• Aucun résultat trouvé

Integral equation for nonequilibrium chemical potential and the Kirkwood diffusion equation: derivation from the generalized Boltzmann equation

N/A
N/A
Protected

Academic year: 2021

Partager "Integral equation for nonequilibrium chemical potential and the Kirkwood diffusion equation: derivation from the generalized Boltzmann equation"

Copied!
31
0
0

Texte intégral

(1)

HAL Id: jpa-00212342

https://hal.archives-ouvertes.fr/jpa-00212342

Submitted on 1 Jan 1989

HAL is a multi-disciplinary open access

archive for the deposit and dissemination of

sci-entific research documents, whether they are

pub-lished or not. The documents may come from

teaching and research institutions in France or

abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est

destinée au dépôt et à la diffusion de documents

scientifiques de niveau recherche, publiés ou non,

émanant des établissements d’enseignement et de

recherche français ou étrangers, des laboratoires

publics ou privés.

Integral equation for nonequilibrium chemical potential

and the Kirkwood diffusion equation: derivation from

the generalized Boltzmann equation

Byung Chan Eu

To cite this version:

(2)

Integral equation

for

nonequilibrium

chemical

potential

and

the Kirkwood diffusion

equation:

derivation from the

generalized

Boltzmann

equation

Byung

Chan Eu

(*)

Department

of

Chemistry

and

Department

of

Physics,

McGill

University, Montreal,

PQ, Canada H3A 2K6

(Reçu

le 13 avril 1989,

accepté

le 8

septembre 1989)

Résumé. 2014 Dans cet

article,

l’équation cinétique

des fluides

simples

denses est

généralisée

et

appliquée

à l’établissement d’une

équation intégrale

pour le

potentiel

chimique

hors

équilibre

et

de

l’équation

de diffusion de Kirkwood des fluides

polyatomiques

denses

(polymères,

par

exemple).

La déduction demande la

généralisation

de

l’équation intégrale

de Kirkwood à des fonctions de distribution de

configuration, l’équation intégrale

du

potentiel chimique

local au cas

des fluides

polyatomiques

hors

équilibre,

et un ensemble

d’ équations

d’ évolution des variables

macroscopiques.

Les

équations

d’évolution des variables

macroscopiques

et la

thermodynamique

des processus irréversibles ont les mêmes structures

mathématiques

que dans le cas des fluides

simples

denses. Les

liquides plus

ou moins

simples

sont traités

séparément

dans cette théorie, par différentes

intégrales

de collision

qui

interviennent dans les

équations

d’évolution

(équations

constitutives)

pour divers flux tels que les tenseurs de contraintes, les flux de chaleur et les

flux

diffusifs.

L’équation

de diffusion de Kirkwood est obtenue pour la fonction de distribution de

configurations

d’une molécule

polyatomique,

en utilisant un ensemble

d’approximations

sur les

équations

des fonctions de distribution et de diffusion du flux de masse. Comme

exemple

d’application

des

équations

d’évolution aux variables

macroscopiques,

on considère la viscosité d’une solution binaire de fluide

polyatomique

et

monoatomique

et on obtient une formule pour la

viscosité

intrinsèque

en termes

d’intégrales

de collision. Cette formule foumit une

expression

de

mécanique statistique

pour la viscosité

intrinsèque.

Abstract. 2014 In

this paper the kinetic

equation

for dense

simple

fluids

reported previously

is

generalized

and

applied

to derive an

integral equation

for

nonequilibrium

chemical

potential

and the Kirkwood diffusion

equation

for dense

polyatomic

fluids

(e.g., polymers).

The derivation

requires

a

generalization

of the Kirkwood

integral equation

for

configuration

distribution

function, the

integral equation

for local chemical

potential

to the case of

nonequilibrium

polyatomic

fluids, and a set of evolution

equations

for

macroscopic

variables. The evolution

equations

for

macroscopic

variables and irreversible

thermodynamics

are found to have the same

mathematical structures as for dense

simple

fluids. The

simple

and

nonsimple

fluids are

distinguished

in the present

theory by

the different collision bracket

integrals appearing

in the evolution

equations

(constitutive equations)

for various fluxes such as stress tensors, heat fluxes Classification

Physics

Abstracts 05.00 - 51.00 - 66.00

(*)

Mailing

address :

Department

of

Chemistry,

McGill

University,

801 Sherbrook Street West, Montreal, Quebec, Canada H3A 2K6.

(3)

and diffusion fluxes. The Kirkwood diffusion

equation

is obtained for the

configuration

distribution function of a

polyatomic

molecule

by using

a set of

approximations

on the distribution function and the mass flux diffusion

equations.

As an illustration of

application

of the evolution

equations

for

macroscopic

variables, the

viscosity

of a

binary

solution of

polyatomic

and monatomic fluids is considered and an intrinsic

viscosity

formula is obtained for it in terms of collision bracket

integrals.

This formula

provides

a statistical mechanical formula for intrinsic

viscosity.

1. Introduction.

Statistical mechanical theories

[1-5]

of dense fluids

consisting

of

complex

molecules

(e.g.,

polymers

and

hydrocarbons)

often

proceed

on the basis of intuitive models and

approxi-mations that

produce

tractable but

only qualitatively

correct results. In such theories

[1-5]

and also in

equilibrium

theories

[6-8]

of dense

simple

fluids

approximations

of various kinds are made in essence for chemical

potentials

for the fluids of interest on the basis of models. As often shown in

equilibrium theory

[6-8]

of dense

simple

fluids,

chemical

potentials play

an

important

role in

calculating thermodynamic

functions,

and there is a

great

deal of

insight

one can

gain by studying approximation

methods for chemical

potentials.

