1: %\documentclass[12pt]{preprint}
2: \documentclass[aps,floatfix,showpacs,twocolumn,prl]{revtex4}
3: \usepackage[dvips]{graphicx}
4: \usepackage{amsfonts}
5: \usepackage{amssymb}\usepackage{times}
6: \usepackage{sub_JP}
7: \usepackage{MHequ}
8:
9: \def\J{{\rm J}}
10: \def\Q{{\rm Q}}
11: \def\L{{\rm L}}
12: \def\R{{\rm R}}
13: \def\aJ{\alpha^\J}
14: \def\aQ{\alpha^\Q}
15: \def\eref#1{(\ref{#1})}
16: \let\rho=\varrho
17: \def\RRR{{\mathcal R}}
18:
19: \begin{document}
20: \title{Magnetically Induced Thermal Rectification}
21: \author{Giulio Casati${}^{1,2,3}$}
22: \author{Carlos~Mej\'\i{a}-Monasterio${}^4$}
23: \author{Toma\v z Prosen${}^5$}
24: \affiliation{Center for Nonlinear and Complex Systems, Universit\`a
25: degli Studi dell'Insubria, Como Italy${}^1$}
26: \affiliation{CNR-INFM and Istituto Nazionale di Fisica Nucleare, Sezione di Milano${}^2$}
27: \affiliation{Department of Physics, National University of Singapore, Republic of Singapore${}^3$}
28: \affiliation{Dipartimento di Matematica, Politecnico di Torino, Italy${}^4$}
29: \affiliation{Physics Department, Faculty of Mathematics and Physics, University of Ljubljana, Ljubljana, Slovenia${}^5$}
30: \date{\today}
31:
32: \begin{abstract}
33: We consider far from equilibrium heat transport in chaotic billiard chains
34: with non-interacting charged particles in the presence of non-uniform
35: transverse magnetic field. If half of the chain is placed in a strong
36: magnetic field, or if the strength of the magnetic field has a large
37: gradient along the chain, heat current is shown to be asymmetric with
38: respect to exchange of the temperatures of the heat baths. Thermal
39: rectification factor can be arbitrarily large for sufficiently small
40: temperature of one of the baths.
41: \end{abstract}
42: \pacs{44.10.+i, 05.70.Ln, 05.45.-a}
43:
44: \maketitle
45:
46:
47:
48:
49:
50: The problem of explaining irreversible macroscopic transport laws, such as the
51: Fourier law of heat conduction, from the reversible microscopic equations of
52: motion is one of the open problems in nonequilibrium statistical mechanics
53: \cite{bonetto,lepri}. The problem is still far from being settled, in
54: particular in the framework of low dimensional systems, e.g. one dimensional
55: particle chains, where different instances of anomalous transport can be
56: identified due to various kinematic or dynamic mechanisms.
57:
58: In systems of non-interacting particles, e.g. quasi 1-d chaotic billiards, the
59: validity of Fourier law has clearly been confirmed \cite{alonso}, even though
60: such non-interacting system cannot exhibit local thermal equilibrium in the
61: non-equilibrium steady state \cite{dhar,mejia}. Furthermore, interesting
62: connection between the Fourier law and particle diffusion has been established
63: \cite{liwang}. These problems are not only interesting for understanding the
64: fundamentals of statistical mechanics, but they may also have straightforward
65: applications, e.g. for connecting dynamics and transport in emerging
66: nano-technology, engineering of molecular motors, understanding and control of
67: energy flow in bio-molecules, etc.
68:
69: In view of these ideas, Terraneo et al. \cite{terraneo} have recently proposed
70: a mechanism for thermal rectification in an anharmonically interacting
71: particle chain. Using an effective phonon approach, they have shown that a
72: chain composed of three different parts may have asymmetric heat transport
73: properties due to (non)matching of the effective phonon bands. This and
74: related ideas have been further elaborated and improved\cite{li-2,li-3}. A
75: further step to devise a thermal transistor has been discussed in \cite{li-4}
76: in terms of the negative differential thermal resistance observed in some
77: anharmonic chains.
78:
79: Since these first works, other different mechanisms leading to thermal
80: rectification have been described. In \cite{segal} it was shown that the
81: nonlinearity of the dynamics of an asymmetric two-level system leads to an
82: asymmetric heat flow. More recently in \cite{EMM-2} it was shown that thermal
83: rectification can be observed in asymmetric billiards of interacting
84: particles. When the billiard is subjected to an external temperature gradient
85: the effective interaction leads to a temperature dependence of the
86: transmission coefficient and thus, it is possible to dynamically controls the
87: transmission probability of the billiard. For this type of billiard systems
88: rectifications as large as $10^3$ were observed. However, the theory presented
89: in \cite{EMM-2} requires the knowledge of the microscopic transmission
90: coefficients, which is, at best phenomenological. Another, simple
91: phenomenological mechanism of thermal rectification, based on strong
92: positional and temperature dependence of local thermal conductivity, has been
93: proposed by Peyrard \cite{peyrard}.
94:
95: A common problem with most of these proposals is that the rectification factor
96: has always been rather limited, i.e. it is very difficult to achieve a
97: situation in which heat flows only in one direction and not in the opposite
98: one. Moreover, the fact that the thermal rectification depends directly or
99: indirectly on the microscopic particle-particle interaction, renders difficult the ability to
100: control the power of rectification.
101:
102: In this letter we propose a novel microscopic mechanism for thermal
103: rectification which works in the absence of particle interactions. Instead,
104: thermal rectification is controlled with an external non-uniform magnetic
105: field, leading to an {\em arbitrarily large} power of rectification. In our
106: model heat is carried by charged particles and the only restriction is that
107: the typical mean-free-paths due to dissipative mechanisms should be much
108: larger than the Larmor radius of the charged particles in the magnetic field.
109: In practice this means that the temperatures of the baths should be quite low,
110: but such regimes are nowadays easily accessible for example in mesoscopic
111: physics, quantum dots/wires, anti-dot lattices etc. The rectification
112: mechanism is very simple, it is based on asymmetric reflection of slow and
113: fast particles off the interface between regions with different magnetic field
114: intensities and predicts arbitrary large rectification factors for sufficiently low
115: temperatures.
116:
117: We consider a gas of noninteracting point particles of mass $m$ and electric
118: charge $e$ that move freely inside a two-dimensional billiard region. The
119: billiard is shaped as a chain of equal cells as depicted in
120: Fig.~\ref{fig:channel}. Let the circular obstacles have radius $R$ and centers
121: of the discs be arranged in a hexagonal array with lattice distance
122: $4/\sqrt{3}$, such that the billiard motion has a closed horizon for $R \ge
123: 1$. Then we cut out a quasi one-dimensional billiard channel of width $\delta
124: y=2/\sqrt{3}$ through the centers of two nearby rows of discs, so that one
125: rectangular $\delta x \times \delta y$ cell of the channel, containing one
126: half disk and two quarter disk obstacles, has length $\delta x=4$. Negative
127: curvature of the billiard boundary ensures that the motion in the absence of
128: the magnetic field is completely chaotic - hyperbolic.
129:
130: \begin{figure}[!t]
131: \begin{center}
132: \includegraphics[scale=0.7]{channel.eps}
133: \caption{
134: Geometry of the model: We consider a billiard composed by $N$ symmetric unit
135: cells, each of which is made of a rectangular boundary of width $\delta y 2/\sqrt{3}$ and length $\delta x = 4$. Inside each cell there are three
136: circulars hard scatterers of radius $R$ disposed in a triangular lattice
137: (see the text). Each cell is subjected to an external perpendicular uniform
138: magnetic field of strength $B_i \in [0,B]$. At the left and right boundaries
139: the billiard is placed in contact with two stochastic thermal baths at
140: different temperatures $T_\L$ and $T_\R$. The figure corresponds to a
141: channel of $N=2$ cells in which $B_1=0$ and $B_2=B$. The dashed lines are
142: drawn as a reference.
