1: %\documentclass[aps,epsfig,twocolumn,showpacs]{revtex4}
2: %\documentclass[aps,prl,twocolumn,floats,showpacs]{revtex4}
3: \documentclass[aps,epsfig,manuscript,showpacs]{revtex4}
4:
5: \usepackage{amsmath,amssymb,graphicx}
6: \usepackage{epsfig}
7: \usepackage{graphicx}
8: \usepackage{enumerate}
9: %\textwidth 150mm \oddsidemargin 5mm \evensidemargin 5mm
10: %\parindent 10mm
11: \newcommand{\beq}{\begin{equation}}
12: \newcommand{\eeq}{\end{equation}}
13: \newcommand{\eq}[1]{Eq. (\ref{#1})}
14: \newcommand{\bea}{\begin{eqnarray}}
15: \newcommand{\eea}{\end{eqnarray}}
16: \newcommand{\eqs}[1]{Eqs. (\ref{#1})}
17: \newcommand {\ket}[1]{|\,{#1}\,\rangle}
18: \newcommand {\bra}[1]{\langle\,{#1}\,|}
19:
20: %\topmargin 0.3cm
21:
22: \begin{document}
23:
24: %\baselineskip 10mm
25:
26: \title{Heat flow in nonlinear molecular junctions}
27:
28: \author{Dvira Segal}
29: \affiliation{Department of Chemical Physics, Weizmann
30: Institute of Science, 76100 Rehovot, Israel}
31:
32: \begin{abstract}
33:
34: We investigate the heat conduction properties of molecular
35: junctions comprising anharmonic interactions.
36: We find that nonlinear interactions can lead to novel phenomena:
37: {\it negative differential thermal conductance} and heat rectification.
38: Based on analytically solvable models we derive an expression for
39: the heat current that clearly reflects the interplay between
40: anharmonic interactions, strengths of coupling to the thermal
41: reservoirs, and junction asymmetry.
42: %
43: This expression indicates that negative differential thermal
44: conductance shows up when the molecule is strongly coupled to the
45: thermal baths, even in the absence of internal molecular
46: nonlinearities. In contrast, diode like behavior is expected for a
47: highly anharmonic molecule with an inherent structural asymmetry.
48: %
49: Anharmonic interactions are also necessary for manifesting
50: Fourier type transport. We briefly present an extension of our
51: model system that can lead to this behavior.
52:
53:
54:
55: \end{abstract}
56:
57: \pacs{63.20.Ry, 44.10.+i, 05.60.-k, 66.70.+f }
58: % 63.22.+m Phonons or vibrational states in low-dimensional structures and nanoscale materials
59: %44.10.+i Heat conduction (see also 66.60.+a and 66.70.+f in transport properties of condensed matter)
60: %05.60.-k Transport processes
61: %66.70.+f Nonelectronic thermal conduction and heat-pulse propagation in solids; thermal waves (for thermal conduction in metals and alloys, see 72.15.Cz and 72.15.Eb63.20.Ry Anharmonic lattice modes)
62: %{\today}
63: \maketitle
64: %------------------------------------------------------------------
65:
66:
67: \section{Introduction}
68:
69: Understanding and controlling heat flow in nanoscale structures
70: is of interest both from the fundamental aspect \cite{Schwab1} and for device
71: applications \cite{Cahill,Shi,Nanos,Inst}. The influential role of quantum
72: effects and geometrical constrictions in low dimensional systems
73: often results in fundamentally interesting behavior \cite{Blencowe}.
74: %
75: Recent theoretical and experimental studies demonstrated that the
76: thermal transport properties of nanowires can be very different
77: from the corresponding bulk properties \cite{Rego,Schwab}. In the
78: low temperature {\it ballistic} regime the phonon thermal
79: conductance of a one-dimensional (1D) quantum wire is quantized,
80: with $g=\pi^2k_B^2T/3h$ as the universal quantum conductance unit
81: \cite{Rego}, where $k_B$ and $h$ are the Boltzmann and Planck
82: constants, respectively, and $T$ is the temperature. Reflections
83: from the boundaries and disorder in the wire can be further
84: treated by considering a Landauer type expression for the heat
85: current ($\hbar=1$) \cite{Rego,Ciraci,Segalheat}
86: %
87: \beq J=\int d\omega \omega\mathcal{T} (\omega)\left[
88: n_L(\omega)-n_R(\omega)\right].
89: \label{eq:Landauer}
90: \eeq
91: %
92: This relationship describes energy transfer between two (left
93: ($L$) , right ($R$)) thermal reservoirs maintained at equilibrium
94: with the temperatures $T_L$ and $T_R$, respectively, in terms of
95: the {\it temperature independent} transmission coefficient
96: $\mathcal T(\omega)$ for phonons of frequency $\omega$. Here
97: $n_K(\omega)=\left(e^{\beta_K\omega}-1 \right)^{-1}$;
98: $\beta_K=1/k_BT_K$, $(K=L,R)$, is the Bose-Einstein distribution
99: characterizing the reservoirs. This expression assumes the absence
100: of inelastic scattering processes, and the two opposite phonon
101: flows of different temperatures are out of equilibrium with each
102: other. This leads to an {\it anomalous} transport of heat, where
103: (classically) the energy flux is proportional to the temperature
104: difference, $\Delta T=T_L-T_R$, rather than to the temperature
105: gradient, $\nabla T$, as asserted by the Fourier law of
106: conductivity,
107: %
108: \beq
109: J= - \mathcal{K}\mathcal{A}\nabla T.
110: \label{eq:Fourier}
111: \eeq
112: %
113: In this equation $\mathcal A$ is the cross section area normal to
114: the direction of heat propagation and $\mathcal K$ is the
115: coefficient of thermal conductivity. An outstanding problem in
116: statistical physics is to find out the necessary and sufficient
117: conditions for attaining this {\it normal} (Fourier) law of heat
118: conductivity in low dimensional systems
119: \cite{Livi,Joel,Prosen,LiWang,Gruber,Schr}. Among the crucial
120: requirements explored is that the molecular potential energy
121: should constitute strong anharmonic interactions.
122:
123: %
124: Heat conductance experiments on short molecules or highly ordered
125: structures provide results consistent with the Landauer
126: expression. A micron length individual carbon nanotube conducts
127: heat ballistically without showing signatures of phonon-phonon
128: scattering for temperatures up to 300 K
129: \cite{MajumdarCNT,ChiuCNT}. Intramolecular vibrational energy flow
130: in bridged azulene-anthracene compounds could be
131: explained by assuming ballistic energy transport in the chain
132: connecting both chromophores \cite{Schwarzer}. In contrast,
133: calculations of heat flow through proteins show substantial
134: contribution of anharmonic interactions leading to an enhancement
135: of the energy current in comparison to the (artificial) purely
136: harmonic situation \cite{Leitner}.
137:
138: %
139: Anharmonic (nonlinear) interactions are also a tool for {\it
140: controlling} heat flow in molecular junctions with potential
141: technological applications, e.g. a thermal diode
142: \cite{Casati1,Casati2,Rectif} and a thermal transistor
143: \cite{Casatixxx}. We have recently demonstrated that when
144: nonlinear interactions govern heat conduction, the heat current is
145: asymmetric for forward and reversed temperature biases, provided
146: the junction has some structural asymmetry \cite{Rectif}.
147:
148: %
149: In this paper we generalize the model developed in Ref.
150: \cite{Rectif}, and present a comprehensive analysis of the heat
151: conduction properties of molecular junctions taking into account
152: anharmonic interactions in the system. We discuss the influence of
153: the following effects on heat flow through the junction: (i)
154: interparticle potential, specifically the degree of molecular
155: anharmonicity, (ii) molecule-thermal reservoirs contact
156: interactions, and (iii) junction asymmetry with respect to the $L$
157: and $R$ ends. We derive an exact analytic expression for the heat
158: current that clearly reflects the role of each of these factors in
159: determining phonon dynamics.
160: %In the present study we therefore generalize the results of
161: %\cite{Rectif} and investigate heat flow through
162: %both harmonic and anharmonic molecular modes coupled either
163: %weakly or strongly to thermal environments.
164: %
165: More specifically, we analyze the necessary conditions for
166: demonstrating negative differential {\it thermal} conductance
167: (NDTC) and diode like behavior. We also follow the transition from
168: the elastic Landauer formula to the Fourier law of conduction as
169: anharmonic interactions are turned on.
170:
171:
172: %
173: The paper is organized as follows: Section II presents our model
174: system. Section III begins with a fully harmonic model and shows
175: that it satisfies the Landauer formula. We then proceed and show
176: that an asymmetric anharmonic molecule linearly coupled to thermal
177: reservoirs, can rectify heat. Section IV further presents
178: strong coupling models that exhibit
179: NDTC. Under additional conditions, the nonlinear
180: models extended to a $l$ sites system satisfy the Fourier law of
181: conductivity as explained in Section V. Section VI provides
182: concluding remarks.
183:
184:
185: %---------------------------------------
186: \section{Model}
187: The model system consists a molecular unit connecting two thermal
188: reservoirs left ($L$) and right ($R$) of inverse temperatures $\beta_L=T_L^{-1}$
189: and $\beta_R=T_R^{-1}$ respectively. Henceforth we take the
190: Boltzmann constant as $k_B=1$.
191: %
192: The general Hamiltonian includes three contributions: the
193: molecular part ($M$), the two reservoirs ($B$) and the system-
194: bath interaction ($MB$)
195: %
196: \beq
197: H= H_M +H_B +H_{MB}.
198: \label{eq:HT}
199: \eeq
200: %
201: %-------------------
202: For simplicity we assume that heat
203: transfer is dominated by a specific single mode.
204: The molecular term in the Hamiltonian is therefore given by
205: %
206: \beq
207: H_M=\sum_{n=0}^{N-1}E_n |n\rangle \langle n|; \,\,
208: E_n=n\omega_0,
209: \label{eq:H1M} \eeq
210: %
211: where $\omega_0$ is the frequency of the molecular oscillator
212: ($\hbar\equiv 1$). We shall consider two situations: harmonic
213: model, and a two-level system (TLS) that simulates a highly
214: anharmonic vibrational mode. For a harmonic molecule $N$ is taken
215: up to infinity. Strong anharmonicity is enforced by limiting $n$
216: to $0,1$. The molecular mode is coupled either linearly (weakly)
217: or nonlinearly (strongly) to the $L$ and $R$ thermal baths
218: represented by sets of independent harmonic oscillators
219: %
220: \beq
221: H_B=H_L+H_R; \ \ H_K=\sum_{j\in K}{\omega_j a_j^\dagger
222: a_j} ;\ \ K=L,R.
223: \label{eq:H1B}
224: \eeq
225: %
226: $a_j^{\dagger}$, $a_j$ are boson creation and annihilation
227: operators associated with the phonon modes of the harmonic baths.
228: The $L$ and $R$ thermal baths are not coupled directly, only
229: through their interaction with the molecular mode.
230: We use the following model for the molecule-reservoirs interaction
231: %
232: \bea H_{MB}&=&\sum_{n=1}^{N-1} \left(B|n-1\rangle \langle n|
233: +B^{\dagger}|n\rangle \langle n-1| \right)\sqrt{n},
234: \label{eq:H1MB}
235: \eea
236: %
237: where $B$ are bath operators. This model assumes that transitions
238: between molecular levels occur due to the environment excitations.
239: Note that in general this interaction does not need to be additive
240: in the thermal baths, i.e. we may consider situations in which $B
241: \neq B_L+B_R$, see Section IV.
242: %
243:
244: The probabilities $P_n$ to occupy the $n$ state of the molecular
245: oscillator satisfy the master equation
246: %
247: \bea
248: \dot P_n&=&-\left[ n k_d + (n+1) k_u \right] P_n
249: \nonumber\\
250: &+&(n+1)k_dP_{n+1}+n k_uP_{n-1},
251: \label{eq:master}
252: \eea
253: %
254: where the occupations are normalized $\sum_n P_n=1$. Within second
255: order perturbation theory the rates are given by
256: %
257: \bea
258: k_d&=& \int_{-\infty}^{\infty} d\tau e^{i \omega_0 \tau}
259: \langle B^{\dagger}(\tau) B(0) \rangle, \,\,\,\
260: \nonumber\\
261: k_u&=& \int_{-\infty}^{\infty} d \tau e^{-i \omega_0 \tau} \langle
262: B(\tau) B^{\dagger}(0) \rangle,
263: \label{eq:kdku}
264: \eea
265: %
266: where the average is done over the baths thermal distributions,
267: irrespective of the fact that it may involve two distributions of
268: different temperatures. In Section IV we demonstrate that these
269: rates also apply in the strong molecule-baths interaction limit.
270:
271: %------------
272: A useful concept in the following discussion is the notion of an effective
273: molecular temperature. It can be defined through the relative
274: population of neighboring molecular levels
275: %
276: \beq T_{M}\equiv -\frac{\omega_0}{\log (P_{n+1}/P_n)}.
277: \label{eq:TM}
278: \eeq
279: %
280: At steady state this ratio does not depend on $n$, see
281: Eq.~(\ref{eq:master}). We show below that the molecular temperature
282: is given in terms of the reservoirs temperatures weighted by the
283: molecule-baths coupling strengths.
284:
285: %------------
286: Given the reservoirs temperatures $T_L$ and $T_R$, we can define
287: two other related parameters: the temperature difference $\Delta
288: T=T_L-T_R$ and the average temperature $T_a=(T_L+T_R)/2$. The
289: temperature difference can be experimentally imposed in various ways. Here we
290: consider two situations: We may fix the temperature at the left
291: reservoir while varying the temperature at the right side,
292: %
293: \bea
294: {\rm (A)} \,\,\, T_L&=&T_s
295: \nonumber\\
296: T_R&=&T_s-\Delta T. \label{eq:modelA} \eea
297: %
298: For the same temperature difference we can also build a
299: symmetric situation where the temperatures of both reservoirs are
300: equally shifted,
301: %
302: \bea
303: {\rm (B)} \,\,\, T_L&=&T_s+\Delta T/2
304: \nonumber\\
305: T_R&=&T_s-\Delta T/2.
306: \label{eq:modelB}
307: \eea
308: %
309: The main difference between these two situations is that the
310: average temperature is decreasing steadily with $\Delta T$
311: in the first case, while it is constant ($T_s$) in (B).
312: We will show below that these boundary conditions determine
313: the effective {\it molecular}
314: temperature which implies on the conduction properties of the system.
315:
316: Next we present the model Hamiltonians in the weak and strong
317: molecule-bath interaction limits for either purely harmonic or a
318: TLS molecular mode, and discuss the implications on the junction
319: thermal conductance.
320:
321: %--------------------------------------
322:
323: \section{Weak system-bath coupling}
324: We begin by analyzing the heat conduction properties of a molecule
325: coupled {\it linearly} to two thermal reservoirs of different
326: temperatures \cite{Rectif}. The Hamiltonian is given by Eqs.
327: (\ref{eq:HT})-(\ref{eq:H1MB}) with linear (harmonic)
328: system-bath interactions
329: %
330: %
331: \bea H_{MB}&=&\sum_{n=1}^{N-1} \left(B|n-1\rangle \langle n|
332: +B^{\dagger}|n\rangle \langle n-1| \right)\sqrt{n};
333: \nonumber\\
334: B&=&B_L+B_R, \label{eq:HMB} \eea
335: %
336: where the bath operators $B_K$ satisfy
337: %
338: \bea
339: B_K&=& \sum_{ j \in K}{\bar{\alpha}_j x_j}; \ \
340: \nonumber\\
341: x_j&=&(2\omega_j)^{-1/2}({a}_j^{\dagger}+a_j) ;\ \ K=L,R.
342: \label{eq:H1in}
343: \eea
344: %
345: In the present linear coupling model, no correlations persist
346: between the thermal baths and the rate constants (\ref{eq:kdku})
347: are additive in the $L$ and $R$ baths
348: %
349: \bea
350: k_d&=&k_L+k_R
351: \nonumber\\
352: k_u&=&k_L e^{-\beta_L \omega_0}+k_R
353: e^{-\beta_R \omega_0},
354: \label{eq:rate1}
355: \eea
356: %
357: with
358: %
359: \beq k_K=\Gamma_K(\omega_0) (1+n_K(\omega_0)) ; \ \ K=L,R.
360: \label{eq:rate2} \eeq
361: %
362: Here $n_K(\omega)=\left( e^{\beta_K\omega}-1\right)^{-1}$,
363: $\Gamma_K(\omega)= \frac{\pi}{2m\omega^2} \sum_{j \in K}
364: \alpha_j^2\delta(\omega-\omega_j)$ and $\alpha_j=\bar\alpha_j
365: \sqrt{2 m\omega_0}$ \cite{Rectif}, where $m$ and $\omega_0$ are
366: the molecular oscillator mass and frequency respectively.
367: %
368:
369: The heat conduction properties of this model are obtained from the
370: steady state solution of Eq.~(\ref{eq:master}) with the rates
371: specified by Eqs.~(\ref{eq:rate1})-(\ref{eq:rate2}). The
372: steady-state heat flux calculated, e.g. at the right contact, is
373: given by the sum
374: %
375: \bea J=\omega_0 \sum_{n=1}^{N-1} n \left( k_R P_n -k_RP_{n-1}
376: e^{-\beta_R \omega_0} \right),
377: \label{eq:Curr} \eea
378: %
379: where positive sign indicates current going from left to right.
380:
381: %-------------------------------------
382:
383: \subsection{Harmonic molecule}
384: For the harmonic model ($N \rightarrow \infty $),
385: putting $\dot{P}_n=0$ in (\ref{eq:master}), % for all $n$
386: and searching a solution of the form $P_n \propto y^n$ we get a
387: quadratic equation for $y$ whose physically acceptable solution is
388: %
389: \beq
390: y=\frac{k_L e^{-\beta_L \omega_0} + k_R e^{-\beta_R \omega_0}
391: }{k_L+k_R}, \label{eq:yy} \eeq
392: %
393: which leads to the normalized state population
394: %
395: \beq P_n=y^n(1-y). \eeq
396: %.
397: Using Eq. (\ref{eq:rate2}) we obtain the heat current
398: (\ref{eq:Curr})
399: %
400: \beq J=\omega_0 \frac{\Gamma_L \Gamma_R}{\Gamma_L+\Gamma_R} \left(
401: n_L-n_R \right).
402: \label{eq:J1}
403: \eeq
404: %
405: In the classical limit, $\omega_0/T_K \ll 1$ ($K=L,R$), it reduces to
406: %
407: \beq J=
408: \frac{\Gamma_L\Gamma_R}{\Gamma_L+\Gamma_R} \left( T_L-T_R \right).
409: \label{eq:J1c}
410: \eeq
411: %
412: This is a special case (with $\mathcal{T}(\omega)= \Gamma_L
413: \Gamma_R(\Gamma_L+\Gamma_R)^{-1} \delta(\omega-\omega_0) $
414: consistent with our resonance energy transfer assumption)
415: \cite{general} of the Landauer expression, Eq.~(\ref{eq:Landauer}).
416: It is also consistent with the standard expression for the heat current
417: through a perfect harmonic chain \cite{Lebowitz}.
418:
419: We emphasize on three important features of this result: (i) The
420: heat current depends (classically) on the temperature difference
421: between the two reservoirs, leading to divergent heat
422: conductivity. Note that there is no need to introduce here the
423: concept of the molecular temperature $T_M$. (ii) The current is
424: the same when exchanging $\Gamma_L$ by $\Gamma_R$, i.e.
425: rectification cannot take place. (iii) The system cannot show the
426: NDTC behavior, i.e. it is impossible to observe a decrease of the
427: current with increasing temperature difference. This is true
428: considering both
429: models for the temperature drop- A and B,
430: (\ref{eq:modelA})-(\ref{eq:modelB}), irrespective of the system
431: symmetry. We can verify it by studying the $\Delta T$ derivative
432: of the current (\ref{eq:J1})
433: %
434: \bea \frac{\partial J}{\partial \Delta T}\propto \frac{\partial
435: n_L}{\partial \Delta T} - \frac{\partial
436: n_R}{\partial \Delta T}
437: =
438: \frac{\partial n_L}{\partial \Delta T} + \frac{\partial
439: n_R}{\partial (-\Delta T)}.
440: \eea
441: %
442: Since the term
443: %
444: \beq
445: \frac{\partial n_{L}}{\partial \Delta T}=
446: \frac { \omega_0 e^{\beta_{L} \omega_0 }}
447: {T_{L}^2 \left( e^{\beta_{L} \omega_0}-1 \right)^2 }
448: \frac{\partial T_L}{\partial \Delta T}
449: \eeq
450: %
451: is always positive (or zero), and similarly the second right hand
452: side term, NDTC cannot show up in the fully harmonic
453: model, and the heat current increases monotonically with the
454: temperature difference.
455:
456: %----------------------------------------------------
457:
458: \subsection{Anharmonic molecule}
459:
460: We proceed to the case of a highly anharmonic molecule coupled
461: -possible asymmetrically- but linearly, to two thermal reservoirs
462: of different temperatures. We simulate strong anharmonicity by
463: modeling the molecular mode by a two levels system (TLS). The
464: Hamiltonian for this model and the resulting rates are
465: the same as presented throughout Eqs.
466: (\ref{eq:HT})-(\ref{eq:Curr}), except that we take $n$=0,1 only.
467: Following Eqs. (\ref{eq:master})-(\ref{eq:kdku}) we obtain the
468: steady state levels population
469: %
470: \beq P_1=\frac{k_u}{k_u+k_d}; \,\,\,\, P_0=\frac{k_d}{k_u+k_d}.
471: \label{eq:P12} \eeq
472: %
473: We substitute it into Eq. (\ref{eq:Curr}) with $N$=2 and find the heat current
474: \cite{Rectif}
475: %
476: \beq J=\omega_0 \frac{\Gamma_L \Gamma_R
477: (n_L-n_R)}{\Gamma_L(1+2n_L)+\Gamma_R(1+2n_R)}.
478: \label{eq:J2}
479: \eeq
480: %
481: Next we calculate the molecular temperature $T_M$ in the weak
482: coupling-TLS case by substituting the population
483: (\ref{eq:P12}) into Eq. (\ref{eq:TM}) using Eqs.
484: (\ref{eq:rate1})-(\ref{eq:rate2}). In the classical limit this
485: results in
486: %
487: \beq T_M= \frac{\Gamma_L T_L +\Gamma_R T_R}{\Gamma_L+\Gamma_R}
488: \label{eq:TM1}. \eeq
489: %
490: We can now study the implications of the different models for the
491: temperature bias, Eqs. (\ref{eq:modelA})-(\ref{eq:modelB}), on the
492: conductance: In Model A the molecular temperature decreases
493: monotonically with the temperature difference
494: %
495: \beq T_{M}^{(A)}=T_s-\Delta T \frac{\Gamma_R}{\Gamma_L+\Gamma_R}.
496: \label{eq:TMA} \eeq
497: %
498: In Model B we find
499: %
500: \beq T_M^{(B)}=T_s+\frac{\Delta T}{2}
501: \frac{\left(\Gamma_L-\Gamma_R\right)}{\Gamma_L+\Gamma_R},
502: \label{eq:TMB} \eeq
503: %
504: which implies that for a symmetric ($\Gamma_L=\Gamma_R$) system,
505: the molecular temperature is constant, whereas in the asymmetric
506: situation it can either increase or decrease with $\Delta T$, depending on the
507: sign of $\Gamma_L-\Gamma_R$.
508: %These observations hold also
509: %in the harmonic case, but the concept was not relevant there.
510: %----------------
511:
512: In terms of the molecular temperature (Eq. (\ref{eq:TM1})), going
513: into the classical limit, the heat current (\ref{eq:J2}) reduces into
514: the simple form
515: %
516: \beq
517: J= (T_L-T_R) \frac{\Gamma_L \Gamma_R}{\Gamma_L+\Gamma_R} \frac{
518: \omega_0}{2T_M}. \label{eq:J2c}
519: \eeq
520: %
521: This relationship differs from the harmonic expression,
522: (\ref{eq:J1c}), by its implicit dependence on the internal
523: molecular temperature. As we show next, this opens up the door for
524: heat rectification and can also lead to the applicability of the
525: Fourier law of conduction. If we still try to fit this expression
526: into the Landauer form (\ref{eq:Landauer}), we find that we have
527: to define an effective {\it temperature dependent} transmission
528: coefficient $\mathcal{T}(\omega,T_L,T_R)\propto
529: 1/T_M\delta(\omega-\omega_0)$.
530:
531: In Fig. \ref{FigM2a} we display the current (Eq. (\ref{eq:J2}))
532: for a representative set of parameters. It increases monotonically
533: with $\Delta T$ and saturates at high temperature gaps. We can
534: verify this trend analytically as
535: %
536: \bea
537: \frac{\partial J}{\partial \Delta T}=
538: %\nonumber\\
539: \left[\frac {\partial n_L}{\partial \Delta T}
540: (1+2n_R) - \frac{\partial n_R}{\partial \Delta T}
541: (1+2n_L) \right] \times
542: \nonumber\\
543: \frac{\omega_0 \Gamma_L \Gamma_R(\Gamma_L+\Gamma_R)}{\left[ \Gamma_L(1+2n_L) +\Gamma_R(1+2n_R)\right]^2} >0,
544: \eea
545: %
546: which indicates that NDTC can not take place. However, Eq.~(\ref{eq:J2})
547: implies that the system can rectify heat current, i.e. the current
548: can be different (in absolute values) when exchanging the reservoirs
549: temperatures. Following \cite{Rectif}, defining the asymmetry
550: parameter $\chi$ such that $\Gamma_L=\Gamma(1-\chi)$ ;
551: $\Gamma_R=\Gamma(1+\chi)$ with $-1 \leq \chi \leq 1$ we get
552: %
553: \bea \Delta J
554: &\equiv& J(T_L=T_h; T_R=T_c) +J(T_L=T_c; T_R=T_h)
555: \nonumber\\
556: &=&\frac{\omega_0 \Gamma \chi (1-\chi^2)
557: (n_L-n_R)^2}{(1+n_L+n_R)^2-\chi^2(n_L-n_R)^2}. \label{eq:J2R} \eea
558: %
559: Here $T_c$ ($T_h$) relates to the cold (hot) bath.
560: Eq.~(\ref{eq:J2R}) implies that for small $\Delta T=T_L-T_R$$,
561: \Delta J$ grows like $\Delta T^2$, and that the current is larger
562: (in absolute value) when the cold bath is coupled more strongly to
563: the molecular system. We exemplify this behavior at the inset of
564: Fig. \ref{FigM2a}.
565:
566: %
567: We found therefore that a system consisting of an anharmonic
568: molecular mode coupled linearly (harmonically) and asymmetrically
569: to two thermal reservoirs of different temperatures can rectify
570: heat, though it cannot manifest the NDTC effect. NDTC requires
571: anharmonic interactions with the thermal baths, which may result
572: in an effective nonlinear temperature dependent molecule-bath
573: coupling term, see section IV. Therefore, there is no direct
574: correspondence between these two phenomena.
575:
576: %------------------------
577: \begin{figure}
578: {\hbox{\epsfxsize=80mm \epsffile{1.ps}}}
579: %\vspace*{53mm}
580: \caption{
581: Conduction properties of a TLS system in the weak coupling limit.
582: $\omega_0$=150 meV (full), 100 meV (dashed), 25 meV (dotted).
583: $T_s$=400 K (Model A), $\Gamma_K$=1.2 meV.
584: Inset: Rectifying behavior of this model,
585: $\omega_0$=25 meV, $\chi$=0.75 and $T_L$=400 K, $T_R=T_L-\Delta T$
586: (full); $T_R$=400 K, $T_L=T_R-\Delta T$ (dashed). }
587: \label{FigM2a}
588: \end{figure}
589: %------------------------
590:
591:
592: %---------------------------------
593: \subsection{General expression for the heat current}
594:
595: We can generalize the harmonic (\ref{eq:J1c}) and anharmonic (\ref{eq:J2c})
596: results and revise the current in the weak coupling limit (${W}$) as
597: %
598: \beq
599: J_W= \omega_0 \frac{\Gamma_L \Gamma_R}{\Gamma_L+\Gamma_R} \frac{ (T_L-T_R)
600: }{T_M} f_{S,B}, \label{eq:J2l}
601: \eeq
602: %
603: where
604: %
605: \bea f_{S,B}=
606: \begin{cases}
607: 1/2, & \rm { Anharmonic \,\, TLS \,\ case\,\, ( "Spin")} \\
608: T_M/\omega_0. & \rm { Harmonic\,\, case\,\, ("Boson")} \\
609: \end{cases}
610: \eea
611: %
612: For an intermediate anharmonicity we expect this function to
613: attain an intermediate value, $1/2<f_{S,B}<T_M/\omega_0$. Note
614: that $f_{S,B}$ can be retrieved by going into the classical limit
615: of $f_{S,B}=\left[\exp( \omega_0/T_M)\pm1\right]^{-1}$. Here the
616: "spin" case takes the plus sign, and the "boson" situation
617: acquires the minus. It can be therefore interpreted as an effective
618: molecular occupation factor.
619:
620: We can now clearly trace the influence of the different factors on
621: the heat conductance. The thermal current is given by multiplying
622: three terms: (i) A {\it symmetric} prefactor that includes the
623: influence of the system-baths coupling, (ii) the factor $\omega_0/
624: T_M$ which includes internal molecular properties: frequency and
625: effective temperature, and (iii) the molecular occupation factor
626: $f_{S,B}$ that varies between $1/2$ for the strictly anharmonic
627: system and $T_M/\omega_0$ in the harmonic case. As we show next,
628: the energy current has the same structure when system-bath
629: interactions are strong.
630:
631: %------------------------------------------------
632: \section {Strong system-bath coupling}
633:
634: We turn now to the situation where the molecular mode is {\it
635: strongly} coupled to the thermal reservoirs. As before, we
636: discuss two limits, the harmonic case, and the anharmonic TLS situation.
637: In both limits the model Hamiltonian includes the following terms,
638: as in Eqs. (\ref{eq:HT})-(\ref{eq:H1MB}),
639: %
640: \bea
641: {H}&=&
642: \sum_{n=0}^{N-1}E_n|n\rangle \langle n|
643: \nonumber\\
644: &+& \sum_{n=1}^{N-1}\sqrt n V_{n-1,n}|n-1\rangle \langle n|e^{i
645: (\Omega_{n}-\Omega_{n-1})}+ c.c.
646: %\nonumber\\
647: %&+& V_{n+1,n}|n+1\rangle \langle n|e^{-i(\Omega_{n+1}-\Omega_n)}
648: \nonumber\\
649: & +& \sum_{j \in L,R} \omega_j
650: a_j^{\dagger}a_j,
651: %+H_{shift},
652: \label{eq:H3Tn}
653: \eea
654: %
655: where $E_n=n\omega_0$,
656: $\Omega_n=\Omega_n^L+ \Omega_n^R$ and $\Omega_n^K=i \sum_{j \in
657: K} \lambda_{n,j}\left( a_j^{\dagger} -a_j \right)$ ($K=L,R$).
658: %
659: In Appendix A we demonstrate that this model Hamiltonian
660: equivalently represents a displaced molecular mode coupled
661: nonlinearly to two thermal reservoirs. The coefficients
662: $\lambda_{n,j}$ are the effective system-bath interaction
663: parameters that depend on the level index and the reservoir mode,
664: see Appendix A. The Hamiltonian (\ref{eq:H3Tn}) is similar to that
665: defined in Eqs. (\ref{eq:H1M})-(\ref{eq:H1in}), except that the
666: $L$ and $R$ system-baths couplings appear as multiplicative rather
667: than additive factors in the interaction term, implying
668: non-separable transport at the two contacts \cite{Rectif}. The
669: dynamics is still readily handled. For small $V$ (the
670: "non-adiabatic limit") the Hamiltonian (\ref{eq:H3Tn}) leads again
671: to the rate equation (\ref{eq:master}) with
672: %
673: \beq k_d=|V|^2 C(\omega_0);\ \ k_u=|V|^2
674: C(-\omega_0), \label{eq:kdM3}
675: \eeq
676: %
677: where
678: %
679: \beq C(\omega_0)=\int_{-\infty}^{\infty}dt e^{i\omega_0 t}
680: \tilde{C}(t), \eeq
681: %
682: and
683: %
684: \bea \tilde{C}(t)&=& \left< e^{i[\Omega_{n}(t)-\Omega_{n-1}(t)
685: ]}e^{-i[\Omega_{n}(0)-\Omega_{n-1}(0)]} \right> =\left<
686: e^{i[\Omega_{n}^L(t) - \Omega_{n-1}^L(t) ]} e^{-i[ \Omega_{n}^L
687: -\Omega_{n-1}^L] } \right>_L
688: \nonumber\\
689: & \times & \left< e^{i[\Omega_{n}^R(t) - \Omega_{n-1}^R(t) ]}
690: e^{-i[ \Omega_{n}^R -\Omega_{n-1}^R] } \right>_R.
691: \eea
692: %
693: This may be evaluated explicitly to produce
694: %
695: \beq \tilde{C}(t)=\tilde{C}_L(t)\tilde{C}_R(t); \ \
696: \tilde{C}_K(t)=\exp(-\phi_K(t)),
697: \eeq
698: %
699: with
700: %
701: \bea \phi_K(t)=\sum_{j \in K}( \lambda_{n,j}- \lambda_{n-1,j})^2 [
702: \left( 1+2n_K(\omega_j) \right)
703: \nonumber\\
704: - \left(
705: 1+n_K(\omega_j) \right)e^{-i \omega_j t} -
706: n_K(\omega_j)e^{i\omega_j t} ].
707: \eea
708: %
709: Note that we have omitted the $n$ dependence from the rates above.
710: This is supported by (i) taking all the inter-levels couplings to
711: be equal, i.e. $|V_{n-1,n}|=V$, and (ii) assuming that
712: $(\lambda_{n,j}-\lambda_{n-1,j})^2$ is the same for all $n$ , e.g.
713: $\lambda_{n,j}\propto n$, see Appendix A.
714:
715: Explicit expressions may be obtained using the short time
716: approximation (valid for $\sum_{j\in K}(\lambda_{n,j}-
717: \lambda_{n-1,j})^2 \gg 1$ and/or at high temperature) whereupon
718: $\phi(t)$ is expanded in powers of $t$ keeping terms up to order
719: $t^2$. This leads to
720: %
721: \beq C(\omega_0)=\sqrt{ \frac{2 \pi}{(D_L^2+D_R^2)}} \exp\left[
722: \frac{-(\omega_0-E_M^L-E_M^R)^2}{2(D_L^2+D_R^2)} \right],
723: \label{eq:C3}
724: \eeq
725: %
726: where
727: %
728: \bea E_M^K&=&\sum_{j \in K} (\lambda_{n,j}- \lambda_{n-1,j} )^2
729: \omega_j, \,\,\,
730: \nonumber\\
731: D_K^2&=&\sum_{j \in K} (\lambda_{n,j}- \lambda_{n-1,j} )^2
732: \omega_j^2\left( 2n_K(\omega_j)+1 \right). \label{eq:DK2} \eea
733: %
734: $E_M^K$ can be considered as the reorganization energy associated
735: with the structural distortions of reservoirs modes around the
736: isolated molecular vibration. In the classical limit
737: ($\omega_0/T_K\rightarrow0$), $D_K^2=2T_KE_M^K$.
738:
739: Following Ref. \cite{Rectif} we calculate the steady state heat
740: current utilizing
741: %
742: \begin{eqnarray}
743: J=|V|^2 \sum_{n=1}^{N-1}\int_{-\infty}^{\infty} d\omega \omega [
744: C_R(\omega)C_L(\omega_0-\omega )P_n
745: \nonumber\\
746: -C_R(-\omega)C_L(-\omega_0+\omega)P_{n-1} ]n,
747: \label{eq:JM3}
748: \end{eqnarray}
749: %
750: where
751: \bea
752: C(\omega_0)&=&\int_{-\infty}^{\infty}d\omega
753: C_L(\omega_0-\omega)C_R(\omega),
754: \nonumber\\
755: C_K(\omega)&=&\frac{1}{\sqrt{2E_M^KT_K}}e^{-(\omega-E_M^K)^2/4T_KE_M^K}.
756: \eea
757: %
758: Eq. (\ref{eq:JM3}) views the process $|n \rangle
759: \rightarrow |n-1 \rangle $ in which the molecular mode looses
760: energy $\omega_0$ as a combination of processes in which the
761: system gives energy $\omega$ (or gains it if $\omega<0$ ) to the
762: right bath and energy $\omega_0-\omega $ to the left one, with
763: probability $nC_L(\omega_0-\omega)C_R(\omega)$. A similar analysis
764: applies to the process $|n-1\rangle \rightarrow |n\rangle $.
765:
766:
767: %----------------------------------------
768: \subsection{Harmonic molecule}
769:
770: The levels population of an harmonic molecule ($N\rightarrow
771: \infty$) are calculated from the steady state solution of Eq.
772: (\ref{eq:master}), leading to $P_n=y^n(1-y)$, $y=k_u/k_d$, with
773: the rates conveyed by Eqs. (\ref{eq:kdM3})-(\ref{eq:DK2}). The
774: heat current (\ref{eq:JM3}) is computed by first making the
775: summation over $n$
776: %
777: \bea \sum_{n=0}^{\infty} nP_n=
778: \frac{C(-\omega_0)}{C(\omega_0)-C(-\omega_0)},
779: \nonumber\\
780: \sum_{n=0}^{\infty} nP_{n-1} =\frac{C(\omega_0)}{C(\omega_0)-C(-\omega_0)}.
781: \eea
782: %
783: Next, performing the integrals over frequency yields
784: %
785: \bea J&=&\frac{2\sqrt{\pi} |V|^2 E_M^LE_M^R
786: (T_L-T_R)}{(E_M^LT_L+E_M^RT_R)^{3/2}}
787: \nonumber\\
788: &\times&
789: e^{-(\omega_0-(E_M^L+E_M^R))^2/4(E_M^LT_L+E_M^RT_R)} \times f_B,
790: \label{eq:JC3}
791: \eea
792: %
793: where
794: \beq
795: f_B=\left[e^{\omega_0(E_M^L+E_M^R)/(E_M^LT_L+E_M^RT_R)}-1\right]^{-1}.
796: \eeq
797: %
798: Before we discuss the heat conduction properties of this model
799: we examine the anharmonic system.
800:
801: %------------------------------------------------
802: \subsection {Anharmonic molecule}
803: The anharmonic model is described by the Hamiltonian
804: (\ref{eq:H3Tn}) with $n=0,1$. The steady state current is
805: therefore obtained by reducing Eq.~(\ref{eq:JM3}) to
806: %
807: \begin{eqnarray}
808: J=|V|^2 \int_{-\infty}^{\infty} d\omega \omega [
809: C_R(\omega)C_L(\omega_0-\omega )P_1
810: \nonumber\\
811: -C_R(-\omega)C_L(-\omega_0+\omega)P_0 ].
812: \label{eq:JM4}
813: \end{eqnarray}
814: %
815: Here $P_0=C(\omega_0)/\left( C(\omega_0) + C(-\omega_0) \right)$
816: and $P_1=1-P_0$ are established from the steady state solution of
817: (\ref{eq:master}) with the rates given by (\ref{eq:kdM3}). By
818: following the same steps as for the harmonic system, the heat
819: current (\ref{eq:JM4}) is obtained as \cite{Rectif}
820: %
821: \bea J&=&\frac{2\sqrt{\pi} |V|^2 E_M^LE_M^R
822: (T_L-T_R)}{(E_M^LT_L+E_M^RT_R)^{3/2}}
823: \nonumber\\
824: &\times&
825: e^{-(\omega_0-(E_M^L+E_M^R))^2/4(E_M^LT_L+E_M^RT_R)} \times f_S,
826: \label{eq:JC4}
827: \eea
828: %
829: with the occupation factor
830: %
831: \beq
832: f_S=\left[e^{\omega_0(E_M^L+E_M^R)/(E_M^LT_L+E_M^RT_R)}+1\right]^{-1}.
833: \eeq
834: %
835: %--------------------------------------------------------
836: \begin{figure}
837: {\hbox{\epsfxsize=80mm \epsffile{2.ps}}}
838: %\vspace*{53mm}
839: \caption{
840: Conduction properties of the TLS system in the strong coupling limit
841: $T_R=T_L-\Delta T$ (model A),
842: $E_M^K$=300 meV, ($K=L,R$) $V$=1 meV, $\omega_0$=10 meV.}
843: \label{FigM4a}
844: \end{figure}
845:
846: \begin{figure}
847: {\hbox{\epsfxsize=80mm \epsffile{3.ps}}}
848: %\vspace*{53mm}
849: \caption{Rectification in the strong coupling limit for a TLS system.
850: $E_M$=300 meV, $V$=1 meV, $\omega_0$=10 meV,
851: $T_L$=300 K, $T_R=T_L-\Delta T$ (full); $T_R$=300 K, $T_L=T_R-\Delta T$
852: (dashed).}
853: \label{FigM4b}
854: \end{figure}
855:
856: %--------------------------------------------
857: \subsection{General expression for the heat current}
858:
859: Next the harmonic (\ref{eq:JC3}) and anharmonic results
860: (\ref{eq:JC4}) are reduced into a common form. We begin by
861: evaluating the internal molecular temperature (\ref{eq:TM}). In
862: the present strong coupling case, for both harmonic and anharmonic
863: molecular modes, it is given by (\ref{eq:kdM3})
864: %
865: \beq e^{-\omega_0/T_M}\equiv
866: P_{n+1}/P_n=\frac{C(-\omega_0)}{C(\omega_0)}. \eeq
867: %
868: Using Eq.~(\ref{eq:C3}) we obtain the explicit expression
869: %
870: \beq T_M= \frac{D_L^2+D_R^2}{2(E_M^L+E_M^R)} \stackrel
871: {\omega_0/T_K \rightarrow 0}{\longrightarrow}
872: \frac{(E_M^LT_L+E_M^RT_R)}{(E_M^L+E_M^R)}. \eeq
873: %
874: The effective temperature in the strong coupling limit is therefore given by
875: the algebraic average of the $L$ and $R$ temperatures weighted by
876: the coupling strengths, here conveyed by the reservoirs reorganization
877: energies.
878:
879: In terms of this quantity, we write
880: a general expression for the current in the strong ($S$) coupling limit as
881: %
882: \bea J_{S}&=& |V|^2 \sqrt {
883: \frac{4\pi}{T_M(E_M^L+E_M^R)} }
884: e^{-(\omega_0-E_M^L-E_M^R)^2/4T_M(E_M^L+E_M^R)}
885: \nonumber\\
886: &\times& \frac{E_M^LE_M^R}{E_M^L+E_M^R} \frac{T_L-T_R}{T_M} \times
887: f_{S,B}, \label{eq:JCg} \eea
888: %
889: where
890: %
891: \bea f_{S,B}&\equiv&
892: \left[e^{\omega_0(E_M^L+E_M^R)/(E_M^LT_L+E_M^RT_R)}\pm1\right]^{-1}
893: \nonumber\\
894: &=& \left(e^{\omega_0/T_M} \pm 1\right)^{-1}. \eea
895: %
896: The plus sign relates to the anharmonic "spin" case, the minus
897: stands for the harmonic "boson" situation.
898:
899: We analyze next the conduction properties of this model. Diode
900: like behavior is expected when $E_M^L\neq E_M^R$, since then the
901: resulting molecular temperature $T_M$ is not the same when
902: exchanging $T_L$ with $T_R$. Note that in the present strong
903: (nonlinear) coupling limit the molecule does not need to be strictly
904: anharmonic for demonstrating this behavior, in contrast to the weak
905: coupling situation.
906:
907: NDTC can also take place in the system,
908: depending on the system asymmetry and the specific model for the applied
909: temperature gradient.
910: When the temperature bias is applied
911: symmetrically at the $L$ and $R$ sides (model B, Eq.
912: (\ref{eq:modelB})), NDTC occurs for an {\it asymmetric}
913: $E_M^L\neq E_M^R$ system. In model A the molecular temperature
914: depends on $\Delta T$ even for a symmetric junction, providing NDTC.
915:
916: %
917: Figure \ref{FigM4a} depicts an example of NDTC behavior in the
918: system. The left reservoir is held at a constant temperature,
919: while the temperature of the $R$ reservoir is decreasing. We find
920: that up to $\Delta T=T_L-T_R \sim 100 $ K the current increases
921: with the temperature bias, while above it, i.e. for lower $T_R$,
922: the current goes down, and even diminishes (dotted line).
923:
924: We can also investigate the effect of asymmetrical contacts.
925: We define the asymmetry parameter $\chi$ such as
926: $E_M^L=E_M(1-\chi)$, $E_M^R=E_M(1+\chi)$, $0<\chi<1$. Figure
927: \ref{FigM4b} presents the heat current when
928: $\chi \neq 0$. (a) For small $\chi$ the current is almost the same
929: for both forward and reversed operation modes. (b) At
930: intermediate $\chi$ values we find that for $T_L=100$ K, $T_R$=300 K
931: there is a maximal heat flow (dashed), while for the reversed
932: operation ($T_R=100$ K, $T_L$=300 K) heat current is blocked
933: (full). (c) For a highly asymmetric system heat flows predominantly
934: in one direction.
935:
936: %\section{Overview of results}
937:
938: We can further formulate a general expression for the current that
939: holds in {\it both} strong and weak interaction regimes and for
940: either harmonic or anharmonic systems. For convenience, we copy
941: here the weak ($W$) {\it linear} coupling result (\ref{eq:J2l})
942: %
943: \beq
944: J_{W}= \omega_0 \frac{\Gamma_L \Gamma_R}{\Gamma_L+\Gamma_R} \frac{ (T_L-T_R)
945: }{T_M} f_{S,B}.
946: \eeq
947: %
948: Comparing it to Eq. (\ref{eq:JCg}) guides us to the compact
949: expression
950: %
951: \beq J=C \frac{f_{S,B}}{T_M} \Delta T .
952: \label{eq:JG}
953: \eeq
954: %
955: Here $C$ includes the contact contribution, which is different in
956: the weak and strong coupling regimes. It may depend on the
957: molecule-baths microscopic couplings, molecular vibrational
958: frequency and the reservoirs temperatures. It is not influenced by
959: the degree of molecular harmonicity which affects only the
960: "Spin-Boson" factor $f_{S,B}$. The temperature $T_M$ provides
961: the effective temperature of the molecular system that is
962: irrelevant in the fully harmonic case. We can therefore clearly
963: distinguish in this expression between the role of the system
964: harmonicity and the effect of molecule-baths interactions.
965: %
966:
967: %------------------------------------------
968:
969:
970: %-----------------------
971: \begin{figure}
972: %\vspace{0mm} \hspace{0mm}
973: {\hbox{\epsfxsize=90mm \epsffile{4.ps}}}
974: \vspace{0mm} \caption{(Color online) A schematic representation of
975: the $l$=4 molecular modes system with three intermediate thermal
976: baths.} \label{scheme}
977: \end{figure}
978: %------------
979:
980: \section{Fourier law of conduction}
981:
982: The validity of Fourier's law of heat conduction
983: (\ref{eq:Fourier}) in 1D lattices is an open issue \cite{Prosen}.
984: This law is a macroscopic consequence of ordinary diffusion at the
985: microscopic level. Here we extend our single mode model and
986: present a realization of a molecular chain that can lead to normal
987: Fourier conduction, provided that the molecule is highly
988: anharmonic, independently of the molecule-baths coupling
989: strengths. We emphasize that we bypass the difficult task of
990: showing how normal diffusion emerges in the system \cite{Schr},
991: and simply assume non-correlated hopping motion between molecular
992: units. Our sole mission here is to construct from the single mode
993: result (\ref{eq:JG}) a model that supplies normal conduction.
994:
995: Figure \ref{scheme} depicts the system: We envision an array of
996: $I+2$ local heat baths where heat transfers along the chain of
997: bath, single mode, bath, single mode...: $L\rightarrow B_1
998: \rightarrow B_2 \rightarrow ... \rightarrow B_I \rightarrow R$.
999: The intermediate thermal baths $B_i$ might be realized by large
1000: molecular groups where full thermalization to a local temperature
1001: occurs. This implies that anharmonic interactions govern the
1002: dynamics within the intermediate baths. We also assume that the
1003: temperature difference $T_L-T_R$ falls linearly on the system,
1004: i.e. the temperature at the $i$ bath is
1005: %
1006: \bea T_i=T_s+\delta T[i-(I+1)/2],\,\,\ \Delta T=\delta T (I+1),
1007: \eea
1008: %
1009: where $i=(0,1,2...I,I+1)$. The $L$ and $R$ bath are indiced by 0
1010: and $I+1$ respectively, and $\delta T$ is the temperature
1011: difference between neighboring reservoirs. The resulting effective
1012: temperature of a molecular mode located in between each two baths
1013: $i-1$ and $i$ is
1014: %
1015: \beq T_M^{(i)}=T_s+\delta T[2i-1-(I+1)]/2,
1016: \eeq
1017: %
1018: assuming equal coupling along the chain ($\Gamma_i=\Gamma$ or
1019: $E_M^i=E_M$).
1020:
1021: The conductance $\kappa_i$ of the unit $B_{i-1}\leftrightarrow
1022: B_{i}$ is defined through the relation
1023: $J=\kappa_i(T_{i}-T_{i-1})$. In the diffusional hopping regime,
1024: %i.e. when correlations between the units diminishes,
1025: the total conductance $\kappa_T$ is established by inversing the
1026: sum of all units resistances, $\kappa_T=\left(\sum_i
1027: 1/\kappa_i\right)^{-1}$. In the linear coupling- TLS model
1028: (\ref{eq:J2l}) it becomes
1029: %
1030: \beq \kappa_T=\frac{\omega_0\Gamma
1031: }{4}\left(\sum_{i=1}^{I+1}T_M^{(i)}\right)^{-1}= \frac{\Gamma
1032: \omega_0}{4T_sl},
1033: \eeq
1034: %
1035: where $l=I+1$ is the actual length of the system, given by the
1036: number of internal molecular modes. This equation implies the
1037: validity of the Fourier law of heat conduction, $J\propto
1038: (T_L-T_R)/l$. In addition, the prefactor depends on the inverse
1039: temperature $T_s^{-1}$ as expected in the high temperature limit
1040: \cite{Kittel} where interactions among phonons are dominant.
1041: %These interactions are
1042: %included here effectively by assuming that local thermal equilibrium
1043: %exists at the intermediate baths.
1044: The same result holds when the temperature falls mainly on the contacts
1045: whereas the temperature distribution along the central baths is almost constant.
1046: Strong molecule-bath interactions contribute a
1047: temperature dependent prefactor, but do not modify the $l^{-1}$ form assuming
1048: $\Delta T/T_s \ll 1$.
1049: A possible realization of this system is a fullerene polymer \cite{fullerene}.
1050: %-----------------------
1051:
1052: \section{Conclusions}
1053:
1054: Using a simple theoretical model we have investigated the effect
1055: of anharmonic interactions on heat flow through molecular
1056: junctions. Our general expressions for the heat current
1057: (\ref{eq:J2l}) and (\ref{eq:JCg}) clearly manifest the interplay
1058: between the system anharmonicity, system-bath coupling and
1059: junction asymmetry. We have also extended our model into a chain
1060: of $l$ local molecular sites, and indicated that anharmonicity is
1061: a crucial requirement for achieving normal conductance.
1062: %
1063: We have found that nonlinear interactions can lead to novel
1064: phenomena: negative differential thermal conductance and heat
1065: rectification. NDTC takes place when the molecular mode is
1066: strongly coupled to the thermal environment. In contrast, diode
1067: like behavior originates from the combination of
1068: substantial molecular anharmonicities with a structural asymmetry.
1069: Figure \ref{Figo} presents an overview of
1070: the different regimes studied, and the nonlinear effects observed in each case.
1071:
1072: %------------------------
1073: \begin{figure}
1074: {\hbox{\epsfxsize=70mm \epsffile{5.ps}}}
1075: %\vspace*{53mm}
1076: \caption{An overview of the parameter ranges providing negative differential
1077: thermal conductance (NDTC) and diode like behavior. The $x$ axis,
1078: (weak and strong coupling) relates to the system-bath interaction model.
1079: In order to obtain rectification the system should be asymmetric with respect
1080: to the $L$ and $R$ ends.}
1081: \label{Figo}
1082: \end{figure}
1083: %------------------------
1084:
1085: We would also like to draw an analogy between the nonlinear
1086: behavior discussed in this paper and some nonlinear effects
1087: discovered in molecular level {\it electron} carrying systems: The
1088: negative differential resistance observed in molecular films of
1089: C$_{60}$ could be explained due to a voltage dependent tunneling
1090: barrier \cite{Crommie}. Rectification of electron current was
1091: theoretically exhibited in one-dimensional asymmetric electronic
1092: conductors with screened electron-electron interactions
1093: \cite{Brend}.
1094: %
1095:
1096: Control of heat flow through molecules by employing nonlinear
1097: interactions might be useful for different applications. In
1098: molecular electronic local heating of nanoscale devices might
1099: cause structural instabilities undermining the junction integrity
1100: \cite{Heat}. Engineering good thermal contacts and cooling of the
1101: the junction \cite{SegalPump} are necessary for a stable and
1102: reliable operation mode. Control of vibrational energy transfer in
1103: molecules affects chemical processes, e.g. reaction pathways, bond
1104: breaking processes, and folding dynamics \cite{Leitner2}. Finally,
1105: we propose building technological devices based on heat flow, in
1106: analogy with electron current devices \cite{CasatiR}.
1107:
1108: \begin{acknowledgments}
1109: The author would like to thank Professor A. Nitzan for his comments.
1110: This project was supported by the Feinberg graduate school of
1111: the Weizmann Institute.
1112: \end{acknowledgments}
1113:
1114: %--------------------------------------------------------------------
1115:
1116: \renewcommand{\theequation}{A\arabic{equation}}
1117: \setcounter{equation}{0} % reset counter
1118: \section*{APPENDIX A: Microscopic model for the strong coupling Hamiltonian }
1119: % use *-form to suppress numbering
1120:
1121: The strong coupling Hamiltonian Eq.~(\ref{eq:H3Tn}) can be derived
1122: from the following microscopic model
1123: %
1124: \begin{eqnarray}
1125: H&=& \sum_{n=0}^{N-1}n\omega_0|n\rangle \langle n|+
1126: \sum_{j \in L,R}\frac{\omega_j^2}{2}\left( x_j- \sum_{n=0}^{N-1}\frac{
1127: n\alpha_{n,j}}{\omega_j^2}|n\rangle \langle n| \right)^2
1128: \nonumber\\
1129: &+& \sum_{n=1}^{N-1}\sqrt{n}V_{n,n-1}|n\rangle \langle n-1|
1130: + \sqrt{n}V_{n-1,n}|n-1\rangle \langle n|
1131: \nonumber\\
1132: &+&\sum_{j \in L,R} \frac{p_j^2}{2},
1133: \label{eq:AH3T}
1134: \end{eqnarray}
1135: %
1136: which describes a forced oscillator of frequency $\omega_0$
1137: strongly interacting with the $L$ and $R$ thermal baths. The
1138: nonlinear contributions are concealed in the second element of
1139: (\ref{eq:AH3T}) providing high order terms such as $\propto x_j
1140: x^2$, with $x$ as the molecular coordinate. Here $x_j$ and $p_j$
1141: are the displacement and momentum of the reservoirs harmonic modes
1142: with frequency $\omega_j$, $\alpha_{n,j}$ is the system-bath
1143: coupling parameter and $V_{n,n-1}$ is the effective inter-level
1144: matrix element. We can expand the quadratic term in
1145: (\ref{eq:AH3T}) and obtain
1146: %
1147: \begin{eqnarray}
1148: H&=& \sum_{n=0}^{N-1}n\left( \omega_0- \sum_{j\in L,R}x_j\alpha_{n,j}\right)
1149: |n\rangle\langle n|
1150: \nonumber\\
1151: &+& \sum_{n=1}^{N-1}\sqrt{n}V_{n,n-1}|n\rangle \langle n-1|
1152: + \sqrt{n}V_{n-1,n}|n-1\rangle \langle n|
1153: \nonumber\\
1154: &+&\sum_{j \in L,R} \omega_j a_j^{\dagger}a_j
1155: +\sum_{n=0}^{N-1}\sum_{j\in L,R}\frac{n^2\alpha_{n,j}^2}{2\omega_j^2}|n\rangle
1156: \langle n|.
1157: \end{eqnarray}
1158: %
1159: Here
1160: $x_j=(a_j^{\dagger}+a_j)/\sqrt{2 \omega_j}$
1161: and $p_j=i\sqrt{\omega_j/2}(a_j^{\dagger}-a_j)$.
1162: Use of the small polaron transformation \cite{Mahan},
1163: $\tilde{H}= UHU^{-1}$, leads to
1164: %
1165: \bea
1166: \tilde{H}&=&
1167: \sum_{n=0}^{N-1}n\omega_0|n\rangle \langle n| +\sum_{n=0}^{N-1}\sum_{j\in
1168: L,R}\frac{n^2\alpha_{n,j}^2}{2\omega_j^2}|n\rangle \langle
1169: n|+H_{shift},
1170: \nonumber\\
1171: &+& \sum_{n=1}^{N-1}\sqrt n V_{n-1,n}|n-1\rangle \langle n|e^{i
1172: (\Omega_{n}-\Omega_{n-1})}+ c.c.
1173: %\nonumber\\
1174: %&+& V_{n+1,n}|n+1\rangle \langle n|e^{-i(\Omega_{n+1}-\Omega_n)}
1175: \nonumber\\
1176: & +& \sum_{j \in L,R} \omega_j
1177: a_j^{\dagger}a_j,
1178: \label{eq:AH3Tn}
1179: \eea
1180: %
1181: where
1182: %
1183: \beq U=\Pi_{n=0}^{N-1} U_n, \,\,\
1184: U_n=\exp(-i\Omega_n|n\rangle \langle n|),
1185: \eeq
1186: %
1187: and where
1188: %
1189: \bea
1190: \Omega_n&=&\Omega_n^L+ \Omega_n^R, \,\,\ \Omega_n^K=i \sum_{j
1191: \in K} \lambda_{n,j}\left( a_j^{\dagger} -a_j \right),\,\ (K=L,R),
1192: \nonumber\\
1193: \lambda_{n,j}&=&(2\omega_j^3)^{-1/2}n\alpha_{n,j}.
1194: \eea
1195: %
1196: The term
1197: %
1198: \beq
1199: H_{shift}=-\frac{1}{2}\sum_{n=0}^{N-1}\sum_{j}\frac{ n^2 \alpha_{n,j}^2}{\omega_j^{2}}
1200: |n\rangle \langle n|
1201: \eeq
1202: %
1203: exactly cancels the $\propto n^2$ term in Eq. (\ref{eq:AH3Tn}),
1204: and we finally recover the strong coupling Hamiltonian (\ref{eq:H3Tn}).
1205: %-------------------------------------------------------------------
1206:
1207: \begin{thebibliography}{99}
1208:
1209: \bibitem{Schwab1}
1210: K. C. Schwab, M. L. Roukes,
1211: %Putting Mechanics into Quantum Mechanics
1212: Phys. Today {\bf 58}, 36 (2005).
1213:
1214: \bibitem{Cahill} D. G. Cahill, K. Goodson, A. Majumdar,
1215: J. Heat Transfer {\bf{124}}, 223 (2002).
1216: % Thermometry and Thermal Transport in Micro/Nanoscale Solid-State Devices and Structures
1217:
1218: \bibitem{Shi} L. Shi, A. Majumdar, J. Heat Transfer {\bf{124}}, 329 (2002).
1219: % Thermal Transport Mechanisms at Nanoscale Point Contacts
1220:
1221: \bibitem{Nanos}
1222: %Nanoscale thermal transport
1223: D. G. Cahill, W. K. Ford, K. E. Goodson, G. D. Mahan, A. Majumdar,
1224: H. J. Maris, R. Merlin, S. R. Phillpot, J. App. Phys. {\bf 93},
1225: 793 (2003).
1226:
1227: \bibitem{Inst}
1228: K. L. Ekinci, M. L. Roukes,
1229: %Nanoelectromechanical systems
1230: Rev. of Sci. Instrum. {\bf 76}, 061101 (2005).
1231:
1232: \bibitem{Blencowe}
1233: % Quantum electromechanical systems
1234: M. Blencowe,
1235: Phys. Rep. {\bf 395}, 159 (2004).
1236:
1237: \bibitem{Rego} L. G. C. Rego, G. Kirczenow,
1238: Phys. Rev. Lett. {\bf{81}}, 232 (1998).
1239:
1240: \bibitem{Schwab} K. Schwab, E. A. Henriksen, J. M. Worlock, M. L. Roukes, Nature {\bf{404}}, 974 (2000).
1241:
1242: \bibitem{Ciraci}
1243: %Quantum effects of thermal conductance through atomic chains
1244: A. Ozpineci, S. Ciraci,
1245: Phys. Rev. B {\bf 63}, 125415 (2001).
1246:
1247: \bibitem{Segalheat} D. Segal, A. Nitzan, P. H\"{a}nggi,
1248: J. Chem. Phys. {\bf {119}}, 6840 (2003).
1249:
1250: \bibitem{Livi} R. Livi, A. Politi, S. Lepri,
1251: %Thermal Conduction in Classical Low- Dimensional Lattices,
1252: Phys. Rep. {\bf 377}, 1 (2003).
1253:
1254: \bibitem{Joel}
1255: %Fourier's Law for a Harmonic Crystal with Self-Consistent Stochastic Reservoirs
1256: F. Bonetto1, J. L. Lebowitz, J. Lukkarinen,
1257: J. Stat. Phys. {\bf 116}, 783 (2004).
1258:
1259: \bibitem{Prosen}
1260: %Normal and anomalous heat transport in one-dimensional classical lattices
1261: T. Prosen, D. K. Campbell, Chaos {\bf 15}, 015117 (2005).
1262:
1263: \bibitem{LiWang}
1264: %anomalous heat conduction and anomalous diffusion in nonlinear lattices, single walled nanotubes, and billiard gas channels
1265: B. Li, J. Wang, L. Wang, G. Zhang,
1266: Chaos {\bf 15}, 015121 (2005).
1267:
1268: \bibitem{Gruber}
1269: C. Gruber, A. Lesne,
1270: %Hamiltonian model of heat conductivity and Fourier law
1271: Physics A- Statistical and Theoretical Physics {\bf 351}, 358
1272: (2005).
1273:
1274: \bibitem{Schr}
1275: M. Michel, G. Mahler, J. Gemmer,
1276: %Fourier's law from Schrodinger dynamics
1277: Phys. Rev. Lett. {\bf 95}, 180602 (2005).
1278:
1279: \bibitem{MajumdarCNT}
1280: %Thermal conductance and thermopower of an individual single-wall carbon nanotube
1281: C. Yu, L. Shi, Z. Yao, D. Li, A. Majumdar, Nano Lett.
1282: {\bf 5}, 1842 (2005).
1283:
1284: \bibitem{ChiuCNT}
1285: %Ballistic Phonon Thermal Transport in Multiwalled Carbon Nanotubes
1286: H.-Y. Chiu, V. V. Deshpande, H. W. Ch. Postma, C. N. Lau, C. Miko, L. Forro,
1287: M. Bockrath,
1288: Phys. Rev. Lett. {\bf 95}, 226101 (2005).
1289:
1290: \bibitem{Schwarzer}
1291: %Intramolecular vibrational energy redistribution in bridged azulene-anthracene compounds: Ballistic energy transport through molecular chains
1292: D. Schwarzer, P. Kutne, C. Schr\"oder, J. Troe,
1293: J. Chem. Phys. {\bf 121}, 1754 (2004).
1294:
1295: \bibitem{Leitner}
1296: D. M. Leitner,
1297: J. Phys. Chem. A {\bf 106}, 10870 (2002);
1298: %Anharmonic decay of vibrational states in helical peptides, coils, and one-dimensional glasses
1299: X. Yu, D. M. Leitner, J. Phys. Chem. B {\bf 107}, 1698 (2003);
1300: %Vibrational Energy Transfer and Heat Conduction in a Protein
1301: %X. Yu, D. M. Leitner,
1302: %Heat flow in proteins: Computation of thermal transport coefficients
1303: J. Chem. Phys. {\bf 122}, 054902 (2005).
1304:
1305: \bibitem{Casati1}
1306: % Controlling the Energy Flow in Nonlinear Lattices: A Model for a Thermal Rectifier
1307: M. Terraneo, M. Peyrard, G. Casati,
1308: Phys. Rev. Lett. {\bf 88}, 094302 (2002).
1309:
1310: \bibitem{Casati2}
1311: % Thermal Diode: Rectification of Heat Flux
1312: B. Li, L. Wang, G. Casati,
1313: Phys. Rev. Lett. {\bf 93}, 184301 (2004).
1314:
1315: \bibitem{Rectif}
1316: D. Segal, A. Nitzan, Phys. Rev. Lett. {\bf 94}, 034301 (2005);
1317: J. Chem. Phys. {\bf 122}, 194704 (2005).
1318: % Heat rectification in molecular junctions
1319:
1320: \bibitem{Casatixxx}
1321: %Negative differential thermal resistance and thermal transistor
1322: B. Li, L. Wang, G. Casati, cond-mat. 0410172.
1323:
1324: \bibitem{general} The general formulation of Ref. \cite{Segalheat}
1325: yields $\mathcal{T}(\omega)=(2/\pi) \omega^2 \Gamma_L \Gamma_R
1326: \left[ (\omega^2- \omega_0^2)^2+ (\Gamma_L+\Gamma_R)^2
1327: \omega^2)\right]^{-1}$.
1328:
1329: \bibitem{Lebowitz}
1330: Z. Rieder, J. L. Lebowitz, E. Lieb,
1331: %Properties of a Harmonic Crystal in a Stationary Nonequilibrium State,
1332: J. Math. Phys. {\bf 8}, 1073 (1967).
1333:
1334: \bibitem{Kittel}
1335: C. Kittel,
1336: {\it Introduction to Solid State Physics}, 7th. Ed. Wiley, New York, (1996).
1337:
1338: \bibitem{fullerene}
1339: {\it Fullerene Polymers and Fullerene Polymer Composites.}
1340: P. C. Eklund and A. M. Rao (eds.), Springer, Berlin, (1999).
1341:
1342: \bibitem{Crommie}
1343: % Tuning negative differential resistance in a molecular film
1344: M. Grobis, A. Wachowiak, R. Yamachika, M. F. Crommie,
1345: App. Phys. Lett. {\bf 86}, 204102 (2005).
1346:
1347: \bibitem{Brend}
1348: %Rectification in one-dimensional electronic systems
1349: B. Braunecker, D. E. Feldman, J. B. Marston, Phys. Rev. B {\bf
1350: 72}, 125311 (2005).
1351:
1352: \bibitem{Heat}
1353: D. Segal, A. Nitzan,
1354: J. Chem. Phys. {\bf 117}, 3915 (2002).
1355: %Heating in current carrying molecular junctions.
1356:
1357: \bibitem{SegalPump}
1358: D. Segal, A. Nitzan, Phys. Rev. E, To be published.
1359:
1360: \bibitem{Leitner2}
1361: J. K. Agbo, D. M. Leitner, D. A. Evans, D. J. Wales,
1362: %Influence of vibrational energy flow on isomerization of flexible molecules: Incorporating non-Rice-Ramsperger-Kassel-Marcus kinetics in the simulation of dipeptide isomerization
1363: J. Chem. Phys. {\bf 123}, 124304 (2005).
1364:
1365: \bibitem{CasatiR}
1366: %Controlling the heat flow: Now it is possible
1367: G. Casati, Chaos {\bf 15}, 015120 (2005).
1368:
1369: \bibitem{Mahan}
1370: G. D. Mahan, {\it Many-particle physics},
1371: Plenum press, New York, (2000).
1372:
1373: \end{thebibliography}
1374:
1375: %-----------------------------
1376:
1377: %----------------------------------------------------------
1378:
1379: \end{document}
1380: