Entropy 2010, 12, 996-1005; doi:10.

3390/e12050996
OPEN ACCESS
entropy
ISSN 1099-4300
www.mdpi.com/journal/entropy
Article
The Maximum Entropy Production Principle and Linear
Irreversible Processes
Pa sko

Zupanovi c
1,
, Domagoj Kui c
1
,

Zeljana Bona ci c Lo si c
1
, Dra zen Petrov
1
, Davor Jureti c
1
and Milan Brumen
2
1
Faculty of Science, University of Split, Teslina 12, 21000 Split, Croatia
2
Faculty of Natural Sciences and Mathematics, University of Maribor, Koro ska cesta 160, SI-2000
Maribor, Slovenia

Author to whom correspondence should be addressed; E-Mail: [email protected];

Tel.: +385-21-385-133; Fax: +385-21-384-086.
Received: 13 March 2010 / Accepted: 23 April 2010 / Published: 27 April 2010
Abstract: It is shown that Onsagers principle of the least dissipation of energy is equivalent
to the maximum entropy production principle. It is known that solutions of the linearized
Boltzmann equation make extrema of entropy production. It is argued, in the case of
stationary processes, that this extremum is a maximum rather than a minimum.
Keywords: entropy production; linear nonequilibrium thermodynamics; linearized
Boltzmann equation
Classication: PACS 65.40.gd
1. Introduction
It was Ehrenfest (Enzykl. Math. Wissensch, IV, 2(II) fasc.6, p82, note23, 1912) who rst asked
whether a function exists which, like entropy in the equilibrium state of an isolated system, achieves its
extreme value in a stationary non-equilibrium state.
There are various results and formulations of irreversible thermodynamics based on the extremum of
entropy production. They are related either to the minimum or maximum entropy production. The well
known example based on minimum entropy production is Prigogine theorem [1].
Entropy 2010, 12 997
There are several different results that concern maximum entropy production (MEP). Ziegler applied
the principle of maximum entropy production in thermomechanics [2]. It states that the rate of entropy
production under prescribed forces should be maximum. Paltridge [3] has formulated empirically
MEP principle in order to describe the Earth climate. Apart from details of his model (like albedo
or cloudiness) the main features of his model are steady incoming sun radiation, outgoing radiation
proportional to the fourth power of the local annual average temperature and horizontal heat ux from
equator to poles. There are no boundary conditions on atmosphere. Paltridge has proposed that steady
state is the state of maximum entropy production due to the latitudinal heat transport. His predictions of
the annually average distribution of the temperature of atmosphere (and some other climatic parameters)
t very well measured values. We note that Paltridge MEP principle is applied to nonlinear process and
is different from the linear problems considered in this paper. Chavanis [4,5] has applied MEP principle
to the problem of two dimensional turbulence and stellar systems in order to nd optimal currents.
Connection between MEP principle and Fokker-Planck equation is established, too.
Kohler [6] starting from the Boltzmanns transport equation and using the denition of the entropy as
the negative integral of product of distribution function and its natural logarithm in molecular velocity
space has found that entropy production can be expressed via scattering integral. It comes from this
denition that molecular collisions in nonequilibrium state of rare gas are the sources of entropy
production. He has shown that the stationary state close to the equilibrium is the state of extremum
entropy production. If one requires that entropy production dened via scattering integral is equal to
the entropy production dened as the product of the thermal ux and thermal force then this extremum
is maximum one. The another constraint of xed thermal ux, also used by Kohler, that leads to the
conclusion that stationary state of rare gas is the state of minimum entropy production is put in question
in this paper.
Recently Jaynes principle of maximum information entropy (MaxEnt) has been exploited to derive
the MEP principle. Dewar [7,8] has introduced paths as possible trajectories of the system in phase
space. Then using the MaxEnt procedure he has shown, under some assumptions, that the most probable
development of the system is accompanied with MEP. Niven [9] has applied MaxEnt using the values of
uxes as variables and he found that the stationary state of the system is the MEP state.
At rst sight, minimum and maximum entropy production results seem to contradict each other. But
if one considers starting assumptions one nds that these assumptions are different. Thus, these results
are independent of one another [10,11]
In this paper we focus on Onsagers principle of least dissipation of energy and showthat this principle
is equivalent to MEP principle. Secondly, as we have already noted, we show that Kohler interpretation
of stationary state of rareed gas close to the equilibrium should be restricted only on the constraint that
it is consistent with the interpretation of this state as the MEP state.
This is the rst of three papers in a series and are referred to as papers I, II and III. In paper II we
discuss whether stationary or relaxation processes are suitable for the formulation of principles. We
argue in favor of relaxation processes. In paper III we apply MaxEnt formalism to relaxation processes
and we derive the MEP principle as its corollary.
Entropy 2010, 12 998
2. The Principle of the Least Dissipation of Energy and Linear NonequilibriumThermodynamics
Onsagers famous papers [12,13] examine linear nonequilibrium thermodynamics. The rst sections
of these papers examine reciprocal relations. The following sections are devoted to the formulation
of linear nonequilibrium thermodynamics. Onsager has employed the principle of the least dissipation
of energy. In the second part of his second paper [13], Onsager has applied the principle of the least
dissipation energy to the general linear nonequilibrium process.
A linear relationship between n independent thermodynamic forces {X
i
} and their conjugated
uxes {j
i
},
X
i
=

