Physica A 341 (2004) 165 170
www.elsevier.com/locate/physa
Possible forceentropy correlation
Enrique Canessa
The Abdus Salam International Centre for Theoretical Physics, Trieste, Italy
Received 7 February 2004; received in revised form 13 March 2004
Abstract
A statistical thermodynamic approach of moving particles forming an elastic body is presented
which leads to reveal molecularmechanical properties of classical and nonextensive dynamical
systems. We derive the BoltzmannGibbs (BG) entropy form and relate it to Newtons law of
motion in relation to a distinct tensile force acting on the systems at constant volume and number
of particles. Tsallis generalization of the BG entropy is deduced assuming the thermal energy of
the particles to be proportional to their energy states by the nonextensivity factor q 1.
c 2004 Elsevier B.V. All rights reserved.
PACS: 05.70.Ce; 05.70.Ln; 05.70.a
Keywords: Statistical physics and thermodynamics; Probability theory; Nonlinear dynamics; Nonextensive
statistical mechanics
1. Overview
It was pointed out long ago by Einstein that the foundations of statistical mechanics
should lie on the underlying dynamics [1]. All statistical mechanics results should
derive from the dynamics if the systems considered are systems consisting of moving
particles.
Within the classical BoltzmannGibbs (BG) statistics, it is still unknown how to
derive from say Newtons law of motion the remarkable expression for the thermodynamic entropy [2]
W
1
SBG =kB =
;
(1)
pi ln pi = ln
pi
i=1
c 2004 Elsevier B.V. All rights reserved.
0378-4371/$ - see front matter
doi:10.1016/j.physa.2004.03.099
166
E. Canessa / Physica A 341 (2004) 165 170
where mean values ( ) correspond to i ( )pi . Even if this functional form has
been profusely veriDed on isolated systems in thermal equilibrium, a rigorous derivation
on how SBG descends from dynamics has been limited to a dilute gas only [1]. It is
usually approximated via the statistics of a large ensemble of identical replicas (see,
e.g., Ref. [3]). Only recently, there has been computational evidence on the fact that
BG thermal equilibrium distribution descends from Newton law [4].
Also unclear at present [1,2], is the connection between the dynamics of a particle
system and the Tsallis nonextensive generalization of the BG entropy
W
W
1 i=1 piq
1
1
Sq =kB = lnq
=
pi lnq =
;
(2)
pi
pi
q1
i=1
where
pi1q 1
;
1q
lnq pi
(3)
which for q = 1, it coincides with SBG giving positive Lyapunov exponents [5]. The
nonextensivity is given by the pseudo-additivity property Sq (A + B) = Sq (A) + Sq (B) +
((1 q)=kB )Sq (A)Sq (B); A and B being two-independent systems.
In order to shed some light on this physics dilemma, we propose a simple statistical
thermodynamic model where moving particles of known type are assumed to form
an elastic specimen (or body). It is shown that this type of approach leads to reveal
intrinsic molecularmechanical properties on classical and nonextensive dynamical systems in relation to a distinct tensile force acting on these systems at constant volume
and number of particles. In particular we derive the entropy of Eq. (1) and relate it
= ma. A new scenario for the entropic index q in Tsallis
to Newtons second law F
statistics in terms of the energy of the system is proposed.
