Definition of Nonequilibrium Entropy of General Systems

arXiv:physics/9903029v2 [physics.data-an] 25 Mar 1999

Title: Definition of Nonequilibrium Entropy of General Systems

Authors: Xiaochun Mei

Subj-class: Statistics

The definition of nonequilibrium entropy is provided for the general nonequilibrium processes by connecting thermodynamics with statistical physics, and the principle of entropy increment in the nonequilibrium processes is also proved in the paper. The result shows that the definition of nonequilibrium entropy is not unique.

Definition of Nonequilibrium Entropy of General Systems

Mei Xiaochun

Fuzhou Theoretical Physics Institute, No.303, Building 2, Yinghu Garden, Xihong Road, Fuzhou, 350025, P.R.China.

E-mail: zjhzzmgh@mail.hz.zj.cn

PACS unmbers:0520,0570

As well known, though equilibrium state's thermodynamics had matured last century, the general theory for the nonequilibrium systems has not yet been established up to now. Though some special theories for several special nonequilibrium processes have been built up, for example, the irreversible thermodynamics based on the hypothesis of local equilibrium^⑴, The extended thermodynamics⁽²⁾, the rational thermodynamics⁽³⁾, the entropyfree thermodynamics⁽⁴⁾ and so on, all of them are approximate. The key to establish the general nonequilibrium system’s thermodynamic is to define the correct nonequilibrium entropy. This is an unsolved problem.

In equilibrium thermodynamic, the first law of thermodynamic or the law of energy conservation is:

(1)

In Eq.(1) Q is total heat, E is total internal energy, P is pressure, V is volume, F is external force. For a single component system, the definition of entropy function S is:

(2)

Because the concept of absolute temperature is based on equilibrium states, Eq.(2) is only suitable for equilibrium states.

For nonequilibrium states, there exist same different definitions about nonequilibrium entropy. According to irreversible thermodynamics based on the hypothesis of local equilibrium, a system can be divided into many small cells, and each cell can be regarded as a small equilibrium system, in which local quilibrium temperature T(x, t) can be defined. Let S_m(x,t)，E_m(x,t)，Q_m(x,t) and F_m(x,t) represent entropy density, internal energy density, heat density and external force density individually, the total entropy, total internal energy, total heat and total external force of a system can be written as:

, (3)

The first law of thermodynamics and the definition of entropy can be written as:

(4)

(5)

It is obvious that the definition of local equilibrium entropy depends on local equilibrium temperature. But it is still an unsolved problem how to determine the function’s form of T(x,t) when Sm is unknown. As for general nonequilibrium systems in which the hypothesis of local equilibrium does not hold, Eq.(5) is not considered suitable at present.

For the definition of nonequilibrium entropy, two conditions should be satisfied. the first is that when a system reaches equilibrium state, its nonequilibrium entropy should identify with equilibrium entropy in the current equilibrium thermodynamics. The second is that the definition should satisfy the principle of entropy increment in the nonequilibrium processes. Conversely, if a function can satisfy these two conditions, it can be regarded as nonequilibrium entropy.

Because the concept of temperature is established based on equilibrium states, in order to avoid argument whether it is possible to introduce the concept of nonequilibrium temperature, as well as to distinguish the paper from the theory of local equilibrium, the function R(x,t) is introduced and Eq.(5) is written as:

(6)

 It is proved below that by means of the connection of thermodynamics and statistical physics, the form of function R can be determined, and Eq.(6) can satisfy the principle of entropy increment in nonequilibrium processes. When the systems reach equilibrium, we take R=T=constant, Eq(6) becomes the definition of equilibrium entropy. Therefore, the two conditions above can be satisfied, so Eq.(6) can be considered as the definition of nonequilibrium entropy for general systems actually.

From Eq.(6) we have:

(7)

in which Sm0 is a constant. According to Eq.(3) and Eq.(7), we get:

 (8)

Here S0=Sm0V, V is the volume of system. Eq.(8) is just the formula to calculate nonequilibrium entropy. Then, let's determine the form of R function. For the nonequilibrium processes, the equilibrium equation of entropy is:

(9)

It is noted that the situation of R function is similar to nonequilibrium temperature T(x,t) in the local equilibrium irreversible thermodynamics. Therefore, for the single component systems, entropy flow Jm and entropy production σm are individually:

(10)

(11)

In the formula, jq is heat flow, V is macro-velocity of fluid, Π is the viscosity stress tensor.

