coherent state $|\a\rangle$, is displayed for

844: %  different

845: %values of the mean excitation $|\a|^2$. Different phases of $\a$ do

846: %not matter. The force protocols of the three cases are characterized

847: %by the parameter $|z| = 1$, see eqs. (\ref{FG}) and (\ref{z}). Details

848: %of the protocol which leave this parameter unchanged have no influence

849: %on the weight factors $q_n$.

850: %In accordance with eq. (\ref{ww}) the distribution becomes broader for

851: %larger values of $\langle a^{+}a \rangle_{0} =|\a|^2$.}

852: %\label{f1}

853: %\end{figure}

854: Panel (a) of Fig.~\ref{f2} shows the dependence of the probabilities

855: $q^{\text{cs}}_{r}(\a)$ on the parameters $\a$. Increasing

856: $\a$ also leads to a broadening of the work

857: distribution without influencing its mean value, cf. also eq. (\ref{aw}). 

858: 

859: %In Fig.~\ref{f3} the work distributions are compared for a canonical

860: %and a coherent initial state both having the same first two

861: %cumulants. Yet the two distributions are quite different: The work

862: %distribution for the canonical initial state is unimodal 

863: %resembling a Gaussian distribution whereas the coherent state leads to

864: %a bimodal slightly asymmetric distribution. 

865:  

866:   

867: \begin{figure}

868: \includegraphics[width=8cm]{Fig8.eps}

869: \caption{(color online) The distribution of work is compared for a

870:   canonical and a  

871:   coherent initial state subject to the same force protocol with

872:   rapidity parameter $z=2$. With $\b \hbar \o  =0.1$ and  $|\a|^2\approx

873:   9.51$ the expectation values of the energies agree in the two initial

874:   states such that according to eqs.~(\ref{aw}) and (\ref{ww}) the

875:   first and second moments of the work also coincide. Still the

876:   distributions of work grossly differ from each other. }

877: \label{f3}

878: \end{figure}  

879: In Fig.~\ref{f3} the probabilities $q_r$ are depicted for different

880: initial states. In one case the oscillator is initially prepared in a

881: canonical state at inverse dimensionless temperature $\tilde{\b}=\b

882: \hbar \o =0.1$. 

883: In the other case, the oscillator stays in a coherent state

884: $|\a\rangle$, where the absolute value of $|\a|$ is chosen such that

885: the mean excitation number is the same for both states, i.e. $|\a|^2 =

886: \exp\{ -\b \hbar \o\} / (1- \exp\{ -\b \hbar \o\} )$. For $\b \hbar \o

887: =0.1$ one finds $|\a|^2 \approx 9.51$. The two

888: oscillators then are subjected to protocols with the same

889: rapidity parameter 

890: $|z|=2$. According to eqs. (\ref{aw}) and (\ref{ww}) the first two

891: moments of the work performed on the oscillator coincide. Yet the

892: distribution of weight factors $q^{\text{c}}_r(\tilde{\b})$ and

893: $q^{\text{cs}}_r(\a)$ distinctly differ. Whereas the

894: distribution is pronouncedly bimodal in case of the coherent state, it

895: is unimodal for the canonical state. The weight factors

896: $q^{\text{c}}_r(\tilde{\b})$ almost 

897: perfectly fall onto the Gaussian probability density which has the

898: same first two

899: moments as the discrete distribution given by $q_r$.   

900: 

901: 

902: 

903:  

904: 

905: \section{Conclusions}

906: In this work we studied the statistics of work performed on an

907: externally driven quantum mechanical oscillator by means of a

908: correlation function expression of the work. We demonstrated that this

909: particular expression indeed always represents a proper characteristic

910: function of a random variable, which is the performed work in the present

911: context. The proof given here is based on Bochner's theorem. It holds

912: for general 

913: quantum mechanical systems, not only for harmonic oscillators.

914: 

915: The considered force linearly couples to the position and momentum of

916: the oscillator. It may describe the influence of an electric

917: field on charged particles in a parabolic trap or the external forcing

918: of a single electromagnetic cavity mode. For this

919: type of additive forcing, the frequency of the oscillator remains unchanged and

920: therefore the level spacing of the eigenvalues of the Hamiltonian is

921: not influenced by the force. The spectrum is only shifted as a whole.