j
R
ij
j
j
, (1)
exists for the system close to equilibrium. The density of the entropy production is equal to the product
of the thermodynamic forces and conjugated uxes [1], [14]. The total entropy production is
=

j=1
X
j
j
j
dr (2)
In his second famous paper Onsager [13] has introduced the dissipation function
=
1
2

i,j=1
R
ij
j
i
j
j
dr (3)
and formulated the principle of the least dissipation of energy,
= maximum (4)
The variational procedure
[ ] = 0 (5)
gives the system of equation (1). In other words, the linear irreversible thermodynamics can be inferred
from the principle of the least dissipation of energy.
3. The MEP Principle and Linear Nonequilibrium Thermodynamics
An alternative approach to linear nonequilibrium thermodynamics is based on the phenomenological
fact that nonequilibrium processes are characterized by uxes. Therefore, physical quantities relevant for
a description of the time development of the system must be functions of uxes. The standard approach
to the nonequilibrium thermodynamics is based on the laws of conservation of mechanical physical
quantities mass, momentum and energy. The next step is to add heat as the additional mechanism of
the exchange of energy between systems. Then assuming local equilibrium one comes to the density
of entropy production written as the sum of products of heat ux and thermal thermodynamic force
and viscous pressure tensor and strain tensor divided by temperature. This result serves as the basis
for the canonical form of the second postulate of irreversible thermodynamics that reads: Entropy
production can be always written as the sum of products of thermodynamic forces and corresponding
uxes [14,15]. Thermodynamic forces and uxes are not independent quantities. In principle, uxes
determine forces and vice versa. Having in mind this fact and the second postulate of irreversible
Entropy 2010, 12 999
thermodynamics, we can say that the dynamical state of the system is described only by uxes, or
equivalently by thermodynamic forces.
Entropy production is a basic, characteristic quantity of a nonequilibrium state. If the system is close
to the equilibrium state we can make the Taylor expansion of the density of entropy production d/dV
in uxes up to the second order,
d
dV
= A +

i
B
i
j
i
+

i,j
R
ij
j
i
j
j
(6)
Here j
i
is the mean value of the i
th
ux. The rst term on the right-hand side of this equation vanishes
since there is no entropy production in the equilibrium state. Coefcients B
i
and R
ij
are the property of
the system in equilibrium state. Entropy production does not depend on the direction of the ux ow,
that is to say, it must be invariant under the replacement of {j
i
} {j
i
}. This means that coefcients
that multiply the odd power of uxes vanish, that is, B
i
= 0. In the lowest order, entropy production is
the bilinear function of uxes,
=

i,j

V
R
ij
j
i
j
j
dr > 0 (7)
Comparison of Equations (7) and (3) shows that the dissipation function is in fact entropy production
written in the space of uxes. When a system is close to equilibrium with locally well dened intensive
thermodynamic quantities the entropy production is given by Equation (2), [1,14].
It is pointed in the introductory part of this Section that canonical form of entropy production (2)
comes from the law of conservation of energy. Then equation,

i
X
i
j
i

i,j
R
ij
j
i
j
j

dr = 0 (8)
is the constraint imposed at the entropy production due to the law of conservation of energy.
We seek the maximum entropy production (7) taking into account the constraint (8). The standard
procedure [16] is to nd the maximum of the functional
F =

i,j

V
R
ij
j
i
j
j
dr +

i
X
i
j
i

i,j
R
ij
j
i
j
j

dr (9)
A standard variational calculus of extremum values combined with the constraint (8) gives = 2 and
F becomes
F =