The same would be true

for

nonequilibrium polyatomic

fluids. In this paper we

develop

a formal

theory

for

nonequilibrium

chemical

potentials

and derive the Kirkwood diffusion

equation

for

configur-ation distribution function

closely

related to the

former,

which would also

help

us

develop

some

approximate

kinetic theories of

complex

molecular fluids.

Kinetic

theory

of

transport

processes in dense monatomic fluids and the attendant

theory

of irreversible

thermodynamics

have been studied in recent papers

[9-11]

in which we have made use of the

generalized

Boltzmann

equation

for dense

simple

fluids. The formalisms

developed

therein have been

successfully applied

to

rheological problems

[10,

llb, c]

and

generalized

hydrodynamics

of non-Newtonian fluids

[llb-d].

They

are also

suggestive

of

generalization

to

nonsimple

fluids and condensed matter in

general.

For the stated aims in this paper we first

present

a formal

generalization

of the

theory

to a fluid mixture

consisting

of

polyatomic

molecules or a mixture of monatomic and

polyatomic

molecules. We will show that the evolution

equations

for various

macroscopic

variables such as the

density, velocity,

internal energy, mass

fluxes,

heat

fluxes,

stress tensors, etc. remain

formally

similar to those obtained for dense

simple

fluids

except

for the necessary

changes arising

from the appearance of the internal

degrees

of freedom absent in the case of

simple

(monatomic)

fluids. The

similarity

we observe between the

macroscopic equations

for

nonsimple

and

simple

fluids is not

surprising

since

nonsimple

and

simple

fluids are, from the continuum

theory

standpoint,

distinguishable

only

in terms of

transport

coefficients and related molecular

parameters.

The situation is therefore similar to the one in the

equilibrium

thermodynamics

where,

for

example,

the virial

equations

of state are similar for both

simple

and

nonsimple

fluids

except

for the details of the virial coefficients which reflect the nature of interactions and the

molecularity

of the fluid in

question.

By

using

the formalism

developed,

we show that when

polyatomic

molecules are

interpreted

to include

polymers,

the

present

kinetic

theory

contains in it the well known

theory

of Kirkwood

[12]

on

polymer

solutions,

since the Kirkwood diffusion

equation

for

configuration

distribution function can be obtained for a

polymer

solution if a set of

approximations

is made. We will list the

approximations

in the text. The Kirkwood diffusion

equation

holds for dilute

polymer

solutions,

but its extension into the

higher

concentration

(4)

concentrated solutions if

pair

correlation functions for

polymers

were included in the calculation of conditional diffusion fluxes

appearing

in the

general theory.

(Here

we mean

by

a conditional diffusion flux the mass flux of a

particle, given

a constrained

configuration

of other

particles).

It appears that

although

this kind of

generalization

is feasible

by

proceeding

one

step

further than the Kirkwood

theory requires

in

calculating

the effects of

pair

correlations,

the

resulting theory

would become

unwieldy

and cumbersome. We find it

simpler

to use the stress tensor and other evolution

equations

to calculate

viscosity

and other

transport

properties

of a

polyatomic

fluid. The formalism for

calculating viscosity presented

in this paper

provides

an alternative route distinctive from the Kirkwood line of attack.

With the evolution

equations developed,

we then

investigate

transport

processes of a

binary

dense

mixture,

with a

particular

attention

paid

to viscous

phenomena

for their obvious

utility

in connection with

rheology

of

polyatomic,

e.g.,

polymeric, liquids.

In section 2 the kinetic

equation

and the evolution

equations

are

presented

for

polyatomic

fluids. The latter

equations

are to be

compared

with those for

simple

dense fluids with the

emphasis placed

on the

points

of

difference,

if there is any.

By using

the

macroscopic

evolution

equations,

we

develop theory

of irreversible

thermodynamics

whose structure also remains the same as that for

simple

dense fluids. In section 3 we

present

a « derivation » of

the Kirkwood diffusion

equation

for

configuration

distribution function which was

originally

obtained with the Brownian motion model for motion of

particles

in a

polyatomic

or

polymer

molecule. The

equation

can be obtained without a Brownian motion model.

Moreover,

statistical mechanical formulas for

parameters

in the

original

Kirkwood diffusion

equation

are identified. We also

present

the Kirkwood

integral equation

for correlation

functions,

e.g.,

configuration

distribution functions and

pair

correlation

functions,

for

nonequilibrium,

and an

integral equation

for

nonequilibrium

chemical

potentials.

These two

equations

would facilitate

study

of

nonequilibrium

effects on correlation functions and

thermodynamic

quantities

in

general.

In section 4 formulas for

transport

coefficients are

obtained

in terms of collision bracket

integrals

for dense

polyatomic

fluids. These results reduce to the

Chapman-Enskog

first

approximation

for

transport

coefficients for a dilute fluid as the

density

of the fluid decreases.

Specializing

to a

binary polymeric

solution,

we obtain the intrinsic

viscosity

formula in terms of the collision bracket

integrals.

This formula

provides

an alternative method of calculation for intrinsic

viscosity

which is

basically

different from the

existing

methods. Section 5 is for discussion and conclusion.

2. Kinetic

equation

and

macroscopic

evolution

equations.

2.1 PRELIMINARY. - In view of the

complexity

of notation necessary for

description

of

polyatomic

molecular

systems

it is useful to

systematize

the

symbols

and their usage. We will reserve

subscripts a,

b,

c, ... for

species, subscripts i, j,

k, f,

... for

molecules,

subscripts q, s, t,

... for atoms or groups in a molecule. These

subscripts

will often appear

consecutively.

