Thermodynamics with Internal State Variables

Bernard D. Coleman and Morton E. Gurtin

Citation: The Journal of Chemical Physics 47, 597 (1967); doi: 10.1063/1.1711937
View online: http://dx.doi.org/10.1063/1.1711937
View Table of Contents: http://scitation.aip.org/content/aip/journal/jcp/47/2?ver=pdfcov
Published by the AIP Publishing

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
65.30.29.111 On: Sun, 16 Mar 2014 01:05:33
 CRYSTAL STRrCTURE OF ALPHA-OXYGEN 597

according to Horl,s are 3.307 A center to center, and ACKNOWLEDGMENTS

in 'Y-Oz, which are 3.41 A. The orientation of the mole-
cules helps to explain the fact that the spacing of (010)
planes in the C-cell (3.429 A) is less than the spacing
of the (001) planes (3.748 A), even though the mole-
cular centers are not directly above each other along a
line normal to (001). From the figure, it can be seen
that there are no 0 4 groups; this absence of dimers is
also true of the other phases of Oz, as well as of the
known phases of N 2 , CO, and F z.
A discussion of the packing of molecules in a-02 and
{1-0z, the a-{1 transformation, and possible faults and
twins in a has been submitted for publication else-
where.
This work was supported in part by the National
Science Foundation Grant GP 5859, the U. S. Army
Research Office Grant DA-ARO(D)-31-124-GS42, the
Office of Naval Research Contract Nom 2121 (11),
and facilities provided by the Advanced Research
Projects Agency. The assistance of David Schiferl in
some of the computations is acknowledged with thanks;
also helpful discussions with P. M. de Wolff and W. H.
Zachariasen. We are grateful to the referee of this
paper for suggesting that we give a detailed presentation
of the experimental data and some of our computations
which are now included in the table and figures.

THE JOURNAL OF CHEMICAL PHYSICS VOLUME 47, NUMBER 2 15 JULY 1967

Thermodynamics with Internal State Variables

BERNARD D. COLEMAN
J(ellon Institute, Pittsburgh, Pennsylvania
AND
MORTON E. GURTIN
Department of Mathematics, Carnegie Institute of Technology, Pittsburgh, Pennsylvania
(Received 13 February 1967)

This is a study of the thermodynamics of nonlinear materials with internal state variables whose temporal
evolution is governed by ordinary differential equations. After employing a method developed by Coleman
and Noll to find the general restrictions which the Clausius-Duhem inequality places on response functions,
we analyze various types of dynamical stability that can be exhibited by solutions of the internal evolution
equations. We also discuss integral dissipation inequalities, conditions under which temperatures can
be associated with internal states, and the forms taken by response functions when the material is a fluid.

1. INTRODUCTION No one approach to dissipation is so general as to

In phenomenological theories of the dynamical be-
havior of continua there are several ways of accounting
for the dissipative effects which, in addition to heat
conduction, accompany deformation. The oldest and
simplest way is to introduce a viscous stress which
depends on the rate of strain, as is done in the Navier-
Stokes theory of linearly viscous fluids. Another de-
scription of dissipation assumes that the entire past
history of the strain influences the stress in a manner
compatible with a principle of fading memory.! A
third approach is to postulate the existence of internal
state variables which influence the free energy and
whose rate of change is governed by differential equa-
tions in which the strain appears.
1 Here we have in mind the fading-memory postulate intro-
duced by Coleman and No1l2 . and recently developed from a set
of elementary axioms by Coleman and Mizel. 4
2 B. D. Coleman and W. Noll, Arch. Rational Mech. Analysis
6, 355 (1960).
3 B. D. Coleman and W. Noll, Rev. Mod. Phys. 33, 239 (1961);
36, 1103 (1964).
4 B. D. Coleman and V. J. Mizel, Arch. Rational Mech. Analy-
sis 23, 87 (1966).
include completely all the others, and each of the three
classes of theories has its domain of usefulness. It has,
for example, long appeared that the Navier-Stokes
equations can successfully describe the mechanical
behavior of water under ordinary conditions. On the
other hand, when certain polyatomic gases, such as
carbon dioxide, are studied in shock tubes5-8 and in
sonic absorption experiments,9-11 it is found that, in
an appropriate range of density and temperature, at-
tenuation effects occur which can be well described,
not by the introduction of a single and constant bulk
6 H. A. Bethe and E. Teller, Aberdeen Proving Ground Rept.
No. X-117, 1941.
6 E. F. Greene, G. R. Cowan, and D. F. Hornig, J. Chern.
Phys.19,427 (1951); 21,617 (1953).
7 F. J. Smiley, E. H. Winkler, and T. K. Slawsky, J. Chern.
Phys. 20, 923 (1952).
8 W. Griffith, D. Brickl, and V. Blackman, Phys. Rev. 102,
1209 (1956).
9 G. W. Pierce, Proc. Acad. Sci. Amsterdam 60,271 (1925).
10 K. F. Herzfeld and F. O. Rice, Phys. Rev. 31,691 (1928).
11 A survey is given by K. F. Herzfeld and T. A. Litovitz,
Absorption and Dispersion of Ultrasonic Waves (Academic Press
Inc., New York, 1959). See also the critique by C. Truesdell,
J. Rational Mech. Analysis, 2, 643 (1953).

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
65.30.29.111 On: Sun, 16 Mar 2014 01:05:33
 598 B. D. COLEMAN AND
viscosity, but rather by assuming that the ideal-gas
law pv=RO holds, with the translational (or "active-
mode") temperature, while accounting, in the consti-
tutive equations governing the heat capacity, for the
time required for the transfer of molecular motion
from the translational mode, to one, two, or several
internal modes. This is a classical example of a theory
of dissipation involving a finite number of internal
state variables; the internal variables being the amounts
of energy associated with the various modes of molec-
ular motion. At elevated temperatures the ideal-gas
law may fail due to chemical dissociation or ionization;
the finite rates of the reactions of decomposition and
reassociation can then also give rise to pressure-volume
relaxation phenomena. 12
Recent experience in high-polymer physics shows
that the mechanical behavior of many materials, in-
cluding polymer melts and solutions, is more easily
described within the general theory of materials with
fading memory than by theories of the viscous-stress
type, which do not account at all for gradual stress
relaxation, or by theories which rest on a finite number
of internal state variables and which, therefore, give
rise to discrete relaxation spectra when linearized.
Using an approach to the thermodynamics of con-
tinua proposed by Coleman and Noll,1a Coleman and
Mizel14 have found the restrictions the second law
places on a class of theories of the viscous-stress type.
Later, Coleman15 developed a general theory of the
thermodynamics of materials with fading memory. Here
we discuss phenomenological theories of dissipation
which rest upon evolution equations for internal state
variables; we attempt to unify such theories by using
the methods developed in Refs. 13 and 14 together
with techniques from the stability theory of nonlinear
differential equations. We consider materials for which
the basic local mechanical and thermal variables, such
as the stress tensor T, the heat flux vector q, the
specific Helmholtz free energy 1/;, and the specific en-
tropy 7], are determined through constitutive equations
when the following "state variables" are known: The
temperature 0, the deformation gradient F (or the
"strain"), the temperature gradient gradO, and a set
of N internal or "hidden" state variables (aI, "', aN)'
We allow the rate of change of each of the ai to be
governed by a (generally nonlinear) function Ji of 0,
F, gradO, and the complete internal state (aI, "., aN) :
ix, =Ji(O, F, gradO, aI, ... , aN), i= 1, .. , N. (1.1)
We first seek the general restrictions placed on theories
12 See, for example, the discussion of M. J. Lighthill, J. Fluid
Mech. 2, 1 (1957). Apparently, the first paper expressing this
idea is that of A. Einstein, Sitzber. Akad. Wiss. Berlin Math.
Phys. Kl. 380 (1920).
13 B. D. Coleman and W. Noll, Arch. Rational Mech. Analysis
13, 167 (1963).
14 B. D. Coleman and V. J. Mizel, J. Chern. Phys. 40, 1116
(1964).
lliB. D. Coleman, Arch. Rational Mech. Analysis 17, 1, 230
(1964).
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
65.30.29.111 On: Sun, 16 Mar 2014 01:05:33
M. E. GURTIN
of this type by the second law, in the form of the
Clausius-Duhem inequality. We then give conditions,
in terms of the thermodynamical functions, for dynami-
cal stability of equilibrium solutions of the evolution
equations (1.1). After finding logical relations between
the various types of stability which can occur, we
discuss special aspects of the theory, such as integral
dissipation inequalities, the assignment of tempera-
tures to internal states, and the requirements of in-
variance. We conclude our study by illustrating the
general theory with its simplest nontrivial special case:
ideal gases with a single internal mode of molecular
motion.
2. THERMODYNAMIC PROCESSES
We consider a body CB with material points X and
assume that the mechanical forces acting on CB can
always be resolved into a body force field and a sym-
metric stress field. In particular, body couples, couple
stresses, and multipolar interactions of higher order
are assumed absent. We also assume there is no dif-
fusion of mass in CB, but CB may deform and conduct
heat. Thus, a thermodynamic process for CB is described
by nine functions, of X and the time t, whose values
have the following physical interpretations:
(1) The spatial position X= 2C(X, t) in the motion 2C.
(2) The symmetric Cauchy stress tensor T=T(X, t).
(3) The specific body Jorce b= b(X, t) per unit mass
(exerted on CB at X by the external world, i.e., by other
bodies which do not intersect CB).
(4) The specific internal energy E=E(X, t) per unit
mass.
(5) The heatflux vector q=q(X, t).
(6) The heat supply r=r(X, t) per unit mass and
unit time (absorbed by CB at X and furnished by
radiation from the external world).
(7) The specific entropy 7]=7](X, t) per unit mass.
(8) The absolute temperature O=O(X, t) >0.
(9) The internal state vector 0:= o:(X, t) = (aI, a2, "',
aN); the numbers a,=a,(X, t) are the internal state
variables ,16
Such a set of nine functions, defined for all X in CB
and all t, is called here a thermodynamic process in
CB if and only if it is compatible with the law of balance
oj linear momentum and the law of balance oj energy.
Under sufficient smoothness assumptions, the familiar
integral forms of these laws are equivalent to the field
equations
px-divT=pb (2.1)
and
pe-TL+divq=pr. (2.2)
Here p is the mass density, L= grad! is the velocity
16 Our present pair (F, .. ) forms a vector which Truesdell and
Toupin,'7 pp. 615-647, call the thermodynamic substate.
17 C. Truesdell and R. A. Toupin, "The Classical Field Theo-
ries," in Encyclopedia of Physics, S. Fhigge, Ed. (Springer-Verlag,
Berlin, 1960), Vol. 3, Pt. 1.
 THE R MOD Y N A M I C S WIT H I N T ERN A L S TAT E VAR I ABL E S 599