i,j

V
R
ij
j
i
j
j
dr + 2

i
X
i
j
i
dr
This is an equation of a quadratic surface turned upside down. The corresponding extremum of F is
maximum. We refer an interested reader to reference [20] where it is explicitly shown that this extremum
is maximum. The system of equations that determines the point of maximum , F/j
i
= 0, is just a
system of linearly coupled thermodynamic forces and uxes (1). Similar approach that relates MEP
principle and Onsagers variational formulation of the linear nonequilibrium thermodynamics can be
found in reference [17].
From Equation (7) follows R
ij
= R
ji
i.e., Onsagers reciprocal relations. In short, linear
nonequilibrium thermodynamics and Onsagers reciprocity relations follow from the MEP principle.
Entropy 2010, 12 1000
A problem analogous to this is the problem of biochemical cycle kinetics close to the equilibrium
state. Starting from the assumption that the energy conservation law is valid for a whole network of
biochemical reactions we have shown that uxes are distributed in such a way to produce maximum
entropy [18].
In references [19,20] we have considered a linear planar electric network held at a constant
temperature. There is only a temperature gradient between network and surroundings. Physically this
can be achieved using thermally high and poor conducting material for a network and surroundings,
respectively. There is no coupling between electric and heat currents. Only electric currents are coupled
via electromotive forces. Assuming that the stationary state is the state of maximum possible generated
heat the Kirchhoff loop law is derived [19,20]. The principle of the maximum heat dissipation is closely
related to the MEP principles.
Using mesh currents the law of charge conservation has been taken implicitly. The energy
conservation law is used explicitly as the constraint. If one only takes charge conservation law one
comes to the conclusion that stationary state is state of minimum generated heat [21]. We shell dwell
more on this question in the paper II submitted to this special issue.
We note that two of us (P.

Z. and D.J.) have considered the heat ow in the anisotropic crystal. It is

shown in this special example that the principle of the least dissipation of energy applied by Onsager in
reference [12] is equivalent to the MEP principle [22].
4. The Linearized Boltzmann Equation and the Extremum of Entropy Production
The important inuence of constraints at extrema of entropy production can be seen in the case of
linearized Boltzmann equation.
Here we follow the elegant approach given in reference [11]. The Boltzmann equation is valid
for rareed gas where collisions are very well dened events. One seeks the one-particle distribution
function f(r, v, t) satisfying the equation,
df
dt
=
f
t
+ v
f
r
+
F
m
f
v
= I (10)
Here, the two terms on the left-hand side of the equation describe the change in the number of
molecules in a given element of the phase space due to the collisionless motion of the molecule in the
outer eld F. The right-hand side of the equation describes the net change in the number of molecules
in a given element of the velocity space due to molecule collisions.
Assuming the instantaneous change of molecule velocities in the collisions and taking into account
the conservation laws in collisions the integral can be written in the form [23]
I =

(

f

f

ff

)qd(q, e)dv

(11)
Here, f = f(r, v, t) ; f

= f(r, v, t) and

f = f(r, v, t) ;

f

= f(r, v

, t) are the distribution functions

of particles before and after collision, respectively. In addition, q = v

v is the magnitude of relative

velocity of particles before collisions, e = ( v v)/| v v| is a unit vector parallel to the velocity change
of one particle in collision and d(q, e) is the differential cross section. It is known from collision theory
Entropy 2010, 12 1001
that differential cross section depends on interaction potential, relative velocity q and unit vector e, i.e.,
on the scattering angle [23].
In the case of local equilibrium, intensive variables (temperature, concentration) are well dened
functions of the space. Then one assumes the approximate solution of Equation (10) [6,11].
f = f
0
(1 + (V)) (12)
where
f
0
= n(
m
2kT
)
3/2
exp(
mV
2
2kT
) (13)
is the Maxwell-Boltzmann distribution and V is the velocity relative to the barycentric (centre
of mass) velocity.
Due to the assumption of local equilibrium the perturbed function must not contribute to the
prescribed intensive parameters like temperature, mean velocity and density. Then

f
0
(V)dV = 0,

f
0
(V)V dV = 0,

f
0
(V)
mV
2
2
dV = 0 (14)
Now the collision integral up to the rst order of the perturbed term becomes
I(f) =

f
0
f

0
(

)qd(q, e)dv

=

O (15)
The designation of functions is the same as the distribution functions in Equation (11).
Operator

O [6,11] is linear and has the following properties

O(A + B) =

OA +

OB (16)

A

OBdV =

B

OAdV (17)

A

OAdV 0 (18)
If we designate the left hand side of Boltzmann equation (10) with Z it becomes
Z =

O (19)
Multiplying the linearized Boltzmann equation with k ln f and integrating over velocity space
we get
s
t
+ j
s
= k

OdV (20)
Here s = k

f ln fdV is the density of entropy and j

s
= k

Vf ln fdV is the ux of entropy.

