Lecture Notes on Non-Newtonian Fluids

Part I: Inelastic Fluids

Professor A. M. Robertson
Department of Mechanical Engineering
University of Pittsburgh
Fall 2005
1
2
Contents
1 Background Material on Continuum Mechanics 5
1.1 Description of Motion of Material Points in a Body . . . . . . . . . . . . . . . . . . . 5
1.2 Referential and Spatial Descriptions . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.3 Velocity and Acceleration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.4 Material Derivative . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.5 Deformation Gradient of the Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.6 Deformation Measures. Strain Measures . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.6.1 Right Cauchy Green Deformation Tensor . . . . . . . . . . . . . . . . . . . . 10
1.6.2 Relative Strain Tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.7 Velocity Gradient . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.7.1 Vorticity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.8 Special Motions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.8.1 Rigid Motions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.8.2 Isochoric Motions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.8.3 Irrotational Motions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.8.4 Simple Shear . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.8.5 Uniaxial Extension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.9 Governing Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.9.1 Conservation of Mass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.9.2 Balance of Linear Momentum . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
1.9.3 Balance of Angular Momentum . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.9.4 Mechanical Energy Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.10 Invariance Under Superposed Rigid Body Motions . . . . . . . . . . . . . . . . . . . 22
1.10.1 Relationship between kinematic variables in the original and superposed mo-
tions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
1.10.2 Requirement: Invariance under a superposed rigid body motion . . . . . . . . 28
2 Non-Linear Viscous Fluids 31
2.1 Constitutive Assumption T = T(, L) . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.1.1 Incompressible Inelastic Fluids . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.2 Reiner-Rivlin Fluids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.2.1 Thermodynamic Restrictions on
1
and
2
for Reiner-Rivlin Fluids . . . . . . 35
2.2.2 Restrictions on
1
,
2
based on behavior of real uids in simple shear . . . . 36
3
2.3 Generalized Newtonian Fluids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.3.1 Power-Law Fluid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.3.2 Prandtl-Eyring Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
2.3.3 Powell-Eyring Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
2.3.4 Cross Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
2.3.5 Yasuda Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
2.3.6 Bingham Fluids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
2.3.7 Modied Bingham Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
2.4 Steady, Fully Developed Flow of a Generalized Newtonian Fluid in a Straight Pipe . 45
2.4.1 Power-Law Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
2.4.2 Bingham Fluid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.4.3 Q as a function of and (S) for unspecied generalized Newtonian uid . . 48
2.4.4 Determination of (S) from measured values of driving pressure drop and
ow rate in a straight pipe . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
2.5 Cone and Plate Flow for a Generalized Newtonian Fluid . . . . . . . . . . . . . . . . 50
2.5.1 Small Angle Approximation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
2.5.2 Finite Cone Angle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
A The Relationship between Coordinate Free Notation and Indicial Notation 61
B Isotropic Tensors 63
C Isotropic Tensor Functions 67
C.1 Scalar valued isotropic tensor functions . . . . . . . . . . . . . . . . . . . . . . . . . . 67
C.2 Symmetric Isotropic Tensor Functions . . . . . . . . . . . . . . . . . . . . . . . . . . 68
D Material Line Elements, Area Elements, Volume Elements 69
E Rectangular Coordinates 71
F Cylindrical Coordinates 75
G Spherical Coordinates 79
4
Chapter 1
Background Material on Continuum
Mechanics
1.1 Description of Motion of Material Points in a Body
We will rst consider a physical body, for example a body of uid, which we identify by the symbol
B. We use the symbol Y to identify an arbitrary material particle in the body. The motion of
arbitrary material particles Y can then be described through a function,
x = (Y, t) (1.1)
where x is the vector identifying the spatial position at time t of the material particle identied
by Y . In continuum mechanics, we are interested in describing the position in time of an uncount-
able number of points, which cannot be identied using letters or numbers. With this in mind,
we identify material points in the body by their position in some chosen reference conguration.
Namely, we assume the body is smooth and can be embedded in a three dimensional Euclidean
space. The mapping of each material particle of the body to Euclidean 3-space at an arbitrary time
t will be called the conguration of B at time t, denoted by (t). This mapping is assumed to be
one-to-one, invertible and dierentiable as many times as necessary for all time. The conguration,

0
, is a special conguration, chosen for the purpose of identifying material particles. For example,

0
could be the conguration of the body at time zero.
We denote the region occupied by the entire body in this conguration as R
0
with closed
boundary R
0
. An arbitrary material region within B in the conguration
0
will be denoted as
V
0
(V
0
R
0
) with boundary V
0
, Fig. 1.1. We denote the corresponding volume in the current
conguration (t) as V (V R) with boundary V.
An arbitrary material point identied by Y may also be identied by its position, denoted by
X, in the reference conguration
0
. It is not necessary that the conguration of the body ever
have coincided with the reference conguration. The motion of material points in the body B may
then be described through the relationship,
x =

0
(X, t). (1.2)
5
Figure 1.1: Schematic of notation used to identify regions of an arbitrary body B in the reference
and current congurations .
Figure 1.2: Schematic of position vector in reference and current congurations for a xed material
point of an arbitrary body B.
6
The vector x is the position at time t of an arbitrary material particle will then be identied by
its position X in reference conguration . We will assume that the transformation (1.2) possesses
an inverse,
X =
1

0
(x, t). (1.3)
For our current purposes it suces to dene one reference conguration for the body, so in further
discussions we will drop the subscript
0
and it will be understood that the functions (X, t) and

1
(x, t) depend on the choice of reference conguration. In the discussion of viscoelastic uids
with fading memory, we will use the current conguration as the reference conguration. The
relevant kinematics will be discussed later in the text.
1.2 Referential and Spatial Descriptions
Field variables such as density, , may be written as a function of X and t,
= (X, t). (1.4)
The function, (1.4) is called the referential or Lagrangian description of the density eld. Whereas,
(1.4) is sucient to describe the density of all material points in the body, it is frequently of interest
in classical uid mechanics to have knowledge of eld variables as a function of spatial positions x
and time,
= (x, t). (1.5)
The function in (1.5) is called the spatial or Eulerian description of density and is independent of
information about the position of individual material particles. Rather, the function given in (1.5)
gives the value of the density at a specied position in space at a given time t. The representations
(1.4) and (1.5) are related through,
= (x, t) = ((X, t), t) = (X, t). (1.6)
1.3 Velocity and Acceleration
The velocity of a material particle is dened as
v =
(X, t)
t
(1.7)
and the acceleration,
a =

2
(X, t)
t
2
. (1.8)
7
Figure 1.3: Relationship between Eulerian and Lagrangian representation of density
As for the density eld, the relationship between the Eulerian representation and Lagrangian rep-
resentation of the velocity eld can be obtained using (1.2),
v = v(x, t) = v((X, t), t) = v(X, t). (1.9)
1.4 Material Derivative
The material derivative of a eld variable such as density, is dened as the partial derivative of the
function with respect to time holding the material point xed,
D
Dt
=
(X, t)
t
. (1.10)
Frequently, in uid mechanics, the spatial description of the eld variables is of interest rather than
the material description. Using the chain rule, the material derivative can be written for the spatial
formulation of a eld variable. For example, in the case of the density eld,
D
Dt
=
(x, t)
t
+ v
i
(x, t)
x
i
. (1.11)
Using (1.11), we may obtain the acceleration (1.8) from the spatial representation of the velocity
eld,
a
i
=
v
i
(x, t)
t
+ v
j
v
i
(x, t)
x
j
. (1.12)
8
1.5 Deformation Gradient of the Motion
The deformation gradient or displacement gradient of the motion relative to the reference congu-
ration
0
is denoted by F. It is a second order tensor dened by,
F =
(X, t)
X
. (1.13)
The components of F are then,
F
iA
=

i
(X, t)
X
A
. (1.14)
When the deformation gradient is given as a second order tensor eld dened on R, then the
deformation eld can be determined from (1.14) which is a system of nine partial dierential
equations for three unknowns x
i
. In order that there exists a solution, the integrability condition,
F
iA
X
B
=
F
iB
X
A
(1.15)
must be satised. The deformation can then be determined to within an arbitrary constant, corre-
sponding to an arbitrary rotation,
x
i
=
_
F
iA
dX
A
. (1.16)
To understand the physical signicance of F it is helpful to consider the transformation of two
material points during the motion (1.2), which occupy positions X
o
and X in the reference cong-
uration and positions x
o
and x in the current conguration, respectively. Using a Taylors series
expansion, we see that,
x
i
= x
oi
+

i
(X, t)
X
A
|
X
o
(X
A
X
oA
) + higher order terms. (1.17)
or in index free notation,
x = x
o
+ F(X
o
, t) (X X
o
) + higher order terms. (1.18)
In which the distance between X and X
o
goes to zero,
dx = F dX, dx
i
= F
iA
dX
A
, (1.19)
where dx and dX are the innitesimal vectors (x x
o
) and (X X
o
), respectively in the limit of
innitesimal distance between points
9
1.6 Deformation Measures. Strain Measures
1.6.1 Right Cauchy Green Deformation Tensor
We now consider the change in magnitude dx during the deformation. Well denote the magnitude
of dx and dX as ds and dS, respectively. Let m and M be the unit vectors tangent to dx and dX,
respectively. The square of the magnitude of the innitesimal vector dx is related to the square of
the magnitude of dX through,
ds
2
= |dx|
2
= dx dx = dx
i
dx
i
= F
iA
F
iB
M
A
M
B
dS
2
= C
AB
(X, t) M
A
M
B
dS
2
,
(1.20)
where C
AB
are the components of the right Cauchy-Green deformation tensor:
C F
T
F and C
AB
F
iA
F
iB
. (1.21)
Note that C is symmetric.
Physical Signicance of Diagonal elements of C
Consider an innitesimal element represented by dX in the reference conguration and dx in the
current conguration. If we now consider the special case where dX = dSe
1
, then it follows from
(1.20), that
ds
2
dS
2
= C
11
. (1.22)
The ratio ds
2
/dS
2
is typically called the stretch ratio and denoted by . We can describe the result
(1.22) in words as,

C
11
is equal to the stretch ratio squared of an innitesimal material element
which was aligned with the e
1
axis in the reference conguration.
The other diagonal elements can be interpreted in a similar way.
Physical signicance of o-diagonal elements of C
Now consider two innitesimal material elements corresponding to dX
(1)
= dS
(1)
M
(1)
and dX
(2)
=
dS
(2)
M
(2)
, where M
(1)
and M
(2)
are unit vectors. In the current conguration, these innitesimal
elements are denoted as dx
(1)
= ds
(1)
m
(1)
and dx
(2)
= ds
(2)
m
(2)
, respectively. The angle between
these same material elements in the current conguration will be denoted as . Therefore,
cos =
dx
(1)
dx
(2)
ds
(1)
ds
(2)
=
F
iA
F
iB
M
(1)
A
M
(2)
B
dS
(1)
dS
(2)
ds
(1)
ds
(2)
=
C
AB
M
(1)
A
M
(2)
B

(1)

(2)
. (1.23)
10
Now consider the special case, where the innitesimal elements are perpendicular in the reference
conguration, for example, M
(1)
= e
1
and M
(2)
= e
2
, in which case, (1.23) reduces to,
cos =
C
12
_
C
11
C
22
. (1.24)
We can describe in words, the physical signicance of

C
12
provides information about the angle between two material elements
which were parallel to e
1
and e
2
in the reference conguration.
The other o-diagonal elements can be interpreted in a similar way. For example, if the angle
between the elements remains unchanged (still 90
o
), then C
12
= 0. If the angle decreases from 90
o
,
then C
12
will be a positive number. If the (90
o
, 270
o
), then C
12
will be negative.
Jacobian of the transformation
In future discussions, we denote the the Jacobian of the transformation (1.2) by J. Namely, the
Jacobian is equivalent to the determinant of F,
J det

i
(X, t)
X
A
= det (F) . (1.25)
Since (X, t) is assumed to have an inverse,
0 < J < . (1.26)
Recall from continuum mechanics that the relationship between the innitesimal volume occupied
by material points in the current conguration dv is related to the innitesimal volume occupied
by the same material points in the reference conguration dV is,
dv = J dV. (1.27)
It will be useful in the discussion of the governing equations to note the following result for the
material derivative of the Jacobian of the transformation (1.2),
DJ
Dt
= J divv, (1.28)
where v is the velocity vector.
11
1.6.2 Relative Strain Tensor
When the motion is rigid, C is equal to the identity tensor. It is useful to dene a relative measure
of strain which vanishes for rigid motions:
E
1
2
(C I), (1.29)
where E, is often referred to as the Lagrangian strain. A second measure of relative strain, e,
dened as
e
1
2
(I B
1
) (1.30)
is referred to as the Eulerian strain where B is the left Cauchy-Green tensor
B F F
T
and B
ij
F
iA
F
jA
. (1.31)
1.7 Velocity Gradient
We denote L as the gradient of the spatial form of the velocity vector, so that the components of
L with respect to rectangular coordinates are
1
L
ij
=
v
i
(x, t)
x
j
. (1.32)
Recall that any second order tensor can be decomposed into the sum of a symmetric and skew
symmetric second order tensor. We can represent L in this way,
L
ij
=
1
2
_
v
i
x
j
+
v
j
x
i
_
+
1
2
_
v
i
x
j

v
j
x
i
_
. (1.33)
We dene the rate of deformation tensor, D, as the symmetric part of L and the vorticity tensor,
W, as the skew-symmetric part of L,
D
ij
=
1
2
_
v
i
x
j
+
v
j
x
i
_
, W
ij
=
1
2
_
v
i
x
j

v
j
x
i
_
. (1.34)
Physical Signicance of Diagonal elements of D
The physical signicance of D can be studied by considering the rate of change in magnitude of an
innitesimal material element dx of length ds. We rst consider the rate of chance of ds
2
which is
equal to
D(ds
2
)
Dt
= 2 dx
i
Ddx
i
Dt
. (1.35)
1
Note that in some books, the gradient of a second order tensor is dened as the transpose of that used here.
12
The rate of change in the innitesimal material element dx is,
Ddx
i
Dt
=
D
Dt
F
iA
dX
A
=
F
iA
t
dX
A
=

2

i
(X, t)
X
A
t
dX
A
=
v
i
X
A
dX
A
=
v
i
x
j
F
jA
dX
A
= L
ij
dx
j
(1.36)
and so from (1.35) and (1.36) we nd,
D(ds
2
)
Dt
= 2 D
ij
dx
i
dx
j
. (1.37)
Thus,
Dds
Dt
=
D
ij
dx
i
dx
j
ds
. (1.38)
We see from (1.38) that the rate of change of magnitude of an innitesimal element which at time
t is parallel to the e
1
axis is
Dds
Dt
= D
11
ds. (1.39)

D
11
is the rate of change of ds divided by ds of a material element which at
time t is aligned with the e
1
axis.
The other diagonal elements can be interpreted in a similar way.
Physical Signicance of the o diagonal elements of D
Now consider two innitesimal material elements x and y which intersect at angle with lengths
|dx| and |dy| respectively. Then,
cos =
dx dy
|dx| |dy|
(1.40)
13
and therefore,
D
Dt
cos =
D
Dt
_
dx dy
|dx| |dy|
_
=
D
Dt
_
dx dy
_
1
|dx| |dy|

dx dy
|dx| |dy|
D
Dt
_
|dx| |dy|
_
=
D
ij
dx
i
dy
j
|dx| |dx|

dx dy
|dx| |dy|
D
Dt
_
|dx| |dy|
_
(1.41)
In particular, consider (1.41) for two innitesimal material elements, dx, dy which at time t are
parallel to the base vectors e
1
and e
2
. We see that the rate of change of the angle between these
two vectors at time t is
D
Dt
= D
12
. (1.42)
Therefore,

D
12
(x, t) is the rate of change of angle between the two innitesimal vectors which
at time t are located at position x and parallel to base vectors, e
1
and e
2
.
The other o-diagonal elements can be interpreted in a similar way. Notice that this interpretation
of the components of D does not require knowledge of the behavior of specic material elements.
Rather D(x, t) is related to the rate of change of material elements which at time t are located at
position x.
1.7.1 Vorticity
Recall the denition of the vorticity, is the curl of the velocity vector,
v, (1.43)
or in indicial notation
2
,

i

ijk
v
k
x
j
. (1.44)
The components of the vorticity vector are related to the components of the vorticity tensor through,

i
=
ijk
W
jk
, (1.45)
and
W
ij
=
1
2

ijk

k
. (1.46)
2
Note that alternate denitions of vorticity are sometimes used. For example, sometimes the vorticity is dened
as the negative of that given in (1.43). In other cases, the vorticity is taken to be twice that in (1.43).
14
1.8 Special Motions
1.8.1 Rigid Motions
A rigid motion is one in which the distance between points remains constant. Therefore, for rigid
motions, the material derivative of ds is zero for all points in the body for all time during the rigid
motion. We see from (1.38) that for rigid motions, D must be identically zero at all points in the
body for all time. Alternatively, we see from (1.38) that motions in which D is zero at all points
in the body for all time are rigid motions.
One can show that the most general rigid motion can be written as,
x = x
o
(t) + Q(t) X, (1.47)
where Q is a proper orthogonal second order tensor. Namely,
Q Q
T
= I, and therefore, Q
T
Q = I. (1.48)
In indicial notation,
Q
ij
Q
kj
=
ik
, and therefore, Q
ji
Q
jk
=
ik
. (1.49)
In preparation for considering the form of the velocity eld for a rigid body motion, we will rst
consider the time derivative of Q. If we take the derivative with respect to time of equation, (1.48),
we nd,
dQ
dt
Q
T
+ Q
dQ
T
dt
= 0, (1.50)
and so,
dQ
dt
Q
T
= (
dQ
dt
Q
T
)
T
(1.51)
We see from (1.51) that the quantity dQ/dt Q
T
is a skew symmetric tensor, which we denote as
, so
dQ
dt
= Q. (1.52)
Since is skew symmetric there exists a vector which we will denote as c, where,

ij
=
ijk
c
k
. (1.53)
The velocity eld for a rigid motion can be obtained by taking the material derivative of (1.47),
v = v
o
(t) + c (x x
o
), (1.54)
where v
o
is the time derivative of x
o
. If we take the curl of (1.54), we nd that,
= 2 c. (1.55)
Therefore, the velocity eld for the most general rigid body motion can be written as,
v = v
o
(t) +
1
2
(x x
o
). (1.56)
15
1.8.2 Isochoric Motions
Isochoric motions are those in which the volume occupied by xed material particles is unchanged
during the motion. A material does not have to be incompressible to undergo isochoric motions.
As will be discussed in the next section, all motions undergone by incompressible uids must be
isochoric. We see from (1.27) that if a motion is isochoric then the value of J is one throughout
the motion. In this case, we have from, (1.28) that the divergence of v is equal to zero,
v = 0 (1.57)
and hence the trace of the rate of deformation tensor is zero.
1.8.3 Irrotational Motions
Irrotational motions are those for which the vorticity is zero. From (1.46), the vorticity tensor W
is zero for irrotational motions.
1.8.4 Simple Shear
A ow eld that is of great signicance in uid mechanics is simple shear ow, sometimes called
lineal Couette ow. In this ow, the velocity eld is steady, fully developed and uni-directional.
The magnitude of the velocity depends linearly on the spatial component with axis perpendicular
to the direction of ow. For example, if rectangular coordinates x
i
are chosen such that the ow
directions is parallel to the x
1
axis, then the orientation and origen of the coordinate system can
be chosen such that the velocity eld can be written as,
v = x
2
e
1
. (1.58)
Clearly simple shear is an example of an isochoric motion. It is possible to show (see later chapters)
that simple shear ow can be generated between parallel plates for a wide class of uids called simple
uids (see, e.g. [8]). As such, it has great value in the design of rheometers. For example, consider
simple shear ow between two parallel plates separated by a distance h, when the upper plate is
moving with speed U in the e
1
direction and the origin of the coordinate system coincides with
a point on the bottom plate, Figure 1.4. Using the no-slip boundary conditions at the top and
bottom surface it can be shown that = U/h. Using the result (1.7) and the initial condition
x = X at t = 0, we can show that the motion corresponding to (1.58) is,
x
1
= X
1
+ X
2
t, x
2
= X
1
, x
3
= X
3
. (1.59)
It is helpful for discussions in later chapters, to calculate several kinematic variables for simple shear.
In particular, using the results (1.58) and (1.59) for simple shear, it follows that the corresponding
values of D, W are,
[D] =

2
_
_
0 1 0
1 0 0
0 0 0
_
_
, [W] =

2
_
_
0 1 0
1 0 0
0 0 0
_
_
. (1.60)
16
Figure 1.4: Schematic of velocity eld in simple shear ow.
while F, C and E are
[F] =
_
_
1 t 0
0 1 0
0 0 1
_
_
, [C] =
_
_
1 t 0
t 1 +
2
t
2
0
0 0 1
_
_
, [E] =
1
2
_
_
0 t 0
t
2
t
2
0
0 0 0
_
_
.
(1.61)
1.8.5 Uniaxial Extension
Another relatively simple ow eld which is of particular interest for viscoelastic uids is uniaxial
extension. The Eulerian representation for the velocity eld for this motion is,
v
1
= x
1
, v
2
=

2
x
2
, v
3
=

2
x
3
. (1.62)
As can be seen from (1.62), uniaxial extension is an isochoric motion. Using the result (1.7) and the
initial condition x = X at t = 0, the position of material particles in this motion can be described
through,
x
1
= X
1
e
t
, x
2
= X
2
e
t/2
, x
3
= X
3
e
t/2
. (1.63)
Using the results (1.62) and (1.63), the corresponding values of D, W are,
[D] =
_
_
1 0 0
1 1/2 0
0 0 1/2
_
_
, [W] =

2
_
_
0 0 0
0 0 0
0 0 0
_
_
. (1.64)
while F, C and E are
[F] =
_
_
e
t
0 0
0 e
t/2
0
0 0 e
t/2
_
_
, [C] =
_
_
e
2 t
0 0
0 e
t
0
0 0 e

_
_
, (1.65)
17
and
[E] =
1
2
_
_
e
2 t
1 0 0
0 e
t
1 0
0 0 e

1
_
_
. (1.66)
1.9 Governing Equations
1.9.1 Conservation of Mass
The mass M of a xed subset of material particles of the body occupying a region V(t) of R(t) is
M =
_
V(t)
dv, (1.67)
where is the mass density of the uid. The volume V(t) is a material volume of the uid: the
volume occupied by a xed set of material particles in the body, that may change in time. The
principle of conservation of mass is the postulate that the mass of this xed set of material particles
does not change in time,
d
dt
_
V(t)
dv = 0. (1.68)
Making use of the Transport Theorem, we can write the principle of conservation of mass with
respect to the spatial (Eulerian) representation of the eld variables,
0 =
_
V(t)
_

t
+ v + v
_
dv, (1.69)
where v is the velocity vector. Making suitable assumptions about continuity of the eld variables
and making use of the arbitrariness of the part of the body V(t), we obtain the local form of (1.69),

t
+ v () + v = 0 (1.70)
or in indicial notation,
0 =
(x, t)
t
+ v
i
(x, t)
x
i
+
v
i
(x, t)
x
i
. (1.71)
Implications of Conservation of Mass for Incompressible Fluids
The motion of an incompressible uid is constrained to be isochoric, in which case the divergence
of the velocity vector must be zero for all motions. In this case, the conservation of mass reduces
to,

t
+ v () = 0. (1.72)
18
From (1.72) we see that for an incompressible uid, the material derivative of the density is always
zero,
D
Dt
= 0. (1.73)
Note that the even thought the material derivative of the density it is zero, it is not necessary
that the density be constant in space. For example, a stratied uid with density distribution,
=
0
+ y, where is a constant, can experience simple shear, Section 1.8.4. It is easily seen
that this motion is isochoric, a necessary condition for an incompressible uid to experience this
motion. In addition,
D
Dt
=

t
+

x
i
v
i
= 0 (1.74)
for the stratied uid and given motion and therefore conservation of mass is satised.
1.9.2 Balance of Linear Momentum
The postulate of balance of linear momentum is the statement that the rate of change of linear
momentum of a xed mass of the body is equal to the sum of the forces acting on the body. These
forces can show up as body forces, or as forces due to stress vectors acting on the surface of the
body,
d
dt
_
V(t)
v dv =
_
V(t)
b dv +
_
dV(t)
t da (1.75)
where b is the body force per unit mass, t(x, n, t) is the surface force acting on the body in the
current conguration per unit area of P and n is the unit normal to the surface P at x at time
t. Recall that the stress vector is in general a function of the surface unit normal n. The rst and
second integrals on the right hand side of (1.75) represent the contributions due to body forces
and to surface forces respectively. Note that stress vector depends on position, time and the unit
normal to the surface at x. The rst and second integrals on the right hand side of (1.75) represent
the contributions due to body forces and to surface forces respectively. Recall, Cauchys Lemma
which follows from suitable smoothness assumptions on the function t(x, n, t) and the balance of
linear momentum

t(x, n, t) =

t(x, n, t). (1.76)

In addition, the existence of a second order tensor, T can be shown, where,
t =

t(x, n, t) =

T(x, t)n. (1.77)
The second order tensor T is called the Cauchy stress tensor. Signicantly, T is independent of the
n.
Making use of the divergence theorem as well as the relationship between the stress tensor and
the stress vector, (1.75) can be written as,
d
dt
_
V(t)
v dv =
_
V(t)
(b + T n) dv. (1.78)
19
The local form of the equation of linear momentum can be obtained form (1.78) by using the
transport theorem and making suitable assumptions about the continuity of the eld variables,

_
v
t
+ (v) v
_
= T + b, (1.79)
or, using index notation,

_
v
i
t
+ v
j
v
i
x
j
_
=
T
ij
x
j
+ b
i
. (1.80)
It is often convenient to represent T as the sum of a deviatoric and spherical part.
T = +

t I, (1.81)
where,

ii
= 0, and

t =
1
3
T
kk
. (1.82)
The tensor is often referred to as the deviatoric part of T and

tI as the spherical part. When the
Cauchy stress tensor is decomposed in this way,

t is often called the pressure and denoted by p.

Using the decomposition, (1.81), the balance of linear momentum can be written as,

_
v
t
+ (v) v
_
= p + + b. (1.83)
Alternatively,

_
v
i
t
+ v
j
v
i
x
j
_
=
p
x
i
+

ij
x
j
+ b
i
. (1.84)
For compressible uids, p is a thermodynamic pressure. An equation of state relating pressure to
other thermodynamic variables such as mass density and temperature will be necessary. For in-
compressible uids, p is a mechanical pressure arising from the constraint of incompressibility. No
equation of state is necessary, rather, p will be determined as part of the solution to the governing
equations and boundary conditions.
1.9.3 Balance of Angular Momentum
In the absence of body couples, the integral form of the balance of angular momentum is
d
dt
_
V(t)
x v dv =
_
V(t)
x b dv +
_
dV(t)
x t da. (1.85)
Making use of the balance of linear momentum, it can be shown that (1.85) reduces to the require-
ment that the Cauchy stress tensor must be symmetric,
T
ij
= T
ji
. (1.86)
20
1.9.4 Mechanical Energy Equation
It is sometimes useful to consider the Mechanical Energy Equation which is not derived from the
equation of balance of energy, rather it is obtained from the equation of linear momentum making
use of the equation of conservation of mass. If we take the inner product of the velocity vector and
the equation of linear momentum (1.80) we obtain,
1
2

D
Dt
(v
i
v
i
) =
T
ij
x
j
v
i
+ b
i
v
i
. (1.87)
It is sometimes useful to have an integral form of this equation. This can be obtained by integrating
(1.87) over a xed part of the uid (same material particles) occupying region V(t) with surface
V(t), to obtain,
1
2
d
dt
_
V(t)
v vdv =
_
V(t)
t vda
_
V(t)
tr(T D
T
)dv +
_
V(t)
b vdv. (1.88)
where we have made use of the divergence theorem and the conservation of mass. For future
reference, we introduce the following notation for terms in (1.88):
K
1
2
_
V(t)
v vdv = kinetic energy in V(t)
R
c

_
V(t)
t vda = rate of work done by surface forces on the boundary V(t)
R
b

_
V(t)
b vdv = rate of work done on the material volume V(t) by body forces
(1.89)
The scalar, tr(T D
T
) = T
ij
D
ij
is the rate of work by stresses per unit volume of the body and
is called the stress power. If we consider a part of the body V(t) in the current conguration, we
can hypothesize the existence of a scalar called the specic internal energy, u = u(x, t) (internal
energy per unit mass). The internal energy for the part V(t) of the body, will then be
_
V(t)
udv. (1.90)
Recall that the kinetic energy of the part V(t)of the body is
_
V(t)
1
2
v vdv. (1.91)
Heat may enter the body through the surface dV(t) of the body with outward unit normal n. It
can be shown, that this heat ux can be represented as the scalar product of a vector q and the
normal to the surface n. Where q n positive is associated with heat leaving the surface and q n
21
negative is associated with heat entering the surface. In addition, heat may enter the body as a
specic heat supply, r = r(x, t): the heat entering the body per unit mass per unit time. Therefore
the rate of heat entering the part V(t) of the body is

_
dV(t)
q nda +
_
V(t)
rdv. (1.92)
The balance of energy is a statement that the rate of increase in internal energy and kinetic energy
in the part V(t) of the body is equal to the rate of work by body forces and contact forces plus
energies due to heat entering the body per unit time. We can write this statement as
d
dt
_
V(t)
(u +
1
2
v v)dv =
_
dV(t)
t vda +
_
V(t)
b vdv
_
dV(t)
q nda +
_
V(t)
rdv (1.93)
where the rst integral on the left hand side is the rate of work by contact forces, the second
integral is the rate of work by body forces. Making suitable assumptions about continuity of the
eld variables, we can obtain the local form of (1.93),

_
Du
Dt
+v
Dv
Dt
_
= T : D + v ( T) + v b q + r. (1.94)
Using results from the mechanical energy equation, we can rewrite (1.94) as

Du
Dt
= T : D q + r, (1.95)
or in indicial form

Du
Dt
= T
ij
D
ji

q
i
x
i
+ r. (1.96)
1.10 Invariance Under Superposed Rigid Body Motions
A variety of criteria have been used to assess whether a proposed constitutive equation is rea-
sonable. In addition to assessing how well a constitutive equation models the behavior of a real
uid for a chosen ow, criteria such as invariance of the constitutive equation under a superposed
rigid motion (or, alternatively, material frame indierence
3
) [30, 14]; thermodynamic considera-
tions [28, 29, 18]; stability of the rest state and stability of some steady ows [10, 17, 16, 13]; and
well-posedeness of classes of initial value problems [6, 15, 12] have been used. For example, based
on thermodynamic restrictions, we require that the viscosity of a Newtonian uid be non-negative.
In this section we focus on the assumption that a constitutive equation be invariant under a
superposed rigid body motion to be physically reasonable . In some works, authors apply the re-
lated but dierent physical requirement that the form of the constitutive equation be independent
3
Truesdell and Noll [30] include an interesting discussion of the history of these two principles.
22
of observer. This assumption is often referred to as the principle of material frame indierence.
In both cases, an assumption is necessary to dene how the stress vector (or stress tensor) in the
original and superposed motion (or between the two observers) are related. In the following subsec-
tion, we discuss the kinematics associated with a general superposed motion without introducing
invariance assumptions. In the subsequent subsection, we provide a mathematical statement of the
requirement that a constitutive equation be invariant under a superposed rigid body motion. The
discussion of invariance under superposed rigid body motion follows that of Green and Naghdi, [14]
and Naghdi (1972,pp. 484-486) [20]).
Superposed Rigid Body Motions
In this section, we consider two motions of a body: an arbitrary motion
x = (X, t) (1.97)
and a second motion which diers from the rst by a superposed rigid body motion
x
+
=
+
(X, t). (1.98)
Namely, the position of a material point which at time t is at position x in the rst motion, is at
position x
+
in the second motion, Figure 1.5. The spatial description of (1.98) is then,
x
+
=
+
(x, t). (1.99)
Signicantly, the motion (1.99) is not an arbitrary second motion. Since the second motion diers
Figure 1.5: Identication of points in the body in the current and superposed congurations.
from the rst motion by a rigid body motion, the function
+
(x, t) must belong to a restricted class
23
of motions. In a rigid motion the distances between material points in the body are kept constant.
In this section, we shown that the most general form for this function is,
x
+
=
+
(x, t) = c(t) + Q(t) x, (1.100)
where Q(t) is a proper orthogonal second order tensor,
Q Q
T
= I, det Q = +1. (1.101)
First consider, two material points which we can identify as P and Q. These points can also be
identied by their respective positions, X and Y , in the reference conguration. The positions of
points P and Q in the current conguration, denoted by x and y respectively, are
x = (X, t), y = (Y , t). (1.102)
The positions occupied by these materials points at time t in the superposed motion, (1.99), are
then,
x
+
=
+
(x, t). y
+
=
+
(y, t). (1.103)
By dention, the distance between two arbitrary points in the superposed motion must be the same
as the distance between those in the motion, (1.97),
(x
+
y
+
) (x
+
y
+
) = (x y) (x y), (1.104)
namely,
(
+
(x, t)
+
(y, t)) (
+
(x, t)
+
(y, t)) = (x y) (x y). (1.105)
Equation (1.105) can be written in indicial notation as,
(
+
i
(x, t)
+
i
(y, t))(
+
i
(x, t)
+
i
(y, t)) = (x
i
y
i
)(x
i
y
i
). (1.106)
Recalling that x and y are independent, we can dierentiate (1.106) with respect to x
j
and then
with respect to y
k
to obtain,

+
i
(x, t)
x
j

+
i
(y, t)
y
k
=
jk
. (1.107)
Equation (1.107) can be written in coordinate free notation as,
_

+
(x, t)
x
_
T

+
(y, t)
y
= I. (1.108)
Equivalently,
_

+
(x, t)
x
_
T
=
_

+
(y, t)
y
_
1
. (1.109)
24
Equation (1.109) must hold for arbitrary choices of x and y in the body. Since the left side of
equation (1.109) is indpendent of y and the right side is independent of x, the second order tensors
on the left and right hand side of (1.109) must be functions of t only, which we denote as the
transpose of Q(t),
Q(t) =
_

+
(x, t)
x
_
, or Q
ij
(t) =

+
i
(x, t)
x
j
. (1.110)
Since (1.110) must hold for all x in R,
Q(t) =
_

+
(y, t)
x
_
(1.111)
and therefore from (1.108)
Q
T
Q = I, and det Q = 1, (1.112)
or
Q
li
Q
lj
=
ij
. (1.113)
We see from (1.112) that Q is an orthogonal tensor. Each superposed motion, (1.99), must include
the particular case in which
+
(x, t) = x. It can be seen from (1.110), that for this case, Q = I
and det Q = 1. Since the motions under consideration are continuous, we must always have,
det Q = 1. (1.114)
The equation (1.110), may be integrated with respect to x to yield,
x
+
=
+
(x, t) = c(t) + Q(t) x, or x
+
i
=
+
i
(x, t) = c
i
(t) + Q
ik
(t)x
k
, (1.115)
where c is a vector function of time. Therefore, the most general superposed rigid body motion is
described by (1.100). From (1.115), we also have
x = Q
T
(x
+
c), x
k
= Q
lk
(x
+
l
c
l
). (1.116)
So far, we have shown that necessarily a superposed rigid body motion is represented by (1.100).
We now show that it is also sucient, namely, that all motions of the form (1.100) preserve distance
between points.
|x
+
y
+
|
2
= (x
+
y
+
) (x
+
y
+
) =
_
Q (x y)
_

_
Q (x y)
_
= (x y)
_
Q
T
Q (x y)
_
= (x y) (x y)
= |x y|
2
.
(1.117)
25
1.10.1 Relationship between kinematic variables in the original and superposed
motions
We now consider other kinematical quantities under the superposed rigid body motion, (1.100),
Figure 1.6. The velocity vector v, transforms as,
v
+
=

+
(X, t)
t
=
dc(t)
dt
+
dQ
dt
x + Q v. (1.118)
We dene as
=
dQ
dt
Q
T
. (1.119)
Recall, that Q Q
T
= I, so
dQ
dt
Q
T
+ Q
dQ
T
dt
= 0 (1.120)
and hence, using (1.119) and (1.120), we nd,
+
T
= 0 (1.121)
and therefore, is skew-symmetric. Using the denition of , (1.119), we can write the expression
for v
+
as,
v
+
=
dc(t)
dt
+ Q x + Q v, (1.122)
and
v
+
i
=
dc
i
(t)
dt
+
il
Q
lm
x
m
+ Q
il
v
l
. (1.123)
For future use, we now derive the relationship between L
+
and L,
L
+
ij
=
v
+
i
x
+
j
=
v
+
i
x
k
x
k
x
+
j
=
_

il
Q
lm

mk
+ Q
il
v
l
x
k
_
x
k
x
+
j
= [
il
Q
lk
+ Q
il
L
lk
] Q
jk
=
ij
+ Q
il
L
lk
Q
jk
(1.124)
or in coordinate free notation,
L
+
= + Q L Q
T
. (1.125)
Using (1.125), it can be shown that
26
Figure 1.6: Representative kinematic variables for the original motion (F, D) and the superposed
motion (F
+
, D
+
) as well as the relationship between these kinematic variables.
D
+
= Q D Q
T
,
W
+
= + Q W Q
T
.
(1.126)
Recalling (1.14) and (1.98), it is clear that
F
+
=

+
(X, t)
X
, or F
+
iA
=

+
i
(X, t)
X
A
, (1.127)
Using (1.127) and (1.100),
F
+
iA
=

+
i
(x, t)
x
j
x
j
X
A
,
= Q
ij
F
jA
(1.128)
and therefore,
F
+
= Q F. (1.129)
Using (1.129) it can also be shown that,
J
+
= J. (1.130)
Additionally, the density and the normal to a surface n, can be shown to transform as,

+
=
n
+
= Q n.
(1.131)
27
A scaler, vector or second order tensor which transform as (1.131)
1
, (1.131)
2
and (1.126)
1
, respec-
tively are called objective.
1.10.2 Requirement: Invariance under a superposed rigid body motion
In this section, we consider the mechanical behavior of a xed material undergoing two dierent
motions: x = (X, t) and a second motion which is related to that at any arbitrary time t diers
from the rst by only a superposed rigid body motion. Since the material is the same for both
motions, the form of the constitutive equation will remain the same. For example, suppose the
material is linear in the symmetric part of the velocity gradient,
T = I + D. (1.132)
Then for the superposed rigid body motion,
T
+
= I + D
+
. (1.133)
More generally, if T =

T(D), then
T
+
=

T(D
+
). (1.134)
Note that the functional form does not change.
In this section, we are going to impose a requirement on the relationship between t and t
+
and
hence on T and T
+
. In particular, we might expect that the magnitude of the stress vector is
unchanged by a superposed rigid body motion ( t
+
and t to have the same magnitude) and that t
+
and t to have the same orientation relative to n
+
and n,respectively. Based on these expectations,
we introduce the following assumption
t
+
= Q t. (1.135)
Using (1.135) and (1.131), it can be shown that,
|t
+
| = |t|, t
+
n
+
= t n. (1.136)
The next step is to determine the relationship between T and T
+
. Recall that,
t = t(x, t; n) = T(x, t) n. (1.137)
Therefore,
t
+
= t
+
(x
+
, t; n
+
) = T
+
(x
+
, t) n
+
. (1.138)
In addition, recall the relationship between the outward unit normal n to a material surface in
the current conguration and the corresponding outward unit normal in the superposed motion,
n
+
= Q n. Therefore,
t
+
= T
+
n
+
= T
+
Q n. t
+
i
= T
+
ij
n
+
j
= = T
+
ij
Q
jk
n
k
. (1.139)
28
Using the assumption (1.135) and (1.137), it then follows that,
t
+
= Q t = Q T n, and t
+
i
= Q
ij
t
j
= Q
ij
T
jk
n
k
. (1.140)
Combining (1.139) and (1.140), we have that
_
T
+
QQ T
_
n = 0, and
_
T
+
ij
Q
jk
Q
ij
T
jk
_
n
k
= 0. (1.141)
Since (1.141) must hold for all n and the expression in brackets is independent of n, we can conclude
that the expression in brackets is equal to zero. Hence,
0 = T
+
QQ T
= T
+
Q Q
T
Q T Q
T
= T
+
Q T Q
T
or
_
_
_
_
0 = T
+
ij
Q
jk
Q
ij
T
jk
= T
+
ij
Q
jk
Q
lk
Q
ij
T
jk
Q
lk
= T
+
il
Q
ij
T
jk
Q
lk
(1.142)
and therefore,

T
+
= Q T Q
T
, and T
+
il
= Q
ij
T
jk
Q
lk
. (1.143)
29
30
Chapter 2
Non-Linear Viscous Fluids
In this chapter, we start with the constitutive assumption that the stress tensor depends on the
uid mass density and the velocity gradient L.
T = T(, L). (2.1)
Inherent in the form of the stress tensor, (2.1) is the assumption that the current state of stress
depends only on the velocity gradient at the current time and not on any previous deformation the
uid might have undergone. Using invariance of the stress tensor under a superposed rigid body
motion and a representation theorem for symmetric isotropic tensor functions, it is shown that the
most general form of (2.1), which does not violate invariance requirements is
T =
0
I +
1
D +
2
D
2
(2.2)
where in general
0
,
1
,
2
depend on and the three principal invariants of D.
1
Similarly, we can start by considering incompressible uids for which the extra stress tensor
is a function of the velocity gradient. After imposing invariance requirements and using the same
representation theorem, we can deduce that the most general form of the Cauchy stress tensor that
satises invariance can be written as,
T = I +
1
D +
2
D
2
(2.3)
where is a Lagrange multiplier arising from the incompressibility constraint and
1
and
2
depend
only on the second and third principal invariants of D. Such incompressible uids are typically
called Reiner-Rivlin uids. The Navier-Stokes uid is a special Reiner-Rivlin uid with
2
equal
to zero and
1
constant. The behavior of Reiner-Rivlin uids with non-zero values of
2
in simple
shear does not match experimental results on real uids [1]. In addition, the dependence on the
value of III
D
is often considered negligible [1]. For this reason, attention is typically conned to
a special class of Reiner-Rivlin uids called generalized Newtonian uids. These are Reiner-Rivlin
uids with
2
zero and
1
dependent only on the second principal invariant of D. In the latter part
of this chapter, specic generalized Newtonian uids are considered and the behavior is contrasted
and compared with experimental results.
1
Recall the principal invariants of a second order tensor A are I
A
= tr A, II
A
= 1/2((tr A)
2
tr (A
2
)), III
A
=
det A.
31
2.1 Constitutive Assumption T = T(, L)
In this section, we start with the constitutive assumption that the stress tensor depends on the
uid mass density and the velocity gradient L. Namely, we assume, T =

T(, L). We then obtain
the most general function which satises invariance requirements. Using the relationship
L = W + D (2.4)
we can rewrite the constitutive assumption as
T =

T(, D, W). (2.5)
If the constitutive equation (2.5) is a valid one, then it must hold for all motions, in particular, it
must hold for a superposed rigid motion,
T
+
=

T(
+
, D
+
, W
+
). (2.6)
Recalling that,
T
+
= Q T Q
T
, (2.7)
and using (2.6)

T(
+
, D
+
, W
+
) = Q

T(, D, W) Q
T
. (2.8)
Recalling,

+
= , D
+
= Q D Q
T
, W
+
= Q W Q
T
+ , (2.9)
we can write (2.8) as

T(
+
, Q D Q
T
, Q W Q
T
+ ) = Q

T(, D, W) Q
T
. (2.10)
Now consider a superposed motion, for which at some arbitrary time t, Q = I, and = 0. In this
case, (2.10) becomes,

T(, D, W + ) =

T(, D, W), (2.11)
for arbitrary choices of the skew-symmetric tensor . It then follows that the response function T
cannot depend on W, and so (2.5) is reduced to
T =

T(, D), (2.12)
32
and must satisfy

T(, Q D Q
T
) = Q

T(, D) Q
T
, (2.13)
for all proper orthogonal Q. Then, from (2.13), we can conclude that

T(, D) must be a symmetric
isotropic tensor function of D, (see, for example, Appendix C). A representation theorem for
tensors (Appendix C), states that the most general second order isotropic tensor function

T(, D)
for symmetric second order tensor D, has the form
T =
0
I +
1
D +
2
D
2
, (2.14)
where in general
0
,
1
,
2
are functions of the three principal invariants of D as well as the mass
density .
2.1.1 Incompressible Inelastic Fluids
In the case, where the uid is incompressible, the density of a xed material point is constant. The
stress tensor is therefore independent of the density ,
T =

T(D). (2.15)
The Cauchy stress tensor is determined by D only to within a stress N which does no work in
any motion satisfying the constraint of incompressibility. We can rewrite (2.15) as the sum of the
indeterminant stress N and an extra stress , (see, for example, Truesdell [31], pp. 70-73),
T = N + (D). (2.16)
As discussed previously in Section 1.9, the expression for the rate at which stresses do work per
unit volume, P, is
P = tr(T D
T
). (2.17)
Recall, the scalar quantity tr(A B
T
) = A
ij
B
ij
satises the requirements for an inner product of
two arbitrary second order tensors, A, B. We will use the notation (A, B) to denote this inner
product. Namely,
(A, B) tr(A B
T
). (2.18)
Using this notation, the condition of incompressibility can be written as (D, I) = 0. Since the
indeterminate stress does no work on the constrained motions,
0 = (N, D) for all motions in which(D, I) = 0. (2.19)
We can state the conditions on N given in (2.19) as, N must be orthogonal to all D which are
orthogonal to I. It therefore follows, that N must be a scalar multiple of I,
N = I, (2.20)
33
where is a Lagrange multiplier. Therefore the Cauchy Stress, for an incompressible uid takes
the form,
T = I + (D). (2.21)
The most general form for (D) can be deduced in essentially the same way we deduced the
form of the stress tensor for a compressible uid. For an incompressible uid, the most general
representation for the extra stress, (D) is
=
0
I +
1
D +
2
D
2
. (2.22)
The rst principal invariant of D is 0 for isochoric motions, so
0
,
1
,
2
are functions of only II
D
and III
D
. Without loss in generality, we can absorb
0
into the Lagrange multiplier and write
(2.22) as

Reiner Rivlin Fluids

T = I +
1
D +
2
D
2
where
1
=

1
(II
D
, III
D
) and
2
=

2
(II
D
, III
D
)
is a Lagrange multiplier arising from the incompressibility
constraint.
(2.23)
It remains to specify the two material functions

1
(II
D
, III
D
) and

2
(II
D
, III
D
). The Lagrange
multiplier will be determined from solving the governing equations subject to the boundary con-
ditions. Incompressible uids for which the stress tensor is related to the velocity gradient through
(2.23) are called Reiner-Rivlin uids.
2.2 Reiner-Rivlin Fluids
As just stated, Reiner-Rivlin uids are incompressible uids with constitutive equation,
T = I +
1
D +
2
D
2
, (2.24)
where

1
(II
D
, III
D
) and

2
(II
D
, III
D
). We have not yet considered specic forms for the material
functions,

1
(II
D
, III
D
) or

2
(II
D
, III
D
). In this section, we discuss some restrictions on these
material functions which can be motivated by thermodynamic considerations. In addition, we will
compare the behavior of the Reiner-Rivlin uid in simple shear with the behavior of real uids to
motivate restricting our attention in the remainder of this chapter to a special case of Reiner-Rivlin
uids with
2
equal to zero and
1
a function only of II
D
. Fluids in this subclass of Reiner-Rivlin
uids are called generalized Newtonian uids.
34
2.2.1 Thermodynamic Restrictions on
1
and
2
for Reiner-Rivlin Fluids
From the Second Law of Thermodynamics we require that the stress power be non-negative,
T : D 0. (2.25)
This requirement can be used to deduce restrictions on the choices of the material functions
1
,
2
,
[28, 29, 7, 18]. For a Reiner-Rivlin uid,
T : D tr (T D) = tr
_
D +
1
D
2
+
2
D
3
_
,
= tr
_
D +
1
D
2
+
2
(II
D
D + III
D
I)
_
,
=
1
tr (D
2
) + 3
2
III
D
,
=
1
tr (D
2
) + 3
2
detD,
(2.26)
where we have made use of the Cayley-Hamilton Theorem and the condition of incompressibility.
Therefore from thermodynamics, we require,
0
1
tr (D
2
) + 3
2
detD. (2.27)
The rst term on the right hand side of (2.27) is non-negative while the second term may be
positive, negative or zero. In order to deduce further information from (2.27), we consider some
special cases of Reiner-Rivlin uids.
Case I:
2
= 0,
1
not necessarily constant.
For the special case of a Reiner-Rivlin uid with
2
equal to zero, we see from (2.27) that necessarily

1
0.
Case II:
1
,
2
both constant.
We now consider the particular case where
1
and
2
are both constants. In this case,
1
and
2
will be independent of the invariants of D, so any conclusions we make based on specic choices of
D will hold for all D.
First, consider the following choice for the rectangular components of D,
[D] =

2
_
_
0 1 0
1 0 0
0 0 0
_
_
. (2.28)
For the choice of D dened in (2.28), tr (D
2
) =
2
/2 and det D = 0. Therefore the inequality
(2.27) takes the form
0

2
2

1
, (2.29)
35
and therefore for this case
1
must be greater than or equal to zero. Since
1
is a constant, this
result holds not only for the choice (2.28) for D but for all D. Now, consider the following two
choices for the rectangular components of D,
[D] =
_
_
2 0 0
0 1 0
0 0 1
_
_
, [D] =
_
_
2 0 0
0 1 0
0 0 1
_
_
. (2.30)
For both cases, tr (D)
2
= 6
2
. For the rst case, det D = 2
3
while for the second case det D =
2
3
. Therefore the inequality (2.27) must satisfy both of the following,
0
1
+
2
,
0
1

2
.
(2.31)
Since
1
is always constant and is arbitrary, we see from that (2.31) that
2
must be zero cor-
responding to an incompressible Newtonian uid with kinematic viscosity equal to 1/2
1
. If
2
is
zero, then from (2.27),
1
must be greater than zero. This is equivalent to the familiar requirement
that the kinematic viscosity be positive. Therefore, we can conclude that the only physically rea-
sonable case with
1
and
2
both constant is the case of an incompressible Newtonian uid.
2.2.2 Restrictions on
1
,
2
based on behavior of real uids in simple shear
Consider ow of an incompressible Reiner-Rivlin uid between two innite parallel plates which is
driven by the motion of the upper plate moving at a constant speed U, Fig. 1.4. We have selected
a rectangular coordinate system with the x
1
axis parallel to the plates and x
2
perpendicular to the
plates. The location of the x
1
axis is chosen such that x
2
= 0 is coincident with the bottom plate.
As discussed in Section 1.58, we look for solutions of the form,
v = v(x
2
) = x
2
e
1
(2.32)
where is a constant. Recall that for (2.32) to be a solution it must satisfy the boundary con-
ditions for the problem, the incompressibility condition, the equation of linear momentum, and
the constitutive equation for the stress tensor. The boundary conditions for this problem are the
no-slip condition at the two plate surfaces,
v(x
2
= 0) e
1
= 0, v(x
2
= h) e
1
= U, (2.33)
and the condition of impermeability of both plates,
v(x
2
= 0) e
2
= 0, v(x
2
= h) e
2
= 0. (2.34)
36
We see that boundary conditions, (2.33) and (2.34) are met when = U/h. For a velocity eld
of the form, (2.32), the components of the rate of deformation tensor D are
[D
ij
] =
_
1
2
(
v
i
x
j
+
v
j
x
i
)
_
=

2
_

_
0 1 0
1 0 0
0 0 0
_

_
ij
. (2.35)
It is clear from (2.35) that for the motion, (2.32), the deformation tensor D is a constant second
order tensor. The components of D
2
are,
_
D
2

= [D
ij
D
jk
] =

2
4
_

_
1 0 0
0 1 0
0 0 0
_

_
, (2.36)
and the three principal invariants of D are,
I
D
= tr D = 0,
II
D
=
1
2
_
(tr D)
2
(tr (D)
2
)
_
=

2
4
,
III
D
= det D = 0.
(2.37)
Using (2.37), we see that the incompressibility condition
0 =
v
i
x
i
= tr D (2.38)
is identically satised.
The incompressible Reiner-Rivlin constitutive equation (2.24) can be split into the constraint
response and response that does work,
T = I + (D), where (D) =

1
(II
D
, III
D
)D +

2
(II
D
, III
D
)D
2
. (2.39)
Using the results (2.36) with (2.39), it follows that,
[] = =
1

2
_

_
0 1 0
1 0 0
0 0 0
_

_
+
2

2
4
_

_
1 0 0
0 1 0
0 0 0
_

_
, (2.40)
Therefore, the only non-zero components of are,

12
=
21
=
1

2

11
=
22
=
2

2
4
. (2.41)
37
where
1
and
2
are both even functions of ,

1
=

1
(
1
4

2
, 0),
2
=

2
(
1
4

2
, 0). (2.42)
It follows from (2.41) and (2.22), that the only non-zero components of T are,
T
12
= T
21
=
1

2
, T
11
= T
22
= +
2

2
4
, T
33
= 0. (2.43)
We see from (2.43) that T
12
is an odd function of while T
11
and T
22
are even functions of .
It follows from (2.43) that to fully determine the solution for the Cauchy stress tensor, we need
to solve for , which will in general require consideration of the balance of linear momentum as well
as the boundary conditions. We will consider this issue at a later time. Here we focus attention
on the viscometric functions. As will be discussed in more detail in the following sections, the
following three functions, determine the behavior of a certain category of uids in class of ows
called viscometric ows:
( ) = T
12
N
1
( ) = T
11
T
22
N
2
( ) = T
22
T
33
(2.44)
where , N
1
, N
2
are referred to as the shear stress, rst normal stress and second normal stress.
Three material functions commonly used in the literature are dened relative to , N
1
, N
2
,
( ) = T
12
/

1
( ) = N
1
/
2
= (T
11
T
22
)/
2

1
( ) = N
2
/
2
= (T
22
T
33
)/
2
(2.45)
where ,
1
,
2
are called the viscometric functions, referred to as the viscosity, rst normal stress
coecient and second normal stress coecient. Most rheometers are based on viscometric ows
and are directed at measuring the viscometric functions.
It follows from (2.43) and (2.45), that for a Reiner-Rivlin uid,
=

1
2
,
1
= 0,
2
=
1
4

2
. (2.46)
and therefore,

1
= 2,
2
= 4
2
. (2.47)
However, there is no evidence of real uids exhibiting a zero value for
1
and a non-zero value for

2
[1]. For this reason, attention is typically conned to Reiner-Rivlin uids with
2
equal to zero.
Therefore, no constitutive equation of the form T =

T(D) will be suitable for describing real uids
with nonzero normal stress coecients. In summary, based on invariance requirements, thermo-
dynamic considerations and the behavior of real uids, the most general reasonable constitutive
equation of the form T =

T(D) is,
T = I +

1
(II
D
, III
D
)D where
1
0. (2.48)
38
Recalling that the mechanical pressure is dened as p = tr(T)/3, it follows from (2.48), that
= p. (2.49)
In addition, in simple shear (and other viscometric ows), the quantity III
D
is identically zero
and therefore the functional dependence of
1
on this variable cannot be determined from most
rheometers. In the remainder of this chapter, we will only consider uids for which
1
is assumed
independent of III
D
. As discussed in [1], page 54, there is some evidence that this may be reasonable
for real uids. After imposing this restriction on (2.48), we obtain the equation for a generalized
Newtonian Fluid (incompressible),
T = pI +

1
(II
D
)D where
1
0. (2.50)
If we rewrite this last results in terms of the viscometric functions we obtain,
T = pI + 2 (II
D
)D where 0. (2.51)
Recalling that II
D
= 1/2tr(D
2
), it is clear that it will be convenient to replace II
D
with a
non-negative quantity such as,
S = 4II
D
(2.52)
where the factor of two is introduced in order that S =
2
in simple shear. Using this new notation,

Incompressible, generalized Newtonian Fluid

T = pI + 2 (S)D
where S 4II
D
, p is the mechanical pressure,
From Thermodynamics: 0.
(2.53)
2.3 Generalized Newtonian Fluids
In this section we rst consider some commonly used examples of generalized Newtonian uids.
We then consider the behavior of these models in several important ow elds: (i) steady, fully
developed ow in a straight pipe and (ii) steady cone and plate ow and (iii) steady ow between
concentric cylinders.
39
2.3.1 Power-Law Fluid
The Power-Law Fluid is an example of one of the simpler generalized Newtonian uids. It is
particularly popular because of the number of exact solutions which can be obtained for this model.
The Cauchy Stress tensor for a Power-Law uid is dened as,
T
ij
= p
ij
+ 2 K|4II
D
|
(n1)/2
D
ij
(2.54)
and therefore, (S) is dened as
(S) = K S
(n1)/2
(2.55)
where n is called the power law index and K is called the consistency. When n is equal to one, the
Navier-Stokes constitutive equation is recovered. When n is less than one, the constitutive equation
Power Law Fluids for n less than or equal
to 1 (shear thinning or Newtonian)
Power Law Fluids for n greater than or
equal to 1 (shear thickening or Newtonian)
is shear thinning while when n is greater than one the behavior is shear thickening. The viscosity
in the limit of vanishing S and in the limit of S tending to innity are,
n < 1 : lim
S0
(S) = , lim
S
(S) = 0,
n > 1 : lim
S0
(S) = 0, lim
S
(S) = .
(2.56)
The unboundedness of the viscosity function and the lack of a non-zero viscosity at zero shear rate
does not match experimental results for real uids and so limits the applicability of the Power-Law
model.
40
2.3.2 Prandtl-Eyring Model
An alternative to the Power-Law model is the Prandtl-Eyring Model which tends to a constant
viscosity in the limit of going to zero. However, the viscosity function tends to zero as tends
to innity,
(S) =
0
sinh
1
(

S)

S
(2.57)
where
0
and are material constants and
lim
S0
(S) =
0
, lim
S
(S) = 0. (2.58)
2.3.3 Powell-Eyring Model
The Powell-Eyring Model is a three constant model which displays a non-zero bounded viscosity
at both the upper and lower limits,
(S) =

+ (
o

)
sinh
1
(

S)

S
(2.59)
where
0
,

and are material constants and

lim
S0
(S) =
0
, lim
S
(S) =

. (2.60)
It is convenient to write rewrite (2.59) as,

=
sinh
1
(

S)

S
(2.61)
2.3.4 Cross Model
The Cross Model [9] is a four constant model which also displays a non-zero bounded viscosity at
both the upper and lower limits

=
1
1 + (K
2
|S|)
(1n)/2
(2.62)
The notation used in (2.62) is chosen to t the shear thinning case where n < 1,
lim
S0
(S) =
0
, lim
S
(S) =

. (2.63)
This model is also suited for shear thickening, though in this case, the notation should be modied,
since, while for n > 1,
lim
S0
(S) =

, lim
S
(S) =
0
. (2.64)
41
Ellis Model
Frequently, the ow regimes of interest are such that it is not essential to specify a nonzero viscosity

infty
. Setting

to zero in (2.62) yields the Ellis model [?],

=

0
1 + (K
2
|S|)
(1n)/2
. (2.65)
A major advantage of the Ellis model is that it is possible to obtain analytical solutions for some
simple ows (e.g. [3], [27]).
2.3.5 Yasuda Model
Yasuda proposed a model similar to the Cross model but with an extra material constant a with
which to t the data [32],

=
1
[1 +
a
S]
(1n)/a
. (2.66)
Several special cases of the Yasuda model are notable including the Carreau model.
Carreau Model
If a is set to 2 in (2.66), the four constant Carreau model is obtained [4]. It also displays a non-zero
bounded viscosity at both upper and lower limits,

=
1
_
1 +
2
S
_
(1n)/2
. (2.67)
Comments regarding the limiting behavior for the Cross model hold for the Yasuda model as
well.
2.3.6 Bingham Fluids
There are many real uids which if tested in simple shear, would appear not to ow at all until
the magnitude of the applied shear stress surpassed a xed nite value termed the yield stress.
E. C. Bingham (1916) observed this type of behavior in paints. Many other real uids show an
apparent yield stress as well. One type of constitutive equation which is used to describe uids
which appear to display a yield stress is the Bingham model (see, e.g. [2] the review article
by Barnes, for a discussion of experimental evidence supporting the lack of existence of a true
yield stress in real materials.). In simple shear of a Bingham uid, there is no motion unless the
magnitude of the applied shear stress is above a threshold,
Y
, in which case, the material behaves
like an incompressible Newtonian uid, Fig. (2.1). It is a simple matter to provide a mathematical
statement of this response,

12
<
Y
= = 0

12

Y
= =
1

(
12

Y
).
(2.68)
42
Figure 2.1: Shear stress
12
versus shear rate in simple shear for (i) Newtonian uids (ii) Bingham
uids with yield stress
Y
.
When the ow is more complex than simple shear, a more general measure of the magnitude
of the shear stress is used. Namely, we cannot simply use
12
=
Y
as a criterion. Note that
this expression is even incorrect for simple shear under a change of basis. The condition, must
be rephrased in terms of an invariant of the stress tensor. For a Bingham uid, the more general
measure of the shear stress is the second invariant of the extra stress tensor, II

. The three
dimensional representation of the Bingham model can be represented as (see, e.g. [11],[25]),
_
|II

| <
Y
= D
ij
= 0
_
|II

|
Y
=
_

_
D
ij
=
1
2
_
_
1

Y
_
|II

|
_
_

ij

ij
= 2
_
_
+

Y
2
_
|II
D
|
_
_
D
ij
(2.69)
where
Y
and are material constants. Recall that in simple shear, I

, III

are zero for a Bingham

uid. In simple shear, 2.69 can be used to obtain .
Early studies of this constitutive equation were carried out by [22, 23, 26, 25]. Other materi-
als which seem to be well modeled by the Bingham constitutive model include pastes, margarine,
mayonnaise, ketchup. and concentrated suspensions of particles. In simple shear, the shear stress
is constant throughout the ow eld. In more complex ows, such as fully developed ow in a
straight pipe or ow between two coaxial cylinders driven by the relative rotation of the cylinders,
we will see that the shear stress is not constant through out the ow eld, and as a result, there
may be regions where the yield criterion is reached ( and the uid is owing) while in other regions
the value of II

is below
Y
and the uid does not ow (D = 0 in that region). Note that D = 0
does not necessarily imply v = 0. A collection of analytical solutions for the Bingham model and
43
extensive reference list can be found in a review article by Bird, Dai and Yarusso [5].
2.3.7 Modied Bingham Model
The Bingham uid equation is important because it appears to model some real uids well and
yet it is possible, in some cases, to obtain analytical solutions for the ow eld, (for example,
Poiseuille and Couette ow). However, this constitutive equation is in general challenging to model
numerically due to the diculty in tracking the yield surfaces in the ow eld. Papanastasiou [24]
modied the Bingham model (2.69) to eliminate the yield stress. Recall, that for Bingham uids, at
zero shear rate, the viscosity function (II
D
) abruptly changes from an innite value to a constant
(nite) viscosity . Papanastasiou dened the viscosity function as an increasing function of II
D
,
as
(S) = +

Y
_
1 e
nS/2
_
S
, (2.70)
where n, ,
Y
are material constants. The value of n determines the rate of exponential increase
Figure 2.2: Shear stress versus shear rate for simple shear of a Bingham uid and modied Bingham
uid.
to the constant viscosity . For simple shear, it follows from (2.70),

12
=
_
_
+

Y
_
1 e
n

_

_
_
. (2.71)
In the limit of n tending to innity the shear stress for the Bingham uid is obtained.
44
2.4 Steady, Fully Developed Flow of a Generalized Newtonian
Fluid in a Straight Pipe
In this section, we consider steady, fully developed ow in a straight pipe of circular cross section of
radius R driven by a pressure drop per unit length in the axial direction equal to . In particular,
we look for velocity elds of the form,
v = w(r)e
z
, (2.72)
where we have used standard cylindrical coordinates in (2.72) such that the z-axis is aligned with
the pipe centerline. The volumetric ow rate, Q, for velocity elds of the form (2.72) is
Q = 2
_
R
0
w(r)rdr. (2.73)
In many engineering applications, it is of interest to obtain a relationship between Q and p/z
(denoted as -). If a closed form solution for the velocity eld can be calculated for a given
viscosity function (S), then it is relatively straightforward to calculate the corresponding ow rate
as a function of from (2.73). In the rst part of this subsection, we consider the Power-Law,
Cross and Bingham uids for which it is possible to obtain closed form solutions for w(r). For most
Generalized Newtonian uids, it is not possible to obtain a closed form solution for the velocity
eld for steady, fully developed ow in a straight pipe. In the second part of this subsection we
consider the more general case and obtain an equation relating Q as a function of and (S).
In the last part of this subsection, we discuss the inverse problem of using measured relationships
between driving pressure drop and volumetric ow rate to obtain the viscosity function (S) for a
generalized Newtonian uid.
Velocity elds of the form (2.72) identically satisfy the incompressibility condition and the
components of D and D
2
in cylindrical coordinates are,
[D] =

2
_

_
0 0 1
0 0 0
1 0 0
_

_
,
_
D
2

=

2
4
_

_
1 0 0
0 0 0
0 0 1
_

_
, (2.74)
where we have denoted dw(r)/dr by . It follows from (2.74) that the invariants of D are,
I
D
= 0, II
D
=
2
/4, III
D
= 0. (2.75)
From these results and (2.53), we see that the only nonzero components of the extra stress tensor
for a Generalized Newtonian Fluid in simple shear are,

rz
=
zr
= (S) where S 4 II
D
=
2
. (2.76)
45
Note that and S depend at most on r. If we use the results (2.76) in the equation of linear
momentum in cylindrical coordinates we nd,
0 =
p
r
0 =
1
r
p

0 =
p
z
+
1
r
(r
rz
)
r
.
(2.77)
It follows directly from (2.77), that p is independent of both r and and p/z is a constant which
we will denote as . Hence, the pressure is p = z + f(t). We can now integrate (2.77)
3
with
respect to r to obtain,
rz
= r/2+ constant/r. In order that the shear stress be bounded at the
center of the pipe, we set the constant to zero.

rz
=
r
2
. (2.78)
Combining this with the result (2.74), we obtain
(
2
) =
r
2
. (2.79)
We cannot proceed further without specifying the viscosity function.
2.4.1 Power-Law Model
From (2.79) and (2.55), we obtain,
K
n
=
r
2
(2.80)
and upon integration and use of the no-slip boundary condition at the wall, we have,
w(r) =
_
R
2K
_
1/n
_
R
1 + 1/n
_
_
1 (
r
R
)
1+1/n
_
. (2.81)
After substituting (2.83) in (2.73) and integrating, we obtain the volumetric ow rate as a function
of the axial components of the pressure gradient , the materials constants K and n and the pipe
radius, R.
Q =
_
R
2K
_
1/n
R
3

3 + 1/n
. (2.82)
Using (2.82), we can write the axial velocity (2.83) as,
w(r) =
Q
R
2
3 + 1/n
1 + 1/n
_
1 (
r
R
)
1+1/n
_
. (2.83)
For the same ow rate, the velocity eld in the shear thinning power-law model is atter than for
the constant viscosity case, Fig. 2.4.1.
46
Figure 2.3: Nondimensional axial velocity prole, w(r)/(Q/R
2
, as a function of nondimensional
radius r/R for steady, fully developed ow of power-law uids in a straight pipe. The proles shown
are for the same ow rate but with dierent values of n.
2.4.2 Bingham Fluid
Recall that for a Bingham Fluid, D is zero at all points where
_
|II

| is less than the yield stress.

For the ow under consideration
_
|II

| is equal to the magnitude of

rz
which is equal to r/2. It is
clear from the form of the solution for
rz
that the magnitude of
_
|II

| will be largest at the largest

radial position: r = R. When R/2 <
Y
, the yield criterion will not be met anywhere in the uid
and hence D will be zero throughout the uid. From the boundary conditions, it then follows that
the velocity is zero everywhere. If the driving pressure gradient increased so that R/2 >
Y
, there
will be a region r [0, 2
Y
/], where the yield criterion is not met, and r (2
Y
/, R] where it is
met. We will denote this critical radius as r
y
2
Y
/. From (2.69),
D
rz
=
_

_
0 for r r
Y

r
4
+

Y
2
for r r
Y
(2.84)
As previously discussed, for the velocity eld under consideration, D
rz
is equal to 1/2 dw/dr and
so, from (2.84), we can obtain the axial velocity as a function of the radial variable r,
w(r) =
_

_
C
1
for r r
Y

r
2
4
+

Y
r

+ C
2
for r r
Y
(2.85)
47
where C
1
and C
2
are constants of integration. If we require that the value of w(r) be zero at the
pipe boundary and be continuous throughout the ow eld (including r equal to r
Y
), then
w(r) =
_

_
R
2
4
_
1
r
Y
R
_
2
for r r
Y
R
2
4
_
1
r
2
R
2
_

Y
R

_
1
r
R
_
for r r
Y
(2.86)
The ow rate as a function of can be calculated directly from (2.73) and (2.86) and is left as an
exercise.
2.4.3 Q as a function of and (S) for unspecied generalized Newtonian uid
For most viscosity functions it will be dicult, if not impossible to obtain a closed form solution
for the velocity eld. In this section, we consider the possibility of directly obtaining an expression
for the ow rate as a function of without rst obtaining an explicity expression for the velocity
eld. While in most cases we will still need to use numerics to obtain the relationship, we will be
able to avoid the intermediate step of rst calculating the velocity eld.
Beginning with (2.73) and the denition of and performing integration by parts twice, we can
write the volumetric ow rate as
Q =

3
_
R
3

w

_
w
0
r( )
3
d
_
, (2.87)
where
w
denotes the value of at the pipe wall. Eq. (2.79) provides and expression for r as a
function of ,
r =
2

(
2
). (2.88)
Using the relationship (2.88) in (2.87), we obtain
Q =

3
_
R
3

w

8

3
_
w
0

3
(
2
)
3
d
_
. (2.89)
We can obtain
w
as a function of and R by solving the equation resulting from evaluating (2.88)
at the wall,

w
(
2
w
) =

2
R. (2.90)
Once this value is obtained, (2.89) can be integrated to obtain volumetric ow rate as a function
of .
Exercise: Use (2.89) to obtain the owrate as a function of for the Power-Law model.
Exercise: Is it possible to evaluate (2.89) in closed form for the Carreau model?
48
2.4.4 Determination of (S) from measured values of driving pressure drop and
ow rate in a straight pipe
Capillary viscometers are commonly used to measure the viscosity function (S). These rheometers
are designed to reproduce, with some level of experimental error, the velocity eld for steady fully
developed ow in a pipe of constant radius v = w(r)e
z
. The general idea is that using results from
balance of linear momentum for this velocity eld, and measurements of ow rate and pressure drop
at dierent ow rates, we can back out and expression for (S). Note that we want this expression
to be for an unspecied generalized Newtonian uid (in fact, we will show in later chapters that the
capillary viscometer can in principal be used to measure (S) for all simple uids). The arguments
in this section, follow that of [1].
As will be clear at the end of this section, it is useful to rephrase the problem as one of obtaining
as a function of a characteristic shear rate
a
and the shear stress at the wall,
w
. We will dene

a
as,

a

4Q
R
3
, (2.91)
which is the shear rate at the pipe wall if the uid is Newtonian. Using measured values of Q it is
a simple matter to obtain
a
. Using the denition of ow rate, integration by parts and (2.91), we
can obtain an expression for the characteristic shear rate

a
=
4
R
_
u
s
+
1
R
2
_
R
0
r
2
dr
_
, (2.92)
where the notation u
s
denotes the slip velocity at the wall of the pipe. Here we focus attention
on uids for which the slip velocity is assumed to be zero. However, there is evidence that this
assumption is not appropriate for some real uids. See [1] for further discussion of using capillary
rheometers in cases where u
s
is not negligible.
Recall, that for a generalized Newtonian uid undergoing a velocity eld v = v
z
(r)e
z
, that
necessarily

rz
=
r
2
, (
2
) =
rz
/ , (2.93)
based on balance of linear momentum and the constitutive equation. Evaluating (2.93)
1
at the
wall,

w

rz
|
r=R
=
1
2
R. (2.94)
It follows that
w
can easily be calculated from measured values of . As mentioned above, we are
interested in obtaining an expression relating
w
and
a
. We can do so, by rewrite the integral in
(2.92) over r as an integral over
rz
. It follows from (2.93)
1
and (2.94) that
r =
rz
R/
w
. (2.95)
When (2.93)
2
is invertible,
= (
rz
). (2.96)
49
After using (2.95) and (2.96) in (2.92), we nd

a
=
4

3
w
_
w
0

2
rz
(
rz
) d
rz
. (2.97)
If we take the derivative of the expression, (2.97) with respect to
w
, we nd,
d
a
d
w
=
12

4
w
_
w
0

2
rz
(
rz
) d
rz
+
4

w
(
w
),
= 3

a

w
+
4 (
w
)

w
.
(2.98)
Rearranging (2.98) and using the chain rule, we obtain
4

w

a
=
_
3 +
ln
a
ln
w
_
(2.99)
where
w
= (
w
). Dening,
n

=
ln
w
ln
a
(2.100)
we can rewrite (2.99) as,

w
=
a
3n

+ 1
4n

. (2.101)
Equation (2.101) is the Mooney-Rabinowitsch equation. Experiments can be run at dierent pres-
sure drops to obtain a curve of
a
as a function of
w
(recall, (2.91) and (2.94)). If these results
are plotted on a log scale, the value n

can be obtained from the slope of the curve, which will in

general be a function of
w
. From the denition of the viscosity function, applied at the wall,
(
w
)
2
=
w
/
w
. (2.102)
Using (2.101) with this last result, enables us to determine results for (S) for this ow. Astarita
and Marrucci provide the details of applying this result to a specic data set (page 73, [1]).
2.5 Cone and Plate Flow for a Generalized Newtonian Fluid
The cone and plate rheometer is one of the most commonly used instruments for measuring viscosity
and normal stress coecients. In particular, it can be used to measure the rst normal stress
coecient
1
in addition to the viscosity function
2
. The implementation of the rheometer is
based on an idealized ow of uid between a cone and plate, driven by the rotation of the cone about
the z-axis, Figure 2.5. In this section, we focus on the derivation of the fundamental relationships
2
While the focus in this chapter is on generalized Newtonian uids, this rheometer can also be used for viscoelastic
uids. Recall that 1 is zero for generalized Newtonian uids.
50
for this rheometer including the implementation of the small angle approximation at the basis of
cone and plate rheometers. A detailed discussion of the assumptions involved in the use of this
rheometer can be found in [19], pages 205-217.
It is convenient to use standard spherical coordinates for which the the solid walls of the rheome-
ter are coordinate surfaces. The angle between the cone and the plate is denoted by and the
magnitude of the angular velocity by . The corresponding applied torque necessary to generate
this angular velocity as m = m
z
e
z
. Using this notation, the boundary conditions on the velocity
eld are,
v|
=/2
= 0, and v|
=(/2)
= r cos e

. (2.103)
The boundary condition on the stress vector at the free surface, denoted by
1
, is
t = p
atm
n, on
1
(2.104)
where n is the normal to the free surface and p
atm
denotes the atmospheric pressure. At this point,
Figure 2.4: Schematic of cone and plate geometry with spherical coordinates. Note: The angle
is typically less that 6 degrees for cone and plate rheometers.
we have not specied the shape of the free surface, which in general should be determined as part
of the solution. Motivated by the boundary conditions (2.103) and (2.104), we look for a velocity
eld of the form,
v = v

(r, ) e

. (2.105)
Moreover, the boundary conditions will be met if v

depends linearly on r, so we look for v

of the
particular form,
v = r () e

. (2.106)
51
From the assumed form of the velocity eld, (2.106), it is straightforward to calculate several
kinematic variables (see, e.g. Appendix G),
[D] =

2
_
_
0 0 0
0 0 1
0 1 0
_
_
,
_
D
2

=

2
4
_
_
0 0 0
0 1 0
0 0 1
_
_
, (2.107)
where
sin
d
d
_

sin
_
. (2.108)
For this ow eld, we can evaluate (2.52) to show S =
2
. Therefore, the only non-zero components
of the extra stress tensor for a Generalized Newtonian uid are (2.53),

= (
2
) . (2.109)
Since is only a function of ,

is also only a function of . From (2.106) and (2.109) the

equation of linear momentum reduces to,
r
2
=
p
r
,

2
r
2
cot =
p

,
0 =
1
sin
p

+
1
sin
2

sin
2
).
(2.110)
From (2.110) and because

depends only on , we can conclude that p/ is a constant.

Furthermore, in order that the boundary condition (2.104) be met for all , necessarily, this constant
is zero and hence p is at most dependent on r and .
Neglect Inertial Eects
By dierentiating (2.110)
1
with respect to and (2.110)
2
with respect to r we obtain,

(
p
r
) = 2 r
d
d

r
(
p

) = 2 r
2
cot .
(2.111)
It then follows that either = 0 or = Asin where A is a constant of integration. Neither of these
solutions can satisfy both boundary conditions. Therefore, our assumed form of the velocity eld
is only meaningful in the limit of negligible inertial eects in which case, the rst two components
of the balance of linear momentum reduce to the condition that p is independent of r and and
hence p is constant throughout the ow eld. It then follows from the condition at the free surface
(2.104), the fact that
r
=
r
are the only non-zero components of the extra stress, the free
surface must be a surface of constant radius and p = p
atm
everywhere in the uid.
52
Relationship between
r
and m
z
.
It follows from the third component of linear momentum (2.111)
3
, that

=
C
1
sin
2

, (2.112)
where C
1
is a constant of integration. It is convenient to write this constant in terms of measurable
quantities, in particular the torque to the uid by the cone. m = m
z
e
z
. Recall, that in general,
the applied torque on a surface with outward unit normal n is related to the stress vector t acting
on that surface through
m =
_
A
x t|

1
da (2.113)
where x is the lever arm, da is the elemental area. At the uid surface adjacent to the cone, is
equal to /2 and
x = r cos (sine

+ cos e
r
),
da = r cos dr d,
n = e

t|

1
= T n
= pe

1
e

(2.114)
and therefore,
x t|

1
= p r cos
2
e

1
r cos e
z
, (2.115)
where the base vector,
e
z
= sine
r
cos e

(2.116)
is constant in space and therefore simplies the evaluation of the integral in (2.113). Noting that,

1
=
C
1
cos
2

(2.117)
if follows from (2.113) -(2.117) that
m =
C
1
2R
3
o
3
e
z
. (2.118)
Using the condition on the applied torque m, we can determine C
1
from (2.118),
C
1
=
3m
z
2R
3
o
. (2.119)
Therefore,

=
3m
z
2R
3
o
sin
2

. (2.120)
We now have an expression for the shear stress in terms of the applied torque and geometry. To
obtain a relationship between the function (S), m
z
and it remains to use (2.120) in conjunction
with the expression for

for a particular constitutive equation.

53
2.5.1 Small Angle Approximation
In this section, we consider ow between a cone and plate for cone angle much less than one.
Using this approximation, we will be able to obtain a function form for (m
z
, R
0
, , ) without
specifying a particular form for the constitutive equation. This result is the theoretical basis for
the use of the cone and plate geometry in rheometers.
In considering the form of the governing equations under the small angle approximation, it is
convenient to introduce an angle dened relative to through the relation,
=

2
. (2.121)
In the uid domain, is always less than or equal to which is much less than one. Using the
denition (2.121), we see that,
cos = sin = + 0(
3
) sin = cos = 1 + 0(
2
). (2.122)
Therefore,

=
C
1
sin
2

=
C
1
cos
2

= C
1
+ 0(
2
). (2.123)
For small angles, the shear rate (2.108) becomes,
=
_
1 + 0(
2
)
_

_
(1 + 0(
2
))
_
. (2.124)
Recalling that (
2
) is equal to

/ which is a constant. It follows from (2.123) that for small

either is proportional to 1/ or is a constant. It then follows from (2.124) that
= C
2
+C
3
+ 0(
2
). (2.125)
We can obtain the constants of integration, C
2
and C
3
from the boundary conditions (2.103), so
that,
=

+ O(
2
). (2.126)
We can determine an approximation for , denoted by
A
by evaluating (2.124) using (2.126),

A
=

+O(
2
). (2.127)
Combining these results, we see that the approximate expression for , which well denote by
A
is:

A
(
2
A
) =
3m
z
2R
3
o

(
A
)
=


A
assuming
1. Velocity eld of the form v = r()e

2. Negligible body forces

3. Negligible Inertial Eects
4. Free Surface of Fluid at r = R
5. Small Cone Angle ( << 1)
(2.128)
54
If m
z
and are measured for a given geometry (R
0
, ), the functional form
A
(
A
2
) can be
determined for the range of measured values. Note that for the small angle approximation (ignoring
inertial eects) both

and
A
are constant through out the uid domain.
2.5.2 Finite Cone Angle
The relationship, (2.128), relating
A
to of the applied torque, , and R
o
was derived for small
cone angle and negligible inertial eects. It is natural to wonder how much the solutions change
when these assumptions are not imposed. While it is more complicated to consider the eects
of inertia, we can obtain some closed form solutions even when we do not make the small angle
approximation. To obtain the corresponding relationship to (2.128) for a nite cone angle, we must
consider particular constitutive equations. In the following section, we obtain solutions for cone
and plate ow for nite angle for Bingham uids and some Power-Law uids. We still impose the
negligible inertia assumption.
Power-Law uid in cone and plate rheometer at nite angle
For a Power-Law uid undergoing steady ow of the form, (2.106), the only non-zero component
of the extra stress tensor is,

= (
2
) = K
n
. (2.129)
Combining this result with the form for

obtained from the equation of linear momentum for a

generalized Newtonian uid

= C
1
/sin
2
, we have that
= sin
d
d
_

sin
_
=
_
C
1
K
_
1/n
sin
2/n
(2.130)
and therefore
v

= r sin
_
C
1
K
_
1/n
_
/2

sin
12/n
d (2.131)
We can obtain an expression for by evaluating (2.131) at the cone surface and making use of the
boundary condition (2.103). The integral in (2.131) can be easily evaluated when (1 + 2/n) is an
integer. If we dene

f() as,

f() =
_
/2
/2
sin
12/n
d (2.132)
then,
=
_
C
1
K
_
1/n

f(), (2.133)
where the function

f() depends on the particular Power-Law uid, in particular, on the choice of
n. From (2.130) and (2.133), we have that
=

f() sin
2/n

,
w
=

f() cos
2/n

. (2.134)
55
If we now evaluate (2.129) at the surface of the cone (denoting the value of at equal to (/2)
by
w
), we obtain
(
2
w
) =

(/2 )

w
=
C
1

f()(cos )
2(1/n1)
. (2.135)
We now consider specic Power-Law uids, and calculate the values of

f() for these uids.
n = 1/2
We rst consider a shear thinning uid with n = 1/2, In this case,

f() =
_
/2
/2
sin
5
d.
(2.136)
From (2.135)
(
2
w
) =
C
1

f() cos
2
. (2.137)
and therefore, at the wall,
(
2
w
) =
3 m
z
2 R
3
o

_
3 sin
3
+ 5 sin
8 cos
2

3
16
cos
2
ln(
1 sin
1 + sin
)
_
. (2.138)
In the limit of much less than one, the expression in brackets in (2.138) is equal to plus higher
order terms. As expected, the expression for ( ) can be approximated by
A
, where

A
=
3 m
z

2 R
3
o

, (2.139)
as deduced in the preceeding section on small cone angle.
n = 2
For a shear thickening uid with n = 2,

f() =
_
/2
/2
sin
2
d = tan. (2.140)
Therefore at the wall of the cone, is,
(
2
w
) =
3 m
z
2 R
3
o
sin
cos
2

. (2.141)
In the limit of much less than one, we can approximate by
A
, (2.139).
We see from Fig. 2.5.2 that the error for both shear thickening and shear thinning uids is on
the order of a few percent for typical values of cone angles of 5-6 degrees.
56
Figure 2.5: Eect of small angle approximation: /
A
for Power-Law uids with n=1/2 and n=2.
Bingham Fluid in a cone and plate rheometer with nite cone angle
In this subsection, we consider the ow of a Bingham uid in a cone and plate rheometer. Recall
that the only non-zero components of the extra stress tensor are

as given in Eq. 2.112.

It then follows that the square root of the magnitude of the second invariant of the extra stress
tensor is,
_
|II

| = |

| =
3m
z
2R
3
o
sin
2

for [/2 , /2], (2.142)

which is a maximum at = /2 . We therefore expect that as the torque is increased from
zero, the yield criterion
_
|II

| =
Y
will rst be met at the surface of the cone. The critical torque
necessary for this condition to be met is,
m
Y
=

Y
2R
3
o
cos
2

3
. (2.143)
For m
z
< m
Y
, the cone will not move. For m
z
greater than (2.143), the uid will ow in some
regions of the uid domain. The yield surface is dened by the angle where
_
|II

| =
Y
,

Y
= sin
1
_
3m
z
2R
3
o
_
. (2.144)
57
Therefore, for a Bingham uid,
D

=
_

_
1
2
_
_
1

Y
_
|II

|
_
_

, for [/2 ,
Y
]
0 for [
Y
, /2]
(2.145)
From (2.112) and (2.119) we see that

is negative and therefore, we can rewrite (2.145) as

D

=
_

_
1
2
_

C
1
sin
2

+
Y
_
for [/2 ,
Y
]
0 for [
Y
, /2].
(2.146)
From (2.107), (2.146) and (2.108), we can now solve for ().
sin
d
d
_

sin
_
=
1

C
1
sin
2

+
Y
_
for [/2 ,
Y
]
sin

(

sin
) = 0 for [
Y
, /2].
(2.147)
Integrating (2.147) we nd,
=
sin

_
C
1
2
_
cos
sin
2

ln(tan(

2
))
_
+
Y
ln(tan(

2
)) +C
2
_
for [/2 ,
Y
]
= C
3
sin for [
Y
, /2].
(2.148)
The constants of integration C
2
and C
3
must be chosen such that the no-slip condition is met at the
lower plate ((/2) = 0) and so that the velocity is continuous throughout the ow eld. From the
condition at the lower plate, we nd that C
3
is equal to zero and therefore v

is zero at all points

in the uid where is greater than or equal to
Y
, in particular it is zero at the yield surface. We
can obtain an expression for C
2
from the condition that the velocity eld be continuous at the yield
surface. Hence,
v

=
r sin

_
C
1
2
_
cot
sin

cot
Y
sin
Y
_
+ ln(
tan(

2
)
tan(

Y
2
)
)
_

C
1
2
+
Y
_
_

_ for [/2 ,
Y
]
v

= 0 for [
Y
, /2].
(2.149)
58
The boundary condition on the upper plate, (2.103), then provides a relationship between , and
m
z
:
=
1

_
C
1
2
_
sin
cos
2

cot
Y
sin
Y
_
+ ln(
tan(
/2
2
)
tan(

Y
2
)
)
_

C
1
2
+
Y
_
_

_. (2.150)
(recall that C
1
= 3m
z
/(2R
3
o
). Equation (2.150) holds when
Y
is greater than or equal to /2.
If not, the value of is zero. Namely, the applied torque must be suciently large that the uid
begins to yield at the surface of the cone.
Once the uid is yielding throughout the domain, (
Y
= /2),
=
1

_
C
1
2
_
cot
sin
_
+ ln(tan(

2
))
_

C
1
2
+
Y
__
. (2.151)
For small , (2.151) becomes,

_
3m
z
2R
3
o

Y
_
where m
z
m
Y
= 2R
3
o

Y
cos
2
/3.
(2.152)
59
60
Appendix A
The Relationship between Coordinate
Free Notation and Indicial Notation
It is sometimes convenient to write expressions in terms of coordinate free notation rather than
with respect to a particular coordinate system. For example, in coordinate free notation, we would
refer to a vector by the symbol v where as in indicial notation, we would write the vector as the
sum of components relative to a particular basis, e
i
,
v = v
i
e
i
(A.1)
where in (A.1) it is understood that repeated indices are summed from one to three. The following
is a table containing some commonly used relationships between coordinate free and index notation.
Consider vectors v and w and second order tenors A and B,
Coordinate Free Notation Indicial Notation
v = v
i
e
i
v w = v
i
w
i
v w =
ijk
v
i
w
j
e
k
A = A
ij
e
i
e
j
A v = A
ij
v
j
e
i
A
T
v = v A = A
ji
v
j
e
i
A B = A
ij
B
jk
e
i
e
k
A
T
B = A
ji
B
jk
e
i
e
k
A B
T
= A
ij
B
kj
e
i
e
k
trA = A
ii
A : B = A
ij
B
ij
(A.2)
61
The Denition of Some Frequently Used Operators
In this section, we recall the denitions with respect to a xed basis e
i
of some commonly used
operators. We will consider the result of operations on scalar elds (x), vector elds v(x) and
w(x) as well as second order tensor elds A(x). Unfortunately, the denition of these operators
is not universal in the literature. In this class, we will use the denitions given in the following
table. Signicantly, in many uid mechanics books these operators are dened as the transpose of
Operator Symbol Definition
grad

x
j
e
j
grad v v
v
i
x
j
e
i
e
j
div v v
v
i
x
i
div A A
A
ij
x
j
e
i
curl v v
ijk
v
k
x
j
e
i
div grad v (v)or
2
v

2
v
i
x
j
x
j
e
i
(w) v v
j
w
i
x
j
e
i
(A.3)
those operations given here. It is important to take care when reading and writing publications
which use these operators. For example, in uid mechanics literature, the operator (w) v is often
written as v w. In addition, sometimes the curl of a vector is dened as the negative of that
given above.
62
Appendix B
Isotropic Tensors
The discussion in this section follows from [21]. Consider a rectangular Cartesion coordinate system
with basis, (e
i
), and a second Cartesion coordinate system which is rotated relative to the rst,
with basis (e

i
). The relationship between (e
i
), and (e

i
) can be represented as
e
i
= A
ij
e

j
(C.1)
where A
ij
are components of a proper orthogonal tensor,
A
ij
A
kj
=
ik
, det A = 1. (C.2)
Therefore, we also have
e

j
= A
ij
e
i
(C.3)
Consider an arbitrary vector v. We refer to the components of this vector with respect to basis
(e
i
), and (e

i
) as v
i
and v

i
, respectively. Namely,
v = v
i
e
i
= v

i
e

i
. (C.4)
Using this notation,
v = v
i
e
i
= v

j
e

j
= v

j
A
kj
e
k
(C.5)
Therefore,
v
l
= v

j
A
kj
e
k
e
l
= A
lj
v

j
(C.6)
Similarly, it can be shown that, the components M
ijk...t
of a tensor M will transform as
M
ijk...t
= A
ia
A
jb
A
kc
...A
th
M

abc...h
. (C.7)
63
A tensor M is called isotropic if its components retain the same value regardless of the coordinate
axis rotation,
M
ijk...t
= M

ijk...t
(C.8)
for all proper orthogonal A
ij
. Alternatively,
M
ijk...t
= A
ia
A
jb
A
kc
...A
th
M

abc...h
. (C.9)
Theorem 1.: All scalar invariants are isotropic tensors of order 0.
Theorem 2.: The only isotropic tensor v of order 1 is the zero tensor.
Proof: If v is an isotorpic tensor of order 1, then the components of v are unchanged under an
proper orthogonal transformations:
v
i
= A
ij
v
j
(C.10)
for all proper orthogonal A. Consider the case where the components of A are dened as
[A] =
_
_
1 0 0
0 1 0
0 0 1
_
_
. (C.11)
Upon consideration of equation (C.10), we see that this choice of A requires that
(v
1
, v
2
, v
3
) = (v
1
, v
2
, v
3
). (C.12)
This implies that v
1
= 0, v
2
= 0. Similarly, if we now consider the case where the components of A
are equal to
[A] =
_
_
1 0 0
0 1 0
0 0 1
_
_
. (C.13)
then necessarily (v
1
, v
2
, v
3
) = (v
1
, v
2
, v
3
). so that v
3
must also be zero. Therefore, if v is an
isotropic tensor of order 1, then necessarily v must be the zero vector. It is clear that this is also a
sucient condition.
Theorem 3.: Any isotropic tensor of order 2, can be written as a scaler multiple of the iden-
tity tensor. Namely, if B is a second order isotropic tensor, then the most general form of B can
be written as
B = I, B
ij
=
ij
(C.14)
where is a scalar.
Theorem 4.: The components of the most general isotropic tensor B of order 3 can be written as
B
pqr
=
ijk
(C.15)
64
where is a scalar.
Proof: If B is an isotropic tensor of order 3, then the components of B are unchanged under an
proper orthogonal transformations:
B
ijk
= A
ia
A
jb
A
kc
B
abc
(C.16)
for all proper orthogonal A. Consider the following three choices for the components of A,
[A] =
_
_
0 0 1
1 0 0
0 1 0
_
_
, [A] =
_
_
0 0 1
1 0 0
0 1 0
_
_
, [A] =
_
_
0 0 1
0 1 0
1 0 0
_
_
. (C.17)
If we consider (C.10) for these three proper orthogonal choices for A, we ned that There are only
six non-zero components of B, and they are related as
B
123
= B
231
= B
312
= B
321
= B
213
= B
132
, (C.18)
and therefore necessarily B
ijk
must be of the form
B
ijk
=
ijk
(C.19)
where lambda is a scaler. The form (C.19) is also sucient:
A
ia
A
jb
A
kc
B
abc
= A
ia
A
jb
A
kc

abc
= det(A
ab
)
ijk
=
ijk
= B
ijk
(C.20)
Theorem 5.: The components of the most general isotropic tensor, B of order 4 can be written
as
B
ijkl
=
ij

kl
+
ik

jl
+
il

jk
. (C.21)
Exercise B.0.1 Prove Theorem 2, making sure to show both necessary and sucient conditions.
65
66
Appendix C
Isotropic Tensor Functions
C.1 Scalar valued isotropic tensor functions
A scalar valued function of a second order tensor =

(B) is said to be an isotropic tensor function
if
(B
ij
) = (B

ij
) = (A
ik
B
kl
A
jl
) (D.1)
for all proper orthogonal A. Such a scalar valued function is also called simply an invariant. The
following isotropic scalar functions I
B
, II
B
III
B
are called the Principal Invariants of a second order
tensor B,
I
B
= trB = B
ii
II
B
=
1
2
_
tr(B)
2
tr(B B)
_
=
1
2
(B
ii
B
jj
B
ij
B
ji
)
III
B
= det B
(D.2)
The set of invariants (D.2) arise frequently as the coecients of the characteristic equation for a
second order tensor. Another set of invariants are,

I
B
= trB = B
ii

II
B
= trB
2
= B
ij
B
ji

III
B
= trB
3
= B
ij
B
jk
B
ki
(D.3)
Representation of Invariant Scalar Valued functions of a Symmetric Tensor
A scalar valued function of a symmetric tensor (B) is an invariant if and only if it can be written
as a function of the principal invariants of the symmetric tensor B.
(B) =

(I
B
, II
B
, III
B
) (D.4)
67
C.2 Symmetric Isotropic Tensor Functions
Consider a second order tensor valued function F =

F(B
1
, B
2
, ...B
N
) where B
i
are second order
tensors. The second order tensor function F is said to be an isotropic tensor function if
A

F(B
1
, B
2
, ...B
N
) A
T
=

F(A B
1
A
T
, A B
2
A
T
, ...A B
N
A
T
) (D.5)
for all proper orthogonal A and all B
i
in the domain of denition of the function F.
Representation of Symmetric Isotropic Tensor Functions
A tensor function F =

F(B) for which both F and B are symmetric second order tensors, is
isotropic if and only if it has a representation of the form,
F =

F(B) =
o
I +
1
B +
2
B
2
(D.6)
where
1
,
2
,
3
are functions of the principal invariants of B.
This theorem can be generalized for tensors which are not of second order (see, for example,
Truesdell, Non-Linear Field Theories of Mechanics, Handbuch Der Physik, Volume III/3, 1965, p
32), [31].
68
Appendix D
Material Line Elements, Area
Elements, Volume Elements
Some useful results for material line elements, area elements and volume elements are summarized
below:
ds
2
= F
iA
F
iB
dX
A
dX
B
. (E.1)
da
i
= J X
B,i
dA
B
. (E.2)
dv = J dV (E.3)
Dv
Dt
=
DJ
Dt
dV = J v
i,i
dV (E.4)
Proofs will be added to this section at a later time.
69
70
Appendix E
Rectangular Coordinates
In this appendix we consider rectangular components (x, y, z) with corresponding orthonormal
basis, (e
x
, e
y
, e
z
).
An arbitrary vector may be written with respect to rectangular coordinates as,
v = v
i
e
i
= v
x
e
x
+ v
y
e
y
+ v
z
e
z
. (F.1)
The following are some commonly used operators written with respect to rectangular components.
Gradient of a scalar function f
f =
f
x
i
e
i
=
f
x
e
x
+
f
y
e
y
+
f
z
e
z
. (F.2)
Laplacian of a scalar function f

2
f
x
2
+

2
f
y
2
+

2
f
z
2
. (F.3)
Gradient of a Vector, v
The gradient of a vector v, often denoted as L, is dened as
[v] =
_
v
i
x
j
e
i
e
j
_
=
_

_
v
x
x
v
x
y
v
x
z
v
y
x
v
y
y
v
y
z
v
z
x
v
z
y
v
z
z
_

_
(F.4)
Divergence of a Vector, v
v =
v
i
x
i
=
v
x
x
+
v
y
y
+
v
z
z
(F.5)
71
Curl of a Vector, v
v =
ijk
v
j
x
i
e
k
=
_
v
z
y

v
y
z
_
e
x
+
_
v
x
z

v
z
x
_
e
y
+
_
v
y
x

v
x
y
_
e
z
. (F.6)
Divergence of a Second Order Tensor, A
A =
A
ij
x
j
e
i
=
_
A
xx
x
+
A
xy
y
+
A
xz
z
_
e
x
+
_
A
yx
x
+
A
yy
y
+
A
yz
z
_
e
y
+
_
A
zx
x
+
A
zy
y
+
A
zz
z
_
e
z
.
(F.7)
Balance of Linear Momentum
The balance of linear momentum can be written with respect to rectangular coordinates, as

_
v
x
t
+ v
x
v
x
x
+ v
y
v
x
y
+ v
z
v
x
z
_
=
T
xx
x
+
T
xy
y
+
T
xz
z
+ g
x
,

_
v
y
t
+ v
x
v
y
x
+ v
y
v
y
y
+ v
z
v
y
z
_
=
T
yx
x
+
T
yy
y
+
T
yz
z
+ g
y
,

_
v
z
t
+ v
x
v
z
x
+ v
y
v
z
y
+ v
z
v
z
z
_
=
T
zx
x
+
T
zy
y
+
T
zz
z
+ g
z
.
(F.8)
If the stress tensor , T is decomposed into the sum of the spherical and deviatoric parts,
T = pI + , (F.9)
then we can rewrite (F.8) as,

_
v
x
t
+ v
x
v
x
x
+ v
y
v
x
y
+ v
z
v
x
z
_
=
p
x
+

xx
x
+

xy
y
+

xz
z
+ g
x
,

_
v
y
t
+ v
x
v
y
x
+ v
y
v
y
y
+ v
z
v
y
z
_
=
p
y
+

xy
x
+

yy
y
+

yz
z
+ g
y
,

_
v
z
t
+ v
x
v
z
x
+ v
y
v
z
y
+ v
z
v
z
z
_
=
p
z
+

xz
x
+

yz
y
+

zz
z
+ g
z
.
(F.10)
72
For the special case of a Navier-Stokes (incompressible Newtonian) uid, the balance of Linear
momentum can be written as,

_
v
x
t
+ v
x
v
x
x
+ v
y
v
x
y
+ v
z
v
x
z
_
=
p
x
+
_

2
v
x
x
2
+

2
v
x
y
2
+

2
v
x
z
2
_
+ g
x
,

_
v
y
t
+ v
x
v
y
x
+ v
y
v
y
y
+ v
z
v
y
z
_
=
p
y
+
_

2
v
y
x
2
+

2
v
y
y
2
+

2
v
y
z
2
_
+ g
y
,

_
v
z
t
+ v
x
v
z
x
+ v
y
v
z
y
+ v
z
v
z
z
_
=
p
z
+
_

2
v
z
x
2
+

2
v
z
y
2
+

2
v
z
z
2
_
+ g
z
.
(F.11)
73
74
Appendix F
Cylindrical Coordinates
In this appendix we consider cylindrical coordinates (r, , z), dened relative to rectangular com-
ponents (x, y, z) through,
Figure G.1: Cylindrical Coordinate System
x = r cos , y = r sin , z = z. (G.1)
The corresponding orthonormal base vectors (e
r
, e

, e
z
) are related to the rectangular basis (e
x
, e
y
, e
z
),
through
e
x
= cos e
r
sine

, e
y
= sine
r
+ cos e

, (G.2)
and e
z
is identical for both coordinate systems. An arbitrary vector may be written with respect
to cylindrical coordinates as,
v = v
r
e
r
+ v

+ v
z
e
z
. (G.3)
Gradient of a scalar function f
f =
f
r
e
r
+
1
r
f

+
f
z
e
z
. (G.4)
75
Laplacian of a scalar function f
f =
1
r

r
(r
f
r
) +
1
r
2

2
f

2
+

2
f
z
2
. (G.5)
Gradient of a Vector, v
[L] = [v] =
_

_
v
r
r
1
r
v
r

r
v
r
z
v

r
1
r
v

+
v
r
r
v

z
v
z
r
1
r
v
z

v
z
z
_

_
(G.6)
Divergence of a Vector, v
v =
1
r
(rv
r
)
r
+
1
r
v

+
v
z
z
(G.7)
Curl of a Vector, v
v =
_
1
r
v
z

z
_
e
r
+
_
v
r
z

v
z
r
_
e

+
1
r
_
(rv

)
r

v
r

_
e
z
. (G.8)
Divergence of a Second Order Tensor, A
A =
_
A
rr
r
+
1
r
A
r

+
A
rz
z
+
A
rr
A

r
_
e
r
+
_
A
r
r
+
1
r
A

+
A
z
z
+
A
r
+ A
r
r
_
e

+
_
A
zr
r
+
1
r
A
z

+
A
zz
z
+
A
zr
r
_
e
z
.
(G.9)
76
The operator A v for a second order tensor A
[A v]
rr
=
_
v
r
A
rr
r
+
v

r
A
rr

+ v
z
A
rr
z

v

r
(A
r
+ A
r
)
_
[A v]
r
=
_
v
r
A
r
r
+
v

r
A
r

+ v
z
A
r
z
+
v

r
(A
rr
A

)
_
[A v]
rz
=
_
v
r
A
rz
r
+
v

r
A
rz

+ v
z
A
rz
z

v

r
A
z
_
[A v]
r
=
_
v
r
A
r
r
+
v

r
A
r

+ v
z
A
r
z
+
v

r
(A
rr
A

)
_
[A v]

=
_
v
r
A

r
+
v

r
A

+ v
z
A

z
+
v

r
(A
r
+ A
r
)
_
[A v]
z
=
_
v
r
A
z
r
+
v

r
A
z

+ v
z
A
z
z
+
v

r
A
rz
_
[A v]
zr
=
_
v
r
A
zr
r
+
v

r
A
zr

+ v
z
A
zr
z

v

r
A
z
_
[A v]
z
=
_
v
r
A
z
r
+
v

r
A
z

+ v
z
A
z
z
+
v

r
A
zr
_
[A v]
zz
=
_
v
r
A
zz
r
+
v

r
A
zz

+ v
z
A
zz
z
_
(G.10)
Balance of Linear Momentum
The balance of linear momentum can be written with respect to cylindrical coordinates, as

_
v
r
t
+ v
r
v
r
r
+
v

r
v
r

+ v
z
v
r
z

v
2

r
_
=
T
rr
r
+
1
r
T
r

+
T
rz
z
+
T
rr
T

r
+ g
r
,

_
v

t
+ v
r
v

r
+
v

r
v

+ v
z
v

z
+
v
r
v

r
_
=
T
r
r
+
1
r
T

+
T
z
z
+ 2
T
r
r
+ g

_
v
z
t
+ v
r
v
z
r
+
v

r
v
z

+ v
z
v
z
z
_
=
T
rz
r
+
1
r
T
z

+
T
zz
z
+
T
zr
r
+ g
z
.
(G.11)
77
Using the decomposition of the tensor T into its spherical and deviatoric parts, we can rewrite
(H.10) as

_
v
r
t
+ v
r
v
r
r
+
v

r
v
r

+ v
z
v
r
z

v
2

r
_
=

p
r
+

rr
r
+
1
r

+

rz
z
+

rr

r
+ g
r
,

_
v

t
+ v
r
v

r
+
v

r
v

+ v
z
v

z
+
v
r
v

r
_
=

1
r
p

+

r
r
+
1
r

+

z
z
+ 2

r
r
+ g

_
v
z
t
+ v
r
v
z
r
+
v

r
v
z

+ v
z
v
z
z
_
=

p
z
+

rz
r
+
1
r

+

zz
z
+

zr
r
+ g
z
.
(G.12)
For the special case of a Navier-Stokes (incompressible Newtonian) uid, the balance of linear
momentum can be written as,

_
v
r
t
+ v
r
v
r
r
+
v

r
v
r

+ v
z
v
r
z

v
2

r
_
=

p
r
+
_

r
_
1
r
(rv
r
)
r
_
+
1
r
2

2
v
r

2
+

2
v
r
z
2

2
r
2
v

_
+ g
r
,

_
v

t
+ v
r
v

r
+
v

r
v

+ v
z
v

z
+
v
r
v

r
_
=

1
r
p

+
_

r
_
1
r
(rv

)
r
_
+
1
r
2

2
v

2
+

2
v

z
2
+
2
r
2
v
r

_
+ g

_
v
z
t
+ v
r
v
z
r
+
v

r
v
z

+ v
z
v
z
z
_
=

p
z
+
_
1
r

r
_
r
v
z
r
_
+
1
r
2

2
v
z

2
+

2
v
z
z
2
_
+ g
z
.
(G.13)
78
Appendix G
Spherical Coordinates
In this appendix we consider spherical coordinates (r, , ), dened relative to rectangular compo-
nents (x, y, z) through,
Figure H.1: Spherical Coordinate System
x = r sincos , y = r sin sin, z = r cos . (H.1)
The corresponding orthonormal base vectors (e
r
, e

, e

) are related to the rectangular basis (e

x
, e
y
, e
z
),
through
e
x
= cos
_
sin e
r
+ cos e

_
sine

,
e
y
= sin
_
sine
r
+ cos e

_
+ cos e

,
e
z
= cos e
r
sin e

.
(H.2)
An arbitrary vector may be written with respect to rectangular coordinates as,
v = v
r
e
r
+ v

+ v

. (H.3)
79
Gradient of a scalar function f
f =
f
r
e
r
+
1
r
f

+
1
r sin
f

. (H.4)
Laplacian of a scalar function f

2
f =
1
r
2

r
(r
2
f
r
) +
1
r
2
sin

(sin
f

) +
1
r
2
sin
2

2
f

2
. (H.5)
Gradient of a Vector, v
[L] = [v] =
_

_
v
r
r
1
r
v
r

r
1
r sin
v
r

r
v

r
1
r
v

+
v
r
r
1
r sin
v

r
cot
v

r
1
r
v

1
r sin
v

+
v
r
r
+
v

r
cot
_

_
(H.6)
Divergence of a Vector, v
v =
1
r
2
(r
2
v
r
)
r
+
1
r sin
(v

sin )

+
1
r sin
v

(H.7)
Curl of a Vector, v
v =
1
r sin
_
(v

sin)

_
e
r
+
_
1
r sin
v
r

1
r
(rv
)
r
_
e

+
1
r
_
(rv

)
r

v
r

_
e

.
(H.8)
Divergence of a Second Order Tensor, A
A =
_
A
rr
r
+
1
r
A
r

+
1
r sin
A
r

+
2A
rr
A

+A
r
cot
r
_
e
r
+
_
A
r
r
+
1
r
A

+
1
r sin
A

+
2A
r
+ A
r
+ (A

) cot
r
_
e

+
_
A
r
r
+
1
r
A

+
1
r sin
A

+
2A
r
+ A
r
+ (A

+ A

) cot
r
_
e

.
(H.9)
80
Balance of Linear Momentum
The balance of linear momentum can be written with respect to cylindrical coordinates, as

_
v
r
t
+ v
r
v
r
r
+
v

r
v
r

+
v

r sin
v
r

(v
2

+ v
2

)
r
_
=
T
rr
r
+
1
r
T
r

+
1
r sin
T
r

+
2T
rr
T

+T
r
cot
r
+ g
r
,

_
v

t
+ v
r
v

r
+
v

r
v

+
v

r sin
v

+
(v
r
v

v
2

cot )
r
_
=
T
r
r
+
1
r
T

+
1
r sin
T

+
3T
r
+ (T

) cot )
r
+ g

_
v

t
+ v
r
v

r
+
v

r
v

+
v

r sin
v

+
(v
r
v

+ v

cot )
r
_
=
T
r
r
+
1
r
T

+
1
r sin
T

+
3T
r
+ 2T

cot
r
+ g

.
(H.10)
Using the decomposition of the tensor T into its spherical and deviatoric parts, we can rewrite
(H.10) as

_
v
r
t
+ v
r
v
r
r
+
v

r
v
r

+
v

r sin
v
r

(v
2

+ v
2

)
r
_
=

p
r
+

rr
r
+
1
r

+
1
r sin

+
2
rr

+
r
cot
r
+ g
r
,

_
v

t
+ v
r
v

r
+
v

r
v

+
v

r sin
v

+
(v
r
v

v
2

cot )
r
_
=

1
r
p

+

r
r
+
1
r

+
1
r sin

+
3
r
+ (

) cot
r
+ g

_
v

t
+ v
r
v

r
+
v

r
v

+
v

r sin
v

+
(v
r
v

+ v

cot )
r
_
=

1
r sin
p

+

r
r
+
1
r

+
1
r sin

+
3
r
+ 2

cot
r
+ g

.
(H.11)
81
For the special case of a Navier-Stokes (incompressible Newtonian) uid, the balance of linear
momentum can be written as,

_
v
r
t
+ v
r
v
r
r
+
v

r
v
r

+
v

r sin
v
r

(v
2

+v
2

)
r
_
=
p
r
+

r
_
1
r
2
(r
2
v
r
)
r
_
+
1
r
2
sin

(sin
v
r

) +
1
r
2
sin
2

2
v
r

2

2
r
2
sin
_

(v

sin ) +
v

__
+g
r
,

_
v

t
+ v
r
v

r
+
v

r
v

+
v

r sin
v

+
(v
r
v

v
2

cot )
r
_
=
1
r
p

_
1
r
2

r
_
r
2
v

r
_
+
1
r
2

_
1
sin

(v

sin)
_
+
1
r
2
sin
2

2
v

2
+
2
r
2
v
r

2 cot
r
2
sin
v

_
+ g

_
v

t
+ v
r
v

r
+
v

r
v

+
v

r sin
v

+
(v
r
v

+ v

cot )
r
_
=
1
r sin
p

_
1
r
2

r
_
r
2
v

r
_
+
1
r
2

_
1
sin

(v

sin)
_
+
1
r
2
sin
2

2
v

2
+
2
r
2
sin
v
r

+
2 cot
r
2
sin
v

_
+ g

.
(H.12)
82
Bibliography
[1] G. Astarita and G. Marrucci, Principles of Non-Newtonian Fluid Mechanics, McGraw-
Hill Book Company (UK) Limited, 1974.
[2] H. A. Barnes, The yield stress- a review of , - everything ows?, J. Non-Newtonian
Fluid Mech., 81 (1999), pp. 133178.
[3] R. B. Bird, Useful non-Newtonian models, Ann. Rev. Fluid Mech., 8 (1976), pp. 1334.
[4] R. B. Bird, R. C. Armstrong, and O. Hassager, Dynamics of Polymeric Liquids, Volume
I, John Wiley & Sons, second ed., 1987.
[5] R. B. Bird, G. C. Dai, and B. J. Yarusso, The rheology and ow of viscoplastic materials,
Reviews in Chemical Engineering, 1 (1983), pp. 170.
[6] B. D. Coleman, Kinematical concepts with applications in the mechanics and thermodynam-
ics of incompressible viscoelastic uids, Arch. Rational Mech. Anal., 9 (1962), pp. 273300.
[7] B. D. Coleman, R. J. Duffin, and V. J. Mizel, Instability, uniqueness, and nonexistence
theorems for the equation u
t
= u
xx
u
xtx
on a strip, Arch. Rational Mech. Anal., 19 (1965),
pp. 100116.
[8] B. D. Coleman, H. Markovitz, and W. Noll, Viscometric Flows of Non-Newtonian
Fluids, Springer-Verlag, 1966.
[9] M. M. Cross, Rheology of non-Newtonian uids- A new ow equation for pseudoplastic sys-
tems, J. Colloid. Sci., 20(5) (1965), pp. 417437.
[10] J. E. Dunn and R. L. Fosdick, Thermodynamics, stability and boundedness of uids of
complexity 2 and uids of second grade, Arch. Rational Mech. Anal., 56 (1974), pp. 191252.
[11] G. Duvant and J. L. Lions, Inequalities in Mechanics and Physics, Springer-Verlag, 1982.
[12] G. P. Galdi, M. Grobbelaar-Van Dalsen, and N. Sauer, Existence and uniqueness of
classical solutions of the equations of a second-grade uid, Arch. Rat. Mech. Anal., 124 (1993),
pp. 221235.
[13] G. P. Galdi and B. D. Reddy, Well-posedness of the problem of ber suspension ows, J.
Non-Newtonian Fluid Mech., 83(3) (1999), pp. 205230.
83
[14] A. E. Green and P. M. Naghdi, A note on invariance under superposed rigid body motions,
J. Elasticity, 9 (1979), pp. 18.
[15] C. Guillop`e and J. C. Saut, Existence results for the ow of viscoelastic uids with a
dierential constitutive law, Nonlinear Analysis, Theory, Methods and Applications, 15 (1990),
pp. 849869.
[16] C. Guillop`e and J.-C. Saut, Existence and stability of steady ows of weakly viscoelastic
uids, Proc. Royal Soc. Edinburgh, Section A-Mathematics, 119 (1991), pp. 137158.
[17] D. D. Joseph, Instability of the rest state of uids of arbitrary grade greater than one, Arch.
Rational Mech. Anal., 75 (1981), pp. 251256.
[18] D. C. Leigh, Non-Newtonian uids and the second law of thermodynamics, Physics of Fluids,
5 (1962), pp. 501502.
[19] C. W. Macosko, Rheology: Principles, Measurements and Applications, VCH Publishers,
Inc., 1994.
[20] P. M. Naghdi, Mechanics of solids, in Handbuch der Physik, C. Truesdell, ed., vol. VIa/2,
Springer-Verlag, 1972, ch. The Theory of Shells and Plates, pp. 425640.
[21] , Lecture notes for Continuum Mechanics, ME 185, Department of Mechanical Engineer-
ing, University of California at Berkeley, 1987.
[22] J. G. Oldroyd, A rational formulation of the equations of plastic ow for a Bingham solid,
Proc. Camb. Philos. Soc., 43 (1947), pp. 100105.
[23] , Two-dimensional plastic ow of a Bingham solid, Proc. Camb. Philos. Soc., 43 (1947),
pp. 383395.
[24] T. C. Papanastasiou, Flows of materials with yield, J. of Rheology, 31 (1987), pp. 384404.
[25] W. Prager, Introduction to Mechanics of Continua, Ginn, 1961.
[26] E. L. Reiner, Handbuch der Physik, vol. 6, Springer-Verlag, 1958.
[27] Z. Tadmor and C. G. Gogos, Principles of Polymer Processing, John Wiley and Sons,
1979.
[28] C. Truesdell, A new denition of a uid. II. The Maxwellian uid, J. Math. Pure Appl.,
30 (1951), pp. 11155.
[29] , The mechanical foundations of elasticity and uid dynamics, J. Rational Mech. Anal.,
1 (1952), pp. 125300.
[30] C. Truesdell and W. Noll, Non-linear eld theories of mechanics, in Handbuch der Physik,
S.Flugge, ed., vol. III/3, Springer-Verlag, 1965, pp. 4147,.
84
[31] , Non-linear eld theories of mechanics, in Handbuch der Physik, S.Flugge, ed., vol. III/3,
Springer-Verlag, 1965.
[32] K. Yasuda, R. C. Armstrong, and R. E. Cohen, Shear ow properties of concentrated
solutions of linear and star branched polystyrenes, Rheol. Acta, 20(2) (1981), pp. 163178.
85