Você está na página 1de 5

week ending

PRL 108, 210602 (2012) PHYSICAL REVIEW LETTERS 25 MAY 2012

Efficiency of Isothermal Molecular Machines at Maximum Power


Christian Van den Broeck,1 Niraj Kumar,2 and Katja Lindenberg2
1
Hasselt University, B-3590 Diepenbeek, Belgium
2
Department of Chemistry and Biochemistry and BioCircuits Institute, University of California San Diego,
9500 Gilman Drive, La Jolla, California 92093-0340, USA
(Received 30 January 2012; published 22 May 2012)
We derive upper and lower bounds for the efficiency of an isothermal molecular machine operating at
maximum power. The upper bound is reached when the activated state is close to the fueling or reactant
state (Eyring-like), while the lower bound is reached when the activated state is close to the product state
(Kramers-like).

DOI: 10.1103/PhysRevLett.108.210602 PACS numbers: 05.70.Ln, 05.20.y, 05.40.a

According to thermodynamics, different forms of work between the two states will typically follow a preferred
can be transformed into one another, with an efficiency of pathway that connects these states via the lowest lying
at most 100% [1]. This lossless limit is achieved with a saddle point, the so-called activated state. One can pro-
reversible process, i.e., an infinitely slow process. The ject the motion on this pathway and introduce a one-
corresponding power output is therefore zero and thus of dimensional reaction coordinate x with corresponding
limited interest from a practical standpoint. One of the effective free energy potential U0 x. The two rest states
early discussions about efficiency at finite power is attrib- of the machine, that is, the minima in the absence of
uted to Moritz von Jacobi around 1840. He realized that the external forces, correspond to, say, locations x 0 and
output power of an electrical device operating in the linear x L. The activated state lies at an intermediate position
response regime is maximum when the internal and exter- xa L, 0    1. In the unperturbed phase there are no
nal resistors are the same, yielding an efficiency of 50%. net transitions, and the states 1 and 2 have the same base-
The Jacobi theorem can easily be reproduced in the much line potential value, U0 0 U0 L 0. The potential has
more general context of linear irreversible thermodynam- a maximum Ua U0 L at the activated state, whose
ics: in any engine operating in the linear response regime, value is typically much larger than the thermal energy
maximum power is achieved when the loading force is 1 kB T (T being the temperature and kB the
equal to half of the stopping force; the corresponding Boltzmann constant). In this rest state, the rates, k0 from
efficiency (output power over input power) is equal to 1 to 2 and k 0 from 2 to 1, are equal and given by an
1=2. A similar result has been proven for the transforma- Arrhenius law, k 0 k  k  expU . We as-
0 0 a
tion of heat into work, where the maximum efficiency, the sume a constant preexponential factor .
Carnot efficiency, is again achieved under reversible op- In the operational regime, that is, in the presence of
eration, with zero power output. In the regime of linear external forces, states 1 and 2 can be identified as fuel
response the efficiency at maximum power is again 50% of (or reactant) and product states, respectively. To
the Carnot efficiency [2]. More recently, in this latter case transform fuel into product, the machine is subject to a
various explicit results, including bounds for efficiency at driving force F1 which allows it to overcome an opposing
maximum power, have also been obtained in the nonlinear but weaker loading force F2 , F2  F1 . These forces can
regime [311]. Whether similar results can be obtained for be of various physical origins, including chemical (differ-
isothermal machines has been recently questioned in [12]. ences in chemical potentials), electrical (internal or exter-
In the present Letter we show that this is in fact the case. nal electric fields), or mechanical (e.g., optical tweezers,
In particular, we derive upper and lower bounds for atomic force microscope, or optical rotational torque). The
the efficiency at maximum power, and we show that the
coefficient of the quadratic term in an expansion around
equilibrium vanishes for systems possessing left-right
symmetry.
Generic model for a molecular motor.We first consider
a generic model for a molecular motor, namely, a two-state
machine operating along a one-dimensional reaction coor-
dinate, see Fig. 1. The states correspond to two minima of
an appropriate free energy landscape. While a physical FIG. 1. Schematic free energy potential U0 x for a two-state
energy landscape is expected to be very complicated molecular engine described by a reaction coordinate x under the
and high dimensional, the thermally activated transitions net load force F F1  F2  0.