Entropy looks like a uid. The left-hand side of Equation (20) is the total time change of the density of
entropy, i.e. it is the density of the entropy production. Then, according to the aforementioned equation
collisions between molecules described by function are the sources of entropy production. There
are a lot of functions obeying conditions (14). We choose those distribution functions that obey the
additional condition

Y ZdV =

Y

OY dV (21)
Entropy 2010, 12 1002
Function Y is not the solution of the Boltzmann equation, i.e.,
Z =

OY (22)
Y can be interpreted as a trial distribution function. The corresponding entropy production due to this
distribution is
s
Y
t
+ (j
s
)
Y
= k[Y,

OY ] (23)
If the temperature gradient in x direction is the only thermodynamic force, the left-hand side of the
equation (21) can be written as [6]

Y ZdV =
(q
x
)
Y
T
2
T
x
(24)
Here, (q
x
)
Y
is the heat ux associated with the distribution function Y . This equation gives a physical
meaning to condition (21). Varied distribution functions are selected in such a way that the entropy
production due to the molecular collisions is equal to the entropy production due to the heat conduction.
In other words entropy produced by molecular collisions is equal to the product of the heat ux and
conjugated thermodynamic force. This condition is in accordance with the starting assumption of a
stationary process.
The multiplication of the Boltzmann equation with the perturbative distribution function and
integration over molecular velocity space V gives

ZdV =

OdV (25)
Using conditions (1618) we get,

(Y )

O(Y )dV 0 (26)

OdV +

Y

OY dV

OY dV

Y

OdV =

OdV+

Y

OY dV2

Y

OdV 0 (27)

OdV +

Y

OY dV + 2

Y ZdV =

OdV+

Y

OY dV2

Y

OY dV =



OdV

Y

OY dV 0 (28)