For

example,

maq

means the mass of

particle

or group q in molecular

species

a, and

raiq

means the coordinate vector of

particle q

in the i-th molecule of

species

a. With this code for usage of the

subscripts pertaining

to

particles

in a molecule of a

species,

we now define

symbols :

Maq =

mass of the

q-th particle

(or group)

in

species

a.

raiq =

position

vector, relative to a fixed coordinate

origin,

of the

q-th particle

in the i-th

composite particle

(molecule)

of

species

a.

Rai =

center of mass vector of the i-th

particle

of

species

a.

aiq

= distance of the

q-th particle

in

species a

from its center of mass located at

(5)

=

velocity

vector of the

q-th particle

in molecule i of

species

a.

= center of mass

velocity

vector of molecule i of

species a

relative to the coordinate

origin.

=

maq

Valq

= momentum

conjugate

to

raiq.

=

ma

V ai

= momentum

conjugate

to

Rai .

- E

Maq,

the total mass of molecular

species

a.

qEa

-

maq

gaiq,

the momentum

conjugate

to

taiq-The dot over a

symbol

means the time derivative :

e.g., g

=

de/dt.

In the above

system

of coordinates there

clearly

holds the relation

for every

position

vectors

of q

E i and

consequently

This

identity

is easy to

verify

if the definition of the center of mass of a molecule is used

along

with

(2.1).

This

implies,

of course, the relation

If the

system

consists of r

species

which may be

polyatomic

or monatomic

molecules,

the Hamiltonian may be written

or, with the center of mass kinetic energy and the internal energy of the molecules made

explict,

Here

where

Vai

is the

potential

energy of an isolated molecule i E a. The

potential

energy

(6)

where

For dense

polyatomic

fluids in which we are

interested,

expressing

the Hamiltonian in the

center of mass energy, the energy for internal

degrees

of freedom and the intermolecular interaction energy is not

particularly advantageous

mode of

representation

for energy from the mathematical

standpoint.

The

phase

of

Na, Nb,

...

particles

will be abbreviated with

X(N)=X(Na,Nb’’’.):

where {Paiq}

and

{raiq}

respectively

stand for the momenta and coordinates of a set of

particles

in N molecules of

species

a, etc., and

2.2 KINETIC EQUATION AND MACROSCOPIC EQUATIONS. - We consider

an

r-component

fluid mixture

containing Na, Nb, N c’

... molecules for

species a,

b,

c, ... that may be

polyatomic,

diatomic or monatomic. The mixture contains at least one

polyatomic species.

The volume of the

system

is V. The evolution of the

system

is

statistically

described

by

the distribution function

obeying

a suitable kinetic

equation.

In the case of monatomic

fluids,

either of a

single

component

or of

multiple

components,

we have shown that the kinetic

equation

may be assumed to follow a

generalized

Boltzmann

equation

[9].

This

generalized

Boltzmann

equation

has been deduced

by using

the

similarity principle

[9e]

which may be stated as follows : there is a set

of

equations

of

motion

for

distribution

functions

at

different

levels

of

correlation which are similar in their basic mathematical structures and

properties

and

particularly

in their broken time-reversal symmetry. These

equations

are

structurally

similar to

the Boltzmann

equation

for dilute gases. This

principle together

with the Boltzmann

equation

for dilute gases has led us to a kinetic

equation

for distribution function

¡(s) (x (s) ;

t )

for cluster

a

consisting

of s monatomic

particles

where

Ls

is the Liouville

operator,

TSI S2 ... SN

(z )

is the collision

operator

for N clusters each of which consists of s

particles (thus

sl, s2, etc.

meaning

the

first,

second,

etc. clusters of s

particles),

ctl is the renormalized

(i. e. ,

scaled

up)

volume of the

system

lu = NV with N

standing

for the number of

clusters,

and X is the number of

particles,

X = sN. This manner

of renormalization leaves the bulk

density

invariant : n =

lim s /V

= lim

X 1 crI .

In

(2.8)

the

limit z - + 0 must be understood. This

parameter

z i e (e :::. 0)

gives

a measure of duration of molecular collisions which take the

system

of molecules from an initial to a final state. This collision time is assumed to be much shorter than the time span

(kinetic

time

scale)

over which the distribution functions evolve.

Therefore,

the kinetic

equation (2.8)

must be

regarded

as a

time-coarse-grained

equation

for distribution functions that are averages of

fine-grained

distribution functions over the duration of collision. The collision time span is difficult to

(7)

The distribution functions

i,,(’)

are normalized as follows :

and if there is a molecular

quantity

A (x (’) ;

t )

such that

where Aj

are

single-particle

variables and

Ak

are

two-particle

variables,

then the average for A at r is

given by

We will henceforth use

angular

brackets to denote the average of mechanical

quantity

A(x(’»

over the

phase

space with the statistical

weight

In

(2.10)

the second line follows from the first

by

virtue of the normalization condition

(2.9),

and the third line from the second

by

the

identity

and

symmetry

of the clusters

(subsystems).

The last line is

simply

the definition of

F (X)(x (X) ;

t ),

but it may be also construed as the usual definition of A which is

commonly

encountered in the dense fluid kinetic

theory

based on the Liouville

equation :

see

Irving

and Kirkwood

[13].

The kinetic

equation (2.8)

with the rule of

averaging

(2.10)

provides

a means to calculate necessary

macroscopic

variables to

study

irreversible processes in dense

simple

fluids. The

argument

[9e]

used to obtain the kinetic

equation

(2.8)

and the rule of

averaging

(2.10)

are

general

since no reference is made to the

molecularity

of the substance

involved,

and hence can be so

generalized

as to describe irreversible processes in dense

polyatomic

fluids.

r

We

imagine

N clusters of s

= L sa

particles

where Sa

stands for the number of molecules of

a = 1

species

a. Then with the

meaning

of the

phase

for

particles

extended as in

(2.7)

we can

(8)

The collision

operator

TS(l),

s(2), ...,

s(N)(z) = TS1,

S2, ..., SN

(z)

for X

particles

broken up into N

clusters of s molecules is determined

by

the classical

Lippmann-Schwinger equation

[9a]

where

c(N)== {s(l),

s (2 ),

..., s(N)}

stands for N clusters of s

particles,

£C’(N)

denotes the

intercluster interaction Liouville

operator

and

R(O ) (z)

is the resolvent

operator

for N

independent

correlated clusters

Here

Ls(a)

is the Liouville

operator

for the o’-th cluster and has the form as

given

in

(2.11).

Then

by using

the Boltzmann

equation

for a

polyatomic

gas mixture and the

similarity

principle

it is

possible

to arrive at the kinetic

equation

where

Here

sa ( a )

stands for the number of

particles

of

species a

in cluster a. The kinetic

equation

is

a

postulate

for the

polyatomic

fluid in

question.

It

clearly

remains

structurally

the same as the dense

simple

fluid kinetic

equation

(2.8)

except

for the increased dimension of the

phase

space to accommodate the internal

degrees

of freedom for

polyatomic

molecules. The

significance

of the

parameter

z in

(2.12a), (2.12b)

and

(2.13)

is the same as discussed in connection with

(2.8).

In

(2.13)

the limit £ -. + 0 must be understood as taken. We thus must

regard

the distribution functions in

(2.13)

as

coarse-grained

ones which evolve on a kinetic

time scale much

longer

than the collision time scale on which the collision

operator

TS(l), s(2), ..., s(N)

is defined.

Appearance

of the collision

operator

in the kinetic

equation

presumes that there are well

separated

time scales of molecular collisions and kinetic

evolution of distribution functions. Existence of such time scales is

generally

assumed in modern works

[14]

on kinetic

theory

and dense fluids in

particular.

As a matter of

fact,

a

binary

collision

operator,

a

particular example

of such collision

operator,

can be used to write the Boltzmann collision

operator

as is often done in the derivation of the Boltzmann

equation ;

see references

[9a]

and

[15].

In this case it is

possible

to show

unambiguously

that

£-1

1

(9)

coarse-grained

in space so that

they

do not

change

over the collision volume. We remark that the

Laplace

transform

[9a]

associated with the definition of collision

operators

must be

interpreted

as a time coarse

graining

over the collisional time scale.

Because of the

similarity

of the kinetic

equation

(2.13)

to

(2.8)

we obtain

macroscopic

equations

which are also

structurally

similar to those derived from

(2.8).

Since

they

may be derived from

(2.13)

in the same manner as from

(2.8)

we

simply

list the

equations

below. We will use

angular

brackets to denote the

averaging

over the

phase

space of a

polyatomic

system :

It is

important

to recall the rule of

averaging

described in

(2.10)

when we derive the evolution

equations given

below. We first define various statistical formulas for

macroscopic

observables

appearing

in the

present

theory.

To this end it is necessary to define the

following

molecular

expressions

for various moments :

where with the notation

(10)

It is useful to note that the definitions of

haB)

given

above are correct to the

approximation

in which the terms of

0(À )

or

higher

in the

expansion

of the

operator

exp (-

’k raiq ;

bks

V r)

are

neglected.

This latter

operator

appears in the terms

containing

the

forces gaiq;

bles. Inclusion of the

neglected

terms would be

mathematically

correct, but very difficult to handle in

practice.

Moreover,

they

appear to be

nonphysical

since

they

arise from the delta functions which may

have too

sharp

a distribution. For

p a, ha

and pa

see

(2.17), (2.33)

and

(2.49b)

below. With the definitions of

h (’)

we now calculate various

macroscopic

variables as follows :

(11)

We construct an

antisymmetric

tensor S =

p5

with

angular

momentum vector S as follows :

These

macroscopic

variables

obey

the

following

conservation laws and evolution

equations :

In the evolution

equations

above,

i.e.,

the

enthalpy

per unit mass

of a, ta

is the internal energy

density

of a, p,, is the

hydrostatic

pressure and

where

/ïJ

are the molecular formulas for the stress tensor, heat

fluxes,

mass

fluxes,

etc.

(12)

defined in table I. We call

A«)

dissipative

terms since

they

are

intimately

related to the

entropy

production

due to

dissipative

processes such as viscous

flow,

heat conduction and

mass diffusions

occurring

in the

system.

The conservation

equations

and the evolution

equations

listed

above,

with the

exception

of the

angular

momentum conservation

equation,

have the same structure as those

appearing

in the dense

simple

fluid kinetic

theory

[9b-g]

except

for the fact that the

macroscopic

variables consist of the center of mass and internal contributions and the

dissipative

terms must reflect

the involvement of the internal

degrees

of freedom in the molecular collision processes in the

system.

Whatever the molecular collisional

details,

the

dissipative

terms may be

expressed

in terms of stress tensors, heat

fluxes,

mass

fluxes,

etc. which

incidentally

are henceforth called

simply

fluxes,

and the flux

dependence

of the

dissipative

terms is similar to the

simple

fluid

counterpart except

that the coefficients reflect the

molecularity

of the fluid in

question.

These

coefficients consists of collision bracket

integrals

which indicates the details of collisions between the

particles

in the

system.

Therefore

they

are the source of information on molecular

properties

of the

system

in the

present

theory.

It is

important

to note that the evolution

equations

(2.29)-(2.32)

are in a fixed

frame,

but

they

can be converted to the

corotating

frame version

by using

the rules

reported

in reference

[9g].