gradient, and TL is the inner product of T and L; off the following implications of the Clausius-Duhem
i.e., TL=Tr TLT with LT the transpose of L. Since inequality:
grad and div are the gradient and divergence with
respect to x keeping t fixed, these operators refer to
7j~0 whenever F=O, i=O, and g=O, (3.4)
spatial derivatives. A superposed dot denotes the ma- i~O whenever F=O, 7j=0, and g=O. (3.5)
terial time derivative, i.e., the derivative with respect
to t computed keeping X fixed. The specific free energy I/; (also called the "Helm-
To specify a thermodynamic process it suffices to holtz free energy per unit mass") is defined by
prescribe the seven functions ~, T, , q, 7], 0, and a. (3.6)
The remaining functions band r are then determined Since
by (2.1) and (2.2).
In continuum physics one usually identifies each
material point X with its position X in a fixed reference (3.3) can be written
configuration ffi and writes x = X (X, t). The gradient 0"1= -~-1]O+SF_(p8)-lq.g, (3.8)
F of 2C(X, t) with respect to X,
and hence (3.2) yields also
F=F(X, t) = GRAD2C(X, t), (2.3)
~~O whenever F=O, 0=0, and g=O. (3.9)
is called the deformation gradient at X (or at X)
relative to the configuration ffi. We consistently use 4. CONSTITUTIVE ASSUMPTIONS
the notation GRAD for the gradient in ffi, i.e., the
gradient with respect to X. For a scalar field over (lS, In the present theory the material at the point ~X is
such as 0, an elementary chain rule yields characterized by five response functions, if;, ii, S, q,
and f, which give 1/;, 'T}, S, q, and Ii at X when F, 0, g,
GRADO=FT gradO. (2.4) and a are known at X:
Since gradO occurs frequently in thermodynamics it is 1/;=if;(F, 0, g, a), ( 4.1)
convenient to have a single symbol for this vector.
We use the abbreviation 1]=fJ(F, 0, g, a), (4.2)

g=gradO. (2.5) S=S(F, 0, g, a), (4.3)

Because we assume that x(X, t) is smoothly in-
vertible in its first variable, the inverse F-l of F exists,
and the velocity gradient L=grad:ic is given by
L=FF-I.
Let us call the tensor S, defined by
S=p-lT(FT)-r,
the Piola-Kirchoff stress tensor. Using this tensor we
can write (2.2) in the form
pi-pS.F+divq=pr.
3. THE CLAUSIUS-DUHEM INEQUALITY
When q/O is regarded to be a flux of entropy due to
heat flow and r/O a supply of entropy from radiation,
the specific rate "I of production of entropy is given by
P'Y=p7j-[(pr/O) -div(q/O)]'
The Clausius-Duhem inequality asserts that the rate
of production of entropy is not negative:
'Y~o.
For each thermodynamic process the energy-balance
equation (2.8) permits us to write (3.1) in the form
'Y=7j- (i/O) +o-lS.F- (1/p82) q. g.
From this equation and the inequality (3.2) we read
q=q(F, 0, g, a), ( 4.4)
li=f(F, 0, g, a). (4.5)
The superposed caret in if;, fJ, 8, and q serves to dis-
(2.6) tinguish these functions from their values. Since for a
given process the value of F depends not only on X
and I, but also on the choice of the reference configura-
(2.7) tion ffi, the form of each of the response functions
depends on ffi. In a materially inhomogeneous body
the functions if;, fJ, S, q, and f will also vary with the
material point Xj although we do not render explicit
(2.8) this possible dependence of response functions on X,
all the arguments we give here are valid equally for
materially homogeneous and materially inhomogeneous
bodies.
We say that a thermodynamic process is admissible
if it is compatible with the constitutive equations
(4.1)-( 4.5).
(3.1) Of course, in assuming (4.1) -( 4.5) we are using
Truesdell's principle of equipresence, which asserts that
a quantity present as an independent variable in one
constitutive equation of a material should be so present
in all, unless its presence contradicts some general
(3.2)
law of physics or the assumed symmetry of the ma-
terial,l8,19 In the following section we follow the ap-
C. Truesdell, J. Pure App!. Math. 30,111 (1951).
18
C. Truesdell, App!. Mech. Rev. 12, 75 (1959), reprinted
19
(3.3) with additions in Applied Mechanics Surveys (Spartan, Washing-
ton, D.C., 1966), pp. 225-236; see also C. Truesdell and R. A.
Toupin,'7 pp. 703-704.

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
65.30.29.111 On: Sun, 16 Mar 2014 01:05:33
 600 B. D. COLEMAN AND M. E. GURTIN

proach of Coleman and MizeI14 20 who regarded the
Clausius-Duhen inequality as included among the laws
of physics which can cause a separation of effects,
i.e., which can be used to show that certain response
functions must be independent of certain variables.
In fact, we shall show that the derivatives of f, ~,
and S with respect to g must be zero. 21
5. CONSEQUENCES OF THE SECOND LAW
Within our present framework the second law of
thermodynamics is given a precise mathematical mean-
ing by the following postulate of positive production of
entropy22:
Every admissible thermodynamic process in ill must
obey the inequality (3.2) at each time t and for all ma-
terial points X in ill.
We now seek the restrictions this postulate places
on the response functions in our basic constitutive
equations (4.1)-(4.5).
It follows from (4.1) that in each admissible process
f= (OFf) .F+(oef)()+(ogf) .g+(oaf) a,
where
oe.j;(F, e, g, a) = (oloe).j;(F, e, g, a);
the components of ogf and OFf, relative to an ortho-
normal basis el, e2, e3, are given by
[Ogf(F, e, g, a)];= (oldg.)f(F, e, g, a),
i=1,2,3,
solution a= a(X, t) of (4.S) for all t in some interval
[to, to+rJ with a(X, to) = ao(X). Then, using the con-
stitutive equations (4.1) -( 4.S) we can compute 1/;, 'T/,
f=1/;+e'l], S, and q at each X for all t in [to, to+r]'
Thus to each sufficiently smooth choice of ao, 2C, and e
there corresponds a unique admissible thermodynamic
process in ill for some time interval [to, to+r].
Now, let (Fo, eo, go, ao) be an arbitrary point in
the domain of the response functions f, n, S, q, and f
for that material point X' of ill which occupies the
place X' in CR. Choosing arbitrarily a time to, a tensor A,
a vector a, and a scalar a, we consider the motion 2C,
and the time-dependent temperature distribution e de-
fined by
2C(X, t) = X'+[Fo+(t-to) AJ[X-X'J,
e(X, t) =eo+(t-to)a+[go+(t-to)a}Fo[X-X'J,
(S.7)
for all X in CR and all f?:. to sufficiently close to to. It
follows from the observation made in the previous
(5.1) paragraph that the fields shown in (S.7) and the
initial condition a(X, to) = ao determine an admissible
thermodynamic process in ill for t in some time interval
(5.2) [to, to+r]. Furthermore, this process has the following
properties:
F(X', to) =Fo, e(X', to) =eo,
g(X', to) = go, a(X', to) = ao, (S.8)
and

[oFf(F, e, g, a) ];j= (dloFij)f(F, e, g, a), F(X', to) =A, ()(X', to) =a, g(X ' , to) =a. (5.9)

i,j=1,2,3, (5.3) Hence, by (4.1)-(4.5), for this process at the point

X' and the time to the equation (5.6) becomes
with gi and Fij the components of g and F; d"f is the
N-vector eo'Y = [S(Fo, eo, go, ao) -dFf(Fo, eo, go, ao)]A
d"f= [oalf, da2f, "', d".vfJ, (5.4) -[def(Fo, eo, go, an) +~(Fo, eo, go, ao) Ja
and, of course, -ogf(Fo, eo, go, ao)a
daf a=da1fal+d"2fa2+" +daNfaN' (5.5) -daf(Fo, eo, go, ao) f(Fo, eo, go, an)
Substituting (5.1) into (3.8) we obtain the following - (poOO)-lq (Fo, eo, go, ao) go, (5.10)
expression for the specific rate of production of entropy
where po is the mass density at X' corresponding to the
in an admissible process:
deformation gradient Fo. It clearly follows from (5.10)
e'Y= (S-dFf) F- (def+'I])() that in order to have 'Y~O at X' and to for every choice
of A, a, and a, we must have
-dgfg-daf a- (11 pO) q. g. (5.6)
Suppose that we are given an initial time to, an
S(F, e, g, a) =dF.j;(F, e, g, a), (5.11)
initial internal state vector ao(X) for each X in CR, ~(F, e, g, a) = -def(F, fJ, g, a), ( 5.12)
a motion 2C of ill, and a time-dependent temperature
distribution e over ill. Suppose further that ao, F=
dgf(F, e, g, a) =0, (5.13)
GRAD2C, e, and g=grade, as functions of X and t, d".j;(F, fJ, g, a) f(F, e, g, a) + (pe)-lq(F, e, g, a) g~O,
are smooth enough to ensure the existence of a unique
(5.14)
20 B. D. Coleman and V. J. Mizel, Arch. Rational Mech. Analy-
sis 13, 245 (1963). where, for convenience, we have dropped the subscript 0
21 A similar result holds in the general theory of materials with
in (Fo, eo, go, (1'0). Since this point was arbitrarily
fading memory; d. Coleman,' Theorem 1, p. 19.
22 See Coleman and Noll.' chosen in the domain of our response functions, we

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
65.30.29.111 On: Sun, 16 Mar 2014 01:05:33
 THE R 1\1 0 D Y N A 1\1 I C S WIT H I N T ERN A L S TAT E VAR I AB L E S 601

have the following four consequences of our postulate holding for nonzero 0.. Such a resolution is valid,
of positive entropy production22a : however, when the following conditions are met: that
f(F, (J, g, a) be independent of g and q(F, (J, g, a)
(i) The response functions -i-, n, and Sare independent
be independent of a. These conditions do not follow
of g; i.e.,1/;, T/, and S are given by functions of F, e, and
from our general assumptions. Yet, the first condition
a alone,
does often follow from special assumptions of linearity
T/=n(F,fl, a), and material symmetry; therefore, it is worth noting
that
(5.15)
agf(F, (J, g, a) =0 => u(F, (J, g, a) ~O
(ii) -i- determines nthrough the entropy relation (5.12) ;
i.e., for all F, (J, g, and a. (5.25)
(5.16)
It is observed that the role played by u in our present
(iii) -i- determines S through the stress relation (5.11); theory is nearly identical to that played by the quantity
i.e., which Coleman in his articles on materials with fading
(5.17) memoryI5 also denotes by u and calls the internal dis-
sipation. For example, our Eqs. (5.21) and (5.15)-
(iv) -i-, f, and q obey the general dissipation inequality (5.17) here yield the equation
(5.14) .
Jj;=S.F-T/8-(Ju, (5.26)
Conversely, using (5.6) it is easily verified that when
the conditions (i)-(iv) hold every admissible process which has the same form as Eq. (8.1) on p. 22 of
in ffi obeys (3.2). Ref. 15. However, in our present theory, if agf is not
It follows from (2.7) and (5.17) that the symmetric identically zero, u may be negative for certain values
Cauchy stress tensor T is given by the equation of g, albeit (5.22) tells us that u~O whenever g=O.
Using (3.6) we can write (5.26) in the form
(5.18)
(5.27)
and hence is a function of F, e, and a. therefore,
The general dissipation inequality (5.14) implies that
when g=O the internal dissipation inequality, u= -e/(J whenever F=O and ~=O, (5.28)
and
aa$(F, e, a) f(F, e, 0, a) ~O, (5.19)
u=~ whenever F=O and e=O. (5.29)
holds, and when aa1/;f=O the heat-conduction inequality,
Since (2.8) and (5.27) imply
q(F, fl, g, a) g~O, (5.20)
u=~-e-l(r-p-l divq) , (5.30)
holds.
If we define the internal dissipation u by
we also have
u=~ whenever pr-divq =0. (5.31)
u=u(F, e, g, a)
= -e-1aa-i-(F, e, a) f(F, fl, g, a), (5.21) Thus, we can assert that in general u measures the
rate ~ at which the entropy of a material point would
we can write the internal-dissipation inequality (5.19; change if the total rate pr-divq of addition of heat were
in the form set equal to zero. In particular, u=~ whenet'er the strain
u(F, e, 0, a) ~O. (5.22) and energy are held constant momentarily.
Further, we may conclude from (4.5), (5.1), and (5.13) 6. STABILITY OF EQUILmRIUM STATES
that
Throughout this section we focus our attention on one
u= -e-1Jj; whenever F=O and 8=0, (5.23) arbitrary material point X in ffi. A triplet (F*, (J*, a*)
and this result when combined with (3.8) yields with
f(F*, (J*, 0, a*) =0 (6.1)
U='Y whenever F=O, 8=0, and g=O. (5.24)
is called an equilibrium state for the material at X.
One cannot generally resolve the inequality 'Y ~ 0 into The domain of attraction at constant strain and tem-
an internal dissipation inequality aa-i- o.~O holding for perature of an equilibrium state (F*, (J*, a*) is the
nonzero g and a heat-conduction inequality qg~O set D(F*, e*, a*) of all ao such that the solution
22. Note added in proof: We have recently seen a report by K. C. a= aCt) of the initial value problem,
Valanis [Iowa State University Bulletin No. 52J giving a theorem
analogous to this. See also the article by E. T. Onat which will a=f(F*, 8*, 0, a) ( 6.2)
appear in the Proceedings of the JUT AM Symposium on Irre-
versible Aspects of Continuum Mechanics, Vienna, June 1966 (to with
be published). . a(O) = It'Q, ( 6.3)

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
65.30.29.111 On: Sun, 16 Mar 2014 01:05:33
 602 B. D. COLEMAN
exists for all t~O and tends to a*,
a(t)-a* as l_oo.
An equilibrium state (F*, 0*, a*) is said to be (locally)
asymptotically stable at constant strain and temperature
if D(F*, 0*, a*) contains a neighborhood of a*, i.e.,
if there exists a r>O such that every internal state
vector ao with I ao- a* I <r is in D(F*, 0*, a*).
Let (F*, 0*, a*) be an equilibrium state for the
material at X, and let ao be in D(F*, 0*, a*). It follows
from the argument containing Eqs. (5.7) that there
exists at least one admissible thermodynamic process
with the property that at X
a(O) = ao, while for all t~O, F(t) =F*,
OCt) =0*, and get) =0.
It follows from (3.9) that for this process I{;(t) ~I{;(O),
or, by (5.15h,
~(F*, 0*, aCt) )~~(F*, 0*, ao).
But, since ao is assumed to be in D(F*, 0*, a*), it
follows from (6.4) and the continuity of ~ that
lim~(F*, 0*, a( t) ) =~(F*, 0*, a*).
1-+00
Combining (6.6) and (6.7) we obtain
~(F*, 0*, a*) ~~(F*, 0*, ao).
In words: If the internal state vector a is in the domain
of attraction of the equilibrium state (F*, 0*, a*) at
constant strain and temperature, then the free energy
corresponding to (F*, 0*, a) cannot be less than that
corresponding to (F*, 0*, a*).
The following theorem follows at once from (6.8)
and the definition of asymptotic stability: If (F*, 0*, a*)
is an equilibrium state that is asymptotically stable at
constant strain and temperature, then
~(F*, 0*, a) ~~(F*, 0*, a*)
for aU internal state vectors a in some neighborhood of a*,
and, consequently,
aa~(F*, 0*, a) la~a.=O.
The equation (6.10) may be called the equation of
internal equilibrium. In deriving this equation we have,
of course, assumed that the variables aI, , aN have
been chosen so as to be free from constraints of the
type h( a) =0. Thus, when changes in the internal
state are due to chemical reactions, a.
is not a con-
centration but rather the degree of advancement of the
ith reaction, in the sense in which the term is used by
De Donder.23 In this case -aa,I{; is the affinity of the
ith reaction, and the relation aa,I{;=O is just De
23 Th. De Donder, Le~ons de Thermodynamique et de Chimie
Physique (F. H. van den Dungen and G. van Lerberghe,
Paris, 1920), p. 117.
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
65.30.29.111 On: Sun, 16 Mar 2014 01:05:33
AND M. E. GURTIN
Donder's form of the equation of chemical equilibrium
for that reaction. 24 .25
(6.4)
We say that an equilibrium state (F*, 0*, a*) is
Lyapunov stable at constant strain and temperature if for
each e> 0 there exists a 5,> 0 such that every solution
aCt) of the differential equation (6.2) with
I a(O) - a* I <5, (6.11)
has the property that
I aCt) - a* I <e ( 6.12)
for all t~O.
Roughly speaking, (F*, 0*, a*) is Lyapunov stable
at constant strain and temperature if holding F and 0
fixed for all t and initially setting the internal state
vector a sufficiently near to a* guarantees that a will
(6.5) remain near to a* for all t.
Since it follows from (5.19) that for all a
aa~(F*, 0*, a) f(F*, 0*, 0, a) ~O, ( 6.13)
(6.6) ~(F*, 0*, a), as a function of a, can serve as a Lyapunov
function 26 for the equation (6.2), and familiar results
in the theory of ordinary differential equations27 yield
the following theorem: Let (F*, 0*, a*) be an equilibrium
(6.7) state and suppose that, in some neighborhood of a*, the
following strengthened form of (6.9) holds:
~(F*, 0*, a) >~(F*, 0*, a*) if a~ a*. (6.14)
(6.8) Then (F*, 0*, a*) is Lyapunov stable at constant strain
and temperature.
If
'Y>O whenever F=O, 8=0, g=O, but ci~O, (6.15)
then we say that the material under consideration
is strictly dissipative. It is not difficult to show that
(4.5), (5.21), and (5.24) imply the following proposi-
tion: A material is strictly dissipative if and only if the
internal dissipation inequality (5.19) is strict in the sense
that
(6.9) aa~(F*, 0*, a) f(F*, 0*, 0, a) <0 (6.16)
for every triplet (F*, 0*, a) that is not an equilibrium
state.
Suppose that (F*, 0*, a*) is an equilibrium state
(6.10) and that (6.16) holds for all a~ a* in some neighbor-
hood m: (a*) of a*; if (F*, 0*, a*) is asymptotically
stable at constant strain and temperature then the
argument which gave us (6.8) also gives us the stronger
24 Th. De Donder, Bull. Classe Sci. Acad. Roy. Belg. 5, 7, 197,
205 (1922).
25 See, for example, 1. Prigogine and R. Defay, Chemical Ther-
modynamics, Translated by D. H. Everett (Longmans Green and
Co., Ltd., London, 1954), Chaps. 1-4, particularly Eq. (3.29)
on p. 40 and Eq. (4.29) on p. 52.
26 J. Wei, J. Chern. Phys. 36, 1578 (1962), has pointed out that
thermodynamic quantities, such as the free energy, can yield
Lyapunov functions for investigations of stability properties of
chemical reactions.
27 See, for example, P. Hartman, Ordinary Differential Equations
(John Wiley & Sons, Inc., New York, 1964), pp. 37-40, particu-
larly Theorem 8.1.
 THERMODYNAMICS WITH INTERNAL STATE VARIABLES 603

result (6.14) for all a in ;n(a*). On the other hand, Now,

it follows from Lyapunov's theorem on asymptotic
stability 28 that if (6.14) and (6.16) hold for all a;t' a* ooif;*(F, 0) =ooif;(F, 0, a)
in a neighborhood of a*, then (F*, 0*, a*) is asymp- +aaif;(F,O, a) aoa(F, 0), (6.19)
totically stable. Combining these observations with the
remark containing (6.14), we obtain the following and a similar formula holds for aFif;. Therefore, the
theorem: Let (F*, 0*, a*) be an equilibrium state with general entropy relation (5.16) and stress relation
the property that (6.16) holds for all a;t' a* in some (5.17), when combined with the equation of internal
neighborhood of a*. Then equilibrium (6.10), yield the equations
(i) (F*, 0*, a*) is asymptotically stable at constant 7]*= -ao.if;*(F*, 0*), (6.20)
strain and temperature if and only if (6.14) holds in
some neighborhood of a*. S* =oF.if;*(F*, 6*), ( 6.21)
(ii) If (F*, 0*, a*) is asymptotically stable at constant
strain and temperature then (F*, 0*, a*) is also Lyapunov and since (6.10) holds at every asymptotically stable
stable under the same conditions. 29 equilibrium state we can assert the following proposi-
tion: If the equilibrium state corresponding to (F*, 0*)
If (F*, 0*, a*) is asymptotically stable, then,
throughout some neighborhood of a*, the only triplet
is asymptotically stable, then ~*(F*, 0*) ix
given by the
equilibrium entropy relation (6.20) and S* (F*, 0*) is
of the type (F*, 0*, a) that is an equilibrium state given by the equilibrium stress relation (6.21).
is the one for which a= a*. [If this were not the case, It follows from the sentence containing (5.20) that
then in each neighborhood of a* we could find an whenever (6.10) holds we have, for all g,
ao such that aCt) == ao satisfies (6.2) and (6.3) but
not (6.4).J Thus we have the following corollary of q(F*, 0*, g, a*) g:$;O. (6.22)
the preceding theorem: Suppose that the material under
consideration is strictly dissipative and let (F*, 0*, a*) Hence this inequality holds for all g whenever (F*, 0*, a*)
be an equilibrium state that is asymptotically stable at is asymptotically stable. The tensor
constant strain and temperature. Then
(6.23)
(i) The inequality (6.14) holds in a neighborhood of a*.
(ii) (F*,O*, a*) is Lyapunov stable at constant strain may be called the equilibrium thermal-conductivity tensor
and temperature. corresponding to (F*, 0*). Since (6.23) and the last
equation of (6.18) yield
For the remainder of this section we assume that
corresponding to each strain-temperature pair (F*, 0*) gq(F*, 0*, g, a*) =g.q*(F, 0*)
there is exactly one internal state vector a* such that
2
(6.1) holds, i.e., such that the triplet (F*, 0*, a*) -gK(F*, O*)g+o(1 g 1 ), (6.24)
forms an equilibrium state for the material at X. The
(6.22) holds for all g only if30
function
a*= a*(F*, 0*) ( 6.17) q*(F*, 0*) =0 (6.25)
determined by this correspondence may be called the and
equilibrium response function for a. Using (5.15) and gK(F*, O*)g~O (6.26)
(4.4) we may construct equilibrium response functions
if;*, ~*, 8*, and q*. These functions give the equilib- for all g. This proves the following theorem: At an
rium free energy if;*, entropy 7]*, stress S*, and heat asymptotically stable equilibrium state the equilibrium
flux q* as functions of only F* and 0*: heat flux vanishes and the thermal-conductivity tensor is
if;* = if;*(F*, 0*) ==if;(F*, 0*, a* (F*, 0*) ), positive semidefinite.
It is a corollary to this theorem that
7]*=~*(F*, 0*) ==~(F*, 0*, a*(F*, 0*,
q =q(F*, 0*, g, a*) =K(F*, O*)g+o(l g\). (6.27)
S*=8*(F*, 0*) ==S(F*, 0*, a*(F*, 0*,
q*=q*(F*, O*)==q(F*, 0*, 0, a*(F*, 0*. (6.18) Thus, at an asymptotically stable equilibrium state
Fourier's law holds to within an error 0 (I g\).
28 See Ref. 27, Theorem 8.2.
29 Usually when one defines asymptotic stability one presup-
poses that the singular point under consideration is Lyapunov 30 This equation can also be obtained using either an argument
stable. Here, however, we have been considering a concept of given by Pipkin and Rivlin 31 or one given by Coleman and Noll13
asymptotic stability that is separate from Lyapunov stability. (pp. 175 and 176).
Thus it is worthwhile for us to observe that Lyapunov stability 31 A. C. Pipkin and R. S. Rivlin, Tech. Rept. No. DA 4531/4,
is actually implied by our weak concept of asymptotic stability from Brown University to the U.S. Army Ordnance Corps,
when (6.16) holds. 1958.

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
65.30.29.111 On: Sun, 16 Mar 2014 01:05:33
 604 n. D. COLEMAN AND
7. ENTROPY AS AN INDEPENDENT VARIABLE
The heat capacity c is defined by 32
c=ae(F,8, a), (7.1)
where is the response function for the internal energy:
(F,8, a) =f(F, 8, a) +8~(F, 8, a). (7.2)
It follows from the entropy relation (5.16) that
(7.3)
Henceforth we assume that c is >0 for all (F, 8, a).
Since 8 is >0, this implies that ~(F, fJ, a) is smoothly
invertible in 8 for each choice of F and a, and we may
rewrite our basic constitutive equations (4.1)-(4.5) in
the forms
e= E(F, 'f/, a), (7.4)
8= O(F, 'f/, a), (7.5)
S=S(F, 'f/, a), (7.6)
q=q(F, 'f/, g, a), (7.7)
a=f(F, 'f/, g, a). (7.8)
ThefunctionO(F,., a) in (7.5) is just the inverse of
the functi,sm ~(F, ., a) in (5.15)2; the functions E,
S, q, and f are given by
E(F, 'f/, a) = (F, O(F, 'f/, a), a}
=f(F, O(F, 'f/, a), a}+'f/O(F, 'f/, a),
S(F, 'f/, a) = S(F, Ii(F, 'f/, a), },
q(F, 'f/, a) =q(F, O(F, 'f/, a), a),
f(F, 'f/, g, a) =f(F, O(F, 'f/, ), g, ex}. (7.9)
It follows from (7.9)1 and familiar chain rules that
aqE= (aef) (aqO) +o+'f/a~O,
aFE=aFf+(aef) (aFO) +'f/aFO,
aa E= (aef) (aaO) +aaf+'f/aaO. (7.10)
Hence, the entropy relation (5.16) implies that
a~E(F, 'f/, ) = O(F, 'f/, ),
aFE(F, 'f/, a) =aFf(F, 8, a),
a"E(F, 'f/, a) =a",f(F, 0, a). (7.11)
Of cours~ the number 8 in (7.11h,s is just the tem-
perature 8(F, 'f/, a) corresponding to F, 'f/, and a.
Equation (7.11h is the expected temperature relation
8=avE(F, 'f/, a), (7.12)
while (7.11h may be combined with (5.17) to obtain
32 Since both F and IX are held fixed in computing the derivative
m (7.1), the number c may also be called the instantaneous specific
!zeat at constant strain (and per unit mass).
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
65.30.29.111 On: Sun, 16 Mar 2014 01:05:33
M. E. GURTIN
the following form of the stress relation:
S=aFE(F, 'f/, a). (7.13)
It follows from (7.11)s and (7.9)3.4 that the general
dissipation inequality (5.14) may be written in the form
po(F, 'f/, a)aaE(F, 'f/, a) f(F, 'f/, g, a)
+q(F, 'f/, g, a) g::;O, (7.14)
and, in particular,
q=O::::}oaE(F, 'f/, ex)l(F, 'f/, g, a)::;O (7.15)
which is useful for materials which do not conduct
heat. Of course we also have the following counterpart
of (5.19) :
oaE(F, 'f/, a) f(F, 'f/, 0, a) ::;0. (7.16)
According to the definition laid down in Sec. 6, a
triplet (F*, 8*, a*) is called an equilibrium state of
the material at X if and only if it obeys (6.1) at X.
It follows, however, from (7.9)4 that this definition
is equivalent to asserting that the triplet (F*, 'f/*, a*)
with
'f/*=~(F*,8*, a*), 8*=O(F*, 'f/*, a*), (7.17)
characterizes an equilibrium state if and only if
f(F*, 'f/*, 0, a*) =0. (7.18)
When (7.18) holds we define the domain of attraction of
(F*, 'f/*, a*) at constant strain and entropy to be the
set D(F*, 'f/*, a*) of internal state vectors ao such that
the solution a= aCt) of the initial value problem
a=f(F*, 'f/*, 0, a), a(O) = ao, (7.19)
exists for all t?,O and approaches a* as t-H/J. We say
that (F*, 'f/*, a*) is asymptotically stable at constant
strain and entropy if D(F*, 'f/*, a*) contains a neighbor-
hood of a*.
Since it follows from (7.16) that E(F*, 'f/*, a(t)} is
a decreasing function of t along all solutions of (7.19),
the argument which gave us (6.8) here tells us that
for each a in D(F*, 'f/*, a*)
E(F*, 'f/*, a)?, E(F*, 'f/*, a*). (7.20)
In particular, if (F*, 'f/*, a*) characterizes an equilib-
rium state that is asymptotically stable at constant strain
and entropy, then (7.20) holds for all a in some neighbor-
hood of a*, and
(7.21)
It of course follows from (7.11)s that (7.21) and
(6.10) are equivalent, i.e., are just different forms of
the same equation for a*. This does not, however,
imply that the concepts of stability at constant strain
and temperature and stability at constant strain and
entropy are logically equivalent: The equation (7.21)
 THE R MOD Y N A M I C S WIT H I N T ERN A L S TAT E V A R I A B L E S 605

or (6.10) is a necessary, but not a sufficient, condition It follows from the first two of these relations that
for these two types of stability. (7.12) and (7.13) can be written in the forms
We say that a triplet (F*, T/*, a*) obeying (7.18)
8=[a,ii(F, , a)]-I, (8.8)
characterizes a state that is Lyapunov stable at constant
strain and entropy if for each >O there exists a 0,>0 s= -aFii(F, , a)/a,ii(F, , a) = -8a Fii(F, , a). (8.9)
such that every solution a= aCt) of (7.19h having
[ a(O) - a* I <li, also has I aCt) - a* I < for all t';?O. From (8.7)3, (8.9), and (7.11h we obtain
The following proposition is an immediate consequence a"ij(F, , a) = -t11a",E(F, T/, a)
of (7.16) and the theory27 of Lyapunov functions: If
(F*, T/*, a*) obeys (7.18) and has the property that for = -t11a",-f(F, 8, a), ( 8.10)
all a in some neighborhood of a* and hence (7.16) becomes
E(F*,T/*, aE(F*,T/*, a*) whenever arfa*, (7.22) a",ii(F, , a)f(F, , 0, a)';?O. (8.11 )
then (F*, T/*, a*) characterizes a state that is Lyapunov
It follows from (8.2)4 that a triplet (F*, T/*, a*)
stable at constant strain and entropy.
characterizes an equilibrium state, i.e., obeys (7.18)
It follows from (7.11)3, (7.9)4, and the proposition
if and only if the triplet (F*, *, a*) with
containing (6.16) that a material is strictly dissipative
if and only if *=i:(F*, T/*, a*), T/*=ij(F*, *, a*), (8.12)
a"E(F*, T/*, a) 'f(F*, T/*, 0, a) <0 (7.23) obeys the relation
whenever (F*, T/*, a) does not characterize an equilibrium f(F*, *, 0, a*) =0. (8.13)
state.
The last proposition of this section follows from When such is the case, the domain of attraction of
arguments completely analogous to arguments used (F*, *, a*) at constant strain and energy is the set
in Sec. 6: If (F*, T/*, a*) obeys (7.18) and if (7.23) D(F*, *, a*) of all internal-state vectors ao with
holds for all arf a* in some neighborhood of a*, then the property that the solution a= aCt) of the initial
the inequality (7.22) is a necessary and sufficient con- value problem
dition for the state characterized by (F*, T/*, a*) to be n=f(F*, *, 0, a), ( 8.14)
aCO) = ao,
asymptotically stable at constant strain and entropy.
exists for all t';?O and converges to a* as t-HJJ. We
8. INTERNAL ENERGY AS AN INDEPENDENT say (F*, *, a*) is asymptotically stable at constant
VARIABLE strain and energy if D(F*, *, a*) contains a neighbor-
The temperature relation (7.12) and our assumption hood of a*. If for each >O there exists a 0,>0 such that
that 8 is strictly positive imply that the function i: in I a(O) - a* I <li, implies I aCt) - a* I < for all t';?O,
(7.4) is smoothly invertible in its second variable T/. then we say that (F*, t*, a*) is Lyapunov stable at
Hence there exists a function ij such that constant strain and energy.
The next two theorems follow from (8.11).
T/=ij(F, , a), (8.1) If a is in the domain of attraction of (F*, *, a*)
and putting then
O(F, , a) = (j(F, ij(F, , a), a), ij(F*, *, a)::; ij(F*, t*, a*), (8.15)
S(F, , a) =S(F, ii(F, , a), a), Hence, if (F*, *, a*) is asymptotically stable at con-
q(F, , a) =q(F, ii(F, , a), a), stant strain and energy, (8.15) must hold throughout
some neighborhood of a* and
f(F, , a) =f(F, ii(F, , a), a), (8.2)
(8.16)
we can write the constitutive equations (7.5)-(7.8)
as follows: [It follows from (8.10) that the equations (6.10),
O=O(F, , a), (8.3) (7.21), and (8.16), giving necessary conditions for
S= S(F, , a), (8.4) three distinct types of stability that an equilibrium
state may manifest, are equivalent.]
q=q(F, , g, a), (8.5) If (F*, *, a*) obeys (8.13) and if for all a in some
n=f(F, , g, a). (8.6) neighborhood of a*
The equations (7.4) and (8.1) imply that ii(F*, *, a) <ii(F*, *, a*) whenever arf a*, (8.17)
a,ij= (a~E)-I,
then (F*, *, a) is Lyapunov stable at constant strain
aFij+ (a,ii) (aFE) =0, and energy.
(a,ij) (a",E) +a",ij=o. (8.7) We may conclude from (8.10), (8.2)4, and the re-

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
65.30.29.111 On: Sun, 16 Mar 2014 01:05:33
 606 B. D. COLEMAN AND M. E. CtJRTIN
mark containing (7.23) that the material under con-
sideration is strictly dissipative if and only if
a"ii(F*, E*, a) f(F*, e*, 0, a) >0
whenever f(F*, e*, 0, a) r!0.
Using arguments given in Sec. 6 one can easily
prove the following proposition which is similar to
the last assertion of Sec. 7: If (F*, e*, a*) obeys (8.13)
and if (8.18) holds for_all ar! a* in a neighborhood of a*,
then (8.17) supplies a necessary and sufficient condition
for (F*, E*, a*) to be asymptotically stable at constant
strain and energy.
9. RELATION OF ISOTHERMAL TO ISENTROPIC
AND ISOENERGETIC STABILITY
Let (F*, 8*, a*) be an equilibrium state for the
material at X, and let
7]*=~(F*, 8*, a*),
e* = f(F*, 7]*, a*).
Assuming that the strong forms (6.16), (7.23), and
(8.18) of the internal dissipation inequality hold, we
here show that asymptotic stability of the state
(F*,8*, a*) at constant strain and temperature implies
asymptotic stability of the same state at constant
strain and entropy, and that this latter type of sta-
bility is equivalent to stability at constant strain and
energy. To prove the first implication we make use
of our assumption that the heat capacity c of Eqs.
(7.1) and (7.3) is positive, i.e., that
a8~(F, 8, a) >0
for all F, 8, and a. Using the entropy relation (5.16),
we can write (9.3) in the form
a8Z/t(F,8, a) <0,
which implies that, for each pair (F, a), $(F, 8, a)
must be a concave function of 8:
$(F,8, a) -$(F, 8', a)
- (8-8')a8$(F, 8', a) ::;0,
or, equivalently,
$(F,8', a) ?:.$(F, 8, a) +(8-8'H(F, 8', a).
Using (9.6) we now prove the following lemma:
If 7]* is given by (9.1) then
f(F*,7]*, a) -f(F*, 7]*, a*)
?:.$(F*,8*, a) -$(F*, 8*, a*).
It follows from (7.9)1 that
e(F*,7]*, a) -f(F*, 7]*, a*) =$(F*, e(F*, 7]*, a), a)
-$(F*,8*, a*) +[e(F*, 7]*, a) -8*J7]*.
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
65.30.29.111 On: Sun, 16 Mar 2014 01:05:33
By (9.6),
$(F*, e(F*, 7]*, a), a)'2.$(F*,8*, a)
(8.18)
+[8*-O(F*, 7]*, a) J~(F*, O(F*, 'YJ*, a), a). (9.9)
But
?)(F*, O(F*, 'YJ*, a), a)=7]*, (9.10)
and, therefore, when (9.9) is added to (9.8) the result-
ing inequality reduces to (9.7).
Using (9.7) we now establish the following theorem:
Let (F*, 8*, a*) be an equilibrium state, let 7]* be given
by (9.1), and assume that the strong forms (6.16) and
(7.23) of the internal dissipation inequality hold for all
ar! a* in some neighborhood of a*. If the state (F*, 8*, a*)
is asymptotically stable at constant strain and tempera-
ture, then this state is asymptotically stable at constant
strain and entropy.
Proof: Let (F*, ()*, a*) be asymptotically stable at
(9.1) constant strain and temperature. Since we here assume
(9.2) that (6.16) holds for all ar! a* in a neighborhood of a,
it follows from conclusion (i) of the theorem on p. 603
(Sec. 6) that for a in a neighborhood m( a*) of a*
a,t.a*~$(F*,()*, a$(F*, 8*, a*). (9.11)
It is clear from our lemma (9.7), that (9.11) has the
consequence that for a in m( a*)
ar!a*~(F*,7]*, a(F*,7]*, a*). (9.12)
But, since we here also assume (7.23), we may observe
that the last sentence of Sec. 7 tells us that (9.12)
is a sufficient condition for asymptotic stability at
(9.3) constant strain and entropy of the state characterized
by (F*, 7]*, a*), i.e., the state (F*, ()*, a*), and this
completes the proof.
To show the equivalence of isentropic and isoener-
(9.4) getic stability we first show that if e*=e(F*, 7]*, a*),
then
e*<e(F*,7]*, a) if and only if ii(F*, e*, a) <7]*. (9.13)
Indeed, it follows from (8.1) and the positivity of
() that ii(F, e, a) is a strictly increasing function of e
and hence
(9.5)
e*<e(F*,7]*, a) if and only if ii(F*, e*, a*)
(9.6) <ii(F*, f(F*, 'YJ*, a), a). (9.14)
But since the function ii(F, " a) is the inverse of the
function E(F, ., a), we have, for each value of a,
ii(F*, e(F*, 7]*, a), a)=1)*, (9.15)
(9.7) and therefore (9.14) implies (9.13).
Our desired result is now an immediate consequence
of (9.13) and the concluding sentences of Sees. 7 and 8:
When the strong forms (7.23) and (8.18) of the internal
dissipation inequality hold for all a,t. a* in a neighborhood
(9.8) of a*, an equilibrium state is asymptotically stable at
 THE R MOD Y N A M I C S WIT H I N T ERN A L S TAT E V A R I A B L E S
constant strain and entropy if and only if it is asymp-
totically stable at constant strain and energy.
10. INTEGRAL DISSIPATION INEQUALITIES
In a body undergoing an arbitrary thermodynamic
process, we may consider, at a given material point,
the following general dissipation integrals,33 defined for
any two times to<h:
(10.1)
(10.2)
t1
X(tl, to) =
1 to
tr1 {SF-E-(pO)-lqgldt. ( 10.3)
It follows immediately from the law (3.2) of positive
production of entropy and the expressions (3.3) and
(3.8) for'Y that
d (tl, to) ~1J-(tl) -1J-(to) ,
~(tl, to) ~E(h) -E(to),
X(tl, to) ~'I/(to) -'l/(h). (lOA)
The theory we have been developing here can be
brought closer to the theory of materials with fading
memory2-4,15 by adding assumptions about the sta-
bility of solutions of the differential equation governing
the evolution of the internal state vector a. For the
remainder of this section we assume the following
stability postulate:
Corresponding to each pair (F*, 0*) there is an internal
state vector a* which makes (F*, 0*, a*) an equilibrium
state that is asymptotically stable in the large at constant
strain and temperature.
In other words, we assume that given (F*, 0*) there
exists at least one a* such that (F*, 0*, a*) obeys
(6.1) and also such that every solution of (6.2) obeys
(6.4).
This postulate has the following consequences.
(i) Corresponding to each pair (F*, 0*) there is exactly
one equilibrium state (F*, 0*, a*).
(ii) At each equilibrium state (F*, 0*, a*), the in-
equalities (6.9), (7.20), and (8.15) hold for all a with
'1/* and E* given by (9.1) and (9.2).
To demonstrate Conclusion (i) we need merely note
that if to a given pair (F*, 0*) there corresponded two
equilibrium states (F*, 0*, a*) and (F*, 0*, at),
both obeying (6.1) with a*~at, then a(t)=a t would
33 Our present X, d, and .g differ only in the terms containing q
from the functionals which Coleman (on pp. 27, 35, and 39 of
Ref. 15) calls "dissipation integrals."
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
65.30.29.111 On: Sun, 16 Mar 2014 01:05:33
607
be a solution of (6.2), and, by our stability postulate,
we would have a(t)~a*, which is impossible since
aCt) =const~ a*. Conclusion (ii) is an immediate con-
sequence of (9.7), (9.13), and the sentence following
(6.8) .
The stability postulate is not a consequence of the
Clausius-Duhem inequality but is rather an inde-
pendent hypothesis expected to hold for some materials
but not all. Under this new postulate a single-valued
function a* relates (F*, 0*) to a*, as in (6.17), and
equilibrium response functions ~*, ~*, etc., may be
defined as in (6.18). Of course (6.20)-(6.27) hold,
and the remark following (6.8) here yields
~(F, 0, a) ~~*(F, 0) (10.5)
for all a.
Consider now a material point X and a time to,
and suppose that at X the triplet (F, 0, g) has the
following sort of time dependence:
(F (t), 0(t), g (t) ) = arbitrary values for t < to,
= (Fo, 00, 0) for t~to. (10.6)
At the time to, the internal state vector a will have a
value ao, determined, through the differential equation
(4.5), by the histories of F, 0, and g up to time to and
some initial data for a. Although we may not know
the value of ao, we do know that for t>to, aCt) will
be the solution of (6.2) corresponding to the initial
condition aCto) = ao. Since, by our present stability
hypothesis, all solutions of (6.2) must approach a* =
a*(Fo, ( 0) we have
lima(t) = a*(Fo, ( 0 ), (10.7)
t_co
Therefore, it follows from the (tacitly assumed) con-
tinuity of ~, ~, T, q, and the equations (6.18) and
(6.25), that at the point X
lim1J-(t) =lim~(Fo, 00 , a(t) )=~*(Fo, ( 0 ),
t-ro t-.m
lim'l/(t) =lim71(Fo, 00, aCt) )=~*(Fo, ( 0),
limT(t) =limT(Fo, 00 , aCt) )=T*(Fo, ( 0 ),
t-m t-co
limq(t) =limq(Fo, 00 , 0, aCt) )=0. (10.8)
t-co t-co
When a history up to some time to is extended beyond
to in the manner illustrated in (10.6), we call the result-
ing extended history the isothermal static continuation
of the original history. Equation (10.8) shows, in
effect, that in any isothermal static continuation of an
arbitrary history, the free energy, entropy, stress, and
heat flux all approach their equilibrium values as t
increases without limit. Moreover, the first of these limits
is approached monotonically, for (3.9) here yields
fCt) =:; 0, for all t> to. (10.9)
 608 B. D. COLEMAN AND M. E. GURTIN

The relations (10.5) and (10.8h tell us that the
theorem on minimum free energy, originally proven for
materials with memory,34 has the following form in
the present theory, provided we assume that equilib-
rium solutions of (6.2) are stable in the large: Of all
histories ending with given values of F and 0, the constant
history with g held fixed at zero and F and 0 held fixed
at the given values gives rise to the least free energy.
Thus, our new stability postulate gives the present
theory many, but not all, the qualitative properties of
Coleman'sl5 theory of materials with memory.
Let us now return to the dissipation integrals defined
in (10.1)-(10.3) and consider processes which start
from an equilibrium state, i.e., processes for which we
have at X,
F(to) =F*, O(to) =e*,
o:(to) = 0:*= a*(F*, 0*),
but for which F, 0, and g at X are given by arbitrary
functions F(t), O(t), g(t) for t>to. For each such
process, at X
following implications holds for processes which start at
to from an equilibrium state of the type (10.10), but which
are otherwise arbitrary:
F(lt) =F(to), e(to) =e(tr)~t1(tl, to) '2 0,
F(tl) =F(to) , 1/(to) =1/(tl) =} ~(tl, to)'20,
F(tl) =F(to), f(t O) = f(tl) =}X(tl, to)'20. (10.20)
We call a process isothermal at X if 8=0 at X,
isentropic if 1]==0 at X, and isoenergetic if f==O at X.
It is clear from (10.1)-(10.3), that the theorem just
proven has the following corollary yielding inequalities
for integrals of the stress power in special cyclic
processes36 : Consider a process starting from an equilib-
rium state of the type (10.10) and for which the tempera-
ture gradient g is zero for to 5, ts. tl . If the process is
( 10.10) either isothermal or isentropic, then
tl
Ft(I)=F(to)=}
I to
SFdt'20. (10.21)

if;(to) =tf(F*, e*, 0:*), (10.11) If the process is isoenergetic, then

if;(t) =tf(F(t), e(t), o:(t)) for t>to, (10.12) ( 10.22)

where o:(t) is to be obtained by solving the differential
equation (4.5) with the initial condition o:(to) = 0:*.
11. TEMPERATURES ASSOCIATED WITH
If F and (} return to their initial values F* and e* at
INTERNAL STATES
some time tl > to, then
We here suppose that the first M of the N internal
I/;(tl) =tf(F*, e*, o:(tl)), (10.13)
state variables ai have the same dimension as the
and this equation, when set alongside of (10.11) and specific internal energy . Putting
the consequence (ii) of our stability postulate, implies .'If
that I/; (t1) is greater than if; (to) , fA=f- L:::: a'i, (11.1)
i=l
F(tl) =F*, e(lt) =e*~if;(tO)5,I/;(tl). (10.14)
and
Of course, (10.10) implies that ( 11.2)
(to)=(F*, 0*, 0:*) == *=E(F*, 1/*, 0:*), (10.15) we call fA the specific internal energy of the active mode
and ai, i= 1, 2, M, the specific internal energy of
000,

1/(to) =i)(F*,e*, 0:*) ==1/* = jj(F*, *, 0:*). (10.16)

the ith internal (or "hidden") mode. Defining the func-
If it should happen that F(tl) =F* and 1/(tI) =1/*, then tion 1/# by
M
(tl) =E(F*, 1/*, o:(tr)), (10.17) 1/#(F, A, aI, 0", aM, ~) =jj(F, fA+ L::::a" 0:), (11.3)
i=l
and, therefore, by (10.15) and the consequence (ii)
of the stability postulate, we put
F(tl) =F*, 1/(tI) =1/*~(tO)5,(tl). (10.18) eA = [a'A1/#(F, fA, aI, 000, a']f, ~)]-r, (11.4)
Similarly, ei =[a".1/#(F, fA, aI, 000, a.lf, ~)J\ (11.5)
F(tl ) =F*, f(t 1) =f*=}1/(tO)'21/(tl). (10.19) for i=l, 2, M. We refer to eA as the translational
000,

Combining (10.13), (10.18), and (10.19) with the
inequalities (lOA), we obtain the following theorem
on dissipation integralSl5 :
Under the stability postulate stated above each of the
34 Reference 15, Theorem 3, p. 26.
36 For the corresponding theorem for materials with fading
memory see Remarks 17, 24, and 32 on pp. 27,36, and 39 of
Rl;f.1.5,
temperature (or the temperature of the active mode) and
to ei as the temperature of the ith mode. We may observe
that the "temperature of the ith mode" need not be
positive. In fact, as there is at this point no reason to
suppose that a".1/#7"'O, we must not rule out the pos-
sibility of ei being infinite.
36 For materials with fading memory, analogous results are
~iven in Theorem 4/ Remark 2.5, and Remark 3,3 of Ref. 15.

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
65.30.29.111 On: Sun, 16 Mar 2014 01:05:33
 THERMODYNAMICS WITH J~TERNAL
Since (11.3) yields
( 11.6)
the temperature relation (8.8) implies that "the tem-
perature" 8 is equal to the translational temperature 8A:
8=8A. (11.7)
It also follows from (11.3) that for i=l, 2, "', M,
(11.8)
and hence we have the following expression for each 8i :
(11.9)
This formula when combined with (8.16) implies the
following theorem: At an asymptotically stable equilib-
rium state the temperature 8. of each internal mode is
equal to the translational temperature 8.'K1
Since 8>0, we conclude from the foregoing proposi-
tion that each 8; is strictly positive when the material
is close to a state of stable equilibrium.
By (11.1),
M tion '1/* defined in (11.3) has the special form (11.12),
i=O if and only if iA+ L a.=O; (11.10)
i=l
hence (8.6), (8.11), and (11.9) yield
a functions CP.* such that
-EA8 + L}.f
;=1
~ ~O whenever i=O,
8;
g=O, and ~=O.
(11.11)
The inequality (11.11) appears to be in accord with
intuitive prejudices about properties to be expected
of temperatures associated with internal modes.
Suppose now that the function '1/* defined in (11.3)
has the special form
M of CPi* at an arbitrary pair (F, ai) the temperature of
'I/*(F, tA, aI, "', aM, ~) ='l/A(F, tA, ~) + L
1'=1
'l/i(F, ai).
(11.12)
Then 8=8A is given by a function 0* of (F, tA, ~),
while each Oi is given by a function Oi* of only (F, ai) ;
for (11.4), (11.5), and (11.12) yield
= [O'A'l/A(F, EA, ~)]-\
O=O*(F, EA, ~) (11.13)
Oi=Oi*(F, a,) = [o",'l/i(F, ai) J-I. (11.14)
Let us assume, in addition, that there exists a function
a* which assigns to each pair (F*, 0*) an (*, as in (6.17),
such that the triplet (F*, 0*, 0:*) constitutes an equi-
librium state for the material point under considera-
tion. As we saw in the conclusion (i) of the first theorem
of Sec. 10, this assumption is a consequence of the
stability hypothesis of Sec. 10. If we write (6.17) in
37 This result should be compared with Truesdell's theorem
about the compatibility of total with partial caloric equations of
state for mixtures: C. Truesdell, Rend. Lined 22,33, 158 (1957),
Theorem IV.
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
65.30.29.111 On: Sun, 16 Mar 2014 01:05:33
STATE VARIABLES 609
the form
i=1,2,"',N, (11.15)
then the fact that O=8 i at states of stable equilibrium
implies
O;*(F*, &i*(F*, 0*) )=0*, i=1,2,",M. (11.16)
Since (11.16) must hold for all (F*, 0*), O;*(F*, .)
is a left inverse of &i* (F*, .). If we assume, as we shall,
that for every pair (F, ai) we can find a 0 such that
ai=&i*(F, 8), then we have the following theorem:
If the stability postulate of Sec. 10 holds and if the caloric
equation of state (8.1) can be cast in the special form
(11.12), then for each i, i=l, 2, "', M, and for each
fixed F, the function Oi*(F, .) is the inverse of the function
&i*(F, .).
We mav use the foregoing theorem to motivate an
alternativ~ definition of internal mode temperatures
which is free from the assumption that the a/s have
the same dimension as the specific internal energy.
To do this let us drop the assumption that the func-
but suppose that the first M of the functions &,* of
(11.15) are such that, for each value of F, &;*(F, 0)
is an invertible function of 0, i.e., that there exist M
cpi*(F, &;*(F, 0) )=0 (11.17)
for all 0>0. When this condition is fulfilled the func-
tion CPi* is uniquely determined by the function &i*
and hence by the function f on the right side of the
constitutive equation (4.5). We call the value
(11.18)
the ith mode corresponding to (F, ai). We continue
to refer to the number entering the constitutive equa-
tions (5.15), (4.4), and (4.5) as the translational tem-
perature. As an immediate consequence of (11.17) we
have the following proposition: Each CPi shares with Oi
the property of being equal to the translational temperature
oin states of stable equilibrium.
Unfortunately, the function CPi* is not, in general,
identical to the function (0",'1/*)-1: Although Oi=CPi=O
at equilibrium, we can expect to have O'i~CPi away
from equilibrium, except, of course, when '1/* has the
special form (11.12). As a matter of fact, if CPI* were
equal to [0"1'1/*]-\ then we could conclude from (11.18)
that 0"1'1/* must be independent of tA and 012,' " a.v,
and hence that
'I/*(F, EA, aI, "', OIN) ='1/1 (F, Oil) +'l/2(F, EA, 012, " OIN).
(11.19)
Thus we have the following result: A necessary condition
that
(11.20)
 610 B. D. COLEMAN AND M. E. GURTIN

is that the function 7]* have the special form (11.19). tensor Q, the following identities in F, 0, a, and g:
A similar result holds for a2, a3, etc.
In view of the fact that temperatures of internal l/-(F,O, a) =l/-(QF, 0, a),
states are often regarded as definable only in terms of n(F,O, a)=n(QF,O, a),
statistical concepts, we think it worthwhile to make the
following observation. The results of this section sug- QT(F,O, a)QT=T(QF, 0, a),
gest that if we accept the general assumptions made Qq(F, 0, g, a) =q(QF, 0, Qg, a),
here and in Secs. 4 and 5, then the question of whether
we can associate a temperature Oi with a particular f(F, 0, g, a) =f(QF, 0, Qg, a). (12.3)
internal state variable ai becomes one that can be These identities can be used to derive the following
settled on purely phenomenological grounds, namely by reduced forms of our constitutive equations:
inspection of the caloric equation of state and the
equation governing the evolution of ai in a thermo- if;=if;t(C, O, 0:),
dynamic process. 7J=7]t(C,O, a),
12. OBJECTIVITY T=FTt(C,O, a)FT,

In Sec. 5 we found th~ restrictions placed on the q=Fqt(C, 0, FTg, a),

response functions l/-, n, S, q, and f of (4.1)-(4.5) a=ft(C, 0, FTg, a), (12.4)
by the postulate of positive production of entropy. where
The principle of material frame indifference,38 which (12.5)
states that an admissible process must remain ad-
missible after a change of frame, imposes further re- is the right Cauchy-Green tensor; of course, by (2.4)
strictions on response functions. and (2.5), FTg is just GRADO. In terms of the response
In our pres en t theory a change of frame is charac- functions introduced here, the entropy relation (5.16)
terized by a time-dependent orthogonal tensor Q. The may be written
scalars 0, , 7], and if; are unaffected by a change of 7]t=-aeif;t (12.6)
frame, but F, g, q, and T transform as follows:
and the stress relation (5.17) becomes
F~QF,
Tt=2pa cif;t. (12.7)
g~Qg,
One might imagine that in some applications a
q~Qq, could represent a vector of dimension three transforming
as a spatial position vector under a change of frame,
T~QTQT. ( 12.1) but such an interpretation for a is not compatible
The manner in which the internal state vector a= with our constitutive equation (4.5). For, if
(ai, " ' , aN) transforms under a change of frame a~Qa, ( 12.8)
depends, of course, on the physical interpretation given
to the components ai of a and therefore varies with then for each time-dependent orthogonal tensor Q, f
the application in mind. For definiteness we assume must obey the identity
here that a is an ordered N-tuple of scalars, each of
which remains invariant upon a change of frame; i.e., (Qa)'=f(QF,O, Qg, Qa), (12.9)
we assume that simultaneously with (12.1) we have
where Q is given by (4.5). Since
a~a. (12.2)
(Qa)'=Qa+Qa=Qf(F, 0, g, a)+Qa, (12.10)
(This is the case, for example, when each ai represents when Q (t) = 1, (12.9) yields
the degree of advancement of a chemical reaction, the
fraction of the total energy in a certain internal mode Qa=O, (12.11)
of molecular motion, or the probability that an internal
degree of freedom is in a particular quantum state.) which equation must hold for all choices of Q at time t.
It follows from (12.2) that the response functions of the But, when Q(t) = 1, the only restriction on Q(t) is that
present theory are compatible with material frame this tensor be skew. Hence (12.11) states that Wa=O
indifference if and only if they obey, for each orthogonal for all skew tensors W, which implies that a=O. In
other words, if a behaves as a spatial three-vector under
38 W. Noll, Arch. Rational Mech. Analysis 2, 197 (1958), who
changes of frame, then (4.5) must be replaced by the
called it the "principle of objectivity of material properties." trivial equations f=O, a=O.

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
65.30.29.111 On: Sun, 16 Mar 2014 01:05:33
 THERMODYNAMICS WITH INTERNAL STATE VARIABLES 611

13. MATERIAL SYMMETRY, FLUIDS gases with internal molecular relaxation forms an im-
portant special case of our present theory, we give the
The symmetry group S of a material is the set of all reduced forms taken by our general constitutive equa-
nonsingular changes of local reference configuration tions when the material under consideration is a fluid.
that leave the response functions of the material un- We say that a material is a fluid if its symmetry
altered.39
If we assume that the internal state variables are
quantities unaffected by the choice of reference con-
a. group S is the group 'U of all unimodular tensors. The
property of being a fluid is intrinsic to the material
in the sense that S='U for one reference configuration
figuration, then in our present theory the symmetry only if S='U for all reference configurations. Assuming
group S may be identified with the group of all m- that the internal state variables a, obey the "scalar
vertible tensors H for which the identities transformation rule" (12.2) under changes of frame,
$(F, 8, a) =$(FH, 8, a), and using arguments now familiar in continuum
mechanics, one can easily show that for a fluid the
n(F, 8, a) =~(FH, 8, a), following representation formulas must hold for the re-
T(F,8, a) =T(FH, 8, a), sponse functions occurring in the identities (12.3) and
(13.1) :
q(F, 8, g, a) =q(FH, 8, g, a),
(13.3)
f(F, 8, g, a) =f(FH, 8, g, a), (13.1)
(13.4)
hold for all F, 8, a, and g. This group S depends on
the choice of reference configuration, but the sym- T=-pl, p=p(u,8, a), ( 13.5)
metry group corresponding to two different reference q=-Kg, K=R(u,8,g, a), ( 13.6)
configurations of the same material are conjugate and
hence isomorphic. A theorem of Gurtin and Williams40 ci=f(v, 0, g, a). (13.7)
tells us that the postulate of positive production of
entropy requires that S must be a subgroup of the uni- Here u = 1/p is the specific volume; g= (g. g) 1/2 is the
modular group; i.e., that each H in S must have magnitude of gradO; $, n, p, R, and the components
I detH 1=1. ji of f are all scalar-valued response functions; p
The similarity between the equations (13.1) and the and K are called, respectively, the pressure and the
identities used to define the symmetry group of an thermal conductivity. It follows from (13.3)-(13.5) that
elastic material with heat conduction is striking,41 and for fluids the entropy relation (5.16) and the stress
one can immediately apply here many results originally relation (5.17) become
obtained within the framework of the theory of elastic
TI=-a8~(V, 0, a),
materials. For example, using an argument given by
Coleman and Nollia one can easily show that if the (13.8)
inversion transformation -1 is in S then
When it is more convenient to take the specific entropy
q(F, 8, 0, a) =0 (13.2) of the fluid, rather than the temperature, as an inde-
for all F, 8, and a. Thus, even if (F, 8, a) is not an pendent variable, we have
equilibrium state, the presence of -1 in S implies =E(V, TI, a),
that the heat flux q vanishes when grad8 vanishes.
We do not discuss here the symmetry groups ap- O=O(u, TI, a) =a~E(U, TI, a),
propriate to the various types of solids. The interested
reader will have little difficulty in extending to our p=p(u, TI, a) = -avE(U, TI, a), (13.9)
present subject the known theory of symmetry in finite
and when the specific internal energy is the preferred
elastic deformations.42 However, since the theory of
independent variable
39 This is essentially a definition due to Noll3s who defined the
local isotropy group 9 of a material to be the set of density-pre- TI=71(u, , a),
serving changes of local reference configuration which leave the
response unaltered. It has recently been shown40 that our present 8= B( u, , a) = [a.71 ( u, , a)]-I,
S must reduce to Noll's g.
40 M. E. Gurtin and W. O. Williams, Arch. Rational Mech. p=p(v, , a) =oav71(v, , a). (13.10)
Analysis 23, 163 (1966).
41 See, for example, Ref. 13, Eq. (3.7). It is clear that for fluids all the remarks we made
42 For some pertinent results in that theory see the articles of
Smith and Rivlin43 and Coleman and Noll.44 about stability in Secs. 6-10, remain valid if the modi-
43 G. F. Smith and R. S. Rivlin, Trans. Am. Math. Soc. 88, fier "at constant strain" is replaced by "at constant
175 (1958). volume"; i.e., if the condition that F be constant is
4' B. D. Coleman and W. Noll, Arch. Rational Mech. Analysis
15, 87 (1964). replaced by the weaker condition that u be constant.

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
65.30.29.111 On: Sun, 16 Mar 2014 01:05:33
 612 R. D. COLEMAN AND
Equation (13.5h asserts that, under the assumption
that each ai transforms as a scalar upon changes of
frame, for a fluid the stress is always a hydrostatic
pressure. This observation, when combined with the
comments made at the end of Sec. 12, suggests that
the present theory lacks the breadth of applications
enjoyed by the theory of the thermodynamics of
materials with fading memory. I.; In that theory a fluid
out of equilibrium can exhibit shearing stresses. It
should not, however, be concluded that the present
theory is a trivial special case of the theory of Ref. 15.
Three reasons for this are the following: (a) The ma-
terials we consider here need not obey the principle of
fading memory; in fact, that principle follows from
our present assumptions only after the addition of very
strong stability assumptions for the differential equa-
tion (4.5). (b) Since g appears in (4.5), for the ma-
terials we consider here the past history of g may in-
fluence the stress. (c) Those special materials which
are covered by both the present theory and the fading
memory theory can be described in the present theory
with a mathematical language that is somewhat
simpler and which, at the present time, appears closer
to kinetic-theory interpretations (a good example of
such a special case is given in our concluding section).
It is clear that by adding F to the set (F, (J, g, 0:) of
independent variables used in (4.1) -( 4.5), one can
generalize our present development so as to obtain a
theory in which fluids exhibit shearing stresses and in
which 0: can be a string of tensors of arbitrary tensorial
rank under changes of frame. 45 However, such a theory,
since it would mix two distinct but specialized concepts
of internal dissipation, would lack both the simplicity
of the present theory and the breadth of the general
theory of materials with fading memory.
14. IDEAL GASES WITH VffiRATIONAL
RELAXATION
We here illustrate our theory with a simple special
case: a dilute nondissociating diatomic gas which, al-
though it does not conduct heat and obeys the ideal gas
law in the form
pv=R(J, (14.1)
with R a constant, can, nonetheless, exhibit dissipa-
tive phenomena due to the finite time required for the
transfer of molecular motion from the translational
mode, where it contributes to the pressure, to the
vibrational mode, where it does not. To place this
familiar theory in our present framework, we first
assume that the entropy of the gas can be written in
the form
(14.2)
46 A simple and particularly elegant special case of a fluid of
this type, with the internal-state vector a transforming as a
spatial position vector, has been studied by J. L. Ericksen, Kol-
loid-Z. 173, 117 (1960).
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
65.30.29.111 On: Sun, 16 Mar 2014 01:05:33
M. E. GURTIX
with
(14.3)
a is called the vibrational energy, fA is called the active-
mode energy,46 the value '1v of ;jv is called the vibrational
entropy, and the value '1A of ;jA is called the active-mode
entropy. The number
( 14.4)
IS the translational (or active-mode) temperature, while
(14.5)
is the vibrational temperature. We now add the usual
assumption that a obeysA a scalar differential equation
of the type (13.7) with f independent of g and linear
a=A (v, e) -B(v, (J)a. (14.6)
If we assume that the temperature e obeys a con-
stitutive equation of the form e= o( v, f, a), then the
present theory based on the constitutive relations
(14.1)-(14.3) and (14.6) clearly falls as a special case
of the theory of fluids obeying (13.3)-(13.10). (Of
course, here the internal-state vector 0: has just the
one component a.) We may therefore conclude from
(13.1Oh.3, (14.2), (14.3), and (14.4) that
p=eOv;jA(V, fA), (14.7)
(J=eA. (14.8)
Thus in this theory "the temperature" (J must be identi-
fied with the translational temperature eA. This fact
has already been established in a much more general
context; indeed, (14.8) is an immediate consequence of
the remark containing (11.7). We may conclude from
the theorem following (11.9) that in a stable equilib-
rium state the vibrational temperature ev is equal to
the translational temperature e.
Noting that (14.1) and (14.7) imply the partial
differential equation
(14.9)
46 Or, at length, "the specific internal energy of the active
mode."
47 This hypothesis, which differs only slightly from a postulate
studied by Herzfeld and RicelO in 1928, has been discussed by
Rutgers,48 Kneser,9 Landau and Teller,60 Bethe and Teller,a
Rubin and Shuler,61 Montroll and Shuler,"' and others. The studies
of Rubin and Shuler6 ] and Montroll and Shuler62 indicate that
since (14.6) holds precisely for a harmonic-oscillator model, it
should hold near to equilibrium for more general models. See
also Ref. 53.
'8 A. J. Rutgers, Ann. Physik 16, 350 (1933).
'9 H. O. Kneser, Ann. Physik 16,360 (1933).
60 L. Landau and E. Teller, Physik. Z. Sowjetunion 10, 34
(1936).
61 R. J. Rubin and K. E. Shuler, J. Chern. Phys. 25,59 (1956).
62 E. W. Montroll and K. E. Shuler, J. Chern. Phys. 26, 454
(1956).
63 W. G. Vincenti and C. H. Kruger, Introduction to Classical
Gas Dynamics (John Wiley & Sons, Inc., New York, 1965).
Chap. 7.
 THE R MOD Y N A M I C S WIT H
we see that iiA must have the special form
1]A =R Inv+h(EA).
Assuming that h is smoothly invertible, we can write
(14.10) in the form
EA=WA(V exp( -1]A/R)
=WA(V exp{[iiv(a) -1]J/R})
=~(v, 1], a) -a.
It follows from (13.9h and (14.11) that
0= -R-1WA ' (v exp( -1]A/R)v exp( -1]A/R)
with the prime indicating the derivative; hence 0 is a
function of only vexp( -1]A/R). Assuming that this
function is invertible, we may solve (14.12) for
v exp( -1]A/R) in terms of 0 and write (14.11) in the
form
( 14.13)
Thus, the active-mode energy is here a function of the
translational temperature alone.
We now add the natural assumptions that a is
always positive and that the differential equation (14.6)
obeys the stability postulate discussed in Sec. 10.
These assumptions imply that (14.6) can be written
in the form
( 14.14)
a=k(v, 0) (/(v, 0) -a),
with the functions k and I positive. Furthermore, the
function I in (14.14) here plays the role of the equilib-
rium response function a* in (6.17): For each pair
(v*, 0*) the vibrational energy 0'* given by
a*=/(v*, 0*)
makes (v*, 0*, 0'*) an equilibrium state.
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
65.30.29.111 On: Sun, 16 Mar 2014 01:05:33
I K T E R K AL S TAT E V A R I "\ B L E S 613
Let us now assume that for each a and v we can
find a 0 such that a=/(v, 0). It is then a direct conse-
(14.10) quence of the theorem following (11.16) that for each
fixed v the function Ov defined in (14.5) is the inverse
of the function I( v, .). Hence I( v, 0) must be inde-
pendent of v,
I( v, 0) = 1(0), ( 14.16)
and
(14.17)
(14.11)
with the function 1-1 the inverse of I.
Equations (14.3), (14.13), (14.15), and (14.16)
(14.12) imply that the value E* of the specific internal energy
of the gas when it is in equilibrium with temperature
0* and specific volume v* is independent of v*; in fact,
e* = fA (0*) +/(0*). (14.18)54
In view of (14.16), we can write (14.14) in the form
a=k(v, 0)[/(0)-0'], (14.19)
where
k(v, 0) >0, 1(00. (14.20)
ACKNOWLEDGMENTS
We thank J. Kestin, J. Meixner, and B. H. S. Rarity
for stimulating discussions about thermodynamical
systems with internal degrees of freedom. This re-
search was supported in part by the U. S. Air Force
Office of Scientific Research under Grant AF-AFOSR
728-66 and by the Office of Naval Research under
Contract Nonr-51031.
54 Since the equilibrium pressure p* is just R()* /v*, to prove
.*
that is a function of ()* alone we could have used an argument
similar to that giving us (14.13). In this instance the proof would
( 14.15) proceed exactly as it does for an ideal gas without relaxation
effects, in which context the proof is spelled out in detail by R.
Courant and K. O. Friedrichs, Supersonic Flow and Shock Waves
(Interscience Publishers, Inc., New York, 1948), pp. 8, 9.