i=1
n
j=1
T
ij
e
i
e
j
.
Hereafter, we shall always use the summation convention for any pair of repeated in
dices, as shown in the above example, for convenience and clarity. Moreover, although
sometimes the super and subindices are deliberately used to indicate the contravariant
and the covariant components respectively (and sum over a pair of repeated super and
subindices only), in most theoretical discussions, expressions in Cartesian components
suce, for which no distinction of super and subindices is necessary.
The inner product of two tensors A and B is dened as
A B = tr AB
T
,
where the trace and the transpose of a tensor are involved. In particular in terms of
Cartesian components, the norm [A[ is given by
[A[
2
= A A = A
ij
A
ij
,
which is the sum of square of all the elements of A by the summation convention.
The Kronecker delta and the permutation symbol are two frequently used notations,
they are dened as
ij
=
_
0, if i ,= j,
1, if i = j,
2 1. KINEMATICS
and
ijk
=
_
_
1, if i, j, k is an even permutation of 1,2,3,
1, if i, j, k is an odd permutation of 1,2,3,
0, if otherwise.
One can easily check the following identity:
ijk
imn
=
jm
kn
jn
km
.
Let c be a threedimensional Euclidean space and the vector space V be its translation
space. For any two points x, y c there is a unique vector v V associated with their
dierence,
v = y x, or y = x +v.
We may think of v as the geometric vector that starts at the point x and ends at the
point y. The distance between x and y is then given by
d(x, y) = [x y[ = [v[.
Let T be an open region in c and W be any vector space or an Euclidean space.
A function f : T W is said to be dierentiable at x T if there exists a linear
transformation f(x) : V W, such that for any v V ,
f(x +v) f(x) = f(x)[v] + o(v),
where o(v) denotes the higher order terms in v such that
lim
v0
o(v)
[v[
= 0.
We call f the gradient of f with respect to x, and will also denote it by
xx x
f, or more
frequently by either grad f or Grad f.
1.1 Conguration and deformation
A body B can be identied mathematically with a region in a threedimensional Euclidean
space c. Such an identication is called a conguration of the body, in other words, a
onetoone mapping from B into c is called a conguration of B.
It is more convenient to single out a particular conguration of B, say , as a reference,
: B c, (X) = X. (1.1)
We call a reference conguration of B. The coordinates of X, (X
, = 1, 2, 3) are
called the referential coordinates, or sometimes called the material coordinates since the
point X in the reference conguration is often identied with the material point X of
the body when is given and xed. The body B in the conguration will be denoted
by B
.
1. CONFIGURATION AND DEFORMATION 3
`
`
/
/
.
.
.
.
..
r
X


`
`
/
/
.
.
.
.
..
r
X
>
>
`
`


`
`


 
r
x
B
B
=
1
: B
, x =
(X) =
(
1
(X)), (1.2)
is called the deformation of B from to
(Fig. 1.1). In terms of coordinate systems
(x
i
, i = 1, 2, 3) and (X
is given by
x
i
=
i
(X
), (1.3)
where
i
are called the deformation functions.
The deformation gradient of
relative to , denoted by F
is dened by
F
=
XX X
. (1.4)
Here we have assumed that
exists
and is also dierentiable. Therefore the determinant of F
,= 0. (1.5)
When the reference conguration is chosen and understood in the context, F
will be
denoted simply by F.
Relative to the natural bases e
(X) and e
i
(x) of the coordinate systems (X
) and
(x
i
) respectively, the deformation gradient F can be expressed in the following component
form,
F = F
i
e
i
(x) e
(X), F
i
i
X
. (1.6)
Let dX = X X
0
be a small (innitesimal) material line element in the reference
conguration, and dx =
(X)
(X
0
) be its image in the deformed conguration,
then it follows from the denition that
dx = FdX, (1.7)
4 1. KINEMATICS
since dX is innitesimal the higher order term o(dX) tends to zero.
Similarly, let da
and n
da
, dv = [J[ dv
. (1.8)
1.2 Strain and rotation
The deformation gradient is a measure of local deformation of the body. We shall in
troduce other measures of deformation which have more suggestive physical meanings,
such as change of shape and orientation. First we shall recall the following theorem from
linear algebra:
Theorem (polar decomposition). For any nonsingular tensor F, there exist unique
symmetric positive denite tensors V and U and a unique orthogonal tensor R such that
F = RU = VR. (1.9)
Since the deformation gradient F is nonsingular, the above decomposition holds.
We observe that a positive denite symmetric tensor represents a state of pure stretches
along three mutually orthogonal axes and an orthogonal tensor a rotation. Therefore,
(1.9) assures that any local deformation is a combination of a pure stretch and a rotation.
We call R the rotation tensor, while U and V are called the right and the left stretch
tensors respectively. Both stretch tensors measure the local strain, a change of shape,
while the tensor R describes the local rotation, a change of orientation, experienced by
material elements of the body.
Clearly we have
U
2
= F
T
F, V
2
= FF
T
,
det U = det V = [ det F[.
(1.10)
Since V = RUR
T
, V and U have the same eigenvalues and their eigenvectors dier
only by the rotation R. Their eigenvalues are called the principal stretches, and the
corresponding eigenvectors are called the principal directions.
We shall also introduce the right and the left CauchyGreen strain tensors dened by
C = F
T
F, B = FF
T
, (1.11)
respectively, which are easier to be calculated than the strain measures U and V from a
given F in practice. Note that C and U share the same eigenvectors, while the eigenvalues
of U are the positive square root of those of C; the same is true for B and V .
3. LINEAR STRAIN TENSORS 5
1.3 Linear strain tensors
The strain tensors introduced in the previous section are valid for nite deformations in
general. In the classical linear theory, only small deformations are considered.
We introduce the displacement vector from the reference conguration (see Fig. 1.2),
u =
(X) X,
and its gradient,
H =
XX X
u.
Obviously, we have F = 1 + H.



`
`
/
/
.
.
.
.
.
.
.
.
r
X
>
>
>
`
`



`
`





r
x =
(X)
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
..
u(X)
(B)
B
F
T
F = 1 +
1
2
(H + H
T
) + o(2) = 1 + E +o(2),
R = FU
1
= 1 +
1
2
(H H
T
) + o(2) = 1 +
R + o(2),
(1.12)
where
E =
1
2
(H +H
T
),
R =
1
2
(H H
T
), (1.13)
are called the innitesimal strain tensor and the innitesimal rotation tensor, respec
tively. Note that innitesimal strain and rotation are the symmetric and skewsymmetric
parts of the displacement gradient.
We can give geometrical meanings to the components of the innitesimal strain tensor
E
ij
relative to a Cartesian coordinate system. Consider two innitesimal material line
segments dX
1
and dX
2
in the reference conguration and their corresponding ones dx
1
and dx
2
in the current conguration. By (1.7), we have
dx
1
dx
2
dX
1
dX
2
= (F
T
F 1)dX
1
dX
2
= 2E dX
1
dX
2
, (1.14)
6 1. KINEMATICS
for small deformations. Now let dX
1
= dX
2
= s
o
e
1
be a small material line segment in
the direction of the unit base vector e
1
and s be the deformed length. Then we have
s
2
s
2
o
= 2s
2
o
(Ee
1
e
1
) = 2s
2
o
E
11
,
which implies that
E
11
=
s
2
s
2
o
2s
2
o
=
(s s
o
)(s +s
o
)
2s
2
o
s s
o
s
o
.
In other words, E
11
is the change of length per unit original length of a small line
segment in the e
1
direction. The other diagonal components, E
22
and E
33
have similar
interpretations as elongation per unit original length in their respective directions.
Similarly, let dX
1
= s
o
e
1
and dX
2
= s
o
e
2
and denote the angle between the two line
segments after deformation by . Then we have
s
2
o
[Fe
1
[ [Fe
2
[ cos s
2
o
cos
2
= 2s
2
o
(Ee
1
e
2
),
from which, if we write = /2 , the change from its original right angle, then
sin
2
=
E
12
[Fe
1
[ [Fe
2
[
.
Since [E
12
[ 1 and [Fe
i
[ 1, it follows that sin and we conclude that
E
12
2
.
Therefore, the component E
12
is equal to onehalf the change of angle between the two
line segments originally along the e
1
 and e
2
directions. Other odiagonal components,
E
23
and E
13
have similar interpretations as change of angle indicated by their numerical
subscripts.
Moreover, since det F = det(1 + H) 1 + tr H for small deformations, by (1.8)
2
for
a small material volume we have
tr E = tr H
dv dv
dv
.
Thus the sum E
11
+ E
22
+ E
33
measures the innitesimal change of volume per unit
original volume. Therefore, in the linear theory, if the deformation is incompressible,
since tr E = Div u, it follows that
Div u = 0. (1.15)
In terms of Cartesian coordinates, the displacement gradient
u
i
X
j
=
u
i
x
k
x
k
X
j
=
u
i
x
k
(
jk
+o(1)) =
u
i
x
j
+ o(2),
in other words, the two displacement gradients
u
i
X
j
and
u
i
x
j
4. MOTIONS 7
dier in second order terms only. Therefore, since in the classical linear theory, the
higher order terms are insignicant, it is usually not necessary to introduce the reference
conguration in the linear theory. The classical innitesimal strain and rotation, in the
Cartesian coordinate system, are dened as
E
ij
=
1
2
(
u
i
x
j
+
u
j
x
i
),
R
ij
=
1
2
(
u
i
x
j
u
j
x
i
), (1.16)
in the current conguration.
1.4 Motions
A motion of the body B can be regarded as a continuous sequence of deformations in
time, i.e., a motion
of B is regarded as a map,
: B
IR c, x =
(X, t). (1.17)
We denote the conguration of B at time t in the motion
by B
t
.
In practice, the reference conguration is often chosen as the conguration in the
motion at some instant t
0
, =
(, t
0
), say for example, t
0
= 0, so that X =
(X, 0).
But such a choice is not necessary in general. The reference conguration can be chosen
independently of any motion.
For a xed material point X,
(X, ) : IR c
is a curve called the path of the material point X. The velocity v and the acceleration a
are dened as the rst and the second time derivatives of position as X moves along its
path,
v =
(X, t)
t
, a =
2
(X, t)
t
2
. (1.18)
The velocity and the acceleration are vector quantities. Here of course, we have assumed
that
(X, t) is twice dierentiable with respect to t. However, from now on, we shall
assume that all functions are smooth enough for the conditions needed in the context,
without their smoothness explicitly specied.
A material body is endowed with some physical properties whose values may change
along with the deformation of the body in a motion. A quantity dened on a motion
can be described in essentially two dierent ways: either by the evolution of its value
along the path of a material point or by the change of its value at a xed location in
the trajectory of the body. The former is called a referential description and the later a
spatial description. We shall make them more precise below.
For a given motion
and a xed reference conguration , consider a quantity, with
its value in some space W, dened on the motion of B by a function
f : B
IR W. (1.19)
8 1. KINEMATICS
Then it can also be dened on the current conguration at any time t,
f(, t) : B
t
W,
by
f(x, t) =
f(
f =
f(X, t)
t
, Grad f =
XX X
f(X, t), etc.
In the spatial description, the time derivative is the usual
t
and the dierential operators
beginning with lowercase letters, grad, div and curl:
f
t
=
f(x, t)
t
, grad f =
xx x
f(x, t), etc.
The relations between these notations can easily be obtained. Indeed, let f be a scalar
eld and u be a vector eld. We have
f =
f
t
+ (grad f) v, u =
u
t
+ (grad u)v, (1.21)
and
Grad f = F
T
grad f, Grad u = (grad u)F. (1.22)
In particular, taking the velocity v for u, it follows that
grad v =
FF
1
, (1.23)
since Grad v = Grad x =
F.
We call
f the material time derivative of f, which is the time derivative of f following
the path of the material point. Therefore, by the denition (1.18), we can write the
velocity v and the acceleration a as
v = x, a = x,
and hence by (1.21)
2
,
a = v =
v
t
+ (grad v)v. (1.24)
5. RATE OF DEFORMATION 9
1.5 Rate of deformation
Whereas the deformation gradient measures the local deformation, the material time
derivative of deformation gradient measures the rate at which such changes occur at
a xed material point. Another measure for the rate of deformation relative to the
current conguration at a xed location, more commonly used in uid mechanics, will
be introduced here. To do it, we choose the current conguration
(, t) as the reference
conguration so that past and future deformations can be described relative to the present
conguration.
`
`
/
/
.
.
.
.
..
r
X


`
`
/
/
.
.
.
.
..
r
x
>
>
`
`


`
`


 
r
(t)
()
B
B
t
Figure 1.3: Relative deformation
We denote the position of the material point X B
at time by ,
=
(X, ) =
(X).
Then
x =
t
(X), =
(t)
(x, ) =
1
t
(x)), (1.25)
where
(t)
(, ) =
1
t
: B
t
B
(t)
(x, ), (1.26)
that is, the deformation gradient at time with respect to the conguration at time t.
Of course, if = t,
F
t
(x, t) = 1,
and we can easily show that
F(X, ) = F
t
(x, )F(X, t). (1.27)
The rate of change of deformation relative to the current conguration is dened as
L(x, t) =
F
t
(x, )
=t
.
10 1. KINEMATICS
From (1.27), F
t
() = F()F(t)
1
, by taking the derivative with respect to , we have
F
t
() =
F()F(t)
1
= (grad v())F()F(t)
1
= (grad v())F
t
(),
and since F
t
(t) = 1, we conclude that
L = grad v. (1.28)
In other words, the spatial velocity gradient can also be interpreted as the rate of change
of deformation relative to the current conguration.
Moreover, if we apply the polar decomposition to the relative deformation gradient
F
t
(x, ),
F
t
= R
t
U
t
= V
t
R
t
, (1.29)
by holding x and t xed and taking the derivative of F
t
with respect to , we obtain
F
t
() = R
t
()
U
t
() +
R
t
()U
t
(),
and hence by putting = t, we have
L(t) =
U
t
(t) +
R
t
(t). (1.30)
If we denote
D(t) =
U
t
(t), W(t) =
R
t
(t), (1.31)
We can show easily that
D
T
= D, W
T
= W. (1.32)
Therefore, the relation (1.30) is just a decomposition of the tensor L into its symmetric
and skewsymmetric parts, and by (1.28) we have
D =
1
2
(grad v + grad v
T
),
W =
1
2
(grad v grad v
T
).
(1.33)
In view of (1.31) the symmetric part of the velocity gradient, D, is called the rate of strain
tensor or simply as the stretching tensor, and the skewsymmetric part of the velocity
gradient, W, is called the rate of rotation tensor or simply as the spin tensor.
Since the spin tensor W is skewsymmetric, it can be represented as an axial vector w.
The components of the vector w are usually dened by w
i
=
ijk
W
kj
in the Cartesian
coordinate system, hence it follow that
w = curl v. (1.34)
The vector w is called the vorticity vector in uid dynamics.
CHAPTER 2
Balance Laws
2.1 General balance equation
Basic laws of mechanics can all be expressed in general in the following form,
d
dt
_
P
t
dv =
_
P
t
nda +
_
P
t
dv, (2.1)
for any bounded regular subregion of the body, called a part T B and the vector eld
n, the outward unit normal to the boundary of the region T
t
in the current conguration.
The quantities and
is a tensor eld of
order m+1, say m = 0 or m = 1 so that is a scalar or vector quantity, and respectively
the ux of and
the supply of .
We are interested in the local forms of the integral balance (2.1) at a point in the
region T
t
. The derivation of local forms rest upon certain assumptions on the smoothness
of the tensor elds ,
, and
t
_
f(t)
g(t)
(x, t) dx =
_
f(t)
g(t)
t
dx + (f(t), t)
f(t) (g(t), t) g(t).
Theorem (transport theorem). Let V (t) be a regular region in c and u
n
(x, t) be the
outward normal speed of a surface point x V (t). Then for any smooth tensor eld
(x, t), we have
d
dt
_
V
dv =
_
V
t
dv +
_
V
u
n
da. (2.2)
In this theorem, the surface speed u
n
(x, t) needs only to be dened on the boundary
V . If V (t) is a material region T
t
, i.e., it always consists of the same material points of
12 2. BALANCE LAWS
a part T B, then u
n
= x n and (2.2) becomes
d
dt
_
P
t
dv =
_
P
t
t
dv +
_
P
t
x nda. (2.3)
Now we shall extend the above transport theorem to a material region containing a
surface across which may suer a jump discontinuity.