(*)

These definitions of

f/Ja)

and

ya)

are correct to the

approximation neglecting

the terms of

o (À)

or

higher

in the operator

exp (- À r aiq ; bks . V r )

which also appears in the definitions of

(13)

Table 1

(continued).

{ABC}

= sum of nonreduntant

symmetrized products

of the vectors or tensors.

2.3 ENTROPY AND THE EQUILIBRIUM SOLUTION TO THE KINETIC EQUATION. - The kinetic

equation

(2.13)

satisfies the H-theorem if the

solutions,

distribution

functions,

belong

to the zero class functions

[9a]

for which the total

eigenvalue

of the Liouville

operator

N

£0

= Y

Ls(a)

equals

zero. The

equilibrium

distribution

function,

for

example,

has this

a = 1

property.

The

nonequilibrium

distribution function must be chosen in such a way that the H-theorem is

satisfied,

and the modified moment method

[16]

we make use of in this series of work ensures that the

requirement

be met.

The

entropy

of the

system

is defined

by

the formula

and for zero class functions

[9a, e]

there holds the

inequality

(14)

equilibrium

distribution function in fact is

uniquely given by

the

vanishing

collision

integral

of

(2.13) :

We note here

again

that the limit s -. + 0 should be understood in

(2.42).

Equation

(2.42)

means that the

operand

is a collisional invariant. It can be

shown,

by

following

the

procedure

described elsewhere

[9a],

that the solution to

equation

(2.42)

subject

to the conservation of energy and momentum is the canonical distribution function

where

Hs)

is the Hamiltonian of s

particles

and is

proportional

to the

reciprocal

temperature.

Therefore we obtain

with q

denoting

the normalization factor

. the number of internal

degrees

of

freedom ,

and h is the Planck constant. It is

put

into the

integral

to make the latter dimensionless. The factor N! is also

put

in there to

compensate

for the

overcounting

of the

frequency

of states.

The

nonequilibrium

distribution function is now looked for in a form to conform to the H-theorem

by taking

an

exponential

form

(15)

and the normalization factor

.ae (X)

is defined

by

the

expression

with

/3

=

1 IkB

T. The

temperature

T is defined

by

the statistical formula

That

is,

the

temperature

is defined

by

the average kinetic energy. It is consistent with the notion of

temperature

being

a measure of motions of

particles

in the

system

and also with the definition of T in the case of dense

simple

fluids. The reader is referred to reference

[9e]

for a more detailed discussion on

temperature.

The factors N! and

h f in

(2.48)

appear for the same reason as for the

equilibrium

normalization factor

(2.45).

The pressure of the

system

is defined in a manner similar to that in the dense

simple

fluid

theory :

where

It is to be noted that the

temperature

is taken not with the

equilibrium

temperature,

but with the

temperature

as defined in

(2.49a).

Therefore

FI(Z)

is different from

F(x)

by

the

temperature

factor. Often in kinetic

theory,

the

hydrostatic

pressure is defined

by

[18]

the

trace of stress tensor : pa = Tr

Pa/3.

Since the stress tensor is

generally

time

dependent

for a

nonequilibrium

system,

this definition makes

hydrostatic

pressure a

nonequilibrium

quantity.

Moreover,

it leads to the conclusion that

da

=

pâ,,

in

(2.36)

vanishes and as a consequence

the dilatational

part

of the

entropy

production accompanying

the

expansion

or

compression

of the fluid vanishes

together

with the bulk

viscosity.

Therefore,

the author believes that it is not

appropriate

for dense fluids to define

hydrostatic

pressure with the trace of stress tensor as mentioned earlier. The definition in

(2.49b)

does not have the

difficulty

mentioned and

easily

reduces to the

equilibrium

statistical mechanical formula for

hydrostatic

pressure

given

in

terms of the virial tensor.

If

H{N’)

is

decomposed

into

components

related to various fluxes in the

system

(16)

which is

positive

for all values of

xJa}:

The

entropy

flux

J,

is

given by

the formula

where

la

is the chemical

potential

of a per unit mass

and

The unknown

xJa}

are determined

by

the

consistency

condition

which may be looked upon as a differential

equation

for the

entropy

density

since

Hère

and

Z

are defined in

(2.29)-(2.32).

Attendant to

(2.56),

there holds the extended Gibbs relation

Equation

(2.57)

is a consequence of

(2.58).

The

entropy

density,

entropy

flux and

entropy

production

form a balance

equation

where

This

inequality

is

equivalent

to the second law of

thermodynamics

as was shown elsewhere

(17)

The

macroscopic

evolution

equations

(2.24)-(2.32), (2.58)

and

(2.59)

have the same form as those

appearing

in the dense

simple

fluid

theory

[9]

except

for the statistical definitions of

macroscopic

variables and the details of the

dissipative

terms which reflect the

molecularity

of the

system

in

question.

This is

quite

understandable since these

macroscopic

equations

are

meant for gross

description

of behavior

universally

exhibited

by macroscopic

systems

and,

in this

particular

case,

by

a

polyatomic

fluid.

3. The Kirkwood diffusion

equation.

Kirkwood

[12]

proposed

a Brownian motion model for

polymeric

solutions which has been

extensively

used

by

Kirkwood himself and later workers

[19-22]

for

study

of

viscoelasticity

of

polymeric

solutions. In the model a diffusion

equation

is obtained for the

configuration

distribution function of a

polymer

when a set of

approximations

and

assumptions

is made. The Brownian motion of beads

making

up a

polymer

chain is one of the

assumptions.

In this section we show that the Kirkwood diffusion

equation

can be obtained within the framework of the

present

polyatomic

fluid kinetic

theory.