2. Formalism
Let us consider a system of noninteracting moving particles and let pi be the probability that the system is in the microstate i(=1; : : : ; W ) given by the product of two
positive functions ui and wi satisfying the normalization condition
W
pi =
i=1
W
ui w i = 1 :
(4)
i=1
Derivative of this relation with respect to changes in particle displacements x, considering particles of mass m to move at velocity v = 9x=9t and acceleration a = 9v=9t,
gives
W
9pi
i=1
9x
W
i=1
pi
9
ln pi = 0
9x
(5)
E. Canessa / Physica A 341 (2004) 165 170
or, alternatively,
W
W
9ui pi
9wi pi
=
:
9x ui
9x wi
167
(6)
i=1
i=1
(9pi =9x) ln ui it is straightforward to show that
W
9pi
9
9
ln pi :
(7)
ln ui = ln wi +
9x
9x
9x
i
Summing up both sides by
i=1
Let us also consider the purely thermodynamic relation for the Helmholtz free energy
A = E TS, assuming the temperature T and volume V to be independent variables
with E the internal energy, A the Helmholtz free energy and S the entropy of the
system. Therefore from Eq. (7) we can readily identify (or, alternatively, deDne) three
quantities
A=kB T ln ui =
W
pi ln ui ;
(8)
i=1
E=kB T ln wi =
W
pi ln wi ;
(9)
i=1
W
9
(S=kB )
9x
i=1
9pi
9x
ln pi ;
(10)
with kB a constant added to have reduced units. The choice of the sign for E is to
approximatethe system energy as the sum of individual contributions (or mean energy)
E E= i pi i , where i =kB T =ln wi represents their energy states, and thus obtain
wi = exp(i =kB T ) 0 :
(11)
This relation is also found to be the solution of the diIerential equation: (9 ln(pi =ui )=
9(i =kB T )) = 1.
Furthermore, from Eqs. (5) and (10) it follows immediately that the entropy satisDes
W
9
9
(12)
pi ln pi ;
(S=kB ) =
9x
9x
i=1
which is the basic result stated by the BG entropy of Eq. (1).
Since the moving particles are assumed to form an elastic body, then the total work
along an inDnitesimal elongation d is given by dW = dA = f d for a change at
constant T , with f a (compressive 0; or expansive 0) tensile force in the direction
of . This relation is obtainable directly from the Drst and second laws of thermodynamics in a reversible isothermal process [6], assuming that there is not another force
acting on the system in addition to f as for example due to hydrostatic pressure P
(dW = PdV 0; recall volume V here is an independent variable). Therefore, the
168
E. Canessa / Physica A 341 (2004) 165 170
dynamics of our particle system relating thermodynamic quantities reduces to
W
9A
9S
9
9E
F=
T
:
= kB T
pi ln ui =
9x T
9x T
9x T
9x
i=1
(13)
Thus through Eqs. (8) and (9) the system entropy can be correlated to Newtons law
of motion in relation to a distinct tensile force acting on the system.
3. Remarks
At a Drst glance the choice for pi in Eq. (4) may be seen quite arbitrary, but in
view of the results derived for S, A and E starting only from this quantity it seems
to be reasonable. As shown below, this choice can further be justiDed in relation with
the measure for an inverse temperature for the system. Using Eqs. (4), (9), (11) and
(12), we obtain at once
(S=kB )
kB T
ln uW
ln uW ;
=1
(14)
=1+
W
(E=kB T )
ln wW
where represents functions diIerence between consecutive microstates Kuctuations.
Although we do not know how each of our ui and wi functions may depend explicitly
on the displacement x, we can still estimate all thermodynamic relations from given
average values of either u or w at each state i. For example, interesting to note are the
results obtained when settings ui = 1=Z, a sort of mean value for u(i) and wi variable.
From Eqs. (4) and (11), it follows then that pi =(ei =kB T =Z), which corresponds to the
Gibbs distribution in thermal equilibrium and satisDes dpi =di = pi =kB T
. Using the
normalization condition for pi we derive in turn the partition function Z = i ei =kB T ,
where the summation is over all microstates of the system. From this result and that
of the energy states i plus Eq. (9), we obtain a relation for the mean energy of a
system at temperature T as E = (9 ln Z=9(1=kB T )). For the Helmholtz free energy
of Eq. (8) we can write A=kB T = ln Z, which is also a well-known classical statistical
mechanics result [3].