Suppose the system has N particles, in the 6N-dimension phase space, the probability function is f_n(x₁…x_N,p₁…p_N,t),the nonequilibrium statistical entropy density or micro-entropy density is S_N(x₁…x_N,p₁…p_N,t); in the 6-dimension phase space, the probability function is f₁(x₁,p₁,t)=f(x,p,t), the nonequilibrium statistical entropy density is S₁(x₁,p₁,t)=Sp(x,p,t). Let F represent the external force, f satisfies the formula of the BBGKY series of the Liouville's equation:

(12)

Defining S_p as:

(13)

the total entropy of the system is:

(14)

Comparing Eq.(14) with Eq.(3), we have the relation between nonequilibrium thermodynamic entropy density and statistical entropy density:

(15)

From Eq.(14) and (12), we have:

(16)

here:

(17)

(18)

Comparing them with Eq.(10) and (11), let J_m=J_p, σ_m=σ_p, we have:

(19)

(20)

Let us first examine Eq.(19), in which j_q and V can be expressed as by the method of statistical physics:

, (21)

(22)

here is heat flow caused by kinetic energy, and are heat flows caused by potential energy^[5]

(23)

(24)

(25)

Eq.(22) can be written as:

(26)

(27)

Putting Eq.(15) and (26) into Eq.(19),removing integral about d³p, we have:

(28)

(29)

On the other hand, in Eq.(20), the viscosity stress tensor can be written as:

(30)

in which P is pressure and is known in general,ε_ij is the viscosity tensor and can also be calculated if f is known. So for a system, Π can be regarded as known quantity. Putting Eq.(28) into (20), we get:

(31)

in which:

(32)

(33)

Therefore, as long as the probability function f is known by means of statistical physics, we can get the form of R function by solving Eq.(31) in principle.

Meanwhile, if the probability function f is known, we can obtain the form of energy density with⁽⁵⁾:

(34)

(35)

(36)

in which is kinetic energy density. is potential energy density. The form of external force F_m is also known in general. So according to Eq.(4), the form of Q_m is also known for a process. Therefore, nonequilibrium entropy can be determined if f is known.

Now, Let’s discuss the equilibrium states. If we define the states with R=constant as the equilibrium states, from Eq.(31),the equilibrium condition is B=0. From Eq.(33), considering the fact that the equilibrium states are with V=constant, we have the equilibrium condition:

(37)

This is just equal to σ_p=0 in Eq.(18) when R=constant. If R has nothing to do with x but is relative to t, let R=R(t)=T(t), T is absolute temperature, suppose Q_m is also relative to t but has nothing to do with x, that is, Q_m=Q_m(t), we have:

(38)

Eq.(8) becomes:

(39)

This is just equilibrium entropy in the current equilibrium thermodynamics. Therefore, the temperature and the heat of equilibrium state can be the function of time (this point has not been noted in the current thermodynamics), as long as the system is even, we can consider it as equilibrium state thought it changes with time. This is just so-called quasi-stationary process.

Now let we prove the principle of entropy increment in adiabatic processes according to Eq.(8):

(40)

In principle, we can do it by two ways. The first is by the method of statistical physics. After the probability function f is obtained, Eq.(40) can be calculated. If it can be proved that for any process Eq.(40) is tenable, the second law of thermodynamics would be proved. Then, the second law would become a theorem, not a law. But this is difficult because of the difficulty in mathematics.

The second is the method of thermodynamics shown as follows. Firstly, because Q_m and R are continuous functions, S(t) is also a continuous function. Then, suppose an isolated system is in the equilibrium state at time t₀ and reaches another equilibrium state at time t_n through nonequilibrium evolution. There is no other equilibrium states between these two equilibrium states. According to the theory of equilibrium thermodynamics, we have △S=S(t_n)-S(t₀)>0, that is, △S increases. It can be proved by the reduction to absurdity that during whole time t₀→t_n, △S must be a monotonously increasing function. If △S does not increase monotonously, when △S changes from △S>0 to △S<0, or from △S<0 to △S>0, it would appear the situation with during a certain time . According to equilibrium theory, it means that the system must be in the equilibrium states at time and , and the evolution from to must be reversible. When -→0 we have from Eq.(8):

(41)

In the formula, it has been considered that F_m=0 for an isolated system so that we have according to Eq.(4).There would exist two ways to make Eq.(41) possible. The first is or E_m=constant. However, this is impossible. Because E_m is relative to time t and x in the nonequilibrium processes, E_m can not be a constant. The second way is that in some region , in the other region , so that integral (41) may becomes zero at a certain time t. However, this is also impossible. Let:

(42)

the integral region is the volume of the isolated system with a(y,z)≤x≤a′(y,z), b(z)≤y≤b′(z), c≤z≤c′, we have:

(43)

If W₅ is the single-valued function of x, as long as a≠a′, we have W₃≠0. Taking integral again, we get:

(44)

The result shows that as long as W₅ is the single-valued function of x,y,z or x(this condition can be ensured in general.), we always have W(t)≠0. So Eq.(41) is not equal to zero in general in the nonequilbrium processes. Therefore, △S must be a monotone function. Because △S=S(t_n)-S(t₀)>0, so during t₀<t₁<t₂<t_n, we must have △S=S(t₂)-S(t₁)>0, that is, △S is a monotone increasing function. Thus, the principle of entropy increment Eq.(40) is proved for the nonequilibrium systems.

The problem of uniqueness of nonequilibrium entropy is discussed below. Suppose we have obtained the concrete forms function R and S_m by means of above method, now we introduce new functions R′, and G which satisfy the following relation

(45)

or:

(46)

Similar to Eq.(8), we get:

(47)

If we can find proper R′ and G, so that we have R′=T， =0 when the system reaches equilibrium state, and when the system is in nonequilibrium state, the function and S′in Eq.(46) and (47) still satisfies the two conditions of nonequilibrium entropy, so they can still be regarded as nonequilibrium entropies, but their forms are different from the definition shown in Eq.(6) and (8). Therefore, the definition of nonequilibrium entropy is not unique. In fact, most of the definitions of nonequilibrium entropy in the current nonequilibrium can be transformed into the form of Eq.(46) in principle. This is the reason why we prefer to introduce the R function than to use the concept of nonequilibrium temperature T, for temperature is a measurable physical quantity and should be unique. As mentioned above, the R function is not unique, though if we take Eq.(6) and (8) as the definition of entropy, it seems as if R may be regarded as nonequilibrium temperature T.

As for the nonequilibrium entropy defined in the current nonequilibrium statistical physics, though they satisfies the principle of entropy increment, they do not represent the entropy in the practical processes indeed, for they are independent of thermodynamics. For example, the famous Boltzmann's entropy is

, (48)

For the Boltzmann's process, the function K in Eq.(12) is when F=0:

(49)

For the equilibrium state, the entropy function shown in Eq.(48) is the same as that in canonical ensemble. But for canonical ensemble, the probability density function is:

(50)

in which E is the total energy of the system with:

(51)

Here U is the total interaction energy between parties, β=(kT)^-1. But for the Boltzmann's process, the probability density function is determined by the Boltzmann's differential-integral equation. In the equilibrum state, we have relation: . It can be proved⁽⁶⁾ in this case that only the Maxwell’s speed distribution function can satisfy this relation, that is to say, we should have in the equilibrium state:

(52)

in the formula v₀ is the even velocity of particles. It is obvious that only when U=0 in Eq.(51), Eq.(50) and Eq.(52) can be equivalent to each other. However, for the general processes, it is impossible to have no interaction between microparticles. So the Boltzmann's nonequilibrium entropy is only suitable for the equilibrium system composed of free particles or nearly equilibrium system, so Eq.(48) can not be regarded as the nonequilibrium entropy of the system composed of non-free particles actually. As for the other nonequilibrium statistical entropies in the current nonequilibrium statistical physics, the problems are the same. The second law of thermodynamics, at present, has only the expressive form of thermodynamics but has no statistical form. For this reason, if the definition of entropy is not connected to thermodynamics, any entropy defined in the pure statistical physical category can not represent the real entropy in the meaning of thermodynamics, though their rates of changes with time may not decrease forever.

Reference

[1] De Groot and Mazur, Nonequilibrium Thermodynamics, 1962.

P.Glansdorff and I.Prigogine, Thermodynamics Theory of Structure, Stability and fluctuations, 1971.

[2] S.Simons, J.phys. A: Math., Nucl. Gen., 6(1973), 1934.

G.Lebon, et al., J.phys. A, 13(1980), 275.

D.Jou, J.Casas-Vazquez, G.Lebon, Rep. Prog. Phys. 51(1988), 1105.

[3] C.Trusdell, Rational Thermodynamics, 1969.

B.p.Coleman, J.Chem. Phys., 47(1967), 597.

W.Noll,Arch. Rational Mech. Anal., 17(1973), 85.

[4] Meixner Arch. Rational Mech. Anal., 33(1969) 33.

[5] Luo Laohu, Nonequilibrium Statistical Theory, (1990), 365,354.

[6] Wang Zhuxi, Introduce to Statistical Physics, (1965), 152.