0031-9007=12=108(21)=210602(5) 210602-1 2012 American Physical Society


week ending
PRL 108, 210602 (2012) PHYSICAL REVIEW LETTERS 25 MAY 2012

combined effect of driving and loading is a tilting of the irreversible thermodynamics of efficiency at maximum
potential towards the product state 2, Ux U0 x  Fx, power equal to 1=2. The associated relation between the
with F F1  F2  0. In a transition from state 1 to state forces, 2F2 F1 is obtained from the first term in expan-
2, a (scaled) input energy 1 F1 L is transformed into a sion (3), 2   1  2 .
(scaled) output energy 2 F2 L. The efficiency of this Turning to the next order corrections in Eqs. (3) and (5),
transformation is given by the coefficients vanish in the symmetric case  1=2,
2 F2 reminiscent of a similar property for thermal machines
 : (1) [10]. Note also that the coefficient of the term proportional
1 F1 to  in (5) goes from a maximum value 1=8 at  0 to the
Equation (1) quantifies how efficiently the energy spent on minimum value 1=8 for  1, switching from positive
driving the system is utilized in the process of doing work to negative values at  1=2, again reminiscent of an
against the load. Its maximum value,  1, is reached analogous feature in thermal machines [11]. Note that the
when the loading force F2 approaches the driving force F1 , first two terms of the expansion (5) were also derived in
and the transition from 1 to 2 becomes infinitely slow. In [12], while the relevance of the parameter  was pointed
this reversible lossless limit the power vanishes. out in [14].
In the case of finite and, in particular, of maximum The efficiency ? at maximum power is a function of 
power, the location of the activated state plays a crucial and . One easily verifies that ? is a monotonically
role. For a so-called Eyring-like process [13] the activated decreasing function of  for given   0. The upper limit
state is very close to the fuel state 1; i.e.,  is close to zero. is the efficiency ?E  ? 0;   1 of the extreme Eyring-
The perturbation Fx barely affects the height of the like scenario, and the lower limit is the efficiency 0 
activation barrier that needs to be crossed to go from state ? 1;   ?K of the extreme Kramers-like case:
1 to 2. The rate also remains essentially unaffected,
k  k0 . However, a maximum barrier increase of FL 1  e 1  e
?K    ?   ?E : (6)
occurs for the backward transition, resulting in a rate k  1e  1  e e
k0 expFL (assuming FL  Ua ). On the other
hand, in the Kramers-like scenario   1 [13], k  We next consider the  dependence, starting with the
k0 expFL, while k  k0 remains essentially unaf- variation of the bounds. The Eyring-like efficiency E
fected. More generally, for a barrier at xa L, one has increases monotonically from 1=2 when  ! 0 to E
k k0 exp and k k0 exp1  , where  1 when  ! 1. The Kramers-like efficiency K decreases
1  2 FL is the net energy loss or net load. F is monotonically from 1=2 when  ! 0 to K 0 when
a proper thermodynamic force (net force divided by the  ! 1. The variation of  between these bounds depends
temperature) that appears in the entropy production and is on . When   1=2,  decreases monotonically from
thus a measure of the distance from equilibrium [1]. 1=2 when  0 to 0 when  ! 1. The system is in the
With these explicit expressions for the rates, we turn to product regime (Kramers-like), and  behaves much like
the output power  given by the output energy 2 multi- the Kramers-like limit, never rising above the linear re-
plied by its net rate of production,  k2 , with k sponse value 1=2 [dotted curve, short-dashed curve, and
k  k k k0 e  e1 . To specify the con- filled circles in Fig. 2(a)]. On the other hand, when 0 <
dition of maximum power we set @=@2 0, which   1=2 the system is in the fuel regime (Eyring-like), and
yields the unique solution  starts at 1=2 when  0, rises to a unique maximum,
and then decreases to 0 as  ! 1 [long-dashed curve, solid
1  e curve, and open circles in Fig. 2(a)]. The optimal value
2 (2)
1  e e of the efficiency at maximum power occurs at the net load
value  which solves the transcendental equation obtained
   
1 1 by setting the derivative of (4) with respect to  equal to
   2   2 3 O4 : (3) zero, 1  e 1  e  1 e
  . Each point along the
2 6
curve in the inset of Fig. 2(a) is associated with a different
This result in Eq. (1) yields one of the central results of this value of . High efficiencies at maximum power require the
Letter, namely, the efficiency at maximum power: system to operate very near the fuel state. Thus, for in-
e  1 stance, referring to the figure, the maximum of the  0:1
? (4) curve (solid) is  0:69 and occurs when the net load is
 1e  1 
 3:19. A maximum efficiency of say  0:9 requires
1 1  2 1  12 122 2 that the net load be  5:68 and that the motor operate at
  O3 : (5)  0:016.
2 8 96 We can repeat our analysis for   0, with net transi-
We point to a number of revealing observations. The tions going from state 2 to state 1. Indeed, many motors,
first term of the expansion (5) is the prediction of linear including ATPase, can operate in reverse. The interchange

210602-2
week ending
PRL 108, 210602 (2012) PHYSICAL REVIEW LETTERS 25 MAY 2012

coefficient and the temperature by the Einstein relation


D kB T=. This description relates transition rates to
more fundamental parameters than the earlier phenome-
nology. It has been shown to be a very useful tool to
describe the response of molecular machines [15], and it
has the further advantage of being analytically tractable. It
also does not require extremely high activation barriers.
Indeed, when the potential barriers are comparable to the
thermal energy kB T, it is no longer appropriate to identify
the minima of the potential (formerly x 0 and x L) as
states between which transitions take place. We therefore
replace the two-state scenario by an extended coordinate
x 2 1; 1, and consider a baseline potential U0 x
periodic in x with period L.
Upon application of a driving force F1 and a load force
F2 , the net force F F1  F2  0 induces a steady state
current in the tilted potential Ux U0 x  Fx with
average velocity V hdx=dti along the positive
x coordinate. The transformation of driving energy 1
FIG. 2. Efficiency at maximum power ? as a function of net F1 L per period into loading energy 2 F2 L per
load , indicating lower (Kramers-like) and upper (Eyring-like) period takes place at a net rate k V=L, with output power
bounds for  in both the two-state model and in the extended  k2 V2 =L. The efficiency of the transformation is
state scenarios. In all panels the open circles represent E and
again given by 2 =1 , see Eq. (1). We note in passing that,
the filled circles ?K . Panel (a): two-state model [cf. Eq. (4)] with
 0 (long dashes), 0.1 (solid), 0.5 (short dashes), and 0.9
besides being a natural model for cyclic molecular motors
(dots). The solid and long-dashed curves exhibit a maximum, such as ATPase, overdamped Brownian motion in a tilted
clearly seen on this scale in the solid curve, while the two lower periodic potential also provides a relevant description in a
curves are monotonic. The inset shows the maximum value of  large number of other physical situations [16], including
as a function of , with 0    1=2. The two panels in (b) are Josephson junctions, rotating dipoles in external fields,
for the periodic square-well potential, and those in (c) are for the particle separation by electrophoresis, transport in tubes
sawtooth potential. For the former  0:1 in the left panel, 0.5 of varying cross section, and biophysical processes such as
in the right panel. In both panels, ? is shown for three barrier neural activity and intracellular transport.
heights: Ua 1 (solid), 3 (dashes), and 9 (dots). For the Turning to the issue of efficiency at maximum power, we
sawtooth potential the curves in the left panel are all for Ua first derive results that hold for arbitrary potential. We
10 and in the right panel for Ua 100. The four curves are for
suppose that the average velocity V can be written as a
different values of the potential asymmetry parameter :  0
(long dashes), 0.1 (solid), 0.5 (short dashes), and 0.9 (dots).
power series in F. Since V vanishes for F 0, there is no
constant term in the expansion. For comparison with the
previous results, we consider the rate k V=L for moving
of 1 and 2 corresponds to a replacement of  by 1  . over one period L, and write the corresponding power
Hence the above theory indicates that, when considering series in terms of  FL, k k V=L
both modes of operation, at least one of them has   1=2, a1  a2 2 a3 3 O4 . Maximization of the output
with a corresponding efficiency at maximum power less power  k2 with respect to the loading energy 2
than 50%. For  1=2 the engine works equally well at gives the following expansion for the output yield at maxi-
maximum power in forward and reverse modes, at 50% mum power [compare with Eq. (3)]:
efficiency.
 2 
Generalized model.The above generic model assumes a2 2 a a
2    2 22  3 3 O4 : (8)
exponentially difficult crossing of the transition state. With a1 a1 a1
a more general analysis, we suppose that the motion
projected on reaction coordinate x can be described as an The corresponding efficiency reads [see Eq. (5)]:
overdamped one-dimensional diffusion process in a poten-
tial Ux U0 x  Fx: 1 a 3a22  4a1 a3 2
?  2   O3 : (9)
dxt dUxt p 2 4a1 8a21
  2kB T t: (7)
dt dx This expansion features the familiar 50% efficiency in
Here  is the viscous friction coefficient and  is Gaussian the regime of linear response. Turning to the nonlinear
white noise, hti 0, and htt0 i t  t0 . The regime, we note that just as in the generic two-state model,
diffusion coefficient is related to the viscous friction the next order correction vanishes (a2 0) whenever