OdV

Y

OY dV (29)
The left-hand side is the entropy production of the rareed gas. It comes from Equation (29) that
solutions of the Boltzmann equation are in accordance with MEP principle.
There is another approach to this problem which has been proposed by Kohler [6]. This approach
assumes xed heat ux,
q
x
= constant. (30)
We note that due to the already xed temperature eld the entropy production is also xed. In this
Kohlers approach the entropy generated by molecular collisions need not be equal to the entropy
Entropy 2010, 12 1003
production calculated as the product of thermal force and thermal ux. Since the entropy production
calculated via scattering integral is not equal, in the variation process, to the entropy production dened
as the product of the thermal force and thermal ux, it comes out that the constraint of the constant
thermal ux (30) violates the equality between these two denitions of entropy production. If one does
not take care of this violation the conclusion can be reached, by using the calculus of variation, that
solutions of the Boltzmann equation are those that generate minimum entropy production.
5. Conclusions
The Onsager principle of the least dissipation of energy is valid for processes close to the equilibrium
state. In this paper, the equivalence between the Onsager principle of the least dissipation of energy and
the MEP principle is established. Starting from the fact that uxes are the main phenomenological
characteristic of irreversible processes, we have expanded the density of entropy production as the
function of uxes up to the second order. We have found that the dissipation function introduced by
Onsager is the entropy production in the space of the uxes. Invoking the rst law of thermodynamics,
the equivalency between the Onsager principle of the least dissipation energy and the MEP principle
is established.
It follows from the Boltzmann equation that entropy production is quantitatively related to the
collision integral. Collisions between molecules in nonequilibrium state produce entropy. In this
paper the solutions of the linearized Boltzmann equations are considered. It is found that these
solutions correspond to the extremum of entropy production. The nature of the extremum depends
on the constraints. Assuming that the entropy produced by molecular collisions is equal to the entropy
productions due to the heat conduction or/and viscosity, we nd that the solutions of the Boltzmann
equation satisfy the MEP principle.
The principle of minimum entropy production is valid if one grants xed uxes. However, the starting
assumption of the xed thermodynamic forces and the additional assumption of xed uxes leave no
room for variation. Although both approaches are equivalent from the mathematical point of view one
has to notice that the assumption of xed uxes violates the equality between entropy production dened
via collision integral and as the sum of products of thermodynamic forces and conjugated uxes.
In short, revisiting the linear nonequilibrium thermodynamics and linearized Boltzmann equation
shows that both approaches are in accordance with the MEP principle. Generally we can conclude that
the MEP principle is valid for processes close to the equilibrium state.
Acknowledgements
The present work was supported by the bilateral research project of the Slovenia-Croatia Cooperation
in Science and Technology, 2009-2010 and Croatian Ministry of Science grant No. 177-1770495-0476
to DJ.
Entropy 2010, 12 1004
References
1. Prigogine, I. Introduction to Thermodynamics of Irreversible Processes; Interscience Publishers:
New York, NY, USA, 1967.
2. Ziegler, H. An Introduction to Thermomechanics; North-Holland publishing company:
Amsterdam, The Netherlands, 1983.
3. Paltridge, G.W. The steady-state format of global climate. Q.J.R. Meteorol. Soc. 1978, 104,
927945.
4. Chavanis, P.H.; Sommeria, J.; Robert, R. Statistical mechanics of two-dimensional vortices and
collisionless stellar systems. Astrophys. J. 1996, 471, 385399.
5. Chavanis, P.H. Generalized thermodynamics and Fokker-Planck equations. Applications to stellar
dynamics and two-dimensional turbulence Phys. Rev. E, 2003, 68, 036108.
6. Kohler, M. Behandlung von Nichtgleichgewichtsvorg angen mit Hilfe eines Extremalprinzips.
Z. Phys. 1948, 124, 772789.
7. Dewar, R. Information theory explanation of the uctuation theorem, maximumentropy production
and self-organized criticality in non-equilibrium stationary states. J. Phys. A: Math. Gen. 2003,
36, 631641.
8. Dewar, R.C. Maximum entropy production and the uctuation theorem. J. Phys. A: Math. Gen.
2005, 38, L371L381.
9. Niven, R.K. Steady state of a dissipative ow-controlled system and the maximum entropy
production principle. Phys. Rev. E 2009, 80, 021113-021127.
10. Ziman, J.M. The general variational principle of transport theory. Can. J. Phys. 1956, 34,
12561273.
11. Martyushev, L.M.; Seleznev, V.D. Maximum entropy production principle in physics, chemistry
and biology. Phys. Rep. 2006, 426, 145.
12. Onsager, L. Reciprocal Relations in Irreversible Processes. I. Phys. Rev. 1931, 37, 405426.
13. Onsager, L. Reciprocal Relations in Irreversible Processes. II. Phys. Rev. 1931, 38, 22652279.
14. De Groot, S.R.; Mazur, P. Non-Equilibrium Thermodynamics; North-Holland Pub. Co.:
Amsterdam, The Netherlands, 1962.
15. Evans, D. J.; Morris, G.P. Statistical Mechanics of Nonequilibrium Liquids; ACADEMIC PRESS:
London, UK, 1990.
16. Krasnov, M.I.; Makarenko, G.I.; Kiselev, A.I. Problems and Exercises in the Calculus of
Variations; Mir Publishers: Moscow, Russia, 1975.
17. Chavanis P.H. Nonlinear mean eld Fokker-Planck equations. Application to the chemotaxis of
biological populations. Eur. Phys. J. B 2008, 68, 179208.
18.

Zupanovi c, P.; Jureti c, D. The chemical cycle kinetics close to the equilibrium state and electrical
circuit analogy. Croat. Chem. Acta 2004, 77, 561571.
19.

Zupanovi c, P.; Jureti c, D.; Botri c, S. Kirchhoffs loop law and the maximum entropy production
principle. Phys. Rev. E 2004, 70, 056108: 15.
Entropy 2010, 12 1005
20. Botri c, S.;

Zupanovi c, P.; Jureti c, D. Is the stationary current distribution in a linear planar electric
network determined by the principle of maximum entropy production? Croat. Chem. Acta 2005,
78, 181184.
21. Jaynes, E.T. The minimum entropy production principle. Ann. Rev. Phys. Chem. 1980, 31,
579601.
22.

Zupanovi c, P.; Jureti c, D.; Botri c, S. On the equivalence between Onsagers principle of the
least dissipation of energy and maximum entropy production principle: Conduction of heat in
an anisotropic crystal. FIZIKA A 2005, 14, 8996.
23. Rumer, Yu.B.; Ryvkin, M.S. Thermodynamics, Statistical Physics, and Kinetics; Mir Publishers:
Moscow, Russia, 1980.
c 2010 by the authors; licensee MDPI, Basel, Switzerland. This article is an open-access
article distributed under the terms and conditions of the Creative Commons Attribution license
(http://creativecommons.org/licenses/by/3.0/).