922: As a consequence the work performed on the oscillator is, as a positive

923: or negative integer of the level spacing, 

924: a discrete random variable. We determined the first few cumulents of

925: the work for arbitrary force protocols and initial states. A

926: complementary study for a parametrically forced oscillator was

927: recently 

928: performed by Deffner and Lutz \cite{DL}.     

929: 

930: It turns out that for the harmonic oscillator 

931: the statistics of work depends on the force

932: protocol $\{f(t)\}_{t_{f},t_{0}}$ only through two real parameters, 

933: which are (i) the shift of the spectrum, given by $L(t_{f})=

934: |f(t_{f})|^{2}/(\hbar \o)$, 

935: and (ii) the absolute value of the dimensionless quantity $z =

936: \i_{t_{0}}^{t_{f}}\dot{f}(s) \exp \{i \o s \}$. 

937: This parameter vanishes for all quasi-static processes

938: and therefore presents a measure of the rapidity of the force

939: protocol.

940: While the presence of $L(t_{f})$ only causes an overall shift of the

941: possible values of the work, the rapidity parameter $|z|$ also

942: influences its distribution. Typically, the distributions move towards 

943: larger values of work $w$ and become broader with increasing rapidity $|z|$.

944: 

945: We also demonstrated that different initial states of the system such

946: as microcanonical, canonical or coherent states, have a large influence on

947: the work statistics. We further note that two different initial

948: density matrices with the 

949: same diagonal elements with respect to the energy eigenbasis of the

950: Hamiltonian $H(t_{0})$ lead to identical work distributions

951: even though the two density matrices may be very different in other

952: respects. For example, the coherent pure state $|\a\rangle \langle \a|$

953: and the mixed state $\exp\{-|\a|^{2 n}\}\sum_{n} |\a|^{2}/n! |n\rangle

954: \langle n| $ cannot be distinguished by means of their respective work

955: statistics. This statistics is also insensitive to the phase of a

956: coherent state.  

957: 

958: \acknowledgments

959: This work has been supported by the Deutsche Forschungsgemeinschaft

960: via the Collaborative Research Centre SFB-486, project A10. Financial

961: support of the German Excellence Initiative via the {\it Nanosystems

962:   Initiative Munich} (NIM) is gratefulle acknowledged as well. 

963: 

964: 

965: 

966: \appendix

967: \section{Proof of the properties of $G_{t_f,t_0}(u,v)$}

968: \label{AA}

969: We prove that the conditions of Bochner's theorem are fulfilled, and

970: consequently $G_{t_{f},t_{0}}(u)$ is a proper characteristic function.

971:   

972: Proof of property (i): $G_{t_{f},t_{0}}(u)$ is a continuous function

973: of $u$. The Hamiltonian operators at the two times of measurement

974: $t_{0}$ and $t_{f}$ are selfadjoint operators. According to the

975: theorem of Stone \cite{Yosida}, each of the exponential operators $\exp \left

976:   \{ -iu H(t_{0})\right \}$ and  $\exp \left

977:   \{ iu H_{H}(t_{f})\right \}$ forms a strongly continuous one-parameter 

978: group of unitary operators with parameter $u$. As the trace of a

979: product of two strongly continuous operator valued functions of $u$

980: with the density operator $\bar{\r}(t_{0})$, which is a trace class

981: operator and independent of $u$, the

982: characteristic function (\ref{GC}) is a continuous function of $u$.

983: 

984: Proof of property (ii): $G_{t_{f},t_{0}}(u)$ is a positive definite

985: function of $u$.

986: Using the cyclic invariance of the trace and the  fact that $H(t_{0})$

987: and $\bar{\r}(t_{0})$ commute with each other, we can rewrite the left

988: hand side of the inequality (\ref{pd}) as

989: \begin{widetext}

990: \be

991: %\begin{split}

992: \sum_{i,j}^{n} G_{t_{f},t_{0}}(u_{i}-u_{j}) z^{*}_{i} z_{j} =

993: \sum_{i,j}^{n} \Tr \:e^{i(u_{i}-u_{j})H_{H}(t_{f})} \:

994: e^{-i(u_{i}-u_{j})H(t_{0})} \bar{\r}(t_{0}) z_{i}^{*} z_{j}

995: = \Tr A^{+} A \:\bar{\r}(t_{0}) \geq 0\;,  

996: %\end{split}

997: \ee{pos}   

998: \end{widetext}

999: where 

1000: \be

1001: A= \sum_{i}z_{i}^{n} e^{-iu_{i}H_{H}(t_{f})} e^{iu_{i} H(t_{0})}

1002: \ee{A}

1003: is a bounded operator and $A^{+}$ its adjoint. The last inequality in 

1004: (\ref{pos}) immediately 

1005: follows with the positivity of $A^{+}A$ and of the density matrix

1006: $\bar{\r}(t_{0})$.   

1007: 

1008: Proof of property (iii): $G_{t_{f},t_{0}}(0) =1$. 

1009: For $u=0$ the exponential operators $\exp \left

1010:   \{ -iu H(t_{0})\right \}$ and  $\exp \left

1011:   \{ iu H_{H}(t_{f})\right \}$ become unity. The trace over the

1012: density matrix $\bar{\r}(t_{0})$ reduces by means of

1013: eqs. (\ref{br}), (\ref{P1}) to the trace of the initial

1014: density matrix $\r(t_{0})$, which is one.

1015: \section{The matrix element $\langle n| \exp \{iu H_H(t_f)\} |n\rangle

1016:   $}

1017: \label{B} 

1018: The total time rate of change of the Hamiltonian $H_H(t)$ coincides

1019: with its partial derivative with respect to the time which for the

1020: driven oscillator becomes, cf. eq. (\ref{hoH}),

1021: \be

1022: \frac{d H_H(t)}{dt} = \dot{f}^* (t) a_H(t) - \dot{f}(t) a^+_H(t),

1023: \ee{dH}

1024: where $a_H(t)$ and $a^+_H(t)$ denote annihilation and creation

1025: operators, respectively, in the Heisenberg picture, which are given by

1026: \be

1027: \begin{split}

1028: a_H(t) &=   e^{-i \o t} a -\frac{i}{\hbar} \i_0^t ds e^{-i\o(t-s)} f(s)\\

1029: a^+_H(t) &=  e^{i \o t} a^+ +\frac{i}{\hbar} \i_0^t ds e^{i\o(t-s)} f^*(s)  

1030: \end{split}

1031: \ee{aH}

1032: This yields for $H_H(t_f)$

1033: \be

1034: H_H(t_f) = \hbar \o a^+ a + B^*(t_{f})a + B(t_{f}) a^+ + C(t),

1035: \ee{HtH}

1036: where

1037: \be

1038: \begin{split}

1039: B(t_{f}) &= \i_0^{t_{f}} ds \dot{f}(s) e^{i \o s}\\

1040: C(t_{f})& = \frac{i}{\hbar} \i_0^{t_{f}} ds \i_0^s ds' \left [ \dot{f}(s)

1041:   f^*(s') e^{i\o(s-s')} \right .\\

1042: &\quad \left . - \dot{f}^*(s) f(s') e^{-i\o (s-s')} \right ].

1043: \end{split}

1044: \ee{FG}

1045: The unitary operator

1046: \be

1047: V= e^{z a^+ -z^* a}

1048: \ee{V}

1049: with 

1050: \be

1051: z = \frac{B(t_{f})}{\hbar \o}

1052: \ee{z}

1053: transforms $H_H(t)$ into

1054: \be

1055: V H_H(t_{f}) V^+ = \hbar \o a^+ a + L(t_{f}),

1056: \ee{VHV}

1057: where 

1058: \be

1059: L(t_{f}) = C(t_{f})-\frac{|B(t_{f})|^2}{\hbar \o}

1060: =\frac{|f(t_{f})|^2}{\hbar \o} .

1061: \ee{L}

1062: Note that $V$ induces a shift of the creation and annihilation

1063: operators

1064: \be

1065: \begin{split}

1066: V a V^+ = a - z, \quad

1067: V a^+ V^+ = a^+ - z^*

1068: \end{split}

1069: \ee{Va}

1070: and further note that, when acting on the groundstate $|0\rangle$ with

1071: $a|0\rangle =0$, the operator $V$ yields the coherent state $|z\rangle$, i.e.