210602-3
week ending
PRL 108, 210602 (2012) PHYSICAL REVIEW LETTERS 25 MAY 2012

the system has left-right symmetry for the velocity, VF e  1


VF. ? ;
e  13  
The average steady state velocity for overdamped (13)
e  1 c1 e e1  e
motion in a tilted periodic potential is given by [17] ;
1    e c1  e1   1e 
DL1  e where c  4sinh2 U
=2. The efficiency ? for the
V RL RL : (10)
0 dx 0 dyeU0 xU0 xyy=L square-well and sawtooth potentials are shown in
Figs. 2(b) and 2(c). We stress again that the bounds are
This expression in principle allows us to find the power the same for the discrete and continuous models, pointing
 V2 =L, and hence makes it possible to explicitly to the universality of these bounds [18].
identify the regime of maximum power and its correspond- Finally, we mention a generic behavior for any potential
ing efficiency. with finite maxima. For very large driving, the barrier(s) of
First, we rederive the results for the two-state model the potential U0 , indeed the entire potential, become ir-
with a potential U0 with a dominant high maximum in each relevant and one goes back to a linear model with V
period, say at x xa L (modulo L), and a unique F= a1  with a1 D=L, implying that ? 1=2. This
minimum at x 0 (modulo L). The dominant contribution return to the linear scenario can be seen in some cases in
to the double integral in Eq. (10) comes from the region Fig. 2; in others one must go to higher values of  than
around the (x, y) point for which U0 x y reaches a those shown in the figure.
maximum and U0 x a minimum. This point lies at x Closing perspective.Is the issue discussed in this
y  xa L and x 0, and consequently y L. The Letter, maximizing power with respect to the load, a rele-
 dependence of the denominator is therefore of the form vant criterion in practice? In the case of thermal motors, this
exp. It then follows directly that V k f1  seems to play a role, at least from an engineering point of
expg exp, and the resulting power is identical to view, since power plants operate under conditions in gen-
that for the two-state model. eral agreement with this criterion [11]. We hope that the
Second, we note that universal conclusions can be drawn present Letter will lead to a reexamination, from the per-
using Eq. (10) even without an explicit evaluation of the spective of maximum power, of the much larger class of
integrals. In particular, we can identify the coefficients ai isothermal engines, including the important class of mo-
in Eq. (9), lecular motors. Experimental verification of our theoretical
predictions should be possible with experiments focused on
D 1 D 2I1  I0 the question of efficiency at maximum power.
a1 ; a2 ;
L2 I0 L2 2I02 (11) This work was partially supported by the U.S. National
D I 2 6I12  3I0 I1 2I2 Science Foundation under Grant No. PHY-0855471.
a3 2 0 ; Note added in proof.Recently, we became aware of
L 6I03 another paper related to the present problem [19].
where

