Home Search Collections Journals About Contact us My IOPscience

Coefficient of performance under optimized figure of merit in minimally nonlinear irreversible

refrigerator

This content has been downloaded from IOPscience. Please scroll down to see the full text.

2013 EPL 101 10005

(http://iopscience.iop.org/0295-5075/101/1/10005)

View the table of contents for this issue, or go to the journal homepage for more

Download details:

IP Address: 148.204.17.219
This content was downloaded on 20/03/2015 at 16:39

Please note that terms and conditions apply.

 January 2013
EPL, 101 (2013) 10005 www.epljournal.org
doi: 10.1209/0295-5075/101/10005

Coecient of performance under optimized gure of merit

in minimally nonlinear irreversible refrigerator
Y. Izumida1,2 , K. Okuda3 , A. Calvo Hernandez4 and J. M. M. Roco4
1
Department of Applied Physics, The University of Tokyo - 7-3-1 Hongo, Bunkyo, Tokyo 113-8656, Japan
2
Department of Physics, The University of Tokyo - 7-3-1 Hongo, Bunkyo, Tokyo 113-0033, Japan
3
Division of Physics, Hokkaido University - Sapporo 060-0810, Japan
4
Departamento de Fsica Aplicada, and Instituto Universitario de Fsica Fundamental y Matematicas (IUFFyM),
Universidad de Salamanca - 37008 Salamanca, Spain, EU

received 14 July 2012; accepted in nal form 11 December 2012

published online 16 January 2013

PACS 05.70.Ln Nonequilibrium and irreversible thermodynamics

Abstract We apply the model of minimally nonlinear irreversible heat engines developed
by Izumida and Okuda (EPL, 97 (2012) 10004) to refrigerators. The model assumes extended
Onsager relations including a new nonlinear term accounting for dissipation eects. The bounds
for the optimized regime under an appropriate gure of merit and the tight-coupling condition
are analyzed and successfully compared with those obtained previously for low-dissipation Carnot
refrigerators in the nite-time thermodynamics framework. Besides, we study the bounds for the
nontight-coupling case numerically. We also introduce a leaky low-dissipation Carnot refrigerator
and show that it serves as an example of the minimally nonlinear irreversible refrigerator, by
calculating its Onsager coecients explicitly.

Copyright

c EPLA, 2013

Introduction. Nowadays, the optimization of ther-
mal heat devices is receiving a special attention because
of its straightforward relation with the depletion of energy
resources and the concerns of sustainable development.
A number of dierent performance regimes based on dier-
ent gures of merit have been considered with special
emphasis in the analysis of possible universal and unied
features. Among them, the eciency at maximum power
(EMP) maxP for heat engines working along cycles or
in steady state is largely the issue most studied indepen-
dently of the thermal device nature (macroscopic, stochas-
tic or quantum) and/or the model characteristics [114].
Even theoretical results have been faced with an exper-
imental realization for micrometre-sized stochastic heat
engines performing a Stirling cycle [15].
For most of the Carnot-like heat engine models analyzed
in the nite-time thermodynamics (FTT) framework [5,6]
the EMP regime allows for valuable and simple expres-
sions of the optimized eciency, which under endore-
versible assumptions (i.e., all considered irreversibilities
coming from the couplings between the working system
and the external heat reservoirs through linear heat
transfer laws) recover the
paradigmatic Curzon-Ahlborn
value [16] maxP = 1 CA using Tc /Th , where dissipation Carnot heat engine model to refrigerators and
Tc and Th denote the temperature of the cold and hot heat
reservoirs, respectively. A signicant conceptual advance
was reported by Esposito et al. [17] by considering a low-
dissipation Carnot heat engine model. In this model the
entropy production in the hot (cold) heat exchange process
is assumed to be inversely proportional to the time dura-
tion of the process (see eqs. (30) and (31) for more details).
Then the reversible regime is approached in the limit of
innite times and the maximum power regime allows us to
recover the Curzon-Ahlborn value when symmetric dissi-
pation is considered. This derivation does not require the
assumption of any specic-heat transfer law nor the small
temperature dierence. Considering extremely asymmet-
ric dissipation limits, these authors also derived that EMP
is bounded from the lower and upper sides as
C C

maxP
, (1)
2 2 C
where C 1 is the Carnot eciency. The Curzon-
Ahlborn value CA (i.e., the symmetric dissipation limit)
is located between these bounds. Additional studies on
various heat engine models [1820] conrmed and gener-
alized the above results.
The work by de Tomas et al. [21] extended the low-
besides proposed a unied gure of merit (-criterion

10005-p1
Y. Izumida et al.

described below) focusing the attention in the common by proposing a model of minimally nonlinear irreversible
characteristics of every heat energy converter (the working heat engines described by extended Onsager relations with
systems and total cycle time) instead of any specic nonlinear terms accounting for power dissipation. The
coupling to external heat sources which can vary according validity of the theory was checked by comparing the results
to a particular arrangement. For refrigerators, we denote for EMP and its bounds with eq. (1). Now, the main
the heat absorbed by the working system from the cold goal of the present paper is to extend the application of
heat reservoir by Qc , the heat evacuated from the working this nonlinear irreversible theory [34] to refrigerators and
system to the hot heat reservoir by Qh and the work analyze their COP under maximum -condition. To get
input to the working system by W Qh Qc . For heat this we rst present the model and analyze the results
engines, we may inversely choose natural directions of heat when the tight-coupling condition is met. Then we also
transfers and work. That is, we denote the heat absorbed study the nontight-coupling case numerically. Finally, we
by the working system from the hot heat reservoir by extend the low-dissipation Carnot refrigerator model [26]
Qh , the heat evacuated from the working system to to a leaky model and show that it serves as an example of
the cold heat reservoir by Qc and the work output by the minimally nonlinear irreversible refrigerator by calcu-
W Qh Qc . Then we can dene this -criterion as lating its Onsager coecients explicitly.
the product of the converter eciency z times the heat
Qin exchanged between the working system and the heat Minimally nonlinear irreversible refrigerator
reservoir, divided by the time duration of cycle tcycle : model. Refrigerators are generally classied into
two types, that is, steady-state refrigerators and cyclic
z Qin ones. Our theory can be applied to both cases. Since
= , (2) the working system stays unchanged for steady-state
tcycle
refrigerators and comes back to the original state after
where Qin = Qc , z = Qc /W for refrigerators and Qin = one cycle for cyclic refrigerators, the entropy production
Qh , z = W/Qh for heat engines. It becomes power rate of the refrigerators agrees with the entropy increase
as = W/tcycle when applied to heat engines and also rate of the heat reservoirs as
allows us to obtain the optimized coecient of perfor-

mance (COP) max under symmetric dissipation condi- Qc + W Qc W 1 1
= = + Qc , (5)
tions when applied to refrigerators [21]: Th Tc Th Th Tc

max = 1 + C 1 CA , (3) where the dot denotes the quantity per unit time for
steady-state refrigerators and the quantity divided by the
where C /(1 ) is the Carnot COP. This result time duration of cycle t
cycle for cyclic refrigerators. From
could be considered as the genuine counterpart of the the decomposition of into the sum of the product of
Curzon-Ahlborn eciency for refrigerators. It was rst thermodynamic ux J and its conjugate thermodynamic
i
obtained in FTT for Carnot-like refrigerators by Yan and force X as = J X + J X , we can dene J x, X
i 1 1 2 2 1 1
Chen [22] taking as target function Qc , where Qc is the F/T and J Q , X 1/T 1/T for steady-state
h 2 c 2 h c
cooling power of the refrigerator, later and independently refrigerators, where W = F x with F and x denoting
by Velasco et al. [23,24] using a maximum per-unit-time the time-independent external generalized force and its
COP and by Allahverdyan et al. [25] in the classical conjugate variable, respectively. Likewise we can also
limit of a quantum model with two n-level systems dene
interacting via a pulsed external eld. Very recent results 1 W
by Wang et al. [26] generalized the previous results for J1 , X1 , (6)
tcycle Th
refrigerators [21] and they obtained the following bounds
of the COP considering extremely asymmetric dissipation Qc 1 1
J2 Qc = , X2 , (7)
limits: tcycle Th Tc
9 + 8C 3
0
max
. (4) for cyclic refrigerators, where X and X play the role of
2 1 2
1
These bounds for refrigerators play the same role as that driver and driven forces, respectively . Anyway, we assume
of eq. (1) for heat engines. 1 To establish the above election of the thermodynamic uxes
A shortcoming of the all the above results is the and forces from eq. (5) we mainly focus on the specic function of
model dependence. Thus additional research work [2730] the refrigerator system (the extracted cooling power Qc of the low-
has been devoted to obtain model-independent results temperature reservoir) as a consequence of the input of an external
power W [30]. An alternative starting point is to express the entropy
on eciency and COP using the well-founded formal- production rate in eq. (5) in terms of W and Qh . If this is done
ism of linear irreversible thermodynamics (LIT) for both the thermodynamic uxes and forces are exactly the same as those
cyclic and steady-state models including explicit calcu- obtained for a heat engine in [34] with the sole dierence coming from
sign convention: W is given in (delivered to) for a refrigerator
lations of the Onsager coecients using the molecular the (heat engine) while Qh is delivered to (given in) for a refrigerator
kinetic theory [3133]. Beyond the linear regime a further (heat engine). We have checked that this alternative election of uxes
improvement was reported by Izumida and Okuda [34] and forces does not change the main results in the present paper.

10005-p2
 COP under optimized gure of merit in minimally nonlinear irreversible refrigerator

L21 Th
J3 = J1 + L22 (1 q 2 )X2 + h J12 , (13)
L11 Tc
where q LL12L is the usual coupling strength para-
11 22
meter (|q|
1 from eq. (10)) and h denotes the strength
of the power dissipation into the hot heat reservoir as
Th
h c > 0. (14)
L11
Equations (12) and (13) allow a clearer description of
the refrigerator than eqs. (8) and (9) by considering J1
as the control parameter instead of X1 at constant X2 .
Fig. 1: Schematic illustration of the minimally nonlinear So, each term in eqs. (12) and (13) has explicit physical
irreversible refrigerator described by eqs. (12) and (13). meanings (see also [34]): the two rst terms mean the
reversible heat transport between the working system and
the heat reservoirs; the second ones account for coupling
that the refrigerators are described by the following
eects between the heat reservoirs; and the third ones
extended Onsager relations [34]:
account for the power dissipation ( J12 ) into the heat
reservoirs, which inevitably occurs in a nite-time motion
J1 = L11 X1 + L12 X2 , (8)
(J1 = 0).
J2 = L21 X1 + L22 X2 c J12 , (9) Using eqs. (12) and (13) instead of eqs. (8) and (9), the
where Lij s are the Onsager coecients with reciprocity power input P = J3 J2 and the entropy production rate
J3 J2
2
L12 = L21 . The nonlinear term c J1 means the power = Th Tc are written as


dissipation into the cold heat reservoir and c > 0 denotes L21 1 Th 2
P = 1 J1 + J , (15)
its strength. We also disregard other possible nonlinear L11 L11 1
terms by assuming that their coecients are too small
when compared with the power dissipation term. and

2 2 2 c h
In the limit of Xi s 0, the thermodynamic forces = L22 (1 q )X2 + J1 + . (16)
Tc Th
can be approximated as X1  W/T (X1 = F/T ) and X2 
T /T 2 for the cyclic (steady-state) refrigerators, where For the refrigerator, the converter eciency z in eq. (2)
T Th Tc and T (Th + Tc )/2. These expressions of is the COP dened as
L21 2 2
the forces are consistent with [27,31]. Then the extended Qc J2 L11 J1 + L22 (1 q )X2 c J1
Onsager relations in eqs. (8) and (9) recover the usual = = 
L21 1
 Th 2
, (17)
P P L11 1 J1 + L11 J1
linear Onsager relations such as in [27,31]. Thus our
minimally nonlinear irreversible model naturally includes where we used eqs. (12) and (15). We also introduce the
the linear irreversible model as a special case. The non- -criterion as a product of the COP and the cooling
negativity of the entropy production rate = J1 X1 + power J2 , identifying the cooling power Qin /tcycle in eq. (2)
J2 X2 , which is a quadratic form of Xi s in the linear as Qc = J2 :
irreversible model, leads to the restriction of the Onsager  2
L21 2 2
coecients [31,35,36] to J 2 J
L11 1 + L22 (1 q )X 2 J
c 1
= 2 = L
1  T
, (18)
P L11 1 J1 + L11 J1
21 h 2
L  0, L  0, L L L L  0.
11 22 11 22 12 21 (10)
where we used eqs. (12) and (15). Though eq. (2) using
Although our model includes the power dissipation as the Qin /tcycle appears to be valid only for cyclic heat energy
nonlinear term, we also assume eq. (10) holds. This idea converters, eq. (18) using J2 is valid even for steady-
that the dissipation is included in the scope of the linear state refrigerators. In order to get a positive cooling power
irreversible thermodynamics is also found in [3739]. J2 > 0, we see from eq. (12) that J1 should be located in
The power input P is given by P W = J1 X1 Th > 0. a nite range as
Then, the heat ux into the hot heat reservoir Qh
P + Qc J3 is given as L21 D L21 + D
< J1 < , (19)
2L11 c 2L11 c
J3 = P + Qc = J1 X1 Th + J2 . (11)
where the discriminant D is given by
By solving eq. (8) with respect to X1 and substituting it D L221 + 4L211 L22 c (1 q 2 )X2 > 0. (20)
into eqs. (9) and (11), we can rewrite J2 and J3 by using
J1 instead of X1 as (see g. 1): We consider the optimization of the -criterion with
respect to J1 in this range. In the following, we rst
L21 2 2 analyze the case |q| = 1 called the tight-coupling condition.
J2 = J1 + L22 (1 q )X2 c J1 , (12)
L11 Afterward we consider the nontight-coupling case |q| = 1.

10005-p3
Y. Izumida et al.

Tight-coupling case |q| = 1. In the tight-coupling case, as O(c1/2 ). Then we can conrm J2max O(c1/2 )
eq. (19) becomes and Pmax O(c1 ) from eqs. (12) and (15), respectively,
1/2
L21 which leads to max = J2max /Pmax O(c ) 0. This
0 < J1 < . (21) is an anomalous situation where the refrigerator transfers
c L11
the innite amount of heat from the cold heat reservoir to
We nd J1 = J1max which gives the maximum of the hot one, consuming the innite amount of work.
by solving the equation /J1 = 0. Noticing that J1 = Nontight-coupling case |q| = 1. Comparing with the
L21 /(c L11 ) as the solution of J2 = 0 is always one of the tight-coupling case, we found that the analytic solution
solutions of /J1 = 0, we can simplify that equation to of the nontight-coupling case becomes too complicated to
the following quadratic one: write down explicitly. Thus we here solve the optimization
problem numerically. We rst note that eq. (20) implies
3L21 L221
J12 + J1 = 0. (22) the restriction
2C Th 2c L11 C Th
L212
Solving this equation, we obtain the sole physically accept- c < 2 )L2 L X
c+ , (27)
2 2
4(1 q 11 22 2
able solution which satises /J1 < 0 and locates in
the range given by eq. (21) as for given Onsager coecients Lij s and X2 . Such restric-
tion does not appear in the tight-coupling case. Besides,
2L21 we note that we also have the restriction 0 < c < Th /L11
J1max =    . (23)
3c L11 + c L11 8C 1 + c + 9h due to the assumption of the non-negativity of c and h
as in eq. (14). Thus, by combining these two inequalities,
we obtain the restriction of the dissipation ratio for the
Substituting eq. (23) into eq. (17), we can obtain nontight-coupling case, depending on the value of + as
c
the general expression of the COP under maximum
-condition for the model of minimally nonlinear irre- h
0< < (28)
versible refrigerators with the tight-coupling condition: c
   in the case of Th /L11
c+ < , and
9 + 8C 1 + hc 3
max =   . Th +
L11 c
(24) h
2 1 + hc + < < (29)
c c
It is a monotonic decreasing function of the dissipation in the case of 0 < c+ < Th /L11 , respectively.
ratio h /c . Considering that the assumption c > 0 and We can obtain max numerically as follows: we x the
c < Th /L11 due to the non-negativity of h as in eq. (14), values of L11 , L22 , q, Th and Tc . Choosing c so as to
the range of the dissipation ratio for the tight-coupling satisfy 0 < c < Th /L11 and eq. (27), which also determines
case is restricted to h from eq. (14), we can nd at the maximum value of
h by calculating eq. (18) as J1 is varied in the range of
0< < . (25)
c eq. (19).
We rst consider the typical dependence of max on
Then we can obtain the lower and the upper bounds by the dissipation ratio h /c under given temperatures for
considering asymmetric dissipation limits h /c and some qs. By changing c , we plotted max as a function of
h /c 0, respectively, as h /c in g. 2(a). As one of the interesting characteristics,
we can see that the dependence can become nonmonotonic
9 + 8C 3
max 0 < max < +
max , (26) and max attains the maximum value at a nite dissipa-
2
tion ratio for each q. This contrasts to the tight-coupling
in full agreement with the results reported in [26] and case in eq. (26), where its upper bound + max is always
expressed in above eq. (4). Here we stress that eq. (26) attained in the asymmetric dissipation limit h /c 0.
has been derived in a model-independent way and is valid Next we determine the upper bound q+ max and the
for general refrigerators, regardless of the types of refrig- lower bound q max for the nontight-coupling case under
erator (steady state or cyclic). It is also interesting to various temperatures. By changing C , we evaluated the
see from eq. (26) that the lower bound for max is 0, maximum value of max as q+ max for some qs (see
which is in contrast to the case of the heat engines (see g. 2(b)). We found that q+ max is always lower than
eq. (21) in [34]). One may feel that this is a counterin- +
max in eq. (26) and monotonically decreases as q is
tuitive result at a rst glance since the limit of the zero decreased for each C . Meanwhile we also found that the
dissipation c 0 in the cold heat reservoir seems to be an lower bound q max agrees with 0 for each q (data not
advantageous condition for a refrigerator. This behavior shown), which can always be attained in the asymmetric
can be understood as follows: from eq. (23), J1max diverges dissipation limit h /c as in the tight-coupling case

10005-p4
 COP under optimized gure of merit in minimally nonlinear irreversible refrigerator

inversely proportional to the duration of the process ti :

Th h
Qh = Th S + , (30)
th
Tc c
Qc = Tc S , (31)
tc
where we denote by S > 0 the quasistatic entropy change
of the working system during each isothermal process,
and by i /ti the corresponding entropy production with a
constant strength i > 0 [21,26]. Then we formally modify
this low-dissipation Carnot refrigerator to a leaky model
by adding a heat leak term as follows:
Th h
Qh = Th S + (Th Tc )(th + tc ), (32)
th
Tc c
Qc = Tc S (Th Tc )(th + tc ), (33)
tc
where we assume the linear Fourier-type heat transfer
from the hot heat reservoir to the cold one as the heat
leak per cycle, denoting by > 0 the eective thermal
conductivity. We here reinterpret Qh (Qc ) as the net heat
transfer into (from) the heat reservoir per cycle, instead of
the heat transfer in each isothermal process, considering
the heat leak contribution which always continues during
one cycle.
Fig. 2: Numerical solutions for the nontight-coupling case
We show that this leaky low-dissipation Carnot refriger-
|q|
= 1: (a) dependence of max on the dissipation ratio h /c
ator is an example of the minimally nonlinear irreversible
under given temperatures (Th = 1 and Tc = 0.7) for some qs.
We also plot eq. (24) for the tight-coupling case |q| = 1 for refrigerator by rewriting eqs. (32) and (33) into the forms
of eqs. (12) and (13): since it works as a cyclic refrigerator
comparison. We always set
L11 = 1, L22 = 1. Giving q, we can
determine L21 as L21 = q 2 L11 L22 . (b) dependence of the and according to eqs. (6) and (7), we can dene its thermo-
upper bound q+ max on C for some qs. We also plot the upper
dynamic uxes and forces as J1 1/tcycle = 1/(th + tc )
bound + max for the tight-coupling case |q| = 1 in eq. (26) for = 1/(( + 1)th ), J2 Qc = Qc /tcycle , X1 W/Th and
comparison. We used the same values for Lij s as in (a). We X2 1/Th 1/Tc , with tc /th . By using these deni-
xed Th = 1 and changed Tc . tions and eqs. (32) and (33), we can calculate the Onsager
coecients Lij s and the strength of the dissipation i s
as
in eq. (26). We also note that in the case where h /c
  Th Th Tc S

satises eq. (29), the lower bound q
max = 0 can be realized
L11 L12
= Th TYc S Tc2 Th S 2
Y
, (34)
also at the lower endpoint of eq. (29), that is, in the limit L21 L22 + ThYTc
Y Y
of h /c (Th /L11 c+ )/c+ . (See the lines of q = 0.3,
0.5 and 0.7 in g. 2(a).) In this case, q max becomes 0
since J2max vanishes: in the lower endpoint of eq. (29), the h = Th h ( + 1), (35)
discriminant D in eq. (20) becomes 0. Thus the solution of
the inequality J2 > 0 in eq. (19) approaches this multiple Tc c ( + 1)
root in that limit, where J2 in eq. (12) is identically 0. This c = , (36)

contrasts to the case of h /c , where J2max diverges.
respectively, where the common factor Y in eq. (34) is
given as

Example: leaky low-dissipation Carnot refrig- Tc c
erator. As an example of the minimally nonlinear Y ( + 1) T
h h + . (37)

irreversible refrigerator, we here introduce a leaky
low-dissipation Carnot refrigerator model. The original From eq. (34) we see that the Onsager reciprocity
low-dissipation Carnot refrigerator without heat leak L12 = L21 holds as expected. In general, the Onsager
studied in [21,26] is an extension of the quasistatic Carnot coecients show the nontight-coupling condition
L12
refrigerator by assuming that heat transfer accompany- |q| = L11 L22 < 1, but recover the tight-coupling condition
ing nite-time operation in each isothermal process is |q| 1 in the zero leak limit of 0. Thus our model

10005-p5
Y. Izumida et al.
eqs. (12) and (13) under the tight-coupling condition
includes the low-dissipation Carnot refrigerator [21,26] in
eqs. (30) and (31) as a special case.
Finally, we note that we obtained eq. (24) for the tight-
coupling case maximizing only by one parameter J1 ,
whereas in the original low-dissipation model without
the heat leak the optimization uses the two parameters
th and tc [21,26]. Thus the parameter space for the
maximization is dierent. In the formalism presented here,
a further optimization of = () by is still possible [11] Gaveau B., Moreau M. and Schulman L. S., Phys.
after the optimization by J1 and it corresponds to the
two-parameter maximization done in [21,26]. But, that
further optimization does not change the bounds in eq. (4).
In particular, under symmetric conditions
(h = c ) the
further optimization by gives max = 1 + 1, from
which we can straightforwardly obtain the symmetric
optimization COP in eq. (3) [21].
Summary. We introduced the extended Onsager
relations as the model of minimally nonlinear irreversible
refrigerators, which can be applied to both steady-state
and cyclic refrigerators. We analytically derived the
general expression for the coecient of performance
(COP) under maximum -condition when the tight-
coupling condition is met and determined its upper
and lower bounds which perfectly agree with the result
derived previously in the low-dissipation Carnot refriger-
ator model in [26]. We also studied the nontight-coupling
case numerically and found that its upper bound is
always lower than that of the tight-coupling case, whereas
all the cases have the same lower bound. Moreover, we
introduce the leaky low-dissipation Carnot refrigerator.
Calculating the Onsager coecients and the strength of
the power dissipation explicitly, we proved that this leaky
low-dissipation Carnot refrigerator is an example of the
minimally nonlinear irreversible refrigerator.
[29] Jimenez de Cisneros B. and Calvo Hernandez A.,
YI acknowledges the nancial support from a Grant-in-
Aid for JSPS Fellows (Grant No. 22-2109). JMMR and
ACH acknowledge nancial support from Ministerio de
Educacion y Ciencia of Spain under Grant No. FIS2010-
17147 FEDER.
REFERENCES
[1] Esposito M., Lindenberg K. and Van den Broeck
C., Phys. Rev. Lett., 102 (2009) 130602.
[2] Benenti G., Saito K. and Casati G., Phys. Rev. Lett.,
106 (2010) 060601.
[3] Sanchez-Salas N., Lopez-Palacios L., Velasco S.
and Calvo Hernandez A., Phys. Rev. E, 82 (2010)
051101.
[4] Tu Z. C., Chin. Phys. B, 21 (2012) 020513.
10005-p6
[5] Wu C., Chen L. and Chen J., Advances in Finite-
Time Thermodynamics: Analysis and Optimization (Nova
Science, New York) 2004.
[6] Durmayaz A., Sogut O. S., Sahin B. and Yavuz H.,
Prog. Energy Combust. Sci., 30 (2004) 175.
[7] Tu Z. C., J. Phys. A: Math. Theor., 41 (2008) 312003.
[8] Schmiedl T. and Seifert U., Phys. Rev. Lett., 98 (2007)
108301.
[9] Schmiedl T. and Seifert U., EPL, 81 (2008) 20003.
[10] Seifert U., Phys. Rev. Lett., 106 (2011) 020601.
Rev. Lett., 105 (2011) 230602.
[12] Esposito M., Kawai R., Lindenberg K. and Van den
Broeck C., EPL, 89 (2010) 20003.
[13] Esposito M., Kawai R., Lindenberg K. and Van den
Broeck C., Phys. Rev. E, 81 (2010) 041106.
[14] Abe S., Phys. Rev. E, 83 (2011) 041117.
[15] Blickle V. and Bechinger C., Nat. Phys., 8 (2012)
143.
[16] Curzon F. and Ahlborn B., Am. J. Phys., 43 (1975)
22.
[17] Esposito M., Kawai R., Lindenberg K. and Van den
Broeck C., Phys. Rev. Lett., 105 (2010) 150603.
[18] Yan H. and Guo H., Phys. Rev. E, 85 (2012) 011146.
[19] Wang Y. and Tu Z. C., Phys. Rev. E, 85 (2012) 011127.
[20] Wang Y. and Tu. Z. C., EPL, 98 (2012) 40001.
[21] de Tomas C., Calvo Hernandez A. and Roco
J. M. M., Phys. Rev. E, 85 (2012) 010104(R).
[22] Yan Z. and Chen J., J. Phys. D, 23 (1990) 136.
[23] Velasco S., Roco J. M. M., Medina A. and Calvo
Hernandez A., Phys. Rev. Lett., 78 (1997) 3241.
[24] Velasco S., Roco J. M. M., Medina A. and Calvo
Hernandez A., Appl. Phys. Lett., 71 (1997) 1130.
[25] Allahverdyan A. E., Hovhannisyan K. and Mahler
G., Phys. Rev. E, 81 (2010) 051129.
[26] Wang Y., Li M., Tu Z. C., Calvo Hernandez A. and
Roco J. M. M., Phys. Rev. E, 86 (2012) 011127.
[27] Van den Broeck C., Phys. Rev. Lett., 95 (2005) 190602.
[28] Jimenez de Cisneros B., Arias-Hernandez L. A. and
Calvo Hernandez A., Phys. Rev. E, 73 (2006) 057103.
Phys. Rev. Lett., 98 (2007) 130602.
[30] Jimenez de Cisneros B. and Calvo Hernandez A.,
Phys. Rev. E, 77 (2008) 041127.
[31] Izumida Y. and Okuda K., Phys. Rev. E, 80 (2009)
021121.
[32] Izumida Y. and Okuda K., Eur. Phys. J. B, 77 (2010)
499.
[33] Izumida Y. and Okuda K., EPL, 83 (2008) 60003.
[34] Izumida Y. and Okuda K., EPL, 97 (2012) 10004.
[35] Onsager L., Phys. Rev., 37 (1931) 405.
[36] de Groot S. R. and Mazur P., Non-Equilibrium Ther-
modynamics (Dover, New York) 1984.
[37] Callen H. B. and Welton T. A., Phys. Rev., 83 (1951)
34.
[38] Apertet Y., Ouerdane H., Glavatskaya O., Goupil
C. and Lecoeur P., EPL, 97 (2012) 28001.
[39] Apertet Y., Ouerdane H., Goupil C. and Lecoeur
Ph., Phys. Rev. E, 85 (2012) 041144.