1072: \be

1073: V|0\rangle = | z \rangle.

1074: \ee{V0}

1075: One finds with these properties

1076: %\begin{widetext}

1077: %\be

1078: %\begin{split}

1079: \begin{align}{\allowdisplaybreaks}

1080: \langle n | e^{iu H_H(t_{f})}|n\rangle &= \frac{1}{n!} \langle

1081: z|(a-z)^n \nonumber \allowdisplaybreaks\\

1082: &\quad\times  e^{iu\hbar \o a^+a +i u L(t_{f})} (a^+-z^*)^n |

1083: z \rangle \nonumber \allowdisplaybreaks \\

1084: &=\frac{1}{n!} e^{iu L(t_{f})} \frac{\partial^{2n}}{\partial x^n \partial

1085:   y^n} \nonumber \allowdisplaybreaks \\

1086: &\quad \langle z| e^{x(a-z)} e^{iu \hbar \o a^+a} e^{y(a^+-z^*)} | z

1087: \rangle|_{x=y=0}

1088: \label{nz}

1089: \end{align}

1090: %\end{split}

1091: %\ee{nz}

1092: %\end{widetext}

1093: Here we have introduced the auxiliary variables $x$ and $y$ which allow

1094: to represent the $n$th powers of shifted creation and annihilation

1095: operators by derivatives of respective order. The scalar

1096: function $e^{-i(xz + y z^*)}$ can be taken out of the scalar product

1097: and the remaining operator can be brought into normal order. It then

1098: becomes \cite{Wi}

1099: %\begin{widetext}

1100: \be

1101: \begin{split}

1102: e^{xa} e^{iu \hbar \o a^+ a} e^{ya^+} &= \mathcal{N}\left \{

1103:   \exp\left [(e^{iu\hbar \o} -1) a^+a \right . \right .\\

1104: &\quad \left . \left . + e^{iu\hbar\o} \left (x a + y

1105:       a^+ + x y \right ) \right ] \right \}\; ,  

1106: \end{split}

1107: \ee{N} 

1108: %\end{widetext}

1109: where  under the normal ordering operator $\mathcal{N}$ all

1110: creation operators stand left of the annihilation operators. The

1111: matrix element with respect to the coherent state $|z\rangle$ can be

1112: read off, yielding,

1113: \begin{widetext}

1114: \be

1115: \begin{split}

1116: \langle n | e^{iu H_H(t_{f})} |n\rangle =&\: \frac{1}{n!}  e^{iu L(t_{f})} \exp

1117: \left \{ \left ( e^{iu \hbar \o}-1 \right ) |z|^2 \right \}

1118: \frac{\partial^{2n}}{\partial x^n \partial y^n} \exp \left \{\left 

1119:     (e^{iu \hbar \o} -1 \right )(xz +y z^*) +e^{iu\hbar \o} xy \right

1120: \}|_{x=y=0}\\

1121: =&\:\frac{1}{n!}  e^{iu L(t_{f})} \exp

1122: \left \{ \left ( e^{iu \hbar \o}-1 \right ) |z|^2 \right \} 

1123: \frac{\partial^n}{\partial y^n} \left [ \left (e^{iu\hbar \o} -1

1124:   \right )z + e^{iu \h \o} y \right ]^n \exp \left \{ \left ( e^{iu\hbar

1125:       \o}-1 \right ) \left ( |z|^2 + y z^* \right ) \right \}|_{y=0}\\

1126: =& \: e^{iu |f(t_{f})|^{2}/(\hbar \o)} \exp

1127: \left \{ \left ( e^{iu \hbar \o} -1 \right ) |z|^2 \right \} 

1128: \sum_{k=0}^n \binom{n}{k} \frac{|z|^{2(n-k)}}{(n-k)!}

1129: e^{iu\hbar \o k} \left ( e^{iu\hbar \o} -1 \right )^{2(n-k)} \; .

1130: \end{split}

1131: \ee{nxy}

1132: \end{widetext}

1133: \section{Work distribution for a canonical initial state}

1134: \label{C}

1135: To determine the expression (\ref{qc2}) for the work

1136: distribution $q^{\text{c}}(\tilde{\b})$ we start from 

1137: the general expression given in the first line of 

1138: eq. (\ref{q}). Interchanging 

1139: the summation over the indices $n$ and $k$ we obtain

1140: %\begin{widetext}

1141: %\be

1142: %\begin{split}

1143: \begin{align}

1144: q^{c}_{r}(\tilde{\b})& = e^{-|z|^{2}}\sum_{m,k=0}^{\infty} \sum_{l=0}^{2k}

1145: (-1)^{l}\frac{|z|^{2(k+m)}}{m!\:k!} \binom{2k}{l}\allowdisplaybreaks

1146: \nonumber \\

1147: &\quad \times \d_{l+m,k+r}

1148: \sum_{n=k}^{\infty} \frac{e^{-\tilde{\b} n}}{1-e^{-\tilde{\b} }}

1149: \binom{n}{k} \allowdisplaybreaks\nonumber \\

1150: &\stackrel{(1)}{=}e^{-|z|^{2}} \sum_{m,k=0}^{\infty} \sum_{l=0}^{2k}

1151: (-1)^{l}\frac{|z|^{2(k+m)}}{m!\:k!} \binom{2k}{l}\nonumber \allowdisplaybreaks\\

1152: &\quad \times  \left

1153:   ( \frac{1}{e^{\tilde{\b} }-1}\right )^{k}

1154: \d_{l+m,k+r}\allowdisplaybreaks \nonumber \\

1155: &\stackrel{(2)}{=}(-1)^{r}e^{-|z|^{2}}

1156: \sum_{m=0}^{\infty}\frac{(-|z|^{2})^{m}}{m!}\nonumber \allowdisplaybreaks\\

1157: &\quad \times \sum_{k=|m-r|}^{\infty}  

1158: \frac{\left (-|z|^{2}/(e^{\tilde{\b}}-1) \right )^{k}}{k!}

1159: \binom{2k}{k+r-m}\allowdisplaybreaks\nonumber \\

1160: &\stackrel{(3)}{=}(-1)^{r}e^{-|z|^{2}}

1161: \sum_{m=0}^{\infty}\frac{(-|z|^{2})^{m}}{m!}

1162: e^{-2|z|^{2}/(e^{\tilde{\b}}-1)} \nonumber \allowdisplaybreaks\\

1163: &\quad \times I_{|m-r|}\left (- \frac{2

1164:     |z|^{2}}{e^{\tilde{\b}}-1}\right )\nonumber \allowdisplaybreaks\\ 

1165: & \stackrel{(4)}{=}e^{-|z|^{2} \coth (\tilde{\b}/2)}

1166: \sum_{m=0}^{\infty} \frac{|z|^{2m}}{m!} I_{|m-r|} \left (\frac{2

1167:   |z|^{2}}{e^{\tilde{\b}}-1} \right )\nonumber \allowdisplaybreaks\\

1168: & \stackrel{(5)}{=} e^{-|z|^{2} \coth (\tilde{\b}/2)} \left \{ 

1169:  \sum_{m=0}^{\infty} \frac{|z|^{2}}{m!} I_{r-m} \left (\frac{2

1170:   |z|^{2}}{e^{\tilde{\b}}-1} \right ) \right .\nonumber\allowdisplaybreaks\\

1171: &\quad \left . +\sum_{m=r+1}^{\infty} 

1172: \frac{|z|^{2}}{m!}\left [I_{m-r} \left (\frac{2

1173:   |z|^{2}}{e^{\tilde{\b}}-1} \right ) \right . \right

1174: .\nonumber\allowdisplaybreaks\\ 

1175: &\left . \left .  \quad -I_{r-m} \left (\frac{2

1176:   |z|^{2}}{e^{\tilde{\b}}-1} \right ) \right] \right \}

1177: \allowdisplaybreaks\nonumber\\ 

1178: & %\stackrel{(6)}{=}  

1179: = e^{-|z|^{2} \coth (\tilde{\b}/2)}

1180: e^{\tilde{\b}r/2} I_{r}\left(\frac{|z|^{2}}{\sinh (\tilde{\b}/2)} \right).