1 ZL ZL
In n2
dxeU0 x dyyn eU0 xy : (12) [1] H. B. Callen, Thermodynamics and an Introduction
n!L 0 0
to Thermostatistics (Wiley, New York, 1985), 2nd ed.
[2] C. Van den Broeck, Phys. Rev. Lett. 95, 190602
Since 0  I1  I0 , the coefficient a2 =4a1 I1 =I0  (2005).
1=2=4 of the linear term lies between 1=8 and 1=8, as [3] F. Curzon and B. Ahlborn, Am. J. Phys. 43, 22 (1975).
was the case for the 2-state model, see Eq. (5). [4] T. Schmiedl and U. Seifert, Europhys. Lett. 81, 20003
Furthermore, the coefficient is zero for a potential with (2008).
left-right symmetry, that is, when there exists a point x0 for [5] Z. C. Tu, J. Phys. A 41, 312003 (2008).
which U0 x  x0 U0 x0  x. [6] A. E. Allahverdyan, R. S. Johal, and G. Mahler, Phys. Rev.
To proceed further, it is in general necessary to invoke E 77, 041118 (2008).
numerical calculations because the integrals in (10) cannot [7] Y. Izumida and K. Okuda, Europhys. Lett. 83, 60003
be performed analytically for a general potential U0 x. (2008).
[8] M. Esposito, K. Lindenberg and C. Van den Broeck,
However, analytic results can be obtained in some limits
Europhys. Lett. 85, 60010 (2009).
(such as the dominant high maximum case considered [9] M. Esposito, R. Kawai, K. Lindenberg, and C. Van den
above), or for some specific shapes of the potential. Broeck, Phys. Rev. E 81, 041106 (2010).
Examples include the square well, U0 x Ua , x 2 [10] M. Esposito, K. Lindenberg, and C. Van den Broeck, Phys.
0; L modulo L, and U0 x 0, x 2 L; L modulo Rev. Lett. 102, 130602 (2009).
L, and a sawtooth potential. To illustrate, we quote the [11] M. Esposito, R. Kawai, K. Lindenberg, and C. Van den
efficiency at maximum power for the square-well potential, Broeck, Phys. Rev. Lett. 105, 150603 (2010).

210602-4
week ending
PRL 108, 210602 (2012) PHYSICAL REVIEW LETTERS 25 MAY 2012

[12] U. Seifert, Phys. Rev. Lett. 106, 020601 (2011). [16] P. Reimann, Phys. Rep. 361, 57 (2002).
[13] J. Howard, Mechanics of Motor Proteins and the [17] R. L. Stratonovich, Topics in the Theory of Random Noise
Cytoskeleton (Sinauer Associates, Sunderland, MA, (Gordon and Breach, New York, 1963), Vol. 1.
2001). [18] C. Van den Broeck, N. Kumar, and K. Lindenberg (to be
[14] T. Schmiedl and U. Seifert, Europhys. Lett. 83, 30005 published).
(2008). [19] N. Golubeva, A. Imparato, and L. Peliti, Europhys. Lett.
[15] O. K. Dudko, T. G. W. Graham, and R. B. Best, Phys. Rev. 97, 60005 (2012).
Lett. 107, 208301 (2011).

210602-5

Você também pode gostar