143: \label{fig:channel}}
144: \end{center}
145: \end{figure}
146:
147: The simplest model that we consider is composed of two cells
148: (Fig.~\ref{fig:channel}). The left cell contains no magnetic field, whereas the
149: right cell is subjected to a perpendicular uniform magnetic field of strength
150: $B$. In what follows we will refer as \emph{step} configuration, to a channel
151: of $N$ cells for which the $N/2$ left cells contain no magnetic field and the
152: $N/2$ right cells contain a magnetic field of density $B$. Let
153: $\lambda$ denote the smallest length scale in the problem, namely the width of
154: the opening between the neighboring cells in our model,
155: \begin{equ} \label{eq:lambda}
156: \lambda = \frac{2}{\sqrt{3}} - R ~.
157: \end{equ}
158:
159: The transmission probability of the billiard is controlled by the strength of
160: the magnetic field. Consider the particles that cross the junction between the
161: two cells from \emph{left to right}. There exist a critical velocity $v_c$:
162: fast particles of velocity $v > v_c$, always enter the right cell, and thus
163: contribute to the left to right energy flow provided they reach the right end
164: of the system which is coupled to a heat bath as explained below.
165: Instead, slow particles of velocity $v < v_c$, such that the
166: gyro-magnetic radius $\rho(v) = mv/(eB)$ is less than $\lambda/2$ will be
167: reflected or transmitted depending on the position at which they reach the
168: interface.
169:
170: Using a statistical ensemble of trajectories, the condition for the critical velocity
171: $\rho(v_c) = \lambda/2$ can be rewritten as the condition giving a critical
172: temperature
173: \begin{equ} \label{eq:Tc}
174: T_{\mathrm{c}} = \frac{(eB_{\mathrm{c}} \lambda)^2}{8m k_{\rm
175: B}} ~.
176: \end{equ}
177: such that particles which are colder than $T_{\rm c}$ will be reflected in
178: their majority.
179:
180: However, for the particles that cross the junction from \emph{right to left}
181: there is no condition on their velocity and they always enter the left cell. The
182: above qualitative argument makes it clear that the transport of heat will be
183: strongly asymmetric with respect to exchange of effective temperatures of
184: particles on different sides of magnetic field boundary, provided the
185: temperatures are strongly different, one being larger and the other smaller
186: than $T_{\rm c}$. In the rest of the letter we measure the temperature in units
187: of $T_{\rm c}$, and denote it as $\tau = T/T_{\rm c} = \frac{m}{2 k_{\rm B}
188: T_{\rm c}} \langle v^2 \rangle$.
189:
190: We couple the left cell with a stochastic heat bath \cite{alonso} of
191: dimensionless temperature $\tau_{\rm L}$ and the right cell with a heat bath
192: of temperature $\tau_{\rm R}$, namely if a particle reaches the end boundary of
193: the billiard that is in contact with a heat bath it is reflected with a
194: velocity chosen from a distribution with probability densities
195: \begin{equa}[3] \label{dist}
196: &P_n(v_x) & \ \ = \ \ & \frac{1}{\tau}|v_x|\exp\left(-\frac{v_x^2}{2\tau}\right) \ ,\\
197: &P_t(v_y) & \ \ = \ \ & \frac{1}{\sqrt{2\pi\tau}}\exp\left(-\frac{v_y^2}{2\tau}\right) \ ,
198: \end{equa}
199: where $\tau$ is the temperature of the respective heat bath in dimensionless
200: units.
201:
202: \begin{figure}[!t]
203: \begin{center}
204: \includegraphics[scale=0.65]{fig-2a.eps}
205: \caption{(Color online).
206: Rectification index $\Delta$ as a function of the minimal temperature
207: $\tau_{\min}$ for a $2$-cell channel in a \emph{step} configuration. The
208: maximal temperature was set to $\tau_{\max}=33.4275$. The dashed line is for
209: the no rectification value $\Delta=1$. The solid line corresponds to
210: $\tau_{\min}^{-1/2}$. Inset: Color density plot for $\Delta$ as a function
211: of the temperatures of both baths. Note that by construction the matrix is
212: symmetric. For the sake of presentation the temperatures are rescaled by
213: $\alpha=1.33711$.
214: \label{fig:rect-2}}
215: \end{center}
216: \end{figure}
217:
218: If the left bath is cold, i.e. $\tau_{\rm L} \ll 1$, then the particles will
219: most of the time remain in the left cell and there will be no heat current
220: between the baths, irrespective of the temperature of the right bath, which we
221: assume is larger than $1$. If we exchange the temperatures of the baths, then
222: we will in general have some non-small heat current flowing, since cold
223: particles have no problem in leaving the region with a magnetic field.
224:
225: Let us measure the heat currents per particle in the steady state, and denote
226: them as $J^+$ if $\tau_{\rm L} < \tau_{\rm R}$ and $J^{-}$ if the temperatures
227: are exchanged, i.e. $\tau_{\rm L} > \tau_{\rm R}$. Comparing the magnitude of
228: these two currents we quantify the thermal rectification as
229: \begin{equ} \label{eq:rect}
230: \Delta = \frac{\max\{|J^+|,|J^-|\}}{\min\{|J^+|,|J^-|\}} ~.
231: \end{equ}
232:
233: From our argument it is clear that rectification will be effective if one of
234: the temperatures is very small, say $\tau_{\rm L} \ll 1$ and the other is
235: simply above the critical, $\tau_{\rm R} > 1$. It is possible to make a
236: quantitative prediction on the scaling of rectification factor $\Delta$ with
237: temperatures. When the energy current through the magnetic field interface is
238: very weak, namely if $\tau_{\rm L} < \tau_{\rm R}$, the current is proportional
239: to the transmission coefficient at the interface, i.e. one minus the
240: probability of reflection.
241:
242: From our previous discussion it is clear that if $\tau_{\rm L} < 1 <
243: \tau_{\rm R}$ the particle density will be larger at the left cell (with zero
244: magnetic field). This is mainly because the cold particles in the left
245: cell spend long time before being able to cross the interface. A particle of
246: velocity $v$ is transmitted (not reflected) to the right cell if it crosses
247: the interface at a distance from the upper boundary shorter than $2\rho(v)$
248: \cite{sign}. Therefore, invoking the ergodicity of the dynamics in the left
249: cell, we can simply estimate the transmission coefficient $t$ as
250: \begin{equ} \label{eq:t}
251: t^+ \sim \frac{2\rho(v)}{\lambda} ~,
252: \end{equ}
253: where $\rho(v(\tau))=\sqrt{2mk_B\tau_{\rm min}}/eB$ and we denote by
254: $\tau_{\rm min} = {\rm min}\{ \tau_{\rm L},\tau_{\rm R}\}$.
255:
256: However, in the reverse situation (exchanging $\tau_{\rm L}$ and $\tau_{\rm
257: R}$) we have $t^- \sim 1$, so we can estimate the rectification
258: \begin{equ} \label{eq:scaling}
259: \Delta = t^-/t^+ \propto \frac{1}{\sqrt{\tau_{\rm min}}} ~.
260: \end{equ}
261:
262: \begin{figure}[!t]
263: \begin{center}
264: \includegraphics[scale=0.7]{profiles-a.eps}
265: \caption{
266: Energy density ($a$) and particle density ($b$) profiles for a $10$-cell
267: channel in a {\em step} configuration. The bath temperatures are
268: $\tau_\mathrm{min}=3.34277\times10^{-2}$,
269: $\tau_\mathrm{max}=33.4277$. In both panels profiles are shown for the
270: positive and negative gradient for which $\tau_\R=\tau_\mathrm{max}$
271: (triangles) and $\tau_\R=\tau_\mathrm{min}$ (circles) respectively.
272: \label{fig:T}}
273: \end{center}
274: \end{figure}
275:
276: In Fig.~\ref{fig:rect-2} we show the rectification index $\Delta$ as a
277: function of $\tau_{\rm min}$ for fixed value of the maximal temperature
278: $\tau_{\rm max}$. We clearly confirm the scaling (\ref{eq:scaling}),
279: indicating also that the rectification index is only very weakly depending on
280: the maximal temperature (as long as $\tau_{\rm max} > 1$. This can be seen in
281: the inset of Fig.~\ref{fig:rect-2}) where the rectification index $\Delta$ is
282: shown as a function of the two temperatures of the heat baths $\tau_\L$ and
283: $\tau_\R$. The correctness of the scaling \eref{eq:scaling} shows that the
284: magnetically induced rectification power is arbitrarily large for sufficiently
285: small temperature $\tau_{\rm min}$.
286:
287:
288: Even though a non-interacting system cannot reach local thermal equilibrium,
289: and therefore the concept of (local) temperature is not well defined, it is
290: instructive to compute temperature and density profiles, as long-time averages
291: of kinetic energy density and particle density as a function of the horizontal
292: coordinate along the chain. In Fig.~\ref{fig:T} we plot the kinetic energy
293: density and the particle density measured in each cell of a channel of $10$
294: cells, the right half of $5$ cells being in a uniform magnetic field. We plot
295: the profiles for the positive gradient, $\tau_\L < \tau_\R$ (triangles), and
296: the inverted negative gradient (circles). The positive gradient for which
297: $\tau_\L$ is $\tau_{\rm min}$, corresponds to the insulating case, \emph{i.e.}
298: to the situation in which the heat current is very small. In Fig.~\ref{fig:T}
299: we see that the insulating case corresponds to a large gap in the kinetic
300: energy profile. In contrast, for the negative gradient the gap in the kinetic
301: energy profile is much smaller and this coincides with the observation of a
302: larger heat current. However, note that the energy profile for the negative
303: gradient is not a monotonous function due to the lack of local thermal
304: equilibrium. Moreover, the density profile confirms our prediction that the
305: low current situation is characterized by a very small density of particles
306: on the side of the magnetic field.
307:
308: \begin{figure}[!t]
309: \begin{center}
310: \includegraphics[scale=0.7]{Delta-10-grad-a.eps}
311: \caption{
312: $10$-cell channel in a \emph{gradient} configuration: ($a$) Rectification
313: index $\Delta$ as a function of the minimal temperature $\tau_{\min}$, with
314: $\tau_{\max}=33.4275$. The temperatures are given in units of
315: $T_\mathrm{c}(B)$, where $B=100$ is the strength of the magnetic field at
316: the rightmost cell. The dashed line is for the no rectification value
317: $\Delta=1$. The solid line corresponds to $\tau_{\min}^{-1/2}$. In the
318: panels at the right the profiles of kinetic energy ($b$) and particle
319: density ($c$) for the positive (triangles) and negative (circles) gradient
320: are shown.
321: \label{fig:grad}}
322: \end{center}
323: \end{figure}
324:
325: We discuss now a slightly different situation in which the magnetic field does
326: not change abruptly from one half of the system to the other, but instead
327: changes gradually, forming a uniform gradient of the magnetic field. We refer
328: to this situation as the \emph{gradient} configuration for which the magnetic
329: field in each cell has an intensity given by $B_i=B(i-1)/(N-1)$ for
330: $i=1,2,\ldots,N$. Note that for $N=2$ the {\em gradient} and the {\em step}
331: configurations coincide.
332:
333: \begin{figure}[!t]
334: \begin{center}
335: \includegraphics[scale=0.7]{vsL-step.eps}
336: \caption{
337: Dependence of the thermal rectification index $\Delta$ as a function of the
338: size of the chain $L$ for a chain in a \emph{step}-configuration with fixed
339: $\tau_{\min}=3.34275\times10^{-3}$, $\tau_{\max}=3.34275\times10^{2}$ and
340: fixed particle density, namely one particle per cell. The dashed line corresponds to $L^{-1}$. Inset:
341: heat current $J$ as a function of the size of the chain $L$ for the
342: positive (triangles) and negative (circles) gradient.
343: \label{fig:vsL-step}}
344: \end{center}
345: \end{figure}
346:
347: We have performed numerical simulations for the {\em gradient} configuration,
348: and we found no qualitative differences with respect to the {\em step}
349: configuration. In Fig.~\ref{fig:grad} we show the dependence of the rectification
350: index on the minimal temperature (panel $a$). As for the {\em step}
351: configuration we have found that for sufficiently low temperature $\tau_{\rm
352: min}$, the rectification index again grows as $\Delta\sim1/\sqrt{\tau_{\rm min}}$.
353: In the panels ($b$) and ($c$) we show the profiles of the kinetic energy and
354: particle density respectively. As expected the profiles for the {\em
355: gradient} configuration are more smooth than for the {\em step}
356: configuration.
357:
358: From our analysis it follows that our rectification effect is a
359: phenomenon which exists only in far from equilibrium situation.
360: In the thermodynamic limit, a vanishingly small temperature
361: gradient is established across the system.
362: %For a finite but large
363: %system of length $L$, a relative difference of local temperatures on
364: %the two sides of the magnetic interface is expected to decrease as
365: %$1/L$.
366: In Fig.\ref{fig:vsL-step} we show the results of numerical
367: simulations which indicate that, for fixed bath temperatures,
368: the rectification index scales as $\Delta \sim 1/L$. In the
369: inset of Fig.\ref{fig:vsL-step} the time averaged heat currents
370: for the positive and negative gradients are also shown.
371:
372: Finally, it is interesting to give a quantitative estimate of the critical
373: temperature in Eq.~\eref{eq:Tc} in physical units. Let us suppose that the gas of particles
374: inside the billiard of Fig.~\ref{fig:channel} consists of electrons.
375: Assuming that the dimension of the opening is $\lambda = 100 {\rm nm}$ and
376: a magnetic field of $B = 1 {\rm T}$ is applied, then the critical temperature is
377: $T_c\sim0.5 {\rm K}$. Thus, a rectification power of $\Delta\sim10$ would
378: be measurable for thermal gradient given by $T_{\rm min}\sim10^{-3} {\rm
379: K}$ and $T_{\rm max}\sim10 {\rm K}$ that appears accessible to nowadays
380: experiments.
381:
382: In this letter we have presented a novel mechanism for thermal rectification.
383: This mechanism is fairly simple. It results from the asymmetric behavior of
384: the dynamics at the magnetic interface, that leads to a simple temperature
385: dependence of the transmission coefficient. Moreover, this mechanism for
386: thermal rectification is not based on the, macroscopically unaccessible,
387: microscopic particle interaction, but on the interaction with an external
388: field, making possible an easy control of the power of rectification. We have
389: shown that the thermal rectification power is arbitrarily large for
390: sufficiently small temperature of one of the heat baths.
391:
392: % this mechanism is particularly adapted to mesoscopic devices.
393: Furthermore, the physical scales needed for optimal implementation of our
394: theoretical model are realizable in present nano-scale experiments with
395: meso-scopic devices. We believe that it is precisely at these scales of
396: meso-scopic physics where such a thermal rectifier would find interesting
397: applications. This fact makes an experimental verification of the mechanism
398: presented here very desirable.
399:
400:
401:
402: \begin{acknowledgments}
403: CM-M acknowledge support by the Institute for Scientific Interchange.
404: TP acknowledges support by the program P1-0044 and project J1-7347
405: of Slovenian Research Agency.
406: \end{acknowledgments}
407:
408:
409: \begin{thebibliography}{}
410:
411:
412: \bibitem{bonetto}
413: F. Bonetto, J.L. Lebowitz and L. Rey-Bellet,
414: in {\it Mathematical Physics 2000} (London: Imp. Coll. Press, 2000)
415: pp. 128.
416:
417: \bibitem{lepri}
418: S. Lepri, R. Livi and A. Politi,
419: {\em Phys. Rep. \/} {\bf 377} (2003), 1.