Denition. An oriented smooth surface o in a material region 1 is called a singular
surface relative to a eld dened on 1, if is smooth in 1 o and suers a jump
discontinuity across o. The jump of is dened as
[[]] =
+
, (2.4)
where
+
and
are the one side limits from the two regions of 1 separated by o and
designated as 1
+
and 1
respectively.





`
`
`
\
\
/
/
.
.
. 

 .
.
.
.
.
.
. 


\
\
/
/
/
Z
Z
.
.
.
.
.
.
.
/
/
'
n
o
u
n
1
1
+
(V)
+
(V)
= 1
1.
Since both (1)
+
and (1)
, 1 = (1)
+
(1)
. (2.5)
Since o need not be a material surface, 1
+
and 1
t
dv +
_
(V)
+
x nda +
_
S
+
(u
n
) da, (2.6)
d
dt
_
V
dv =
_
V
t
dv +
_
(V)
x nda +
_
S
u
n
da. (2.7)
Adding (2.6) and (2.7), we obtain by the use of (2.5) the following transport theorem in
a material region containing a singular surface:
2. LOCAL BALANCE EQUATION AND JUMP CONDITION 13
Theorem. Let V (t) be a material region in c and o(t) be a singular surface relative to
the tensor eld (x, t) which is smooth elsewhere. Then we have
d
dt
_
V
dv =
_
V
t
dv +
_
V
x nda
_
S
[[]] u
n
da, (2.8)
where u
n
(x, t) is the normal speed of a surface point x o(t) and [[]] is the jump of
across o.
2.2 Local balance equation and jump condition
For a material region 1 containing a singular surface o, the equation of general balance
in integral form (2.1) becomes
_
V
t
dv +
_
V
x nda
_
S
[[]] u
n
da =
_
V
nda +
_
V
dv. (2.9)
Denition. A point x is called regular if there is a material region containing x in which
all the tensor elds in (2.1) are smooth. And a point x is called singular if it is a point
on a singular surface relative to and
.
We can obtain the local balance equation at a regular point as well as at a singular
point from the above integral equation. First, we consider a small material region 1
containing x such that 1 o = . By the use of the divergence theorem, (2.9) becomes
_
V
_
t
+ div( x
_
dv = 0. (2.10)
Since the integrand is smooth and the equation (2.10) holds for any 1, such that x 1,
and 1 o = , the integrand must vanish at x. Therefore we have
Theorem (local balance equation). At a regular point x, the general balance equation
(2.9) reduces to
t
+ div( x
= 0. (2.11)
The notation div( x) in (2.11) should be understood as div( x) when is a scalar
quantity.
Next, we consider a singular point x, i.e., x o. Let 1 be an arbitrary material
region around x, and s = 1 o (Fig. 2.2). We shall take the limit by shrinking (1)
+
and (1)
down to s in such a way that the volume of 1 tends to zero, while the area
of s remain unchanged. If
t
and
n]] [[]] u
n
da = 0. (2.12)
Since the integrand is smooth on s and (2.12) holds for any s containing x, the integrand
must vanish at x. We obtain
14 2. BALANCE LAWS
r
'
~
~
~
~
~


  .
.
. .
.
.
.
.
.
x
>
>.
>
>
V
o
s
n
(V)
+
(V)
n]] [[]] u
n
= 0, (2.13)
if in addition,
t
and
]] n = 0, (2.14)
where
U
= u
n
x
n (2.15)
are called the local speeds of propagation of o relative to the motion of the body. If o is
a material surface then U
= 0.
2.3 Balance equations in reference coordinates
Sometimes, for solid bodies, it is more convenient to use the referential description. The
corresponding relations for the balance equation (2.11) and the jump condition (2.13) in
the reference coordinates can be derived in a similar manner. We begin with the integral
form (2.1) now written in the reference conguration ,
d
dt
_
P
dv
=
_
P
da
+
_
P
dv
. (2.16)
In view of the relations for volume elements and surface elements (1.8), the corresponding
quantities are dened as
= [J[ ,
= J
F
T
,
= [J[
. (2.17)
The transport theorem (2.2) remains valid for
dv
=
_
V
dv
+
_
V
da
, (2.18)
4. CONSERVATION OF MASS 15
where U
(X, t) is the outward normal speed of a surface point X V (t). Therefore for
singular surface o
dv
_
S
[[
]] U
da
=
_
V
da
+
_
V
dv
, (2.19)
where U
Div
= 0, [[
]] U
+ [[
]] n
= 0, (2.20)
where U
and n
are the normal speed and the unit normal vector of the singular surface.
2.4 Conservation of mass
Let (x, t) denote the mass density of B
t
in the current conguration. Since the material
is neither destroyed nor created in any motion in the absence of chemical reactions, we
have
Conservation of mass. The total mass of any part T B does not change in any
motion,
d
dt
_
P
t
dv = 0. (2.21)
By comparison, it is a special case of the general balance equation (2.1) with no ux
and no supply,
= ,
= 0,
= 0,
and hence from (2.11) and (2.13) we obtain the local expressions for mass conservation
and the jump condition at a singular point,
t
+ div( x) = 0, [[ ( x n u
n
)]] = 0. (2.22)
The equation (2.21) states that the total mass of any part is constant in time. In
particular, if
dv
=
_
P
t
dv, (2.23)
which implies that
= [J[, (2.24)
where J = det F. This is another form of the conservation of mass in the referential
description, which also follows from the general expression (2.20) and (2.17).
We note that if the motion is volumepreserving, then
d
dt
_
P
t
dv = 0,
16 2. BALANCE LAWS
which by comparison with (2.1) again leads to
div x = 0,
which in turn implies that = 0 by (2.22)
1
. Therefore, for an incompressible motion,
the mass density is timeindependent and the divergence of velocity eld must vanish.
2.5 Equation of motion
For a deformable body, the linear momentum and the angular momentum with respect
to a point x
) xdv.
In laying down the laws of motion, we follow the classical approach developed by
Newton and Euler, according to which the change of momentum is produced by the
action of forces. There are two type of forces, namely, one acts throughout the volume,
called the body force, and one acts on the surface of the body, called the surface traction.
Eulers laws of motion. Relative to an inertial frame, the motion of any part T B
satises
d
dt
_
P
t
xdv =
_
P
t
b dv +
_
P
t
t da,
d
dt
_
P
t
(x x
) xdv =
_
P
t
(x x
) b dv +
_
P
t
(x x
) t da.
We remark that the existence of inertial frames (an equivalent of Newtons rst law)
is essential to establish the Eulers laws (equivalent of Newtons second law) in the above
forms. Roughly speaking, a coordinate system (x) at rest for c is usually regarded as an
inertial frame. Eulers laws in an arbitrary frame can be obtained through the notion of
change of reference frames considered in Chapter 3.
We call b the body force density (per unit mass), and t the surface traction (per unit
surface area). Unlike the body force b = b(x, t), such as the gravitational force, the
traction t at x depends, in general, upon the surface T
t
on which x lies. It is obvious
that there are innite many parts T B, such that T
t
may also contain x. However,
following Cauchy, it is always assumed in the classical continuum mechanics that the
tractions on all likeoriented surfaces with a common tangent plane at x are the same.
Postulate (Cauchy). Let n be the unit normal to the surface T
t
at x, then
t = t(x, t, n). (2.25)
An immediate consequence of this postulate is the wellknown theorem which ensures
the existence of stress tensor.
6. CONSERVATION OF ENERGY 17
Theorem (Cauchy). Suppose that t(, n) is a continuous function of x, and x, b are
bounded in B
t
. Then Cauchys postulate and Eulers rst law implies the existence of a
second order tensor T, such that
t(x, t, n) = T(x, t) n. (2.26)
The tensor eld T(x, t) in (2.26) is called the Cauchy stress tensor. The proof of the
theorem can be found in most books of mechanics. The continuity assumption in the
theorem is often too stringent for the existence of stress tensor in some applications, for
example, problems involving shock waves. However, it has been shown that the theorem
remains valid under a much weaker assumption of integrability, which would be satised
in most applications [7, 6].
With (2.26) Eulers rst law becomes
d
dt
_
P
t
xdv =
_
P
t
b dv +
_
P
t
Tnda. (2.27)
Comparison with the general balance equation (2.1) leads to
= x,
= T,
= b,
in this case, and hence from (2.11) and (2.13) we obtain the balance equation of linear
momentum and its jump condition,
t
( x) + div( x x T) b = 0, [[ x( x n u
n
)]] [[T]]n = 0. (2.28)
The rst equation, also known as the equation of motion, can be rewritten in the following
more familiar form by the use of (2.22),
x div T = b. (2.29)
A similar argument for Eulers second law as a special case of (2.1) with
= (x x
) x,
n = (x x
) Tn,
= (x x
) b,
leads to
T = T
T
, (2.30)
after some simplication by the use of (2.29). In other words, the symmetry of the stress
tensor is a consequence of the conservation of angular momentum.
2.6 Conservation of energy
Besides the kinetic energy, the total energy of a deformable body consists of another part
called the internal energy,
_
P
t
_
+
2
x x
_
dv,
where (x, t) is called the specic internal energy density. The rate of change of the total
energy is partly due to the mechanical power from the forces acting on the body and
partly due to the energy inow over the surface and the external energy supply.
18 2. BALANCE LAWS
Conservation of energy. Relative to an inertial frame, the change of energy for any
part T B is given by
d
dt
_
P
t
( +
2
x x) dv =
_
P
t
( x Tn q n) da +
_
P
t
( x b + r) dv. (2.31)
We call q(x, t) the heat ux vector (or energy ux), and r(x, t) the energy supply
density due to external sources, such as radiation. Comparison with the general balance
equation (2.1), we have
= ( +
2
x x),
= T x q,
= ( x b +r),
and hence we have the following local balance equation of total energy and its jump
condition,
t
_
+
2
x x
_
+ div
_
( +
2
x x) x +q T x
_
(r + x b) = 0,
[[( +
2
x x)( x n u
n
)]] + [[q T x]] n = 0.
(2.32)
The energy equation (2.32)
1
can be simplied by substracting the inner product of the
equation of motion (2.29) with the velocity x,
+ div q = T grad x + r. (2.33)
This is called the balance equation of internal energy. Note that the internal energy is
not conserved and the term T grad x is the rate of work due to deformation.
Summary of basic equations:
By the use of material time derivative (1.21), the eld equations can also be written as
follows:
+ div v = 0,
v div T = b,
+ div q T grad v = r,
(2.34)
and the jump conditions in terms of local speed U of the singular surface become
[[ U]] = 0,
[[ Uv]] + [[T]]n = 0,
[[ U( +
1
2
v
2
)]] + [[Tv q]] n = 0,
(2.35)
where v
2
stands for v v.
7. BASIC EQUATIONS IN MATERIAL COORDINATES 19
2.7 Basic Equations in Material Coordinates
It is sometimes more convenient to rewrite the basic equations in material description
relative to a reference conguration . They can easily be obtained from (2.20)
1
,
= [J[
1
x = Div T
b,
+ Div q
= T
F +
r,
(2.36)
where the following denitions have been introduced according to (2.17):
T
= J TF
T
, q
= JF
1
q. (2.37)
T
is not
symmetric and it must satisfy
T
F
T
= F T
T
. (2.38)
The denition has been introduced according to the relation (1.8), which gives the rela
tion,
_
S
Tnda =
_
S
da
. (2.39)
In other words, Tn is the surface traction per unit area in the current conguration,
while T
is the surface traction measured per unit area in the reference conguration.
Similarly, q n and q
are the contact heat supplies per unit area in the current and
the reference congurations respectively.
We can also write the jump conditions in the reference conguration by the use of
(2.20)
2
,
[[
]] U
= 0,
[[
x]] U
+ [[T
]] n
= 0,
[[
( +
1
2
x x)]] U
+ [[T
T
x q
]] n
= 0,
(2.40)
where n
and U
are the unit normal and the normal speed at the singular surface in
the reference conguration.
2.8 Boundary conditions of a material body
The boundary of a material body is a material surface. Therefore, at the boundary
U
= f
, (2.43)
and if the boundary is adiabatic,
q
= 0, (2.44)
where f
and q
are the force and the heat ux per unit area acting on the xed boundary
B
respectively.
The relations (2.41) and (2.42) are the usual boundary conditions in elasticity and
heat conduction.
CHAPTER 3
Constitutive Relations
The balance laws introduced in Chapter 2 are the fundamental equations which are com
mon to all material bodies. However, these laws are not sucient to fully characterize the
behavior of material bodies, because physical experiences have shown that two samples of
exactly the same size and shape in general will not have the same behavior when they are
subjected to exactly the same experiment (external supplies and boundary conditions).
Mathematically, the physical properties of a body can be given by a description of
constitutive relations. These relations characterize the material properties of the body. It
is usually assumed that such properties are intrinsic to the material body and, therefore,
must be independent of external supplies as well as of dierent observers that happen to
be measuring the material response simultaneously.
By an observer we mean a frame of reference for the event world and therefore is
able to measure the position in the Euclidean space and time on the real line. The
conguration of a body introduced so far should have been dened relative to a given
frame of reference, but since only a given frame is involved, the concept of frame has
been set aside until now when a change of frame of reference becomes essential in later
discussions.
3.1 Frame of reference, observer
The spacetime J is a fourdimensional space in which events occur at some places and
certain instants. Let J
t
be the space of placement of events at the instant t, then in
Newtonian spacetime of classical mechanics, we can write
J =
_
tIR
J
t
,
and regard J as a product space c IR, where c is a three dimensional Euclidean space
and IR is the space of real numbers, through a oneto one mapping, called a frame of
reference (see [21]),
: J c IR,
such that it gives rise to a onetoone mapping for the space of placement into the
Euclidean space,
t
: J
t
c. (3.1)
The frame of reference can usually be referred to as an observer, since it can be
depicted as taking a snapshot of events at some instant t with a camera, so that the image
22 3. CONSTITUTIVE RELATIONS
of
t
is a picture (threedimensional conceptually) of the placement of events, from which
the distance between dierent simultaneous events can be measured physically, with a
ruler for example. Similarly, an observer can record a sequence of snapshots at dierent
instants with a video camera for the change of events in time.
Motion of a body
The motion of a body can be viewed as a sequence of events such that at any instant t,
the placement of the body B in J
t
is a onetoone mapping
t
: B J
t
,
which relative to a frame of reference
t
: J
t
c can be described as
t
: B c,
t
:=
t
t
, x =
t
(X) X B,
which identies the body with a region in the Euclidean space. We call
t
a conguration
of B at the instant t in the frame , and a motion of B is a sequence of congurations of
B in time,
t
, t IR [
t
: B c.
Reference conguration
As a primitive concept, a body B is considered to be a set of material points without
additional mathematical structures. Although it is possible to endow the body as a
manifold with a dierentiable structure and topology,
1
for doing mathematics on the
body, usually a reference conguration, say
: B c, X =
(X), B
:=
(B),
is chosen, so that the motion can be treated as a function dened on the Euclidean
space c,
(, t) : B
c c, x =
(X, t) =
t
(
1
(X)) X B
.
For examples, in previous chapters, we have already introduced
x(X, t) =
(X, t)
t
, x(X, t) =
2
(X, t)
t
2
,
and
F
(X, t) =
XX X
(X, t),
called the velocity, the acceleration, and the deformation gradient of the motion, respec
tively. Note that no explicit reference to
1
,
: c IR c IR, (x, t) = (x
, t
),
a change of frame (observer), where (x, t) and (x
, t
= Q(t)(x x
0
) +c(t)
t
= t + a,
(3.2)
for some x
0
, c(t) c, Q(t) O(V ), a IR, where O(V ) is the group of orthogonal
transformations on the translation space V of the Euclidean space c. In other words, a
change of frame is an isometry of space and time as well as preserves the sense of time.
Such a transformation will be called a Euclidean transformation.
In particular,
t
1
t
: c c is given by
x
t
(
1
t
(x)) = Q(t)(x x
o
) +c(t), (3.3)
which is a timedependent rigid transformation consisting of an orthogonal transforma
tion and a translation. We shall often call Q(t) as the orthogonal part of the change of
frame from to
.
Objective tensors
The change of frame (3.2) on the Euclidean space c gives rise to a linear mapping on
the translation space V , in the following way: Let u() = x
2
x
1
V be the dierence
vector of x
1
, x
2
c in the frame , and u(
) = x
2
x
1
V be the corresponding
dierence vector in the frame
) = Q(t)u().
24 3. CONSTITUTIVE RELATIONS
Any vector quantity in V , which has this transformation property, is said to be objective
with respect to Euclidean transformations. The Euclidean objectivity can be generalized
to any tensor spaces of V . In particular, we have the following denition.
Denition. Let s, u, and T be scalar, vector, (second order) tensorvalued functions
respectively. If relative to a change of frame from to
) = s(),
u(
) = Q(t) u(),
T(
= f(
).
The denition of objective scalar is selfevident, while for objective tensors, the
denition becomes obvious if one considers a simple tensor T = u v, by dening
(u v)() = u() v() and hence
(u v)(
) = u(
) v(
(x
, t
xx x
(x, t) = Q(t)
T
xx x
(x
, t
) or (grad )(
(x
, t
) = f
(x
(x, t), t
(t)) = f(x, t) = 0.
2. CHANGE OF FRAME AND OBJECTIVE TENSORS 25
By taking gradient with respect to x, we obtain, from (3.4) that
Q
T
xx x
f
=
xx x
f,
from which, since Q is orthogonal, [Q
T
xx x
f
[ = [
xx x
f
[, it follows that
n
= Qn. (3.5)
.
Transformation properties of motion
Let
,
x =
t
(X) =
(X, t), x
t
(X) =
(X, t
), X B.
Then from (3.3), we have
(X, t
) = Q(t)(
(X, t) x
o
) +c(t), X B. (3.6)
Consequently, the transformation property of a kinematic quantity, associated with the
motion in c, can be derived from (3.6). In particular, one can easily show that the
velocity and the acceleration are not objective quantities. Indeed, it follows from (3.6)
that
x
= Q x +
Q(x x
o
) + c,
x
= Q x + 2
Q x +
Q(x x
0
) + c.
(3.7)
The relation (3.7)
2
shows that x is not objective but it also shows that x is an objective
tensor with respect to transformations for which Q(t) is a constant tensor. A change of
frame with constant Q(t) and c(t) linear in time t is called a Galilean transformation.
Therefore, we conclude that the acceleration is a Galilean objective tensor quantity but
not a Euclidean objective one. However, from (3.7)
1
, it shows that the velocity is neither
a Euclidean nor a Galilean objective vector quantity.
Transformation properties of deformation
Let : B J
t
0
be a reference placement of the body at some instant t
0
, then
=
t
0
and
t
0
(3.8)
are the two corresponding reference congurations of B in the frames and
at the
same instant (see Fig. 3.1), and
X =
(X), X
(X), X B.
Let us denote by =
to
in
the change of frame, then it follows from (3.8) that =
t
0
1
t
0
and by (3.3), we have
X
= (X) = K(X x
o
) +c(t
0
), (3.9)
26 3. CONSTITUTIVE RELATIONS
J
t
0
B
B
c B
.
`
`
`
`
`
`
`
`
`
`
t
0
t
0
Figure 3.1: Reference congurations
and
where K = Q(t
0
) is a constant orthogonal tensor.
On the other hand, the motion in referential description relative to the change of
frame is given by
x =
(X, t), x
(X
, t
),
and from (3.3) we have
(X
, t
) = Q(t)(
(X, t) x
o
) +c(t).
Therefore we obtain for the deformation gradient in the frame
, i.e., F
=
XX X
, by
the chain rule,
F
(X
, t
) = Q(t)F(X, t) K
T
,
or simply,
F
= QFK
T
, (3.10)
where, by (3.9), K = Q(t
0
), is a constant orthogonal tensor due to the change of frame
for the reference conguration.
2
The deformation gradient F is not a Euclidean objective tensor. However, the prop
erty (3.10) also shows that it is frameindierent with respect to Galilean transformations,
since in this case, K = Q is a constant orthogonal transformation.
Now we consider transformation properties of some other kinematic quantities related
to the deformation gradient. With polar decompositions of F and F
, (3.10) gives
R
= QRUK
T
, V
= QVRK
T
.
By the uniqueness of polar decompositions, we conclude that
U
= KUK
T
, V
= QV Q
T
, R
= QRK
T
, (3.11)
and also
C
= KCK
T
, B
= QBQ
T
. (3.12)
Therefore, R, U, and C are not objective tensors, but the tensors V and B are objective.
2
The transformation property (3.10) stands in contrast to F
= Q
FK
T
+
QFK
T
.
With L = grad x =
F F
1
by (1.23), we have
L
= QLFK
T
+
QFK
T
= QLQ
T
F
+
QQ
T
F
,
and since F
is nonsingular, it gives
L
= QLQ
T
+
QQ
T
. (3.13)
Moreover, with L = D + W, it becomes
D
+ W
= Q(D + W)Q
T
+
QQ
T
.
By separating symmetric and skewsymmetric parts, we obtain
D
= QDQ
T
, W
= QWQ
T
+
QQ
T
, (3.14)
since
QQ
T
is skewsymmetric. Therefore, while the velocity gradient grad x and the spin
tensor W are not objective, the rate of strain tensor D is an objective tensor.
Objective time derivatives
Let and u be objective scalar and vector elds respectively,
(X
, t
) = (X, t), u
(X
, t
) = Q(t)u(X, t).
It follows that
=
, u
= Q u +
Qu.
Therefore, the material time derivative of an objective scalar eld is objective, while the
material time derivative of an objective vector eld is not longer objective.
Nevertheless, we can dene an objective time derivatives of vector eld in the following
way: Let P
t
() : V V be a transformation which takes a vector at time t to a vector
at time and dene the time derivative as
D
P
t
u(t) = lim
h0
1
h
(u(t + h) P
t
(t + h)u(t)), (3.15)
so that the vector u(t) is transformed to a vector at (t + h) through P
t
(t + h) in order
to compare with the vector u(t + h) at the same instant of time. This is usually called
a Lie derivative.
Now, we shall take P
t
() to be the relative rotation tensor of the motion R
t
() (see
(1.29)) and denote
u= D
R
t
u.
Since
R
t
(t +h) = R
t
(t) +
R
t
(t)h + o(2)
28 3. CONSTITUTIVE RELATIONS
and from (1.31)
R
t
(t) = 1,
R
t
(t) = W(t),
we have
u(t) = lim
h0
1
h
(u(t + h) u(t) hW(t)u(t)),
or
u= u Wu. (3.16)
Note that by (3.14)
2
,
(
u)
= u
= Q u +
Qu (QWQ
T
+
QQ
T
)Qu
= Q( u Wu) = Q
u .
Therefore,
u is an objective time derivative of u. This derivative is called the corotational
time derivative, which measures the time rate of change experienced by material particles
rotating along with the motion.
Similarly, if we take P
t
() to be the relative deformation gradient of the motion F
t
()
(see (1.29)) and denote
u = D
F
t
u.
It is an called the convected time derivatives, which measures the time rate of change
experienced by material particles carried along with the motion. It follows that
u = u Lu,
which is also an objective time derivative.
For an objective tensor eld S, we can dene
D
P
t
S(t) = lim
h0
1
h
(S(t + h) P
t
(t + h)S(t)P
t
(t + h)
T
), (3.17)
and let
S= D
R
t
S,
S= D
F
t
S,
which dene the objective corotational and convective time derivative of an objective
tensor eld S as
S =
S WS + SW,
S =
S LS SL
T
,
respectively.
3. CONSTITUTIVE RELATIONS 29
3.3 Constitutive relations
Physically the thermomechanical behavior of a body is characterized by a description
of the elds of density (X, t), motion
(X, t) and temperature (X, t), called the basic
thermodynamic elds. The material response of a body generally depends on the past
history of its thermomechanical behavior.
Let us introduce the notion of the past history of a function. Let h() be a function
of time. The history of h up to time t is dened by
h
t
(s) = h(t s), (3.18)
where s [0, ) denotes the timecoordinate pointed into the past from the present
time t. Clearly s = 0 corresponds to the present time, therefore
h
t
(0) = h(t).
Mathematical descriptions of material response are called constitutive relations or con
stitutive equations. We postulate that the history of the behavior up to the present time
determines the response of the body.
Principle of determinism (in material description). Let be a frame of reference, and
( be a constitutive quantity, then the constitutive relation for ( is given by a functional
of the form,
3
((X, t) = T
Y B
0s<
(
t
(Y, s),
t
(Y, s),
t
(Y, s), t ; X), X B, t IR. (3.19)
We call T
(
t
,
t
,
t
, t ; X),
where the rst three arguments are functions:
t
: B [0, ) IR,
t
: B [0, ) c,
t
: B [0, ) IR.
30 3. CONSTITUTIVE RELATIONS
formulation of a mathematical model in general, we shall rst impose these requirements
on the proposed model. The most important universal requirements of this kind are
principle of material frameindierence,
material symmetry,
thermodynamic considerations.
The principle of material frameindierence and material symmetry will be discussed in
this chapter and a brief account of thermodynamic considerations will be given later.
These requirements impose severe restrictions on the model and hence lead to a great
simplication for general constitutive relations. The reduction of constitutive relations
from very general to more specic and mathematically simpler ones for a given class of
materials is the main objective of constitutive theories in continuum mechanics.
3.4 Euclidean objectivity
In addition to the conservation laws in an inertial frame stated in the previous chapter,
we should have also postulated that the mass density, the internal energy (but not the
total energy) and the temperature are objective scalar quantities, while the heat ux and
the surface traction are objective vector quantities, so that the conservation laws relative
to an arbitrary frame of reference can be deduced (see [14]).
In particular, under a change of frame, we have t
(X, t
(X, t
(X, t
) = (X, t),
(3.20)
where Q(t) is the orthogonal part of the change of frame.
Note that this postulate is a universal requirement in order to formulate the conservation
laws of material bodies in an arbitrary frame of reference, and no constitutive assumptions
are involved in the postulate of Euclidean objectivity.
4
Let ( = T, q, be constitutive quantities and be a frame of reference. Then from
(3.19), the constitutive relations can be written as
((X, t) = T
(
t
,
t
,
t
, t ; X), X B, t IR, (3.21)
4
In early writings of Truesdell and Noll (see [22]), this postulate was stated as part of the principle
of material frameindierence. It would be much clear conceptually, if it were stated independently,
because constitutive equations are irrelevant in this assumption.
5. PRINCIPLE OF MATERIAL FRAMEINDIFFERENCE 31
where T
t
,
t
, dened on B[0, ), stand for
the past histories up to the instant t. Similarly, relative to the frame
, the corresponding
constitutive relation can be written as
(
(X, t
) = T
_
(
t
)
, (
t
)
, (
t
)
, t
; X
_
, X B, t
IR. (3.22)
The two constitutive functions T
and T
((
t
)
, (
t
)
, (
t
)
, t
; X) = Q(t) T
(
t
,
t
,
t
, t ; X) Q(t)
T
, X B, (3.23)
for any histories
(
t
)
(Y, s) =
(Y, t
s) = (Y, t s),
(
t
)
(Y, s) =
(Y, t
s) = (Y, t s),
(
t
)
t
(Y, s) x
o
) +c(t s),
Y B, (3.24)
where Q(t) O(V ), x
o
, c(t) c are associated with the change of frame from to
.
The rst two relations of (3.24) state that the density and the temperature are ob
jective scalar eld as assumed and the last relation follows from (3.3).
The relation (3.23) will be referred to as the condition of Euclidean objectivity
of constitutive functions, and can be regarded as the denition of the T
once the
constitutive function T
is given
5
.
3.5 Principle of material frameindierence
The essential idea of the principle of material frameindierence is that material proper
ties are frameindierent, i.e., material properties must be independent of observers. This
statement only makes sense when applies to constitutive quantities that are themselves
frameindierent, i.e., they must be objective quantities.
Since any intrinsic property of a material should be independent of observers, we
postulate that for any objective constitutive quantity, its constitutive function must be
the same in any frame.
Principle of material frameindierence. The response function of an objective
(frameindierent with respect to the Euclidean transformations) constitutive quantity (,
in material description dened by (3.21) and (3.22), must be independent of the frame,
i.e., for any frames of reference and
( ; X) = T
( ; X), X B. (3.25)
5
For example, consider T = T
= , T
= QTQ
T
, and from
(3.13), L
= QLQ
T
+
QQ
T
, it follows that
T
= QT
(, L) Q
T
= QT
, Q
T
L
QQ
T
Q) Q
T
:=
T
, L
; Q,
Q) := T
, L
),
where T
, dened in the above relation, depends explicitly on Q(t) of the change of frame
.
32 3. CONSTITUTIVE RELATIONS
We emphasize that the forminvariance (3.25) of constitutive function stated above
is valid only when the material description is used so that the constitutive functions
characterize the material properties independent of any reference conguration. The
implication in referential descriptions will be considered later. Moreover, for a non
objective constitutive quantity, such as the total energy, since the quantity itself is frame
dependent, it is obvious that its response function can never be independent of the frame.
In our discussions, we shall often consider only the constitutive function of stress, an
objective tensor quantity, for simplicity. Similar results can easily be obtained for any
other vector or scalar objective constitutive quantities.
Thus, from the condition of Euclidean objectivity (3.23) and the principle of material
frameindierence (3.25), we obtain the following condition:
Condition of material objectivity. The response function of an objective tensor
constitutive quantity, in material description, satises the condition,
T
((
t
)
, (
t
)
, (
t
)
, t
, X) = Q(t) T
(
t
,
t
,
t
, t ; X) Q(t)
T
, (3.26)
for any histories related by (3.24) where the change of frame from is arbitrary.
Since the condition (3.26) involves only the constitutive function in the frame , it
becomes a restriction imposed on the constitutive function T
.
We emphasize that in the condition of Euclidean objectivity (3.23), Q(t) is the or
thogonal part of the change of frame from to
= x, t
= t + a,
for arbitrary constant a IR. By (3.24), the condition (3.26) implies that
T
(
t
,
t
,
t
, t + a ; X) = T
(
t
,
t
,
t
, t ; X).
Since this is true for any value of a IR, we conclude that T
(
t
,
t
,
t
; X), X B, t IR.
5. PRINCIPLE OF MATERIAL FRAMEINDIFFERENCE 33
Conditions in referential description
Now let us consider the implication of forminvariance and the condition of material
objectivity in referential description. Let : B J
t
0
be a reference placement of the
body at some instant t
0
, then
=
t
0
: B c and
t
0
: B c are the two
corresponding reference congurations of B in the frames and
(X) c, X
(X) c, X B.
Let us denote by =
to
in
the change of frame, then it follows that =
t
0
1
t
0
and by (3.3), we have
X
= (X) = K(X x
o
) +c(t
0
), (3.27)
where K =
XX X
= Q(t
0
) is a constant orthogonal tensor.
The motion in referential description relative to the change of frame is given by
x =
(X, t) =
(
1
(X), t) =
(X, t),
=
,
x
(X, t
) =
1
(X
), t
) =
(X
, t
),
.
From (3.21) and (3.22), we can dene the corresponding constitutive functions with
respect to the reference conguration,
T
t
; X) = T
; X) := H
; X),
T
((
t
)
; X) = T
((
; X) := H
((
; X
),
where for simplicity, only one argument function
t
is written out symbolically.
Since the forminvariance (3.25) requires T
= T
and H
((
t
)
; X
) = T
((
; X) = T
((
; X)
= T
((
; X) = H
((
t
)
; X),
where =
t
0
1
t
0
. Therefore, they are not forminvariant in general, i.e.,
H
,= H
( ; X
) = H
( ; X), (3.28)
The Euclidean objectivity relation (3.20) can now be written in the form,
H
((
; X
) = Q(t) H
; X) Q(t)
T
. (3.29)
Moreover, by combining (3.28) and (3.29), we obtain the condition of material objectivity
in referential description,
H
((
; X) = Q(t) H
; X) Q(t)
T
, Q(t) O(V ). (3.30)
6
If the reference conguration is assumed to be unaected by the change of frame, then is an
identity map on c, and H is forminvariant, H
= H
0s<
(
(X, s), F
t
(X, s),
t
(X, s), g
t
(X, s); X), X B
, (3.31)
where g = grad is the spatial gradient of temperature. Note that since
(X, t) =
(X)
[det F(X, t)[
,
the functional dependence of (3.19) on
t
is absorbed into the dependence of (3.31) on X
and F
t
(X, s).
A material body with constitutive relation dened by (3.31) is called a simple material
body. The class of simple materials, introduced by Noll [18], is general enough to include
most of the materials of practical interests, such as: the elastic solids, thermoelastic
solids, viscoelastic solids as well as elastic uids, NavierStokes uids and viscous heat
conducting uids.
An immediate consequence of the condition of material objectivity (3.30) can be
obtained by the following choices of change of frame. Consider a change of frame given
by (Q(t) = 1, a = 0)
x
= x +c(t) x
0
, t
= t.
Clearly, we have
((X), t) =
(X, t) + (c(t) x
0
),
and the condition (3.30) implies that
H
+c
t
x
0
, F
t
, X) = H
, F
t
, X).
Since (c
t
(s) x
0
) V is arbitrary, we conclude that H
(X, s).
Therefore, the constitutive relation of a simple material body in general is given by
((X, t) = H
0s<
(F
t
(X, s),
t
(X, s), g
t
(X, s); X). (3.32)
It is important to point out that constitutive functions can not depend on the position
x nor on the velocity x, and indeed this can be proved for any materials in general as a
consequence of the principle of material objectivity (3.26).
7. MATERIAL SYMMETRY 35
For simple materials, the consequence of forminvariance (3.28) by the use of (3.9)
takes the form
H
((F
t
)
; X
) = H
((F
t
)
K; X), (3.33)
and the Euclidean objectivity condition (3.29) becomes
H
((F
t
)
; X
) = Q(t) H
(F
t
; X) Q(t)
T
.
Combining the above two conditions and knowing the relation F
K = (QFK
T
)K = QF,
we obtain the following condition,
7
H
(Q
t
F
t
; X) = Q(t) H
(F
t
; X) Q(t)
T
. (3.34)
More generally, we have the following conditions for objective constitutive functions
of simple materials:
Condition of material objectivity. Constitutive functions for simple materials must
satisfy the following conditions, for objective tensor, vector and scalar constitutive quan
tities respectively,
T (Q
t
F
t
,
t
, Q
t
g
t
; X) = QT (F
t
,
t
, g
t
; X) Q
T
,
Q(Q
t
F
t
,
t
, Q
t
g
t
; X) = QQ(F
t
,
t
, g
t
; X),
c (Q
t
F
t
,
t
, Q
t
g
t
; X) = c (F
t
,
t
, g
t
; X),
Q O(V ). (3.35)
These conditions are the restriction imposed on the constitutive functions by the require
ment of material objectivity.
Note that in condition (3.35), no mention of change of frame is involved, and Q(t)
can be interpreted as a superimposed orthogonal transformation on the deformation.
This interpretation is sometimes viewed as an alternative version of the principle of
material objectivity, and is called the principle of invariance under superimposed rigid
body motions.
3.7 Material symmetry
We shall consider homogeneous simple material bodies in this section. A body is called
homogeneous if there is a reference conguration such that the constitutive function
(3.32) does not depend on the argument X explicitly and we can write
((x, t) = H
0s<
(F
t
(X, s),
t
(X, s), g
t
(X, s)), x =
and H
,
H
(
F
t
,
t
, g
t
) = H
(
F
t
G,
t
, g
t
). (3.38)
A material body subjected to the same thermomechanical history at two dierent
congurations and may have dierent results. However, it may happen that the
results are the same if the material possesses a certain symmetry. For example, one can
not distinguish the response of a material body with a cubic crystal structure before and
after a rotation of 90
( ) = H
( ).
By (3.38), the above condition is equivalent to
H
(F
t
,
t
, g
t
) = H
(F
t
G,
t
, g
t
), (F
t
,
t
, g
t
). (3.39)
We call a transformation G L(V ) which satises (3.39) a material symmetry transfor
mation with respect to .
We assume that a material symmetry transformation is volumepreserving. Since,
otherwise, if G is a material symmetry transformation, so is G
n
for any n = 1, 2, , and
therefore, the material could suer arbitrarily large dilatation or arbitrary contraction
with no change in material response a conclusion that seems physically unacceptable
(for another justications see [8, 11]). Therefore, we must require that G (V ), where
(V ) = G L(V ) : [det G[ = 1 is called the unimodular group. It is easy to verify
that
Proposition. The set of all material symmetry transformations
(
= G (V ) [ H
(F
t
,
t
, g
t
) = H
(F
t
G,
t
, g
t
), F
t
L(V ),
is a subgroup of the unimodular group.
We call (
the material symmetry group of the material body in the reference cong
uration . We have
Condition of material symmetry. For any material symmetry transformation G
(
, the condition
H
(F
t
G) = H
(F
t
) (3.40)
holds for any deformation history F
t
.
Since by (3.39) the symmetry condition is a condition on the dependence of the rst
argument F
t
only, the other argument are not written out here.
7. MATERIAL SYMMETRY 37
It is clear that the symmetry group depends on the reference conguration. Suppose
that is another reference conguration such that P =
XX X
(
1
). Then for any
G (
(F
t
P) = H
(F
t
PG)
= H
(F
t
(PGP
1
)P) = H
(F
t
(PGP
1
)),
which implies that PGP
1
(
. Therefore we have proved the following proposition.
Proposition (Nolls rule). For any and , such that P =
XX X
(
1
), the following
relation hold,
(
= P(
P
1
. (3.41)
Physical concepts of real materials such as solids and uids, are usually characterized
by their symmetry properties. One of such concepts can be interpreted as saying that a
solid has a preferred conguration such that any nonrigid deformation from it alters its
material response, while for a uid any deformation that preserves the density should not
aect the material response. Based on this concept, we shall give the following denitions
of solids and uids due to Noll.
Denition. A material body is called a simple solid body if there exists a reference
conguration , such that (
O(V ).
Denition. A material is called a simple uid if the symmetry group is the full uni
modular group, i.e., (
= (V ).
For a simple uid, the Nolls rule (3.41) implies that the symmetry group is the
unimodular group relative to any conguration. In other words, a uid does not have a
preferred conguration.
A material which is neither a uid nor a solid will be called a uid crystal. In other
words, for a uid crystal, there does not exist a reference conguration for which either
(
O(V ) or (
O(V ).
Physically, we can interpret the above denition as saying that any rotation from
an isotropic conguration does not alter the material response. The following theorem
characterizes isotropic materials. For the proof of the theorem see [17].
Theorem. The orthogonal group is maximal in the unimodular group, i.e., if g is a
group such that
O(V ) g (V ),
then either g = O(V ) or g = (V ).
Therefore, an isotropic material is either a uid, (
T (Q
t
F
t
t
Q
T
, QF) = Q
T (F
t
t
, F) Q
T
,
T (F
t
t
, FG) =
T (F
t
t
, F),
Q O(V ),
G (.
(3.46)
We can obtain general constitutive relations for a simple uid, whose material sym
metry group is the unimodular group, satisfying the conditions of material objectivity
and symmetry.
Theorem. For a simple uid, the constitutive function
T satises the conditions of
material objectivity and symmetry (3.46) if and only if it can be represented by
T (F
t
t
, F) =
T (U
t
t
, ), (3.47)
where U
t
is the relative stretch tensor and is the mass density in the present congu
ration. Moreover,
T is an arbitrary isotropic symmetric tensorvalued function, i.e., it
satises the condition,
T (QU
t
t
Q
T
, ) = Q
T (U
t
t
, ) Q
T
, (3.48)
for any orthogonal tensor Q.
Proof. Since for a simple uid, the symmetry group ( = (V ), by taking the uni
modular tensor G = [det F[
1/3
F
1
the condition (3.46)
2
gives
T (F
t
t
, F) =
T (F
t
t
, [det F[
1/3
1).
Since [det F[ =
/, where
T (F
t
t
, F) =
T (F
t
t
, ).
The condition (3.46)
1
now becomes
T (F
t
t
, F) = Q
T
T (Q
t
F
t
t
Q
T
, ) Q.
40 3. CONSTITUTIVE RELATIONS
With the decomposition F
t
t
= R
t
t
U
t
t
, taking Q
t
(s) = R
t
t
(s)
T
, hence Q(t) = R
t
t
(0) = 1,
we obtain (3.47). Similarly, choosing Q
t
(s) =
Q(t) R
t
t
(s)
T
, hence Q(t) =
Q(t), for any
T (F
t
t
, F) =
T (U
t
t
, B), (3.49)
where U
t
is the relative stretch tensor and B = FF
T
is the left CauchyGreen tensor.
Moreover,
T is an arbitrary isotropic symmetric tensorvalued function, i.e., for any
orthogonal tensor Q, it satises the condition,
T (QU
t
t
Q
T
, QBQ
T
) = Q
T (U
t
t
, B) Q
T
.
Proof. For isotropic solids, ( = O(V ), by taking F = RU and G = R
T
, (3.46)
2
becomes
T (F
t
t
, F) =
T (F
t
t
, RUR
T
).
Since V = RUR
T
and B = V
2
, the function
T can be rewritten as
T (F
t
t
, F) =
T (F
t
t
, B).
The rest of the proof is similar to that of the previous theorem. .
CHAPTER 4
Constitutive Equations of Fluids
4.1 Materials of grade n
We have seen that the constitutive functions for isotropic materials are isotropic functions
satisfying the conditions (3.42). The main diculty in nding general solutions for
these conditions is due to the fact that they generally depend on all past values of
thermomechanical histories. However, in most practical problems, memory eects are
quite limited. In other words, most materials have only short memories, in the sense
that we can assume that the history be approximated by a Taylor series expansion,
h
t
(s) = h
t
(0) +
h
t
(s)
s
s=0
s +
1
2
2
h
t
(s)
s
2
s=0
s
2
+
n
k=0
1
k!
k
h
t
(s)
s
k
s=0
s
k
.
and consequently, the dependence of the constitutive functions on the history h
t
(s) re
duces to the dependence on the derivatives of h
t
(s) up to the nth order at the present
time, s = 0. A material body with such an innitesimal memory will be called a material
of grade n relative to the history variable h.
For a simple uid, from the representation (3.47) the constitutive relation can be
written as
T(x, t) = T (C
t
t
(x, s), (x, t)), (4.1)
where C
t
() is the relative right CauchyGreen tensor dened by C
t
= (U
t
)
2
. By dening
A
n
=
n
C
t
()
=t
, (4.2)
called the RivlinEricksen tensor of order n, we have
C
t
t
(s) 1 A
1
s + +
(1)
n
n!
A
n
s
n
.
Therefore, from (4.1), we can write the constitutive relation for a simple uid of grade n
in the form
T = T (, A
1
, , A
n
). (4.3)
Furthermore, it is also required that the constitutive function must be isotropic, i.e., it
must satisfy the condition (3.48), which now takes the form,
T (, QA
1
Q
T
, , QA
n
Q
T
) = QT (, A
1
, , A
n
) Q
T
. (4.4)
General solutions in the form of explicit representations for isotropic functions are known
in the literature.
42 4. CONSTITUTIVE EQUATIONS OF FLUIDS
4.2 Isotropic functions
We shall give a more general discussion on isotropic functions in this section. Let , h and
S be scalar, vector and tensorvalued functions dened on IRV L(V ) respectively.
Denition. We say that , h, and S are scalar, vector, and tensorvalued isotropic
functions respectively, if for any s IR, v V , A L(V ), they satisfy the following
conditions:
(s, Qv, QAQ
T
) = (s, v, A),
h(s, Qv, QAQ
T
) = Qh(s, v, A),
S(s, Qv, QAQ
T
) = QS(s, v, A) Q
T
,
Q O(V ). (4.5)
Isotropic functions are also called isotropic invariants. The denition can easily be
extended to any number of scalar, vector and tensors variables.
Example. The following functions are
1) isotropic scalar invariants: v u, det A, tr(A
m
B
n
), A
m
v B
n
u;
2) isotropic vector invariants: A
m
v, A
m
B
n
v;
3) isotropic tensor invariants: A
m
, A
m
v B
n
v;
for any u, v V and A, B L(V ). .
Example. The vector product v u is not an isotropic vector invariant, but it is a
vector invariant relative to the proper orthogonal group O
+
(V ). To see this, we have for
any w V ,
Qv Qu Qw = (det Q) v u w,
by the denition of the determinant. Hence we obtain
(Qv Qu) = (det Q) Q(v u).
Therefore it is not a vector invariant because det Q = 1. .
Before giving representation theorems for isotropic functions let us recall the Cayley
Hamilton theorem, which state that a tensor A L(V ) satises its characteristic equa
tion,
A
3
I
A
A
2
+II
A
A III
A
1 = 0, (4.6)
where I
A
, II
A
, III
A
are called the principal invariants of A. They are the coecients
of the characteristic polynomial of A, i.e.,
det(1 A) =
3
I
A
2
+II
A
III
A
= 0. (4.7)
Since eigenvalues of A are the roots of the characteristic equation (4.7), if A is symmetric
and a
1
, a
2
, a
3
are three eigenvalues of A, distinct or not, then it follows that
I
A
= a
1
+a
2
+ a
3
,
II
A
= a
1
a
2
+ a
2
a
3
+ a
3
a
1
,
III
A
= a
1
a
2
a
3
.
(4.8)
2. ISOTROPIC FUNCTIONS 43
It is obvious that I
A
= tr A and III
A
= det A. Moreover, I
A
, II
A
and III
A
are respectively
a rst order, a second order and a third order quantities of [A[. Let the space of symmetric
linear transformations on V be denoted by Sym(V ) = A L(V ) [ A = A
T
.
Theorem (Rivlin & Ericksen). Let S : Sym(V ) Sym(V ), then it is an isotropic
function if and only if it can be represented by
S(A) = s
0
1 + s
1
A + s
2
A
2
, (4.9)
where s
0
, s
1
and s
2
are arbitrary scalar functions of (I
A
, II
A
, III
A
), the three principal
invariants of A.
Proof. The proof of suciency is trivial, and we shall only prove the necessity of the
above representations. We need the following lemma.
Lemma. Let S(A) be an isotropic tensorvalued function of a symmetric tensor A, then
every eigenvector of A is an eigenvector of S(A).
To prove the lemma, let e
1
, e
2
, e
3
be a principal orthonormal basis of A. We
consider a rotation Q of 180
about e
1
axis, therefore,
Qe
1
= e
1
, Qe
2
= e
2
, Qe
3
= e
3
.
Clearly, we have QAQ
T
= A and since S(A)is isotropic, it follows that
QS(A) = S(QAQ
T
)Q = S(A)Q.
Therefore, we have
QS(A)e
1
= S(A)Qe
1
= S(A)e
1
,
i.e., S(A)e
1
remains unchanged under Q. By our choice of Q, this can happen only if
S(A)e
1
is in the direction of e
1
. In other words, e
1
is an eigenvector of S(A). For other
eigenvectors, we can use similar arguments. Hence the lemma is proved. .
From the lemma, if we express A as
A = a
1
e
1
e
1
+ a
2
e
2
e
2
+ a
3
e
3
e
3
, (4.10)
then S(A) can be expressed as
S(A) = b
1
e
1
e
1
+ b
2
e
2
e
2
+ b
3
e
3
e
3
, (4.11)
where b
1
, b
2
, and b
3
are functions of A.
Suppose that the three eigenvalues a
1
, a
2
, and a
3
are distinct. We consider the
following simultaneous equations for s
0
, s
1
, and s
2
,
s
0
+ a
1
s
1
+ a
2
1
s
2
= b
1
,
s
0
+ a
2
s
1
+ a
2
2
s
2
= b
2
,
s
0
+ a
3
s
1
+ a
2
3
s
2
= b
3
.
44 4. CONSTITUTIVE EQUATIONS OF FLUIDS
Since the determinant of the coecient matrix does not vanish,
det
_
_
1 a
1
a
2
1
1 a
2
a
2
2
1 a
3
a
2
3
_
_ = (a
1
a
2
)(a
2
a
3
)(a
3
a
1
) ,= 0,
we can solve for s
0
, s
1
, and s
2
which are functions of A of course. Then from (4.10),
(4.11) can be written in the form
S(A) = s
0
(A)1 + s
1
(A)A +s
2
(A)A
2
. (4.12)
Furthermore, since S(A) is isotropic,
S(QAQ
T
) = QS(A)Q
T
, Q O(V ),
it follows from (4.12) that
s
0
(QAQ
T
)1 + s
1
(QAQ
T
)A + s
2
(QAQ
T
)A
2
= s
0
(A)1 + s
1
(A)A + s
2
(A)A
2
,
and we conclude that s
i
are all scalar isotropic functions,
s
i
(QAQ
T
) = s
i
(A), i = 0, 1, 2, Q O(V ).
Now it suces to show that s
i
(A) = s
i
(B) whenever A and B have the same principal
invariants, or equivalently they have the same eigenvalues by (4.8). Thus let A and B
be two symmetric tensors and assume that their eigenvalues are the same. Then by the
spectral theorem, there exist orthonormal bases e
i
and d
i
such that
A =
i
a
i
e
i
e
i
, B =
i
a
i
d
i
d
i
.
Let Q be the orthogonal tensor carrying the basis d
i
into the basis e
i
, Qd
i
= e
i
.
Then since
Q(d
i
d
i
)Q
T
= (Qd
i
) (Qd
i
) = e
i
e
i
,
it follows that QBQ
T
= A. But since s
i
is isotropic,
s
i
(A) = s
i
(QBQ
T
) = s
i
(B),
which proves that s
i
= s
i
(I
A
, II
A
, III
A
).
With similar arguments one can show that when A has exactly two distinct eigen
values, S(A) admits the representation (4.11)) with s
2
= 0. For the case when A = a1,
every vector is an eigenvector, and hence by the lemma, we must have S(A) = s
0
1 which
is a special case of the representation (4.9). The proof of the Corollary is straightforward.
.
2. ISOTROPIC FUNCTIONS 45
Corollary. If S(A) is an isotropic and linear function of A, then
S(A) = (tr A)1 + A, (4.13)
where and are independent of A.
Proof. Since S(A) is isotropic and linear in A, we have, for any ,
S(A) = s
0
(A)1 + s
1
(A)A + s
2
(A)A
2
, S(A) = S(A),
It follows that
s
0
(A)1 + s
1
(A)A + s
2
(A)
2
A
2
=
_
s
0
(A)1 + s
1
(A)A +s
2
(A)A
2
_
,
which implies that
s
0
(A) = s
0
(A), s
1
(A) = s
1
(A), s
2
(A) = s
2
(A).
By taking = 0 in the last two relations, we have
s
1
(A) = s
1
(0), s
2
(A) = 0.
Therefore, by setting s
1
(0) = , we obtain
S(A) = s
0
(A)1 + A.
On the other hand, by isotropy, we have
s
0
(A) = s
0
(I
A
, II
A
, III
A
),
and from s
0
(A) = s
0
(A), it follows that
s
0
(I
A
, II
A
, III
A
) = s
0
(I
A
, II
A
, III
A
).
Since
I
A
= I
A
, II
A
=
2
II
A
, III
A
=
3
III
A
,
by taking the derivative with respect to , we obtain
s
0
(I
A
, II
A
, III
A
) =
s
0
I
A
A
I
A
+ 2
s
1
II
A
A
II
A
+ 3
2
s
2
III
A
A
III
A
,
where the partial derivatives are evaluated at A i.e., (I
A
, II
A
, III
A
). Consequently,
by setting = 0, it follows that
s
0
(I
A
, II
A
, III
A
) =
s
0
I
A
A=0
I
A
= I
A
,
where is independent of A, i.e.,
s
0
(A) = tr A,
and the corollary is proved. .
Representations for isotropic functions of any number of vector and tensor variables
have been extensively studied and the results are usually tabulated in the literature
[14, 24, 19, 1]. We shall give here without proof another theorem for isotropic functions
of one vector and one symmetric tensor variables.
46 4. CONSTITUTIVE EQUATIONS OF FLUIDS
Theorem. Let , h and S be scalar, vector and symmetric tensorvalued functions
respectively of a vector and a symmetric tensor variables. Then they are isotropic if and
only if
= f(I
A
, II
A
, III
A
, v v, v Av, v A
2
v),
h = h
0
v + h
1
Av + h
2
A
2
v,
S = s
0
1 + s
1
A + s
2
A
2
+ s
3
v v + s
4
(Av v +v Av) + s
5
Av Av,
(4.14)
where the coecients h
0
through s
5
are arbitrary functions of the variables indicated in
the scalar function f.
4.3 NavierStokes uids
A simple uid of grade 1 is dened by the constitutive equation
T = T (, D),
where the rate of stretch tensor D is related to the RivlinEricksen tensor A
1
by A
1
= 2D.
Since it is an isotropic function, by the representation theorem (4.9) the constitutive
relation is given by
ReinerRivlin uid. The constitutive equation is given by
T =
0
1 +
1
D +
2
D
2
,
i
=
i
(, I
D
, II
D
, III
D
), i = 0, 1, 2.
(4.15)
This is the most general form of constitutive equations for simple uids of grade 1.
Even though this seems to be a general model for nonlinear viscosity, it has been pointed
out that this model is inadequate to describe some observed nonlinear eects in real uids
(see Sect. 119 [22]). Nevertheless, by the use of (4.13), the special case, when only up
to linear terms in the stretching tensor D are considered, leads to the most wellknown
model in uid mechanics.
NavierStokes uid. The constitutive equation is given by
T = (p + tr D) 1 + 2D. (4.16)
The coecients and are called the coecients of viscosity, while and ( +
2
3
)
are also known as the shear and the bulk viscosities respectively. The pressure p and the
viscosities and are functions of . A NavierStokes uid is also known as a Newtonian
uid in uid mechanics. It is usually assumed that
0, +
2
3
0. (4.17)
The nonnegativeness of the shear and bulk viscosities can be proved from thermodynamic
considerations. When the bulk viscosity vanishes identically, it is known as a Stokes
uid, a model adequate to describe some real uids and frequently used in numerical
calculations.
3. NAVIERSTOKES FLUIDS 47
Stokes uid. The stress tensor is given by
T = p 1 + 2
D, (4.18)
where
D is the traceless part of D, which in component forms is given by
D
ij
=
1
2
_
v
i
x
j
+
v
j
x
i
_
1
3
v
k
x
k
ij
.
A NavierStokes uid is governed by the system of equations consists of the conser
vation of mass (2.22),
t
+
x
k
( v
k
) = 0, (4.19)
and the equation of motion (2.29),
_
v
i
t
+ v
k
v
i
x
k
_
+
p
x
i
x
i
_
v
k
x
k
_
x
k
_
v
k
x
i
_
x
k
_
v
i
x
k
_
= b
i
. (4.20)
The last equation (4.20) is known as the NavierStokes equation. The pressure p and the
viscosities, and , are in general functions of the density . Equations (4.19) and (4.20)
form a system for the elds ((x, t), v(x, t)). For Stokes uids, the governing equations
are obtained from above by substitution of =
2
3
.
We remark that the linear constitutive relation in (4.16) need not be regarded as an
approximation of the more general constitutive relation (4.15). Any particular form of
a constitutive relation (4.15) characterizes a particular class of simple uids. Thus it is
conceivable that there are some uids which obey the constitutive equation (4.16) for
arbitrary rate of deformation. Indeed, water and air are usually treated as NavierStokes
uids in most practical applications with very satisfactory results.
The simplest constitutive equation used in continuum mechanics is that of an elastic
uid, which is inviscid, i.e., no viscosities. In the following constitutive relation we have
included the dependence of temperature in general.
Elastic uid. The constitutive variables are (, ), therefore the we have
T = p(, ) 1, q = 0, = (, ). (4.21)
For an elastic uid, from the balance equations of mass, linear momentum and energy,
we have the following governing equations in component forms:
t
+
x
j
( v
j
) = 0,
t
( v
i
) +
x
j
( v
i
v
j
+ p
ij
) = b
i
,
t
( +
1
2
v
2
) +
x
j
_
v
j
+
1
2
v
2
v
j
+p v
j
_
= r.
(4.22)
The internal energy = (, ) and the pressure p = p(, ) have to be specied, for
instance by the ideal gas laws, and the equations becomes a hyperbolic system for the
elds ((x, t), v(x, t), (x, t)), usually known as the Euler equations for compressible
ows of ideal gases.
48 4. CONSTITUTIVE EQUATIONS OF FLUIDS
4.4 Viscous heatconducting uids
We consider a simple uid of grade 1 relative to mechanical histories and of grade 0
relative to thermal histories, i.e.,
( = T(, , g, D). (4.23)
This class of simple uids can describe ow properties with viscosity and heat conduc
tion. From the representation (4.14), one can immediately write down the most general
constitutive equations for the stress, the heat ux and the internal energy.
Viscous heatconducting uid. The constitutive equations are given by
T =
0
1 +
1
D +
2
D
2
+
3
g g +
4
(Dg g +g Dg) +
5
Dg Dg,
q =
1
g +
2
Dg +
3
D
2
g,
= (, , I
D
, II
D
, III
D
, g g, g Dg, g D
2
g),
(4.24)
where the coecient
i
and
j
as well as are scalar functions of eight variables indicated
in the arguments of .
The special case, when only up to linear terms in both D and g are considered, gives
the following wellknown model for viscosity and heat conduction in common use.
NavierStokesFourier uids. The constitutive equations for the stress tensor and
the heat ux are given by
T = (p + tr D) 1 + 2D,
q = g,
(4.25)
where the coecients as well as the internal energy are functions of (, ).
These are the classical NavierStokes theory and the Fouriers law of heat conduction.
The material parameters , are the viscosity coecients and is called the thermal
conductivity. The thermal conductivity is usually assumed to be nonnegative, which
can also be proved from thermodynamic considerations. The governing equations for the
elds ((x, t), v(x, t), (x, t)) now consist of the mass conservation (4.19), the Navier
Stokes equation (4.20) and the equation for energy (2.33) with the NavierStokes stress
and the Fouriers law (4.25).
4.5 Incompressibility
A motion is called incompressible if it is volumepreserving, which can be characterized
by det F = 1. We call a body incompressible material body if it is capable of undergoing
only incompressible motions.
In the formulation of constitutive relations discussed in the previous chapter, it is
assumed that a material body is capable of undergoing any compressible or incompressible
5. INCOMPRESSIBILITY 49
motions. Obviously, for incompressible bodies, some constitutive assumptions must be
modied. Indeed, in order to maintain the constant volume some internal stress is needed
to counter the eect of change of volume due to applied forces on the body. This is called
the reaction stress and it will not do any real works in the motion. Since the rate of work
in the motion due to the reaction stress N can be expressed as (N grad v), we shall
require that
N grad v = N
FF
1
= 0, (4.26)
for any incompressible motion. Taking the material time derivative of the equation
det F = 1, we obtain
F
T
F = 1
FF
1
= 0.
By comparison, we conclude that the reaction stress N must be proportional to the
identity tensor, so let us write,
N = p 1.
Therefore for an incompressible body, the stress tensor can be expressed as
T = p 1 +S, (4.27)
where p, called the indeterminate pressure, is a function that can not be uniquely de
termined by the motion, and the extra stress S is a constitutive quantity subject to
the general requirements of material objectivity and material symmetry discussed in the
previous sections.
Incompressible NavierStokesFourier uid. The constitutive equation is given by
T = p 1 + 2() D, tr D = 0,
q = () g,
(4.28)
where p is an indeterminate pressure.
The condition tr D = 0 ensures that the ow is incompressible, and hence the mass
density is constant, =
0
, and
div v = 0. (4.29)
The NavierStokes equation can now be written as
v + grad p div(() grad v) = b. (4.30)
In an isothermal process =
0
, the equation (4.30) reduces to
v + grad p
2
v = b, (4.31)
where = () is a constant. The two equations (4.29) and (4.31) become the governing
equations for the elds (p(x, t), v(x, t)).
50 4. CONSTITUTIVE EQUATIONS OF FLUIDS
4.6 Viscometric ows
When a particular class of motions is concerned the generality of constitutive relations
could be severely restricted resulting in a much simpler representation of constitutive
equations. In this section we shall consider a general class of motions, called viscometric
ows, for simple uids. This class of motions has a considerable importance in both
theoretical and experimental investigations of nonlinear viscosities.
We consider a viscometric ow dened by the following velocity eld,
v(x, t) = u(y) e
x
, (4.32)
where x = (x, y, z) and e
x
, e
y
, e
z
is the natural basis of a coordinate system such that
e
x
and e
y
are constant unit vector elds. This class of ows includes simple shearing
channel ows in a Cartesian coordinate system, as well as Poiseulle and Couette ows in
a cylindrical coordinate system (r, , z) by regarding the r and the zaxes as the y and
the xaxes in (4.32) respectively.
One can check easily that [ det F[ = 1, therefore the ow is incompressible. From
(1.27) the relative deformation gradient is given by
F
t
(t s) = F(t s)F(t)
1
= 1 sN,
where we have dened
=
du
dy
, N = e
x
e
y
. (4.33)
With C
t
= F
T
t
F
t
it follows that
C
t
(t s) = 1 s A
1
+
1
2
s
2
A
2
, (4.34)
where
A
1
= (N + N
T
), A
2
= 2
2
N
T
N (4.35)
are the RivlinEricksen tensors (see (4.2)). Therefore, from (4.1) and (4.3) we conclude
that as long as viscometric ows are concerned the most general simple uids are simple
uids of grade 2 and the constitutive relation can be expressed as
T = T (A
1
, A
2
), (4.36)
and by (4.4) the function T must satisfy
T (QA
1
Q
T
, QA
2
Q
T
) = QT (A
1
, A
2
) Q
T
, (4.37)
for any Q O(V ).
An immediately consequence follows from the condition (4.37) by choosing Q as the
180
_
1 0 0
0 1 0
0 0 1
_
_ +
_
1
0
2
0
0 0 0
_
_ , (4.38)
where the pressure p and the three other parameters ,
1
and
2
are scalar functions
of A
1
and A
2
, or by (4.35) they are functions of only. We call the shear stress
viscometric function, and call
1
and
2
the normal stress viscometric functions.
Again, by the condition (4.37), we have
T (0, 0) = QT (0, 0) Q
T
, Q O(V ),
which implies that
T (0, 0) = 1,
for some constant . Therefore, from (4.38) it follows that
= p(0),
and
(0) =
1
(0) =
2
(0) = 0. (4.39)
Moreover, by choosing Q as the 180
C=1
is a fourth order tensor which is a linear transformation of the space of symmetric tensors
into itself. If we further assume that the reference conguration is a natural state, i.e.,
2. LINEAR ELASTICITY 55
T (1) = 0 and so is o(1) = 0, then by neglecting the second order terms in (5.4) we obtain
the linear stressstrain law,
T = L[E], (5.7)
since L[E] is of order o(1) and F = 1 + o(1). This linear stressstrain relation is also
known as the Hookes law and L is called the elasticity tensor. By denition the elasticity
tensor has the following symmetry properties in terms of components:
L
ijkl
= L
jikl
= L
ijlk
, (5.8)
and an additional symmetry,
L
ijkl
= L
klij
, (5.9)
if the material is hyperelastic, i.e., there exists a function (F), called a stored energy
function, such that
T =
E
. (5.10)
The existence of a stored energy function will be proved later from thermodynamic con
siderations.
Moreover, the conditions of material objectivity and material symmetry (5.2) imply
that
T (QFQ
T
) = QT (F) Q
T
, G Q O(V ).
Since
T(F) = L[E(F)], E(F) =
1
2
(H + H
T
) =
1
2
(F + F
T
) 1,
it follows immediately that
L[QEQ
T
] = QL[E] Q
T
, Q ( O(V ). (5.11)
This relation can be written in component forms,
L
ijkl
= Q
im
Q
jn
Q
kp
Q
lq
L
mnpq
.
Finally, by (1.16) and the symmetry condition (5.8), the equation of motion (2.29)
for linear elasticity in component forms is given by
2
u
i
t
2
=
x
j
_
L
ijkl
u
k
x
l
_
+ b
i
. (5.12)
This is the governing equation for the displacement vector u.
Remark. We should point out that the linear law (5.7) does not satisfy the condition
of material objectivity (5.2)
1
for arbitrary orthogonal tensor Q. Therefore, unlike the
NavierStokes theory, which does satisfy the condition of material objectivity, the theory
of linear elasticity is meaningless for large deformations.
56 5. CONSTITUTIVE EQUATIONS OF SOLIDS
Indeed, if we choose F = 1, which is a natural state by assumption, then the condition
(5.2)
1
implies that
T(Q) = QT(1)Q
T
= 0,
for any orthogonal tensor Q. On the other hand, since
T(F) = L[E(F)], E(F) =
1
2
(H + H
T
) =
1
2
(F + F
T
) 1,
if we choose Q as a rotation about zaxis,
Q =
_
_
cos sin 0
sin cos 0
0 0 1
_
_ ,
we have
E(Q) =
_
_
cos 1 0 0
0 cos 1 0
0 0 0
_
_ ,
and for ,= 0
T(Q) = L[E(Q)] ,= 0.
Hence, the condition of material objectivity is not satised in general. Nevertheless, one
can show that it is approximately satised when both the displacement and the rotation
are small (in this case, cos 1). .
5.3 Isotropic elastic solids
For an isotropic solid with no memory of past histories, from the constitutive relation
(5.6) and the representation theorem (4.9), we have
Isotropic elastic solid. The constitutive equation can be written as
T = t
0
1 + t
1
B + t
2
B
2
, (5.13)
where t
i
for i = 0, 1, 2 are functions of (I
B
, II
B
, III
B
), the principal invariants of B.
This is the general constitutive equation for isotropic nite elasticity, the theory of
elastic solids for nite deformations. It can also be obtained as a special case from the
constitutive equation (5.4) with o being an isotropic function of C by (5.5), since the
symmetry group ( is the orthogonal group O(V ).
For small deformations, we have
B = FF
T
= 1 + 2E +o(2),
and hence the linear approximation of T (B) at B = 1 can be written as
T (B) = 2
B
T (1)[E] + o(2),
4. INCOMPRESSIBLE ELASTIC SOLIDS 57
where the reference conguration is assumed to be a natural state as before. By explicitly
carrying out the gradient
B
T from (5.13) and neglecting the second order terms, we
obtain
T = (tr E)1 + 2E, (5.14)
where and are called the Lame elastic moduli and they are related to the material
parameters t
0
, t
1
, and t
2
of (5.13) by
= 2
_
t
I
B
+ 2
t
II
B
+
t
III
B
_
(3,3,1)
t = t
0
+t
1
+t
2
,
= t
1
(3, 3, 1) + 2 t
2
(3, 3, 1).
(5.15)
The equation (5.14) is the constitutive relation of the classical theory of isotropic linear
elasticity. It is a special case of (5.7) with the elasticity tensor given by
L
ijkl
=
ij
kl
+ (
ik
jl
+
il
jk
). (5.16)
For an isotropic linear elastic body, the equation of motion becomes
2
u
i
t
2
=
x
i
_
u
k
x
k
_
+
x
k
_
(
u
i
x
k
+
u
k
x
i
)
_
+ b
i
. (5.17)
If the body is homogeneous, then and are constants, and the governing equation
becomes
u = ( + ) grad(div u) +
2
u + b. (5.18)
The linear stressstrain relation (5.14) can also be derived directly from (5.7) for the
symmetry group ( = O(V ). Indeed, if we dene
S(E) = L[E],
then by (5.11) the symmetric tensorvalued function S is an isotropic function and it is
linear in the symmetric tensor variable E. Therefore, from the representation theorem
(4.13), we can represent S(E) as an expression linear in E in the following form,
S(E) = (tr E)1 + 2E.
5.4 Incompressible elastic solids
Incompressible elastic bodies can be similarly formulated. It has been shown that the
reaction stress for incompressibility is a hydrostatic pressure and the stress tensor takes
the form (4.27).
58 5. CONSTITUTIVE EQUATIONS OF SOLIDS
Incompressible elastic material. The constitutive equation for the stress tensor is
given by
T = p1 +T (F), det F = 1, (5.19)
where p is the indeterminate hydrostatic pressure.
For incompressible isotropic elastic solids, (5.19) reduces to T = p1 +T (B), where
the function T is an isotropic function. Hence from the representation (5.13) the consti
tutive equation can be written as
T = p 1 + t
1
B +t
2
B
2
,
or equivalently, by the use of CayleyHamilton theorem,
T = p 1 + s
1
B + s
1
B
1
, (5.20)
where B is required to be unimodular, i.e., det B = 1 and in general the parameters s
1
and s
1
are functions of (I
B
, II
B
), since III
B
= 1 for incompressibility. Experimental
data seem to indicate that
s
1
> 0, s
1
0. (5.21)
This assumption is known as (empirical) Einequalities in elasticity (see Sect. 55 [22] for
more discussions on experimental data).
Two special cases are of practical interest for nite elasticity, namely, the simple
models for which the parameters s
1
and s
1
are constants.
NeoHookean material. The stress tensor takes the form,
T = p 1 +s
1
B, det B = 1, (5.22)
where s
1
is a positive constant parameter.
MooneyRivlin material. The stress tensor takes the form,
T = p 1 + s
1
B + s
1
B
1
, det B = 1, (5.23)
where s
1
> 0 and s
1
0 are constants.
These incompressible material models are often adopted for rubberlike materials.
NeoHookean materials are also predicted by the kinetic theory of rubber from molecular
calculations in the rst approximation [20]. It provides a reasonable theory of natural
rubber for modest strains.
5. THERMOELASTIC MATERIALS 59
5.5 Thermoelastic materials
Heat conduction in elastic materials is usually taken into account by including the tem
perature and the temperature gradient g as constitutive variable.
Thermoelastic material. The constitutive equations are given in the form,
( = T(F, , g). (5.24)
Beside the requirements of material objectivity and material symmetry, for thermoe
lastic materials, thermodynamic considerations are essential to obtain a great simplica
tion to constitutive equations. We shall state here the main results for further discussions
and leave the proof in the next chapter
1
.
Proposition. For thermoelastic materials, there exists a function = (F, ), such
that the stress tensor T and the internal energy are given by
T =
F
F
T
, =
, (5.25)
and the heat ux q satises the condition,
q g 0. (5.26)
Consequently, the stress T and the internal energy are independent of the tem
perature gradient g and they are completely determined by a scalar function (F, ),
called the free energy function. The relation (5.25)
1
also states that the material body is
hyperelastic.
For isotropic thermoelastic solid bodies, by (5.6) and the representation theorem
(4.14) the most general constitutive functions can now be written down immediately.
Isotropic thermoelastic solid. The constitutive equations for the stress tensor and
the heat ux vector are given by
T = t
0
1 +t
1
B + t
2
B
2
,
q = k
1
g + k
2
Bg + k
3
B
2
g,
(5.27)
The internal energy and the coecients t
i
are scalar functions of (, I
B
, II
B
, III
B
), while
the coecients k
i
are functions of (, I
B
, II
B
, III
B
, g g, g Bg, g B
2
g).
Further restrictions on the coecient functions can be obtained from the relations
(5.25).
1
It will be proved for isotropic elastic materials and elastic materials of Fourier type based on the
general entropy inequality. However, it can also be proved for elastic materials in general based on the
more restricted ClausiusDuhem inequality.
60 5. CONSTITUTIVE EQUATIONS OF SOLIDS
5.6 Linear thermoelasticity
It is more convenient to introduce a function called the entropy density, which will be
considered in more details later, by = + . The relations (5.25) becomes,
T =
F
F
T
, =
. (5.28)
From these relations we can obtain
= +
1
TF
T
F,
and hence the energy equation (2.33) for thermoelastic materials can be written as
+ div q = r. (5.29)
For the classical linear theory, the displacement gradient H and the temperature
increment
=
0
are assumed to be small quantities, where
0
is the reference
temperature of the body. On the other hand, similar to (3.44) the principle of material
objectivity requires that the dependence of the free energy function on the deformation
gradient F must reduces to the dependence of the right stretch tensor U. Since U =
1+E+o(2) by (1.12), in the linear theory we have = (E, ). Therefore let us express
the function up to the second order terms in E and
in the following form,
=
0
+M
ij
E
ij
1
2
c
v
0
P
ij
E
ij
+
1
2
L
ijkl
E
ij
E
kl
. (5.30)
The relations (5.28) now take the form,
T =
0
E
, =
,
and if the reference state is assumed to be a natural state, then M
ij
in (5.30) must vanish
and we obtain
T
ij
= L
ijkl
E
kl
P
ij
,
=
0
+
c
v
+
1
0
P
ij
E
ij
.
(5.31)
The fourth order tensor L is the elasticity tensor (see (5.7)) and c
v
is called the specic
heat because from (5.25)
2
it follows that
c
v
=
.
For the linear theory, we shall assume in addition that the Fouriers law of heat
conduction holds, so that the linear expression for the heat ux is given by
q
i
= K
ij
g
j
, (5.32)
where K is the thermal conductivity tensor.
6. LINEAR THERMOELASTICITY 61
Therefore, we can summarize the constitutive equations of linear thermoelasticity for
anisotropic materials in the following:
T = L[E] P
,
q = Kg,
=
0
+
c
v
+
P
0
E.
(5.33)
Moreover, the coecients satisfy the following conditions:
L
ijkl
= L
jikl
= L
ijlk
= L
klij
, P
ij
= P
ji
,
K is positive semidenite,
c
v
> 0.
(5.34)
The rst two conditions follow from the denition in (5.30) and the condition for the
thermal conductivity tensor follows from (5.26). The last inequality is a consequence of
thermal stability which will be discussed later.
If the material is isotropic, then we have, for any orthogonal tensor Q,
T(QFQ
T
, ) = QT(F, ) Q
T
,
q(QFQ
T
, , Qg) = Qq(F, , g).
Since
E =
1
2
(H + H
T
) =
1
2
(F + F
T
) 1,
from (5.33) it follows immediately that
L[QEQ
T
] = QL[E] Q
T
,
P = QPQ
T
,
KQ = QK.
Therefore, we conclude that in addition to the relation (5.16) we also have
P
ij
=
ij
, K
ij
=
ij
.
In summary, the constitutive equations of linear thermoelasticity for isotropic mate
rials are given in the following:
T = tr E 1 + 2E
1,
q = g,
=
0
+
c
v
0
tr E.
(5.35)
The eld equations for thermoelasticity consist of the momentum equation and the
energy equation (or the equivalent equation (5.29)) for the displacement u(x, t) and the
62 5. CONSTITUTIVE EQUATIONS OF SOLIDS
temperature (x, t). In component forms, from (1.16) and the constitutive equations
(5.33), we have the following eld equations for anisotropic thermoelastic materials:
2
u
i
t
2
x
j
_
L
ijkl
u
k
x
l
_
+
x
j
_
P
ij
(
0
)
_
=
0
b
i
,
0
c
v
t
+
0
P
ij
t
_
u
i
x
j
_
x
i
_
K
ij
x
j
_
=
0
r.
(5.36)
In particular, if the body is homogeneous, i.e., the material coecients are independent
of X (in the linear theory, it is the same as x), then the eld equations (5.36) become
2
u
i
t
2
L
ijkl
2
u
k
x
j
x
l
+ P
ij
x
j
=
0
b
i
,
0
c
v
t
+
0
P
ij
t
_
u
i
x
j
_
K
ij
x
i
x
j
=
0
r,
(5.37)
or if the body is rigid and xed, then P
ij
= 0 and the only eld equation is the classical
equation of heat conduction,
0
c
v
t
+
x
i
_
K
ij
x
j
_
=
0
r. (5.38)
For isotropic thermoelastic materials, from (5.35) the eld equations become
2
u
i
t
2
x
i
_
u
k
x
k
_
x
j
_
(
u
i
x
j
+
u
j
x
i
)
_
+
x
i
_
(
0
)
_
=
0
b
i
,
0
c
v
t
+
0
t
_
u
i
x
i
_
x
i
_
x
i
_
=
0
r.
(5.39)
If the body is homogeneous then the eld equations become
2
u
i
t
2
( + )
2
u
k
x
i
x
k
2
u
i
x
k
x
k
+
x
i
=
0
b
i
,
0
c
v
t
+
0
t
_
u
k
x
k
_
x
k
x
k
=
0
r.
(5.40)
5.7 Fading Memory
In this section, we shall give a brief discussion of memory eects of simple materials and
the linear approximation which gives rise to the BoltzmannVolterra theory of viscoelas
ticity.
From the general solution (3.44) for simple materials, the stress tensor can be repre
sented by
T(t) = T (F
t
) = R(t) T (U
t
) R(t)
T
, F = RU.
7. FADING MEMORY 63
The dependence on X is not indicated for simplicity. It is useful to put the functional
T (U
t
) in a slightly dierent form by writing it as a sum of a term at the present value
of deformation and a term which vanishes when the body has always been at rest,
T (U
t
(s)) =
T (U(t)) +/(G
t
(s); U(t)), (5.41)
where G
t
is dened as
G
t
(s) = U
t
(s) U(t), and /(0; U(t)) = 0.
We shall now make an additional physical assumption that the memory fades in time. In
order to do this we shall rst introduce the concept of an inuence function which is used
to characterize the memory of past histories the material can recall (for more general
discussions on fading memory see [3, 4, 23]).
Denition. We call h(s), dened for 0 s < , an inuence function if it is monoton
ically decreasing and
h(s) > 0,
_
0
h(s) ds < . (5.42)
An inuence function is a material property. h(s) = (s +1)
p
for p > 1, and h(s) = e
s
for > 0 are some examples of such functions. We can then dene a norm to measure
the recollection of past histories.
Denition. The norm G of a history G(s) relative to an inuence function h(s) is
dened as
G =
_
_
0
[G(s)[
2
h(s) ds
_
1/2
, (5.43)
where [G[ is the norm of the tensor G, i.e., [G[
2
= tr G
T
G.
The value of G is also referred to as the recollections of G. The inuence function
h(s) is regarded as the weight in computing the norm of G(s). Since h(s) decreases
monotonically, the values of G(s) for small s (recent past) have a greater weight than
the values for large s (distant past). Note that by (5.42) all the constant histories, have
nite recollections. A constant history A
c
(s) is a history such that A
c
(s) = A(0) for
0 s < .
The set of all histories with nite recollections forms a Hilbert space IH with the inner
product dened by
(G, H)
IH
=
_
0
tr(G(s)
T
H(s)) h(s) ds (5.44)
for any G, H IH.
A deformation history which has always been nearly at a constant history, or that may
have suered some large departures from a constant history only in the distant past, has
a small recollection in the norm (5.43). These circumstances are those we should expect
from our experiences that should not much aect the stress in a material with fading
memory.
64 5. CONSTITUTIVE EQUATIONS OF SOLIDS
Material with fading memory. We call a material dened by the constitutive equa
tion
T(t) = T (F
t
) = R(t) T (U
t
) R(t)
T
, F = RU,
where
T(U
t
(s)) =
T (U(t)) +/(G
t
(s); U(t)), G
t
(s) = U
t
(s) U(t), (5.45)
a material with fading memory if both the function
T and the functional / are smooth
in some proper sense in the usual norm in L(V ) and the fading memory norm (5.43) for
past histories in IH.
5.8 Linear viscoelasticity
For small deformations, the displacement gradient H = F 1 is an innitesimal quantity
with [H(t)[ of the order 1 for all time t. We use the order symbol o(n) in the sense
of the usual norm as well as the fading memory norm,
[o(n)[ < k
n
, o(n) < k
n
,
for some constants k and k
K[G])
ij
=
K
ijkl
G
kl
, and hence, for any indices i and j
xed,
K
ij . .
must belong to IH. It follows that
K must have the property:
_
0
[
K(s, U)[
2
h(s) ds < .
Therefore, we have the following linear approximation,
/(G
t
; U) =
_
0
K(s; U(t))[G
t
(s)] ds + o(2), (5.47)
where K(s) =
K(s)h(s) have the property:
_
0
[K(s; U)[
2
h(s)
1
ds < .
Since
G
t
(s) = U
t
(s) U(t) = E(t s) E(t) + o(2),
it follows that
/(G
t
; U) =
_
0
K(s; 1)[E(t s)] ds
_
0
K(s; 1)[E(t)] ds + o(2),
which on dening the stress relaxation function M,
M(s) =
_
s
K(; 1) d,
M(s) = K(s; 1), (5.48)
becomes
/(G
t
; U) = M(0)[E(t)] +
_
0
M(t )[
E()] d, (5.51)
66 5. CONSTITUTIVE EQUATIONS OF SOLIDS
This is the classical equation of linear viscoelasticity of the BoltzmannVolterra theory.
The special case M(s) 0 corresponds to the classical theory of linear elasticity.
Both the elasticity tensor L and the stress relaxation function M are fourth order
tensors and their components satisfy the following symmetry conditions:
L
ijkl
= L
jikl
= L
ijlk
,
M
ijkl
= M
jikl
= M
ijlk
,
since both the stress and the strain tensors are symmetric.
CHAPTER 6
Thermodynamic Considerations
6.1 Second law of thermodynamics
In this chapter we shall give a brief consideration of thermodynamic restrictions imposed
on constitutive equations. We have already mentioned the rst law of thermodynamics,
i.e., the energy balance, in Chapter 2. Now we are going to consider the second law for
which the essential quantity is the entropy,
_
P
t
dv, (6.1)
where (x, t) is called the specic entropy density. Unlike the total energy, the rate of
change of total entropy of a body can not be given completely in the form of a balance
equation (2.1). There are internal entropy productions in nonequilibrium processes.
Entropy production. For any part T B, the entropy production (T, t) is given by
(T, t) =
d
dt
_
P
t
dv +
_
P
t
nda
_
P
t
s dv.
We call (x, t) the entropy ux and s the external entropy supply density. Although
entropy is not a quantity associated with some easily measurable physical quantities, its
existence is usually inferred from some more fundamental hypotheses concerning thermal
behaviors of material bodies, usually known as the second law of thermodynamics. We
choose to accept the existence of entropy and state the consequence of such hypotheses
directly by saying that the entropy production is a nonnegative quantity.
Second law of thermodynamics. The following entropy inequality must hold for any
part T B:
d
dt
_
P
t
dv +
_
P
t
nda
_
P
t
s dv 0. (6.2)
Comparing with the general balance equation (2.1) by setting
= ,
= ,
= s,
we have the following local form of the entropy inequality,
+ div s 0. (6.3)
2. ENTROPY PRINCIPLE 69
6.2 Entropy principle
One of the principal objectives of continuum mechanics is to determine or predict the
behavior of a body once the external causes are specied. Mathematically, this amounts
to solve initial boundary value problems governed by the balance laws of mass, linear
momentum and energy,
+ div x = 0,
x div T = b,
+ div q T grad x = r,
(6.4)
when the external supplies b and r are given.
The governing eld equations are obtained, for the determination of the elds of
the density (X, t), the motion
(X, t), and the temperature (X, t), after introducing
the constitutive relations for T, , and q, into the balance laws (6.4). Any solution
(X, t),
q, s =
1
r. (6.5)
The resulting entropy inequality is called the ClausiusDuhem inequality,
+ div
q
0. (6.6)
Exploitation of entropy principle based on the ClausiusDuhem inequality has been widely
adopted in the development of modern continuum thermodynamics following the simple
ColemanNoll procedure. The main assumptions (6.5) while seem to be plausible in
all classical theories of continuum mechanics, are not particularly well motivated for
materials in general. In fact, the relation (6.5) is known to be inconsistent with the
results from the kinetic theory of ideal gases and is also found to be inappropriate to
account for thermodynamics of diusion. Therefore, we shall not abide this assumptions
70 6. THERMODYNAMIC CONSIDERATIONS
in the present considerations. Exploitation of the entropy principle based on the entropy
inequality in its general form (6.3) has been proposed by M uller [15, 16] and the method
of Lagrange multipliers proposed by Liu [10] greatly facilitates its procedure.
6.3 Thermodynamics of elastic materials
We shall now exploit the entropy principle for thermoelastic materials, following the
M ullerLiu procedure. Although the results will not be dierent from those following the
much simpler ColemanNoll procedure in this case, it is a typical example to illustrate
the method of Lagrange multipliers for the exploitation the thermodynamic restrictions
in general.
First of all, M uller proposed that if the body is free of external supplies, the en
tropy supply must also vanish, which is certainly much weaker than the assumptions
(6.5). Since constitutive relations do not depend on the external supplies, in exploiting
thermodynamic restrictions it suces to consider only supplyfree bodies.
It is more convenience to use referential description for elastic bodies. For supplyfree
bodies, from (2.36) we have the following balance laws,
= [J[
1
x Div T
= 0,
+ Div q
F = 0,
(6.7)
and the entropy inequality,
+ Div
0, (6.8)
where T
= JTF
T
and q
= JF
1
q (see (2.37)) are the PiolaKirchho stress tensor
and the material heat ux respectively. In analogy to the material heat ux, the material
entropy ux
is similarly dened.
The constitutive relations for thermoelastic materials (5.24) can be written as
T
=
T(F, , g
),
q
=
q(F, , g
),
= (F, , g
),
= (F, , g
),
=
(F, , g
).
(6.9)
where g
= and F =
, q
+ Div
v
(
x Div T
+ Div q
F) 0 (6.10)
is valid under no constraints, i.e., it must hold for any elds
a
S
a
X
a
+ 0, a = 1, , 40. (6.11)
where S
a
and are functions of (, ,
), while X
a
= (
,
,
,
, (), (
)).
Here, we have use the notation, F = (
), as well as (
,
,
,
, (), (
v
, therefore, we
conclude that
v
= 0,
and the inequality (6.10) becomes
) + (Div
Div q
) +
F 0. (6.12)
Since both and as well as
) = H
+ H
g
+H
F
F,
where H
, H
g
and H
F
are functions of (F, , ). The linearity of the inequality (6.12)
in (
,
F) then leads to
H
= 0, H
g
= 0, H
F
=
. (6.13)
The inequality (6.12) now reduces to Div
Div q
_
+ G () + M (
) 0, (6.14)
1
For general information on the applicability of the method of Lagrange multipliers, please refer to
[10] or Chap. 7 of [14].
72 6. THERMODYNAMIC CONSIDERATIONS
where in component forms,
G
,
M
i
i
,
i
,
.
By the linearity of (6.14) in () and (
+ G
= 0, M
i
+ M
i
= 0. (6.15)
In the remaining of this section, we shall prove further that M
i
= 0. In order to do
this, we need to invoke the condition of material objectivity (see (3.35)) of the heat ux,
which for elastic materials, can be expressed as
q(QF, , Qg) = Qq(F, , g) Q O(V ),
where g = grad = F
T
g
by (1.22). Let
q(F, , g) = q(F, , g
), (6.16)
then it is an easy exercise to show that
q(QF, , g
) = Q q(F, , g
) Q O(V ).
Taking Q = R
T
, with R being the rotational part of the polar decomposition F = RU,
we obtain
q(F, , g
) = R q(U, , g
). (6.17)
By the denition of material heat ux and using the right CauchyGreen tensor C = F
T
F,
the constitutive relation for the material heat ux (and similarly for material entropy ux)
can be written as
q
=
q
(C, , g
),
(C, , g
), (6.18)
where C = F
T
F is the right CauchyGreen tensor (see e.g. [14, 16, 22]).
On the other hand, since C = F
T
F, we have, for any tensor A,
F
[A] =
C
[A
T
F + F
T
A] = 2F
C
[A],
because C is symmetric and accordingly the gradient must be symmetrized. We shall
write the gradient after being symmetrized in components simply as
q
for
1
2
_
q
+
q
_
,
and hence
q
F
i
= 2 F
i
.
Likewise, similar relations are valid for the material entropy ux
.
3. THERMODYNAMICS OF ELASTIC MATERIALS 73
Let
,
then the relation (6.15)
2
, M
i
+ M
i
= 0, becomes
F
i
(
M
+
M
) = 0.
Since F is nonsingular, it implies that
M
=
M
, it follows that
M
=
M
. Therefore, we have
=
M
=
M
=
M
=
M
=
M
=
M
,
and hence
M
= 0, i.e.,
C
= 0 or
F
= 0. (6.19)
Summary of thermodynamic restrictions
The restrictions imposed by the entropy principle on elastic materials in the above ex
ploitation can be summarized below.
The inequality of (6.10) has been reduced successively to (6.12) and (6.14), and nally
to the remaining one, which gives the entropy production density as
=
_
_
g
0. (6.20)
The material entropy ux and heat ux must satisfy the relations (6.15)
1
and (6.19)
2
,
which can now be written as
_
_
sym
=
_
sym
,
F
=
F
, (6.21)
where (A)
sym
denotes the symmetric part of the tensor A. Finally, the relations (6.13)
can be summarized in the following dierential expression, similar to the Gibbs relation
in classical thermodynamics,
d =
_
d
1
dF
_
. (6.22)
The relations (6.20), (6.21) and (6.22) are the thermodynamic restrictions for elastic
bodies in general. Note that these relations contain one remaining Lagrange multiplier
= (1/)q
= K(F, ) g
= P(F, ) g
. (6.23)
We shall further assume that the thermal conductivity tensor K and the tensor P be
symmetric.
It follows from the relation (6.21)
1
that P =
. (6.24)
Taking the gradient of (6.24) with respect to g
= tr
_
_
= 0,
from which it implies that
must be independent of g
, since q
= (),
and by (6.23)
1
, the entropy production density (6.20) becomes
=
_
_
(g
Kg
) 0.
Since the entropy production does not vanish identically in heat conducting bodies, we
require that
() =
1
. (6.25)
Therefore, we have the following Gibbs relation,
d =
1
_
d
1
dF
_
, (6.26)
from which we obtain
=
1
,
g
=
1
.
The integrability condition, i.e., by taking the mixed partial derivative with respect to
and g
from the above two relations, implies that and hence are independent of g
.
By the use of the free energy function (F, ) dened by
= ,
4. ELASTIC MATERIALS OF FOURIER TYPE 75
the Gibbs relation (6.26) can be rewritten as
d =
1
dF d,
which implies that the following constitutive relations are determined completely by a
single scalar function (F, ):
T
F
, =
, =
.
Finally, the remaining entropy production inequality takes the form,
=
1
2
(q
) =
1
2
(g
Kg
) 0,
which implies that the thermal conductivity K(F, ) is a positive semidenite tensor
(see (5.34)).
Changing the PiolaKirchho stress tensor T
_
d
1
TF
T
dF
_
, (6.27)
and summarize the above general thermodynamic restrictions in the following proposi
tion:
Proposition. The stress tensor, the internal energy and the entropy are related to the
free energy function = (F, ) by
T =
F
F
T
, =
, =
, (6.28)
and the heat ux satises the Fourier inequality,
q(F, , g) g 0. (6.29)
Moreover, the following entropy ux relation is valid,
=
1
q.
An elastic material is called hyperelastic if there exists a function (F), called a stored
energy function, such that the stress tensor is given by
T =
F
F
T
or T
F
. (6.30)
From the relation (6.28)
1
, it follows that in an isothermal process, a thermoelastic ma
terial is a hyperelastic material, with free energy function served as the stored energy
function. In particular, for the linear theory considered in Section 5.2, the relation (6.30)
2
reduces to (5.10).
76 6. THERMODYNAMIC CONSIDERATIONS
6.5 Isotropic elastic materials
In the previous section, we have seen that further evaluations from the general restrictions
(6.20), (6.21) and (6.22) rely essentially on the ux relation (6.24), namely,
,
which can easily be veried for elastic materials of Fourier type. Now, we shall prove
that the ux relation is also valid for isotropic elastic materials.
For isotropic materials, the material symmetric group is the full orthogonal group, in
particular, from (3.39) for the heat ux q = q(F, , g), we have
q(FG, , g) = q(F, , g) G O(V ),
or equivalently from (6.16) with g
= F
T
g and Q = G
T
,
q(FQ
T
, , Qg
) = q(F, , g
) Q O(V ),
which combined with the material objectivity condition (6.17) by use of the polar decom
position, FQ
T
= (RQ
T
)(QUQ
T
), leads to the following condition for material symmetry
of isotropic elastic materials,
q(QUQ
T
, , Qg
) = Q q(U, , g
) Q O(V ).
In terms of material ux (6.18), the material symmetry condition for the material uxes
q
and
can be expressed as
(QCQ
T
, , Qg
) = Q
q
(C, , g
),
(QCQ
T
, , Qg
) = Q
(C, , g
),
Q O. (6.31)
In other words, they are isotropic vectorvalued functions of (C, , g
).
An isotropic function is restricted in its dependence on the independent variables as
we have seen in Sect. 4.2. Here, instead of using the explicit representation formula given
in (4.14), we shall consider restrictions in the form of dierential equations. We need the
following lemma:
Lemma. Let T(A) be a scalarvalued function of a tensor variable and suppose that
T(Q) = 0 for any orthogonal Q O. Then the gradient of T(A) at the identity tensor
is symmetric, i.e., for any skew symmetric tensor W,
A
T(1)[W] = 0. (6.32)
Proof. For any skew symmetric tensor W, we can dene
2
a timedependent orthogonal
tensor Q(t) = exp tW. Therefore, T(Q(t)) = 0. By taking the time derivative, we obtain
0 =
d
dt
T(Q(t)) =
A
T(exp tW)
_
d
dt
exp tW
_
=
A
T(exp tW)[W exp tW],
2
Let A L(V ), the exponential of A is dened as
exp A = 1 +A+
A
2
2!
+
A
3
3!
+ =
n=0
A
n
n!
, exp A L(V ).
It follows that (exp A)
T
= exp A
T
and if AB = BA then (exp A)(exp B) = exp(A+B).
5. ISOTROPIC ELASTIC MATERIALS 77
which, at t = 0, proves the lemma. .
For the material heat ux q
=
q
(C, , g
(ACA
T
, , Ag
) A
q
(C, , g
)).
Then from the condition (6.31), it follows that T(Q) = 0 for any Q O and hence the
above Lemma implies that
A
T(1)[W] = a
C
[WC CW] +a
Wg
a W
= 0.
In analogy, we have a similar equation for the material entropy ux
(C, , g
).
Now, let k =
and
q
, and the
relation (6.19)
1
, we have
a Wk = a
GWg
, (6.33)
where
G =
.
Since it holds for any vector a and any skew symmetric tensor W, in particular, by taking
alternatively,
a = e
1
, W = e
2
e
3
e
3
e
2
,
a = e
2
, W = e
3
e
1
e
1
e
3
,
a = e
3
, W = e
1
e
2
e
2
e
1
,
and knowing from (6.15)
1
that
G is skew symmetric, we obtain
G
12
g
3
+
G
31
g
2
= 0,
G
23
g
1
+
G
12
g
3
= 0,
G
31
g
2
+
G
23
g
1
= 0.
Summing of the last three equations gives
G
12
g
3
+
G
23
g
1
+
G
31
g
2
= 0, which implies
immediately that
G
12
g
3
= 0,
G
23
g
1
= 0,
G
31
g
2
= 0.
Since g
= (g
1
, g
2
, g
3
) does not vanish in general, it follows that
G
12
=
G
23
=
G
31
= 0
or the skew symmetric tensor
G = 0. Therefore, by (6.33), since W is arbitrary, k must
vanish and we have proved the ux relation
3
,
().
The second relation is a consequence of (6.19) and G = 0. .
3
For similar ux relations involving more than one vector and one tensor variables, see [13].
78 6. THERMODYNAMIC CONSIDERATIONS
The remaining evaluations can be done in the same manner given in the previous
section. In particular, we can also identify the Lagrange multiplier
as the reciprocal
of the absolute temperature,
=
1
,
and the general results given in the Proposition (on page 75) as well as the Gibbs relation
(6.27) are valid for isotropic elastic materials in general, except that the heat ux q
is
not restricted to the case of Fourier type, instead, it is an isotropic vectorvalued function
of (C, , g
).
Remark Suppose the entropy ux relation
is valid for each of the two bodies I and II in thermal contact, then at the contact surface,
the wall, the following conditions hold,
q
I
= q
II
,
I
=
II
,
where n
is the unit normal at the wall. Since from the entropy ux relation,
=
I
q
I
,
II
=
II
q
II
,
it follows that the Lagrange multiplier
I
=
II
.
This continuity property duely suggests that the Lagrange multiplier
be interpreted
as a measure of thermodynamic temperature. It is referred to as the coldness by M uller
[16].
Elastic uids
An elastic uid (see (4.21)) is an isotropic material, which depends on the deformation
gradient F only through its dependence on det F, or equivalently on the density .
Consequently, the relation (6.27) reduces to the wellknown Gibbs relation for uids in
classical thermostatics:
d =
1
_
d
p
2
d
_
, (6.34)
and
T = T(, ), q = q(, , g), = (, ).
From the representation formula for isotropic functions (as a special case of (4.14)), we
have
T = p(, ) 1, q = (, ) g,
where p is the pressure and the (scalar) thermal conductivity is nonnegative.
6. THERMODYNAMIC STABILITY 79
6.6 Thermodynamic stability
Another important concept associated with the entropy inequality is the thermodynamic
stability of a material body. To illustrate the basic ideas let us consider a supplyfree
body occupying a region 1 with a xed adiabatic boundary. We have
v = 0, q n = 0, n = 0 on 1,
and hence the entropy inequality (6.2) and the energy balance (2.31) become
d
dt
_
V
dv 0,
d
dt
_
V
( +
1
2
v v) dv = 0. (6.35)
In other words, the total entropy must increase in time while the total energy remains
constant for a body with xed adiabatic boundary. Statements of this kind are usually
called stability criteria as we shall explain in the following example.
Stable equilibrium state. We say that an equilibrium state is stable if any small
disturbance away from it will eventually disappear and thus the original state will be
restored.
Suppose that the region 1 is occupied by an elastic uid in an equilibrium state at
rest with constant mass density
o
and internal energy density
o
. Now let us consider a
small disturbance from the equilibrium state at the initial time such that
(x, 0) = (x), (x, 0) = (x), v(x, 0) = 0,
and [
o
[ and [
o
[ are small quantities. If we assume that the original state is
stable then the perturbed state will eventually return to the original state at later time.
Therefore since the total entropy must increase we conclude that
_
V
o
dv
_
V
dv, (6.36)
where
o
= (
o
,
o
) and = ( , ) are the nal equilibrium entropy and the perturbed
initial entropy. Expanding in Taylor series around the equilibrium state, we obtain
from (6.36)
_
V
_
o
(
o
) +
o
(
o
) +
o
(
o
) +
1
2
o
(
o
)
2
+
o
(
o
)(
o
) +
1
2
o
(
o
)
2
_
dv + o(3) 0.
(6.37)
Since total mass and total energy remain constant, we have
_
V
(
o
) dv = 0,
_
V
(
o
o
) dv = 0,
80 6. THERMODYNAMIC CONSIDERATIONS
and hence
_
V
(
o
) dv =
_
V
(
o
o
) dv
o
_
V
(
o
) dv = 0,
_
V
(
o
) dv =
_
V
(
o
)
2
dv =
_
V
o
(
o
)
2
dv + o(3).
Therefore, up to the second order terms (6.37) becomes
_
V
_
1
2
o
(
o
)
2
+
2
o
(
o
)(
o
) +
_
1
2
2
+
1
o
(
o
)
2
_
dv 0.
By the mean value theorem for integrals, it reduces to
_
o
(
o
)
2
+ 2
2
o
(
o
)(
o
) +
_
2
+
2
o
(
o
)
2
_
V
2
0,
where V is the volume of the region 1 while
= (x
) and
= (x
in 1.
Since the nonpositiveness of the above quadratic form must hold for any small dis
turbance and
o
and
o
are arbitrary, it follows that the matrix
_
2
+
2
_
must be negative semidenite, or equivalently
2
0,
2
2
+
2
0, (6.38)
2
_
2
+
2
_
2
_
2
0. (6.39)
In these relations partial derivatives are taken with respect to the variables (, ).
In order to give more suggestive meanings to the above conditions for stability, we
shall reiterate them in terms of the independent variables (, ). In making this change
of variables we must admit the invertibility of (, ) with respect to the temperature,
i.e. / ,= 0. To avoid confusions variables held constant in partial dierentiations
will be indicated. Two relations frequently used in thermostatics for change of variables
are given below.
Lemma. Let u, v, and w be three variables and there is a relation between them so that
we have the functional relations: u = u(v, w), v = v(u, w), and w = w(u, v). Then
u
v
w
v
u
w
= 1,
u
v
w
v
w
u
w
u
v
= 1. (6.40)
6. THERMODYNAMIC STABILITY 81
The proof follows immediately by writing u = u(v(u, w), w) = f(u, w) and computing
the partial derivatives of f(u, w) with respect to u and w respectively.
From the Gibbs relation (6.34), we have
=
1
=
p
2
,
which further imply the integrability condition for ,
_
1
=
1
_
p
.
By the use of (6.40) and the above relations, we then obtain
=
1
=
1
+
2
=
1
.
Therefore, in terms of the variables (, ), the stability condition (6.38)
1
reduces to
> 0, (6.41)
since we have already admitted that / ,= 0. The condition (6.39) after simplications
then reduces to
p
0, (6.42)
The remaining condition (6.38)
2
is merely a consequence of (6.42). Obviously, these
conditions are restrictions on the constitutive functions (, ) and p(, ).
We have shown that the stability of equilibrium requires that: (1) the specic heat
at constant volume (equivalently at constant density), c
v
= /, must be positive (see
(5.34)); (2) the isothermal compressibility,
T
= ( p/)
1
, must be nonnegative.
The condition (6.42) can also be expressed in a dierent form. Let = 1/ be the
specic volume, the volume per unit mass. Then the Gibbs relation (6.34) can be written
as
d = d p d.
and the condition (6.42) implies
0.
Therefore, the stability of equilibrium requires that the free energy be a concaveupward
function of the specic volume, a wellknown result for stability. .
82 6. THERMODYNAMIC CONSIDERATIONS
To establish a stability criterion for a material system under a dierent condition, one
may try to nd a decreasing function of time /(t) from the balance laws and the entropy
inequality in integral forms. Such a function is called the availability of the system, since
it is the quantity available to the system for its expense in the course toward equilibrium.
Such a function is also known as a Liapounov function in the stability theory of dynamic
systems (see [9]). In the present example, from (6.35)
1
one may dene the availability /
of the system as
/(t) =
_
V
dv,
d/
dt
0.
As a second example, we consider a supplyfree body with a xed isothermal boundary,
v = 0, =
o
on 1,
and assume that the relation = q/ holds. Then the energy balance (2.31) and the
entropy inequality (6.2) lead to
d
dt
_
V
( +
1
2
v v) dv +
_
V
q nda = 0,
d
dt
_
V
dv +
1
o
_
V
q nda 0.
Elimination of the terms containing surface integrals from above, gives
d/
dt
0, /(t) =
_
V
(
o
+
1
2
v v) dv. (6.43)
In this manner we have found a decreasing function of time, the availability /(t), which
characterizes the stability for this system. Note that
_
V
(
o
) dv
is the total free energy if =
o
throughout the body. Therefore, it follows that for a
body with constant uniform temperature in a xed region the availability / reduces to
the sum of the free energy and the kinetic energy.
Summarizing the above two situations, we can state the following criteria for the
stability of equilibrium.
Criteria of thermodynamic stability.
1) For a body with xed adiabatic boundary and constant energy, the entropy tends to
a maximum in equilibrium.
2) For a body with xed boundary and constant uniform temperature, the sum of the
free energy and the kinetic energy tends to a minimum in equilibrium.
We have seen in this section that thermodynamic stability criteria, like the entropy
principle, impose further restrictions on properties of the constitutive functions, namely,
specic heat and compressibility must be positive. On the other hand, such criteria,
besides being used in analyzing stability of solutions, they are the basic principles for
the formulation of equilibrium solutions in terms of minimization (or maximization)
problems.
Bibliography
[1] Boehler, J. P.: On irreducible representations for isotropic scalar functions, ZAMM
57, 323327 (1977).
[2] Coleman, B. D., Markovitz, H., Noll, W.: Viscometric Flows of NonNewtonian
Fluids, springer (1966).
[3] Coleman, B. D., Mizel, V. J.: On the general theory of fading memory, Arch. Ra
tional Mech. Anal. 29, 1831 (1968).
[4] Coleman, B. D., Noll, W.: Foundations of linear viscoelasticity, Rev. Modern Phys.
33, 239249 (1961).
[5] Coleman, B. D., Noll, W.: The thermodynamics of elastic materials with heat con
duction and viscosity, Arch. Rational Mech. Anal. 13, 167178 (1963).
[6] Gurtin, M. E., Martins, L. C.: Cauchys theorem in classical physics, Arch. Rational
Mech. Anal. 60, 305324 (1976).
[7] Gurtin, M. E., Mizel, V. J., Williams, W. O.: A note on Cauchys stress theorem,
J. Math. Anal. Appl. 22, 398401 (1968).
[8] Gurtin, M. E., Williams, W. O.: On the inclusion of the complete symmetry group
in the unimodular group, Arch. Rational Mech. Anal. 23, 163172 (1966).
[9] Knops, R. J., Wilkes, E. W.: Elastic Stability, Handbuch der Physik Vol. VIa/3, ed.
by S. Fl ugge, Springer, Berlin (1973).
[10] Liu, IShih: Method of Lagrange multipliers for exploitation of the entropy principle,
Arch. Rational Mech. Anal., 46, 131148 (1972).
[11] Liu, IShih: On the requirement that material symmetries shall preserve density,
Arch. Rational Mech. Anal. 68, 1926 (1978).
[12] Liu, IShih: On representations of anisotropic invariants, Int. J. Engng Sci. 20,
10991109 (1982).
[13] Liu, IShih: On entropy uxheat ux relation in thermodynamics with Lagrange
multipliers, Continuum Mech. thermody. 8, 247256 (1996).
84 BIBLIOGRAPHY
[14] Liu, IShih: Continuum Mechanics, SpringerVerlag, BerlinHeidelberg (2002).
[15] M uller, I.: On the entropy inequality, Arch. Rational Mech. Anal. 26, 118141 (1967).
[16] M uller, I.: The coldness, a universal function in thermoelastic bodies, Arch. Rational
Mech. Anal. 41, 319332 (1971).
[17] Noll, W.: Proof of the maximality of the orthogonal group in the unimodular group,
Arch. Rational Mech. Anal. 18, 100102 (1965).
[18] Noll, W.: A mathematical theory of the mechanical behavior of continuous media,
Arch. Rational Mech. Anal. 2, 197226 (1968).
[19] Smith, G. F.: On isotropic functions of symmetric tensors, skewsymmetric tensors
and vectors, Int. J. Engng Sci. 9, 899916 (1971).
[20] Treloar, L. R. G.: The Physics of Rubber Elasticity, Clarendon Press, Oxford (1975).
[21] Truesdell, C.: A First Course in Rational Continuum Mechanics, Vol. 1, Second
Edition, Academic Press, Boston (1991).
[22] Truesdell, C., Noll, W.: The NonLinear Field Theories of Mechanics, Handbuch
der Physik, Ed. by S. Fl ugge, Vol. III/3, Springer (1965).
[23] Wang, C.C.: Stress relaxation and the principle of fading memory, Arch. Rational
Mech. Anal. 18, 117126 (1965).
[24] Wang, C.C.: A new representation theorem for isotropic functions, Part I and II,
Arch. Rational Mech. Anal. 36, 166197, 198223 (1970); Corrigendum, ibid. 43,
392395 (1971).
[25] Wang, C.C.: Mathematical Principles of Mechanics and Electromagnetism, Part A:
Analytical and Continuum Mechanics, Plenum Press, New York and London (1979).
[26] Wang, C.C., Truesdell, C.: Introduction to Rational Elasticity, Noordho Interna
tional Publishing, Leyden (1973).
Index
Acceleration, 7
Anisotropic elastic solid, 54
Anisotropic function, 54
Availability function, 82
Balance laws, 18, 19
Balance of
internal energy, 18
linear momentum, 17
mass, 15
Body force, 16
Boundary conditions, 19
Cauchys theorem, 17
CauchyGreen stain tensor, 4
CayleyHamilton theorem, 42
ClausiusDuhem inequality, 69
Condition of
Euclidean objectivity, 31, 33, 35
forminvariance, 33, 35
material objectivity, 32, 33, 72
Conguration, 2, 22
reference, 2, 25
Conservation of
angular momentum, 17
energy, 18
linear momentum, 17
mass, 15
Constitutive function, 29
Convected time derivative, 28
Corotational time derivative, 28
Deformation, 3
incompressible, 6
relative, 9
Deformation gradient, 3
relative, 9
transformation property, 26
Description
material, 32
referential, 7, 33
spatial, 7
Displacement, 5
gradient, 5
Elastic uid, 47, 78
Elastic material, 53
incompressible, 58
isotropic, 76
thermodynamics of, 70
Elastic solid
anisotropic, 54
isotropic, 56
Elasticity
nite, 58
linear, 54
tensor, 55
Entropy principle, 69
Equation
heat conduction, 62
NavierStokes, 47
of motion, 17
RankineHugoniot, 14
Euclidean objectivity, 30
Euclidean transformation, 23
Eulers law of motion, 16
External supply, 11
Fading memory, 62
Fluid, 37
elastic, 47, 78
incompressible, 49
NavierStokes, 46
NavierStokesFourier, 48
86 INDEX
Newtonian, 46
ReinerRivlin, 46
Stokes, 47
viscous heatconducting, 48
Fluid crystal, 37
Flux, 11
energy, 18
Forminvariance, 32
Fouriers law, 48, 60
Frame of reference, 21
change of, 23
Galilean transformation, 25
General balance equation
at a regular point, 13
at a singular point, 14
in integral form, 11
in reference coordinates, 14
General jump condition, 14
in reference coordinates, 14
Gibbs relation, 81
elasitc uid, 78
elastic solid, 75
Heat ux, 18
material, 19
Hookes law, 55
Hyperelastic material, 55, 59, 75
Incompressibility, 48
Incompressible elastic material, 58
Incompressible NavierStokesFourier uid,
49
Invariant
anisotropic, 54
isotropic, 42
principal, 42
Isothermal compressibility, 81
Isotropic elastic solid, 56
incompressible, 58
Isotropic function, 42
Isotropic invariant, 42
Isotropic material, 37
Isotropic thermoelastic solid, 59
Jump condition, 18, 19
Jump discontinuity, 12
Lagrange multiplier, 74, 78
Method of, 71
Lame elastic moduli, 57
Law
Eulers, 16
Fouriers, 48, 60
Hookes, 55
Linear elasticity, 54
isotropic, 57
Linear thermoelasticity
anisotropic, 61
isotropic, 61
Linear viscoelasticity, 64
Local speed of propagation, 14
Material
elastic, 53
hyperelastic, 75
incompressible elastic, 58
isotropic, 37
MooneyRivlin, 58
neoHookean, 58
of grade n, 41
simple, 34
thermoelastic, 59
with fading memory, 64
Material heat ux, 19
Material symmetry, 35
group, 36
transformation, 36
Material time derivative, 8
MooneyRivlin material, 58
Motion, 7
incompressible, 16, 48
NavierStokes equation, 47
NavierStokes uid, 46
NavierStokesFourier uid, 48
NeoHookean material, 58
Objective tensor, 24
Observer, 21
Change of, 23
PiolaKirchho stress tensor, 19
INDEX 87
Polar decomposition theorem, 4
Principal invariants, 42
Principal stretch, 4
Principle of
determinism, 29
entropy, 69
material frameindierence, 31, 32
material objectivity, 32
RankineHugoniot equation, 14
Rate of strain tensor, 10
Reference conguration, 2
RivlinEricksen tensor, 41
Rotation tensor, 4
innitesimal, 5
Simple uid, 37
Simple material, 34
Simple solid, 37
Singular surface, 12
Solid, 37
anisotropic elastic, 54
incompressible isotropic, 58
isotropic elastic, 56
isotropic thermoelastic, 59
Specic heat, 60
at constant volume, 81
Spin tensor, 10
Stability
of equilibrium, 79
thermodynamic, 79
thermodynamic criteria, 82
Stokes uid, 47
Stored energy function, 55, 75
Strain tensor
CauchyGreen, 4
innitesimal, 5
rate of, 10
Stress tensor
Cauchy, 17
PiolaKirchho, 19
Stretch tensor, 4
Supplyfree bodies, 70
Surface traction, 16
Theorem
Cauchy, 17
CayleyHamilton, 42
polar decomposition, 4
transport, 11
Thermal conductivity, 48
tensor, 60
Thermodynamic stability, 79
Thermoelastic material, 59
Time derivative
convected, 28
corotational, 28
Transport theorem, 11
Velocity, 7
gradient, 8
Viscoelasticity
linear, 64
Viscosity
bulk, 46
shear, 46
Viscous heatconducting uid, 48
Vorticity vector, 10