The result itself is not new, but its derivation from the

generalized

Boltzmann

equation

demonstrates the power of the latter for kinetic

phenomena.

Furthermore,

there are a

couple

of intermediate results necessary for the derivation which turn out to be new for

nonequilibrium polyatomic

fluids but also can be useful for

calculating

some of their

nonequilibrium

properties.

The intermediate results in

question

are the

generalized

Kirkwood

integral equation

for correlation functions and chemical

potentials

when there exist

nonequilibrium

fluxes in the

system.

These

equations,

when

appropriately

solved,

would

yield

the

magnitude

of the effects on correlation functions of

nonequilibrium

processes, e.g., shear-induced distortion of

pair

correlation

functions,

etc.

They

have, therefore,

a

potential

for

application

to studies of

nonequilibrium

fluid

properties.

3.1 GENERALIZED KIRKWOOD INTEGRAL EQUATION. - To carry out this

part

of discussion we assume that the interaction energy of the

system

consists of

pairwise

additive

potentials.

We therefore may write

Now

following

Kirkwood

[23],

we introduce

charging

parameters

[6]

which indicate the

strengths

of interactions between

particles

in the

system.

It is sufficient for our purpose to

designate

a

particular

molecule,

say, al and consider its correlation with the rest of the

system.

Thus

denoting

the

charging

parameter

by e,

we may write the

potential

V for the

system

in the form

Here the

charging parameter e

ranges

from e

= 0

to e

= 1. Thus

at e

= 0 the

particle

al is

completely decoupled

from the rest of the

system

and

at e

= 1 it is

fully coupled

with all

(18)

The

potential

energy

part

h (’)

of the molecular formulas for fluxes

h,,q

can be also written

similarly :

These

decompositions

(3.2)

and

(3.3)

will be

presently

used in the derivation of the

generalized

Kirkwood

integral equation.

,

We now define the

configuration

distribution function

1JI’a(lR, t)

where R stands for the set

of coordinates for

particles

in a

polyatomic

(e.g., polymer)

molecule,

IIB =

{Raiq} :

This distribution function

gives

the

probability

of

finding particles q

E a at

Raq

for all q E a at time t. If the

charging

parameter e

is not

equal

to

unity,

then

for the

configuration

distribution function at

arbitrary

The

configuration

distribution function

1/’ a

is a

mass-weighted

reduced distribution function

describing

the evolution of the

configuration

of a

single polyatomic

molecule

interacting

with the rest of the

system.

In

(3.5)

we have inserted the

parameter e

in F (oN’) to indicate that

particle

al is

partially

«

charged

».

We may write

It is also useful to define

pair

correlation functions

and the

symbol

where

The reduced distribution function

V(2)

describes the correlation between a molecule of

species a

and

particle s

of

species

c. This is the

counterpart

of the

pair

correlation function for a

simple

fluid.

Differentiation of

(3.5)

with

respect to e

and use of

(3.6)

together

with the

appropriate

definitions of

H(N)(e), H1(x)(e)

and

A (x)(e)

yield,

after some

algebraic manipulations,

the

(19)

for which we made use of the

identity

and the definition

Equation

(3.9)

is the

generalized

Kirkwood

integral equation

for

1/’a(lR, t ).

If

(3.7)

is differentiated

with e

and a similar calculation is

made,

there follow

integral equations

for

â 2 cs

and

so on, that

is,

essentially

a

nonequilibrium

Bogoliubov-Born-Green-Kirkwood-Yvon

(BBGKY)

hierarchy

for correlation functions. It is of

nonequilibrium

since there are

nonequilibrium

fluxes

appearing

in the

equations.

These

equations

reduce to those in the

equilibrium

BBGKY

hierarchy

[14a-d]

as the fluxes or

xJa)

vanish for all a and a The terms

containing

xJa)

are the new features in the

generalized

Kirkwood

integral equation. They

should in

principle

be able to account for

nonequilibrium

effects on the

configuration

distribution of

polyatomic

(e.g., polymer)

molecules in

nonequilibrium

flux fields. Since

(3.9)

is sufficient for the purpose of the

present

work we will not work out the

equation

for

1/’ Js

here. It is

straightforward

to derive the

integral equation

for

P (2 )

and the

subsequent

members of the

hierarchy.

3.2 LOCAL CHEMICAL POTENTIALS. - The

equation

for a local chemical

potential

can be

obtained

by following

the same

procedure

as for dense monoatomic fluids when

(3.2)

and

(3.3)

are used

along

with the definition of A

(e).

Since it will be

repetitive

if we go

through

the

derivation in

detail,

we will

simply

present

the result below. When the volume and the

temperature

are

kept

constant the chemical

potential

1£a may be written as

[6, 23]

The

one-particle

chemical

potentiel

F£o a

is defined

by

the formula

(20)

for

particle a1, na

=

X a/CU’,

and

xi(n’ ’)

collectively

stands for the

phase

of the internal

degrees

of freedom for molecule al. With this definition of

we may write

and thus

By using

this

identity

we find

in

the form

Eliminating

the second term on the

right

hand side of

(3.15)

by using

(3.9),

we obtain the

equation

for local chemical

potential

J.L a :

where qa is such that

If the

nonequilibrium

contribution to qa is

neglected,

then we find

where

qâint )

is the internal molecular

partition

function for the

polyatomic

molecule in hand. In the last term in

(3.16)

the sum over s is restricted to those of s = q if c = a. That

is,

P 121

is the «

pair

correlation function » for a molecule of

species a

and

particle s

in another

species

if c =/= a or another

polyatomic

molecules of the same

species

if c = a.

Let us assume that a stands for a

polyatomic species

and the concentrations of

polyatomics

are

sufficiently

low. Then since the «

pair

correlation function »

IP 121

may be

neglected,

we may

approximate

ILa to the lowest order as follows :

This is the

approximate

local chemical

potential

which we will use in the derivation of Kirkwood’s diffusion

equation.

This

approximation

limits the Kirkwood diffusion

equation

to

the

regime

of

sufficiently

dilute solutions.

3.3 KIRKWOOD DIFFUSION EQUATION. - With the

preparation

above,

we are now

ready

to

(21)

partition

function qa with

(3.17’)

we find

where

and thus

F .

is the force on

particle y

in al

by

other

particles

in the molecule

- 1. e. ,

it is the so-called

spring

force if we borrow the

commonly

used

terminology

in

polymer

solution theories.

In the case of uniform pressure the

thermodynamic

force

X y4

may

then be written as

The evolution

equation

for diffusion flux

Jaq

may be obtained from

(2.30)

by

decomposing

the latter

into q

components

where

To the lowest order

approximation

taking

a linear term in the

thermodynamic

force or the fluxes and also to a linear

approximation

for

LBz >,

where Àaq; bs

are collision bracket

integrals

for diffusion fluxes which are

given by

the formula

where

with m,

standing

for a mass

(e.g.,

reduced

mass)

and d for a size

parameter

of a molecule

(or

a

monomer unit in the case of a

polymer),

and

(22)

faster time scale than conserved

hydrodynamic

variables.

Therefore,

it is reasonable to set

That

is,

Substituting

(3.21)

and

(3.22)

into

(3.23),

we find

Inverting

this

relation,

we

finally

obtain the linear

flux-thermodynamic

force relations

It is now convenient to define the

transport

coefficients

These are diffusion coefficients. Substitution of

(3.19)

into

(3.25)

yields

the diffusion flux in the form

We have shown that

1JI’a

obeys

the

generalized

Kirkwood

integral equation.

The same function also

obeys

the

following

differential

equation

which can be derived from the definition for

1JI’a,

(3.4),

and the kinetic

equation

(2.13) :

where

This flux factorizes

approximately

to the form

This can be shown

by following

the

procedure

described in a

previous

paper

[9f]

on

electrolyte

solutions. This

approximation

is valid if the heat flux

part

of the

nonequilibrium

contribution

to the distribution function is assumed to consist of the kinetic energy

part

depending

on the

(23)

where

the self-diffusion coefficient. This is related to the friction coefficient

appearing

in the Brownian motion model. When

AJj

is

neglected

and the

approximate

Jaq

so obtained is substituted into

(3.28),

there follows the

equation

which is the Kirkwood diffusion

equation

for

configuration

distribution function

qf,, (R, t )

for a

single polyatomic

(e.g., polymer)

molecule in a flow field u.

Neglect

of

OJâq

is

justifiable

if the diffusion of a bead is uninfluenced

by

other beads since the

approximation

amounts to the

assumption

that

Cùaq; as

= 0

if q

s. Therefore the

approximation

should

yield

reasonable results if the chain is not

tangled by

itself and others. This fact and the manner in which

(3.32)

is derived indicate

clearly

that it is an

equation holding

for a dilute solution.

It is useful to collect the

assumptions

and

approximations

made to obtain the Kirkwood diffusion

equation

in the

present

kinetic

theory.

1. Factorization

ofj(2)

This is due to the

assumption

on the heat flux term

appearing

in the distribution function

by

which the

potential

energy contribution to the heat flux is

neglected.

If the processes do not include heat flows in the

system

this

assumption

is not

required

and the factorization of

Jâq

becomes exact within the thirteen moment

approximation

which is

generally satisfactory

for the usual

hydrodynamic description

of a

fluid ;

cf. reference

[9f].

2.

Neglect of

« pair

correlation

function »

ll,(2)

- This

approximation

holds if the solute concentration is

sufficiently

low,

and

permits

an

approximate

calculation of local chemical

potentials

for solutes and

corresponding thermodynamic

forces.

3.

Steady

state

Îaq

in the substantial

frame of reference.

- This

assumption

is

justifiable

since diffusion fluxes should relax faster than conserved variables. Since the evolution

equation

(3.20)

for

Jaq

is the average of Newton’s

equation

of motion for

particle q

e a, this

assumption

is

equivalent

to

neglecting

the inertia term in the

Langevin equation

for

particle

q e a in the Brownian motion model

and,

in

addition,

also

assuming

the

homogeneity

of diffusion fluxes in space,

i. e. ,

If the diffusion occurs in a flow

field,

this

assumption

is

generally

violated,

and one must take the effect of flow into account. This may be done if an

approximate

formula for u is

employed.

For

example,

if the shear is not

large,

then we may take

where y is the shear rate tensor assumed to be constant.

(It

is useful to note here that the

(24)

even further and take the Oseen tensor

approximation generally

taken in

polymer

solution

theories

[12, 19-22].

To be more

rigorous,

it is necessary to solve the diffusion flux evolution

equation

together

with the momentum balance

equation

and the stress tensor evolution

equation.

This

approach,

however,

is rather difficult to

implement

in

practice.

Neither is it the

most convenient of the methods for

obtaining

transport

properties

of

polyatomic

(e.g.,

polymer)

solutions.

4.

Neglect

of

àj (2)

This

assumption

is

justifiable

if the

configuration

of the chain is

relatively simple,

i.e.,

untangled,

and if the concentration of

polymers

is low. Therefore the retention of this term

probably

would

improve

the accuracy for

1JI’a

when the solute concentration is not

sufficiently

low.

Finally,

a note on the flux evolution

equation

(3.20).

As mentioned

before,

it is

equivalent

to the

Langevin equation

for motion of a

particle

immersed in the medium of other

particles.

However,

there are some

important

differences which endow

(3.20)

with a more

powerful

capability.

First of

all,

it can deal with nonlinear situations since the

dissipative

term

llaq

is

generally

nonlinear for processes removed far from

equilibrium.

The

dissipative

terms

can be

explicitly

calculated in terms of molecular

properties

if the

present

molecular

approach

is

pursued

to its ultimate end.

Equation

(3.20)

does not

require

an

assumption

on diluteness of the solution.

Therefore,

it can handle concentrated solutions as well. All these features are vested in the

dissipative

terms in the

present

theory.

4.

Viscosity

of a

binary

mixture.

The kinetic

equation

used in this work

provides

a formalism with which to calculate various

transport

coefficients for

polyatomic

fluids and we take this

opportunity

to show some formal results for shear

viscosity.

The

procedure

to

employ

for the purpose is

basically

the same as the one used for

simple

dense fluid mixtures

reported previously

[9e, 24].

Therefore we shall

not go into the

subject

in

depth

here.

Instead,

we shall pay a close attention to

viscosity only,

since it is better studied

experimentally especially

in connection with

polymer

solutions which are relevant to the

present

kinetic

theory

and also related to the friction coefficients

À aq ; bs

presented

in section 3.

4.1 STEADY STATE SOLUTION OF STRESS EVOLUTION EQUATIONS. - To be more

specific,

in our

investigation

we shall confine the discussion to a

binary

mixture in which a

polyatomic

fluid is dissolved in a solvent

consisting

of a

spherical

molecular

(simple)

fluid. If there is no

oscillatory

shear

acting

on the

system

the stress relaxes

generally

on a faster time scale than the fluid

velocity

of the

density.

It is then sufficient to consider the

steady

state stress

evolution

equation.

Furthermore we will assume that the stress is

homogeneous

so that it does

not

depend

on

position.

We also assume that the shear rate is

sufficiently

small so that it is sufficient to consider the linear shear rate

dependence

of the stress tensor. That

is,

we assume

that the flow is Newtonian. We will later

develop

a

theory

for non-Newtonian flow

along

the line taken in the

previous

paper

[9]

on dense

simple

fluids.

Under these

assumptions

the stress tensor evolution

equations

become linear

algebraic

equations

for

Rc,

(c = a, b ; a

=

solute,

b =

solvent) :

(25)

term

A a (1)

for

the solute

species

as follows :

and

similarly

for the solvent

component

b. In

(4.1a, b)

the coefficients

Raa,

Rab,

etc. are the collision bracket

integrals

related to the stress tensors. Their

explicit

forms are

given by

the formula

where the

subscripts

c and d are either

a (solute )

or b

(solvent ).

These coefficients are related

to the Newtonian viscosities of the solvent and solute

species.

The linear

algebraic

set

(4.1a, b)

can be solved for

ee=

p P c,

(c

= a,

b )

to

yield

where

These solutions can be used to calculate the viscosities of the solvent and the solute.

4.2 INTRINSIC VISCOSITY OF THE SOLUTION. - Since the viscosities

are defined

by

the relations

we

identify

71 g

and

q§,

respectively,

with the coefficients in

(4.3a, b) :

The formulas in

(4.6a, b)

give

the Newtonian viscosities in terms of statistical formulas

which,

(26)

Since in

practice

the

viscosity

of a mixture is measured in the

laboratory,

the actual observable is the total

viscosity 17 0 :

where

The intrinsic

viscosity

of a solution is defined

by

the

limiting

formula

where c stands for the concentration of the solute in the units of the

weight

per volume

(e.g.,

g/deciliter).

Therefore the intrinsic

viscosity

is the relative

viscosity

of the solute in the infinite dilution limit and has the dimension of volume/mass. It is of interest to us here.

In order to have a better idea of the intrinsic

viscosity

we use

(4.4a-d)

and

(4.6a, b)

and examine their mass fraction

dependence

a

step

beyond

the formal structure. For this purpose

we introduce a

parameter

Te of dimension time defined

by

the formula

where mr is a mean mass and d is a mean size

parameter

of the molecules

making

up the

system.

This

parameter

rc is

roughly

a measure of collision time between two

particles.

We then scale the collision bracket

integrals Raa,

etc. in the

following

manner :

These

scalings

leave

Rbb, Raa,

etc. dimensionless. The choice for the

density

factors is made on the basis of

analysis

for the collision bracket

integrals

and the collision

operator

in

particular.

The

analysis

in

question

is made

by

means of a cluster

expansion

for the collision operator

appearing

in the collision bracket

integrals,

for details of which the reader is referred to

Références

Documents relatifs

0 (the Boltzmann-Grad limit), the prob- ability density of a test particle converges to a solution of the linear (uncutoed) Boltzmann equation with the cross section relative to

Many different methods have been used to estimate the solution of Hallen’s integral equation such as, finite element methods [4], a general method for solving dual integral equation

The use of the matrix density formalism, and of a multiscale expansion, allows the derivation of a macroscopic nonlinear evolution equation for a short light pulse (a

The following lemma allows us to compare the results at the global scale (in variables (t, x)) and at the local scale (in variables (s, y)) in order to carry the desired result from

In fact, using our results, Ramalho’s argument for the existence of precisely two positive solutions could

The question of diffusion approximation of kinetic processes (limit λ/L 1 and `/L ∼ 1) has motivated a lot of works, with various fields of applications: neutron transport by

As mentioned, the existence and uniqueness of the solutions of the chemical master equation (1.1), (1.2) for given initial values is then guaranteed by standard results from the

This paper addresses the derivation of new second-kind Fredholm combined field integral equations for the Krylov iterative solution of tridimensional acoustic scattering problems by