Let us analyse next a second case, i.e., the case of having an averaged w(i) and
ui as the variable. To derive Tsallis statistics we assume the thermal energy of the
particles kB T to be proportional to their energy states by the nonextensivity (integer)
factor q 1 for all the i-microstates. From Eq. (11) and q = 1 this is equivalent to
consider
ln wi =
1
:
1q
(15)
Using the lnq deDnition of Eq. (3) in conjunction with Eqs. (4) and (15), we then get
the relation lnq (1=pi ) ln(pi =ui ) = (piq1 =q 1). For q 1 and by Taylor expansion
[2] we can also approximate piq = pi pi(1q) = pi exp(ln pi(1q) ) pi [1 + (1 q)
E. Canessa / Physica A 341 (2004) 165 170
169
ln pi1 ]. Thus in this limit lnq pi1 ln pi1 . These results imply that
ln ui
piq1
:
q1
(16)
Therefore within our statistical thermodynamic approach it follows that
W
W
piq1
1
; E=kB T
= =kB T ;
pi
pi
A=kB T
q1
1q
i=1
(17)
i=1
whence
S=kB = E=kB T A=kB T
W
i=1
q1
piq
(18)
which corresponds to the entropy term of Eq. (2) introduced by Tsallis. As in
Ref. [2], if we assume equiprobability pi = 1=W (i) we also obtain S = lnq W .
Possible values of the q index are constrained by the logarithm functions of
Eqs. (15) and (16). The condition 0 6 pi ui wi 6 1(i) implies that 0 6 exp(1=(1
q)) exp(piq1 =(q 1)) 6 1. Hence we can deduce that !i (q) (piq1 1)=(q 1) 6 0,
due to the positive and continue nature of the exponential function. Then the limiting behaviour of this function as q 1 is obtained by LHopitals rule because
this is an indeterminate form of type =. We found limq1 !i (q) = ln pi , hence
0 6 exp(ln pi ) 6 1. From this result, we also have 0 6 piq1 6 1(i; q), which via
Eq. (15) it implies that ln wi 6 !i (q) 6 0. This, in turn, implies a positive wi in agreement with Eq. (11) such that 0 wi 6 1. Therefore, we can conclude that the range
of validity of our statistical thermodynamic model for Tsallis statistics is q (; 1].
On the other hand, using Eqs. (15) and (16) for (u; w) at the upper microstate i =W ,
Eq. (14) for q = 1 takes the form
(S=kB )
q1
:
(19)
1 pW
(E=kB T )
This implies that the thermodynamic deDnition for the temperature at constant volume and number of particles T = (9E=9S)V; N follows for q1 since, in physical
terms, pW 6 1 for large subextensive systems. For superextensive systems with integers q 1, this relation becomes negative. For q = 1, from Eq. (14) we readily
found ( (S=kB )= (E=kB T )) = 1 as should be expected since Sq=1 recovers SBG .
Finally for the force on the system within Tsallis statistics we have
W
q
9
i=1 pi
F = kB T
= ma ;
(20)
9x
q1
T
which gives a physical signiDcance to the q-values representing complex systems and
correlates them to the system dynamics.
In summary, the new insights gained by our statistical thermodynamic approach of
moving particles forming an elastic body are on the molecularmechanical properties
derived for classical and nonextensive dynamical systems including a possible correlation between force and entropy in the system. Our initial factorization for pi = ui wi
may still be seen as rather arbitrary, but we have shown that this simple multiplicative
170
E. Canessa / Physica A 341 (2004) 165 170
form leads to interesting connections between an applied stretching force (tension) and
thermodynamics quantities of dynamical systems (this work was inspired on preliminary ideas discussed by Canessa [7] where set of numbers played the role of classical
particles). Such class of normalized product of positive functions appears formally,
e.g., in the analysis of stochastic processes on graphs according to the Hammersley
CliIord Theorem.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
E.G.D. Cohen, Physica A 305 (2002) 19.
C. Tsallis, arXiv:cond-mat/0312500.
F. Mandl, Statistical Physics, 2nd Edition, Wiley, New York, 1988.
F. Baldovin, L.G. Moyano, C. Tsallis, arXiv:cond-mat/0402635.
C. Tsallis, J. Stat. Phys. 52 (1988) 479.
L.R.G. Treloar, The Physics of Rubber Elasticity, Clarendon Press, Oxford, 1975.
E. Canessa, Physica A 328 (2003) 44.