420:
421: \bibitem{alonso}
422: D. Alonso, R. Artuso, G. Casati and I. Guarneri,
423: {\em Phys. Rev. Lett.} {\bf 82} (1999), 1859.
424:
425: \bibitem{dhar}
426: A. Dhar and D. Dhar, {\em Phys. Rev. Lett.} {\bf 82} (1999), 480.
427:
428: \bibitem{mejia} C. Mejia-Monasterio, H. Larralde, and F. Leyvraz, {\em Phys.
429: Rev. Lett.\/} {\bf 86} (2001), 5417; H. Larralde, F. Leyvraz, and C.
430: Mejia-Monasterio, {\em J. Stat. Phys.\/} {\bf 113} (2003), 197.
431:
432: \bibitem{liwang}
433: B. Li and J. Wang,
434: {\em Phys. Rev. Lett.} {\bf 91} (2003), 044301.
435:
436: \bibitem{terraneo}
437: M. Terraneo, M. Peyrard and G. Casati,
438: {\em Phys. Rev. Lett.\/} {\bf 88} (2002), 094302.
439:
440: \bibitem{li-2}
441: B. Li, L. Wang and G. Casati,
442: {\em Phys. Rev. Lett.\/} {\bf 93} (2004), 184301.
443:
444: \bibitem{li-3} B. Li, J.-H. Lan and L. Wang, {\em Phys. Rev. Lett.\/} {\bf 95}
445: (2005), 104302.
446:
447: \bibitem{li-4}
448: B. Li, L. Wang and G. Casati,
449: {\em Appl. Phys. Lett\/} {\bf 88} (2006), 143501.
450:
451: \bibitem{segal}
452: D. Segal and A. Nitzan,
453: {\em Phys. Rev. Lett.\/} {\bf 94} (2005), 034301.
454:
455: \bibitem{peyrard}
456: M. Peyrard,
457: {\em Europhys Lett.\/} {\bf 76} (2006), 49.
458:
459: \bibitem{EMM-2}
460: J.-P. Eckmann, C. Mejia-Monasterio
461: {\em Phys. Rev. Lett.\/} {\bf 97} (2006) 094301.
462:
463: \bibitem{sign} The particles that are transmitted to the right cell bounce
464: repeatedly with either the upper or the lower boundary of the billiard and
465: this depends on the sign of the magnetic field.
466: \end{thebibliography}
467:
468: \end{document}
469:
470:
471:
472:
473:
474:
475:
476:
477:
478:
479:
480:
481:
482:
483:
484:
485: Similar ideas has been discussed in \cite{li-3} for dissimilar anharmonic
486: lattices in terms of the thermal resistance at the interface.
487:
488: They considered an open billiard consisting of two cells of different geometry
489: connected through an opening and filled with a gas of effectively interacting
490: particles.
491:
492:
493: \bibitem{nanoscopics}
494: D. G. Cahill, {\em J. Appl. Phys. \/} {\bf 93} (2003), 793.
495:
496: \bibitem{rectifier}
497: P. W. O'Callaghan, S. D. Probert and A. Jones,
498: {\em J. Phys. D: Appl. Phys.} {\bf 3} (1970), 1352.
499:
500: \bibitem{mejia}
501: C. Mejia-Monasterio, H. Larralde, and F. Leyvraz,
502: {\em Phys. Rev. Lett.\/} {\bf 86} (2001), 5417.
503:
504: \bibitem{larralde}
505: H. Larralde, F. Leyvraz, and C. Mej\'\i{a}-Monasterio,
506: {\em J. Stat. Phys.\/} {\bf 113} (2003), 197.
507:
508: \bibitem{ey}
509: J.-P. Eckmann and L.-S. Young,
510: {\em Commun. Math. Phys.\/} {\bf 262} (2006), 237.
511:
512: \bibitem{em-1}
513: J.-P. Eckmann, C. Mej\'\i{a}-Monasterio and E. Zabey,
514: {\sf ArXiv:cond-mat/0510698}.
515: