8XTjICF2ER Elements of Continuum Mechanics PDF

This page intentionally left blank
Elements of Continuum Mechanics and Thermodynamics
This text is intended to provide a modern and integrated treatment of the founda-
tions and applications of continuum mechanics. There is a significant increase in
interest in continuum mechanics because of its relevance to microscale phenomena.
In addition to being tailored for advanced undergraduate students and including
numerous examples and exercises, this text also features a chapter on continuum
thermodynamics, including entropy production in Newtonian viscous fluid flow and
thermoelasticity. Computer solutions and examples are emphasized through the
use of the symbolic mathematical computing program MathematicaÒ.
Joanne L. Wegner is a Professor of Mechanical Engineering at the University of

Victoria, Canada. Her research interests include wave propation in elastic and
viscoelastic materials with applications to the dynamic behavior of polymers and
elastomeric materials and earthquake engineering. Professor Wegner has been
recognized for her outstanding contributions to the engineering profession by the
Canadian Society of Mechanical Engineers. In addition to her archival publica-
tions, she has coedited two books on nonlinear waves in solids.
James B. Haddow received the Canadian Congress of Applied Mechanics award

in 2003. He has published papers on plasticity, nonlinear elasticity, wave propaga-
tion, and thermodynamics (exergy, thermoelasticity).
Elements of Continuum Mechanics
and Thermodynamics
Joanne L. Wegner
University of Victoria
James B. Haddow
University of Victoria
CAMBRIDGE UNIVERSITY PRESS
Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore, São Paulo
Cambridge University Press

The Edinburgh Building, Cambridge CB2 8RU, UK
Published in the United States of America by Cambridge University Press, New York
www.cambridge.org
Information on this title: www.cambridge.org/9780521866323
© Joanne L. Wegner and James B. Haddow 2009
This publication is in copyright. Subject to statutory exception and to the

provision of relevant collective licensing agreements, no reproduction of any part
may take place without the written permission of Cambridge University Press.
First published in print format 2007
ISBN-13 978-0-511-53375-4 eBook (EBL)
ISBN-13 978-0-521-86632-3 hardback
Cambridge University Press has no responsibility for the persistence or accuracy

of urls for external or third-party internet websites referred to in this publication,
and does not guarantee that any content on such websites is, or will remain,
accurate or appropriate.
Contents
Preface Page ix
1. Cartesian Tensor Analysis .................................................................... 1

1.1 Introduction 1
1.2 Rectangular Cartesian Coordinate Systems 3
1.3 Suffix and Symbolic Notation 4
1.4 Orthogonal Transformations 6
1.5 Concept of a Second-Order Tensor 11
1.6 Tensor Algebra 19
1.7 Invariants of Second-Order Tensors 20
1.8 Cayley-Hamilton Theorem 24
1.9 Higher-Order Tensors 25
1.10 Determinants 28
1.11 Polar Decomposition Theorem 32
1.12 Tensor Fields 38
1.13 Integral Theorems 44
EXERCISES 46
SUGGESTED READING 48
2. Kinematics and Continuity Equation ................................................ 50

2.1 Description of Motion 50
2.2 Spatial and Referential Descriptions 53
2.3 Material Surface 54
2.4 Jacobian of Transformation and Deformation Gradient F 56
2.5 Reynolds Transport Theorem 59
2.6 Continuity Equation 63
v
vi Contents
2.7 Velocity Gradient 64

2.8 Rate of Deformation and Spin Tensors 65
2.9 Polar Decomposition of F 70
2.10 Further Decomposition of F 74
2.11 Strain 75
2.12 Infinitesimal Strain 78
EXERCISES 81
3. Stress ...................................................................................................... 84
3.1 Contact Forces and Body Forces 84
3.2 Cauchy or True Stress 87
3.3 Spatial Form of Equations of Motion 90
3.4 Principal Stresses and Maximum Shearing Stress 91
3.5 Pure Shear Stress and Decomposition of the Stress Tensor 94
3.6 Octahedral Shearing Stress 95
3.7 Nominal Stress Tensor 98
3.8 Second Piola-Kirchhoff Stress Tensor 100
3.9 Other Stress Tensors 101
EXERCISES 102
REFERENCES 104
4. Work, Energy, and Entropy Considerations .................................. 105

4.1 Stress Power 105
4.2 Principle of Virtual Work 110
4.3 Energy Equation and Entropy Inequality 114
EXERCISES 119
REFERENCES 120
5. Material Models and Constitutive Equations ................................ 121

5.1 Introduction 121
5.2 Rigid Bodies 121
5.3 Ideal Inviscid Fluid 123
5.4 Incompressible Inviscid Fluid 124
5.5 Newtonian Viscous Fluid 125
5.6 Classical Elasticity 126
5.7 Linear Thermoelasticity 130
5.8 Determinism, Local Action, and Material
Frame Indifference 138
Contents vii
EXERCISES 147
REFERENCES 149
6. Finite Deformation of an Elastic Solid ........................................... 150

6.2 Cauchy Elasticity 150
6.3 Hyperelasticity 155
6.4 Incompressible Hyperelastic Solid 160
6.5 Alternative Formulation 162
EXERCISES 166
REFERENCES 167
7. Some Problems of Finite Elastic Deformation .............................. 168

7.2 Strain Energy Functions and Stress-Strain Relations 169
7.3 Simple Shear of a Rectangular Block 170
7.4 Simple Tension 172
7.5 Extension and Torsion of an Incompressible
Cylindrical Bar 175
7.6 Spherically Symmetric Expansion of a
Thick-Walled Shell 180
7.7 Eversion of a Cylindrical Tube 185
7.8 Pure Bending of an Hyperelastic Plate 189
7.9 Combined Telescopic and Torsional Shear 195
EXERCISES 200
REFERENCES 201
8. Finite Deformation Thermoelasticity ............................................. 202

8.1 Principle of Local State and Thermodynamic Potentials 202
8.2 Basic Relations for Finite Deformation Thermoelasticity 202
8.3 Ideal Rubber-Like Materials 209
8.4 Isentropic Simple Tension of Ideal Rubber 214
8.5 Isentropic Simple Tension of Compressible Rubber 215
EXERCISES 218
REFERENCES 219
9. Dissipative Media .............................................................................. 220

9.1 Newtonian Viscous Fluids 220
viii Contents
9.2 A Non-Newtonian Viscous Fluid 223

9.3 Linear Viscoelastic Medium 224
9.4 Viscoelastic Dissipation 237
9.5 Some Thermodynamic Considerations in
Linear Thermoelasticity 240
9.6 Simple Shear Problem 244
EXERCISES 246
REFERENCES 248
APPENDIX 1: Orthogonal Curvilinear Coordinate Systems .............. 249

A1.1 Introduction 249
A1.2 Curvilinear Coordinates 249
A1.3 Orthogonality 252
A1.4 Cylindrical and Spherical Polar Coordinate Systems 253
A1.5 The = Operator 256
A1.6 The =2 Operator 261
APPENDIX 2: Physical Components of the Deformation Gradient

Tensor ............................................................................... 263
APPENDIX 3: Legendre Transformation .............................................. 266
APPENDIX 4: Linear Vector Spaces ..................................................... 269
Index 272
Preface
This book is based on notes prepared for a senior undergraduate or begin-

ning graduate course that we taught at the universities of Alberta and
Victoria. It is primarily intended for use by students of mechanical and
civil engineering, but it may be of interest to others. The mathematical
background required for the topics covered in the book is modest and
should be familiar to senior undergraduate engineering students. In partic-
ular it is assumed that a reader has a good knowledge of classical vector
mechanics and linear algebra. Also, a background of the classical thermo-
dynamics usually taught in undergraduate engineering courses is desirable.
One motivation for the book is to present an introduction to continuum
mechanics that requires no background in certain areas of advanced math-
ematics such as functional analysis and general tensor analysis. The treat-
ment of continuum mechanics is based on Cartesian tensor analysis, but
orthogonal curvilinear coordinates and corresponding physical coordinates
are considered in appendices.
A list of books that consider tensor analysis and applications is given at
the end of chapter 1. Several of the books are out of print but may be useful
to students if they can be obtained from libraries. Mathematica is used for
symbolic manipulation, numerical computation, and graphs where appro-
priate, and its use is encouraged.
Chapter 1 is a detailed introduction to Cartesian tensor analysis. It
differs from some other treatments of the topic, for example, the early texts
by Jeffreys [1] and Temple [2], by emphasizing both symbolic and suffix
notation for first-(vectors) and second-order tensors. Many tensor relations
are given in both symbolic and Cartesian suffix notation. The introduction
ix
x Preface
of general tensor analysis involving contravariant and covariant compo-

nents of tensors and Christoffel symbols is avoided. However, two appen-
dices are included, which introduce the concept of physical components, of
first- and second-order tensors, with respect to orthogonal curvilinear
coordinates.
The kinematics of a deformable medium is the subject of chapter 2,
along with the continuity equation and Reynold’s transport theorem.
Cauchy, nominal (transpose of Piola-Kirchhoff I), Piola-Kirchhoff II, and
Biot stress tensors are introduced in chapter 3, along with equations of
motion. In chapter 4 the concept of stress power is related to the first and
second laws of thermodynamics. Continuum thermodynamics involves ir-
reversible processes, and what is known as the thermodynamics of irrevers-
ible processes (TIP) is assumed. Several classical models for continuous
media are discussed in chapter 5, and conditions that should be satisfied for
the admissibility of constitutive relations are discussed. Thermodynamic
considerations also appear in chapter 5, including a discussion of the con-
stitutive relations for linear thermoelasticity. Nonlinear isotropic elasticity
is the subject of chapters 6 and 7. The treatment is for isothermal static
deformation, and one purpose is to illustrate the application of certain
principles introduced in earlier chapters. Solutions to several boundary
value problems are given in chapter 7. Chapter 8 considers nonlinear ther-
moelasticity. The final chapter (9) is concerned with dissipative materials.
A brief discussion of viscous fluids is given, in addition to a discussion of
certain aspects of linear viscoelasticity. An elementary example of an in-
ternal variable is introduced in chapter 9 in connection with the three-
parameter standard model of linear viscoelasticity.
1 Cartesian Tensor Analysis
1.1 Introduction
In this chapter we present an elementary introduction to Cartesian tensor

analysis in a three-dimensional Euclidean point space or a two-dimensional
subspace. A Euclidean point space is the space of position vectors of points.
The term vector is used in the sense of classical vector analysis, and scalars
and polar vectors are zeroth- and first-order tensors, respectively. The
distinction between polar and axial vectors is discussed later in this chapter.
A scalar is a single quantity that possesses magnitude and does not depend
on any particular coordinate system, and a vector is a quantity that pos-
sesses both magnitude and direction and has components, with respect to
a particular coordinate system, which transform in a definite manner under
change of coordinate system. Also vectors obey the parallelogram law of
addition. There are quantities that possess both magnitude and direction
but are not vectors, for example, the angle of finite rotation of a rigid body
about a fixed axis.
A second-order tensor can be defined as a linear operator that oper-
ates on a vector to give another vector. That is, when a second-order tensor
operates on a vector, another vector, in the same Euclidean space, is gen-
erated, and this operation can be illustrated by matrix multiplication. The
components of a vector and a second-order tensor, referred to the same
rectangular Cartesian coordinate system, in a three-dimensional Euclidean
space, can be expressed as a ð3 3 1Þ matrix and a ð3 3 3Þ matrix, respec-
tively. When a second-order tensor operates on a vector, the components of
the resulting vector are given by the matrix product of the ð3 3 3Þ matrix of
components of the second-order tensor and the matrix of the ð3 3 1Þ
1
components of the original vector. These components are with respect to

a rectangular Cartesian coordinate system, hence the term Cartesian tensor
analysis. Examples from classical mechanics and stress analysis are as follows.
The angular momentum vector, h, of a rigid body about its mass center is
given by h ¼ Jv where J is the inertia tensor of the body about its mass
center and v is the angular velocity vector. In this equation the components
of the vectors h and v can be represented by ð3 3 1Þ matrices and the tensor
J by a ð3 3 3Þ matrix with matrix mutiplication implied. A further example
is the relation t ¼ sn, between the stress vector t acting on a material area
element and the unit normal n to the element, where s is the Cauchy stress
tensor. The relations h ¼ Jv and t ¼ sn are examples of coordinate-free
symbolic notation, and the corresponding matrix relations refer to a partic-
ular coodinate system.
We will meet further examples of the operator properties of second-
order tensors in the study of continuum mechanics and thermodynamics.
Tensors of order greater than two can be regarded as operators oper-
ating on lower-order tensors. Components of tensors of order greater than
two cannot be expressed in matrix form.
It is very important to note that physical laws are independent of any
particular coordinate system. Consequently, equations describing physical
laws, when referred to a particular coordinate system, must transform in
definite manner under transformation of coordinate systems. This leads to
the concept of a tensor, that is, a quantity that does not depend on the
choice of coordinate system. The simplest tensor is a scalar, a zeroth-order
tensor. A scalar is represented by a single component that is invariant
under coordinate transformation. Examples of scalars are the density of
a material and temperature.
Higher-order tensors have components relative to various coordi-
nate systems, and these components transform in a definite way under
transformation of coordinate systems. The velocity v of a particle is an
example of a first-order tensor; henceforth we denote vectors, in sym-
bolic notation, by lowercase bold letters. We can express v by its com-
ponents relative to any convenient coordinate system, but since v has no
preferential relationship to any particular coordinate system, there must
be a definite relationship between components of v in different
1.2 Rectangular Cartesian Coordinate Systems 3
coordinate systems. Intuitively, a vector may be regarded as a directed

line segment, in a three-dimensional Euclidean point space E3 , and the
set of directed line segments in E3 , of classical vectors, is a vector space
V3 . That is, a classical vector is the difference of two points in E3 . A
vector, according to this concept, is a first-order tensor. A discussion of
linear vector spaces is given in Appendix 4.
There are many physical laws for which a second-order tensor is an oper-
ator associating one vector with another. Remember that physical laws
must be independent of a coordinate system; it is precisely this indepen-
dence that motivates us to study tensors.
1.2 Rectangular Cartesian Coordinate Systems
The simplest type of coordinate system is a rectangular Cartesian system,

and this system is particularly useful for developing most of the theory to be
presented in this text.
A rectangular Cartesian coordinate system consists of an orthonormal
basis of unit vectors ðe1 , e2 , e3 Þ and a point 0 which is the origin. Right-
handed Cartesian coordinate systems are considered, and the axes in the
ðe1 , e2 , e3 Þ directions are denoted by 0x1 , 0x2 , and 0x3 , respectively, rather
than the more usual 0x, 0y, and 0z. A right-handed system is such that a 908
right-handed screw rotation along the 0x1 direction rotates 0x2 to 0x3 ,
similarly a right-handed rotation about 0x2 rotates 0x3 to 0x1 , and a right-
handed rotation about 0x3 rotates 0x1 to 0x2 .
A right-handed system is shown in Figure 1.1. A point, x 2 E3 , is given in
terms of its coordinates ðx1 , x2 , x3 Þ with respect to the coordinate system
0x1 x2 x3 by
x ¼ x1 e1 þ x2 e2 þ x3 e3 ,
which is a bound vector or position vector.

If points x; y 2 E3 , u ¼ x y is a vector, that is, u 2V3 . The vector u is
given in terms of its components ðu1 , u2 , u3 Þ; with respect to the rectangular
x3
e3
0 e2
x2
e1
x1
Figure 1.1. Right-handed rectangular Cartesian coordinate system.
coordinate system, 0x1 x2 x3 by
u ¼ u1 e1 þ u2 e2 þ u3 e3 :
Henceforth in this chapter when the term coordinate system is used,

a rectangular Cartesian system is understood. When the components of vec-
tors and higher-order tensors are given with respect to a rectangular Carte-
sian coordinate system, the theory is known as Cartesian tensor analysis.
1.3 Suffix and Symbolic Notation
Suffixes are used to denote components of tensors, of order greater than

zero, referred to a particular rectangular Cartesian coordinate system.
Tensor equations can be expressed in terms of these components; this is
known as suffix notation. Since a tensor is independent of any coordinate
system but can be represented by its components referred to a particular
coordinate system, components of a tensor must transform in a definite
manner under transformation of coordinate systems. This is easily seen
for a vector. In tensor analysis, involving oblique Cartesian or curvilinear
coordinate systems, there is a distinction between what are called contra-
variant and covariant components of tensors but this distinction disappears
when rectangular Cartesian coordinates are considered exclusively.
Bold lower- and uppercase letters are used for the symbolic rep-
resentation of vectors and second-order tensors, respectively. Suffix nota-
tion is used to specify the components of tensors, and the convention that
1.3 Suffix and Symbolic Notation 5
a lowercase letter suffix takes the values 1, 2, and 3 for three-dimensional

and 1 and 2 for two-dimensional Euclidean spaces, unless otherwise
indicated, is adopted. The number of distinct suffixes required is equal to
the order of the tensor. An example is the suffix representation of a vector u,
with components ðu1 , u2 , u3 Þ or ui , i 2 f1, 2, 3g. The vector is then given by
3
u ¼ + ui e i : ð1:1Þ
i¼1
It is convenient to use a summation convention for repeated letter suffixes.

According to this convention, if a letter suffix occurs twice in the same term,
a summation over the repeated suffix from 1 to 3 is implied without a sum-
mation sign, unless otherwise indicated. For example, equation (1.1) can be
written as
u ¼ ui ei ¼ u1 e1 þ u2 e2 þ u3 e3 ð1:2Þ
without the summation sign. The sum of two vectors is commutative and is
given by
u þ v ¼ v þ u ¼ ðui þ vi Þei ,
which is consistent with the parallelogram rule. A further example of the

summation convention is the scalar or inner product of two vectors,
u v ¼ ui vi ¼ u1 v1 þ u2 v2 þ u3 v3 : ð1:3Þ
Repeated suffixes are often called dummy suffixes since any letter that does
not appear elsewhere in the expression may be used, for example,
ui vi ¼ uj vj :
Equation (1.3) indicates that the scalar product obeys the commutative law
of algebra, that is,
u v ¼ v u:
The magnitude juj of a vector u is given by
pffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffi
juj ¼ u u ¼ ui ui :
Other examples of the use of suffix notation and the summation convention
are
Cii ¼ C11 þ C22 þ C33

Cij bj ¼ Ci1 b1 þ Ci2 b2 þ Ci3 b3 :
A suffix that appears once in a term is known as a free suffix and is un-
derstood to take in turn the values 1, 2, 3 unless otherwise indicated. If
a free suffix appears in any term of an equation or expression, it must
appear in all the terms.
1.4 Orthogonal Transformations
The scalar products of orthogonal unit base vectors are given by
ei ej ¼ dij , ð1:4Þ
where dij is known as the Kronecker delta and is defined as

1 for i ¼ j
dij ¼ : ð1:5Þ
0 for i 6¼ j
The base vectors ei are orthonormal, that is, of unit magnitude and mutu-
ally perpendicular to each other. The Kronecker delta is sometimes called
the substitution operator because
uj dij ¼ u1 di1 þ u2 di2 þ u3 di3 ¼ ui : ð1:6Þ
Consider a right-handed rectangular Cartesian coordinate system 0x9i with

the same origin as 0xi as indicated in Figure 1.2. Henceforth, primed quan-
tities are referred to coordinate system 0x9i .
x2
e2
x2′
e1′
x1′
e2′
e1
0 x1 Figure 1.2. Change of axes.
x3 e3
e3′
x3′
The coordinates of a point P are xi with respect to 0xi and x9i with
respect to 0xi . Consequently,
xi ei ¼ x9j e9j , ð1:7Þ
where the e9i are the unit base vectors for the system 0xi . Forming the inner
product of each side of equation (1.7) with e9k and using equation (1.4) and
the substitution operator property equation (1.6) gives
x9k ¼ aki xi , ð1:8Þ
where
aki ¼ e9k ei ¼ cosðx9k 0xi Þ: ð1:9Þ
Similarly
xi ¼ aki x9k : ð1:10Þ
It is evident that the direction of each axis 0x9k can be specified by giving its
direction cosines aki ¼ e9k ei ¼ cosðx9k 0xi Þ referred to the original axes
0xi . The direction cosines, aki ¼ e9k ei , defining this change of axes are
tabulated in Table 1.1.
The matrix ½a with elements aij is known as the transformation matrix;
it is not a tensor.
Table 1.1. Direction cosines for

rotation of axes
e1 9 e2 9 e3 9
e1 a11 a21 a31
e2 a12 a22 a32
e3 a13 a23 a33
It follows from equations (1.8) and (1.10) that
@x9k @xi
aki ¼ ¼ , ð1:11Þ
@xi @x9k
and from equation (1.7) that
@xi @x9j
ei ¼ ej9 ¼ e9k , ð1:12Þ
@x9k @x9k
since @xj =@xk ¼ @jk , and from equations (1.11) and (1.12) that
e9k ¼ aki ei , ð1:13Þ
and
ei ¼ aki e9k : ð1:14Þ
Equations (1.13) and (1.14) are the transformation rules for base vectors.
The nine elements of aij are not all independent, and in general,
aki 6¼ aik :
A relation similar to equations (1.8) and (1.10),
u9k ¼ aki ui , and ui ¼ aki u9k ð1:15Þ
is obtained for a vector u since ui ei ¼ u9k e9k , which is similar to equation

(1.7) except that the ui are the components of a vector and the xi are
coordinates of a point.
The magnitude juj ¼ ðui ui Þ1=2 of the vector u is independent of the

orientation of the coordinate system, that is, it is a scalar invariant;
consequently,
ui ui ¼ u9k u9k : ð1:16Þ
Eliminating ui from equation (1.15) gives
u9k ¼ aki aji u9j ,
and since u9k ¼ dkj u9j ,
aki aji ¼ dkj : ð1:17Þ
Similarly, eliminating uk from equation (1.15) gives
aik ajk ¼ dij : ð1:18Þ
It follows from equation (1.17) or (1.18) that
2
det aij ¼ 1, ð1:19Þ

where det aij denotes the determinant of aij . A detailed discussion of
determinants is given in section 10 of this chapter. The negative root of
equation (1.19) is not considered unless the transformation of axes involves
a change of orientation since, for the identity transformation xi ¼ x9i ,
aik ¼ dik and det½dik ¼ 1. Consequently, det½aik ¼ 1, provided the trans-
formations involve only right-handed systems (or left-handed systems).
The transformations (1.8), (1.10), and (1.15) subject to equation (1.17)
or (1.18) are known as orthogonal transformations. Three quantities ui are
the components of a vector if, under orthogonal transformation, they trans-
form according to equation (1.15). This may be taken as a definition of
a vector. According to this definition, equations (1.8) and (1.10) imply that
the representation x of a point is a bound vector since its origin coincides
with the origin of the coordinate system.
If the transformation rule (1.10) holds for coordinate transformations

from right-handed systems to left-handed systems (or vice versa), the vec-
tor is known as a polar vector. There are scalars and vectors known as
pseudo scalars and pseudo or axial vectors; there have transformation rules
that involve a change in sign when the coordinate transformation is from
a right-handed system to a left-handed system (or vice versa), that is, when

det aij ¼ 1. The transformation rule for a pseudo scalar is

/9 ¼ det aij /, ð1:20Þ
and for a pseudo vector

u9i ¼ det aij aij uj : ð1:21Þ
A pseudo scalar is not a true scalar if a scalar is defined as a single

quantity invariant under all coordinate transformations. An example of
a pseudo vector is the vector product u 3 v of two polar vectors u and v.
A discussion of the vector product is given in section 9 of this chapter. The
moment of a force about a point and the angular momentum of a particle
about a point are pseudo vectors. The scalar product of a polar vector and
a pseudo vector is a pseudo scalar; an example is the moment of a force
about a line. The distinction between pseudo vectors and scalars and polar
vectors and true scalars disappears when only right- (or left-) handed co-
ordinate systems are considered. For the development of continuum me-
chanics presented in this book, only right-handed systems are used.
EXAMPLE PROBLEM 1.1. Show that a rotation through angle p about an

axis in the direction of the unit vector n has the transformation matrix

aij ¼ dij þ 2ni nj , det aij ¼ 1:
SOLUTION. Referring to Figure 1.3, the position vector of point A has

components xi and point B has position vector with components x9i :
B c A
Figure 1.3. Rotation about axis with direction n.
Since AC ¼ CB, it follows from elementary vector analysis that
x9i ¼ xi þ 2ðxs ns ni xi Þ ¼ dij xj þ 2ðdsj ns ni Þxj ¼ dij xj þ 2ðnj ni Þxj
so that
aij ¼ dij þ 2ni nj :
1.5 Concept of a Second-Order Tensor
The tensor or indefinite product, sometimes known as the dyadic or tensor

product of two vectors u, v 2 V 3 is denoted by u v.
This product has no obvious geometrical interpretation, unlike the
scalar or vector products, but it is a linear operator that satisfies the
relation
ðu vÞc ¼ uðv cÞ, ð1:22Þ
where c is any vector. The tensor product has nine components obtained as
follows:
ðu vÞij ¼ ui vj , ð1:23Þ
or, in matrix notation,
2 3 2 3
u1 ½v1 v2 v3 u1 v1 u1 v 2 u1 v3
½u vij ¼ 4 u2 5 ¼ 4 u2 v1 u2 v 2 u2 v3 5, ð1:24Þ
u3 u3 v1 u3 v 2 u3 v3
and can be expressed as follows in terms of the basis vectors, ei ej :
u v ¼ ui vj ei ej : ð1:25Þ
The basis vectors ei ej are not vectors in the sense of elementary vector
analysis but belong to a nine-dimensional vector space.
A second-order tensor S is a linear transformation from V3 into V3 that
assigns to each vector t a vector
s ¼ St: ð1:26Þ
It follows that the tensor product of two vectors is an example of a second-

order tensor.
Equation (1.26) is in symbolic form, that is, it has no preferential re-
lationship to any coordinate system, and it may be interpreted in compo-
nent form by the following matrix equation:
2 3 2 32 3
s1 S11 S12 S13 t1
4 s2 5 ¼ 4 S21 S22 S23 54 t2 5, ð1:27Þ
s3 S31 S32 S33 t3
where the square matrix on the right-hand side of equation (1.27) is the matrix
of the nine components Sij, of S referred to the rectangular Cartesian coordi-
nate axes 0xi . Equation (1.26) or (1.27) can be written in suffix notation as
si ¼ Sij tj , ð1:28Þ
which, unlike equation (1.26), has a preferential relationship to a partic-

ular coordinate system. At this point, it is useful to make a clear distinc-
tion between a column matrix and a vector and between a square matrix
and a second-order tensor. The components of a vector or second-order
tensor may be put in matrix form as in equation (1.27), but a column or

square matrix does not necessarily represent the components of a vector
or tensor.
The tensor S can be expressed in terms of the basis vectors in the form
S ¼ Sij ei ej ; ð1:29Þ
and this shows that a second-order tensor is the weighted sum of tensor
products of the base vectors. The left-hand side of equation (1.29) is in
coordinate free symbolic form; however, the right-hand side involves the
components Sij , which are with respect to a particular coordinate system.
Since a tensor is invariant under coordinate transformation,
Sij ei ej ¼ S9kl e9k e9l ,
and with the use of equation (1.14) the transformation rules
S9kl ¼ aki alj Sij , ð1:30Þ
and
Sij ¼ aki alj S9kl ð1:31Þ
for the componts of S are obtained. It follows from equation (1.29) that the
Cartesian components of S are given by
Sij ¼ ei Sej , ð1:32Þ
that is, Sij is the scalar product of the base vector ei with the vector Sej .
It is convenient at this stage to introduce the transpose of a vector and
second-order tensor. The components of the transpose uT of a vector u
form a row matrix ðu1 , u2 , u3 Þ. The transpose ST of a second-order tensor S
has components STij ¼ Sji so that
S T ¼ Sji ei ej :
It may be shown that, for any two vectors u, v 2 V3 ,
u S T v ¼ v ðSuÞ:
Transposition also applies in general to square matrices as well as the

component matrices of second-order tensors.
A tensor S is symmetric if S ¼ S T , that is, Sij ¼ Sji . Second-order
symmetric tensors in E3 have six independent components. A second-order
tensor T is antisymmetric if T ¼ T T , that is, Tij ¼ Tji . Second-order
antisymmetric tensors in E3 have three independent components, and di-
agonal elements of the component matrix are zero. A second-order tensor
C can be decomposed uniquely into the sum of a symmetric part and an
antisymmetric part as follows:
ðsÞ ð Þ
C ¼ C þC a ,
where
ðsÞ 1
C ¼ C þ CT
2
is the symmetric part and
ðaÞ 1
C ¼ C CT
2
is the antisymmetric part.
EXAMPLE PROBLEM 1.2. Decompose the tensor C into its symmetric

and antisymmetric parts, where the components of C are given by
2 3
1 2 3
Cij ¼ 4 4 2 3 5:
5 2 1
SOLUTION. The components of C T are given by
2 3
h i 1 4 5
CTij ¼ 4 2 2 2 5:
3 3 1
Let the tensors CðsÞ and C ðaÞ denote the symmetric and antisymmetric parts
of C, respectively; then
2 3 2 3
h i h i 2 2 8 1 1 4
ðsÞ 1 14 4
C ¼ CþC T
¼ 2 4 5
5 ¼ 1 2 2:5 5
2 2
8 5 2 4 2:5 1
and
2 3 2 3
h i h i 0 6 2 0 3 1
ð Þ 1 1
Ca ¼ C CT ¼ 4 6 0 1 5 ¼ 4 3 0 0:5 5:
2 2
2 1 0 1 0:5 0
Note that this decomposition is unique.

A second-order tensor S is positive definite if Sij ui vj > 0 for all vectors
u, v 2 V 3 . Other conditions for positive definiteness are introduced later in
this chapter.
As already mentioned, it is important to note that ð3 3 1Þ and ð3 3 3Þ
matrices do not in general represent the components of first- or second-
order tensors, respectively. The nine quantities Sik are the components of
a second-order tensor if, under orthogonal transformation of coordinates
given by equations (1.8) and (1.10), they transform according to equations
(1.30) and (1.31). If the transformation matrix ½a with elements aik and the
square matrix ½S of tensor components Sij are known, the matrix ½S9 with
elements S9kl can be obtained from the matrix form ½S9 ¼ ½a½S ½aT of
equation (1.31) by using Mathematica or a similar package. Matrix nota-
tion is sometimes useful in carrying out algebraic manipulations that in-
volve components of vectors and second-order tensors. In Table 1.2 we list
a number of examples of vector and tensor relations expressed in symbolic,
suffix, and matrix notation.
Table 1.2. Examples of symbolic, suffix, and matrix notation
Symbolic notation Suffix notation Matrix notation
uv ui vi ½uT ½v

uv ui vj ½u½vT
Au Aij uj ½A½u
AT u ui Aij ½AT ½u
uT Av ui Aij vj ½uT ½A½v
AB Aik Bkj ½A½B
ABT Aik Bjk ½A½BT
The inner or scalar product of the vectors w and Su obeys the com-
mutative rule and is denoted by
w Su or Su w: ð1:33Þ
We must also note that
ðSuÞ w 6¼ S ðu wÞ, ð1:34Þ
since the left-hand side is a scalar and the right-hand side is a tensor. Also
we note that
w Su ¼ Sij wi uj : ð1:35Þ
The product, not to be confused with the dyadic product, of two second-
order tensors S and T is a second-order tensor with components
ðST Þij ¼ Sik T kj : ð1:36Þ
This is essentially matrix multiplication. It is easily shown that the product

of two second-order tensors is a second-order tensor. It should be noted
that in general
ST 6¼ TS,
that is, the product of two second-order tensors does not, in general, satisfy
the commutative law.
The components of the unit tensor I are given by the Kronecker delta
so that
I ¼ dij ei ej : ð1:37Þ
The unit tensor is an isotropic tensor, that is, it has the same components
referred any coordinate system since
d9ij ¼ aik ajl dkl ¼ aik ajk ¼ dij ,
where the substitution property of the Kronecker delta has been employed.
It may also be shown that
I ¼ ei ei : ð1:38Þ
We note that in symbolic notation equation (1.6) becomes
u ¼ Iu, ð1:39Þ
and
S ¼ IS ¼ SI: ð1:40Þ
The inner product of two second-order tensors is a scalar defined by

T
tr S T ¼ Sij T ij , ð1:41Þ
where tr denotes the trace and has the same significance as in linear algebra
so that trS ¼ Sii , that is, it is the sum of the diagonal elements of the
component matrix. The magnitude of a second-order tensor is a scalar de-
fined by,
1=2 n o1=2
jS j ¼ Sij Sij ¼ tr ST S : ð1:42Þ
It is shown in section 11 of this chapter that S T S is symmetric and positive

or positive semidefinite. Consequently, tr ST S > 0.
It may be shown by using suffix notation that
ðST ÞT ¼ T T S T : ð1:43Þ
A second-order tensor C has an inverse denoted by C1 , defined by
C1 C ¼ CC 1 ¼ I, ð1:44Þ
unless C is singular, that is, det½C ¼ 0. If det½C 6¼ 0, the tensor is said to

be invertible. A further discussion of the inverse of a second-order tensor is
deferred until section 10 of this chapter.
The inner product of ei ej and ek et is
h
T i
tr ei ej ðek et Þ ¼ dik djt , ð1:45Þ
and a simple application of the tensor transformation rule shows that the
sum of second-order tensors is a second-order tensor. Consequently, the set
of all second-order tensors is a nine-dimensional linear vector space with
orthonormal basis ei ej , where the term vector is taken in its more general
sense as belonging to a linear vector space.
It is often convenient to decompose a second-order tensor T into two
parts, a deviatoric part and an isotropic part, as follows
T ij ¼ T ij þ Td
^ ij , ð1:46Þ
where T^ ¼ Tkk =3 ¼ trT=3. The deviatoric part with components T ij has
^ ij has the same
the property, trT ¼ T ¼ 0, and the isotropic part Td
ii
components in all coordinate systems. The decomposition equation
(1.46) is unique and is important in linear elasticity and plasticity theory.
The concept of tensor multiplication can be extended to more than
two vectors or to higher-order tensors. For example, the tensor product
T S of two second-order tensors T and S is a fourth-order tensor with
1.6 Tensor Algebra 19
components T ij Skl . A detailed discussion of higher-order tensors is given in

section 9 of this chapter.
1.6 Tensor Algebra
The following results are easily verified by using the transformation rule.
1. Tensor addition. The sum or difference of two tensors of the same

order is a tensor, also of the same order. For example, let Aij and Bij
be the components, referred to the same coordinate system, of two
second-order tensors. Then
Cij ¼ Aij 6 Bij ð1:47Þ
are the components of a second-order tensor. It follows from equation

(1.47) that addition of two tensors is commutative.
2. Tensor multiplication. We have already seen an example of this,

namely, the tensor product of two vectors that results in a second-
order tensor. A generalization of this is that the tensor product of
a tensor of order m and a tensor of order n is a tensor of order m þ n.
3. Contraction. We have already met an example of contraction. If we

take the trace of u v, this is the same as contracting ui vj and a scalar
ui vi is obtained so that
u v ¼ trðu vÞ: ð1:48Þ
Equation (1.48) is a result of a special case of contraction. In general,

contraction involves two suffixes equating and summing over the resulting
repeated suffix. A further special case is the contraction of a second-order
tensor S, with components Sij , to obtain the scalar trS ¼ Sii . Consider the
third-order tensor with components Aijk . It follows that Aiik , Aijj , and Aiji
represent the components of vectors. The concept can be extended to
consider higher-order tensors. Contraction of two suffixes of an mth -order
tensor ðm > 2Þ results in a tensor of order m 2.
Test for tensor character. If ui is a set of three quantities possessing
the property that ui vi is an invariant scalar for an arbitrary vector v, then ui
are the components of a vector. The reader should prove this. If Cij is a set
of nine quantities possessing the property that Cij uj is a first-order tensor
for an arbitrary vector u, then Cij are the components of a second-order
tensor. Extension to consider higher-order quantities is straightforward.
1.7 Invariants of Second-Order Tensors
The theory in this section is closely related to the corresponding theory of

square and column matrices in linear algebra.
The direction of the unit vector n is a principal direction of the symmet-
ric tensor A if the vector An with components Aij nj is parallel to n, that is, if
Aij nj ¼ kni : ð1:49Þ
If k ¼ 0, the unit vector can only be obtained if A is singular, that is, if

det½A ¼ 0. Equation (1.49) can be written as

Aij kdij nj ¼ 0, ð1:50Þ
which represents a system of three linear homogeneous equations for the

components n1 , n2 , and n3 of n. The trivial solution, n1 ¼ n2 ¼ n3 ¼ 0, of
(1.50) is not compatible with the condition ni ni ¼ n n ¼ 1 ; consequently,
the determinant of the coefficients in equation (1.50) must vanish if the
system has a solution. That is, the requirement for a solution is
det½A kI ¼ 0, ð1:51Þ
where det½A kI denotes the determinant of the matrix of the coefficients

of the tensor A kI. Equation (1.51) is known as the characteristic equa-
tion and is a cubic in k, which may be written as
k3 J1 k3 þ J2 k J3 ¼ 0, ð1:52Þ
where
1h i 1h i
J 1 ¼ Aii ¼ tr A, J2 ¼ ðAii Þ2 Aik Aki ¼ ðtrAÞ2 trðA2 Þ ; J 3
2 2
¼ det½ A:
The coefficients J 1 , J 2 , and J 3 , of the cubic equation (1.52), are scalar

invariants and are known as the basic invariants of the tensor A. It
follows that the roots ðk1 , k2 , k3 Þ of equation (1.52) are independent of
the coordinate system and are the principal values of the symmetric
tensor (or eigenvalues), and the corresponding directions are called
the principal directions (or eigenvectors). The eigenvalues of the matrix
of components are intrinsic to the tensor A. If A is symmetric, the eigen-
values are real; this is sufficient but not necessary for real eigenvalues,
and if the eigenvalues of A, ðk1 , k2 , k3 Þ, are all positive, then A is a pos-
itive definite tensor. If A is symmetric, and if the eigenvalues are distinct,

then the normalized eigenvectors nð1Þ , nð2Þ , nð3Þ are mutually orthogo-
nal. These two axioms are shown in the following manner. At least one
of the roots of a cubic equation with real coefficients must be real;
consequently, one of the principal values, say, k1 must be real. Let nð1Þ
be the unit vector associated with k1 , which is obtained from the
equations

ðÞ
Aij dij k1 nj 1 ¼ 0, ð1:53Þ
and the condition nð1Þ nð1Þ ¼ 1. If the x91 axis of a new coordinate system is
in this direction, the tensor A has the primed components
A911 ¼ k1 , A912 ¼ A913 ¼ 0, A922 , A913 , A923 :
Consequently, the characteristic equation can be written as
2 3
k1 k 0 0
det4 0 A922 k A923 5 ¼ 0
0 A923 A933 k
or
h i
ðk1 kÞ k2 ðA922 þ A933 Þk þ A922 A933 A9232 ¼ 0: ð1:54Þ
This is known as the characteristic equation. The remaining principal val-

ues are the roots of
h 2
i
k2 ðA922 þ A933 Þk þ A922 A933 A923 ¼ 0, ð1:55Þ
and the discriminant of this quadratic equation is
2
ð A922 þ A933 Þ2 4 A922 A933 A9232 ¼ ðA922 A933 Þ2 þ4 A923 > 0: ð1:56Þ
Consequently, the roots k2 and k3 are also real.

We will now show that if A is symmetric, and if k1 , k2 , and k3 are
distinct, then the corresponding eigenvectors are mutually orthogonal. It
follows from equation (1.53) that
ð Þ ð Þ
Aij nj 1 ¼ k1 ni 1 . ð1:57Þ
Similarly
ð Þ ð Þ
Aij nj 2 ¼ k2 ni 2 : ð1:58Þ
ð2Þ ð1Þ
Scalar multiplication of equation (1.57) by ni and equation (1.58) by nj
and subtracting gives
ðk1 k2 Þnði 1Þ nði 2Þ ¼ 0,
which shows that nð1Þ is orthogonal to nð2Þ since k1 6¼ k2 .

If the tensor is referred to axes 0x 0i that coincide with the principal
directions nð1Þ the transformation matrix is given by
ðiÞ ðiÞ
a9ij ¼ nj ¼ n ej ,
and x 0i ¼ a9ij xj . The components A 0i referred to 0x0i are then
ðiÞ ð jÞ
A0i ¼ nk nt Akt , ð1:59Þ
and it can be deduced that
2 3
k1 0 0
A0ij ¼ 4 0 k2 0 5.
0 0 k3
This is known as the canonical form of the matrix of components. For many
problems and proofs, it is desirable to refer components of second-order
symmetric tensors to the principal axes. Equation (1.59) can be expressed
in symbolic form
diagðAÞ ¼ NAN T , ð1:60Þ
where diagðAÞ is the canonical form of the matrix of components of A and

N is a proper orthogonal second-order tensor with components
ðÞ
Nij ¼ nj i :
The rows of the matrix N are the eigenvectors of A. Usually, in trea-

tises on linear algebra, the diagonalization matrix is the transpose of N, that
is, the columns are the eignvectors of A. The motivation for the above
definition of N is that Mathematica gives a matrix whose rows are the
eigenvectors.
If exactly two eigenvalues are equal, that is, k1 6¼ k2 ¼ k3 , then the
discriminant equation (1.56) is zero, that is, A922 ¼ A933 , and A923 ¼ 0.
Equation (1.54) then has roots k2 ¼ k3 ¼ A922 ¼ A933 , and the matrix of
components referred to the 0x9i axes has a diagonal form regardless of the
choice of the orthogonal pair of axes, 0x92 and 0x93 if the 0x91 axis is in the nð1Þ
direction. It follows that any axis normal to nð1Þ is a principal axis (eigen-
vector). As an exercise consider the case k1 ¼ k2 ¼ k3 .
If the ki are distinct, the symmetric tensor A may be expressed, with

the summation convention suspended, in the form,
3
ðÞ ðÞ
A ¼ + ki n i n i ,
i¼1
which is known as the spectral form. If two eigenvalues are equal, that is, if
k1 ¼ k2 ¼ k3 , then

ð Þ ð Þ ð Þ ð Þ
A ¼ k1 n 1 n 1 þ k2 I n 1 n 1 :
Further, if all three eigenvalues are equal, that is, if k1 ¼ k2 ¼ k3 ¼ k,

then
A ¼ kI,
and A is an isotropic tensor.

The eigenvalues and eigenvectors of a second-order tensor can be
obtained using Mathematica and routine but tedious computations are
avoided.
1.8 Cayley-Hamilton Theorem
The Cayley-Hamilton theorem has important applications in continuum

mechanics. According to this theorem, a square matrix satisfies, in a matrix
sense, its own characteristic equation. We will consider this theorem for
a second-order symmetric tensor C, although it is valid for any second-
order tensor or square matrix. As shown in the previous section, the tensor
C has three real eigenvalues k1 , k2 , and k3 , which are the roots of the
characteristic equation (1.54). Consequently, we have the matrix equation
½k3 J1 ½k2 þJ2 ½k J ½I ¼ ½0,
where ½I is the unit ð3 3 3Þ matrix and ½k is the diagonal matrix of the
components of C referred to the principal axes. Since a tensor equation
referred to a particular coordinate system is true referred to any other

coordinate system, it may be deduced from equation (1.52) that
C 3 J1 C2 þ J2 C J3 I ¼ 0, ð1:61Þ
which is the Cayley-Hamilton theorem for a second-order symmetric

tensor. The Cayley-Hamilton theorem is also valid for any ð3 3 3Þ matrix.
1.9 Higher-Order Tensors
The concept of tensor mutiplication can be extended to more than two

vectors. For example, ðu v wÞijk ¼ ui vj wk , which transforms under
coordinate transformation as
u9r v9s w9t ¼ ari asj atk ui vj wk , ð1:62Þ
ui vj wk ¼ ari asj atk u9rv9s w9t : ð1:63Þ
A third-order tensor with components Dijk may be expressible in the form
Dijk ei ej ek ,
which is invariant under coordinate transformation. The components trans-

form according to
D9ijk ¼ air ajs akt Drst , Dijk ¼ air ajs akt D9rst :
The extension to higher-order tensors is obvious. The tensor product S T

of the two second-order tensors S and T is a fourth-order tensor, with
components
ðS T Þijrs ¼ Sij Trs : ð1:64Þ
We meet fourth-order tensors in the study of classical elasticity. It is some-

times convenient to consider a fourth-order tensor to be a linear
transformation that assigns to each second-order tensor another second-

order tensor. If Bijkt are the components of a fourth-order tensor, then
Sij ¼ Bijkt Tkt ,
where T is any second-order tensor and S is the result of B operating on T

as indicated.
Symbolic notation is not particularly useful for tensors of higher-order
than two and henceforth we use only suffix notation for these.
The alternating tensor, which is actually a pseudo tensor, is a useful
third-order isotropic tensor if referred to right- (or left-) handed co-
ordinate systems exclusively. The alternating symbol eijk is defined as:
eijk ¼ 1 if (i, j, k) is an even permutation of (1, 2, 3), that is, if
e123 ¼ e231 ¼ e312 ¼ 1;
eijk ¼ 1 if (i, j, k) is an odd permutation of (1, 2, 3), that is, if
e321 ¼ e132 ¼ e213 ¼ 1;
eijk ¼ 0 if any two of i, j, k are equal (e.g., e112 ¼ 0, e333 ¼ 0 ).

The vector product of two base vectors is given by
ei 3 ej ¼ eijk ek ,
which follows from the definition of vector product in elementary vector

algebra and the orthogonality of the unit base vectors. It then follows that
the vector product of two vectors u and v can be expressed in the form
u 3 v ¼ ui vj ei 3 ej ¼ eijk ui vj ek : ð1:65Þ
Since u 3 v is invariant under right-handed to right-handed (or left-handed

to left-handed) coordinate transformations, it may be deduced from equa-
tion (1.65) that eijk may be regarded as the components of a third-order
isotropic tensor for these transformations.
A useful identity is
eijk eipq ¼ djp dkq djq dkp : ð1:66Þ
In order to prove equation (1.66) we note that
ðu 3 vÞ ðs 3 t Þ ¼ eijk eipq uj vk sp tq ,
and
u 3 v ¼ ðu1 e1 þ u2 e2 þ u3 e3 Þ 3 ðv1 e1 þ v2 e2 þ v3 e3 Þ
¼ e1 ðu2 v3 v2 u3 Þ þ e2 ðu3 v1 v3 u1 Þ þ e3 ðu1 v2 v1 u2 Þ:
Similarly
s 3 t ¼ e1 ðs2 t3 t2 s3 Þ þ e2 ðs3 t1 t3 s1 Þ þ e3 ðs1 t2 t1 s2 Þ:
It then follows that
ðu 3 vÞ ðs 3 t Þ ¼ ðu2 v3 u3 v2 Þðs2 t3 s3 t2 Þ þ ðu3 v1 u1 v3 Þðs3 t1 s1 t3 Þ

þ ðu1 v2 u2 v1 Þðs1 t2 s2 t1 Þ

¼ djp dkq djq dkp uj vk sp tq :
Comparison of the above two expressions for ðu 3 vÞ ðs 3 t Þ gives

equation (1.66). The identity equation (1.66) can be used to obtain the
relation
a 3 ðb 3 cÞ ¼ bða cÞ cða bÞ:
This is obtained from equations (1.65) and (1.66) as follows,
a 3 ðb 3 cÞ ¼ eijk aj ðelmk bl cm Þei

¼ dil djm dim dlj aj bl cm ei
¼ ðbi am cm ci al bl Þei
¼ bða cÞ cða bÞ:
It should be noted that, in general, a 3 ðb 3 cÞ 6¼ ða 3 bÞ 3 c.

Identity equation (1.66) involves fourth-order tensors, and it may be
deduced that this is an isotropic tensor. This tensor is a particular case of
the most general fourth-order isotropic tensor that has components
adij dkl þ bdik djl þ cdil djk , ð1:67Þ
where a, b, and c are scalars.
1.10 Determinants
The Laplace expansion of a determinant can be expressed in the form
2 3
A1 A2 A3
det4 B1 B2 B3 5 ¼ eijk Ai Bj Ck , ð1:68Þ
C1 C2 C3
where eijk are the components of the alternating tensor as given in

section 9.
It follows from the Laplace expansion that det½I ¼ 1 and from equations
(1.65) and (1.68) that the vector product of two vectors u and v can be
expressed as a determinant,
2 3
e1 e2 e3
4
u 3 v ¼ det u1 u2 u3 5,
v1 v2 v3
The triple scalar product u v 3 w is given by,
2 3
u1 u2 u3
4
u v 3 w ¼ det v1 v2 v3 5 ¼ eijk ui vj wk : ð1:69Þ
w1 w2 w3
A physical significance of the triple scalar product is that equation (1.69)

represents the enclosed volume of the parallelpiped with contiguous sides
of the three noncoplanar vectors, u, v, and w. We will use this interpretation
in later chapters.
It follows from equation (1.69) that
2 7
C11 C12 C13 7
7
6 7
det 4 C21 C22 C23 5 ¼ eijk C1i C2j C3k expanded by rows
ð1:70Þ
C31 C32 C33
¼ eijk Ci1 Cj2 Ck3 expanded by columns:
Further important relations are,
eijk Cir Cjs Ckt ¼ erst det½C, ð1:71Þ
eijk Cri Csj Ctk ¼ erst det½C, ð1:72Þ
1
eijk erst Cir Cjs Ckt ¼ erst det½C: ð1:73Þ
6
In equations (1.70) to (1.73) we assume that Cij are the components of the
second-order tensor C, although the results are valid for any ð3 3 3Þ ma-
trix. The determinant of a second-order tensor is a scalar and is known as
the third invariant of the tensor, as already mentioned in section 7. If
det½C ¼ 0, the tensor is said to be singular. The coefficient of Cij in the
expansion of the determinant is called the cofactor of Cij and is denoted by
Cijc , so that expansion by rows gives
c c c
det½C ¼ C1j C1j ¼ C2j C2j ¼ C3j C3j ,
which in turn gives
Cij Cckj ¼ dik det½C: ð1:74Þ
c
It may be deduced from equation (1.74) that Ckj are the components of
a second-order tensor; consequently, we can express equation (1.74) in
symbolic notation as,
CCcT ¼ Idet½C, ð1:75Þ

where the matrix of components of CcT is the transpose of the matrix of the
cofactors of C. It then follows from equation (1.74) that
trCC cT ¼ Cij Cijc ¼ 3det½C :
EXAMPLE PROBLEM 1.3 Show that
c 1
Cmi ¼ eijk eimst Csj Ctk :
2
SOLUTION. Multiply each side of equation (1.71) by emst to obtain
emst eijk Cri Csj Ctk ¼ emst erst det½C:
It follows from equation (1.66) that emst erst ¼ 2dmr ; consequently,
emst eijk Cri Csj Ctk ¼ 2drm det½C:
c
Compare with Cri Cmi c
¼ drm det½C to obtain Cmi ¼ 12 eijk emst Csj Ctk .
The inverse of C is a second-order tensor denoted by C 1 and
CC1 ¼ I or Cij Cjk

1
¼ dik : ð1:76Þ
If the tensor C is singular, the inverse does not exist. The term inverse
is meaningful for scalars and second-order tensors but not for tensors of
any other order. The relation ðCBÞðCBÞ1 ¼ I follows from equation
(1.76), and premultipying each side first by C1 and then by B1 gives
the the result
ðCBÞ1 ¼ B1 C1 :

c
Cji
C1
ij ¼ : ð1:77Þ
det½C
Since the value of a determinant is unchanged if the rows and columns

are interchanged.
h i
T
det½C ¼ det C : ð1:78Þ
EXAMPLE PROBLEM 1.4. Prove that
Au Av 3 Aw ¼ det½Aðu v 3 wÞ:
SOLUTION.
Au Av 3 Aw ¼ eijk Air ur Ajs vs Akt wt

¼ erst det½ Aur vs wt ¼ ðu v 3 wÞdet½A:
It may be deduced from equation (1.70) that
det½aC ¼ a3 det½C,
and from the result of example problem (1.4) that
det½CB ¼ det½C det½B, ð1:79Þ
since replacing A with BC gives
CBu CBv 3 CBw ¼ det½CBðu v 3 wÞ

¼ det½CðBu Bv 3 BwÞ
¼ det½Cdet½Bðu v 3 wÞ,
where C and B are nonsingular second-order tensors. It then follows from

equation (1.79) by replacing B by C1 that
h h ii1
det½C ¼ det C 1 , ð1:80Þ
1.11 Polar Decomposition Theorem
We have already seen that an arbitrary second-order tensor can be decom-

posed, uniquely, into the sum of symmetric and antisymmetric parts. A
further decomposition of an arbitrary nonsingular second-order tensor is
given by the polar decomposition theorem. Before introducing this theo-
rem we define the terms orthogonal second-order tensor and positive def-
inite second-order tensor.
An orthogonal second-order tensor Q satisfies the condition,
QQT ¼ QT Q ¼ I, ð1:81Þ
or, in suffix notation,
Qji Qki ¼ Qij Qik ¼ djk :
QT ¼ Q1 ,
det½Q ¼ 61,
and Q is described as a proper (improper) orthogonal tensor and when the

plus (minus) sign is valid. We are concerned with proper orthogonal tensors
and henceforth use the term orthogonal tensor to denote a second-order
tensor, which satisfies condition (1.81) and det½Q ¼ 1. A proper orthog-
onal tensor is similar, in a sense, to the transformation matrix for a proper
orthogonal transformation. This is evident by comparing equations (1.19)
and (1.20) with equation (1.81). The scalar product of two arbitrary vectors
u, v 2 V3 is unchanged if the vectors are operated on by an orthogonal

tensor. That is,
u v ¼ Qu Qv, ð1:82Þ
ui vi ¼ Qij Qik uj vk :
This is sometimes given as the definition of an orthogonal tensor. Equation

(1.82) follows from equation (1.77).
When an orthogonal tensor operates on a vector u 2 V3 , a vector of
equal magnitude results, that is,
juj ¼ jQuj: ð1:83Þ
Equation (1.83) is obtained from equation (1.82) by putting u ¼ v. The

unit tensor I is a special case of an orthogonal tensor.
A second-order tensor S is said to positive semidefinite if, for an ar-
bitrary vector, u,
u Su ¼ Sij ui uj > 0, ð1:84Þ
and positive definite if the equality sign holds if and only if u is the null
vector. The antisymmetric part of S does not contribute to u Su since, if W
is an antisymmetric second-order tensor,
u Wu ¼ Wij ui uj ¼ 0:
When we consider second-order positive definite tensors we usually

assume the tensor is symmetric, and henceforth, the term positive
definite tensor is used for a positive definite symmetric second-order
tensor.
The second-order tensor S ¼ AT A, where A is an arbitrary second-

T
order tensor, is symmetric since AT A ¼ AT A. It then follows that S is
either positive definite or semidefinite since

T
u Su ¼ u A A u ¼ ðAuÞ ðAuÞ > 0: ð1:85Þ
Inequality (1.85) is evident if it is expressed in suffix notation since

u AT A u ¼ uj Aij Aik uk ¼ Aij uj ðAik uk Þ:
If the tensor A is singular, that is, det½A ¼ 0, it may be deduced from

equation (1.85) that S is positive semidefinite since Au ¼ 0 has a nontrivial
solution for u. If S can be expressed in the form S ¼ AAT or S ¼ AT A, it
follows from equation (1.85) that a necessary and sufficient condition for S to
be positive definite is det½S > 0. If S is an arbitrary second-order symmetric
tensor, the condition det½S > 0 is necessary but not sufficient condition for
positive definiteness. The necessary and sufficient condition for a symmetric
second-order tensor to be positive definite is that the eigenvalues (or prin-
cipal values) are all positive. This can be shown by expressing equation
(1.84) in terms of the components with respect to the principal axes of S,
that is, axes, that are parallel to the eigenvectors (principal directions) of S,
u Su ¼ k1 u^21 þ k2 u^22 þ k3 u^23
where k1 , k2 , k3 are the eigenvalues of S and u^1 , u^2 , and u^3 are the compo-
nents of u referred to the principal axes of S. When all the eigenvalues of
S are negative, it follows from equation (1.84) that u Su < 0 with the
equality sign holding only when u is the null vector. The tensor S is then
said to be negative definite.
The square root S1=2 of a positive definite tensor S is defined as the
symmetric second-order tensor whose spectral form is
3
ð Þ ð Þ
S 1=2 ¼ + ki1=2 n 1 n 1 ð1:86Þ
i¼1
if the eigenvalues are distinct,

1=2 1=2 ð1Þ ð1Þ 1=2 ð1 Þ ð1 Þ
S ¼ k1 n n þ k2 In n
if k2 ¼ k3 and
S1=2 ¼ k1=2 I
if the eigenvalues are equal, that is, k1 ¼ k2 ¼ k3 ¼ k.

The summation convention is suspended and k1 and nðiÞ , i 2 f1, 2, 3g,
are the eigenvalues and eigenvectors, respectively, of S. We adopt the
convention that positive roots of the k1 are taken in equation (1.86) so that
S 1=2 is positive definite and unique.
If the matrix of components of the positive definite tensor S is known,
with respect to an arbitrary system of Cartesian axes, the matrix of compo-
nents of S1=2 , with respect to the same system of axes, can be obtained as

follows. Firstthe eigenvalues kð1Þ , kð2Þ , kð3Þ of S are determined, then the
1=2 1=2 1=2
square roots k1 , k2 , k3 . It then follows from equation (1.60) that the
matrix of components of S 1=2 is given by
2 1=2
3
k1 0 0
T6 7
½N 4 0 k21=2 0 5½N ,
0 0 k1=2
3
where the first, second,

and third rows of N are the eigenvectors of S
1=2 1=2 1=2
corresponding to k1 , k2 , k3 , respectively. It should be noted that
the eigenvectors of S1=2 are the same as those of S. The square root of
a matrix of components of a second-order positive definite tensor can be
determined directly by using Mathematica.
We are now ready to introduce the polar decomposition theorem,
which states that an arbitrary nonsingular second-order tensor A with
det½A > 0
can be expressed uniquely in the forms,
A ¼ QU, ð1:87Þ
A ¼ VQ, ð1:88Þ
where Q is an orthogonal tensor and U and V are positive definite second-

order tensors. If det½A < 0, then U and V are negative definite tensors;
however, we do not consider this case since it is not important in the
applications that follow in later chapters. Equations (1.87) and (1.88) are
known as the right and left polar decompositions, respectively. In order to
prove the theorem, we note that AT A and AAT are positive definite tensors,

1=2
1=2
so that AT A and AAT are also positive definite. The tensors

T 1=2

T 1=2
Q ¼ A A A and P ¼ AA A are orthogonal since
1=2 1=2 T 1
T T
A A A A A A ¼ A AT A AT
1
¼ AA1 AT AT ¼ I,

1=2
and det½P ¼ det½Q ¼ 1, so that A ¼ QU ¼ VP, where U ¼ AT A

1=2
and V ¼ AAT .
The square root of a positive definite tensor is unique, when the pos-
itive roots are taken in equation (1.83), so that the decomposition A ¼ QU
is unique. We now have to show that Q ¼ P to complete the proof of the
theorem. This follows since

A ¼ VP ¼ P P T VP ,
and it can be shown that P T VP ¼ U. Consequently, Q ¼ P and

U ¼ QT VQ.
The positive definite tensors U and V have the same eigenvalues, and
the eigenvectors l ðiÞ of U are related to the eigenvectors mðiÞ of V by
ðiÞ T ðiÞ
l ¼ Q m : ð1:89Þ
EXAMPLE PROBLEM 1.5. Determine the polar decomposition equation

(1.87) of the matrix of components, with respect to a particular co-
ordinate system,
2 3
1:1 0:5 0:3
½A ¼ 4 0:4 1:2 0:5 5,
0:2 0:3 0:9
of a nonsingular second-order tensor A.

SOLUTION. The matrix of components of
C ¼ AT A
is given by
2 32 3
1:1 0:4 0:2 1:1 0:5 0:3
6 76 7
½C ¼ 4 0:5 1:2 0:3 54 0:4 1:2 0:5 5
0:3 0:5 0:9 0:2 0:3 0:9
2 3
1:41 0:97 0:35
6 7
¼ 4 0:97 1:78 1:02 5
0:35 1:02 1:15
and the matrix of components of
C1=2 ¼ U
by
2 3
1:11294 0:406455 0:0785375
6 7
½U ¼ 4 0:406455 1:18478 0:459452 5,
0:0785375 0:459452 0:965783
which can be obtained by using Mathematica or the procedure already

described.
The matrix of components of the inverse of U
2 3
h i 1:036 0:3956 0:104
1 6 0:532 7
U ¼ 4 0:3956 1:186 5
0:104 0:532 1:280
can be obtained using Mathematica or a standard procedure given in most

texts on linear algebra.
The matrix of components of the tensor
Q ¼ AU 1
is
2 32 3
1:1 0:5 0:3 1:036 0:3956 0:104
6 76 7
½Q ¼ 4 0:4 1:2 0:5 54 0:3956 1:186 0:532 5
0:2 0:3 0:9 0:104 0:532 1:280
2 3
0:9726 0:018 0:2324
6 7
¼ 4 0:0085 1:0 0:04314 5:
0:2322 0:4393 0:9717
It is easily verified that ½A ¼ ½Q½U .
1.12 Tensor Fields
If the value of a tensor is assigned, at every point x of a domain D of a

Euclidean point space E3 , we have a tensor field. The tensor, which may
be of any order, may then be expressed as a function of x 2 D, and
perhaps time. We are mainly concerned with scalar (zeroth-order ten-
sor), vector (first-order tensor), and second-order tensor fields’’. These
fields may have discontinuities, but for the present we will assume that
the tensor fields are continuous with continuous first derivatives. The
treatment of derivatives of tensor components, with respect to spatial
coordinates, is greatly simplified when rectangular Cartesian coordinates
are adopted. In this section both suffix and symbolic notation are used to
develop the theory.
First we consider a scalar field /ðxÞ. Then @/=@xi , sometimes written
as /;i , are the Cartesian components of a vector since,
@/ @/ @/
0 ¼ 0 ,
@xj @xi @x
j
which, with the use of equation (1.11) becomes,
@/ @/
0 ¼ aji : ð1:90Þ
@xj @xi
Equation (1.90) is the transformation rule for the vector ð@/=@xi Þei 2 V 3 ,
which is known as the gradient of / and is often denoted by =/ or grad/. At
any point =/ is normal to the surface /ðx1 , x2 , x3 Þ ¼ constant passing
through the point and is in the direction of / increasing. The scalar
=/ n, where n is a unit vector, is the rate of change of / with respect to
the distance in the direction n.
We use the symbol = preceding an operand to denote either the dif-
ferential operator
@ðÞ
= [ ei ð1:91Þ
@xi
or
@ðÞ
=[ ei , ð1:92Þ
@xi
and we will carefully note which definition is intended. It is immaterial

which definition of = is used for the gradient of a scalar or the divergence
of a vector. When = operates scalarly on a vector, it gives the divergence of
the vector,
@vi
=v ¼ :
@xi
When = operates scalarly, with either definition, on itself, the result is

a scalar differential operator
2
@
ð= =Þ [ =2 [ :
@xi @xi
According to the definition of curl v, where v is a vector field, in elementary

vector analysis,
@vk
curl v ¼ eijk ei ; ð1:93Þ
@xj
so that
curl v [ = 3 v,
with = given by definition (1.91). However, with definition (1.92),

curl v [ = 3 v.
Henceforth, the form
=3v
is not used and curl v is understood to be given by equation (1.93).

The nine quantities @vi =@xj where vi are the components of a vector
are the components of a second-order tensor field since it is easily shown
that they transform according to the tensor transformation rule,
0
@vt @vn
0 ¼ atn amj :
@xm @xj
The second-order tensor = v with components @vi =@xj is known as the

gradient of vector v, and the definition (1.92) is implied, so that
@vk
=v ¼ ek ei : ð1:94Þ
@xi
The matrix of components of = v is then given by
2 3
@v1 @v1 @v1
6 @x1 @x2 @x3 7
6 7
@vi 6 7
6 @v @v2 @v2 7
¼ 6 2 7: ð1:95Þ
@xj 6 @x1 @x2 @x3 7
6 7
4 @v3 @v3 @v3 5
@x1 @x2 @x3
It follows from equation (1.95) that = v ¼ trð= vÞ. The notation vk;i is
often used for @vk =@xi , and this is compatible with equations (1.94) and
(1.95). When definition (1.92) is used we obtain the transpose of the right-
hand side of equation (1.95). With definition (1.92),
dv ¼ ð= vÞdx;
which is analogous to the result
df
df ðxÞ ¼ dx
dx
in elementary calculus, whereas with definition (1.91), we obtain
dv ¼ dx ð= vÞ:
In spite of the difficulty with the curl it is usually more convenient to adopt
definition (1.92). The gradient of the vector field vðxÞ may also be defined as
the linear transformation that assigns to a vector a the vector given by the
following rule:
vðx þ saÞ vðxÞ

ð= vÞa ¼ lim ;
s!0 s
which is compatible with definition (1.92) and is analogous to a definition of

the gradient of a scalar field /ðxÞ, namely, that the gradient of /ðxÞ assigns
to a vector a the scalar given by the following rule
/ðx þ saÞ /ðxÞ

ð=/Þ a ¼ lim :
s!0 s
The divergence of a second-order tensor T is a vector given by
@Tij
divT ¼ ej ð1:96Þ
@xi
if definition (1.91) is used and
@T ij
divT ¼ ei ð1:97Þ
@xj
if definition (1.92) is used.

With definition (1.92),
@T ij
=T ¼ ei ej ek ,
@xk
which is a third-order tensor. When @T ij =@xk is contracted with respect

to the suffixes i and k, the components of divT defined by equation (1.91)
are obtained, and if contracted with respect to j and k, the components
defined by equation (1.92) are obtained. The gradient of a tensor of
higher order than two with components @Sij :::=@xm can be contracted
with respect to m and any one of the suffixes i,j,k ... l or any two of the
suffixes i,j,k ... l.
There is an important relationship between curl v and = v. The
symmetric part of = v,
1 T
D ¼ = v þ ð= vÞ Þ
2
has components
1

Dij ¼ vi;j þ vj;i ,
2
where we have adopted the notation vi;j for @vi =@xj , and the antisymmetric
part
1
V ¼ = v ð= vÞT
2
has components
1

Vij ¼ vi;j vj;i :
2
Now consider the vector w ¼ 12 curl v, which in component form is,
1

w ¼ e1 v3;2 v2;3 þ e2 v1;3 v3;1 þ e3 v2;1 v1;2 : ð1:98Þ
2
We note that the independent components of Vij are related to the com-
ponents of the vector w as follows,
Xij ¼ eijk wk , ð1:99Þ
1
wk ¼ eijk Xij : ð1:100Þ
2
It follows from equation (1.99) that w 3 a ¼ Va where V ¼ Vij ei 3 ej .

We can obtain a vector b from any second-order antisymmetric tensor B as
follows, bi ¼ eijk Bjk . Actually, a vector c can be obtained from any second-
order tensor C as follows, ci ¼ eijk Cjk , ci ¼ eijk Cjk ; however, c is the null
vector if C is symmetric, and if C is not symmetric, the symmetric part does
not influence c since
ðsÞ
eijk Cjk ¼ 0, ð1:101Þ
where C ðsÞ is the symmetric part of C.

The relation
trðABÞ ¼ Aji Bij ¼ 0;
where A is symmetric and B is antisymmetric, is similar to equation

(1.101).
1.13 Integral Theorems
The integral theorem, which is most frequently encountered in the study of

continuum mechanics, is the divergence theorem. In what follows we present
the divergence theorem and Stokes’ theorem without proof, since proofs of the
basic theorem are presented in numerous texts on elementary vector analysis.
We consider a bounded regular region of E3 , that is, the union of the set of
points on a closed piecewise smooth boundary or boundaries and the interior
points. It may be simpler to think of a volume bounded by an outer surface, or
by an outer surface and one of more inner surfaces. If / is a continuous scalar
field, then
Z Z
=/dv ¼ /nda,
V @V
Z Z
@/
dv ¼ /ni da, ð1:102Þ
@xi
V @V
where v denotes the region, @v is the boundary with area a, and n is the
outward unit normal to the boundary. The divergence theorem
Z Z
divudv ¼ u nda,
v @v
Z Z
@ui
dv ¼ ui ni da ð1:103Þ
@xi
v @v
is obtained from equation (1.97) when / is replaced by the components

ui of a vector u, and the summation convention applied. Equation
(1.03) is what is usually known as the divergence theorem, although
equation (1.02) is more fundamental. The theorem can also be applied
1.13 Integral Theorems 45
to second- and higher-order tensors. For example, if T is a second-order

tensor,
Z Z
divTdv ¼ T ij nj da
@V
V
where divT is given by equation (1.97), so that in suffix notation we have,

Z Z
@Tij
dv ¼ Tij nj da:
@xj
V @V
EXAMPLE PROBLEM 1.6. Use the divergence theorem to prove that
Z Z
n 3 uda ¼ curludv,
@R R
where u 2 C1 is a vector field in region R and n is the unit normal to the

boundary @R of R.
SOLUTION.
Z Z Z Z
@uk
n 3 uda ¼ eijk nj uk da ei ¼ eijk dv ei ¼ curludv:
@R @R R @xj R
The second theorem that frequently occurs is Stokes theorem. Let dA be

a part of a surface A bounded by a closed curve C, and u be a continuous
vector field in the domain containing A, then Stokes theorem is
I Z
u dx ¼ curlu ndA
A
H
where denotes the integral around the curve C, dx is a vector element
of C, and n is the unit normal to A taken in the sense of the right-hand
rule.
EXERCISES
1.1. Derive the vector relations
(a) curl = a ¼ 0:
(b) divcurl u ¼ 0:
(c) divðauÞ ¼ adivu þ u grada:
(d) curlðauÞ ¼ acurlu þ =a 3 u:
(e) curlcurlu ¼ =divu =2 u:
(f) a=2 b ¼ divða=bÞ =a =b:
(g) u 3 curlu ¼ 12 =ðu uÞ u =u:

1.2. Consider the tensor I n n ¼ dij ni nj ei ej , where n is a unit
vector. Show that ðI n nÞu is the projection of u onto a plane
normal to n and that
ðI n nÞ2 ¼ ðI n nÞ:
1.3. Show that if S is a second-order positive definite tensor
h i
det S 1=2 ¼ fdet½S g1=2 :
1.4. Let Tij be the components of the deviatoric part of the second-order
symmetric tensor T : Prove that
1
det½T ¼ trðT 3Þ:
3
1.5. Consider the nine quantities Aij referred to a given coordinate system.
If Aij Bij transforms as a scalar under change of coordinate system for
an arbritrary second-order symmetric tensor B, prove that Aij þ Aij
are components of a second-order tensor.
1.6. Show that for a proper orthogonal tensor Q, there exists a vector u
such that Qu ¼ u.
1.7. Show that det½u v ¼ 0.

Exercises 1.1–1.12 47
1.8. Consider the positive definite matrix
2 3
3 2 1:5
½B ¼ 4 2 4 1 5:
1:5 1 2
Obtain the square root,
2 3
h i 1:57443 0:542316 0:476527
B1=2 ¼ 4 0:542316 1:91127 0:230106 5
0:476527 0:230106 1:31148
and check this result using the Mathematica command MatrixPo-

wer[B,1/2].
1.9. Show that for any proper orthogonal tensor Q,
det½Q I ¼ 0:
It is perhaps easier to show this using symbolic notation, noting that

QQT ¼ I and the result follows.
1.10. Let V be a region of E3 with boundary @V and outward unit normal n.

Let s be a Cauchy stress field and u a vector field both of class C1 .
Prove that
Z Z
u snda ¼ fu divs þ ð= uÞsgdv:
@V V
1.11. Show that
Z Z
ð= uÞdv ¼ ðn uÞda,
v @V
where n is the unit normal to the suface @v.

Assume definition (1.91) for =:
1.12. If b ¼ /=w, where / and w are scalar functions, show that
Z Z
/=2 w þ =/ =w dV ¼ /=w ndA;
V @V
where n is the unit normal to the suface A.
SUGGESTED READING
1. Jeffreys, H. (1931). Cartesian Tensors. Cambridge University Press. This

was the first text on Cartesian tensor analysis. Does not introduce base
vectors or symbolic notation.
2. Temple, G. (1960). Cartesian Tensors. Methuen. Presents a different
approach to the definition of a tensor. Does not introduce base vectors
or symbolic notation.
3. Prager, W. (1961). Introduction to Mechanics of Continua. Ginn. Introdu-
ces base vectors and limited use of symbolic notation. Gives proofs of
divergence and Stokes theorems.
4. Chadwick, P. (1976). Continuum Mechanics. George Allen and Unwin.
Uses Symbolic notation exclusively. Many worked problems.
5. Gurtin, M.E. (1981). An Introduction to Continuum Mechanics. Academic
Press. Uses symbolic notation almost exclusively.
6. Pearson, C.E. (1959). Theoretical Elasticity. Harvard University Press.
Has a useful introduction to Cartesian Tensor analysis. Does not introduce
base vectors or symbolic notation.
7. Stratton, J.A. (1941). Electro-Magnetic Theory. McGraw-Hill. Has a good
introduction to Cartesian Tensor analysis. Introduces base vectors. Has an
excellent treatment of orthogonal curvilinear coordinate systems. Does not
use summation convention.
8. Leigh, D.C. (1968). Nonlinear Continuum Mechanics. McGraw-Hill.
Introduces suffix notation, for both Cartesian and curvilinear coordinates,
along with symbolic notation.
9. Lichnerowicz, A. (1962). Tensor Calculus. Methuen. Has an excellent
treatment of linear vector spaces and of general tensor analysis.
10. Strang, G. (1993). Introduction to Linear Algebra. Wellesley Cambridge
Press. An excellent text which provides the background in linear algebra
required for the further study of tensor analysis.
Suggested Reading 49
11. Kellogg, O.D. (1953). Foundations of Potential Theory. Dover. This is

a classic text, first published in 1929. It contains a detailed discussion of
the integral theorems.
12. Flügge, W. (1972). Tensor Analysis and Continuum Mechanics. Springer-
Verlag. Contains an excellent treatment of general tensor analysis and its
applications to continuum mechanics.
2 Kinematics and Continuity Equation
2.1 Description of Motion
In this chapter we are concerned with the motion of continuous bodies

without reference to the forces producing the motion. A continuous body
is a hypothetical concept and is a mathematical model for which molecular
structure is disregarded and the distribution of matter is assumed to be
continuous. Also it may be regarded as an infinite set of particles occupying
a region of Euclidean point space E3 at a particular time t. The term particle
is used to describe an infinitesimal part of the body, rather than a mass point
as in Newtonian mechanics. A particle can be given a label, for example, X,
and there is a one-one correspondence between the particles and triples of
real numbers that are the coordinates at time t with respect to a rectangular
Cartesian coordinate system.
There are four common descriptions of the motion of a continuous
body:
1. Material description. The independent variables are the particle X and

the time t.
2. Referential description. The independent variables are the position
vector X of a particle, in some reference configuration, and the time
t. The reference configuration could be a configuration that the body
never occupies but it is convenient to take it as the actual unstressed
undeformed configuration at time t ¼ 0. The term natural reference
configuration is used to describe the unstressed undeformed configu-
ration at a uniform reference temperature. A difficulty arises when
a body has an unloaded residually stressed configuration due perhaps
50
2.1 Description of Motion 51
to lack of fit of component parts. This is not discussed in this chapter.

The reference configuration is often called the material, or Lagrangian
configuration and the components XK , K 2 f1, 2, 3g, with respect to
a rectangular Cartesian coordinate system, are the material coordi-
nates of X. In many applications, for example, in the theory of elas-
ticity, the reference configuration is taken to be the natural reference
configuration.
3. Spatial description. The independent variables are the current position
vector x and the present time t. This is the description used in fluid
mechanics. The components xi of x are known as the spatial coordi-
nates and sometimes as the Eulerian coordinates.
4. Relative description. The independent variables are the current posi-
tion vector x and a variable time t. The variable time t is the time at
which the particle occupied another position with position vector 1.
The motion is described with 1 as the dependent variable, that is,
1 ¼ X t ðx, t Þ,
where the subscript t indicates that the reference configuration is the

configuration occupied at time t or that x is the position vector at time t.
That is, we have a variable reference configuration.
Descriptions (2) and (3) are emphasized in the applications that follow and
(1) and (4) will play only a minor role in the topics considered in this book.
We will take the reference configuration for description (2) to be the con-
figuration at time t ¼ 0, usually the undeformed state of the body, and will
refer the material and spatial coordinates to the same fixed rectangular
Cartesian coordinate system.
It is convenient to use lowercase suffixes for the spatial configuration,
for example, the position vectors of a particle in the spatial and reference
configurations are given by
x ¼ xi ei ,
and
X ¼ XK EK ,
respectively. Since the same coordinate system is considered, the unit

base vectors ei and EK coincide, that is,
ðe1 , e2 , e3 Þ ¼ ðE1 , E2 , E3 Þ:
The motion of a body may be represented by
^
x ¼ xðX, tÞ: ð2:1Þ
Equation (2.1), for a particular particle with position vector X at t ¼ 0, is

the parametric equation with parameter t, of a space curve passing through
point X, and this curve is the particle path of the particle. A stream line is
a curve whose tangent at a fixed time at each of its points is in the direction
of the velocity vector.
Although the motion is completely determined by equation (2.1), it is often
desirable to consider the state of motion at fixed point in space, described
by the velocity field
^
v ¼ vðx, tÞ: ð2:2Þ
Equation (2.2) represents the velocity at time t of the particle occupying the
point with position vector x. The stream lines for a fixed time t0 are
given by the solution of the three, in general nonlinear, ordinary
differential equations,
^
dxi vi
¼ , i 2 f1, 2, 3g,
ds jvj
where s is the distance measured along the curve so that dx=ds is the unit
vector tangential to the curve. A stream
. line and particle path coincide
^
in steady flow, that is, if @vðx, tÞ @t [ 0, but the converse is not
necessarily true.
2.2 Spatial and Referential Descriptions 53
2.2 Spatial and Referential Descriptions
The velocity can also be given as a function of X and t,
v ¼ V ðX, tÞ: ð2:3Þ
Equation (2.3) represents the velocity at time t of the particle that occupied
the point X in the reference configuration.
Similarly it follows from equation (2.1) that any quantity f that may be
expressed as a function
^
f ¼ f ðx, tÞ
of the variables x and t may also be expressed as a function
^
f^ðX, tÞ ¼ f ðxðX, tÞ, tÞ
of X and t. These are known as the spatial and referential descriptions,

respectively. There are two different partial derivatives of f with respect to
^
time, @f ðx, tÞ=@t, which is the rate of change of f with respect to time at
a fixed point, and @ f^ðX, tÞ=@t, which is the rate of change of f with respect to
time of the particle with initial position X and is known as the material
derivative. The material derivative can also be expressed as
1 n^ o
^
df ^
¼ lim f ðx þ vDt, t þ DtÞ f ðx, tÞ : ð2:4Þ
dt t!0 Dt
Using suffix notation and expanding the right-hand side of equation (2.4) as
a Taylor series gives1
" ^ #
^
df 1 @f
^
@f ^ ^
2 ^ @f @f
¼ lim vi Dt þ Dt þ O Dt , ¼ vi þ , ð2:5Þ
dt Dt @xi @t @xi @t
1
The meaning of the O symbol is as follows: f ðxÞ ¼ OðgðxÞÞasx ! 0 implies that
limjf ðxÞj=jgðxÞj ¼ K where K is bounded. When K = 0, we use the small o symbol,
x!0
that is, f ðxÞ ¼ ogðxÞ implies that limjf ðxÞj=jgðxÞj ¼ 0.
x!0
or, in symbolic notation,
^ ^
df ^ @f
¼ v =f þ : ð2:6Þ
dt @t
Since v = is a scalar differential operator, equation (2.6) is valid if f is

a scalar or tensor of any order. An alternative notation for the material
derivative dðÞ=dt is DðÞ=Dt: The velocity of a particle is a particular case of
a material derivative since
^ dx @ ðxðX, tÞÞ
v¼ ¼ ¼ x:
_ ð2:7Þ
dt @t
The motion is called steady flow if the velocity is independent of time, that
is, if @v=@t ¼ 0.
2.3 Material Surface
A material surface is a surface embedded in a body so that the surface is

occupied by the same particles for all time t. In most problems the bound-
ary of a continuous body is a material surface. An exception is when abla-
tion occurs, as in a burning grain of propellant.
A necessary and sufficient condition for a surface /ðx1 , x2 , x3 , tÞ ¼ 0 to
be a material surface is
d/ @/ @/
¼ þ vi ¼ 0, ð2:8Þ
dt @t @xi
where v is the particle velocity and d/=dt is evaluated on / ¼ 0. The proof

is as follows. Suppose a typical point on the surface, /ðx, tÞ ¼ 0, is moving
with velocity u, not necessarily equal to v, the velocity of the particle in-
stantaneously occupying that point on the surface. The velocity u must
satisfy the condition
@/ @/
þ ui ¼ 0 ð2:9Þ
@t @xi
2.3 Material Surface 55
since
/ðx; tÞ ¼ 0:
The normal component of u on the exterior normal to / ¼ 0 is
us ð@/=@xs Þ
uðnÞ ¼ u n ¼ ,
@/ @/ 1=2
@xk @xk
and with the use of equation (2.9) this becomes
ð@/=@tÞ
uðnÞ ¼ : ð2:10Þ
@/ @/ 1=2
@xk @xk
The normal component of the particle velocity v on the exterior normal to

/ ¼ 0 is
vs ð@/=@xs Þ
vðnÞ ¼ : ð2:11Þ
@/ @/ 1=2
@xk @xk
Substituting for @/=@t from equation (2.10) and for vs ð@/=@xs Þ from equa-
tion (2.11) in
d/ @/ @/
¼ þ vs
dt @t @xs
gives

d/ @/ @/ 1=2
¼ vðnÞ uðnÞ : ð2:12Þ
dt @xk @xk
If the surface always consists of the same particles, uðnÞ ¼ vðnÞ , and it follows
from equation (2.12) that d/=dt ¼ 0, evaluated on / ¼ 0, is a necessary
condition for / ¼ 0 to be a material surface. This proof of the necessity of
the condition can be given in words. If the surface is a material surface, the
velocity, relative to the surface, of a particle lying in it must be wholly
tangential (or zero); otherwise we would have flow across the surface.
It follows that the instantaneous rate of variation of / for a surface parti-
cle must be zero, if /ðx, tÞ ¼ 0 is a material surface. To prove the sufficiency
of the condition we put
UðX, tÞ ¼ /ðxðX, tÞ, tÞ ¼ 0, ð2:13Þ
so that d/=dt ¼ 0 implies that
@UðX, tÞ
¼ 0:
@t
This further implies that U = 0 is the equation of a surface fixed in the

reference configuration, and it then follows from equation (2.13) that
/ðx, tÞ always contains the same particles.
2.4 Jacobian of Transformation and Deformation Gradient F
Let dV be the volume of an infinitesimal material parallelepiped ele-

ment of a continuous body in the reference configuration, taken as the
undeformed configuration, and let dv be the volume of the parallelepiped
in the current or spatial configuration. Contiguous sides of the parallelepi-
ped are the material line elements dX ð1Þ , dX ð2Þ , dX ð3Þ in the reference con-
figuration, and dxð1Þ , dxð2Þ , dxð3Þ in the spatial configurations are as shown
in Figure 2.1.
The volume of a parallelepiped is given by the triple scalar product of the
vectors, representing contiguous sides, taken in the appropriate order,
consequently,
ð1Þ ð2 Þ ð3 Þ ð1 Þ ð2Þ ð3Þ

dV¼dX 3dX dX ¼ eKLM dXL dXM dXK , ð2:14Þ
ð Þ ð Þ ð Þ
dv ¼ dx 1 3dx 2 dx 3 ¼ eijk dx1i dx2j dx3k : ð2:15Þ
2.4 Jacobian of Transformation and Deformation Gradient F 57
dX(3) dx(3)
dX(2)
dx(2) dx(1)
dX(1)
Figure 2.1. Deformation of infinitesimal parallelepiped.
Substituting

ðsÞ @xi ðsÞ ðsÞ 2
dxi ¼ dXL þ O dX , s 2 f1, 2, 3g ð2:16Þ
@XL
in equation (2.15) gives
@xi @xj @xk ð Þ ð Þ ð Þ

dv ¼ eijk dXL1 dXM2 dXK3 , ð2:17Þ
@XK @XL @XM
2
neglecting terms O dX ðsÞ .
It follows from equations (1.71), (2.14), and (2.17) that
dv ¼ JdV, ð2:18Þ
where

@xi
J¼det , ð2:19Þ
@XK
or, using a common notation,
@ ðx1 , x2 , x3 Þ
J¼ :
@ ðX1 , X2 , X3 Þ
The determinant J is the Jacobian of the transformation (2.1) and is some-

times known as the dilatation because of equation (2.18). It is evident from
physical considerations that
0 < J < ‘,
and this also follows from the invertibility of equation (2.1). A motion for
which there is no volume change for any element of the body is described as
isochoric. If a body is assumed to be incompressible, it can only undergo
isochoric motions, otherwise thermodymamic anomalies arise. It follows
from equation (2.18) that J ¼ 1 for an isochoric motion.
The elements Fik ¼ @xi =@XK of the matrix in equation (2.19) are the
components of a second-order two-point tensor,
F ¼ FiK ei EK , ð2:20Þ
known as the deformation gradient tensor. It is implied by equation (2.20)

that F transforms as a vector for coordinate transformations in one config-
uration. This shows that F is related to both configurations and it is a linear
operator that operates on a vector in the reference configuration to give

a vector in the spatial configuration. If ðX, xÞ and X O , xO are the position

vectors of two adjacent particles in the reference and spatial configurations,
respectively, then
2
O @xi O

O
xi xi ¼ XK XK þ O XK XK : ð2:21Þ
@XK
When X and X O are an infinitesimal distance apart we can put

dX ¼ X X O and dx ¼ x xO so that
dx ¼ FdX, dxi ¼ FiK dXK : ð2:22Þ
Equation (2.21) should be compared with equation (2.16). The term two-
point tensor indicates the relationship to both the spatial and reference
configurations.
The deformation gradient tensor F can also be expressed in the form
F ¼ Gradx
where
@ðÞ
Grad [ EK :
@XK
The inverse F 1 of the deformation gradient tensor occurs frequently, and

its components are given by
1 @XK 1 @XK
FKi ¼ ,F ¼ E K ei ð2:23Þ
@xi @xi
and the inverse of equation (2.22) is
dX ¼ F 1 dx:
2.5 Reynolds Transport Theorem
We have already considered a material surface that is a surface that con-

tains the same particles for all time. Material lines and material volumes are
similar concepts. A material volume is bounded by a material surface or
surfaces and is frequently referred to as a system volume, especially in the
literature of thermodynamics.
There are certain theorems concerning the material derivatives of line,
surface, and volume integrals of a field /ðx, tÞ where / may be a scalar,
vector, or higher-order tensor. The most important of these theorems is
that for the volume integral that is usually known as Reynolds Transport
Theorem. This theorem can be expressed in the form
Z Z Z
d @/
/ðx, tÞdv ¼ dv þ /v ndS, ð2:24Þ
dt @t
vðtÞ vðtÞ @vðtÞ
where v is the velocity field, n is the outward unit normal to the surface S of
the material volume vðtÞ. According to equation (2.24), the rate of change
of the integral of f over the material volume vðtÞ is equal to the integral of
@/=@t over the fixed volume that coincides instantaneously with vðtÞ at time
t plus the flux /v through the surface S of this fixed volume.
In order to prove equation (2.24) we need the result
dJ
¼ J= v, ð2:25Þ
dt
which is due to Euler’s Equation. The proof of equation (2.25) is as follows.

First we note that the Jacobian J is given by
@xi
Jdij ¼ AjK , ð2:26Þ
@XK
where AiK is the cofactor of FiK ¼ @xi =@XK in the determinant J. Since the
derivative of J with respect to t is the sum of the three determinants formed
by replacing the elements of a different row (or column) of each determi-
nant by their time derivatives,

dJ d @xi @vi @vi @xj
¼ AiK ¼ AiK ¼ AiK ,
dt dt @XK @XK @xj @XK
and with the use of equation (2.26) it follows that
dJ @vi
¼J ¼ J= v: ð2:27Þ
dt @xi
Using equations (2.1) and (2.18) we obtain
Z Z
d @
/ðx, tÞdv ¼ J /^ðX, tÞ dV, ð2:28Þ
dt @t
vðt Þ V
^ ðX, tÞ ¼ /ðxðX, tÞ, tÞ. Since the integral on the right-hand side of
where /
equation (2.28) is over a fixed volume we can differentiate inside the in-
tegral sign to obtain
Z Z !
d @ /^ ^ dJ
/ðx, tÞdv ¼ J þ/ dV,
dt @t dt
vðtÞ V
and with the use of equation (2.27) this becomes

Z Z
d d/
/ðx, tÞdv ¼ þ /divv dv: ð2:29Þ
dt dt
vðtÞ vðt Þ
Equation (2.24) can be obtained from equation (2.29) with the use of
the identity
= ð/vÞ ¼ /= v þ v =/,
^
relation (2.6), with f replaced by /, and the divergence theorem.
Equations (2.24) and (2.29) are valid when /ðx, tÞ is continuous. Sup-
pose /ðx, tÞ has a discontinuity across a surface +ðtÞ, that moves with
velocity wðx, tÞ, not necessarily equal to the field velocity v, and divides
the volume v into two parts, vþ and v , as indicated in Figure 2.2. The unit
normal m to +ðtÞ points from the –side to the +side.
Then, equations (2.24) and (2.29) must be replaced by equivalent forms
Z Z Z Z
d @/
/ðx, tÞdv ¼ dv þ /v ndS ½½ðv wÞ m/da,
dt @t
vðtÞ vðtÞ @vðtÞ RðtÞ
ð2:30aÞ
Z Z Z
d d/
/ðx, tÞdv ¼ þ /divv dv ½½ðv wÞ m/da, ð2:30bÞ
dt dt
vðtÞ vðt Þ RðtÞ
respectively, where ½½ðv wÞ m/ ¼ ððv wÞ m/Þþ ððv wÞ m/Þ

and the superscripts + and – denote the values on the + and –side of
+ðtÞ, respectively. A proof of equation (2.30a) is as follows. The rates of
a+(t)
v (x,t)
m
w(x,t)
v+
(t)
v-
a-(t)
Figure 2.2. Discontinuity across surface RðtÞ:
change of the integral of / over the volumes vþ ðtÞ and v ðtÞare equal to the
integral of @/=@t over the fixed volumes that coincide instantaneously with
vþ ðtÞ and v ðtÞat time t plus the flux /v through the surfaces of the fixed
volumes. It then follows that
Z Z Z
d @/
/ðx, tÞdv ¼ dvþ /v nda
dt @t
þ þ þ
v ðt Þ v ðt Þ a ðt Þ
Z

ðv wÞ m/þ da
RðsÞ
Z Z Z
d @/
/ðx, tÞdv ¼ dvþ /v ndaþ
dt @t
v ðt Þ v ðtÞ a ðt Þ
Z
ððv wÞ m/ Þda,
RðsÞ
and addition of these equations gives equation (2.30a)

Equation (2.30) has applications in wave propagation problems that in-
volve shock waves.
The one-dimensional form of equation (2.24) is,
2 2
Zx ðtÞ Zx ðtÞ
d @/
/ðx, tÞdx¼ dx þ x_ 2 /ðx2 , tÞ x_ 1 /ðx1 , tÞ,
dt @t
x1 ðtÞ x1 ðtÞ
where a superposed dot denotes differentiation with respect to t. This is just

Leibniz’s rule in elementary calculus for differentiation of an integral with
variable limits.
2.6 Continuity Equation 63
The theorems that involve material differentiation of line and surface

integrals are given in the problem set at the end of this chapter.
2.6 Continuity Equation
Conservation of mass requires that
Z
d
qðx, tÞdv ¼ 0,
dt
vðtÞ
where q is the density in the spatial configuration and vðtÞ is a material

volume. It then follows from the form (2.29) of the transport equation that
Z
dq
þ q= v dv ¼ 0:
dt
v ðt Þ
Since this result must hold for an arbitrary material volume and the in-
tegrand is continuous,
dq
þ q= v ¼ 0, ð2:31Þ
dt
which is the spatial form of continuity equation and which can also be
expressed in the form
@q
þ = ðqvÞ ¼ 0,
@t
by using the expression for the material derivative of q and the relation
= ðqvÞ ¼ q= v þ v =q. The material form of the continuity equation is
qO ¼ qJ, ð2:32Þ
where qO is the density in the reference configuration. Equation (2.32) is

obtained from equation (2.18) and qO dV ¼ qdv, which follows from
conservation of mass. Equation (2.31) can also be obtained by taking the

material derivative of equation (2.32) and using equation (2.22).
A useful result
Z Z
d du
qudv ¼ q dv ð2:33Þ
dt dt
vðtÞ vðt Þ
can be obtained from equations (2.29) and (2.31).
2.7 Velocity Gradient
The velocity gradient tensor, = v, is the gradient of the velocity vector v;

consequently, it may be expressed in terms of its components vi;j relative to
a rectangular Cartesian coordinate system as
@vi
=v¼ ei ej :
@xj
A common notation is
L ¼ = v,
and henceforth, L denotes the velocity gradient.

Material differentiation of the deformation gradient gives

d @xi @vi @vi @xj
¼ ¼ ,
dt @XK @XK @xj @XK
and, in symbolic notation,
F_ ¼ LF ð2:34Þ
or
_ 1 ,
L ¼ FF ð2:35Þ
which gives the relation between the velocity gradient and the deformation
gradient.
and let the velocities of P and Q
Consider a material line element PQ,
be vO and v, respectively. It then follows from a Taylor series expansion
that the velocity of Q relative to P is

v vO ¼ = vðx xO Þ þ O jx xO j2 ,
where xO and x are the position vectors of P and Q, respectively. When PQ

is an infinitesimal line element we can replace this expression by
dv ¼ = vdx, ð2:36Þ
@vi
dvi ¼ dxj :
@xj
We must note the meanings of dx and dv, namely, that dx is the vector of
and dv is the velocity of Q
the infinitesimal directed line segment PQ
relative to P. The material time derivative of dx is dv, since

d d @xi @vi @vi @XK @vi
ðdxi Þ ¼ dXK ¼ dXK ¼ dxj ¼ dxj ¼ dvi :
dt dt @XK @XK @XK @xj @xj
ð2:37Þ
2.8 Rate of Deformation and Spin Tensors
The symmetric and antisymmetric parts D and W , respectively, of the

velocity gradient tensor, L ¼ = v ¼ D þ W , are known as the rate of
deformation and spin tensors, respectively, so that
1 T

D¼ LþL ð2:38Þ
2
and
1
W¼ L LT : ð2:39Þ
2
We use notations dij and wij for the components of D and W, respectively,
so that

1 @vi @vj
dij ¼ þ
2 @xj @xi
and

1 @vi @vj
wij ¼ :
2 @xj @xi
The material derivative of the square of the length ds of an infinites-

¼ dx is
imal material line element PQ
d 2 d d @vi
ds ¼ ðdxi dxi Þ ¼ 2dxi ðdxi Þ ¼ 2 dxj dxi ¼ 2 dij þ xij dxi dxj
dt dt dt @xj
¼ 2dij dxi dxj ,
ð2:40Þ
since
xij dxi dxj ¼ 0: ð2:41Þ
Also note that
d 2 d
ds ¼ 2ds ðdsÞ,
dt dt
and using this along with equation (2.40) gives
1 d dxi dxj
ðdsÞ ¼ dij ¼ dij ni nj : ð2:42Þ
ds dt ds ds
Equation (2.42) gives an insight into the physical significance of the rate of
deformation tensor D, which is often called the strain rate tensor especially
in plasticity theory.
It follows from equation (2.40) or (2.42) that a necessary and sufficient
condition for a rigid body motion, that is, a motion for which the distance
between any two particles remains invariant, is dij ¼ 0. This is known as
Killing’s theorem. It was shown in the previous section that the velocity of
Q relative to P is dv ¼ = vdx for the material line element PQ ¼ dx,
and it may be deduced from Killing’s theorem that the part W dx of
= vdx ¼ Ldx corresponds to a rigid body motion.
In order to see the significance of the spin tensor W , it is desirable to
introduce the vector

1 1 @v3 @v2 @v1 @v3 @v2 @v1
v ¼ curlv ¼ e1 þ e2 þ e3 :
2 2 @x2 @x3 @x3 @x1 @x1 @x2
In fluid mechanics the vector 1 ¼ 2x ¼ curlv is known as the vorticity

vector.
Comparison of the components of the vorticity vector 2x with the compo-
nents of the spin tensor shows that x is the vector of the antisymmetric spin
tensor W . Consequently,
1
xi ¼ eijk wjk ,wij ¼ eijk xk : ð2:43Þ
2
The significance of the vector x and the tensor W then becomes evi-
dent from the following theorem. The angular velocity of the triad of
material line elements that instantaneously coincides with the principal
axes of the rate of deformation tensor D is given by v ¼ 12 curlv. In order
to prove this theorem we need to show that
d ðaÞ ð Þ
n ¼ x3n a , ð2:44Þ
dt
where nðaÞ ; a 2 f1, 2, 3g are the principal directions or eigenvectors of D.

This can be shown as follows:

1 d ðaÞ
ðaÞ
dnk d dxk 1 d
¼ ¼ ðdxk Þ ds dxk , ð2:45Þ
dt dt dsðaÞ ð Þ
ds a dt ð Þ2
ds a dt
where dsðaÞ is an element of length in the corresponding principal direction.

Using equations (2.37), (2.42), (2.43), and

ða Þ ðaÞ
dik d dki ni ¼ 0,
where dðaÞ are principal values of the rate of deformation tensor D, it

follows from equation (2.45) that
ðaÞ
dnk @vk ðaÞ ð Þ ðaÞ ð Þ ð Þ
¼ n d a nk ¼ wki ni a ¼ ekji xj ni a ¼ ðv 3 nÞk ,
dt @xi i
and the result (2.44) is proved.

Another interpretation of the vorticity 2v is as follows. If an element
of a continuum with its mass centre at P is instantaneously made rigid
without change of angular momentum, then its angular velocity immedi-
ately after it has been made rigid is v ¼ 12 curlv, if and only if the principal
axes of inertia for the resulting rigid element lie along the principal axes of
the rate of deformation tensor evaluated at P immediately before the ele-
ment becomes rigid. The proof of this depends on certain results in rigid
body dynamics and is as follows. Let r be the position vector of a particle of
the element with respect
to P. The element is assumed to be sufficiently
small that terms O jrj2 may be neglected. At the instant immediately
before becoming rigid the angular momentum H of the element about its
mass center is
Z
H¼ qr3v dv,
v
where v is the volume of the element and the velocity relative to P of

a particle, of the element, is given by
@vi
vi ¼ rs þ O jr j2 ;
@xs
and @vi =@xs is evaluated at P. The angular momentum H of the element is

then given by
Z Z
@vk
Hi ¼ eijk qrj rs dv ¼ eijk ðdks þ wks Þ qrj rs dv: ð2:46Þ
@xs
v v
At the instant immediately after becoming rigid without change in angular

momentum,
Hi ¼ Iij x9j : ð2:47Þ
where v9 is the angular velocity of the rigid element and
Z
Iij ¼ q r2 dij ri rj dv, ð2:48Þ
v
where Iij are the components of the inertia tensor of the element. Also at
the instant immediately after the element becomes rigid, the rate of de-
formation tensor is zero, so that
Z
Hi ¼ eijk x9ks qrs rj dv, ð2:49Þ
v
0
where xks are the components of the spin tensor. Equation (2.49) is equiv-
alent to equation (2.47), since substituting equation (2.38)2 in equation
(2.43) gives
Z Z

Hi ¼ eijk estk x9t qrs rj dv ¼ dij djt dit djs x9t qrs rj dv
Z v v
¼ x9j q r 2 dij ri rj dv:

v
R
Referring to equation (2.46), dks rs rj dv are the components of a second-
v R
order symmetric tensor if and only if drs and qrs rj dv are coaxial, that is, if
v
and only if they have the same principal axes or eigenvectors. Also the
R
principal axes of Iij and qrs rj dv coincide. Comparison of equations (2.46)
v
and (2.49) shows that if and only if dij and Iij are coaxial,
wij ¼ w9ij
since
Z
eijk dks qrs rj dv¼0,
v
R
if and only if drs qrs rj dv is symmetric.
v
2.9 Polar Decomposition of F
The polar decomposition theorem given in chapter 1 can be applied to the

deformation gradient tensor F since det½F > 0, so that we have the unique
decompositions,
F ¼ RU, F ¼ VR, ð2:50Þ
FiK ¼ RiK UKL ¼ Vij RjK :

T T
U ¼ R VR,V ¼ RUR : ð2:51Þ
According to the Polar Decomposition Theorem given in chapter 1, U

and V are symmetric positive definite second-order tensors with the same
eigenvalues, ka , a 2 f1, 2, 3g, that are known as the principal stretches, and
R is a proper orthogonal tensor so that
det½R ¼ 1 and RT ¼ R1 :
If R ¼ I, F ¼ U ¼ V, and the deformation is pure strain with no rotation.

The spectral forms of U and V are
3 3
ða Þ ðaÞ ð Þ ð Þ
U ¼ + ka N N , V ¼ + ka n a n a , ð2:52aÞ
a¼1 a¼1
if the ka are distinct, where N ðaÞ and nðaÞ are the eigenvectors of U and V,
respectively, and

ð3Þ ð3 Þ
ð3Þ ð3Þ
U ¼ k3 N N
þk IN N ,
ð2:52bÞ
ð Þ ð Þ ð Þ ð Þ
V ¼ k3 n 3 n 3 þ k I n 3 n 3 ,
if k1 ¼ k2 ¼ k 6¼ k3 , and
U ¼ kI, V ¼ kI: ð2:52cÞ
It follows from equations (2.51) and (2.52a) that
ð Þ ðaÞ ðaÞ ð Þ
n a ¼ RN ,N ¼ RT n a :
Since R is an orthogonal tensor, a vector is unchanged in magnitude

when operated on by R. It follows that for a rigid body motion F = R and
U = I, V = I. This may be shown otherwise since, for a rigid body motion,
dx dx ¼ dX dX
for all material line elements at a point, and using equation (2.20) and the
substitution operator property of the Kronecker Delta we obtain
FiK FiL ¼ dKL
for rigid body motion, which implies F = R and U = I, V = I.

The principal components or eigenvalues, ka ; a ¼ f1, 2, 3g, of U and V
are called the principal stretches of U and V, the right and left stretch
tensor, respectively. The form ð2:50Þ1 corresponds to a pure deformation
followed by a rigid body rotation given by R that is sometimes called the
rotation tensor. Similarly the form ð2:50Þ2 corresponds to a rigid body
rotation, given by R, followed by a pure deformation. A pure deformation
is as follows. Consider a material rectangular parallelipiped with sides of
ðaÞ
length lO , a ¼ f1, 2, 3g that is uniformly deformed into a rectangular par-
allelipiped with corresponding sides in the same directions and of length
ðaÞ
lðaÞ . Each side undergoes a stretch ka ¼ lðaÞ =lO .
As an example of polar decomposition we consider simple shear de-
formation given by
x1 ¼ X1 þ KX2 , x2 ¼ X2 , x3 ¼ X3 , ð2:53Þ
as indicated in Figure 2.3.

The components of F are given by
2 3
1 K 0
½F ¼ 4 0 1 0 5: ð2:54Þ
0 0 1
This is a problem of plane motion where all the particles move parallel to
the 0x1 x2 plane and the motion does not depend on x3 . Consequently, the
rotation tensor R is of the form
2 3
cos h sin h 0
½R ¼ 4 sin h cos h 0 5, ð2:55Þ
0 0 1
X2
0
X1
Figure 2.3. Simple shear deformation.
where R produces a clockwise rotation, when h is measured in the clock-

wise direction and R operates on a vector in the X1 X2 plane.
We consider the form (2.50)2 so that
V ¼ FRT
and it then follows from equations (2.54) and (2.55) that
2 3 2 32 3
V11 V12 V13 1 K 0 cos h sin h 0
4 V21 V22 V23 5 ¼ 4 0 1 0 54 sin h cos h 0 5
V31 V32 V33 20 0 1 0 0 1 3
cos h þ K sin h sin h þ K cos h 0
¼4 sin h cos h 0 5:
0 0 1
Since V21 ¼ V12 ,
sin h þ K cos h ¼ sin h:
Consequently,
K
tan h ¼ :
2
The components of V and R are then given by
1=2
V11 V12
¼ 4þK
2 2 þ K2 K þ 2K
V21 V22 K 2
and
V33 ¼ 1, V31 ¼ V32 ¼ 0,

1=2 2
R11 R12 2 K
¼ 4þK :
R21 R22 K 2
R33 ¼ 1, R31 ¼ R32 ¼ 0:
Components of U can be obtained similarly.
2.10 Further Decomposition of F
The deformation gradient F can be decomposed into the product of a

dilatational part: and an isochoric or distortional part, as follows

F ¼ J 1=3 I F : ð2:56Þ
Since det½F ¼ 1, F is the isochoric part and is known as the modified

deformation gradient.
Other related decompositions that follow from equation (2.56) are

U ¼ J 1=3 I U and C ¼ J 2=3 I C : ð2:57Þ
It may be deduced from equation ð2:57Þ1 that the stretches ka , a 2 f1, 2, 3g,
that are the principal components of U can be similarly decomposed as
1=3
ka ¼ J ka ,
where k1 k2 k3 ¼ 1 and the ka are known as the modified stretches.
2.11 Strain 75
2.11 Strain
The right and left stretch tensors U and V, respectively, which result from
the polar decomposition of the deformation gradient tensor F, may be
1=2 1=2
U ¼ F TF , V ¼ FF T , ð2:58Þ
so that the squares of U and V are given by the positive definite symmetric
tensors
C ¼ U 2 ¼ F T F and B ¼ V 2 ¼ FF T , ð2:59Þ

@xi @xi @xi @xj
CKL ¼ and Bij ¼ :
@XK @XL @XK @XK
Tensors C and B are known as the right and left Cauchy strain tensors,
respectively, and are a measure of the deformation at a material point in
a body since the square of the length of a material line element in the spatial
configuration is given by
dxi dxi ¼ CKL dXK dXL ¼ dX CdX ð2:60Þ
and in the reference configuration by
1 1
dXK dXK ¼ Bij dxi dxj ¼ dx B dx: ð2:61Þ
It follows that if C = I, which implies that B = I, for every material point in

the body, the body has undergone a rigid body displacement.
Another measure of deformation is given by the tensor E known as
Green’s strain tensor and defined by
dx dx dX dX ¼ dX EdX ð2:62Þ
dxi dxi dXL dXL ¼ 2EMN dXM dXN :

It follows from equations (2.59), (2.60), and (2.62) that
1 1 T 1
2

E ¼ ðC I Þ ¼ F FI ¼ U I , ð2:63Þ
2 2 2
and, in suffix notation,

1 @xi @xi
EKL ¼ dKL :
2 @XK @XL
For a rigid rotation dx dx ¼ dX dX, consequently C ¼ I and E ¼ 0.

The displacement vector j of a particle is given by
j ¼ x X: ð2:64Þ
In suffix notation this equation presents a problem if we use the convention

of an uppercase suffix referring to the reference configuration and a lower
case suffix for the spatial configuration. This can be resolved as follows by
expressing equation (2.64) in suffix notation as
ni ¼ xi XK dKi , ð2:65Þ
where
EK ei ¼ dKi :
It follows from equations ð2:63Þ2 and (2.65) that E may be expressed in the
component form
" #
1 @ n^K @ n^L @ n^J @ n^J
EKL ¼ þ þ , ð2:66Þ
2 @XL @XK @XK @XL
where n^K ¼ dKi ni , and in this form E is sometimes called the Lagrangian
strain.
A further measure of deformation is given by the tensor h defined by
dxi dxi dXK dXK ¼ 2glm dxl dxm ¼ 2dx hdx, ð2:67Þ
2.11 Strain 77
and it follows from equations (2.61) and (2.67) that

1 1
h¼ I B1 ¼ I F T 1 F , ð2:68Þ
2 2

1 @XK @XK
gij ¼ dij ,
2 @xi @xj
and with the use of equation (2.65) this can be put in the form

1 @ni @nj @ns @ns
gij ¼ þ :
2 @xj @xi @xi @xj
The tensor h is known as Almansi’s strain tensor or the Eulerian strain

tensor. It is less useful than E. It is easily shown that
T 1 1
h¼F EF :
It is useful to note that the components of the tensors C and E are

independent of any rigid body rotation of the spatial configuration and the
components of B and h are independent of any rotation of the reference
configuration.
The material derivative of E, Green’s strain tensor, is related to D, the
rate of deformation tensor by
E_ ¼ F T DF, ð2:69Þ
@xi @xj
E_ KL ¼ dij :
@XK @XL
This result is obtained from equations (2.34) and (2.63) as follows, using
symbolic notation,
1 T
1 d T
_
E¼ F FI ¼ F_ F þ FT F_
2 dt 2
1 T T
¼ F L F þ F T LF ¼ F T DF:
2
The tensor E_ is known as the strain rate tensor, although sometimes D

is called the strain rate tensor especially in plasticity theory. It is shown in
the next section that D ’ E_ when j@ui =@XK j << 1.
There are several other strain tensors in addition to the Green and
Almansi strain tensors, for example, the Biot strain tensor,
Eb ¼ U I,
and the logarithmic strain tensors lnU and lnV whose principal components
are lnki .
2.12 Infinitesimal Strain
A tensor known as the infinitesimal strain tensor is a linearization of equa-

tion (2.66), given by
T
1 ^ ^
e¼ =X j þ = X j ð2:70Þ
2
where
@ n^
=X j^ ¼ K EK EL
@XL
is the displacement gradient tensor.

For the present we will not express e in component form. The infinitesimal
strain is related to Green’s strain tensor by the relation obtained from
equations (2.66) and (2.70)
T
1
E ¼eþ =X j^ =X j:
^
2
2.12 Infinitesimal Strain 79

Consequently, if =X jîs, in some sense, small, it is reasonable to make the
approximation E ’ e since

E ¼ e þ O e2 ð2:71Þ
where

e ¼ =X n^,
so that as e ! 0, E ! e. It follows that if e << 1, e is a good approximation

for E. As e ! 0, the difference between the spatial and reference config-
urations becomes negligible. Consequently, a linearization of equation
(2.64a) is also a suitable infinitesimal strain tensor.
The tensor e is not a suitable measure for finite strain since e does not,
in general, vanish for a rigid body motion. This may be shown as follows,
suppose we have an infinitesimal rigid body rotation given by
F ¼ Q ¼ I þ a, jaj << 1
^ The orthogonality condition QQT ¼ I gives

where a ¼ =X j.

a ¼ aT þ O e2 ,
where e ¼ jaj. Green’s strain tensor vanishes for a rigid displacement of the
spatial configuration so that from equation (2.67) we have, for a rigid body
rotation,

e ¼ O e2 :
That is e vanishes to within an error O e2 when F = Q so that

unless the rotation is infinitesimal e is not a suitable approximation
for E.
We now introduce the polar decomposition F = RU and put
R ¼ ðI þ aÞ and U ¼ ðI þ bÞ
noting that a and b are both of OðeÞ. It follows from the orthogonality of R
and the symmetry of U that

a ¼ aT þ O e2 , b ¼ bT :
Then we have

2
=X j^ ¼ F I ¼ a þ b þ O e ,
that is, for infinitesimal deformation, when terms O e2 can be neglected,

the tensor b is the symmetric part of the displacement gradient tensor and
is identical to e given by equation (2.61). The antisymmetric tensor a can be
regarded as a measure of the infinitesimal rotation.
It should be noted that the use of infinitesimal strain tensor e is justified
when j=X U j ¼ 1, and this implies that both the rotation and E are small not
just the E. When the use of the infinitesimal tensor is justified when terms

O e2 can be neglected, the distinction between spatial and material coor-

dinates can be neglected and we can express e and a in coordinate form as

1 @ui @uj 1 @ui @uj
eij ¼ þ , aij ¼ : ð2:72Þ
2 @xj @xi 2 @xj @xi
The theory for the infinitesimal strain and rotation tensors is analogous to
the theory for the rate of deformation tensor D and the spin tensor W with
the infinitesimal displacement analogous to the velocity.
When the use of infinitesimal strain approximation (2.69) is justified it
is evident that
e_ ij dij :
This is perhaps the reason why D is often called the strain rate tensor.
EXERCISES
2.1. If da and dA are the vectors representing an element of area in the

spatial and reference configurations, respectively, show that
q0 @XK
dai ¼ dAK ,
q @xi
or, in symbolic notation,
q0 1T
da ¼ F dA:
q
This relation is known as Nanson’s formula

(a) Use Nanson’s formula to show that Div JF 1 ¼ 0.

R
Note that, for a closed surface enclosing volume V ¼ da ¼ 0.
@V
2.2. Prove the following results
(a)

d @XK @vj @XK
¼ :
dt @xi @xi @xj
(b)
@ 1
J FiK ¼ 0:
@xi
(c)
@UK @ 1
¼J J FiK UK :
@XK @xi
2.3. Let dxð1Þ and dxð2Þ be material line elements emanating from a particle
P. They specify a material surface element given by
ð Þ ð Þ
da ¼ dx 1 3dx 2 :
Show that
d @vj @vj
ðdai Þ ¼ dai daj :
dt @xj @xi
2.4. If F = RU is the polar decomposition of the deformation gradient

show that
T
_ Tþ 1 _ 1 RT RUU
_ 1 RT
W ¼ RR RUU
2
and
1 n _ 1 o
D ¼ R UU þ U 1 U_ RT :
2
2.5. If e is the isotropic part of the infinitesimal strain tensor show that

J ¼ 1 þ 3e þ 0 e2 ,
where e ¼ j=X U j.
2.6. Show that the acceleration vector can be expressed in the form
@v 1
a¼ þ ð= vÞv ð= vÞT v þ =v v:
@t 2
2.7. Show that

@v @ ðqvÞ
q þ v gradv ¼ þ divðqv vÞ:
@t @t
2.8. Prove that, for each particle of a deforming continuum there is at least
one material line element whose direction is instantaneously
stationary.
2.9. Obtain the polar decompositions F = RU and F = VR for the

deformation
2 3
3
1 2 0
½F ¼ 4 0 1 0 5:
0 0 1
3 Stress
3.1 Contact Forces and Body Forces
In order to introduce the concept of stress we first consider the spatial

configuration of a body that occupies a region v of Euclidean point space
E3 . At an interior point PðxÞ, of the body, there is an infinite number of
plane elements that contain P. Let n be the unit normal to one of these
elements with area Da as indicated in Figure 3.1.
The infinitesimal force DF acts on a surface element Dan ¼ Da. Denote by
mþ the material adjacent to the element on the side from which n is direc-
ted and m the material on the other side. In general mþ exerts a force DF
on m , and m exerts an equal in magnitude but opposite in direction force
on mþ . The stress vector at P acting on the plane element with unit vector n
is defined as
DF
t ðn, x, tÞ ¼ lim :
Da!0 Da
For convenience we suppress the dependence on x and time t. The stress

vector t is known as a contact force. At present we assume that t depends
continuously on n so that
t ðnÞ ¼ t ðnÞ:
This condition is not satisfied when we have singular surfaces across which
discontinuities of certain field variables can occur, for example, for shock
waves.
84
3.1 Contact Forces and Body Forces 85
m+
P
a
m_
Figure 3.1. Force on area element at point Pð xÞ:
The resultant moment exerted by mþ on m is denoted by DM and we

assume that at P,
DM
lim ¼ 0,
Da!0 Da
so that contact forces are present but not contact torques. In certain media,
known as polar media, contact torques can exist; however, in this book we
will not be concerned with this aspect of continuum mechanics.
The stress vector at points on the surface @v of a body, with n the
outward unit normal to the surface, is known as the surface traction.
The stress vector can be resolved into a normal component
N ðnÞ ¼ n tðnÞ, ð3:1Þ
and a shearing component
n o1=2
SðnÞ ¼ jt ðnÞj2 N ðnÞ2 . ð3:2Þ
When N is positive the normal component is said to be tensile and

when N is negative it is said to be compressive, although the opposite
convention is often used in soil mechanics.
86 Stress
Forces acting on interior particles, due to external agencies such as

gravity, are known as body forces, and we denote the body force per unit
mass by f . It is assumed that there are no body torques.
Henceforth, some equations are given in both symbolic and Cartesian
tensor notation.
In what follows the density, denoted by q, is a scalar property with
physical dimensions of mass per unit volume. As usual the mass of a body is
a measure of the amount matter of the body.
According to a generalization of Newton’s second law the resultant
force acting on a body, due to surface tractions and body forces, is equal to
the rate of change of momentum of the body. Consequently,
Z Z Z
d
tda þ qf dv ¼ qmdv,
dt
@v v v
Z Z Z ð3:3Þ
d
ti da þ qfi dv ¼ qvi dv,
dt
@v v v
where vðx, tÞ is the velocity, qðx, tÞ is the density, dv is an element of vol-

ume, and da is an element of surface area. It is shown in chapter 2 that
Z Z
d dv
qmdv ¼ q dv:
dt dt
v v
The resultant moment acting on a body, about a fixed point in an inertial

frame of reference or about the mass center of the body, due to surface
tractions and body forces, is equal to the rate of change of momentum of
the body. Consequently,
Z Z Z
d
x 3 tda þ x 3 qf dv ¼ x 3 qvdv,
dt
@v v v
Z Z Z ð3:4Þ
d
eijk xj tk da þ eijk xj qfk dv ¼ eijk xj qvk dv,
dt
@v v v
where x is the position vector with respect to the fixed point or the mass
center.
3.2 Cauchy or True Stress
Consider a plane element at P whose normal is in the x1 direction.

Then,
t ðe1 Þ ¼ r11 e1 þ r12 e2 þ r13 e3 ¼ r1i e1i , ð3:5Þ
where r11 , r12 , and r13 are the x1 , x2 , and x3 components, respectively, of the
stress vector acting on the plane element; r11 is the normal component, and
r12 and r13 are the shear components. It follows that the resulting shearing
stress on the element is
1=2
Sðe1 Þ ¼ r212 þ r213 :
Similarly for plane elements at P whose normals are in the x2 and x3

directions,
t ðe2 Þ ¼ r21 e1 þ r22 e2 þ r23 e3 , ð3:6Þ

t ðe3 Þ ¼ r31 e1 þ r32 e2 þ r33 e3 , ð3:7Þ
1=2
Sðe2 Þ ¼ r221 þ r223 ,
1=2
Sðe3 Þ ¼ r232 þ r231 :
Equations (3.5) to (3.7) may be expressed in suffix notation as
t ðei Þ ¼ rij ej : ð3:8Þ

The matrix rij specifies the state of stress at P. Component rij is defined as
the stress component in the xj direction of the stress vector acting on the plane
element whose normal is in the xi direction. Some texts define this as rji rather
than rij . The signs of the components rij are obtained as follows. If the normal
to a plane element is in the 6 xi direction and the stress vector component is
88 Stress
in the 6 xj directions, the component rij is positive, otherwise it is negative.

Positive stress components are shown in Figure 3.2 for a plane state of stress.
Now consider an arbitrary area element with unit normal n and stress vector
t ðnÞ acting on it. Let this element be the face ABC of the infinitesimal
tetrahedron PABC with sides PA, PB, and PC in the x1 , x2 , and x3 directions,
respectively, as shown in Figure 3.3.
Also let Da be the area of ABC and h be the perpendicular distance of P from
ABC. The components of t are given by ti ¼ t ðnÞ ei . Then, the equation of
motion of the tetrahedron in the xi direction is

1 1
ð Þ
ti Da þ q hDa fi r1i PCB þ r2i PAC þ r3i PAB ¼ q hDa v_ i , ð3:9Þ
3 3
dv dvi
where v_ ¼ ¼ v_ i ei ¼ ei is the particle acceleration.
dt dt
Substituting
PCB ¼ n1 Da, PAC ¼ n2 Da, PAB ¼ n3 Da,
in equation (3.9) and letting h ! 0 gives
T
ti ¼ rji nj , t ¼ r n: ð3:10Þ

The elements of the matrix rji are the components of a second-order
tensor rT since rji nj are the components of a vector for arbitrary unit vector
x2 σ22
σ21
σ12
σ11 σ11
σ12
x1
σ21
σ22
Figure 3.2. Stress components for plane Cauchy stress.

x3
t
C
n
P
B x2
x1
Figure 3.3. Equilibrium of infinitesimal. Tetrahedron PABC.
n. This follows from the test for tensor character given in chapter 1 since rT
is a tensor so is r. If rij and rij9 are the components of stress referred to the
rectangular Cartesian coordinate systems Oxi and Ox9,
i respectively, it fol-
lows from the tensor transformation rule
rij9 ¼ aik ajl rkl , ð3:11Þ
rkl ¼ aik ajl rij9, ð3:12Þ
where aik are the elements of the transformation matrix ½a. Equations
(3.11) and (3.12) can be expressed in symbolic matrix notation as
½r9 ¼ ½a½r½aT ,
½r ¼ ½aT ½r9½a,
respectively.
Equations (3.1) and (3.10) give the expression
N ðnÞ ¼ rji nj n1 ¼ n rT n ð3:13Þ
for the normal stress acting on a plane element with unit normal n and the
resultant shearing stress is given by equation (3.2).
90 Stress
The stress tensor
r ¼ rij ei ej
is referred to the current or spatial configuration of the body and is

known as the Cauchy stress and is a particular case of the class of Euler-
ian tensors, that is, the class of tensors referred to the spatial configura-
tion. We have yet to show that it is symmetric for nonpolar media.
3.3 Spatial Form of Equations of Motion
Substituting equation (3.10) in equation (3.3) gives

Z Z
rji nj da þ qðfi v_ i Þdv ¼ 0,
@v v
and application of the divergence theorem gives

Z
@rji
þ qðfi v_ i Þ dv ¼ 0: ð3:14Þ
@xj
v
Since equation (3.14) must hold for an arbitrary volume and it is assumed
that the integrand is continuous,
@rji dvi
þ qfi ¼ q ,
@xj dt
ð3:15Þ
dv
divrT þ qf ¼ q :
dt
It should be noted that dv=dt ¼ @v=@t þ v:gradv, dvi =dt ¼

@vi =@t þ vj @vi @xj , is the material derivative of v. For linearized problems
dv=dt is replaced by @v=@t.
Similarly the angular momentum equation with (3.10) substituted in
equation (3.4) is
Z Z
eijk xj rrk nr da þ eijk xj qðfk v_ k Þdv ¼ 0 ð3:16Þ
@v v
in Cartesian tensor suffix notation. With the use of the divergence theorem,
the substitution property of the Kronecker delta and equation (3.15), equa-
tion (3.16) becomes
Z
eijk rjk dv ¼ 0. ð3:17Þ
v
Since equation (3.17) must hold for an arbitrary volume and it is assumed
that the integrand is continuous, eijk rjk ¼ 0. It then follows that
rjk ¼ rkj , r ¼ rT ,
which shows that the stress tensor r is symmetric in the absence of body
moments.
3.4 Principal Stresses and Maximum Shearing Stress
We have shown that the Cauchy stress at a point is completely described by

a second-order symmetric tensor r. The properties of symmetric second-
order tensors have been given in detail in chapter 1, and some of these
properties are repeated here in less detail for the Cauchy stress tensor.
If the direction of the stress vector acting on a plane element is normal
to the element, there is no shearing stress acting on the element and the
stress is a principal stress. It then follows that
rij nj ¼ kni , rn ¼ kn, ð3:18Þ
where k is an eigenvalue and n is an eigenvector.

Since r is symmetric it has three real eigenvalues, ra , a 2 f1, 2, 3g,
which may or may not be distinct. These eigenvalues of r are known as
principal stresses, and three corresponding eigenvectors are known as prin-
cipal directions and are denoted by the orthonormal triad nðaÞ , a 2 f1, 2, 3g.
It should be noted that the superscript a involves no summation. Equation
(3.18) can be put in the form

rij kdij nj ¼ 0, ðr kI Þn ¼ 0, ð3:19Þ
92 Stress
and it follows that rðaÞ , a 2 f1, 2, 3g are the roots of the cubic equation

det rij kdij ¼ 0: ð3:20aÞ
Equation (3.20) can be expanded as
k3 I1 k2 þ I2 k I3 ¼ 0,
where
1n o
I1 ¼ rii , I2 ¼ ðrii Þ2 rij rij , I3 ¼ det rij ,
2
1n 2
o
I1 ¼ trr, I2 ¼ ðtrrÞ2 trr , I3 ¼ det½r
2
are independent of the choice of coordinate system and are known as the
first, second, and third basic invariants, respectively, of the stress tensor r.
If the third basic invariant I3 ¼ 0, then the tensor r is singular and one of
the principal stresses is zero.
The components of nðaÞ , a 2 f1, 2, 3g, corresponding to ra , can be

obtained from any two of equations (3.19) and nðaÞ ¼ 1 when the ra have
been obtained.
When the ra are distinct, the spectral representation of the Cauchy
stress tensor is
3
r= R ra nðaÞ na , a 2 f1, 2, 3g,
a¼1
where the summation convention is suspended. If r1 ¼ r2 6¼ r3 any direc-

tion normal to nð3Þ is principal direction and the spectral representation is

ð3 Þ ð3 Þ ð3 Þ ð3 Þ
r ¼ r3 n n þ r1 I n n :
If r1 ¼ r2 ¼ r3 ¼ p, we have a hydrostatic or isotropic state of stress and

the spectral representation is
r ¼ pI, rij ¼ pdij :

Equation (3.19) can also be obtained by considering the stationary values,

with respect to n, of the normal stress N ðnÞ. In order to determine these
directions we must find the stationary values of rij ni nj subject to the subsidiary
condition ni ni ¼ 1. By introducing the Lagrangian multiplier k we have
@
rij ni nj kni ni ¼ 0,
@nk
and differentiation gives

rij kdij ni ¼ 0: ð3:20bÞ
that is identical to equation (3.19). The Lagrangian multiplier k is identified

with a principal stress and the normal stress N ðnÞ is stationary and is a prin-
cipal stress, when n is an eigenvector of r. It follows that the greatest and
least normal stresses at a point are principal stresses.
The directions associated with the stationary values of the magnitude
of the shearing stress, jSðnÞj, are determined in a similar manner except that
in this case it is desirable to express equation (3.2) with respect to axes
coinciding with the principal directions of r, that is,
2
2 2 2 2 2 2 2
S ðnÞ ¼ ðr1 n1 Þ þðr2 n2 Þ þðr3 n3 Þ r1 n1 þ r2 n2 þ r3 n3 : ð3:21Þ
It is evident from equation (3.21) that if n1 ¼ 61, n2 ¼ n3 ¼ 0;

n2 ¼ 61, n1 ¼ n3 ¼ 0; or n3 ¼ 61, n1 ¼ n2 ¼ 0, jSðnÞj ¼ 0, which is as
expected since the shearing stresses acting on plane elements normal to
the principal directions are zero. We introduce a Lagrangian multiplier k to
obtain the directions associated with other stationary values of S2 ðnÞ given
by equation (3.21) and subject to the constraint ni ni ¼ 1. Then stationary
values, in addition to the zero values on the plane elements whose normals
are in the principal directions, are obtained from
@ 2
S kni ni ¼ 0,
@nk
94 Stress
where S2 is given by equation (3.21). After some computation we obtain the

following components, referred to the principal axes of r, of the corre-
sponding stationary values of jSj,
pffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffi 1
n1 ¼ 0, n2 ¼ 1=2, n3 ¼ 1=2, jSj ¼ jr2 r3 j,
2
n1 ¼ 1=2, n2 ¼ 0, n3 ¼ 1=2, jSj ¼ jr3 r1 j, ð3:22Þ
2
n1 ¼ 1=2, n2 ¼ 1=2, n3 ¼ 0, jSj ¼ jr1 r2 j:
2
It follows from equation (3.22) that the maximum value of jSj is given by
1
jSjmax ¼ ðrmax rmin Þ, ð3:23Þ
2
where rmax and rmin are the algebraically greatest and least principal
stresses, respectively. The result (3.23) has an important application in
the theory of plasticity since the Tresca yield condition states that the onset
of plastic deformation of a metal occurs when jSjmax reaches a certain
critical value, which is equal to the yield point in simple shear.
3.5 Pure Shear Stress and Decomposition of the

Stress Tensor
A state of pure shear stress may be defined as follows. If a rectangular

Cartesian coordinate system exists such that the normal components of
stress at a point are zero, that is, r11 ¼ r22 ¼ r33 ¼ 0, the state of stress at
that point is a state of pure shear. The stress tensor corresponding to a state
of pure shear may be expressed in the form
r ¼ r12 ðe1 e2 þ e2 e1 Þ þ r23 ðe2 e3 þ e3 e2 Þ

ð3:24Þ
þ r31 ðe3 e1 þ e1 e3 Þ,
where the ei are the unit base vectors of a coordinate system for which
r11 ¼ r22 ¼ r33 ¼ 0. This coordinate system is not unique since the form
(3.24) is valid for an infinite number of coordinate systems. The proof of
this is given by Pearson [1]. It is clear that rii ¼ trr is a necessary condition
for a state of pure shear stress and the proof of this is also a sufficient
condition is also given by Pearson [1].
We now introduce the decomposition of the stress tensor into its iso-
tropic or hydrostatic part and deviatoric parts
rkk
rij ¼ sij þ dij ,
3
r ¼ s þ trr , ð3:25Þ
I,
3
trr
where I and s are the isotropic and deviatoric parts, respectively, of the
3
stress tensor. It is easily verified that sii ¼ trs ¼ 0 and that the basic invari-
ants of the tensor s are
1
J1 ¼ 0, J2 ¼ sij sij , J3 ¼ det sij ,
2
Consequently, the deviatoric part of r represents a state of pure shear. The

isotropic part corresponds to a pure hydrostatic stress with the normal
stress the same for all plane elements at a point and no shearing stress.
The decomposition equation (3.25) is very important in plasticity theory
since one interpretation of the Mises yield condition is that yielding of
a metal occurs when jJ2 j reaches a critical value.
A special case of a deviatoric stress is given by
r ¼ se1 e2 ,
as indicated in Figure 3.4.

This stress state is often described as pure shear; however, we prefer to
regard it as a special case of pure shear and call it simple shear.
3.6 Octahedral Shearing Stress
First we consider a rectangular Cartesian coordinate system with the origin

at a point P in a material body and with its axes coinciding with the prin-
cipal axes of the stress tensor r at P. A plane, which makes equal intercepts
96 Stress
x2
τ
τ Figure 3.4. Simple shear stress.

τ
x1
τ
with these three coordinate axes, is known as an octahedral plane. An

octahedral plane, which intercepts the positive directions of the axes, is
shown in Figure 3.5.
The normals to the eight octahedral planes are given by
1
n ¼ pffiffiffi ð6e1 6e2 6e3 Þ: ð3:26Þ
3
Now imagine an octahedron enclosed by the eight elemental octahedral

planes and shrunk to point P and consider eight plane elements whose
normals are given by equation (3.26). Since the axes coincide with the
principal axes of stress, it follows from equations (3.13) and (3.26) that
the same normal stress
1 I1
Noct ¼ ðr1 þ r2 þ r3 Þ ¼ ,
3 3
that is, the isotropic part of the stress tensor, acts on each octahedral plane.
The shearing stress jSoct j acting on a plane element with a normal given by
equation (3.26) is known as an octahedral shearing stress, and it follows from
equations (3.21) and (3.26) that it is the same for all eight octahedral planes
and is given by
1 2 1
S2oct ¼ r1 þ r22 þ r23 ðr1 þ r2 þ r3 Þ2 , ð3:27Þ
3 9
which can be put in the form
2 1n o
Soct ¼ ðr1 r2 Þ2 þðr2 r3 Þ2 þðr3 r1 Þ2 : ð3:28Þ
9
e3
e1 e2
Figure 3.5. Octahedral plane.
In equation (3.28) the principal components of r can be replaced by the

principal components of s so that equation (3.27) becomes
1 2
S2oct ¼ s1 þ s22 þ s23 ,
3
which referred to an arbitrary coordinate system gives
1 2
S2oct ¼ sij sij ¼ J2 : ð3:29Þ
3 3
Equation (3.29) has an important application in the theory of plasticity [2].

The von Mises yield criterion implies that yielding of a metal is not influ-
enced by the isotropic part of the stress tensor, and yielding occurs when
jJ2 j reaches a critical value. This criterion can be expressed in the alterna-
tive forms [2],
2
sij sij ¼ 2k ,
ðr1 r2 Þ2 þðr2 r3 Þ2 þðr3 r1 Þ2 ¼ 6k2 ,

ðr11 r22 Þ2 þðr22 r33 Þ2 þðr33 r11 Þ2 þ6 r212 þ r223 þ r231 ¼ 6k2 ,
98 Stress
where k is the yield stress in simple shear. The reader should show that
pffiffiffi
Y ¼ 3k, where Y is the yield stress in simple tension.
The Tresca yield criterion states that yielding of a metal occurs when the
maximum shearing stress reaches a critical value, that is, when
ðrmax rmin Þ ¼ 2k,
where rmax and rmin are the algebraically greatest and least principal
stresses.
3.7 Nominal Stress Tensor
In order to introduce the concept of the nominal stress tensor we first

consider simple tension of a straight bar of uniform cross section. The
Cauchy or true tensile stress is equal to the tensile force divided by the
current cross-sectional area of the bar. However, when the stress is taken
as the force divided by the cross section of the bar in its undeformed
natural reference configuration, it is said to be the nominal stress. This
can be generalized and made more precise by introducing the nominal
stress tensor R with components RKi , where RKi dA is the ith component of
force currently acting on the plane element, which, in the undeformed
natural reference configuration, is normal to the K th axis and is of infin-
itesimal area dA. The vector form of the element dA is dA ¼ NdA, where
N ¼ NK Ek is the unit normal to dA and the EK are the unit base vectors.
Unit normal N and area dA in the reference configuration become unit
normal n and area da in the current configuration. The force acting on the
element dA is denoted by T ðN ÞdA, where T ðN Þ is the nominal stress
vector acting on dA and is given by
T
Ti ¼ RKi NK , T ¼R N: ð3:30Þ
Equation (3.30) may be obtained in a similar manner to equation (3.10)

with the tetrahedron in Figure 3.3 taken in the reference configuration.
This proof is left as an exercise for the reader. It may be deduced
from equation (3.30) that R is related to both the reference configuration
and the spatial configuration and it is called a second-order two-point
3.7 Nominal Stress Tensor 99
tensor. The transpose of R is known as the first Piola-Kirchhoff stress

tensor.
The relation between the Cauchy stress and the nominal stress R is
obtained as follows. It follows from equation (3.10) and
ti da ¼ Ti dA, tda ¼ TdA
that
rij nj da ¼ RKi NK dA,
and substituting the result (exercise 2, chapter 2)
q0 @XK
ni da ¼ NK dA,
q @xi
where q0 and q are the densities in the reference and spatial configurations,
respectively, gives
q0 @XK q0 1
RKi ¼ rji , R¼ F r: ð3:31Þ
q @xj q
The nominal stress tensor may be expressed in terms of its Cartesian com-
ponents as
R ¼ RKi EK ei :
The form of equation (3.3) in terms of integrals over the reference volume
V and surface @V is
Z Z Z
@v0i ðXK , tÞ
Ti dA þ q0 f0i dV ¼ q0 dV,
@t
@V V V
Z Z Z ð3:32Þ
@vo ðX, tÞ
TdA þ q0 f 0 dV ¼ q0 dV,
@t
@V V V
where T ¼ RT N, f 0 ðX, tÞ ¼ f ðxðX, tÞ, tÞ and vo ¼ vðxðX, tÞ, tÞ. Substituting

equation (3.30) in equation (3.32), applying the divergence theorem, and
100 Stress
noting that equation (3.32) applies to any part of the body gives the equa-
tion of motion:
@RKi @v0i ðX, tÞ

ðX, tÞ þ q0 ðX Þf0i ðX, tÞ ¼ q0 ,
@XK @t
ð3:33Þ
@v0 ðX, tÞ
DivRT ðX, tÞ þ q0 ðX Þf0 ðX, tÞ ¼ q0 ðX Þ ,
@t
where the notation Div is used to denote the divergence operation based on
the operator,
@ð Þ
=X [ EK ,
@XK
but each term of equation (3.33) is a vector referred to the spatial

configuration.
3.8 Second Piola-Kirchhoff Stress Tensor
The second Piola-Kirchhoff stress tensor,
S ¼ SKL EK EL ,
is a symmetric second-order tensor, based on the reference configuration. It

is related to the Cauchy and nominal stress tensors as follows:
@XK @XL T
SKL ¼ J rij , S ¼ JF 1 rF 1 ð3:34Þ
@xi @xj
@XK T
SKL ¼ RLi , S ¼ F 1 RT ¼ RF 1 : ð3:35Þ
@xi
The tensor S unlike the tensor R is symmetric, and this may be advanta-
geous in some applications, but it cannot be directly related to the stress
vector acting on a plane element. It is a particular case of the class of
Lagrangian tensors.
3.9 Other Stress Tensors 101
3.9 Other Stress Tensors
The following stress tensor is obtained from the nominal stress tensor as
follows:
^
S ¼ RR ¼ RKi RiL EK EL , ð3:36Þ
and its symmetric part
1 1
SB ¼ RR þ RT RT , SB
KL ¼ ðRKi RiL þ RiK RLi Þ ð3:37Þ
2 2
is known as the Biot stress tensor. Equation (3.36) gives rise to the polar
^ T , which is analogous to the polar decomposition of
decomposition R ¼SR
the deformation gradient tensor F. However, ^
S is not unique since the
deformation due to a nominal stress R may not be unique [3]. Using equa-
tion (3.35) and the polar decomposition F ¼ RU, equation (3.37) can be
put in the form
1 1
S B ¼ ðSU þ US Þ, SB
KL ¼ ðSKL ULJ þ UKL SLJ Þ: ð3:38Þ
2 2
The Biot stress tensor is useful when certain problems of isotropic elasticity
are considered. For isotropic elasticity the tensors S and U are coaxial so
that the product is commutative. It then follows that
S B ¼ US, ð3:39Þ
which indicates that S B is coaxial with U and S.

The relations R ¼ SF T and RT ¼ FS, obtained from equation (3.35),
and F T F ¼ U 2 give RRT ¼ SU 2 S. It follows from this result, equation
(3.39), and the symmetry of S and U that
pffiffiffiffiffiffiffiffiffiffi
SB ¼ RRT : ð3:40Þ
102 Stress
It should be noted that equations (3.39) and (3.40) are valid only
for isotropic elastic media. It may be shown that RRT is positive definite
or positive semidefinite if R is nonsingular or singular, respectively. If
RRT is positive definite its principal values zð1Þ , zð2Þ , and zð3Þ are positive.
If RRT is positive semidefinite, its principal values are non-negative.
pffiffiffiffiffiffiffiffiffiffi
The principal values of S B ¼ RRT are the positive or negative
square roots of zð1Þ , zð2Þ , and zð3Þ . It is evident that a given R, does not,
in general, determine a unique S B , and this is discussed in detail by
Ogden [3].

It is easily shown that det½R ¼ det SB and that RKi diL ¼ SKL if R ¼ I.
EXERCISES
3.1. Given the state of true stress with the following components
2 3
3 1 3
4 1 2 2 5
3 2 5
Find
(a) the principal Cauchy stresses,
(b) the octahedral shearing stress,
(c) the direction cosines of the normals to the octahedral planes,
(d) the normal stress, shearing stress, and direction of the shearing
stress acting on the plane with unit normal
1 1 1
n¼ e1 þ e2 þ pffiffiffi e3 :
2 2 2
3.2. Show that if the Cauchy stress tensor is singular, that is, det½rij ¼ 0,
one of the principal stresses is zero.
3.3. Prove that if the Cauchy stress tensor r and left Cauchy-Green strain
tensor B are coaxial, the second Piola-Kirchhoff stress tensor S and
the right Cauchy-Green strain tensor C are also coaxial.
3.4. Consider the following state of simple shear
x1 ¼ X1 þ KX2 , x2 ¼ X2 , x3 ¼ X3 :
The components of the Cauchy stress tensor required to produce this

deformation are
2 3
p a 0
4 a p1 0 5:
0 0 p2
Find the matrix of components of the first and second Piola-Kirchhoff

stress tensors.
3.5. Prove that if a body of volume V and surface A is in equilibrium under
surface traction t ðxÞ, x 2 A where x is the position vector of a material
point with respect to a fixed origin the mean hydrostatic part of the
stress tensor, rm , is given by,
Z
1
rm ¼ t Xda:
3V
@V
3.6. Let m and n be unit vectors normal to surface elements Dam and Dan
at a point, and let t ðmÞ and t ðnÞ be the stress vectors acting on these
plane elements. Show that
n tðmÞ ¼ m t ðnÞ:
3.7. Consider a state of plane stress with r33 ¼ r32 ¼ r31 ¼ 0. Let
Nð1Þ , Nð2Þ , and Nð3Þ
be the normal stress on plane elements that are normal to the x3 plane
and form angles of 120° with each other. Find the principal stress and
the matrix of components of the stress tensor if the normal stress N(1)
is in the x1 direction and the direction of Nð2Þ is in the second quadrant
of the 0x1 x2 plane.
104 Stress
xi , i 2 f1, 2, 3g are obtained from Oxi

3.8. If rectangular Cartesian axes O^
by rotating about Ox3 through angle a, using the right-hand rule, show
that
r^11 þ r^22 ¼ r11 þ r22 ,
r12 ¼ e2ia ðr11 r22 þ 2ir12 Þ,

r^11 r^22 þ 2i^
r23 ¼ eia ðr13 þ ir23 Þ,

r^13 þ i^
where rîj and rij are the components of the Cauchy stress with respect
xi and Oxi , respectively.
to O^
3.9. A body of current volume V is in equilibrium under the action of
surface traction rn and body force f. Show that the mean stress
R
r ¼ V1 rdV is given by
V
8 9
Z Z
1< T =
r ¼ r n x da þ f xdv :
v: ;
@v v
3.10. Prove that
SC_ ¼ 2RF:
_
REFERENCES
1. Pearson, C.E. (1959). Theoretical Elasticity. Harvard University Press.

2. Hill, R. (1950). The Mathematical Theory of Plasticity. Oxford University
Press.
3. Ogden, R.W. (1984). Non-Linear Elastic Deformations. Ellis Harwood.
4 Work, Energy, and Entropy
Considerations
This chapter is concerned with some aspects of work, energy, and contiuum
thermodynamics. The treatment is general and reference is not made to any
particular material. Notation for deformation and stress variables is the
same as in chapters 2 and 3, where the lower- and uppercase suffixes refer
to the spatial and reference configurations, respectively. Most of the equa-
tions in this chapter are given in both suffix and symbolic notation.
4.1 Stress Power
Stress power is the rate of work done by the stresses and is a purely mech-
anical quantity, that is, obtained from the rate of work of the body forces
and surface tractions acting on a material body and does not involve any
other forms of energy transfer such as heating. It is a specific quantity per
unit volume or per unit mass and can be expressed in terms of either the
Cauchy, nominal, or second Piola-Kirchhoff stress and the respective con-
jugate rate variable.
The rate of work of the body forces and surface tractions acting on
a material body of volume v and surface a in the spatial configuration is
given by
Z Z
W_ ¼ ti vi da þ qfi vi dv,
@v v
Z Z ð4:1Þ
W_ ¼ t vda þ qf vdv,
@v v
where q is the density in the spatial configuration and the vectors t, v,

and f are the surface traction, velocity, and body force per unit mass,
105
106 Work, Energy, and Entropy Considerations
respectively, and are functions of x and t. Substituting equation

(3.13),
ti ¼ rij nj , t ¼ sn, ð4:2Þ
where r is the Cauchy stress tensor and n is the outward unit normal to the
surface, in equation (4.1) and applying the divergence theorem gives
Z
@rij
W_ ¼ vi þ qfi þ rij Lij dv
@xj
Zv ð4:3Þ
W_ ¼ fv ½divs þ qf þ trðsLÞgdv
v

where Lij ¼ @vi @xj are the components of the velocity gradient tensor
= v. With the use of the equation of motion
@rij dvi
þ qfi ¼ q ,
@xj dt
ð4:4Þ
dv
divs þ qf ¼ q ,
dt
equation (4.3) becomes
Z Z
d vi vi
W_ ¼ q dv þ rij Lij dv,
dt 2
v v
Z Z ð4:5Þ
d vv
W_ ¼ q dv þ trðsLÞdv:
dt 2
v v
The first term on the right-hand side of equation (4.5) is the rate of change
of the kinetic energy of the body, and the second term is integral over the
volume of the specific stress power per unit volume of the spatial configu-
ration of the body. It follows that the specific stress power per unit mass is
1 1
P ¼ rij Lij ¼ trðsLÞ: ð4:6Þ
q q
Decomposition of the velocity gradient tensor into symmetric and antisym-

metric parts is given by
Lij ¼ dij þ xij , L ¼ D þ W, ð4:7Þ
where D ¼ ðL þ LT Þ=2 and W ¼ ðL LT Þ=2 are the rate of deformation

and spin tensors, respectively. Substitution of equation (4.7) in equation
(4.6) gives
1 1
P ¼ rij Dij ¼ trðsDÞ, ð4:8Þ
q q
since xij rij ¼ trðW rÞ ¼ 0.

The specific stress power may also be expressed in terms of the first
and second Piola-Kirchhoff stresses and the respective conjugate rate vari-
ables. In this treatise, the nominal stress tensor R, which is the transpose of
the first Piola-Kirchhoff stress, is used. Component RKj , of R, as noted in
chapter 3, is the jth component of the stress vector acting on the plane
element, referred to the reference configuration and whose normal is in
the K th direction. In order to express P in terms of the nominal stress
tensor, R, equation (4.1) is replaced by
Z Z
W_ ¼ T i m0i dA þ q0 f0i m0i dV,
@V V
Z Z ð4:9Þ
W_ ¼ T m 0 dA þ q0 f 0 m 0 dV,
@V V
where
ðq0 , v0 , f 0 Þ ¼ ðq, v, f ÞðxðX, tÞ, tÞ
and
T i ¼ RKi NK , T ¼ RT N ð4:10Þ
is the nominal surface traction and q0 , V, A, and N denote the density,

volume, surface, and outward unit normal to the surface, respectively, of
a material body in the reference configuration. In equation (4.9) the inte-
grands T v0 and q0 f 0 v0 are the rate of work of the surface tractions per
unit area of the surface of the reference configuration and the rate of work
of the body force per unit volume of the reference configuration, respec-
tively. Substituting equation (4.10) in equation (4.9) and applying the diver-
gence theorem gives
Z
@RKi
W_ ¼ vi _
þ q0 f0i þ RKi FiK dV,
@XK
ZV n n o ð4:11Þ
T o
W_ ¼ v: DivR þ q0 f 0 þ tr RF_ dV,
V
where FiK ¼ @xi =@XK , are the components of the deformation gradient
tensor and F_iK ¼ Lij FjK , F_ ¼ LF. With the use of the equations of motion
@RKi @v0i
þ q0 f0i ¼ q0 ,
@XK @t
@m 0
DivRT þ q0 f 0 ¼ q0 , ð4:12Þ
@t
equation (4.11) becomes
Z Z
@ v0i v0i
W_ ¼ q0 dV þ RKi F_iK dV,
@t 2
V V
Z Z ð4:13Þ
@ v0 v0
W_ ¼ q0 dV þ tr RF_ dV:
@t 2
V V
It follows from equation (4.13) that the specific stress power (per unit mass)
is given by
1 1
P¼ _
ðRKi F_iK Þ ¼ trðRFÞ: ð4:14Þ
q0 q0
_
The stress power in terms of the second Piola-Kirchhoff stress S and E,
the Lagrangian strain rate, cannot be determined directly in the same
manner as equations (4.8) and (4.14) but can be determined from equation
(4.8) or (4.14). Substitution of
q @xi @xj q T
rij ¼ SKL ; s¼ FSF ,
q0 @XK @XL q0
obtained from equation (3.34), and
@XM @XN
dij ¼ E_ MN ; D ¼ F T EF
_ 1 ,
@xi @xj
obtaind from equation (2.29), in equation (4.8) gives
1 1 1 _
P¼ SKL E_ KL ¼ tr S E_ ¼ tr S C : ð4:15Þ
q0 q0 2q0
An alternative determination of equation (4.15) is as follows. Substitution of
RKi ¼ SKL FiL , R ¼ SF T
in equation (4.14) gives
1 T _
P¼ tr SF F : ð4:16Þ
q0
.
E_ KL ¼ ðF_iK FiL þ FiK F_iL Þ 2, E_ ¼ ðF_ T FþF T FÞ
_ 2: ð4:17Þ
Substitution of equation (4.17) in equation (4.16) gives equation (4.15)

since

tr SF_ T F ¼ tr S F_ T F þ F T F_ ¼ tr S E_ :
2
4.2 Principle of Virtual Work
The principle of virtual work applied to static problems involves the spatial
form of the equations of equilibrium,
@rij
þ qfi ¼ 0,
dxj ð4:18Þ
divs þ qf ¼ 0,
or the referential form,
@RKi
þ q0 f0i ¼ 0,
@XK ð4:19Þ
DivRT þ q0 f0 ¼ 0,
rather than the equations of motion.

In order to introduce the principle of virtual work, a boundary value
problem is posed and a statically admissible stress field and a kinematically
admissible virtual displacement field are defined. A statically admissible
stress field is a stress field that satisfies the equilibrium equations and the
stress boundary conditions. Stress discontinuities occur in certain static
quasi-static problems, for example, at the contact surface of two compo-
nent cylinders of a compound cylinder with shrink fit. Equilibrium requires
that the normal component, ti ni ¼ t n, of the stress vector acting on the
discontinuity surface must be continuous, where n is the unit normal on one
side of the discontinuity surface.
A kinematically admissible virtual incremental displacement field,
duðxÞ, is an imagined displacement that is assumed to be infinitessimal,
continuous, and vanishes on the part of the boundary @vu upon which the
displacements are prescribed. It should be emphasized that the stress field
remains unchanged during the application of duðxÞ. Since it is assumed that
duðxÞ is infinitesimal, it is consistent with the concept of virtual displace-
ments in classical mechanics. The distinction between duðxÞ and duðxÞ, that
is, an actual increment of displacement, is worth noting. An actual
infinitesimal incremental displacement field could be regarded as a virtual
incremental displacement field, but the converse is not necessarily true.
First we consider a boundary value problem referred to the spatial

configuration. Then, the body force field acting on a body could be pre-
scribed along with the surface tractions on part @vt of the surface and the
displacements on the remainder @vu of the surface. It is also possible that
the normal component of stress and the tangential component of displace-
ment or vice versa could be prescribed on a part of the surface. Malvern [1]
has considered further combinations of traction and displacement compo-
nents prescribed on part of the surface. It is easy to modify the theory to
consider these cases; however, we will assume the tractions and displace-
ments are prescribed on disjoint parts of the surface so that @vt [ @vu ¼ @v
and @vt \ @vu ¼ 0 where the symbols [ and \ denote the union of the two
surfaces and the intersection of the two surfaces, respectively.
The virtual work of the surface tractions and body force, referred to
the spatial configuration for static problems, is
Z Z
dW ¼ ti dui da þ qfi dui dv, ð4:20Þ
@vs v
where du ¼ 0 on the part @vu of the surface. With the use of equation (3.10)
and the divergence theorem (4.20) becomes
Z
@ rij duj
dW ¼ þ qfi dui dv
@xi
v
Z
@rij @duj
¼ duj þ rij þ qfi dui dv: ð4:21Þ
@xi @xi
v
Then, with the use of the equation of equilibrium (4.18), (4.20), (4.21), and

rij @dui @xj ¼ rij deij ;
Z Z Z
ti dui da þ qfi dui dv ¼ rij deii dv,
@vs v v
Z Z Z ð4:22Þ
t dui da þ qf dudv ¼ trðsdeÞdv,
@vs v v
where

1 @dui @duj 1
deij ¼ þ , de ¼ = du þ ð= duÞT , ð4:23Þ

2 @xj @xi 2
which is a virtual infinitesimal strain increment, based on the current con-

figuration. Equation (4.22) is a form of the principle of virtual work and
indicates that, for any admissible virtual displacement field, the virtual work
(4.20) of the external forces is equal to the virtual work of the internal
forces. It is usually assumed that the d variation and the gradient operator
= commute, and with this assumption equation (4.23) is equivalent to
de ¼ d ð= uÞ þ ð= uÞT , ð4:24Þ
2
which is a virtual increment of the infinitessimal strain tensor. A difficulty

arises when a problem involving finite deformation is considered, then
equation (4.23) is appropriate but not equation (4.24).
The form of the principle of virtual work, for a statically admissible
stress field in terms of the nominal stress R and a virtual displacement field,
_
duðX Þ ¼ duðxðXÞÞ is
Z Z Z
_ _
Ti dui dA þ q0 f0i dui dV ¼ RKi dFiK dV,
@VT V V
Z Z Z ð4:25Þ
_ _
T dudA þ q0 f 0 dudV ¼ trðRdF ÞdV,
@VT V V
referred to the reference configuration, where V is the volume in the ref-

erence configuration and @VT is the part of the reference surface @V upon
which the surface traction given by equation (4.10) is prescribed. The proof
of equation (4.25) is similar to that for equation (4.22) with Ti eliminated
using equation (4.10) followed by use of the divergence theorem and
equation (4.19).
It is evident from equations (4.22) and (4.25) that, for a static problem,
the virtual work of the applied forces, that is, the surface tractions and body
forces, is equal to virtual work of the internal forces, that is, the virtual work
of the internal stresses.
Another form of the principle of virtual work for static or quasi-
static processes arises when we consider a statically admissible stress field
rðxÞ along with a kinematically admissible velocity field vðxÞ in a region v
with surface @v of a material body. A kinematically admissible velocity
field satisfies the velocity boundary conditions on the part of the bound-
ary upon which the velocity is prescribed and any kinematical constraints
such as incompressibility. We then obtain the result for quasi-static
processes
Z Z Z
ti vi da þ qvi fi dv ¼ rij dij dv,
dv v v
Zt Z Z ð4:26Þ
t:vda þ qv f dv ¼ trðsDÞdv,
dvt v v
where D is the rate of deformation tensor obtained from the kinematically

admissible velocity field vðxÞ. It is important to note that there need be no
cause and effect relationship between the statically admissible stress field
and the kinematically admissible velocity field in equation (4.26). In some
applications it is convenient to consider velocities and rates of deformation
with respect to a monotonically varying parameter of the problem rather
than time. This is often the case for quasi-static problems in the theory of
plasticity. In the theory of rigid perfectly plastic solids volume preserving
discontinuities may occur in kinematically admissible velocity fields for
certain problems.
The principle of virtual work may also be applied to dynamic problems
and, referred to the spatial configuration, is then given by
Z Z Z Z
ti dui da þ qfi dui dv qv_ i dui dv ¼ rij deij dv,
@vt v v v
Z Z Z Z ð4:27Þ
t duda þ qf dudv qv_ dudv ¼ trðsdeÞdv,
@vt v v v
where the virtual displacement field duðxÞ is continuous and vanishes on

the part of the boundary upon which the velocity is prescribed and the
deformation of the material body is imagined to be frozen when the virtual
displacement field is applied. Equation (4.27) shows that the virtual work of
the applied forces and inertia forces is equal to the virtual work of the
internal stresses. The derivation of equation (4.27), which is the continuum
form of d’Alembert’s Principle of classical mechanics, is similar to the
derivation of equation (4.20) with the equilibrium equation (4.18) replaced
by the equation of motion (4.4). As an exercise show that the reference
configuration form of equation (4.27) is
Z Z Z Z
_ _ _
Ti dui dA þ q0 f0i dui dV q0 v_ 0i dui dV ¼ RKi dFiK dV,
@V V V V
Z Z Z Z
_ _ _
T dudA þ q0 f 0 dudV q0 v_ 0 dudV ¼ trðRdF ÞdV:
@V V V V
A detailed discussion of the continuum form of d’Alembert’s Principle of

classical mechanics is given by Haupt [2].
4.3 Energy Equation and Entropy Inequality
In this section we consider some aspects of the thermodynamics of defor-

mation. Continuum mechanics involves the study of nonequilibrium states
of continuous media as well as equilibrium states; consequently, in order to
introduce thermodynamical concepts, an extension of classical equilibrium
thermodynamics, more accurately described as thermostatics, is required.
A simple extension of thermostatics to cover irreversible processes is based
on the Principle of Local State [3], which is a plausible hypothesis. This
principle is as follows:
The local and instantaneous relations between the thermodynamic

properties of a continuous system are the same as for a uniform system
in equilibrium.
A treatment of thermodynamics of deformation based on this principle is

referred to as the thermodynamics of irreversible processes [4]. Essentially
what is implied is that infinitesimal elements of a system in a non-equilibrium

state behave locally as if undergoing reversible processes. This gives satisfac-
tory results for most engineering problems.
The energy equation is a postulate that follows from the first law of
thermodynamics (conservation of energy) for a closed system (material
volume). The rate of work of the surface tractions and body forces plus
the rate of heat flow in to the body is equal to the rate of change of the sum
of the kinetic energy and internal energy. Gravitational potential energy is
neglected since it is not significant in most contiuum mechanics problems,
but it could easily be included in the equations that follow. The spatial form
of the energy equation for a material volume, vðtÞ is
0 1
Z Z Z
d@ vi vi A
qð þ uÞ dv ¼ qðr þ fi vi Þdv þ ðti vi qi ni Þda,
dt 2
v v @v
0 1 ð4:28Þ
Z Z Z
d@ vv
qð þ uÞAdv ¼ qðr þ f vÞdv þ ðt v q nÞda,
dt 2
v v @v
where n is the outward unit normal to the surface, u ¼ uðx, tÞ is the specific
internal energy, r is rate of heat supply per unit mass, and q is the heat flux
vector. Fourier’s law for the spatial configuration,
@H
qi ¼ kij , q ¼ k=H, ð4:29Þ
@xj
is a semiempirical relation that relates q and the temperature gradient. In

equation (4.29) k is the heat conduction tensor that is a positive semidefinite
second-order tensor for an anisotropic medium. For an isotropic medium
kij ¼ dij k, where k is a non-negative scalar. The thermal conductivity tensor
is a function of temperature, and the deformation and may also be a function
of position for an inhomogeneous body. For many problems that involve
suitably small temperature differences and changes, the temperature depen-
dence may be neglected and k assumed constant. It should be noted that
equation (4.29) has a defect, when applied to transient problems, since it
predicts an infinite thermal wave speed; however, equation (4.29) is adequate
for many engineering applications, and it will be assumed for what follows in
this text. Modifications of equation (4.29) in order to remove the defect have
been extensively researched and [5] is a useful reference on this topic.
Heat production can arise from various external agencies such as
radiation, or from Joule heating in electrical conductors. Potential energy
is neglected in equation (4.28). Substituting for ti , from equation (4.2), in
equation (4.28), and using the divergence theorem and the equation of
motion (4.4) gives
Z
@qi
qu_ qr rij dij þ dv ¼ 0,
@xi
v
Z ð4:30Þ
ðqu_ qr trðsDÞ þ divqÞdv ¼ 0:
v
Since the integrand is assumed to be continuous and equation (4.30) is valid

for volume v or any part of v, it follows that
@qi
qu_ qr rij dij þ ¼ 0,
@xi ð4:31Þ
qu_ qr trðsDÞ þ divq ¼ 0:
Equation (4.31) is the local form of the energy equation referred to the spatial
configuration. The local form of the energy equation can also be referred to
the reference configuration. In order to obtain this form we note that
QK dAK ¼ qi dai , Q dA ¼ q da, ð4:32Þ
where Q is the heat flux vector referred to the reference configuration and
dA and da are the vector areas of a plane element in the reference and
spatial configurations, respectively. Using equation (4.32) along with
@XK T
dai ¼ J dAK , da ¼ JF dA,
@xi
gives
1 1
qi ¼ J FiK QK , q ¼ J FQ: ð4:33Þ
The divergence of equation (4.33) and the result
@ 1 1
ðJ FiK Þ ¼ 0, divðJ FÞ ¼ 0,
@xi
give
@qi @QK
¼ J 1 , divq ¼ J 1 DivQ, ð4:34Þ
@xi @XK
where, as before,
q0
J¼ ¼ det½F:
q
It then follows from equations (4.8), (4.14), (4.26), and (4.29) that the local
form of the energy equation (4.31), referred to the reference configuration, is
@QK
q0 u_ 0 q0 r0 RKi F_iK þ¼ 0,
@XK ð4:35Þ
_ þ DivQ ¼ 0,
q0 u_ 0 q0 r0 trðRFÞ
where u0 ðX, tÞ ¼ uðxðX, tÞ, tÞ and r0 ðX, tÞ ¼ rðxðX, tÞ, tÞ. Equation (4.35)
can also be obtained from the reference form of equation (4.28),
Z Z Z
@ v0i v0i
q0 ðu0 þ ÞdV ¼ q0 ðr0 þ foi v0i ÞdV þ ðTi v0i Qk NK ÞdA:
@t 2
V V @V
In addition to the energy equation we postulate an entropy production

relation, that follows from the second law of thermodynamics,
Z Z Z
d qr qi ni
qsdv dv þ da > 0,
dt H H
v v @v
Z Z Z ð4:36Þ
d qr qn
qsdv dv þ da > 0,
dt H H
v v @v
referred to the spatial configuration, where s ¼ sðx, tÞ is the specific entropy

and H is the absolute temperature. Equation (4.36) is known as the Clausius-
Duhem inequality. If q ¼ 0 on @v and r ¼ 0, the material volume in equation
(4.36) is thermally isolated, and
Z Z
d
qsdv ¼ _ > 0,
qsdv
dt
v v
so that the entropy is nondecreasing. Since equation (4.36) is valid for v or

any part of v, and the integrand is assumed continuous, it follows, with the
use of the divergence theorem, that
@qi qi @H
qHs_ qr þ > 0,
@xi H @xi
q
qHs_ qr þ divq :=H > 0,
H
and if we substitute for r from equation (4.31) or subtract equation (4.31)

we obtain
qi @H
_ þ rij dij
qðHs_ uÞ > 0,
H @xi ð4:37Þ
q
qðHs_ uÞ
_ þ trðsDÞ :=H > 0:
H
Equation (4.37) is the local form of the dissipation inequality, or Clausius

Duhem inequality, referred to the spatial configuration. As an exercise
show that the local form of the dissipation inequality referred to the refer-
ence configuration is
_
_ __
__ @vi QK @H
q0 ðH s uÞ þ RKi _ > 0,
@XK H @XK ð4:38Þ
_ __ _ Q =X H > 0,
__ _
q0 ðH s uÞ þ trðRFÞ _
H
_
where s ðX, tÞ ¼ sðxðX, tÞ, tÞ, HðX, tÞ ¼ HðxðX, tÞ, tÞ. In equation (4.38) RF_
_
_
can be replaced by SE_ or S C=2.
It is often convenient to introduce the specific Helmholtz free energy

function, f ¼ u Ts, and substituting for u in equation (4.37) gives
qi @H
_ þ f_Þ þ rij dij
qðsH > 0,
H @xi ð4:39Þ
_ þ f_Þ þ trðrDÞ q :=H > 0:
qðsH
H
The equality signs in equations (4.37) to (4.39) hold only for an ideal thermo-
dynamically reversible process. The left-hand sides of equations (4.37) and
(4.38) are equal to q0 Hc_ and qHc,
_ respectively, where c>0
_ is the rate of
entropy production per unit mass. It is interesting to note that the rate of
entropy production per unit mass due to heat conduction alone is given by
qi @H q
c_ ðhcÞ ¼ ¼ :=H, ð4:40Þ
qH2 @xi qH2
referred to the spatial configuration and

_
QK @H Q _
c_ hc _ 2 ¼ _ 2 =X H , ð4:41Þ
@XK
q0 H q0 H
referred to the reference configuration. In the absence of shock waves, heat

conduction is the only source of entropy production in a perfectly elastic
solid. This is discussed in later chapters.
EXERCISES
4.1. Obtain the spatial form of the stress power in terms of the deviatoric
and isotropic parts of the Cauchy stress tensor s and the rate of de-
formation tensor D.
4.2. Show that for an isotropic linear elastic solid the specific stress power is
equal to the rate of change of specific strain energy.
4.3. Obtain the following virtual relations,
1 1
dF ¼ ð= duÞF and dF ¼ F ð= duÞ:
4.4. Use the results of problem 3 to show that the virtual

.variation of Almansi’s
strain tensor, g, and de ¼ = du þ ð= duÞT 2 are related by
@duk @duk
deij ¼ dgij þ gkj þ gik ,
@xi @xj
T
de ¼ dg þ ð= duÞ h þ hð= duÞ:
4.5. Show that the virtual variation of the Lagrangian strain tensor is given
by
1n T o
dE ¼ F ð=X duÞ þ ð=X duÞF :
2
4.6. Express the referential form of the Clausius Duhem inequality in

terms of the second Piola-Kirchhoff stress tensor and the right
Cauchy-Green strain tensor.
4.7. Justify the Clausius-Planck inequality
qðHs_ uÞ
_ þ trðsDÞ > 0
and discuss its significance for an isentropic process.

4.8. Discuss the significance, for a reversible isothermal process, of the
Clausius-Planck inequality in the equivalent form
_ þ f_Þ þ trðsDÞ > 0:

qðsH
REFERENCES
1. Malvern, L.E. (1969). Introduction to the Mechanics of a Continuous

Medium. Prentice-Hall, Inc.
2. Haupt, P. (2000). Continuum Mechanics and Theory of Materials. Springer-
Verlag.
3. Kestin, J.A. (1966). Course in Thermodynamics, Vol. I. Blaisdell.
4. Muller, I., and Ruggiero, T. (1993). Extended Rational Thermodynamics,
2nd Ed. Springer.
5. Chandrasekaraiah, D.S. (1986). Thermoelasticity with Second Sound. App.
Mech. Rev. 39, pp. 355–376.
5 Material Models and
Constitutive Equations
5.1 Introduction
The response of material bodies to applied forces and heating has not been
considered in previous chapters, except that it is assumed that the materials are
nonpolar. In order to obtain relations between stress, deformation, and tem-
perature fields, it is necessary to have constitutive equations. These equations
give the response of idealized models of real materials, to the application of
forces and heating. In this book electromagnetic effects are not considered.
Admissible constitutive equations must not violate the laws of thermo-
dynamics and should satisfy certain additional principles. In this chapter some
classical constitutive equations and other relations for models of actual mate-
rials are considered, before discussing these additional principles in detail.
As in previous chapters, certain relations are presented in both suffix
and symbolic notation; however, relations that involve tensors of higher
order than two are given in suffix notation only.
5.2 Rigid Bodies
A rigid motion of a continuum is given by
xi ðtÞ ¼ QiK ðtÞXK þ ci ðtÞ,

ð5:1Þ
xðtÞ ¼ QðtÞX þ cðtÞ,
where Q is a two-point proper orthogonal second-order tensor and c has

the dimension of length. It follows from equation (5.1) that
T
FiK ¼ QiK , det½F ¼ 1, C ¼ F F ¼ I and D ¼ 0:
121
122 Material Models and Constitutive Equations
A rigid body may be defined as a body for which the distance between any
two particles of the body remains invariant, and the motion given by equa-
tion (5.1) satisfies this definition. To show this, take any two particles of the
body with position vectors X 1 and X 2 and corresponding position vectors
x1 and x2 after a rigid body motion that involves translation and rotation. It
then follows from equation (5.1) that
ðx1 x2 Þ ðx1 x2 Þ ¼ QðX 1 X 2 Þ QðX 1 X 2 Þ
¼ ðX 1 X 2 Þ ðX 1 X 2 Þ,
since QQT ¼ QT Q ¼ I. Consequently, the distance between any two par-

ticles is invariant for a rigid motion. The rigid body model is a useful
idealization for certain real materials in applications where any deforma-
tion is negligible for the problem considered.
A necessary and sufficient condition for a rigid motion is
dij [ 0, D [ 0, ð5:2Þ
which implies that the stress power rij dij is zero. This is Killing’s theorem,
which has already been discussed in chapter 2. Equation (5.2) is a constitu-
tive equation for a rigid body, and it follows that dii ¼ 0 is a necessary but
not sufficient condition for a body to be rigid but is a necessary and suffi-
cient condition for an incompressible fluid. The stresses in a rigid body are
undetermined, the response to the application of forces is governed by
equation (5.1) and the laws of classical rigid body mechanics. Mechanical
and thermal effects are uncoupled. In addition to equation (5.2), there is
a thermal equation of state for a rigid body,
ZH
uðHÞ ¼ ^ dH,
c H ^ ð5:3Þ
H0
where u, as before, is the specific interna1 energy, H is the absolute tem-

perature, H0 is a reference temperature, and cðHÞ is the specific heat. The
energy equation for a rigid body follows from equation (4.12) and is
@qi
qu_ qr þ ¼ 0,
@xi ð5:4Þ
qu_ qr þ divq ¼ 0,
5.3 Ideal Inviscid Fluid 123
where q is the heat flux vector and r is rate of heat supply per unit mass, for
example, Joule heating due the passage of an electric current. A semi-
empirical law, Fourier’s Law of heat conduction, given by
@H
qi ¼ kij ðHÞ , q ¼ kðHÞgradH ð5:5Þ
@xj
is now introduced. In equation (5.5) k is the heat conduction tensor, is

symmetric and positive or positive semidefinite, and in general is a function
of temperature. The positive or positive semidefinite property is consistent
with the Clausius statement of the second law of thermodynamics.
Substituting equation (5.5) in equation (5.4) gives the spatial form of the
energy equation,
@H @ @H
qcðHÞ qr ðkij Þ ¼ 0,
@t @xi @xj
ð5:6Þ
@H
qcðHÞ qr divðk gradHÞ ¼ 0:
@t
For many problems the temperature range is small enough that the tem-
perature dependence of k and c can be neglected. If in addition the rigid
body is isotropic, k ¼ Ik, and equation (5.6) becomes
@H
qc qr k=2 H ¼ 0,
@t
where
.
=2 [ divgrad [ @ 2 @xi @xi :
5.3 Ideal Inviscid Fluid
An ideal inviscid fluid has a simple constitutive equation, namely,
rij ¼ pdij , r ¼ pI, ð5:7Þ
where p is a hydrostatic pressure. Equation (5.7) arises since an inviscid

fluid, which is a mathematical idealization, can sustain no shearing stresses;
consequently, the Cauchy stress must be a pure hydrostatic stress with no

deviatoric part. A thermal equation of state is required in addition to
equation (5.7), for example, if the fluid is a perfect gas, that is, a gas that
obeys Boyle’s law and Charles law, then
p
¼ RH, ð5:8Þ
q
where R is the gas constant for the particular gas, H is the absolute tem-
perature, p is the absolute pressure, and q is the density. Since the internal
energy of a perfect gas can be expressed as a function of temperature only,
equation (5.3) also holds for a perfect gas if cðHÞ is replaced by cv ðHÞ, the
specific heat at constant volume. The stress power of a compressible in-
viscid fluid is P ¼ pv_ where v ¼1=q.
If we consider a viscous fluid, equation (5.7) holds only for equilibrium
states and is not a constitutive relation.
5.4 Incompressible Inviscid Fluid
This is an inviscid fluid model that can only undergo isochoric flow. The
constitutive equation chosen to model a given material may be influenced
by the problem considered. For example, for some problems involving
a liquid, such as gravity water waves, it may be realistic to assume an in-
compressible inviscid fluid, whereas for other problems, such as water
hammer in a pipe, compressibility effects must be considered, and for
boundary layer problems, viscous effects are dominant.
If a fluid is assumed to be mechanically incompressible the coefficient
of volume thermal expansion must be zero so that only isochoric deforma-
tion is possible. It has been shown by Müller [1] that a fluid model that is
mechanically incompressible cannot have a nonzero coefficient of volume
thermal expansion, otherwise the second law of thermodynamics is
violated. Since an incompressible fluid can undergo only isochoric
deformation,
dii ¼ 0, trD ¼ 0:
5.5 Newtonian Viscous Fluid 125
The stress power for an incompressible inviscid fluid is zero and equation
(5.3) holds.
5.5 Newtonian Viscous Fluid
First we consider the compressible case. The Newtonian viscous fluid

model has the following properties. (1) Stress components rij þ pdij , where
p is the thermodynamic pressure, are linear homogeneous functions of dij .
The thermodynamic pressure is the pressure that appears in the thermo-
dynamic equation of state, for example, for a perfect gas p is given by
equation (5.8). (2) The fluid is isotropic. (3) When the rate of deformation
is zero, the stress is hydrostatic, that is, rij ¼ pdij . It follows from (1) to (3)
that rij þ pdij ¼ Cijkl dkl , where Cijkl are the components of a fourth-order
isotropic tensor. Since the most general fourth-order isotropic tensor has
components given by
Cijkl ¼ adij dkl þ bdik djl þ cdil djk ,

where a, b, and c are scalars, rij þ pdij ¼ adij dkl þ bdik djl þ cdil djk dkl , for
a single phase fluid. Since r and D are symmetric, this can be put in the form
rij þ pdij ¼ 2gdij þ 1dkk dij
ð5:9Þ
r þ pI ¼ 2gD þ 1ItrD,
where g and 1 are viscosity coefficients. For isochoric flow it follows from
the constitutive equation (5.9) that
rij þ pdij ¼ 2gdij , dii ¼ 0,
which indicates that g is the viscosity component for simple shear. For pure
dilatation it follows from equation (5.9) that
rkk
þ p ¼ ndkk , ð5:10Þ
3
where n ¼ 2g=3 þ 1 is the bulk viscosity. Substitution of equation (5.9) in

the equation of motion (3.15),
@rij dvi
þ qfi ¼ q ,
@xj dt
with the use of equation (5.10) gives
2
@p @ 2 vi 1 @ vj dvi
þg þ nþ g þ qfi ¼ q : ð5:11Þ
@xi @xj @xj 3 @xi @xj dt
This equation is the Navier-Stokes equation and is nonlinear since

dvi =dt ¼ @vi =@t þ vj @vi @xj is nonlinear. It should be noted that the
mechanical pressure rkk =3 is not equal to the thermodynamic pressure
p unless it is assumed that n ¼ 0. This assumption is known as Stokes’
hypothesis and is a good approximation for monatomic gases at relatively
low pressures. If n ¼ 0, it may be deduced from equation (5.10) that there
is no dissipation of mechanical work due to viscosity when the fluid
undergoes dilatation with no distortion. According to White [2], the bulk
viscosity property has been a controversial concept.
The viscosity coefficients g and n are transport properties and are, in
general, functions of temperature. For a liquid g decreases and for a gas g
increases as the temperature increases.
An incompressible Newtonian fluid is a useful model for many liquids.
This model has only one viscosity coefficient g and equation (5.9) is
replaced by
rij ¼ 2gdij pdij , ð5:12Þ
with dii ¼ 0, where p ¼ rkk =3 is a Lagrangian multiplier that is obtained

from stress boundary conditions. A more extensive treatment of a New-
tonian viscous fluid is given in [2] and some other aspects are given in
chapter 9.
5.6 Classical Elasticity
An elastic solid is a material for which the stress depends only on the
deformation, when the deformation is isothermal. Classical elasticity is
a linear elastic theory based on the assumption that the displacement gra-
dient tensor is infinitesimal so that the strain and rotation tensors are
linearized and the distinction between the spatial and referential forms
of the governing equations disappears. There is a vast literature on classical

elasticity dating from the mid-nineteenth century onward and only a very
brief discussion of the stress strain relations is given here. The book by
Love [3] is an extensive exposition of linear elasticity and contains refer-
ences up to 1926. Two significant later general treatises on the subject are
Goodier and Timoshenko [4] and Sokolnikov [5], and there are many
books that consider specialized areas. Here we briefly discuss some basic
features of the constitutive relations of classical elasticity for isothermal
deformation. The components of Cauchy stress, rij , are linear functions of
the components of the infinitesimal strain tensor

eij ¼ ð1=2Þ @ui @xj þ @uj @xi :
Since rij and eij are the components of second-order symmetric tensors,
rij ¼ cijkl ekl , ð5:13Þ
where the cijkl are the components of a fourth-order tensor known as the
stiffness tensor, with symmetries.
cijkl ¼ cjikl ¼ cjilk ,
since rij ¼ rji and ekl ¼ elk . Equation (5.13) is sometimes known as the
generalized Hooke’s law. It can be shown that there is a further symmetry,
cijkl ¼ cklij if a strain energy function,
1
wðeÞ ¼ cijkl eij ekl , ð5:14Þ
2
exists, such that
@wðeÞ
rij ¼ : ð5:15Þ
@eij
It then follows that a classical linear elastic solid can have at most twenty-
one independent elastic constants. It should be noted that, in equation
(5.14), the symmetry of the infinitesimal strain tensor e is ignored for the
differentiation (5.15). Sokolnikov [5] has given more extensive discussion
of the symmetries of the stiffness tensor. Equation (5.13) can be inverted to
give
eij ¼ aijkl rkl , ð5:16Þ
where aijkl are the components of a fourth-order tensor known as the com-
pliance tensor, which has the same symmetries as the stiffness tensor. The
strain energy function in terms of the stress components, sometimes known
as the complementary energy per unit volume, is
1
wc ¼ aijkl rij rkl ,
2
and equation (5.16) can be obtained from
@wc ðrÞ
eij ¼ :
@rij
We will consider a homogeneous solid so that the components of the stiff-

ness and compliant tensors are constants.
As already mentioned the components of the most general isotropic
fourth-order tensor are of the form
adij dkl þ bdik djl þ cdil djk ,
where a, b, and c are scalars. It then follows that for an isotropic linear
elastic solid, equation (5.13) is of the form,
rij ¼ adij ekk þ beij þ ceji :
Since eij ¼ eji it then follows that
cijkl ¼ 2ldik djl þ kdij dkl , ð5:17Þ
and from equation (5.13) that

rij ¼ 2leij þ kekk dij , ð5:18Þ
where l and k are known as Lamé’s constants for isothermal deformation.

A compressible classical linear elastic solid has two independent elastic
constants, and any other elastic constant can be expressed in terms of l
and k. For example, Young’s modulus, the bulk modulus, and Poisson’s
ratio are given by
lð3k þ 2lÞ 2 k
E¼ , K ¼ k þ l, and m¼ ,
kþl 3 2ðk þ lÞ
respectively. Equation (5.18) can be inverted to give
1 kdij
eij ¼ rij þ rkk : ð5:19Þ
2l 2lð3k þ 2lÞ
A brief discussion of some aspects of anisotropic linear elasticity is now

given. Equation (5.13) can be expressed in the matrix form
½r ¼ ½C½e, ð5:20Þ
where ½r and ½e are ð631Þ column matrices related to the Cauchy stress
and linear infinitesimal strain tensors by
½rT ¼ ½r1 r2 r3 r4 r5 r6 ¼ ½r11 r22 r33 r23 r31 r12 ,

½eT ¼ ½e1 e2 e3 e4 e5 e6 ¼ ½e11 e22 e33 2e23 2e31 2e12 ,
and ½C is a ð636Þ symmetric matrix related to the fourth-order stiffness

tensor by
2 3
C11 ¼ c1111 C12 ¼ c1122
C13 ¼ c1133 C14 ¼ c1123 C15 ¼ c1113 C16 ¼ c1112
6 C22 ¼ c2222
C23 ¼ c2233 C24 ¼ c2223 C25 ¼ c2231 C26 ¼ c2212 7
6 7
6 C33 ¼ c3333 C34 ¼ c3323 C35 ¼ c3331 C36 ¼ c3312 7
6 c2323 þc2332 7
6 7
½C¼ 6 Symmetric C44 ¼ C45 ¼ c2331 C46 ¼ c2312 7:
6 2 7
6 c3131 þc3113 7
6 C55 ¼ C56 ¼ c3112 7
6 2 7
4 c1212 þc1221 5
C66 ¼
2
ð5:21Þ
The matrix ½C is known as the Voigt matrix. It follows from equation (5.21)
that for an isotropic solid,
2 3
2l þ k k k 0 0 0
6 2l þ k k 0 0 07
6 7
6 2l þ k 0 0 07
½C ¼ 6
6
7:
6 Symmetric l 0 077
4 l 05
l
Equation (5.20) can be expressed in suffix notation,
6
rp ¼ + Cpq eq , p 2 f1; 2: . . . 6g;
q¼1
however, it is important to note that this is a matrix equation, not a tensor

equation. The compliance relation (5.16) can also be expressed in matrix
form as
½e ¼ ½ A½r,
where ½ A is the matrix inverse of ½C.
5.7 Linear Thermoelasticity
In this section some constitutive aspects of linear thermoelasticity are con-

sidered. These are restricted to isotropic solids and the assumptions of
isothermal linear elasticity, along with small deviations from a reference
absolute temperature Ho , so that jH H0 j=H0 << 1. When the temperature
of an isotropic unstressed elastic solid is changed, pure dilatation e ¼
av ðH Ho Þ occurs so that the components of the thermal strain etij are
av
etij ¼ ðH H0 Þdij , ð5:22Þ
3
where av is the volume coefficient of thermal expansion, which is assumed

to be constant.
It is clear from equation (5.22) that no shear deformation is produced

by a uniform temperature change in a homogeneous elastic solid. For
homogeneous deformation with strain eij , the part due to the stress is
eij etij , and it follows from superposition and equations (5.18) and (5.22)
that

2
rij ¼ 2leij þ kekk dij l þ k av ðH H0 Þdij , ð5:23Þ
3
which is an equation of state. In equation (5.23) l and k are the isothermal
moduli, assumed constant for infinitesimal deformation and
jH H0 j=H0 << 1. In this section specific thermodynamic properties are
taken per unit mass. The stress and infinitesimal strain r and e are now
decomposed into the deviatoric and isotropic parts to illustrate that thermo-
dynamic effects in an isotropic linear compressible thermoelastic solid are
confined to the dilatational part of the deformation. Substitution of
rij ¼ sij þ rdij and eij ¼ eij9 þ 3e dij , where e ¼ eii , r ¼ rii =3, in equation
(5.23) gives the decoupled shear and dilatational responses,
sij ¼ 2leij9, r ¼ K fe av ðH H0 Þg, ð5:24Þ
where K ¼ 2l=3 þ k is the isothermal bulk modulus.

It is evident from equation (5.24) that the uncoupled shear deforma-
tion is not influenced by temperature change and is both isothermal and
isentropic, that is, for a linear thermoelastic solid the thermodynamic
effects are restricted to the dilatational part of the deformation.
There are four important specific thermoelastic potentials each of
which can be expressed as a function of two of the thermodynamic
properties r, e, s, and H. A potential, which is expressed as a function
of two of the above thermodynanic properties, such that the other two
can be obtained by partial differentiation is a fundamental equation of
state. These fundamental equations of state are related by Legendre
transformations.
First the specific internal energy and Helmholtz free energy are
considered and are expressed as fundamental equations of state in the
form
_

u ¼ u eij9, e, s ð5:25Þ
and
_
f ¼ f eij9, e, H , ð5:26Þ
_
respectively, where uð0, 0, s0 Þ ¼ 0 and f ð0, 0, H0 Þ ¼ 0. The strain tenor e is
replaced by e9 þ ðe=3ÞI in equations (5.25) and (5.26) in order to indicate
the decoupling of the distortional and dilatational effects.
A Gibbs relation for a linear isotropic thermoelastic solid is
rij deij sij deij9 rde

Hds ¼ du ¼ du : ð5:27Þ
q q
It follows from equation (5.27) and
_ _ _
@u @u @u
du ¼ deij9 þ de þ ds
@eij9 @e @s
that
_ _ _
@u @u @u
sij ¼ q , r¼q , H¼q : ð5:28Þ
@eij9 @e @s
Equation (5.28) indicates that the fundamental equation of state for the
_
internal energy is of the form u ¼ u eij9, e, s .
A Gibbs relation
rij deij sij deij9 þ rde

df ¼ sdH þ ¼ sdH þ ð5:29Þ
q q
is obtained from equation (5.27) and the Legendre transformation f ¼

u Hs, and it follows from equation (5.29) and
@f @f @f
df ¼ deij þ de þ dH
@eij9 @e @H
that
_ _ _
@f @f @f
sij ¼ q , r¼q , s¼ : ð5:30Þ
@eij9 @e @H
Equations (5.30) indicate that the fundamental equation of state for the
_
Helmholtz free energy is of the form f ¼ f eij9, e, s .
Substituting equation (5.24) in equation (5.29) and integrating gives

_ 1 1 2
f eij9, e, H ¼ leij9eij9 þ Ke Kav eðH H0 Þ þ F ðHÞ, ð5:31Þ
q 2
where F ðHÞ is a function to be determined. Equation (5.30) is consistent

with equations (5.24) and (5.30)1,2. It follows from (5.30)3 and (5.31) that
Kav e
s¼ F9ðHÞ: ð5:32Þ
q
Since ðHdsÞe¼0 ¼ ce dH, where ce is the specific heat at constant strain e, and
se¼0 ¼ F 0 ðhÞ,
ZH _
dH h
F9ðHÞ ¼ ce _ ¼ ln : ð5:33Þ
H h0
H0
The linearized form of equation (5.33) is

H H0
F9ðHÞ ¼ ce , ð5:34Þ
H0
and this along with equation (5.32) gives

Kav e H H0
s¼ þ ce : ð5:35Þ
q H0
According to equation (5.35) the specific entropy is taken as zero for

a reference state with H ¼ H0 and e ¼ 0.
Substituting the integrated form of equation (5.34) in equation (5.31)
gives the fundamental equation of state

_ 1 1 2 1 ðH H0 Þ2
f eij9, e, H ¼ leij9 eij9 þ Ke Kav eðH H0 Þ ce :
q 2 2 H0
ð5:36Þ
It may be verified that equation (5.30) can be obtained from equation

(5.36). The strain energy per unit volume for isothermal deformation,
1
wðe9, eÞ ¼ qf ðe9, e, H0 Þ ¼ leij9 eij9 þ Ke2 ,
2
is obtained from equation (5.36).

It follows from equation (5.35) that, for isentropic deformation from
the reference state,
Kav eH0
H H0 ¼
qcv
and substitution in equation (5.24)2 gives
!
a2v K 2 H0
r¼ Kþ e:
qce
Consequently, the isentropic (reversible adiabatic) bulk modulus Ka is

given by
a2v K 2 H0
Ka ¼ K þ ,
qce
and the isothermal and isentropic shear moduli are equal. It can be deduced
from the Gibbs relation (5.27) that the following expression for the specific
internal energy

1 1 2
u^ eij9, e, H ¼ leij9 eij9 þ Ke þ Kav eH0
q 2
ð5:37Þ
1 ðH H0 Þ2
þ ce þ ce ðH H0 Þ
2 H0
is obtained from the Legendre transformation u ¼ f þ Hs and equations

(5.35) and (5.36). It should be noted that equation (5.37) is not a funda-
mental equation of state since u^ is a function of eij9, e and H instead of
eij9, e and s. It is easily determined from equation (5.37) that

@ u^
ce ¼ :
@H H¼H0
The fundamental equation of state

1 1 2
u~ eij9, e, s ¼ leij9 eij9 þ Ke þ Kav eH0 þ
q 2
( )
s Kav eHo 1 s Kav eHo 2
ce H0 þ H0 ð5:38Þ
ce qce 2H0 ce qce
is obtained by eliminating ðH H0 Þ from equations (5.35) and (5.37). It

may be verified that equations (5.28) can be obtained from equation (5.38)
with the use of equation (5.35).
The specific enthalpy for a linear thermoelastic solid is related
. to the
internal energy by the Legendre transformation h ¼ u sij eij9 q re=q,
and the differential of this along with equation (5.27) gives the Gibbs
relation

eij9 dsij þ edr
dh ¼ Hds : ð5:39Þ
q
@h @h @h
dh ¼ dsij þ dr þ ds
@sij @r @s
and
@h @h @h
eij9 ¼ q , e¼q , H¼ : ð5:40Þ
@sij @r @s
Equation (5.40) indicates that the fundamental equation of state for the

enthalpy is of the form h ¼ h sij , r, s . Before determining the fundamental

form h sij , r, s it useful to consider the form
^ sij sij 1 K r 2
h sij ,r,H ¼ þ þav ðHH0 Þ þ
2ql 2 q K

Kav H0 r r 1 ðHH0 Þ2
þav ðHH0 Þ þce ðHH0 Þþ ce , ð5:41Þ
q q K 2 H0
that is. determined from equations (5.24), (5.38), and h ¼

u sij eij9 q re=q. It may be deduced from equation (5.39) that, for a re-
versible process at constant stress, the specific heart at constant stress
is given by
^ !
@h
cr ¼ :
@H
H¼H0
It then follows from equation (5.41) that
Ka2v H0
cr ¼ ce þ :
q
A fundamental equation of state for the enthalpy is obtained by eliminat-

ing ðH0 HÞ from equation (5.41) and
, !
av r Ka2v ce
H H0 ¼ s þ ,
q q h0
that is obtained from equations (5.24)2 and (5.35).

The Gibbs free energy is the final thermodynamic potential for a linear
thermoelastic solid that we will consider. It is related to the Helmholtz free

energy by the Legendre transformation, g ¼ f sij sij q re=q,
which gives
eij9 dsij edr

dg ¼ sdH : ð5:42Þ
q
It follows from equation (5.42) and
@g @g @g
dg ¼ dsij þ dr þ dH
@sij @r @H
that
@g @g @g
eij9 ¼ q , e ¼ q , s¼ : ð5:43Þ
@sij @r @H
Equation (5.43) indicates that the fundamental equation of state for

the Gibbs free energy is of the form g ¼ g sij , r, H , which can be obtained
as

1 sij sij K h r i2 h r i
g sij ,r,H ¼ þ þav ðHH0 Þ þav ðHH0 Þ
q 4l 2 K K
1 ðHH0 Þ2
½Kav ðHH0 Þþrg ce , ð5:44Þ
2 H0
.
by using the Legendre transformation g ¼ f sij eij9 þ re q, and equa-
tions (5.24) and (5.36).
As an exercise the reader should verify that equation (5.43) is satisfied
by equation (5.44).
The complementary energy per unit volume for isothermal deforma-
tion is given by wc ðs, rÞ ¼ qgðs, r, h0 Þ, and it follows from equation (5.44)
that
sij sij r2
ws sij , r ¼ , ð5:45Þ
4l 2K
which is equal to the negative of wðe9, eÞ, the strain energy per unit volume
for isothermal deformation. Often the complementary energy per unit
volume is described as the strain energy in terms of stress with the negative
signs in equation (5.45) replaced by positive signs. However, the form
(5.45) with negative.signs is obtained from the Legendre transformation,

g ¼ f sij eij9 þ re q.
Constitutive relations for nonlinear elastic solids are considered in
later chapters.
5.8 Determinism, Local Action, and Material

Frame Indifference
It has already been noted that constitutive equations should not violate the
laws of thermodynamics. In addition they should satisfy the principles of
determinism, local action, and frame indifference. These principles are
sometimes called axioms, and it is possible to formulate more than three.
For example, Eringen [6] lists eight axioms of constitutive theory; however,
the three principles discussed here appear to be sufficient to develop a con-
stitutive theory.
The principle of determinism states that the stress and specific internal
energy at a material point at time t are determined uniquely by the past
history of the motion and temperature field up to and including time t. The
heat flux q is also so determined; however, this is usually of less importance
in our considerations than the stress. It is reasonable to assume that the
history of the motion and temperature fields in the recent past has more
influence on the state of the medium than that of the distant past, that is,
the material has a fading memory. A perfectly elastic solid has a perfect
memory of an undeformed reference configuration and an inviscid fluid or
a Newtonian viscous fluid has no memory of a reference configuration.
The principle of local action states that the stress, specific internal en-
ergy, and heat flux at a material point X are independent of the history of
motion and temperature outside an arbitrary small neighborhood of X.
Materials that obey this principle are known as simple materials. In a simple
material, the stress at time t at a material point depends only on the history of
the deformation gradient and temperature. This implies that all relations
5.8 Determinism, Local Action, and Material Frame Indifference 139
between stress and past history of motion can be obtained from experiments
that involve homogeneous deformation and temperature. These relations
are constitutive equations. In this book we consider only simple materials,
since all the usual material models considered in civil and mechanical engi-
neering refer to simple materials.
Before discussing the principle of frame indifference it is necessary to
introduce the concepts of a frame of reference and an observer that are
often regarded as synonymous. A frame of reference may be regarded as
a hypothetical rigid body to which an infinite number of coordinate systems
may be attached. Right-handed, rectangular Cartesian coordinate systems
are considered for the following discussion.
The concept of a frame of reference involves an observer that can
measure time and the relative position of a point in E3 . Two different
observers are said to be equivalent if they observe the same distances
between pairs of points in E3 and the same interval of time between events.
Transformation relations between coordinate systems attached to the
same frame of reference are time independent. However, transformation
relations between coordinate systems attached to different frames of ref-
erence are, in general, time dependent unless the frames are moving in
relative translation. Equations or physical quantities, which are invariant
under coordinate transformation in the same frame of reference, must be
tensors or tensor equations; however, this does not necessarily ensure
invariance for transformations between coordinate systems attached to
different frames of reference. Physical quantities are said to be objective
if observers on different frames, in relative motion, observe the same phys-
ical quantities, that is, if the quantities are invariant under all changes of
frame. Temperature is an example of an objective scalar, velocity is an
example of a vector that is, intuitively, not objective. Applied to constitu-
tive equations the principle of material frame indifference, sometimes
known as the principle of material objectivity, requires that equivalent
observers observe the same material properties or constitutive equations.
An alternative statement is, constitutive equations are frame indifferent if
they are invariant under all changes of frame. A simple example is the
extension of an elastic spring due to an axial force. Equivalent observers
observe the same extension of the spring and, since the axial force is related
to the extension, the various observers observe the same force. We can
assert that contact forces are objective; however, body forces may not be.
For example, an observer attached to a body in free fall, under the action of
the earth’s gravitational field, observes no body force since no acceleration
is observed. However, an observer fixed in space observes a body force.
A tensor is said to be objective if equivalent observers on different frames
of reference observe the same tensor, so that the tensor is invariant under
change of frame. The precise meaning of this statement is discussed later in
this chapter.
Consider the ordered pair fx, tg, where x is the position vector of
a point in E3 . A change of frame or observer is a one-to-one mapping of
E3 3 t onto itself so that the same distance between any two points in E3 is
observed and time intervals between events are preserved. Let x and t be
the position vector, of a point P in E3 , relative to an origin O, and time,
respectively, in frame of reference F, and let x and t ¼ t a, where a is
a constant, be the position vector of P relative to an origin O , and time,
respectively, in frame of reference F . It should be noted that x is observed
by observer F and x by observer F . If F is moving in translation, relative
to F , we have the trivial relation
x ¼ cðt Þ þ x,
where cðt Þ is the position vector of O with respect to O . However, if F

and F are undergoing relative translation and rotation,
x ¼ cðtÞ þ QðtÞx, t ¼ t a, ð5:46Þ
where Q is a proper orthogonal tensor, which represents a rotation of F

relative to F , and a is a constant so that two time scales have the same unit
and a different origin. Equation (5.46) is known as a Euclidean transfor-
mation and is an observer transformation from F to F since observer F
observes position vector x and observer F observes position vector x for
the same point in E3 . It may be easier to visualize the Euclidean trans-
formation (5.46) when it is expressed in suffix notation related to coordi-
nate systems O xi and Oxi fixed in F and F, respectively,
xi ci ðtÞ ¼ Qij ðtÞxj :
When Q ¼ I and cðtÞ is constant, the transformation is said to be Galilean.

A Galilean frame of reference is a frame with respect to which the laws of
classical mechanics hold.
If x1 and x2 are position vectors in F of two points it follows from
equation (5.46) that

x2 x1 ¼ Qðx2 x1 Þ: ð5:47Þ
It is easily verified from equation (5.47) that the condition

x x
¼ jx2 x1 j,
2 1
for equivalent observers is satisfied and that is recovered when Q ¼ I. It

follows from equation (5.47) that a vector u is objective if and only if it
satisfies the condition
u ðx , t Þ ¼ QðtÞuðx, tÞ, ð5:48Þ
for all F and F , as does the directed line segment ðx2 x1 Þ, so that different
observers observe the same vector. In order to clarify this, let fe1 , e2 , e3 g be

the unit base vectors for coordinate system 0xi , and e1 , e2 , e3 the unit
base vectors for 0 xi , given by

ei ðtÞ ¼ QðtÞei : ð5:49Þ
The relation
ei ei ¼ I
and equation (5.49) give
ei ðtÞ ei ðtÞ ¼ QðtÞ: ð5:50Þ
Eliminating Q from equation (5.48) and equation (5.50), gives
u ðx , t Þ ei ¼ uðx, tÞ ei : ð5:51Þ

It may be deduced from equation (5.51) that the components of an

objective vector u referred to the coordinate system 0xi are the same as
the components of u referred to the coordinate system 0 xi . Similarly,
a tensor T ðx, tÞ referred to F is objective if T ðx , t Þ referred to F , and
T are related by
T ¼ QðtÞTQT ðtÞ: ð5:52Þ
Substitution of equation (5.50) in equation (5.52) gives
ei Tej ¼ ei T ej : ð5:53Þ
It may be deduced from equation (5.53) that the components of an objec-

tive second-order tensor T referred to the coordinate system 0xi in F are the
same as the components of T referred to the coordinate system 0 xi in F .
It may also be deduced from equation (5.48) that the scalar product of
two objective vectors is an objective scalar / ðx , t Þ ¼ /ðx, tÞ and that
a second-order objective tensor T operating on an objective vector u results
in an objective vector since
T
T u ¼ QTQ Qu ¼ QTu:
The following scalar thermodynamic quantities are objective, temperature,

pressure, entropy, internal energy, free energy, and enthalpy, etc. However,
there are scalars that are clearly not objective such as speed and magnitude of
acceleration. Two vectors that are, intuitively, not objective for all observer
transformations are velocity and acceleration. For the velocity vector this can
be verified by differentiating equation (5.46) once with respect to time to obtain

v ¼ Qv þ c_ þ Xðx cÞ, ð5:54Þ
where QQ_ T ¼ QQ_ T ¼ X. It is clear that the velocity vector is objective,

only for transformations with c_ ¼ 0 and Q_ ¼ 0. For the acceleration vector,
differentiation of equation (5.54) with respect to time and the substitution
of v obtained from equation (5.54) gives

a ¼ Qa þ c€ þ X_ X2 ðx cÞ þ 2Xðv c_ Þ: ð5:55Þ
In equation (5.55) the terms X_ ðx cÞ, X2 ðx cÞ, and 2Xðv c_ Þ are
analogous to the terms, which appear in the expressions ar ¼ r€ rh_2 ,
ah ¼ rh€ þ 2r_h,
_ for acceleration components in terms of plane polar coordi-
nates ðr, hÞ. It is clear that the acceleration vector is objective only if

2
c€ þ X_ X ðx cÞ þ 2Xðv c_ Þ ¼ 0,
and transformations that satisfy this condition are known as Galilean trans-
formations. A frame of reference, with respect to which Newton’s laws are
valid, is known as an inertial frame of reference, and all frames obtained
from an inertial frame of reference by a Galilean transformation are also
inertial frames of reference.
Some further examples of physical quantities and aspects of objectivity
of interest in continuum mechanics are now considered. In what follows the
objective is taken to mean objective for all observer transformations.
The material time derivative of any objective vector is not objective
since differentiation of equation (5.48), with respect to time, gives
u_ ¼ Qu_ þ Qu:
_ ð5:56Þ
Also the material time derivative,
@T ðx, tÞ
T_ ¼ þ v =T ðx, tÞ ð5:57Þ
@t
or
@T ðxðX ÞtÞ
T_ ¼ ,
@t
of an objective second-order tensor, T is not objective since it follows

from equation (5.52) that
_ T þ QTQ
_ T T
T_ ¼ QTQ þ QT Q_ : ð5:58Þ
It is evident that the material rate equation (5.57) is not suitable for use in
constitutive equations that involve rates of objective second-order tensors.
In order to obtain an objective rate of an objective second-order tensor it is
desirable to consider first the two-point deformation gradient tensor,
F ¼ Gradx, ðFiK ¼ @xi =XK Þ, and then the velocity gradient tensor,

L¼ gradv, Lij ¼ @vi @xj , and its symmetric and antisymmetric parts.
Since no change in the reference configuration of a deforming body is
observed by observers on F and F , Grad [ Grad . It then follows that
F ¼ Gradx and substitution of x ¼ cðt Þ þ Qðt ÞxðX, tÞ gives
F ¼ QF: ð5:59Þ
According to the condition (5.52), F is not objective; however, it may be

regarded as quasi-objective since the reference configuration does not
change under an observer transformation. The velocity gradient is given
_ 1 and substituting equation (5.59) in L ¼ F_ F 1
in terms of F by L ¼ FF
gives

L ¼ QF
_ þ QF_ F 1 QT ¼ QLQT þ QQ
_ T, ð5:60Þ
which shows that L is not objective. It follows from equation (5.60) that, for
the rate of deformation and spin tensors,
1
D ¼ L þ LT ¼ QDQT ð5:61Þ
2
and
1 T

T _ T,
W ¼ L L ¼ QW Q þ QQ ð5:62Þ
2
respectively, which show that the rate of deformation is objective and the
spin is not objective.
It is interesting to consider the left and right Cauchy-Green tensors,
T T
B ¼ FF and C ¼ F F, ð5:63Þ
respectively. It can be shown from equations (5.59) and (5.62) that
B ¼ QBQT ð5:64Þ
and

C ¼ C: ð5:65Þ
Equations (5.64) and (5.65) indicate that B is objective but C does not
satisfy condition (5.52). However, like F, it may be regarded as quasi-
objective since it does not change under an observer transformation. We
now consider the material rate of change of a second-order objective ten-
sor. Elimination of Q_ and Q_ T from equation (5.58),
Q_ ¼ W Q QW
and
Q_ T ¼ QT W þ W QT ,
that are obtained from equation (5.62), gives the relation

T_ WT þ T W ¼ Q T_ WT þ TW QT : ð5:66Þ
Consequently,
o
T ¼ T_ WT þ TW ð5:67Þ
is a possible objective rate. The objective rate equation (5.67) is known as

the corotational or Jaumann rate and was proposed by Jaumann in 1911. It
is not unique since an infinite number of objective rates can be obtained.
Several other objective rates have been proposed for various constitutive
relations since 1911, for example,
D
T ¼ T_ þ LT T þ TL ð5:68Þ
is a rate proposed by Oldroyd [7].

The rate equation (5.67) is usually applied to the Cauchy stress tensor
r and has a particular physical interpretation that other stress rates do not
have, namely, that it is the stress rate relative to a frame of reference
rotating with the angular velocity of the principal axes of D. To show this
let ui ðtÞ be a right-handed triad of unit base vectors in the principal direc-
tions of D. The Cauchy stress can be expressed in the form
rðtÞ ¼ rij ðtÞui ðtÞ uj ðtÞ,
where the material time derivative is
d drij
r_ ðtÞ ¼ rij ui uj ¼ ui uj þ rij u_ i uj þ rij ui u_ j , ð5:69Þ
dt dt
and the rij are the stress components with respect to triad ui , i 2 f1, 2, 3g.
The dependence of r on x has been suppressed. The rate of change of stress
dr
relative to the system of base vectors xi ðtÞ is dt ij ui uj . It follows from
equation (2.39) that
u_ i ¼ W ui , i 2 f1, 2, 3g: ð5:70Þ
Substituting equation (5.70) in equation (5.69) and using W ¼

Wrs ur us ¼ W T gives, after some manipulation,
drij d
ui uj ¼ rij ui uj Wip rpj ui uj þ sip Wpj ui uj
dt dt ð5:71Þ
o dr
or r ¼ W r þ rW :
dt
The right-hand sides of equations (5.67), with T replaced by r, and (5.71)

are equal, so that
o drij
r¼ ui uj :
dt
An application of the Oldroyd rate equation (5.68) relates Almansi’s strain

tensor
1
g¼ I F T1 F 1
2
to the rate of deformation tensor D since it may be shown that
D

g ¼ g þ LT g þ gL ¼ D:
All the terms of materially objective constitutive equations must be objec-

tive. For example, in equation (5.9) the quantities s, D, and p are objective
and equation (5.9) is materially objective. However, the constitutive equa-
tions for linear elasticity are not materially objective and are valid only for
infinitesimal displacement gradients.
EXERCISES
5.1. Decompose the specific strain energy function W ðeÞ for a linear iso-
tropic elastic solid into a dilatational part and a distortional part. Obtain
the specific strain energy as a function Wc ðrÞ, known as the complemen-
tary energy. Decompose Wc ðrÞ into a dilatational part and a distortional
part and show that the distortional part is related to the octahedral
shearing stress and the second invariant of the deviatoric Cauchy stress.
5.2. The potential energy of a linear elastic body is given by
Z Z Z
V¼ wðeÞdv ti ui da qfi ui dv,
v @vt v
where @vt is the part of the surface @v upon which the surface traction
is prescribed.
Use the result that the strain energy function, wðeÞ ¼ 12 cijkl eij ekl , is
positive definite to show that, V is a minimum when ui ðxÞ is the solu-
tion displacement field in competition with all other members of the
set of kinematically admissible displacement fields.

Hint, show that V ui V ðui Þ > 0, where ui is a kinematically admis-
sible displacement field and ui is the solution displacement field, with
the equality holding if and only if ui ¼ ui
5.3. The complementary energy of a linear elastic body is given by
Z Z
Vc ¼ wc ðrÞdv ti ui da,
v @vu
where @vu is the part of the surface @v upon which the displacement is
prescribed.
Use the result that the complementary energy function,
1
wc ðrÞ ¼ 2 aijkl rij rkl , is positive definite, to show that V c is a minimum
when rij ðxÞ is the solution stress field in competition with all other
members of the setof statically
admissible stress fields.

Hint, show that V c rij V c rij > 0, where rij ðxÞ is a statically ad-

missible displacement field and rij ð xÞ is the solution displacement

field, with the equality holding if and only if rij ¼ rij .
5.4. Show that ðW XÞ, where W is the spin tensor, is objective.
5.5. A proposed constitutive equation has the form
r_ ¼ 2lD þ kItrD,
where l and k are Lamé’s constants of linear elasticity r_ is the mate-

rial derivative of Cauchy stress. Show that this relation is not materi-
ally objective.
5.6. Show that a constitutive of the form
o
r ¼ 2lD þ kItrD
is materially objective.
5.7. Show that a constitutive equation of the form r ¼ f ðDÞ is materially
objective.
5.8. Show that the time-dependent basic invariants of the stress tensor r
are stationary when the Jaumann stress rate vanishes.
References 149
REFERENCES
1. Müller, I. (1985). Thermodynamics. Pitman.

2. White, F.M. (1991). Viscous Fluid Flow, 2nd Ed. McGraw-Hill.
3. Love, A.E. H. (1927). A Treatise on the Mathematical Theory of Elasticity,
4th Ed. Cambridge University Press.
4. Goodier, J.N., and Timoshenko, S. (1970). Theory of Elasticity, 3rd Ed.
McGraw-Hill.
5. Sokolnikov, I.S. (1956). Mathematical Theory of Elasticity, 2nd Ed. McGraw-
Hill.
6. Eringen, A.C. (1962). Nonlinear Theory of Continuous Media. McGraw-Hill.
7. Oldroyd, J.G. (1950). On the Formulation of Rheological Equations of
State. Proc. Roy. Soc. A 200, pp. 523–541.
6 Finite Deformation of
an Elastic Solid
6.1 Introduction
The constitutive equations of the linear theory of elasticity, which have been
discussed in chapter 5, can be obtained as alimitingcase of those of finite
deformation elasticity by neglecting terms O j=X uj2 . However, the consti-
tutive equations of linear elasticity are not objective, although those of finite
deformation elasticity are objective. In this chapter we consider the constitu-
tive equations of finite deformation elasticity, henceforth described as non-
linear elasticity. The emphasis is on isotropic solids and it is evident where the
specialization to isotropy occurs. Symbolic notation is mainly used, although
some equations are given in both symbolic and suffix notation.
6.2 Cauchy Elasticity
An elastic solid is a solid that is rate independent and the components of the
Cauchy stress tensor are single-valued functions of the deformation gradi-
ent tensor, for either isothermal or adiabatic deformation, so that
s ¼ qðF Þ, rij ¼ qij ðFmK Þ: ð6:1Þ
In equation (6.1) any dependence on X and t has been suppressed, that is,
the solid is assumed to be homogeneous in the undeformed reference
configuration. The form of the tensor function q, of the tensor argument
F, depends on whether the deformation is isothermal or adiabatic and also
depends on the reference configuration unless the solid is isotropic. The
difference between the forms of q for isothermal and adiabatic deforma-
tion, of materials for which the nonlinear elastic model is realistic, is very
150
small and is negligible for most applications. In this chapter and chapter 7,
isothermal deformation is assumed. Equation (6.1) is an example of a con-
stitutive model for a simple material, and the function q can be obtained
from experiments involving homogeneous static deformations.
The function in equation (6.1) must be objective. An objective func-
tion is obtained by noting that the stress depends only on the change of
shape and is not influenced by rigid body rotations of a deformed config-
uration. Suppose a deformed configuration of an elastic body is given a rigid
rotation, described by a proper orthogonal tensor Q, about a fixed material
point taken as the origin. The position vector x of a particle before the
rotation becomes
x ¼ Qx
after the rotation. Then the deformation gradient and Cauchy stress tensors
after rotation are given by,

F ¼ QF ð6:2Þ
and
T
s ¼ QsQ , ð6:3Þ
respectively. It follows from equation (6.1) that

s ¼ qðF Þ: ð6:4Þ
Since a rotation of the spatial configuration is equivalent to a rotation of an

observer, it further follows from equations (6.1) to (6.4) that q is observer
independent, that is, objective, if
QqðF ÞQT ¼ qðQF Þ ð6:5Þ
or
qðF Þ ¼ QT qðQF ÞQ:

152 Finite Deformation of an Elastic Solid
for any Q 2 Orthþ where Orthþ is the set of all proper orthogonal tensors.
Substitution of the polar decomposition of the deformation gradient,
F ¼ RU, in equation (6.5) gives
qðF Þ ¼ QT qðQRU ÞQ,
which is valid for any proper orthogonal tensor, in particular for Q ¼ RT .

Consequently,
qðF Þ ¼ RqðU ÞRT , ð6:6Þ
which shows that q can be expressed as a function of six independent

components of U instead of nine components of F. It follows that equation
(6.6) is a necessary condition for equation (6.5) to hold. This condition is
also sufficient for by replacing F by QF, in equation (6.6), where Q is an
arbitrary proper orthogonal tensor, and noting that ðQRÞU is the polar
decomposition of QF, we obtain
qðQF Þ ¼ QRqðU ÞRT QT ¼ QqðF ÞQT ,
and this is equivalent to equation (6.5).

The following relation obtained from equations (6.1) and (6.6),
s ¼ RqðU ÞRT , ð6:7Þ
can be put in a different form, which may be convenient since the decom-
position of F is not required. Using the polar decomposition of F, we obtain
s ¼ FU 1 qðU ÞU 1 F T ,
and then using the right Cauchy Green deformation tensor C ¼ F T F ¼ U 2 ,
T
s ¼ FfðCÞF , ð6:8Þ
where

fðC Þ ¼ C 1=2 q C1=2 C1=2 :
It is easily verified that equation (6.8) is objective. Using equation (3.33)
S ¼ JF 1 sF T ,
where S is the second Piola-Kirchhoff stress, equation (6.8) can be put in

the form
S ¼ JfðC Þ, ð6:9Þ
pffiffiffiffi
where J ¼ det½U ¼ det C ¼ det½F . The Cauchy stress and the second
Piola-Kirchhoff stress can be obtained from equations (6.8) and (6.9),
respectively, for a given deformation if the function f is known. The func-
tion f is a symmetric second-order tensor function of a symmetric tensor
argument that depends on what is known as the material symmetry and must
give rise to a physically reasonable response to the application of stress.
For example, application of a hydrostatic pressure should result in a decrease
in volume, and an axial tension force applied to a bar should result in an
increase in length for either isothermal or isentropic deformation.
According to equations (6.1) and (6.9), the stress in an elastic solid is
determined uniquely from the deformation from the undeformed reference
configuration and does not depend on the deformation path. A solid that
satisfies this condition is described as Cauchy elastic. If the stress cannot be
obtained from a scalar potential function known as the strain energy func-
tion, the elastic solid is not a conservative system and the work done by the
stress could be dependent on the deformation path.
The simplest type of material symmetry is isotropy and the theory for
an isotropic elastic solid is given in this chapter. An isotropic material has
properties that are directionally independent and this is made more precise
in what follows.
If the stress due to an applied deformation is unchanged by a rigid
body rotation of the natural reference configuration, the rotation is called
a symmetry transformation. The rigid body rotation of the reference con-

figuration from configuration B1 to B2 is described by H 2 Orthþ where
Orthþ denotes the set of second-order proper orthogonal tensors, and if the
stress is unchanged 8H, where the symbol 8 means for all, the solid is
isotropic. Consider a homogeneous deformation described by deformation
gradient tensor F, referred to the unrotated reference configuration B1 ,
and by F 0 referred to B2 . Then with rectangular Cartesian coordinate sys-
tems, ð0, EÞ, ð00 , E0 Þ, related to B1 and B2 , respectively, and ðo,eÞ related to
the spatial configuration,
@xi @xi 9 @X9b

F¼ ei Ek , F9 ¼ 0 ei Ea , and H ¼ E9b EK ,
@XK @Xa @XK
and it follows that F 0 ¼ FH 1 . Since F9 is the deformation gradient re-

ferred to B2 ,

q9ðF9Þ ¼ q FH 1 ,
where q9 is the response function relative to B2 . An isotropic solid has

a response function that does not depend on rotation of the reference
configuration so that for isotropy q0 ¼ q, and
qðF Þ ¼ qðFP Þ, for 8P 2 Orthþ , ð6:10Þ
where P ¼ H 1 ¼ H T . The set orthþ is the symmetry group for an isotropic

elastic solid. A group is a set G that satisfies the following conditions: (1) an
operation (multiplication) is defined on G, that is, the product of two
members of the set is also a member; (2) the set G has a unit element;
and (3) every member of G has an inverse in G. It may be verified that
Orthþ is a group. If equation (6.10) is not true for all P 2 Orthþ but is
true for a subgroup of Orthþ , the solid is anisotropic and the subgroup is
known as the symmetry group of the solid. Finite deformation of anisotropic
elastic solids is not considered in this text and details can be found in [1].
The function fðCÞ in equation (6.9) is an isotropic function of C, that

is, by definition, QfðCÞQT ¼ f QCQT , 8Q 2 Orthþ if the solid is isotropic.
It may be shown [2] that f is an isotropic function if and only if it can be

2
f ¼ /0 I þ /1 C þ /2 C , ð6:11Þ
where /o , /1 , and /2 are scalar functions of the basic invariants of C,
1n o
I1 ¼ trC, I2 ¼ ðtrCÞ2 trC2 , I3 ¼ det C:
2
The invariants of C, as the term indicates, are unchanged if C is replaced by

QCQT for 8Q 2 orthoþ .
If the deformation is given and /0 , /1 , and /2 as functions of
I1 , I2 and, I3 are known, the second Piola-Kirchhoff stress can be obtained
from equations (6.9) and (6.11) for an isotropic solid.
6.3 Hyperelasticity
If a strain energy function, often referred to as a stored energy function w,

exists, the solid is known as Green elastic or hyperelastic.
The class of hyperelastic solids is a subclass of the class of Cauchy
elastic solids, and in this book we assume that elastic materials are hyper-
elastic. Strain energy functions correspond to the Helmholtz free energy for
isothermal deformation and the internal energy for isentropic deformation.
It has been shown in chapter 4 that the rate of work done by the body
forces and surface tractions acting on a body is equal to the sum of the
integrated stress power of the body and rate of change of the kinetic
energy of the body. It may be deduced from this that, if the system is
conservative, the integrated stress power is equal to the rate of change
of stored energy. This motivates the definition of a hyperelastic solid
that it is a solid for which the specific stress power is equal to the rate
of change of the specific strain energy [3]. The specific stress power and
specific strain energy, henceforth called the stress power and strain
energy function, respectively, are per unit volume of the undeformed
reference configuration in this chapter, rather than per unit mass as is
customary for specific thermodynamic properties. The specific strain

energy w for finite deformation of a hyperelastic solid can conveniently
be expressed as a function of one of, E, C, U, or F. A dependence on X is
suppressed so that it is assumed that the solid is homogeneous in
the reference configuration. The function w must be unchanged by a rota-
tion of the deformed configuration. This condition is satisfied if w is
expressed as a function of E, C, or U. If it is expressed as a function of
F, then the condition wðF Þ ¼ wðQF Þ must be satisfied 8Q 2 Orthþ in
particular for Q ¼ RT . It then follows that wðF Þ ¼ wðU Þ. Since the

Green strain tensor E ¼ 12 U 2 I ; and the Green deformation tensor
C ¼ U 2 , the strain energy function must also be expressible in the forms
wðEÞ and wðCÞ where we note that at this stage the same symbol, w, has
been used for different functions of different arguments that represent the
same quantity.
In order to obtain the stress strain relations for a hyperelastic solid
we first consider the stress power in terms of the second Piola-Kirchhoff
stress tensor S and the Green strain rate tensor E_ so that the specific stress
power P, is given by

P ¼ tr SE_ ¼ SKL E_ KL , ð6:12Þ
and P ¼ w_ ðEÞ. It is convenient to use suffix notation for the next develop-
ment. Then
@wðEÞ _
w_ ðEÞ ¼ EKL : ð6:13Þ
@EKL
It should be noted that in equation (6.13) all components of EKL are taken
as independent, for the differentiation, that is, the symmetry of EKL is
ignored. Then it follows from equations (6.12) and (6.13) that
@wðEÞ
SLK ¼ : ð6:14Þ
@EKL
In order to relate to equation (6.9), we now consider the stress power in

terms of the second Piola-Kirchhoff stress tensor S and the time rate of
the right Green deformation tensor C. It then follows from equation (6.12)
and E ¼ ð1=2ÞðC I Þ that
1 1
P ¼ trS C_ ¼ SKL C_ KL ð6:15Þ
2 2
and
@wðCÞ
SLK ¼ 2 : ð6:16Þ
@CKL
Equations (6.14) and (6.16) are unchanged by rotation of the deformed

configuration.
It follows from equations (6.9), (6.11), and (6.16) that, for an isotropic
elastic solid,
@wðCÞ
2 ¼ J / 0 þ /1 C þ /2 C 2 ,
@CKL
and this determines wðCÞ. For isotropy, the strain energy function wðCÞ is
a scalar isotropic function of C, that is,

w QCQT ¼ wðCÞ:
This condition is satisfied if w is expressed as a function,
^ ðI1 , I2 , I3 Þ,
w¼w ð6:17Þ
of the invariants of C. Since the strain energy must be zero for no defor-
mation w must satisfy the equivalent conditions,
w ðI Þ ¼ 0 ^ ð3, 3, 1Þ ¼ 0:
or w
The form (6.17) has been used extensively in the literature of hyperelasticity,
and constitutive equations based on this form are now developed as follows:
@wðC Þ 3 @w
^ @Ii
¼+ , ð6:18Þ
@CKL i¼1 i @CKL
@I
where
@I1 @I2 @I3

¼ dKL , ¼ I1 dKL CKL , ¼ I2 dKL I1 CKL þ CKM CML:
@CKL @CKL @CKL
ð6:19Þ
Equations (6.16), (6.18), and (6.19) then give

@w^ @w^ @w^ ^
@w @w^ @w^
SKL ¼2 þ I1 þ I2 dKL þ I1 CKL þ CKM CMK ,
@I1 @I2 @I3 @I2 @I3 @I3
ð6:20Þ

^
@w @w^ @w^ ^
@w @w^ @w^ 2
S¼2 þ I1 þ I2 I þ I1 Cþ C :
@I1 @I2 @I3 @I2 @I3 @I3
The relation between the Cauchy stress and the second Piola-Kirchhoff
stress
1 T
s ¼ J FSF
has already been obtained in chapter 5. Using this relation and FCF T ¼ B2
and FC 2 F T ¼ B3 , gives

@w^ @w^ @w^ ^
@w @w^ @w^ 3
s ¼ 2J 1 þ I1 þ I2 B þ I1 B2 þ B ,
@I1 @I2 @I3 @I2 @I3 @I3
and by using the Cayley-Hamilton theorem to eliminate B3 we obtain

@w^ @w^ @w^ @w^ 2
s ¼ 2J 1 I3 Iþ þ I1 B B : ð6:21Þ
@I3 @I1 @I2 @I2
We note that the invariants of B are the same as those of C. An alternative

form of equation (6.21),

1 @w^ @w^ @w^ @w^ 1
s ¼ 2J I2 þ I3 Iþ B I3 B , ð6:21aÞ
@I2 @I3 @I1 @I2
is obtained by eliminating B2 from equation (6.21) and B2 ¼

I1 B I2 I þ I3 B1 that follows from the Cayley-Hamilton theorem. It is
evident from equations (6.20) and (6.21) that, for an isotropic solid, the
tensors S and C and s and B are coaxial.
Since s ¼ 0 when B ¼ C ¼ I, or ðI1 , I2 , I3 Þ ¼ ð3, 3, 1Þ, it follows from
equation (6.21) that the strain energy function w ^ ðI1 , I2 , I3 Þ must satisfy the
condition
@w^ @w^ @w^

þ2 þ ¼ 0,
@I1 @I2 @I3 C¼B¼I
^ ð3, 3, 1Þ ¼ 0. Examples of physically realistic strain energy

in addition to w
functions are given in chapter 7.
The equality of the specific stress power and the specific strain energy
rate can also be applied to the nominal stress and the time rate of the
deformation gradient tensor to obtain

_
w_ ðF Þ ¼ tr RFÞ: ð6:22Þ
It then follows from
@wðF Þ _
w_ ðF Þ ¼ FIK
@FIK
and equation (6.22)
@wðF Þ
RKi ¼ : ð6:23Þ
@FiK
The relations equations (6.20) and (6.21) can be obtained from equation
(6.23), and this is left as an exercise for the reader.
The pairs of tensors fS, E ðor C=2Þg, fR, F g are said to be conjugate
since the stress power relations (6.12) or (6.15) and (6.22) are the trace of S
and the time rate of E ðor C=2Þ and the trace of R and the time rate of F. The
stress tensor s and the rate of deformation tensor D, which appear in the stress
power per unit reference volume, relation, P ¼ JtrðsDÞ, do not give rise to
conjugate variables since D is not the time derivative of a strain except for
infinitesimal strains where D ’ e,
_ and e is the infinitesimal strain tensor.
6.4 Incompressible Hyperelastic Solid
An incompressible elastic solid may be defined as an elastic solid that can

undergo only isochoric deformation, that is, the deformation is indepen-
dent of the hydrostatic part of the stress tensor. This implies that the bulk
modulus is infinite so that the solid is mechanically incompressible. Müller
[4] has shown that a mechanically incompressible fluid must have zero
coefficient of volume thermal expansion, and this proof can be extended
to consider a mechanically incompressible solid. It then follows that the
volume coefficient of thermal expansion is zero. Most solid rubbers, as
opposed to foam rubbers, have bulk moduli that are orders of magnitude
greater than their shear moduli so that for most applications volume deforma-
tion can be neglected compared with distortional deformation and the in-
compressible model is applicable. The incompressible model is not
applicable when the propagation of longitudinal waves is considered. It
should be noted, however, that solid rubber-like materials have volume
coefficients of thermal expansion that are large compared with those for
metals and several other solids.
The stresses in an incompressible solid can be obtained from the
deformation only to within an isotropic Cauchy stress that must be
obtained from the stress boundary conditions. Since the invariant I3 ¼ 1
for incompressible materials the strain energy functions for incompressible
isotropic hyperelastic solids are of the form
^ ðI1 , I2 Þ:
w¼w ð6:24Þ
Stress-deformation relations, for an incompressible hyperelastic solid, are

obtained by introducing the Lagrangian multiplier p and using the equation
of constraint,
I3 1 ¼ det C 1 ¼ 0: ð6:25Þ
6.4 Incompressible Hyperelastic Solid 161
Equation (6.16) is then replaced by
@wðC Þ @ det C
SLK ¼ 2 p : ð6:26Þ
@CKL @CKL
c
CIJ CKJ ¼ dIK det C,
CC cT ¼ IdetC,
c
where CKJ is the cofactor of CKJ , and it may be deduced that
@ det C c 1
¼ CKL ¼ CKL det C: ð6:27Þ
@CKL
Then from equations (6.25), (6.27), and the method of Lagrangian multi-
pliers we obtain
@ det C 1
¼ CKL : ð6:28Þ
@CKL
Substitution of equation (6.28) in equation (6.26) gives
^ ðC Þ
@w 1
SLK ¼ 2 pCKL , ð6:29Þ
@CKL
for an incompressible hyperelastic solid. Since I3 ¼ 1 and w ¼ w

^ ðI1 , I2 Þ for
incompressibility, equation (6.18) becomes
@wðCÞ 2 @w
^ @Ii
¼+ ,
@CKL i¼1 @Ii @CKL
where
@I1 @I2
¼ dKL , ¼ I1 dKL CKL ,
@CKL @CKL
and substituting in equation (6.29) gives

^
@w @w^ @w^ 1
SKL ¼2 þ I1 dKL CKL pCKL , ð6:30Þ
@I1 @I2 @I2

@w^ @w^ @w^
S¼2 þ I1 I C pC1 :
@I1 @I2 @I2
The relation between s and S for an incompressible solid is s ¼ FSF T and

applying this to equation (6.30) gives

@w^ @w~ @w^
rij ¼ 2 þ I1 Bij Bik Bkj pdij , ð6:31Þ
@I1 @I2 @I2

@w^ @w^ @w^ 2
s¼2 þ I1 B B pI:
@I1 @I2 @I2
Equation (6.31) can be put in a different form

@w^ @w^ 1
s¼2 B B pI, ð6:32Þ
@I1 @I2
by using the Cayley-Hamilton theorem to eliminate B2 : It should be noted

that the Lagrangian multiplier p in equation (6.32) is different from that in
the previous equations and that it is not determined by the deformation but
by stress boundary conditions.
The relation (6.23) for the nominal stress is replaced by
@wðF Þ 1
RKi ¼ pFiK , ð6:33Þ
@FiK
for the incompressible model.
6.5 Alternative Formulation
For some problems involving isotropic hyperelastic solids there is an ad-

vantage in expressing the strain energy as a symmetric function
^ ^ ^
w ¼ wðk1 , k2 , k3 Þ ¼ wðk2 , k1 , k3 Þ ¼ wðk2 , k3 , k1 Þ, ð6:34Þ
of the principal stretches ki , i 2 f1, 2, 3g, which are the eigenvalues of

pffiffiffiffi pffiffiffiffi
U ¼ C or V ¼ B. It is evident that
^
wð1, 1, 1Þ ¼ 0:
If the eigenvalues of C are distinct it can be expressed as the spectral

decomposition
3
ðiÞ ðÞ
C ¼ + k2i l l i ,
i¼1
where the unit vectors l ðiÞ , i 2 f1, 2, 3g are the eigenvectors of C and U,
pffiffiffiffi
corresponding to the ki . If only two eigenvalues of C, k1 and k2 , are dis-
tinct and k2 ¼ k3 , the spectral decomposition of C is

ð1 Þ ð1Þ ð Þ ð Þ
C ¼ k21 l l þ k22 I l 1 l 1 ,
and if k ¼ k1 ¼ k2 ¼ k3 , the spectral decomposition of C is
C ¼ k2 I:
The spectral decomposition of a second-order symmetric tensor is also

considered in chapter 1.
Principal invariants of C or B, in terms of the ki , are
I1 ¼ k21 þ k22 þ k23 , I2 ¼ k21 k22 þ k22 k23 þ k23 k21 , I3 ¼ k21 k22 k23 ,
and substitution in equation (6.17) gives equation (6.34).

It follows from C ¼ U 2 , S B ¼ SU ¼ US and equation (6.15) that the
stress power P can be expressed as

P ¼ tr SB U_ ,

_ ,
so that SB and U are conjugate. Then, since w_ ðU Þ ¼ tr ð@w=@U ÞUÞ
^
B @wðUÞ
S ¼ : ð6:35Þ
@U
The eigenvalues of U are ðk1 , k2 , k3 Þ and it follows from equation (6.35) that
the principal components of the Biot stress are then given by
@wðk1 , k2 , k3 Þ
SB
i ¼ , i 2 f1, 2, 3g, ð6:36Þ
@ki
and it follows that
@wð1, 1, 1Þ
¼ 0, i 2 f1, 2, 3g:
@ki
It may be shown from, s ¼ J 1 FSF T , F ¼ RU and the Biot stress given by
SB ¼ US ¼ SU,
that
JRT sR ¼ S B U: ð6:37Þ
The tensor JRT sR and the weighted Cauchy stress Js have the same
eigenvalues values. Consequently, it follows from equation (6.36) and the
principal values of each side of equation (6.37) that the principal compo-
nents of Cauchy stress are given by
^
@w
ri ¼ J 1 ki , i 2 f1, 2, 3g: No summation: ð6:38Þ
@ki
This result can be obtained otherwise. Let L1 , L2 , and L3 be the lengths of

the contiguous sides of a rectangular block in the reference configuration.
Apply an irrotational, homogeneous deformation such that the lengths of
the corresponding sides and stretches become l1 , l2 , and l3 and k1 , k2 , and k3 ,
respectively. Application of the principle of virtual work gives
^
L1 L2 L3 dwðk1 , k2 , k3 Þ ¼ r1 l2 l3 dl1 þ r2 l1 l3 dl2 þ r3 l1 l2 dl3 ,
3 ^
dki , J ¼ Ll11Ll22lL3 3 , and the dki are arbitrary the result
^ @w
and, since dw ¼ +i @k i
(6.38) follows.
It should be noted that, for an isotropic elastic solid, the right stretch
tensor, U, S, and S B are coaxial and the left stretch tensor, V, and s are coaxial.
The use of the strain energy function in the form (6.34) along with equa-
tion (6.36) or (6.38) could be the preferred method of solution for problems
with the principal directions of s and/or SB known beforehand. Problems that
involve cylindrical or spherical symmetry are examples. For problems such as
simple shear and the torsion of a cylindrical bar the use of form (6.34) is less
desirable since the principal directions of stress have to be found initially.
For incompressibility, k1 k2 k3 ¼ 1. An assumption, known as the
Valanis-Landel hypothesis [5], for an incompressible isotropic elastic
solid is
^
wðk1 , k2 , k3 Þ ¼ w
~ ðk1 Þ þ w
~ ðk2 Þ þ w
~ ðk3 Þ,
~ ð1Þ ¼ 0 and w9
where w ~ ð1Þ þ w0ð1Þ ¼ 2l: Most strain energy functions that
have been proposed for incompressible isotropic elastic solids satisfy this
hypothesis.
For incompressibility equations (6.36) and (6.38) are replaced by
^
@w p
SB
i ¼ , i 2 f1, 2, 3g: No summation,
@ki ki
and
^
@w
ri ¼ ki p, i 2 f1, 2, 3g: No summation, ð6:39Þ
@ki
respectively.
^
Since k1 k2 k3 ¼ 1, it is possible to eliminate k3 from w to obtain

ðk1 , k2 Þ ¼ w k1 , k2 , k1 1
^
w 1 k2 : ð6:40Þ
It follows from equation (6.40) and k3 ¼ ðk1 k2 Þ1 that
^ ^ ^ ^
@w @w @w @k3 @w 2 1 @w
¼ þ ¼ k1 k2 : ð6:41Þ
@k1 @k1 @k3 @k1 @k1 @k3
Then it may be deduced from equations (6.39) and (6.41) that
@w
r 1 r3 ¼ k 1 : ð6:42Þ
@k1
Similarly
@w
r2 r3 ¼ k 2 : ð6:43Þ
@k2
Equations (6.42) and (6.43) are useful when plane stress problems with
r3 ¼ 0 are considered.
Further aspects of strain energy functions are given in chapters 7
and 8.
EXERCISES
6.1. Show that the basic invariants of the left Cauchy-Green tensor B are of
the same form as those of the right Cauchy-Green tensor C, that is,
1n o
I1 ¼ trB, I2 ¼ ðtrBÞ2 trB2 , I3 ¼ det B:
2
6.2. Show that for consistency with classical linear elastic theory the strain
energy function, wðk1 , k2 , k3 Þ, must satisfy the condition
@2w
ð1, 1, 1Þ ¼ k þ 2ldij ,
@ki @kj
where k and l are Lamé’s constants of linear classical elasticity.

6.3. The Mooney-Rivlin strain energy function for an incompressible
hyperelastic solid is given by
l
w ¼ faðI1 3Þ þ ð1 aÞðI2 3Þg,
2
where 0 < a < 1. Show that it can be put in the form
^

~ ðk1 Þ þ w
w k1 , k2, k3 ¼ w ~ ðk2 Þ þ w
~ ðk3 Þ,
~ 0 ð1Þ þ w00 ð1Þ ¼ 2l:

~ ð1Þ ¼ 0 and w
where w
References 167
6.4 Show that, for an incompressible neoHookean elastic solid, the con-
stitutive equation r ¼ pI þ lB is materially objective.
REFERENCES

2. Rivlin, R.S., and Ericksen, J.L. (1955). Stress-Deformation Relations for
Isotropic Materials. J. Rat. Mech. Anal. 4, pp. 329–422.
3. Haupt, P. (2000). Continuum Mechanics and Theory of Materials. Springer.
4. Müller, I. (1985). Thermodynamics. Pitman.
5. Valanis, K.C., and Landel, R.S. (1967). The Strain-Energy Function of
a Hyperelastic Material in Terms of the Extension Ratios. J. Appl. Phys.
38, pp. 2997–3002.
7 Some Problems of Finite
Elastic Deformation
7.1 Introduction
In this chapter problems of finite deformation elastostatics for isotropic

hyperelastic solids are considered. Exact solutions for some problems of
finite deformation of incompressible elastic solids have been obtained by
inverse methods. In these methods a deformation field is assumed, and it is
verified that the equilibrium equations and stress and displacement boundary
conditions are satisfied. There are some elastostatic problems that have de-
formation fields that are possible in every homogeneous isotropic incom-
pressible elastic body in the absence of body forces. These deformation
fields are said to be controllable. In an important paper by Ericksen [1] an
attempt is made to obtain all static deformation fields that are possible in all
homogeneous isotropic incompressible bodies acted on by surface tractions
only. Ericksen further showed that the only deformation possible in all ho-
mogeneous compressible elastic bodies, acted on by surface tractions only, is
a homogeneous deformation [2]. In this chapter solutions are given for sev-
eral problems of finite deformation of incompressible isotropic hyperelastic
solids and two simple problems for an isotropic compressible solid. These
solutions with one exception involve controllable deformation fields and are
based on the physical components of stress, deformation gradient, and the
left Cauchy-Green strain tensor. This is in contrast to the approach of Green
and Zerna [3], who used convected coordinates and tensor components.
Rivlin, in the late 1940’s, was the first to obtain most of the existing
solutions, for deformation of an isotropic incompressible hyperelastic solid.
Several important papers by Rivlin, containing these solutions, are repro-
duced in a volume edited by Truesdell [4].
168
7.2 Strain Energy Functions and Stress-Strain Relations 169
7.2 Strain Energy Functions and Stress-Strain Relations
Solutions are given in this chapter for an incompressible solid based on the
Mooney-Rivlin strain energy function,
l
wðI1 , I2 Þ ¼ fcðI1 3Þ þ ð1 cÞðI2 3Þg, ð7:1Þ
2
where l is the shear modulus for infinitesimal deformation from the natural
reference state and 0 < c < 1 is a constant, or its special case with c ¼ 1, the
neo-Hookean strain energy function
l
wðI1 Þ ¼ ðI1 3Þ: ð7:2Þ
2
It has been found experimentally that, for simple tension, equations (7.1),
with c = 0.6, and (7.2) give results in good results for rubber-like materials
for axial stretches up to about 3.0 and 1.8, respectively. For larger stretches
a strain energy function of the form w ¼ w ~ ðk1 , k2 , k3 Þ, which is not readily
put in the form (6.24), gives results for simple tension that are in good
agreement with experiment. This strain energy function is considered in
the final section of this chapter.
Other strain energy functions for isotropic incompressible solids are dis-
cussed in [5].
The Hadamard strain energy function
l
wðI1 , I2 , I3 Þ ¼ fcðI1 3Þ þ ð1 cÞðI2 3Þ þ H ðI3 Þg, ð7:3Þ
2
where 0 < c < 1, H ð1Þ ¼ 0 and 3 c þ H9ð1Þ ¼ 0, is a compressible gen-

eralization of the Mooney-Rivlin. A compressible generalization of the
neo-Hookean strain energy function (7.2),
m
l 1 2m 12m
wðI1 , I3 Þ ¼ ðI1 3Þ þ I3 1 , ð7:4Þ
2 m
has been proposed by Blatz and Ko [6]. In equation (7.4), m is Poisson’s ratio
for infinitesimal deformation from the natural reference state, and as
170 Some Problems of Finite Elastic Deformation
m ! 0:5 and I3 ! 1, (7.4) ! ð7:2Þ. Several other compressible generaliza-

tions of equations (7.1) and (7.2) have been proposed [7].
Hyperelastic stress-deformation relations used in this chapter are
@w @w
s ¼ pI þ 2B 2B1 ð7:5Þ
@I1 @I2
for an incompressible hyperelastic solid and

2 @w^ @w^ @w^ @w^ 1
s ¼ pffiffiffiffi I2 þ I3 Iþ B I3 B ð7:6Þ
I3 @I2 @I3 @I1 @I2
for a compressible hyperelastic solid.
7.3 Simple Shear of a Rectangular Block
We consider a block that is rectangular in the undeformed reference con-

figuration with three sides coinciding with the coordinate axes as indicated
in Figure 7.1.
The block is subjected to a simple shear deformation as indicated by the
dashed lines in Figure 7.1 and the deformation field is homogeneous and
given by
x1 ¼ X1 þ kX2 , x2 ¼ X2 , x3 ¼ X3 , ð7:7Þ
where the angle of shear is h ¼ arctan k. It follows that the matrices of

components of F, B, and B1 are
2 3
1 k 0
6 7
½F ¼ ½Gradx ¼ 4 0 1 0 5,
0 0 1
2 3 ð7:8Þ
h i 1 þ k2 k 0
6 7
½B ¼ FF T ¼ 4 k 1 0 5,
0 0 1
7.3 Simple Shear of a Rectangular Block 171
X2
θ
X1
X3
Figure 7.1. Simple shear of a rectangular block.
2 3
h i 1 k 0
1
B ¼ 4 k k2 þ 1 0 5,
0 0 1
respectively.
Simple shear is a plane deformation and it is easily verified that
I1 ¼ I2 ¼ 3 þ k2 .
First we consider an incompressible solid and use equations (7.8) and con-
stitutive equation (7.5) to obtain
@w @w
r11 ¼ p þ 2 1 þ k2 2 ,
@I1 @I2
@w @w
r22 ¼ p þ 2 2 1 þ k2 ,
@I1 @I2
ð7:9Þ
@w @w
r33 ¼ p þ 2 2 ,
@I1 @I2

@w @w
r12 ¼2 þ k.
@I1 @I2
It is clear that the normal stresses cannot be obtained from the deformation
only, since the Lagrangian multiplier p depends on a stress boundary con-
dition. If r33 ¼ 0, equations (7.9) give

2@w @w
2 @w @w
r11 ¼ 2k , r22 ¼ 2k , r12 ¼2 þ k, ð7:10Þ
@I1 @I2 @I1 @I2
and for the Mooney-Rivlin strain energy function (7.1)
r11 ¼ lck2 , r22 ¼ lð1 cÞk2 , r12 ¼ lk: ð7:11Þ
This analysis proceeds in the same way if r11 or r22 is taken as zero. It is
evident from equations (7.10) and (7.11) that, in order to maintain the finite
simple shear given by equation (7.7), normal stress components r11 and r22
are required for the Mooney-Rivlin and r11 for the neo-hookean Hookean,
in addition to r12 . This is a second-order effect that is a form of the Poynting

effect [8]. The normal components r11 and r22 are O k2 for the Mooney-
Rivlin solid, whereas there is a linear relation between r12 and k: The
relation between r12 and k is nonlinear for materials for which
@w=@I1 and or @w=@I2 are functions of k.
We now consider the compressible solid and use the constitutive equa-
tion (7.6). The stress components are now completely determined by the
deformation equation (7.7) and for strain energy function (7.4) are
obtained from equations (7.6) and (7.8) as
r11 ¼ lk2 , r22 ¼ r33 ¼ 0, r12 ¼ lk: ð7:12Þ
It is interesting to note that these components do not depend on c and are

the same as those for equation (7.11) with c ¼ 1, that is, for the neo-
Hookean case with r33 ¼ 0:
7.4 Simple Tension
When an isotopic elastic solid is subjected to simple tension, due to

a tensile stress r11 ¼ r and all other rij ¼ 0, the resulting homogeneous
deformation is
x1 ¼ k1 X1 , x2 ¼ k2 X2 , x3 ¼ k3 X3 , k2 ¼ k3 , ð7:13Þ
7.4 Simple Tension 173
where k1 > 1 for simple tension and 0 < k1 < 1 for simple compression. It
follows from equation (7.13) that the matrix of components of the left
Cauchy-Green strain tensor is given by

½B ¼ diag k21 , k22 , k22 , ð7:14Þ
and the principal invariants of B are
2 2 2 2 4 2 4
I1 ¼ k1 þ 2k2 , I2 ¼ 2k1 k2 þ k2 , I3 ¼ k1 k2 : ð7:15Þ
First we consider a compressible solid. The following relations for the

stresses are obtained from equations (7.6), (7.14), and (7.15),

2k1 @w 2 @w @w
r11 ¼ 2 þ 2k2 þ 2k1 k22 , ð7:16Þ
k2 @I 1 @I2 @I 3

2 @w 2 @w @w
r22 ¼ r33 ¼ þ k1 þ k22 þ 2k1 k22 ¼ 0: ð7:17Þ
k1 @I1 @I2 @I3
The lateral stretch k2 is obtained from equation (7.17), in terms of the

extension stretch k1 ; however, the form of wðI1 , I2 , I3 Þ must be known in
order to do this [9]. Substitution of k2 in equation (7.16) gives r11 as a
function of the stretch k1 .
In order to illustrate this, strain energy function (7.4) is considered, so that
@w l @w @w l 2 4 ð1mÞ=ð12mÞ
¼ , ¼ 0, ¼ k k ,
@I1 2 @I2 @I3 2 1 2
which is substituted in equation (7.17). The resulting equation with given

values of k1 and m must be solved numerically for k2 , which is then
substituted in equation (7.16) to obtain r11 .
Numerical results for k1 ¼ 1:25 and m ¼ 0:49, which is a realistic value
for certain rubber-like materials [10], are
r11
k2 ¼ 0:89643, ¼ 0:75552:
l
The nominal stress for simple tension is given by R11 ¼ r11 k22 .
The problem for an incompressible solid is simpler since the incom-
pressibility condition I3 ¼ 1 gives
1=2
k2 ¼ k3 ¼ k1 :
PROBLEM. Show that, for simple tension of an incompressible solid,
@w 1 @w
2 1
r11 ¼ 2 k1 k1 þ ,
@I1 k1 @I2
and in particular for a Mooney-Rivlin solid,
ð1 cÞ

r11 ¼ l k21 k1
1 c þ
k1
and
N ð1 cÞ

¼ k1 k2
1 cþ ,
pA2 l k1
where N is the axial force and A is the radius in the undeformed reference
configuration.
For simple tension of the neo-Hookean solid with k1 ¼ 1:25 it follows
from the results of this problem that
r11
k2 ¼ 0:89443, ¼ 0:7625,
l
which differ negligibly from the corresponding values for the compressible
case already considered. This indicates that, for static problems of rubber-
like materials, the incompressible model is justified unless a hydrostatic
compressive stress of the order of magnitude of the bulk modulus is applied.
The Mooney-Rivlin strain energy function and its special case, the
neo-Hookean, give results, for simple tension, with stretch less than about
two that are in good agreement with experiment. For stretches greater than
about three experimentally obtained nominal stress-stretch curves for most
7.5 Extension and Torsion of an Incompressible Cylindrical Bar 175
rubbers are S shaped with an inflexion point at about k ¼ 3. This cannot be

predicted by the Mooney-Rivlin strain energy function. Ogden [5] has pro-
posed a three-term strain energy function, that is capable of modeling
nominal stress vs. stretch curves for large stretches in simple tension,
3 li ai a a
wðk1 , k2 , k3 Þ ¼ + k1 þ k2i þ k3i 3 , k1 k2 k3 ¼ 1; ð7:18Þ

i¼1 ai
where
3
+ li ai ¼ 2l,
i¼1
and the ai and li are chosen to fit experimental data. For simple.tension of
pffiffiffi
an incompressible solid the axial stretch k1 ¼ k and k2 ¼ k3 ¼ 1 k and it
follows from equation (7.18) that the axial tensile nominal stress is given by
dwðkÞ 3
R¼ ¼ + li kai 1 kai =21 : ð7:19Þ
dk i¼1
Ogden [5] has shown that equation (7.19) gives a close agreement with
experimental data for several rubbers for stretches up to seven when
l1 =l ¼ 1:491, l2 =l ¼ 0:003, l3 =l ¼ 0:0237,

ð7:20Þ
a1 ¼ 1:3 a2 ¼ 5:0 a3 ¼ 2:
Stress-stretch relations, obtained from equations (7.19) and (7.20), and the
Mooney Mooney-Rivlin strain energy function with c ¼ 1 (neo-Hookean)
and c ¼ 0:6, are shown graphically in Figure 7.2.
7.5 Extension and Torsion of an Incompressible

Cylindrical Bar
In this section we consider the extension and torsion of an incompressible

right circular cylinder. Cylindrical polar coordinates are used and the anal-
ysis is based on the physical components of the stress and deformation
gradient tensors. The deformation is a controllable deformation.
Figure 7.2. Nominal stress-stretch curves for simple tension.
The z axis of the cylindrical polar system is chosen to coincide with the
axis of the cylinder and the origin is taken at one end of the cylinder, and the
cylindrical polar coordinates of a particle in the spatial and undeformed
reference configurations are denoted by ðr, h, zÞ and ðR, H, ZÞ, respec-
tively, and the corresponding radii of the cylinder are r ¼ a and R ¼ A: De-
formation of the cylinder is a pure torsion with plane cross sections remaining
plane followed by a homogeneous extension or vice versa, and is given by,
1=2
r ¼ Rk , h ¼ H þ DZ, z ¼ kZ, ð7:21Þ
where D is the angle of twist per unit length of the reference configuration.
It is reasonable to expect that a stress system equivalent to a torque T and
axial force N is required to maintain the deformation (7.21).
The physical components of the deformation gradient tensor, left
Cauchy-Green strain tensor B, and B1 are
2 3
k1=2 0 0
6 7
½F ¼ 4 0 k1=2 rD 5,
0 0 k
2 1
3 ð7:22Þ
h i k 0 0
T 6 1 7
½B ¼ FF ¼ 4 0 k þ r 2 D2 krD 5,
0 krD k2
2 3
h i k 0 0
1 4
B ¼ 0 k rD 5,
2 1 2 2
0 rD k þk r D
respectively.
A solution is now obtained and particularized for the Mooney-Rivlin
material. The procedure given can be adapted to consider any incompress-
ible hyperelastic material, although more complication results when
@w=@I1 and=or @w=@I2 are not constant.
Expressions that involve the Lagrangian multiplier p are obtained
from equations (7.1), (7.5), and (7.22) for the nonzero components of the
Cauchy stress,

1 @w @w
rr ¼ p þ 2 k k , ð7:23Þ
@I1 @I2

1 2 2 @w @w
rh ¼ p þ 2 k þ r D k , ð7:24Þ
@I1 @I2

2 @w 2 1 2 2 @w
rz ¼ p þ 2 k k þk r D , ð7:25Þ
@I1 @I2

@w @w
rhz ¼ rD k þ : ð7:26Þ
@I1 @I2
The only nonzero trivial equilibrium equation is
drr rr rh
þ ¼ 0: ð7:27Þ
dr r
Substitution of equations (7.23) and (7.24) in equation (7.27) gives
drr 2 @w
¼ 2rD ,
dr @I1
which, along with the boundary condition rr ðaÞ ¼ 0, gives
Za
@w
rr ¼ 2 rD2 dr. ð7:28Þ
@I1
r
then p is obtained from equations (7.23) and (7.28) and this enables it to be
eliminated from equations (7.24) and (7.25).
It is clear from the analysis so far that equation (7.21) is a controllable
deformation. If the Mooney-Rivlin strain energy function is considered, the
solution is relatively simple since
@w=@I1 ¼ lc=2 and @w=@I2 ¼ lð1 cÞ=2
are constants and it then follows from equation (7.23) that

@p=@r ¼ @rr =@r. The following results are for the Mooney-Rivlin strain
energy function:
p cD 2
2
¼ a r2 þ ck1 ð1 cÞk,
l 2
rr cD2 2
¼ a r2 , ð7:29Þ
l 2
rh cD2 2 2

¼ a 3r , ð7:30Þ
l 2
rz nc o cD2 a2
¼ D2 r2 ð1 cÞk1
l 2 2 ð7:31Þ

2
þ ð1 cÞ k k þ c k k1 ,
2
rhz
¼ fcðk 1Þ þ 1grD: ð7:32Þ
l
The torque T required is
Za
T ¼ 2p r2 rhz dr,
0
and equation (7.32) and the incompressibility condition, a2 ¼ k1 A2 , give

the nondimensional form,
T
D
3 ¼ 4 ½cðk 1Þ þ 1, ð7:33Þ
plA 2k
¼ DA.
where D
Similarly, the axial force required is given by
Za
N ¼ 2p rrz dr,
0
and with the use of equation (7.28) this gives the nondimensional form
( ! )
N 2
D k2
3 3 3 2
¼ c k k þ ð1 cÞ 1 k þc kk :
plA2 2 2
ð7:34Þ
Equation (7.33) shows that, for the Mooney-Rivlin material, the relation
between torque and angle of twist per unit length is linear for a fixed value
of k. Equation (7.34) shows that if the length of the bar is constrained with
k ¼ 1 an axial force given by
N 2c
D
2 ¼
plA 4
is required, and if the axial force is zero, the length increases. This is a
second-order effect, known as the Poynting effect [8], which does not exist
in linear elasticity theory.
7.6 Spherically Symmetric Expansion of a Thick-Walled Shell
A thick-walled incompressible hyperelastic spherical shell is considered

and a solution is obtained for the Mooney-Rivlin strain energy function.
This solution is similar to a more general solution given by Green and
Zerna [3]. A spherical polar coordinate system with origin at the center
of the shell is used and the deformation is given by
r ¼ f ðRÞ, h ¼ H, / ¼ U ð7:35Þ
where r, h, and / are the spherical polar coordinates as defined in Appen-

dix 1.
As usual in this text lower- (upper-) case letters refer to the spatial
(reference) configurations. The inner radii in the undeformed natural ref-
erence configuration and spatial configuration are A and a, respectively,
and the corresponding outer radii are B and b.
Nonzero components of the deformation gradient tensor are identical

to the principal stretches kr , kh , k/ where it follows from incompressibil-

ity that kr kh k/ ¼ 1. It may be deduced from equation (7.35) that
kh ¼ k/ ¼ r=R and from the incompressibility condition that kr ¼ ðR=rÞ2 .
The components of B and B1 are then given by
h i h i h i
½B ¼ diag Q4 , Q2 , Q2 , B1 ¼ diag Q4 , Q2 , Q2 , ð7:36Þ
where Q ¼ R=r. Expressions, involving the Lagrangian multiplier p, for the

stresses,

4 @w 4 @w
rr ¼ p þ 2 Q Q , ð7:37Þ
@I1 @I2

2 @w 2 @w
rh ¼ r/ ¼ p þ 2 Q Q , ð7:38Þ
@I1 @I2
7.6 Spherically Symmetric Expansion of a Thick-Walled Shell 181
are obtained from equations (7.5) and (7.36). The nontrivial equilibrium
equation is
drr 2ðrr rh Þ
þ ¼ 0, ð7:39Þ
dr r
and it follows from the incompressibility condition r3 a3 ¼ R3 A3 that
dQ 1 2
¼ Q Q :
dr r
drr drr dQ
Substituting equations (7.37), (7.38), and ¼ in equation (7.39)
dr dQ dr
gives
@w @w
drr
¼ 4 Q4 Q2 Q4 Q2 Q2 Q ð7:40Þ
dQ @I1 @I2
Equation (7.40) can be integrated and the constant of integration deter-

mined by a boundary condition. This integration is simplified if
@w=@I1 and @w=@I2 are constants as in the Mooney-Rivlin strain energy
function (7.1), then equation (7.40) becomes
drr n o.
4 2 4 2 2
¼ 2l Q Q c Q Q ð1 cÞ Q Q : ð7:41Þ
dQ
A solution is now obtained for the Mooney-Rivlin strain energy function.

The integral of equation (7.41) is
4 !
rr Q Q2 1
¼ þ 2Q c þ 2ð1 cÞ Q þ K, ð7:42Þ
l 2 2
where K is a constant of integration that is evaluated from the boundary

conditions
rr ðQi Þ ¼ Pi and rr ðQo Þ ¼ P0 , ð7:43Þ
where Qi ¼ A=a and Qo ¼ B=b and Pi and P0 are the internal and
external pressures, respectively. Since the shell is assumed to be
incompressible the deformation due to equation (7.43) is the same as that

due to
rr ðQi Þ ¼ Pî and rr ðQo Þ ¼ 0, ð7:44Þ
where Pî ¼ Pi Po . Consequently, the solution is obtained for boundary

conditions (7.44). From equations (7.42) and ð7:44Þ1
!
Qi 4 Q2i 1 Pî
þ 2Qi c þ 2ð1 cÞ Qi þK ¼ , ð7:45Þ
2 2 l
Eliminating K from equations (7.42) and (7.45) gives

8 9
< 4 4 =
rr Q Q i
¼c þ 2ðQ Qi Þ
l : 2 ;
80 19 ð7:46Þ
< Q2 Q2i = P^
þ 2ð1 cÞ @ Q1 Q1 A i,
i
: 2 ; l
and eliminating p from equations (7.37) and (7.38) to obtain rh rr and

adding rr from equation (7.46) gives
8 9
< 4 4 =
rh r/ Q Q i
¼ ¼ c Q2 þ 2ðQ Qi Þ
l l : 2 ;
8 9 ð7:47Þ
< Q2 Q2i = P^
i
þ 2ð1 cÞ Q1 Q1i :
: 2 ; l
It is desirable to have a relation between Qi ¼ A=a and Pî for a given value
of g ¼ A=B. It follows from equations ð7:44Þ2 and (7.46) that
8 9
Pî < Q4o Q4i =
¼c þ 2ðQ0 Qi Þ
l : 2 ;
8 9 ð7:48Þ
< Q2o Q2i =
1 1
þ 2ð1 cÞ Qo Qi :
: 2 ;
7.6 Spherically Symmetric Expansion of a Thick-Walled Shell 183
3
The incompressibility condition B3 A ¼ b3 a3 gives, after some
manipulation, the relation
Qi
Q0 ¼ n o1=3 : ð7:49Þ
3 3 3
g þ 1 g Qi
Elimination of Qo from equations

. (7.48) and (7.49) gives the non
^
dimensional relation between Pi l and a=A ¼ Q1 for a given value of
1
g. This relation is shown graphically in Figure 7.3 for the Neo-Hookean
model ðc ¼ 1Þ and g ¼ 0:5.
Instability of the shell occurs when
d Pî l
¼ 0:
dða=AÞ
For the above data the onset of instability occurs for

Pî l ¼ 0:8164 and a=A ¼ 1:828:
A thin-walled spherical shell is a limiting case of the problem just

considered. For a thin-walled shell, the wall thickness in the undeformed
reference configuration is T ¼ B A ¼ Aðg1 1Þ < < A and rh ¼
r/ >> rr ðaÞ ¼ Pî . The variation rh ¼ r/ is assumed to be constant across
Figure 7.3. Nondimensional internal pressure-internal radius relation for thick-walled

Neo-Hookean shell with g ¼ 0:5.
the wall thickness and rr ðrÞ are neglected, and the elementary approximate
relation
2rh t
Pî ’ ð7:50Þ
a
is obtained, where t ¼ b a is the wall thickness in the spatial configura-

tion. It also follows from the approximations made that
a
Q1 ’ ,
A
and the thin-walled shell is approximately in a state of equibiaxial stress so

that for the neo-Hookean solid
l
w ¼ ð2k2 þ k4 Þ,
2
pffiffiffiffiffiffiffiffiffi
where, for the thin-walled shell problem, k ’ a=A. Then, it may be
deduced that
1 dwðkÞ
rh ¼ ;
2 dk
which gives

rh a 2 a 4
’ : ð7:51Þ
l A A
It follows from the incompressibility condition that t=T ’ ða=AÞ2 and
t t T a 3 1
¼ ¼ g 1 : ð7:52Þ
a T a A
Substituting equations (7.51) and (7.52) into equation (7.50) gives
Pî a 1 a 7

1
¼2 g 1 : ð7:53Þ
l A A
.
In Figure 7.4, the relation between Pî l and a=A is shown graphically for
the neo-Hookean model and g ¼ 0:975. The dashed curve is obtained
from equation (7.53) and the solid curve is obtained from equations
(7.48) and (7.49). It is evident that, for thin-walled spherical shell, the
approximate theory is in close agreement with the theory for the thick-
walled shell.
7.7 Eversion of a Cylindrical Tube
Eversion of a hyperelastic circular cylindrical tube is the turning inside out.

This is an example of a nonunique solution for the deformation of a hypere-
lastic solid subjected to prescribed surface tractions, in this case zero sur-
face tractions. A formal solution of this eversion problem has been given by
Rivlin [11], and later by Wang and Truesdell [12]. The solution given in [12]
involves two complicated nonlinear algebraic simultaneous equations.
According to [12] it is not known whether or not these equations have an
admissible solution except for the Mooney-Rivlin strain energy function or
its special case, the neo-Hookean. The solution given here for the neo-
Hookean solid is essentially that given in [12] and results for a particular
example are given.
The inside and outside radii of the tube in the (reference, spatial)
configuration are denoted by ðA, aÞ and ðB, bÞ, respectively, and the
Figure 7.4. Comparison of results for g ¼ 0:975 obtained from thick-walled theory and
approximate thin-walled theory.
eversion process is shown schematically in Figure 7.5 for length ðL, lÞ of

a infinite tube in the (reference, spatial) configuration. For a finite tube it is
assumed that it is sufficiently long that warping of the ends can be neglected
and the resultant axial force is zero. It follows that the deformation field is
given by
r ¼ f ðRÞ, h ¼ H, z ¼ kZ, ð7:54Þ
where k ¼ l=L, and the physical components of the deformation gradient

tensor are given by
½F ¼ diag ½dr=dR, r=R, k: ð7:55Þ
Since incompressibility is assumed, det½F ¼ 1, and it follows from equation

(7.55) that
dr R
¼ : ð7:56Þ
dR kr
Integration then gives
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
r ¼ aR2 þ b, ð7:57Þ
where a ¼ k1 < 0 and b is a constant of integration.
0 0′
0′ 0
a B
b A
r R
l L
Figure 7.5. Eversion of a cylindrical tube.
It is useful to introduce nondimensional quantities by dividing quan-

tities with dimension of length by A and stresses by l, so that
ðr, RÞ b ðrr , rh , rz Þ
ðr, RÞ ¼ , b ¼ 2 , ðrr , rh , rz Þ ¼ ,
A A l
¼ 1, B
A B ¼ ¼ g:
A
Henceforth nondimensional quantities are used but the superposed bars

are omitted for convenience. Components
" #
2
2R r2 2
½B ¼ diag a , ,a ð7:58Þ
r2 R2
of the left Cauchy-Green tensor, B ¼ FF T , are obtained from equations

(7.55) and (7.56). The nonzero components of Cauchy stress for the neo-
Hookean solid are
2
aR
rr ¼ p þ , ð7:59Þ
r
r 2
rh ¼ p þ : ð7:60Þ
R
rz ¼ p þ a2 : ð7:61Þ
Substitution of equations (7.59) and (7.60) in the nontrivial equilibrium

equation
@rr rr rh
þ ¼0
@r r
gives
Z !
a2 R2 r2 dr
rr ¼ 2 2 : ð7:62Þ
r R r
Since rr ðaÞ ¼ rr ðbÞ ¼ 0 the integral (7.62) with limits of integration a and b
is zero, that is,
Zb !
a 2 R2 r2 dr
2 2 ¼ 0: ð7:63Þ
r R r
a
Substitution of equation (7.57) in equation (7.63) gives the non

dimensional relation, referred to the reference configuration,
2 3
Zg 2 2
6 aR 17
4 2 2 5RdR ¼ 0: ð7:64Þ
2
aR þ b R
1
The everted tube is assumed to have no resultant axial force applied so that
Zb
2p rrz dr ¼ 0: ð7:65Þ
a
Using equations (7.57) and (7.59) to (7.61), it may be deduced from equa-
tion (7.65) that
Zg " #
a 2 R2 2 aR2 þ b
þ RdR ¼ 0: ð7:66Þ
aR2 þ b a2 R2
1
Equations (7.64) and (7.66) can be solved simultaneously to obtain a and b.

It appears to be very difficult to determine closed-form expressions for
a and b and a numerical determination for a given value of g is desirable.
A numerical example is now considered for the neo-Hookean solid
with g ¼ 1:15. Then, a numerical solution of equations (7.64) and (7.66)
obtained using Mathematica gives
a ¼ 0:995729, b ¼ 2:32722:
Since k ¼ a1 > 1 the tube lengthens when it is inverted. The non
dimensional radii in the deformed configuration
a ¼ 1:1539 and b ¼ 1:00517
are then obtained from equation (7.57). It is easily verified that the
isochoric condition

g2 1 ¼ a1 a2 b2
is satisfied.
The Mooney-Rivlin solid can be considered with some additional
complication.
7.8 Pure Bending of an Hyperelastic Plate
Pure bending of an incompressible isotropic hyperelastic plate is a further

example of a controllable deformation. Cuboid, rather than plate, is the
term used by Rivlin [11], who appears to be the first to have considered this
problem that has also been considered by Ogden [5].
The natural reference configuration of the plate is bounded by the
planes
X1 ¼ S1 , X1 ¼ S2 , X2 ¼ 6L, X3 ¼ 6W=2
and
S2 S1 ¼ H:
It is assumed that the plate is subjected to pure bending by normal stresses,

statically equivalent to couples, on faces X2 ¼ 6L, and it is bent into a
sector of a hollow circular cylinder with inner and outer radii a and b;
respectively, as indicated in Figure 7.6. The bent deformed configuration
X2
r L
a b S1 S2
0 θ
α X1
H
Figure 7.6. Finite deformation pure bending of hyperelastic plate.
is described by the cylindrical polar coordinates ðr, h, zÞ and is bounded by

surfaces
r ¼ a, r ¼ b, h ¼ a, h ¼ a, z ¼ w, z ¼ w:
For the pure bending, planes normal to the X2 axis in the undeformed
reference configuration are plane in the deformed configuration and are
in meridian planes containing the X3 axis. Also the stretch, k, in the
X3 direction is uniform, so that z ¼ kX3 , k ¼ w=W; which implies that
anticlastic bending is constrained.
A material plane X3 ¼ K, where S1 < K < S2 is a constant, becomes
a sector of a cylinder of radius r subtending the angle 2a at the center. It
then follows that the circumferential stretch in the spatial configuration is
given by
ra
kh ¼ : ð7:67Þ
L
Since the solid is incompressible the radial stretch is given by
L
kr ¼ : ð7:68Þ
rak
And since the r, h, and z directions are principal directions, the physical
components of the left Cauchy strain tensor are given by
" #
L2 r2 a2 2
½B ¼ diag 2 2 2 , 2 ,k : ð7:69Þ
kr a L
A further relation

2 2
L k b a
¼ ð7:70Þ
a 2H
is obtained from the incompressibility condition. Relations (7.67) to (7.70)

are valid for any isotropic incompressible strain energy function.
A solution is now outlined for the neo-Hookean material and solutions
can readily be obtained for other isotropic hyperelastic solids, especially
the Mooney-Rivlin, with some additional complication.
The nonzero physical components of Cauchy stress, obtained from
equations (7.5) and (7.68) for the neo-Hookean solid, are given by
rr p L2
¼ þ 2 2 2, ð7:71Þ
l l kr a
2 2
rh p r a
¼ þ 2 , ð7:72Þ
l l L
rz p
¼ þ k2 , ð7:73Þ
l l
to within the Lagrangian multiplier p.

Substitution of equations (7.71) and (7.72) in the equilibrium equation
(7.27)
drr ðrr rh Þ
þ ¼0
dr r
gives
" #,
1 drr L2 r2 a2
¼ 2 2 2 2 r,
l dr kr a L
so that, upon integration,
" #
rr 1 L 2 r2 a2
¼ þ 2 þ c1 , ð7:74Þ
l 2 k2 r2 a2 L
where c1 is a constant of integration. It then follows from the boundary

conditions,
rr ðaÞ ¼ 0, rr ðbÞ ¼ 0,
that
! !
2 2 2 2 2 2
1 L b a 1 L aa
c1 ¼ 2 2 2þ ¼ 2 2 2þ : ð7:75Þ
2 kb a L2 2 kaa L2
The stress rr is obtained by eliminating c1 from equations (7.74)

and (7.75). Solving the second equation of equation (8.75) for L2 a2
gives
2
L
¼ kab: ð7:76Þ
a2
If k and either a or one of the radii of the cylindrical surfaces are given,
equations (7.70) and (7.76) determine the deformation.
!
p 1 r2 a2 L2
¼ þ c1 , ð7:77Þ
l 2 L2 k2 r2 a2
and then from equations (7.72) and (7.73) that
!
rh 1 r2 a2 L2
¼ 3 2 2 2 2 þ c1 , ð7:78Þ
l 2 L kr a
!
2 2 2
rz 1 r a L
¼ k2 þ 2 2 2 2 þ c1 , ð7:79Þ
l 2 L kr a
where c1 is given by equation (7.75).

The resultant force acting on the surfaces z ¼ 6w is given by
Zb
Fz ¼ 2a rrz dr, ð7:80Þ
a
and the distribution of rh acting on surfaces h ¼ 6a is statically equivalent

to the couple
Zb
Mz ¼ 2w rrh dr: ð7:81Þ
a
The length of a material line parallel to the X2 axis in the reference con-
figuration is 2L, and in the spatial configuration, this line is bent into an arc
of a circle of radius r and length ra. If k ¼ 1 and the length of the line are
unchanged due to the deformation, it follows from equation (7.76) that
r ¼ r0 where r0 ¼ ðabÞ1=2 , and then from equation (7.68) that the principal
stretches on surface r ¼ r0 are ð1, 1, 1Þ. Consequently, on the surface r ¼ r0
the stress is isotropic and from equations (7.74) to (7.76), (7.78) and (7.79) is
given by
rr ðr0 Þ=l ¼ rh ðr0 Þ=l ¼ rz ðr0 Þ=l ¼ 1 þ c1 : ð7:82Þ
Numerical examples are now given to illustrate the solution for k ¼ 1.

Nondimensional quantities are used and are denoted by a superposed bar.
Quantities with dimension of length and stress are nondimensional-

ized by dividing by L and l, respectively, so that
. .
¼ H=L,
r ¼ r=L, r ¼ r l, H ¼ W=L,
W F ¼ F lL2 , ¼ M lL3 :
M
The nondimensional forms of equations (7.70) and (7.76), with k ¼ 1, are

a b2 a2
¼ 1 and a2 ab ¼ 1, ð7:83Þ

2H
The
respectively, which are solved simultaneously to obtain a and b.
nondimensional resultant force,
Zb
1n 2 2
o
Fz ¼ 2a r 1 þ rÞ ða
ða rÞ þ c1 d
r, ð7:84Þ
2
a
acting on surfaces z ¼ 6w is obtained from equations (7.79) and (7.80) and

the distribution of rh acting on surfaces h ¼ 6a is statically equivalent to
the nondimensional couple per unit width of the reference configuration,
Zb
1 2 2
Mz =2W ¼ r ð
3 aÞ ð
r ar Þ þ c1 d
r, ð7:85Þ
2
a
obtained from equations (7.78) and (7.81). The stress distribution rz ðrÞ at
z ¼ 6w is statically equivalent to the force Fz and a couple that prevents
anticlastic bending.
Numerical results are now given in nondimensional form, for two sets
of data,
¼ 0:1, a ¼ 0:8
ðaÞ H ¼ 0:2, a ¼ 1:2:
and ðbÞ H
For (a),
a ¼ 1:20104, b ¼ 0:130096, r0 ¼ 1:25,

Fz ¼ 2:1315 3 104 , Mz =2W ¼ 2:66185 3 104 :
For (b)
a ¼ 0:73993, b ¼ 0:938526, r0 ¼ 0:82768,

Fz ¼ 3:8075 3 103 , Mz =2W ¼ 3:14635 3 103 :
It is interesting to compare the above results for Mz with those obtained

from linear bending theory. For the linear theory applied to the problem
just considereded it follows from the theory given in [13] that Mz =2W is
given by
EH 3 1
Mz =2W ¼ , ð7:86Þ
2 r0
12 1 m
where E ¼ 2lð1 þ mÞ is Young’s modulus, m is Poisson’s ratio, and r0 is

radius of curvature of the mid plane. The nondimensional form of equation
(7.86) with m ¼ 0:5 is
H3
Mz =2W ¼ ;
3
r0
which gives
4
Mz =2W ¼ 2:667 3 10
for (a) and
Mz =2W ¼ 3:223 3 103
for (b). It can be shown that the difference between the values of Mz =2W
for linear and nonlinear theories increase as H and a increase.
7.9 Combined Telescopic and Torsional Shear
In this section we consider a deformation, which is not controllable for an

incompressible isotropic hyperelastic solid, although an inverse method of
solution of the problem of combined telescopic and torsional shear is ap-

plicable. The deformation is assumed to be for an incompressible circular
tube, bonded at its outer radius to a fixed rigid sleeve and at its inner radius
to a rigid bar that is subjected to prescribed displacement and twist as
indicated in Figure 7.7.
This deformation is given by
r ¼ R, h ¼ H þ aðRÞ, z ¼ Z þ wðRÞ, ð7:87Þ
in the usual notation for the cylindrical polar coordinates of a material

point in the deformed and undeformed states. Equations (7.87) follows
from the incompressibility assumption. The boundary conditions are
wð AÞ ¼ wi , að AÞ ¼ ai , ð7:88Þ
wðBÞ ¼ aðBÞ ¼ 0, ð7:89Þ
where A and B are the fixed inner and outer radii, respectively, of the tube.
The problem is to determine the nonzero components of Cauchy stress so
that the axial force and torque required per unit length can be obtained.
Instead of equation (7.78) a stress boundary condition
rrz ð AÞ ¼ s, rrh ¼ q
could be considered and then the problem would be to obtain

wðRÞ and aðRÞ. It is assumed that the system is sufficiently long that end
effects can be neglected.
Fixed
B A
Figure 7.7. Combined telescopic and torsional shear.

Deformation field (7.87) is not controllable since, in general, the func-

tions raðRÞ and wðRÞ depend on the strain energy function.
If aðRÞ ¼ 0 deformation field (7.87) represents telescopic shear, some-
times known as axial or antiplane shear, and if wðRÞ ¼ 0, equation (7.87)
represents torsional shear sometimes known as azimuthal shear.
The physical components of the deformation gradient F, left Cauchy-
Green tensor B, and B1 are
2 3
1 0 0
6 7
½F ¼ 4 ra9 1 0 5,
w9 0 1
2 3
1 ra9 w9
6 2 2 7
½B ¼ 4 ra9 1 þ r a9 ra9w9 5, ð7:90Þ
w9 ra9w9 1 þ w92
2 3
h i 1 þ w92 þ r2 a92 ra9 w9
6 7
B1 ¼ 4 ra9 1 0 5,
w9 0 1
where a prime denotes differentiation with respect to the argument, which

can be taken as r since r ¼ R.
A solution is outlined for the Mooney-Rivlin solid and the procedure
can be adapted to consider other isotropic incompressible hyperelastic
solids.
The components of the Cauchy stress tensor obtained from equations
(7.5) and (7.90) are
rr p
¼ þ c ð1 cÞ 1 þ w92 þ r2 a92 , ð7:91Þ
l l
rh p
¼ þ c 1 þ r2 a92 ð1 cÞ, ð7:92Þ
l l
rz p
2
¼ þ c 1 þ w9 ð1 cÞ, ð7:93Þ
l l
rrh
¼ ra9, ð7:94Þ
l
rrz
¼ w9, ð7:95Þ
l
rhz
¼ cra9w9: ð7:96Þ
l
It is interesting to note that rhz ¼ 0 when c ¼ 0 and for telescopic

shear or torsional shear alone. The stresses rrh and rrz do not depend on
the telescopic shear and torsional shear, respectively, so that there is partial
uncoupling; however, this is only true for the Mooney-Rivlin solid. Also
rrh and rrz do not depend on c.
The equilibrium equations are
@rr rr rh
þ ¼ 0, ð7:97Þ
@r r
@rrh 2rrh
þ ¼ 0, ð7:98Þ
@r r
@rrz rrz
þ ¼ 0, ð7:99Þ
@r r
where r can be replaced by R.

Combining equations (7.94) and (7.98) gives
d2 a 3 da
2 þ ¼ 0,
dr r dr
with solution
a ¼ c1 þ c2 r2 ,
where c1 and c2 are constants of integration that are evaluated from the
boundary conditions ð7:88Þ2 and ð7:89Þ2 to give

r2 B2
a ¼ ai : ð7:100Þ
A2 B2
Similarly combining equations (7.75) and (7.79) gives
2
dw 1 dw
2 þ ¼ 0,
dr r dr
with solution
w ¼ c3 ln r þ c4 ,
where c3 and c4 are constants of integration that are evaluated from the
boundary conditions ð7:84Þ1 and ð7:85Þ1 to give

r A
w ¼ wi ln ln : ð7:101Þ
B B
The shearing stresses are then obtained by substituting equations (7.100)

and (7.101) in equations (7.94) to (7.96). Since equations (7.100) and
(7.101) do not depend on the parameter c, the deformation is controllable
for all Mooney-Rivlin solids.
The normal stress rr is obtained by substituting equations (7.100) and
(7.101) in equations (7.91) and (7.92) and then substituting equations (7.91)
and (7.92) in equation (7.97) and integrating. It is assumed that rr ¼ 0 if
wi ¼ ai ¼ 0, so that the constant of integration is zero and the integral gives
8 9
>
< 4 >
rr w 2
a ðr=AÞ =
2
¼ ð1 cÞ 2 i 2 þ i 2 : ð7:102Þ
l >
: 2r ðln gÞ >
1g ;
2
Examination of equation (7.102) indicates that a limited super position

principle holds for rr , that is, rr due to wi applied alone, plus rr due to ai
applied alone is equal to rr due to wi and ai applied together. The form of
the Lagrangian multiplier p that appears in equations (7.92) to (7.93) can
be obtained from equations (7.91) and (7.102). The normal stresses
rr , rh , and rz are second-order effects for the problem considered. It
follows from equation (7.102) that rr ¼ 0 for telescopic shear of the
neo-Hookean solid.
EXERCISES
7.1. Solve the thick-walled spherical shell problem for the neo-Hookean
solid using the strain energy function in the form
^
w ¼ wðk1 , k2 , k3 Þ:
7.2. Solve the thick-walled axially symmetric cylindrical tube problem for
the neo-Hookean solid using the strain energy function in the form
^
w ¼ wðk1 , k2 , k3 Þ:
Assume that the z axis of a cylindrical polar system of axes coincides

with the axis of the tube and that the corresponding stretch is k3 ¼ 1.
7.3. A thin-walled compressible spherical shell whose strain energy func-
tion is given by equation (7.4) has wall thickness T and mean wall
radius R in the undeformed reference configuration. Find an approx-
imate expression for the internal pressure at the onset of instability.
7.4. A thick-walled incompressible tube has inner and outer radii A and B,
respectively, in the undeformed reference configuration. It spins about
its axis with angular velocity x. Find the relation between the de-
formed inner radius a and x, for the neo-Hookean strain energy func-
tion. Assume that the length is constrained to remain constant.
References 201
REFERENCES
1. Ericksen, J.L. (1954). Deformations Possible in Every Isotropic, Com-

pressible Perfectly Elastic Body. ZAMP 5, pp. 466–489.
2. Ericksen, J.L. (1955). Deformations Possible in Every Isotropic, Incom-
pressible Perfectly Elastic Body. J. Math. Phys. 34, pp. 126–128.
3. Green, A.E., and Zerna, W. (1968).Theoretical Elasticity, 2nd Ed. Oxford
University Press.
4. Truesdell, C., Editor. (1965). Continuum Mechanics IV Problems of Non-
Linear Elasticity, International Review Series. Gordon and Breach Sci-
ence Publishers.
6. Blatz, P.J., and Ko, W.L. (1962). Application of Elastic Theory to the De-
formation of Rubbery Materials. Trans. Soc. of Rheology 6, pp. 223–251.
7. Levinson, M., and Burgess, I.W. (1971). A Comparison of Some Simple
Constitutive Relations for Slightly compressible Compressible Rubberlike
Materials. Int. J. Mech. Sci. 13, pp. 563–572.
8. Poynting, J.H. (1909). On Pressure perpendicular to the Shear Planes in
Finite Pure Shears, and on the Lengthening of Loaded Wires When
Twisted. Proc. Roy. Soc. London A82, pp. 546–559.
9. Adkins, J.E. (1961). Large Elastic Deformations. Progress in Solid
Mechanics, Vol. II. Ed. I.N. Sneddon and R. Hill. North Holland.
10. Beatty, M.F., and Stalnacker, D.O. (1986). The Poisson Function of Finite
Elasticity. ASME J. Appl. Mech. 53, pp. 807–813.
11. Rivlin, R.S. (1949). Large Elastic Deformations of Isotropic Materials,
Part VI, Further Results in the Theory of Torsion Shear and Flexure. Phil.
Trans. Roy. Soc. A.242, pp. 173–195.
12. Wang, C.C., and Truesdell, C. (1973). Introduction to Rational Elasticity.
Noordhoff.
13. Timoshenko, S.P., and Woinowsky-Krieger, S. (1959). Theory of Plates
and Shells, 2nd Ed. McGraw-Hill.
8 Finite Deformation Thermoelasticity
8.1 Principle of Local State and Thermodynamic Potentials
The theory in this chapter is based on the Principle of Local State given in
chapter 4. Kestin [1] has noted that this principle implies that ‘‘the local and
instantaneous gradients of the thermodynamic properties as well as their
local and instantaneous rates of change do not enter into the description of
the state and do not modify the equations of state.’’ It follows that the
thermodynamic potentials are assumed to be of the same form as for an
equilibrium state.
The important potentials in finite deformation thermoelasticity are the
Helmholtz free energy and the internal energy, since, as for linear thermo-
elasticity, the Helmholtz free energy at constant temperature is the iso-
thermal strain energy and the internal energy at constant entropy is the
isentropic strain energy. This indicates the relation between hyperelasticity
and thermoelasticity.
Application of thermoelasticity to finite deformation of idealized
rubber-like materials involves the concept of entropic response and ener-
getic response and this is discussed in this chapter.
8.2 Basic Relations for Finite Deformation Thermoelasticity
The theory in this section is presented in terms of the absolute tem-

perature H, the specific entropy s, and the referential variables, S, the
second Piola-Kirchhoff stress, and C, the right Green tensor, or E, Green’s
strain tensor.
202
There are four thermodynamic potentials that are fundamental equa-

tions of state, that is, all the thermodynamic properties can be obtained from
a potential by differentiation if the potential is expressed as a function of the
appropriate thermodynamic properties. The potentials are related by Legen-
dre transformations and are the specific (per unit mass) internal energy,
u ¼ uðC, sÞ, ð8:1Þ
the specific Helmholtz free energy,
f ¼ f ðC, HÞ ¼ u Hs, ð8:2Þ
the specific enthalpy
SKL CKL
h ¼ hðS, sÞ ¼ u , ð8:3Þ
2q0
and the specific Gibbs free energy,
SKL CKL
g ¼ gðS,HÞ ¼ f : ð8:4Þ
2q0
Since C ¼ F T F ¼ 2E þ 1, equations (8.1) and (8.2) can be put in the al-

ternative equivalent forms
_
u ¼ u^ðF, sÞ, f ¼ f^ðF,HÞ,
_
u ¼ uðE, sÞ and f ¼ f ðE,HÞ,
and equations (8.3) and (8.4) in the equivalent forms
h ¼ h^ðR, sÞ and g ¼ g ðR,HÞ:
The thermodynamic relations based on u and f are important for what

follows in this chapter and are now given in terms of the conjugate prop-
erties, S, the second Piola-Kirchhoff stress, and, C, the right Cauchy-Green
strain tensor (or E the Lagrangian strain tensor). A Gibbs relation, based
on u, that can be deduced from the first and second laws of thermodynamics is
SKL dEKL
du ¼ þ Hds
q0
ð8:5Þ
SKL dCKL
¼ þ Hds:
2q0
From equation (8.1) or (8.5),
@u @u
du ¼ dCKL þ ds: ð8:6Þ
dCKL ds
Comparison of equations (8.5) and (8.6) gives
@uðC, sÞ @uðC, sÞ
SKL ðC, sÞ ¼ 2q0 , HðC, sÞ ¼ : ð8:7Þ
@CKL @s
A further Gibbs relation, based on f ,
S~KL ðC,HÞdCKL
df ¼ sdH, ð8:8Þ
2q0
follows from equations (8.2) and (8.5). It then follows from equation (8.2)
or (8.8) that
@f @f
df ¼ dCKL þ dH: ð8:9Þ
@CKL @h
Comparison of equations (8.8) and (8.9) gives
@f ðC,HÞ @f ðC, HÞ
S~KL ðC,HÞ ¼ 2q0 , sðC, HÞ ¼ : ð8:10Þ
@CKL @H
The determination of the Gibbs relations based on h and g is left as an

exercise at the end of the chapter.
From equations (8.2) and ð8:10Þ2 , the internal energy as a function of

C and H is given by
@f ðC,HÞ
u~ðC,HÞ ¼ f ðC,HÞ H .
@H
This is, a useful form, although it is not a fundamental equation of state.

It follows from the first law of thermodynamics that at constant deforma-
u ¼ cc dH ¼ Hds, where cc is the specific heat at constant C;
tion d~
consequently,
@ u~ @sðC, HÞ
cc ¼ ¼H : ð8:11Þ
@H @H
The following relation is obtained from equations (8.2), (8.10), and (8.11)

@ u~ @s
S~KL ðC,HÞ ¼ 2q0 2q0 H :
@CKL @CKL
This shows that the stress consists of two parts, one due to change of in-
ternal energy produced by deformation and the other due to the change of
entropy. The first part is described as energetic and the second as entropic.
Equations ð8:7Þ1 and ð8:10Þ1 indicate that the strain energy w is given
by
q0 uðC, s0 Þ ¼ wisentropic and q0 f ðC, H Þ ¼ wisothermal ,
for isentropic deformation and isothermal deformation, respectively, with

reference temperature and entropy H0 and s0 .
An explicit expression for f ðC, HÞ is obtained as follows. Combining
gives equations ð8:10Þ2 and equation ð8:11Þ2 gives
@ 2 f ðC,HÞ
cc ðC, HÞ ¼ H ,
@H2
and integrating once from H0 to H gives
ZH
@f @f cc ðC, H9Þ
¼ dH9: ð8:12Þ
@H @H H¼H0 H9
H0
Equations (8.2) and ð8:10Þ2 give
@f
f ¼ u~ þ H , ð8:13Þ
@H
and it can be deduced from this that
( )
@f 1
¼ f C; H0 u~ C, H0 ,
@H H¼H0 H0
and substitution in equation (8.12) gives
( ) ZH
@f 1 cc ðC, H9Þ
¼ f C, H0 u~ C, H0 dH9: ð8:14Þ
@H H0 H9
H0
Integration of equation ð8:11Þ1 from H0 to H gives
ZH
u~ðC, HÞ ¼ cc ðC, H9ÞdH9 þ u~ðC, H0 Þ: ð8:15Þ
H0
The relation

H H
f ðC, HÞ ¼ f ðC, H0 Þ 1 u~ðC, H0 Þ
H0 H0
ZH ð8:16Þ
H
1 cc ðC, H9ÞdH9
H9
H0
is then obtained by substituting equations (8.14) and (8.15) in equation

(8.13).
The Clausius-Duhem or entropy inequality (4.38) can be put in the
form
QK @H
_ þ S~KL C_ KL 2
q0 ðHs_ uÞ > 0, ð8:17Þ
H @XK
where Q is the referential heat flux that is related to the spatial heat flux, q,
by
q ¼ J 1 FQ:
The alternative form,
Qk @H
q0 sh_ þ f_ þ S~KL C_ KL 2 > 0, ð8:18Þ
H @XK
is obtained from equation (8.2). Substituting
@u _ @u
u_ ¼ CKL þ _
s,
@CKL @s
which follows from equation (8.6), in equation (8.17) gives the entropy
inequality for a thermoelastic solid,

1 @u _ @u QK @H
SKL q0 CKL þ q0 H s_ > 0: ð8:19Þ
2 @CKL @s h @XK
Substituting
@f _ @f _
f_ ¼ CKL þ H,
@CKL @h
that follows from equation (8.9), into equation (8.18) gives an alternative
form

1~ @f @f _ QK @H
SKL q0 C_ KL q0 s þ H > 0: ð8:20Þ
2 @CKL @H H @XK
Inequality (8.20) must hold for all values of C_ and H,_ and this gives an
alternative derivation of equations (8.10) since the factors of C_ and H
_ in
equation (8.20) must be zero. The same reasoning applied to equations
(8.19) gives an alternative derivation of equations (8.7).
It follows from equations (8.7), and (8.19) or (8.10) and (8.20) that
QK @H @H
c_ ¼ 2 > 0, or QK > 0, ð8:21Þ
q0 H @X K @X K
where c_ is the rate of entropy production per unit mass. As an exercise the
reader should show that the spatial forms of equations (8.21) are
qk @H @H
c_ ¼ 2 > 0 or qK > 0:
qH @xk @xk
Inequality ð8:21Þ2 is consistent with the Clausius statement of the second

law of thermodynamics that states that ‘‘heat can cannot pass spontane-
ously from a body of lower temperature to a body of higher temperature.’’
Entropy is produced in a thermoelastic solid by temperature gradients
resulting in heat conduction, a thermodynamically irreversible process,
with entropy production as indicated by equation (8.21), when shock waves
occur there is an entropy jump across a shock. There is no entropy pro-
duction in an ideal thermoelastic solid due to hysteresis effects.
Usually in elastostatic problems thermal effects are neglected. This
leads to consideration of the piezotropic model for which mechanical and
thermal effects are uncoupled. For this model,
uðC, sÞ ¼ u1 ðCÞ þ u2 ðsÞ,

and it may be deduced that the volume coefficient of thermal expansion is

zero for a piezotropic solid. The zero coefficient of thermal expansion
indicates that, in an exact sense, the piezotropic model is not physical;
however, its use is implied when thermal effects are neglected and the
hyperelastic model is assumed for static problems. Then, since the mechan-
ical effects are uncoupled it may be shown from conservation of mechanical
energy that the specific stress power is equal to the material time derivative
of the strain energy.
8.3 Ideal Rubber-Like Materials
Material in this section is influenced by important papers by Chadwick [2]

and Chadwick and Creasy [3] that should be referred to for a more detailed
treatment of the topic.
Most solid materials that can undergo finite elastic strain are polymers
or rubber-like materials, henceforth described as rubbers. These are com-
posed of entanglements of high–molecular weight chains. This is in contrast
to hard solids such as metals that are essentially crystalline. The stresses in
an ideal rubber result from changes in entropy, only, and these solids are
said to be strictly entropic. Most nonideal rubbers are not strictly entropic
but are partly energetic and are said to exhibit modified entropic elasticity
[3]. In particular this is true for compressible elastic solids. A suitable
model for a compressible isotropic rubber-like elastic solid has entropic
behavior for pure shear and energetic behavior for dilatation so that the
internal energy can be expressed in the form
u~ ¼ u~1 ð J Þ þ u~2 ðHÞ, ð8:22Þ
pffiffiffiffiffiffiffiffiffiffiffi
where J ¼ det½F ¼ detC . Henceforth, equation (8.22) is assumed in this
section.
Equations ð8:11Þ1 and (8.22) imply that the specific heat at constant
deformation is a function of temperature only or constant.
The bulk modulus K for infinitesimal deformation of most rubbers is
up to three orders of magnitude greater than the corresponding shear
modulus l; however, the volume coefficient of thermal expansion is about
three times a typical value for a metal. This has motivated a model that is
mechanically incompressible but has a nonzero coefficient of volume ther-
mal expansion [4]. We avoid the use of this model since it is thermodynam-
ically inconsistent. It may be reasonable in a thermoelastic problem
involving a rubber-like material to neglect the dilatation, due to mechanical
compressibility, compared with that due to volume thermal expansion and
this is essentially the same as assuming the mechanically incompressible
model with a nonzero coefficient of volume thermal expansion. However, it
is possible that the dilatation due to the isotropic part of the stress tensor
could be of the same order of magnitude as that due to thermal expansion
depending on the particular problem.
In this section we consider both incompressible and compressible
isotropic models for rubber-like solids and express the deformation in
terms of the principal stretches ki , i 2 f1, 2, 3g. In the present discussion
an incompressible solid is defined as a solid capable of isochoric deforma-
tion only so that the volume coefficient of thermal expansion is zero. This
model is thermodynamically consistent but not strictly physical. The tem-
perature in the natural reference configuration is denoted by H0 , and it is
assumed that jH H0 j=H0 < < 1 and jJ 1 < < 1j, for the problems
considered.
A physically realistic form for the isothermal strain energy function
per unit volume of the natural reference configuration, for an isotropic
compressible rubber rubber-like solid at temperature H0 is [2],
wðki Þ ¼ l/ðki Þ þ KgðJ Þ, ð8:23Þ
where l and K are the isothermal shear and bulk moduli, respectively, for
infinitesimal deformation from the natural reference configuration, / is
a nondimensional symmetric function of the principal stretches,
ki , i 2 f1, 2, 3g, which vanishes when k1 ¼ k2 ¼ k3 , and g is a nondimen-
sional function of J ¼ k1 k2 k3 that satisfies the conditions
gð1Þ ¼ 0, g9ð1Þ ¼ 0 and g0 ¼ 1, where a prime denotes differentiation with
respect to J. It may be shown that the compressible models given by equa-
tions (7.3) and (7.4) can be put in the form (8.22). For most rubbers the ratio
l=K is of order of magnitude 103 , and it follows that for finite distortion
the dilatation may be considered infinitesimal unless the isotropic part of
the Cauchy stress is very large compared with l, Ogden [5] has proposed an
empirical relation
K 1
p¼ J J 10 , ð8:24Þ
9
where p ¼ rkk =3, and, since, p ¼ K@gð J Þ=@K for hydrostatic

compression,
( )
1 J 9 1
g¼ þ ln J : ð8:25Þ
9 9 9
For jJ 1j of the same order of magnitude as for classical elasticity, equa-

tions (8.24) and (8.25) could be replaced by
p ¼ K ðJ 1Þ, ð8:26Þ
and
1
g ¼ ðJ 1Þ2 , ð8:27Þ
2
respectively, with negligible difference from equations (8.24) and (8.25) for
0:98 < J < 1:02. Consequently, equations (8.26) and (8.27) are assumed in
what follows.
Since, for the problems considered, it is assumed that
jH H0 j=H0 < < 1 the approximation cc ¼ constant is adopted, and from
equations ð8:11Þ1 and ð8:22Þ,
u~2 ¼ cc ðH H0 Þ,
where, as before, cc is the specific heat at constant deformation. Also equa-

tion (8.16) is then approximated in terms of the principal stretches, for
a rubber-like elastic solid, by

H H H
f ðki , HÞ ¼ f ðki , H0 Þ 1 u~1 ð J Þ cc H ln þ cc ðH H0 Þ:
H0 H0 H0
ð8:28Þ
It has been shown, from the statistical theory of rubber elasticity [6], that
an ideal rubber is incompressible, is strictly entropic, that is, the internal
energy can be expressed as a function u ¼ uðHÞ of temperature, only, as for
a perfect gas and has the Helmholtz free energy, at temperature H0 , given, by
l 2
w=q0 ¼ f ðki , H0 Þ ¼ k1 þ k22 þ k23 3 , k1 k2 k3 ¼ 1. ð8:29Þ
2q0
In equation (8.29), w is the neo-Hookean strain energy function; however,

the following theory can be applied to other admissible incompressible
strain energy functions. The neo-Hookean model gives results for simple
tension of most rubber that are in good agreement with experiments for an
axial stretch up to about 1.8. Numerical results for simple tension of the
neo-Hookean model and a compressible generalization are given in this
section for an axial stretch up to 1.6.
In order to consider the consequences of strictly entropic elasticity we
consider simple tension with axial stretch k and axial nominal stress
R ¼ r=k, where r is the corresponding axial Cauchy stress.
For simple tension, the axial nominal stress is identical to the axial Biot
stress SB . For isothermal simple tension equation (8.29) becomes
l 2
w ¼ q0 f ðk,H0 Þ ¼ k þ 2=k 3 , ð8:30Þ
2
and
dw dw
2 2 1
R¼ ¼l kk , r¼k ¼l k k : ð8:31Þ
dk dk
The simple tension stress (8.31) is in good agreement with experimental

results for stretch up to about 1.8.
Since u2 ðH0 Þ ¼ 0 and J ¼ 1, equation (8.28) becomes
H H
f ðk, HÞ ¼ f ðk, H0 Þ cH ln þ cc ðH H0 Þ: ð8:32Þ
H0 H0
The nominal stress obtained from equations ð8:31Þ1 and (8.32) is then
given by
H df ðk,H0 Þ H dw
Rðk,HÞ ¼ q0 ¼ , ð8:33Þ
H0 dk H0 dk
which shows that R is proportional to the absolute temperature for a given

value of k. It follows from equation ð8:33Þ2 that if the temperature of the
reference state is increased and deformation constrained, the stress will
remain zero. This means that, for a strictly entropic model for an elastic
solid the coefficient of volume thermal expansion is zero.
The entropy is given by

@f ðk,HÞ f ðk,H0 Þ H
s¼ ¼ cc ln , ð8:34Þ
@H H0 H0
which shows that s is of the form s ¼ s1 ðkÞ þ s2 ðHÞ. It follows from equa-
tions (8.33) and (8.34) that, for isothermal simple tension,
@s
R ¼ q0 H ,
@k
and this indicates that the stress is obtained from the change of entropy
with stretch for isothermal deformation.
8.4 Isentropic Simple Tension of Ideal Rubber
We now consider isentropic simple tension from the reference state. It may
be deduced from equation (8.34) that

f ðk,H0 Þ
H ¼ H0 exp ; ð8:35Þ
cc H0
since s ¼ 0 and k ¼ 1 in the reference state, and a temperature rise

wðkÞ
H H0 ¼ H0 exp 1
q0 cH0
results from the deformation. The relation between k and H H0 is shown

graphically in Figure 8.1 for wðkÞ given by equation (8.30), and
3
q = 906:5 kg:m , l = 420kPa and
1 1
cc = 1662 J:kg K , H0 = 293:15K,
where K denotes degrees Kelvin. These properties are given in [2].

The rise in temperature of an elastic band due to finite axial strain was
first observed by Gough in 1805 and studied experimentally by Joule in
1859. This temperature rise is in contrast to the decrease in temperature
Figure 8.1. Temperature–stretch relation for isentropic simple tension of ideal

(incompressible) rubber.
that occurs when a classically thermoelastic bar is subjected to isentropic

simple tension where the stress is due to change of internal energy.
The uniaxial nominal (or Biot) stress for isentropic simple tension of
an ideal rubber
h i
w
R ¼ exp l k k2 ,
q0 cc H0
follows from equations (8.33) and (8.34).

It should be noted that the strictly entropic ideal rubber model is a con-
siderable idealization and the temperature rise indicated in Figure 8.1 is
unrealistically high for a given value of k. However, the above use of this
model illustrates some important aspects of entropic thermoelasticity.
8.5 Isentropic Simple Tension of Compressible Rubber
Isentropic simple tension of a compressible model for an elastic rubber-like

material with a positive volume coefficient of thermal expansion exhibits
a decrease in temperature from k ¼ 1 to a minimum, known as the thermo-
elastic inversion point, at approximately k ¼ 1:06. The temperature then
increases monotonically, as indicated in Figure 8.2, with zero temperature
change at approximately k ¼ 1:125. A mechanically incompressible ther-
moelastic model, with a finite positive volume coefficient of thermal ex-
pansion has been used by Holzapfel [4] to analyze this inversion relation. In
the present treatment this model is not considered.
Figure 8.2. Isentropic simple tension of compressible rubber.

In order to investigate the thermoelastic inversion effect in isentropic

simple tension of a rubber, for which equation (8.22) is valid, we consider
a special case of equation (8.23) that is a compressible generalization of the
neo-Hookean strain energy function,
l 2 2 2 2=3
1
2
w ¼ q0 f ðki ,H0 Þ ¼ k1 þ k2 þ k3 3J þ K ðJ 1Þ : ð8:36Þ
2 2
Before considering isentropic simple tension we first consider a general

homogeneous deformation, then, with the use of equation (8.22), equation
(8.28) can be put in the form

H w H H
f ðki ,HÞ ¼ 1 u1 ð J Þ cc H ln þ cc ðH H0 Þ, ð8:37Þ
H0 q0 H0 H0
where w is given by equation (8.36). The entropy is then given by

@f w u1 ð J Þ H
s¼ ¼ cc ln : ð8:38Þ
@H q0 H0 H0 H0
and the principal Biot stresses by

@f H @w H uð J Þ @J
d~
SB
i ¼ q0 ¼ q0 1 : i 2 f1, 2, 3g: ð8:39Þ
@ki H0 @ki H0 dJ @ki
In equations (8.37) to (8.39) the term u1 ð J Þ is approximated by
aKH0
u1 ¼ ðJ 1Þ, ð8:40Þ
q0
where a is the volume coefficient of thermal expansion, assumed to be

constant for the range of temperature involved. Equation (8.40) is obtained
from equation (5.37) and the assumption that the dilatational strain is the
same order of magnitude as in classical elasticity. Instead of the factor

ðJ 1Þ, in equation (8.40), Chadwick [2] has proposed 0:4 J 2:5 1 , but
the difference is negligible for 0:98 < J < 1:02.
We now consider a particular deformation, simple tension, and outline

a procedure that requires numerical implementation for a particular set of
data. For simple tension k ¼ k1 , k2 ¼ k3 ¼ kr , SB
1 is the uniaxial Biot (or
nominal stress) and SB B
2 ¼ S3 ¼ 0.
A relation
/ðkr , k, HÞ ¼ 0 ð8:41Þ
is obtained from SB
2 ¼ 0, where, for isothermal simple tension H is replaced
by H0 . For isentropic simple tension from the natural reference state, with
J ¼ 1 and s ¼ 0,

w u1 ð J Þ H
cc ln ¼ 0, ð8:42Þ
q0 H 0 H0 H0
follows from equation (8.38). Substitution of

w u1 ð J Þ
H ¼ H0 exp cc ð8:43Þ
q0 c c H 0 H0
obtained from equation (8.42) in equation (8.41) gives a relation between kr

and k, for isentropic simple tension, that, for a given value of k, can be
solved for kr : The temperature is then determined for the given value of k
by substituting kr in equation (8.43). The axial stress is determined, for the
given value of k, by substituting equation (8.43) in SB
1 obtained from equa-
tion (8.39), then substituting the already determined value of kr . Imple-
mentation of this procedure was done using Mathematica. Without the use
of Mathematica or a similar package implementation would involve con-
siderable computational effort.
Numerical results for isentropic simple tension that are now presented
are for the following data given in [2],
q0 ¼ 906:5 kg:m3 , l ¼ 420 kPa, cc ¼ 1662 J:kg1 K1 ,

6 4 1
K ¼ 1:95 3 10 kPa, a ¼ 6:36 3 10 K , H0 ¼ 293:15:
The temperature change as a function of k for isentropic simple tension is

shown graphically in Figure 8.2.
The inversion effect, namely, the initial decrease in temperature to a min-
imum followed by a monotonic increase in temperature is evident in Figure
8.2. This effect is well known and has been extensively studied.
There is negligible difference between results from the incompressible
stress-stretch relation (8.31) and those for the isothermal and isentropic
relations for the corresponding compressible strain energy function (8.36)
with K=l ¼ 4642:9 obtained from the above data. For example, for k ¼ 1:6,
R=l ¼ 1:209375, 1:20686, 1:21226,
for equation (8.31), isothermal compressible and isentropic compressible,

respectively. The corresponding values for kr are
kr ¼ 0:790569, 0:790601, 0:790622:
The compressible model based on equation (8.22) is a special case of

a modified entropic elasticity with the dilatational part of the deformation
is energetic and the isochoric part is entropic. A more general modified
entropic model has the isochoric part of the deformation partly entropic
and partly energetic and for this model (8.22) is replaced by
u~ðF,HÞ ¼ u1 ðFÞ þ u2 ðHÞ:
There is some experimental evidence that this model is applicable to cer-

tain rubber-like materials. A detailed discussion of modified entropic elas-
ticity is given in [3].
EXERCISES
8.1. Determine the Gibbs relations for the enthalpy, h ¼ hðS, sÞ, and the
Gibbs free energy, g ¼ gðS,HÞ.
References 219
8.2. Prove that
@h @h @g @g
H¼ , CKL ¼ 2qo , s¼ , CKL ¼ 2q0 :
@s @SKL @H @SKL
8.3. Prove that f u þ h g ¼ 0.
REFERENCES
1. Kestin, J. (1968). A Course in Thermodynamics, 2nd Ed. Blaisdell Publish-

ing Company.
2. Chadwick, P. (1974). Thermo-Mechanics of Rubber-like Materials. Phil.
Trans. R. Soc. A. 276, pp. 371–403.
3. Chadwick, P., and Creasy, C.F.M. (1984). Modified Entropic Elasticity of
Rubberlike Materials. J. Mech. Phys. Solids. 32, No. 5, pp. 337–357.
4. Holzapfel, G.A. (2000). Nonlinear Solid Mechanics. Wiley.
5. Ogden, R.W. (1997). Non-Linear Elastic Deformations. Dover.
6. Treloar, L.R.G. (1956). The Physics of Rubber Elasticity, 2nd Ed. Oxford.
9 Dissipative Media
Ideal elastic solids and inviscid fluids do not exhibit dissipative effects. Con-
sequently, the only irreversible processes possible in these media are heat
conduction and shock wave propagation. Deformation and flow of dissipa-
tive media involve dissipation, due to hysteresis effects, of mechanical energy
as heat and this is an additional contribution to irreversibility. Entropy pro-
duction is a measure of irreversibility, and for a dissipative medium, it has
contributions from dissipation of mechanical energy, heat conduction, and
shock wave propagation. We will not be concerned with shock wave propa-
gation in this chapter. This topic is covered in the classic text by Whitham [1].
The dissipative media discussed in this chapter are Newtonian viscous
fluids, Stokesian fluids, and linear viscoelastic solids. Other dissipative media
are discussed in texts on rheology, for example, Tanner [2].
9.1 Newtonian Viscous Fluids
The constitutive equations of a Newtonian viscous fluid have been obtained

in chapter 5. In the present chapter the dissipation of mechanical energy
and the related entropy production are considered. The constitutive equa-
tion (5.10), obtained in chapter 5, for a compressible Newtonian viscous
fluid can be put in the form

2
r þ pI ¼ 2gD þ n g trD I,
3
ð9:1Þ
2
rij þ pdij ¼ 2gdij þ n g dkk dij ,
3
220
9.1 Newtonian Viscous Fluids 221
where p is the thermodynamic pressure and g and n are the shear and
bulk viscosity coefficients, respectively, which are temperature depen-
dent in real fluids. The shear viscosity g of a gas increases with temper-
ature increase and the viscosity of a liquid decreases with temperature
increase. Temperature dependences of g and n are neglected in what
follows, and this restricts the theory presented to relatively small tem-
perature changes. Constitutive equation (9.1) can be decomposed into
dilatational and deviatoric parts
rkk
þ p ¼ ndkk ð9:2Þ
3
and
sij ¼ 2gdij9: ð9:3Þ
Equation (9.2) is obtained by contracting equation (9.1), and equation (9.3)

is obtained by substituting
rkk
rij ¼ sij þ dij ,
3
ð9:4Þ
dkk
dij ¼ d9ij þ dij ,
3
into equation (9.1) and using equation (9.2).

It follows from equations (9.2) and (9.3) that the stress power per unit
mass consists of a dilatational part and a deviatoric part

nd2kk pdkk 2ldij9 dij9
and , ð9:5Þ
q q
respectively. The stress power per unit mass

rij dij 1 2
¼ pdkk þ 2gdij dij þ n g d2kk
q q 3
222 Dissipative Media
can also be obtained from equation (9.1) and substitution of equation (9.4)
gives, after some manipulation,
rij dij 1 2

¼ pdkk þ 2gdij9 dij9 þ ndkk : ð9:6Þ
q q
In equation (9.6) the stress power has been decomposed into three parts,

a nondissipative dilatational part pdkk =q, and dissipative parts, 2gd9ij d9ij q

due to shear deformation and nd2kk q due to dilatation. This is consistent
with equation (9.5). The dissipative parts are rates of dissipation per unit
mass of mechanical energy as heat. It follows that the rates of entropy
production per unit mass due to shear deformation and dilatation are
2
2gdij9 dij9 nd
c_ s ¼ and c_ d ¼ kk ,
qH qH
respectively. In addition there will be the rate of entropy production due to

heat conduction,
qi @H
c_ c ¼ :
qH2 @xi
If the Stokes’ hypothesis, that is, n ¼ 0, is assumed, the thermodynamic

pressure is p ¼ rkk =3; and there is no viscous dissipation of mechanical
energy as heat in dilatation. According to Schlichting [3], it appears that the
bulk viscosity vanishes identically in gases of low density.
If an incompressible fluid model is considered, equation (9.1) is
replaced by
rij þ pdij ¼ 2gdij , dkk ¼ 0, ð9:7Þ
where p is a Lagrangian multiplier determined from the stress boundary

conditions. The trace of equation (9.7) shows that p ¼ trr=3. Conse-
quently, equation (9.7) can be put in the form
sij ¼ 2gdij , dkk ¼ 0: ð9:8Þ
It is evident that only the shear viscosity coefficient is required.

9.2 A Non-Newtonian Viscous Fluid 223

Equation (9.8) indicates that a dissipation potential, / ¼ gdij dij q,
dkk ¼ 0, exists, that is , half the rate of dissipation of mechanical energy or

stress power 2gdij dij q per unit mass, so that
@/
sij ¼ q : ð9:9Þ
@dij
In equation (9.9), as usual, when an isotropic scalar function of a second-order

symmetric tensor is differentiated with respect to a component of the
tensor, the nine components of the tensor are taken as independent.
9.2 A Non-Newtonian Viscous Fluid
We now consider a model for a nonlinear viscous fluid. This model is an

incompressible variant of one known as the Reiner-Rivlin fluid or Stoke-
sian fluid [4] and is given by
2
r ¼ pI þ a1 D þ a2 D ; trD ¼ 0, ð9:10Þ
rij ¼ pdij þ a1 dij þ a2 dik dkj ; dkk ¼ 0,
where p is determined from the stress boundary conditions. The parame-

ters a1 and a2 are functions of the invariants of D and possibly temperature
and a1 =2 can be regarded as the viscosity and a2 as the cross viscosity. Since
incompressibility is assumed, D is a deviatoric, that is, D ¼ D9. It follows
that the nonzero invariants of D are
1 1
I 1 ¼ 0, I2 ¼ trD92 , I3 ¼ Det½D9 ¼ trD93 :
2 3
Equation (9.10) indicates that r þ pI is due only to viscosity effects. Con-

sequently, the stress power per unit volume
P ¼ a1 trD2 þ a2 trD3
is equal to the rate dissipation of mechanical energy. It then follows that
a1 trD2 þ a2 trD3 > 0
is a necessary condition for equation (9.10) to be admissible.

The incompressible Newtonian fluid is a special case of equation (9.10).
Equation (9.10) predicts second-order effects that are not predicted by
equation (9.8), for example, the normal stresses that are predicted for sim-
ple shear flow. However, certain other phenomena, observed for nonlinear
dissipative fluids, are not predicted by equation (9.10), and a viscoelastic
fluid model is then required.
9.3 Linear Viscoelastic Medium
A viscoelastic medium, usually a solid, exhibits creep under constant

stress and stress relaxation under constant deformation. Similar to a vis-
cous fluid the stress is rate dependent but also elastic energy can be
stored and dissipated. The linear theory is based on infinitesimal strain,
e, as in linear elasticity and e_ [ de=dt @e=@t is described as the strain
rate, rather than the rate of deformation when finite deformation is
considered. In what follows isotropy is assumed. A viscoelastic solid
behaves elastically when the rate of deformation is vanishingly small
and the corresponding elastic moduli are known as the relaxation mod-
uli. The relaxation modulus for a viscoelastic fluid is zero. Elastic behav-
ior is approached as the rate of deformation approaches infinity and the
corresponding elastic moduli are known as the impact moduli. In this
section an outline of linear viscoelasticity is given. Detailed treatments,
of the topic, are given in various specialized texts by Christensen [5],
Flügge [6], and others.
One-dimensional viscoelastic constitutive relations can be obtained
from the study of the response of spring-dashpot models. These models
consist of discrete representations of uniform bars subjected to simple
tension and compression. A linear spring that schematically represents
a uniform elastic bar is shown in Figure 9.1.
e
σ Figure 9.1. Elastic element.
The uniaxial stress and strain are denoted by r and e, respectively, and are
related by Hooke’s law,
r ¼ Ee, ð9:11Þ
where E is constant. A linear viscous element that is modeled schematically

by a dashpot is shown in Figure 9.2.
_ respectively, and
The uniaxial stress and strain rate are denoted by r and e;
r ¼ ge,
_ ð9:12Þ
where g is a viscosity coefficient. One-dimensional viscoelastic models are

obtained by combining spring and dashpot elements. The two simplest
models are the Maxwell fluid model that is represented schematically by
a spring and dashpot connected in series and the Kelvin-Voigt solid model
that is represented by a spring and dashpot in connected in parallel, shown
in Figure 9.3.
The Maxwell model represents a viscoelastic fluid since it undergoes con-
tinuing deformation when a constant stress is applied. Referring to Figure
9.3a,
r
e ¼ e 1 þ e2 ¼ þ e2 ,
E
e
σ Figure 9.2. Viscous element.
σ
e1 e2
(a) Maxwell model.
σ1
σ2
(b) Kelvin-Voigt model.

Figure 9.3. Newtonian viscous fluids. (a) Maxwell model. (b) Kelvin-Voigt model
and the time derivative and r ¼ ge_2 give the constitutive equation for
a Maxwell material,
E
r_ þ r ¼ Ee:
_ ð9:13Þ
g
Suppose that at t ¼ 0þ , e ¼ e0 and thereafter remains constant so that e_ðtÞ = 0,

that is, eðtÞ ¼ e0 H ðtÞ, where H ðtÞ is the unit step function defined by

1, t > 0
H ðt Þ ¼ :
0t<0
It then follows from equation (9.13) and the initial condition rð0þ Þ ¼ Ee0
that

Et
rðtÞ ¼ Ee0 exp H ðtÞ: ð9:14Þ
g
Equation (9.14) indicates relaxation, that is, the stress decays with time and

Et
GðtÞ ¼ Eexp ð9:15Þ
g
is known as the relaxation function. The quantity g=E has dimension of

time and is known as the relaxation time. In general the relaxation
function represents the stress rðtÞ required to produce the strain

eðtÞ ¼ e0 H ðtÞ, that is,
rðtÞ ¼ GðtÞe0 :
Since equation (9.14) indicates that r ! 0, as t ! ‘, the relaxation modu-

lus is zero. As e_ ! ‘ the viscous element approaches rigidity, that is,
e_2 ! 0. Consequently, the impact modulus is E. Suppose now that the
Maxwell model is subjected to a sudden application of axial stress
rðtÞ ¼ r0 H ðtÞ. For t > 0þ , r_ ¼ 0 and r ¼ r0 and
Er
_ t > 0þ :
¼ Ee, ð9:16Þ
g
Then, it follows from equation (9.16) and initial condition eð0þ Þ that

1 t
e ¼ r0 þ : ð9:17Þ
E g
Equation (9.17) indicates creep and that the Maxwell model represents
a fluid since e increases monotonically due to a constant stress and

1 t
J ðtÞ ¼ þ : ð9:18Þ
E g
is known as the creep function. In general the creep function is the strain
eðtÞ that results from the application of stress rðtÞ ¼ r0 H ðtÞ, that is,
eðtÞ ¼ J ðtÞr0 :
The Maxwell model is now used to illustrate the superposition principle for
linear viscoelasticity. Suppose a stress rðtÞH ðtÞ is applied to the Maxwell
element, with rð0Þ ¼ r , as indicated in Figure 9.4.
The response deðtÞ due to stress increment dr applied at t ¼ s,

1 ðt sÞ
de ¼ dr þ , t > s,
E g
σ(t)
dσ(τ)
σ*
0 τ τ + dτ
t
Figure 9.4. Application of stress increment.
follows from equation (9.17). Since the governing equation (9.13) is linear
a superposition principle holds. Consequently,
Zt
1 t drðsÞ 1 t s
eðtÞ ¼ r þ þ þ ds
E g ds E g
þ
0
or
Zt
drðsÞ 1 t s
eðtÞ ¼ þ ds:
ds E g
0
If eðtÞ is prescribed the resulting stress rðtÞ is determined by solving equa-

tion (9.13).
Referring to Figure 9.3b for the Kelvin-Voigt model
r ¼ r1 þ r2 ,
so that, with the use of equations (9.11) and (9.12), the constitutive equa-
tion is
E r
e_ þ e¼ : ð9:19Þ
g g
Suppose the Kelvin-Voigt element is subjected to a sudden application of
axial stress rðtÞ ¼ H ðtÞr0 ; then from equation (9.19),
E r0
e_ þ e¼ : ð9:20Þ
g g
The solution of equation (9.20) with initial condition eð0Þ ¼ 0 is

r0 Et
e¼ 1 exp : ð9:21Þ
E g
Equation (9.21) indicates creep since e ! r0 =E as t ! ‘. The creep func-

tion follows from equation (9.21) and is

1 Et
J ðtÞ ¼ 1 exp : ð9:22Þ
E g
If at time t9 the strain is fixed at constant value eðt9Þ, r immediately drops

and remains constant with value

Et9
r ¼ Eeðt9Þ ¼ r0 1 exp ,
g
that is, when the strain is held fixed, the stress immediately relaxes.
When the strain rate approaches infinity the viscous element
approaches rigidity, consequently the impact modulus is infinite. Substitut-
ing eðtÞ ¼ e0 H ðtÞ in equation (9.19) gives
rðtÞ ¼ ðgdðtÞ þ EÞe0 ,
and the relaxation function is
GðtÞ ¼ gdðtÞ þ E, ð9:23Þ
where dðtÞ is the Dirac delta function. The Dirac delta function dðtÞ, and the
unit step function H ðtÞ belong to a class known as generalized functions.
The unit function has already been defined and the Dirac delta function is
defined by
Z‘ Z‘
dðtÞ ¼ 0; t 6¼ 0, and dðtÞdt ¼ 1or uðtÞdðtÞdt ¼ uð0Þ,
‘ ‘
E2 σ2 η
e2 e3
σ1
E1 e1= e
Figure 9.5. Standard model.
and can be regarded, in a nonrigorous way, as the time derivative of H ðtÞ. The
theory of generalized functions is given in a text by Lighthill [7]. It is evident
from equation (9.23) that the Kelvin-Voigt model has an infinite impact
modulus, and this indicates that it may not be suitable for certain applications.
There are many more complicated spring-dashpot models than the
Maxwell and Kelvin-Voigt models. The so-called standard model consists
of a spring and Maxwell element in parallel as indicated in Figure 9.5 and
involves three parameters, two stiffness parameters, E1 and E2 , and a vis-
cosity g. It has been proposed as a more realistic model than the Kelvin-
Voigt model for a linear viscoelastic solid. The standard model is used to
illustrate the theory in the remainder of this chapter; however, the expres-
sions for the hereditary integrals given are valid in general for isotropic
linear viscoelasticity.
Referring to Figure 9.5,
e2 þ e3 ¼ e, ð9:24Þ
and
r1 þ r2 ¼ r: ð9:25Þ
It follows from equations (9.11) and (9.12) and the time derivative of equa-
tion (9.24) that
r_ 2 r2
þ ¼ e;
_ ð9:26Þ
E2 g
and from equations (9.25) and (9.11) and the time derivative of equation
(9.25) that
r2 ¼ r E1 e, ð9:27Þ
r_ 2 ¼ r_ E1 e:
_ ð9:28Þ
Elimination of r2 and r_ 2 from equations (9.26) to (9.28) gives the differen-

tial constitutive equation
E2 E1 E2
r_ þ r ¼ ðE1 þ E2 Þe_ þ e: ð9:29Þ
g g
Equation (9.29) is equivalent to
r ¼ E1 e þ E2 ðe qÞ ð9:30Þ
and
E2
q_ ðtÞ ¼ ðe qÞ, ð9:31Þ
g
where q ¼ e3 is strain in the dashpot and is known as an internal variable.

Equation (9.31) is known as the evolution equation for q.
Substitution of eðtÞ ¼ e0 H ðtÞ into equation (9.29) gives, for t > 0þ ,
E2 E1 E2
r_ þ r¼ e0 ,
g g
with solution

E2 t
r ¼ bexp þ E1 e0 , ð9:32Þ
g
where b is a constant of integration. At t ¼ 0þ , the viscous element is

instantaneously rigid and this gives the initial condition rð0þ Þ ¼ ðE1 þ E2 Þe0
that in turn gives b ¼ E2 e0 and equation (9.32) becomes

E2 t
r ðt Þ ¼ E2 exp þ E1 e0 : ð9:33Þ
g
It then follows from equation (9.33) that the relaxation function for the
standard model is

E2 t
GðtÞ ¼ E2 exp þ E1 , ð9:34Þ
g
Gð0Þ ¼ E1 þ E2 is the impact modulus and Gð‘Þ ¼ E1 is the relaxation

modulus.
The form of a typical relaxation function is shown in Figure 9.6.
The impact modulus Gð0Þ for the Kelvin-Voigt model is infinite and the
relaxation modulus Gð‘Þ for the Maxwell model is zero.
Referring to Figure 9.7 and using superposition, a strain history eðtÞ, that
has a step discontinuity e at t ¼ 0 gives rise to
Zt
deðsÞ
rðtÞ ¼ GðtÞe þ Gðt sÞ ds, ð9:35aÞ
ds
þ
0
or
Zt
deðsÞ
r ðt Þ ¼ Gðt sÞ ds: ð9:35bÞ

ds
0
G(0)
G(t)
G(∞) Figure 9.6. Typical relaxation function.
t
Equation (9.35) is described as a hereditary integral since rðtÞ depends on

the strain history eðsÞ for s < t.
In order to obtain the creep function for the standard model,
r ¼ r0 H ðtÞ is substituted in equation (9.29). Then for t > 0þ ,
E2 E1 E2
r0 ¼ ðE1 þ E2 Þe_ þ e,
g g
with solution

E1 E2 t r0
e ¼ cexp þ , ð9:36Þ
gðE1 þ E2 Þ E1
where c is a constant of integration. The initial condition is

eð0þ Þ ¼ r0 =ðE1 þ E2 Þ and it follows from equation (9.36) that

1 E2 E1 E2 t
e¼ exp r0 :
E1 E1 ðE1 þ E2 Þ gðE1 þ E2 Þ
and the creep function is

1 E2 E1 E2 t
J ðtÞ ¼ exp : ð9:37Þ
E1 E1 ðE1 þ E2 Þ gðE1 þ E2 Þ
The form of a typical creep function is shown in Figure 9.8.

For the Maxwell model J ðtÞ ! ‘ as t ! ‘ and for the Kelvin-Voigt
model J ð0Þ ¼ 0.
e(t)
de(τ)
e*
τ τ + dτ
t
Figure 9.7. Application of strain increment.
J(∞)
J(t)
Figure 9.8. Typical creep function.
J(0)
It follows from superposition that a strain history rðtÞ with a step

discontinuity r at t ¼ 0 gives rise to another hereditary integral
Zt
drðsÞ
eðtÞ ¼ J ðtÞr þ J ðt sÞ ds ð9:38Þ
ds
0þ
or
Zt
drðsÞ
eðtÞ ¼ J ð t sÞ ds: ð9:38bÞ

ds
0
The forms (9.35) and (9.38), for the hereditary integrals, are valid for all
discrete linear viscoelastic models and are valid for continuous systems.
A continuum is now considered and distortion and dilation relations
are separated by expressing Cauchy stress, r, and infinitesimal strain, e, in
terms of the deviatoric and isotropic parts,
e~
rij ¼ sij þ rdij , eij ¼ eij9 þ dij ,
3
where r ¼ rkk =3 and e~ ¼ ekk .
The relaxation functions for distortion and dilatation described as
the shear and bulk relaxation functions are denoted by Gl ðtÞ and GK ðtÞ,
respectively, and the corresponding creep functions are denoted by Jl ðtÞ
and JK ðtÞ, respectively, and are obtained by replacing E1 and E2 in equa-
tions (9.34) and (9.37) by 2l1 and 2l2 for the distortional deformation and
K1 and K2 for the dilatational deformation. Axial stress and strain in the
relations already obtained for the one-dimensional viscoelastic model are
now replaced by sij and eij9 for distortional deformation and r and e~ for
dilatation.
The hereditary integrals are
Zt
deij9
sij ¼ Gl ðt sÞ ds, ð9:39Þ

ds
0
Zt
d~
e
r ¼ GK ðt sÞ ds, ð9:40Þ

ds
0
Zt
dsij
eij9 ¼ Jl ðt sÞ ds, ð9:41Þ

ds
0
Zt
d
r
e~ ¼ JK ðt sÞ ds: ð9:42Þ

ds
0
The relaxation and creep functions GK and JK in equations (9.40) and

(9.42), respectively, differ from those in [5] by a factor of 3, since they
involve r ¼ rkk =3 instead of rkk . The relaxation and creep functions in
equations (9.39) to (9.42) for the continuum are of the same form as those
for the one-dimensional discrete models, equations (9.34) and (9.38), with
suitable choice of the parameters. For example, for the standard model
each relaxation function involves three parameters and are of the form

2l2 t
Gl ðtÞ ¼ 2l2 exp þ 2l1 , ð9:43Þ
g
with parameters 2l1 , 2l2 , and g, and

K2 t
GK ðtÞ ¼ K2 exp þ K1 , ð9:44Þ
g^
with parameters K1 , K2 , and g^. The parameters l1 and K1 corresponding to

Gl ð‘Þ and GK ð‘Þ, respectively, are the relaxation shear and bulk moduli,
respectively, and correspond to those of linear elasticity as expected.
Similarly, the corresponding impact moduli are 2ðl1 þ l2 Þ and ðK1 þ K2 Þ

and from equation (9.37) the creep functions for the continuum are

1 2l2 4l1 l2 t
Jl ¼ exp , ð9:45Þ
2l1 4l1 ðl1 þ l2 Þ 2gðl1 þ l2 Þ

1 K2 K1 K2 t
JK ¼ exp : ð9:46Þ
K1 K1 ðK1 þ K2 Þ g^ðK1 þ K2 Þ
It follows from equations (9.43) to (9.46) that
Gl ð‘ÞJl ð‘Þ ¼ 1 and GK ð‘ÞJK ð‘Þ ¼ 1 ð9:47Þ
and
Gl ð0ÞJl ð0Þ ¼ 1 and GK ð0ÞJK ð0Þ ¼ 1: ð9:48Þ
It should be noted that equations (9.47) and (9.48) do not imply that
Gl ðtÞJl ðtÞ ¼ 1 and Gl ðtÞJl ðtÞ ¼ 1, for all t. The correct result [5] is obtained
from
a ðsÞJa ðsÞ ¼ s2 ,

G
where the superposed bar denotes the Laplace transform and a is replaced
by l or K for shear or dilatational deformation, respectively. The Laplace
transform fðsÞ of f ðtÞ is defined by
Z‘
fðsÞ ¼ f ðtÞest dt,
0
and a standard text on integral transforms should be consulted for details.

If it is assumed that the relaxation function for shear is proportional to

that for dilatation, that is, Gl ðtÞ GK ðtÞ is constant for all t. Then,
Gl ðtÞ 3ð1 2mÞ

¼
GK ðtÞ 1þm

since Gl ð‘Þ GK ð‘Þ ¼ 2l1 =K1 and, for linear elasticity,
3K ¼ 2lð1 þ mÞ=ð1 2mÞ, where m is Poisson’s ratio.
Similarly, it may be shown that
Jl ðtÞ 1þm
¼ :
JK ðtÞ 3ð1 2mÞ
It may also be shown, for equations (9.43) and (9.44), that the relaxation
times are equal, that is,
g g^
¼
2l2 K2

if Gl ðtÞ GK ðtÞ is constant for all t.
9.4 Viscoelastic Dissipation
A continuum Kelvin-Voigt model, with the dilatation purely elastic, is

perhaps the easiest linear viscoelastic model to obtain a relation for the
rate of dissipation of mechanical energy. Referring to Figure 9.3b, the
continuum form of the constitutive equation (9.19) is
rij ¼ 2ge_ij9 þ 2leij þ kekk dij ,
where k ¼ K 2l=3, K and l are the relaxation bulk and shear moduli,
respectively. It then follows that the stress power per unit volume is
rij e_ij ¼ 2ge_ij9 e_ij9 þ 2leij e_ij þ kekk e_ll :
The stress power is composed of two parts, a dissipative part, 2ge_ij9 e_ij9, and
a part, 2leij e_ij þ kekk e_ll , which is the rate of change of stored elastic energy. It
is reasonable to assume that, in general, the stress power is composed of
a dissipative part and a rate of stored energy part. This is now shown for
a more difficult case, a standard material. Continuum forms of the differential
constitutive equation (9.29), for the shear and dilatational deformation, are
2l2 4l l
s_ij þ sij ¼ 2ðl1 þ l2 Þe_ij9 þ 1 2 eij9 ð9:49Þ
g g
and
K2 K1 K2
r_ þ r ¼ ðK1 þ K2 Þe~_ þ e~, ð9:50Þ
g^ g^
respectively. In equations (9.49) and (9.50) it is not assumed that the vis-
cosity coefficient g is the same for shear and dilatation.
The stress power can be decomposed into a deviatoric or shear part,
Pl ¼ sij e_ij9, and a dilatational part PK ¼ re~_ so that rij e_ij ¼ Pl þ PK . A fur-
ther decomposition of each of these parts into a dissipative part and a rate
of stored energy part is possible. The part for the deviatoric deformation
governed by equation (9.49) is now considered. Equation (9.49) is equiva-
lent to [8],
sij ¼ 2l1 eij9 þ 2l2 eij9 qij9 ð9:51Þ
and
2l2
q_ ij9 ¼ eij9 qij9 : ð9:52Þ

g
Similarly, equation (9.50) is equivalent to
r ¼ K1 e~ þ K2 ðe~ q~Þ ð9:53Þ
and
K2
q~_ ¼ ðe~ q~Þ, ð9:54Þ
g^
where the tensor q with components qij ¼ qij9 þ ðq~=3Þdij , q~ ¼ qkk , is a so-called
internal variable and is analogous to the strain of the dashpot element in
Figure 9.5. More complicated viscoelastic materials whose spring and dash-
pot representations have more than one dashpot have a corresponding
number of internal variables.
The stress power for the deviatoric part of the deformation obtained
from equation (9.51) is
sij e_ij9 ¼ 2l1 eij9 e_ij9 þ 2l2 eij9 qij9 e_ij9,
and with the use of equation (9.52) can be put in the form
sij e_ij9 ¼ 2l1 eij9 e_ij9 þ 2l2 eij9 qij9 e_ij9 q_ ij9 þ gq_ ij9 q_ ij9: ð9:55Þ
If the parameters l1 , l2 , and g, and sij and eij9 are known, qij9 can be obtained
from equation (9.51). The first two terms on the right-hand side of equation
(9.55) represent the rate of storage of elastic strain energy and the third
term represents the rate of dissipation of mechanical energy. This is evident
if we consider the spring-dashpot model shown in Figure 9.5, since elastic
energy is stored in the springs and mechanical energy is dissipated in the
dashpot. Similarly, the part of the stress power for the dilatational defor-
mation can be put in the form
re~ ¼ K1 e~e~_ þ K2 ðe~ q~Þ e~_ q~_ þ g^q~_ 2 :
A model for linear viscoelasticity has been proposed [10], which assumes
that the dilatational deformation is purely elastic, so that equations (9.40)
and (9.42) are replaced by
r ¼ K~
e:
All the relations already obtained for the deviatoric deformation of the
standard model are valid, but for purely dilatational deformation linear
elasticity theory holds. This model is somewhat similar to that for a com-
pressible Newtonian viscous fluid with zero bulk viscosity and would be
completely analogous for the Kelvin-Voigt viscoelastic model. A disad-
vantage of the model is that Poisson’s ratio is not constant. It may be
shown that for the standard model for deviatoric deformation along with
purely elastic dilatation that the equilibrium and impact Poisson’s ratios
are given by
3K 2l1
me ¼
8K þ 2l1
and
3K 2ðl1 þ l2 Þ
m¼ ,
8K þ 2ðl1 þ l2 Þ
respectively.
The present treatment of linear viscoelasticity has so far neglected ther-
mal effects, although the material properties may be temperature dependent.
It is assumed that neglect of thermal effects may be justified if
jH H0 j=H0 << 1, where H0 is the temperature in the undeformed reference
configuration.
9.5 Some Thermodynamic Considerations in

Linear Thermoelasticity
The linear theory of viscoelasticity, governed by equations (9.39) to (9.42),

does not include thermal effects and may be regarded as isothermal. It also
allows uncoupled decomposition into deviatoric and dilatational relations.
Use of the theory implies the assumption of zero volume coefficient of
thermal expansion and infinite thermal conductivity since mechanical dis-
sipation produces temperature increase. It is clear that the usefulness of the
linear theory is restricted to small temperature changes, that is, for
jH H0 j=H0 << 1, fortunately this condition is satisfied in many practical
applications, for example, in structures.
The theory of thermoviscoelasticity has been developed in different
ways by various authors [5]. Thurston [9] has shown that a difficulty arises
when thermal effects are incorporated into linear viscoelasticity. Accord-
ing to Thurston that, in applying linear viscoelastictiy theory to adiabatic
changes, the effect of energy dissipation gets linearized away. However,
9.5 Some Thermodynamic Considerations in Linear Thermoelasticity 241
linear thermoviscoelastic theories have been presented by Christensen [5]

and Hunter [10]. These theories will not be discussed here. In this section
we give a short general outline of thermoviscoelastic theory, and in order
to illustrate aspects of linear thermoviscoelasticity, we consider the stan-
dard model, with a purely elastic dilatational response, that has one in-
ternal variable with components qij such that qkk ¼ 0. The assumption of
purely elastic response for dilatation enables a simple thermoviscoelastic
theory to be developed.
In what follows specific thermodynamic problems are per unit mass.
The energy equation (4.31) with dij replaced by e_ij and u f þ Hs is

@qi
q f_ þ Hs
_ þ sH
_ qr rij e_ij þ ¼ 0, ð9:56Þ
@xi
where qi are the components of the heat flux vector, and the Clausius-
Duhem in equality is
_ þ f_Þ þ rij e_ij qi @H > 0:

qðsH ð9:57Þ
H @xi
It should be noted that equations (9.56) and (9.57) are valid only for in-
finitesimal deformation and ðH H0 Þ=H0 << 1. Consequently, it is reason-
able to assume that the thermal conductivity k is constant and Fourier’s law
of heat conduction is given by
@H
qi ¼ k :
@xi
In order to develop further a thermoviscoelastic theory rij and eij are

decomposed into their deviatoric and isotropic parts, rij ¼ sij þ rdij and
eij ¼ eij9 þ ðe~=3Þdij .
The stress components are given by

r ¼ K e~ aðH H0 Þ , ð9:58Þ
and

sij ¼ 2l1 eij9 þ 2l2 eij9 qij , ð9:59Þ
and the stress power by
rij e_ij ¼ sij eij9 þ re~,
where a is the coefficient of volume thermal expansion assumed constant

for infinitesimal strain and ðH H0 Þ=H0 << 1. Equation (9.58) is the
thermoelastic dilatation relation and equation (9.59) is identical to
equation (9.51).
A possible form of the constitutive relation for the Helmholtz free
energy is

f ¼ f eij9; e~; H; gi ; qij , ð9:60Þ
where gi ¼ @H=@xi are the components of the temperature gradient vector.

Substitution of
@f @f @f _ @f @f
f_ ¼ _ þ e~_ þ
e9 Hþ g_ i þ q_ ij ð9:61Þ
@eij9 e
@~ @H @gi @qij
from equation (9.60) into equation (9.57) gives

1 @f 1 @f _ @f _
sij e_ij9 þ r e~ s þ H ð9:62Þ
q @eij9 q e
@~ @H
@f @f 1
g_ i q_ ij qi gi > 0:
@gi @qij qh
_ and g_ i , and this

Inequality equation (9.62) is valid for all values e_ij9, e~_ , H,
_ and gi must be zero. Consequently,
requires that the coefficients of e_ij , H,
@f @f @f
rij9 ¼ q , r~ ¼ q , s ¼ , ð9:63Þ
@eij9 e
@~ @h
9.5 Some Thermodynamic Considerations in Linear Thermoelasticity 243
and f does not depend on gi . Then, equations (9.62) and (9.58) become
@f 1
q_ ij qi gi > 0, ð9:64Þ
@qij qh
f ¼ f eij9; e~; H; qij , ð9:65Þ
respectively.
The elimination of gi from equation (9.60) is also justified by invoking the
Principle of Local State. In equation (9.64), the first term is independent of
gi ; consequently,
@f
q_ ij > 0, ð9:66Þ
@qij
and this is the rate of dissipation of mechanical energy. It follows from the
mechanical dissipation term that appears in equation (9.55) and the expres-
sion (9.66) for mechanical dissipation that
@f
q q_ ij ¼ gq_ ij q_ ij : ð9:67Þ
@qij
Then, from equations (9.52) and (9.67)
@f
q ¼ 2l2 qij eij9 : ð9:68Þ

@qij
Substituting equations (9.63) and (9.68) into the differential of equation

(9.65) and then substituting equations (9.58) and (9.59) gives

qdf ¼ 2ðl1 þ l2 Þeij9 2l2 qij deij9 þ K fe~ aðH H0 Þgd~
e

ð9:69Þ
þ qsdH þ 2l2 qij eij9 dqij :
In order to integrate equation (9.69), it should be noted that

dseij9 ¼~e¼qij¼0 ¼ c0 dH=H so that
seij9 ¼ e~ ¼ qij¼0 ¼ c0 lnðH=H0 Þ

and linearizing gives

H H0
seij9¼~e¼qij¼0 ¼ c0 ; ð9:70Þ
H0
where c0 is the specific heat at constant deformation and is assumed con-

stant. Using equation (9.70) it follows from equation (9.69) that
2
e~
qf ¼ ðl1 þ l2 Þeij9 eij9 2l2 qij eij9 þ K
2
ð9:71Þ
c0 ðH H0 Þ2
KaðH H0 Þ~
e þ l2 qij qij ;
2 H0
and the linearized form of the entropy is

Ka~e H H0
s¼ þ c0 ,
q H0
which satisfies the relation
@s
c0 ¼ H :
@H
Substituting equations (9.63) and (9.68) in the time derivative of equation

(9.65) or differentiating equation (9.71) with respect to time gives
1 1
f_ ¼ sij e9 _ þ 2l qij eij9 q_ ij < 1 sij e9

_ þ re~_ sH
1 _
_ þ re~_ sH, ð9:72Þ
2
q q q q
where the inequality arises because of equation (9.66). Inequality equation

(9.72) shows that the rate of change of the Helmhotz free energy is less than
or equal to the stress power plus the ‘‘thermal power’’ minus sH _ [8].
9.6 Simple Shear Problem
A problem of adiabatic simple shear involving the standard viscoelastic

material with purely elastic dilatational response is now considered in order
9.6 Simple Shear Problem 245
to illustrate some aspects of thermoelastic behavior. Simple shear is a ho-

mogeneous isochoric deformation and e12 ¼ e21 are the only nonzero com-
ponents of the infinitesimal strain tensor and q12 ¼ q21 is the only nonzero
components of the internal variable tensor. The problem considered is to
determine the rate of dissipation of mechanical energy dissipation and the
mechanical energy dissipated when equilibrium is attained if an application
of strain
e12 ¼ e12 H ðtÞ ð9:73Þ
is applied. Also the rate of dissipation of mechanical energy dissipation and

the mechanical energy dissipated if unloading occurs from the loaded equi-
librium state is determined.

2l2 t
s12 ðtÞ ¼ 2l2 exp þ 2l1 e12 ,
g
and from equations (9.51) and (9.52) that
s12 ¼ 2l2 e12 þ gq_ 12 , ð9:74Þ
then
e12 t
q_ 12 ¼ exp : ð9:75Þ
T T
The rate of mechanical energy dissipation per unit volume, obtained from
equation (9.55), is given by
D ¼ 2gq212 ,
and with equation (9.75) this gives,
he ti2
12
D ¼ 2g exp : ð9:76Þ
T T
The mechanical energy dissipated, per unit volume when equilibrium has
been reached is given by
e 2 Z‘ h ti2 e2
12 2
Wd ¼ 2g exp dt ¼ g 12 ¼ 2l2 e12 , ð9:77Þ
T T T
0
and this does not depend on the relaxation time.

An alternative derivation of the final result in equation (9.77) is as follows.
When the strain (9.73) is applied, the impact modulus is 2ðl1 þ l2 Þ, and this
results in an instantaneous stored elastic strain energy, 2ðl1 þ l2 Þe212 per
unit volume. When the equilibrium state is attained as t ! ‘, relaxation
modulus is 2l1 and the stored elastic strain energy is 2l1 e212 . The difference
between the stored elastic strain energies at tð0Þ and tð‘Þ is 2l2 e212 and this is
in agreement with equation (9.77).
Now suppose the stress, s12 ¼ 2l1 e12 , at tð‘Þ is suddenly removed at
time ~t ¼ 0. A stress-free equilibrium state is then approached as ~t ! ‘ and
the elastic stored energy is zero. Since the stored elastic strain energy at ~t ¼ 0
is 2l2 e212 ; it follows that this is the dissipation of mechanical energy for the
unloading process.
EXERCISES
9.1. Show that equations (9.1) and (9.7) are frame indifferent.
9.2. Show that relation (9.10) is frame indifferent and trr=3 ¼ p if and
only is the fluid is in equilibrium, that is, D[0.
9.3. Determine the nonzero components of the extra stress r þ pI, given by
equations (9.7) and (9.10) for simple shear flow with velocity field
v1 ¼ kx2 , v2 ¼ 0, v3 ¼ 0:
9.4. (a) Verify that J ð0Þ ¼ ðGð0ÞÞ1 for linear viscoelastic solids and fluids,
and J ð‘Þ ¼ ðGð‘ÞÞ1 for linear viscoelastic solids.

(b) Show that J ¼ s2 G 1 , where the superposed bar denotes the

R‘
Laplace transform defined as fðsÞ ¼ f ðtÞ expðstÞdt.
0
9.5. A Kelvin-Voigt element and a spring in series as shown in the figure
is an alternative form of the three parameter model.
σ3 E2
e2 E1 σ
e Exercise 5. Chapter 9.
e1
σ2
η
Determine the differential constitutive relation, the relaxation func-

tion, and the creep function.
9.6. A spatially homogeneous simple shear deformation is given by
e12 ðtÞ ¼ 0, t < 0,

e12 ðtÞ ¼ e12 t t9, 0 < t < t9,
e12 ðtÞ ¼ e12 ,
where e12 is a constant strain. Show that for the standard material
s12 ¼ 0, t < 0,
Tn to
s12 ¼ 2l1 e12 ðtÞ þ 2l2 e12 1 exp , 0 < t < t9,
t9 T
t
T t t9
s12 ¼ 2l1 e12 þ 2l2 e12 exp exp , t > t9,
t9 T T
where T ¼ g=2l2 is the relaxation time.
9.7. A spring-Dashpot element shown in Figure 9.5 is subjected to a strain

rate e_ H ðtÞ where e_ is constant. Show that the stress-strain relation is
given by

E2 e
rðeÞ ¼ E1 e þ ge_ 1 exp :
ge_
9.8. A ð2N þ 1Þ parameter viscoelastic model, that is, a generalization of the

three parameter standard model, is shown schematically in the figure.
E0
E1 η1
E2 η2
σ Exercise 8. Chapter 9.
EN ηN
Determine the set of relaxation times Tk ; k 2 f1; 2; :::N gs, the relaxa-
tion function and the creep function. The set of relaxation functions is
known as the relaxation spectrum.
REFERENCES
1. Whitham, G.B. (1973). Linear and Nonlinear Waves. Wiley-Interscience.

2. Tanner, R.I. (1988). Engineering Rheology. Oxford University Press.
3. Schlichting, H. (1968). Boundary-Layer Theory, 6th Ed. McGraw-Hill.
4. Rivlin, R.S. (1948). The Hydrodynamics of Non-Newtonian Fluids I. Proc.
Roy. Soc. London (A)193, pp. 260–281.
5. Christenen, R.M. (1982). Theory of Viscoelasicity, 2nd Ed. Academic
Press.
6. Flügge, W. (1975). Viscoelasticity, 2nd Ed. Springer-Verlag.
7. Lighthill, M.J. (1958). An Introduction to Fourier Analysis and General-
ised Functions. Cambridge University Press U.P.
8. Haupt, P. (2000). Continuum Mechanics and Theory of Materials.
Springer.
9. Thurston, R.N. (1984). Waves in Solids. Mechanic of Solids, Vol. IV. Ed.
C. Truesdell, Springer-Verlag.
10. Hunter, S.C. (1983). Mechanics of Continuous Media, 2nd Ed. Ellis
Horwood.
Appendix 1
Orthogonal Curvilinear Coordinate Systems
A1.1 Introduction
Rectangular Cartesian coordinate systems are well suited for presenting the
basic concepts of continuum mechanics since many mathematical com-
plications that arise with other coordinate systems, for example,
convected systems, are avoided. However, for the solution of specific prob-
lems, the use of rectangular Cartesian coordinate systems may result in con-
siderable difficulties, and the use of other systems may be desirable in order to
take advantage of symmetry aspects of the problem. For example, in the
analysis of the expansion of a thick-walled cylindrical tube, the use of cylin-
drical polar coordinates has an obvious advantage. We introduce a simple
theory of curvilinear coordinates in this appendix and specialize it for orthog-
onal curvilinear systems, in particular cylindrical and spherical. We avoid the
use of the general theory of tensor components referred to curvilinear coor-
dinates by considering what are known as the physical components of tensors
that are derived for orthogonal coordinate systems. Superscripts are used to
denote curvilinear coordinates.
A1.2 Curvilinear Coordinates
A point y 2 E3 is the intersection of three mutually perpendicular planes

whose equations are
x1 ¼ y1 , x2 ¼ y2 , x3 ¼ y3
249
250 Orthogonal Curvilinear Coordinate Systems
referred to the rectangular Cartesian coordinate system 0xi . Now introduce

three scalar functions of x,
hj ¼ hj ðx1 , x2 , x3 Þ, j 2 f1, 2, 3g, ðA1:1Þ
with single valued inverses,

xi ¼ xi h1 , h2 , h3 : ðA1:2Þ
A superscript, rather than a suffix, is used for hj , since hj are not Cartesian
coordinates. However, the summation convention applies as for suffixes.
A point with rectangular coordinate xi may also be specified by the values
of hj at the point since equation (A1.2) represent three families of surfaces
and the intersection of three surfaces hj ¼ Pj , where each Pj is a constant,
locates a point. The Pj , j 2 f1, 2, 3g are the curvilinear coordinates of the
point. It is assumed that the functions (A1.2) are at least of class C1 , that is,
continuous with continuous first derivatives, and since equation (A1.2)
must be invertible,

@xi
det 6¼ 0,
@hj
except at isolated points or on isolated lines. A set of functions (A1.1) is an

admissible transformation of coordinates if these conditions are satisfied. The
surfaces hj ¼ Pj are called coordinate surfaces and the space curves formed by
their intersections in pairs, as indicated in Figure 1.1A, are called coordinate
curves.
Consider a point P with position vector x, with respect to the origin of the
rectangular Cartesian coordinate system 0xi . At P there are three vectors gi ,
i 2 f1, 2, 3g that are tangent to the coordinate curves and are given by
@x
gi ¼ : ðA1:3Þ
@hi
A1.2 Curvilinear Coordinates 251
θ 3curve
θ 2surface θ 1surface
g3
g1 g2
Figure 1.1A. Curvilinear coordinates.
θ 3surface
θ 2curve
θ 1curve
The base vectors for rectangular Cartesian coordinate systems are given by
ej ¼ @x=@xj . Consequently,
@xs @hi
gi ¼ es , es ¼ gi : ðA1:4Þ
@hi @xs
The vectors gi are not in general unit vectors nor are they in general con-
stants, except for the special case of oblique Cartesian coordinates, that is,
when equations (A1.1) represent three nonorthogonal families of planes.
Also the vectors gi are nondimensional if and only if the corresponding hi
have the dimension of length.
Referring to Figure 1.2A, the point P has position vector x ¼ xi ei and
curvilinear coordinates hj , and the neighboring point has position vector,
x þ dx ¼ ðxi þ dxi Þei , and curvilinear coordinates hj þ dhj .
If PP# ¼ ds,
2
ds ¼ dx dx ¼ ei dxi ej dxj ,
and using equation ðA1:4Þ2 , this becomes

2 i j i j
ds ¼ gi gj dh dh ¼ gij dh dh , ðA1:5Þ
where
@xk @xk
gij ¼ gi gj ¼ , ðA1:6Þ
@hi @hj
x3
dx P′
e3 P
x
x + dx
0 e2 x2
e1
x1
Figure 1.2A. Infinitesimal displacement of point P.
are the components of what is known as the metric tensor.
A1.3 Orthogonality
A curvilinear coordinate system is orthogonal if the vectors gi form an

orthogonal triad. It then follows from equation (A1.6) that
gij ¼ 0 if i 6¼ j,
and from equation (A1.5) that
2 2 2
ds2 ¼ g11 dh1 þg22 dh2 þg33 dh3 , ðA1:7Þ
where
2 2 2
@x1 @x2 @x3
gii ¼ þ þ , ðs:c:sÞ: ðA1:8Þ
@hi @h i
@hi
Henceforth (s.c.s), following an equation means that the summation con-

vention is suspended and curvilinear coordinates only are considered. It is
convenient to put
g11 ¼ h21 , g22 ¼ h22 , g33 ¼ h23 : ðA1:9Þ

Unit base vectors bi tangential to the coordinate curves at P are given by
g g
bi ¼ i ¼ i , ðs:c:sÞ: ðA1:10Þ
gi hi
The orthogonal triad of unit base vectors, bi , at P forms a basis for a local-
ized Cartesian coordinate system; however, in general as P moves, the
orientation of the triad changes. It follows that the vectors bi , are nondi-
mensional and of constant unit magnitude but in general are not constant
vectors.
Consider the vector field uðxÞ where x is the position vector of any
point P as indicated in Figure 1.2A. The vector uðxÞ can be expressed in
terms of its components with respect to the triad of unit base vectors bi , that
is, with respect to the local Cartesian coordinate system at P as
u ¼ ui bi :
The components ui are known as the physical components of the vector

and have the same physical dimension as u . It should be noted that the
vector must be associated with a particular point since two equal vectors
associated with different points in space have in general different physical
components since the unit base vectors, bi , are, in general, not constants.
A1.4 Cylindrical and Spherical Polar Coordinate Systems
The cylindrical polar coordinate system is the simplest orthogonal curvi-

linear coordinate system.
Referring to Figure 1.3A, P# is the orthogonal projection of P in the 0x1 x2
plane, and
h1 ¼ r ¼ 0P#, h2 ¼ / ¼ x1 ^
0P#, h3 ¼ z ¼ PP#:
Consequently,
x1 ¼ r cos /, x2 ¼ r sin /, x3 ¼ z, ðA1:11Þ
where r > 0, 0 < / < 2p, ‘ < z < ‘,

and these equations can be inverted to give
1=2 x2
r ¼ x21 þ x22 , / ¼ arctan , z ¼ x3 :
x1
These equations represent a family of circular cylinders whose axes co-

incide with 0x3 , a family of meridian planes and a family of planes normal
to 0x3 , respectively. It follows from equations (A1.7), (A1.8), and (A1.11)
that
h1 ¼ 1, h2 ¼ r, h3 ¼ 1, ðA1:12Þ
ds2 ¼ dr2 þ r2 d/2 þ dz2 :
The unit base vectors, b1 , b2 , and b3 are in the directions r, /, and z in-
creasing, respectively, as indicated in Figure 1.3A.
The spherical polar coordinates of P are indicated in Figure 1.4A.
x3 b3
z
0
x2
φ Figure 1.3A. Cylindrical polar coordinates.
b2
r
P′
b1
x1
x3
b1
P
b2
θ r
0
x2
φ b3 Figure 1.4A. Spherical polar coordinates.
P′
x1
Referring to Figure 1.4A, P# is the orthogonal projection of P in the

0x1 x2 plane, and
1 2 3
h ¼ r ¼ 0P, h ¼ h ¼ x3 ^0P, h ¼ / ¼ x1 0P#:
Consequently,
x1 ¼ r sin h cos /, x2 ¼ r sin h sin /, x3 ¼ r cos /; ðA1:13Þ
where 0 < h < p, 0 < / < 2p. The angles h and / are the colatitude and
azimuth angles, respectively. Equation (A1.13) can be inverted to give

2 2 2
1=2 x21 þ x22 x2
r ¼ x1 þ x2 þ x3 , h ¼ arctan , / ¼ arctan ;
x3 x1
which represent a family of spheres with centers at the origin, a family of

cones with apexes at the origin, and a family of meridian planes, respec-
tively. It follows from equations (A1.7), (A1.8), (A1.9), and (A1.13) that
h1 ¼ 1, h2 ¼ r, h3 ¼ r sin h, ðA1:14Þ
2 2 2 2 2 2 2
ds ¼ dr þ r dh þ r sin hd/ :
A1.5 The = Operator
Rectangular Cartesian unit base vectors ei are constants, consequently in

applying either form of the gradient operator =, given by
@ðÞ @ðÞ
= [ ei , =[ ei , ðA1:15Þ
@xi @xi
to a nonscalar operand, say a vector or second-order tensor expressed in

component form u ¼ ui ei or S ¼ Sij ei ej , only the components are differ-
entiated. This results in great simplification. At any point in a tensor field,
a tensor can be expressed at that point in terms of its physical components
with respect to an orthogonal curvilinear coordinate system with unit base
vectors bi ; for example,
S ¼ S~ij bi bj ,
where S~ij denotes the physical components of S . Since the unit base vectors,
bi ; are not, in general, constants, this must be considered when S is operated
on by a differential operator such as = .
If C is a scalar field, then at a point in the field
dC
=C bi ¼ ,
dsi
where bi are the unit base vectors given by equation (A1.10) and si is the
distance along the coordinate curve hi increasing.
Henceforth, the summation convention is suspended in this appendix.
It follows from equations (A1.7) and (A1.9) that
dsi ¼ hi dhi , 2 f1, 2, 3g,

It may then be deduced that the orthogonal curvilinear form of equation

ðA1:15Þ2 is
3 1 @ðÞ
=[ + i bi : ðA1:16Þ
i¼1 hi @h
There is no difficulty in obtaining the gradient of a scalar field C,
3
1 @C
=C[ + i bi
i¼1 hi @h
since no base vectors have to be differentiated. The cylindrical polar form

of the gradient operator obtained from equations (A1.12) and (A1.16) is
@ðÞ 1 @ðÞ @ðÞ

=[ br þ b/ þ bz , ðA1:17Þ
@r r @/ @z
where b1 , b2 , and b3 are replaced by br , b/ , and bz , and the spherical polar

form is
@ðÞ 1 @ðÞ 1 @ðÞ

=[ br þ bh þ b/ : ðA1:18Þ
@r r @h r sin h @/
where b1 , b2 , and b3 are replaced by br , bh , and b/ .

In order to obtain expressions for divu, gradu, curlu, divS, etc., in
terms of the physical components of an orthogonal curvilinear coordinate
system the elements of the matrix

@bi
ðA1:19Þ
@hj
are required. It follows from equations (A1.3) and (A1.10) that
@ x @ hj bj @ ðhi bi Þ
i j ¼ i ¼ , i 6¼ j, ðs:c:sÞ
@h @h @h @hj
and differentiating gives
@bj @hj @bi @hi

hj i þ i bj ¼ hi þ bi : ðA1:20Þ
@h @h @hj @hj
Since the corresponding components of equal vectors are equal it follows

from equation (A1.20) that
@bi 1 @hj
j ¼ bj , ðA1:21Þ
@h hi @hi
for the off diagonal elements of equation (A1.19).

The diagonal elements of equation (A1.19),
@bi 1 @hi 1 @hi

i ¼ j bj bk ðA1:22Þ
@h hj @h hk @hk
are obtained from
@bi @ bj 3 bk
¼
@hi @hi
and equation (A1.20). Equations (A1.21) and (A1.22) give
2 3
0 b/ 0
@bi 4
¼ 0 br 05 ðA1:23Þ
@hj 0 0 0
for cylindrical polar coordinates, and
2 3
0 bh b/ sin h
@bi
¼ 40 br b/ cos h 5 ðA1:24Þ
@hj 0 0 br sin h bh cos h
for spherical polar coordinates.

The divergence of a vector u referred to orthogonal curvilinear

coordinates,
3 1 @u
divu ¼ + i bi ,
i¼1 hi @h
is obtained from equation (A1.16) and substitution of u ¼ b1 u1 þ

b2 u2 þ b3 u3 gives

1 @ ðu1 h2 h3 Þ @ ðu2 h1 h3 Þ @ ðu3 h1 h2 Þ
divu ¼ þ þ : ðA1:25Þ
h1 h2 h3 @h1 @h2 @h3
For cylindrical polar coordinates
@ur ur 1 @u/ @uz

divu ¼ þ þ þ
@r r r @/ @z
is obtained from equations (A1.12) and (A1.25), and for spherical polar
coordinates

2
1 @ u r r 1 @ ðuh sin hÞ 1 @u/
divu ¼ 2 þ þ
r @r r sin h @h r sin h @/
is obtained from equations (A1.14) and (A1.25).

The definition of the curl of a vector u is usually based on the form
equation (A1.15)1 of the gradient operator consequently,
curlu ¼ = 3 u,
where = is of the form ðA1:15Þ2 . This is because of the anticommutative

property of the vector product. It then follows that, for orthogonal curvi-
linear coordinates
3 1 @u
curlu ¼ + i 3 bi ,
h
i¼1 i @h
and, after substitution of u ¼ b1 u1 þ b2 u2 þ b3 u3 ,
2 3
h1 b1 h2 b2 h3 b3
1 6 @ @ @ 7
curlu ¼ det6
4 1
7:
5 ðA1:26Þ
h 1 h2 h3 @h @h2 @h3
h1 u1 h2 u2 h3 u3
For cylindrical polar coordinates

1 @uz @u/ @ur @uz 1 @ ru/ 1 @uz
curlu ¼ br þ b/ þ bz
r @/ @z @z @z r @r r @/
is obtained from equations (A1.12) and (A1.265). For spherical polar

coordinates

br @ u/ sin h @uh bh 1 @ur @ ru/
curlu ¼ þ
r sin h @h @/ r sin h @/ @r

b/ @ ðruh Þ @ur
þ
r @r @h
is obtained from equations (A1.14) and (A1.26).

The gradient of the vector u has in general nine components and is
given by
3 1 @u
=u¼ + i bi , ðA1:27Þ
i¼1 hi @h

where a second summation arises when @u @hi is evaluated using equations
(A1.21) and (A1.22). It follows from equations (A1.23) and (A1.27) that,
for cylindrical polar coordinates, the matrix of components of = u is
2 3
@ur 1 @ur @ur
a/
6 @r r
@/ @z 7
6 7
½= u ¼ 6 @u/ 1 @u/
a/
@u/
7,
4 @r r @/ @z 5
@uz 1 @uz @uz
@r r @/ @z
A1.6 The =2 Operator 261
and from equations (A1.24) and (A1.27) that, for spherical polar
coordinates,
2
3
@uR 1 @ur 1 @ur
@r r @h uh u/ sin h
r sin h @/
6
7
6 h 7
½= u ¼ 6 @u 1 @uh
r @h þ uR
1 @uh
r sin h @/ u/ cos h 7:
4 @r 5
@u/ 1 @u/ 1 @u/
@r r @h r sin h @/ þ uh cos h þ ur sin h
A1.6 The =2 Operator
The =2 operator, sometimes known as the Laplacian operator, is a scalar dif-

ferential operator given by
2
3 @2
= [ +
i¼1 @xi @xi
referred to a rectangular Cartesian coordinate system. This operator, which

may operate on a scalar, vector, or higher-order tensor can be expressed as
=2 [ div grad ¼ = =,
and with the use of equation (A1.16),

2 1 @ h2 h3 @ðÞ @ h3 h1 @ðÞ @ h1 h2 @ðÞ
= [ þ 2 þ 3 ,
h1 h2 h3 @h1 h1 @h1 @h h2 @h2 @h h3 @h3
ðA1:28Þ
referred to orthogonal curvilinear coordinates. For cylindrical polar coor-

dinates, it follows from equation (A1.28) that
@2 1@ 1 @2 @2
=2 [ 2þ þ 2 2þ 2: ðA1:29Þ
@r r @r r @/ @z
and for spherical polar coordinates

1 @ 2 @ðÞ 1 @ @ðÞ
=2 [ 2 r þ 2 sin h
r @r @r r sin h @h @h
ðA1:30Þ
1 @ 2 ðÞ
þ 2 2 2 :
r sin h @/
When equation (A1.29) or (A1.30) operates on a vector it is important to

note that the base vectors are not, in general, constants, which results in
some complication, as indicated in this example, for cylindrical polar
coordinates,

2 2 ur
2 @u/ 2 u/ 2 @ur 2
= u ¼ br = ur 2 2 þ b/ = u/ 2 þ 2 þ bz = u z :
r r @/ r r @/
The terms that result from the nonconstant base vectors, br and b/ , are
clearly evident.
Appendix 2
Physical Components of the Deformation

Gradient Tensor
The concept of physical components of vectors and second-order tensors,

with respect to orthogonal curvilinear coordinate systems, was introduced
in Appendix 1. In this appendix, extension to two-point tensors is consid-
ered, in particular, the deformation gradient tensor,
F ¼ =X x,
given in terms of its Cartesian components by
@xi
F¼ ei EK ,
@XK
in the notation of chapter 2. In what follows lowercase letters refer to

the spatial configuration and uppercase letters to the natural reference
configuration as for the rectangular Cartesian expressions. The curvilinear
coordinates and base vectors, referred to the reference configuration,
are ðR, U, ZÞ and ðBR , BF , BZ Þ, respectively, for cylindrical polar
coordinates, and ðR, Q, FÞ and ðBR , BQ , BF Þ, respectively, for spherical
polar coordinates. Then, the gradient operator =X , referred to cylindrical
polar coordinates, is
@ðÞ 1 @ðÞ @ðÞ

=X [ BR þ BF þ BZ , ðA2:1Þ
@R R @U @Z
263
264 Physical Components of the Deformation Gradient Tensor
and referred to spherical polar coordinates is
@ðÞ 1 @ðÞ 1 @ðÞ

=X [ BR þ BH þ BU ðA2:2Þ
@R R @Q R sin Q @U
These are of the same form as the corresponding forms of = given in

Appendix 1.
The position vector x of a material particle in the spatial configuration
is given, in terms of its cylindrical polar coordinates ðr, /, zÞ, by
x ¼ rbr þ zbz , ðA2:3Þ
and in terms of its spherical polar coordinates ðr, h, /Þ, by
x ¼ rbr : ðA:2:4Þ
The component matrix of F for cylindrical polar coordinates is now

obtained. It follows from equations (A2.1) and (A2.3) that F can be
@ 1 @ @
F¼ ðrbr þ zbz Þ BR þ ðrbr þ zbz Þ BF þ ðrbr þ zbz Þ BZ :
@R R @F @Z
ðA2:5Þ
The matrix (A1.22),
2 3
0 b/ 0
@bi 4
j ¼ 0 br 0 5, i, j 2 fr, /, zg,
@h 0 0 0
is required in order to perform the differentiations in equation (A2.5).

Then
2 3 2 3 2 3 2 3 2 3
b b b b b
@ 4 r5 @ 4 r 5 @r @ 4 r 5 @/ @ 4 r 5 @z 4 / 5 @/
b/ ¼ b/ þ b/ þ b/ ¼ br :
@R @r @R @/ @R @z @R @R
bz bz bz bz 0
ðA2:6Þ
Physical Components of the Deformation Gradient Tensor 265
Similarly
2 3 2 3
b b
@ 4 r 5 4 / 5 @/
b/ ¼ br ðA2:7Þ
@F @F
bz 0
and
2 3 2 3
br b/
@ 4 5 4 @/
b/ ¼ br 5 : ðA2:8Þ
@Z @Z
bz 0
Use of relations (A2.6) to (A2.8) along with equation (A2.3) gives
@r 1 @r @r
F¼ br B R þ br B U þ br B Z þ
@R R @F @Z
@/ r @/ @/
r b/ BR þ b/ B U þ r b/ BZ þ
@R R @F @Z
@z 1 @z @z
bz B R þ bz BU þ bz BZ :
@R R @F @Z
Consequently,
2 3 2 @r 1 @r @r
3
FrR FrF FrZ @R R @F @Z
4 F/R 6 @/ @/ 7
F/F F/Z 5 ¼ 4 r @R r @/
R @F r @Z 5: ðA2:7Þ
FzR FzF FzZ @z 1 @z @z
@R R @F @Z
The component matrix of F for spherical polar coordinates
2 3
2 @r 1 @r 1
3
@r
FrR FrH FrU @R R @Q R sin h @F
4 FhR 6 @h 7 ðA2:8Þ
FhH FhU 5 ¼ 4 @h
r @R r @h
R @Q
r
R sin h @F 5
F/R F/H F/U @/ r @/ r @/
r sin h @R R sin h @Q R @F
is obtained in a similar manner.

The physical components of other tensors such B ¼ FF T , C ¼ F T F,
and E ¼ 12 ðC I Þ are readily obtained from equations (A2.7) and (A2.8).
Appendix 3
Legendre Transformation
Consider a function /ðx; yÞ of two variables, x and y, with differential
@/ @/
d/ ¼ dx þ dy: ðA3:1Þ
@x @y
Let
@/ @/
p¼ and q ¼ ; ðA3:2Þ
@x @y
so that equation (A3.1) becomes
d/ ¼ pdx þ qdy: ðA3:3Þ
The purpose of the translation is to change the function / of x and y to

a function of the new variables p and q with its differential expressed in
terms of dp and dq . Let wðp; qÞ be defined by
w ¼ / px qy, ðA3:4Þ
with differential,
dw ¼ d/ pdx xdp qdy ydq: ðA3:5Þ
266
Legendre Transformation 267
Substituting equation (A3.3) in equation (A3.5) gives
dw ¼ xdp ydq,
where
@w @w
x¼ , y¼ : ðA3:6Þ
@p @q
Equation (A3.4) is a dual Legendre transformation. A single Legendre

transformation involves one variable. For example, instead of equation
(A3.4), let wðp; yÞ be defined by
w ¼ / px, ðA3:7Þ
with differential
dw ¼ d/ pdx xdp: ðA3:8Þ
Substituting equation (A3.3) in equation (A3.8) gives
dw ¼ qdy xdp,
where
@w @w
x¼ , q¼ : ðA3:9Þ
@p @y
The potentials of thermoelasticity, which are fundamental equations of

state, are the internal energy uðC, sÞ, the Helmholtz free energy f ðC, HÞ,
the enthalpy hðS, sÞ, and the Gibbs free energy gðS, HÞ . These are inter-
related by Legendre transformations. An example of a single Legendre
transformation in thermoelasticity theory is
f ¼ u Hs:
268 Legendre Transformation
It follows from equation ðA3:9Þ1 that
sðC, HÞ ¼ @f =@H, and qðC, HÞ ¼ @f =@C:
It can be shown otherwise that
q ¼ S=ð2q0 Þ:
An example of a dual Legendre transformation in thermoelasticity theory is
SKL CKL
gðS, HÞ ¼ u Hs:
2q0
It follows from equation (A3.6) that
@g @g
sðS, HÞ ¼ , CKL ðS, HÞ ¼ 2q0 :
@H @SKL
Appendix 4
Linear Vector Spaces
In order to consider linear vector spaces it is useful to first consider a special

case, the set of vectors of elementary vector analysis. The vector sum x þ y
of any two vectors of the set is given by the parallelogram rule, and the set
has the following properties:
a. x þ y ¼ y þ x (commutative property),
b. x þ ðy þ zÞ ¼ ðx þ yÞ þ z (associative property),
c. There exists a zero vector, 0, such that x þ 0 ¼ x,
d. For every x there is a negative, ðxÞ, such that x þ ðxÞ ¼ 0.
When a vector x is multiplied by a real scalar a the result is a vector ax,

and scalar multiplication has the following properties:
e. aðbxÞ ¼ abx (associative property),

f. ða þ bÞx ¼ ax þ bx (distributive property),
g. aðx þ yÞ ¼ ax þ ay (distributive property).
The vector space of elementary vector analysis, denoted by V, is a special

case of a Euclidean space since it has a scalar or inner product
x y ¼ x1 y1 þ x2 y2 þ x3 y3 ¼ xi yi , ðA4:1Þ
in terms of the components of vectors x and y referred to a three-dimensional

rectangular Cartesian coordinate system. The scalar product has properties
h. x y ¼ y x (commutative property),
i. x ðy þ zÞ ¼ x y þ x z (distributive property),
j. x x > 0 (equality holds when x ¼ 0 )
269
270 Linear Vector Spaces
The relation (A4.1) implies that V is a three-dimensional space and the

dimensionality of a vector space is made precise later in this appendix.
A set of vectors that are normal to a particular vector, and includes the
vector 0, is a two-dimensional subspace of V since properties (a) to (i) are
valid with the scalar product given by x y ¼ x1 y1 þ x2 y2 .
Now consider a set quantities x1 , x2 , x3 , xn that have properties
analogous to (a) to (g) and have a scalar product qðxi , xj Þ with properties
analogous to (h) and (i). These set quantities constitute what is known as
a Euclidean linear vector space or inner product space, and the elements
are regarded as vectors in a general sense. That is, the space of vectors of
elementary vector analysis is a special case. The vectors x1 , x2 , x3 , xn
form a linearly independent system of order n if
a1 x1 þ a2 x2 þ a3 x3 þ an xn ¼ 0
holds only when all the numbers a1 ; a2 ; a3 a4 are zero, otherwise the vec-
tors are linearly dependent. The basis of a linear vector space is any linearly
independent system maximum order. For example, for the three-dimensional
space of vectors of elementary vector analysis any three non coplanar vectors
x, y, and z is a basis. This means any vector u of the space can be expressed as
a linear combination
u ¼ a1 x þ a2 y þ a3 z
of the basis vectors.

In Cartesian tensor analysis the orthogonal system of unit base vectors
fe1 ; e2 ; e3 g with ei ej ¼ dij , is linearly independent, and a vector u can be
expressed in rectangular Cartesian component form as
u ¼ u1 e1 þ u2 e2 þ u3 e3
as shown chapter 1.
Linear Vector Spaces 271
Another example of Euclidean linear vector space is the space of

ð1 3 nÞ row matrices with vectors in the general sense given by vector
a ¼ ða1 , a2 , an Þ . A basis for this space is the set
fð1, 0, , 0Þ, ð0, 1, , 0Þ, , ð0, 0, ,1Þg
of n linearly independent vectors, and the scalar product of vectors a and b

is given by
n
qða, bÞ ¼ + ai bi :
i¼1
The space of vectors of elementary vector analysis is a special case with

n ¼ 3 (or n ¼ 2 for a set of coplanar vectors) and each element of a row
matrix regarded as a component of a vector in the sense of elementary
vector analysis.
An important vector space is the set of second-order tensors, and this is
shown to be a Euclidean linear vector space as follows. If X, Y, and Z are
second-order tensors, referred to a three-dimensional point space, the prop-
erties (a) to (g), with x, y, and z replaced by X, Y, and Z are valid. Since
a second-order tensor can be given in terms of Cartesian components by
X ¼ Xij ei ej ,
it follows that the space of second-order tensors is nine dimensional and

ei ej is a possible basis. This basis is orthogonal since the inner product of
ei ej and ek et is
h T i
tr ei ej ðek et Þ ¼ dik djt :
A scalar product

qðX, Y Þ ¼ X Y ¼ tr XY T
with x, y, and z replaced by X, Y, and Z, has properties (h) to (j).

Index
ablation, 54 Clausius-Duhem inequality, 118, 207

adiabatic deformation, 150 Clausius-Planck inequality, 120
Almansi’s strain tensor, 77, 147 coefficient of volume thermal expansion,
alternating tensor, 26, 28 124, 213
axial vector. See pseudo vector cofactor, 29, 161
azimuth angle, 255 co-latitude angle, 255
azimuthal shear, 197 column matrix, 12
commutative rule, 15
base vectors, 250 complementary energy, 148
basic invariants, 21 compliance tensor, 128
basis vectors, 12 compressible generalization of the
Biot strain tensor, 78 Mooney-Rivlin, 169
Blatz P.J., 169 compressible generalization of the
body force per unit mass, 86 neo-Hookean strain energy func-
bound vector, 9 tion, 169, 215
Boyle’s law, 124 compressible isotropic models, 210
bulk modulus, 160, 209, 210 conjugate properties, 203
bulk viscosity, 125, 126, 221, 222, 239 conjugate rate variable, 105
conjugate tensors, 159
canonical form, 23 conservation of energy, 115
Cartesian tensor analysis, 4 conservation of mass, 63
Cauchy elasticity, 153 constant entropy, 202
Cauchy stress tensor. See stress tensors constitutive equation, 122
Cauchy-Green tensors, 144 contact force, 84
Cayley-Hamilton Theorem, 24, 158, 159 continuous bodies, 50
Chadwick P., 209, 216 contravariant, 4
change of entropy, 205 coordinate curves, 250
change of internal energy produced by coordinate surfaces, 250
deformation, 205 covariant, 4
characteristic equation, 20, 22 Creasy C.F.M., 209
Charles law, 124 creep, 224
Christensen R.M., 224, 240 creep function, 227, 247
classical elasticity, 126, 166 current configuration, 90, 112
classically thermoelastic bar, 214 curvilinear coordinate system, 4, 250
Clausius statement of the second law, cylindrical bar, 165
123, 208
272
Index 273
cylindrical polar coordinates, 175, 253, Eringen C., 138

259, 264 Euclidean linear vector space, 271
cylindrical polar form, 257 Euclidean point space, 1
cylindrical symmetry, 165 Euclidean transformation, 140
Euler’s Equation, 60
d’Alembert’s Principle, 114 Eulerian coordinates. See spatial
deformation fields, 168 coordinates
deformation gradient tensor, 58, 59, 70, Eulerian strain tensor. See Almansi’s
74, 108, 152 strain tensor
dilatational part, 74 Eulerian tensors, 90
isochoric part, 74 eversion of a cylindrical tube, 185
density, 86 experimental results, 212
reference configuration, 99
spatial configuration, 63, 99 finite axial strain, 214
determinant, 20, 28, 31, 58 finite deformation elastostatics for
determinism, 138 isotropic hyperelastic solids, 168
dilatation, 58, 210, 222 first law of thermodynamics, 115
Dirac delta function, 229 first order tensor fields. See vector field
displacement gradient tensor, 78 Flügge W., 224
displacement vector, 76 foam rubbers, 160
dissipation inequality, 118 Fourier’s law, 115, 123
dissipation potential, 223 fourth order tensor, 26
divergence, 42 frame of reference, 139
divergence of a vector, 39 free suffix, 6
divergence operation, 100 fundamental equation of state, 136, 203
divergence theorem, 44, 45, 100, 106, 112,
118 Galilean frame of reference, 141
dyadic product, 11, 16 Galilean transformations, 143
generalized functions, 229
eigenvalues, 23, 24, 34, 35, 36, 91, Gibbs free energy, 137
163 Gibbs relation, 132, 135, 204
eigenvectors, 22, 23, 24, 34, 35, 36, 91 Goodier, 127
163 gradient operator, 263
energetic, 205, 218 gravitational potential energy, 115
energetic response, 202 Green A.E., 180
energy equation, 115, 122 Green deformation tensor, 156
enthalpy, 136 Green elastic, 155
entropic, 205, 218 Green strain rate tensor, 156
entropic response, 202 Green’s strain tensor, 75, 77, 78, 156
entropic thermoelasticity, 215
entropy, 118, 213 Hadamard strain energy function, 169
entropy inequality, 207 Haupt P., 114
entropy inequality for a thermoelastic heat conduction, 220
solid, 207 heat conduction tensor, 115, 123
entropy jump across a shock, 208 heat flux vector, 115, 123, 138
entropy production, 208, 222 Helmholtz free energy, 119, 131, 155, 202
equation of equilibrium, 111 hereditary integral, 233
equation of motion, 106 Hill R., 104
equilibrium thermodynamics, 114 Holzapfel G.A., 215
Ericksen J.L., 168 homogeneous deformation, 168
274 Index
Hooke’s law, 127 isotropic tensor, 24, 28

generalized, 127 isotropy, 153, 154, 160
Hunter S.C., 240
hydrostatic state of stress, 92, 95 Jacobian of the transformation, 58
hyperelastic solid, 155, 156, 160, 162 Jaumann rate, 145
hyperelastic stress-deformation relations, Joule heating, 116
170
hyperelasticity, 157 Kelvin-Voigt element, 246
hysteresis effects, 208, 220 Kelvin-Voigt viscoelastic solid, 225, 228,
239
ideal inviscid fluid, 123 Kestin J., 202
ideal rubber, 212 Killing’s theorem, 67
impact modulus, 227, 232 kinematically admissible virtual
improper orthogonal tensor, 32 displacement field, 110
incompressibility, 58, 160 kinetic energy, 106, 155
incompressible fluid, 122 Ko W.L., 169
incompressible Hyperelastic solid, 160 Kronecker delta, 6, 17, 91
incompressible isotropic models, 210
incompressible materials, 160
Lagrangian configuration. See reference
incompressible solid, 210
configuration
inertia tensor of the element, 69
Lagrangian multiplier, 93, 126, 160, 162,
inertial frame of reference, 143
180, 222
infinitesimal rotation, 80
Lagrangian strain, 76
infinitesimal rotation tensor, 80
Lagrangian strain rate, 109
infinitesimal strain tensor, 78, 80, 160
Lagrangian strain tensor, 204
inner product, 5, 17
Lamé’s constants, 129
internal energy, 124, 131, 155, 202, 212
Laplace expansion, 28
internal variable, 231, 238
Laplace transform, 236, 246
invariants, 158, 223
Laplacian operator, 261
inverse, 30, 59
laws of thermodynamics, 121
inverse methods, 168
left Cauchy strain tensor, 75
inversion effect, 218
left Cauchy-Green strain tensor, 172
inviscid fluids, 220
left polar decompositions, 36
isentropic deformation, 155
left stretch tensor, 72, 165
isentropic process, 120
Legendre transformation, 131, 132, 135,
isentropic simple tension, 213, 214, 216,
137, 203
217
Leibniz’s rule, 62
isentropic strain energy, 202
Lighthill M.J., 230
isochoric deformation, 160, 210
linear theory of elasticity, 150
isochoric flow, 124, 125
linear thermoelasticity, 130
isochoric motions, 58
linear vector spaces, 269
isothermal, 150
linear viscoelastic medium, 224
isothermal deformation, 155, 213
linearly independent system, 270
isothermal moduli, 131
local action, 138
isothermal shear, 210
logarithmic strain tensors, 78
isothermal simple tension, 212, 213
Love A.E.H., 127
isothermal strain energy, 202
isotropic elastic solid, 157
isotropic state of stress. See hydrostatic Malvern L.E., 111
state of stress mass point, 50
Index 275
material configuration. See reference octahedral shearing stress, 96

configuration Ogden R.W., 102, 104, 174, 175, 189
material description, 50, 53, 77 Oldroyd J.G., 146
material frame indifference, 138 orthogonal curvilinear coordinate sys-
material lines, 59 tems, 249, 258
material objectivity, 167 orthogonal tensor, 32, 33, 36, 71
material surface, 54, 59 orthogonal transformations, 9, 15
material symmetry, 153 orthogonality, 252
material volume, 59, 63 orthonormal, 6
Mathematica, 24, 35, 37
matrix, 2 parallelogram rule, 5
matrix multiplication, 1 particle, 50
Maxwell fluid, 225 particle path, 52
mechanical compressibility, 210 partly energetic, 209
mechanical dissipation, 243 Pearson C.E., 95
mechanically incompressible fluid, 160 perfectly elastic solid, 119
mechanically incompressible solid, 160 physical components, 253, 256
metric tensor, 252 piezotropic model, 208
modified deformation gradient tensor, 74 plane deformation, 171
modified entropic elasticity, 218 Poisson’s ratio, 169
Mooney-Rivlin material, 179 polar decomposition theorem, 35, 70
Mooney-Rivlin solid, 197 polar media, 85
Mooney-Rivlin strain energy function, polar vector, 1, 10
166, 178 polymers, 209
Müller I., 124 positive definite second order tensor, 21,
mutually orthogonal, 21, 22 32, 33, 34, 35, 36
positive semidefinite, 33, 34
natural reference configuration, 50 potential energy, 116
Navier-Stokes equation, 126 Poynting effect, 172, 179
negative definite, 34 principal axes, 34, 67
negative definite tensors, 36 principal components, 164
neoHookean elastic solid, 167 principal directions, 91
neo-Hookean solid, 188 principal invariants, 163, 173
neo-Hookean strain energy function, 169 principal stresses, 91, 92, 93
Newton’s second law, 86 principal stretches, 71, 72, 163, 210
Newtonian viscous fluid, 125, 220, 239 modified stretches, 74
nominal stress, 98, 112, 162 principle of local state, 114, 202, 243
nominal stress tensor, 98 principle of virtual work, 164
non-ideal rubbers, 209 proper orthogonal tensor, 32, 71, 152,
non-newtonian viscous fluid, 223 154
non-singular second order tensor, 35 pseudo scalar, 10
normal component, 85 pseudo vector, 10
normal stress, 89, 93 pure bending of an hyperelastic plate, 189
normalized eigenvectors, 21 pure dilatation, 125, 130
numerical results, 217 pure hydrostatic stress, 95
pure shear stress, 94, 95
objectivity, 151
oblique Cartesian coordinate system, 4 radiation, 116
observer, 139, 140 rate of change of momentum of the
octahedral plane, 96 body, 86
276 Index
rate of deformation, 107, 113 shearing stress, 89

rate of deformation tensor, 65, 67, 77, 159 shearing stress on the element, 87
rate of entropy production per unit shock wave propagation, 220
mass, 208 shock waves, 208
rate of heat supply, 115, 123 simple materials, 138
rectangular Cartesian coordinate system, simple shear, 125, 165, 171
1, 2, 3, 6 simple shear deformation, 170
reference configuration, 50, 108, 112, 117 simple tension, 172, 212
referential description, 50, 53 simple tension of the neo-Hookean solid,
referential heat flux, 207 174
Reiner-Rivlin fluid, 223 Sokolnikov I.S., 127, 128
relation between the velocity gradient spatial configuration, 51, 84, 90, 111, 113,
and the deformation gradient 118, 154
tensor, 65 spatial description, 51, 53
relative description, 51 spatial form of continuity equation, 63
relaxation function, 226, 232, 246 spatial heat flux, 207
relaxation modulus, 224, 227, 246 specific enthalpy, 203
relaxation time, 226, 237, 246 specific Gibbs free energy, 203
resultant moment acting on a body, 86 specific heat, 122, 133
right Cauchy Green deformation tensor, specific heat at constant deformation, 209
75, 152, 157, 204 specific Helmholtz free energy, 203
right polar decompositions, 36 specific internal energy, 115, 203
right stretch tensor, 72, 165 specific strain energy, 156
rigid body, 122 spectral decomposition, 163
rigid body mechanics, 122 spectral form, 24, 34
rigid body motion, 122 spherical polar coordinates, 180, 254, 259,
rigid motion, 121 264, 265
Rivlin R.S., 189 spherical polar form, 257
rotation, 151 spherical symmetry, 165
rate of deformation, 107, 113 spin tensor, 65, 70, 107, 148
rubber-like materials, 160 spring-dashpot models, 224
rubber-like solids, 210 square matrices, 13
square root of a second order tensor, 35,
scalar, 1, 2, 17 36
scalar differential operator, 40 simple shear, 125, 165, 171
scalar field, 38 spin tensor, 65, 70, 107, 148
scalar function, 223 standard material, 247
scalar invariants, 9, 21 standard model, 230, 239
scalar product, 5, 32 statically admissible stress, 112
Schlichting H., 222 statically admissible stress field, 110, 113
second law of thermodynamics, 117, 124 stationary values, 93
second order tensor field, 38 statistical theory of rubber elasticity,
shear and bulk relaxation functions, 234 212
shear deformation, 222 steady flow, 52, 54
shear modulus, 160, 209 stiffness tensor, 127
shear modulus for infinitesimal Stokes’ hypothesis, 126, 222
deformation, 169 Stokes’ theorem, 44
shear viscosity, 221 Stokesian fluid, 223
shearing component, 85 stored energy, 155
Index 277
stored energy function, 155 third order, 25

strain energy, 239 tensor addition, 19
strain energy function, 156, 157, 159 tensor character, 19
strain energy rate, 159 tensor contraction, 19
strain rate tensor, 78, 80 tensor field, 38
stream line, 52 tensor multiplication, 18, 19
stress boundary conditions, 162 tensor product, 11, 25
stress discontinuities, 110 tensor transformation rule, 18
stress power, 155, 159, 221 thermal equation of state, 122
stress relaxation, 224 thermal strain, 130
stress tensor thermal wave speed, 115
Biot, 101, 164, 214 thermodynamic potentials, 202
Cauchy, 2, 90, 91, 106, 150, 153, 158, thermodynamic pressure, 125, 221, 222
160, 164 thermodynamic properties, 131
deviatoric part, 95 thermodynamics of irreversible
isotropic part, 95 processes, 114
first Piola-Kirchhoff, 107 thermoelastic inversion effect, 215
nominal, 107, 212 thermoelastic potentials, 131
second Piola-Kirchhoff, 100, 107, 153, third invariant, 29
156, 158, 203 third order isotropic tensor, 26
stress vector, 2, 84, 85 Thurston R.N., 240
strictly entropic elasticity, 209, 212, 213, Timoshenko S., 127
215 trace, 17
substitution property, 17, 91 transformation matrix, 7, 15
suffix notation, 4, 18 transformation rules for base vectors, 8
summation convention, 35 transport equation, 63
superposition principle, 227 transport properties, 126
surface traction, 85 transpose, 13
symbolic form, 12 Tresca yield condition, 94, 98
symbolic notation, 4, 17 triple scalar product, 28
symmetric part of the displacement true scalar, 10
gradient tensor, 80 Truesdell C.A., 168
symmetric second order tensor, 34
symmetry group, 154 unit base vectors, 7
symmetry transformation, 154 unit step function, 226, 229
Tanner R.I., 220 valanis-Landel hypothesis, 165

telescopic shear of the neo-Hookean vector, 1, 3, 9, 12
solid, 200 vector field, 38, 41
temperature, 121 vector gradient, 39, 41
tensor, 2 vector magnitude, 9
first order. See vector velocity, 2, 54
fourth order, 25 velocity gradient tensor, 64
higher order, 2 virtual displacements, 110
second order, 1, 12, 13, 34, 42 virtual work, 110
antisymmetric part, 14 viscoelastic, 237
deviatoric part, 18 viscosity coefficient, 125, 238
isotropic part, 18 viscous fluid, 124
symmetric part, 13, 14, 22, 33, 34, 91 Voigt matrix, 130
278 Index
volume coefficient of thermal expansion, White F.M., 126

160, 209, 210, 215, 216, 240 Whitham, 220
volume thermal expansion, 160 Zerna W., 168, 180
von Mises yield criterion, 97 zeroth order tensor field. See scalar
vorticity vector, 67 field

8XTjICF2ER Elements of Continuum Mechanics PDF

Uploaded by

Copyright:

Available Formats

8XTjICF2ER Elements of Continuum Mechanics PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

8XTjICF2ER Elements of Continuum Mechanics PDF

Uploaded by

Copyright:

Available Formats

This page intentionally left blank

Elements of Continuum Mechanics and Thermodynamics

Joanne L. Wegner is a Professor of Mechanical Engineering at the University of

James B. Haddow received the Canadian Congress of Applied Mechanics award

Cambridge University Press

This publication is in copyright. Subject to statutory exception and to the

ISBN-13 978-0-511-53375-4 eBook (EBL)

ISBN-13 978-0-521-86632-3 hardback

Cambridge University Press has no responsibility for the persistence or accuracy

1. Cartesian Tensor Analysis .................................................................... 1

2. Kinematics and Continuity Equation ................................................ 50

2.7 Velocity Gradient 64

4. Work, Energy, and Entropy Considerations .................................. 105

5. Material Models and Constitutive Equations ................................ 121

6. Finite Deformation of an Elastic Solid ........................................... 150

7. Some Problems of Finite Elastic Deformation .............................. 168

8. Finite Deformation Thermoelasticity ............................................. 202

9. Dissipative Media .............................................................................. 220

9.2 A Non-Newtonian Viscous Fluid 223

APPENDIX 1: Orthogonal Curvilinear Coordinate Systems .............. 249

APPENDIX 2: Physical Components of the Deformation Gradient

APPENDIX 3: Legendre Transformation .............................................. 266

APPENDIX 4: Linear Vector Spaces ..................................................... 269

This book is based on notes prepared for a senior undergraduate or begin-

of general tensor analysis involving contravariant and covariant compo-

In this chapter we present an elementary introduction to Cartesian tensor

components of the original vector. These components are with respect to

coordinate systems. Intuitively, a vector may be regarded as a directed

1.2 Rectangular Cartesian Coordinate Systems

The simplest type of coordinate system is a rectangular Cartesian system,

which is a bound vector or position vector.

Figure 1.1. Right-handed rectangular Cartesian coordinate system.

coordinate system, 0x1 x2 x3 by

Henceforth in this chapter when the term coordinate system is used,

1.3 Suffix and Symbolic Notation

Suffixes are used to denote components of tensors, of order greater than

a lowercase letter suffix takes the values 1, 2, and 3 for three-dimensional

It is convenient to use a summation convention for repeated letter suffixes.

which is consistent with the parallelogram rule. A further example of the

The magnitude juj of a vector u is given by

Cii ¼ C11 þ C22 þ C33

1.4 Orthogonal Transformations

The scalar products of orthogonal unit base vectors are given by

where dij is known as the Kronecker delta and is defined as

uj dij ¼ u1 di1 þ u2 di2 þ u3 di3 ¼ ui : ð1:6Þ

Consider a right-handed rectangular Cartesian coordinate system 0x9i with

xi ei ¼ x9j e9j , ð1:7Þ

x9k ¼ aki xi , ð1:8Þ

aki ¼ e9k ei ¼ cosðx9k 0xi Þ: ð1:9Þ

xi ¼ aki x9k : ð1:10Þ

Table 1.1. Direction cosines for

It follows from equations (1.8) and (1.10) that

and from equation (1.7) that

e9k ¼ aki ei , ð1:13Þ

ei ¼ aki e9k : ð1:14Þ

A relation similar to equations (1.8) and (1.10),

u9k ¼ aki ui , and ui ¼ aki u9k ð1:15Þ

is obtained for a vector u since ui ei ¼ u9k e9k , which is similar to equation

The magnitude juj ¼ ðui ui Þ1=2 of the vector u is independent of the

ðu vÞc ¼ uðv cÞ, ð1:22Þ

and can be expressed as follows in terms of the basis vectors, ei ej :

Sij ei ej ¼ S9kl e9k e9l ,

u v ¼ trðu vÞ: ð1:48Þ