1181: \label{qcp}

1182: \end{align}

1183: %\end{split}

1184: %\ee{qcp}   

1185: %\be

1186: %q \stackrel{(5)}{=} e^{-|z|^{2} \coth (\tilde{\b}/2)} \left \{ 

1187: %\sum_{m=0}^{\infty} \frac{|z|^{2}}{m!} I_{r-m} \left (\frac{2

1188: %  |z|^{2}}{e^{\tilde{\b}}-1} \right ) + %e^{-| q| () /2} 

1189: %\sum_{m=r+1}^{\infty} 

1190: %\frac{|z|^{2}}{m!}\left [I_{m-r} \left (\frac{2

1191: %  |z|^{2}}{e^{\tilde{\b}}-1} \right )-I_{r-m} \left (\frac{2

1192: %  |z|^{2}}{e^{\tilde{\b}}-1} \right ) \right] \right \} 

1193: %\ee{test}

1194: %\end{widetext} 

1195: In the first step ($\stackrel{(1)}{=}$) we  performed the sum on $n$

1196: according to 

1197: \be

1198: \sum_{n=k}^{\infty} \frac{x^{k}}{1-x} \binom{n}{k} = \left (

1199:   \frac{x}{1-x}\right)^{k},

1200: \ee{s1} 

1201: cf. Ref.~\cite{PBM}, 5.2.11.3. In the second step $\stackrel{(2)}{=}$

1202: the Kronecker delta is used 

1203: to perform the sum over k. The third step $\stackrel{(3)}{=}$ is based

1204: on the relation 

1205: \be

1206: \sum_{k=|l|}^{\infty} \frac{x^{k}}{k!} \binom{2k}{k+l} = e^{2x}

1207: I_{|l|}(2x)

1208: \ee{s2}

1209: valid for integer $l$. Here $I_{\n}(x)$ denotes the modified Bessel

1210: function of the first kind of order $\n$. 

1211: With $I_{\n}(-x) =(-1)^{\n}I_{\n}(x)$ where

1212: $\n$ is an integer, we come to the right hand side of the equality 

1213: $\stackrel{(4)}{=}$. In the next step the sum on $m$ is rewritten. The

1214: term in the square brackets vanishes because $I_{\n}(x)$ is an even

1215: function of the order $\n$. The remaining sum can be performed by

1216: means of the identity

1217: \be

1218: \sum_{k=0}^{\infty} \frac{t^{k}}{k!} I_{\n -k} (x) = \left ( \frac{2

1219:     t}{x} +1\right )^{\n/2} I_{\n} \left( \sqrt{x^{2}+2 t x}\right )\; ,

1220: \ee{sB}

1221: cf. \cite{PBM} 5.8.3.1.

1222: This leads to the final 

1223: result given in eq. (\ref{qc2}).

1224: \section{Work distribution for a coherent initial state eq. (\ref{qcs})}

1225: \label{D}

1226: Starting from eq.~(\ref{q}) we may proceed in an analogous way as 

1227: in the case of a canonical

1228: initial state, cf. the Appendix~\ref{C}. 

1229: According to eq. (\ref{pcs}) 

1230: a Poissonian average over the binomial

1231: $\binom{n}{k}$ has to be performed

1232: instead of the geometric average in the first step of eq. (\ref{qcp}).

1233: This yields

1234: \be

1235: \sum_{n=k}^{\infty} \frac{|\a|^{2}}{k!} e^{-|\a|^{2}} \binom{n}{k}

1236: = \frac{|\a|^{2k}}{k!} \; .

1237: \ee{qcs1}

1238: Next the Kronnecker delta is used to perform the sum over $l$ leaving

1239: one with two  sums of which the inner one over k can be expressed in

1240: terms of a generalized hypergeometric function, \cite{RG}, to become

1241: \begin{widetext}

1242: \be

1243: \sum_{k=|m-r|}^{\infty} \frac{(-|\a z|^{2})^{k}}{(k!)^{2}}

1244: \binom{2k}{k+r-m} = \frac{(-|\a z|^{2})^{|m-r|}}{(|m-r|!)^{2}}\\

1245: {}_1F_{2} \left( |m-r|+\frac{1}{2};|m-r|+1,2|m-r|+1;-4|\a

1246:   z|^{2}\right) \; .

1247: \ee{qcs2}

1248: \end{widetext}

1249: This immediately leads to the expression in eq. (\ref{qcs}).

1250: \begin{thebibliography}{99}

1251: \bibitem{fwt} G.N. Bochkov, Yu.E. Kuzovlev, Sov. Phys. JETP {\bf 45},

1252:   125 (1977).

1253: \bibitem{ECM}

1254: D.J. Evans, E.G.D. Cohen, G.P. Morriss,

1255:   Phys. Rev. Lett. {\bf 71}, 2401 (1993).

1256: \bibitem{J} C. Jarzynski, Phys. Rev. Lett. {\bf 78}, 2690 (1997).

1257: \bibitem{J07} C. Jarzynsky, C. R. Physique {\bf 8}, 495 (2007).

1258: \bibitem{ex} F. Douarche, S. Ciliberto, A. Petrosyan, I. Rabbiosi,

1259:   Europhys. Lett. {\bf 70}, 593 (2005).

1260: \bibitem{BLR}

1261:  C. Bustamante, J. Liphardt,

1262:   F. Ritort, Physics Today {\bf 58} (7), 43 (2005).

1263: \bibitem{BHSB}

1264: V. Blickle,

1265:   T. Speck, L. Helden, U. Seifert, C. Bechinger, Phys. Rev. Lett. {\bf

1266:   96}, 070603 (2006).

1267: \bibitem{qm} H. Tasaki, cond-mat/0009244.

1268: \bibitem{M}

1269: S. Mukamel,

1270:   Phys. Rev. Lett. {\bf 90}, 170604 (2003). 

1271: \bibitem{RM}

1272: W. De Roeck, C. Maes,

1273:   Phys. Rev. E {\bf 69}, 026115 (2004).

1274: \bibitem{EM} M. Esposito, S. Mukamel, Phys. Rev. E {\bf 73}, 046129 (2006).

1275: \bibitem{TLH} P. Talkner, E. Lutz, P. H\"anggi, Phys. Rev. E {\bf 75},

1276:   050102(R) (2007).

1277: \bibitem{TH} P. Talkner, P. H\"anggi, J. Phys. A {\bf 40}, F569 (2007).

1278: \bibitem{TMH} P. Talkner, M. Morillo, P. H\"anggi, arXiv:0707.2307v2.

1279: \bibitem{DL} S. Deffner, E. Lutz, Phys. Rev. E {\bf 77}, 021128 (2008).

1280: \bibitem{Seif} U. Seifert, J. Phys. A: Math. Gen. {\bf 37}, L517 (2004).

1281: \bibitem{H} K. Husimi, Prog. Theor. Phys. {\bf 9}, 381 (1953).

1282: \bibitem{Boch} E. Lukacs, {\it Characteristic Functions}, Griffin,

1283:   London, 1970.

1284: \bibitem{CBK} B. Cleuren, C. Van den Broeck, R. Kawai,

1285:   Phys. Rev. Lett. {\bf 96}, 050601 (2006).

1286: \bibitem{Yosida} K. Yosida, Functional Analysis, Springer Verlag, Berlin, 1971.

1287: \bibitem{RG} L.S. Gradshteyn, I.M. Ryzhik, {\it Table of Integrals,

1288:     Series and Products}, Academic Press, San Diego (2000).

1289: \bibitem{vK} N.G. van Kampen, {\it Stochastic Processes in Physics and

1290:   Chemistry}, North Holland, Amsterdam, 1992.

1291: \bibitem{Wi} R.M. Wilcox, J. Math. Phys. {\bf 8}, 962 (1967).

1292: \bibitem{PBM} A.P. Prudnikov, Yu.A. Brychkov, O.I. Marichev, {\it

1293:     Integrals and Series}, Vol. 1, Gordon and Breach, New York, 1986.

1294: \end{thebibliography} 

1295: 

1296: \end{document}