1: % ****** Start of file apssamp.tex ******

2: %

3: %   This file is part of the APS files in the REVTeX 4 distribution.

4: %   Version 4.0 of REVTeX, August 2001

5: %

6: %   Copyright (c) 2001 The American Physical Society.

7: %

8: %   See the REVTeX 4 README file for restrictions and more information.

9: %

10: % TeX'ing this file requires that you have AMS-LaTeX 2.0 installed

11: % as well as the rest of the prerequisites for REVTeX 4.0

12: %

13: % See the REVTeX 4 README file

14: % It also requires running BibTeX. The commands are as follows:

15: %

16: %  1)  latex apssamp.tex

17: %  2)  bibtex apssamp

18: %  3)  latex apssamp.tex

19: %  4)  latex apssamp.tex

20: %

21: \documentclass[twocolumn,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}

22: %\documentclass[preprint,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}

23: 

24: % Some other (several out of many) possibilities

25: %\documentclass[preprint,aps]{revtex4}

26: %\documentclass[preprint,aps,draft]{revtex4}

27: %\documentclass[prb]{revtex4}% Physical Review B

28: 

29: \usepackage{graphicx}% Include figure files

30: \usepackage{dcolumn}% Align table columns on decimal point

31: \usepackage{bm}% bold math

32: 

33: %\nofiles

34: 

35: 

36: \newcommand{\be}{\begin{equation}}

37: \newcommand{\en}{\end{equation}}

38: \newcommand{\bea}{\begin{eqnarray}}

39: \newcommand{\ena}{\end{eqnarray}}

40: 

41: 

42: \begin{document}

43: 

44: %%% \draft

45: 

46: \title{Minimal Brownian Ratchet: An Exactly Solvable Model}

47: 

48: \author{Youngki~Lee$^{1,2}$, Andrew Allison$^3$, Derek~Abbott$^3$, and

49: H.~Eugene Stanley$^2$}

50: 

51: \affiliation{

52: $^1$Yanbian University of Science \& Technology,

53: Beishan St. Yanji, Jilin, 133000, China \\

54: $^2$Center for Polymer Studies and Department of Physics,

55: Boston University, Boston, MA 02215 \\

56: $^3$Centre for Biomedical Engineering (CBME) and School of

57: Electrical and Electronic Engineering, The University of Adelaide,

58: SA 5005, Australia

59: }

60: 

61: \date{Last modified: June 19, 2003.  Printed: \today}

62: 

63: \begin{abstract}

64: 

65: We develop an analytically solvable three-state discrete-time minimal

66: Brownian ratchet (MBR), where the transition probabilities between

67: states are asymmetric. By solving the master equations we obtain the

68: steady-state probabilities. Generally the steady-state solution does

69: not display detailed balance, giving rise to an induced directional

70: motion in the MBR. For a reduced two-dimensional parameter space we

71: find the null-curve on which the net current vanishes and detailed

72: balance holds. A system on this curve is said to be balanced. On the

73: null-curve, an additional source of external random noise is

74: introduced to show that a directional motion can be induced under the

75: zero overall driving force. %We also indicate the off-balance behavior

76: %with biased random noise.

77: 

78: \end{abstract}

79: 

80: \maketitle

81: 

82: %\pacs{PACS numbers: 87.10.+e, 87.80.+s, 87.90.+y}

83: 

84: %\begin{multicols}{2}

85: 

86: %%% \section{INTRODUCTION}

87: 

88: The Brownian ratchet and pawl system was first correctly explained by

89: Smoluchowski~\cite{smoluchowski1912} and later revisited by

90: Feynman~\cite{feynman1963} -- this has inspired much activity in the

91: area of Brownian ratchets, despite flaws in Feynman's analysis of the

92: thermal efficiency of the ratchet engine~\cite{parrondo1996} and

93: detailed balance~\cite{abbott2000}.

94: 

95: Interest has revived because molecular motors~\cite{astumian1994} have

96: been described in terms of Brownian

97: ratchet~\cite{doering1995,magnasco1993} models. Another area of

98: interest has been in Parrondo's paradox~\cite{harmer1999a} where losing

99: strategies cooperate to win. This can be illustrated in terms of games

100: that lose when played individually, but win when alternated -- this

101: has been shown to be a discrete-time Brownian ratchet~\cite{harmer1999b},

102: otherwise known as a `Parrondian game'. Parrondo's games have significantly

103: sparked recent interest in the areas of lattice gas automata~\cite{meyer02},

104: spin models~\cite{moraal00}, random walks and

105: diffusions~\cite{key02,pyke02,cleuren02}, biogenesis~\cite{davies01},

106: molecular transport~\cite{kinderlehrer02,heath02}, noise induced

107: patterns~\cite{buceta02}, stochastic

108: control~\cite{allison01a,kocarev02}, stochastic

109: resonance~\cite{allison01b} and quantum game

110: theory~\cite{flitney02,lee02}. Recently, Reimann~\cite{reimann02} has performed

111: an extensive review of the ratchet field.

112: 

113: Jarzynski {\it et~al.}~\cite{jarzynski1999} developed an exactly

114: solvable Brownian ratchet that can be operated as a heating system or

115: refrigerator, depending on the parameters between two heat reservoirs

116: of different temperatures. However this is treated as a six state

117: system and solution is via matrix inversion of coupled linear

118: equations. The derivation is somewhat complex, so the physical picture

119: and key ingredients of the observed properties are obscured.

120: 

121: Westerhoff {\it et~al.}~\cite{westerhoff1986} have analyzed enzyme

122: transport using a four-state model. In this paper, for the first time,

123: we develop a three-state discrete-time Brownian ratchet model that can

124: be solved analytically. We call it the {\it Minimal Brownian Ratchet}

125: (MBR)~\cite{footnote_01}. By setting up and solving the steady-state

126: solution of the corresponding master equations we obtain the

127: probability current and the null-surface, in the parameter space, of

128: the noisy and noise-free MBR. The obtained solution does not show any

129: critical behavior and can be suitably explained in terms of

130: non-singular behaviors.

131: 

132: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

133: % FIGURE 1

134: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

135: \begin{figure}

136:    \resizebox{84.67mm}{!}{\includegraphics*{mbrfig1.eps}}

137:    \caption {State-transition diagram of a 3-state discrete-time Brownian

138:    ratchet with asymmetric transition probabilities $p_0$, $p_1$ and

139:    $p_2$ in the positive direction (counter-clockwise) and $(1-p_0)$,

140:    $(1-p_1)$ and $(1-p_2)$ in the negative direction (clockwise).  Each

141:    transition has two numbers associated with it, $\left\{ p_k , R_k \right\}$.

142:    The first number in the brackets, $p_k$, is the

143:    conditional probability of that transition (given the initial

144:    state). The second number, $R_k$, is the reward associated with that

145:    transition. Note that we have a skip-free process, which means the reward structure

146:    is $+1$ for `winning' transitions and $-1$ for `losing' transitions.}

147:    \label{fig:fig1}

148: \end{figure}

149: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

150: 

151: The minimal ingredients of a Brownian ratchet are an asymmetric potential and

152: random noise. In Fig.~\ref{fig:fig1} we show the state diagram of the

153: MBR. The MBR has three states, $\{ S_0,S_1,S_2 \}$, where the

154: transition probabilities between states are asymmetric. The transition

155: probability that a random walker in state $S_k$ steps in the positive

156: direction is $p_k$. The probability of a shift in the negative

157: direction is $\tilde{p_k}$. This is true for $k \in \{0,1,2 \}$. We

158: define the positive direction as counterclockwise.  The condition of

159: normalization, $p_k+\tilde{p_k}=1$, is automatically enforced by our

160: choice of symbols. These ingredients comprise a three-state random walk model with

161: generalized asymmetric potential and we call it a noise-free MBR.

162: 

163: It is straight forward to set up the following difference equations for

164: the probability distributions of the noise-free MBR model:

165: \be

166:   P_k(t+1)= P_{k+1}(t) \tilde{p}_{k+1} + P_{k+2}(t) p_{k+2}

167:   \label{eq:master}

168: \en

169: for all cyclic (modulo-3) state indices $k$.

170: $P_k(t)$ is the probability for the random walker at time $t$ to

171: be on the state of $S_k$. This can be written in matrix form as ${\bf

172: P }_{t+1} = {\bf P }_{t} B $, where ${\bf P}_{t}$ is the time varying

173: probability (row) vector at time $t$ and $B$ is the transition

174: probability matrix. We can write:

175: \be

176:   \left[B_{i,j} \right] =

177:   \left[

178:   \begin{array}{ccc}

179:   0            & p_0           & \tilde p_0  \nonumber \\

180:   \tilde{p_1}  & 0             & p_1                   \\

181:   p_2          & \tilde{p_2}   & 0           \nonumber

182:   \end{array}

183:   \right].

184:   \label{eq:transition_matrix}

185: \en

186: The steady-state probability, after a sufficiently long time,

187: $\lim_{t\rightarrow \infty} {\bf P_t} = {\bf P}$ is simply given as

188: \be

189:   {\bf P} = {\bf P} B

190:   \label{eq:sss}

191: \en

192: which is a characteristic value problem.

193: A partial probability current, $I$, can be defined as

194: \be

195:   I = P_kp_k - P_{k+1} \tilde{p}_{k+1}.

196:   \label{eq:current-definition}

197: \en

198: If $I=0$ there is no net current and detailed balance

199: \cite{onsager_1931} is satisfied, otherwise there exists a net current

200: and the system will assume a non-equilibrium steady-state.

201: 

202: Solving the Eq.~\ref{eq:sss} together with the normalization condition,

203: $   P_0+P_1+P_2=1$ ,

204: is again straightforward. Using the standard methods for

205: characteristic value or eigenvalue problems, we obtain

206: \be

207:  P_k = (\tilde{p}_{k+1} +p_{k+1}p_{k+2})/D

208:   \label{eq:steady-state-solutions}

209: \en

210: for all $k$. The denominator $D$ is given as

211: \be

212:   D = 2 + p_0p_1p_2 + \tilde{p_0} \tilde{p_1} \tilde{p_2}.

213: \en

214: These expressions are consistent with the results of

215: Pearce~\cite{pearce2000}. It is easy to check that they are the

216: solution to Eq.~\ref{eq:sss} by direct substitution.

217: 

218: We can substitute the results from Eqs.~\ref{eq:steady-state-solutions}

219: into Eq.~\ref{eq:current-definition} to solve for the net current,

220: $I$,

221: \be

222:   I = (p_0 p_1 p_2  - \tilde{p_0}\tilde{p_1} \tilde{p_2}) /D.

223:   \label{eq:noise-free-current}

224: \en

225: The condition for detailed balance $I=0$  is then

226: \be

227:   p_0 p_1 p_2 = \tilde{p_0} \tilde{p_1}\tilde{p_2},

228:   \label{eq:detailed-balance}

229: \en

230: which is the equation of a two-dimensional surface in the three

231: dimensional parameter space, $\{ p_0 , p_1 , p_2 \}$.  Note that the

232: Eq.~\ref{eq:noise-free-current} is independent of state index $k$ as

233: is required by its definition given in

234: Eq.~\ref{eq:current-definition}.

235: 

236: The second part of the MBR is to introduce  additional random noise

237: to the system.  To the noise-free MBR, we add more noise, controlled by the

238: parameter $\gamma$, to the MBR as follows.  With a probability of

239: $\tilde{\gamma}$, a random walker follows the dynamic rule of the

240: noise-free MBR otherwise, with the probability of $ \gamma~\left(

241: =1-\tilde{\gamma} \right)$, the walker randomly takes a right or left

242: step with the equal probability of a half. For $\gamma = 0$ the model

243: is exactly same with the noise-free MBR.  In the other limit, for

244: $\gamma = 1$, the randomizing process dominates and the system reduces

245: to a simple unbiased random walk where the net current remains zero.

246: It is important to note that $\gamma$ {\it influences} the level of

247: noise in the ratchet but is not {\it identical} with the noise itself.

248: We refer to $\gamma$ as a `noise parameter'. With this modification

249: the transition probability matrix $B$ changes as

250: \be

251:   \left[B_{i,j} \right] =

252:   \left[

253:   \begin{array}{ccc}

254:   0 & \tilde{\gamma}p_0 + \gamma /2 & \tilde{\gamma} \tilde{p_0} + \gamma /2   \nonumber \\

255:   \tilde{\gamma}\tilde{p_1} + \gamma /2 & 0 &  \tilde{\gamma} p_1 +\gamma /2             \\

256:   \tilde{\gamma}p_2 + \gamma /2 & \tilde{\gamma}\tilde{p_2} + \gamma /2   & 0  \nonumber

257:   \end{array}

258:   \right].

259:   \label{eq:transition-mbr}

260: \en

261: 

262: From the transition matrix, we know that adding the random noise with

263: parameter $\gamma$ effectively changes the existing parameters as

264: \be

265:   p_k \rightarrow \tilde{\gamma}p_k +\gamma /2

266:   \label{eq:parameter-change}

267: \en

268: and same holds for the $\tilde{p_k}$'s. The steady state solution and net

269: current for noisy MBR can be obtained by exchanging all the

270: $p_k$'s in Eqs.~\ref{eq:steady-state-solutions} and

271: \ref{eq:noise-free-current} according to

272: Eq.~\ref{eq:parameter-change}. The expression for current is given as

273: \be

274:   I_\gamma =[ \tilde{\gamma}^3A_- + \tilde{\gamma}(\gamma/2)(1-\gamma/2)B]/D_\gamma

275: \en

276: where

277: \bea

278:   D_\gamma &=& 2 + \tilde{\gamma}^2A_+ + \tilde{\gamma}(\gamma/2)(1+\gamma/2) \nonumber  \\

279:   A_{\pm}&=&p_0p_1p_2 \pm \tilde{p_0}\tilde{p_1}\tilde{p_2}                   \\

280:   B&=&p_0+p_1+p_2-\tilde{p_0}-\tilde{p_1}-\tilde{p_2}.                        \nonumber

281: \ena

282: 

283: It is possible to further restrict the choices of $\{p_0,p_1,p_2\}$

284: without losing the important properties of the ratchet. Parrondo's

285: original definition imposed the further constraints $p_0=q$ and

286: $p_1=p_2=p$.  This reduced the parameter space to a two dimensional

287: space with parameters $\{p,q\}$.  In two dimensional $\{p,q\}$

288: parameter space, the condition of detailed balance, i.e.,

289: Eq.~\ref{eq:detailed-balance}, gives the equation for a curve that we

290: call the null-curve:

291: \be

292:   q = {\tilde{p}^2 \over p^2 +\tilde{p}^2} = {(1-p)^2 \over p^2 + (1-p)^2}.

293:   \label{eq:null-curve}

294: \en

295: The null-curve is a special case of the more general

296: null-surface or null-hypersurface, in higher dimensions.

297: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

298: % FIGURE 2

299: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

300: \begin{figure}

301:   \resizebox{84.67mm}{!}{\includegraphics*{mbrfig2.eps}}

302:   \caption

303:   {The null-surface of a 3-state discrete-time Brownian ratchet.  On the

304:   null-surface, $q = (1-p)^2/(p^2+(1-p)^2)$, the current vanishes.

305:   Above the curve, the system has positive net current. Below the

306:   curve, the system has negative net current.}

307:   \label{fig:fig2}

308: \end{figure}

309: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

310: Fig.~\ref{fig:fig2} shows the `positive' and `negative' net current

311: regions of the noise-free MBR. Note that as expected from the symmetry

312: of the system the curve is invariant under the transformations $q

313: \rightarrow (1-q)$ and $p \rightarrow (1-p)$. This also apparent from

314: a consideration of Eq.~\ref{eq:detailed-balance}.

315: 

316: On the null-surface, we introduce additional random noise to the system

317: by controlling the value of $\gamma$. For $\gamma = 0$ the model is

318: exactly same as the noise-free MBR and the net current remains zero

319: since we are on the null-surface. In the other limit, for $\gamma =

320: 1$, the randomizing process dominates and the system reduces to a

321: simple unbiased random walk where the net current is also zero.

322: However, counter-intuitively, for $0<\gamma<1$ non-zero current is

323: induced by introducing random noise controlled by $\gamma$.

324: 

325: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

326: % FIGURE 3

327: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

328: \begin{figure}

329:   \resizebox{84.67mm}{!}{\includegraphics*{mbrfig3.eps}}

330:   \caption

331:   {The probability current, $I_\gamma$, versus noise parameter,

332:   $\gamma$, on the null-surface. For values of $p \neq 0.5$,

333:   additional noise induces a net current that increases, in

334:   magnitude, with increasing $\gamma$ and then decreases, in

335:   magnitude, to zero after $\gamma$ exceeds an optimum value. The

336:   bottom curve corresponds to $p=0.1$. All the other curves represent

337:   increments of $\Delta p = 0.1$. The middle curve corresponds to $p

338:   = 0.5$. The top curve corresponds to $p = 0.9$. Parrondo's original

339:   games had $p = 0.75$.}

340:   \label{fig:fig3}

341: \end{figure}

342: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

343: 

344: In Fig.~\ref{fig:fig3} we show the current versus noise parameter,

345: $\gamma$, for different values of parameters $p$ and

346: $q=\tilde{p}^2/(p^2+\tilde{p}^2)$. As $\gamma$ is increased from zero,

347: the current increases to a maximum and then falls off, which has the

348: form of stochastic resonance~\cite{allison01b}. The position of this

349: extremum can be obtained from the condition, $\partial I_\gamma

350: /\partial \gamma = 0$. $\gamma$ varies a little from $0.408$ for

351: $p=0.1$ to $0.423$ for $p=0.5$.

352: %~\cite{referee-comment}.

353: 

354: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

355: % FIGURE 4

356: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

357: \begin{figure}

358:   \resizebox{84.67mm}{!}{\includegraphics*{mbrfig4.eps}}

359:   \caption

360:   {The probability current $I_\gamma$, versus noise parameter,

361:   $\gamma$, off the null-surface. The bottom curve corresponds to

362:   $p=0.73$ and $q=0.08$ and the top to $p=0.77$ and $q=0.12$.}

363:   \label{fig:fig4}

364: \end{figure}

365: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

366: 

367: Fig.~\ref{fig:fig4} shows the net current, $I_\gamma$, versus noise $\gamma$

368: when the system is not on the null-surface any more.  In the off-balance

369: region, $q \ne \tilde{p}^2 / (p^2 + \tilde{p}^2)$, the net current is not

370: zero for $\gamma=0$ but still should be zero for $\gamma=1$ and the

371: intermediate behavior is qualitatively the same as the balanced behavior. The

372: actual values of $p$ and $q$ for the various curves in Fig.~\ref{fig:fig4}

373: are in linear increments of $0.01$ for $p$ and $q$. The top curve has

374: parameters $p=0.77$ and $q=0.12$. The bottom curve has parameters $p=0.73$

375: and $q=0.08$.

376: 

377: We generalize the MBR by introducing a bias into the added noise. The walker

378: takes a right step with probability of $0.5-\epsilon$ and a left step with

379: probability of $0.5+\epsilon$. For $\epsilon \ne 0$ this noise introduces

380: non-zero net current. The new parameter, $\epsilon$, is essentially a measure

381: of the bias in the added noise.

382: 

383: %In Fig.~\ref{fig:fig5} we show the effect of the bias on the

384: %current, $I_\gamma$, versus noise, $\gamma$, curve for differenct

385: %$\epsilon$ values on the null-surface at the point ($p=0.75, q=0.10$).

386: %The biased noise essentially produces the same effect with the

387: %off-balanced behavir.

388: 

389: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

390: % FIGURE 5

391: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

392: %\begin{figure}

393: %   \resizebox{84.67mm}{!}{\includegraphics*{mbrfig5.eps}}

394: %   \caption

395: %   {The probability current, $I_\gamma$ versus noise parameter,

396: %   $\gamma$, on the null-surface with biased noise. The bottom curve

397: %   corresponds to $s=0.46$. All the other curves represent increments of

398: %   $\Delta s = 0.01$. The top curve corresponds to $s = 0.54$.}

399: %   \label{fig:fig5}

400: %\end{figure}

401: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

402: 

403: We can generalize this model to a system of size $N $ by repeating the

404: unit cell of modulo-3 $N$ times with a periodic boundary condition. In

405: this case, the periodic potential ensures

406: %\be

407: $ p_{k}(t)=p_{k+3n}(t)~\forall~n=0,\pm1,\pm2, \cdots.   $

408: %\en

409: Because of the normalization condition, $\sum_{k=1}^{N} P_k(t) =1$,

410: the current will be reduced by factor of $N$. Otherwise, the

411: corresponding master equations and solutions are exactly same as the

412: minimal model.

413: 

414: For different moduli, in principle, we can also set up the master

415: equations and solve them exactly by matrix inversion for the set of

416: linear equations. It can be shown that these results have

417: qualitatively the same statistical behavior as the 3-state MBR. Note

418: that for even number moduli there are oscillatory non-stationary

419: solutions.

420: 

421: The transformation in Eq.~\ref{eq:parameter-change} tells us

422: effectively that the MBR gives the same results as a biased random walk, where

423: the transition probability is not symmetric but biased. Although this

424: analogy can be used to investigate the characteristics of MBR, it is

425: absolutely impossible to determine whether the system is itself a biased

426: random walk or an MBR by analyzing the result of measurements, without prior

427: knowledge that the model is a combination of a balanced unbiased random walk

428: and added random external noise. This makes the MBR valuable for

429: understanding the minimal features of the discrete-time Brownian ratchet.

430: %The MBR is analogous to combining two dice, which individually give you overal

431: %zero winning or losing probabilties, but when combined give your more

432: %winning probability.

433: % Since MBR is a discrete time model it is

434: %difficult to find a direct application to physical or a biochemical

435: %systems.

436: The MBR has applicability in discrete-time processes where

437: the transition probabilities do not fluctuate in time, such as in game and

438: computation theory where transitions occur at precisely defined times.

439: 

440: Funding from GTECH Australasia, Sir Ross and Sir Keith Smith Fund and

441: the Australian Research Council (ARC) is gratefully acknowledged.

442: 

443: \begin{thebibliography}{999}

444: 

445: \bibitem{smoluchowski1912}

446: M.~von Smoluchowski, Phys.~Zeitschrift, {\bf XIII,} 1069 (1912).

447: 

448: \bibitem{feynman1963}

449: R.P.~Feynman, R.B.~Leighton and M.~Sands, {\it The Feynman Lectures on

450: Physics,} {\bf 1,} 46.1, Addison-Wesley, (1963).

451: 

452: \bibitem{parrondo1996}

453: J.M.R.~Parrondo and P.~Espa\~{n}ol,  Am.~J.~Phys., {\bf 64,} 1125

454: (1996).

455: 

456: \bibitem{abbott2000}

457: D.~Abbott, B.R.~Davis and J.M.R.~Parrondo, in {\it Proc. 2nd Int. Conf. on

458: Unsolved Problems of Noise and fluctuations (UPoN'99),} Adelaide, Australia,

459: 12-15 July 1999, American Institute of Physics (AIP), {\bf 511}, 213 (2000).

460: 

461: \bibitem{astumian1994}

462: R.D.~Astumian and M.~Bier, Phys.~Rev.~Lett., {\bf 72}, 1766 (1994).

463: 

464: \bibitem{doering1995}

465: C.R.~Doering, Nuovo Cimento, {\bf 17D,} 685 (1995).

466: 

467: \bibitem{magnasco1993}

468: M.O.~Magnasco, Phys.~Rev.~Lett., {\bf 71}, 1477 (1993).

469: 

470: \bibitem{harmer1999a}

471: J.M.R.~Parrondo, in {\it EEC HC\&M Network on Complexity and Chaos}

472: (\#ERBCHRX-CT940546), ISI, Torino, Italy, (1996);

473: G.P.~Harmer and D.~Abbott, Nature, {\bf 402,} 846 (1999);

474: J.M.R.~Parrondo, G.P.~Harmer and D.~Abbott, Phys.~Rev.~Lett.,

475: {\bf 85}, 5226, (2000);

476: G.P.~Harmer and D.~Abbott, Fluctuation and noise Lett., {\bf 2}, R71 (2002);

477: G.P.~Harmer, D.~Abbott, P.G.~Taylor and J.M.R.~Parrondo, in {\it Proc.~2nd

478: Int. Conf.~on Unsolved Problems of Noise and fluctuations (UPoN'99),}

479: Adelaide, Australia, 12-15 July 1999 (Ref. \cite{abbott2000}, p. 189).

480: 

481: \bibitem{harmer1999b}

482: G.P.~Harmer and D.~Abbott, Statistical Science, {\bf 14}, 206

483: (1999).

484: 

485: \bibitem{meyer02}

486: D.A.~Meyer and H.~Blumer, J.~Stat.~Phys., {\bf 107}, 225, (2002).

487: 

488: \bibitem{moraal00}

489: H.~Moraal, J.~Phys.~A, {\bf 33}, L203, (2000).

490: 

491: \bibitem{key02}

492: E.S.~Key, M.M.~K{\l}osek and D.~Abbott, math.PR/0206151.

493: 

494: \bibitem{pyke02}

495: R.~Pyke, math.PR/0206150.

496: 

497: \bibitem{cleuren02}

498: B.~Cleuren and C.~Van den Broeck, Phys.~Rev.~E, {\bf 65}, 030101,

499: (2002).

500: 

501: \bibitem{davies01}

502: P.C.W. Davies, Lecture in honour of Abdus Salam, in

503: {\it Proc.~Sixth Trieste Conference on

504: Chemical Evolution,} eds. J. Chela-Flores, T. Owen and F. Raulin,

505: Trieste, Italy 18-22 September, 2000 (Kluwer, Dordrecht, 2001) pp.~13-20.

506: 

507: \bibitem{kinderlehrer02}

508: D.~Kinderlehrer and M. Kowalczyk,

509: {Arch. Ration. Mech. An.}, {\bf 161}, 299 (2002).

510: 

511: \bibitem{heath02}

512: D.~Heath, D. Kinderlehrer and M. Kowalczyk, {Discrete and Continuous

513: Dynamical Systems-Series B,} {\bf 2}, 153, (2002).

514: 

515: \bibitem{buceta02}

516: J.~Buceta, K.~Lindenberg and J.M.R.~Parrondo, {Fluctuation and Noise

517: Lett.,} {\bf 2}, L21, (2002); J.~Buceta, K.~Lindenberg and

518: J.M.R.~Parrondo, {Phys.~Rev.~Lett.}, {\bf 88}, 024103 (2002).

519: 

520: \bibitem{allison01a}

521: A.~Allison and D.~Abbott, {Chaos,} {\bf 11}, 715, (2001).

522: 

523: \bibitem{kocarev02}

524: L.~Kocarev and Z.~Tasev, {Phys.~Rev.~E.}, {\bf 65}, 046215, (2002).

525: 

526: \bibitem{allison01b}

527: A.~Allison and D.~Abbott, {Fluctuation and Noise Letters}, {\bf 1},

528: L239, (2001).

529: 

530: \bibitem{flitney02}

531: A.P.~Flitney, J.~Ng and D.~Abbott, {Physica A,} {\bf 314}, 35 (2002),

532: {quant-ph/0201037}.

533: 

534: \bibitem{lee02}

535: C.F.~Lee and N.F.~Johnson, {quant-ph/0203043}/

536: 

537: \bibitem{reimann02}

538: P.~Reimann, {Phys.~Rep.} {\bf 361}, 57 (2002).

539: 

540: \bibitem{jarzynski1999}

541: C.~Jarzynski and O.~Mazonka, {Phys.~Rev.~E} {\bf 59}, 6448 (1999).

542: 

543: \bibitem{westerhoff1986}

544: H.V.~Westerhoff, T.Y.~Tsong, P.B.~Chock, Y.~Chen and R.D.~Astumian,

545: {PNAS,} {\bf 83}, 4734 (1986).

546: 

547: \bibitem{footnote_01} 

548: Although a two-state system is possible, it cannot model the class of

549: skip-free discrete-time ratchet systems described by Parrondo's games. This

550: is because two states do not provide the necessary asymmetry for the Parrondo

551: effect. We can obtain a winning game from two losing games in a two-state

552: system, if we violate the skip-free condition. A skip-free process implies

553: payoffs in steps of +1 or -1. The significance is that particles in a

554: continuous-time Brownian ratchet generally do not skip over multiple teeth at

555: once -- our discrete-time model reflects this by imposing the skip-free

556: condition. This is why our model can be regarded as a {\it minimal}

557: discrete-time ratchet.

558: 

559: %{So far the minimal model of the Brownian ratchet involves at least two

560: %coupled two-state systems described by Parrondo's games. So the total

561: %degree of freedom is four except the random noise exterted on it. Also

562: %a two state cycle system with fluctuating rate coefficients, for which

563: %an even simpler analytic solution can be given, also works. However,

564: %the degree of freedom of such sytems are bigger than our system, which

565: %is three in our case, and the coupling between those degree of

566: %freedoms are often of unknown.  In that sense, we are comfortable to

567: %call our model 'minimal' showing all the feature of Browninal Rachet

568: %in its simplest way.}

569: 

570: \bibitem{onsager_1931}

571: Notice that Eq.~\ref{eq:detailed-balance} is sufficient condition for

572: detailed balance only when the transition probabilities do not depend

573: on time, which is our case. Eq.~\ref{eq:detailed-balance} is analogous to

574: a corollary of the Onsager condition for circular chemical reactions at

575: equilibrium: L.~Onsager, {Phys.~Rev.}, {\bf 37}, 405 (1931). Also see

576: {Phys.~Rev.~A}, {\bf 39}, 6416 (1989).

577: 

578: \bibitem{pearce2000}

579: C.E.M.~Pearce, in {\it Proc. 2nd Int. Conf. on Unsolved Problems of Noise and

580: fluctuations (UPoN'99),} Adelaide, Australia, 12-15 July 1999, American

581: Institute of Physics (AIP), {\bf 511}, 207 (2000).

582: 

583: \end{thebibliography}

584: 

585: \end{document}

586: